Rural Landscapes along the Vardar Valley: Two site-less surveys near Veles and Skopje, Republic of Macedonia

Cover Page The handle http://hdl.handle.net/1887/28917 holds various files of this Leiden University dissertation. Author: Donev, Damjan Title: Rural landscapes along the Vardar Valley: two site-less surveys near Veles and Skopje, the Republic of Macedonia Issue Date: 2014-09-24 Rural Landscapes along the Vardar Valley: Two site-less surveys near Veles and Skopje, the Republic of Macedonia Proefschrift ter verkrijging van de graad van Doctor aan de Universiteit Leiden, op gezag van Rector Magnificus prof. mr.Carel Stolker, volgens besluit van het College voor Promoties te verdedigen op 23.09.2014 door Damjan Donev geboren te Skopje, the Republic of Macedonia in 1979 1 Promoter: Other committee members: Prof. J.L. Bintliff, BOŽIDAR SLAPŠAK PROFESSOR LJUBLJANA MAARTEN JANSEN PROFESSOR LEIDEN PETER ATTEMA PROFESSOR GRONINGEN HANS KAMERMANS DR. LEIDEN TESSE STEK DR. LEIDEN MIHAILO POPOVIĆ DR. AUSTRIAN ACADEMY OF SCIENCES/UNIVERSITY OF VIENNA 2 Contents Acknowledgements 6 Chapter I: Introduction, theoretical approaches and method of fieldwork 7 I.1.1 Archaeological surveys in the Republic of Macedonia 7 I.1.2 Theoretical definitions; the study of human settlement as habitation practices; from settlements to landscapes 11 I.1.3 The principle aims of the survey, definition of the survey areas and a couple of old-new concepts 19 I.2 Method of fieldwork 26 I.2.1 Division of the study areas into field walking blocks, the factor of the field block’s size 26 I.2.2The field block and the individual transect, problems of varying survey intensity 28 I.2.3 The factor of ground visibility 31 I.2.4 The impact of modern human activities 32 I.2.5 Categories of counted material, other types of field documentation 35 I.2.6 The regular grid survey and collection of surface material 37 I.2.7 Conclusions 43 Chapter II: Description of the first survey area 46 II.1 The valley of Sopot: its drainage basin, size and limits, geo-morphology, communication, natural conditions and land-use 46 II.2 The 11 survey sectors or the main topographic components of the survey area54 II.3 The geo-strategic importance of the valley of Sopot in its wider geographic context 56 Chapter III: Chronology of the collected material and spatial distribution by periods 60 III.1.1 Introduction 60 III.1.2 The Middle Neolithic Period (6th millennium BC?) 62 III.1.3 The Late Neolithic (5th and first half of the 4th millennium BC?) 63 rd nd III.1.4 The Bronze Age (3 and 2 millennium BC, until ca. 1200 BC) 65 III.1.5 The Late Bronze Age (1600-1200 BC) 68 III.1.6 The Dark Ages/Early Iron Age (1200-1000; 1000-800 BC) 69 III.1.7 The Early Iron Age? (1000-800BC) 72 III.1.8 The Late Iron Age (7th and 6th century BC) 75 III.1.9 The problem of the Early Antique layer (Classical Period, 5th and 4th century BC) 79 III.1.10 The Hellenistic Period (late 4th – late 1st century BC) 81 III.1.11 The Roman to Late Roman Period (AD 1st-end of 6th century) 84 th th III.1.12 The Late Byzantine and Early Ottoman Period (14 -18 century)102 III.1.13 The Late Ottoman and Early Modern Period (1800-1950 AD) 107 3 III.2 Conclusions 113 III.2.1 History of habitation in the survey area in relation to the regional context and major historical circumstances 113 III.2.2 The dynamic of settlement in the survey area, the long-term trends and cycles, the carrying capacity 145 Chapter IV: The region of Montenegro near Skopje and the second survey area159 IV.1 Name and geographic location of the wider study region 159 IV.2 Limits and geography of the wider study region 160 IV.3 Geomorphology and geology 165 IV.4 Land-use, agricultural and mineral resources; vegetation, soils and climate168 IV.5 Modern day settlement pattern, settlement locations and settlement territory171 IV.6 The survey area, limits and description 177 Chapter V: Chronology of the collected surface finds and spatial distribution by periods 182 V.1 Introduction 182 V.2 The problem of the prehistoric settlement in the survey area 184 V.3.1 The Roman Period: overall and on-site distribution (2nd-4th c AD?) 188 V.3.2 The off-site zone 214 V.3.3 Analysis of the integral network of Roman sites and land-use 218 V.4 The Late Roman Period, 4th-6th century AD 232 V.5 The Middle and Late Byzantine Period (early 10th-early 14th century) 236 V.6 The Late Byzantine-Early Ottoman Period (late 14th- early 18th century)238 V.7 The Late Ottoman-Early Modern Period 1800-1950 244 V.8 Conclusions: the history of settlement in the second survey and the wider study region 249 Chapter VI: General conclusions 280 VI.1 Patterns of habitation and land-use in the two survey areas 280 VI.2 The relation between local, micro-regional and broader regional dynamics; the importance of the geographic setting 289 Referrences 302 List of maps 319 List of tables 319 List of graphs 319 List of maps (on CD) 319 List of photos (on CD) 325 Contents of the appendices (on CD) 325 Contents of appendix I: Analysis of the survey results; the distribution of the total surface record in the first survey area 325 List of tables in Appendix I (on CD) 326 List of graphs in Appendix I (on CD) 326 4 List of photosin Appendix I (on CD) 326 List of maps in Appendix I (on CD) 327 Contents of appendix II:Surface material distribution by periods from survey area 1 328 List of tables in Appendix II (on CD) 328 List of graphs in Appendix II (on CD) 329 Contents of Appendix III: Analysis of the survey results; the distribution of the total surface record in the second survey area 330 List of tables in Appendix III (on CD) 330 List of graphs in Appendix III (on CD) 330 List of photosin Appendix III (on CD) 330 List of maps in Appendix III (on CD) 331 Contents of Appendix IV: Surface material distribution by periods from survey area 2 332 List of tables in Appendix IV (on CD) 332 List of graphs in Appendix IV (on CD) 333 Abbreviations 334 A short CV 335 Abstract 336 5 Acknowledgements Many a times I have spoken wrong when saying that this research was self-funded. It would’ve been impossible without the unselfish support of my family and friends. They invested a great deal of money, time and energy in a venture, which from the majority’s point of view was difficult to understand or simply absurd. My wholehearted indebtedness goes to my grandparents, parents and brother for their endless patience and trust and to the friends and colleagues who voluntarily decided to spend hours of hard work in the open. The basis of this work rests on their good will, love and respect. The list of names of all those who directly or indirectly supported the project over the last 5 years is too long and I have probably lost track. But I should single out the most constant members of the team, A. Ačkovski, J. Gajtanoska, P. Daljev, K.Ivanovski and D. Ristevski who taught me the basics of GIS. Finally, we mustn’t forget the hospitality and friendliness of the local people from Sopot, Kučevište and Pobužje. I would’ve still been helpless was it not for the pottery course organized by the universities of Leuven, Amsterdam and Leiden. I express my deepest gratitude to Prof. Jeroen Poblome, Prof. Vladimir Stissi and Philip Bes who introduced me to the world of ceramic studies. I also gladly thank my senior colleagues and professors in Macedonia, Trajanka Jovčeva, Prof. Dragi Mitrevski, Zlatko Videvski, Branislav Ristevski, Marina Ončevska-Todorovska whom I consulted about the dating of the fabric groups. All misinterpretations of their comments on the material go on my account. It would be equally wrong saying that this was a fully independent research project. From its very conception in 2007 up until its final completion, I was guided by the quite hand of my mentor, Prof. John Bintliff. Chapter I: Introduction, theoretical approaches and method of fieldwork 6 I.1.1Archaeological surveys in the Republic of Macedonia The appearance and the development of archaeological research and surveys in particular in the Republic of Macedonia is closely related to the main historical developments in the country: 1) the resilient survival of Ottoman rule and way of life deep into the twentieth century, the slow emergence of national identity and civic society; 2) the 20th century wars and the imperialistic attempts of the young neighboring nation states; 3) stabilization within the frame of the Yugoslav Federation, then new crisis and consolidation as an independent republic. Respectively these are the periods of the second half of the 19th and the first decade and a half of the 20th century; the first half of the 20th century and the period after the Second World War. As elsewhere in the Ottoman part of the Balkans, modern archaeological research appeared only after capitalist civic society was established, along with a nation state ready to protect it. Its development is basically parallel to the appearance and the slow establishment of modern society and state, of institutions as we know them. Changes, particularly in a field like archaeology have normally come very slow, with the incentive or the causes often lying outside the field. In view of organizational and technical advancements the development of archaeological survey went hand in hand with other forms of archaeological research. There are however certain differences when surveys and excavations are looked at separately. This is because in the country’s archaeological tradition, like elsewhere on the Balkans and in Central Europe, surveys were seen as a supplement to excavations, as a first step in the archaeological study of a landscape. Technically the history of archaeological surveys is a series of independent and poorly related, extensive survey campaigns or, at best, waves of extensive surveys, organized by smaller regions. Nonetheless the scale and character of archaeological surveys and the ways in which they are perceived and valued by the local archaeological community varied considerably between the suggested general periods of development. The factors of geography and natural conditions also bear a direct effect on the type and intensity of archaeological research. The territory of the Republic of Macedonia is situated in the southern central parts of the Balkan Peninsula, very close to the Aegean, less than 100km and the Ionian coast, less than 200km. But the country is typically continental. Influences from the sea are blocked by long mountain ranges, rising over 2 000 meters high. The Vardar-Morava Valley corridorties the territory of modern Macedonia firmly to the central parts of the Balkans. It is the back-bone of the entire region. Natural conditions are alike between the two valleys and communication is much easier. The coastal areas on the other hand are connected only by long and difficult mountain passes or narrow valleys (map I_1). 7 Map I_1:The central parts of the Balkan Peninsula. In the regional context the country is an important cross-road. It is traversed by the two shortest communication lines in the region: the one between the Aegean and the Danube (the Vardar-Morava corridor) and between the Ionian and the Aegean (the Via Egnatia). On a larger scale however, the country lies outside the main routes between Asia Minor and Central Europe, passing through Istanbul, Sofia and Belgrade. Like the other countries in the western Dinarid half of the Peninsula, Macedonia is to a certain degree isolated, protected from the violent currents coming from Asia Minor and from beyond the Lower Danube. This geographical position also explains the tendency of retardation and the resilience of certain traits or practices, throughout the entire history of these lands. A similar dichotomy can be observed within the territorial limits of the modern Macedonian state. Although measuring only about 25 000 square kilometers, there are very sharp regional differences in relief, climate, human development and population (map I_2). The parts of the country lying along the lines of the major interregional roads, the Vardar-Morava Valley line and the Thessalonica-Dhurres or the ancient Via Egnatia are the most developed sections of the country, hosting the greater part of the total archaeological research, both surveys and excavations. These are the large river valleys and plains in the central and the southwest parts of the country.The “interior”, particularly the western half of the country, but also extensive mountainous and hilly regions in the east, have been left deserted and environmentally impoverished during the 20th century. These regions along with the high mountain ranges amount to almost two-thirds of Macedonia’s territory. Relatively large areas feature only few inhabitants per square kilometer. Consequently certain parts of this land have never received archaeological studies of any kind until the present day. 8 Map I_2:The Republic of Macedonia with the main interregional corridors. The archaeological tradition in the Republic of Macedonia is short-lived, even when compared to the other countries of the central Balkans. The prevailing historical and geographic conditions have not been very inclined to the development of the archaeological discipline in this country. It emerged very late and changed very slowly. Early travelers and archaeologist from most of the old European powers worked on the territory of the Republicof Macedonia. The most lasting however was the impact of the archaeologists that worked in the early institutions founded around the mid 20th century, in the old pre-war Yugoslav Federation. Most of them were trained within the school of the Classical Archaeology tradition. The influence of this “school” continued to operate throughout the 20th century, through Macedonian archaeologists studying at the Universities of Belgrade or Zagreb and through other historical or scientific disciplines, historical geography in particular. These influences have largely determined the theoretical and the basic fieldwork methodology that marks contemporary Macedonian archaeology: archaeology as culture history, as the prevailing theoretical paradigm and an unsystematicand non-quantitative approach to the archaeological remains in the practice of fieldwork. In the last couple of decades influences from other traditions will probably incite certain changes, but in general the basic paradigm and a good deal of the fieldwork methods have remained unchanged. The culture group approach is still fully relevant for the prehistoric periods, while for the historic periods, the main concern remains the revealing of buried ancient treasures or identifying names and events mentioned in the historical records, that is archaeology in the service of history and historical geography. The archaeologists’ interests have rarely ever stretched beyond the late 9 14thcentury (the exception being scholars interested in the archaeology of churches or other public edifices). Whetherin prehistoric, classical or medieval archaeology, the site was the central term; a concept that shaped not only the way archaeological research was performed in the field, but also formulated interpretation of archaeological data and dictated solutions regarding heritage preservation and management. But the achievements of a certain tradition cannot be fairly presented if only the negative side is considered. Archaeological research in the Republic of Macedonia and surveys in particular had and still has aims and methods that made a considerable advance in the knowledge of the archaeological topography of the country, despite the very unfavorable practical and technological conditions. Particularly fruitful and interesting were the numerous debates between ancient or medieval historians and archaeologists involved in field surveys. Within half a century, a great advance was made in unraveling the geography and the history of the country during Antiquity and the Middle Age1. Extensive, unsystematic archaeological surveys were used not unsuccessfully as contributions to historical and historical geographic problems. Surveys were also used during preparatory stages, prior to excavation, both on a site and on a regional level2. Unfortunately as surveys were assigned “a secondary” role in the ideal process of field archaeological research, survey results are most of the time poorly published and very often, only unpublished reports are available. Surveys are normally considered to be a rough, exploratory and orientational method of field research. By themselves they have never been accepted as a self-sufficient or even fully relevant method of field research. In the regional archaeological tradition, this type of fieldwork is usually related to pioneering studies in unstudied areas or for inspecting the preservation state of known sites. There have been at least two, unofficially accepted explanations for the reluctance to use surveys as a method of fieldwork, clearly emanating from the traditional concept of archaeological sites. First is the suspicion about the potential of surveys to register and especially date and “culturally determine” archaeological sites. There is a very strong distrust in the coherence of the archaeological record in general and especially when dealing with disturbed surface remains. The other remark often raised against archaeological surveys is its ability to recognize lower cultural strata on multilayered sites. Surveys are therefore logically limited to the study of later architectural remains and can only record the predominant phases on multi-period sites. In general surveys were performed throughout the history of archaeological research in the country.They were used both as independent study projects and on a couple of occasions, for the preparations of archaeological atlases. In fact one type of study closely related to surveys has received a good deal of respect among domestic archaeologists over the past decades. These are regional, archaeological and historical syntheses, based on compiled information from R. Grujić, Pološko-tetovska eparhija i grad Lješek, Glasnik Skopskoga Naučnoga Društva d.n. XII, 1933; Sv. Radojčić, Menada iz Tetova, Glasnik Skopskoga Naučnoga Društva d.n. XII, 1933; N. Vulić, Teritorija rimskog Skoplja, Glasnik Skopskoga Naučnoga Društva. d.n I/1 1924; N. Vulić, Geografija Južne Srbije u antičko doba; Glasnik Skopskoga Naučnoga Društva d.n. XIX, 1938; I. Mikulčić, Nepoznat antički grad kaj s. Mojno, Godišen Zbornik na Filozofskiot Fakultet 22, 1970 ; I. Mikulčić, Topografija na Eudarist, 173-197,Macedoniae Acta Archaeologica 1, 1975; B. Josifovska, Prilog lokalizovanja grada Argosa u Peoniji, Živa Antika XV, 1964; V. Sokolovska, Stadion stone from Isar-Marvinci, Archaeologica Iugoslavica XXII-XXIII, 1982-1983; T. Tomoski, Dali postoel grad Polog,Glasnik na Institutot za Nacionalna Istorija I-1, 1959; V. Lilčić, Fauces Pelagoniae,4-53, Macedonian Heritage IV, 1997. 2 The material is compiled in the Arheološka Karta na Republika Makedonija,vol. II, Skopje 1996. 1 10 excavations, but mostly from surveys3. This type of regional study is naturally reserved for the best researched regions of the country. In most cases they cover larger micro-regions, occupying areas of several hundred square kilometers. On the other hand there are limitations regarding the type and the date of the archaeological remains: for instance, a survey of Early Christian basilicas or Prehistoric forts or inscriptions, very popular in the early days of Macedonian archaeology. In accordance with the Classical tradition, archaeological survey is primarily used to survey monumental building remains, architectural sculpture or inscriptions; those types of remains that unambiguously point to the phenomenon of site, that are “visible”, describable and conceivable as “real”, archaeological phenomena. Excavation data and stray finds are also incorporated. In regional synthesis usually the basic aim is to relate the monumental, the more attractive archaeological finds to the written sources and basically support or complement the existing historical narratives. As an independent method of fieldwork, the scope of archaeological surveys is severely underestimated. This is predictably related to the prevalent method of extensive and unsystematic reconnaissance, but far more profound is the theoretical outlook and the attitude towards archaeology in general. Issues such as size and locations of rural settlements, problems of demography and economy or human-environment interaction have only been introduced in recent decades, even in Mediterranean archaeology4.Partly because of the difficult socioeconomic conditions during the past several decades, partly because of an inherent conservativism and reluctance to face the new global trends, the development of survey archaeology is even slower in the countries of the Balkan interior, its potential to contribute to the overall archaeological knowledge being completely underestimated. I.1.2 Theoretical definitions; the study of human settlement as habitation practices; from settlements to landscapes The major incentive for the following study is the almost complete absence of evidence for settlement positioning, distribution and dynamics on regional and especially on microregional level in the country. As mentioned in the preceding section, there is a strong tradition of very precise documentation of architectural remains. There were even isolated attempts of regional, long-term syntheses. But these studies are limited by their extensive, unsystematic character to the monumental remains, to locations close to modern habitation centres and communication and to certain time-periods. Indeed the present study was preceded by a number of attempts to explore the spatial distribution of published fortified locations or other types of monumental archaeological remains, but the quality of the published data, along with the usual form of fieldwork, always thwarted these study efforts in their very early phase of assembling reasonably complete chronological maps of settlements. Fortified settlements are particularly frustrating in this respect, as they often feature more than one occupation phase and this was rarely clearly distinguishable. Visits to a number of fortified hilltops in the past couple of years showed that the majority of these sites lacked sufficient amounts of datable surface material, the 3 I. Mikulčić, Pelagonija u svetlosti arheoloških nalaza, Beograd-Skopje 1966; I. Mikulčić, Staro Skopje so okolnite tvrdini, Skopje 1982; V. Lilčić, Docnežnoantičkite tvrdini vo Tikveš i Vitačevo, 115-136, Godišen Zbornik na Filozofskiot Fakultet 41-42, 1988-89; Aleksova, B. Bargala i sredniot tek na Bregalnica, 61-71, Glasnik na Institutot za Nacionalna Istorija 3, 1983 4 I. Morris, Archaeologies of Greece, 8-48, ed. I. Morris, Classical Greece: Ancient Histories and Modern Archaeologies, Cambridge 1994. 11 building technique providing only very general chronological terms. It was thus impossible to continue any further with the present body of knowledge. Not only because there lacked the basis for the chronological determination of the fortified sites, but primarily because the location, the form and the size of human settlements in most of the archaeological periods remained completely unknown. The lack of surface material on a very large number of fortified sites signaled that these could not have been normal civic settlements; there simply lacked the usual traces of long-term human habitation on the surface within or immediately around the fortification line. Similarly for a number of periods, the monumental sepulchral or other sacral remains still await the discoveries of the settlement centres to which they belonged. There can be no doubts that the main reason behind this situation rests in the traditional technique of fieldwork and the lack of systematic and intensive archaeological survey campaigns. It is almost certain that the great majority of human settlements in the past were the open, more or less agglomerated types of settlement, preserved as vaguely discrete clusters of surface finds. The traditional method of archaeological surveys didn’t have the means to document thistype of archaeological remains.The likelihood that this category of sites will be registered with the traditional way of fieldwork is minimal. In order to advance and contribute to the knowledge of past human settlement on the territory of the Republic of Macedonia, it became necessary to apply not only different methods of field survey, but also to promote a shift in the general perspective on the problem of past human settlement. The prevailing tradition in archaeology has usually dealt with the problem of settlements on larger, interregional levels and it remained focused on the formal aspects of the settlements, seen in isolation from their immediate environmental context – the method of fieldwork determined the principle research problems. This site-centered, overall perspective have so far given very limited results concerning problems such as rural habitation practices, the dynamics of individual settlements in the long run and especially, the extent and organization of settlements at the micro-regional level. In other words, this approach has left untouched a number of very complex and important issues that have been the subject of research for decades in World Prehistoric and Classical Archaeology. We stillargue that the study of fortifications has made a great contribution not only to archaeological, but also to problems of wider socio-historical and cultural interest. However it is far from adequately addressing the problem of settlements in general and through most periods of human habitation in this land. The continuing studies of fortification plans in understudied regions and revisits of earlier documented forts has certainly more to contribute to the archaeology of the country and the wider region, but if we are to have more complete settlement maps, to answer a great number of problems of purely archaeological and wider, socio-historical significance, it is necessary to intensify archaeological surveys and shift the research focus from the conspicuous and known archaeological sites to the blank, un-researched countryside. In local archaeology as in the archaeologies of the surrounding countries, the study of the settlement remains mainly refers to the positioning, the type and rarely the size of a handful of settlements, most often the largest or at least those featuring conspicuous physical remains. Usually under the subheading of “settlements”, one finds information about vernacular architecture, household and settlement level organization, research topics traditionally related to excavations5. To be sure, the complete story of human settlement can never be told without these components of settlement life. As will be shown during the interpretation of the survey results, it is often difficult to infer conclusions about the socio-historical character of the surface remains 5 The massive and very ambitious edition entitled Praistorija Jugoslovenskih Zemalja, vol.1-5, Sarajevo 1977-1989 exemplifies this normally adopted approach. 12 when lacking information about the size of individual dwellings, the existence of subsidiary buildings, pits and other similar manifestations. This is why in the past couple of decades, regional survey projects combine intensive and systematic field survey with various methods of geophysical and geochemical prospection6 and work on improving techniques of artifact collection.7 This is finally a point where survey and excavation data can complement each other. The problem in regional archaeology however is that settlement has never been researched in the light of its spatial distribution, from a regional perspective. For example, unless there are visible, non-movable surface remains, the size of the settlement area remains unknown. Excavations, inevitably focused on a limited number of non-representative sites and further limited to unearthing very small portions of these sites, can be of little help concerning this problem. But even when it comes to vernacular architecture and inner organization, with the exception of a few larger prehistoric and antique settlements, very little has been learnt 8. This is particularly problematic for prehistoric periods, but rural settlement in Antiquity presents a no lesser mystery. However there is a necessary step further to be made. The study of settlement has so far been synchronic; like most other general categories of material culture, the study of human settlements was limited to certain time-periods. In purely theoretical terms a certain time-period is being studied, rather than a certain region or even a certain settlement. In this theoretical perspective the specific settlement and more rarely, the specific region are seen as mere data repositories, as the physical limits of a concrete study material. A certain region is surveyed or certain sites are excavated for the purpose of studying formal categories of material culture belonging to certain time-periods, not for the sake of studying the region or even the site itself. This is another fundamental difference between the traditional approach and the one advocated in the present study. If one wishes to study habitation practices in a certain area, it is inevitable that we adopt a long-term perspective. Habitation practices can never be fully understood if studied synchronically. The distribution pattern and hierarchy of settlements during certain archaeological epochs is hardly comprehensible if nothing is known about earlier and even later settlements in the same region9. In essence the difference is again related to the way in which the topic of human settlement is defined: for the traditional excavation-oriented archaeologist, settlements are categories of material culture, identified solely with the settlement proper that vary in form, type and size through different time-periods and regions. In the perspective of the 6 The Leiden-Ljubljana Ancient Cities of Beotia Projects, annual reports published in Pharos; the Nicopolis Project, A. Sarris et al, The Nicopolis Project – the integration of geophysical prospection, satellite remote sensing, and GIS techniques in the study of Epirus, Greece, Archaeometry conference, 1996; J.Bintliff, B. Davis et al, Trace metal accumulations in soils on and around ancient settlements in Greece, 9-24, ed. P. Spoerry, Geoprospection in the Archaeological Landscape, Oxford 1992. 7 P. Bes, J. Poblome, J. Bintliff, Puzzling over pottery. Thespiae, Tanagra and methodological approaches towards surface pottery, 339-345, eds. D. Malfitana, J. Poblome, J. Lund, Old pottery in a new century: Innovating perspectives on Roman Pottery Studies,Ibam, National Museum of Denmark, Leuven, Icrates2006. 8 The few exceptions are the systematic, long-term excavations on larger urban sites, J. Wiseman, et al.Studies in the Antiquities of Stobi, vol. I-III, 1973,1975, 1979; D. Koračević, Skupi- gradska teritorija, Skopje 2004, D. Mitrevski et al, Vardarski Rid vol.I, Skopje 2004. 9 Discussions in, D. R. Keller, D.W. Rupp eds. Archaeological Survey in the Meiterranean Area, Oxford 1983; J.F. Cherry, Frogs around the pond: Perspectives on current archaeological survey projects in the Mediterranean region, 383-417, the same volume; J.L. Bintliff, A.M. Snodgrass, The Cambridge/Bradford Boeotian Expedition: The first four years, 123-161, Journal of Field Archaeology 12, 1985; G. Barker, Approaches to Mediterranean Landscape History, 1-16; ed. G. Barker, A MediterraneanValley: Landscape archaeology and Annales History in the BifernoValley,Leicester 1995. 13 now predominant strand of LandscapeArchaeology in the Mediterranean, the topic of human settlements has a much broader meaning. It refers neither to individual settlements nor to certain formal categories of settlements, but to human settlement as a continuous and dynamic long-term process; a theoretical shift that goes back to the 1960’s and the emergence of the New or Processual Archaeology10. Hence the turn towards the long-term regional studies; if human settlement is conceived of in broader anthropological terms, the long-term regional approach is an inevitable theoretical implication. In this perspective the human settlement in a certain region is the history of habitation practices and strategies in the long run, not a dis/continuous sequence of styles of vernacular or defensive architecture. The distribution of settlements during a certain period is to a large degree predetermined by the situation in the preceding periods and preconditions the settlement pattern of subsequent periods. The currently prevailing theoretical approach in Mediterranean settlement archaeology has broadened the concept of settlement to include elements of settlement practices other than the settlement proper or the various military installations. The continuously perfected method of intensive, systematic surveys allowed archaeologists to study a wider range of smaller or seasonal features of settlement practices. After almost four decades of experience, practitioners of this method of field survey are able to recognize a series of anthropogenic installations that form an inextricable part of the human settlement, in most historical and geographic conditions. Open settlements of minor size, groups of hamlets or individual farmsteads, rural shrines and cemeteries are now regularly appearing on reconstructed settlement maps 11. Non-residential, less intensive habitation practices (ancient zones of intensive agriculture, industrial areas) are also documented through the study of off-site scatters12. As mentioned earlier, these and similar categories could hardly be recognized by the traditional method of field survey and even if they were accidentally discovered, there simply lacked an adequate documentation technique. Consequently the new method of fieldwork offers a far more complete picture of human habitation practices.The maps of settlements are no longer simply indicating locations of major habitation centres in certain time periods. The systematic quantification of surface finds can produce surprisingly detailed and nuanced reconstructions of population distribution, productive and religious foci, landscape modifications and so forth. These elements of habitation practices are of no lesser importance than the interior elements of the settlements proper. Researching these “secondary” features, modern settlement archaeology makes a valuable contribution to the understanding of past agricultural economy, demography and landscape planning. It is by these means that human settlement is studied as a long-term, anthropogenic process. Inseparably connected to the intensive survey projects in the Mediterranean and to the general shifts in the theoretical approach was the increasing interest and appreciation of the past 10 L.R. Binford, Archaeology as Anthropology, 217-225, American Antiquity 28-2 1962,; Ibid, A Consideration of Archaeological Research Design, 425-441,American Antiquity 29-4 1964,; T.K. Earle, and R. Preucel, Processual Archaeology and the radical critique, 501-512, Current Anthropology 28-4 1987; B. Trigger, A History of Archaeological Thought, Cambridge 1989. 11 J.L. Bintliff, P. Howard, Studying needles in haystacks-Surface Survey and Rural Landscapes of Central Greece in Roman Times, 51-91, Pharos 1999; S.E. Alcock and, J.E. Rempel, The More Unusual dots on the map: Special purpose sites and the texture of landscape, 27-46, eds. P. Guldager Bilde, V.F. Stolba, Surveying the Greek Chora: The Black Sea Region in a Comparative Perspective, Aarhus 2006. For concrete examples, M.H. Jameson, et al, A Greek Countryside: The southern Argolide from Prehistory to present day, Stanford 1994, tab. 4.5. 12 J. Bintliff, A. Snodgrass, Off-site pottery distributions: A regional and interregional perspective, 506513, Current Anthropology 29-3, 1988; T.J. Wilkinson, Extensive shard scatters and land-use intensity: some recent results, 31-46 Journal of Field Archaeology 16-1, 1989. 14 environmental context. The study of cultural change and stability was closely related to environmental factors since the early days of Processual Archaeology and the dynamics of settlement pattern was the obvious candidate topic to test the newly discovered interdisciplinary field of research. The significance of the changing environmental conditions needs little elaboration. The possible impacts on human settlement and economy, as well as on the formation of the surface archaeological record were realized immediately and human-environment interaction ranked high in the agenda of all major research projects influenced by the new tendencies in Anglo-American archaeology. The human-environment relations present a very wide and complex research topic, involving the expertise of a number of different disciplines, most prominently, geomorphology, soil science and palynology. The fruitful cooperation between these various disciplines was best reflected in the large, interdisciplinary regional survey projects in the Mediterranean13. Even the earliest of these studies already included geomorphologic surveys, soil and vegetation mapping and often, coring for pollen samples. The aim of these large research projects was a complete environmental and landscape reconstruction. The study of settlement and environmental change went not only hand in hand, but were deeply interwoven; changes in one of the spheres were regularly related to changes in the other. Implications on a theoretical plan were once more unavoidable. Even without involving the study of past environment, there was already a fundamental shift in research strategy. Actually the abovementioned shift in the research focus, from individual sites and certain timeperiods to the region as a whole is two times underlined. First, by redefining the study of settlements in archaeology, the study of settlement as a continuous, long-term process and secondly, by broadening the concept of settlement with an array of features and activities for which traditional survey archaeology lacked the appropriate methods of field study. Human settlement is not simply an agglomeration of houses with their defenses and communications; it is also seasonal and auxiliary establishments, satellite settlements, water supply, agricultural fields and terraces, ritual locations and cemeteries etc. This vast range of human activities can only be studied on a supra-settlement level, by looking at a region or a micro-region in its entirety. It is very logical then to see landscape archaeologists joining their forces with natural scientists in an effort to understand the dynamics of human-environment interaction. The study of human settlement as defined by this approach can never be complete without the integration of environmental data.In fact the correct interpretation of the data gathered by intensive field surveys is itself greatly dependent on the understanding of past and present sedimentation and erosion processes. The departure from the traditional approach of studying settlements as discontinuous and formal categories of material culture, unstoppably led the way to the present developments in the field of Landscape Archaeology. These were the basic changes in the theoretical premises; the refocusing of the study interest from specific settlement centres to the process of settlement, to the long-term habitation practices taking place on a supra-settlement, regional or micro-regional level. Among the implications of this fundamental change, there were also realignments in the traditional relations with history and geography. Certain schools of thought within these traditional disciplines have developed advanced and elaborate methods and theories by the mid 20th century that became 13 For instance the Pylos Regional Archaeological Project, E. Zangger et al. The Pylos Regional Archaeological Project, part II: Landscape Evolution and Site Preservation, 549-641, Hesperia66-4, 1997; M.H. Jameson, C.N. Runnels, T.H. van Andel , 173-194, 1994. In fact, environmental reconstruction was also one of the goals of the Messenia survey project, G.R. Rapp, A.W. McDonald eds. The MinnesotaMessenia Expedition: Reconstructing a Bronze Age Regional Environment, Minneapolis 1972. 15 very influential among the new settlement and landscape archaeologists. The influences of the French, Annales School or the New Geography movement was actually far more significant than the newly introduced theoretical concepts of Processual Archaeology14 (although we stress that in general the split between the old and the new approaches in settlement archaeology can be traced along the same axis that divided Old and New Archaeology in the 1960’s). The overall impression is that the theoretical concerns of the “new wave” of surveys remained to a certain degree unaffected by the later Processual - Postprocessual debate. Other mostly nonarchaeological traditions have been for a century busy theorizing the landscape as a research subject or elaborating models for intra and interregional interaction. Their efforts were far more relevant to the newly discovered field of archaeological research and the models offered proved practically applicable for the interpretation of survey archaeological data. Particularly promising were F. Braudel’s concept of the tri-fold structure of historical change (for examining settlement dynamics in the long run) and I. Wallerstein’s core-periphery model (for understanding transregional and global developments)15. Predecessors and inspiration was also found in the work of the German Anthropogeographic tradition, the Landeskunde, very influential throughout continental Europe in the early twentieth century16. Somewhat less prominent are the applications of ecological, Darwinian or Malthusian models and the spatial analysis models, borrowed from the New Geography17. Finally, the Postprocessual critique during its apex in the late 1980’s-early 1990’s attempted to build its own approach to the problem of settlement and landscape archaeology, insisting on the study of ideational landscapes or authentic landscape perceptions.18 Apart from problems and concepts borrowed from these major theoretical traditions, the “new wave” of regional survey archaeology was generating its own set of theoretical and methodological topics. The very practice of the method of intensive, systematic survey was delineating the range of questions that could be adequately addressed on the base of the gathered set of data. First and foremost was the question of defining known or hypothetical forms of cultural/human activity on the basis of the field records. Traditional survey and excavation archaeology operated with a universal set of terms designating the usual manifestations in the archaeological record revealed by these fieldwork techniques. They mostly borrowed from standard architectural terminology or referred to past socio-historic phenomena: the various 14 D. Clarke ed. Spatial archaeology, New York 1977; J.L. Bintliff and C.F. Gaffney, eds. Archaeology at the interface; Studies in archaeology’s relationships with history, geography, biology and the physical sciences, Oxford 1986; T.K. Earle and R. Preucel, 501-512, 1987; J.L. Bintliff, ed. The AnnalesSchool and Archaeology, Leicester 1991; A.B. Knapp, ed. Archaeology, Annales and Ethnohistory, Cambridge 1992; G. Barker ed. 1995. 15 C.K. Chase-Dunn, T.D. Hall, Rise and Demise: Comparing World – Systems, Boulder 1997; E.M. Shortman, A.Urban, Resources, power and interregional interaction, Springer1992. For archaeological applications, A. B. Knapp, Independence and Imperialism: politico-economic structures in the Bronze Age Levant, 83-98, A.B. Knapp, ed, 1992; J. L. Bintliff, Regional Survey, Demography, and the Rise of Complex Societies in the Ancient Aegean: Core-Periphery, Neo-Malthusian, and other Interpretive Models, 1-38,Journal of Field Archaeology 24 1997. 16 J.L. Bintliff, History and Continental Approaches, 147-164, eds. R. A. Bentley, H. D. Maschner, Handbook of Archaeological Theories, Altamira Press 2008. Closely related to this approach are a group of Czech Archaeologists, E. Neustupny, ed. Space in Prehistoric Bohemia,Prague 1998; M. Godja, ed. Archaeology, Ancient Landscape, Settlement dynamics and Non-destructive Archaeology, 2004. 17 C. Orton, I. Hodder, Spatialanalysis in archaeology, Cambridge 1976, K. Flannery ed, The Early Mesoamerican village, New York 1976; J.L. Bintliff, 23-28, 1997. 18 C. Tilley, A phenomenology of Landscape, Oxford 1994, B. Bender ed. Landscapes, politics and perspectives, Berg 1993. 16 categories of architectural remains or the levels of urban settlement hierarchy. Standard architectural terms such as walls, floors, basilicas or terms such as a polis, a roman villa and a Medieval castle have a centuries-long usage in archaeology. But the majority of these apply to monumental, architectural remains, immediately recognized as known architectural or sociohistorical phenomena. There lacked criteria for interpreting sites without visible architectural remains.The occasionally registered clusters of surface finds were roughly designated as small, open types of settlement. As explained in preceding paragraphs, traditional survey archaeology recognized and documented manifestations in the archaeological record as perceivable, isolated phenomena. Only with the application of intensive, systematic surveys did there arise the need to define quantitative criteria for site definition and categorization, always specific to the surveyed region. The task is far from simple, as sheer quantity and extent are not always direct indicators of the type and intensity of past human activity19. Not only because traces of human activities vary across and within different time-periods, but also because surface remains are further transformed under the work of the various post depositional factors. Moreover since the study focus is now on the entire surveyed area and since most of it is continuously covered with a carpet of broken pottery, it becomes also necessary to find criteria to distinguish between the site and the off-site; or inversely, to interpret the distribution of surface material outside the limits of the traditional categories of sites-settlements, cemeteries, farms etc20. These are clearly problems of interpretation that can only be adequately approached through accumulation and careful study of sufficient field data, separately for each studied region. Another set of hotly debated issues, intrinsic to the methodology of regional, intensive surveys was the reliability and congruence of the surface data and consequently, the very limits of interpretation based on this type of data21. The very detailed and controlled surface coverage is the key advantage of intensive surveys; it is an attempt to register and map the smallest traces of past human activities in the studied area. But even if coverage is 100%, the map of past habitation practices is neither complete, nor equally representative for each period of human settlement in the studied area. The problem is two-fold: Is the surface record truly reflecting the amount and character of the buried material and how much of the perished landscapes has actually survived in the surface archaeological record22? These questions must be considered seriously before the final word is said on the history of settlement based on the surface record. Absence of certain periods or certain categories of sites can equally be explained as the result of destructive erosive forces, low visibility or as evidence of low population density and sparse permanent settlement. The problem is hardly trivial, because the preservation and visibility of the surface record vary not only region-wise, but also vary for chronologically and typologically different categories of surface artifacts. Thus the rate of preservation of surface remains is 19 J. L. Bintliff et al. Classical farms, hidden prehistoric landscapes and Greek rural survey: a response and an update, 259-265, Journal of Mediterranean Archaeology 15, 2002; J. L. Bintliff, P. Howard, 57-67, 1999. 20 J.F. Cherry, et al, Archaeological survey in an artifact rich landscape: a Middle Neolithic example from Nemea, 159-176, American Journal of Archaeology 92-2, 1988; J.; S. E. Alcock, J.F. Cherry, J.L. Davis, Intensive survey, Agricultural practice, and the Classical Landscape of Greece, 135-168, ed. I. Morris, Classical Greece: Ancient Histories and modern Archaeologies, Cambridge 1994; A.M. Snodgrass, The Archaeological aspect, 197200, ed. I. Morris, 1994; J. L. Bintliff, The concepts of 'site' and 'offsite' archaeology in surface artifacts survey, 200215, eds. M. Pasquinucci, F. Trément Non-Destructive Techniques Applied to Landscape Archaeology, Oxford 2000. 21 J.F. Cherry, 483-489, eds. D. Rupp and D.W. Keller, 1983, contra R. Hope-Simpson, Mycenaean Greece, Noyes Publishers 1981, for instance. 22 J. Bintliff, P. Howard, A. Snodgrass, The hidden landscape of prehistoric Greece, 139-168, Journal of Mediterranean Archaeology 12-2, 1999. 17 naturally lower as their age increases, simply because of the accumulative effect of postdepositional processes through longer periods of time23. However certain categories of surface material and in some cases, remains from entire archaeological periods are poorly represented in the surface record because of poor physical qualities or equally possible, due to decreased usage and discard rate24. It is thus rather difficult, if not impossible to project a general loss-rate for all archaeological periods. Each period is represented in the surface record to a degree limited by the intensity and longevity of habitation and the durability of material culture among other factors. The obvious problem is that these period and region-specific factors can only be known through the study of the truncated surface record; the above-posed questions can only be addressed through a range of empirical observations: On relations between visibility conditions, land usage (both seasonal and short-term) and the variations in the amount of documented material, on relations between the type and amounts of surface and sub-surface material, relations between conventional and other forms of traces of human presence and activity (element traces in soils, geo-magnetism )25. It is only certain that these sets of relations are highly variable and that consequently, there is always a gap of unknown expanse, separating the visible (surface) record and the (always presumed) original amount of discarded artifacts. It has to be underlined though that the latter is always in the hypothetical realm, regardless of whether one predicts the height of a building on the basis of the quantity of building rubble or extrapolates population levels on the basis of the amount and distribution of surface finds. In both cases the pre-depositional context is not lying somewhere hidden, awaiting its discovery but is gone for good. The uncertainties of field data are not the sole property of intensive surveys. Excavation data are equally incomplete and difficult to interpret, (especially when the very small, unrepresentative sample of excavated data is considered) and yet, archaeologists do not seem to have refrained from far-reaching inferences in its interpretation26. Later in this chapter, under the subheading of field method, I’ll return in a greater detail to this issue. For now it suffices to remark that the method of intensive, regional surveys has opened up a new dimension in the study of human settlement.The detailed data it offers may not be “complete”, but they certainly offer a better, far more accurate picture of past human settlement than the one offered by the centuries of extensive, non-systematic surveys. Even after a generation of scholars employing this technique and interpreting its results, there still remains much to be learned about the potential and limits of intensive survey data. Perhaps the best, the most logical way of delineating the legitimate limits of a certain method of research is through continual practice and experiment. 23 J.L. Bintliff, A.M. Snodgrass, 137-138, 1985; J. F. Cherry et al, Landscape archaeology as long-term history: northern Keos in the Cycladic Islands from earliest settlement until modern times, Los Angeles 1991. 24 J.L. Bintliff, E. Farinetti, Landscape and Early Farming: Settlement Dynamics in Central Greece, 665674, Geoarchaeology 21 2006; M. Millett, Pottery: population or supply pattern? The Ager Tarraconensis approach, 18-26, eds. G. Barker, J. Lloyd, Roman Landscapes: Archaeological surveys in the Mediterranean Region, London 1991. 25 W.G. Cavanagh, S. Hirst, C. Litton,Soil Phosphate, Site Boundaries and Change Point Analysis, 67-83, Journal of Field Archaeology 15 1988; G. Barker, 51-54, ed. G. Barker, 1994; J. L. Bintliff, E. Farinetti, et al. The Tanagra project: investigations at an ancient city and its countryside (2000-2002), 541-606, Bulletin de Correspondence Hellènique 2004-5;W. Cavanagh, C. Mee and P. James, et al. The Laconia Rural Sites Project, Athens 2005. 26 J.F. Cherry, 382-385, eds. D. Keller, D.W. Rupp, 1983; A.M. Snodgrass, J.F. Cherry, On not digging up the past, 9-13, The Cambridge Review 109, 1988. 18 I.1.3 The principle aims of the survey, definition of the survey areas and a couple of oldnew concepts It seemed appropriate to insert the present study into a wider historical and theoretical context; to consider the situation with surveys and settlement studies in regional archaeology, as well as the context of contemporary tendencies in Mediterranean and world survey archaeology. Further references will be made throughout the text, as the present research is modeled after the regional survey projects in the Mediterranean. Limitations regarding financial means and expertise are naturally reflected in the very small scale of the survey, as in the limited focus of field research. It is therefore necessary to define research goals that can be realistically pursued with the present means. Some of these limitations require modifications and adjustments in the basic theoretical premises and particularly, in the practice of fieldwork. The small size of the survey team, consisting of not more than 4 or 5 persons in the field, significantly narrowed the extent of the survey area. With a team of maximum five inexperienced field walkers it was clearly impossible to survey more than 1 to 1.5 square kilometers per year. This meant that the idea of surveying representative samples of certain types of terrains was out of the question27. The only option to obtain meaningful and interpretable results from an area of such a miniature scale was to ensure that the limits of the survey correspond to the narrower territory of at least a single settlement – a Siedlungskammer or a settlement niche in the terminology of early 20th century Anthropogeography28. A settlement niche is a geographic concept, particularly applicable to the circumstances of the fragmented discontinuous reliefs, very common for the lands surrounding the Mediterranean Basin. It refers to the settlement and its immediate physical surroundings, the parish or the village area in administrative terms and thus, it is to a certain degree equivalent to the more widespread concept of catchments. It is a simple but fascinating theoretical connection between the settlements’ areas and the hydro and orographical units: the small lateral valleys, the little plateaus, the gentle mountain slopes, all present natural, pre-given human habitats, often displaying very clear topographic limits. Unlike the geometric methods employed by the New Geography and later by a certain number of archaeologists29, the concept of the Siedlungskammer has the advantage of recognizing the natural divisions of the terrain. The ingeniousness of this concept was confirmed through simple observations on the locations of still standing, Late Ottoman-Early Modern villages in the geography of the Republic of Macedonia. The territories of most villages, at least those situated along the Middle Vardar Valley occupied the drainages of minor streams, flowing directly into the Vardar. In principle every stream featuring a more developed relief hosted a village of a minor or a medium size (mapI_3). 27 Sampling has figured high on the agenda in almost all regional survey project, for obvious reasons, M.B. Schiffer, A.P. Sullivan, T.C. Klinger, The design of archaeological surveys, 1-28, World Archaeology, 10-1, 1978; J.F. Cherry, A Preliminary definition on site distribution on Melos, 16-23, eds. C. Renfrew, M. Wagstaff, An Island Polity: the archaeology of exploitation on Melos, Cambridge 1982; D. R. Keller, D.W. Rupp eds. 1983; J.L. Bintliff, A.M. Snodgrass, 129-130, 1985. 28 J.L. Bintliff, 158-159, eds. R. A. Bentley & H. D. Maschner, 2008; for similar, though not identical concepts, E. Neustupný, 45-61, ed. E. Neustupný, 1998; M. Kuna, Surface artifact studies in the Czech Republic 2944, eds. J. Bintliff, M. Kuna, N. Venclová, The future of archaeological field survey in Europe, Sheffield2000. For criticism see R.E. Blanton, Mediterranean myopia, 627-629, Antiquity 75, 2001. 29 C. Vita-Finzi and E.S. Higgs, Prehistoric economy in the Mount Carmel area of Palestine: Site catchment analysis, 1-37, Proceedings of the Prehistoric Society 36, 1970; K.V. Flannery ed, 1976; summarized in B. Erdoğu, The Prehistoric Settlement of Eastern Thrace, Oxford 2005. 19 Map I_3: Section of the mid-Vardar Valley showing borders of topographic entities, red line and 1.5 km buffers, black circles, around existing settlements. Their territories clearly correspond to the micro-hydrographic entities.Communication with the neighboring villages, not more than 4-5 kilometers away is possible via a limited number of points, usually mountain passes and small valley floors. It is thus possible to identify with a great ease the basic settlement units in the region of the Middle Vardar. As these are primarily natural micro-regional units, it seemed reasonable to believe that human settlement followed these frames throughout its history in this region. It was therefore decided to set the limits of the first survey area within the limits of one of these natural settlement chambers, hoping to discover at least some degree of settlement continuity. However the identified settlement chambers usually extend over an area of several or a dozen square kilometers, if the natural limits of the watershed are followed. As the entire village hinterland was too large for a complete, intensive coverage, it was necessary to further narrow the survey area, within the limits of the settlement chamber. This proved to be a minor difficulty, as only a smaller percentage of the village areas in the hilly country of the Middle Vardar presented flat or gently sloping terrains; the rest were mainly steep slopes or bare rocky ridges, where systematic field walking was technically impossible. It was further possible to define elements in these microregions that were most likely to host traces of past human settlement. When the territories of settlement chambers from a certain micro-region are looked at in greater details and compared, it 20 becomes possible to observe a number of repetitive components such as a central surface (hosting the settlement and the bulk of arable land), higher plateaus (usually, pastures and cultivated fields, but also locations of alternative or satellite settlement), the valley floor (gardens, watermills, but also communication and defenses), the dominant hilltops (military and religious installations) and so forth. All permanently inhabited settlement chambers in the Middle Vardar Valley display some sort of inner differentiation related to topographic, functional and symbolic divisions. Of course it is wrong to ignore the possibility of changes in this inner organization during the various periods of human settlement (for example, what one sees at present is clearly a deformed and contracted land-usage pattern dating to Late OttomanEarly Modern times, the late 18th - early 20th century), but as these divisions are primarily of a topographic nature, it seems reasonable enough to expect a good deal of landscape and land-use stability, naturally leaving aside the major geomorphologic transformations. The idea is that within the usual territory of 10 to 15 square kilometers it is possible to distinguish between “central” and “marginal” surfaces and locations, on the basis of the physical configuration of the area and on the basis of the present day land-use. One of the principle aims of the present study will be to explore the distribution of surface archaeological finds in relation to the observed – “anthropo-topographic” fragmentation of the surveyed area. The changing locations of the focus of human habitation within the narrower settlement chamber will hopefully aid in the understanding of the inner dynamics of the settlement chamber: the changing of settlement and production foci, the changing importance of the local communication lines, even the changing limits of the entire settlement chamber30. Much is expected at least for later historical periods, the Late Roman through the Late Ottoman, from which more substantial remains can be observed on the surface. In this context it is a particular handicap that there lack published micro and messo-regional geomorphologic studies in the country.There is little interest and information about the potential of interdisciplinary regional studies, both among local archaeologists and environmental scientists. Geomorphologic and other environmental data are equally crucial for understanding the dynamics on a supra-settlement level and of the taphonomic processes in the surveyed areas. In the present study the local geo-pedological layer can barely be tied within the larger regional series and facies. It should be noted however that even on large-scale geologic maps, the correlations between the geo-pedologic layer and the actual land-use patterns are striking. In a way the proposed approach goes a step further in theorizing the area of the settlement chamber (and landscape in general) as a fragmented, composite, but relatively integrated entity. In this perspective the territories of settlement chambers consist of a number of recurrent anthropo-topographic components: the settlement proper nested into certain physical features, the narrow valley floor with its small gardens and local pathways, the flat foothills and plateaus hosting the arable fields, to number only the most apparent ones. An anthropo-topographic component is a relief feature transformed and modified through a long-term, repetitive usage.It is assumed that the observed divisions of the terrain are primarily a response to the natural, topographic configuration. Just as a certain region is broken up into a number of naturally defined settlement niches, so the inner topographic fragmentation of the settlement chambers 30 Cf. J.L. Bintliff, Archaeological survey of the Valley of the Muses and its significance for Beotian history, 193-224, eds. A. Hurst, A Schachter, La montagne des Muses, Geneva 1996. 21 antecedes and determines the anthropogenic divisions. As in so many other aspects nature offers archetypes for material culture forms and human behavior in general31. Adjusting to the limited means and potential of the present research, it was decided to survey and compare human settlement in two settlement chambers. We hoped to see if certain regularities can be observed in the long-term settlement dynamic. To ensure that there will be sufficient evidence on the surface, the limits of the first study area were drawn along the limits of the narrower territory of a still standing village, whose existence can be traced back into the Early Ottoman (and most likely, the Late Byzantine?) Period on the basis of the written evidence. The tactics gave results, thanks to the facts that in the first study area settlement dynamics comforted to the so called, restrictive mobility model: in most periods of human habitation history, the settlement proper was somewhere within a radius of one kilometer or less32. There were however considerable problems with this choice, which became apparent at the very beginning of fieldwork (the extreme variations in ground conditions between neighbouring fields) and especially later, during the analysis of the results (much of the surface material distribution had to be attributed to modern anthropogenic activities, while the surface was covered with a considerable layer of modern and Early Modern artifacts). Therefore a different tactics was adopted in deciding on the location and the limits of the second survey area. Bearing in mind the settlement chamber concept and the peculiar overlapping between the basic settlement units and the topographic fragmentation on a regional level, it is possible to identify potential, but vacant settlement chambers within the limits of the general region and avoid the background noise created by present-day human activities. Obviously this approach is not without pre-assumptions. How to be sure that a certain valley is a potential settlement chamber? The only clue is offered by the size and the degree of complexity of the stream. In reality however it is impossible to put all hydrographic units in one of the two categories of spacious, developed valleys and small, inarticulate streams. Hydrographic units range from bare fissures, a couple of hundred meters long, to deep inaccessible ravines, to complex little valleys with terraces and even minor tributary streams. Moreover the fairly isolated valley presents the basic micro-regional unit and basic settlement niche only in certain regions of the country, most prominently in the Mid-Vardar Valley. In other topographic settings, such as the extensive mountain plateaus or the large basins, there lack a pronounced hydrographical fragmentation of the region. These types of regions are usually drained by a single larger river, fed by small streams with undeveloped hydrography. As will be shown in these geographic conditions other types of topographic units take over the role of settlement loci. Various orographic units, river or lake terraces become the focus of human settlement. Although these micro-topographic units also feature more or less clear boundaries, they lack the physical integrity and the size of the “normal” settlement chamber, the small, isolated valley. Theoretically they correspond to the micro-topographic units that constitute the drainage basins along the Mid-Vardar. The micro-topographic units that constitute a larger plain or mountain plateau can accommodate a small or medium sized nucleated settlement, along with the fields and gardens in its immediate surroundings, but the full extent of the settlement’s territory had to extend beyond the faint, micro-topographic boundaries. This circumstance has important implications, because the territorial boundaries between two neighbouring settlements 31 A very similar theoretical approach has been explicated by J. Benesh, M. Zvelebil, A historical interactive landscape in the heart of Europe: the case of Bohemia, 73-93, ed. P. Ucko, R. Layton, The archaeology and anthropology of landscape, New York1994. 32 B. Erdogu, 33-34, 2005. 22 cannot be drawn along the lines of natural, geographic divisions. Admittedly it is always possible to locate faint topographic barriers, often reinforced by anthropogenic alterations, such as marker stones or temples. However apart from the natural, physical barriers there are other constraining factors that determine the extent of settlement territories in regions lacking pronounced fragmentation into separate micro-regional units. It is these factors that are largely accounted for by concepts borrowed from the field of New Geography, the site-catchment analysis being amongst the most relevant for the present study33. The other significant consequence springing from the absence of clear topographic barriers (which appears to be irrelevant for the wider regional context of the first survey) is the high integrity of the settlement pattern in larger regional units. A change in the location of one major settlement will inevitably have effect on all other settlements sharing the same geographical unit. On the other hand in the well-defined, micro-regional entities along the Mid-Vardar Valley, it seems that the settlements’ locations could shift within the limits of a settlement’s chamber with no apparent effect on settlement location in the neighbouring chambers. We’ll return to these issues in the concluding chapter of this study. For the moment, it suffices to admit that largely unaware of the importance of some of the abovementioned concepts it was decided to situate the second survey area in a geographic context different from the one surrounding the first survey. Survey area number two was going to be a potential, but presently vacant settlement locus, situated on a larger Tertiary plain. In order to ensure comparability, it will be of a roughly equal size as the first survey area. The decision to conduct the second survey in a different geographic setting was instigated by two principle reasons. Firstly, these two smallscale survey projects introduced a new method of field survey for the regions along the Vardar Valley; hence inevitably one of the basic goals of the study was to explore their potential in geographic conditions different than those prevailing in the narrower Mediterranean belt. Knowing next to nothing about settlement patterns on a micro-regional scale in this part of the Balkan Peninsula, it seemed appropriate to focus the survey on areas featuring different environmental conditions. Secondly, we wanted to test the importance of micro-regional barriers in dictating the distribution of settlement, the significance of the concept of natural settlement chambers. For this purpose the first survey was going to cover a well-defined micro-region, while the second survey area, although coinciding with a separate micro-topographic entity will feature less accentuated topographic limits and consequently, a lesser degree of regional integrity. The two survey areas are also supposed to differ in another respect. Study area number one is situated in an economically passive and administratively and politically, marginal region. The present day village of Sopot is situated in a region that lacked a major political, tribal or urban centre throughout known history. Even at present it stands on the very border of two, major administrative units, the territories of the cities of Skopje on the northwest and Veles on the south34. In contrast the second survey area is going to be chosen from the vacant settlement chambers situated in the immediate to medium vicinity of a known, major settlement and political centre, most likely near the Roman colony Scupi, in the Skopje Basin. The aim is to 33 C. Vita-Finzi, E. Higgs, 1-37, 1970; K. Flannery, Empirical determination of Site Catchments in Oaxaca and Tehuacán, 103-116, ed. K. Flannery, 1976; J.L. Bintliff, Territorial behavior and the natural history of the Greek polis, 207-249, eds. E. Olshausen, H. Sonnabend, Stuttgarter Kolloquium zur Historischen Geographie des Altertums, Amsterdam1994. 34 A close parallel is to be sought in the NemeaValley, on the Peloponnesus, J. Wright et al, The Nemea Valley Archaeological Project, a preliminary report, 579-659, Hesperia 59-3, 1990. 23 attempt an examination of the impact of the vicinity of major political and economic centres on settlement on the micro-regional level. Among other topics, the impact of the changing political and socio-economic conditions on human settlement is surely one of the most challenging subjects of regional survey studies 35. In essence it ought to be linked to the general problem of scale and reliability of intensive survey data. Over the past few decades it’s been acknowledged that only through synthesis of data from several regions can intensive regional surveys hope to contribute to issues such as colonization, regional or trans-regional migration and deliberate demographic policies in general 36. The last two decades have seen a growing concern of producing comparable, standardized data in intensive regional survey projects. A number of very successful attempts at synthesis of regional data confirmed the great potential of this approach, especially for addressing issues of wider socio-historical significance37. Does this automatically mean that by focusing on micro-regions, one is inevitably hampered in inferring conditions pertaining to the distant, socio-historical reality38? Surely the nameless local situations are not the best place to look for the agency of known historical factors and policies. It suffices to mention the problem of equifinality: observed changes in the extent and distribution of human habitation practices can equally be the result of deliberate and planned political acts or a consequence of some unknown global demographic tendencies or climatic changes or the initiative of the local community. (Needless to stress, ignorance on the subject of environmental history is again detrimental to the final analysis.) Focusing on certain micro-regions, it is impossible to reveal even a segment of the settlement hierarchy and the local dynamic can never be unambiguously related to external factors. Evidently the questions of the effect of major socio-economic and demographic processes and events on micro-regional local trends are chiefly beyond the scope of this study. In the end however, it will be necessary to try and insert the local developments into a broader regional context. It has to be remembered that some of these issues have been studied by survey archaeologists working in this region. As explained in the previous section, much of the country’s ancient topography has been illuminated. The major urban centres of Antiquity and the Middle Ages are located and many have been subjected to decades-long systematic excavations. This will allow us to position the survey areas in relation to known larger settlements, political and administrative boundaries and major communication lines. In many regions of the country, especially those along the Vardar Valley, major archaeological monuments (fortifications, prehistoric mounds) have long since been put on the archaeological map. Finally, during the last several decades, modern construction and mechanized ploughing has brought to light a large number of open settlements or necropoleis from various time-periods. Although only a very small segment of this archaeological evidence has received proper publication and analysis, it will hopefully help establish at least a vague connection between the local and regional 35 C. Renfrew, M. Wagstaff, eds. An Island Polity: The Archaeology of Exploitation on Melos, Cambridge 1982; A. B. Knapp ed. 78-93, 1992; S.E. Alcock, Graecia Capta: The Landscapes of Roman Greece, Cambridge 1993; J.L. Bintliff, 1-38, 1997; J. Wiseman, K. Zachos ed, Landscape Archaeology in southern Epirus, Greece, vol. I, Hesperia Supplements 2003. 36 P. Attema, Two challenges for Landscape Archaeology, 18-26, eds. P. Attema et al, New developments in Italian Landscape Archaeology, Oxford 2002; S.E. Alcock, J.F. Cherry eds. Side-by-side survey: comparative regional studies in the Mediterranean World, Oxford 2004. 37 For example the Nemea Valley Survey, J. Wright et al. 579-675, 1990; S.E. Alcock, Archaeology and imperialism: Roman Expansion and the Greek City, 87-135, Journal of Mediterranean Archaeology 2, 1989; J.L. Bintliff, 1997. 38 R.E. Blanton, 629, 2001. 24 developments. As pointed out, it is at the micro-region level where local archaeology is most deficient in systematically gathered evidence. At a certain point of time it will be necessary to shift the research focus on the lowest tier of settlement hierarchy. The present research is barely beginning to fill-in this gap in the archaeological atlas of the Republic of Macedonia. In the end, studying developments on a local level is in itself a legitimate subject of research. This is especially true in a situation where literally nothing is known about the basic forms of human settlement in the countryside, the small village or hamlet. Even when observed in isolation from the broader regional context, the micro-location and the size of a settlement reflect strategies of adaptations to certain types of environments, whose relevance exceeds the micro-regional limits. Preference for certain topographic units, dispersion or nucleation and the settlement size are closely related to local economic and social organization. Not being limited to a specific time-period, the study will attempt to reveal the entire history of human settlement in the survey areas. Thus local processes of displacement, contraction and expansion can be observed in the long-run, hopefully underlying persistent tendencies and limitations or repetitive cycles of dispersion and nucleation, growth and decline39. In this perspective, local long-term trends are of no lesser significance than regional or supra-regional developments. These broad theoretical positions have set the basic course of the study. In their greater part, they are borrowed from the predominant theoretical strand in Mediterranean Landscape Archaeology and ultimately, from other related disciplines. It is now time to turn to the methodology of field work. Naturally the study presented here strives to follow, at least in their basic principles, the current regional surveys in the Mediterranean area. But even here one can predict a number of inevitable modifications, not simply because of the small-scale of the present project, but because of the specific conditions in the land of the Mid-Vardar. There are a number of difficult problems associated either with the planning and execution of the survey or with the delicate analysis of the results. It is on this point that the richness and complexity of the survey data become truly apparent. This is the realm where survey archaeology has a plenty of room to experiment, suggest its own interpretative models and build a genuine theory. Conversely it is also the stage where this method and approach reveals its deepest weaknesses and is at risk of producing mistaken, incongruent data and of giving misleading, essentially incorrect interpretations. 39 J.L. Bintliff, ed.1992; A.B. Knapp, 83-98, ed. A.B. Knapp, 1993; G. Barker, 1-16, ed. G. Barker, 1995. 25 I.2 Method of fieldwork I.2.1Division of the study areas into field walking blocks, the factor of the field block’s size As in all modern surveys, one of the basic goals of the present study is to measure the quantity of surface material in the study areas and to suggest thresholds in the density of surface finds for the site and the off-site. It was to be done regardless of the known archaeological sites in the areas, of all pre-knowledge about the archaeology of the micro-regions studied. Since nothing is known about the quantity and distribution of surface material on the various categories of archaeological sites in the country, the areas of the known forts, of the Late Ottoman-Early Modern villages were surveyed with equal intensity as the open fields. The imperative was to achieve a maximum coverage, as far as ground conditions allowed. Considering the smallness of the overall study area, there was a natural strive to increase the survey intensity. The surveyed territories were divided into unequal field walking units or field blocks, using the existing divisions of the terrain into arable parcels, stretches defined by the micro-relief or the local vegetation patterns40. These divisions are easy to observe and identify on 1: 2500 horizontal geodetic maps and especially on rectified aerial or satellite photographs. It was thus very easy to navigate through the field blocks and the survey process went relatively swiftly: a team of four covered up to 20 field blocks daily, each measuring about 2500 square meters on average. In the field the block was identified, numerated and field walkers were set at roughly equal distances of about 10 meters. Only when the parcel was wider than 50 meters and more than 70-80 meters long (larger than 4 000 square meters) was it divided either along the short or the long axis41. Using the ready divisions of the terrains greatly accelerated the course of fieldwork, but complicated matters in the immediate analysis of the quantification results. In the second survey featuring much gentler relief, the size of the field walking unit was a lesser problem. But in the hilly or extremely fragmented landscape of the first survey area the disproportion in the size of the various field blocks was so great, it was thought it will severely bias the computer estimates. Narrow valley floors or small isolated terraces often measured less than 1000 square meters and stood isolated, not bordering on any other field block of a similar or average size (map I_4). It was thus impossible to join them with neighboring field blocks, without sacrificing their integrity as separate depositional units. When field blocks were compared on the basis of absolute counts of pottery fragments, the smaller field blocks, even when featuring large amounts of pottery on the surface, ranked average or even lower than average. On the other hand when field blocks were compared on the basis of artifact density, the same units ranked far ahead of the field blocks with average size and quantity of surface material. The principle problem was to determine the blocks’ area as a factor in the distribution of surface finds. But this is almost 40 Cf. J.F. Cherry, et al, 22-23, 1991. For comparison, the average size of large field blocks or transects in the regional survey projects of the Mediterranean vary from 0.2 hectares, to as much as one or two square kilometers; D. Keller, D.W. Rupp, eds. 1983; J.L. Bintliff, A.M. Snodgrass, 128-136, 1985; M.H. Jameson et al, A Greek Countryside: The Argolide from Prehistory to the Present Day, Stanford 1994; C. Mee, H. Forbes, Survey Methodology, 33-41, eds C. Mee, H. Forbes, A rough and rocky place: The landscape and settlement history of the Methana Peninsula, Greece, Liverpool 1997; W.R. Caraher, D. Nakassis, D.K. Pettegrew, Siteless Survey and Intensive Data Collection in an Artifact-rich Environment: Case Studies from the Eastern Corinthia, Greece, 7-43, Journal of Mediterranean Archaeology 19-1, 2006. 41 26 impossible as there is always a number of other factors, (visibility, variations in the intensity of survey, taphonomic processes and others) simultaneously affecting the distribution of surface finds. Some of these factors could in fact weigh far more than the size of the field blocks’ areas. Map I_4: 1: 2 500 vertical map of the first survey area with the field walking units The experience of this survey project showed that substantial modifications of the survey results in attempting to eliminate certain biases are actually more likely to confuse than to clarify the picture. In the case of the field blocks’ unequal size, radical adjustments in the initial divisions of the study area will only blur the actual distribution of surface finds, particularly in fragmented landscapes, such as the first survey area. This is because each of the parceled units of land surrounding the village has existed independently for decades.It had a specific history of usage and its own, discrete history of deposition and clearance. Imposing arbitrary divisions over the pre-existing ones will not only make field walking awkward and on many places impossible, but it will greatly distort the original distribution pattern of the surface finds. Because of the nature of the fieldwork method, one often forgets that a certain landscape is never a bare and compact area. The present natural or artificial divisions are more than arbitrary and in a number of cases, they are not so recent to simply see them as an obstacle hiding the original exploitation pattern. These are the constitutive components of the landscape; 27 they make up its real anthropo-geography and model the map of surface material distribution42. An old or presently cultivated vineyard for example, is more than an arbitrarily enclosed surface.It’s become an integral entity by means of repetitive, long-term usage, deposition and clearance practices. Analogous to the way in which the specific parts of the settlement chamber are transformed into various anthropo-topographic elements: a hilltop becomes a citadel, old terraces are repaired and the ancient field network revived or transformed into a settlement and so forth. One looks in vain for an ideal, blank territory underneath. The landscape is essentially discontinuous and this is inevitably reflected in the distribution of the surface finds. Consequently the patterns revealed usually appear irregular and unpredictable, provoking the researcher to believe that the presently existing divisions or the variable size of the parcels in particular, mask a regular and continuous distribution of surface artifacts. Modifications in the shape or the size of the field walking units are therefore warranted only for a few extreme instances, where it is obvious that the size of the field block’s area greatly affects the number of surface finds. But insisting on field blocks of equal size in the conditions of the first survey area, disregarding the actual fragmentation of the terrain is technically difficult and methodically wrong. I.2.2 The field block and the individual transect, problems of varying survey intensity The field block was obviously going to be the basic quantitative unit of the study. Conditions regarding vegetation, soils and exploitation varied greatly, even between neighboring field blocks and this was yet another reason why the given division of the area were followed. These conditions were in fact barely equal within the limits of a single field. Each field block was surveyed by between one and four field walkers, depending on its width. The distance between neighboring field walkers was decided to vary between 7 and 12 meters. The lower limit is slightly above the maximum sight range of the field walker in optimal conditions. On ground with good surface visibility and low artifact density, the surveyors reported finds up to five meters on both sides of their trajectories! On fields with lower visibility or greater quantity of surface finds, the sight range was narrowed to the standard 1-2 meters on both sides of the surveyors’ trajectories43. Thus in principle, each field block was longitudinally subdivided into two, three or four sections, called individual transects. The individual transects were labeled with the surveyor’s personal name and the number of the field walking unit and the artifact counts and other recorded parameters for each individual transect were kept separate (map I_5). J. Benesh, M. Zvelebil, 73-78, eds. P. Ucko, R. Layton, 1994; J.L. Bintliff et al, Deconstructing “the sense of place”? Settlement systems, field survey and the historic record: a case study from Central Greece, 123149, Proceedings of the Prehistoric Society 66, 2000. 43 J.L. Bintliff, A.M. Snodgrass, 130, 1985; C. Mee, H. Forbes, 36, eds. C. Mee, H. Forbes, 1997. 42 28 Map I_5: Field blocks and pottery counts per individual transects. Although complicating the data base, this practice was found useful for a number of reasons. First, it became quickly obvious that there were going to be differences in the distribution of surface finds within every field block and particularly on those with lower artifact densities44. Keeping the records for the individual transects separate revealed the structure of the distribution of surface finds on an individual field block level. Other variations concerning visibility conditions and usage were also recorded for every individual transect. It enabled integration of the visibility factor on a level of individual transect, which proved far more accurate than summing up the counts for the entire field block and then correcting for the visibility factor. Later, during the phase of surface material collection, individual transects were used as collection units, especially on field blocks with low densities of surface finds. Finally, the individual transects counts demonstrated the abilities of each field walker in detecting and recognizing surface material. This source of bias was checked at the beginning of each survey campaign, by the means of a little experiment: the same field walking unit was surveyed repeatedly, each time changing the field walkers’ trajectories. Maintaining the same team over the course of the survey campaign, it was possible to determine the variable ability of each team member of detecting surface artifacts. The same set of parameters was recorded for both the individual transects and the field walking blocks, the latter being merely cumulative of the individual transects’ records: number of ceramic fragments, building material, bones and metal or glass artifacts, the presence/absence of building remains or modern debris and finally, the ground visibility conditions45. In principle 44 M.B. Schiffer, A.P. Sullivan, T.C. Klinger, 4-6, 1978. Usually a number of other parameters are recorded in the large regional survey project in the Mediterranean, J.F. Cherry, et al, 162-163, 1988; C. Mee, H. Forbes, 34-5, eds. C. Mee, H. Forbes, 1997; W. Caraher et al, 10-13, 2006. 45 29 the field walking block is but a spatial frame for the individual transects. The sum of surface finds per field walking block is the sum of the counts on individual transects; its ground visibility is the average of the ground visibility values for the individual transects etc. But by means of individual transects one is never covering 100% of the field block area. Each field walker covers only a sample of his/her transects and thus the sum of the counts per individual transects are always showing only a portion of the total amount of surface finds per field unit (roughly between 25 and 30%)46. This is the coverage parameter.Ina way it is an index of the degree of survey intensity. Again it all depends on the ground conditions, visibility and the actual amount of material on the surface. The more surface finds and higher the visibility, the smaller the portion counted by individual field walkers. Thus even if surveyors were equally spaced on all field blocks, the intensity of survey would not have been equal over the various sections of the study areas. This is why it was of a great imperative to impose a fixed limit on the sight-range of individual field walkers, regardless of the surface conditions. Failing to do so in the first survey area created a certain number of problems in analyzing the large block survey results and as a consequence, in deciding the focus of the regular grid survey. Most typically, on field units where the large block survey recorded higher artifact densities, the total collection by grid units revealed only average amounts of surface material and vice versa; on field units featuring artifact densities below the survey’s average, the collection by regular grid units revealed some of the highest artifact concentrations in the survey area. As discussed further in the text, other factors can also contribute to these discrepancies. It is nevertheless clear that on some field blocks in the first survey area, where the large block or the transect survey recorded artifact densities higher than the survey’s average, the results were biased due to the fluctuating sight-range of the individual surveyor. More precisely, the large block survey underestimated the real quantity of surface material on field units characterized by higher artifact density and lower visibility (including vineyards, which inevitably limited the sight-range to the more usual 1.5-2 meters on both sides of the surveyor’s trajectory). Luckily the relatively small number of field walking units and the close familiarity with the local conditions allowed us to compensate against this source of bias in the first survey. The entire area was essentially resurveyed during the collections of the surface material the following year and during the process, we managed to correct some of the results of the quantification campaign. It was realized that the greatest overestimates (widest sight-range) were made on field units with average amounts of surface finds and good visibility conditions, mostly fields with larger amount of off-site debris situated near the modern village or by the local roads. At the other extreme, the most underestimated field units (where the sight range was normal) were fields with larger concentrations of freshly unearthed material and vineyards in particular. In the second survey the sight-range of the individual field walker was always limited to 1.5-2 meters on both sides of the trajectory. Although inevitably there were minor discrepancies between the transect and the grid surveys, the quantification results were in general much more accurate. I.2.3 The factor of ground visibility The most obvious source of bias in intensive surveys is the ground visibility condition. In reality ground visibility is affected by a number of factors (type and density of the vegetation 46 J.L. Bintliff, A.M. Snodgrass, 133, 1985; C. Mee, H. Forbes, 36, eds. C. Mee, H. Forbes, 1997. 30 cover, modern rubbish disposal, depth of ploughing and so forth), but normally this factor is related to the vegetation cover47. In most regions the type and the density of vegetation varies through different parts of the landscape and through the yearly seasons. In some instances the vegetation cover can greatly reduce the amount of finds visible on the surface and in others it completely covers the surface making systematic surveying completely pointless. The extreme fragmentation of the landscape, particularly in the first study area featuring a pastiche of cultivated and abandoned, overgrown fields, required a close control over the vegetation type and density on every individual transect. The scale of one to ten applied in most survey projects in Greece to express the ground visibility factor was found too wide for the limited varieties of ground encountered during the two survey campaigns. Initially a narrower scale was applied, ranking ground visibility from 1 (standing for best visibility, plowed or cleared fields) to 4 (worst visibility where survey was feasible, abandoned fields with tall grasses and scrubs, with occasional bare spots). It was soon discovered that it is going to be very difficult to standardize the criteria for grading ground visibility on the field walking forms. The agreed grading scheme was often misapplied (e.g. fields with rare scrubs and 50% bare surface were sometimes graded 3 or 4) and more importantly, the surveyors had difficulties in agreeing on a unified perception of this parameter. Participants in the survey were therefore asked to give a descriptive record of the ground visibility conditions for every field walking unit. These were later translated into numeric variables by the author. These records were also found useful for inferring current land use and vegetation, though the very small size of the survey area and the personal involvement in field walking ensured that current conditions on each field were well remembered. Following the usual method of correcting for the ground visibility factor48, the number of counted surface finds was simply multiplied with an integer expressing the visibility factor value. The results however were not always satisfactory. On field blocks with high or average number of surface finds and covered with sparse, short grasses, (ground visibility factor of 2) doubling the artifact counts often produced over-inflated figures. The reason was apparent: the difference in ground visibility conditions between these and recently plowed and cleared fields (ground visibility factor of 1) was surely less than 100%, although such is the assumption behind this simple formula. At the other end of the scale on fallow or abandoned fields covered with dense and tall vegetation, the number of surface finds was regularly so low that even when multiplied by a visibility factor of 8, the corrected figures still remained far below the counts for the neighbouring fields with better ground conditions. As with the problem of the unequal size of the field walking units, it proved very difficult to isolate the impact of the visibility factor. And the greater danger lurked not in underestimating ground visibility conditions, but in overestimating them, in confusing the visibility conditions and the actual incidence of surface finds 49. As with the field blocks’ varying size, ground visibility must not become a determinant of the incidence of surface finds. Correcting for the factor of ground visibility is indeed necessary and useful, but only as long as it strictly addresses the objective conditions of ground visibility. The simple scale of integral numbers, ranging from one to four didn’t express adequately the finely nuanced differences in ground visibility conditions between the different categories of cultivated fields. On the other extreme, its range proved too limited to express the ground visibility factor on M.B. Schiffer et al, 7-8, 1978; D. R. Keller, D.W. Rupp eds, 59-72, 1983; T.W. Gallant, “Background noise” and site definition: A contribution to survey methodology, 403-418,Journal of Field Archaeology 13-4, 1986. 48 J.L. Bintliff, A.M. Snodgrass, 131 1985. 49 J.F. Cherry, 384, eds. D.R. Keller, D.W. Rupp, 1983. 47 31 fallow and abandoned fields. In other words, the sharp discontinuum of ground visibility conditions in the study area couldn’t be adequately expressed through a continuous scale of integers. Fearing that the ground visibility conditions were becoming overemphasized, it was decided to apply a slightly different method. Instead of categorizing ground visibility conditions into discrete classes, based on impressions or vegetation types and expressed through integers, it was decided to roughly grade the percentage of the surveyed surfaces covered with vegetation. Thus a field block with best ground visibility (e.g. cleared field) had 0% vegetation cover, those featuring slightly worse visibility (cultivated fields with turf and sparse grasses or vineyards) had 25% vegetation cover, fields with medium visibility (fields with stubbles or meadows with short grasses) featured 50% vegetation cover and fields with low ground visibility (not cultivated stretches, fields dotted with sparse scrubs or trees) had 75% of the surface covered with vegetation. Fields with lowest ground visibility (completely overgrown, abandoned parcels) had 100% vegetation cover and the number of counted artifacts in these instances was doubled.In all other conditions the number of counted artifacts was increased by 25, 50 or 75 percents, respectively50. In effect this model neatly expressed the fine differences between the various types of cultivated surfaces, but failed to express the seemingly vast difference between the categories of cultivated and abandoned fields. As has been suggested, in conditions of very low ground visibility it is impossible to arrive at a realistic estimate of the true amount of surface material using this or similar methods51. Understandably the final aim of these corrections is to produce relative figures for the amounts of surface material on fields with variable visibility conditions rather than to predict absolute quantities. I.2.4 The impact of modern human activities Unlike the factors discussed previously, the effects of modern human agency can hardly be translated into a simple numeric variable. Naturally these effects vary across different study areas and they can merely be observed and described in their specifics. For now it suffices to mention the most common effects of modern human agency and roughly assess their potential impact on the distribution of surface finds52. When discussing ground visibility conditions, it was noted how vegetation cover is not the only determinant of ground visibility. Returning to fields featuring high densities of surface finds for the purpose of artifact collection, on a number of occasions it was noted that even though the field was recently ploughed, there were hardly any artifacts visible on the surface. Despite the fact that a higher than average amount of artifacts was registered on the surface in the quantification campaign, immediately after the ploughing, the greater portion of the surface material was deeply embedded in large, heavy soil lumps. Thus although there was no vegetation cover on these fields, the visibility of surface artifacts was almost completely minimized after deep ploughing. Luckily the quantification of surface material took place in an agriculturally inactive season, when most fields have been ploughed for at least a month or crops were harvested and the stubble cleared. This finding warns us against surveying fields in different 50 Cf. B. Erdogu,10, 2005, though the author has retained the scale of integers ranging from one to ten. J.L. Bintliff, 204, eds. M. Pasquinucci, F. Trément, 2000. 52 D.R. Keller, D.W. Rupp, eds. 59-62, 1983; B. Slapšak, The 1982-1986 Ager Pharensis survey. Potentials and limitations of “wall survey” in karstic environments, 145-149, eds. J.C. Chapman, et al. Recent developments in Yugoslav archaeology, Oxford 1988; M.H. Jameson et al, 222-223, 1994. 51 32 stages of cultivation or surveying over a period of more than one agricultural season. To be more precise, it is not advisable to put together on the same density map the counts made over different agricultural seasons. Closely related and equally disturbing was the factor of the varying depth of the topsoil. Its significance didn’t become absolutely clear until the second stage of the survey was underway or the collection of the surface material by regular grids. On a number of occasions, while closely surveying locations with high quantities of surface finds, it was discovered that it is impossible to follow the extent of the suspected sites in their entirety. Crossing over to a neighbouring parcel, the material suddenly disappeared from the surface, although it was found in great quantities only a few meters away. There were no doubts that the cluster of surface finds continued to spread over the neighbouring parcel but there was simply no evidence on the surface, not a single artifact from the related time-period; as if the entire surface was cemented or covered by a modern building. The cause behind the problem was fairly trivial, though its implications can be quite disconcerting. We quickly realized that the surface on which artifacts were registered stood twenty centimeters to half a meter lower than the neighbouring, “sterile” surface. There was also a difference in land-use: the former was a deeply ploughed field, while the latter featured a house with gardens and animal huts. The problem recurred on all locations where there was a sudden breach in land-uses; essentially, whenever surfaces exposed at different depths meet. Field blocks used as house-yards or gardens, not cultivated or ploughed with hand tools or light machines tended to feature more Late Ottoman and Early Modern artifacts, while most artifacts dating to earlier human activities appeared on deeply ploughed fields or on vineyards. While the vegetation cover, unless extremely dense and impassible could not completely mask all traces of surface material, the different land use leaves absolutely no base to roughly project the amount of surface material in normal visibility conditions. To complicate matters further, the usually observed correlation between the occurrence of surface finds and ground visibility is not simply technical. Normally fields with best ground visibility are the presently cultivated ones and the great number of artifacts usually encountered on cultivated field blocks is not simply a result of the good ground visibility conditions, but is simply related to the fact that the local peasants tend to bring unsorted manure on the most intensively cultivated surfaces.53However there wasn’t a strict rule and in neither of the two survey areas could we observe a clear relation between the presence of debris from the recent centuries. It wasn’t rare to discover cultivated fields that were nearly sterile. More to the point, fields with low visibility, usually those abandoned or left fallow, lacked the obtrusive carpet of Late Ottoman-Early Modern finds not because of the unfavourable ground conditions, but because manure was applied less frequently. On certain locations, especially along local roads and in small ravines or valleys, the amount of modern trash on the surface was such that it completely covered traces of older activities or heavily distorted the survey record54. Particularly inconvenient was the case of modern brick and tile, easily blending with the ceramic material from earlier periods and 53 I could find only a few, sporadic accounts on Early Modern fertilizing practices in the region, although it was confirmed by local farmers, S. Tomić, Skopska Crna Gora, 409-509, Srpski Etnografski Zbornik III 1905. On manuring and its significance for landscape archaeology see, J.L. Bintliff, A.M. Snodgrass, 506-510, 1988; T. J. Wilkinson, 31-46, 1989; S.E. Alcock et al, 137-170,ed. I. Morris, 1994; C. Mee, H, Forbes, 40, eds. C. Mee, H, Forbes, 1997; see also H. Forbes, Lost souls: Ethnographic observations on manuring practices in a Mediterranean community, 159-172, ed. R. Jones, Manure Matters: Historical, Archaeological and Ethnographic Perspectives, Ashgate 2012. 54 Cf. H. Forbes, 162, ed. R. Jones, 2012. 33 consequently, inflating the survey counts. The large, modern brick breaks into hundreds of tiny pieces that are difficult to recognize unless picked up and examined. Therefore a graph asking the surveyor to indicate the presence or absence of modern rubbish and building material was added in the field walking forms. Another group of modern human activities that can distort the distribution of surface finds are modifications in the relief, whether for the purpose of securing new arable land or for building modern constructions. The most typical, especially for hilly regions prone to soil erosion is the terracing for agricultural purposes. Roads carved in the gentle slopes produce an almost identical effect. Roads and agricultural terraces act as little barriers, retaining all surface material washed by erosion. The result became evident during the large block survey: the amount of surface material suddenly rose by several times on individual transects that were nearest to the terrace edge. Unless these anthropogenic features date to earlier periods, the original distribution of surface material is irretrievably lost on these locations. Modern development affects the original distribution of surface material in a number of other ways. Whenever surveying terrains in the close vicinity of modern settlements one is inevitably faced with the problem of the rapidly changing micro-relief. The use of heavy ploughing machinery levels the cultivated terrain, erasing traces of past human activities along with the features of the original micro-relief: low ridges and hillocks are literarily wiped out, while ravines and small valleys are completely filled with sediments torn from elsewhere. Entire sites of minor size can easily be removed by bulldozers and the material used to prepare terraces for new buildings or roads. Recently on large farming estates, owners insist on removing all inorganic waste from the fields and one often finds large quantities of archaeological material thrown along the edges of the fields, along with stones and other waste55. Ultimately intensive modern development results in a complete destruction of all traces of earlier human activity or at best, leaves the archaeological record in a badly truncated condition, dislocated from its original context and location. Needless to say it is very difficult if not impossible to survey in such conditions. Thankfully both study areas have so far escaped the spread of modern constructions; the largest anthropogenic feature of the studied landscapes is the modern highway in the first study area, covering not more than a 25 meters wide tract. Heavy agricultural machinery and building of modern houses is also very limited. In fact the opposite was the case, at least for the first survey area. The lack of extensive agricultural exploitation, the large areas occupied by overgrown, abandoned fields presented a greater problem for the survey than the excesses of modern development. In this respect the second survey area featured much more favorable ground visibility conditions. It’s impossible to translate the impact of modern human activities into a simple numeric variable as was the case with the parameters discussed in the preceding sections. The range of modern day human activities is wide and varied and their impact is specific not merely to certain landscapes, but to various parts of the same landscape. This group of factors operates simultaneously with the natural conditions and the specifics of fieldwork, sometimes reinforcing, sometimes counterbalancing their effects. It is therefore very difficult to clearly distinguish the work of each of the relevant conditions and predict its impact on the distribution of the surface material. But unlike the ground visibility conditions or the degree of coverage, present-day human activities are not merely distorting the surface archaeological record, they are utterly destroying it. Because of their specificity and uniqueness, the impact of present day human activities can only be assessed idiosyncratically, case by case. 55 B. Slapšak, 46, eds. J. Chapman, et al, 1988. M.H. Jameson, et al, 218, 1994. 34 I.2.5 Categories of counted material, other types of field documentation The principle study subject of intensive field surveys is the surface material. This is a rather broad category and technically it includes all traces of past (and recent) human activities visible on the surface. The results of all survey projects are to a certain degree shaped by the way in which various categories of surface material are defined. It was therefore deemed important to briefly present the categories of surface material used in this survey. As with the estimation of the visibility factor or the degree of intensity, a couple of modifications were made in the survey’s course. The way in which one defines categories of surface material also affects the type of field documentation. In principle there are always two broad categories of surface material: quantifiable and non-quantifiable types of finds.Our primary aim for this study was the first component, but we also tried to make field records of the architectural remains and earthworks. Although seemingly complementary and compatible, the studies of these two components are not always easy to combine56. Particularly in projects with limited funds and expertise, compromises are often inevitable. At the beginning of the survey in the first study area, four categories of quantifiable surface finds were listed: ceramic vessel fragments, ceramic building material, bone fragments and a composite category of metal/glass/flint artifacts (table I_1). After only a few days of fieldwork the first modification was made: tile and brick, along with other types of building material (mud-brick, hewn stone) were moved to the non-quantifiable category of finds. Participants in the survey were asked to simply indicate the presence of building material, as it became clear that most had difficulties in distinguishing tile or even brick fragments from coarse pottery. Similarly hewn stone blocks were often confused with natural rock. Counting this category of material would have naturally produced nothing but confusion. In this respect the small size of the survey team turned to be advantageous, because it was possible to individually inspect all finds collected by the survey team. TableI_1: The field walking forms F P B B one ottery uild. uild. fragments Material Remains y/n. y/n ield block no. Mo etal lass dern rubbish y/n Vegeta tion and land use No major problems were observed in the quantification of bone and metal or glass artifacts. On a number of occasions field walkers were confused by the occurrence of both metal and glass artifacts and bones amidst the piles of modern debris, but this was a purely terminological problem.These were mostly modern artifacts and belonged to the non-quantifiable 56 In this respect exemplary are the surveys of the towns of Phlius, by S.E. Alcock, Urban survey and the city of Phlius, 421-463, Hesperia 60-4 1991; and the cities surveyed in the Boeotia survey, J.L. Bintliff, A.M. Snodgrass, Mediterranean survey and the city, 57-71, Antiquity 62, 1988; J.L. Bintliff, B. Slapšak, The LeidenLjubljana Ancient Cities of Boeotia Project: season 2006, 15-27, Pharos 14, 2007. 35 category of modern debris. A source of greater concern at least in the early phase of fieldwork, was the conspicuous absence of this type of finds in the individual transects records. It was feared that focused on the most prevalent category of surface finds, the pottery fragments, field walkers would unwillingly miss the more isolated and often more concealed metal or glass artifacts. However in the later phase of fieldwork, when total collection of surface material was carried out, the almost complete absence of these categories of finds was confirmed. Isolated bone fragments, mostly sheep were slightly more numerous. The sub-category of flint was quickly dropped from the list, as we lacked the expertise to recognize it57. Pottery fragments are usually the predominant category of surface finds in all surveys in the Mediterranean world. It is by far the most numerous type of surface material and most often, it is the most characteristic one. It is therefore reasonable to expect that this category of finds will be overemphasized at the expense of other categories, though during the later phase of fieldwork, it turned out that the effects of this bias were not as detrimental as was initially thought. Far more problematic was the uneven detection of the various classes of pottery fragments 58. The color and the texture of the fragments appear to be the decisive factors. Participants in the field survey were able to spot and recognize red and especially glazed pottery with great ease, even in unfavorable visibility conditions. Grey or brownish ware, fired on lower temperatures was on the other hand often missed, particularly on freshly ploughed surfaces. This created considerable difficulties, the significance of which was realized only in the later phase of fieldwork. On a certain number of sites in both survey areas, only after total systematic collection of surface material did the full extent of brownish and grayish ware become truly apparent. The non-quantifiable category of surface finds included, besides building material, in situ traces of building remains (houses, sepulchral objects, terraces and fortification walls) and modern debris. Prior to the first survey campaign there was a doubt as to whether we should attempt to quantify individual building remains, but it quickly became clear that this would achieve little in terms of enriching the data gathered, but even more importantly in practice, it often proved very difficult to count building traces as separate, individual units. For instance, it was impossible to confidently say how many tumuli were still visible on the ground in the large necropolis, stretching along the eastern border of the first study area without actually clearing the top layer and even the loose stones in-between the tumuli. The mounds were often built one against another and in most instances oak trees were growing out of the tumular mass. Similar problems were encountered during the survey of the area occupied by the modern village houses and its cemetery. It was almost impossible to count all individual tombs in the village cemetery because the cemetery area was not expanded, but there were either multiple interments in single constructions or new tombs were dug in the narrow space between the existing tombs. Most building remains in the study areas had undergone radical transformations or were almost completely obliterated. A good, traditional documentation with precise ground-plan sketches and photographs would’ve consumed a large amount of time. Instead their presence was simply indicated on the field walking forms, alongside a general identification (e.g. a tomb, a building). Later in the campaign, the related field blocks were re-visited and either individual architectural 57 In the Boeotia Survey Project a flint specialist was employed and assigned to count only stone artifacts, walking transects parallel to the main team; J. Bintliff, P. Howard, A. Snodgrass, 139-168, 1999. In the Nicopolis survey, a separate team carried out the survey on areas suspected to have Paleolithic remains on the surface; C. N. Runnels, T.H. van Andel, The Early Stone Age of the nomos of Preveza: Landscape and Settlement, 47-133; eds. J. Wiseman, K. Zachos, 2003. 58 J.B. Rutter, Some thoughts on the analysis of ceramic data generated by site surveys, 137-142, eds. D. Rupp and D. Keller, 1983. 36 units were mapped or, when a larger construction or a cluster of numerous units were encountered, the ground-plan and the contours were roughly sketched. A basic set of data was thus hopefully secured; that is the location, the extent, the form and/or the spatial organization of the discovered architectural remains. It is interesting to note at this point that some of these earthworks and architectural remains wouldn’t have been discovered when walking in straight lines, holding strictly to the field block divisions. Many of these sites occupied inaccessible locations, such as narrow terraces, isolated outcrops or, as was the case with the mentioned mound necropolis, overgrown tracts in-between the cultivated parcels. These types of location, either because of access difficulties, low visibility or contracted space are often omitted from the field block divisions. On the contrary, in traditional topographic surveys precisely these types of locations are searched with a particular attention. I.2.6 The regular grid survey and collection of surface material The survey projects basically consisted of two general phases, roughly corresponding to the concepts of site-less or off-site and intra-site surveys59. In most parts of Greece and in many other regions of the wider Mediterranean, the land is covered with a more or less continuous carpet of fragmented artifacts, mostly broken pottery and dislocated building material 60. Even when appearing focally, surface archaeological material usually appears over substantial areas, in relatively large quantities. In most cases it is impossible to collect material from all quantitative units in a study area, even if the strictest collection criteria were applied. Highly controlled counting and total collection would have to be limited to certain locations and for concrete purposes. The most immediate goal of the first phase of the survey, the quantification of surface material by large unequal blocks was to reveal locations with higher densities of surface finds, assuming that they indicate sites of past human activities, most commonly some form of permanent or seasonal settlement, but also the overall distribution and profile of the archaeological surface material. The underlying logic is fairly straightforward: a prolonged and/or intensive human presence would leave a many times greater amount of surface material than normally encountered on uninhabited parts of the landscape. But this is only the ideal case; in most instances the surface material in a certain area was produced over a longer period of time. Traces from periods of less intense human presence are hidden amidst a mass of material from other periods and a chronologically homogenous material displays differences regarding function, formal and technical qualities etc61. Crucial to the understanding of human habitation practices in the long term is the systematic and controlled collection of representative samples of surface material. The large block or the off-site survey essentially serves to reveal the overall distribution pattern and point to the various thresholds of surface material density. It is the more intensive, site-centered survey that draws the extent and the location of sites and determines their chronological profile. 59 J.L. Bintliff, A. Snodgrass, 130-137, 1985; J.F. Cherry et al, 162-163, 1988. J.L. Bintliff, A. Snodgrass, 506, 1988; T.J. Wilkinson, 31-32, 1989. 61 J.F. Cherry et al, 159-160, 159-160, 1988; E. Neustupný, 51-53, ed. E. Neustupný, 1998; J. Bintliff, P. Howard, A. Snodgrass, 141-145, 1999. 60 37 A great number of surface material collection techniques have been tried over the past couple of decades in various parts of the Mediterranean62. Experiments were made with systematic random samples, the so called, grab samples and total collections. Naturally it all depends on the concrete conditions regarding quantity and quality of the surface finds, but also the actual extent of knowledge about the chronological development of the various categories of finds. The most efficient and most commonly used is the collection of the so called featureshards, but in conditions of low density or poor quality of the material this tactic would hardly work63. For instance a collection of feature shards in the first study area would barely yield 20% of the counted surface finds and probably less than 15% in the second survey. Moreover so little is known or published on pottery production from certain periods in the country and especially on local, rural production, that it seems that only total collection (with the exception of very small, amorphous or badly damaged fragments) will secure a record of less intensive, local settlement activities. It must be stressed however that total collections were only made possible by the relatively low artifact density in the survey areas (and probably, in the general region). When confronted with the mass of surface material typically found in many regions of Greece or the Near East, some form of sampling is inevitable. In the survey projects presented here, material was gathered from three various types of collection units. As total collection was performed on most field blocks in the survey areas, the collection units are at the same time quantitative units. This circumstance revealed a very important regularity concerning the method of field survey in general. The number of material counted on field blocks with high quantities of finds is sometimes several times lower than the actual number of artifacts on the surface.Even on average ranking field blocks, the actual amount of surface finds is usually two or three times the counted. Consequently the two basic components of the field surveys, the counting and gathering of the surface finds, although complementary shouldn’t be used alternately nor are the results of the two directly comparable. Participants in the survey searched the surface with a far greater scrutiny when material was collected; similarly, much more attention was paid on phenomena that were counted than on those requiring simple indication in the field walking forms. It can be thus claimed that the degree of survey intensity relates not only to the parameters of sight range, obtrusiveness of the finds and the distance between field walkers, but also depends on the documentation techniques for the various categories of surface finds. This creates a problem when trying to interpret the overall distribution of the surface material. Because of the variable degree of survey intensity, it is necessary to correct the records to account for this factor, before combining the results of the transect and the grid survey on the same map. Comparing solely the raw data will inevitably result in higher artifact densities on gridded areas or field units subjected to more intensive transect collection, drawing artificial site limits. For the purposes of this study, the solution adopted was to multiply the transect survey records by a factor of 2.564. Obviously depending on the class of ceramic material in question, this figure sometimes proved too high, sometimes too low, but it does give a fairer image of the 62 The evolution of these techniques in the Boeotia Survey is given in J.L. Bintliff, A.M. Snodgrass, 130137, 1985. For earlier surface collection techniques exploited in the Near East, R. Whallon, Methods of Controlled Surface Collections in Archaeological Surveys, 73-83, D. Rupp, D. Keller, eds. 1983. Most other survey projects used combination of various random sampling techniques and collection of diagnostic shards: the Nemea Valley project, the Pylos Regional Project, the Methana Survey. 63 For an alternative method of collecting samples of surface finds, W. Caraher et al, 12-13, 2006. 64 J. Bintliff, P. Howard, A. Snodgrass, et al. Testing the hinterland: the work of the Boeotia survey (19891991) in the southern approaches to the city of Thespiai, Cambridge 2007. 38 overall surface record. In both survey areas in certain instances it proved equally revealing to focus solely on the transect survey results when determining the limits of the surface clusters. In these cases however there emerged a closely related problem, because the collection of surface artifacts by individual transects weren’t equal on all field blocks. Following the analysis of the large block survey results, three basic thresholds of surface material quantity/density were identified: lower than average, higher than average and very high. These simple, relative categories worked well in a survey area with an equal distribution of surface material over its entire territory, but in conditions of sharp differences in the quantities of finds, it may be necessary to divide the surveyed area into several sectors, enclosing field blocks with comparable densities of surface finds and ground visibility conditions. Such was the case with the first study area, subdivided into 11 sectors, most of which corresponded to certain topographic entities in the micro-relief, but primarily featured comparable conditions regarding ground visibility, exploitation and density of surface finds. It was a necessary step, for otherwise locations that were obviously exhibiting traces of past human activities ranked average or even below the average when all field blocks were compared. On the contrary the gentle, continuous relief of the second survey area and the relatively even surface conditions required an integral interpretation. On field blocks with very high densities of surface material, regular grids were laid out and all surface material was gathered by equal grid units, measuring between 5x8 and 10x15 meters; the size depending on the location and the density of finds. The grids had the same orientation as the field blocks, so as to cover the maximum of a field block’s area. They were in other words inscribed into the field blocks’ irregular perimeters. The grid was expanded as far as the respective material appeared on the surface. Once the quantity of surface finds dropped to an average level or visibility conditions drastically worsened, we stopped expanding the grid and continued the survey using the individual field walking transects as collection units. It became clear at least for the first study area that more than a dozen and a half locations would have had to be surveyed by imposing regular grids, if the initially adopted criteria were to be consistently followed. This meant that relatively large quantities of surface material had to be collected and processed. Fearing that this would be too great a burden, at the beginning of the on-site collections in the first survey, it was decided to collect only diagnostic material. In addition a regular grid was imposed over the selected field blocks, but artifacts were gathered from every second row of units (map I_6). Though relieving the mass of gathered finds, adopting this technique inevitably sacrificed the advantages offered by the quantification of surface material over a continuous surface. Even if we continued to count surface finds on all units and collect from every second row of units, the results would hardly be representative because of the explained difference in survey intensity when quantifying and collecting surface finds. Finally, the little experience earned during the previous survey campaigns taught us that the distribution of surface finds is focal rather than continuous, especially in the case of clusters of minor size and density. Collection by alternately spaced units will therefore always carry the risk of missing a number of artifact concentrations65. After a few trial attempts it was decided to carry out total surface collection by continuous grid units66. 65 M.B. Schiffer, A. Sullivan, T. Klinger, 4-5, 1978. This on-site collection procedure allows for a much greater spatial control, J. Bintliff, A. Snodgrass, 5859, Antiquity 62, 1988; J. Cherry et al, 163, 1988; J.L. Bintliff, P. Howard, A. Snodgrass, et al. 9-13, 2007. 66 39 Map I_6: Regular grids one and two over field blocks in the first survey As mentioned in the preceding paragraphs, the most commonly used technique of gathering surface material is the so called diagnostic sample or the collection of feature fragments (rims, bases, handles, decorated fragments, plus coarse ware examples) from the quantification units. Experience has shown that this tactic secures a good amount of diagnostic material, while greatly reducing the total amount of gathered finds 67. In essence the global tendency is to collect less and document more on the field 68. This is surely a growing tendency, but in conditions where so little is publicly known about local, rural pottery production, it was feared that much will be missed if we attempted sample collections. When grab collections were attempted at the beginning of the campaign, the results were at best disappointing. Most of the survey participants lacked sufficient fieldwork experience and had difficulties in recognizing certain types of ware. Early Modern and Late Antique red ware, as well as glazed fragments prevailed in the collections, although on the surface these categories were clearly the minority69. 67 J.L. Bintliff, A. Snodgrass, 131-132, 1985; W. Caraher et al, 12-13, 2006; the latter approach clearly presupposes a solid prior knowledge of the material. 68 T.E. Gregory, Less is better: the quality of ceramic evidence from archaeological survey and practical proposals for low-impact survey in a Mediterranean context, 15-35, eds. E.F. Athanassopoulos, L-A. Wandsnider, Mediterranean Archaeological Landscapes: current issues, Philadelphia 2004; see however, J. Bintliff, Contemporary issues in surveying complex urban sites in the Mediterranean region: the example of the city of Thespiai (Boeotia, Central Greece), 44-52, eds. F. Vermeulen et al. Urban Landscape Survey in Italy and the Mediterranean, Oxford 2012. 69 J.B. Rutter, 137-139, eds. D.R. Keller, D.W. Rupp, 1983; this is the so called obtrusiveness of the surface record, M.B. Schiffer, A.P. Sullivan, T.C. Klinger, 6, 1978; L. Wandsnider, E.L. Camilli, The character of surface archaeological deposits and its influence on survey accuracy, 169-188, Journal of Field Archaeology 19-2, 1992. 40 Brown or grey fired shards were more difficult to spot, particularly on freshly ploughed fields.They remained unnoticed even when displaying diagnostic features. Eventually it was decided to continue with the initially proposed technique of total surface material collection. Given the present state of knowledge of local pottery production, the prospect of sampling still appears distant and insecure. It can only be hoped that the slow accumulation of data on local ceramic production will eventually allow the application of sampling techniques, enabling quicker, more efficient and less damaging survey campaigns. Actually as the survey progressed, we were beginning to recognize certain categories of finds with a greater confidence (particularly, modern and ancient tile), allowing us to count the total number on the field and take only samples. The relatively low artifact density in both survey areas meant that the basic processing of the total collections wouldn’t present an insurmountable challenge. More importantly it was quickly realized that about the same amount of time was spent on counting and collecting surface artifacts per grid unit. Counting total surface records and conducting sample collections separately would have consumed more time and energy and the records would have certainly been less accurate than the records obtained by total collections. In effect total collection became an instrument of measuring on-site densities. It should be repeated that this approach is only possible thanks to the relatively small amounts of surface material in the survey areas70. The imposition of regular grids is an arduous and time-consuming task, especially on rugged and fragmented terrains. Although it ensures very close control over the spatial distribution of surface finds, it’s clearly impossible to cover even a representative sample of the study area in this manner, unless aided by very accurate GPS receivers. On field blocks featuring average or lower than average quantity of surface finds for their respective sectors, it was decided to use the basic field walking units for the purpose of material collections. Thus on field blocks with average or lower than average quantities of surface finds, artifacts were gathered by individual transects, while field blocks with very low densities of finds were searched unsystematically and all finds were collected. On the majority of the field blocks with low or very low artifact density, the disproportion between the counted and the gathered artifacts was minimal and in some cases, we actually retrieved the same number of artifacts as recorded during the large block survey. In conditions of low artifact density, it was also possible to roughly map individual and smaller clusters of finds. In a number of other instances however, when fields were revisited for the purpose of transect collections we encountered a very different situation on the surface than that indicated by the field block survey. We either discovered larger concentrations of material that went completely unnoticed or we could locate only a small fraction of the material counted during the quantification campaign. As will be shown this creates considerable problems during the analysis of the transect survey results and particularly when trying to estimate artifact density on the basis of the collections by individual transects. It would be obviously misleading to compare artifact densities inferred from transect collections representing 15 and 65% of the material counted during the quantification campaign. Yet because of data loss or a deliberate decision to save the surface record for the more detailed regular grid survey, in some cases the transect collections represented only a very small fraction of the artifact counts. At the same time, whether because of difficult ground conditions or inadvertently, on certain field blocks the transect collections were far more intensive. The problem was particularly pronounced in the first survey area where we often collected a greater 70 These amounts are small in relation to the artifact density encountered in the Aegean or the Near East, but in reality the total collections from both survey areas produced over 20000 potshards. 41 number of artifacts than indicated by the individual transect counts. Clearly in order to make use of these data, it will be necessary to standardize the transect collections across the survey area. For the purposes of this study, we decided to adjust the number of finds collected per individual transect unit to 100% of the material counted and predict the artifact density per period on the basis of the number of finds included in the transect collections 71. The procedure is open to criticism on the grounds that not all of the ceramic categories have an equal chance of entering the transect collections. Indeed in small collections the presence of a few highly obtrusive classes of ceramic artifacts could result in very high theoretical densities for a given period, while the deliberate exclusion of certain categories will inevitably underestimate their true share in the total surface record72. This is none the less a better alternative to simply using the data of the unadjusted transect collections. While collection by individual field walking units can offer a rough preview of the location and the extent of the distinct surface clusters, it can never reveal the exact size, density and the inner on-site distribution. As will be shown in the chapters dealing with the analysis of the survey results by period, even the total grid coverage doesn’t always succeed in recording the full extent of the surface cluster. A number of factors are involved, among the most prominent being that the presence of certain chronological classes of material is not necessarily reflected in the distribution of the overall surface record;73 and (if they do affect the total surface record) they are not necessarily forming continuous zones of higher artifact density74. Blindly following the results of the large block survey and limiting the total grid collections to field units featuring higher overall artifact densities, we often ended up spending precious time in conducting total grid collections on fields covered exclusively with Early Modern off-site debris, while genuine archaeological scatters were lurking on the neighbouring field units. In order to avoid this problem in the second survey it was decided to collect feature shards during the quantification of surface material by individual transects. This decision saved us a great deal of time and effort, although it only cured a smaller part of the problem75. At the same time it created another unperceived difficulty. Due to the low quantities of surface material on certain categories of sites, there was the danger of depleting the surface clusters through transect collections, prior to the total grid survey. However the collection of diagnostic material during the quantification campaign was seen as a more convenient method of probing the chronological profile of the total surface record than returning to the fields after the overall distribution is established. It not only saved us much time and energy, offering a better guidance for the total collections by regular grids, but it also produced less discrepancy between the number of counted and collected finds. Despite all of the deficiencies, the transect survey records are the only source of information for the segments of the survey area that we excluded from the regular grid survey. It is therefore of utmost importance to have them carried out as systematically and consistently as possible. Even so, a certain number of adjustments are necessary during the analysis of the material distribution by periods. These were much easier to implement on the transect survey record from the second survey area, because there the collections by individual transects were 71 Cf. J. Bintliff, P. Howard, A. Snodgrass, et al. 18-19, 2007. T.E. Gregory, 15-35, eds. E.F. Athanassopoulos, L-A. Wandsnider, 2004. 73 This is in essence the “hidden landscape” phenomenon, J.L. Bintliff, P. Howard, A. Snodgrass, 145, 72 1999. 74 Perhaps reflecting certain taphonomic processes or the various components of the settlement areas, S.E. Alcock, et al, 137-141, ed. I. Morris, 1994; E. Nesutupný, 48, ed. E. Nesutupný, 1998. 75 For example, vestigial prehistoric remains would in most instances pass unnoticed unless a total grid collection is carried out, J.L. Bintliff, P. Howard, A. Snodgrass, 145-146, 1999. 42 carried out more consistently and simultaneously with the quantification campaign. In the first survey area when using the transect collections for estimating artifact density by periods, the specific conditions on each field block had to be considered, both at the time of the counting and the gathering of surface finds. I.2.7 Conclusions One must remember that the adopted method of field survey is still in a relatively early phase of development. Most of the problems discussed in this chapter are actually typical for many survey projects in the Mediterranean. Despite the great number of very interesting and revealing studies, there still lack definite cures for problems such as the rapidly changing conditions of fieldwork or the idiosyncratic differences between the surfaces that comprise the survey area. On certain methodological points, and especially in the inferences and conclusions based on data from this type of surveys, there will probably follow further improvements and experiments. Indeed in many archaeological communities even today, the methods presented are treated with a great deal of suspicion or a complete distrust. In its greatest part the method of fieldwork applied for the purpose of the present study was borrowed from the experience of the large regional survey projects in the eastern Mediterranean. It was deemed fair to present the application of this method in some detail, so that the reader can judge which shortcomings are intrinsic to the method of intensive field surveys in general and which were born from the faults of this particular survey project. The survey experience in both study areas brought many methodological challenges and more significantly, it revealed a number of interesting regularities that should apply across a wide range of survey conditions. Hopefully the fruits of this effort were of some general methodological value. Four principle sources of bias generally typical for this method of field work were suggested: the unequal size of the field blocks, the variable ground visibility, the unequal degree of coverage and the extremely variable forms of ground usage or the unequal surface depths. All of these except the last one can be recorded as numeric variables affecting the artifacts counts in some definite way. In other words, their impact as a factor can be approximated. In the preceding sections it was demonstrated how these factors were dealt with during the survey result analysis and why those particular tactics were adopted. It was also pointed to the intricate connectedness between these various groups of factors and the fundamental difficulty of weighing their separate impacts. Here lies the danger of overestimating the reach of these technical factors and of leaving a much narrower space for interpreting the distribution of the surface material in archaeological terms. It seems most reasonable to believe that, unless large scale human alteration is in question or major episodes of erosion and sedimentation, the surface record can provide important information about habitation practices and land-use in the past. Certain corrections in the initial field records were nonetheless necessary to obtain a clearer picture of the surface material distribution. It is hardly surprising that they often appear grossly imperfect. One is attempting to deal with very local and specific conditions using simple formulae. One of the greatest problems of intensive surveys and this has often been stressed by critics among archaeologists working in the region are the extreme ground conditions and the seemingly whimsical appearances and disappearances of the surface finds76. Surfaces exposed at 76 J.F. Cherry, 398-399, eds. D.R. Keller, D.W. Rupp, 1983; citing J. Lloyd, G. Barker, Rural settlement in ancient Molise: problems of archaeological survey, 289-304, eds. G.W. Barker, R.A. Hodges, Archaeology and Italian society: Prehistoric, Roman and Medieval studies, Oxford 1981. 43 different depths thwarted our efforts to follow continuously the dispersal of the surface finds clusters, offering not a trace of evidence to the possible situation “underneath”. Moreover while resurveying parts of the study areas in a different season, it was noticed that not only were conditions drastically changed, but also the amount of material visible on the surface. It was changed to such an extent, that it displayed a substantially different distribution pattern. Understandably these processes will bear a lesser effect on larger sites featuring high artifact density, than on small rural sites, especially if dating to prehistory77. Because of the different ground conditions, both survey areas posed specific set of challenges, which will be discussed in more details in the separate appendices. The first survey area was a pastiche of fields with various land uses, displaying not only contrasting visibility conditions, but frequent and abrupt interruptions in the distribution of surface material. In effect it wasn’t always possible to exactly determine the extent and size of the sites, but this disadvantage couldn’t undermine the survey’s achievements. Even in relatively complicated conditions it was possible to make a fairly accurate record of the surface material encountered and systematically collect sufficient amounts of surface artifacts. The second survey area, although characterized by a gentler relief, with cultivated fields representing nearly 90% of the surveyed terrain generated a different type of problem. While in the first survey the great majority of surface clusters clearly stood apart from their surrounding by the sheer quantity of material, in the second survey, the dense off-site carpet accumulated during the last two centuries effectively obscured most of the unearthed clusters. Consequently we were often misguided in the decision of where to focus the total collections. Nonetheless as in the first survey, it was possible to map the distribution of the surface material by periods and define areas of concentrated human activity in the past. The sources of bias related to the technical shortcomings of the surveys or to external, environmental factors weren’t completely detrimental to the principle aims of the survey, although they inevitably created problems of interpretation. But as pointed out almost three decades ago, analogous problems of interpretation are inherent to all types of archaeological studies, including excavations.78 Another serious weakness of the surface material in general, also often emphasized by most archaeologists working in the region is the low or complete lack of chronological sensitivity of the collected finds79. Indeed some archaeologists are in principle reluctant to discuss the chronology of non-stratified material. Admittedly the chronological framework of the survey data is often very raw. Even in regions where the study of local pottery production has a long tradition, based on material from stratified contexts, surface finds are often dated only to very broad chronological periods80. As might be expected, this circumstance prevents finer historical interpretation: it is mostly impossible to relate observed changes in the surface record to known historical events or decide whether certain sites were truly contemporary or existed successively within the same, broad time-period. In this respect the criticism is well-grounded, but to argue that surface material has no chronological value what so ever is clearly an exaggeration. Even in cases when it is impossible to roughly date certain categories of ceramic material, they can still be associated with other more recognizable categories on the basis of their spatial distribution. The material from the first survey is a particularly good example, because 77 J.L.Bintliff, P. Howard, A. Snodgrass, 157, 1999. J.F. Cherry, 383-384, eds. D.R. Keller, D.W. Rupp, 1983. 79 J.F. Cherry, 379, 383-384, eds. D.R. Keller, D.W. Rupp, 1983; J.L. Bintliff, A.M. Snodgrass, 127, 1985. 80 W. Caraher et al, 12-13, 2006. 78 44 here the great variety of fabric groups were distributed into groups of several different categories, mostly limited to one or a few locations. Thus they formed discrete ceramic assemblages, only certain categories of coarse ceramics appearing alongside different assemblages. These assemblages defined on the basis of fabrics, form and decoration and spatial distribution can be dated through parallels with material known from stratified contexts. In this way survey archaeology can actually contribute to the study of local pottery production, without making presumptions about the chronology of the collected finds 81. Although omitted from the list of basic research goals, the classification of the collected tile and pottery into separate fabric categories and assemblages could be one of the major achievements of this study. All in all the method of intensive and systematic survey is giving results, even when applied in more continental areas and carried out in circumstances of relatively limited research potential and experience. A couple of dozen new sites dating to various time-periods were discovered, definitely justifying the suspicions concerning the accuracy of the data presently available on the extent and distribution of human settlement in local archaeology. For the first time in this region the overall distribution of surface material was studied, revealing important facts about its quantity and composition. It is a crucial step towards understanding past habitation practices on a micro-regional level and from an alternative theoretical perspective. Though somewhat truncated and lacking the fines attributed to excavation data, the findings of these micro-regional surveys can hardly be achieved by the methods of traditional extensive survey or excavations. Nor can they be denied their full significance, solely on the grounds of an a priori mistrust in the adopted method of fieldwork. At this point it seems appropriate to let the results of the surveys speak for themselves, even if the reader disagrees with some of the suggested interpretations. 81 Cf. J. Vroom, After Antiquity. Ceramics and Society in the Aegean from the 7th to the 20th centuries A.C. A Case Study from Boeotia, Central Greece, Leiden, 2003. 45 Chapter II: Description of the first survey area Let us first describe the survey area in its two basic aspects: as a separate composite and as a part in the mosaic of the wider region. Looking at the study area from the first perspective involves its definition as a micro-regional entity and the definition of its inner components. In other words, it is necessary to explain the drawn limits of the surveyed area and the suggested inner divisions. It is a question of finding a topographic or a hydrographic entity and observing and mapping its inner variations. In doing so we had to acknowledge that the way in which we read the present day landscape is to a large degree predetermined by the modern patterns of settlement and land use. These are the parameters that draw the limits of the area, both its outer and inner limits. In the hilly, dry and desolate conditions of the Mid-Vardar Valley, with very scarce resources, one has little choice but to at least roughly follow the limits of the zone of modern agricultural exploitation. Later we’ll see that the ‘anthropo-geography” of the present is not of such a recent date after all and that it chiefly followed the path cleared by its more distant predecessors. Once the outer and the inner limits of the area are defined, we’ll try to understand its role and place in the wider geographic context and determine its geostrategic value. It is a matter of understanding how is this region situated in the mosaic of similar and contrasting micro-regional entities. We’ll also briefly turn to the geographic history of the wider region, to see what place it’s occupied in the political, ethnic and administrative divisions of the past and examine its importance in the regional and interregional communication network. The point is to prepare a background for the survey results. Earlier we’ve insisted that no landscape is a blank territory. Even when in pristine conditions, all landscapes display physical fragmentations and asymmetries in the distribution of resources. Not all locations in a certain area are inhabitable, nor are the basic natural resources equally accessible to its every corner. Every micro-region has its centre and periphery, just as it occupies a central or peripheral place in the wider regional context. We often worry that much of this had been determined or greatly modified by the hand of modern civilization, when it is far more likely that people have always tended to install their settlements, fields and roads onto the natural surroundings with the least disturbance and minimum efforts. This is true for all pre-modern civilizations; people have made optimal use of the fragmentary, asymmetrical character of the natural surrounding: shaping that ridge into a cart-road, that flattened hilltop into a fort or a sacred platform etc. Complete reshaping of landscapes was rare and exceptional, unnecessary and beyond the capacities of the small communities of peasants and shepherds that settled the regions along the Mid-Vardar Valley. And yet their share in shaping the landscapes of the present is both fascinating and monumental. It is the result of a millennia-long repetitive usage and re-usage of the various landscape components82. Resiliently and quietly they’ve been furbishing the natural surroundings to their own ends, both side by side and against the forces of nature. II.1 The valley of Sopot: its drainage basin, size and limits, geo-morphology, communication, natural conditions and land-use The small stream that flows by the present-day village Sopot is one of the literally hundreds of small streams that drain the hilly and rugged banks of the Mid-Vardar Valley. This 82 J. Benesh, M. Zvelebil, 73-93; eds. P. Ucko, R. Layton, 1994; J. Bintliff, 123-149, 2000. 46 section of the valley is called the Taor Canyon, stretching from the village Taor, 15 kilometers southeast from the centre of Skopje to the plain of Veles (maps II_1, II_2). The majority of the streams that drain the rocky banks of the Vardar are but dry, narrow and often, very deep ravines. Only a few display features of true rivers, featuring well articulated hydrographies. The valley of Sopot is one of these. On a regional physical map, we can observe its headwaters, upper, lower course and mouth (maps II_2a, 2b). Yet the creek is basically nameless; the name the valley of Sopot or the central valley is given here for convenience. The local inhabitants have various names for the different parts of its course, mainly referring to the surrounding hillsides, the dominant vegetation or some local event or person. Truly it is difficult to speak of a river valley with floor and sides. For the greater part of its course, the stream is a barely visible channel cutting across the foothills of the surrounding massifs. From around mid-course, this small channel is suddenly transformed into a deep, V-shaped ravine. Its sides become steep, rising almost vertically for tens of meters. Along certain sections the sides of the small valley merge unnoticeably with the slopes of the surrounding hills. The valley of Sopot is formed almost 3 kilometers northeast of the present-day village, at the foot of Vranov Rid (point 339 on map II_2b). Two major sleeves, coming from the north and northwest meet at the feet of Vranov Rid, Radičica and Goli Rid. The toponomy is vague, often imprecise because of the ruggedness of the terrain. These hills are a part of the large Tertiary Basin of the central parts of the Republic of Macedonia. Erosion has carved numerous small ravines into the soft sediments consisting mainly of sand and pebbles. The hilly masses are mostly dissected along the NW-SE axis, parallel to the course of the Vardar Valley. The microrelief is very complex. It is difficult to follow the main ridge line, broken into numerous offshoots and small flattened hilltops. Particularly elusive is the watershed line on the north side, the one that basically separates the drainages of the Vardar and its major tributary, the Pčinja. Along the rest of the valley’s course the limits of its drainage are clear. The watershed line follows the ridges of Prisoj and Goli Rid on the west, Radičica on the east and Gaber on the south. It is higher on the western and northern sides, reaching 550 meters above sea level; on the south and east, the limits of this drainage basin are below the line of 400 meters above sea level. All the hill masses that surround this basin, though insignificantly high in absolute terms, have very steep, hardly traversable sides. They present very effective barriers, leaving only two comfortable exits from the area of the valley: one near its confluence with the Vardar and the other, at the point where the massifs of Radičica and Gaber meet, the southwest and southeast corners of the survey area. The easiest way to grasp the drainage of the valley of Sopot is to visualize a right-angled triangle, where Gaber and Radičica are the shorter sides and the PrisojGoli Rid hill-chain the longer side. The valley of Sopot is both the back-bone of the survey area and the main axis of the administrative territory of Sopot. Measured from the centre of the headwater to its mouth, it is 3.6 kilometers long, draining an area of roughly 6.5 square kilometers. The once powerful stream has cut its course deep at the foot of the hills on its left. For the greater part of its course, the right bank is wider and gentler. The stream flows due south until its mid-course, a point where the masses of Prisoj and Radičica almost meet, forming a narrow strait (point of Bodleva Češma on map II_2a). Downstream from this point the small stream gradually turns westwards, making a slow, 90 degrees turn. From roughly the same point its channel is quickly transformed into a deep corridor, with steep sides and well-defined floor. During the last 400 meters, the valley is cut deep at the northern foot of Gaber, following an east-west direction. Its floor is about 30 meters wide at this point, 15 to 20 meters below the edge of its right bank, the southern foot of 47 Prisoj. In its lower course the valley of Sopot presents a significant physical barrier, splitting the micro-region into very different western and eastern halves. This transformation of the central valley neatly coincides with the change in the local geologic substrate: Tertiary sand and pebbles in the upper course, Quaternary deposits made of sediments eroded from the slopes of Prisoj and Gaber in the lower course (map II_3). In its lower course the central valley is cut nearly along the eastern limit of the Quaternary deposits, so that its left bank is entirely made of Neogene sediments, while the opposite bank of Quaternary colluvium. Thus the central valley is both a topographic and a geologic frontier. Although part of its drainage, the left bank of the stream belongs to the large Tertiary Plateau of Ovče Pole, not to the Vardar Valley proper. To the west the barren hills that separate the Vardar Valley from the drainage area of its tributary are made of older, metamorphic rocks, overlaid with interrupted chains of whitish limestone. The limestone belt crowns the ridges of Prisoj and Gaber, at places emerging into massive, rocky outcrops. Particularly imposing is the stretch from the mouth of the Pčinja to the mouth of the valley of Sopot: a 3 kilometers long and over 200 meters tall wall, crowned with a whitish crown of limestone blocks (photo II_ 1). The limestone belt follows the line of the ridges, descending into the mouth of the central valley and climbs the northwest edge of Gaber. It separates the mouth of the valley of Sopot from those of the neighbouring smaller streams and at the same time, it is a physical and geologic frontier between the drainage of Sopot and the Vardar Valley. Already a century ago the founder of the geology, human and physical geography of the Balkan Peninsula, Jovan Cvijič observed that the Vardar in its mid-course flows at the foot of hard metamorphic rocks, not through the softer Neogene sediments spreading barely one or two kilometers to the east83. The explanation of the famous natural historian is still valid today: the canyons of the Vardar are younger, epigenetic formations. The hard metamorphic rocks of a Paleozoic date were exposed later, through processes of uplift and erosion. The prevailing geomorphologic tendency along the Vardar is one of valley incision and backwards expansion in the region of the headwaters. Cvijić used the term regressive erosion to designate this tendency84. Its effects are particularly striking further down the river course, south of the chain of old metamorphic rocks that form the Taor and the Veles Gorge. But even along the stretches of these canyons, the erosive forces were powerful enough to scar the Neogene sediments kilometers to the east of the Vardar. Two smaller ravines carved in the eastern bank of the valley of Sopot, just before it meets the Gaber Massif are probably the result of this tendency (map II_2a). The northern one is basically a micro-replica of the central valley; it splits the southern foot of Radičica along a north-south axis, turning westward in its lower half. The southern one is twice as short; it is cut in a west-east direction, at the foot of Gaber. Although very small it is an important element in the local topography, opening an easy link between the basins of Sopot and the neighbouring Vranov Stream. The slopes of Prisoj and Gaber are barren and rocky (photo II_2) and erosion has taken its toll even on the upper portions of the Radičica Massif. In the nearer past, when life and agricultural exploitation was more intense in the area, the slopes drained by the valley of Sopot must have appeared even more desolate than now. At present young oak forest covers large portions of the hills east of the central valley (photo II_3). 83 84 J. Cvijić, Osnovi za Geologiju i Geografiju Južne Makedonije i Stare Srbije, Beograd 1906. J. Cvijić, 275, 1906. 48 There is another peculiarity related to the geology of this area. The contact zone between the Neogene sediments and the rocks that form the canyons of the Mid-Vardar is marked by a chain of outcrops made of isolated red conglomerate (map II_3). They appear in the central valley’s mid-course and in the southeast corner of the area as low offshoots of the surrounding hill-masses. All major gullies in the area, including the two mentioned earlier, were carved close to the line where the softer Tertiary sediments meet the older Paleozoic rocks, flanking the Vardar Valley. The prominence of these small rocky outcrops is partly because of the color and the texture of the rock, but their appearance is equally accentuated by the work of erosion, by the recesses cut in the softer sediments surrounding the red conglomerate. They are naturally the most dominating and strategically positioned locations in the survey area and both were adapted into forts, during certain periods of the past (photo II_4). The basin of Sopot is a separate, geographically well defined micro-region. Along most of its perimeter it is clearly separated from the neighbouring micro-regions. Basically there are only two comfortable exits from the area; points where the watershed line is broken and where the surrounding ridges gently descend into the plains. The one leading westwards into the Vardar Valley is the mouth of the central valley, one of the rare points offering easy communication between the left and the right banks of the Vardar. The gently descending ridges of the Taor Gorge are both crossed by a local road and the modern Skopje-Thessalonica highway. On the opposite southeast corner of the area, a low flattened ridge, spurring from the southern slopes of Radičica presents barely a symbolic barrier between the drainage of Sopot and the drainage of the neighbouring Vranov Stream. Both the highway and the local road leave the area at this point, passing under the low plateau of red conglomerates, at the northeast foot of Gaber. The local road follows the small gully, carved at the foot of this hill, while the highway follows the shortest line, cutting across the middle of the foothills and over the central valley (map II_2a, the highway is marked by a full red line; the local roads, black, interrupted lines) Two other roads spring from this central axis, leading northwards. One follows the course of the central valley until the point of Bodleva Češma, where it parts into a branch leading up the steep hillside of Radičica and a branch that heads across the gentler right bank for a local pass leading towards the mouth of the Pčinja and the village Vetersko. The other local road leads northwest, cutting across the western slopes of the Prisoj-Goli Rid hill-chain, parallel to the modern road and the Vardar Valley. It is drawn several hundred meters away from the edge of the canyon cliffs, avoiding the deep impenetrable ravines that dissect the surface into barren and isolated ridges. On a regional physical map this natural communication is shown as an alternative path to the village Vetersko (map II_2b). The passer-by can also continue up the Vardar and arrive at the mouth of the river Pčinja, 5 to 6 kilometers northwest of Sopot. If there ever was an interregional road passing through the area of Sopot, it must have followed this line of communication rather than the overgrown, at places impenetrable floor of the Vardar Valley. There is an ambiguous relation between the present day village and the Vardar Valley floor. Sopot is situated only 250 meters from the Vardar’s left bank and yet, the village and its land belong to the “interior”, like all other villages situated along this section of the Vardar Valley. Even the most exposed houses are hidden from sight from the valley floor. The settlements in the region are either situated on the banks of some of the larger streams that flow into the Vardar or in the more spacious foothills, hundreds of meters above the valley floor. There is a plain and apparent explanation for this pattern. Along this stretch, the Vardar flows through a chain of steep, inhospitable canyons. Their sides are rocky, often rising vertically for hundreds of meters. The valley floor is narrow, reduced to small and often isolated terraces 49 (photo II_5). Larger stretches of flat land appear only around the mouths of its larger tributaries. These too present rare points of exit from the narrow valley floor. During its course through the narrow Taor Gorge, the Vardar is fast and treacherous. It is bridged over on two locations, two times near the mouths of its larger tributaries, the Pčinja and 5 kilometers downstream near the confluence with the valleys of Sopot and Solp. The scarcity of spacious arable land is not the sole problem on the valley floor. There are no freshwater sources in the canyon; the polluted waters of the Vardar are barely good for irrigation. A number of small monasteries and isolated chapels situated either on the edge of the high cliffs of the Taor Gorge or on the valley floor signal that prior to the building of the SkopjeThessalonica railway in the late 19th century, this stretch of the Vardar Valley was a secluded and peripheral corridor. The small and isolated hamlets, mostly appearing on the gentler right bank would hardly survive without the railway line. Nevertheless access to the banks of the Vardar was important for the inhabitants of the villages in the region. The villagers of Sopot for instance, occupied a narrow isolated terrace, to the right of the mouth of Sopot’s valley. It is a thirty meters wide stretch, clasped between the descending limestone ridge and the fast river. A small, now barely visible path links this isolated corner with the village. At present this small terrace is like the rest of the valley floor, densely overgrown. Sheets of modern rubbish brought by the river cover the small sandy pockets. Surface remains and testimonies of the older inhabitants revealed that once, this was an exploited and gladly visited part of the village area. A number of small gardens, supported by terrace walls made of finely cut limestone blocks can still be seen preserved in very good condition at the foot of the ridge. People came to swim and fish here, to rest and escape the hot, treeless foothills. It was a discrete and pleasant corner of the local landscape. Over one hundred meters up the river course, there is another terrace, several times larger than the one belonging to Sopot. It is an important location for this micro-region, because it is one of the rare points where the Vardar can be crossed. At present a hanging bridge connects the river banks. On both sides of the river there are comfortable exits from the bottom of the canyon, upstream through the valleys of Sopot and Solp. It is also possible to continue up the course of the Vardar, along its left bank, but not downstream, towards the plain of Veles. The steep, southwest side of Gaber falls sharply into the river bed, effectively blocking downstream land communication (photo II_5). Therefore the small river terrace acts as a major local crossroad. Too important to belong exclusively to some of the surrounding villages, it became a monastic land. A wide, but badly maintained dirt road leads from the monastery to Sopot. It climbs the descending ridges of the Taor Gorge and continues eastwards, parallel to the SkopjeThessalonica highway; this is the main west-east communication line in the study area. Although geographically it is not a part of the drainage basin of Sopot’s central valley, the young terraces by the Vardar were included in the survey area. They were after all a part of the village territory, a place where a number of everyday activities took place. There are no permanent streams on the surface in the village area. The floor of the valley of Sopot is dry for decades and access to the waters of the Vardar must have been of great importance. It is another question if this was the case in the more distant past. Until the early 20th century there was another, smaller village just across the monastic complex, around the mouth of the Solpski Potok, on the Vardar’s right bank. According to early 20th century ethnographers, the small complex dedicated to St. George started to thrive in the early decades of the 20 th century, after the demise of the village85. It is possible that in the more distant past, the land of the monastic 85 M. Filipović, Severna Veleška Sela, 489-573, Srpski Etnografski Zbornik XXIX 1935. 50 complex was part of another settlement’s territory. Over the last centuryit has become a “forecourt” of Sopot’s hinterland and it was included in the survey area. As a micro-regional entity, the valley of Sopot is the obvious core of the village administrative area. It makes up to between 60 and 70% of Sopot’s hinterland, which also includes the rugged terrain between the left edge of the Taor Gorge and the ridge of the PrisojGoli Rid hill-chain (mapsII_2a, 2b). The village area measures approximately between 9 and 10 square kilometers. However the bulk of the village productive territory, the great majority of the agricultural fields are concentrated on barely 10% of this land. If we look at the agricultural divisions of the land, we’ll notice a multitude of field borders, longitudinal sub-divisions and dozens small terraces on the steep sides of the lower valley (map II_2c). These are clearly signs of a long-term, intensive agricultural exploitation, concentrated on a relatively small portion of the village hinterland. The plough-zone didn’t expand but became “denser”. The rest of the parish land are steep and barren slopes and were mostly used for extensive herding. The inhospitable hillsides sum up to between 80 and 90 percent of the village area. On the other hand, only very small portions of the hillside are flat: the narrow ridges of the hills and the small flattened hilltops. With rare exceptions, these types of locations are most affected by soil erosion and the strong northerlies. The entire hillside is rugged and rocky. Soil layer and vegetation are thin. Only grasses and low scrubs grow on Prisoj. Young oak forest has spread over most of the hills to the east of the central valley, but only recently, in a time of almost complete abandonment of the village and the surrounding fields. The difference in the vegetation cover between the hills east and west of the valley of Sopot is naturally related to the local geo-pedologic substrate. The Prisoj-Goli Rid hill-chain is made of porous, limestone. The small amount of water falling from the atmosphere quickly sinks beneath the surface, allowing the growth of only the toughest, least demanding plants. East of the central valley on the slopes of Radičica, the oak grows on brighter clayish soils that seem to retain water for longer. They have suffered less from erosion because the terrain is gentler and less fragmented. But along steeper stretches and along the watershed line, the typical soil layer is suddenly replaced by small and medium-sized rounded rocks and sparse grasses and scrubs replace the dense oak groves. The orientation of the hillside is also a factor: the northern faces are regularly covered with much denser vegetation than the southern slopes. The most obvious example is the Gaber Massif; its northeast face is almost completely covered with oak, while the southern side is totally desolate. Finally, there are still finer variations regarding plant cover between different micro-relief forms. Thus on Prisoj grazing-land was organized by small and shallow ravines, called simply the “holes” by the local shepherds. On these micro-depressions the soil layer is expectedly thicker and the grasses higher and denser. Each flock had its own “hole” and these were most likely rotated between the village families at certain intervals. Most of the small paths that traverse the hillside were cleared for the daily commuting of flocks avoiding the flat land at the foot of the hills. They are either drawn across the gentler slopes of the hills or along the very edge of the foothills, to avoid intrusions into the plough-zone. The asymmetrical distribution of resources, typical for most of the land along the Middle Vardar Valley has greatly confined the inhabitable zone, particularly on its left bank. It explains the conspicuous disproportion between the number of inhabitants in the villages and the size of their administrative territories. In its prime Sopot had up to 30 households, which sums up to between 100 and 150 inhabitants sharing a territory of almost 10 square kilometers! Even the size of the actively cultivated zone, measuring nearly a square kilometer is more than sufficient to feed a village of this size. After the demise of the Ottoman Empire and the çiftlik system in the 51 first decades of the 20th century, the dependant peasants suddenly become owners of over 5 hectares of arable land on average. According to early 20th century ethnographers, this amount of land was exceeding the productive capacities of the traditional, individual farmer and large portions of the individual landholdings were left fallow or simply, unexploited86. But the relatively oversized village area is also related to the predominant economy of the village, the sedentary herding. Large portions of the arable land are actually reserved for fodder, most commonly clover and rye. In fact according to some of the farmers, these cultures were grown on the fields closest to the village houses, while wheat was grown on the more distant fields, half an hour walk from the village centre. In the summer, the numerous flocks of sheep and goats required extensive pastures and most of the hillside of this micro-region was used for grazing. All villages in the wider region have disproportionally large hinterlands, consisting of smaller “cores”, where the houses and the bulk of the agricultural land are concentrated and several times larger “peripheries”, used as extensive pastures and woodland. To all the scarcities of flat arable land, water and other natural resources, we must add the relatively unfavorable climatic conditions in the study area87. The Mid-Vardar and the neighbouring plateau of Ovče Pole are among the driest regions in the Central Balkans, with an annual precipitation of barely 500 cubic mm88. Most of this falls in the form of strong, often porous rain, in the months between December and May, followed by a long period of drought, with only a few rainy days in the period between early June and late September. This is a very unfavorable regime for most cultures grown in the region. The temperature amplitudes during the year are extreme: the medium monthly temperatures range from several degrees below zero in the months of January and December, to almost 15 degrees Celsius by late April and in some years, to over 25 in the months of June, July and August. In fact relatively high temperatures have been measured as late as mid-October.89 The winters are brief and not very cold in absolute terms, but in reality the cold northerlies and the fog, clogging the valley in early winter, create rather harsh outdoor conditions. By midFebruary, the number of sunny days steadily increases, inciting the brief cycle of nature. As in all other regions along the Vardar, the warm currents from the Mediterranean protrude quietly and slowly; the cold continental currents on the other hand arrive suddenly, followed by strong winds and storms. The local regime of precipitation has pre-conditioned a rather brief cycle of nature. It starts in the last month of winter, stirred by the unusually warm periods in the month of February. By late spring the fields are ready for harvest and grapes and walnuts are harvested as early as September. During most of the year the landscape appears dry and barren. Green belts are limited to the valley’s floor and the north-facing slopes. Only certain grasses in the hillside were seen repeating the cycle in the second half of the year. During most of the year the rest of the landscape is brown and gold, marked by the cleared agricultural fields and the oak forest. It is important to acknowledge the possibility that in the past, the area of Sopot looked very different from today. Thus far, the only historical accounts referring to this landscape in the past are the observations of ethnographers made in the early 20th century90. As far as we know M. Filipovič, 531, 1935. Gj. Tanev, Počvite pomegju Pčinja i Vardar, sekcija Skopje-Veles II, Skopje 1969. 88 Gj. Tanev, 4, tab. 2, 1969. 89 Gj. Tanev, 4, tab. 1, 1969; It is noteworthy that these measurements were made at a station near Skopje. Local amplitudes in temperature and precipitation could be slightly more extreme. 90 J. Cvijić, 200, 1906; M. Filipović, 493-494, 1935. 86 87 52 this micro-region was never treated separately by earlier travellers. A century ago, the landscape around Sopot was far drier than the one described in this study. This is simply the result of the almost complete abandonment of the village in the mid decades of the last century. Its effects were the gradual abandonment of the agricultural fields east of the central valley and an almost complete retreat from the hillsides, especially the more distant Radičica. Thus the cultivated area has contracted mainly on the lower, right bank of the central valley, in a radius of about 500 meters around the present day village, at the southern foot of Prisoj. According to the meaning of the name Sopot and according to local narratives, waters were plentiful in the study area in the more distant past. The frequent toponym of Sopot is an Old Slavic word for a powerful spring, a strong gush of water.91 In the Ottoman census for the years 1467-1468, no less than three watermills are recorded in the village area 92. All older inhabitants will tell the story of the young Ottoman landlord, who used the skins of 100 sheep to calm the powerful springs of Bodleva Češma, in the valley’s mid-course. At present the village fountain tapping water from this spring is dry. Two other springs are still active. Both are found on the southern slope of Prisoj, close to the village.The first is lower, standing only 380 meters to the northwest of the village, by the local northwest road, close to the village cemetery (blue dots on map II_2a). The second is further up the hillside, more than 700 meters from the village, in the same direction. These springs feed the two fountains of the present-day village. However these sources are not particularly prolific, especially during the later summer months and most people are forced to rely on water tapped from wells. These faint clues scattered across the local toponomy and oral tradition indicate that water was perhaps much more abundant in the past. Although drier, the hillsides west of the central valley are much richer in freshwater springs. The lower hill-chain east of the central valley is on the other hand more abundant with vegetation, especially along its upper course. However no water sources are shown on the topographic map nor are there any mentions in the local narratives and the local toponomy. As explained earlier, the abundance of vegetation compared to Prisoj is due to a number of other factors, including the geological and soil layers, relief and agricultural exploitation. A toponym broadly referring to the headwaters of the neighbouring stream, the Vranov Dol, suggest that the area was specifically used for tree cutting. These forests belonged to the inhabitants of Novačani, 3 kilometers to the southeast of Sopot. It is most reasonable to expect that the inhabitants of Sopot harvested wood from the upper course of their valley. These areas are presently returned almost to a pristine condition and are rarely frequented by humans. The intensively surveyed area roughly corresponds to the ploughed zone of the village territory. It basically consists of the foothills of the massifs surrounding the valley, its floor and sides and the two terraces on the Vardar Valley floor. As explained in the preceding paragraphs, this sums up to about 10% of the village territory and about 15 % of the territory of the valley’s drainage basin. In absolute terms a total of 1 square kilometer was intensively surveyed. This is surely not a representative sample of Sopot’s administrative area. The survey blocks spread mostly over Quaternary and Tertiary deposits, covering only a tiny fraction of the masses built of older Paleozoic and Mesozoic rocks. Expressed in percentages, more than 75 percent of Sopot’s hinterland is made of Neogene sediments, 20 percent are the rocky ridges of the Taor Gorge and only a minor fraction belongs to the colluvial deposit in the central part of the area. On the other hand, the latter sediments occupy nearly 40% of the area intensively surveyed. The surveyed sample is even less representative of the micro-region regarding land use and topograpgy. Only 91 92 Lj. Stankovska, Makedonska Onomastika, Skopje 2002. M. Sokolovski, Turski Dokumenti; Opširen Popisen Defter, 4 (1467-1468), Skopje 1971. 53 very small stretches of the inhospitable hillside were surveyed, although this topographic zone occupies almost 90% of the micro-region studied. Even if we attempted to survey a representative sample of the micro-region, it would have been very difficult, if not impossible, to conduct the survey practically. We saw that most of the hillside is very rugged and large stretches are simply impenetrable: basically all sides of Prisoj, the western side of Radičica and the almost vertical sides of the Gaber Massif. A larger portion is covered with dense, impermeable oak forest: the hill-slopes on the valley’s left side, as well as the gentler slopes around its headwaters. Systematic field walking along these stretches is impossible. Ground inclination and visibility allow intensive survey only along the narrow ridges, i.e. the watershed line. This is a narrow, 20-30 meters wide tract of land, mostly barren with scattered pockets of thin soil. Where the soil layer is thicker, as in the shallow recesses around the heads of the steep ravines, the tall and dense grasses completely obscure the surface. We did however manage to survey a remote corner around the headwaters of the central valley, deep into the Neogene Basin. Though topographically it is not a part of the hillside, this locality fairly resembles the unexploited, untamed conditions along the hill ridges. Intensive survey of a representative sample of the various topographic and geological units of the region was never the chief aim of this research. The idea was rather to survey the territory of a geographic and administrative entity as a separate whole and to trace the history of its central settlement. Technically the survey covered a small, unrepresentative part of the microregion, but a great percentage of what was left outside the survey area was simply inaccessible, overgrown or barren rock. Furthermore it was clear that the central settlement and the basic practices of sedentary life took place in the foothills, where all the essential natural resources are concentrated. It made no point to pretend that the two basic zones of the area, the hillside and the foothills offer equally favorable conditions for the development of sedentary life. But as a consequence of this purposive sampling we cannot claim a complete reconstruction of the human landscape in the studied micro-region. In all likelihood based on the findings of other surveys and the extensive surveys in the periphery of the Sopot, there are isolated remains of past human activities even in the hillside, in-between the drainage basins. In fact most of the sites discovered by earlier researchers, hill-forts and religious buildings belong to the hillside and the watershed zone, in particular. But these are mostly the so called, specialpurpose sites, not the settlements proper. The hillside, at least the hills around the Sopot couldn’t sustain a larger settled community, although they were an integral part of the central settlement’s hinterland. They were important for communication and contained important natural resources, primarily wood and pastures. They also played an important role in the local myths and religious practices. The centre of sedentary life however was in the foothills. It is an oasis of flat and fertile land in otherwise rugged and desolate surroundings. In that sense, we did manage to completely survey the inhabitable portion of the studied micro-region and the data and material gathered allow reconstruction of the history of the area’s central settlement. II.2 The 11 survey sectors or the main topographic components of the survey area Although small and relatively homogenous regarding topography and land use, the surveyed area is not a compact, blank piece of territory. The prevalent characteristic of the region along the Mid-Vardar is extreme topographic fragmentation. In every geographic entity we can observe a number of topographic components or sub-units. These are more or less physically 54 separate surfaces, occupying particular locations in the local geography, displaying various ground conditions, lack or abundance of natural resources and different land-use. As elaborated in the previous chapter it was decided to divide the survey area into separate sectors for both methodological and theoretical purposes. The limits of most of the sectors in the study area were drawn along natural borders; only the southern foothills of Prisoj were artificially divided into four sectors because of its size and the large amount of surface finds (map II_4). Sector I covers the two terraces on the left bank of the Vardar, outside the valley of Sopot (photo II_1, map II_5). A large portion of the surface is overgrown, while the central parts of the larger western terrace are occupied by the monastery. Despite the unfavorable ground visibility we managed to survey about 15000 sq meters, divided into 13 field blocks. Because it is incomparably larger than all other sectors of the survey area, the southern foothills of Prisoj was subdivided into four parts: sectors SW-II and NW-III and sectors SE-IV and NE-V. The artificial borders were drawn along the line of the highway and the local, northsouth road, connecting the valley’s floor and Prisoj’s southern slopes; these are basically the longer and the shorter axis of the foothills (map II_6). All four sectors totaled 96 field walking units or 262 000 square meters. This is more than one quarter of the entire survey area. Apart from abandoned locations and dumping yards, surface conditions were for the greater part excellent. Visibility is good and considerable portions of the sub-surface are exposed. The floor and the sides of the central valley in its lower course were included in sector VI. Only a small portion of the surface is cultivated, with most of the fields belonging to the lower half of the valley, south of the Skopje-Thessalonica highway. This narrow but fertile stretch of land shows traces of intensive agricultural exploitation in the recent past(map II_2a, photo II_7). The area was divided into 17 field blocks, measuring about 35 000 sq meters. The fields on the right bank of the central valley, stretching between a rocky outcrop on the north (photo II_11)to a low and barely accentuated ridge that symbolically separates them from the plain at Prisoj’s southern foot belong to sector VII or the eastern foothills of Prisoj (map II_7). The entire area is intensively cultivated and features good visibility conditions (photo II_8). The survey covered approximately 75 000 sq meters, divided into 25 field blocks. Sector VIII covers the floor and the banks of the southern of the two ravines, east of the central valley and cut at the northern foot of Gaber. It is much wider and gentler than the gully to the north. It is 50 meters wide at the floor and 450 meters long, measured from the point where it merges with the central valley to the western foot of the low plateau jutting out of the northeast face of Gaber (map II_7). The sector also includes the narrow, disintegrating bank that separates the two ravines and the low plateau protruding from the northeast face of Gaber (photo II_4). Low ground visibility has confined systematic intensive survey to an area of 38000 square meters, barely a half of the sector’s actual area. Opposite the eastern foothills of Prisoj spreads the wider and the slightly taller foothills of Radičica. The area is roughly rectangular, measuring nearly 900 meters in a north-south direction and 250 across the shorter axis (map II_7, photo II_8). This gentle ridge called Ramnište was covered by sector IX. It features worse ground visibility in comparison to the sectors west of the central valley, but conditions were more even within the sector’s limits. 33 field blocks occupying an area of almost 100000 square meters were intensively surveyed(map II_7). The neighbouring ridge to the east, along with the western foot of Radičica was included in sector X. This ridge runs parallel to Ramnište and at about the same height, its tip ending slightly further south. Its maximum length is 750 meters, the maximum width, 250 meters. The 55 western foot of Radičicahas almost the same size, but a large portion of its surface is covered with the young oak forest. This is a peripheral part of the studied landscape, lacking the usual field terraces and hedges. 24 fields blocks were drawn on the southern half of the sector, on Jakupica proper, plus 9 in the northern half, on Radičica’s slopes. An area of almost 145 000 square meters was intensively surveyed (maps II_7, photo II_9). The upper course of the central valley is a rugged and irregular area, 1.3 kilometers long and 400 meters at its widest. Only a few smaller stretches of land are cleared. They are arranged theatrically, on two levels on the right bank of the valley (maps II_2b, II_9, photo II_10). Each stretch of cleared ground is divided into several agricultural fields. As usual we followed the given divisions of the agricultural land. A total of 31 field blocks, measuring over 133 000 square meters were intensively surveyed. II.3 The geo-strategic importance of the valley of Sopot in its wider geographic context Although the survey area described in the preceding sections is only the inner, central part of this natural micro-region, there is a fascinating diversity regarding natural conditions, local topography and land-use. The differences between the eleven sectors are neither barely perceived nor overemphasized. They are real and the full understanding of the distribution of surface finds is only possible through separate analysis of each sector independently. The preceding detailed description of the study area by sectors had the double goal of presenting the outer and the inner limits of the intensively surveyed zone and of presenting the studied microregion as a partly integrated entity consisting of a number of uneven sub-units, each with specific conditions and varying potential for the growth of human settlements. It is the background against which we’d like to analyze the findings of the survey. But before we turn to the survey result, it is important to broaden the perspective and examine the place of this micro-region in a wider geographic context. One general reason to survey this particular micro-region was its geo-strategic importance and its place in the historical geography of the lands along the Mid-Vardar Valley. As explained at the beginning of this chapter, the small valley of Sopot offers a natural exit from the Taor Gorge, the second largest canyon in the Vardar Valley. It is the most exposed point on Vardar’s left bank, offering access to the plateau of Ovče Pole on the east and south, towards the small plain of Veles. It is impossible to travel by land downstream from the mouth of Sopot’s valley and avoid the village. It is moreover one of the rare points where the Vardar is bridgeable, linking the mountains on the river’s west bank with the main north-south corridor. Because of its “central” location, Sopot is a major local crossroads and a station on the interregional road along the Vardar Valley. The modern Skopje-Thessalonica highway literally passes through the village. It is very likely that in the more distant past, there was also an arterial interregional road, closely following the valley of the Vardar. The chain of visually connected hill-forts in the furthest corners of the study area and along the Vardar, upstream from Sopot is most likely the remnant of a highly developed road infrastructure. They are not only visually interconnected, but they also cover different sections of the surrounding area. Earlier researchers thought that the ancient road along the Vardar, known as the Via Aksiaor at least one of its branches, 93 passed through the area of Sopot, closely following the trajectory of the modern highway. In fact one of the stations on this road, Kephalon, has been identified with the small hill-fort over the 93 This road is mentioned in the Tabula Peutingeriana, in the section, Naissus-Scupi-Stobi-Thessalonica and in few other itineraries of the Late Roman Period, I. Mikulčić, 1982. 56 monastery of St. George, a kilometer to the west of the village. 94 The fort indeed occupied an ideal location for a road station, standing only fifty meters above the point where the modern road leaves the Taor Gorge, turning east towards the village. However there are still major disagreements among historians and archaeologists concerning the exact trajectory of the Via Aksia.95 Particularly problematic are the discordances between the distances recorded in the ancient itineraries and the lengths of modern roads. This is a clear indication that for the greater part, the modern Skopje-Thessalonica road doesn’t follow the line of the ancient Via Aksia.96 The steep, narrow canyons of the Vardar Valley were rarely used for long-distance communication prior to the building of the railway in the 1870’s. They are neither comfortable nor safe and were often avoided, although the valley floor provides the most direct link between the central Balkans and the Aegean. In the Ottoman Era for instance, the main SkopjeThessalonica road passed over 3 kilometers to the east of Sopot, avoiding the narrow straits in the Vardar Valley. Likewise west of the Vardar, a caravan route leading south from Skopje to the towns of Prilep and Veles followed a chain of low mountain passes, crossing over the shoulders of the mountain ridges rather than through the Taor Gorge.97 The Taor Gorge is part of an important physical barrier. It is a segment of a long mountainous chain stretching perpendicularly to the course of the Vardar and linking the mountains of the central Republic of Macedonia with the mountain ranges in the eastern part of the country (map II_1). It thus separates the upper and the middle course of the Vardar, as well as the plains of the northern and the central Republic of Macedonia. This chain consists of low hills, like the ones surrounding the valley of Sopot. It is nonetheless an effective barrier, a kilometer-wide belt of rugged, inhospitable terrain. It effectively blocks not only communication, but the movement of air-masses, particularly those coming from the south. Although only 20 kilometers apart, the plains and the plateau to the north of these hills have slightly colder and longer winters98. Today as in the past, state and administrative borders followed this natural frontier line. Annexing the ancient Macedonian Kingdom in 167 BC, the Romans drew the border of the newly acquired province north of the modern towns of Veles and Sveti Nikole. In fact during the Late Roman Period, the provincial border between the provinces of Macedonia II and Dardania follows exactly this chain of low hills. This is plainly indicated by the language used on the Roman inscriptions: Latin to the north of the mountain chain, Greek in the lands to the south. The road station Kephalon, identified by some archaeologists with the hill-fort near the Sopot, is the northernmost Greek toponym on the Via Aksia. The next station towards Scupi I. Mikulčić, 99-101, 1982. F. Papazoglu, Les villes de Macédoine à l'époque romaine, Bulletin de Correspondence HellèniqueSupplement XVI, 1988; T. Tomoski, Beitrage zu einer Rekonstruktion der Strecke Scupi-Stobi auf der Tabula Peutingeriana, 113-125, Ziva Antika VI-1 1961; B. Georgievski,The ancient road Scupi-Stobi after the Tabula Peuntigerianna, 117-134, Macedoniae Acta Archaeologica 12 1991. 96 This is far more likely than to assume gross errors in the historical records as I. Mikulčić, 99, 1982 and elsewhere, following F. Papazoglu, 87, 1957. 97 The road networks of the Ottoman Period are an unfamiliar topic among historians and archaeologists from the region. There is I believe plentiful indirect evidence related to communication and transport in this period, ranging from various historical documents, local narratives and toponyms, to archaeological evidence consisting of the remains of special purpose settlements, caravan-sarays etc. For the mentioned caravan routes, M. Filipović, 510, 1935; A. Stojanovski, Dervenđistvoto vo Makedonija,Skopje 1974. 98 J. Cvijić, 198-199, 1906. 94 95 57 bears a Latin name, Praesidium.99 Like the supposed interregional road, the border-line was marked by a chain of roughly contemporary fortifications.100 It is very probable that the Romans simply followed an older border that separated the land of the later Macedonian dynasts and the tribal territory of the Dardanians. For a brief period of time in the early 14th century, this border was reactivated, separating the expanding Medieval Serbian Kingdom and the Byzantine Empire101. After the Ottoman conquest by the middle of the 15th century, this natural barrier became the administrative border between the regions of Skopje and Veles. All villages south of this line, until the present-day remained under the administrative authority of Veles, those to the north fall under the jurisdiction of Skopje. The surveyed area is thus part of a zone that is peripheral in a multitude of aspects. It lies at the end of a geographic, geological and climatic zone and during long periods in the past, it had a political, administrative, ethnic and linguistic border passing by in its immediate vicinity. But how important is this circumstance as a factor in the history of its settlements? In what ways does the proximity of a regional frontier or a major interregional road play their roles in the local settlement dynamics? These are the major theoretical difficulties facing the present study. In essence, it is necessary to locate parameters in the local settlement pattern that somehow reflect the role of regional and supra-regional landscape phenomena. In other words we need to establish concrete relations between the historical dynamic on regional or supra-regional and local levels. One still has only a vague idea of what the nearness of an arterial road brings to a settlement. We saw in the case of Sopot that the modern highway brought little advantage to the village. On the contrary, it only took away its share in the regional and interregional road network. After the building of the modern highway, the village and its local roads remained a bypassed, forgotten corridor. Lying on the fringes of major administrative entities in the last century also made life only more difficult for the local peasant; the closest markets or hospital are ten kilometers away in Veles. The village remained trapped and un-integrated in the highly centralized, modern network of settlements and roads. Nonetheless it has to be emphasized that without an arterial road closely following the Vardar Valley, the village and its territory have all the chances to become one of the most isolated corners in the wider region of the Middle Vardar. This is the case with the two neighbouring villages, Vetersko on the north and Novačani on the south. Both occupy locations from where it is either impossible or very difficult to reach the floor of the Vardar Valley and to travel beyond. Their only connections to the outside world are the local dirt roads that link them to the line of the highway, branches that are essentially dead-ends in the regional communication network. These villages are situated on the fringe of the Neogene Basin, along the foot of the impenetrable ridges of the Taor Gorge, where communication across or along the Vardar Valley is possible only at certain points. Yet despite this advantage the village of Sopot differs little from its neighbours, whether in terms of settlement size or life-style and economy. Nothing indicates that Sopot and its immediate surroundings were traversed by one of the most important roads in the wider region of the Central Balkans. I. Mikulčić, 100, 1982; the same theses are later repeated in I. Mikulčić, Spätantike und frühbyzantinische Befestigungen in Nordmakedonien, Munich 2002. Earlier this was suggested by the historian F. Papazoglu, 87-96, 1988, and N. Vulić, 1-5, 1928. 100 I. Mikulčić, 22, 2002. 101 T. Tomoski, Ovče Polje vo Sredniot Vek, 243-265, Godišen Zbornik na Filozofskiot Fakultet 30, 1978. 99 58 The difficulty of finding parameters in the surface archaeological record concretely related to the wider geo-strategic value of certain micro-regions rests in the very nature of the problem. To securely determine relations between the realms of the main historical narratives for the wider region and the developments inferred from the surface archaeological record in one of its miniature parts, it is necessary to have a far greater amount of comparable data than the one obtained during this small-scale research. This was briefly discussed in the “Method and Theory” chapter: the only path to understanding the impact of wider, interregional on local developments is through the comparative examination of the results of a number of standardized regional studies102. Only thus would it be possible to locate certain patterns in the size, character and distribution of archaeological sites in relation to the developments and events inferred from other types of sources. This is the only means by which we can hope to archaeologically recognize the impact of wider social and historical phenomena, such as main arterial roads or provincial borders on local developments. In that sense a major problem for this and similar studies is the absence of comparative material from the region of the Vardar Valley. Subjects such as major roads or political frontiers have hardly been treated as research problems in archaeology in general, outside the highly specialized field of the archaeology of the Roman army. For the region in question there are numerous studies on the historical geography of the country, written by both historians and archaeologists, but apart from supporting purely historical theses, there was very little advance in the gathering and use of archaeological evidence related to this type of issues. In the conclusion to the next chapter we will attempt to synthesize the archaeological evidence available from surrounding micro-regions and observe the local developments in the first survey area within the broader regional context. But addressing the issue of the possible impact of the supra-regional phenomena (roads, borders, colonies) on the history of human settlement in the basin of Sopot remains beyond the reach of this study. Apart from the absence of comparative data, one could say that the definition of this supra-regional phenomenon is itself problematic. Last but not least, while trying to determine the date of the surface material gathered we quickly became aware that finer dating was going to be impossible. We had to work with very broad chronological periods, which obviously won’t allow us to relate the local history of settlement with major historical events and processes. 102 P. Attema, 18-26, 2002; S.E. Alcock, J.F. Cherry, eds. 2004. 59 CHAPTER III. Chronology of the collected material and spatial distribution by periods III.1.1 Introduction Before we begin with the discussion of the distribution of surface finds by periods, it is important to look at the immediate survey results. In Appendix I we presented the total quantity of surface finds in terms of individual fragments and weight and discussed their spatial distribution at several levels. The overall distribution of finds and particularly their distribution on sector and site levels are analyzed in a greater detail. This analysis reveals a very focal and discontinuous distribution pattern on every level, often with very sharp differences between the zones of various artifact densities. A particular attention is given to the various post-depositional processes and events that affected the surface record and within the same context, to the relation between the local physical environment and the overlying carpet of surface finds. The prominently focal character of the distribution of surface material, but also the fragmented character of the surveyed terrain, seemingly allows the definition of discrete site and off-site zones. In some cases, it is likewise possible to define separate zones of artifact density within the limits of a site or observe the transition between the site and the off-site. But the careful reader will quickly realize that these “sites” are but greater concentrations or clusters of surface finds. In Appendix I they are primarily seen and analyzed as quantitative and physical phenomena, rather than as the remains of past human installations in the landscape. The main goal of this discussion is to provide the sympathetic reader with an account of the fieldwork, of how much of the denser artifact concentrations were covered by the regular grid survey and of the physical conditions potentially shaping the surface archaeological record. The total archaeological surface record was naturally formed over a long period of time, often accumulated in discontinuous intervals and with variable intensity103. It is most conveniently envisioned as consisting of many layers, each representing a separate archaeological epoch. But in reality, the total surface record of an area is an agglomerate of the remains of all periods of local human occupation. It is a palimpsest obscuring both chronological and functional interpretation104. In this respect the estimated areas of these clusters of finds and their artifact densities tell us little about the actual size and intensity of human settlement in various periods of the past. Before the collected material is chronologically sorted, it can simply point to certain locations in the landscape which humans in general chose to inhabit or exploit more intensely regardless of the particular time-period. The principle aim of this chapter is to distinguish the input of every period of human occupation in the surface record of the survey area. This should ultimately explain the formation of the clusters of surface finds described in Appendix I. More importantly it will permit us to observe the distribution of surface finds by broader chronological periods and shed light on the history of human habitation over the long term. In order to synthesize these newly acquired data with earlier observations on the physical environment and the location of sites, frequent references will be made to the discussions in chapter II and in Appendix I. Once the material is chronologically sorted, for some periods it becomes difficult to retain the concepts of site and 103 M. Kuna, Method of surface artifact survey, 77-83, ed. E. Neustupný, Space in Prehistoric Bohemia,Prague 1998. 104 T. Whitelaw, Reconstructing a Classical Landscape with Figures: some interpretative explorations in NW Keos, 227-243, eds. R. Francovich, H. Patterson, Extracting meaning from ploughsoil assemblages, The Archaeology of Mediterranean Landscapes 5, Oxford 2000. 60 off-site, at least in their original meaning (although the physical limits are often very clearly defined)105. The material from most prehistoric periods was often found limited to one or a few field units, with little or no finds in the rest of the surveyed area. On the other hand, certain periods were characterized by single large and dense cluster, surrounded by zones of lesser density and sporadic appearance of isolated, tiny clusters further away from the main cluster. One or two assemblages were found dispersed in yet thinner scatters, but limited to the sectors where the main cluster was situated. Especially in the latter case, it is sometimes very difficult to draw even a very flexible site limit, because the finds collected outside the main concentrations are not dispersed in an even carpet. They rather appear in small scattered groups, representing much smaller concentrations than the main cluster, but still several times greater than on the surrounding field units. In principle one can still determine a site threshold, but it is very uncertain if the material scattered on the surrounding fields represents traces of field manure or other types of human occupation. It seems that off-site material resulting from field manure was characteristic only for certain time periods, such as the Late Ottoman and the Early Modern Period and perhaps, the Roman and Late Roman Period106. Despite all this the concepts of site and off-site will be retained during the analysis of the material distribution by periods. As will be shown, many of the phenomena observed by surveyors in various parts of the Eastern Mediterranean, such as site core and site halo can also be observed in this survey area107. The neutral term of satellites will be used for the surrounding, “secondary” clusters. We’ll see that in most instances these are truly smaller and isolated concentrations, though we often failed to decisively document their extent. They simply went unnoticed during the field survey, precisely because of their small size or the predominance of Early Modern off-site finds. Some of the assemblages are difficult to date with certainty. Indeed it has to be admitted that many fabric groups, especially the coarse ware are hardly datable. Only in cases where it was possible to distinguish a separate coarse fabric and where it appears consistently with more datable fabrics could a tentative dating be suggested. Surprisingly or not, the communities that settled the studied micro-region had a local and rather inert tradition of pottery production. Very few changes were made regarding modelling and decoration during long time-periods. Being locally produced, the pottery from different periods was made of similar raw materials and fabric classification was inevitably a very delicate and tricky process. Above all and in relation to the latter fact, there are hardly any close parallels among the material gathered from the surrounding regions. In order to make the main text more legible, the detailed descriptions of the surface material distribution by periods and the alternative readings will also be given in a separate appendix, in Appendix II. There the reader can find more explicit information about the ways in 105 M. Kuna, 78, 1998; E. Neustupný, The transformation of community areas into settlement areas, 45-61, ed. E. Neustupný, 1998. 106 Indeed once a quantitative approach to the surface archaeological record is adopted, the concepts of site and off-site do become primarily heuristic devices and not true theoretical concepts. T.J. Wilkinson, The definition of ancient manured zones by means of extensive shard-sampling techniques, 323-333, Journal of Field Archaeology, 9 1982; J.L. Bintliff, A.M. Snodgrass, 506-513, 1988; J.L Bintliff, 200-215, eds. M. Pasquinucci, F. Trément, 2000. On the other hand, while theoretically more sound, it is very difficult to operate with the concept of settlement area and its components solely on the base of surface material collections. 107 J. Bintliff, P. Howard, A. Snodgrass, 15-37, 2007; for different views see, S.E. Alcock, 135-168, ed. I. Morris 1994;D.K. Pettegrew, Chasing the Classical farmstead: assessing the formation and signature of rural settlement in Greek landscape archaeology, 189-209, Journal of Mediterranean Archaeology 14, 2001. 61 which we arrived at the interpretations given in the following paragraphs. These will include definitions of the site location, size and inner structure, definitions and interpretations of the offsite zone and the satellite clusters and brief considerations of locational preferences. Finaly, because of the complex distributional patterns and the possibility of different interpretations, this chapter will end with a longer summary of the history of habitation in the first survey area, where we’ll attempt to make a condensed overview of the long-term developments and insert the local into the broader regional dynamics. III.1.2 The Middle Neolithic Period (6th millennium BC; tables 1 and 2, Appendix 2) The earliest community to occupy the survey area is represented by slightly over 800 potshards, exclusively limited to 4 field walking units on the southern tip of sector IX (map III_1b). This assemblage constitutes the greater portion of site 11 (see Appendix I). It is spread over two terraces on the left bank of the central valley, overlooking the junction between the valley and a small ravine that delimits the sector from the east. The material dated to the Middle Neolithic is limited to the western half of the upper and the southern portion of the lower terrace, spreading over field blocks 157, 159, 192 and 193. On these field blocks the transect survey recorded artifact densities ranging from 5 to 26 Mid-Neolithic shards per 1000 sq meters. In contrast, the rest of the 270 field blocks in the survey area featured zero density of MiddleNeolithic finds. These finds were accompanied by a group of later period fabrics, which are mostly concentrated to the east of the Mid-Neolithic cluster, but the two are also largely overlapping. They determine the size and the shape of site 11. There are also negligible amounts of Hellenistic, Roman and Early Modern material (graph 1 in Appendix II). Total collection by grid units was carried out only on the upper terrace, because of the higher overall artifact density and the visibility conditions. If we focus only on the Mid-Neolithic finds collected by the grid survey, their estimated mean density is 12 per 100 sq meters or nearly half the mean density of all surface finds on this site (map III_2). The maximum density of midNeolithic pottery recorded in the southwest portions of the site reaches over 70 fragments per 100 sq meters. From this core they dwindle rather abruptly to less than 2 fragments per 100 sq meters, with a second peak in the central part of the grid. Because of the absence of this material on the rest of the field units, the site limits on the upper terrace are clear. However as discussed in the appendices, it is certain that the Neolithic site also occupied the lower terrace, as well as most of the remaining area of the upper terrace, west of the site core. It is also quite possible that it spread a little bit further to the east, given that its surface presence is smothered by the later assemblage of fabrics, mostly concentrated in the eastern half of the upper terrace. The character of the finds, the presence of storage vessels, flint tools and an almost entirely preserved quernstone indicate a wide range of occupational activities. In addition, the total collections included a fragment of a seated female figurine and a fragment of an obsidian scraper. The most obvious and likely interpretation of the survey findings is that the Middle Neolithic site consisted of two roughly equal cores, positioned near the edges of the two terraces. Together they occupied an area of about 8 500 sq meters or nearly 90% of site 11. In terms of quantity they comprise almost 55% of all finds on this site and 4.8% of the total collection in the first survey area (table 1, Appendix II, see table 2 for residual analysis). This is one of the larger nucleated settlements in the studied micro-region. As discussed in Appendix I, it occupies a central position in the survey area providing it with an equally easy access to both the Neogene 62 soils in the eastern half of the survey and the colluvial deposits on the western bank and on the valley floor. Not surprisingly there are absolutely no traces of off-site activity from this period. Few other contemporary sites have been discovered so close to the Vardar Valley, especially along the stretches of the valley south of Veles108. Most of the known Middle Neolithic sites are located in the Neogene basins to the north and east. This could suggest that the prime resource for this community were the lighter Neogene soils in the survey’s eastern sectors, more suitable for hoe agriculture109. The steep eastern and southern sides of the ridge provided relative protection. Nevertheless this settlement occupies a rather exposed location, easily accessed from the floor of the valley or the Ramnište Ridge. It overlooks one of the central crossroads in this micro-region, where the main east-west artery is joined by a local road leading north, towards the confluence of the rivers Vardar and Pčinja. As will be shown, during some later periods of Sopot’s history the local settlements occupied far less exposed locations. At least in the more recent past, the nearest freshwater sources were situated on the valley floor. There remains the problem of the approximate longevity of this settlement. The assemblage was roughly dated to the Middle Neolithic based on general parallels found on sites in the neighboring regions. There lack the finds characteristic for the Early or the Late Neolithic, although it has to be stressed that these chronological distinctions are mostly based on the analysis of fine painted ware and terracotta figurines110.Judging by the relatively large number of finds and the production and use of at least five different pottery fabrics, it seems plausible to argue that this settlement existed over a longer time-period. III.1.3 The Late Neolithic (5th and first half of the 4th millennium BC?Tables 3 and 4 in Appendix II) The most likely candidate for a direct successor of the Mid-Neolithic settlement is a community that produced an undated assemblage of fabrics, mostly overlapping with the MidNeolithic assemblage, but also spreading in a thinner carpet over most of sector IX. It consisted of slightly over 380 fragments classed into two basic fabric groups. This assemblage is less than half the size of the Middle Neolithic assemblage, but its dispersal area is nearly 10 times larger (table 3 in Appendix II). The great majority of the finds or nearly 80% were collected from site 11. Already at the time of the transect survey it became evident that this was a site location. The combined quantities of the Middle and the possible Late Neolithic assemblage clearly stood out from the low background densities in sector IX. However after the collected finds were divided into separate assemblages, there arose serious doubts concerning the initial interpretation of the surface record as explicated in Appendix I. 108 An up to date summary of the known settlements from the Upper Vardar is given in D. Zdravkovski Sreden Neolit vo Gorno-Vardarski Region, 55-63, Macedonia Acta Arheologika 18, 2008. 109 A. Bogaard, The Nature of Early Farming in Central and Southeast Europe, 49-58, Documenta Praehistorica XXXI 2004. 110 Radiocarbon data from the Neolithic sites in the regions of the Central Balkans are generally scarce, M. Garašanin, The Stone Age in the Central Balkan Area, 75-186, Cambridge Ancient History, volume III, Cambridge 1983; P. Biagi, M. Spataro, New observations on the Radiocarbon Chronology of the Starčevo-Criş and Kōrōş Cultures, 35-50, eds. L. Nikolova, J. Fritz, J. Higgins, Prehistoric Archaeology & Anthropological Theory and Education. RPRP 6-7, 2005; D. Orton,Herding, Settlement and Chronology in the Balkan Neolithic, 5-40, European Journal of Archaeology 15-1 2012 63 The possible Late Neolithic assemblage succeeds the Middle Neolithic on site 11 (map III_3). As mentioned in the preceding section the two are mostly overlapping, with the core of the undated assemblage situated a couple of dozen meters to the east of the Middle-Neolithic site, in the eastern third of the gridded area. Here the total grid survey recorded a maximum artifact density of nearly 30 fragments per 100 sq meters. There lack a regular concentric on-site pattern. The site on the upper terrace consists of two cores defined by artifact densities higher than 6 fragments per 100 sq meter or more than 3 fragments per grid unit. A peripheral zone of between 1 and 3 fragments per grid unit stretches over most of the gridded area, except for grid units along the northern and western peripheries of the grid. Similarly to the Middle Neolithic cluster on the upper terrace, the two cores are not symmetrical. The smaller and thinner concentration is situated at the southern edge of the terrace, while the much larger one is situated between 5 and 10 meters to the north. The same arguments supporting the existence of a Middle Neolithic cluster on the lower terrace apply to the distribution of the undated assemblage. That this site spread over to the lower terrace is demonstrated by the high artifact densities recorded by the transect collections, at least 2-3 times as highas on the rest of the field walking units, including field block 157. Compensating for the low ground visibility and the lesser degree of survey intensity will elevate the artifact density on field units 192-193 to an on-site level, but not on the rest of the field units where this material was collected. The total collection on the small cleared parcel on field block 159 also revealed on-site artifact densities. Like the Mid-Neolithic site, this settlement spreadboth over the upper and the lower terrace, possibly occupying a slightly larger area, nearing 1 hectare. The collections from the other field units in this sector, from field blocks 160-162 or 168171 must also be treated as distinct site locations. What remains unclear is their exact size and extent. After the analysis of the results, we adopted the interpretation that these are series of small and thin clusters featuring artifact densities not greater than 1.5 fragments per 100 sq meters, when corrected for the visibility factor and the lower degree of survey intensity (map III_4a). This is well-above the district mean densities recorded by the grid surveys, but it barely equals the densities recorded on the periphery of site 11. Nevertheless if we focus only the transect collections, there are absolutely no grounds to distinguish between site 11 and the rest of the field blocks on which this material was encountered (table 4, Appendix II). This observation remains valid even if we assume that the transect collections were equally thorough on all field units (map III_4c). Because of the low resolution of the field block survey their exact extent remains unknown, but assuming that these were separate settlement locations would produce a nearly impossible pattern of three settlements each measuring roughly one hectare, all situated within an area of slightly over 15 hectares. However one shouldn’t exclude the possibility that we are dealing with the remains of a shifting settlement and that these locations weren’t occupied at the same time. Judging by the quality of the material outside the limits of site 11 and their inconspicuousness during fieldwork, it is more likely that during this unknown prehistoric period, just as during most other periods when the surveyed basin was occupied, there was one central settlement, in this case situated on the two terraces covered by field blocks 157 and 192193, accompanied by a series of smaller satellites situated hundreds of meters from the main settlement site. This peculiar distribution pattern once again raises the question of the chronology of the assemblage. In regional archaeology it is usually assumed that during the three millennia of the Neolithic Period, all settlements were nucleated. But it must be stressed that the survey method 64 used in this survey has no precedence in the surrounding regions. The methods of landscape exploration used by the majority of local archaeologists are ill-suited for discovering and documenting dispersed sites. Although far more widespread, the two concentrations north of site 11 are thinner and probably consist of a number of tiny and isolated clusters. Unless resulting from some unknown post-depositional processes, the amounts recorded on these locations suggest much less intensive occupation, on the order of extensive burial ground or industrial activities. At most these satellite clusters could represent the remains of isolated dwellings or huts, representing a strikingly different pattern of habitation in comparison to the highly nucleated Mid-Neolithic settlement. It is difficult to bring forward even a highly speculative interpretation, in the absence of comparable data from the nearby regions111. Nevertheless it is indicative that most, but not all prehistoric assemblages from this survey area exhibit similar distribution pattern. The nearly total nucleation exhibited by the Middle Neolithic assemblage will never be repeated in the later periods of human occupation in the first survey area. Most decisive for the tentatively proposed dating were certain fabric and formal features, such as the rim shapes and decoration techniques. Diagnostic fragments of this group feature simple vertical rims, very similarly shaped to those seen in fabrics dated to the Mid-Neolithic, but also fragments decorated with barbotine. On some examples it was possible to observe very small amounts of grass temper, preserved as linear voids on the surface. This feature is also characteristic for Neolithic pottery production and rarely occurs in later periods. It must be stressed that there lack direct parallels from sites dated through excavation. However very similar pottery was found on a Middle to Late Neolithic site, situated at the northern entrance of the Taor Canyon. At the same time it has to be admitted that even if this assemblage truly belongs to early prehistory, in general it shows little resemblance to the prevalent Middle Neolithic fabric groups. This circumstance suggests that there was probably no direct continuity between the two overlapping settlements. The community that produced this assemblage occupied the location of the Middle Neolithic site, making use of its central position, command over the local roads and easy access to the floor of the central valley and the easily arable Neogene deposits of the eastern half of this micro-region. In this respect the other two locations where clusters of this material were found are to a certain degree less favorable. The nearest freshwater springs are hundreds of meters away and the humus layer is probably thinner than on the tip or the lower terraces of the ridge. Although withdrawn from the central section of the area, these are open and unprotected locations. Nevertheless a similar location, site 12, on field block 167 was certainly fully occupied sometime during the Hellenistic and perhaps the Early Roman Period. III.1.4 The Bronze Age (3rd and 2nd millennium BC, until ca. 1200 BC, tables 5, 6 and graph 2, Appendix II) The settlement on site 11 was definitely deserted by the time of the Eneolithic, at the latest. There are no traces of human occupation in the survey area during the second half of the 111 A similarly dispersed settlement, with one central and a few satellite clusters dated to the Late to Final Neolithic have been recorded in far-away Laconia, on the southern Peloponnese, W. Cavanagh, Joost Crouwel, The Survey Area in the Prehistoric Periods, 121-150, W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al. The Laconia Survey: Continuity and Change in a Greek Rural Landscape, Volume I, London 2002. 65 4th millennium BC. It seems that this episode of complete or near complete desertion of the will be repeated only in a few other periods of the past. The following Early and especially the Middle Bronze Age are among the least understood periods in domestic archaeology. Comparatively a rather small number of sites are dated to this period and only a few were excavated. Our knowledge about the pottery production at the time is particularly limited112. Luckily a very small assemblage with a distinct and characteristic assemblage was recognized as a Bronze Age material by a colleague from the National Museum, though it wasn’t possible to date it more precisely113. It consists of at least three distinct fabric groups, each represented by several or a dozen finds. But despite the small size of the collection, the finds are nicely preserved and a number of fragments have preserved traces of matt reddish slip on the surface. As discussed in Appendix I, this could be related to the fact that these artefacts were probably unearthed not long before the grid collection. The bulk of the finds datable to this period were collected from sector I, from a garden in the yard of the monastic complex (mapsIII_5, 6). This is field block 5b covered by grid 16b. Discussing the distribution of the overall surface record, this cluster was defined as site 3 (see Appendix I). Because of the peculiar conditions surrounding its location and discovery, especially the presence of uncultivated parcels and monastic outbuildings along the edges of the field, it is very probable that only a portion of the site area was revealed by the grid survey. The couple of dozen fragments collected from site 3 occupied an area of 800 sq meters, including the site periphery. The site core is defined by artifact densities higher than 8 fragments per 100 sq meters or 4 fragments per grid unit. It was discovered in the southeastern end of the grid, on the right bank of a small seasonal creek that drains the rugged terrain on the Vardar’s left bank. Taking into account only the zone of over 2 fragments per grid unit, the Bronze Age site measures at least 300 sq meters. Sites of this size can only be interpreted as the remains of individual farmsteads. The farm exploited the narrow terraces on the floor of the Taor Canyon. A larger nucleated settlement can hardly be expected on this location as cultivable land on the Vardar’s left bank is limited to an area of a few hectares. But the individual transect collections indicate that the Bronze Age material also spread over to the other side of the creek, on field block 6 (table 6, Appendix II). The small collection comprising several fragments from the northernmost individual transect on this field unit probably came from a separate core. Because of the low ground visibility and survey intensity, it is impossible to determine its size and relation to site 3. But knowing that the same collection technique from field 5b yielded not a single fragment datable to the Bronze Age, one shouldn’t exclude the possibility that the two cores were of a similar size. Only on field block 6 did the transect collections included finds datable to the Bronze Age. This assemblage is not limited to sector I. A couple of certain and another couple of possible Bronze Age finds were recognized among the grid collections from later period sites in sectors III, V and VII (mapsIII_5, 7). The couple of securely identified fragments came from the southeast corner of site 5a, about 150 meters west of the modern village. Possible Bronze Age shards were collected from sites 6 and 8; the first situated immediately northeast of the village, the second approximately half a kilometer to the northeast, at Prisoj’s eastern foot. When these isolated finds are translated into density values, they hardly approach the limit of 1 fragment per 100 sq meters. This is surely above the survey’s and sector’s mean values for the Bronze Age, 112 Z. Videvski, Bronzenoto Vreme na Vardarski Rid, 91-113, ed. D. Mitrevski, Vardarski RidStaromakedonski grad, vol I, Skopje 2005. 113 Courtesy of Z. Videvski, a custodian at the Archaeological Museum of Macedonia, in Skopje. 66 but they are well below the densities recorded on the periphery of site 3. In this respect they are analogous to the satellite clusters of the undated assemblage in sector IX, although in the case of the small Bronze Age assemblage they appear at much greater distances from the “central” cluster. The circumstances in which these satellite clusters were discovered are also different. The larger satellite clusters of the undated assemblage were found in a peripheral and underexploited survey sector, with little background noise. In contrast, the isolated Bronze Age fragments were discovered on later period settlement sites, some of which were multi-period. Intensive survey projects in Greece have shown that the occurrence of prehistoric material in such tiny amounts is not untypical, particularly on sites of larger settlements from later historical periods114. Later building and agricultural activities ensure that only a tiny fraction of the older remains survive in the surface record. It needs to be pointed out that sectors III, V and VII are going to become the most intensely cultivated and inhabited portion of the survey area in later periods. Therefore we allowed the possibility that these were the vestigial remains of sites of similar size and rank to those of site 3 in the monastic complex. Nevertheless we are more inclined to view the more substantial cluster found in the monastic complex as truly representing a larger settlement. Only the finds collected from sites 3 and 5a and from field block 6 can securely be attributed to the Bronze Age. If the isolated shards from sites 6 and 8 are nonetheless treated as possible Bronze Age finds and plotted alongside the securely dated material, there emerges a pattern of one central site in sector I and a series of tiny find-spots marking the locations of smaller or even similarly sized establishments spaced at roughly regular intervals of 300-400 meters along Prisoj’s southern and eastern foot (map III_7). The possibility that settlement was more widespread in the survey area during the Bronze Age shouldn’t be dismissed. As will be shown, a more recognizable fabric category datable to the Late Bronze Age exhibits a very similar distribution pattern of tiny clusters spaced at equal distances on the west bank of the central valley. Even if the possible existence of a network of small, short-lived farms is left aside and we concentrate on the securely dated evidence, the size and the location of the two Bronze Age clusters on sites 3 and 5a indicate a preference for dispersed rather than nucleated settlement. This extremely dispersed scheme of isolated farmsteads or huts is clearly different from that of the period represented by the undated assemblage when a larger hamlet-sized settlement was surrounded by satellite clusters, all situated within a distance of a few hundreds meters from the central cluster. Understandably because of the low chronological resolution we cannot be sure if these clusters formed a contemporary network or if we are seeing the collapsed image of a number of subsequent phases. Acknowledging that this assemblage could theoretically span over nearly 2 millennia, the latter option seems more likely115. Site 3 clearly occupied a strategically important location in a local context, but as will be shown it is much more sheltered in comparison to sites 5a or 6. The importance of this sector in the local communication network was discussed in chapter II. Its main disadvantage is the sheer scarcity of fertile and cultivable stretches of land on the canyon floor and especially on the Vardar’s left bank. It is difficult to imagine a larger nucleated community settling on this location unless it’s assumed that it cultivated the land on the opposite bank. On the other hand, the faint traces found in the southern half of site 5a occupy a type of location that will become highly favored in later periods. This small installation is perhaps the earliest that made use of the 114 Cf. J.L. Bintliff, P. Howard, et al, 1999; C. Mee, H. Forbes, eds. A rough and rocky place: The landscape and settlement history of Methana Peninsula, Greece, Liverpool 1997; J.C. Wright, et al., 579-659, 1990. 115 J.F. Cherry, 19 1982. 67 colluvial sediments covering most of the surface of the survey’s western half. It was positioned on an open and exposed location, in the centre of the agricultural land of the modern village. 150 meters to the north, there is an active freshwater spring. Regarding settlement location, it is the small vestigial remains at Prisoj’s southern foot that will set the precedent for the location of later settlements in the survey area, not the farmstead on site 3. During the following three millennia and up until the present-day, the flat and fertile stretches between the foot of Prisoj and the central valley will remain the most intensely inhabited section of the studied micro-region. III.1.5 The Late Bronze Age (1600-1200 BC, tables 7-9 in Appendix II) Only one fabric group among the collected finds can be associated with this period, though there is an equal possibility that it is later. The chief problem with the dating of this group is that it appears alongside different assemblages. It is thus very uncertain if it represents a separate archaeological period. The greatest concentration of these finds was found at site 3, in the monastic complex (maps III_8b and 9, tables 8, 9, Appendix II). Here it accounts for about 13% of the total collection. It doesn’t follow closely the distribution of the Bronze Age material spreading slightly further to the northwest. If these finds are to be treated as representing a separate phase of occupation on site 3, then evidently the site core has moved to the north. It is even smaller than the Bronze Age site measuring only about 200 sq meters, with a maximum artifact density of 5 fragments per 100 sq meters. The possible core on the opposite bank of the seasonal creek was definitely abandoned. The other half of these finds was found dispersed in several tiny clusters on sites 7, 8 and at the foot of site 9 (map III_10). On these locations the predominant material can be dated either to the transitional Late Bronze-Early Iron Age (1200 to 1000 BC) or the Late Iron Age (7th - 6th centuries BC). There is a possibility that they were accompanied by a small amount of less recognizable Bronze Age finds. However considering that the ratio of fabrics dated to the Bronze Age and the Late Bronze Age at site 3 is nearly 3 to 1, one would expect to find at least a similar situation in the other sectors of the survey. Instead the Late Bronze Age fabric appears more consistently alongside fabrics dated to the Late Bronze-Early Iron Age. Finds that belong to this fabric group at least four times appear on two or three contingent grid units. The artifact density never exceeds the limit of 2 fragments per 100 sq meters or 1 fragment per grid unit. In addition individual fragments appear on isolated grid units in-between the larger clusters. On the south, this series of closely spaced find-spots terminated with a pair of fragments collected by individual transects on field block 94. Apart from field block 5b in the monastic complex, this is the second largest concentration of Late Bronze Age finds collected by the individual field walking transects. It is thus possible that there was a larger concentration immediately to the south of site 7 that wasn’t included in the grid survey. On the north at the foot of site 9, individual fragment were collected from field blocks 113 and 114, but the total grid survey covering the eastern half of these two field blocks didn’t detect examples of this fabric group. This pattern is evidently similar to the distribution of the Bronze Age material and its interpretation is even more difficult given the possibility that this fabric group was a part of a larger assemblage. The fact that these scattered finds appear on later period sites points to an interpretation identical to that applied to the isolated Bronze Age finds. These are the vestigial remains of a dispersed network of small farms or more likely, a single or a few farmsteads shifting their locations along the valley’s western bank. To be sure, the fact that roughly equal 68 quantities were collected from both off-site and later periods’ sites is slightly confusing. As it is possible that this fabric group was accompanied by other fabric groups, one should be reserved about the authenticity of the pattern presented on map III_10. In the case of the Late Bronze Age fabric group, the distance between the more concentrated find-spot on site 3 and the ultra-thin carpet along the valley’s right bank is over two kilometers. If this fabric group truly represents a separate period, then there were at least two separate settlement units in the survey area during the Late Bronze Age. The finds dated to the Late Bronze Age are the first to appear in larger quantities on sectors west of the central valley. They don’t follow the distribution of the few Bronze Age finds, but appear several hundreds of meters to the northeast, in sector VII. These are the narrow, but protected eastern foothills of the Prisoj. During the Late Iron Age, this entire stretch will be permanently inhabited. It is a far less exposed location compared to that occupied by the small hypothetical Bronze Age farm on site 5a. At the same time, it is centrally positioned in the studied micro-region, with an easy access to the fields at the southern foot of Prisoj and on the floor of the central valley. As mentioned in the previous chapter, there are freshwater springs on several locations along the valley floor. It has to be remembered that it is quiet possible that this group of finds is only a part of some other assemblage. Understandably in such a case, the clusters discussed in the preceding paragraphs are artificial phenomena and shouldn’t be analyzed in isolation. It seemed worthwile to carry out a separate analysis, because from its pattern of distribution it appears more like a predecessor than a component of the succeeding Late Bronze-Early Iron Age assemblage. III.1.6 The Dark Ages/Early Iron Age (1200-1000; 1000-800 BC, tables 10-12, graphs 3 and 4, Appendix II) Two rather different assemblages can be dated to this long epoch. Both comprised several different fabric groups, some of which were clearly function specific. Though not lacking in diversity, it is difficult to observe function specific groups in the earlier prehistoric assemblages. These mainly consist of two broad categories of coarse ware; one characterized by the presence of numerous and poorly sorted grains of a black volcanic rock, the other by the large amounts of silvery mica leafs. The few diagnostic pieces indicate that cooking pots and portable stoves were probably made in these fabrics. We’ll find them appearing side by side on a number of sites from different periods, which naturally creates problems of interpretation on multi-period sites. Once introduced, the production of these fabrics will continue with few visible changes for a period of over one millennium. The first of the two assemblages, the one that accompanies the Late Bronze Age fabric group, consists of several fabric groups characterized by the firing technique and the surface finish. It again resembles some aspects of the Late Bronze Age pottery production. Over 400 fragments were classed into some of these fabric groups. To these, one should add about 100 fragments of a coarse ware that were found alongside the rest of the assemblage, but can’t be clearly distinguished from a similar group used in later periods. In total this assemblage comprises slightly over 3% of the total collection (table 10, Appendix II). It is larger than all previous assemblages except for the Mid-Neolithic, but one has to account for the fact that this material was produced over a much shorter period of time. A thin, discontinuous carpet spreads over the greater part of the western bank, from the houses of the modern village to site 9, situated over a kilometer to the north (map III_11a). It is interrupted around the middle of this stretch by the large concentration of Late Iron Age finds on site 8. This is an even more extensive carpet 69 than the one formed by the possible Late Neolithic assemblage on the eastern bank of the central valley. The transect survey recorded very high artifact densities or between 14 and 19 fragments per 1000 sq meters on two pairs of field units, 103-95a and 102a/b situated only about 100 meters apart in the central parts of sector VII (table 11, Appendix II). On the neighbouring field blocks the artifact density decreases to less than 4 fragments per 1000 sq meters, rising again over this limit on field blocks couple of hundreds meters to the north and south of the zone of maximum density. On a field block south of the Skopje-Thessalonica highway and on the eastern periphery of the modern village, it nearly reaches 1 fragment per 100 sq meters. Similar densities were recorded on two field blocks at the foot of site 9, almost half a kilometer to the north of field block 103. The difference between the densities recorded on the pair of field units 103-95a and 102a/b and those on field units to the south and north was at least twofold and it seemed obvious that we were again dealing with a dispersed network of dwellings and outbuildings, gravitating around one or two larger foci. At the same time, one shouldn’t exclude the possibility that there was a continuous carpet of finds, stretching south of the cluster on field block 103 and into sector IV, on some of the fields east of the village. A slightly different pattern emerges when the transect collections are adjusted to represent 100% of the counted material (map III_11b). In case all counted material was collected by individual transects, the concentration on field block 103 gains further prominence, clearly spreading over the neighbouring field block to the north. Density decreases on the presumed second focus on field blocks 102a/b, bringing it closer to the clusters on field blocks to the north of the central site. At the same time another potential focus emerges on field block 80. Along with the neighbouring field block 87, they practically equal in density the cluster on field blocks 95a, 103 and 104. The difference between the two is that the latter appear at the southern periphery of the total dispersal area of the Late Bronze-Early Iron Age material and are not surrounded by field blocks featuring intermediary artifact density. In other words, the concentrations on field blocks 80 and 87 do not produce a site halo and in this respect they are analogous to the satellite clusters of the undated assemblage. The results of the total grid survey chiefly confirmed this pattern. The main settlement during this phase was discovered on the pair of field blocks 95a-103. According to the total grid survey there is a small but very dense concentration of finds in the southwest corner of the grid, on the border between field blocks 95a and 103. It occupies an area of about 1250 sq meters featuring a maximum artifact density of over 50 fragments per 100 sq meters. The artifact density declines rather sharply towards the eastern and northern periphery of field block 103 and it was thought that this marks the settlement limits. But as suggested by the analysis of the transect collections record, the grid survey on the neighbouring field block to the north revealed a dense network of smaller and thinner cores with artifact densities higher than 10 fragments per 100 sq meters, well above the intermediary densities recorded on the periphery of “site 7” (grid 9 on table 12, Appendix II). It extended over the greater portion of field block 104, gradually becoming sparser on the eastern third of the gridded area, but not towards its northern periphery where some of the denser cores were discovered. This extensive area of high and intermediary density must be interpreted as a continuation of “site 7”. Excluding it from the site area would imply that this is a halo zone interspersed with outbuildings, burials and ancient manure. This is highly unlikely because it also implies that a small farmstead occupying a mere 0.1 hectare can produce such an extensive and uneven site halo. We rather adopted the view that these smaller and sparser satellite cores are isolated dwellings, each measuring not more than 150 sq meters. 70 As shown on the thematic map, they are regularly surrounded by at least one or two grid units with average or low artifact densities, repeating the inner distribution on the central cluster (map III_12). In all likelihood they formed an integral but dispersed settlement, with “site 7” representing the focus of the network. The similarities with the pattern exhibited by the more restricted Late Bronze Age fabric group are unmistakable. This network of isolated dwellings and huts almost certainly continues to spread to the north and south of the central cluster. The total grid collections in this part of the survey mainly aimed at documenting the much larger cluster defined as site 8 in AppendixI. Luckily we were rather confident that this site continued over field block 104 and gridded the entire field. This is how we discovered the continuation of “site 7”. However a number of field blocks where the transect survey recorded higher than average artifact density in the southern half of sector VII were left out of the grid survey, because of the lower overall density and the absence of feature shards. Compensating for the lesser degree of survey intensity and the low ground visibility conditions would barely elevate the density records to a site-halo level and this is true only for a few field blocks, including the pair 102a-b about 100 meters east of “site 7” and field blocks 80 and 87, 250 to 300 meters to the south. But knowing that field block 104 featuring an even lower than average density produced almost a dozen small clusters after being gridded, one would expect to see a similar pattern behind the increased densities on field blocks 102a-b, 95a, 80 and 87. This interpretation was seemingly challenged by the total grid surveys to the north of field block 104. Although we gridded an area nearly 4 hectares large, the total collections included but a handful of fragments, dispersed mostly along the periphery of the later site 8. Only on field block 113a, at the southern foot of site 9 and almost 500 meters away from “site 7” do we see another one of the small clusters of Late Bronze-Early Iron Age material. It too behaves like a residential site in miniature, with a maximum of 10 fragments per 100 sq meters and a gradual transition to the off-site, marked by a narrow belt where artifact density ranges between 2 and 6 fragments per 100 sq meters (map III_13). The sudden disappearance of these clusters perfectly coincides with the very large concentration of Late Iron Age finds that constitutes site 8. Because of its size and later chronology, but also because of the fact that the Late Bronze-Early Iron Age finds appear consistently along the periphery of the larger site 8, it was concluded that parts of the dispersed settlement were completely obscured by later activities and the sheer quantities of the Late Iron Age finds. As the results of the grid survey on field blocks at the southern and western foot of site 9 demonstrate, the network of isolated dwellings was predictably becoming sparser as one moves away from the central cluster on “site 7”. The few scattered fragments collected from the grid at the western foot of site 9, on field blocks where the transect survey indicated average and higher than average artifact densities warns us against giving too much weight to the increased densities recorded by the transect survey. Interestingly not a single fragment was collected from field units in sector V, at the southern foot of Prisoj. Because of its dispersed character, the Late Bronze-Early Iron Age settlement was only partly documented and it is difficult to give a precise estimate of the settlement area. Adding up the dozen small clusters revealed on field block 104 to the central cluster on “site 7” would more than double the occupied area. We further must allow that there were at least as many unrevealed clusters on field blocks to the south and to the north, beneath the dense layer of Late Iron Age material. If this interpretation of the survey record is correct, we would have to assume that this settlement occupied a minimum of 5000 sq meters. In all likelihood there are more than a dozen 71 unrecorded satellites, although one also has to allow for possible burial remains, industrial and agricultural activities. By the end of the 2nd millennium BC, settlement in the survey area has almost returned to the pre-Bronze Age level. The settlement that produced these remains was only slightly smaller than the Mid-Neolithic settlement and its descendant on site 11. It was focused on the same area where fragments of the Late Bronze Age fabric group were found. This is sector VII or the eastern foothills of the Prisoj, the narrow stretch between its slope and the central valley. Although not very spacious, this location offers an easy access both to arable land and fresh water. At the same time it is less exposed than the insolated and spacious southern foothills. Focusing on the micro-locations of the revealed clusters, it is possible to observe a preference for slightly elevated and protected positions. This is particularly evident at site 9, where one of the discovered foci was positioned at the very foot of a small hillock that was fortified probably sometime during the Roman Period. “Site 7” too occupies the low ridge that separates the eastern and southern foothills of Prisoj, standing slightly higher than the surrounding fields. III.1.7 The Early Iron Age? (1000-800BC) Tables 13-14, graphs 5-6, Appendix II) The second assemblage datable to the period between 1200 and 800 BC occupies a completely different location in the survey area and features a very different distribution pattern. It consists of several fabric groups, again very unlike the previous assemblage. In fact they are so similar to the fabrics of the Late Iron Age assemblage that it is often impossible to distinguish between certain examples. Yet the core locations of the two assemblages are hundreds of meters apart: the possible Early Iron Age assemblage is concentrated on sites 5a, 5b and 6, in the western survey sectors, while the Late Iron Age material was found concentrated on site 8, partly overlapping with the material datable to the Late Bronze-Early Iron Age (maps III_15b). If the proposed chronology of these assemblages proves correct, it would imply rather dramatic displacements during the long period of transition from the Bronze to the Iron Age. It suggests that the eastern foothills of Prisoj, probably occupied during the Late Bronze Age were almost completely abandoned for a new location, about 350 meters west of “site 7”. This was again however a relatively short-lived settlement, because during the Late Iron Age or the 7th and 6th centuries, the eastern foothills of the Prisoj are again occupied by a settlement far larger than any of its predecessors or successors. As with the assemblage discussed previously, it is possible that a small fraction of the finds collected from the latter site belonged to some of the Early Iron Age fabric groups, but went unnoticed because of the great similarity between the two assemblages. It has to be noted though, that not a single fragment belonging to these fabric groups was found in the fields surrounding the Late Iron Age site. The dating of this assemblage remains problematic. Most of the Late Iron Age fabric groups have been securely dated thanks to the excavations of parts of the mound necropolis, dispersed over most of sector X and parts of sector IX116. The fabric groups for which an Early Iron Age date is proposed, sometimes feature shapes and decorative patterns more characteristic for the first two centuries of the 1st millennium BC. On the other hand, it is not very likely that this material post-dates the Late Iron Age assemblage, as there lack examples of some fabrics characteristic for this period, such as Gray Paionian or Black Gloss ware. Both of these groups D. Petački, Dabici-nekropola od postaro Železno Vreme 71-73, Arheološki Pregled 26, 1986; D. Mitrevski, 1997 92-96; A. Papazovska-Sanev, Keramika od Železnoto Vreme po dolinata na Vardar 19-20, unpublished MA thesis, 2006. 116 72 were present on the Late Iron Age site and examples of Gray Paionian were particularly numerous. Nevertheless the proposed Early Iron Age dating is merely seen as most reasonable at the moment; it is far from definitive. As it appears alongside Roman-Late Roman material on site 5a, it is impossible to determine if certain coarse fabrics belong to this or to the Roman assemblage. In the case of this assemblage, the transect survey records point rather unambiguously to the possible site locations. This is also illustrated by the statistical distribution of the finds (table 13, Appendix II). Only 6% of all finds datable to this period were collected by individual field walking transects and they all came from 10 field units in sectors III and V, at Prisoj’s southern foot. The maximum dispersal area of this material measures about 1.4 hectares, only slightly larger than the entire carpet of Middle Neolithic finds. After the Mid-Neolithic material this is probably the most concentrated group of finds in the survey area. The ten field units where we found Early Iron Age material are not contingent and they exhibit significant differences in the densities recorded by the transect survey. This circumstance eliminates most of the problems encountered during the interpretation of the more dispersed finds, such as those of the undated or the Late Bronze-Early Iron Age assemblage. Thus field blocks 66 and 44 clearly stand out with over 8 and 6.5 shards per 1000 sq meters, 50 to 100% higher than on the rest of the field blocks where this material was present (table 14, Appendix II). This difference is even more pronounced if we calculate the fraction of the collected Early Iron Age finds in the total sum of counted finds (map III_15b). Lower quantities were discovered on field block 31, situated at about equal distance from field blocks 66 and 44, as well as on field blocks neighbouring the latter two, 65 and 53 and 44b-45a. It quickly became clear that the main focus of activity was on field blocks 52 and 66, but we were surprised to discover substantial quantities among the transect collections from field blocks 31 and 44, after the processing of the finds. The few finds collected from field blocks 31, 68 and 69, about 100 meters east and west of the main focus represent the extremely scant site halo, characteristic for this phase. The total grid surveys on field blocks 44-44a, 45a-45b (defined as site 5a-b in Appendix I) and field blocks 66, 52-53 (site 6) confirmed and clarified the interpretation proposed on the basis of the transect survey records (maps III_16, 17). As for most prehistoric assemblages in this survey area, the differences between the quantities picked up by the transect and the total grid collections are worryingly high. This discrepancy must be related to the low obtrusiveness of this material in the overall surface record. Only field block 31 escaped the close scrutiny of the total grid collections. However it has to be emphasized that on this field unit, individual transect collections were more thorough because we noted the presence of a small concentration of Roman pottery of unusual quality for this survey area. Although the presence of an Early Iron Age activity on this field unit is beyond any doubt, the density of slightly over 4 fragments per 1000 sq meters is artificially enhanced by the more intensive collections by individual transects. If collections from this field block were similar to collections from the rest of the field units, the recorded density would havecertainly been lower (map III_15b). Site 6 was already recognized during fieldwork stage and the aim of gridding field blocks 66, 52 and partly 53 was to document the extent of this cluster. Although it is situated less than 100 meters from the last of the village houses and swamped by the large amounts of Late Ottoman-Early Modern off-site debris, the site limits and inner structure came out very clearly. The site area is defined by the threshold of 6.5 fragments per 100 sq meters, with very high densities of over 20 fragments per 100 sq meters, concentrated along a narrow strip on field block 66 and partly extending over field block 52. The maximum density of slightly over 40 73 fragments per 100 sq meters was recorded in the centre of this strip. The entire site area has an elongated shape and measured about 4000 sq meters, excluding the zone of less than 6.5 fragments per 100 sq meters. This is the site halo forming a continuous peripheral belt, at least 15 to 20 meters wide. The rare fragments recognized amidst the total collections from grid 6 located about 100 meters northeast of site 6 mark the limits of the impact zone of this site. The Early Iron Age phase on site 5a-b was only recognized after the study of the collected finds. Field blocks 44-44a, 45-45a were included in the total grid survey because of the visible presence of Roman and Late Roman material, particularly architectural ceramics. The discovery of a small collection of Early Iron Age material from these sites came as a great surprise and only after repeated study of the collected material. Although the finds attributed to the Early Iron Age assemblage form a discontinuous and a relatively thin cluster barely equalling the densities recorded on the periphery of site 6, the fact that they are superimposed by a later period site must be taken into account. Despite the seemingly irregular pattern, the Early Iron Age finds are clearly concentrated in the northeast quarter of the gridded area. It is even possible to observe a larger concentration of finds on a few grid units in the northern half of field block 44, surrounded by a ring of grid units featuring average artifact densities. Into the southern half of field block 44 and over most of field block 44a, the density of the Early Iron Age finds remains low but constant, with only a few sparse fragments collected from field blocks 45a-b, on the other side of the ravine that delimits site 5a from the west. This roughly concentric pattern, along with the fact that the collection comprises a full domestic assemblage indicates that site 5a had another pre-Roman phase dating to the Early Iron Age. Despite the fairly low on-site densities, this cluster extends over a considerable area and it was suggested that it represents the remains of a settlement of similar size to that on site 6. Both sites were nucleated settlements, most probably of the rank of small hamlets. In terms of occupied area they are close to the preceding Late Bronze-Early Iron Age settlement and smaller than the Middle Neolithic and the possible Late Neolithic phases on site 11. This material was evidently highly concentrated on the site areas, with scatters of low density and lacking apparent focus on field blocks 68, 31 and 45a. They exhibit one quality that distinguishes them from all previous settlements in the survey area. On both sites it was possible to observe a continuous peripheral belt measuring a minimum width of 15-20 meters and gradually disappearing into an even more extensive zone of less than 1 fragment per 100 sq meters. To be sure, the phenomenon of site halos could be observed even on the smallest Bronze Age sites or on the individual clusters of Late Bronze-Early Iron Age finds, but they were usually more limited and rarely formed a continuous peripheral rings around the site area. Most of the settlement sites from later periods will feature a similarly extensive site halos. The advantages of the location occupied by site 5a were discussed earlier. This is an exposed location, dominating over the surrounding fields. It is situated at the natural entrance into this micro-region and exerts a close control over access to the Vardar’s Valley floor. In its immediate vicinity, at the very foot of Prisoj there is an active freshwater spring. Site 5a is situated at the western edge of the fertile stretch of colluvium lying at the southern foot of Prisoj. Site 6 on the other hand occupies the very centre of this stretch. It is only 250 meters north of the centre of the modern village, on a slightly higher ground, just outside the village periphery. This is in practice the same location, the modern village being situated closer to the central valley. The occupants of site 6 obviously showed little concern for matters such as protection or even a symbolic, physical delimitation of the settled area. Nothing in the micro-relief suggests that this 74 was a potential archaeological site. Most other sites in this survey area, although lying on flat ground occupy less exposed locations, with access usually limited to one or two directions. The fact that two, at least roughly contemporary settlements occupied different locations within a distance of 500 meters seems difficult to accept. Despite the great similarities in the pottery production, it is very unlikely that a site of a similar size would pass unnoticed amidst the Late Iron Age material. Against the assumption that the two sites are part of a network of farms or hamlets stands their relative size. Both sites spread over an area of at least 4 000 sq meters. In earlier periods settlements in this micro-region feature only slightly greater size and they were the sole focus of activity during the respective time periods. The clusters of contemporary material were usually many times smaller than the central focus. In this case we have two settlements of a similar size, situated at a distance of less than 500 meters measured from the site cores. Moreover site 5a is bounded by an uncultivable, rocky stretch to the west. Both settlements were thus inevitably looking towards the fertile stretch at the southern foot of Prisoj or sectors III and V. The amount of the finds gathered, as well as the character of the material eliminates the possibility that the two sites were in fact, a settlement and its communal cemetery. Admittedly it took a great effort to realize that part of the collection from site 5a consisted of fabrics identical to those encountered on site 6. The material on the surface of site 6 is fairly well preserved and there are many examples of fine, sometimes decorated table ware. In contrast to this, the finds from site 5a are for the greater part badly worn and appear visibly coarser. Only a very small percentage could be classed as fine pottery and exhibited surface treatment and decorative techniques identical to those seen on site 6. Perhaps this reflects certain functional or chronological differences between the two sites, but it seems more likely that it is simply a result of the different preservation states of the two assemblages. The fact that the settlement on site 5a was reoccupied during the Roman-Late Roman Period further contributed to the worn character of the collected finds. III.1.8 The Late Iron Age (7th and 6th century BC, table 15-16, graphs 7-9, Appendix II) The 1980’s rescue excavations on the mound necropolis occupying the Jakupica Ridge (sector X) and partly stretching over the neighbouring ridge of Ramnište117, firmly placed Sopot into the network of Late Iron Age sites in the regions of the Middle and Upper Vardar Valley and the adjacent regions to the east. As was the case with most other mound necropoleis from the period, the location of the contemporary settlement remained unknown. There was no decisive evidence on the two known sites in the immediate vicinity. In both cases these were fortified hilltops dating to the Middle and Late Roman Period. As nearly all known Iron Age sites were found on hill-tops, no one expected to reveal that fabrics datable to the Late Iron Age were by far the most predominant category in the surface record along the western bank of the valley of Sopot, opposite the mound necropolis. Nearly one third or over 5 500 of the 16 525 collected shards can be dated to this period. Moreover they were found almost exclusively in sectors west of the central valley, where their share in the total collection rises to nearly 38%. Even if the material of all previous periods was summed up, it would still comprise less than 15% of the total collection. The second most numerous material in the surface record, the debris discarded in the course of the last two centuries, comprises about 26.5%, although it’s spread over a much larger area. Most of the collected finds matched the material excavated from the mound necropolis, confirming 117 Gj. Petački 71, 1986; D. Mitrevski 92-96, 1997. 75 that the contemporary settlement was situated on low and open terrain, on the west side of the central valley118. Comparing the mean densities recorded by the grid and the transect survey (table 15, Appendix II), it is clear that this material isn’t evenly distributed across the landscape, despite its large dispersal area. The bulk of the material is concentrated on a certain number of field blocks. Only about 150 fragments were collected by individual transects, including both field blocks within and outside the limits of the gridded area. Like the Early Iron Age finds, this material is not very obtrusive in the surface record, regardless of its appearing in very large quantities. As explained in the appendices, it was the character of the finds rather than their quantity that drew our attention. Almost none of the field blocks where the Late Iron Age finds were discovered belonged to the group with very high overall densities. Looking at the distribution of the Late Iron Age finds by field walking blocks, one observes greater concentration of finds on at least four separate locations (map III_18a). These are field blocks 31, 87, 80 and the group of contingent field blocks, 107 and 124-126. The latter group, situated on the right bank of the central valley at the eastern foot of Prisoj was already defined as “site 8”in Appendix I. The other three clusters came to light only after the processing of the gathered finds. Situated on fields adjacent to the village, they were completely swamped by the large mass of Late Ottoman to Early Modern finds, far more visible on the surface. The one on field block 31 is situated at the northern periphery of the village, only 200 meters west of site 6. This location is particularly interesting, because apart from the small group of Late Iron Age finds there are traces of Early Iron Age activity and slightly larger quantities of fine Roman pottery. The other two concentrations are in sector IV, on the adjacent fields east of the village. These field blocks were not covered by the total grid survey, because when revisited the quantity of surface finds was disappointingly low and a considerable portion consisted of material from the past century. Nevertheless even the couple of dozens shards collected from all three field blocks give densities of nearly 1 fragment per 100 sq meters. In fact field block 80 featured an artefact density slightly greater than the density recorded on field block 107 (table 16, Appendix II)! A thin, discontinuous carpet of this material spreads over many field blocks west of the central valley. The artifact density usually ranges between 0.5 and 5.7 fragments per 1000 sq meters. Most of this material was collected from field blocks east of the village, in sectors 4 and 7, but isolated field blocks featuring densities of over 5 shards per 1000 sq meters also appear on the fields west of the village. Interestingly in the sectors east of the valley, on fields over which the mound necropolis was built, only a couple of fragments were collected. In fact the layer of Late Iron Age material rarely crosses into the valley floor, which contributes to the fairly low overall mean density of 1.2 fragments per 1000 sq meters. The overlap between the dispersal of the Late Iron Age material and that of the Late Bronze-Early Iron Age is evident from the results of the transect survey. Even though the thin carpet of Late Iron Age finds spreads over most of the field blocks in the survey’s western sector, there is a visible decline on field blocks to the north and west of the modern village, in sectors II, III and V. Only on field blocks to the east of the modern village in sectors IV and VII do we see field blocks featuring artifact densities higher than the district average. As explained earlier, the increased density on field block 31 is chiefly caused by the intensified individual transect collections. Moreover it is clear that the continuous carpet featuring between 0.5 and 5.7 fragments per 1000 sq meters becomes sparser to the west 118 A. Papazovska-Sanev, 19-20, 2006. Many thanks to Professor D. Mitrevski for positively identifying the survey finds. 76 of the dirt road connecting the villages Sopot and Vetersko. Thus for the second time in the history of habitation in the first survey area, the southern foothills of Prisoj were abandoned in favor of the eastern foothills and the western bank of the central valley. However this observation loses its validity once we adjust the individual transect collections to 100% of the counted material (map III_18b). While the artificial character of the cluster on field block 31 is unmasked, the concentrations revealed on field units 80 and 87, 18a and 18b at the southern foot of Prisoj become more substantial, equaling or exceeding field blocks 107, 124-126 in artifact density. The implication is that apart from site 8, there were at least 3 other, equally dense or denser clusters: one on the Early Iron Age site 5a, two on previously uninhabited locations in sectors II and IV, in the lower part of the valley. The reasons for which this interpretation was rejected were elaborated in Appendix II. Most problematic of all is the fact that the counts by individual transect units grossly underestimated the true quantity of surface material on field blocks in sector VII, where site 8 is situated. This explains the high sample fraction of the collections from these units in comparison to the transect collections from field blocks surrounding the modern village. Site 8 spreading over field blocks 104, 107, 126 and 125 doesn’t stand out by the greater density of finds when compared to the clusters found on field blocks adjacent to the village. However it became immediately clear that it consisted mostly of Iron Age finds and that it spread over a much larger area. The clusters of Late Iron Age finds on field blocks 18a, 18b, 45a, 80 and 87 are either limited to the field boundaries or continue to spread on the neighboring fields in very small quantities. Even if it is assumed that they occupied most of the field blocks’ area, which is very unlikely, they’ll still occupy less than 3000 sq meters. In comparison, the detailed analysis of both the transect and the grid survey results show that site 8 occupies over 3.5 hectares, including the intervening overgrown stretches, but not the surrounding zones of lower artifact density. This pattern of one larger site core accompanied by several satellite clusters greatly resembles the distribution of the assemblage datable to the Late Bronze-Early Iron Age, spreading roughly over the same part of the surveyed landscape. The only obvious difference between the two assemblages is one of scale; the clusters of finds datable to the Late Iron Age are several times larger. To illustrate, the smaller satellite Late Iron Age clusters are almost as large as “site 7”. The total collections by regular grid units clarified the limits and the size of site 8 (map III_19). It is defined by artifact densities higher than 10 fragments per 100 sq meters. This zone covers most of field blocks 107 and 126 and smaller portions of field blocks 104 and 124. Thus the core of the site was not on field blocks 124-126 as suggested by the transect survey collections, but on field block 107. On the core of this site, the total grid survey recorded over 100 fragments per 100 sq meters, with a maximum density of 220 fragments per 100 sq meters. This is an almost continuous zone mostly concentrated in the southwest half of field block 107, with another smaller peak on field block 126. It is most certain that the cluster continued to spread under the overgrown stretch separating field blocks 107 and 126. To the southeast and into the valley floor ground visibility conditions precluded a total grid survey, but the individual transect collections confirmed the absence of Late Iron Age finds. To the northeast across the dirt road connecting Sopot and Vetersko, ground visibility conditions were better and we conducted total grid survey on field blocks at the southern and western foot of site 9. Although the transect collections recorded artifact density close to the mean overall value, the regular grid survey revealed only a thin, discontinuous off-site scatters giving artifact densities lower than 2 fragments per 100 sq meters. To the northeast and southwest Late Iron Age material decreases 77 more gradually and defining the site limits is not so straightforward. The threshold of 10 fragments per 100 sq meters was chosen because such densities rarely appear on the more distant field units in sectors III and V. We took the total dispersal area of the Late Iron Age material to represent the impact zone of the settlement on site 8. It is possible to distinguish a narrower zone with artifact densities ranging between 8 and 10 fragments per 100 sq meters and a farther off-site zone characterized by a discontinuous carpet where artifact densities rarely exceed the overall mean value of 7.5 Late Iron Age shards per 100 sq meters. The first zone extends asymmetrically to the northeast and southwest of the site area for about 100 and 150 meters from the site edges. As we saw it cannot be followed along the shorter axis of the site area. We identified this zone with the site halo; it is analogous but far more extensive than the peripheral zone surrounding the Early Iron Age site. Unlike the latter, the Late Iron Age settlement produced a far more extensive off-site, stretching over most of the western survey sectors. The furthest isolated Late Iron Age fragments were collected from grid units west of the modern village, at a distance of over 900 meters from site 8. Such an extensive off-site carpet will only be produced during the RomanLate Roman and the Late Ottoman-Early Modern periods. Similar but far less extensive scatters were produced by the smaller settlements on sites 7 and 11. The total grid surveys on sites 5a-b, 6 and on field blocks 62-63 and 68 demonstrated that this farther off-site zone is not a continuous carpet of surface finds. It rather consists of ultra-thin scatters of less than 2-3 fragments per 100 sq meters separated by sterile stretches, plus small satellite clusters on which artifact density barely approached the values on the peripheral belt of site 8. Two such clusters were discovered on the settlements from the preceding Early Iron Age period, on sites 5a and 6. Although it is possible that these small clusters are slightly inflated by the inclusion of coarse pottery that belonged to the Early Iron Age assemblage, it is highly probable that there was either continued or renewed activity on these two sites. We suspect that clusters of similar size elevated the artifact densities on field blocks 80 and 87. With the means presently at our disposal, we can only speculate about the potential sources of this material. A certain portion must have originated from enclosed deposits, especially in the case of satellite clusters, such as those on sites 5a and 6 and the potential clusters on field blocks 80 and 87. In this respect the distribution pattern of the Late Iron Age material repeats the distribution of the undated assemblage. But on the other hand, it is less likely that the entire corpus of this material comes from enclosed deposits. The total dispersal area is simply too extensive and along certain sections, the finds are highly dispersed. One wonders if these are not the traces of intense manure. It is symptomatic that they are mostly limited to the most fertile stretches of land, at Prisoj’s southern foot. To sum up this interpretation of the statistical and spatial distribution of the Late Iron Age finds, it can be concluded that the main focus of occupation during this period was the narrow stretch of land between the eastern foot of Prisoj and the floor of the central valley. It is possible that there was another focus of occupation further downstream, east of the modern village, but with the exception of fields 80 and 87, there is very little evidence on the rest of the field units surrounding the area omitted from the total survey. The Late Iron Age settlement was of a much greater rank and size than all previous and later periods’ settlements in this microregion. It stretched over an area of almost 3.6 hectares, several times greater than the size of some earlier prehistoric settlements. By local standards this is a settlement of average size. It should be put in the rank of small to medium size villages, while most of the earlier settlements were of the rank of farms or hamlets. The difference in scale is also made apparent by the sheer 78 amount of the material and the wide dispersal area. For the first time in this region, there appears an extensive off-site zone, spreading almost continuously along the entire length of the valley’s western bank. This is a thin carpet of badly worn fragments, sometimes completely disappearing, sometimes becoming slightly denser, as on fields 87 and 80. Whether resulting from certain nonresidential activities or as agricultural manure, it is a clear indicator of the relative size of the local community and its mark on the surrounding landscape. A wider dispersal was also observed for some earlier assemblages, but in these cases, the carpet of off-site finds was much thinner and patchier. They were moreover mainly limited to one sector of the surveyed area, while the Late Iron Age finds spread over much of the western half of the survey. Considering the fact that there were only a very few Late Iron Age finds in the sectors east of the central valley, it becomes evident that most of the habitation activities were concentrated on the opposite west bank. However the less fertile stretches on the eastern bank were not simply excluded from the area occupied by this community. To complete the picture of the Late Iron Age landscape, one must include the extensive mound necropolis, occupying most of sector X and parts of sectors VIII and IX. Recall that the groups of mounds were mostly concentrated on the Jakupica Ridge, the eastern watershed line of the central valley and the boundary separating the areas of the modern villages Sopot and Novačani. It is a fairly imposing location and the line of mounds atop the low crest would have presented a clear message to anyone entering the area from the east. The mound necropolis marked the eastern limit of the territory claimed by the Late Iron Age community119. Thus during the Late Iron Age, the entire lower half of the surveyed valley was incorporated into the communal territory. Furthermore there was a clear separation between the spaces for the living and the dead, the western, more humid and fertile and the eastern, drier and less fertile bank. The settlement itself was wisely positioned in the northern corner of this triangular territory, away from the spacious, but exposed southern foothills of Prisoj, which must have been given away to cultivation. This is a relatively sheltered, withdrawn position especially when compared to that of site 5a or site 6. At the same time however, it is closer to the geometric centre of the study region, offering quick access to both banks of the central valley. Roughly the same location was occupied by an earlier settlement, datable to the Late Bronze and Early Iron Age, but the micro-locations of the two are not identical. The much larger and nucleated Late Iron Age settlement was positioned on a slightly lower and gentler terrain, closer to the valley’s floor. III.1.9 The problem of the Early Antique layer (Classical Period, 5th and 4th century BC) The Late Iron Age assemblage consists of a considerable number of fabric groups. Including the coarse wares, there are nearly 20 different fabric groups! A few of these fabrics consisted of a relatively small number of finds and judging by the fabric properties,they were most probably imported. But even so, the number of different fabric groups is extremely large for such a short period of time of less than two and half centuries. Assemblages from earlier, equally long or much longer time periods consisted of not more than several fabric groups. The proposed dating of this large group of fabrics is based on the prominence of certain shapes and Cf. S. E. Alcock, J. E. Rempel, 27-46, eds. P.G. Bilde, V.F. Stolba, 2006; P. Novaković, Detecting territoriality and social structure in the Bronze and Iron Ages. GIS and the hillforts in the Kras region, 101-115, ed. B. Slapšak, On the Good Use of Geographic Information Systems in Archaeological LandscapeStudies, COST Action G2, Ancient Landscapes and Rural Structures, 2001. 119 79 techniques but above all, on the results of the rescue excavations on the mound necropolis. As mentioned before, these mounds were dated not earlier than the late 8th and not later than the end of the 6th century BC. The small number of diagnostic or decorated shards have close parallels among the material from other sites dated to this period and can hardly be dated earlier or later than the Late Iron Age. Nevertheless it is difficult to accept that such a variety of fabrics was produced within the suggested chronological frame. It is somewhat problematic to group all these fabrics into wider categories, because the differences between two fabric classes are often very subtle. But focusing on certain fabric properties, such as surface treatment, firing and texture of the paste, it is possible to distinguish between two very general categories of plain fabrics. The first, possibly earlier group is characterized by a deep spatula polish, compact but brittle paste and uneven firing. The second, possibly later group, lacks the characteristic deep surface polish, has a more granular texture, while firing is generally more even. The fragments of this group of fabrics have slipped and superficially smoothed surfaces. Although the suggested differences are not always clear, they aren’t insignificant and it is quite possible that they truly reflect chronological differences. Unfortunately the two groups of fabrics are completely overlapping and spread over most of the site area, the presumed later group being only slightly more confined and slightly less numerous (map III_20). It never appears north of the site limits and it is mostly confined to the fields east of the modern village. If this suggests a possible contraction of the settlement during a later phase, it was of a rather small scale. Nearly the same amount of finds can be classed into either groups of fabrics or more precisely, 1 239 in the group of granular, slipped fabrics and 1434 in the presumably earlier group of polished and compact fabrics. The size of the settlement didn’t change significantly, although the total dispersal area evidently shrank, as examples of the granular slipped fabrics rarely appear on the fields west of the village. Graph III_1: Distribution of the Late Iron Age assemblage by fabrics 60.00% 50.00% 40.00% Compact 30.00% 20.00% Grainy 10.00% Coarse 0.00% Other -10.00% Grid units Field blocks It is another question if the presumed later phase actually represents the Early Antique period or if it simply marks the later part of the Late Iron Age production. One of the more characteristic pottery types, known from excavations on a number of 5th and 4th century sites in the country is Black Gloss ware120. Only a few black gloss examples were found in the survey area, all collected from the central sections of “site 8” and it is equally possible that these too date to the late 6th century BC. Another type of pottery characteristic for this period is Gray D. Vučković-Todorović, Antička Demir Kapija, 229-267, Starinar XII, 1961; I. Mikulčić, Iliro-Arhajski grobovi is okoline Štipa 197-209, Starinar XIII-XIV, 1965; V. Sokolovska, Isar Marvinci i Povardarieto vo Antičko vreme, Skopje 1986; Rescue excavation in 2009 on the latter site unearthed considerable amounts of Late Archaic and Classical black gloss pottery, most probably locally produced. 120 80 Paionian ware, deeply rooted in the local, prehistoric traditions. On larger sites gray-fired pottery imitating shapes from Greece and the Aegean have been attested in 6th century layers121. Although present in respectable amounts on the Late Iron Age site, the gray-fired fabrics exhibited typically Iron Age features, such as deep surface polish and incised or notched decoration. The fabrics classed into the granular, slipped group almost never exhibit features that can be particularly related to the Early Antique Period. As with the Gray Paionian group, there lack the typical Classical shapes and techniques. Most of the diagnostic fragments featured shapes characteristic for the local prehistoric traditions. Examples of strap handles are an exception, but these have been attested in 6th century deposits during the excavation of the mound necropolis. Thus there is in general very little positive evidence of occupation after the end of the 6th century BC on site 8. The principle problem with the collected material is the long persistence of certain shapes and techniques that were probably already well-established in the local Iron Age production122. Pottery made of the same or very similar raw material appears alongside typical Late Roman tile featuring only very subtle changes in the shapes and production techniques. As will be shown, fabrics datable to the Hellenistic Period are also unrelated to the wares known from excavations on larger sites. In regards to this, one should allow the possibility that occupation on site 8 continued into the 5th and 4th century BC. At the moment it is impossible to arrive at a nonarbitrary conclusion. It is noteworthy to mention that research in the past several decades has shown that, although many aspects of material culture have changed, life continued on most of the known Iron Age settlements after 500 BC, particularly in the middle and upper stretches of the Vardar Valley. It is in any case certain that the by the late 4th century BC at latest, there was another dramatic shift in the settlement history of the survey area. What remains unclear is whether site 8 and the entire micro-region were completely abandoned by around 500 BC or sometime during the ensuing period of Early Antiquity. The later Hellenistic settlement was of a very different scale and seemingly shows little respect for the Late Iron Age landscape. III.1.10 The Hellenistic Period (late 4th – late 1st century BC, table 17-18, graph 10, Appendix II) According to the survey results, the studied micro-region experienced dramatic cultural and demographic changes by the beginning of the Hellenistic Period, if not earlier. The amount of finds datable to the Hellenistic Period is sharply reduced and they feature a very different distribution pattern. Although many characteristics of the local pottery production survive, the differences with the Late Iron Age material are more than apparent. Actually one can confidently argue that there was an obvious decline in the overall quality of pottery production. Certain techniques for example, such as fully reduced firing are almost completely forgotten, while firing and surface treatment is much humbler. In total only about 225 fragments can be broadly dated to the Hellenistic Period or 1.35% of the total collection. Apart from the Bronze Age, this is the smallest of the assemblages discussed so far. It is at least several times less numerous than the much earlier prehistoric assemblages, indicating drastic decline of settlement and population. V. Sokolovska, 40-54, 80, 1986; for the “phenomenon” of local gray-fired pottery, V.StojanovićAnderson, Stobi, the Hellensitic and Roman Pottery 14-18, Princeton 1992. 122 M. Šurbanovska et al. Arheološki istrazuvanja-Stobi 2001, 271-282, Macedonia Acta Arhaeologica. 18, 2008. 121 81 Assuming that survival rate diminishes with the passage of time, this dwindling of the quantity of Hellenistic finds becomes even more significant. But despite the small size of the collection, the Hellenistic material is spread over a considerable area (table 17, Appendix II). According to the field block survey, it occupies a number of field blocks in the central parts of the Ramnište Ridge and along the left bank of the central valley. It even partly spreads into sector X, on the neighbouring Jakupica Ridge (map III_21a). Moreover the transect survey collections indicate a relatively even distribution. On at least seven field units the artifact density is at least twice as high as the mean district value of 2 fragments per 1000 sq meters (table 18, Appendix II). The highest density of 6.5 fragments recorded on field block 167 is not much higher than the densities recorded on field blocks 164 and 188 with 5.3 and 5.7 fragments per 1000 sq meters. Thus the collections by individual field walking transects point to at least three potential clusters: on field blocks 167-168, but also on field blocks 186a-188 and 163-164. At a first sight, not much is changed when the number of finds collected by individual transect units is adjusted to represent 100% of the counted material (map III_21b). However two of the concentrations of Hellenistic material not only retain their integrity, but also feature artifact densities much higher than the rest of the field units with Hellenistic finds. On field blocks 167 and 188, the artifact density could increase to over 17 and nearly 25 fragments per 1000 sq meters, between 4 and 5 times the mean district value. On the rest of the field units, especially along the crest of the Ramnište Ridge, the artifact density rarely exceeds 5 fragments per 1000 sq meters. There is a somewhat more substantial increase in sector VI, along the left bank of the creek, but note the fairly small size of these field units. Field blocks 186a-b and184 feature between 7 and 8 fragments per 1000 sq meters, considerably higher than the district average. Whether analyzing the adjusted or the raw records of the collections by individual transect units, the Hellenistic assemblage is characterized by a fairly contracted zone of lower than average artifact density. It was documented on field blocks along the eastern and northern periphery of the area over which this material is dispersed. Obviously the focus was on the crest of the ridge, in its central part and on the valley’s floor. Unfortunately only after the processing of the collected material did we become aware of this distribution pattern. Prior to the study of the material, only field blocks 167-168 stood apart by the higher overall artifact density and the grid survey was limited to these field units. The total collections covering most of field block 167 and the southwest quarter of field block 168 confirmed the existence of a larger cluster dating to the Hellenistic Period (map III_24). It was situated in the central part of field block 167, possibly continuing for a short distance to the south. This was a compact cluster measuring at least 2000 sq meters, but if we consider only the zone featuring over 2 fragments per 100 sq meters, the occupied area will shrink to about 1800 sq meters. The cluster consisted of two cores featuring artifact densities of over 20 fragments per 100 sq meters. One was situated in the central part of the grid, the other on its southern edge, partly cut by an artificial terrace. They are surrounded by a narrow belt of artifact densities lower than the district average of 1.5-2 fragments per 100 sq meters of total grid survey. This peripheral belt wasn’t revealed in its entirety, though it certainly spread for a distance of 10 to 20 meters from the site edge, mostly to the north and east. It seems as if the distribution of the Hellenistic material is a mirror-image of the distribution pattern of the possible Late Neolithic assemblage. It too consists of one central cluster and at least three smaller, satellite clusters. As already concluded, the Hellenistic assemblage is slightly less concentrated and less numerous. The main focus of occupation 82 spreads over an area of 1800 sq meters, excluding the peripheral zone and the possibility that it extended slightly further to the south. Clearly this settlement was of the rank of a farm indicating a strong demographic decline from the period of the Late Iron Age. In fact site 12 is smaller even in comparison to some of the earlier prehistoric settlements. There remains however the difficulty of interpreting the “satellite” clusters. The main issue is whether they represent residual remains of clusters of similar size to that of site 12 or if these are traces of less intensive non-residential activities, such as dispersed burials, field-sheds, rubbish pits, votive deposits or traces of ancient manure. If the former case was true, it would imply that the Hellenistic settlement was at least two or three times larger, approaching the size of the earlier prehistoric settlements. Judging solely by the results of the transect survey this is absolutely plausible. At the same time, it has to be stressed that on most of the field units where the satellite clusters of Hellenistic material were found, the visibility conditions and the recent history of land-use were similar to that on field block 167. Even after compensating for the worse visibility conditions on some of the field blocks and the lesser degree of survey intensity, the density of Hellenistic finds on these clusters would barely equal the artifact density on the periphery of site 12. This was to a certain degree confirmed on similarly dispersed assemblages in the western survey sectors, such as the LBA-EIA or the Late Iron Age assemblage. In these cases, total grid survey was carried out on both the central and the satellite clusters and the results clearly confirmed that the satellite clusters are indeed much smaller. This supports the initial interpretation that during the Hellenistic Period settlement in the surveyed basin consisted of one central focus and a number of satellite clusters or an extensive and discontinuous off-site carpet marking the community’s impact territory or the settlement area, a zone where apart from residential, agricultural, industrial, sepulchral and other practices took place. However it is exactly at this point that we run into a considerable problem: if the total dispersal area marks the communities’ inner territories, then how to explain the fact that the many times larger site 11 produced an impact zone of roughly equal size to that of the Hellenistic farmstead? It is mainly because of this reason that we remain doubtful about the completeness of the map of Hellenistic clusters in the survey area. It is possible that there existed at least one other residential site on field blocks 163-164 or on the valley floor, on field blocks 186a-188. It is certain that these potential clusters weren’t larger than the cluster on site 12. In this respect the Hellenistic settlement is much more similar to the dispersed Late Bronze-Early Iron Age settlement on the opposite bank of the central valley. Note than in both cases the “central” clusters measured less than 2000 sq meters. The quantity and the quality of the finds on these locations indicate that these were truly larger clusters, perhaps representing the foci of dispersed settlements. Even accepting the possibility that the settlement on site 12 was accompanied by a few other residential sites, it can hardly challenge the observation that there was a serious population decline during this period. The Late Iron Age village on site 8 was abandoned and replaced by an individual or a small group of farmsteads. Apart from the great demographic reversal, the Hellenistic Period brings about another, no less significant change in the local settlement history. After a longer period of continuous occupation of the western bank, the new Hellenistic settlement moved to the opposite eastern bank. There are no securely identified Hellenistic finds in sectors west of the central valley. Like most of the previously discussed assemblages, they are completely confined to a single micro-topographic unit, in this case the Ramnište Ridge and the lower terraces on the valley’s left bank. As explained in the previous chapter, this part of the studied micro-region is drier and less fertile. In most periods of the past, the main focus of occupation was on the higher western bank, covered with Quaternary deposits and richer in 83 freshwater springs. The choice to settle on the eastern bank, entirely covered with Neogene deposits is understandable for the Middle-Neolithic settlement. It is in accord with the known settlement pattern during this period and the preference for the lighter Neogene soils can be explained bearing in mind that they’re easier to cultivate. Moreover the Mid-Neolithic settlement and its successor on site 11 occupied the tip of the Ramnište Ridge, which represents the centre of the surveyed basin and offers a ready access to the floor of the valley. But it is somewhat more difficult to understand the perspective of the Hellenistic community. Focused around site 12 and the central portions of the Ramnište Ridge, it occupied a peripheral part of this micro-region, hundreds of meters away from the main natural artery and the most fertile stretches of land. This is an open unprotected location, lacking any topographic integrity. The only protection it offered is its relatively withdrawn position. Judging by the appearance of certain fabric groups, occupation on site 12 continued into the Roman Period. With such a low chronological resolution, it is impossible to know if this small settlement existed continuously or if it was abandoned sometime during the Hellenistic Period and briefly reoccupied in Roman times. Noteworthy is the similarity between the predominant plain fabrics of the two periods. III.1.11 The Roman to Late Roman Period (AD 1st-end of 6th century, tables 19-27, graphs 12-21, Appendix II) Regarding local pottery production, the transition from the Hellenistic to the Roman Period is known from the excavations on larger urban centers, such as Stobi or Heraclea Lyncestis123. It is unfortunate that nearly nothing has been published on the subject of pottery from the Early and Middle Imperial periods since the publication of Stojanović-Anderson’s study of the Stobi material. In the meantime more pre-4th century finds have been unearthed from Stobi, Heraclea Lyncestis and other urban centres in the south, but also from Scupi where small parts of the Early and Mid-Imperial colony are finally coming to light124. However local pottery production in the survey area rarely followed the more recognizable Roman shapes and decorative techniques. In fact in many aspects it follows the local traditions established during the Iron Age. There were onlya few close parallels with the material from Stobi or Scupi, mostly Late Roman examples. Often the only indicator that one is dealing with material from the Roman Period was the accompanying material, primarily brick and tile. In the southern half of the Republic of Macedonia, architectural ceramics appear already in the Early Hellenistic Period, but they are attested mostly on larger urban centers. Only in the Roman Period do they become widespread, often being by far the most predominant type of surface find125. Hence brick and tile were a very useful chronological indicator and greatly helped in the dating of certain fabric groups. V. Stojanović-Anderson 1992; E. Maneva, Nekolku tipovi i formi na Elinistička i Rimska keramika od Herakleja, Bitola 1979; M. Šurbanovska, Rimska keramika od Stobi 151-162, Macedonia Acta Arhaeologica 17, 2006. 124 J. Wiseman, et al. 1973, 1975, 1979; I. Mikulčić, Stobi, an ancient city, Skopje 2003; E. Nikolova, Stobi-zapadna nekropola 247-270, Macedonia Acta Arhaeologica 18, 2008, M. Šurbanovska et al. 271-282, 2008; L.Ðidrova, Heraklea Lyncestis; problem na hronologija i interpretacija 269-291, Macedonia Acta Arhaeologica 15, 1999; L. Jovanova, Skupi-Colonia Flavia Scupinorum: topografija i urbanistitchka struktura na gradot od I do III vek 197-222, Macedonia Acta Arhaeologica 18, 2008. 125 V. Lilčić, Antička gradežna keramika vo R. Makedonija, Skopje 1996. 123 84 In local production more recognizable shapes and decorations begin to take hold during the Late Roman Period. In fact most of the Roman fabric groups are dated between the 4th and the 6th centuries AD. Excluding the coarse fabrics, which were only broadly dated to the Roman Period, only a few fabric groups can be dated earlier than the 4th century. But this could quite possibly reflect the fact that Late Roman pottery is more recognizable than plain pottery from the Early and Mid-Imerial phases126. As the majority of Roman fabrics from the first survey lack horizontal stratigraphy, it was found impossible to define a separate Early to Mid-Imperial assemblage. Only on one or two site locations did the more recognizable Late Roman fabrics appear unaccompanied by other fabric groups broadly datable to the Roman Period. Needless to say the coarse wares are particularly troublesome in this respect, for they hardly change even when compared to later prehistoric periods. Similarly there lacks even a most basic chronology for the architectural ceramics. Perhaps it is not chronologically insignificant that the various fabric groups of brick and tile exhibit a clear horizontal stratigraphy. Unfortunately the pottery fabrics don’t necessarily follow the divergent distribution patterns of the architectural ceramics. It is nevertheless apparent that almost all of the architectural ceramics are accompanied by Late Roman material. Given the circumstances, the Roman material was grouped in two chronological classes: a very broad category of Roman material datable anywhere between the 1st and late 6th century AD and a narrower category of Late Roman material datable between the 4th and the 6th centuries. The latter group includes fabric categories that can be dated through direct parallels with known Late Roman material, while the rest of the fine fabrics and all coarse fabrics were grouped into the broader class of Roman finds. Again the greater portion of the coarse fabric groups is almost certainly Late Roman. In fact only three fabric groups exhibit a consistently divergent distribution pattern from the Late Roman finds and only for two of these was it possible to find some parallels among material from excavated Mid-Late Roman deposits. As a result of this crude chronology, at least two highly important problems of the local settlement history remain unresolved. Firstly, it remains unclear if life continued uninterrupted after the Roman conquest. The earlier Hellenistic settlement was clearly occupied sometime during the Roman Period, but it’ll remain uncertain if the site was reoccupied or continued its existence after the Roman conquest. The second issue remaining unanswered due to the poor understanding of the chronology of the finds is the settlement and population dynamics during the various phases of the Roman Period. In this respect it is important to bear in mind that even if there is a more substantial Early Imperial layer, it will be diminished by the super-imposed and more recognizable Late Roman phase. A total of about 1450 fragments can be dated both to the Roman and the Late Roman Period. This is the third most numerous chronological group in the survey area, after the Late Ottoman and Early Modern and the Late Iron Age material. It represents about 8.75% of the total collection. However when comparing the amounts of Roman and the material from earlier periods, a number of additional factors need to be considered. The Roman Period lasted two to three times longer than the Hellenistic Period or the two phases of the Iron Age. If one counts only the finds that can be strictly dated to the Late Roman Period, their sum will barely exceed 1000, only slightly greater than the amount of material dated to some of the prehistoric periods. There is also the factor of decreasing survival rate of the surface finds with the passage of time127. Being at least half a millennium later than the prehistoric periods, the material from the 126 127 Cf. D.K. Pettegrew, The Busy Countryside of Late Roman Corinth, 743-784, Hesperia 76, 2007. J. Bintliff, P. Howard, A. Snodgrass 146-147, 1999; E. Neustupný 51-57, 1998. 85 Roman Period should have better chances of survival in the surface record and even obscure the presence of earlier layers. Finally one has to bear in mind that around 700 or nearly 50% of all Roman finds comprised brick and tile. This was a novelty in the local material culture, mainly introduced during the Roman Period. The architectural remains of earlier periods including the Hellenistic, almost never survive in the surface record. Thus the number of Roman finds is not only increased due to the temporal factor, but also because of the more monumental character of the material culture during this period. Put in simpler terms, more goods of fairly imperishable materials were being produced during the Roman Period. A Roman structure, even if only partly built of firm materials would predictably produce a larger surface cluster than a prehistoric dwelling made of wood and mud. Now, if the architectural ceramics are excluded from the Roman assemblage, though in certain circumstances it is possible to argue that the tile actually conceals the pottery finds, the quantity of all Roman finds will actually equal the quantity of the Middle Neolithic or the Early Iron Age assemblage. Further in the text, while discussing the relations between Roman and earlier material on multi-period sites, we’ll reveal another indicator of the actual sparseness of the Roman material in this survey area. For the moment, we can only comment on the very modest range between the minimum and maximum on-site density recorded by total grid survey (table 19, appendix II). The maximum on-site density of only 20 fragments per 100 sq meters is strikingly low. In fact it stands higher only to the densities recorded on the vestigial Bronze and Late Bronze Age sites! It is difficult to attribute this fact to the slightly unfavorable ground visibility conditions on site 5a at the time of the total grid collections, because we’ll see that these low on-site densities appear persistently on all Roman sites in this survey area. Understandably this small difference between the densities on the site core and periphery complicated the definition of site limits, especially when clusters of intermediary densities were involved. Despite their relative sparseness, the finds datable to the Roman Period are among the most widespread class of finds in the first survey area. In contrast to the assemblages dating to earlier periods, this material was found virtually in every survey sector. Because of the large total area of dispersal, instead of analyzing the transect survey results block by block we grouped them into three basic zones of various artifact density (map III_25a). Of the 104 field units on which the transect survey recorded Roman material, about 80 featured artifact densities lower than the mean overall value of 3.45 fragments per 1000 sq meters. They are found dispersed across the entire survey area with a tendency of clustering around field blocks featuring higher artifact density. Depending on their micro-locations and the character of the collected finds, they were interpreted as the remains of the off-site or in some cases, site halos. The second zone featuring artifact densities higher than the mean overall value, but lower than 39 fragments per 1000 sq meters comprises about 20% of all field blocks with Roman material. They appear isolated or in pairs in most survey sectors, but they are particularly concentrated in sectors west of the village, near the monastic complex and on the easternmost survey sector, in the area of the Late Iron Age mound necropolis. Here it is useful to make a further distinction between field blocks featuring between 3.5 and 7 fragments per 1000 sq meters and those featuring over 10 fragments per 1000 sq meters. The latter group is characterized by artifact densities at least 3 times the overall mean value and approach the minimum on-site densities recorded by the grid surveys (tables 20-21, Appendix II). Only 4 field blocks feature artifact densities higher than 40 fragments per 1000 sq meters, but it has to be noted that in these cases the increased density could very well result from the small size of the field units (field blocks 4a, 4b in sector I) or the more thorough individual transect collections (field block 31). The fact that they appear 86 alongside or close to field blocks belonging to the second zone deepens the overall distribution pattern of Roman material as revealed by the transect survey. Considering only the field blocks with densities higher than 10 fragments per 1000 sq meters as potential site locations, there is a clear concentration of activity in the peripheral parts of the basins: on field blocks in the central part of sector I at the foot of the fort “Kale”, Sopot, on field blocks 44-45 in the northwest corner of the survey, but also on the terraces southwest of the modern village and finally, on the pair of field blocks 133-134 and 142-143 along the survey’s eastern limit. The central portions of the basin are conspicuously vacant. This was rarely the case in earlier periods and it could be related to profound changes in the local settlement pattern and agrarian relations. We saw however that the records of the individual transect collection need adjustments, if we are to compare the artifact density on field blocks across various survey sectors. This is especially necessary for the Roman to Late Roman assemblage, where on a few field blocks, individual transects collections amounted to total collections from larger units or were repeated in more than one occasion. If we assume that they were limited to the number of finds counted during the quantification campaign, we’ll end up with a density map that is visibly modified (map III_25b). On certain field units in the western survey half, there is a considerable increase in the density of Roman to Late Roman material, while on others, including the clusters of field blocks in the easternmost sector, it declines below the overall average. The range between the minimum and the maximum density increases considerably and it becomes necessary to introduce separate mean density values for each survey sector. The concentrations on field blocks in sectors VII and X become invisible if the overall density is compared and Roman material becomes visibly concentrated on the field blocks in sectors I through III. What this analysis confirms is that the clusters on field blocks 133-134 and 142a/b-143 in sector X are thinner and also probably smaller than the clusters by the monastic complex or the northwest corner of the survey area (graph 11, Appendix II). Although we lack definitive evidence, it is nearly certain that the network of potential Roman sites was at least partly contemporary with the three fortifications built at the corners of the surveyed basin. The one above the monastic complex known in the literature as “Kale”, Sopot has been published some decades ago and dated to the Roman Period with a possible prehistoric phase128. The other two labeled site 9 and 10 were discussed in Appendix I. The few finds found on the barren surfaces were identified as Late Roman tiles and a similar date was suggested by the masonry and the layout. It is noteworthy that each of the field blocks featuring higher density of Roman material is positioned in the immediate vicinity of some of the forts. Only at the foot of site 9 did there lack evidence for occupation during the Roman Period, although even here, the density of Roman finds was slightly above the overall mean value. The detailed sector by sector analysis of the transect survey results and the total grid surveys on some of the field units with artifact densities higher than the survey’s average largely confirmed these preliminary observations. A few unperceived but important details came to light, but in general the existence of genuine sites was confirmed on most field blocks featuring artifact densities at least 2 to 3 times the mean overall values recorded by the transect survey. But at the same time, we regretted the fact that some of the potential sites were left out of the total grid survey. It has to be recalled that at the time of the total grid survey we had no idea about the chronological composition of the finds and we were chiefly guided by the overall artifact densities. As a result some field blocks in the eastern survey periphery, where the off-site carpet I. Mikulčić, 99-100, 1982; as mentioned, it is very probable that the ceramic finds that suggested a prehistoric phase on this site were in fact examples of the local conservative production during the Roman Period. 128 87 dating to the last couple of centuries became sparser escaped closer scrutiny. This inevitably creates the familiar problem of defining the rank, size and character of the potential sites that weren’t documented by the regular grid survey. As we saw, the lack of detailed data can only partly be compensated for with the resolution of the transect survey records. The presence of Roman material in sector I was recognized early in the fieldwork process, thanks to the larger quantities of brick and tile on the meadows to the east of the monastic complex, on field blocks 4a and 4b (map III_26b). The location was defined as “site 2” in Appendix I on the basis of the high overall artifact density. The study of the material showed that the great majority of the finds on this location are Late Roman. Indeed these field blocks feature the highest density of Roman material in the survey area of over 6 fragments per 100 sq meters of transect survey. However we remained reserved about the original location of this material pointing to the possible dislocations from “Kale”, Sopot, as well as from the monastic complex built in the early decades of the 20th century. The total collections on grid 16a-b and the analysis of the individual transect collections showed that Roman material spread over most field blocks in sector I, with an evident concentration on field blocks 3-5 covering the monastic yard and the surrounding field blocks. There is a gradual decrease towards the northwest and southeast extremes of the sector. Thus it is very unlikely that this extensive scatter was eroded from the hill-fort above the monastery, although local dislocations (such as piling up larger fragments along the edges of the fields) are not improbable129. The total collections on grids 16a-b revealed an absence of clear patterning (map III_26a). Rather there were a few small “cores” featuring between 8 and 12 fragments per 100 sq meters, not much higher than the density recorded by the transect survey. Two such “cores” limited to single grid units were discovered on field block 4a and on a pair of grid units in the monastic yard, near the Bronze Age site. They are surrounded by a fairly continuous carpet with artifact densities ranging between 2 and 6 Roman fragments per 100 sq meters. As the analysis of the adjusted transect collections shows, this layer certainly spread over the yard of the monastery, over field blocks 3, 5a-b and possibly, over part of field block 6, on the other side of the seasonal creek (cf. map III_26b and 26c). Although indicated by the transect survey results, the total grid survey failed to detect even a roughly concentric pattern of distribution. As a consequence it is difficult to offer a rough estimate of the size of these scatters. They could spread over an area of almost two hectares, but considering the fact that individual clusters are rarely larger than 200 sq meters, the extent of the total dispersal area is hardly a useful indicator. Judging by the predominantly multi-focal and discontinuous distribution pattern, but also by the character of the finds (mostly tile and plain pottery with little coarse ware, graph 13 in Appendix II), these could be the remains of a small communal necropolis. This is chiefly supported by the two inscribed tomb-stones, either found in the monastic yard or brought from the nearby fields. It remains unclear if this necropolis belonged to the inhabitants of the small fortification on the top of the cliff or to a hypothetical open settlement situated on the opposite bank of the Vardar130. The difficulty of accepting the necropolis thesis is that the Roman Period isn’t known to feature ceramic assemblages with a specifically sepulchral character or function. Alternatively these could be the remains of a residential site, a “suburb” of “Kale”, Sopot; the lack of a compact and dense surface cluster resulting from recent disturbances inflicted during This phenomenon was also encountered in sector III and in the second survey area, cf. B. Slapšak, 145149, eds. J. Chapman et al. 1988. 130 One of the tomb-stones has been summarily published, N. Vulić, Antički spomenici naše zemlje, Spomenik Srpske Kraljevske AkademijeLXXI, Belgrade 1931. 129 88 the construction of the monastery. But as explained in the previous chapter and in the section discussing the remains of the Bronze Age, there is very little arable land in sector I and the location is isolated from the fields at Prisoj’s foot. If this was a more substantial settlement quarter, it must have exploited the fields on the opposite bank of the Vardar. Unlike with the scattered and disturbed remains in sector I, there were no doubts about the character of the clusters on field blocks 44-44a and 45-45a. It immediately became clear that these were site locations and they were defined as sites 5a and 5b in Appendix I. These field units stood out by the quantity of Roman finds, even though individual transect collections were less thorough than on the rest of the field blocks. The great majority of Roman material in the survey area came from the total grid collections on sites 5a and 5b. Site 5a is most probably the largest Roman settlement in the first survey area (map III_27a). The cluster revealed by the total grid survey measured at least 8000 sq meters and probably spread over the neighbouring field blocks to the east, on field blocks 41 and 42. On these field units we were not allowed to carry out individual transect collections, though the overall artifact density suggests that even if site 5a did extend over field blocks 41-42, it was only for a short distance. In all likelihood these field blocks hide the site halo, the zone featuring between 2 and 4 fragments per 100 sq meters. Only on the south side on field block 44a did we manage to reveal a larger portion of the site halo using regular grid units. It probably spread for over 80 meters from the southern edge of the site occupying the entire area of field block 44a. Along the northern and western limits, it is impossible to observe a transitional zone. Artifact densities ranging between 6 and 12.6 fragments per 100 sq meters were recorded across the entire width of field block 44 and the northern half of field block 44a. It is possible that site 5a produced a halo to the north on field block 40; but as this field was occupied by the cemetery of the modern village and ground visibility was very low, it was decided to limit the total grid survey to field blocks 44 and 44a. To the west site 5a is delimited by a steep ravine. The extent of the halo of site 5a can roughly be determined on the basis of the individual transect collections. Except on field blocks 41-43, east of the site, where we were only allowed to count surface artifacts, site 5a is surrounded with field blocks featuring average artifact density (map III_25a). On the east, after the artificial gap on field blocks 41-43, Roman material reappears in small quantities on field blocks 31, 33 and 34, about 180 meters measured from the site’s eastern edge. A modest amount was also collected from the village cemetery, covered by field block 40, while to the west on field blocks 45-45b, artifact density returns to an on-site level. This pattern is only reinforced when the transect collections are adjusted to represent 100% of the material counted (map III_27b). The adjusted collections by individual transect units also suggest an increased density on field blocks in sector II, south of the site. Some of these, for example field blocks 20 or 72b are situated over 200 meters from the site’s southern edge. But as shown in Appendix II, this is not a carpet that spreads continuously from the site’s southern edge, while the finds from sector II, although datable to the Roman Period clearly stand apart from the predominant fabric classes on site 5a. Although the maximal artifact density on this site is not much higher than the densities recorded on “site 2” in sector I, densities higher than 6 shards per 100 sq meters are continuously spread over an area many times larger than the tiny clusters in sector I. Site 5a is a large and compact cluster with one major and one minor core in the western half of the site area. The finds collected by individual transect and regular grids constitute a full domestic assemblage including architectural ceramics, kitchen ware, storage jars and plain table ware, but only a few fragments of fine ware (graph 13, Appendix II). One of the more characteristic classes of finds from this 89 site was a large group of amorphous chunks of coarse ceramics, featuring very large quantities of poorly sorted volcanic rock. The size of these fragments as well as the occurrence of fine, funnelshaped imprints on some examples suggest an architectural function, probably daub. Looking at the on-site distribution pattern and especially the distribution of architectural ceramics, it was concluded that this was a nucleated settlement ranking as a small hamlet (map III_28). The importance of the micro-location occupied by site 5a was explained in Appendix 1 and in the section discussing the distribution of Early Iron Age finds. If there are truly remains of a residential quarter in sector I near the monastic complex, it occupied a far less favourable location than the settlement on site 5a. The individual transect collections clearly indicated that this material spread over to field blocks 45, 45a and 45b, on the western side of the ravine. However the total grid survey revealed a rather contracted and discontinuous cluster and total collections were limited to field block 45 and parts of field block 45a (maps III_27a, 28). More precisely there were two separate clusters. One larger in the eastern half of the grid measured about 1200 sq meters, the other smaller and certainly disturbed in the western end of the grid, at the border between field blocks 45 and 45a. The more extensive core in the eastern half featured a maximum artifact density of about 10 fragments per 100 sq meters, while the smaller western core was far denser. On one grid unit we counted over 70 fragments of tile, but as explained in the appendices, this is certainly the result of recent dislocation. Originally this western core was probably larger but thinner. The two clusters are bounded together by a zone of 2 to 3 fragments per 100 sq meters characteristic for the halo zone to the south of site 5a. The material collected from site 5b is identical to that collected from its larger neighbour. The finds constitute a full domestic assemblage, although there are differences in the presence of certain basic categories. There is a larger amount of tile and less coarse ware, but it is evidently different from the assemblage from sector I consisting almost exclusively of architectural ceramics and only rare fragments of plain, domestic fabrics (graph 13, Appendix I). It remains unclear if “site 5b” was a continuation of the larger settlement on the eastern side of the ravine or traces of other non-residential activities. A very similar pattern can be seen in the modern landscape, with animal sheds and the village cemetery situated a few hundred meters from the village, on the less fertile stretches to the west. Sites 5a and 5b occupy twin locations, the smaller site occupying the less favorable and more isolated location. The two appear as an almost exact replica of the modern village and cemetery. The difficulty of accepting this interpretation rests primarily in the high artifact density and the character of the material collected from “site 5b”. Concerning both aspects this cluster differs little from its larger neighbour on the other side of the ravine. And yet if it was a continuation of the settlement on site 5a, its positioning is rather inconvenient. The deep ravine that separates the two sites prevents direct communication between the main cluster and its satellite. The majority of the finds on sites 5a and 5b can be dated to the Late Roman Period. However it is possible that both locations were occupied prior to the 4th century AD. Fragments of a locally produced fabric group characterized by fully oxidized firing and plain surfaces (the so called soft-orange fabric group) were discovered on both sites, usually following the distribution of the predominant Late Roman fabrics (map III_30). As discussed in appendix II, the distribution of this material and its similarities to the material excavated from Early and Middle Imperial layers point to a possible earlier date, although it is equally possible that we are dealing with a fabric group produced over a longer period of time. Either way an earlier, pre- 90 4thcentury phase on site 5a shouldn’t be excluded. Unfortunately it wasn’t possible to define a separate Early to Mid-Imperial assemblage and the exact extent of this phase remains unknown. The second largest concentration of the soft-orange fabric came from the intensified transect collections on field block 31, on the very edge of the halo zone, 170 meters east of site 5a. Here this material was found accompanied by equally large or larger quantities of other fabric groups with a possible pre-4th century date and a handful of examples of the local Late Roman fabrics that predominate in the collections from site 5a-5b. The entire collection of Roman material on field block 31 gave an artifact density of almost 4 fragments per 100 sq meters. This equals the artifact densities recorded on the periphery of site 5a-5b; because the individual transect collections were more thorough, there is no compensation for the lesser survey intensity on field block 31. The analysis showed that in case the individual transect collections were limited to the number of finds counted, field block 31 wouldn’t differ from other field blocks surrounding site 5a-5b, encompassing its halo. The composition of the collection as well as the character of the finds also indicates that this cluster doesn’t represent the remains of a residential site. There is little coarse pottery and more significantly, architectural ceramics is nearly absent. Moreover nearly 50% of the finds collected from this field were fragments of fine pottery that wasn’t producedlocally. The examples of fine red-slipped pottery or a peculiar soft-buff ware decorated with grayish wash have no parallels among the predominant local fabrics on site 5a-5b, but show great similarity to the material excavated from Early and Mid-Imperial deposits in larger urban centres. It should be noted that most of these finds were relatively well preserved, especially in comparison to the material from site 5a-5b. This circumstance suggests that they came from recently disturbed, enclosed deposits. Field block 31 also yielded a pair of perfectly preserved pyramidal loomweights. They were made in the local Late Roman fabric indicating a continued presence on this site throughout the Roman Period. It should be stressed that loom-weights are frequently found on Roman necropoleis. This unusual concentration of fine pottery isn’t limited exclusively to field block 31. At least three smaller clusters were collected from field blocks and grid units in sector II, west of the modern village (map III_25a). Adjusting the transect collections to represent 100% of the material counted, these field blocks are ranked among the highest in the survey area in terms of artifact density (map III_27b). As pointed out in Appendix II, the high values predicted for field blocks in sector I and II are to a large degree inflated by the more accurate pottery counts, coupled by the low ground visibility on these fields. In addition on certain field blocks the high artifact density was greatly enhanced by the small size of the field unit. The detailed analysis showed that the maximum density barely exceeds the overall mean value and it is lower than the threshold of 2-3 fragments per 100 sq meters (table 22, Appendix II). The average density recorded by the grid survey for this sector is far below the densities recorded in sectors I and III. These are in practice tiny clusters of surface material consisting of between a pair and a few fragments dispersed on field blocks 16a, 20, 72 and on a couple of units on grids 2 and 4. Not a single fragment came from grid 1, although it is situated closer to site 5a-5b (III_maps 28, 30). Thus these clusters appear only after an intervening sterile stretch. As on field block 31, they consist almost exclusively of fragments of fine table ware, in all likelihood imported from some larger centre. Among the dozen collected examples it was possible to recognize fragments of jugs, table jars and a nicely preserved lid. Coarse fabrics and architectural ceramics are poorly represented. 91 Because of the relative proximity to the settlement on site 5a-5b and the special character of the finds, it was suggested that these clusters including the more substantial one on field block 31 are the remains of a necropolis. Admittedly the fact that they were found dispersed across a fairly large area of about 300 meters is to certain degree confusing. One would expect to find the remains of a communal necropolis concentrated over a smaller area, although it is possible that we are dealing with dispersed family burial-plots. In terms of artifact density these clusters produce densities similar or slightly higher than those recorded on the periphery of site 5a-5b. It is the character and the quality of the finds that clearly distinguish them from the rest of the offsite material, although it has to be repeated that unlike during some other periods, the material from Roman necropoleis doesn’t exhibit a particularly distinct characteristics. The total grid survey on field blocks in sector V and the southern half of sector VII confirmed the absence of Roman or Late Roman sites in the eastern half of Prisoj’s southern foothills (map III_31a). In these survey sections artifact density is even lower than on field blocks in sector II, only rarely exceeding the mean overall of 1.5 fragments per 100 sq meters. This situation was made apparent by the individual transect collections.Across this entire stretch, artifact density remained below the mean overall value and this pattern remains unaltered after adjusting the transect collections to represent 100% of the material counted (map III_31b). But it is the character of the material and the complete absence of focalization that were decisive in distinguishing this zone from the thin clusters of fine pottery to the south of site 5a-5b and interpreting them as off-site debris generated by the small hamlet 300 to 700 meters from the fields. The local fabric groups that predominated the surface record on site 5a-5b reappear on these fields, but now they are regularly badly worn and with eroded surfaces. Unlike in the supposed necropolis area, the compositions of the off-site collections are more balanced with roughly equal amounts of architectural ceramics and plain fabrics and very little coarse pottery (graph 14, Appendix II). After the Late Iron Age, the Roman-Late Roman Period is the second period for which we have evidence for off-site discard. In the case of the Roman-Late Roman assemblage, this off-site carpet is more extensive, not showing even the slightest sign of clustering. This ultra-thin but persistent carpet spreads for a maximum distance of 750 meters from the centre of site 5a, before running into the impact zone of a contemporary, smaller settlement on site 8. According to the transect collections, it is even possible that it partly spread into the valley floor. Thus although generated by a three times smaller settlement, this extensive impact zone is almost as large as that produced by the Late Iron Age village. Moreover while the Late Iron Age off-site showed considerable discontinuities and clustering suggesting that it could at least partly come from disturbed satellite deposits, the Roman-Late Roman carpet is continuous with a slight increase on field blocks surrounding site 5a-5b, probably related to the site halo or disturbed burials. These differing patterns of dispersal indicate that different mechanisms created the Late Iron Age and the Roman-Late Roman off-site. The pattern of the former points to more concentrated industrial, agricultural or ritual activities, while that of the latter looks more like the diminished remnants of a once denser off-site carpet created by prolonged spreading of unsorted manure on the fields at Prisoj’s southern foot. Studying the large collections of Late Iron Age finds from the southern half of site 8 situated about 800 meters east of site 5a, we came across a small group of finds that could be dated to the Late Roman Period (graph 14, Appendix II). It all seemed as if the off-site carpet documented on grids 6, 8 and 9 continued to spread across the fields at Prisoj’s eastern foot. However the character of these finds was completely different from the worn fragments collected 92 from sectors V and the southern half of sector VII. Particularly prominent was a group of fine, gray-fired pottery. It was possible to distinguish them from the more dominant Paionian gray ware by the presence of a fine slip, slightly darker than the color of the paste. This material finds parallels on the majority of excavated Late Roman settlements and it is dated to the 4th century or later131. It was accompanied by small quantities of locally produced plain fabrics and very little tile and coarse pottery. Apart from this radical change in the composition of the collections of Roman material from grid 10, there is also a slight increase in the artifact density from the surrounding fields. These sudden changes in the off-site zone became more comprehensible after the surprising discovery of a small Late Roman site in the hearth of the Late Iron Age village, on field block 126 covered by grid 12 (map III_32). The individual transect collections on this field unit recorded an artifact density only slightly higher than the overall mean density, which was related to the very low quality of the surface finds. Indeed if we assume that the collections were limited to the number of finds counted, field block 126, like the rest of the field blocks in sectors VI and VII becomes part of the low density zone, undistinguishable from the off-site at Prisoj’s southern foot (cf. maps III_32a and 32b). As discussed in Appendix II, there is no doubt that this merely reflects an underestimate of the true amounts of surface material on certain field blocks in sector VII during the quantification campaign. In fact the total grid survey recorded the highest densities of Roman material in the survey area precisely on field block 126, reaching 18 and 20 fragments per 100 sq meters in the southern half of the grid (table 23, Appendix II). As the detailed analysis of the grid survey results showed, this was a larger farmstead measuring a maximum of 3500 sq meters. Being surrounded by thickly overgrown stretches on three sides, parts of the site periphery and the site halo remain hidden. It is nonetheless certain that the site halo wasn’t very extensive. It probably spread mostly to the south of the site core and onto the valley floor, but not on field blocks to the north where the total grid survey (grids 13 and 14) recorded a very sparse off-site carpet. Despite the fact that the Late Roman cluster on site 8 was the densest in the survey area, it was almost completely overwhelmed by the much denser Late Iron Age assemblage. It is another stark illustration of the relative sparseness of the Roman material in this survey area. Although overlaying the earlier Late Iron Age settlement and characterized by a wider spectrum of ceramic artifacts, the Late Roman phase failed to obscure the remains of the earlier settlement. On the contrary until the study of the collected finds was completed, it remained totally invisible. There are apparent similarities between the Roman collections from sites 8 and 5a (graph 15, Appendix II). The same types of brick and tile are present in both assemblages, though certain groups predominant in one of the assemblages appear in smaller quantities in the other. This is also true for the finer fabric categories and the coarse ware. Most of the fabric groups found on site 8 can also be seen among the collection from site 5a. Nevertheless certain fabrics are clearly more numerous on site 8, particularly a semi-reduced fabric group, often decorated with incised linear or wavy patterns. This is one of the favorite decorative patterns in the local pottery production between the late 4th and late 6th century AD132. Examples decorated in this V. Stojanović-Anderson, 14-18, 1992. M. Ončevska-Todorovska,Docnoantička grnčarija od Skupi, Skopje 2004, especially tables 34-36; the author explicitly relates these decorative patterns with stew-pots. It should be stressed that very similar decorative techniques and patterns characterize the pottery of the Early Middle Age, B. Ristevski, Medieval ceramics from the Skopje area, 49-71, Macedonian Heritage 21, 2004; A.K. Vionis, J. Poblome, M. Waelkens, The hidden material culture of the Dark Ages. Early Medieval ceramics from Sagalassos (Turkey): new evidence (ca AD 650-800), 147165, Anatolian Studies 29, 2009. 131 132 93 manner rarely appear among the material from site 5a-5b, although the fabric is clearly present. Possibly this reflects a slight chronological difference between the two sites, the assemblage found on site 8 being slightly later. It is evident that on this site there is a narrower spectrum of fabric groups than on site 5a. This could also reflect a shorter life-span of site 8, though if this is truly the case, the much higher artifact density becomes even more baffling. One is tempted to relate it to the fragmented character of the finds on field 126. The two sites are also similar in another aspect: both are associated with scatters of fine imported pottery situated about 150 meters from the site cores. These satellite clusters are comprised of imported table ware that although possibly contemporary with the settlement sites, doesn’t appear among the on-site collections. To the north of the Late Roman cluster on site 8, both the quantity and the character of the material change yet again (table 23, mapsIII_32, 32a and 32b). On field blocks 125 and 113ab covered by grids 13 and 14, the density of Roman finds drops to an off-site level. Individual badly worn fragments were collected from a dozen grid units randomly dispersed across the gridded area. They give densities lower than 2 fragments per 100 sq meters. The composition of the collection also changes with roughly equal presence of architectural ceramics and plain fabrics and a considerable percentage of unrecognizable fragments. This composition is almost identical to the composition of the off-site collections from grids 6, 8 and 9 in sectors V and VII (graph 15, Appendix II). About 275 meters to the north of the Late Roman farm on site 8 rises a low hillock dominating over the middle section of the small valley. Like the larger hillock at the southeast corner of the survey area it was made of pinkish conglomerates and like the latter, it was fortified sometime during the Roman Period. The faint architectural remains discovered on the top of the hillock and at its western foot were defined as site 9 and described in Appendix I. On field blocks surrounding this small hill-fort, the transect survey recorded average overall quantities and an average amount of Roman-Late Roman material (map III_32a). In fact the highest density of 6.8 fragments per 1000 sq meters was discovered on field block 122, to the east of the hillfort, but this concentration came to light only after the processing of the finds. However as argued in the preceding paragraphs and in Appendix II, on some of the field blocks in sector VII, transect collections were more intensive, representing at least 100% of the material counted. Assuming that only counted finds were collected would rank this entire sector below the survey average (map III_32b). Wanting primarily to explore the relation between the small fortification and the surface finds on fields at its western foot, we limited the grid survey to field blocks 114 and 115 covered by grid 15. The total collections by regular grid units confirmed the results of the individual transect collection. There was only a slight increase of Roman finds on these field blocks. The maximum density barely reached 4 fragments per 100 sq meters and the average density per gridded area is much lower, although still higher than on grids 6 or 13 (table 23, Appendix II). Predictably the majority of the Roman finds came from grid units closer to the foot of the hillfort and the abovementioned architectural remains. But their quantity, character and inner distribution suggested that there was no permanent settlement on this location. The slightly increased artifact density could result from less intensive, non-residential activities. Indeed site 9 and the surrounding fields offer a perfect setting for a refugium for the small community that inhabited the surveyed basin. This is one of the best sheltered locations in the survey area, supplied with a freshwater spring only a few hundred meters to the north. 94 On the eastern bank of the central valley, the overall quantity of Roman finds roughly equals the amounts collected by field walking units in sectors VII and II (graph 11, Appendix II). The average density in sector IX is slightly over 2 fragments per 1000 sq meters, although there lack larger concentrations. The density of Roman material recorded by the transect survey never exceeds 1 fragment per 100 sq meters, even if we assume that all counted material was collected (table 24, Appendix II). However the amount of the total surface record is nearly twice as low as in the western survey sectors, while Roman finds comprise a significant portion, though they are still in the shadow of earlier periods, such as the Hellenistic or the possible Late Neolithic. As explained in Appendix I, the Late Ottoman-Early Modern off-site material from the village comprising a large fraction of the surface record in sectors west of the valley, spreads in a very thin carpet on the opposite bank. The decrease in artifact density is nearly ten-fold. In such circumstances it is much easier to recognize clusters of surface finds from earlier periods, but at the same time they can appear fairly insignificant and escape closer attention. According to the results of the individual transect collections there were no major concentrations of Roman material in sector IX, on the Ramnište Ridge (maps III_32a and 32b). We see the zone of less than 10 Roman shards per 1000 sq meters of transect survey spreading over the majority of the field units that belong to this sector. On the southern half of the ridge, on field blocks covering site 11 and its immediate surroundings the density of Roman finds is slightly higher fluctuating between 1 and 4.8 fragments per 1000 sq meters. After a brief sterile interval in the central part of the sector, there is a slight recovery on field blocks in the northern end of the sector with artifact densities lower than 1.5 fragments per 1000 sq meters. It all seemed as if this was a continuation of the off-site carpet covering the western half of the survey, the quantity of the Roman finds decreasing towards the more peripheral field units at the foot of the Radičica Massif. But the careful study of the collected material and the more detailed total grid collections on sites 11 and 12 proved otherwise. The total grid collections on site 11 were primarily concentrated on the dense clusters of prehistoric material. They only confirmed the presence of very small quantities of Roman material on field block 157, already indicated by the field block survey (map III_33). According to the individual transect collections, the Roman material was mostly concentrated on field blocks 159, to the north of site 11 and on field block 192, on the lower terrace. Assuming that the individual transect collections were limited to the number of finds counted, the zone of higher artifact density is completely limited to field block 159, while field blocks 192 and 193 merge into the off-site. It is no doubt unfortunate that field block 159 was left out of the total grid survey. However if there was a larger concentration of Roman finds on these two field blocks approaching the on-site thresholds on sites 5a-5b or 8, one would expect to see at least a portion of the halo of this hypothetical site on the grid covering site 11. Instead the density of Roman material on this location remains well below the on-site threshold recorded in the western survey sectors (cf. tables 22 and 24, Appendix II). In terms of artifact density it is not different from the distribution in the off-site zone at Prisoj’s southern foot, but the character of the finds and the composition of the collection change significantly. Only a very small percentage of the Roman collection could be dated more specifically to the Late Roman Period. The Late Roman plain fabrics commonly encountered in the survey’s western sectors nearly disappear, while the percentage of coarse and plain fabrics broadly datable to the Roman Period remains stable. Most significantly there is a sharp increase in the amount of architectural ceramics and the tile collected from this entire sector didn’t belong to the same fabrics as those in the survey’s western sectors. This increased concentration of architectural ceramics on the southern tip of the 95 Ramnište Ridge was eventually interpreted as the remains of less intensive, non-residential activities. The suggestion was mostly based on the large size and the nearly intact surfaces of the tile fragments collected from field blocks 159 and 192. Fragments that are preserved so well that a number of them could be fitted together must have originated from recently disturbed deposits. That is why it was surmised that there was a smaller, non-residential site near the border between field blocks 192 and 159 that was left just outside the total grid survey on site 11. On the other hand, the nearly 10% of coarse ware attached to the Roman collection from site 11 and comprising nearly 20% of all Roman finds in sector IX must be treated with suspicion. As explained earlier, this class of pottery is strongly conservative and one can hardly distinguish between coarse fabrics accompanying Roman and pre-Roman assemblages. Therefore it is possible that most of the examples of coarse pottery from this sector belong to the predominant Hellenistic assemblage. In the central parts of sector IX, the results of the transect survey weren’t confirmed by the total grid collections (map III_34). The study of the material collected from the predominantly Hellenistic site 12 revealed an equally large but thinner cluster of fragments, belonging to the soft-orange fabric group. This is the same fabric group that signaled an earlier pre-4th century phase on site 5a-5b and was found in larger quantities in the possible necropolis area to the south of this site. On site 12 it forms a fairly compact cluster situated immediately west of the core of the Hellenistic site (map III_35). Judging solely by the dispersal area of this group of finds, the Hellenistic farmstead experienced only a slight contraction during the Roman Period measuring about 1200 sq meters. It has also evidently shifted a couple of dozens meters to the west. Remarkably the maximum density of this material equals the maximum density recorded on site 5a. As on site 5a it is certain that an unknown portion of the coarse ware on site 12 accompanied the soft-orange fabric group (graph 17, Appendix II). But even if assuming that all coarse ware discovered on grid units covering the core of the Roman cluster belonged to the Roman assemblage wouldn’t increase the size of this site. It would merely elevate the on-site density drawing a clearer distinction between the site core and the site halo. It must be emphasized that the very small amount of architectural ceramics from site 12 is distributed independently of the soft-orange fabric group. The off-site impact of this small farm remains even more elusive. It is nonetheless clear that it didn’t produce a very extensive off-site carpet. Indirectly this circumstance supports the thesis that the Hellenistic farm on site 12 wasn’t an isolated establishment on the Ramnište Ridge. The clusters of Hellenistic material discovered on a number of locations in sector IX can hardly be explained as satellites or off-site debris emanating from the small farmstead on site 12. On the other hand the soft-orange fabric group in sector IX was found almost completely limited to site 12. With the exception of brick and tile fragments, the great majority of possible Roman finds collected by field walking units in sector IX either belong to the soft-orange fabric group or to one of the coarse fabric groups broadly datable to the Roman Period. The overall density of off-site Roman finds is lower than in sectors on the western bank and the character of the finds is obviously different. The Late Roman finds are present in very small quantities. It is noteworthy that outside the limits of site 12, the few examples of the soft orange fabric or the accompanying coarse ware mostly follow the smaller clusters of Hellenistic material, such as those on field blocks 160-164 and 168. The distribution of this material doesn’t follow the distribution of tile and it is rarely accompanied by the Late Roman fabrics. Spatially it seems rather more related to the Hellenistic assemblage, which could also reflect its actual chronology. This clearly implies 96 that the off-site carpet of Roman finds documented in the western survey sectors didn’t spread in equal intensity on the opposite bank. Knowing the size of the settlements in the western survey half, this is hardly a surprise. The increased average density for sector IX is chiefly due to the presence of a thin scatter of brick and tile collected from field units in its southern part, near site 11 and by the cluster on site 12. Unless the architectural ceramics is dated Late Roman, this period nearly disappears from sector IX. The individual transect collections from sector X, the Jakupica Ridge clearly indicate a significant presence of Roman material on at least three locations (map III_36a). On two pairs of field blocks on the tip of the Jakupica Ridge, on field blocks 133-134 and 142a-142b, the transect survey recorded over 10 fragments per 1000 sq meters (table 26, Appendix 2). These field units are surrounded by field blocks featuring average or lower than average artifact densities, as if forming halos over the two closely spaced clusters. To the north, on field blocks covering the central parts of the sector, Roman finds nearly disappear from the surface. They reappear about 750 meter to the north of field blocks 142a-142b, on a group of contingent field blocks at the western foot of Radičica, field blocks 152a-155. Thus to a large degree, the pattern revealed in sector IX is repeated, with the crucial difference that the concentrations at the northern and the southern end of sector X are considerably larger and denser. Failing to carry out total grid survey at least on some of these locations largely deprived us of the possibility to offer more precise definition of the site limits and its inner structure. On the basis of the individual transect collections, it was merely possible to conclude that the pair of Roman sites didn’t spread across the entire field block area and they probably didn’t form compact clusters. We rather encountered a series of smaller clusters, often standing more than 10 meters apart. This was especially pronounced on field blocks 142a-142b or site 13b. Although it can’t be demonstrated through the results of the individual transect collections, site 13b was made up of a series of smaller clusters, especially concentrated in the southern halves of the field blocks (map III_37a). Only rare fragments spread over the northern halves of field blocks 142a and 142b and over field block 143 to the east. Because of this focalized and discontinuous onsite distribution, it would have been difficult to determine the size of this site even if a total grid survey was carried out. It was reckoned that the maximum site area is not larger than 5-6000 sq meters, but considering the sum of the separate clusters only, it is certainly much smaller. Site 13a, about 110 meters to the west and on a lower terrace comprised a more compact and continuous cluster, with an evident focus in the central parts of field block 133 and the eastern half of field block 134. There is a gradual decrease along a south-north axis, though as on site 13b, we observed that small groups of finds were separated by sterile stretches. Beyond the limits of fields 133 and 134, Roman material becomes even sparser. Site 13a was clearly limited to these two field units. Excluding their westernmost quarters where the transect collections included but isolated fragments, site 13a could spread over an area of nearly 8000 sq meters. It is thus potentially as large as the settlement on site 5a. Why were these relatively large sites underestimated during the transect survey? Earlier it was pointed to the sparseness of the material on this location relative to the large quantities encountered in the sectors closer to the modern village. If the Roman-Late Roman material on site 5a-5b was found unaccompanied by finds from other periods, it would have hardly appeared more substantial than sites 13a and 13b. Lacking the high resolution of the total grid survey complicates the direct comparison between the two sites regarding artifact density. Comparing the results of the individual transect collections, the two pair of clusters look equally substantial, site 13b being even slightly denser than site 5a (table 26, Appendix II). But as the individual 97 transect collections on these field block were not equally thorough, this comparison is misleading. It is equally deceptive to increase the transect records on sites 13a and 13b by a factor of 2.5 and compare them with the grid survey records on site 5a-5b, because transect collections on the former were more thorough amounting to a total collection by 80x3 meters large strips. Estimating the artifact density for each of these individual transects and correcting for the ground visibility factor only, we see that the artifact densities on sites 13a and 13b are indeed lower than on sites 5a-5b and 8. It is both the low overall artifact density and the sparseness of Roman material that contributed to the inconspicuousness of the clusters on sites 13a and 13b. In fact if we adjust the individual transect collections so that they represent 100% of the material counted on all field blocks, the density of Roman material on field blocks in sector X is slightly decreased and they join the zone of average artifact density (map III_36b). This puts them in the same rank as the field blocks covering portions of the halo of site 5a-5b or the intermediary density scatters in sector IX. Locally however the concentrations on site 13a-13b still present a considerable increase from the surrounding field units, with density 5 to 6 times the sector’s average. Apart from indicating that the clusters on sites 13a and 13b are less substantial than those in the western survey sectors, the adjusted record of the individual transect collections revealed only slight changes in the on-site distribution pattern, further emphasizing its multi-focal, discontinuous character (map III_37b). Another circumstance that distinguishes sites 13a and 13b from the main residential sites in the western survey sectors is the character and the composition of the surface material. The clusters that constitute sites 13a and 13b are almost exclusively made up of architectural ceramics (graph 19, Appendix II). Fragments of brick and tile made in fabrics different than those encountered on sites 5a-5b and 8 were accompanied by very small amounts of badly worn pottery fragments. Plain pottery is particularly scarce, while coarse fabrics are exclusively represented by pithos fragments. This is not untypical for rural sites of the Roman Period133. Roman Period clusters predominantly made of architectural ceramics are known from regional projects in Greece and we’ll also encounter them in the second survey area. The differences between the assemblages collected from sites 5a-5b and 13a-13b could merely result from the different taphonomic processes on the two locations. It has long-since been acknowledged that the amount and the quality of surface material are largely determined by specific post-depositional processes and events, as well as the agricultural season in which the survey took place134. In addition it is highly probable that the different clusters of Roman to Late Roman material discovered in this survey area experienced different histories of recycling and removal, both prior to and after deposition and abandonment135. We should nevertheless consider the possibility that sites 13a and 13b are special-purpose sites. The relatively large amount of overfired tile wasters on site 13b could point to industrial activities or tile kilns, though in such a case it is strange that these tile fabrics almost never found their way among the collections from sites 5a-5b and 8. One also thinks of other forms of non-residential activities, primarily the storage and processing of agricultural goods. It must be stressed that only a few of the fragments 133 J.L. Bintliff, P. Howard, A.M. Snodgrass, 2007; W. Cavanagh, C. Mee and P. James, et al. 2005; H. Bowden, D. Gill, Late Roman Methana, 84-91, C. Mee, H. Forbes eds, 1997. For a summary of this phenomena see now J. Bintliff, The Complete Archaeology of Greece: From Hunter-gatherers to the 20th century AD, WileyBlackwell 2012. 134 J.F. Cherry, 398-399, D.R. Keller, D.W. Rupp,eds. 1983 J.L. Bintliff, A.M. Snodgrass, 123-135, 1985. 135 D. K. Pettegrew, 189-201, 2001; J.T. Peňa, Roman Pottery in the Archaeological Record, Berkeley 2007. 98 of plain pottery were identical to the predominant Late Roman fabrics on sites in the western survey sectors, implying that sites 13a and 13b were at least partly contemporary with the rest of the Late Roman sites in the survey area. Even if we allow for a certain chronological difference between these two groups of clusters, the presence of two completely different sets of tile fabrics on field blocks east and west of the central valley is surely striking. This sharp divergence in the distribution of the most prevalent classes of Roman material must be related to local patterns of production and distribution. Either the establishment on sites 13a-13b had its own local production of brick and tile or the material was brought from elsewhere. Apart from the similarities in the shape of tiles and the few examples of pottery fabrics identical to those found on the sites in the survey’s western sectors, the two assemblages appear almost unrelated. Interpreting the possible character and rank of sites 13a and 13b, one also needs to consider the regional context surrounding the surveyed area. We return to this problem in the conclusion to this section. For the moment, it suffices to mention the proximity of these two sites to the main east-westcommunication axis and the possibility that during the Roman Period, this was an active section of the Scupi – Stobi - Thessalonica road. The border-line between the provinces Macedonia and Moesia Superior (Dardania in the Late Roman Period) also passed nearby the survey area. In such circumstances one shouldn’t exclude the possibility of a statesponsored construction along this important corridor. At least some of the numerous Late Roman forts in the wider region may be viewed in this light, especially the very strategic 2 hectare large fort on site 10, situated 300 meters to the south of sites 13a and 13b. But turning back to the transect survey evidence and the location of the two clusters, we note two important details suggesting that site 13a-13b could be settlement locations after all. This is firstly indicated by the extent of the Roman cluster covering most of the southern tip of the Jakupica Ridge. Although we lack a finer spatial resolution, it is evident that the main clusters on field blocks 133-134 and 142a-142b are surrounded by a zone of lower artifact density spread over the neighbouring field blocks to the east and west. It represents a nearly identical pattern to that discovered on field blocks surrounding site 5a-5b and we’ll see further parallels from the second survey area136. It is difficult to see this fairly extensive peripheral zone as an exclusive result of site weathering and post-depositional dislocations. It is indicative that it spreads over field boundaries which can’t be of a very recent date and only on certain sides of the site areas. But if this zone truly represents the phenomenon of site halo resulting from the combined effects of site weathering and intensive in-field cultivation, then we have to allow that sites 13a and 13b represent the remains of a residential site. Industrial and non-residential sites in general aren’t expected to leave very extensive impact zones, though it has to be admitted that we still don’t know the signatures of these site categories in the surface record. The second important observation concerns the great similarity between the pair of sites 5a-5b and 13a-13b, their positioning relative to each other and their location in the surrounding landscape. In both cases we have two closely spaced clusters of similar size; one larger and relatively continuous positioned on a more fertile stretch of land, the other smaller and discontinuous with a less favorable micro-location. Moreover both pairs of sites are situated on similar micro-topographic units, on the very edges of the surveyed basin. These locations are strategically important in the local context for they control the western and eastern entrances in the basin of Sopot. This tendency to occupy twin locations is characteristic for settlements of all periods. In the case of the Roman settlements in the first survey area, we see them concentrating 136 Also cf. J. Bintliff, The Leiden University Ancient Cities of Boeotia Project: 2005 season at Tanagra, 29-38, Pharos 13, 2006; J.L. Bintliff, P. Howard, A.M. Snodgrass eds. 23-24, 2007. 99 on the edges of the local basins and avoiding the inner part of the small valley. As argued, site 5a-5b was evidently exploiting the southern foothills of Prisoj, while the possible settlement on site 13a-13b could either focus on the Ramnište Ridge or on the neighbouring valley to the east of the survey area. In our final interpretation we see the clusters on sites 13a-13b as the remains of a fully or seasonally inhabited residential site, of a lower rank than the hamlet on site 5a-5b. It was located in the drier half of the landscape, surrounded on all sides by the less fertile, Neogene soils. Micro-topographically and strategically this may have been a location similar to that occupied by site 5a-5b, but it is situated in a less favourable pedological and hydrological context. The last definite cluster of Roman material featuring artifact densities higher than the onsite threshold was discovered about 750 meters to the north of site 13b, in the northern end of sector X (map III_38a). This is a different micro-topographic unit, the western foot of the Radičica Massif. After almost completely disappearing from the central parts of the Jakupica Ridge, Roman finds reappear on the surface reaching densities higher than the mean overall values on field blocks 152a-155 (table 26 in appendix II). According to the transect collections, the highest artifact density was on field block 152a, but we saw that this was determined by the unequal size of the field blocks and the uneven distribution of the finds within the field block area. The largest concentration of relatively well preserved finds came from the eastern half of field block 155. This fact becomes more apparent once we assume that the individual transect collections were limited to the number of finds counted during the quantification campaign (map III_38b). The collection from the easternmost third of field block 155 revealed a tiny cluster, consisting of not more than a couple of dozen finds and occupying a maximum area of a couple of hundreds square meters. On the surrounding field units, although featuring similar or even higher artifact densities, we couldn’t detect distinct cores. The poorly preserved finds were found randomly scattered across a larger area. However if the small cluster on field block 155 was the sole focus of activity in this part of the survey area, then how to explain the fairly extensive carpet of Roman material spreading over several contiguous field units and over the entire northern end of sector IX? We therefore suspect that at least one or two other clusters of similar size went unnoticed or were reduced to thin inconspicuous scatters by post-depositional disturbances. A likely location is field block 152a, where although we failed to notice a definite concentration, the transect collections clearly indicate increased density of Roman material. As almost 100% of the finds on these field blocks date to the Roman-Late Roman Period, the observations about the distribution of the total surface record presented in Appendix I are still valid. The clusters on field block 155 and 152a or sites 14 and 15 represent similarly composed assemblages to those collected from sites 13a and 13b. They too comprise a very limited array of forms and fabrics. There are nevertheless certain differences (graph 19, Appendix II). Tile is not as predominant as on sites 13a, 13b and in sector I. It comprises 50% of the collection from field block 155 and less then 70% of all Roman finds from field blocks 152a-155. Pithos fragments feature even more prominently than on site 13b. The majority were collected from field block 155, though a pair came from field block 152a, again pointing to the second potential cluster of Roman finds in this part of the survey. On field blocks 152b and 153, we mostly collected worn fragments of architectural ceramics. There is very little plain pottery mostly datable to the Late Roman Period, while examples of cooking fabrics were virtually absent. 100 These small and isolated clusters were certainly contemporary with the establishments on sites 13a and 13b. They were made up of the same tile fabrics and formats. The lower artifact density, the smaller size of the clusters and their extremely isolated location define them as an analogous phenomenon of satellite clusters, frequently encountered during the analysis of the distribution of prehistoric assemblages. To some degree it offers a further support for the view that sites 13a and 13b are indeed settlement remains. The clusters on field units 152a-155 occupy the western foot of the Radičica Massif, almost a kilometer away from the Skopje-Thessalonica road. This micro-topographic unit stands between 30 and 40 meters above the floor of the small valley. It overlooks the valley’s middle course and the small hill-fort on site 9. The western foot of Radičica is one of the most isolated corners of the surveyed basin. There is no direct communication with the valley floor and there is little flat land in the foothills. The local soils are thin and stony and there is no water on the surface. This is not a likely settlement location. It is not only withdrawn from all major resources in the area (arable land, water, communications), but it offers a very limited living space. There are almost no Roman finds in sector VIII, at Gaber’s northern foot (maps III_38c and 38d). Only a couple of worn tile fragments were collected from the westernmost transects in this sector; possibly a continuation of the dispersed cluster on the southern tip of the Ramnište or an infiltration from the off-site carpet on the opposite bank of the valley. We expected to discover more finds from the fortified area on site 10, knowing its fairly large size, elaborate layout and strategic positioning. But we only came across another couple of tile fragments. Interestingly they were made in fabrics similar to the fabrics found on sites in the western sectors and were different from the predominant fabric groups on sites 13a and 13b. The general scarcity of surface finds in this survey sector is surprising, especially the absence of residential remains on site 10 (table 27, Appendix II). A small village could comfortably fit into the fortified area. The location offered not only a close control over the eastern pass linking the basin of Sopot with the northern half of the Veles Basin, but it also had access to larger stretches of fertile land, both in the surveyed basin and in the neighbouring valley to the east. It is possible that the area was avoided because of the lack of freshwater springs and its northern aspect. During most of the year, the Gaber casts a shadow over this part of the landscape, a circumstance that is neither favorable for human habitation nor for the growth of cereals. Predictably the overall density of surface finds sharply decreases in sector XI, in the upper course of the surveyed valley (maps III_39a and 39b). Most of the area is covered by dense wild vegetation, although there were a number of cultivated fields. However we were surprised to discover that more than half of the collected finds dated to the Roman-Late Roman Period and only a quarter to the Late Ottoman-Early Modern Period (graph 20, Appendix II). The maximum density of slightly over 2 fragments was recorded on a field unit in the sector’s southern end. This is a tiny collection of a few fragments of tile and a pithos fragment, though in the context of the very low background density in this sector, it is possible that these are the scant remains of non-residential buildings, similar to those discovered on field blocks 152a-155 or on the southern tip of sector IX. Equally significant, the composition of this tiny collection repeats the composition of the collections from field blocks 152a-155. They consist entirely of worn tile and pithos fragments. It is unlikely that this thin scatter is a continuation of the off-site carpet covering the western survey sectors. This survey sector is situated over 1200 meters from the small farmstead on site 8 and nearly 2.5 kilometers from the hamlet on site 5a-5b. Both sites are too small to produce an off-site carpet of such an extent. Alternatively the scant surface 101 remains in sector XI could signal the presence of yet another residential site that the transect survey failed to detect. Although presently overgrown and deserted, this area presents at least 50 hectares of arable land. It is a smaller but geographically well-defined settlement niche that could comfortably sustain a settlement of the rank of a farmstead or a small hamlet. III.1.12 The Late Byzantine and Early Ottoman Period (14th-18th century, tables 28-29, graphs 22-23 in Appendix II) As with the majority of the periods represented in the surface record of the first survey, there are very few published studies on pottery production from the post-antique era and this applies to most countries on the Balkan Peninsula. These studies are further limited to the luxurious and table ware from the Middle and the Late Byzantine Periods137. The situation is even worse when it comes to pottery from the Ottoman Period in the Central Balkans, for which we lack even a very general introductory study. In such circumstances it proved impossible to work with a finer chronology. It was merely possible to distinguish the fabric groups produced after the late 18th century; basically because they’re characterized by higher firing temperature resulting in very good solidity and because of the frequent use of vitreous, monochrome glaze, usually poorly fused with the body. Evidently these finds are predominant among the postAntique surface material. The other chronological category consisted of less widespread fabric groups, almost exclusively concentrated in the southern part of sector II, southwest of the modern village. They are characterized by less sophisticated modelling, thicker walls and unstable firing conditions. Only a handful of fragments had traces of lead glaze, in most cases poorly fused with the paste through a thin white slip or engobe. This feature was one of the rare chronological indicators. It is characteristic for the 14th and 15th century, when glazed pottery was produced more massively resulting in a visible decline in quality138. Only one fragment had traces of fine sgraffitto decoration. The dating of the rest of the finds from this assemblage relied on the experience and expertise of D-r B. Ristevski. These were fragments of plain local pottery, which could roughly be dated to the end of the Byzantine and the Early Ottoman Period. There are no finds earlier than the 14th century among the assemblage collected west of the village. Only one or two fragments collected from other locations in the survey area could possibly date prior to the 14th century, but this is far from certain. In any case it suggests that if existent, the pre-14th century settlement was most probably situated elsewhere. It also remains unclear if the small assemblage found by the modern village represents this entire period of four centuries. Published studies of the Ottoman censuses and the local oral traditions revealed many cases of deserted and reoccupied villages or settlement displacement due to various historical events and processes. The small community of peasant serfs could be dislocated with little fuss on the order of the local landlord. B. Ristevski, 2004; B. Babić, Trouvaille scellée de poterie de table Byzantine à Skopsko Kale 45-54, Archaeologica Iugoslavica 13, 1971; I. Mikulčić, M. Bilbia, Markovi Kuli, Vodno, Skopje, 1977-1980, 205-220, Macedoniae Acta Archaeologica 7-8, 1981-1982; M. B. Hađi-Pešić, Keramika u Srednovekovnoj Srbiji, Belgrade 1981. 138 B. Ristevski, 2004; D. Minić, Prilog datovanju srednjovekovne keramike iz Kruševca, 43-49, Starinar XXX, 1980; J. Vroom, 2005, the chapter on Late Byzantine/Frankish period; A. Vionis, “Crusader” and “Ottoman”Material Life: The Archaeology of Built Environment and Domestic Material Culture in the Medieval and post-Medieval Cyclades, Greece, Leiden 2006. 137 102 The proposed dating also finds support in the written documents. The name Sopot appears for the first time in one exhaustive census for the region of Veles, roughly dated in the second quarter of the 15th century139. This is the first written evidence relating specifically about the survey area. The fact that this document doesn’t emphasize that Sopot was a newly established settlement, probably indicates that the village existed prior to the Ottoman conquest. It provides invaluable information about the size of the settlement and its main agricultural products in the centuries prior to the introduction of American cultures 140. Nearly half a century later the village of Sopot is included in another exhaustive census, for the year 1467-1468141. In this case the identity of the village is confirmed by the editors of the document. But comparing the data from the two censuses, one wonders if this is the same village, as there are considerable differences in the number of households and the amounts of taxes paid in cash or kind. According to the earlier census, Sopot had 21 households and produced a fairly limited array of agricultural goods; most prominently wheat, barley and grapes and smaller quantities of linen and honey. One generation later, the settlement has experienced a dramatic transformation. In the extensive census for the year 1467/168, Sopot had 30 households including a priestly family and produced a different and wider range of cultures. Wine was the most prominent product, accounting for over one third of the total amount of taxes paid in kind or in cash. It has to be stressed that wine appears rarely on the list of products that were taxed by the Ottomans 142. Only a few villages from the entire administrative region of Veles paid tax in wine. Apart from wheat and barley, taxes were paid for growing rye, lentils and fruits, for bee-keeping and breeding pigs. Linen is not mentioned and there was no separate tax for the vineyards. Interestingly neither of the censuses mention sheep-herding, though it played an important role in many villages along the Mid-Vardar in later centuries. The importance of this evidence reaches beyond the chronological limits of the Late Byzantine and Early Ottoman periods. We saw that in earlier periods, the size of the local settlement remained stable, its area rarely exceeding 1 hectare. As will be shown Late Byzantine and Early Ottoman Sopot didn’t alter this pattern. Thus the data pertaining to this village also points to the potential population size of most of its predecessors. It is indicative that the majority of the villages in the surrounding micro-regions have between 15 and 30 households. The situation didn’t change significantly until the middle of the 20th century. According to the later summary censuses dated to the 16th and 17th century, the number of households in Sopot and in most surrounding villages rarely exceeded 30 or fell below 15143. Despite the fairly rich historical information, Late Byzantine and Early Ottoman Sopot remained archaeologically elusive. It wasn’t the pottery finds that revealed its location, but rather 139 M. Sokolovski, A. Stojanovski, Turski dokumenti za istorijata na makedonskiot narod, 52, vol. II, Skopje 1973. 140 In the Republic of Macedonia, as in most other former Yugoslav republics, plus Greece and Albania, the archaeological remains from the Ottoman Period have almost never been studied in their own right. Most references pertain to rescue excavations on more elaborate architectural monuments, for example the brief reports by B. Babić, Suli-an (Late Middle Ages-Ottoman Period), 163-164, Arheološki Pregled 9, 1967, B. Babić, I. Mikulčić, Skopje-a prehistoric and Medieval fortress, 162-163, Arheološki Pregled 9, 1967. The potential of this neglected field have been elucidated in a number of recent publications, S. Davies, J.L. Davis, eds. Between Venice and Istanbul: Colonial Landscapes in Early Modern Greece, Hesperia Supplement 40, 2007; J. Bintliff, H. Stoeger, eds. Medieval and post-Medieval Greece, the Corfu papers, Oxford 2009. 141 M. Sokolovski, 149-150, 1971. 142 Apparently because only certain landlords had the monopoly over the selling of wine on the territory of administrative units, M. Sokolovski, A. Stojanovski, 15, 1973. 143 D. Georgiev, A. Šerif, Turski dokumenti za istorijata na makedonskiot narod, Skopje 1994. 103 the poor traces of building remains discovered on field blocks 15b, 20 and 21, 120 to 160 meters south-southwest of the village (III_map 42, photo III_1). These remains were discussed in Appendix I. However prior to the processing of the ceramic material, there were doubts about their character and date. In fact they were so meager it was thought they were the remnants of abandoned animal folds. Moreover we believed that the Late Byzantine and Early Ottoman settlement was somewhere in the immediate vicinity, perhaps under the houses of the present day village. Although evidence from other known villages of the çiftlik type, clearly suggested that the quarters of the land-owners and the peasant serfs were always built at a certain distance144. The difficulties in detecting the surface remains of Late Byzantine-Early Ottoman Sopot become understandable when we consider the very small amount of finds that could be dated to this period. Only about 100 fragments or 0.61% of the total surface record belong to some of the fabric groups which could date to the Late Byzantine-Early Ottoman Period (table 28, Appendix II). This could very well result from the fact that the bulk of this material came from intensified transect collections and not from total collections by regular grid surveys. But even if we suppose that in reality the number of Late Byzantine-Early Ottoman finds is 2.5 times greater, they would still represent one of the humblest assemblages in the first survey area, amounting to less than 1.5% of the total surface record. However it has to be stressed that in the case of the Late Byzantine-Early Ottoman assemblage, the correction for the lesser degree of survey intensity will probably produce slightly enhanced figures, not least because collections by individual transects were more thorough than normally. At the same time, this sparseness of the Late Byzantine-Early Ottoman surface cluster cannot be completely attributed to external factors, ground visibility conditions or collection technique. There is evidence suggesting that the ceramic assemblages from this period were genuinely humbler in comparison to some other periods, because of dining habits and the availability of cheap metal ware, most commonly copper alloys or iron145. Only after the careful study of the individual transects collection did we become aware of the relation between the weak building remains southwest of the village and the small group of finds datable to the Late Byzantine-Early Ottoman Period. For the greater part the scatters of building rubble and Late Byzantine-Early Ottoman pottery are overlapping. The former extends over field blocks 14-18b, while the latter was mostly concentrated on the narrow terraces to the south, on field blocks 19-20 and 21-72. The Late Byzantine-Early Ottoman scatter was thus left out of the total grid survey covering field blocks in the northern half of the sector and field blocks 15-16, south of the dirt road linking the village with the monastic complex on the Vardar Valley floor (map III_44). Because of the ground configuration in the southern end of sector II, narrow discontinuous terraces separated by relatively steep stretches, it was decided to carry out total collections by individual field transects. The laying out of a regular grid in these conditions would have been impractical and time consuming. The detailed transect collections revealed a fairly large cluster of finds, extending from the western periphery of the modern village on the east, to a small limestone outcrop, about 250 meters to the west. These limits were already suggested in Appendix I. This site occupies one of the rare, physically well-defined micro-locations in the first survey. However there is no visible M. S. Filipović, 511, 1935; very close parallels have been encountered and studied in Greece, J. Bintliff, chapter 21, 2012. 145 A. Vionis, Material culture studies: the case of the Medieval and post-Medieval Cyclades, 177-197, eds. J. Bintliff, H. Stoeger, Medieval and post-Medieval Greece, the Corfu papers, Oxford 2009. 144 104 physical limit to the north; on this side the site limits are indicated by the sudden disappearance of the scatter of building remains and loose stone rubble. The site of the Late Byzantine-Early Ottoman village is further defined by the artifact densities recorded by the intensified transect survey (map III_44a and 44b, table 29, Appendix II). The more detailed collection of surface material defined a clear site threshold of over 6.5 fragments per 1000 sq meters. It includes the two southernmost rows of field blocks in sector II. From west to east, these are field blocks 19-20, 15b, 20a-20b, 18b, 21 and 72b-73. There are two separate cores characterized by artifact densities higher than 1 fragment per 100 sq meters. They correspond to the eastern and western scatters of building rubble, closely following the configuration of the local terrain. The eastern core is on field blocks 19 and 20 with 16.9 and 27 Late Byzantine-Early Ottoman shards per 1000 sq meters; the western core is situated about 130 meters to the west, on field block 21 with 22 fragments per 1000 sq meters. Given that all counted material was collected, this pattern is slightly changed. The western core is confined to field block 20, while the eastern core expands over field blocks 18b and 73. At the same time, both the maximum and the district average density increase, but the limits of the cluster remain the same. The fields occupied by this site are narrow artificial terraces, presently uncultivated and covered with stone rubble and artifacts. It is possible that we are dealing with original units of deposition, although their dating remains unknown. It is noteworthy that similarly built terraces, presumably gardens abandoned by the middle decades of the last century, were discovered at the valley floor and even on the Vardar Valley floor, near its confluence with the studied valley (field blocks 10 and 11 in sector I). Moreover the mosaic of terraces continues to the north of site 4, subdividing the sloping terrain west of the village as described in Appendix I. Some of these terraces, such as those in the northern half of the sector are still being used as separate agricultural parcels, but on fields south of the dirt road modern agricultural divisions show no regard for the old terrace system. The eastern group of building remains and the terrace wall delimiting them from the north were completely ploughed over on field blocks 17 and 18a-18b. It is quite probable that we are dealing with an older modification of the landscape, only partly incorporated into the modern land-use system. According to the individual transect collections there are no major concentrations of Late Byzantine-Early Ottoman material outside the southernmost terraces on site 4. However this material appears in smaller quantities on nearly all of the terraced fields. On field blocks north of the site the transect collections didn’t include Late Byzantine-Early Ottoman finds, but total grid collection on grid 4 recorded a density of 6.6 fragments per 1000 sq meters on a few find-spots. Low quantities of the Late Byzantine-Early Ottoman material were also found among the grid and the transect collections from the field blocks in the northern half of the sector and among the heaps of material in the western half of the sector, on field blocks 8a-9b. However apart from the very small field blocks 8e and 8d, artifact density never exceeds the limits of 2.7 fragments per 1000 sq meters of transect survey and 6.6 fragments per 1000 sq meters of grid survey. To the south on the floor of the studied valley, visibility conditions precluded systematic survey, but the complete absence of material other than modern rubbish and the micro-topographic location make it unlikely that settlement spread near the old river-bed. Unlike most of the sites from the Roman Period, in the case of the Late Byzantine-Early Ottoman settlement there is a clear coincidence between the distribution of building remains and the movable surface record. Moreover the original micro-topography of the site location and its immediate surroundings is relatively well preserved. Thanks to these factors and the relatively 105 sharp differences in the site and off-site artifact densities, we’re rather confident that Late Byzantine-Early Ottoman Sopot occupied an area of about 1 hectare. It is nevertheless questionable if this entire area was occupied by dwellings or portions of it were left to open space or meadows; an arrangement that one can still witness in villages where the traditional layout has survived. It is also possible that isolated dwellings or maybe even a smaller quarter went unnoticed on the overgrown fields in the western half of the sector. But in general the size of the site doesn’t contradict the information in the Early Ottoman censuses in that Sopot was a dependent hamlet with not more than 30 families in its heyday. Late Byzantine-Early Ottoman Sopot generated one of the more extensive continuous off-site carpets in this survey area. Rare fragments datable to the Late Byzantine-Early Ottoman Period were found even among some of the total collections from sectors III and V, half a kilometer to the north of the settlement. It is nevertheless evident that the main impact zone of the village was limited to sector II.As in most other periods of human settlement in the survey area, the settlement and its impact zone were limited to a single micro-topographic unit. In the case of the Late Byzantine-Early Ottoman village, the site halo extended for a maximum of 200 meters from the site limits. This was a fairly compact carpet with artifact densities ranging between 1.5 and 2.7 fragments per 1000 sq meters. It is important to note that this material was found on the terraced fields above the settlement, clearly indicating that it was originally discarded on these fields and not dispersed through site weathering and other secondary dislocations. This phenomenon known from other regional survey projects has been interpreted as resulting from outbuildings, rubbish disposal and intensive agricultural activities, primarily gardening146. In terms of settlement size, the Late Byzantine and Early Ottoman periods didn’t bring significant changes in the local history of human settlement. The central settlement retained the rank of a hamlet or a small village and as in many other periods, it was the sole focus of human habitation in the studied micro-region. The most obvious change from the previous epochs is the occupation of a previously uninhabited location. The earliest traces of human activity in sector II date to the Iron Age and there are also small amounts of Roman material, but these finds are either too scarce or comprise untypical assemblages. They were interpreted as off-site material or traces of non-residential activities. The location occupied by the Late Byzantine and Early Ottoman village features a number of particularities that distinguish it from the locations of earlier sites. Like site 5a-5b it lies on the edge of the fertile stretch at Prisoj’s southern foot, the watershed between the Vardar and the central valley and at the contact zone between the Quaternary deposits and the barren Mesozoic limestone rocks crowning the Taor Gorge. The Late Byzantine and Early Ottoman village was established on a lower ground, literally hanging over the edge of the foothills. This location is hidden from sight, at least for those following the modern lines of communication or passing over the higher stretches along Prisoj’s southern foot. It is practically its only defensive quality; otherwise the village could be easily accessed from literally any direction. It seems that the major disadvantage in comparison to site 5a-5b is the absence of a freshwater spring in the immediate vicinity. As explained in the preceding section, drinking water is still carried by pipes from the spring by site 5a, about 500 meters from the village centre. There are a number of possible explanations why the local community gave up this advantage of site 5a-5b. Apart from offering a better shelter, site 4 had an immediate access both to the banks 146 A suggestion that finds support in historical sources pertaining to the Late Roman and Byzantine periods, J. Bintliff, P. Howard, A. Snodgrass, 23, 2007. 106 of the central valley and the floor of the Vardar Valley. A narrow path running parallel to and below the modern dirt-road links the village with the small terraces in the eastern half of sector I, underlining the importance of this zone for the local community. These presently deserted stretches, apart from offering immediate access to running water are suitable for gardening.147 The dating of site 1, the small rectangular tower by the banks of the Vardar, next to field block 7 remains problematic, though it is quite possible it coexisted with the Ottoman village. A narrow path slowly falling into oblivion connects the tower and the narrow terraces on the Vardar’s left bank with the modern village, passing directly through the site of the Late Byzantine-Early Ottoman settlement. Local oral tradition recognizes it as the “Turkish guarding tower”, presumably guarding a crossing over the Vardar. Presently there is a wooden hanging bridge 120 meters upstream, but at least until the late 19th century transport of goods across the river was mostly carried out by rafts148. One can envisage a chain of similar towers along the banks of the Vardar, particularly in the narrow canyons along the river’s middle and lower course. There were no traces of the suspected platform, as this entire stretch is silted by sand and modern debris deposited by the river. The architecture of the tower described in Appendix I is indeed similar to the two standing towers in the modern village. Not surprisingly there were only a few fragments of worn bricks or roof tiles amidst the large amount of roughly hewn stone blocks. Their fabric and probable format are clearly different from the Early Modern and Roman tile fabrics. There is unfortunately no other datable evidence on the surface. III.1.13 The Late Ottoman and Early Modern Period (1800-1950 AD; tables 30-31, graphs 23-24, Appendix II) Because of the chronological proximity and the fact that this period is still very much present through many aspects of material culture, it was never recognized as a veritable subject of archaeology. Specifically in the Republic of Macedonia, as in most other former Yugoslav republics, the material culture and the living traditions of this period are studied by folklorists and ethnographers. Unfortunately with the exception of some earlier studies, few scholars from these disciplines have dealt with issues such as settlement size and location or rural economy. On the other hand, although there are studies of the various categories of material culture relevant to archaeology and architecture in particular, there is practically no communication between the two disciplines. As a result not too many archaeologists can recognize pottery or building remains from the 19th or the early 20th century. This situation presents a practical problem for the student of surface archaeology, even if one wishes to deliberately ignore this period. The surface of all regions inhabited during the past two centuries is usually always littered with vast amounts of debris produced by the Late Ottoman and Early Modern settlements. In many cases architectural remains are well preserved. In certain landscapes there are visible traces even of minor constructions, such as terrace walls, watermills or old cemeteries. It is not always easy to distinguish the remains of this and earlier architecture and the same problem surrounds the dating of certain pottery classes. But more importantly, it is theoretically unjustified to exclude the archaeological remains of the Late 147 The use of the power of running water could also have been a locational factor, J.L. Bintliff, The Ottoman era in the context of long-term settlement history. A case-study: the archaeological survey of the Valley of the Muses, Boeotia, Central Greece,203-229, Arab Historical Review for Ottoman Studies 19-20, 1999. 148 J. Cvijič, 241, 1906, this practice survived until the last decade of the 19 th century. 107 Ottoman and Early Modern period from the long-term study of a local settlement history149. Many villages have been completely deserted by the middle decades of the last century. Except for rare mentions in historical documents and in earlier studies, the architectural remains and the carpet of surface debris are the only tangible evidence for the study of these settlements. On no ground can one exclude this material from the rest of the surface archaeological record. Finally by studying the size of the settlement, the dispersal area and the distribution of off-site material, we reveal one of the possible ways in which a landscape was adapted to the purposes of human settlement and exploitation. This could greatly help in the understanding of patterns in the past, usually featuring far more poorly preserved surface remains and uncertain social and historical contexts. One must bear in mind that although there were important changes in the general living conditions, production and economy, for the local communities the subsistence base and the technological capacities changed little prior to the industrialization of the country in the middle of the 20th century150. The material dated to the Late Ottoman and Early Modern period was set apart thanks to a number of specific fabric features. Most prominent were the great solidity of the paste and the application of a vitreous glaze, usually of a poor quality and applied directly to the paste. Certain shapes and decorative techniques are also exclusively related to this time-period. Nevertheless in certain aspects this material can closely resemble earlier pottery production, particularly certain classes of Late Roman pottery. It also proved problematic to recognize at least one class of cooking ware fabric. An exception is the characteristic bread-baking pan, but this pottery class has been in use since the Middle Age. The same was the case with the material dated to the Late Byzantine and Early Ottoman periods and this is probably related to the local or regional specifics of pottery production or the increased usage of metal vessels 151. Distinct categories of coarse fabrics have been recently recognized among the material excavated from the Skopje citadel152 and among the Late Ottoman and Early Modern finds from the second survey area. Like the assemblages from most other periods, the various fabric groups dated to the Late Ottoman and Early Modern Period for the greater part had overlying distribution patterns. This greatly helped in the determination of one or two fabric categories of uncertain date. The similarity of the distribution pattern of the various fabric classes not only indicates that they are contemporary, but also reveals that they were all discarded in a similar fashion. This material doubtlessly consists of debris generated by the village during the last two centuries. Fragments of tile and pottery could find their way on the surface of surrounding fields either as a part of unsorted manure or simply as a result of the local habits of waste disposal153. Large amounts of modern building material, clothes and plastics can regularly be seen, especially on uncultivated fields close to the modern settlement and the local roads. The off-site debris produced by the Late Ottoman and Early Modern village is the second most numerous chronological group on the surveyed surface. The actual percentage is probably even greater, as not all tile fragments were collected during the transect survey. Although less numerous than the Late Iron Age assemblage, this still seems as a rather substantial amount considering the size of the settlement and the short duration of the period in question. These 149 J. Bintliff, 2012 , S. Davies, J.L. Davis, eds. 2007. S. B. Sutton, Settlement patterns, settlement perceptions: Rethinking the Greek village, 313-336, ed. P.N. Kardulias, Beyond the Site: Regional Studies in the Aegean, 1994. 151 A.K. Vionis, 177-197, eds. J. Bintliff, H. Stoeger, 2009. 152 Brief reports are available at www.skopskokale.com.mk 153 H. Forbes, 159-172, 2012. 150 108 finds are more than twice as numerous as the material broadly dated to the Roman Period, produced over a three times longer time-period. Moreover one should take into account that the Late Ottoman and Early Modern assemblage doesn’t include the artifacts that are still on standing monuments or into everyday use. This large discrepancy must be related to the temporal factor and the declining preservation rate, particularly of fragments discarded on the surface of fields. One can imagine that in 2000 years, worn fragments of tile and rare potshards will be all that remain of the continuous carpet of over 4500 shards collected from this survey area (see table 30 in Appendix II). The context and the origin of this material were already clear at the time of the field survey. There were no doubts that it was primarily accumulated through the repeated spreading of unsorted manure on the fields. This is also indicated by the pattern of dispersal revealed by both the transect and the regular grid surveys. The highest artifact densities were recorded in sectors II and IV, on the fields to the east and west of the modern village (maps III_46a and 46c). These fields are most easily accessed from the village and many are given to labour-intensive garden cultures, although there are also vineyards, cereal fields and meadows. Often the highest artifact densities were recorded on the fields adjacent to the village houses, such as field blocks 25 or 30. Although the artifact density is above the mean overall value of 10 fragments per 1000 sq meters on the majority of the field blocks in these sectors, on some of the units the amount of the Late Ottoman-Early Modern material is considerably diminished. In most cases these are abandoned fields or meadows, although small amounts of off-site debris were sometimes discovered even on cultivated fields. In sectors III and V on the fields to the north of the village, finds datable to the Late Ottoman-Early Modern Period never approach the very high densities recorded in sectors II and IV. Indeed along certain stretches at Prisoj’s foot, at a distance of only about 250 meters from the village periphery, this material completely disappears from the surface record (maps III_46a and 46c). These fields featured poor soils and were often abandoned or left fallow for longer periods of time. Quite appropriately the village cemetery is situated at the western end of this stretch, covered by field block 40. Here artifact density rose again to slightly over 1 fragment per 100 sq meters, although the origin of the finds is obviously different. These were table jugs and dishes left by the graves and often, intentionally broken. On the rest of the field blocks in these sectors the artifact density is close to the mean overall value and in comparison to sectors II and IV, there is little variation between the neighbouring field units. This zone continues uninterrupted into Prisoj’s eastern foothills (maps III_46b and 46d). The most distant field blocks featuring densities higher than the mean overall of 10 fragments per 1000 sq meters were located at a distance of over 1 kilometer from the centre of the modern village. In this sector there are evidently greater fluctuations in the artifact densities between the neighbouring field blocks, which again corresponds to the greater variety of land-use. Apart from overgrown fields there are meadows, parcels planted with cereals and vineyards. Note however that once it is assumed that the individual transect collections were limited to the number of finds counted during the quantification campaign, the variations in artifact densities between field blocks is less dramatic and most of the sectors belong to the zone of average density. Only in the northern end of sector VII, at a distance of about 1200 meters from the village, does the carpet of Late Ottoman-Early Modern finds become sparser, featuring consistently below 5 fragments per 1000 sq meters. The last survey sector where the transect survey recorded artifact densities higher than the survey’s average was sector VI, on the valley floor. In this part of the survey area there is an 109 apparent difference between the sector’s northern and southern half. On field blocks in the lower part of the valley situated below the modern village, the density of the Late Ottoman-Early Modern material usually exceeds the mean overall value and on one field unit, we recorded nearly 5 fragments per 100 sq meters or over 25 fragments per 100 sq meters, in case all counted finds were collected (maps III_46b and 46d). As in sector VII, there are considerable variations in land-use betweenthe neighbouring field units. On the field blocks north of the SkopjeThessalonica highway the density of Late Ottoman-Early Modern finds suddenly plummets well below the mean overall value, even on cultivated fields. These fields lie at approximately equal distance from the village as the field blocks in sector VII, but they are more difficult to access and agricultural exploitation is less intensive. Predictably the quantity of the Late Ottoman-Early Modern material is even lower in sectors east of the central valley, as well as in sectors I and XI (maps III_46c and 46e). On field blocks in these peripheral parts of the survey area, the artifact density rarely exceeds the limit of 5 fragments per 1000 sq meters. Large sections in sectors IX, X and XI are completely sterile. As in the rest of the survey area, off-site debris from the last couple of centuries was mostly discovered on cultivated fields. Thus in general the transect survey records point to two zones concerning the dispersal of the Late Ottoman-Early Modern material. One characterized by artifact densities close to or higher than the mean overall value of 10 fragments per 1000 sq meters is clearly limited to Sopot’s inner agricultural territory, the southern and eastern foothills of Prisoj. It extends for a maximum of 1 kilometer from the village centre in a northeast direction, along the western bank of the valley. In the other directions its dispersal is limited by topographic barriers: the steep slopes of Prisoj to the north and the V-shaped valleys and ravines to the west and east. Although the southern tip of the Ramnište Ridge is closer to the village than field blocks in the northern end of sector VII, it is more difficult to access, a factor that certainly played a decisive role in the distribution of field manure154. The other zone characterized by an ultra-thin and discontinuous carpet features a maximum density of 5 fragments per 1000 sq meters. More commonly however, field units feature not more than 1-2 shards per 1000 sq meters. This zone is spread over the peripheral parts of the studied landscape. These fields lie at distances greater than 1 kilometer or are less accessible from the village and cultivation is less intensive. We believe that such extremely low amounts of material datable to the last couple of centuries can be expected even in the most peripheral parts of all landscapes inhabited by stable agricultural communities. Despite the seemingly random fluctuations between neighbouring fields (table 31, Appendix II), the spatial distribution of the Late Ottoman-Early Modern finds in the survey area features a roughly concentric pattern. Emerging from the site of the modern village, these finds form a continuous carpet, covering the entire western half of the survey and the floor of the small valley. In the eastern survey sectors and along the upper course of the central valley, their quantity sharply decreases, but they are still present on the surface. This pattern can be linked to two closely related factors; distance and topographic configuration and to the type and intensity of land-use155. The former factor or the accessibility of the fields certainly determines the overall concentric pattern of distribution, while the localized variations between neighbouring field units 154 P. Howard, Spatial analysis of Boeotia field-walking survey data, 111-128, eds. J. Bintliff, P. Howard, A. Snodgrass, 2007. 155 P. Howards, 120-124, J.L. Bintliff, P. Howard, A.M. Snodgrass, 2007. 110 must be related to the differences in land-use, the varying strategies and wealth of individual farmers and the type of cultures grown156. In general we see a striking parallel to the infield-outfield system of cultivation known from historical sources and from intensive surface surveys of rural sites in Greece and the Near East157. The nearer and denser off-site zone marks the most intensively cultivated fields in the immediate vicinity of the village (III_map 45). Depending on the topography and the distribution of soils, it extends for between 500 meters to 1 km from the village centre. The cultures grown on these fields include gardens, vines and certain sorts of wheat. The local farmers regarded them as their most highly prized possessions and during the course of fieldwork we observed that much care was devoted in tending these fields. The outer off-site carpet coincided with the zone of outfield agriculture, but also with abandoned fields or fields with poorer soils. These fields are planted with less labour demanding cultures, such as barley or rye and are often left fallow every second year. Therefore the bringing of manure was probably deemed unnecessary by the local farmers, resulting in only sporadic finds that gave the low densities of 1-2 fragments per 1000 sq meters or a few fragments per field block158. Understandably other mechanisms could also contribute to the presence of off-site material, including casual loss, outdoor activities or specialpurpose sites. But the extensive and relatively structured carpet of surface material produced during the last couple of centuries must have been created through a prolonged and intensive process, such as the regular bringing of manure interspersed with household debris on the fields. Casual loss or non-residential activities could only have played a very minor part in the formation of this layer of surface material. Hence the density of the Late Ottoman-Early Modern off-site carpet is a good indicator of the character and intensity of agricultural exploitation across various parts of the surveyed landscape. Analyzing the presence of the basic functional categories in the collections of Late Ottoman-Early Modern material, we pointed to the large amount of architectural ceramics, mostly small-format, roof tile fragments (graph 24, Appendix II). They comprised around one half of the assemblage and their share in the off-site record even rose to over 60% in the more peripheral parts of the landscape. This is contrary to the suggestion that brick and tile are the least likely ceramic categories to find their way in the loads of manure carried to the fields159. Admittedly this suggestion pertains to the period of Antiquity, but the arguments presented also apply to the Late Ottoman-Early Modern Period. Tile was present in such quantities in the offsite record that recycling is very unlikely. It is possible that this prominence of roof-tile fragments even on the more distant fields signals a change in material culture standards. Except for a few ruined buildings, the old traditional roof tiles cannot be seen on the roofs of standing village houses. They are replaced by industrially produced tile and non-ceramic materials. It is clear that the Late Ottoman and Early Modern settlement is a direct successor of the pre-19th century village mentioned in the written sources and located on the narrow terraces, about 100 meters to the west. Although there was a slight displacement, the two settlements occupied essentially the same location. In terms of the proximity to various resources, it is evident that the floor of the central valley and the Vardar were important for both communities. The chief difference between the locations of the Late Byzantine and Early Ottoman and the Late Ottoman and Early Modern village is that the latter is larger and more spacious. Indeed 156 H. Forbes, 168-169, ed. R. Jones, 2012. J.L. Bintliff, P. Howard, A.M. Snodgrass, 23-26, fig. 4.3, 2007; T.J. Wilkinson, 31-46, 1989. 158 H. Forbes, 167, ed. R. Jones, 2012. 159 D.K. Pettegrew, 199-200, 2001. 157 111 measuring from the 1950’s map (maps III_42, 46a), it turns out that the area occupied by the Late Ottoman and Early Modern village is nearly 3 hectares or almost three times the are occupied by its Early Ottoman predecessor. In fact the need for space could be one of the causes for the displacement. However this casts a doubt over our earlier interpretations of the size of the Late Byzantine and Early Ottoman village. According to the Ottoman censuses for the year 1467-1468, 15th century Sopot is actually considerably larger than the 1930s village. Yet the distribution of the surface remains suggested that the Late Byzantine and Early Ottoman settlement occupied not more than 1 hectare. Two possible explanations come to mind. It is firstly possible that Late Byzantine and Early Ottoman Sopot was a more dispersed settlement and that the survey only managed to capture its central quarters160. The isolated farms and houses scattered across the surrounding fields would have left only faint surface remains, easily missed amidst the large amounts of material from several different periods. If this is the case, it would imply reinterpretation of the rank of earlier settlements as well, but also a significant increasein population during the Late and Post-Byzantine period. Recall that the area occupied by the houses and barns of the Late Ottoman and Early Modern village approaches the size of the Late Iron Age settlement. It also implies that the settlement remains that spread over an area of one hectare or less were of the rank of very small hamlets or even more likely, individual farms. But according to this reading of the data, it turns out that the large mound necropolis consisting of at least several dozen mounds was constructed by a larger hamlet or a small village. The other explanation assumes that the expanding settlement area of the Late Ottoman and Early Modern settlement has more to do with changes in agricultural economy, vernacular architecture and also perhaps in social relations, rather than simply reflecting increased population. The impact of the wider introduction of American cultures during the late 18th and throughout the 19th century has never been assessed, but considering their importance in the 20th century agriculture, they probably brought certain changes in local agricultural production and nutrition. Specifically in this region, tobacco was more important, but garden products, especially peppers and tomato are also grown. These cultures are most commonly grown on small parcels within the house-yards or very close to the houses, which has certainly expanded the space between neighbouring houses. Another important factor was the gradual dissolution of the old çiftlik system and the appropriation of larger tracts of arable land by the Christian peasants. In the old çiftlik type of settlement, only the houses of the landowners had spacious yards, usually surrounded by tall walls. The peasant-serfs lived together with the animals in small, single storied houses, often built one next to other and leaning against the wall of the landowner’s estate. By the beginning of the 20th century this type of settlement completely disappeared, at least from the regions along the upper Mid-Vardar. Two storied houses surrounded with separate yards have become the norm in many villages by the late 19th century161. It seems more likely that it was these developments in local rural society and economy and the related changes in living standards that brought about the expansion of the area of the Late Ottoman and Early Modern village, rather than a supposed population increase during the 14th and 15th centuries. In terms of population size and basic subsistence, Late Ottoman and Early 160 Cf. J. Bintliff, 469-470, 2012. M.S. Filipović, 500-501, 1935; J. Cvijić, Balkansko Polustrovo i Južnoslovenske Zemlje, I, Beograd 1918. Identically organized çiftlik settlements existed in Greece, J. Bintliff, 459-460, 2012. 161 112 Modern Sopot differed little from its predecessor and from most other earlier settlements in the survey area. III.2 Conclusions III.2.1 History of habitation in the survey area in relation to the regional context and the major historical circumstances As explained in the previous chapter, the survey area is an enclosed and well-defined micro-regional entity. It is one of the dozens small, lateral valleys draining directly into the Vardar. In the dry and hilly region of the Mid-Vardar, they present small oases of flat, cultivable land. The upper Mid-Vardar and particularly the stretch along the Taor Gorge lack larger, arable plains that could sustain larger communities. The nearest larger plains lie 10 and 15 kilometers to the south and east, the plain of Veles and the Ovče Pole Plateau. In such conditions one could predict that the small settlement niches in the rugged terrain of the upper Mid-Vardar were a secondary choice of habitat, occupied after the capacities of the larger basins were becoming fully exploited. Bearing in mind the present state of research of this kind, this will remain an untested hypothesis for a longer period of time. The existing archaeological data, mostly coming from small-scale excavations and extensive surveys rarely relates to issues such as settlement size, positioning in relation to the surrounding environment and are usually limited to certain time periods. As a consequence, it is rather difficult to compare them with the data from the Sopot survey, especially because quality and extent vary greatly across periods. Yet in order to obtain a fuller understanding of the developments in the survey area, it will be necessary to take into account the data available from the surrounding regions. Doubtless much more can be extracted from the published material in combination with targeted field surveys, but this by itself will require a separate study. Presently it will suffice to bring together the published data and shed a little light on the broader regional context, while summarizing the history of human settlement in the survey area. The earliest remains of permanent human habitation in the surveyed valley date back to the 6th millennium BC, during the period of the Middle Neolithic. There were no traces of Early Neolithic occupation. This period is known from excavations in the Ovče Pole region, where a number of single and multi-phase settlements have been attested162. In regional archaeology, the Middle Neolithic is usually seen as a period of stabilization and expansion, though the dating is far from clear163. The majority of the Neolithic sites from the Vardar Valley and the Skopje Basin are dated to the Middle and the Late Neolithic. Depending on the topography, these are settlements of the “tell” or toumba type, found in plains or open settlements, positioned on river terraces, in hilly regions. The Middle Neolithic settlement near Sopot was of the latter type. It was positioned on two river terraces, overlooking the turn of the central valley, near the geometric centre of the studied micro-region and the contact zone between the Tertiary and Quaternary sediments (map III_47). M. Gimbutas et al. Neolithic Macedonia vol. I, Los Angeles 1976; M. Garašanin, Centralnobalkanska zona 82-106, ed. A. Benac Praistorija Jugoslovenskih Zemalja vol. II, Sarajevo 1979;M. Garašanin, 87-100, 1983. 163 D. Zdravkovski, 56-57, 2008; D. Mitrevski, Prehistory in FYRO Macedonia 13-73, ed. D.V. Grammenos, Recent research in the prehistory of the Balkans Thessalonica-Athens 2003; V. Sanev et al. Urgeschichte Makedoniens, Skopje 1976. 162 113 There is an almost identically positioned, Middle and Late Neolithic settlement about 15 kilometers to the north164. Situated on a terrace on the Vardar’s right bank, it is surrounded by alluvial deposits. But it seems that the majority of the Early and Middle Neolithic sites have opted for the lighter lacustrine soils, covering the floors of all major basins in the country. At present the regions belonging to this geo-pedological zone have a rather dry and desolate appearance. This wasn’t necessarily the case in the deeper past; in fact some of these plains were marshy until the middle of the last century. Their light soils were particularly well suited for hand cultivation165. It should be stressed that this once fertile and lush plain was the setting for some of the earliest Neolithic settlements in the central parts of the Balkan Peninsula166. The second nearest known neighbour of the Middle Neolithic settlement near Sopot is situated about 15 kilometers to the east-northeast, deep into the Ovče Pole Plateau167. It too is positioned on a terrace, possibly occupying an area of slightly over 1 hectare. 20 kilometers to the east of the Middle Neolithic site near Sopot is one of the earliest Neolithic settlements north of Thessaly and the eponymous site for the Anza cultural group168. Again it is situated on a river terrace, at the contact zone between Quaternary colluvium and the vast sea of Neogene sediments. Back to the Vardar Valley, we already mentioned the Middle and Late Neolithic site at the northern end of the Taor Gorge. Unlike the settlements mentionedpreviously, it is focused on alluvial deposits, exploiting the wide river terraces of the Vardar just before it enters the Taor Canyon. Excavations on some of these sites have revealed that they were long-lived settlements, sometimes existing over a period of two millennia. For the regions south of Sopot and the Veles Basin, along the lower-mid Vardar Valley, information mostly comes from extensive surveys and the site locations are often very vaguely described 169. The nearest settlements are situated about 25 kilometers downstream from Sopot. Judging from the limited information, they too either occupied river terraces or were focused on the Tertiary lacustrine sediments, covering much of the regions along the lowerMid-Vardar. For the majority of settlements mentioned in the archaeological atlas, the chronology remains unclear (map III_48). Thus the Middle Neolithic settlement near Sopot was part of an extensive network of roughly contemporary settlements featuring a similar positioning and similar size170. It is near certain that we are only seeing a small part of the map of Neolithic settlements. The density of settlements at least in certain regions of the country must have approached the density observed in northern Greece. According to the scant data relating to their size, the sites discussed in the M. Garašanin, G. Spasovska, Novi Iskopuvanja vo Zelenikovo kaj Skopje, 85-117,Macedoniae Acta Archaeologica 1 1976. 165 In southern and central Greece the more stable, early farming communities are clearly concentrated on flood-plains or spring out-flow zones allowing for hoe cultivation with stone tools, J.L. Bintliff, E. Farinetti, 665674, 2006; A. Sherratt, Water, soil and seasonality in early cereal cultivation, 313-330, World Archaeology 11, 1980; different tendencies have been observed in Thessaly with very dense regular grid of settlements occupying a wider range of geo-pedological units, C. Perlès, The distribution of Magoules in Eastern Thessaly, 42-56, ed. P. Halstead, Neolithic society in Greece, Sheffield 1999. 166 M. Gimbutas ed. 283-286, 1976; M. Garašanin, 87-94, 1983. 167 V. Sanev, Neolitskata naselba Rug-Bair kaj s. Gorobinci, 203-247, Zbornik Naroden Muzej Štip IV-V, 1976;M. Garašanin, 89, 1983. 168 M. Gimbutas et al. 1976; V. Sanev et al. 24-26, 1976. 169 Most information comes from the archaeological atlas, Arheološka Karta na Republika Makedonija, 6085, vol. II, Skopje 1996. 170 Similar but much denser patterns have been observed in parts of northern Greece, Thessaly and Macedonia, D. Grammenos ed. Recent research in the prehistory of the Balkans Thessalonica-Athens 2003; C. Perlès, The early Neolithic in Greece: the first farming communities in Europe 121-152, Cambridge 2001. 164 114 preceding paragraphs occupied between 1 and 3 hectares. It has to be stressed however that these estimates refer to the size of the terraces, rather than to the size of the sites themselves. Nevertheless if we assume that the site area was equal to the topographic unit occupied, it puts them alongside the Mid-Neolithic settlement near Sopot in the rank of small to medium sized villages. The settlement near Sopot was only slightly smaller, measuring nearly 0.9 hectares. It is therefore highly predictable that future intensive regional surveys will reveal more sites of similar size and possibly other higher or lower ranking site categories171. In terms of population size, settlements measuring between 1 and 3 hectares were probably inhabited by between 80 and 300 inhabitants172. The Middle Neolithic settlement in Sopot was at the lower end of this range. This is not untypical even for later historic periods. As we saw in the first half of this chapter, in certain aspects the Middle Neolithic will lay the foundation for the local settlement pattern in later periods. During most of the history of habitation in the survey area, the local settlement will be of a similar rank and size, though its position will often shift considerably across the survey area. Later in this chapter we will try to examine if this continuity in the size of the local community is primarily related to demographic or environmental constraints. It is finally interesting to note that despite their small size and rank, Neolithic settlements along the Vardar Valley often occupy very open and exposed locations. The settlement near Sopot is perhaps one of the more extreme examples. As explained it occupies a central position in the small valley, poorly defended and very close to the main local west-east transversal. In this aspect too, the Middle Neolithic Period sets the precedent for a number of other prehistoric settlements of similar size and rank, which occupied equally open and unprotected locations. On most of the Neolithic settlements along the Vardar Valley and along its eastern tributaries life continued throughout the Late Neolithic, although a certain number of sites are known to have been abandoned towards the beginning of the 4th millennium BC. In the case of Sopot, the situation remains uncertain. The study of the pottery gathered recognized a separate assemblage, whose distribution mostly overlaps with the area of the Middle Neolithic site. These are doubtlessly the remains of a similarly sized settlement, exploiting the same resources as its predecessor. However it is uncertain if this is the direct successor of the Mid-Neolithic settlement or traces of a much later settlement. The problem arises from the character of the pottery, which is only vaguely related to the Mid-Neolithic production. Unlike the Mid-Neolithic material, limited to the site area and the surrounding site-halo, small scatters of the undated pottery were found dispersed across the entire eastern bank of the surveyed basin, hundreds of meters from the site core. Lacking secure dates for this assemblage, it is impossible to offer a reasonable explanation for these scatters. The fact that these are isolated groups of small numbers of fragments perhaps indicates that they came from enclosed deposits and were not part of a continuous off-site carpet. The implications are that these are either the remains of isolated dwellings, burials or some sort of votive offerings (map III_49). Phenomena such as these are known, but from excavations on later period sites. In A few years ago, rescue excavations along the newly built aqueduct channel east of Ovče Pole revealed a string of Neolithic settlements set at intervals of only a couple of kilometers, T. Jovčevska, Museum of Veles, personal communication. 172 Cf. J.C. Chapman, The early Balkan village, 33-53, Varia Archaeologica Hungarica 2, 1989; P. Halstead ed. Neolithic society in Greece, Sheffield 1999; C. Perlès, 173, 2001; for a synthesis see J. Bintliff, 55-56, 2012. 171 115 regionalarchaeology Late Neolithic sites are usually regarded as nucleated, although it has to be stressed that attention was rarely given to the areas surrounding the site core173. Regardless of the date of this assemblage, its distribution reflects a different attitude towards the physical surroundings. The occurrence of the abovementioned clusters of finds on several locations along the left bank of the valley, hundreds of meters from the site core marks the Ramnište Ridge as an integral part of the wider settlement area. This is the settlement’s impact zone or in the terminology of Czech landscape archaeology, the settlement’s area. 174 This will become the norm for the majority of later settlements in the survey area. Apart from the main concentration of finds marking the settlement, one often finds smaller clusters of contemporary material spread across the entire topographic unit on which the settlement was founded. By the middle of the 4th millennium BC at the latest, the site of the Middle Neolithic settlement and the entire basin of Sopot were completely abandoned. Compared to the Neolithic, very little is known about the following Eneolithic Period, particularly in the regions along the Middle Vardar Valley. Not a single Eneolithic site is listed in the archaeological atlas for the regions of Veles and Ovče Pole. In general the transition from the Neolithic to the Eneolithic remains a complete mystery. It is nevertheless evident that most of the known Eneolithic sites, at least in the northern parts of the country, occupied completely different location types. For example, the two known Eneolithic sites in the Skopje Basin occupy low, but well defended hillocks, overlooking larger alluvial tracts, which offered access to fertile land and control over natural lines of communication175. Globally this shift in the settlement focus away from the alluvial and lacustrine plains and onto the rugged, interfluve zone has been related to the so called Secondary Products Revolution: introduction of milk and diary products, wool, the introduction of ox-driven plough and more advanced molding techniques176. The complete abandonment of the Neolithic settlement pattern was also confirmed through excavations; the site of “Slatina”, on the right bank of the Vardar, 15 kilometers north of Sopot was certainly abandoned by the middle of the 4th millennium BC, as were most of the tellsettlements in the Skopje Basin177. In this respect the survey area fits well into the broader regional pattern. Even more limited is the information about human settlement in the Middle Vardar during the Bronze Age. Most of the information comes from reconnaissance and contains only brief remarks about the locations of sites178. Actually very little of this information is published, except for the short entries in the archaeological atlas (map III_50). Most problematic of all is 173 Evidence for dispersed Late and even Middle Neolithic settlement has come from Lower Macedonia, though the very large size of these sites is incomparable to our example, K. Kotsakis, What tells can tell – social space and settlement in the Greek Neolithic, 66-76, ed. P. Halstead, 1999. Far more comparable examples come from distant Crete and Laconia, K. Branigan, Late Neolithic Colonization of the Uplands of Eastern Crete, 57-65, the same volume; W. Cavanagh, Joost Crouwel, 121-150, W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al., 2002. 174 M. Kuna, M. Zvelebil, P. Foster, D. Dreslerová, Field survey and landscape archaeology research design: methodology of a regional field survey in Bohemia, 110-130, Památky Archeologické LXXXIV-2, 1993; E. Neustupný ed. 1998. 175 B. Babić, I. Mikulčić, 162, 1967; R. Pašić, Gradište kao tip naselja u Skopskoj Kotlini, 145-153, Materijali: Naseljavanje i naselja u praistoriji, X kongres arheologa Jugoslavije, Prilep 1976. 176 A. Sherratt, Plough and pastoralism: Aspects of the secondary products revolution, 261–305, eds. I. Hodder,G. Isaac, N. HammondPatterns of the past. Studies in honor of David Clarke, Cambridge 1981. 177 V. Sanev, 241, 1976a, M. Garašanin, G. Spasovska, 85, 1976. 178 Arheološka Karta na Republika Makedonija, 65-80, 344-361, 1996. 116 the absence of a finer chronology. Apart from a small number of Late Bronze Age sites or sites dated to the period of transition between the Late Bronze and the Early Iron Age, the rest are simply dated to the Bronze Age. In effect one wonders if the Bronze Age sites listed in the archaeological atlas are actually contemporary. Following the chronology for the region of the Central Balkans, the Bronze Age begins around 3000 BC and lasts until 1200 BC179. Sometime during this long period of nearly two millennia, the survey area was reoccupied. The Bronze Age settlement was unlike the Middle-Neolithic and it was focused on the western end of the survey area. In fact, the major concentration of finds was discovered outside the limits of the Sopot Basin, on site 3. This site was situated on the floor of the Vardar Valley on its left bank, by the confluence with one of its smaller tributaries. It was a tiny establishment, spreading over an area of less than 1000 sq meters or ten times smaller than the Middle Neolithic settlement and its undated successor. This individual farm occupied a very different type of settlement niche. It is much smaller and better defined than the open ridges drained by the valley of Sopot. The study of the material collected suggests that this was not the only site that could be broadly dated to the Bronze Age (map III_51). Faint traces of Bronze Age activity were also discovered on the later period sites 5a, 6 and 8. These locations are much more exposed and they are positioned in the midst of the most fertile and spacious stretch of land in the survey area, the foothills of Prisoj. The meager remains point to two significant changes in the local settlement during the Bronze Age. For the first time in the local history of habitation, the western bank was occupied, as well as the floor of the Vardar Valley. This shift in settlement focus is difficult to insert into a wider context, given the paucity of archaeological data and the very general chronology. One is tempted to view it as a belated effect of the advances in metallurgy, the so called Secondary Product Revolution and the introduction of animal traction180. This allowed the farmers of the late 4th millennium BC to expand their agricultural territories outside valley bottoms and flood-plains and exploit difficult but fertile terrains181. During most subsequent periods the main local settlement will always be situated in the western half of the basin, though not as close to the Taor Canyon as the Bronze Age farms. Also for the first time in this area, a fully dispersed type of settlement appears. Because of the small number of finds and the poor understanding of local and regional Bronze Age chronology, it remains unclear if these were contemporary farms, exploiting different parts of the landscape or if only one farm was active at a time. If the vestigial remains on site 6 and 8 truly belong to the Bronze Age, it reflects a careful dividing of the area, with settlement foci at intervals of 400-500 meters (map III_51). This pattern of extreme dispersal explains why the Bronze Age is so poorly represented in the archaeological atlas182. The great majority of Bronze Age sites known from the regions of Skopje, Ovče Pole and Veles were found only thanks to the presence of later period remains. These sites show considerable diversity regarding the types of location they occupy. The majority were discovered on forts from later periods, which is doubtless related to the predominant method of field survey. But a few sites mostly found by accident were positioned 179 W. Heurtely, Prehistoric Macedonia, Oxford 1939; V. Sanev et al, 29-30, 1976; A.F. Harding, European Societies in the Bronze Age, Cambridge 2000, especially 9-14; for the regions of the Central Balkans, the absolute chronologies is closer to that of Greece and the Aegean, although radiocarbon or dendro-chronological sequences are lacking. 180 A. Sherratt, 261, eds. I. Hodder,G. Isaac, N. Hammond, 1981. 181 J.L. Bintliff, E. Farinetti, 665-667, 2006; for Greece in general, see J. Bintliff, 2012, chapters 3 and 4. 182 Cf. J. Bintliff, P. Howard, A. Snodgrass, 139-168, 1999. 117 on open terrain, close to streams. Unfortunately there is no mention of their size, either because only a small probe was opened or because of the multi-period character of the sites. In any case this type of open locations by streams was also occupied by the small farms found in the Sopot survey. The closest parallel comes from the shores of a small artificial lake, 5 kilometers southeast of Sopot183 (map III_50). As in most other cases, the size was not documented, partly because the cluster of surface finds was covered by large amounts of Late Roman tile and pottery. Like the small farms near Sopot, it occupies an open terrain at the junction of two streams. The Middle Neolithic and the Bronze Age settlement represent the two extreme modes of habitation in the survey area: a completely nucleated settlement, with no traces of off-site activities and a network of isolated farms, positioned at roughly equal distances. In most other time-periods, the local settlement will neither be completely nucleated, nor completely dispersed. It must be stressed however that the low chronological resolution prevents us from directly contrasting these two patterns. As mentioned in preceding paragraphs, it is impossible to know if all Bronze Age sites were actually contemporary or if the revealed pattern was created in successive stages of abandonment and relocation184. The community that re-occupied the survey area during the Bronze Age was evidently smaller than the Middle Neolithic community or their successors on site 11. Even assuming that the revealed network of farms was contemporary and incomplete, their combined size would still be much smaller than the nucleated settlements on site 11. In fact low population could be the factor that allowed settlement dispersal to such a degree. During the Bronze Age each of the settlement foci occupied a separate microtopographic unit, sufficiently large to sustain an extended family or a clan. On the other hand, the larger nucleated Mid-Neolithic settlement occupied a more exposed, central location giving access to various sections and resources in the studied landscape. A small group of finds belonging to a single fabric group could be dated more narrowly to the Late Bronze Age. This is a problematic category because it appears alongside different assemblages, including the small Bronze Age assemblage on site 3, in the monastic complex. Isolated fragments were also found scattered across sites 7 and 8, but not on sites 5a and 6. It would be groundless to suggest any elaborate interpretation, as we don’t know if this material represents a separate epoch or if it is a part of some larger assemblage. In any case, the distribution of these finds traces the way of future developments in the local settlement history. Obviously there was some activity on site 3, on the Vardar Valley floor, but the exposed southern foothills of Prisoj were completely abandoned. The majority of the finds of this group were collected from the more sheltered eastern foothills, 700-900 meters upstream from site 5a. This part of the landscape will become the most favored settlement location in the coming centuries (mapIII_52). As in most other parts of the Balkan Peninsula and the Aegean, the period between 1200 and 800 BC is poorly understood. Interestingly almost all sites dated to these centuries are necropoleis. The few settlements dated to this period consist almost exclusively of hillocks fortified in the Hellenistic or the Late Roman Period185. Particularly problematic is the chronology of the pottery material. Pottery production of the time retains many of the Late Arheološka Karta na Republika Makedonija, vol. II, 60-61, 1996. According to researchers working in Greece and the Aegean the latter was the likelier scenario, J. Bintliff, P. Howard, A. Snodgrass, 139-140, 1999; J.F. Cherry, 19, eds. C. Renfrew, M. Wagstaff, 1982. 185 D. Mitrevski, Protoistoriskite Zaednici vo Makedonija preku pogrebuvanjeto i pogrebnite običai, Skopje 1997; D. Mitrevski, 13-73, ed. D.V. Grammenos, 2003. 183 184 118 Bronze Age traditions and it is often difficult to distinguish between genuine Late Bronze Age material and the Early Iron Age retentions. In regional archaeology it is generally accepted that these were turbulent times, marked by instability and violent migrations. It has to be stressed though that these conclusions are mostly based on the evidence from excavations on sites in the Lower Vardar Valley186. In stark contrast to the scarcity of evidence from the surrounding regions, in the survey area settlement was returning to the pre-Bronze Age level. Two very different assemblages were dated to this period. The presumably earlier, featuring some typical Late Bronze Age characteristics, mostly came from site 7. Smaller, isolated clusters were collected from a wider area, stretching from Prisoj’s eastern foot to the fields east of the modern village. It greatly resembled the distribution pattern of the possible Late Neolithic assemblage, spread on the opposite eastern bank. Site 7 is smaller than site 11, but the satellite clusters are more substantial and more numerous (map III_53). Moreover at least one or two clusters were almost certainly swamped by the much larger Late Iron Age cluster on site 8. Even if not all of these clusters were dwelling remains, it is evident that population increased during this period. A new nucleated settlement was established, approaching the rank of the Mid-Neolithic settlement and its successor on site 11. The closest parallel comes from the lower Mid-Vardar, 30 kilometers south-southeast of Sopot187. A site called “Stolot” near the village Ulanci is situated on a low flattened top and spreads over to a lower terrace of similar size, on the Vardar’s left bank (map III_50). This is an old eroded terrace, cut by streams on the eastern and western side. It has an excellent control over the alluvial plain and a direct visual communication with Stobi, one of the main cross-roads in the region, also inhabited at the time. Excavations revealed traces of what is still considered the earliest stone architecture in the country. This was doubtless a site of some importance, but it barely measures 5000 sq meters, slightly larger than the central settlement near Sopot. 200 meters to the north, on a higher and similarly shaped plateau, the researchers found a highly organized necropolis, stretching over a distance of at least 60 meters. It was estimated that it contained about 100 cists. This fairly close proximity to the settlement indicates that at least some of the clusters surrounding site 7 were possibly necropolis remains. The problem lies in the much greater dispersal of the clusters documented by the Sopot survey in comparison to the communal nescropolis near “Stolot”. 188 It should be stressed that site 7 was a newly occupied location in the survey area. By the turn of the 1st millennium BC, all activity seems to have shifted to Prisoj’s eastern foothills and the fields east of the village, closer to the western bank of the central valley. Not a single fragment was found west of the dirt-road leading north, along the right bank of the valley. Thus as in the case of the possible Late Neolithic settlement, the settlement’s impact zone was limited to a single micro-topographic unit, Prisoj’s eastern foothills. It is likely that apart from isolated dwellings and various outbuildings, this area also marks the most intensively cultivated part of the landscape. Interestingly the main concentration of finds on site 7 was located at the very edge of the dispersal area of this assemblage. Both this and the settlement on “Stolot” are located in the Vardar Valley, the main natural corridor linking the Aegean with the Central Balkans. It I. Mikulčić, 11, 1966; M., Garašanin, Contributions à la chronologie de l’àge du fer en Macédoine, 173186, Živa Antika X 1-2, 1960, B. Hänsel, Kastanas. Die Grabung der Baubefund, Prähistorische Archäologie im Südosteuropa, Band 7-2, Berlin 1989; W. Heurtely, 217-218, 1939. 187 D. Mitrevski 37-39, 44, 1997. 188 Ibid. 35-37. 186 119 should be stressed that most other known sites from this period are situated along major natural corridors, though this could merely reflect the better state of research in these regions189. If the adopted dating is correct, this was a relatively short-lived settlement. The first typically Iron Age finds were collected from two different sites at Prisoj’s southern foot. Sometime during the first couple of centuries of the first millennium BC, sites 5a and 6 were reoccupied. The chronology of this material is not exactly determined. It is evidently very similar to the Late Iron Age fabrics and bears little resemblance to the assemblage collected from site 7. It is thus quite possible that there was a brief abandonment of the entire micro-region, towards the end of the II millennium BC190. It seems that by the time Prisoj’s southern foothills were reoccupied, site 7 was completely deserted, along with the entire eastern foothills. To a certain degree the Bronze Age pattern was restored, but the two clusters of Iron Age finds were much larger. Each occupies an area of about 4 000 sq meters. Individually they are smaller than site 7, but their combined areas are slightly larger, nearly equalling the size of site 11. As in the case of the Bronze Age farms, it is impossible to decide if the two settlements were contemporary or successive. Another obvious characteristic of the Early Iron Age settlement is that there lack traces of satellite clusters. We were able to identify only a single satellite situated in-between the settlement sites and in addition, a few isolated finds were collected from the fields at the southern foot of Prisoj, east of site 6. In other words this period sees a return to a more nucleated pattern of settlement, similar to the Middle Neolithic. Because almost all finds datable to this period are confined within the site limits, it is difficult to determine the impact zone. Given that the central sites were truly contemporary settlements, it is likely that the settlement on site 5a exploited the area of the modern village and the fields to its west, while site 6, the eastern half of Prisoj’s southern foothills(map III_54). Unfortunately there lack comparative data from the surrounding regions of Veles, Ovče Pole and Skopje. The few known sites dated to the Early Iron Age are almost exclusively fortified hill-tops, while there is no mention of the existence of a corresponding necropolis191. On the other hand, only a few of the Late Iron Age mound necropoleis have burials earlier than 800 BC192. While it remains uncertain if population remained stable or slightly dwindled during the first two centuries of the first millennium BC, it is clear that there was absolutely no continuity with the settlement at Prisoj’s eastern foot. There was not only a dramatic shift in settlement location, but also greater nucleation of settlement and an almost complete break with earlier pottery production. This development doesn’t contradict the generally accepted view of the historical conditions at the time. According to many scholars, this is the period of the Great Migrations and the surveyed area being situated by one of the main interregional corridors would have certainly felt the impact of the supposed large-scale invasions. Similar drastic changes in D. Mitrevski, 45, 1997; I. Mikulčić 9-10, 1966. Some authors tend to date similarly described material slightly later, towards the 9 th and the 8th centuries BC. Z. Georgiev, Basarabi stilot i Makedonija 81-96, Macedoniae Acta Archaeologica 13, 1992; J. Stankovski, Tumulot I od tumularnata nekropola kaj s. Strnovac, opština Staro Nagoričino, 135-153, Macedoniae Acta Archaeologica 18, 2008. 191 B. Georgievski, Praistoriski gradišta od kumanovsko 51-68, Macedoniae Acta Archaeologica 13, 1992. 192 D. Mitrevski 71-81, 1997. 189 190 120 pottery production have been observed on excavated sites in the Lower Vardar, in layers dated around the turn of the millennium193. There are two main reasons to be reserved about this scenario. Firstly, unlike in Kastanas or Wardarofca (Axiochorion) the chronology of the finds from the Sopot survey is far from clear. The assemblages from sites 5a and 6 are barely determined as earlier than the securely dated Late Iron Age assemblage from site 8 and later than the assemblage on site 7. Little else can be done at present, as only future research can help establish a more refined chronology. Secondly, we saw that dramatic shifts in settlement locations occurred even during relatively peaceful and prosperous periods. One always has to bear in mind that these were fairly small communities, which could easily abandon their old settlement sites in the face of local events and processes. In fact even in the heart of the Mycenaean territory it is difficult to link the destruction layers on the major palatial centres to discrete historical or environmental events194. Towards the end of the 9th century BC there sets in a longer period of stability in most regions of the Balkan Peninsula, along the Danube and in the Eastern Mediterranean 195. Locally it is known as the “Developed” or the Late Iron Age196. The term “developed”, actually reflects the view that by this time, a distinct local culture was established, with its own characteristic pottery production, metallurgy and burial rites. Although some of the Late Bronze Age traditions survived, the pottery of the 7th and 6th century BC is much easier to distinguish than material from the Dark Ages. Most of the larger settlements often continuing their existence into the Hellenistic and Roman Period were established by the time of the Late Iron Age. As elsewhere this is a much better understood period. The first excavations on the Late Iron Age sites date back to the 1950’s197. However this is greatly assisted by the fact that the predominant type of burial rite during the Late Iron Age was under mounds. Usually they are not particularly large, but because they appear in groups, often on deliberately chosen and exposed locations, they were easily spotted during the first regional archaeological surveys in the 1950’s and 1960’s. Since then much has been learned about the burial rites and other aspects of the Late Iron Age material culture, but almost no progress has been made in the study of settlements and settlement patterns. In fact for most mound necropoleis, including the one in the survey area, the locations of the corresponding settlements were unknown. Due to the prevailing survey method, the known settlements from 193 W. Heurtely, 98-99, 217-218, 1939; A. Hochstetter, Kastanas: die Handgemachte Keramik 276-283, Prähistorische Archäologie in Südosteuropa, Band 3, Berlin 1984. 194 O. Dickinson, The Aegean from Bronze Age to Iron Age, continuity and change between the 12 th and 8th century BC, London-New York 2006. 195 R. Vasić, Kulturne grupe starijeg gvozdenog doba u Jugoslavii, Belgrade 1973; A.M. Snodgrass, The Dark Age of Greece, Edinburgh 1971; O. Dickinson, 98, 216-217, 2006. 196 In this study we followed the periodization proposed by D. Mitrevski 1997; D. Mitrevski, Prilog kon poznavanjeto na Donovardarskata-Paionska grupa na Železnoto Vreme 145-159, Macedoniae Acta Arhaeologica 12, 1990-1991, in the broader context of the central Balkans, this is Iron Age II, R. Vasić, Gevgeliska grupa starijeg Gvozdenog Doba, 637-38, ed. A. Benac, Praistorija Jugoslovenskih Zemalja V, Sarajevo 1987; M. Garašanin, 173186, 1960. 197 M. Garašanin, D. Garašanin, Arheološka iskopavanja u selu Radanju, kod lokaliteta “Krivi Dol” 9-61, Zbornik na Štipskiot Naroden Muzej I, 1959; I. Mikulčić, Izveštaj sa probnog iskopavanja halštatske nekropole kod s. Star Karaorman 95-112, Zbornik na Štipskiot Naroden Muzej I, 1959; V. Sanev, Mogilite of Kunovo Čuki kaj s. Orizari-Kočansko, 7-27, Zbornik na Arheološkiot Muzej VIII-IX, 1978. 121 this period almost exclusively consist of fortified hillocks198. It is somewhat paradoxical that this type of Iron Age sites is completely absent from the regions featuring higher concentrations of mound necropoleis; even if present, they were almost never related to a nearby mound necropolis. Looking at the distribution of mound necropoleis and hill-forts in the regions of Veles and Ovče Pole, it is possible to get some idea about the density and distribution of the Late Iron Age settlements, although we can only speculate about their size and exact location (map III_55). It has to be stressed that this information mostly comes from extensive surveys, which is to a great extent problematic, because of the re-introduction of mound burials during the Roman Period. The latter can be distinguished by their construction and much greater size, but due to the fact that the majority of these mounds have suffered greatly from agricultural activity, one needs to be cautious with the chronology proposed in the survey reports or in the archaeological atlas. Finally, it has to be stressed that mound burials were not the sole type of funerary rite during the Late Iron Age. By the 6th century BC, skeletal burial in flat necropoleis became the norm for the regions of Pelagonia and the Lower Vardar Valley, but they also sporadically appear in the areas of the Middle Vardar and the Bregalnica199. Mound burials are also conspicuously absent from the Upper Vardar, despite the great similarities with the region of the Middle Vardar in other aspects of the burial rite and material culture in general. The mound necropolis near Sopot has been securely dated to the 7th and 6th century BC. The material excavated during the 1980’s finds very close parallels among the material from earlier excavations on the mound necropoleis along the Bregalnica River, the largest eastern tributary of the Vardar200. These sites are located 40-45 kilometers east of Sopot, as the crow flies. However, similarly constructed mounds have been registered on a number of sites in the intervening regions. At least three separate mound necropoleis have been documented along the eastern edge of the Ovče Pole Plateau and at least a couple in the region to the north 201. Further west, in the central parts of the plateau, they become scarcer. Only isolated mounds are reported in the archaeological atlas202. The next mound necropolis to the west is much more similar to the Sopot necropolis, consisting of a larger number of smaller mounds, constructed entirely of rounded stones. It is situated less than 8 km northeast of the mound necropolis near Sopot, in the hilly region separating the Veles Plain and the Ovče Pole Plateau, along its western edge 203. Though the exact location of the settlement remains unknown, the surrounding landscape is very similar to the Sopot Basin. This is a minor stream that presently drains into a small, artificial lake, 5 kilometers southeast of Sopot. It is one of the typical settlement niches that dot the region of the Middle Vardar. I. Mikulčić, 17-35, 1982; Z. Georgiev, Tri pred-rimski naselbi kraj Pčinja 91-101, Godišen Zbornik na Filozofski Fakultet 17-18, 1990-1991; B. Georgievski, Praistoriski gradišta od kumanovsko 51-68, Macedoniae Acta Archaeologica 13, 1992. 199 Z. Georgiev, Praistoriski naodi od Kočani i prašanje na ramnite nekropoli po Bregalnica 65-79, Godišen Zbornik na Filozofski Fakultet 7, 1981; D. Mitrevski, 146, 1997. 200 Gj. Petački, 71, 1986; D. Mitrevski 86-93, 1997. 201 Arheološka Karta na Republika Makedonija, vol. II, 204-224, 344-362, 1996. 202 Although isolated mounds are more typical for the Roman Period, one should allow the possibility that only meager traces survived from the original group. There are a number of examples of completely levelled mound necropoleis from the Iron Age, but in most cases, they consisted of a smaller group of several to a dozen mounds; D. Mitrevski, 90-91, 1997. 203 Arheološka Karta na Republika Makedonija, vol. II, 77 1996. 198 122 Apart from the mound necropolis near Sopot, Iron Age mounds are rare along the Vardar Valley. Only three sites are mentioned in the archaeological atlas for the entire Veles Region, all three being isolated mounds204. In contrast the mound necropolis near Sopot consists of nearly 100 mounds, approaching the size of some of the large necropoleis in the Bregalnica Valley205. It is very unlikely that this is a simple result of better preservation. Rather this seems to indicate the greater rank and size of the Late Iron Age centre near Sopot, but primarily its slightly later date206. Earlier mound necropoleis tend to consist of not more than several mounds. The individual mounds are larger and display a more formal and elaborate construction. It has been pointed out both by scholars working in the country and in the region that these tendencies of popularization and simplification of the sepulchral ritual reflect profound changes in social relations207. Specifically in the Republic of Macedonia, they have been related to the gradual demise of the old social order based on strict division into clans and lineages208. A further problem in judging the significance of Late Iron Age Sopot in a regional context is the existence of a number of sites with flat necropoleis. Some of these are partly contemporary with the mound necropolis near Sopot and they evidently belonged to larger settlement centres. Such is the case of Byla Zora, the later Paionian capital, situated less than 20 kilometers east of Sopot, in the centre of the Ovče Pole Plateau. It’s been justly observed that this divergence in burial customs during the 7th and the 6th century within a relatively small region reflects socio-cultural, rather than chronological or regional differences209. The mound burials near Sopot keep the old Iron Age tradition, while the cist burials in flat necropoleis pave the way for future developments. It is surely indicative that the majority of the known mound necropoleis are concentrated in hilly and peripheral regions. Both Sopot and its nearest neighbour to the east, as well as the supposed centres along the eastern edge of Ovče Pole belong to small, lateral drainage basins. The excavated mound necropoleis on the left bank of the Bregalnica belong to similar landscapes. On the other hand, the central parts of the Ovče Pole Plateau or the Veles Basin, either lack major mound necropoleis or lack any remains from the Late Iron Age. Historical evidence pertaining to this period is rather scanty, but compared to earlier periods, we have some notion about the people that inhabited the regions along the Vardar Valley and about the political and economic relations with their neighbours. Although the earliest direct historical sources date to the 5th century BC, it is possible to infer some conclusions about the situation in the preceding centuries. Thus the archaeological findings from the 7th and 6th centuries are the earliest that can be put into a historical context210. Arheološka Karta na Republika Makedonija, vol. II, 77, 79-80, 84. I. Mikulčić, 95-103, 1959; M. Garašanin, D. Garašanin, 9-51, 1958-1959. 206 The salvage excavation on the mound necropolis near Sopot was limited to its southern end, but the researchers were rather confident that the entire necropolis dates to the 7 th-6th century. D. Mitrevski, 107, 1997; Ð. Petački, personal communication. 207 I. Morris, Burial and Ancient Society: the rise of the Greek city-state, Cambridge 1987; O. Dickinson, 174-177, 2006. 208 I. Mikulčić, 61-62, 1966; D. Mitrevski, 106-107, 1997; it has to be pointed out though that although larger, the earlier mounds could hardly provide enough burial space for the communities that built the later mound necropoleis, such as the one near Sopot. A likelier explanation is that during this early phase, the privilege of mound burial was limited only to a very narrow section of society. 209 D. Mitrevski, 103-105, 1997. 210 D. Mitrevski, 72-73, 1997; E. Petrova, Paeonia, 3-26, Skopje 1999. 204 205 123 The mound necropoleis, like all other remains datable to the Late Iron Age have been related to the Paionian tribes211. There are a number of historical records relating the Paeonians with the territory of the present-day Republic of Macedonia. A later historical source from the Hellenistic Period mentions that future Paionian kings performed ritual bathing in the river Breglanica prior to their crowning212. But much earlier sources also clearly point to the Vardar Valley and the Pelagonian Plain as the home of the Paionian tribes213. The Paionians were an important political factor in the regions of the central and southern parts of the Balkan Peninsula. Sometime in the sixth century BC, they were organized enough to lay siege to Perinthos, on the east Thracian Coast214. There were also intensive political and trade relations with the Greek city-states, both in mainland Greece and in Asia Minor. Certain Paionian tribes are known to have minted silver coins, using Hellenic standards215. Silver and wood were most probably the main export goods, the Vardar certainly playing an important role in the trade. Another testimony for the close relations (probably trading) with the Hellenic world is a series of peculiar bronze pendulums, discovered as votive offerings at the sanctuaries in Delphi and Olympia, in the mid decades of the last century216. Although lying at its very edge, Paionia was an integral part of the Aegean and this seems to have been established by the time of the Late Bronze Age, at the latest217. The much larger number of known Late Iron Age sites can’t be solely explained through research bias or their better state of preservation compared to remains of earlier prehistoric periods. There is an evident expansion, both in numbers and in monumentality of the sites, in most regions along the Vardar and its tributaries 218. Essentially the Sopot survey confirmed these developments on a micro-regional level. The Late Iron Age settlement was at least 3.5 times larger than the possible Late Neolithic hamlet, the second largest settlement in the surveyed area and almost 4 times larger than its Early Iron Age predecessor. This near quadrupling of size was accompanied by a radical shift in the settlement focus. For the second time, the focus of settlement moved from Prisoj’s southern foothills to the west bank of the valley, at Prisoj’s eastern foot. As in the case of the assemblages from most other periods, the Late Iron Age finds were found outside the narrower site area, either isolated or in small clusters. But for the first time we find them spread over most of the western half of the valley. Smaller clusters were found even on sites 5a and 6, suggesting that these earlier settlement locations were not completely abandoned (map III_56). The mound necropolis, stretching almost along the entire length of the easternmost ridge drained by the valley of Sopot completes the local Late Iron Age landscape. It further illustrates the expansion of settlement during the Late Iron Age. The entire lower half of the basin was clearly demarcated, its natural limits coinciding with the territory of the local community. It has to be admitted that the Late Iron Age doesn’t appear as a simple episode of local settlement expansion. This settlement is of a completely different rank and scale, unprecedented in the local history of human habitation. Although covered only by extensive surveys, it is 211 D. Mitrevski, 187-188, 1997; E. Petrova, 47-50, 1999. E. Petrova, 20, 1999 (Polyaen IV-12, 31). 213 E. Petrova, 3, 1999, Iliad X, 428-429 214 E. Petrova, 6, 1999, Herodotus V, 1. 215 E. Petrova, 93-97, 1999. 216 E. Petrova,21, 1999. 217 D. Mitrevski, New aspects on the Bronze Age sites on the northern periphery of the Mycenaean World, 449-457, ed. B. Hänsel, Man and environment in the European Bronze Age, Kiel 1998. 218 D. Mitrevski 72, 1997. 212 124 difficult to suspect that there are similarly-sized mound necropoleis in the neighbouring microregions, still awaiting their discovery. If this is taken as a measure of its social and demographic rank, Late Iron Age Sopot clearly stood apart from its immediate neighbours. While in most other periods of the past, the survey area was alternately occupied by small farms and hamlets or completely abandoned, the Late Iron Age brought local settlement to the rank of a medium-sized village. If the proposed interpretation of the survey results is correct and given that a settlement between 15 and 30 households occupies an area of about one hectare, a settlement area of at least 3.6 hectares could accommodate between 60 and 90 households219. This can also be inferred from the summarily published results of the excavations on the mound necropolis. Although smaller than those on earlier necropoleis, each of these mounds contained between 3 and 7 graves and many featured reuse of earlier graves or later interments, dug into the tumular mass after the mound construction was completed220. This evidence of frequent reopening of the mounds and modifications of the original construction indicates that the necropolis didn’t expand. If credit is given to the claims that the entire necropolis consisted of nearly a hundred mounds, it is very likely that it belonged to a settlement far more substantial than its predecessors221. Even though the tumular necropolis near Sopot belongs to the later group of larger necropoleis, featuring smaller and simpler mounds, it is very unlikely that the right of mound burial was granted to all sections of society222. If this was the case, the supposed 100 mounds could accommodate not more than 2, 2.5 generations, even if we assume that each mound contained up to 10 burials223. An alternative explanation is that the community was smaller consisting of about 40-50 households, all members being given the right to mound burial. But the thick ritual layer discovered in the tumular mass and surrounding the individual graves points to elaborate funerary rites. It is difficult to accept that such ceremonies were carried out for all members of the local community. Because of the size of the settlement revealed through surface artifact survey and because of the ideological connotations surrounding funerary mounds, we believe that the right to formal burial on the necropolis was reserved only for certain sections of society and that the total population size is greater than suggested by the extent and the hypothetical capacity of the mound necropolis. Late Iron Age Sopot was both quantitatively and qualitatively different from the farms and hamlets of the Late Bronze-Early Iron Age or the Middle Neolithic. The suggested figure of between 60 and 90 households roughly equals a population range of 300-450. This is certainly a very dramatic increase from earlier periods. In the concluding section we’ll see that even a population of slightly over 150 can cause problems in maintaining social relations on a face-toface basis. This implies that the Late Iron Age community near Sopot featured some form of 219 Estimating population on the basis of settlement size is still regarded as highly problematic, not least because of variations across different regions and periods, different economic and social backgrounds, J.L. Bintliff, Further considerations on the population of ancient Boeotia, 231-252, ed. J. Bintliff, Recent Developments in the History and Archaeology of Central Greece, Oxford 1997; see also the excurse on the rural population in the southern hinterland of Thespiai, J. Bintliff, P. Howard, A. Snodgrass, 144-145, 2007; for other regions in Greece, W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al. 133, 147, 205-211, 2002; M.H. Jameson et al, 539-567, 1994; for a critical overview see R. Osborne, Demography and survey, 63-72, eds. J.F. Cherry, S.E. Alcock, Side – by - side survey: comparative regional studies in the Mediterranean World,Oxford 2004. 220 D. Mitrevski, 92-94, 1997. 221 D. Mitrevski, 92, 1997. 222 I. Morris, 61-62, 1987. 223 This is not very realistic however, because the excavations showed that in certain mounds only half of the available space was used; D. Mitrevski, fig. 24, 1997; unfortunately the salvage excavations on the mound necropolis still await their proper publication. 125 horizontal, if not vertical stratification224. It is possible that these social arrangements are actually reflected in the organization of the mound necropoleis, with separate clusters of between several and a dozen mounds. One is tempted to suggest that the mound necropolis was used exclusively by the local leading clans, competing for prestige through the construction of higher, more massive mounds and performing lavish funerary feasts. The terminal date of this settlement remains unknown. In most regions of the Vardar Valley, settlements established during the Iron Age continue their existence at least until the late 4th century BC225. There are apparent and significant changes in many aspects of culture and society during the following two centuries. Excavations on the majority of the mound necropoleis indicate that this burial rite was almost completely abandoned by the middle of the 5th century BC226. In many cases however, the Iron Age necropoleis continued to be in use until the Hellenistic Period. On some necropoleis the newly introduced rite of cremation was combined with small mound-like constructions, while on flat necropoleis, cists built in the 6th centuries were emptied of their contents and reused, the old relics carefully placed in the corner of the cist227. All of this indicates that on most known sites there was a strong continuity from the Late Iron Age through the Late Archaic and Classical Periods. Although insufficient the survey and excavation data pertaining to settlements of this period, also confirm that life went on uninterrupted on most known settlements in the Vardar Valley. There are no reasons to believe that the large settlement near Sopot contradicted this pattern. The absence of the characteristic Hellenized shapes and decorative techniques among the pottery finds could rather reflect the strong, conservative character of the local tradition of pottery production, but also reduced contacts with the south and perhaps a certain decline in importance of the old tribal centre. Given the large number of different fabric groups collected from site 8, it is quite possible that the settlement on site 8 continued its existence at least into the 5th century BC, though its exact size and relation to its predecessor remains an unknown. Research has so far revealed that the regions along the Upper and the Lower Vardar experienced divergent socio-historical developments after the late 4th century BC. Most of the sites known from the regions of Skopje and Polog, as well as from further north in the Morava Basin, were certainly abandoned by this time228. Survey and excavation data confirm that these were substantial and wealthy settlements, maintaining close relations with the Aegean 229. All of this came to a sudden end by the late 4th century BC. Sites from the Hellenistic Period are very rare in the regions of Skopje or the Pologs and the local Hellenistic material is almost completely 224 J. Bintliff, Settlement and territory, 505-545, ed. G. Barker, Routledge Companion Encyclopedia of Archaeology, London 1999. 225 V. Sokolovska, 1986; I. Mikulčić, Topografija na Stenae-Prosek-Demir Kapija 65-78, Godišen Zbornik na Filozofkski Fakultet 15, 1988-1989; D. Mitrevski et al. 2005; www.tafhr.org information on both, Vardarski Rid and Byla Zora. 226 D. Mitrevski 146-151, 1997. 227 I. Mikulčić, 197-209, 1965; Z. Georgiev, Karakterot i značenjeto na vongrobnite naodi vo južnopovardarskite nekropoli od Železno Vreme, 37-53, Macedoniae Acta Archaeologica 6, 1980; S. Karpuzova, N.P. Pecovska, Iskopuvanje vo selo Korešnica-Demir Kapija, 77-86, Macedoniae Acta Archaeologica 14, 1996. 228 This was confirmed both through excavations and extensive surveys; I. Mikulčić, M. Jovanović, Helenistički opidum u Krševici, kod Vranja 355-375, Vranjski Glasnik IV, 1968; I. Mikulčić, 20, 1982; V. Sokolovska, 26, 1986. 229 I. Mikulčić, V. Sokolovska, Grobnicata vo Brazda, kaj Skopje 79-92, Macedoniae Acta Archaeologica 11, 1990; It came as a great surprise when a small-scale excavation on a site in the South Morava Valley, 300 km north of the Bay of Thessalonica produced evidence of Red Figured pottery, antefixes, and classical jewelry; I. Mikulčić, M. Jovanović, 355-56, 1968. 126 unrelated to earlier traditions or to the material culture of the deeply Hellenized regions in the southern parts of the country. In contrast to the situation in the north, life continued during the Hellenistic Period on the majority of larger sites established in earlier centuries, along the Middle and the Lower Vardar, along the Bregalnica and in the Pelagonian Plain. But here too, it is possible to observe certain changes in the settlement pattern230. Specifically in the regions of the Mid-Vardar Valley and on the Ovče Pole Plateau a number of new settlements were established or gained greater importance. The mentioned Paionian capital of Byla Zora, dominating the Ovče Pole Plateau continues to flourish231. On the other hand there is very little evidence coming from the areas with a greater concentration of mound necropoleis during the Late Iron Age, probably indicating decline and contraction. The earliest substantial habitation levels from Stobi also date to the Hellenistic Period232. Further north along the Vardar, two new centres emerge. There were no excavations or systematic surveys on these sites, but there is clear evidence of occupation during this period233. One is situated about 15 kilometers to the north of Stobi, on the Vardar’s right bank, the other, another 15 kilometers further upstream, on the southern shore of the artificial lake, less than 5 kilometers southeast of Sopot (map III_57). Unlike the older Late Iron Age centres, these new settlements occupy very exposed, central locations, controlling key communication routes, but also larger tracts of fertile land. They are either focused on the alluvial soils along the Vardar Valley or on the lacustrine sediments of the larger basins. It is difficult to accept the view that the hilly, peripheral drainage basins were completely abandoned, but they certainly lost their importance by the late 4th century BC. Archaeological remains of a more monumental character will return to these areas only in the Late Roman Period, although as the survey results showed, they weren’t completely abandoned. Similar tendencies have been observed in the Pelagonian Plain and in the regions along the Bregalnica234. The divergent developments in the northern and southern parts of the country and the changing settlement dynamics in the regions along the Middle and the Lower Vardar, the Ovče Pole Plateau and the Pelagonian Plain are usually related to the known historical circumstances during the 3rd and 2nd century BC235. The sudden disappearance of the Hellenized centres in the regions of the Upper Vardar Valley towards the end of the 4th century is commonly explained through a number of related demographic processes236. The main event marking the beginning of this period is the appearance of the Celts on the Mid-Danube and the subsequent invasions of the Greek mainland. Historical sources confirm that both Paionia and Macedonia were overrun and plundered on the way. This major destabilization was accompanied by two related demographic 230 231 The most detailed evidence still comes from the Pelagonian Plain, I. Mikulčić, 80, 1966. I. Mikulčić, Ubikacija na Bila Zora, 147-165, Godišen Zbornik na Filozofkski Fakultet 2, 1976; www. tfahr.org I. Mikulčić, Stobi vo pred-rimsko vreme, 101-117,Godišen Zbornik na Filozofkski Fakultet 13, 1985; M. Šurbanovska et al. 271-282,2008. 233 A. Keramitčiev, Beleški od rekognosciranjata vo strumičkiot basen i po dolinata na Vardar 115, Zbornik na Arheoloshki Muzej III, 1961; I. Mikulčić, Antički Gradovi vo Makedonija, 152-154, Skopje 1999; for the process of Hellenization of material culture in general, D. Mitrevski, On the Hellenization process of the Iron Age culture in Macedonia, or on the early contacts between Macedonia and the Aegean regions 191-203, Thracia 11, 1995; V. B. Grozdanova, Spomenici od Helenističkiot Period vo S.R. Makedonija, Skopje 1987. 234 I. Mikulčić, 80-81, 1966, I. Mikulčić, 148-149, 1999, mentions the possibility of colonization in both regions during the 3rd century BC; for the region of Bregalnica see Z. ,, Bregalničkiot Bazen vo Rimskiot i Ranosrednovekoven Period 9-41, Zbornik na Štipskiot Naroden Muzej VI, 1990. 235 N.G. Hammond, T.G. Griffith, A history of Macedonia vol. II, Oxford 1979; E. Petrova, 1999. 236 I. Mikulčić, 20-21, 1982; V. Sokolovska, Pajonskoto pleme Agrijani i vrskite so Damastion 9-32, Macedoniae Acta Archaeologica 11, 1990. 232 127 processes. It’s been assumed that after Alexander’s conquests in the East and during the ensuing clashes between his successors, a considerable portion of the population of Upper Macedonia and Paionia moved to the East, whether to serve in the Macedonian armies or as colonizers237. This assumption is supported by the attempts of later Antigonid dynasts to establish new or recolonize existing settlements in Upper Macedonia and Paionia, especially during the 3rd and the 2nd century BC. It is surely no accident that town names such as Antigonea or Perseida suddenly appear in the written sources relating to the 3rd and 2nd century BC238. Further evidence comes from their readiness to grant land to large contingents of foreign tribesmen in exchange for military and political alliance, during the decades between the Second and the Third RomanMacedonian Wars239. The cause behind these desperate measures lies in the constant inroads of Dardanian and Thracian raiding parties into former Paionian territory, a condition that became particularly precarious with the arrival of the Romans on Macedon’s western border. The last Antigonids were certainly aware of the upcoming perils and the final assimilation of the Paionian territories marked by the conquest of Byla Zora in 217 BC must have had the ultimate aim of imposing greater control over the northern borders of the Kingdom. By the middle of the 2nd century BC, the Paionians were obviously not in a position to fulfill the role of a buffer against the tribes of the Central Balkans240. The differences in the archaeology of the Hellenistic Period in the northern and southern parts of the Republic of Macedonia are thus clearly emphasized by a political and territorial frontier. The regions along the Middle Vardar Valley, the Ovče Pole Plateau, the valleys of the rivers Bregalnica (ancient Astibo) and the Crna (ancient Erigon), all former Paionian territories, remained under the control of Macedon, while the regions along the Upper Vardar and the Kumanovo Pass were presumably left in Barbarian hands or were turned into a sort of no-man’s land, if we’re to trust some ancient sources241. This line of division could be one of the earliest political frontiers in the region, defended by fortifications at strategic points and well-connected to bases in the interior to enable coordinated military actions242. Its efficiency is best recognized by the fact that it survived the Kingdom of Macedon for over 150 years. The Romans will retain the same line of defense until the time of the Early Empire, when the frontier was pushed to the Danube Valley243. Can we follow the impact of these ethnic and political divisions on the regional and local settlement pattern? Regarding infrastructure this early frontier was a far cry from the later Roman limes. In this respect it is impossible to claim if the survey area belonged to the sphere of control of Macedon’s kings and the later Roman governors, solely on the basis of distribution of standing monuments. But considering the fact that the former Paionian capital was only about 16 237 E. Petrova 24, 1999. I. Mikulčić, 80, 1966; I. Mikulčić, Problemot na Antigonea 111-135, Godišen Zbornik na Filozofkski Fakultet 11, 1983; F. Papazoglu, 1988. 239 F. Papazoglu, Srednjobalkanska plemena u pred-Rimsko doba, 147-150, Sarajevo 1969. 240 I. Mikulčić, 149-165, 1976; E. Petrova 25, 1999. 241 I. Mikulčić, map 1, 1976; F. Papazoglu, 24-133, 1969. 242 There is plenty of evidence in the chronicle of Titus Livius, particularly in the sections covering the years of the Second Roman-Macedonian War, book 31-40; Livy, The dawn of the Roman Empire, translated by J.C. Yardley, Oxford 2000. 243 F. Papazoglu 133-34, 1969. Due to a lack of intensive and systematic survey data, the evidence comes from the distribution of coins, particularly the autonomous series of Macedon’s towns and the coins minted by the first Roman governors. Very frequently they were collected from the same hill-forts, guarding the northern entrances of Macedon. N. Šeldarov, Macedonia and Paeonia, Skopje 2003. D. Donev, Some observations on the northern frontier of the ancient MacedonianKingdom 29-63, Macedonian Heritage 26, 2005. 238 128 kilometers to the east and taking into account the natural geographic divisions in the region, it seems more likely that this area remained within the limits of ancient Macedon. Situated at the southern exit of the Taor Canyon, the physical barrier that separates the upper and the middle course of the Vardar and providing quick and easy access to the larger settlement centres in the Ovče Pole Plateau and the Mid-Vardar, the survey area was surely of a great strategic value to anyone ruling the Mid-Vardar Valley. However the intensive survey revealed only the remains of a tiny settlement, a larger farm or a small hamlet, unrelated to all earlier settlements and withdrawn from the main natural corridor connecting the Vardar Valley with the Veles Basin and the Ovče Pole Plateau. The main focus of occupation was obviously site 12, stretching over an area of less than 2000 sq meters, although as in many earlier periods, much smaller, isolated clusters were found dispersed across the entire topographic unit, hundreds of meters from the site core (map III_58). Despite the small size of the settlement, the distribution of the satellite clusters and the total dispersal area of the Hellenistic pottery suggest a fairly extensive impact zone, equal to that of the possible Late Neolithic settlement. This could indicate that the majority of the satellite clusters are remains of dispersed residential quarters. Both the character of pottery production and the location of this site, suggests that there was little or no continuity with the Late Iron Age traditions. In all likelihood the area was abandoned for some period of time after the demise of the Late Iron Age settlement and prior to the establishment of the small Hellenistic settlement on site 12. An equally dramatic change is the reoccupation of the eastern bank or the Ramnište Ridge. During most of the local history of habitation, the main settlement was on the western bank of the central valley, exploiting more fertile stretches of land and exerting closer control over the exit from the Taor Gorge. In this context it would be very interesting to establish at least a rough dating for the post – Middle Neolithic assemblage on site 11. The size and location of site 12 certainly don’t reflect the military or political presence of ancient Macedon. It is possible that some of the Roman fortifications in the area have Hellenistic foundations; particularly the small fort over St. George’s monastery, at the very exit of the canyon, but there lacks decisive evidence244. In fact the nearest site where occasional finds from the Hellenistic Period are reported is the abovementioned hill-fort on the southern shore of an artificial lake, 5 kilometers to the southeast of Sopot. Although there were no excavations or systematic surveys, rare fragments of Hellenistic pottery and a number of coins minted in the early decades of the 2nd century BC are sure signs that the site was occupied prior to the Roman conquest245. This too is a strategically important location, as it guards the entrance to the plain of Veles and blocks the route that leads parallel to the Vardar, by-passing the Taor Canyon and the survey area. This site was also in control of a plain that had far greater agricultural potential than the small valley of Sopot. In essence this is the northern half of the Veles Basin, measuring over 15 sq kilometers of arable land. It represents a very similar type of location to that of the fort “Gradište”, Knežje, the supposed Paionian capital of Byla Zora. It too is situated in the central part of the Ovče Pole Plateau, rather than in the immediate vicinity of the mountain passes coming from the north. Both cases seem to illustrate the shift from the peripheral valleys to the central plains246. Unfortunately we lack even the most basic data for the Late Iron Age mound necropolis centres along the eastern edge of Ovče Pole or the one near the village Ivankovci, 8 kilometers northeast of Sopot. I. Mikulčić 102, 1982. N. Šeldarov, 260, 648, 670, 690, 2003. 246 D. Donev 51-52, 2005. 244 245 129 The overall character of site 12 and its very presence indicate something else. It is unusual for such small and undefended settlements to appear in a peripheral region, with a constantly impending threat of invasion. One would rather expect establishments of such rank in conditions of social and political security and uninterrupted access to larger markets. This could mean that the Dardanians and the other invading tribes didn’t use the valley of Sopot and the corridor through the Vardar Valley, but the more open routes, running parallel to the Vardar and avoiding its narrow canyons247. When describing the surveyed basin and its place in the regional geographic context, it was mentioned that without an active transversal along the Vardar Valley, it loses its strategic value, becoming but a small and withdrawn outlet of the Veles Basin. Only similarly designed research in the small valleys to the east of Sopot can confirm or reject this thesis. Even if the survey area wasn’t directly exposed to Barbarian raids, the small size and the location of the Hellenistic settlement suggest that living conditions in the region were nevertheless precarious. Otherwise it is difficult to explain the withdrawn location of site 12, as well as the abandonment of the more fertile western bank, in favor of the poorer soils of the Ramnište Ridge. The local settlement dynamic clearly indicates withdrawal and contraction during the Hellenistic Period, although it is very likely that this was more related to changes in the settlement pattern in the former Paionian territory, rather than to external instability. It wasn’t possible to point to a more narrow date for the Hellenistic assemblage from Sopot. This is locally produced pottery, bearing only faint resemblance to the more recognizable Hellenistic pottery classes. It is nevertheless evident that it shows even less similarity to the pottery of the Roman Period. It is quite possible that site 12 and the entire survey area were briefly abandoned in the first centuries of Roman occupation. But on the other hand, the total grid survey on this location did reveal finds datable to the Roman Period along the western periphery of the Hellenistic settlement. One must bear in mind the historical circumstances in the regions of the Upper and the Middle Vardar Valley during the Late Hellenistic and Early RomanPeriod248. While the territory under control of the Antigonid dynasty fell under Roman control as early as 168 BC, immediately after the end of the Third Roman-Macedonian War, the regions to the north were only turned into a Roman province towards the middle of the 1st century AD. Although the Roman presence was certainly felt among the population of the Central Balkans by the middle 2nd century BC, the local material culture becomes fully Romanized only after the Roman conquest and the establishment of the first colonies249. Thus Scupi situated less than 60 km to the north of the survey area became a Roman colony almost two centuries after the Roman conquest of Macedonia250. Being located by the turbulent northern frontier of the newly created province and in the vicinity of the main regional corridor, it is quite probable that the studied micro-region became more intensely inhabited only towards the end of the Early Imperial phase. There is not a single example of the characteristic red slipped pottery or the imitations of Eastern Sigillata, Also suggested by I. Mikulčić, 150, 1976. F. Papazoglu, 138-143, 1969; also F. Papazoglu, 1988. 249 The history of the modern regions of Skopje and Kumanovo in the Early Roman Period, the southern end of Upper Moesia at the time, along with the epigraphic evidence are studied in B. Josifovska-Dragojević, Inscriptions de la Mésie supérieure, Regions de Skopje et Kumanovo, vol. VI, Belgrade 1982. For the region of the central Balkans in general, see A. Mócsy, Pannonia and Upper Moesia, London 1972. 250 B.Josifovska-Dragojević 3, 1982; archaeological evidence mostly comes from excavations in Scupi, L. Jovanova 197-198, 2008, I. Mikulčić, From the topography of Scupi, 29-35, Archaeologica Iugoslavica XIV, 1973. 247 248 130 featuring so prominently in the local production at Stobi. On the other hand the more familiar Late Roman wares, such as Gray Macedonian are well represented251. It was already explained that the period of the Late Republic and the Early Roman Empire are solely known from excavations on large urban centres, such as Stobi or Heraclea Lyncestis, in the south of the Pelagonian Plain. We know very little about what happened to the rest of the Hellenistic towns. Current excavations on the acropolis of Byla Zora confirmed the chronology proposed by earlier researchers. The site was abandoned by the 2nd century BC252. However survey and accidentally excavated material show that the great majority of Hellenistic settlements along the Middle Vardar continued their existence after the Roman conquest, despite the strong expansion of Stobi253. Only on one site, 15 kilometers north of Stobi, is there clear evidence of decline and possible abandonment. It has to be stressed however that without exception, these are larger and fortified settlements. Only the enclosed area often measures several hectares. There is no or very little archaeological data pertaining to settlements of minor rank. This raises the familiar dilemma of whether the Roman Period brought nucleation and urban expansion or an overall population increase254. Unfortunately the situation is not much different when the Early and Mid-Imperial periods are in question. Although over 70% of the sites listed in the archaeological atlas for the regions of Veles and Ovče Pole are broadly dated to the Roman or the Late Roman Period, a gigantic increase, especially in comparison to the Hellenistic Period, only a small number are dated more narrowly to the Early Roman Period. Moreover the majority of these “sites” are inscribed or decorated tombstones and statues, often found dislocated, with no certain information about their origins. But in cases when it is possible to relate them to certain sites, they are useful chronological indicators. It’s been long accepted by scholars of the period that the erection of inscribed or decorated stones for various purposes almost completely ceased by the beginning of the 4th century AD255. In fact in the Republic of Macedonia most of these monuments can be dated more narrowly between the late 1st and the late 3rd century AD256. Thus the period of the 1st century BC and the 1st century AD remains problematic, most of the information coming from excavations on larger civic centres and necropoleis257. Another phenomenon marking the Early Imperial Period, especially in the eastern parts of the country, but also along parts of the Middle Vardar is the re-introduction of mound burials. Only a few of these mounds have been excavated and the earliest were dated to the late 1 st – V.Stojanović-Anderson tab. 4.3-4.12, 1992; E. Maneva 21-32, 1979; M.Ončevska-Todorovska, tables 34-35, 2004. 252 I. Mikulčić, 163-164, 1976; www.tfahr.org. 253 I. Mikulčić, 1999; although the author’s opinion was rather the contrary in some earlier publications; I. Mikulčić, 133, 1983. 254 Cf. S.E. Alcock, 1993; J. Bintliff, 315-318, 2012. 255 This epigraphic decline is characteristic for most regions of the Balkan Peninsula, except for a few larger urban centres on the coast, A. G. Poulter, The transition to Late Antiquity, 1-50, especially 11-12, ed. A.G. Poulter, The Transition to Late Antiquity on the Danube and Beyond, Oxford 2007; J. Bintliff, 360, 2012. 256 B. Josifovska-Dragojević,1982; F. Papazoglu, Istoriski priliki: Lihnid vo vremeto na ranoto Carstvo 101-113, Ohrid i ohridsko niz istorijata, Skopje 1985. 257 A. Wessolowski, Burial customs in the west cemetery 97-141, eds. J. Wiseman, Dj. Mano-Zissi, et al. Studies in the Antiquities of Stobi I, T. Veles 1973; I. Mikulčić, 48-52, 2003; E. Nikolova 247-271, 2008; I. Mikulčić, 29-35, 1973, I. Mikulčić, Ranorimski skeletni grobovi iz Skupa 89-102, Starinar XXIV-XXV, 1975; E. Maneva,1979; I. Mikulčić, 55-72, 1999. 251 131 early 2nd century AD258. Some of the few scholars that have dealt with the problem of these mounds attribute them to certain Thracian tribes, settling in the Bregalnica Valley by the 1st century BC. It is symptomatic that they often appear in the areas of earlier, Late Iron Age mound necropoleis, sometimes incorporating the older mounds. This conservative tradition has actually survived into the 4th century AD259. As in the case of the Iron Age mound necropoleis, the corresponding settlements have not been located. It has to be noted that these were most probably individual or family burial chambers, an alternative to the Hellenistic-Roman traditions of erecting family mausolea in the countryside (map III_59)260. The distribution pattern of these two categories of archaeological sites is obviously not representative of the actual settlement pattern during the Early and Mid-Imperial periods. A large portion of the sites broadly determined as Roman in the archaeological atlas were probably established by the late 1st, early 2nd century AD. The increased number of sites datable to the Middle and especially the Late Roman Period, to a large degree reflects the monumental character of archaeological remains in the period between the 2nd and 6th century AD. It doesn’t mean that this evidence needs to be downgraded. Evidently there was an expansion of building activity, both in the towns and in the countryside, but the basis for this development could go back at least to the 1st century AD. Looking at the distribution of sites dated more narrowly to the Early and Mid Imperial periods, it has to be remembered that these are nearly always remains of necropoleis or more rarely sanctuaries261. Moreover they are often securely characterized by researchers as small family mausolea, which explains their frequent appearing in fairly large concentrations in certain areas. Nevertheless one cannot deny that by the 2ndcentury AD, many of the smaller peripheral basins in the regions of the Middle Vardar and the Bregalnica were reoccupied. In a number of cases, the sites of the local settlements were roughly determined. According to the cursory descriptions, the majority is situated on gentle, lower terraces, close to the arable land, but there are also examples of occupied hill-tops, not necessarily fortified. In some cases they occupy areas of at least 2 or 3 hectares, which is a fairly large rural settlement by local standards. Perhaps one of the most noticeable features of the Early to Mid-Imperial landscapes is the unusually high density of sites on relatively small areas. In certain parishes along the Middle Vader Valley, extensive or architectural surveys have recorded up to 4 sites per 1 sq kilometer262. Often there are combinations of one or two possible settlement sites and a number of smaller, burial or other special-purpose sites, obviously resembling the distribution of Roman sites in the survey area. Admittedly the number of finds in the survey area datable prior to the 4th century is too small, but they are clearly distributed on at least two different sites. The farm on site 12 was reoccupied or continued to exist for some time in the Roman Period and at the same time, site 5a was reoccupied (map III_60). To these finds, one should add the two funerary stelai with inscriptions, brought (or found) in the monastic complex. It is also possible that the small M. Garašanin, D. Garašanin, Arheološke beleške sa rekognosciranja u Istočnoj Makedonji 69-95, Zbornik na Štipskiot Naroden Muzej II 1961; M. Garašanin, D. Garašanin, Iskopuvanja na mogilata Krst, kaj selo Tarinci 65-69, Zbornik na Štipskiot Naroden Muzej II 1961; Z. Beldedovski, 17-27, 1990. 259 V. Sanev, 7-20, 1978. 260 Information is taken from the archaeological atlas and surveys of Early Roman architectural sculpture and inscriptions, V. Lilčić, Makedonskiot kamen za bogovite, Hristijanite i za život po životot, vol. I, Skopje 2001; B.Josifovska-Dragojević, map1, 1982. 261 V. Lilčić, 461-463, 2001. 262 V. Lilčić, 245-246, 351-361, 2001; B. Josifovska-Dragojević, Izveštaj za arheološkoto rekognosciranje po dolinata na Sreden Vardar 106-127, Zbornik na Arheološkiot Muzej IV-V, 1966. 258 132 fort above the monastery was built at the same time. We should recall that the border between the provinces of Macedonia and Moesia followed the line of the northern frontier of the Kingdom of Macedon, probably passing very close to the survey area263. As it proved impossible to define a separate Early to Mid-Imperial pottery assemblage, there is no basis to speculate about the size or the rank of these sites. It is in any case evident that this was a period of stabilization and expansion of local settlement. In a sense the situation was slowly returning to the pre-Late Iron Age level. The main settlement was re-instated on site 5a for the third time in the history of habitation in the survey area. The reoccupation of this exposed location and the possible building of the small fort at the very exit of the Taor Gorge indicate a renewed security within the broader region, but also perhaps a reactivation of the Vardar Valley as the main interregional corridor. This is after all the shortest line connecting Thessalonica, Stobi and Scupi and the stable conditions brought by the Pax Romana enabled its normal functioning264. This reorganization of the local settlement during the period of the Principate laid down the foundations for further settlement expansion during the period of the Late Roman Empire. There are almost no mentions of the regions along the Vardar Valley and the province of Macedonia in general, during the Early and the Mid-Imperial periods. The situation changes drastically after the end of the 3rd century AD. Apart from the historical narratives of various chroniclers, a lot of valuable information comes from official documents, such as itineraries or the lists of imperial dignitaries and ecclesiastical sources265. On the one hand, they inform us about developed urban life and road networks. Most of the earlier larger towns grew into Episcopal sees and we hear of a number of new urban centres, especially in the eastern regions of the country, along the Middle and the Upper Bregalnica. On the other hand, the historical narratives covering the period between the 4th and the late 6th century abound with violent events. Most prominent are the records of Barbarian invasions, especially the Gothic invasion of 378 AD, the Hunnic invasions of the 440’s, the raids of the Ostrogoths in 479 AD and the repetitive incursions of the Slavs and Avars during the 6th century266. In some of these incidents the Barbarian invaders used the Vardar Valley, either on their way to the Plain of Thessalonica or on their way back to the Danube. The sack of Stobi and Scupi is mentioned by a number of reliable sources and it is confirmed by archaeological evidence on both sites267. But despite this chronic instability, buildings were repaired or built anew, not only in Scupi and Stobi, but also in other urban centers. There is however evidence of considerable contraction of the urban cores. This is especially evident in Stobi and in Heraclea Lyncestis, in the south Pelagonian Plain268. The Late Roman walls of Stobi enclosed an area of about 15 hectares, nearly three times smaller than the I. Mikulčić, 210-212, 1999; I. Mikulčić, 20-21, Munich 2002. Which doesn’t mean that the parallel routes were closed; it is quite possible that a cart-road functioned through the canyons of the VardarValley, while the caravan road followed the gentler but slower route through the low mountain passes, 4 km east of the canyon. A similar situation has been documented on the hostile terrain of Demir Kapija (ancient Stenae), 70 km south of the TaorCanyon. I. Mikulčić, 65-78,1988-1989. 265 I. Mikulčić, 9-19, 2002; F. Papazoglu, 8-21, 1988. 266 A more recent overview of the period, with a particular emphasis on the eastern part of the Balkan Peninsula is given by J.H.W.G. Liebeschuetz, The Lower Danube under pressure: from Valens to Heraclius, 101134, ed. A.G. Poulter, 2007; for the central and western Balkans see J-P. Sodini, The transformation of cities in Late Antiquity within the provinces of Macedonia and Epirus, 311-336, the same volume. 267 I. Mikulčić, Some new factors in the history of Stobi 205-229, Eds. J. Wiseman, B. Aleksova, Studies in the Antiquities of Stobi, vol. III, T. Veles, 1981; D. Koračević, 2002; J-P. Sodini, 313, ed. A. Poulter, 2007. 268 I. Mikulčić, 254-255, 1999; I. Mikulčić, 108-112, 2003; the phenomenon is much more widespread, J-P. Sodini, 320, 2007; cf. the example of Thespiai in Boeotia J. Bintliff, A. Snodgrass, 57-61, 1988. 263 264 133 fortified area of the Early Imperial town. In Heraclea Lyncestis the fortified area was limited to the former town-square and its Christian basilicas. The case of this town vividly illustrates the fate of most urban centres that have survived the end of the 4th century; they literally became identical with the Episcopal see. In many Late Roman towns most of the urban core consisted of churches and palaces of the church dignitaries269. Nevertheless excavations in Stobi, Scupi and Heraclea Lyncestis have revealed that the civilian populations continued to dwell in these settlements, at least until the late 6th century AD. Living conditions have evidently declined, especially towards the end of the 5 th and throughout the 6th century. Most of the dwellings from this period were made of mud-brick and spolia from deserted public edifices. In Heraclea Lyncestis an entire quarter of humble houses was revealed atop the theater. Similar dwellings appear in some of Stobi’s former palaces and the Early Christian basilica in Scupi270. Thus although life continued on the majority of the larger Hellenistic-Roman centres, the contraction and the drastic decline of living standards in the ancient urban centres are more than evident, especially during the late 5th and throughout the 6th century. But at the same time there is more archaeological information from the countryside, especially when compared to the Early and Middle Imperial phases. The monumental family mausolea and mounds marking the countryside landscape during the period of the Principate completely disappear by the end of the 4th century. They are replaced by two other phenomena that will become typical for the Late Roman Period. These are the fortified hill-tops271 and the Early Christian chapels272. The majority of the Roman sites listed in the archaeological atlas are dated more narrowly to the Late Roman Period or between the early 4th and late 6th century AD. This is largely due to the fact that these types of monumental remains have drawn the attention of some of the earliest field researchers in the country. There is nonetheless a considerable number of Late Roman sites that occupy flat and open locations, not unlike those of early prehistoric sites. Indeed we saw that a large number of open prehistoric settlements were reoccupied during the Roman or Late Roman periods. In the archaeological atlas, these sites are variably called rural settlements or villas. Rarely a rough estimate of their size is given, ranging from several thousand sq meters to 3-4 hectares. The dating remains slightly problematic, because for only a smaller number of sites is there a more substantial body of evidence. In some cases the proposed dating is based on stray coin finds or presumably, on the building materials. It 269 This is another feature of Late Antique towns recurring throughout the Balkan provinces and beyond, A.G. Poulter, The transition to Late antiquity on the Lower Danube: the city, a fort, and the countryside, 51-97, ed. A.G. Poulter, 2007; B. Bavant, Caričin Grad and the changes in the nature of urbanism in the central Balkans in the sixth century, 337-374, the same volume; B. Aleksova, C. Mango, Bargala: a preliminary report, 269-281, Dumbarton Oaks Papers 25, 1971; for Greece, J. Bintliff, 360-363, 2012; for a broader perspective see, L. Lavan ed. Recent research in Late Antique urbanism, Journal of Roman Archaeology supplement 42, 2001. 270 I. Mikulčić, 209, 1981; T. Janakievski, Docnoantička mikrostanbena celina nad teatarot vo Heraclea Lyncestis, Bitola 2001; D. Koračević 157-173, 2002. 271 There is an extensive bibliographic list on the topic of Late Antique fortifications in the Balkan Peninsula. More recent treatise of the subject can be found in J.H.W.G. Liebeschuetz, 107-110, ed. A.G. Poulter, 2007; I. Mikulčić, 2002; V. Lilčić, Late Antique communications and fortification in the Mid-Vardar Region 13-53, Macedonian Heritage 7, 1998; V. Dinčev, Ranovizantiite kreposti v B’lgaria i s’sednite zemi, Razkopki i proučvania XXXV, 2006; M. Milinković, Stadt oder “Stadt”: Frühbyzantinische Siedlungsstukturen im nördlichen Illyricum, 159-192, ed. J. Henning, Post-Roman towns, trade and settlement in Europe and Byzantium vol. 2, Berlin 2007. 272 I. Mikulčić, Frühchristlicher Kirchenbau in der S.R. Makedonien 221-251, Corso di cultura sull’arte ravennate et bizantina XXXIII, Ravenna 1986; B. Aleksova, Loca sanctorum macedoniae, Skopje 1995; I. Mikulčić, 2002; V. Lilčić, vol. II, 2002. 134 is noteworthy that in many entries, it is explicitly stated that the surface clusters consisted of tile, building material and pithoi fragments, sometimes stressing the scarcity of pottery fragments. At least two or three sites from the area of Sopot closely match this description, sites 2, 13a-b and 14. Burial in isolated monumental tombs continues well into the 4th century, but there are also examples of small groups of later, cist burials273. As usual the corresponding settlement sites have not been located, but the appearance of small, isolated groups of tombs in the countryside could indicate that isolated agricultural estates continued to dot the landscape, at least throughout the 4th century AD. All this seemingly leads to the conclusion that the Late Roman was a period of considerable settlement expansion in the countryside, not only along the Vardar Valley, but in most other regions of the country and in the wider region 274. Even if we allow that many of the open sites dated to the Late Roman Period have an earlier, pre-4th century phase, it doesn’t cancel the fact that the great majority of over 500 known fortifications and hundreds of Christian basilicas were either founded or completely built anew during this period. However as discussed in the first half of this chapter, a recent study has rightly doubted the extent of the Late Roman revival, especially in the countryside275. Unlike the plain pottery of the Early and Middle Imperial Periods, Late Roman storage and transport vessels (and one may add tile) are far more recognizable in the surface record. It is thus quite possible that a considerable portion of the finds broadly dated Roman in survey and excavation publications actually date to the Early and Middle Imperial Periods. This implies that the Late Roman recovery could have began by the Middle Imperial phase, if not earlier276. The developments during the Late Roman Period in the survey area are not an isolated phenomenon. There is an evident expansion of settlement, especially in the countryside, in most parts of the country. Examples of Late Roman architectural sculpture begin to appear even in the most isolated, mountainous parts of the river basins. Their appearance in these hostile environments, poor with resources has often puzzled researchers277. Some scholars have speculated about an intensified exploitation of metal ores, related to the increased presence of the Roman army on the Danube during the 4th century. Extensive surveys of the mountainous regions in the Upper Bregalnica, on the high plateaux south of the river Crna and elsewhere have indeed revealed substantial evidence of mining activities, mostly datable between the 4 th and 6th century AD278. But there is an evident expansion of building activity even in areas poor with mineral resources. At the same time one cannot fully accept the claim that there was a general shift in settlement focus, from the lower sections of the major valleys to their upper peripheral stretches, simply incited by the increased insecurity in a period marked by massive and often violent invasions279. The decline and contraction of the former urban centres is undeniable. There 273 G. Dimitrioska, Grupa Docnorimski grobovi od Kamnik, 131-138, Macedoniae Acta Archaeologica 5, 1979; Ð. Petački, A complex of Late Antique vaulted tomb chambers from Veles, 79-88,Macedonian Heritage5, 1997;D. Kalamagdeska-Mihailova, Late AntiqueAREA MACERIA CINCTA from Novo Selo-Zelenikovo23-28, Macedonian Heritage 20, 2002. 274 Cf. J. Bintliff, 2012, chapter 15. 275 D.K. Pettegrew, 743-784, 2007. 276 D.K. Pettegrew, 778, 2007. 277 I. Mikulčić, 193, 1976. 278 I. Mikulčić, 285-290, 1999; A. Keramitčiev, Za staroto rudarstvo vo Mariovo I, 89-104, Macedoniae Acta Archaeologica 2, 1976, A. Keramitčiev. Za staroto rudarstvo vo Mariovo II, 67-81, Macedoniae Acta Archaeologica 4, 1978. 279 I. Mikulčić, 267, 1999; I. Mikulčić. 57, 1982; V. Lilčić, 533, 2002. 135 is also clear evidence for desertion of some of the Early to Mid-Imperial rural estates. However on many sites there is evidence of continued occupation during the Late Roman Period and there are even a greater number of sites that were probably established or at least prospered after 300 AD. What remains uncertain is the true extent of this Late Roman expansion in the rural areas, as well as its underlying cause280. It is clear that there was a certain increase in the number of sites compared to the periods of the Early and Middle Empire, even if allowing that the Late Roman remains are more recognizable in the surface record. It is another question if all of these sites were actually permanently inhabited. Doubtless many of the Late Roman fortification have plenty of surface material, indicating permanent occupation. But an equally large or possibly even a larger number of forts lack substantial surface remains, apart from building rubble and tile281. For example on only two of the eight hill-forts documented along the southern end of the Taor Canyon were there fragments of pottery for domestic purposes. Erosion or ground visibility conditions cannot always account for this fact. A similar situation was encountered on a number of “flat” Late Roman sites in the survey area and we saw in the archaeological atlas that many of the Late Roman sites are described as “clusters of stone rubble, tile and pithoi fragments and very little pottery”. There is little ground to speculate about the character of these sites, but they are evidently different from the typical remains of domestic occupation 282. Apart from small familial necropoleis, sanctuaries and various types of agricultural estates, one has to take into account the possible state-sponsored establishments for the purposes of road security and maintenance, the postal and customs service or the extraction and processing of metal ore283. The towns and the countryside seem to experience divergent developments during this period. It is difficult to ascribe this tendency solely to the constant threat of Barbarian invasions. They are better understood in the context of internal socio-economic and demographic developments, such as the demise of the old city-based landowning elite, the growing importance of the small estate run by soldier-farmers tied to their land284 and the increased presence of the state apparatus. Finally, one cannot exclude the possibility of an actual population growth in these regions, at least during the 4th and early 5th century AD285. As in most other micro-regions, the Late Roman Period left considerable traces in the surface archaeology of the survey area (map III_61). Apart from the three fortifications marking the corners of the lower half of the valley of Sopot and delimiting the settlement’s territory, the hamlet on site 5a reached its maximum, possibly spreading over to the neighbouring field on the west. This was obviously the main settlement centre in the immediate surroundings. Occupation was also renewed on site 8, at Prisoj’s eastern foot, though on a much smaller scale. On the eastern bank of the valley, site 12 is completely abandoned by this time, but the two sites 13a280 J. Banaji, Agrarian Change in Late Antiquity: Gold, Labour and Aristocratic Dominance, Oxford 2001; D.K. Pettegrew, 775-777, 2007, J. Bintliff, 357-358, 2012. 281 J.H.W.G. Liebeschuetz, 106, ed. A.G. Poulter, 2007, points out that many of these fortifications could have been constructed by the local rural populations to serve as refugia, see note 190. 282 Similarly composed assemblages on Late Roman rural sites have been discovered by the regional survey projects in Boeotia and Laconia, J.L. Bintliff, P. Howard, A.M. Snodgrass, fig. 5.1, 2007; W. Cavanagh, C. Mee and P. James, et al. 2005; C. Mee, H. Forbes, 1997. 283 I. Mikulčić, 58-64, 1982. 284 See however M.H. Jameson et al, 400-415, 1994; J. Bintliff, P. Howard, A. Snodgrass, et al. 158-159, 2007; for evidence of large estates during the Late Roman Period in central and southern Greece. 285 A far less optimistic picture is painted in the studies in A.G. Poulter ed. 2007, possibly because the focus is on the eastern Balkan provinces, far more exposed to events on the Danube frontier. 136 13b on the Jakupica Ridge and their satellites 14 and 15 at the western foot of Radičica were most probably newly established or at least continued their existence into the Late Roman Period. There was at last a continued activity in sector I, on the Vardar Valley floor. Thus for the first time in the history of local settlement there are traces of activity in nearly all microtopographic units that constitute the studied landscape. Every potential arable piece of land was occupied by an establishment of an according rank. In fact small amounts of material datable to the Late Roman Period were documented on over two thirds of the surveyed territory. But regardless of these positive signs there are a good number of reasons to be skeptical about the actual growth in terms of population and settlement rank. Most important of all, and this is something for which we completely lack comparative data, is the small amount of Late Roman remains on the surface, especially when compared to the several times larger concentrations of prehistoric finds. Although compounded by the presence of brick and tile, the Late Roman assemblages never fully obscured the superimposed remains of earlier epochs and in some cases, the opposite happened. In fact two of the three forts featured an artifact density only slightly higher than in the off-site zone. The isolated clusters of finds in sector I consist almost exclusively of tile and very similar assemblages were discovered in sector X, on the Jakupica Ridge. Only on two sites did the collected finds form a full domestic assemblage, including architectural ceramics, table ware, cooking and storage vessels. These are sites 5a-5b and 8, whose combined territory is slightly over 1 hectare. This is an evident increase from the Hellenistic and probably the Early Roman Period, but from a long-term perspective, it is essentially a return to the settlement rank of earlier, prehistoric periods. We need to consider one final issue related to the expansion of settlement during the Late Roman Period, both in the survey area and in the broader region. Apart from the increased number of sites of various types and sizes, the careful recording of individual surface artifacts revealed an ultra-thin carpet of (predominantly?) Late Roman material spreading across the entire landscape, covering even abandoned areas. Unlike the ubiquitous off-site carpets discovered in certain parts of Greece and the Near East, this manifestation is invisible prior to the highly intensive survey, collection and study of the surface artifacts. One is dealing with densities on the order of 0.1-0.5 shards per 100 sq meters or between 1 and 5 fragments, scattered across an area of 1000 sq meters. In the first survey area it was possible to observe a general pattern in the composition of this material. Zones closer to the settlement sites featured not only slightly higher densities, but also a roughly equal percentage of tile and pottery, while in the more peripheral parts of the landscape (excluding sectors where the presence of sites of intermediary density is suspected), the thin scatter of Roman to Late Roman finds consists almost exclusively of worn fragments of brick and tile. Evidently this is not a homogenous offsite carpet: the thin scatters of tile in the peripheral survey sectors could have hardly originated from the same source as the material in the western survey half. The latter is clearly identical to the material collected from the settlements on sites 5a-5b and 8. Considering its extreme sparseness and wide dispersal area chiefly overlapping with the most fertile stretches of land in the basin, this ultra-thin layer of Roman to Late Roman finds was interpreted as the inorganic remains of ancient manure. On the other hand, the equally thin scatters of brick and tile and the occasional pithos fragments in the more distant survey sectors were seen as the remains of minor farms, field-sheds or non-residential activities. At this point however we run into a problem. If the increase in population during the Roman-Late Roman Period was slight in comparison to other periods of settlement in the survey area, it becomes difficult to explain the need to intensify agricultural exploitation. And yet this is 137 what the existence of both the extensive shards scatter and the peripheral clusters of architectural ceramics suggest. One possible explanation is that the exact size of the Roman to Late Roman settlement is actually underestimated and that we should interpret all major clusters of Roman to Late Roman material as contemporary and fully residential sites, regardless of the composition of the ceramic assemblages. This will increase the total occupied area to slightly over 2 hectares, more than twice the occupied area during the Late Byzantine-Early Ottoman and certain prehistoric periods. But accepting this as a likely scenario doesn’t solve the problem of the relatively low density of Roman to Late Roman material on site locations. This cannot be explained by simply pointing to the peculiarities of the local taphonomic or ground visibility conditions, because we had examples of multi-period sites where artifact densities for earlier periods were much higher than for the Roman to Late Roman assemblages, although they were collected from the same location. One could think of the possibility that this episode of growth was short-lived, but this is again difficult to reconcile with the fairly extensive off-site carpet and the considerable variety of fabric groups. We are therefore inclined to believe that the evidence of intensified agricultural exploitation doesn’t reflect solely demographic tendencies, but also a specific socio-economic environment and agrarian relations. The phenomenon of the busy Late Roman countryside is far from being unique to our study region.Allowing for certain degree of regional variability, it can be observed throughout the countries of the Eastern Mediterranean 286. Apart from the problematic increase in population, scholars in the field have pointed to the impact of the new capital of Constantinople and the booming of the other urban centres in the Eastern Empire, the proximity of the army and the stationing of garrisons in previously demilitarized provinces, the fiscal policy and taxation in the Late Empire and so forth 287. The incentive for increased investments in agriculture doesn’t necessarily need to be related to demographic pressures. Because of the low chronological resolution, it is impossible to pin-point the exact date when these Late Roman settlements were deserted. Most of the finds could only be roughly dated between the 4th and late 6th century AD. It is generally accepted that the great majority of the Late Roman settlements were abandoned by the last quarter of the 6th century. This claim is supported both by evidence from historical sources and from archaeological research on the old urban settlements, but also on a number of larger, well-defended fortifications288. It is possible that some of the low-land sites were abandoned earlier, but there lacks decisive evidence289. By the end of the 6th century AD nearly all known settlements were either completely abandoned or continued to exist on a much humbler scale, leaving traces that we still can’t recognize in the archaeological record. On the entire territory of the Republic of Macedonia, there are only a few incidentally discovered sites datable prior to the late 9th century290. In many cases these are isolated finds, discovered on settlements from the Late Roman Period, such as Scupi and Stobi, but there were also newly established settlements, unrelated to those of the 286 J. L. Bintliff, The contribution of regional survey to the Late Antiquity debate: Greece in its Mediterranean context, 649-678, ed. A.G. Poulter, 2007; D.K. Pettegrew, 760-765, 2007. 287 C.R. Whittaker, Late Roman trade and traders, 163-180, eds. P. Garnsey, K. Hopkins, C.R. Whittaker, Trade in the Ancient Economy, Berkeley-Los Angeles 1983; J. Banaji, 60-65, 2001; C. Kosso, The archaeology of public policy in Late Roman Greece, Oxford 2003; J. Bintliff, 357-358, 2012. 288 C. Snively, Golemo Gradište at Konjuh: Reports on the excavations in 2000 293-302, Dumbarton Oaks Papers 56, 2002. 289 For example, excavations in Scupi revealed a continuous presence on small parts of the town well into the Middle Age, D. Koračević 157-173, 2002. 290 D. Koračević, 157-173, 2002; B. Risteski 49-53, 2004; Z. Beldedovski 45-49, 1990. 138 previous period. In all of these examples there is very little information about their actual size and extent. The scarcity of archaeological remains from this period is matched by the silence of contemporary historical sources. The chroniclers of the time mention the names of at least a dozen different Slavic tribes, but they are all situated on the territory of modern Greece. The northernmost tribe, according to these sources, occupied the lower Strymon Valley291. However it must be noted that in the past couple of decades the increased attention given to plain and coarse fabrics has resulted in the discovery of post-Roman phases on a number of Late Roman sites in the Aegean. This material is often accompanied by fine ware traditionally dated not later than the early 7th century, which makes it very likely that the post-Antique phase was overlooked on at least some of the excavated Late Roman centres292. The earliest mentions of the Vardar Valley and the surrounding regions in the historical records date to the second half of the 9th century, this also being the period to which the first more substantial archaeological remains are dated293. They consist almost exclusively of churches and chapels, often built over Early Christian basilicas and accompanied by cemeteries.Scattered evidence for the period between the 9th and 12th century also comes from nearly all known Medieval fortresses, although the earliest architectural remains begin to appear only towards the end of the 10th and the beginning of the 11th century294. On most of these sites very little is preserved of this period on the surface, due to the fact that many were used well into the Ottoman Period. On the other hand the strong Late Roman walls were often only repaired and slightly modified, leaving very little evidence of the post-Antique phase in the architecture295. This circumstance further diminishes the visibility of the Early and MidByzantine periods in the surface archaeological record. The (re)building of the first churches and fortresses roughly coincides with the emergence of the first Medieval states in the region and the Byzantine re-conquest of the Balkans. The names of many of these towns are recorded in the written sources and the majority will become the core of the later Ottoman and Early Modern towns. Being often occupied for nearly a millennium, very little is known about the inner organization and the social rank of the towns of the Mid-Byzantine Period. Thanks to historical sources, it is possible to conclude that these early towns were primarily administrative and military centres. Civic quarters or “lower towns” developed separately296. It has to be stressed however that most of the evidence comes V. Popović, Aux origins de la slavisation des Balkans 230-257, Comptes rendu des sèances de Acadèmie des inscriptiones et belles lettres 1, 1980. I. Mikulčić 27-28, 1996; though this approach of supporting historical evidence of invasions with coin hoards has been criticized, M.D. Metcalf, Avar and Slav invasions into the Balkan Peninsula (c. 575-625): The nature of numismatic evidence, 140-148, Journal of Roman Archaeology 4, 1991; for the historical sources, F. Curta, The Making of the Slavs: History and Archaeology of the Lower Danube Region (c. 500-700), Cambridge 2001. 292 A.K. Vionis, J. Poblome, M. Waelkens, 147-165, 2009; A.K. Vionis, The Medieval and post-Medieval pottery and Tanagra village history, 570-578, J.L. Bintliff, et al. The Tanagra Project: Investigations at an ancient Boeotian city and in its countryside, B.C.H. 128-129, 2004-2005. 293 S. Antoljak, Srednovekovna Makedonija, vol. I, Skopje 1985; B. Aleksova 20-22, 1989; B. Babić, Materijalnata kultura na makedonskite Sloveni, vo svetlina na arheoloshkite iskopuvanja vo Prilep, Prilep 1986. 294 E. Maneva 186-208, 1992; I. Mikulčić Srednovekovni Gradovi vo Makedonija, 135-356 (catalogue of forts), Skopje 1996. In recent years, systematic excavations were initiated on the citadels of most major towns in the country. According to the preliminary reports, the findings of earlier researchers have been mostly confirmed, www.mav.mk. 295 I. Mikulčić, 68-69, 1996. 296 A few researchers have dealt with the problem of the genesis of Medieval urban centres in the southern Balkans; particularly for the region of Macedonia, S. Antoljak 524, 1985; I. Mikulčić, 39-41, 1996; A. Dunn, The 291 139 from the 14th century, from towns that have preserved the main lines of their Medieval topography and from historical sources. Not surprisingly even less archaeological evidence has come from the countryside. This is not only the case with the Mid-Byzantine Period, but also with the subsequent Late Byzantine and Early Ottoman periods. The data presented in the archaeological atlas are to a great extent useless, because with rare exceptions, the sites are very broadly determined as Medieval. Moreover the great majority of them consist of churches and adjoining cemeteries. For example, of about 20 identified sites for the region of Veles, only two are identified as settlements. However one wonders if these sites are truly Medieval or perhaps Early or even Late Ottoman. The general impression is that the term “Medieval” has been applied to remains datable anywhere between the end of Antiquity and the Late Ottoman Period. It is thus rather difficult to put the findings from the Sopot survey into a broader, regional context (map III_62). The nearest sites datable to the Middle and the Late Byzantine Period are two fortifications, situated 7 and 11 kilometers to the north and south of Sopot. Both forts were built by the late 12th century and controlled sections of the Vardar Valley road 297. Their appearance in this rugged, inhospitable terrain above the narrow canyons of the Vardar, perhaps reflects the continued importance of the natural corridor along the river. It should be stressed that they both have Late Roman phases. The two forts occupy very steep hillocks, offering little room for a “lower town”. These were castles, enclosing areas not larger than a couple of hectares and were most probably reserved for a local landlord and a small military force 298. There is very little arable land in the surroundings; the nearest fertile stretches lie at distances of over 1 kilometer from the forts. This is where presently and in the recent past, the main settlements were situated. In the case of the fort to the north of Sopot, this is a small village, situated 1200 meters to the west, on the right bank of the Pčinja. The fort to the south of Sopot, Medieval Veles, gained much greater importance during the Ottoman Period, developing into a larger town. The castle and the possible “lower town” in the narrow canyon at the foot of the fort were gradually abandoned and the new civic centre moved an entire kilometer and a half to the north, giving the name to the entire basin (maps III_63 and 64). In both cases a ring of chapels delimits the narrower territory of the castles. Compared to the Late Roman Period, the fortifications network of the Middle Age is obviously much sparser, but one has to take into account the fact that these were not the sole centres of the landowning class. There is ample historical evidence that by the late 12th century, the greater portion of the land belonged to the numerous monastic centres299. In rare cases these too can develop into fortified complexes300, but most commonly they leave very humble traces in the surface archaeological record. Constant renovations in later centuries further contribute to the low state of preservation of surface remains from earlier phases. Of the four monastic complexes transition from polis to kastron in the Balkans (III-VII c): General and regional perspectives, 60-80, Byzantine and modern Greek studies 18, 1994; for Greece, see J. Bintliff, 394-395, 409-411, 2012. 297 I. Mikulčić, 133-135, 1982; I. Mikulčić. 340-344, 1996; www.mav.mk. 298 I. Mikulčić, 133-134, 1982, the author interprets the fort to the north of Sopot as a military castle, while the one to the south, Veles as a town. Although there were probably differences in status between the two, they occupy absolutely identical positions, at an equal distance from agricultural land. 299 V. Mošin et al. Spomenici za srednovekovnata i ponovata istorija na Makedonija, Skopje 1975. The castle of “Markovi Kuli” to the north of Sopot and the monastic church of St. Nicholas are particularly illustrating examples. According to a mid-14th century edict, the entire land surrounding the castle actually belonged to the nearby monastery of St. Nicholas, I. Mikulčić, 134, 1982. 300 V. Lilčić, The canyon of Matka, 47-59, Macedonian Heritage 3, 1997. 140 situated along the southern portion of the Taor Canyon, only the one near the castle “Markovi Kuli” was certainly established by the 14th century. The rest are relatively recent foundations, like St. George’s monastery near Sopot, though it is possible that some of them have earlier foundations. The network of Medieval castles, towns and monasteries must have been complemented by a much larger number of small, rural settlements, which for the greater part have been ignored as a research subject, both by archaeologists and Medievalists. From an archaeological point of view, the greater part of the problem lies in the humble character of the material remains; not only the near absence of larger architectural monuments, but also the relatively low level of distinctiveness of locally produced, plain pottery. Unlike the pottery assemblages from the urban settlements, regularly comprising a certain amount of very distinctive glazed pottery, it seems that this category is very scarce in the countryside, at least until the end of the Byzantine or even the Early Ottoman Period301. A caution is necessary however, because the published studies deal with material that comes almost exclusively from high-ranking settlements or monasteries. Another possible circumstance contributing to the low visibility of Medieval rural settlements in general is the size and degree of dispersal.We saw that some of the periods that are underrepresented in the archaeological atlas (the Bronze Age, the Hellenistic Period) were indeed characterized by very small and dispersed scatters. Thus the fact that there are no finds securely dated prior to the 14th century in the area of Sopot could result from our inability to recognize this material, either during the course of fieldwork or during the processing of the finds. But even in case the survey area was inhabited in the centuries between the end of Antiquity and the 14th century, it is nearly certain that this long period was marked by a decline and contraction. The possible settlement was either very small or short-lived or it is situated on a presently inaccessible location. Recall that even the tiny assemblages dating to the Bronze Age and the Hellenistic Period left a recognizable pattern of distribution. This implies that the potential Medieval settlement is either small enough to remain hidden amidst the ceramic material from other periods or it is not represented in the surface record. Despite the negative effects of the Black Death, the constant wars and political instability and the imminent threat of Ottoman invasion, the 14th century was a period of intensified building activity and settlement expansion302. Many of the earlier towns received new fortification walls, incorporating novel architectural conceptions, imported from the Eastern Adriatic and Italy. These are the citadels of the modern towns of Skopje, Ohrid, Prilep and many others. Little was changed in their main architectural conception during the Ottoman Period, as they began to loose their military importance soon after the arrival of the Ottomans. Many new churches were built or completely renewed during the 14th century, even in its later decades, on the very eve of the Ottoman conquest303. More importantly these were not necessarily town churches. There were a number of newly built or fully reconstructed monastic churches in the countryside, lavishly dedicated with people and property. Numismatic finds from this century also indicate increased prosperity and intensified trade relations, especially with the Eastern 301 B. Ristevski, 49-71, 2004; even on the Skopje citadel and the surrounding castles, glazed pottery is rare prior to the late 12th century and appears alongside the adoption of certain shapes of dining and drinking vessels. Cf. G.D.R. Sunders, Recent developments in the chronology of Byzantine Corinth, 385-399, eds. C.K. Williams II, N. Bookidis, Corinth: the Centenary, 1896-1996, Athens 2003. 302 I. Mikulčić, 1996; in general, this is a period of decline in most regions of Greece, J. Bintliff, 395-398, 2012 303 A. Serafimova, The Middle Ages 93-174, Macedonia; cultural heritage, Skopje 1995. 141 Adriatic304. Other categories of archaeological finds, such as jewelry and certain types of table ware point to a developed local production305. Finally, and although pertaining to the northern parts of the Skopje Basin, there is a direct historic evidence of re-colonization in the early decades of the 14th century, under the Serbian King Milutin306. For some decades the survey area lied at the border between the receding Byzantine Empire and the expanding Serbian Kingdom. But by the second quarter of the 14th century, it certainly became a part of the Medieval Kingdom of Serbia and it shouldn’t be excluded that similar actions were undertaken by Milutin’s successors in the regions along the Middle Vardar Valley. These rather general positive developments at least provide a vague context for the establishment of Late Medieval Sopot. As in many earlier periods, it seems that developments in the survey area resonate with broader, regional tendencies. Understandably until more comparative data are available, it is impossible to relate the local circumstances in one small micro-region to the general socio-historical climate. It is nevertheless certain that by the early 15th century, settlement in the survey area rose back to the rank of a larger hamlet or a small village. Like most of its predecessors, it spread over an area of about 1 hectare and its related material was largely confined to a single topographic unit (map III_65). In the case of Late Byzantine and Early Ottoman Sopot, this was the southwest corner of Prisoj’s southern foothills. Along with the majority of the villages in the neighbouring micro-regions, Sopot continued its existence and most probably prospered during the first few centuries of Ottoman rule. The published exhaustive censuses from the 15th century mention nearly all of the presently existing villages in the region surrounding the southern end of the Taor Gorge307. Sopot and many other villages were actually more populous than in the early decades of the 20th century. This fairly dense network of settlements must have been for the greater part established prior to the Ottoman conquest, for not too many new establishments are mentioned in the earliest censuses. With a few exceptions these were communities of peasant-serfs, usually featuring between 15 and 30 households. In the Early Ottoman Period, the village territories along with their entire communities were given to a feudal lord in exchange for military service, while in later centuries, they were gradually turned into hereditary properties308. There is no contemporary parallel for the actual appearance of Late Byzantine and Early Ottoman Sopot. In later times villages of the çiftlik type usually consisted of two separate quarters; one for the peasant serfs, the other for the landlords or the estate manager 309. At the same time this was also a religious segregation. The villages in the rugged landscapes surrounding the southern exit of the Taor Canyon were almost 100% Christian during the Early Ottoman Period. It is very unlikely, especially in the first couple of centuries of the Ottoman reign that the constantly warring feudal lords ever stayed in these villages. Local affairs were probably left to the village elders and the landlord visited the villages only to collect his levy310. It seems that only in later centuries when the feudal land became hereditary did it become more common for the petty Ottoman landlords to build their houses near the Christian villages. Some 304 J. Kondijanov ed. Numismatic collection of the National Bank of the Republic of Macedonia, Skopje 1999. 305 E. Maneva, Srednovekoven Nakit vo Makedonija, Skopje1992; B. Risteski 58-71, 2004. T. Tomoski, Skopskata oblast of XI do XIV vek, 57-75, ed. V. Mošin, 1975. 307 M. Sokolovski, A. Stojanovski, 1973; M. Sokolovski, 1971. 308 A. Stojanovski, Makedonija vo turskoto srednovekovie (XIV-XVIII vek), Skopje 1989; cf. J. Bintliff, 438-443, 467-469, 2012. 309 M. Filipović, 511, 1935; J.Bintliff, 467-469, 2012. 310 A. Stojanovski, 66-67, 1989. 306 142 of these estates have survived until the beginning of the 20th century. Sopot too was a village of the çiftlik type and according to local oral tradition the two taller, tower-like structures in the modern village were actually the houses of the Ottoman landlords311. It is impossible to know if these or similar structures were actually contemporary with the Late Byzantine and Early Ottoman village, situated between 150 and 250 meters to the west-southwest. Nevertheless the positioning of the living quarters of the landlord on slightly higher ground and at a certain distance from the houses of the peasant serfs fits the descriptions of the çiftlik villages from later centuries312. The great majority of the villages mentioned in the Early Ottoman censuses survived until the middle of the last century. However it’s been generally accepted, especially among early 20th century ethnographers, that a number of villages changed their locations sometime during the late 18th or the early 19th century313. This latest major shift in the settlement pattern is usually related to the serious decline of the state authority, especially felt by the communities in the more peripheral regions and along the major roads314. It has to be stressed however that this process is espoused mostly on the bases of local oral tradition recorded by ethnographers in the early 20th century. The disappearance, relocation and breaking up of communities can also be related to the expansion of the çiftlik, the inherited estate at the expense of land that belonged to the fairly autonomous villages or land leased by the state in return for military service315. In the case of Sopot we observed only a minimal displacement, which couldn’t be dated precisely. At an unknown point of time, the local community merely moved for about 150-250 meters to a more open and spacious location, occupying the living quarters of their former masters. But in principle the settlement remained within the limits of the same micro-topographic unit. A logical, albeit highly hypothetical explanation for this shift in settlement focus is the final demise of the Ottoman landlords and the appropriation of the land by the peasantry. However this would date the displacement not later than the mid-19th century, merely a few generations before the ethnographer Milenko Filipović visited the village. It is improbable that such an event passed unrecorded and it is even less likely that it was left out of the local spoken narratives. Either the abandonment of the old village took place much earlier and was forgotten by the inhabitants or more likely, it was a long-term, gradual process. Indeed the surface architectural evidence seems to indicate a gradual eastward movement of the village houses. A more precise dating of the ceramics will certainly help in resolving this issue. The causes behind the relocation remain unknown. Apart from social transformations, this could also reflect changed standards of living and new agricultural practices. Finally one shouldn’t exclude the possibility of a local population growth in the last century of the Ottoman Period. In this context it is unfortunate that we lack specific textual evidence for the period of the 18th and the 19th century. Some researchers have pointed to the possibility that the introduction of New World cultures and the slow development of a capitalist, market economy could have brought increased prosperity to the countryside316. Indeed regional studies in southern and central Greece have revealed a period of increased demographic growth, especially during the M.S. Filipović, 535, 1935. Cf. map III_22 and J. Bintliff, fig. 21.5, 2012. 313 M. Filipović, 509, 1935; a similar process has been observed in eastern Boeotia, but here it is confirmed through the study of later Ottoman censuses and it is dated much earlier, to the mid-17th century; J.L. Bintliff, 193224, eds. A. Hurst, A Schachter, 1996. 314 A. Stojanovski, 1974. 315 J. L. Bintliff, 207-208, eds. A. Hurst, A Schachter, 1996. 316 M.H. Jameson et al, 280, 1994. 311 312 143 second half of the 19th century317. Although the general social and political conditions in the regions that remained under Ottoman rule were different, one shouldn’t exclude the possibility that this was a major regional tendency. In fact the surface archaeological evidence provides support for this thesis. The last couple of centuries were a period of intensified agricultural activity and a greater influx of goods produced in more distant towns. This cannot be solely related to the temporal proximity of the Late Ottoman-Early Modern village. The relatively dense and highly obtrusive carpet of ceramic fragments, including tile, a wide range of table, storage and transport vessels and spread across the entire basin of Sopot, must signal an increased wealth and investment in agricultural production. The distribution pattern of this material brought to the fields alongside organic manure bears a striking resemblance to the distribution of the Roman to Late Roman off-site carpet. In both examples the settlements were located on the western fringes of the colluvium at Prisoj’s foot and the core of the productive land and the bulk of the off-site finds were located in the western survey half, only an ultra-thin carpet spreading east of the central valley (map III_66). This distribution pattern is not simply determined by the greater accessibility of the fields west of the central valley; it also marks the inner zone of intensive cultivation and gardening and the outer zone, characterized by a more extensive regime of exploitation318. It is uncertain if this agricultural expansion is related to a population increase or to the slow commercialization of production and increased living standards. In most of the published Early and Mid-Ottoman censuses, Sopot has an equal or a greater number of families than in the 1930’s. But because we lack population data for the 18th and the 19th century and a finer chronology for the surface finds, one shouldn’t exclude the possibility that this gap conceals an episode of contraction and decline. The final episode in the settlement history of the first survey area, the rapid decline and near abandonment of the village in the second half of the last century is not an isolated, local event. It is a process that can be followed in most regions of the country and one can see it as a part of a much wider migratory movements from villages to towns, characteristic for many parts of the Mediterranean in the late 19th and 20th century319. Specifically in the Republic of Macedonia, as in most other former socialist countries on the Balkan Peninsula, it can be related to a number of aspects introduced through the radical social and economic reforms in the mid20th century. One possible incentive to leave the rural sector was the abolishment of private property and the introduction of collective farming, though the issue can hardly be addressed without serious sociological studies. At the same time, towns and cities were being modernized at a much faster rate than the countryside and offered secure and more attractive job opportunities in the administrative and industrial sectors, better education, social and health care. By the last quarter of the 20th century, the depopulation of the micro-regions along the southern end of the Taor Canyon was nearly completed. Most of the local inhabitants moved to the larger towns or left the country. Living conditions in this particular region are further aggravated by the bad communication and the lack of even the most basic supplies. At the time of the survey the smaller villages on the Vardar’s left bank, including Sopot and the 317 H. Forbes, Turkish and modern Methana, 101-117, eds. C. Mee, H. Forbes, 1997; B. Forsen, A. Karivieri, The Roman-Early Modern Periods: conclusions, 307-331, eds. J. Forsen, B. Forsen, et al. The Asea Valley Survey: An Arcadian Mountain Valley from the Paleolithic Period until Modern Times, Stockholm 2003; M. Wagstaff, The formation of the modern landscape of the survey area, 403-420, W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al. 2002; J. Bitnliff, 478-497, 2012. 318 Cf. J. Bintliff, P. Howard, A. Snodgrass, eds. 23-24, 2007. 319 P. Horden, N. Purcell, The CorruptingSea: A study of Mediterranean history, Oxford 2000. 144 neighbouring Vetersko and Novačani had but a few older inhabitants. There are several active households in the larger village Solb on the opposite bank of the Vardar, but these represent less than 10% of the village population in the first half of the 20th century320. In fact it is possible that the last couple of decades witness the emergence of a new pattern of settlement. People begin to return to the countryside, though not on the sites of the abandoned villages. New houses can be seen in most of the emptied settlement niches, usually standing isolated and unrelated to the old village. During fieldwork we managed to establish contact with the newcomers, some of which informed us that they stay on a seasonal basis and for the purposes of cultivation of cash-crops (vines, walnuts) or bee-keeping. Likewise the owner of the sheep-farm located in the northwest corner of the old village territory is not a full-time resident of the village and the herds are chiefly looked after by shepherds, often hired from more distant regions. One could say that this is just another episode in the millennia-long history of settlement in the survey area, characterized by cycles of growth and contraction, nucleation and dispersal. III.2.2 The dynamic of settlement in the survey area, the long-term trends and cycles, the carrying capacity One characteristic feature of the local settlement history in the hinterland of Sopot was the constant dislocation of the central settlement321. This dynamic was analyzed against the background of the topographic, geologic and pedologic divisions in the survey area and the distribution of freshwater sources. We also attempted to take into account the possible effect of local and regional communication lines and the proximity of political and ethnic frontiers. The shifting of the settlements’ locations between the eastern and western half of the valley was related to the global changes in the technological, economic and political environment. During the Neolithic the central settlement was located on the eastern bank of the valley, showing an obvious preference for the light, lacustrine soils that cover the eastern survey sectors. The introduction of animal traction and copper tools allowed the local communities to settle the more fertile western bank, while the re-occupation of the eastern bank during the Hellenistic Period was possibly caused by the insecure political conditions. But it proved trickier to offer a reasonable interpretation for the settlement displacement within the area of the western bank. The location of site 5a seems to have been particularly favored. It was occupied during at least three periods in the past. This isn’t a noticeable point in the landscape; it is merely a corner of Prisoj’s southern foothills. The settlements that occupied this location were practically clinging on to the edge of the arable territory. Looking at the locations of other settlement sites, such as site 6 or 7, it becomes clear that in principle, they differ little from site 5a. Site 7 occupies the opposite, northeast corner of Prisoj’s southern foot, while site 6 was built in the very centre of this micro-topographic unit. One may view the location of site 8 in the same context. Because of its size and better physical articulation, it admittedly differs from the sites at Prisoj’s southern foot. It too however gravitates around the southern foothills, although it occupies a different micro-topographic unit. Here one has to distinguish between the Late Iron Age settlement and the small Late Roman farm, occupying a small part of the same site. In the M. Filipović, 544-546, 1935. Cf. H. Lehmann, Die Siedlungsraeume Ostkretas, 212-238, Geographische Zeitschrift 45, 1939, after M. Gkiasta, The Historiography of Landscape Research on Crete, 72-75, Leiden 2008. 320 321 145 former case, the entire eastern foothills are turned into a settlement, while the main focus of agricultural production was at the southern foot and also possibly on the valley floor. The Late Roman farm on the other hand, features a location analogous to that of site 6. It occupies the centre of the neighbouring, smaller micro-topographic unit, exploiting the narrow strip of fertile land at Prisoj’s eastern foot, in effect the area occupied by the dwellings of the Late Iron Age settlement. Evidently all settlement sites on the western bank, with the exception of the Late Roman farm on site 8, gravitate around Prisoj’s southern foothills. The advantage of site 5a perhaps lies in the presence of a freshwater spring in its immediate vicinity, although sites 8 and 7 also had a quick access to a source on the valley floor, 150-200 meters away. But the advantages of site 5a become apparent, once we look at the settlements relations to local and regional lines of communication that pass through the survey area. Understandably all settlements were aligned by local roads; the main difference being that some of these were sections of larger regional roads, while others were merely local lateral roads that joined the village with the fields in the peripheral parts of the micro-region. In this respect site 5a stands apart from the rest of the sites, because it is located at the very exit of the micro-region, by the road that leads northwards, passing over the top of the rocky, eastern side of the Taor Canyon. This is difficult and treacherous terrain, cut by numerous ravines with vertical sides, but it is the shortest route connecting the plain of Veles and the southern periphery of the Skopje Basin. An easier, but longer road passes over three kilometers to the east, avoiding the narrow canyons and the survey area. But even without a state-sponsored interregional road, the location of site 5a is strategically important. It overlooks the western gate of the surveyed micro-region. Whether going northwards along the top of the canyon or westwards, into the Vardar Valley floor, one has to pass by site 5a. Site 7 and especially site 8 are much more sheltered in this respect. They are aligned on a local road that leads along the central valley, towards its upper course and ultimately to Vetersko. In a sense sites 7 and 8 are situated in a recess, keeping a certain distance from the main east-west axis of this micro-region (map III_67). Following this line of reasoning it is possible to group nearly all discovered sites into two categories, depending on their proximity to the main transversal in the survey area. On the one hand, sites 5a-5b, 6, 11, 13a-13b are all situated in the immediate vicinity of this line; on the other, sites 7, 8, 12 and 14 were either deliberately avoiding this alignment or were simply indifferent to it. The site of the Late Byzantine to Early Ottoman village (site 4) is to a certain degree ambiguous. It too is aligned along the east-west transversal, but unlike the first group, it occupies lower ground and a more sheltered location. More specifically it stands by a local road, leading eastwards from the floor of the Vardar and passing at Gaber’s northern foot. This route must have had some kind of wider regional importance, because it connected the villages on the Vardar’s right bank with the Veles Basin. The remains of the small tower on the Vardar’s left bank testify that this was one of the rare points in the Taor Canyon where it was possible to cross the river. In any case this wasn’t a major interregional transversal, but a regional road, offering a direct link between the Vardar’s right bank and the Veles Basin. In this context it should be stressed that the location of the Late Ottoman-Early Modern site is obviously more exposed and belongs to the first group of sites. The three forts that delimit the survey area also differ regarding their relation to the eastwest axis. Apparently the fort over the monastic complex and the newly discovered one on site 10 stand at the extremes of the main axis. Their very presence could be explained by the significance of this transversal. The small fort on site 9 on the other hand, guards the local northsouth road, leading along the central valley. Hidden in the small valley between the two massifs 146 of Prisoj and Radičica, it could effectively act as a refugium for the community that inhabited this micro-region. To sum up, there is one group of sites that forms a chain almost perfectly aligned by the main east-west axis, occupying very open and exposed locations and another group, almost randomly dispersed to the north and south of the main axis and usually occupying sheltered and topographically more articulate locations. The small Bronze Age farm discovered in the yard of the monastic complex definitely belongs to the latter group. Up until the construction of the first Skopje-Thessalonica railway in the 1870’s, this section of the Vardar Valley could hardly have been used for travelling longer distances along the river banks. Because long stretches of the Taor Canyon are impassible, the narrow river terraces are in effect isolated islands of flat, cultivable land, communicating only with their hinterlands and in certain cases, with the opposite bank. The different positioning of the sites in relation to natural communication lines puts the dynamic of settlement displacement in an interesting perspective. While in certain periods of the past proximity to the main, east-west route was apparently preferred, in others it was either avoided or was perceived as irrelevant. During the Roman and especially the Late Roman Period, when there is an obvious diversification of sites regarding function, both types of locations were occupied. Given that site 9 was truly acting as a refugium at this time, the main settlements could readily be moved away from their exposed locations on the east-west axis; the two alternate patterns of settlement were virtually coexistent. Thus along with the distribution of the basic natural resources and population size, major, interregional lines of communication were the third determining factor in the local settlement history. Proximity to the main east-west transversal would firstly indicate that it was possibly functioning as a section of a major interregional road and secondly, that living conditions in the wider region were relatively stable. Avoidance or disconnection from this transversal would either indicate that it lost its significance, other factors becoming more pressing or that living conditions were becoming more precarious. The fact that the local settlement shifted close to and away from the main axis, almost alternately from period to period, probably reflects the changing importance of the Vardar Valley road through the past (table 1). Understandably the maintenance of an active transversal along this section of the Vardar would by itself require stable social conditions or a stronger central authority, though its inactivity wouldn’t necessarily imply the opposite. 147 Table III_1: Location of the central settlements in relation to the main east-west axis Period Mid-Neolithic Exposed/Sheltered Exposed Late Neolithic Bronze Age L. Bronze Age LBA-EIA E Iron Age L Iron Age Hellenistic R-L. Roman LByz-EOtt LO-EM Exposed Sheltered Sheltered Sheltered? Exposed Sheltered Sheltered Exposed Sheltered Exposed But a cautionary remark is necessary at this point, because the existence of major interregional roads doesn’t have to be to solely reflected in the positioning of the settlement sites in relation to natural corridors. Settlement sites are but one component of the human landscape and while settlement location can indicate retreat from the main natural corridor, further implying precarious living conditions or decline of the corridor’s importance in the interregional road network, the locations of other components of the landscape such as military posts, sanctuaries or cemeteries can reflect a very different attitude. Apart from the Roman-Late Roman period when the two patterns existed side by side, the Late Iron Age remains present another instructive example. By its location site 8, the main settlement of this period, clearly belongs to the group of settlements situated away from the main east-west axis. This would imply that security was an important concern for the local community or that the road leading through the Vardar Valley lost its importance during this period. However the positioning of the mound necropolis on the top of the Radičica Ridge, the eastern limit of the basin, clearly demonstrates that this was hardly the case. The string of mounds reaches to the eastern entrance into the Sopot Basin and it would have been visible to anyone leaving or entering the region. Thus while not fully undermining the usefulness of this type of analysis, this example calls for a careful weighing of the existing evidence prior to jumping into premature conclusions. Unlike in the case of the Roman-Late Roman and the Late Iron Age settlements, non-residential sites are not always visible in the surface archaeological record. Moreover it is more likely than not that there will be cases of ambiguously located settlement sites, which can neither be classed as sheltered, nor as exposed. Site 7 nicely illustrates this case: although situated at a certain distance from the main transversal and classed into the category of “sheltered” sites, it has a fairly imposing location while some of its satellite clusters were aligned along the same east-west axis that joined sites 5a-5b, 6, 11 and 13a-13b. Clearly if we are to study the impact of interregional communications, political and administrative frontiers on local settlement, it will be essential to secure independent archaeological evidence for the existence of such supra-regional phenomena. The evidence presently available is simply too scant to establish the exact route of the Via Axia or the provincial frontier between Macedonia and Moesia. They are not described in the historical records and we are ignorant about their associated manifestations in the archaeological record. Until more substantial data is secured, we can only analyze the locational preferences of 148 settlements in relation to local and regional roads and the geometry of the surrounding landscape. In this sense the analysis elaborated in the preceding paragraphs is essentially valid, although the constant shifting of the central settlement on and off the main east-west axis doesn’t necessarily reflect the activity of a major interregional road or general social and economic conditions322. Another one of the more constant features of the local history of habitation was the size of the settlement (graph III_1). To be sure, determining the size and the character of the sites proved to be a rather elusive affair. Whether because of inconsistencies during fieldwork, potential faults in the processing of the finds or because of technical problems, it wasn’t always possible to draw the exact limits of the sites. In these cases the reading of the survey results was inevitably ambiguous and we had to offer alternative interpretations. Closely related to the determining of the size of the local settlements on the basis of the surface record is their degree of dispersal, but also the sharpness of the chronological resolution. As we saw, during most periods in the past the remains of local settlement consisted of one central cluster and a number of smaller satellites. Only the Middle Neolithic material was found completely limited to the area of the central settlement. The assemblage datable to the Early Iron Age also behaves differently, because the material was found distributed in two, almost equally large clusters. Also somewhat specific is the distribution of the Late Bronze-Early Iron Age assemblage. Although it too consisted of one central and several satellite clusters, the latter were slightly larger and denser than the satellite clusters dating to other periods. In other periods despite the dispersal, there was an evident occupational focus, a central scatter, usually much larger and denser than its satellites. Understandably this makes the task of drawing the limits of the actually occupied zone rather delicate. Periods characterized by a more nucleated settlement present a no lesser problem when it comes to determining settlement size. In cases where the material is distributed in more than one, equally large clusters, there emerges the question of their exact chronology. Because of the poor understanding of the chronology of the material, it is impossible to know if these were contemporary or consecutive establishments. Such was the case with the finds from the Bronze Age, the Early Iron Age and the Roman Period. As shown on graph III_1, this makes a considerable difference when discussing settlement size. For example, given that the small Late Roman farm on site 8 was established after the central settlement on site 5a-5b was abandoned, it would imply a completely unperceived episode in the local settlement history at the end of Antiquity, marked by a drastic contraction of settlement and a new relocation away from the main transevrsal. But regardless of all the uncertainties, a number of facts were clearly established. In most periods when the survey area was inhabited, the main settlement varied between the ranks of a farm and a hamlet. The size of the intensely occupied areas could range from several hundreds square meters to slightly over 1 hectare. Nevertheless the prevalent tendency is to have settlements of the rank of a hamlet, measuring between 7-8 000 sq meters and 1 hectare. Only in certain periods, such as the Bronze Age and possibly the Hellenistic, was the area occupied by isolated farms, stretching over not more than 1-2 000 sq meters. At the other extreme stands the Late Iron Age settlement. By the 7th century BC, the central settlement in the survey area moved to a new location, nearly quadrupling in size. At least 3.6 hectares were occupied and there were surface remains in all western sectors of the survey area. It is impossible to understand this sudden expansion of settlement and the subsequent and equally drastic contraction solely on the 322 Cf. H. Bowden, D. Gill, 77-83, eds. C. Mee, H. Forbes, 1997; where the shifting of settlement on and off the coast line is seen as signaling increased or deteriorating security on the see. 149 basis of internal developments. Similar to the regional pattern of the Middle Neolithic, the Late Iron Age centre near Sopot was lying on the edge of a wider network of contemporary settlements, but its appearance in this small and peripheral micro-region remains a mystery. There is also an apparent increase of the total site area during the Roman-Late Roman and the Late Ottoman-Early Modern periods. As explained in a number of occasions, the remains from these periods represent a special case. A considerable portion of the combined areas of the Roman sites is comprised of scatters of brick and tile. They don’t comprise full domestic assemblages and this feature distinguished them from sites with clear traces of full residential activities. Excluding these sites from the total occupied area and assuming that the central settlement on site 5a-5b was contemporary with the smaller satellite on site 8, the Roman Period brought only a slight expansion of settlement. The combined areas of sites 5a-5b and 8 measure about 1.3 hectares. However if we add sites 13a-13b, 14 and 15, the total occupied area will increase to slightly over 2 hectares. Taking into account the evidence for intensified agricultural exploitation, this isn’t unlikely. It must be stressed however that as in all earlier periods, the individual Roman to Late Roman settlements didn’t exceed the ranks of farms and hamlets. On the other hand, the increased size of the Late Ottoman-Early Modern settlement doesn’t agree with the written records relating to the number of households in the village. According to the textual evidence, Late Ottoman-Early Modern Sopot had roughly the same number of inhabitants as its Early Ottoman predecessor, but it occupied a three times greater area. It was suggested that this doesn’t necessarily reflect an increase in population, but rather a change in social relations, living standards and agricultural economy. Graph III_1: total site areas by period, in hectares (the area of the castra excluded). SpecPur p 5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 Satellite CentSett MidNeo Bronze Age LBA-EIA EIA LIA Hell Rom-Lrom LB-EO LO-EM Undated It is important to briefly examine the significance of the various conditions that favoured this stability concerning settlement and population size. Throughout its long history of human habitation, the basin of Sopot was either inhabited by a single extended family or clan or by tiny, closely knit communities of not more than 30 households. The considerable expansions during the Late Iron Age and also possibly during the Roman-Late Roman Period seem to suggest that environmental factors alone couldn’t have constrained the development of a larger settlement in the survey area. Although the Late Iron Age expansion was a unique and relatively brief episode, the fact that this settlement existed for at least two centuries indicates that during most periods represented in the surface record, the population size was below the maximum carrying capacity of the surveyed area. It is nevertheless possible that during the Late Iron Age, Sopot was a major regional centre, partly relying on the agricultural products from neighbouring micro-region. It is therefore important to try and examine the agricultural potential of the survey area and its impact on the size of local settlement. 150 This subject was partly addressed in the description of the survey area, in Chapter II. There are two major problems standing in our way. First is the absence of published studies dealing specifically with the agricultural potential of the Sopot Basin or the wider region of the upper Mid-Vardar. We therefore have to rely on the scattered evidence collected by early ethnographers, the reports (mostly unpublished and difficult to find) of regional soil studies and traces of earlier agricultural activities visible on the ground or on aerial photographs. But while it is possible to roughly determine the size of the cultivable land, we are not in a position to estimate the agricultural potential of the area in terms of kg of grain per hectare. Nevertheless one can predict that the productivity of the agricultural land along the Vardar Valley is equal or slightly higher in comparison to the regions of central and southern Greece 323. The region is characterized by greater humidity, though soil stability must have presented a serious problem, especially along certain sections of the Middle Vardar Valley. For the purposes of this analysis, we will assume that the minimum a family needed to secure its basic subsistence needs were about 3.5 hectares of farming land324. Normally however, (and in view of the mixed agropastoral economy of later historic periods, when a certain portion of the agricultural land must have been reserved for animal fodder) the individual estates of independent families were certainly larger. The second problem is related to the more general issue of what proportion of the territory actually exploited by the settlement falls within the limits of the survey area. Because of the character of the local topography, settlements aren’t necessarily positioned in the centre of their territories, nor are their territories equivalent to an orographic or a hydrographic entity. The Roman to Late Roman sites situated in the eastern periphery of the survey area are particularly problematic, because more than half of their land could belong to the neighbouring drainage of the Vranov Dol. It is also nearly certain that the agricultural territory of the settlements in sector I didn’t belong to the surveyed basin and was probably spread on the Vardar’s right bank. In this context we further need to stress the fact that we remain ignorant of the specifics of the local agricultural economy in nearly all periods of occupation represented in the surface record. Analyzing the distribution of the Ottoman to Early Modern settlement in the broader region of the second survey area, we will see that this peripheral positioning of settlements could reflect a careful economic strategy of securing equal access to a variety of natural resources. In order to avoid this problem, it might be useful to draw theoretical boundaries around the peripherally positioned settlement sites. We can do this following the principles established by the site catchment concept325. The central tenet of this theory is that there is a natural limit to the territory exploited by all pre-industrial societies. Depending on the general type of the economy, hunting and gathering, pastoralist or farming, the radius of the settlement’s or the camp’s territory can range between 10 and 5 kilometers. The exploitation of natural resources situated beyond these limits becomes economically unviable, because of the increased time-costs of traveling to and back from the production locus on a daily basis. In practice when farming communities in Greece and the Balkans are considered, the empirical evidence suggests that 323 M.H. Jameson et al. 280-285, 1994; van Berghem, J-W Fiselier, J. Soils and Land Use Potential, 57-71; W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al. 2002; R. Shiel, A. Stewart, The soils and agricultural potential of the Thespiai area, 95-109, J. Bintliff, P. Howard, A. Snodgrass et al, 2007. 324 M. H. Jameson et al. 283, 1994; J. Bintliff, P. Howard, A. Snodgrass et al, 107-108, 143-151, 2007; cf. R.W.V. Catling, The Survey Area from the Early Iron Age to the Classical Period (1050-300), 151-256; W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al. 2002. 325 E.S. Higgs, C. Vita-Finzi, 1-37, 1970; J. Bintliff, 505-513, ed. G. Barker, 1999. 151 shorter radii of 2-3 kilometers or half an hour walk on flat terrain are more common 326. It is been estimated that a territory with a radius of 2-3 kilometers is sufficiently large to sustain a community of a few thousand people or settlements of the rank of large villages or small towns. This already implies that the farms and hamlets in the basin of Sopot required much smaller territories to secure their subsistence needs. Even in the rugged landscape of the upper MidVardar Valley, where it is very likely that a large portion of the theoretical territory is not cultivable, an area of over a dozen square kilometers will by far exceed the needs of the small farming communities. But simply drawing buffer zones around the settlements in the first survey area will hardly be of any help, because of the extremely fragmented and rugged terrain. For example, it is obvious that the settlements on sites 4 or 5a-b didn’t exploit with equal intensity the land on the left and the right bank of the Vardar, although the latter consumed over 50% of their theoretical territories. This portion of the potential hinterland was cut off by the deep Taor Canyon and the walking distance to the Vardar’s right bank could extend to over 1 km. Obviously these small hamlets could comfortably procure their needs from the land east of the Vardar, but again, merely drawing buffer zones with smaller radii doesn’t solve the problem, because we saw that the great majority of the sites were located on the periphery of the farmland (map III_68). In order to carry out this analysis, it will be necessary to work with time-distance rather than with linear distances and create a 3-dimensional model of the survey area. As this is unavailable, for the present purpose we will have to follow the obvious topographic divisions of the area. In practice this means treating all hill-sides and ravines as simple physical barriers, although it is known that even these uncultivable sections of the landscape were exploited for certain resources. In this particular case however, it is evident that the dry and barren hill-side was of little economic value. Apart from being used as extensive grazing grounds during the months prior to the harvest season, the hills separating the drainages of the Sopot, the Vardar and the Vranov Dol offered few other resources, including wood, various herbs and wild fruits327. Because of the specific distribution of the settlements in the first survey, we will have to establish the approximate carrying capacity of three different areas (map III_69). The first is located outside the drainage of the central valley, on the Vardar Valley floor. It includes the terraces on the Vardar’s left bank (our sector I) and the low-land area on the opposite bank, around the confluence of the Solpski Stream and the Vardar. This niche was occupied by a small Bronze Age settlement, while in the more recent past there was a small hamlet on the Vardar’s right bank and the monastic complex on the opposite eastern bank. It is also certain that there were activities during the Roman to Late Roman Period, but these weren’t necessarily residential. The second area coincides with the central valley and its drainage. It is much larger and encompasses the territories of the great majority of settlements discovered in this survey area. Finally, the third area is in part theoretical, spreading over the eastern half of the drainage of Sopot and over the western half of the neighbouring Vranov Dol. This is the hypothetical territory of the Roman to Late Roman settlement on site 13a-13b. But even here a more flexible approach is needed, because given that this settlement was contemporary with site 5a-5b, its potential territory on the west will be limited by the hinterland of its larger neighbour. Situated at a distance of about 1.5 kilometers and surrounded by extensive stretches of inhospitable and barren terrain on the north and west, the only agricultural land available to the communities that 326 J. Bintliff, 513-518, ed. G. Barker, 1999. O. Rackham, Observations on the historical ecology of Laconia, 73-119, W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al. 2002. 327 152 occupied site 5a-5b would have been the surveyed basin. This implies that the bulk of the agricultural land belonging to the inhabitants of site 13a-13b fell within the drainage of the neighbouring stream, the Vranov Dol and was left outside of the survey area. The tiny island of fertile land spreading on both sides of the Vardar, near its confluence with the Solpski Stream is covered with fertile soils derived from flysch and alluvial sediments. In addition this land has some regional strategic importance, because it is one of the rare spots where the Vardar can be crossed in the Taor Canyon. However the arable land available, including the low hills on the Vardar’s right bank totals a maximum of about 25 hectares. This niche could sustain not more than several families, assuming that they owned humble properties, occupying slightly over 3 hectares of farmland. This suggestion is confirmed both by the census data for the Ottoman village Vlahčani situated on the Vardar’s right bank 328 and by the surface survey results. According to the official records of the Early Ottoman Period, Vlahčani was a small hamlet, consisting of not more than 10 households. It should be stressed however that the territory of these settlement was probably confined to the western bank of the Vardar, as the narrow terraces on the opposite bank were a monastic land. Thus it is possible that at least during certain periods, the left and the right bank of the Vardar didn’t form a single settlement niche. The narrow terraces on the left bank could barely sustain a single household. Even if assuming that 100% of the cultivable land was exploited and deducing 0.1 hectares for the occupied area, the maximum amount of farmland in this sector measures about 3.7 hectares. The surface artifact survey revealed that in the Bronze Age and possibly the Late Bronze Age, the site of the modern monastery was indeed occupied by a small farm, measuring less than 0.1 hectares. As explained earlier the basin of Sopot is a well enclosed hydrographic entity, surrounded on all sides by low, but rather extensive and rugged hill masses. Because this is a geologically dynamic region, it is often difficult to draw the exact limits of the watershed line, especially along the northern periphery of the basin. Nevertheless in chapter II, on the basis of very detailed military topographic maps, we estimated that the valley drains about 10 sq kilometers. Only a very small fraction of this land is cultivable, almost entirely concentrated on the terraces and the floor of the central valley. The modern plough-zone spreads over half of this potentially arable land and measures about 80 hectares, roughly overlapping with the intensively surveyed area. But judging by the vegetation patterns and the remains of old agricultural terraces visible on the ground and on aerial photographs, at certain points in time, the agricultural potential of this region was more fully exploited. There are clear traces of agricultural activities (old hedges and terraces) along the eastern periphery of the survey area, on the western foot of Radičica and in the upper course of the central valley. Taking into account these peripheral zones of the landscape, the size of the potentially cultivable land increases to about 230 hectares: 143.5 along the middle and the lower course of the valley, 83.3 hectares along the upper course. In addition there are also traces of agricultural activity on the upper portion of Prisoj, but these are scattered fields amounting to 11 or 12 hectares of poor land. Thus the potentially arable land in the basin of Sopot totals about 240 hectares or 2.4 sq kilometers. It has to be stressed that in terms of agricultural potential, there are considerable variations across different sections of the landscape. The land east of the central valley and in its upper course was apparently less productive than the land west of the central valley, at Prisoj’s foot and on the valley floor. The thin soils covering the upper portions of Prisoj or the western slope of Radičica were probably the least attractive, not least because of the difficulty of access and the dangers of failed harvests due to low 328 M. Sokolovski, 183, 1971. 153 precipitation. It is no doubt unfortunate that there is no precise information concerning the agricultural productivity of the different sections of the agricultural land. Like the (contemporary?) settlement on site 5a-5b, the Roman to Late Roman sites 13a13b, 14 and 15 stand on a watershed line, at the opposite eastern end of the basin. They occupy the Jakupica Ridge, which separates the basins of Sopot and the neighbouring stream to the east, the Vranov Dol. Theoretically these sites had an equal access to both valleys, but given that most of the lower half of the basin of Sopot was exploited by site 5a-5b, the bulk of their agricultural territory had to belong to the drainage of the Vranov Dol. This stream flows through a slightly larger valley, with more arable land. Only on the right bank of the Vranov Dol there is over 1 sq kilometer of low-lying terrain, covered with relatively fertile Tertiary deposits. These are the same sediments that cover the eastern half of the surveyed basin, although the local soils are better preserved than in the basin of Sopot329. At present the entire lower half of the Vranov Dol is under vineyards. The terrain lies within 15 to 20 minutes walking distance from sites 13a-13b and its direct exploitation doesn’t present a particular problem. With over 100 hectares of good arable land, it could sustain a settlement as large as the one on site 5a-5b. In fact only the northern end of this stretch, the fields north of the Skopje-Thessalonica highway and located in the immediate vicinity of sites 13a-13b, offer nearly 35 hectares of productive land. Thus a very small hamlet consisting of 6-7 families, each cultivating between 5 and 6 hectares or a substantial farm could comfortably live off the immediate surroundings of sites 13a-13b. We find it unlikely that the settlement on this site exploited the entire western bank of the Vranov Dol. The exact size of this site remained undefined, although it is clear that it didn’t occupy an area larger than site 5a-5b. The view that this site didn’t exploit the entire western bank of the lower Vranov Dol primarily derives from its location in the surrounding landscape, although this was also indicated by the relatively low artifact density and the character of the material. These sites are fairly detached from the lower half of the neighbouring valley and although they occupy the watershed line between the basin of Sopot and the Vranov Dol, they evidently gravitate towards the former, especially the component on site 13a, sites 14 and 15. The eastern foot of the hillock occupied by site 10 or the ridge on the opposite bank, represent far more suitable bases for the exploitation of the Vranov Dol. What are the implications of this analysis for the population dynamic in our survey area? Given that a property of about 3.5 hectares is the minimum required to feed a single family, the estimated agricultural potential of the basin of Sopot can sustain a maximum of 66-67 households. This however hardly leaves any space for fallowing or the cultivation of fodder. If we allow that a portion of the cultivable land was given away to animal fodder or was left uncultivated and increase the minimum agricultural estate to about 5.5 hectares, the carrying capacity of this micro-region drops to not more than 45 families. These generally accepted minimum quanta of farmland for traditional agriculturalists are actually much lower in comparison to the scanty information pertaining to the region of the upper Mid-Vardar in the first half of the 20th century330. According to this ethnographic account, the average peasant farmers in the neighbouring villages owned about 50 dunums or 8 hectares of farmland. As the author himself admits however, the particular timing of the ethnographic survey, immediately after World War I was probably misleading, because the majority of the peasant-farmers in the country amassed substantial expanses of arable land from the fleeing Ottomans in the early decades of the 20th century. Indeed as the author remarks, a great portion of this land laid 329 330 Gj. Tanev, 1969. M. Filipović, 503, 1935. 154 uncultivated, because the local farmers simply lacked the labour force and the means to exploit it. In some of the villages, the informants remembered that prior to the retreat of the Ottomans, the poorer farmers owned five times smaller estates or barely 2 hectares of arable land. It is unlikely that an estate of this size could sustain a family of four and this information must relate to the Late Ottoman Period, when the majority of the peasants worked on the estates of the local landlords. Turning to the size of the local communities inferred from the survey results, we see that during most periods in the past population levels were below the maximum carrying capacity of the area. Obviously the over 200 hectares of potential agricultural land was far above the needs and the production capacity of the communities that inhabited the Bronze Age and Hellenistic farmsteads. It is certain that during these periods, only a small fraction of the cultivable land was being exploited. But even the predominant settlement rank in this survey area, the hamlet measuring between 0.5 and 1 hectare didn’t stretch the agricultural potential of the basin to its limits. Assuming that these settlements housed between 15 and 30 households331, their basic subsistence needs could easily be met by focusing the agricultural production only on the lower and the middle course of the central valley. Even if we accept the upper limit of 30 households (which was the maximum recorded in the mid-15th century Ottoman censuses), each family could cultivate over 4.5 hectares. Given that the entire basin was under cultivation (which according to the site catchment theory was technically possible), the estates of the individual households will increase to 8 hectares. This sharply matches the figures reported by farmers from the neighbouring villages in the period between the two world wars. It is thus clear that during most of the local settlement history, the small farming communities could comfortably live off the land available in the immediate vicinity of the central settlements, i.e. the lower and the middle course of the central valley or the western bank of the Vranov Dol. Even here a portion of the land available could be given away to a less intensive regime or pastures. Indeed this is what the survey data suggest: evidence of less intensive, off-site activity was almost always limited to a single survey section, in the immediate surroundings of the settlement. If these halos spreading for a couple of hundreds of meters beyond the site peripheries mark an area of intensive agriculture and gardening, it could be that the remaining land was less intensively exploited. A similar pattern of land-use was documented by M. Filipović in the early decades of the last century: wheat, vines and garden cultures in the immediate vicinity of the village houses and on the valley floor, rye, barley and occasionally millet on the more distant fields and along the upper portions of the surrounding hills332. The only period when the agricultural resources of the surveyed basin were becoming perilously strained was the Late Iron Age. Again the data inferred from the surface artifact survey and the estimates concerning the agricultural potential of the surveyed basin aren’t contradictory. Assuming that the entire basin, including the poorer soils in the hillside was under cultivation and allowing for very small estates, not larger than 3 hectares, a maximum of 80 household could live off this land during the Late Iron Age. However this would be an impossible regime, for it not only consumes the entire productive land leaving no room for fallow and pastures, but it also doesn’t take into account the fact that the area of the extensive mound necropolis and the 3.6 hectares occupied by the settlement were left out of the plough331 This is a rather generous estimate, at least for certain historic periods, cf. J. Bintliff, P. Howard, A. Snodgrass, et al. 144-145, 2007; cf. R.W.V. Catling, 205-211, W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al. 2002. 332 M. Filipović, 502-507, 1935. 155 zone. If we deduce the 15 hectares occupied by the settlement and the mound necropolis from the sum of the potentially arable land and allow for larger individual estates, closer to the 5.5 hectares limit, the maximum number of households will shrink to about 50; a figure which is still substantially higher than in all other periods of human occupation in the survey area, but slightly lower than suggested by the survey results. It is possible that we’ve slightly overestimated the size of the Late Iron Age community, but even if we agree that it comprised not more than 50 households, it is evident that the agricultural potential of the area was becoming exhausted. In fact a community of 50 households could still live off the land without overstraining its resources, though it would imply that the entire agricultural potential of the basin was used. One shouldn’t exclude the possibility that this limit was actually crossed at a certain point of time during the Late Iron Age and that for a generation or two, the natural resources of this microregion were becoming truly overexploited. This could indeed be one of the central reasons behind the sudden demise of the Late Iron Age village. The Roman to Late Roman Period also requires a brief reconsideration in the light of the estimated carrying capacity of the survey area and its surroundings. If we assume that the majority of the possible settlement sites from this long period were at least partly contemporary, the total occupied area increases to slightly over 2 hectares. In terms of population size, this is still below the productive limits of the basin of Sopot, even if we make a further assumption that these were all nucleated settlements housing a total of between 30 and 40 households. However the distribution of the Roman to Late Roman settlement sites is not ideally suited for the exploitation of the surrounding landscape. They are all situated in the lower half of the surveyed basin and even site 13a-13b is closer to the central valley than to the fertile eastern bank of the Vranov Dol. Agricultural exploitation of the peripheral upper course of the central valley or the banks of the Vranov Dol is still viable, although it would be logical to discover at least one of the settlement sites in these zones. We believe that this clustering of sites in the lower and middle course of the central valley, once again reinforces the impression that the Roman to Late Roman Period didn’t bring a particularly strong population increase. It merely saw a return to the earlier, pre-Late Iron Age population levels. The increased settlement area could reflect a change in the standards of living in the countryside; we know that during the Roman and Late Roman Periods, individual farms could easily occupy several thousand square meters 333. With the possible exception of site 5a-5b, the rest of the sites were almost certainly individual farms or outbuildings with agricultural functions. In fact this could be one of the factors contributing to the relatively low on-site densities, even on locations where prehistoric material was present in much larger quantities. It also implies that the intensified agricultural exploitation evidenced by the appearance of a very thin off-site carpet wasn’t necessarily incited by local demographic pressures. The calculating of the survey area’s agricultural potential basically confirms what was anticipated in our earlier discussions concerning the size of the local community; in particular, the evident tendency to stabilize around the limit of 100 to 150 individuals. As shown on graph III_1, over half of the settlements discovered in this survey area measured between 0.5 and 1 hectare. Further growth was evidently unconstrained by the local agricultural potential, but there was another underlining factor, identified by physical anthropology a couple of decades ago 334. These small farming communities could continue their existence as largely egalitarian (that is 333 Cf. J. Bintliff, P. Howard, A. Snodgrass, et al. sites THS 3, THS 4, 2007. R.I.M.Dunbar, Coevolution of Neo-cortex size, group size and language in humans, 681-735,Behavioral and Brain Sciences 16 (4)1993; J.L. Bintliff, 526-532, ed. G. Barker, 1999. 334 156 lacking formal vertical or horizontal subdivisions) and exogamous, as long as their population remained below or close to the threshold of 150 individuals. The analysis of the carrying capacity of the wider study region demonstrated that the small agricultural potential of the basin of Sopot precluded the emergence of a larger, corporate community. Its land could barely sustain a population of 300 and it is now clear that even the Late Iron Age settlement didn’t exceed the hamlet-rank. But although the limited agricultural potential of the region prevented the emergence of a larger settlement, it was social factors that kept the population at an even lower level. Without the more refined pottery chronology, we remain ignorant about the settlement dynamic during the separate periods of occupation. For example, we don’t know if the Late Bronze-Early Iron Age hamlet developed gradually from a settlement of a lower rank or if its foundation consisted of one or more episodes of colonization. Likewise we can only guess what happened with the local settlements at the end of nearly all periods of occupation, the only exception being the Late Ottoman-Early Modern Period. Apart from the Late Iron Age, there is no evidence in the surface archaeological record that would suggest the existence of phases during which the settlements grew beyond the threshold of 15-30 households. This means that the local communities had fairly efficient mechanisms of controlling population growth and that environmental and demographic pressures weren’t the decisive factors for the demise and relocation of subsequent settlements. Evidently there aren’t too many gaps in the local history of habitation, a fact which nicely demonstrates the stability of the surveyed basin as a settlement niche. Nevertheless certain time periods such as the Early Neolithic, the Eneolithic, but also much of the Middle Ages, left no recognizable traces in the surface record. Even allowing the possibility that some evidence was overlooked or misinterpreted, it seems likely that during these periods, settlement in the survey area experienced considerable contraction. In fact it is theoretically possible that brief episodes of abandonment marked the ends of almost all settlement phases – though this could again stem from the low chronological resolution. Yet the frequent discontinuities observed in local pottery production between two subsequent periods and the lack of evidence of gradual horizontal displacements seem to indicate that the inhabitants of the survey area showed little respect for the landscape of the preceding periods, at least concerning the central settlement locus. To a certain degree, this supports the thesis that most periods of occupation where separated by brief intervals of radical transformation of the local societies or even complete abandonment of the basin. Adopting a long-term perspective it is possible to observe four, possibly five asymmetric cycles of growth and decline of settlement in the survey area, at surprisingly equal intervals. The earliest and probably longest cycle covers the period of the Middle Neolithic and possibly, the Late Neolithic. Naturally it all depends on the chronology of the latter assemblage. But even if it truly dates to the Late Neolithic, it is possible that there was a longer period of abandonment before the Middle Neolithic site was re-occupied, especially bearing in mind the differences between the fabric groups of the two assemblages and their different distribution patterns. If this was the case, there were two separate cycles during the integral period of the Neolithic, bringing them closer to later cycles regarding longevity. The next cycle of growth and decline begins towards the end of the Late Bronze Age, culminating with the Late Iron Age unparalleled expansion and abruptly ending by the early Hellenistic Period, at the latest. It is interesting that within this long-term perspective an abandonment of the area by the early 5th century BC would be in a greater accord with the overall dynamic. The third or the fourth cycle begins sometime in 157 the Hellenistic Period and ends with the collapse of Late Roman authority on the Balkan Peninsula, towards the end of the 6th and the early decades of the 7th century AD. The last cycle covers the period between the 14th and mid-20th century, ending with the sudden decline and abandonment of the old village and the gradual emergence of a new, dispersed pattern over the past couple of decades. The coarse chronological framework prevents us from determining the exact duration of every cycle, but it is evident that the periods during which the local settlement retained the rank of a hamlet or small village didn’t last longer than 5-6 centuries. Although there is very little evidence, one can imagine that in most periods of the past, the local settlement was a part of a wider network of settlements of similar rank and size. This is hardly surprising knowing the geographic conditions in the region of the upper Mid-Vardar. It is unexpected episodes such as the sudden rise of Sopot during the Late Iron Age or the lack of evidence dating to the Eneolithic or the Early and Mid-Byzantine Period that call for similarly designed research in at least some of the surrounding micro-regions. The small communities, consisting of not more than a couple of dozen households were necessarily bound to maintain close relations with their neighbours. Thus although geography carved a separate niche for each of these communities, their survival depended on the normal functioning of a highly integrated social, economic and demographic network that surpassed micro-regional limits. To obtain a better understanding of the developments in one component of this network, it is necessary to have at least some idea about developments in the neighbouring components. By the same principle, the rugged region of the Taor Canyon shouldn’t be seen in isolation. Settlement pattern and dynamics in this peripheral area must be sensitive to developments in the larger settlement centres in the basins of Skopje and Veles and on the Ovče Pole Plateau. Knowing more about the history of the nearest larger centres will doubtlessly shed more light on the developments in the “periphery” and vice-versa. 158 Chapter IV: The region of Montenegro near Skopje and the second survey area IV.1 Name and geographic location of the wider study region Around 15 kilometers to the north of modern Skopje rises a low, but extensive mountain range, known as Montenegro near Skopje or Skopian Montenegro, to differentiate it from the later kingdom of Montenegro on the Adriatic Coast (map IV_1a and 1b). Its main ridge stretches for over 20 kilometers in a NW-SE direction, roughly parallel to the course of the Vardar. Thanks to its peculiar arched shape, it effectively encloses the plain of Skopje from the north and northeast. The mountain mass has an irregular, triangular shape. It is bounded by rivers on all four sides: the Vardar on the south, its tributaries the Lepenec and the Pčinja on the west and east, while on the north it is drained by the river Binačka, belonging to the basin of the Morava and ultimately to the Danube. Located at the watershed between the drainages of the Vardar and the Morava, Mount Montenegro presents one of the larger crossroads on the Balkan Peninsula. The main arterial road coming from the Aegean splits in two directions at the southern foot of this mountain, one leading towards the Adriatic, the other continuing north towards the Danube. The main mountain ridge stretches asymmetrically across the southwest half of the mountain and has a gentle southeast inclination. Its southern slopes are much steeper than its eastern or northern sides. In fact the bulk of the mountain mass falls to the north and northeast from the main ridge. Consequently the southern foothills are well articulated in the relief, while to the north and east the mountain gradually merges into the surrounding valleys. Particularly well defined is the northwest half of Montenegro’s southern foothills, the portion situated directly beneath the highest peak Ramno, at 1658 meters above sea level. Geographically the southern foothills of Montenegro belong to the Skopje Basin, but this particular portion stands nearly 200 meters above the basin and it is also delimited on the west and east by two lower offshoots that stretch southwards from the main mountain ridge. It thus presents a separate micro-region; in essence a small oval plateau roughly measuring 5x6 kilometers, excluding the more extensive mountainside. Seen on the map it appears as a smaller outlet of the Skopje Basin. Interestingly in the past the name Montenegro referred exactly to this part of the mountain. The name extended to designate the entire mountain much later, probably not before the late 19th century. An ethnographic record dating from the first years of the 20th century stresses that the local people still distinguished themselves from their closest neighbours on the mountain by their clothes and their name, Montenegrins335. This micro-regional and ethnocultural entity is also one of the oldest known administrative units in the country. As early as the mid 13th century the name Montenegro designates the land stretching from the highest peaks of the mountain on the north to the plain of Skopje on the south and between the ridges of Markov Kamen on the west and Buzalak on the east336. Thus defined it is one of the oldest administrative units in the wider area, measuring approximately 120 square kilometers. Thanks to the rapid growth of Skopje during the last century, most of the micro-regions that comprise the Skopje Basin have either lost their integrity or came to be seen merely as peripheral parts of the city’s greater area. A similar fate awaits the region of old Montenegro. S. Tomić, 409-509, 1905. There is an extensive list of bibliography on this region, covering a period of over one century. Early studies are nicely summarized by J. Trifunovski, Skopska Crna Goraanthropogeographische Forschungen,Skopje 1971. 336 The area is first mentioned in an edict issued by the Bulgarian king Constantine Asen in the late 1250’s. V. Mošin et al. Gramoti na manastirot Sv. Gorgi-Gorg Skopski, 181-209, eds. V. Mošin, et al, 1975. 335 159 Situated in a distant corner of the Skopje Basin and embraced by the mountain on three sides,so far it has escaped the urban development that eclipsed the micro topographic divisions in the rest of the basin. But the advances in technology and living standards have not bypassed the villages in the area. Only a small percentage of the people still inhabit the traditional houses in the old cores of the villages. Most live in modern houses surrounded by spacious yards, built outside the limits of the old settlement337. Modern roads lead to all of the villages in the area and some even have access to public transportation. Many of the local inhabitants work in Skopje or have relatives and friends in the city. But there are also movements in the opposite direction. Every year parts of agricultural land are sold and transformed into villas for the richer city dwellers. Over the past few decades, the mountain with its lush nature and a number of well preserved monastic churches dating to the early 14th century has become one of the most visited resorts in the vicinity of Skopje. The region of Skopian Montenegro has certainly maintained close relations with Skopjeby the Middle-Byzantine Period, but its administrative and economic integrity came under threat only with the rapid expansion of Skopje in the second half of the 20th century. It is a good illustration of the impact of the current settlement pattern on our reading of the landscape. Modern geography sees the area discussed as an integral part of the Skopje Basin, but in the past this was probably never the case. Indeed both geologically and administratively, the area of Montenegro belongs to the mountain bearing the same name rather than to the plain of Skopje. Early geographic researchers of the Skopje Basin have rightly defined a separate habitation zone consisting of the settlements located in the foothills and on the lower ridges of the mountains that surround the plain of Skopje338. The region of old Montenegro ideally illustrates the main characteristics of this habitation zone. All but 2 of its 11, currently existing settlements are situated exactly along the foot of the mountain, where the mountain massif meets the plain. Clearly the aim of this settlement pattern is to provide an equal access to the mountainside and the plain at the mountain foot, for all major settlements in the area. Further to the east, along the southern foot of mount Montenegro or on the opposite side of the plain, at the northern foot of mount Vodno, there are a number of villages similarly located at the edge that separates the mountain from the plain. In all these cases the settlement territories extend over part of the mountainside and over part of the flat land in the foothills. Moreover the very topographic entities that they occupy, the ridges and the small valleys, often run across both geographic zones. This peculiar geographic positioning naturally had an important effect on the socio-economic character and the history of these settlements. They are of a different kind from the villages in the central parts of the Skopje Basin and from those situated in the mountain interior. IV.2 Limits and geography of the wider study region No other part of the Skopje Basin unites such contrasting types of environments as the region of Montenegro. It encompasses the steepest, southwest section of the mountain and a small ovoid plateau, gently inclined towards the inner Skopje Basin. It thus occupies an area where the plain extends furthest into the mountain massif. This is the principle factor that distinguishes the region of Montenegro from both the rest of the mountainous area and the plain of Skopje. Studied by a number of naturalists and ethnographers and existing as a separate 337 338 J. Trifunovski, 24-25, 1971. J. Trifunovski, Seoska Naselja Skopskog Polja, 345-517, Srpski Etnografski Zbornik LXIX 1955. 160 administrative unit for a very long period of time, the borders of this region are well established339 (map IV_2). On the west it follows the watershed line between the river Lepenec and the streams that drain the southern slope of mount Montenegro. On the north and northeast, the border coincides with the flattened main mountain ridge. In its westernmost part to the point of its highest peak Ramno, it stretches in an east-west direction. Roughly a kilometer to the southeast of Ramno, a new ridge springs out in a southeast direction, running parallel to the longitudinal axis of the Skopje Basin. It encloses the area from the northeast. The regional borders to the south and southeast are much vaguer. On the southeast side, immediately to the east of the village Ljubanci, there is a low mountain ridge called Buzalak. It is a low offshoot of the mountain with gentle sides, but it reaches slightly deeper into the plain, partly enclosing the southeast side of the area. Even less clear is the border on the south, towards the Skopje Basin. Earlier researchers have drawn this border somewhere between the contour lines of 390 and 400 meters above sea level. There is no other physical barrier separating the two, but a series of low hillocks after which the ground falls sharply for about fifty meters into the plain of Skopje. It is the differences in micro-relief and in the geo-pedological substrate that have set apart the small plateau at the foot of Mt. Montenegro from the rest of the Skopje Basin, rather than a particular topographic barrier. 75% of the territory of Skopian Montenegro belongs to the mountainside. The width of this mountainous belt varies from between 3 and 5 kilometers on the western and eastern flanks to nearly 10 kilometers on the north. The relief is broken up into a number of ridges separated by narrow, but very deep valleys. The widest is the northern, central part. It features a denser hydrology and rises more gently than the flanks. It seems nearly impossible to summarize the complicated web of steep ridges and narrow valleys that make up the geography of the mountainside. The mountainous part of this region wasn’t included in the intensive survey. At present it is thickly overgrown and except for the few famous monastic churches, it is rarely frequented by humans. However only a few centuries ago, to the local inhabitants the resources of the mountainside would have been as important as the small plain at its foot. Ethnographic accounts from the early 20th century reveal that during the summers, often half of the families moved to seasonal camps in the mountain to stay with their herds340. A number of abandoned settlements mentioned in the Early Ottoman records were probably situated somewhere in the vast mountain interior341. Earlier extensive surveys have revealed three Late Roman forts in the area342. Finally, a brief description of the local geography is necessary, because the geography of the flat land at the mountain foot is basically a continuation of the geography of the mountainside. Mount Montenegro is unusually opulent with streams and freshwater springs. There are at least a dozen small streams issuing from the main mountain ridge. Needless to say these are all minor streams. Their valleys are very narrow, not wider than several meters across the bottom and with very steep sides, rising hundreds of meters above the valley floor. These minor basins are separated by an equal number of narrow and steep mountain ridges, usually with flattened tops and terminating with low, rounded peaks. They all spring from the main mountain ridge, 339 J. Trifunovski, 7, 1971. S. Tomič, 447-450, 1905; J. Trifunovski, 50-54, 1971. 341 J. Trifunovski, Isčeznati naselbi na Skopska Crna Gora, 145-149 Godišen Zbronik na PrirodnoMatematički Fakultet 18, 1970. 342 N. Čausidis, Novootkrieni docnoantički tvrdini na Skopska Crna Gora, 183-196, Macedoniae Acta Archaeologica 10 1989. 340 161 gradually converging towards the mountain foot. In the eastern half of the area they are mostly orientated southwest, while in the western half they spread in a southeast direction. Local paths leading across the mountain normally follow the crests or the upper portions of these ridges. They all lead to the main mountain ridge and beyond, into the basins of the Pčinja and the Binačka. It is possible to reach the summit from the foot of the mountain by following virtually any of these mountain ridges. Roads that follow the valley floors on the other hand are usually dead-ends. They connect the villages with the monasteries situated on the valley floors and with the numerous freshwater springs dotting the sides and the floors of the valleys. Though small, the water channeled by these valleys is fast and powerful. The difference in height between their headwaters and the point where they leave the mountain is nearly 1000 meters and the distance about 7 kilometers. Dozens and dozens watermills can still be seen along the courses of almost all streams that drain this portion of the mountain. It is difficult to point to any principle difference between the ridges and the valleys that comprise the local relief. Except for the ridges that flank the micro-region from the west and east, all exhibit not only similar proportions, but similar micro-topographies. They all rise very gently towards the main mountain ridge and at roughly the same height. Even the small rounded peaks, often marking their ends look very similar and have almost equal heights. One of the first naturalists that studied this area remarked that looked at from the flanks, the summit lines of these ridges overlap, forming a continuous arch-shaped stretch running almost parallel to the edge of the foothills343. The entire terrain on Montenegro’s southwest slope basically has a terraced structure. This is difficult to see because of the complex hydrology. It is nevertheless possible to follow a series of “floors” at approximately the same heights, along the greater part of the mountain slope. This provides an important clue to the geomorphology of the mountain slope and of the entire micro-region. Although easily accessible the mountainside offers only very small and narrow patches of flat, inhabitable land. Some of them coincide with the aforementioned terraces, at 690, 740 and 900 meters above sea level. These are mostly concentrated on the summits and the upper portions of the ridges and only rarely on the floors of the narrow valleys. The latter are usually reserved for the monasteries, while the former type of locations were mostly turned into pastures and summer camps, at least until the last couple of centuries. Some of them bear the names of abandoned villages and it is most certain that they were indeed locations of earlier settlements. In fact Brodec, the only existing village in this part of the mountain is located on one such location. There are a few other locations in the mountainside sufficiently spacious to receive a settlement of minor size. The majority of them are not more than 2 to 3 kilometers away from the foot of the mountain. These are the summits of the ridges that rise above the modern villages and the medieval monasteries, slightly further into the mountain interior. Spacious pastures with summer camps are also mentioned along the summit of the main mountain ridge, on the top or at the foot of the highest peaks. The peak of Gnoino Ramno, a kilometer and a half to the south of the highest peak is basically a small plateau, measuring over 15 hectares of flat land. Like the rest of the mountainside, these locations were abandoned long ago. They are either covered by the regenerating forest or by a tall and very dense grass cover. In the western end of the area, running across the mountain’s western shoulder and parallel to the river Lepenec is a shallow pass, flanked by a series of low, rounded peaks. It is over 3 kilometers long and at certain points, nearly 200 meters wide. Throughout its length it rises for not more than 50 meters, opening an easy path between the plain of Skopje and the 343 P.S. Jovanović, Reljef Skopske Kotline, 62-116, Glasnik Skopskoga Naučnoga Društva X, Skopje 1931. 162 valley of the Lepenec, an alternative to the modern motorway and railway that run along the valley floor through the canyon of Kačanik. It is the main road linking the plain of Skopje and the rest of the Vardar Valley with the plain of Kosovo and the Adriatic Coast. This pass could actually represent the remnants of the old river bed of the Lepenec, presently broken up by erosion and pulled into the new drainage basin of the river. The geography of the rugged plain in the foothills is a continuation of the geography of the mountain. The line marking the edge of the foothills runs almost parallel to the line of the main mountain ridge. It bears the shape of a deformed horse-shoe, bent at the western end (map IV_3). Measured from the low hillock Gradište near the village Brazda, to the springing point of the Buzalak Ridge, southeast of Ljuboten, it is 13.6 kilometers long. The vague southern border stretches between these two points, approximately along the line that separates the transitional hilly zone from the inner Skopje Basin. The apex of this arch is off-centered in its northwest corner, almost in the centre of the triangle formed by the villages Gornjane, Banjane and Čučer. This is also the point where the pass described in the preceding paragraph enters the plain at the foot of Mt. Montenegro. On the map it appears as a deltoid shaped recess, characteristic for areas where a river leaves a narrow canyon and enters a plain. It clearly defines the westernmost third of the plateau. Standing nearly 100 meters lower than the rest of the foothills, it presents a separate basin shared by three major streams descending from the mountain, the Kučeviški, the Banjanski and the Čučerski. The first two are larger. They are formed under the central ridge of the mountain and flow from the north, while the third is a minor ravine that drains the low hills above the villages Čučer and Gornjani. It practically issues from the end of the pass that links the foothills with the valley of the Lepenec and it is the only stream that enters the foothills from the west. Two and a half kilometers from its spring, it flows into the Banjanski Stream. The basins of the streams Banjanski and Kučeviški are separated by the first of the seven low ridges that compose the wider study region (map IV_4). It stretches for over two kilometers running southwards, parallel to the streams. The main asphalt road connecting this region with the city of Skopje follows the summit of this ridge. It is very possible that it also follows the trace of an ancient road linking Roman Skupi with the mountain. The Kučeviški Stream flows along its eastern flank. Its valley measures between 30 to 40 meters at the bottom and 1200 meters between the watershed lines, i.e. between the first and the second ridge, counting from the west. Much wider is the basin of the Banjanski Stream, west of the first ridge. It basically drains the entire northwest corner of the region, probably inheriting the old estuary of the Lepenec. Apart from the Čučerski Stream, it also gathers the waters coming from a powerful spring at the mountain foot, located between the villages Banjani and Mirkovci. The two streams meet in the southwest corner of the foothills, at the northern foot of Gradište near Brazda. Together they enter the Lepenec River, nearly 4 kilometers to the west of their confluence. Drawn by the powerful erosive forces that pushed the course of the Lepenec to the west, these two streams extended their course carving a small canyon between the hill Gradište and the western shoulder of Mount Montenegro. The western third of the foothills of Mt. Montenegro features a gentler terrain and is probably richer in arable land than the other two thirds. Five of the eleven villages in the region belong to this part of the plateau. Four major streams enter the foothills on the opposite, eastern end. Three of these flow from the north, converging almost at the geometric centre of the plain. From west to east, these are the Pobuški, the Turčevski and the Ljubanski Stream. Around 2 kilometers from their entrance in the foothills, the Turčevski and the Ljubanski Stream merge and are joined by the Pobuški 600 meters downstream. Coming from the slightly recessed and fragmented central 163 section of the mountain slope, the latter stream is larger than its tributaries on the east. The ridges that separate the streams are rather small, with gentle sides, while those that flank their threepartite basin are somewhat steeper and higher. Counting from the west these are the third and the sixth ridge in the region, spreading below the villages Pobužje and Ljubanci. As on the first ridge, the asphalt roads leading to Skopje follow their summits. The distance between neighboring streams is about 500 meters, decreasing towards their confluence. At its widest this triangular basin measures 2700 meters. 1200 meters south of the point of convergence of the three streams, a fourth enters the basin. Together they form a larger, more articulated valley floor. Before it enters the Skopje Basin, this newly formed Radiška River measures over 200 meters across the valley floor. It is covered with Quaternary sediments similar to those covering the Inner Skopje Basin. The area drained by the easternmost, fourth stream is almost as large as the area drained by its three tributaries on the north. It flows from the northeast, springing from the peak Pupljak at 1628 meters above sea level, the northeast corner of the region of Skopian Montenegro. Like most other streams entering the foothills it carries the name of the village built by its side, the Ljubotenski Stream. It doesn’t stand apart by its size or length, but the ridges that flank this stream are several times wider and nearly 100 meters higher than their neighbours to the west. As one might expect their relief is far more developed, particularly on the sides drained by the easternmost of the streams, the Ljubotenski. The outer slopes are steeper and less fragmented. In general the eastern part of this region is far more rugged and higher than the western. The mountain slope is also steeper, particularly above the southeast corner of the area. Two kilometers from the villages Ljuboten and Ljubanci into the mountain, the ground rises for over 1000 meters. Consequently this corner of the foothills is slightly more isolated from the main roads leading across the mountain. This is not the case with the upper part of the basin. The area drained by the Pobuški, the Turčevski and the Ljubanski Stream is lower, offering an easier access to the mountain interior. It is therefore more similar to the basin in the western half of the foothills. These two groups of streams don’t drain the entire area of the foothills. They are separated by another basin, shallow and dry on the surface in its upper part. It is the only basin fully belonging to the foothills. It begins as a shallow ravine immediately under the road between the villages Kučevište and Pobužje and spreads southwards, gradually descending into the plain of Skopje. Water flows on the surface in its lower half and only during the rainy periods. The stream is too weak to cut a fixed bed and the water dissipates over the surface or sinks into the ground, at places creating small swamps. Hence in the lower course, it bears the name Bara or swamp. It is 30 to 50 meters wide at the bottom, while the width of its entire drainage is approximately 1300 meters. Along its course leads another, recently built asphalt road establishing a direct link between Kučevište and Skopje. This small valley plunges like a wedge between the basins draining the eastern and western half of the plain at the foot of Mt. Montenegro. It is flanked by the second and the third ridge; the former running southwards, parallel to the course of the Kučeviški, the latter southeast, following the course of the Pobuški Stream. They are similar to the other ridges that constitute the area of the foothills, measuring only 100 meters across their summits and running at heights between 460 and 540 meters above sea level. Their sides are gentle, except for the eastern slope of the third ridge drained by the Pobuški Stream. Interestingly both ridges spring from the centre of the line that defines the mountain foot. They branch out from a higher ridge that once separated the villages Kučevište and Pobužje. Today the two settlements are merged, but the old 164 settlement cores are positioned on the opposite sides of the foot of this ridge, Kučevište on the western, Pobužje on the eastern side. The small, centrally positioned valley was uninhabited. Similarly to the neighboring village Radišani, a minor settlement developed only recently on the southern periphery of the area, where the valley enters the Skopje Basin. Despite the proximity to the older villages near the mountain foot and despite the fact that both settlements were mostly made up of the former inhabitants of these villages, they are considered a part of the Skopje Basin, not of Montenegro. IV.3 Geomorphology and geology Looking more closely at the topography in the foothills, one observes that it closely follows the relief of the mountain. The arch shape of the main mountain ridge is repeated along the mountain foot and less clearly, by a series of flattened surfaces appearing at roughly equal heights on the summits of the low ridges. These discontinuous “terraces” are better preserved in the eastern half of the area. Two appear at the heights of 580 and 540 meters and a third one, running at a height of 470 meters above sea level can be followed from west to east, across the entire plain. When connected these low summits form arches, roughly concentric to the one along the line of the mountain foot. Recall that the relief of the mountainside was similarly structured: it consists of a concentric series of flattened summits positioned at similar heights. These are the remnants of the terraces of the Central Balkan Lake that filled most basins along the Vardar Valley after the last regression of the sea, during the Upper Oligocene344. Finer dating of the formation and the draining of this lake is still lacking, chiefly because most of the fossils found are from endemic species. Nevertheless it was clear even to the first researchers of the Skopje Basin that it existed by the beginning of the Miocene and on certain lower sections of the Basin, it survived into the Holocene Era. This lake is the prime factor in the shape and the structure of the geography of the entire Skopje Basin, including this section of Mount Montenegro and its foothills. The wider study region was basically a small outlet of the Skopje Lake and it was among the first portions of land that emerged after it began to retreat 345. Hence the faintly preserved structure of the local relief: the series of concentrically positioned arches mark the former levels of the Skopje Lake. The fact that they are relatively closely spaced suggests that the lake was unstable and that it drained away rather quickly. A more stable phase was reached, when the water level stayed at 600 and 620 meters above the sea, coinciding with the eastern half of the mountain foot. In fact geomorphologists were able to distinguish a relatively well preserved section of this lake terrace, running for hundreds of meters between the villages Ljuboten and Raštak, in the eastern end of the region 346. The next more stable phase was achieved when the lake surface retreated to 390-400 meters above the sea. As mentioned earlier in the text, this is the line that separates the foothills of Mt. Montenegro from the plain of Skopje. The reason behind this poor preservation of the old abrasive relief is the powerful erosive forces of the Vardar347. This is the second most important factor in the geomorphology of the basin and one that still operates today. These are the same erosive forces that have contributed to the shape of the first study area. In the case of the Skopje Basin and particularly on the southern J. Cvijić, 91-118, 1906; M. Luković, Tektonika i geološka istorija skopskog bazena, 4-55, Glasnik Skopskoga Naučnoga Društva X, 1931. 345 P.S. Jovanović, 81-84, 1931. 346 P.S. Jovanović, photo 7, 1931. 347 P.S. Jovanović, 99, 1931. 344 165 slopes of mount Montenegro, the old lacustrine terraces were heavily disintegrated and dissected by a great number of small valleys. These streams were formed after the retreat of the lake and like the Vardar and its major tributaries the Lepenec and the Pčinja, originally they flowed directly into the Skopje Lake. When the lake retreated to the southeast corner of the Skopje Basin, a new erosion base was carved by the Vardar. In effect the lateral streams extended their courses in the areas of the headwaters, incising deeper into the relief and creating new tributaries348. At the same time, their courses were diverted towards the newly carved erosion base and began to flow directly into the Vardar. The streams that come down the southern slopes of Mount Montenegro are typical examples of this process. Mount Montenegro along with the Skopje Basin is a part of a larger geo-tectonic unit called the Vardar Zone349. The first survey area was also a part of this zone. Its geology is particularly complex and diverse, often featuring discordant layers and signs of extreme disturbances, with older formations mounting over younger rocks. It is still one of the tectonically most active regions in the Balkan interior. The present-day geography of the studied area is also partly related to its geologic substrate. In its southwest corner, Mount Montenegro is predominantly made up of Paleozoic metamorphic rocks brought to the surface through tectonic uplift350 (map IV_5). These mostly consist of different types of slates and schist, amphibolites, chlorites and muscovites, with insertions of marbles, quartzite and gneiss. The later are exploited on two locations along the mountain foot, between the villages Banjani and Kučevište and in the southwest corner of the region near the village Brazda. The transgression of the sea during the Early Mesozoic left behind an extensive area covered with a mighty layer of flysch. It spreads in a 10 kilometers wide belt across the entire mountain range: from the villages Kučevište and Pobužje in the southern foothills to the basin of the Binačka, at the opposite northern foot of the mountain. It thus occupies the northern, central part of the region of Skopian Montenegro, flanked by a series of older, metamorphic rocks on both sides. The flysch is a softer sediment and is more prone to lacustrine and fluvial erosion than the Paleozoic rocks that surround it. As a result this central section of the mountain slope is slightly recessed and gentler than the flanking sections. Consequently it features more developed relief and hydrology. During the next episodes of sea transgression and during the lacustrine phase, water will follow the line of the least resistance penetrating deep into the mountain massif through the area covered with flysch. Little is preserved of the sediments formed during the Oligocene transgression. All that was left of this layer were a number of small and isolated patches along the northern section of the foothills, stretching between the villages Banjani and Ljubanci. They consist of sandstone, clay, conglomerates and limestone. The next phase in the geologic history of the area was the period of the Skopje Lake. Most of the area in the foothills is still covered with thick lacustrine sediments, reaching several hundreds of meters in depth. They fill the entire plain, stretching approximately to the contour line of 390-400 meters above sea level to the south. Thus the faint topographic border between the foothills of Mt. Montenegro and the Skopje Basin is further underlined by a change in the geologic substrate. This layer encircles the entire Skopje Basin, spreading along its perimeter. It roughly coincides with the transitional zone of lower mountain offshoots that separates the mountains from the plain. It has also survived on the upper portions of the isolated hillocks on the left higher bank of the Vardar. Although the lake penetrated deep This tendency was first observed and explained by J. Cvijić, 130-131, 1906. M. Luković, 4-15, 1931. 350 Osnovna Geološka Karta S.F.R. Jugoslavije, 1:100 000, sekcija K34-55, Uroševac, Belgrade 1984. 348 349 166 into the mountain masses, reaching the level of 900 meters above the sea, there are no remains of these sediments on the mountain slopes. Like the sediments left by the sea during the Oligocene, they have been completely washed away by erosion. The lake sediments in the foothills of Montenegro are mostly made of sandstone, clays and marls. Only in the eastern end of the foothills on a higher ground around the village Ljuboten, did they consist of sand and pebbles. The bulk of the agricultural land in the study region lies on this geologic base. It is an important factor in the soil formation process, greatly determining their type and fertility. The last and current phase in the formation of the Skopje Basin is dominated by fluvial erosion. Most of the interior surface of the Skopje Plain is made of sediments brought by the rivers during the last geological phase. These sediments are limited to the lower portions of the basin and only sporadically do they appear in the transitional and mountainous zones. In the case of our study area, Holocene sediments are found only at the floors of the small valleys that run across the foothills and at the foot of the steep ridges near the village Ljuboten. Even the bottoms of the smallest valleys are filled with sediments torn from the upper sections of the mountain slope. It appears that the work of the erosive forces doesn’t present a threat to the settlement houses or to the agricultural fields. An exception is the village Ljuboten, where occasional landslides have been reported. Nevertheless most of the villages in the area are positioned at least 10-20 meters above the streams. Interestingly the largest stretch of land covered with Quaternary sediments in the study area belongs to the mountainside. It is an almost continuous belt, 800-900 meters wide and over 9 kilometers long. It stretches in a northwest-southeast direction parallel to the river Lepenec, perfectly overlapping with the low pass, which we assumed was the ancient river bed. The character and the date of these sediments flanked by a series of old metamorphic rocks provide further support for this assumption. The wider study region has two principal characteristics; the first regards its position in the wider context of the Skopje Basin, the other the regular and symmetric character of its geography. This area is an extreme example of the so called transitional zone of the Skopje Basin351. It unites two contrasting landscapes: the relatively gentle foothills and an extensive and rugged mountainside. This is literally a place where the plain and the mountain meet. Although the edge of the foothills is seemingly clear, it is very difficult to decide whether the small plain at the foot of Mount Montenegro belongs to the plain or to the mountain mass. Partially enclosed and far removed from the Vardar Valley, it is only technically a part of the Skopje Basin, not more than any other portion of the mountain slope submerged by the Neogene Lake. But unlike other regions at the foot of Mt. Montenegro, the geography of the study area is not dominated by larger, isolated basins, like those at the eastern foot of the mountain. Instead a number of minor, mountain streams converge towards a wider plateau, where they further converge into two larger streams before entering the inner Skopje Basin. There is no single predominant stream with a well articulated drainage basin, linking the mountain interior with the plain. The continuity in the relief is too faint and in practice, insignificant. These peculiar natural arrangements are of a cardinal importance for the local economy and settlement pattern. In the past couple of centuries, the people that inhabited this area have sought to make optimal use of the two contrasting landscapes that comprise their land. In the wider region of the Vardar Valley, most commonly the areas of rural settlements coincide with the drainage basins of local streams. In the case of our study area this is almost impossible, because of the character of the local geography. On the mountain slopes, the basins 351 J. Trifunovski, 360-366, 1955. 167 of the local streams are too steep and narrow to receive and feed even the smallest of nucleated settlements. In the foothills the valleys of these streams are only slightly wider and the watershed lines that separate them are too vague. Most modern settlements are raised tens of meters above the valley floors, closer to the watershed lines. But although insignificant as geographic entities, these streams form a surprisingly regular web, dividing both the mountain slope and the foothills into a number of roughly equal drainage basins. This hydrographic constellation served as a basis for the present day settlement pattern in the region. Despite the fact that their articulation in the relief is either too extreme or too faint, these basins divide the region into a number of units and subunits. More importantly their courses run at almost equal intervals, consequently allowing for an equal access to different resources, for all settlements located in the foothills. The majority of the local streams drain approximately equal portions of the mountain slope and they enter the area of the foothills at surprisingly regular intervals. Nevertheless the streams of Mt. Montenegro offer little room for the development of human settlement. As physical topographic units they don’t provide sufficiently large surfaces. In the conditions posed by this particular geography, larger stretches of flat, inhabitable land are to be found only along the watershed lines, at the summits of the ridges that separate the streams. In a sense this is exactly the opposite from the case of the first survey area where the settlement niches were always within the limits of a certain drainage basin. In the second study area, another type of topographic element comes to the fore. This is the flattened or gently sloping mountain ridge; the remnants of the old terraces of the Skopje Lake. These topographic features are not only the most convenient and logical locations for human settlement, but they are also carrying the bulk of the productive agricultural land and present a basis for the local and interregional road network. Parallel to the longitudinal divisions of the region drawn by its hydrography, they draw vertical, latitudinal divisions creating concentrically arranged terraces. Although heavily disintegrated and broken up into separate ridges, they have played an important role in the organization of human settlement and agricultural production in the region. IV.4 Land-use, agricultural and mineral resources; vegetation, soils and climate The contrast in the geography between the two principal parts of the wider study region is naturally reflected in the modern land-use and the vegetation cover. Today the mountainside is almost completely abandoned. Apart from the weekend resorts and a number of monastic complexes, there is little activity on the mountain slope. Its forests are still visited for the purposes of tree cutting and hunting, but the bulk of the old productive surface, the pastures in particular, are long since abandoned. As a result most of the mountainside is covered with forests and tall grasses. The most commonly found tree species are the elm, the oak and the wild pear352. Like the rest of the Vardar Valley zone, Mount Montenegro is poor in mineral resources. At present there are stone quarries on two or three locations along the foot of the mountain and in the western half of the region. These are the quarries mentioned earlier, located between Banjani and Kučevište and the one southwest of Brazda, in the southwest corner of the region. On both locations quartzite, marbles and limestone are exploited. In addition ethnographers that worked in the area in the early decades of the last century mention extraction of iron ore on locations deeper into the mountain interior, but without pointing to specific locations353. 352 J. Trifunovski, 15-16, 1971. S. Tomić, 463, 1905; N. Čausidis, 185-186, 1989, in his survey of forts in the mountain interior also notes a number of toponyms indicating ore extraction and melting. 353 168 Mount Montenegro is also very rich with freshwater springs. All major streams that drain the southwest slopes of the mountain are fed by a number of powerful springs, usually positioned on the steep valley sides. In contrast there are only a few major springs in the area of the foothills. An equally important resource is the very fertile flysch sediment covering the entire central section of the mountainous part of the study area. These two basic resources have preconditioned the mountain’s riches in lush pastures. According to earlier ethnographic researchers, these mountain pastures were probably the principle resource for the entire region354. Until the great changes brought about during the Late Ottoman and Early Modern periods, the semi-nomadic pastoralism played an important role in the local economy. At that time the mountainside was not the mere background, but the very centre of production for the region of Skopian Montenegro. Only the current toponyms testify to the intensive exploitation of the mountainous zone in the past. Many of the ridges, at least their upper portions bear the name of the families that once grazed their flocks there. There are only vague remarks about the organization and the character of local herding. As in many other parts of the Balkan Peninsula, it involved seasonal movement of the flocks between summer and winter pastures355. In the case of Skopian Montenegro, at least until the beginning of the 19th century, entire families moved with their flocks and spent the summer in specially built summer camps356. For the winters they returned to the foothills where the flocks stayed in pens, near the villages. It is important to note that these movements were gradual and carefully timed to avoid destruction of the cultivated surfaces and make optimal use of the pastures. The terraced terrain of the mountain meant that the pastures were distributed at different heights, featuring optimal grazing conditions during different phases of the season. Over the course of the last couple of centuries, the herding economy almost completely disappeared from this region. The entire agricultural potential is now focused on the fields in the foothills. Here an area measuring approximately 30 square kilometers is almost completely under cultivation. Only the stretch along the line of the mountain foot, where modern settlements and villages are situated and the narrow valley floors are left uncultivated. The rest of the land in the foothills is divided into hundreds and hundreds of agricultural fields (photo IV_1). On average they measure around 2000 square meters. Larger fields are usually divided along the longitudinal axes. They mostly follow the natural configuration of the terrain, oriented perpendicularly to the sloping sides, though on some slopes there are groups of fields stretching across the contour lines. Many fields are supported with artificial terraces and are further delimited by tall hedges. For the greater part, the terrain is relatively gentle and soil erosion shouldn’t be a pressing concern for the local farmers, though the lack of vegetation and the exhausted soils could potentially incite erosion. It seems that overexploitation and the lack of fresh arable land present a greater problem. In sharp contrast to the first survey area, barely 1 in 10 fields is left fallow or completely abandoned. Literally every available piece of arable land is turned into an agricultural parcel. The substantial efforts of delimiting the agricultural fields with terraces or hedges are most likely signs of demographic pressure and limited resources. The cultures grown in this area are typical for the transitional habitation zone of the Skopje Basin. In principle, “dry” cultures are grown on fields positioned on the summits and the upper portions of the ridges. These are the traditional cereals, wheat, barley, rye and millet and S. Tomić, 460-461, 1905. Following a number of papers presented at a symposium entitled “Prirodni i socio-geografski karakteristiki na zonite na nomadsko i polu-nomadsko stočarstvo vo Makedonija”, Skopje 1981. 356 S. Tomić, 429-433, 1905. 354 355 169 the vine. “Wet” cultures, mainly maize, vegetables and fruits are cultivated on fields on the lower sections of the ridges and on the valley floors357. During the survey however, it was noticed that this general rule wasn’t always followed and fields with maize or onions were seen on the dry upper portions of the ridges. Closer to the mountain, orchards with walnut and chestnut trees were occasionally spotted. The distribution of the predominant farming cultures is chiefly patterned by the access to irrigation water. In this region it is either channeled from the local streams or it is tapped from underneath the surface. Written documents dating to the middle of the 14th century reveal very elaborate rules regarding the usage of irrigation water from local streams358. Each village was given the right to irrigate its fields only on certain days of the week. In sharp contrast to the mountainside this part of the micro-region appears rather barren and dry. This is mainly a result of the complete transformation of the area into agricultural land. The local climate is similar to the rest of the Skopje Basin. Located in the upper part of the Vardar Valley, over 200 kilometers north of the Bay of Thessalonica, the Skopje Basin feels only faint echoes of the Mediterranean climate. Compared to the lower parts of the valley, the region of Skopje is cooler and more humid, in particular the area at the foot of Mount Montenegro359. Access to water is not the sole factor in the organization of local agricultural production. Equally important are the local soil types. In this aspect the region of Montenegro has been studied as a part of the Skopje Basin360. There are no pedological studies focused specifically on this area, which is a pity because the variation in soil types on a relatively small piece of territory is surprising. The local farmers have developed their own nomenclature for the soils in the area, each suited for different cultures361. A detailed soil map of the region would open a very interesting and unique insight into the agricultural knowledge and production in this region. An early study of the soils in the lower parts of the Skopje Basin has identified four principle soil types, three of which are found in our study region362. The predominant soil type was derived from the lacustrine sediments. They are called smonlnicas (or vertisoils) and are distinguished by their intensive black color. This is the oldest and originally, the most widespread type of soil. The other two types, called crvenica or red soil and ganjača or colluvial soilsare derived from the smolnicas. Namely the former are formed on the substrate of eroded smolnicas and in the surrounding landscape, crvenicas are normally found along steeper section and above the smolnicas. Ganjačas on the other hand are either derived from eroded material or on alluvial surfaces, protected from further flooding. Clearly in the studied region ganjačas on eroded material are more common. They are mostly found at the foot of the ridges and on their lower slopes. Ideally the distribution pattern of the various soil types should consist of ganjačas at the foot of the ridges; smolnicas on the gentler slopes or on the flattened tops and the least represented crvenicas, usually following the ganjačas on the upper portions of the slopes. But whether this distribution corresponds to reality and how it affects the local agricultural production remains unclear. It is evident from historical and ethnographic sources that the region 357 J. Trifunovski, 360, 1955. R. Grujić, Gusari na Svetoj Gori i Hilandarski pirg Hrusija od XII-XIV veka, 1-32, Glasnik Skopskoga Naučnoga Društva XIV, 1934. 359 P. Vujević, Klima skopskog bazena, 125-240, Glasnik Skopskoga Naučnoga Društva X 1931. 360 D.B. Todorović, Pedološka ispitivanja u skopskog bazena, 242-268, Glasnik Skopskoga Naučnoga Društva X 1931. 361 S. Tomić, 454, 1905. 362 D.B. Todorović, 250-265, 1931; 358 170 of Skopian Montenegro is indeed rich with crvenicas or red soils. One of the earliest researchers of this region remarks that most villages were situated close to, or on this very type of soils 363. Perhaps this fact is hidden behind the original name of the area, in translation the Red Mountain. Only in later times did it become Montenegro, because of the similarity of the words for black and red in the Old Slavonic language. In general the soils in the area exhibit various macroscopic qualities, sometimes changing from field to field. Most have fine structure and can retain water for longer periods of time. Their colours can vary from nearly white to dark brown and black. On certain locations and especially on the top of the steeper ridges, the upper horizons have been washed away by erosion and the soil layer is stony and drier. IV.5 Modern day settlement pattern, settlement locations and settlement territory Today the parish of Skopian Montenegro consists of 11 villages. Most of them are medium to large size, but there are a few very small ones, which have most likely been formed by families and clans who moved out from some of the larger neighbouring villages 364. Such are the cases of Brodec, a former summer camp of the neighbouring Kučevište and of Gornjani, formed by a dozen families who have moved away from the neighbouring Banjani. The two had less than 20 families and presently, Brodec consists only of weekend houses, with no permanent residents. On the other end of the scale are the villages Mirkovci and Kučevište. Mirkovci had over 100 houses in the early part of the 20th century, while Kučevište, the largest settlement in the region, nearly 200 houses365. The rest of the villages have between 50 and 100 families. The total number of inhabitants in the area, at the beginning of the 20th century, was over 8000, making it one of the most densely inhabited rural areas in the country (tab. 1). TableIV_1: List of the villages in Skopian Montenegro and the number of houses and inhabitants at the end of the 19th century (after S. Tomić, 1905, 507-508) Village name Mirkovci Brazda Brodec Ljubanci Gornjani Pobužje Kučevište Čučer Banjani Gluvo Ljuboten Total Number of houses 114 50 28 120 24 72 193 76 79 60 94 910 S. Tomić, 411, 1905. J. Trifunovski, 59-62, 1971. 365 S. Tomić, 508, 1905. 363 364 171 Number of inhabitants 1026 (num. of houses x9) 450 252 1080 216 576 1737 684 711 540 846 8190 Over the past several decades these settlements have greatly changed and this has also had effects on the surrounding landscape. In its basic principles however, the settlement pattern established at least a couple of centuries ago has remained unchanged. 9 of the 11 villages are situated exactly along the line that marks the mountain foot, encircling the entire area of the foothills from west to east. Closely following the local hydrology, they are distributed into pairs and groups at roughly equal intervals. The distance between these groups is slightly over 2 kilometers, while the distance between individual villages is 1 kilometer or less. 6 of these villages are grouped into pairs, while the other 3 form a tight group in the northwest corner of the foothills. From west to east, these are: Brazda and Gluvo, Čučer, Gornjani and Banjani, Kučevište and Pobužje and the easternmost pair, Ljubanci and Ljuboten. Almost all these villages are regularly positioned at points where the mountain streams enter the foothills; basically by the very entrances to the small canyons carved by the mountain streams. The exceptions are Brazda and Gluvo, the pair in the lower southwest corner of the area, where there are no major streams coming from the low mountain shoulder. But even in these cases, the villages are positioned at the mouths of minor ravines that drain the hills above. In addition, Brazda occupies a location physically very similar to the recesses occupied by the rest of the villages. It is positioned on the southwest exit of the area, where the main stream leaves the foothills, surrounded by the hills Gradište and Vražanski Rid. Of the villages lying along the mountain foot, only Gluvo is more exposed and unprotected on its sides. As mentioned in the introductory part, the settlement pattern in this micro-region features a remarkable regularity and evenness. Despite the varying size of these villages, they all have an equal access to the two principal groups of resources in the area: the pastures and the forests in the mountainside and the arable land in the foothills. In the hilly lands of the central part of the Balkan Peninsula, it is possible to distinguish between two very general types of geographical regions as bases for settlement patterns. These are the narrow valley floors and the basins of the old Central Balkan lakes. The second study area seems to be a representative of the latter. Here the younger, fluvial relief forms are too vaguely articulated in the relief to become territories of separate settlements,although the local hydrography is still very important for the settlements’ micro locations. Instead the principle agricultural resources are either concentrated on a larger plain, with only faint inner topographic divisions or are dispersed into numerous tiny islands of flat land through an extensive mountainside. This means that a larger nucleated settlement can only develop in the foothills, large enough to accommodate several settlements of similar size or one or two larger settlements, accompanied by minor satellites. In the case of Skopian Montenegro’s current settlement pattern, it is difficult to point to a higher rank settlement. True, 2 of the 11 villages are considerably larger than the rest and another two are considerably smaller, but with the exception of one of the latter, all have equal positioning in the landscape. Understandably this doesn’t mean that all villages were of equal social and economic standing. It is even possible that the larger villages occupying the central section of the region have access to a more extensive mountainside. In principal however, the territories of almost all of the current villages extend over the mountainside and the foothills. This was most likely seen as the optimal arrangement for the needs of the local mixed economy of farmers and herders. This settlement pattern partly rests on the fluvial relief forms, despite their poor articulation and insignificance in terms of size. Not only because all settlements in the area are situated by one of the local streams. We saw that their valleys offer very little space. These streams are rather more important dividing both the foothills and the mountain slope into series of smaller units and subunits discussed earlier in the text. In the mountainside, where the 172 resources are scattered they diverge; in the foothills, where resources are concentrated they converge. This characteristic of the local geography is reflected in the peculiar pairing of villages and their unusually close spacing in general. Ideally following the watershed lines, each of the settlements would have a territory triangular in shape, with the top pointed towards the centre of the foothills (map IV_6). In reality however the little evidence that we have speaks against strict divisions of the land between settlements366. Even in the mountainside where there are physically separate basins, they were divided into larger sections simultaneously or alternately used by a pair or group of villages. In the foothills these physical divisions are so minimized that it is impossible to divide its territory into separate basins, each belonging to a separate settlement. Instead the area of the foothills is divided into a number of larger basins, drained by one or more streams and occupied by a pair of settlements. All of this indicates that the settlements in this region form a strongly integrated union. The geography of their region and perhaps their economy and social organization has forced them into sharing a larger stretch of compact land. This tendency is more pronounced in the lower western part of the foothills, where five settlements basically share the same basin. The regularity of the local settlement pattern is broken by the location of the second largest settlement in the area, Mirkovci. Unlike the rest of the nine villages situated in the foothills, Mirkovci is positioned 1.5 kilometers south of the line marking the mountain foot, at the end of the ridge that separates the basins of the Kučeviški and the Banjanski Stream. Its houses are distributed in two groups, with the centre of the settlement positioned almost exactly at 500 meters above the sea. Mirkovci is the most centrally positioned settlement in this region and the only natural crossroad. Here the main road coming from Skopje splits into a branch leading north, to Kučevište and another, towards the trio of villages in the northwest corner of the foothills. Surprisingly enough however, in the past couple of centuries Mirkovci never really achieved a status higher than its neighbors. Moreover after World War II a new regional seat was established on a neutral location, only half a kilometer to the southwest of Mirkovci367. It is the seat of the municipal government and the main crossroad for the western part of the region. In addition a separate asphalt road was built 1.7 kilometers to the east, establishing a direct link between Kučevište and Skopje. At the other end of Mirkovci’s ridge near the mountain foot, there is at present a stone quarry. It was only opened in the 1960’s and thus, cannot be responsible for this disruption of the local settlement pattern. Mirkovci was occupying its present-day location by the early 19th century, at the latest. It seems however that neighbouring Banjani changed its location368. In the early part of the last century its inhabitants remembered that the “old village” was situated a kilometer to the east, very close to the modern quarry. If we look at the western part of the foothills as a separate basin, than the location of Mirkovci doesn’t defy the logic behind the settlement pattern in the area (map IV_4). Positioned on the southeast corner of the small basin of the Banjanski Stream, it actually complements the other groups of settlements that occupy the northwest and southwest corners of the basin. In the eastern part of the foothills, the string of settlements along the mountain foot is again broken between Pobužje and Ljubanci. Mid-way between these two villages, Turčevski S. Tomić, 429-430, 1905; on the basis of local oral traditions, the author also talks about earlier (preOttoman?) land divisions. Far more direct and reliable testimonies are to be found in the 14 th century imperial edicts, translated and analyzed in V. Mošin et al. 1975. 367 J. Trifunovski, 33, 1971. 368 S. Tomić, 423, 1905. 366 173 Stream enters the foothills. It is about the same size as the neighboring streams, yet the recess at its entrance in the foothills is uninhabited. However there is a mention of a settlement called Turčev Dol in the Ottoman census of 1467-1468369. It was initially decided to survey this empty settlement niche, but unfortunately we found a large portion of this locality occupied by modern villas, surrounded by enclosed yards. These were mostly built by former inhabitants of Pobužje, who had moved to Skopje in the past few decades. The recent phase of settlement displacement in the area has consumed most of the land along the mountain foot. As a result, it is impossible to survey the zone in which most settlements in the area were traditionally located. The other village that defies the distribution of settlements in the region is situated almost 3 kilometers to the north-northeast of Pobužje, by one of the tributaries of the Pobuški Stream. At present it doesn’t have permanent inhabitants; the old village houses were abandoned or transformed into weekend resorts. This is Brodec, the only remaining establishment situated in the mountainside. As mentioned before, this village was a former summer camp of Kučevište, although it is closer to Pobužje than to its mother settlement. Its case warns against assumptions that the settlement territory simply followed the natural limits of the closest topographic entity. Brodec doesn’t belong to the same drainage basin as its mother settlement, Kučevište. If we analyze the character of these settlements’ micro-locations, we’d notice that 10 of the 11 villages are situated on the foot or the lower sections of the ridges that surround the area of the foothills. Only Mirkovci is situated on the summit of the ridge, on the watershed line. The heights of these surfaces closely correspond to the levels of the receding Skopje Lake. The lowest pair of villages is Brazda and Gluvo at 390 meters above the sea, in the southwest corner of the area. Recall that this level marked the border between Skopian Montenegro and the inner part of the Skopje Basin. Čučer, Banjani and Mirkovci are positioned just above the contour line of 500 meters above sea level. Finally, the villages in the central and eastern parts of the foothills, Kučevište, Pobužje, Ljubanci and Ljuboten and the small village Gornjani in the northwest corner of the region are all situated at a height of about 600 meters above sea level. We saw that this line marks the most stable phase of the Skopje Lake and its terrace is consequently wider and better preserved370. Even the small outpost Brodec, situated 3 kilometers into the mountainside is positioned at a height corresponding to one of the lacustrine terraces, perhaps the highest one located at nearly 900 meters above the sea. Thus although barely visible in the present-day relief, these old terraces are the chief substrate on which the settlement pattern in the area is based. Broken up into numerous low ridges, the remnants of these terraces represent potential settlement locations. In the topography of the present, they can appear either at the foot or on the top of the mountain ridges. We saw that by far the most favorite type of locations for human settlement in this area was the mountain foot, where the small streams exit the mountainside. The example of Mirkovci however, shows that there are other types of inhabitable locations, primarily the flattened summits of the low ridges in the foothills. Moreover some of these locations bear the names of abandoned villages, mentioned in the ethnographic studies and in the early Ottoman censuses. These locations are far more spacious and closer to the arable surfaces. But at the same time they are more exposed, consume larger portions of the productive surfaces and are less conveniently positioned regarding the seasonal movements of flocks, crucial for the pastoralist economy that sustained this region in the past. 369 370 M. Sokolovski, 478, 1971. P.S. Jovanović, 82,1930. 174 This settlement pattern appears incredibly stable and highly integrated. It mostly consists of medium to large size settlements with a long history, reaching back to the Late Byzantine Period. A more radical displacement of one of these settlements could hardly happen without causing shifts in the location and related changes for the rest of the settlements in the area. One has to remember however that the stability and the optimality of this scheme maybe originated only in the past two or three centuries. Although a number of the currently existing villages are mentioned in the written sources dating to as early as the 13th and the 14th centuries, it is far from certain that they occupied the same locations or if they were of the same size and character. In these same documents there are also mentions of other settlements, the locations of which are completely forgotten. This fact along with the modern examples of alternative settlement locations indicates that the modern settlement pattern is but a stage of a long and probably very complex history of human habitation. In fact over the past few decades the local settlement pattern has experienced relatively radical changes. Looking at an updated map or a satellite image of the area, one will hardly recognize the location of the traditional settlement cores. There is another element that has survived in the present-day landscape and deserves a separate mention. These are the well known seven larger monasteries and the numerous smaller chapels or isolated crosses, situated at the perimeters of the villages’ inner territories 371. To these we may also add a number of small groves, rare islands of uncultivated land usually reserved for the flocks and with restricted access372. Naturally only the names of the larger establishment are preserved, the oldest dating to at least the end of the 13th and the first decades of the 14th century.The locations and the ancient origins of some of these monasteries and churches suggest that the basis for the settlement pattern that has survived to this day could be at least seven centuries old. Not merely because some of the central village churches date to the Late Middle Ages. The distribution pattern of these monasteries fits into the distribution pattern of settlements strikingly well. This becomes much more evident if we look at the locations of the numerous minor chapels, sacred crosses and groves. They form a fairly regular web, consisting of triangles and polygons surrounding the settlements. Only some of the major monasteries, St. Elijah near Banjane, the Holy Archangels near Kučevište and the Holy Virgin near Pobužje are located at a slightly greater distance, in any case not further than 3 kilometers from the nearest settlement. The rest of the sacred locations and monasteries are usually within a radius of 1 or 2 kilometers from the village centre. Good examples are the villages Kučevište and Mirkovci, the largest two in the region. They are literally surrounded on all sides by small chapels, sacred locations and even smaller monastic convents, such as the female convent of St Elijah, south of Mirkovci and again, a small female convent dedicated to St. Paraskeva, south of Kučevište (map IV_7). These humble establishments date to the Late Ottoman or the Early Modern Periods, but were allegedly built on earlier foundations or on ancient sacred land. Similarly the narrower territory of the smaller village Pobužje is marked by a small grove and an isolated cross, positioned on the summits of the ridges and less than a kilometer south of the village. These features are obviously closely related to the nearby settlement and it’s very likely that they are indicative of a settlement’s status and wealth. 371 J. Trifunovski, Manastirski konaci u Skopskoj Crnoj Gori, 113-124 Glasnik Etnografskog Instituta XXIX 1980. 372 S. Tomić, 431-432, 1905; J. Trifunovski, Les arbres culturelles dans la region de Skopska Crna Gora, 129-137, Bulletin de l’institut ethnographique de l’Academie Serbe XXIV, 1975; for an analogous role of trees in the ancient world, S.E. Alcock, R. Osborne, eds. Placing the Gods. Sanctuaries and sacred space in ancient Greece, Oxford 1994; J. McInerney, Sacred land and the margins of the community, 33-59; eds. R. Rosen, I. Sluiter, City, countryside, and the spatial organization of value in Classical Antiquity, Leiden-Boston, 2006. 175 If we analyze the topography of these sacred locations it becomes evident that they occupy alternative or minor settlement locations. The four larger monasteries north of the villages Banjane, Kučevište, Pobužje and Ljubanci are all situated at the ends of the narrow canyons, deep into the mountainside. Their locations resemble the locations of the villages situated in the foothills: both occupy the banks of the mountain streams, where the valleys widen, entering a lower section of their courses. Thus the villages were positioned at the points where the streams leave the mountainside and enter the foothills, while the monasteries were further upstream, where the streams cross from the upper to the lower sections of the mountainside, roughly at 650 meters above sea level. Simply put they occupy the opposite ends of the same narrow canyons; the locations of the monasteries being a smaller version of the same type of settlement loci occupied by the modern villages. In both cases the installations are positioned on a theatrically shaped terrain, a real physical niche. Indeed one of the monasteries in the region, the one dedicated to the Holy Warrior St. Nikita occupies an empty settlement locus in the northwest corner of the foothills. Like all other settlements in the region, it is situated at the mountain foot, where a small nameless stream enters the foothills, forming a recess identical to those occupied by the rest of the villages. Only much later when even the memory of the past wealth and glory of this monastery completely faded did a couple of clans move from the village Banjani to the left bank of the nameless stream, opposite the monastery. Although far from the foot of the mountain, the female convent of St. Elijah near Mirkovci occupies an analogous location. It was founded on the right bank of the Kučeviški Stream, where the small river enters the low southwest corner of the area. If we turn specifically to the locations of minor complexes, isolated chapels and crosses, we notice that although bound to a different type of topographic features, most of them are nevertheless situated on the alternative type of settlement loci, i.e. the summits of the low ridges that dissect the foothills. Their locations are analogous to the location of Mirkovci, at the ends of the ridges’ gentler sections. Thus again they occupy transitional points in the micro-relief, where the steadily declining terrain is interrupted by sharper escarps. Examples are the chapel of St. Athanasius, 700 meters south of Kučevište and the chapel of the Holy Warrior St. Mercurius, a kilometer to the north of Mirkovci. In both cases the chapels are situated on the same ridge as the villages. Furthermore a great number of these churches or isolated crosses, regardless of the type of their micro-locations are on the same terraces as the majority of local settlements. In other words they too are possibly located on the remains of old lacustrine terraces. The chapels of St. Mercurius and St. Paraskeva are for example on the contour line of 500 meters above sea level, the monastery of St. Nikita at 600 meters above sea level, the monastery of the Holy Virgin north of Gornjane, at 900 meters above sea level and so forth. So far these observations can only be transformed into a tentative hypothesis. In the first study region, we saw that the corners of the wider settlement territory were clearly marked either by forts or small chapels and monasteries. Are we also dealing with the same sort of frontier landmarks in the second study area, with the important difference that here, most of these landmarks are still active and not merely the remnants of some past landscape? The described distribution of monastic churches and chapels at the foot of Mt. Montenegro could also be closely related to the distribution of settlements, at least during the last big phase of the history of human habitation in the region. In this case the sanctuaries are marking the settlements’ “inner” territories, their immediate surroundings or the territory within roughly one kilometer from the settlements’ cores. If we assume that the sanctuaries postdated the settlements, it would imply that by the time of their foundation, the founder settlements must have been firmly established 176 and sufficiently wealthy. But as will be shown later in the text, nothing justifies such an assumption. In some cases at least, the sanctuary predated the nearby settlement or the settlement moved, while the sanctuaries continued to exist or were renewed by another settlement. In any case, in this study region it is obvious that there is an important relation between settlements and sacred locations. It is very probable that in the conditions of faint topographic divisions of the terrain, high population density and limited resources, sacred locations were a key element in the local landscape. They marked the limits of the settlements’ immediate surroundings, justifying and reinforcing the drawn divisions and at the same time providing neutral meeting points, where people from all the surrounding villages would gather on the day of the patron saint. In addition the larger monasteries situated on the narrow valley floors, deep into the mountain interior could represent places of potential refugia for the inhabitants of the villages in the foothills. It was emphasized that the natural niches they occupy are but minor, sheltered version of the niches of the present day settlements. Limited to a territory of one square kilometer, the principle target of the survey was one of the empty or potential settlement loci in the foothills. In the mountainside it is impossible to conduct a survey of such a scale and intensity as in the first study area. Here the largest compact piece of accessible terrain doesn’t exceed 20 or 30000 square meters. These locations are now mostly covered with dense and tall vegetation or are under forest. But choice is limited even in the foothills. The half a kilometer wide belt stretching along the entire perimeter along the foot of the mountain is almost completely consumed by modern housing, infilling the space between the old settlement cores. In addition we wanted to avoid surveying parts of the immediate surroundings of the modern settlements and the large quantities of recently deposited material usually covering the nearest fields. Therefore our focus shifted towards the central parts of the foothills, roughly in the area between the second and the sixth ridge, the points of Krst and Orlovec on the map (map IV_4). This is the geometric centre of the foothills, lying at an almost equal distance from Pobužje, Kučevište and Mirkovci. The southern half of this area, towards the inner Skopje Basin features worse visibility conditions and there is hardly any location that looks like the typical settlement foci in this region. Much more promising were the flattened summits and the gentle sides of the ridges in the central part of this plain. These locations are basically identical to the location of present-day Mirkovci. We wanted to survey at least two of these ridges, with a particular focus on the 500 meters contour line, the height on which most of the villages in the western part of the foothills are situated. There were two principle aims in sight: to check if there are traces of a more dispersed settlement pattern in the past and to try and date the beginnings of the latest settlement pattern in this region, on the basis of the offsite finds. But ultimately we wanted to determine the quantity, the character and the distribution of surface finds in a landscape different from the one in the first survey area and situated in a different geopolitical context. IV.6 The survey area, limits and description One of the more difficult problems of this field survey and in principle of any archaeological field research is drawing the spatial limits of the survey area. In the first survey this appeared as a lesser problem because here the survey area roughly coincides with a given geographic entity, the valley of Sopot. But even in the Sopot survey, we couldn’t evade the question of what the survey area actually represents. Weren’t we simply following the limits 177 imposed by the currently existing settlement pattern and land use, both in our reading of the physical geography and in interpreting the survey results? Basically this is an ever-present problem: even when completely surveying what appears as a perfectly enclosed geographic entity, one has made the assumption that it represents an integral settlement area, rather than just a smaller fragment of a larger unit. Or alternatively, that the limits of the assumed settlement areas follow the limits of hydrographic rather than of orographic entities. In the case of the second survey the problem becomes far more acute and obvious. We have chosen to focus the survey on a minor part of a larger geographic and administrative unit, with a surprisingly high level of integration. The foothills of Mt. Montenegro measure over 30 square kilometers, while the scope of the survey was limited to approximately 1 square kilometer. Moreover unlike the valley of Sopot where the survey limits were largely determined by visibility conditions and access, the foothills of Mt. Montenegro offer equal ground conditions over a relatively large territory. Even after deciding to narrow the choice to the central part of the foothills, the area stretching between the second and the sixth ridge, we were still left with a territory of over 7 square kilometers, featuring equal ground conditions and very faint inner topographic divisions. In a way we tried to follow the approach applied in deciding on the limits of the first survey area; that is to identify the types of topographic units occupied by current or known past settlements and limit the survey to one of these actual or potential settlement loci and their immediate surroundings373. But while in the first region it was easier to identify wider settlement chambers rather than settlement loci, in the second study region, the opposite was the case. Here we immediately observed two basic types of settlement loci: the recesses created by the local streams, usually at the points where they exit the narrow canyons at the foot of the surrounding ridges and the flattened summits of the ridges, along the watershed lines. These types of locations are too small and it was necessary to expand the survey area, including more than one potential settlement micro-location and/or, finding elements in the landscape that mark the narrower or wider territories of local settlements. As already suggested during the description of the geography and the modern settlement pattern in the wider study region, in the absence of pronounced natural landmarks or inner topographic divisions, the inner territories of local settlements could have been delimited by the erection of sacral monuments. This secondary element in the studied landscape at least helped us determine the northern boundary of the survey area, the one facing the currently existing settlements and thus avoid surveying their immediate surroundings. The other criterion taken into account was the character of the micro-topography, the spaciousness and the compactness of the topographic units. Of the five ridges running across the vacant centre of the foothills, the most spacious and gentlest was the terrain between the second and the third ridge, counting from west to east. In between these two ridges, the smallest central basin of the foothills gradually emerges. Its shape becomes more articulate 1.5 kilometers south of the foot of the ridge that separates the villages Kučevište and Pobužje. Acknowledging the fact that the majority of the settlement loci in the area appear at specific altitudes, we decided to focus on the upper section of this small valley, on the area between the contour lines of 450 and 540 meters above sea level. We saw that a number of villages and sacred locations were located at this altitude. Guided by these broad criteria, the limits of the surveyed area run as follows: to the north, the boundary line was drawn between the small female convent of St. Paraskeva on the west and a small grove, roughly coinciding with point 540 on the map (map IV_8). The western and 373 E. Neustupný, , 45-61, ed. E. Neustupný, 1998; J. Bintliff, 505-545, ed. G. Barker, 1999. 178 eastern boundaries were the summits of the second and the third ridge; finally the vaguest, southern boundary stretches just below the contour line of 450 meters above sea level, at a point where the two ridges are slightly drawn towards each other. The survey area is thus situated at an equal distance of around 1.7 kilometers from the three nearest villages, Kučevište and Pobužje on the north and Mirkovci on the west. Drawn on a map the surveyed territory bears the shape of an almost regular square. Its east-west axis measures 1.2 kilometers at its widest, while the north-south axis is only 100 meters shorter. In total it occupies an area of over 1.1 square kilometers and is slightly larger than the first survey area. In reality the limits chosen enclose a theatrically shaped terrain (photo IV_2). The northern boundary is longer and nearly 100 meters higher than the southern along the valley floor; while along the ridges’ summits, the difference in altitude between the northern and southern end of the eastern ridge is less than 70 meters. The ground slopes very gently and evenly, even across the slopes of the ridges, i.e. the valley sides. The eastern ridge holds the highest point in the survey area, but in its southern half, it is lower than the almost flat summit of the western ridge. Both measure slightly over 100 meters across their summits, but the slope on the eastern side is almost twice as wide as the opposite western slope. The former roughly measures over 600 meters, the latter around 330. As a result of this circumstance the eastern slope is much gentler than the western, where along certain sections the ground is stabilized by tall terrace walls. The valley floor is roughly marked by the Skopje-Kučevište road, barely measuring a dozen meters in width. It runs asymmetrically across the survey area, in a northwest-southeast direction. Therefore the western slope gradually widens from north to south, while the eastern slope retains its width, as the main mass of the ridge spreads parallel to the valley floor and the asphalt road. It is obviously difficult to clearly separate the relief elements that comprise the survey area. In this respect too, it is almost completely the opposite from the area of Sopot. For the latter, it was possible and even necessary to define a dozen separate subunits, usually coinciding with certain micro-topographic entities. Although larger, the second survey area technically consists of only five topographic elements: the flattened summits of the two ridges, the valley floor and its western and eastern slopes. But these are not separate surfaces as was the case in Sopot. It is impossible to the draw the boundary between the valley floor and the slopes for instance, neither on topographic grounds nor on the basis of land use or ground visibility. However in the context of the surrounding landscape, they are not totally irrelevant. Three of these subunits could present potential settlement loci: the flattened tops of the two ridges and the valley floor, particularly at the point where it leaves the survey area. We saw in the neighbouring valleys in the foothills of Mt. Montenegro that these types of locations were either occupied by a sacral monument or were inhabited, as was the case with the village Mirkovci situated on the top of the neighbouring ridge. Even in this almost featureless compact landscape not all locations are of the same value. In relation to the current settlement pattern and land-use, the position of the second survey area is again, the opposite of the first. For the first survey area, we followed the limits of an existing settlement and its territory; the second survey area is on the other hand positioned away from the narrower territories of modern settlements, on the periphery of their hinterlands. The lack of natural, topographic fragmentation is further underlined by modern land use. Like most of the foothills, the survey area is parcelled into hundreds of agricultural fields. Their size usually varies between 2 and 4000 square meters and they are often surrounded by hedges and terraces. A network of small, dirt roads gives access to the fields in the various parts of the area. 179 On the steeper western slope, they also play the role of large support terraces. Looking at this landscape, one gets the impression of highly ordered, almost planned agricultural divisions. It is interesting to note however that although the survey area stretches over the hinterlands of two villages, Kučevište and Pobužje, there is no clear boundary separating the territories of the two settlements. This could be a result of the very close connections between the two villages, which in the past few decades are rapidly merging into a larger agglomeration. As in the broader surroundings, only less than 1 in 10 fields is left fallow or abandoned. Cereals and vines are grown along the summits and the upper sections of the ridges, corn and vegetables on the valley floor and on the lower sections. Larger concentrations of uncultivated land appear in the southwest quarter and in the opposite northeast corner of the survey area. The rest of the surveyed land is a monotonous maze of fields of roughly equal size and shape. It is only interrupted by a few enclosed fields. Wild vegetation appears only on the abandoned parcels and on one location on the valley floor, overgrown with tall reeds. Compared to the surroundings of Sopot, ground visibility conditions and accessibility are much better. We were thus finally able to survey an almost continuous stretch of land and follow the distribution of surface finds across fields and across various zones of the survey area. The ground visibility factor was also enhanced by the lack of large amounts of modern debris. In this case waste was usually deposited in the ditches by the road, rarely on abandoned fields. On the cultivated fields modern rubbish appeared only in smaller quantities. There remains the problem of the changing visibility conditions, even in the course of a single agricultural season. It is unfortunate that these variations are most acute during the period of the best ground visibility, in the months between August and November. This is the period when most of the fields are ploughed and prepared for sowing. Between the end of March and late August, the bulk of the cultivated territory is covered with dense and tall vegetation and low ground visibility doesn’t allow systematic survey. However despite the fact that ground visibility is optimal in the second half of the year, the consecutive deep ploughing and harrowing of the fields can result in very dynamic conditions in the topsoil. The second survey areahasn’t suffered major disturbances in the surface layer caused by natural or human induced factors. Except for certain sections on the western slope, the terrain is very gentle. Extreme erosion episodes are very unlikely in such conditions, although the intensive artificial terracing of the terrain is rather baffling. Perhaps the over-exhausted soils are vulnerable even though the potential for erosion is fairly limited. Sedimentation seems as a greater potential problem, though according to the available geologic maps there are no significant concentrations of Holocene sediments on the surveyed terrain or in the wider study region. Because erosion was a much more potent factor in the first survey area, direct comparison of the overall quantity of surface finds between the two could be misleading. Being an inseparable part of a larger micro-regional entity, the second survey area lacks the integrity of the basin of Sopot. We’ll see that this circumstance will greatly influence the local history of settlement and land-use. It will also carry important implications on our reading of the survey results, which become comprehensible only in the light of the developments in the wider study region. Geographically the second survey area along with the entire region of Skopian Montenegro is considered an integral part of the Skopje Basin. Although earlier we stressed that its geographic position on the very edge of the basin and far away from the main rivers is somewhat ambiguous, this micro-region shares the same geo-strategic qualities as most other 180 parts of the Skopje Basin. They are close to two very important interregional corridors and to one of the most important cities of the Central Balkans, but they aren’t as exposed as the valley of Sopot. Compared to the villages in the hilly region of the Middle Vardar, the settlements of the Skopje Basin are much larger and certainly far more receptive of the influences emanating from the central settlement in the plain. Unlike the first study area, situated near ethnic, political and administrative borders throughout most of its history, the second study area was in general always belonging to the central parts of territorial entities, whether political or administrative. As will be shown in the chapters that follow, the developments in the inner Skopje Basin often had a profound impact on the local settlement pattern. 181 Chapter V:Chronology of the collected surface finds and spatial distribution by periods V .1 Introduction As explained in Appendix III unlike in the first survey area, in the Skopian Montenegro survey samples of surface material were collected by individual transects during the quantification campaign. This not only saved us a great deal of time, but it eliminated the potential discrepancies between the number of counted and collected finds; discrepancies which often complicated the interpretation of the individual transect records in the first survey area, where counting and collections were carried out separately. In the second survey the individual transect collections always represented less than 100% of the material counted. Taking place within the period of a single agricultural season, they represent a more coherent and reliable record than the individual transect collections from the Sopot survey. Nonetheless if we are to use them for determining the site limits, it will be necessary to apply the same formula and adjust the transect collections to represent 100% of the material counted. Another positive effect of the slightly modified collection strategy in the second survey area is that we had a raw preview of the chronological profile of the finds prior to the total grid surveys. Many of the suspicions raised in Appendix III proved correct after the processing and basic study of the gathered finds. Most apparent of all and in sheer contrast to the situation in the first survey was the absence of more than a few chronological periods in the surface record. That there lacked the dazzling variety of pottery finds encountered in the Sopot survey became evident by the first year’s campaign, though admittedly there was a hope that the more detailed study of the finds will reveal at least tiny vestigial traces of prehistoric settlement. As will be shown, possible traces of prehistoric settlement were indeed found, but more than 99% of the material belonged to some of the historic periods within the last two millennia. In this respect the dating of the material seemingly posed a lesser challenge, but the fact that the majority of the finds didn’t appear in discrete clusters, comprising integral assemblages was in itself problematic. As a result the dating of a few fabric groups remains uncertain, even within a broader chronological framework. The same low chronological resolution as in the first survey area was used, which as we saw, prevents finer historical interpretation. It seems that traces of prolonged occupational activity date to two, possibly three broader chronological periods. Nevertheless the problems of interpretation encountered in the Sopot survey are no less acute. 15 certain and 4 possible clusters were discovered in the second survey, nearly the same number as in the Sopot survey. Each of these separate sites had to be interpreted, their limits drawn and the character of the material that comprised them analyzed. Although for the greater part these were single-period establishments, they were discovered amidst a fairly dense carpet of off-site material, produced over the past several centuries by the villages Kučevište and Pobužje. Often the quantity of the total surface record on the site locations didn’t surpass the quantity of off-site material on certain field blocks in the western half of the basin. To further complicate matters, the material that comprises the clusters of archaeological finds on the eastern ridge spreads in an uneven carpet across much of the area’s eastern half. Thus these sites were not only difficult to detect during field survey, but they also needed to be defined against the contemporary background scatter. In these cases, as in Appendix II dealing with the surface material from the Sopot survey, we will have to employ a sort of residual analysis to test the on-site status of suspected concentrations374. Like some of the sites discovered in the Sopot 374 For this method see, J. Bintliff, P. Howard, 51-91, 1999. 182 survey, these are not always compact, single-core clusters. We will see that a number of sites consist of two equal, closely spaced cores or one larger, accompanied by a closely positioned, smaller core. Some clusters also feature an intermediary zone, characterized by artifact densities lower than the core but higher than the surrounding background 375. Finally, the interpretation of these findings inevitably raises the question of what thesesites actually represent in sociohistorical terms. Of no lesser importance is the character of the off-site carpet generated by the villages at the mountain foot376. Most important of all, its chronology could point to the approximate date of origin of the present-day settlement pattern. As discussed in chapter IV, it is quite possible that the basis for the local network of villages was established as early as the Late Byzantine Period. Establishing an approximate lower chronological limit for the off-site debris will be an important contribution to the understanding of the history of human settlement in the broader region of Skopian Montenegro. The distribution pattern exhibited by this off-site carpet is itself baffling. Typically for off-site carpets it extends continuously over a large piece of territory with gradual tendencies of decrease along certain axes. At places however this regular distribution is interrupted by sudden peaks, zones of higher artifact density that differ little from the rest of the clusters of archaeological material. As suspected during the total grid survey, most of the clusters that emerged on the western ridge consisted of discarded, rather than unearthed material. Hopefully the analysis of the composition of this material by fabric groups will shed more light on these anomalous concentrations of off-site debris. In the end in order to understand the long-term developments in the surveyed basin, it will be necessary to turn to the broader regional context of the Skopje Basin, with a particular emphasis on the region of Skopian Montenegro. The high degree of integrity of human settlement in this micro-region was stressed in chapter IV. It is the key to understanding why the surveyed basin never grew into a stable settlement niche, although it had the potential to become one. Before discussing the chronology of the collected finds and their spatial distribution, a word is needed on the method of finds processing, which also slightly differed from the one applied to the material from the Sopot survey (see Appendix III). Already at the stage of surface material collection, it became clear that a very large portion of the finds consisted of badly worn, often completely defaced fragments of architectural ceramics. The collections were weighed and counted and all duplicate and badly worn tile or brick specimens were discarded prior to washing and further study. This was possible thanks to the fact that over large zones of the survey area including several dozens field blocks, the bulk of the material consisted of a few, repeating types of brick or tile of the same fabric. Based on brief notes taken during the counting and weighing of the finds and keeping a few total collections, it was possible to observe a consistent tendency in cases when the total collections were studied (graph V_1). The categories of Late Ottoman and Early Modern and unrecognizable material are regularly increased in the total collections. These are simply the categories that we wanted to get rid off: in the case of the Late Ottoman to Early Modern finds, the discarded finds almost exclusively consisted of brick and tile fragments, while the category of unrecognizable material always consisted of completely worn and rounded ceramic fragments that could offer little specific information. The effect of this selection strategy 375 All these phenomena find close parallels among the results of similarly designed research in the Eastern Mediterranean. S.E. Alcock, et al. 137-170, ed. I. Morris, 1994;J. Bintliff, 200-215, eds. M. Pasquinucci, F. Trément, 2000. 376 J. Bintliff, A. Snodgrass, 506-513, 1988; T.J. Wilkinson, 31-46, 1989. 183 is that it slightly minimizes the powerful presence of the finds datable to the past two centuries. This must be taken into account while discussing the artifact densities by period. On most grid collections they are probably greater than indicated on the thematic maps, as these don’t include the discarded categories. In cases where the margin is larger, such as in collection 2 on graph V_1, the actual quantities will be specified in the text. Graph V_1: Chronological profiles of sample and total collections from various grid units 100.00% EM_LO 80.00% Med_Ott 60.00% Roman 40.00% Undated 20.00% Unrec. 0.00% Spl.coll.1 Spl.coll.2 Tot. coll.1 Spl.coll3 Tot. coll.2 Spl.coll.4 Tot. coll.3 Tot.coll.4 As with the first survey area, the detailed descriptions and analysis of the overall surface record, including the architectural remains and the surface material distribution by periods are given in separate appendices III and IV. In this chapter we present the final interpretations including the size and inner structure of the on-site clusters, definition and interpretation of the off-site zone and satellite clusters, analysis of the site locations and socio-historical interpretation of the discovered surface remains. Descriptions of the surface material distribution and fluctuations in artifact densities will be as cursory as possible. V.2 The problem of the prehistoric settlement in the survey area (tables 1-2, Appendix IV) The present day relief and the pedo-geological substrate of the survey area seemingly offer favorable conditions for the emergence of early prehistoric farming communities. The entire survey area is covered by the same or very similar Tertiary sediments to those covering the basins along the Middle Vardar and its eastern tributaries. Recall that the earliest farmers in the Central Balkans occupied this zone by the end of the 7th millennium BC377. Admittedly the settlements from this period tend to concentrate on slightly lower attitudes, in the immediate vicinity of running or still water. But in this respect too, the survey area wasn’t deficient in water close to the surface. At least in the recent past, large portions of the surveyed basin were swampy, while the majority of the streams coming from the mountainside are perennial. Possibly the slightly cooler and more humid climate conditioned growth of forests in this peripheral zone of the Skopje Basin confining the habitation zone of the Early Neolithic communities to the interior of the basin, closer to the Vardar and the last remnants of the Central Balkan Lake378. But even later prehistoric periods were only represented by a very small fraction of the collected surface material. There lacks even a single ceramic assemblage broadly datable to the 377 378 M. Garašanin, 79-114, ed. A Benac, 1979; D. Zdravskovski, 55-63, 2008. A. Sherratt, 261-305, eds. I. Hodder, G. Isaac, N. Hammond, 1981; J. Bintliff, E. Farinetti et al. 665- 674, 2006. 184 Bronze or the Iron Age. This came as a surprise, especially after the realization of the richness of prehistoric material found in the seemingly inhospitable and dry area of Sopot. We expected to find at least half of the prehistoric periods documented in the Sopot survey in the presently far more fertile region of Skopian Montenegro. But after the full study of the collected finds, it was possible to identify only a small group of prehistoric shards and a small fabric group, possibly datable to some late prehistoric epoch. The small number of finds and their undiagnostic character makes the search for direct parallels hardly attainable379. The amount and the pattern of dispersal of these finds recall the situation in the first survey area during the Bronze Age. Admittedly the collections of prehistoric pottery from Skopian Montenegro present an even more extreme case, as we don’t know if they formed a contemporary assemblage. Therefore their distribution should be analyzed separately. It must be emphasized that the very small volume of the collection came to light only after the processing of the finds. These “discoveries” were unintentional and came both from the individual transect and the total grid collections. The first group of securely dated prehistoric fragments didn’t form a coherent fabric category, but the shapes and the fabric features clearly pointed to a prehistoric date. Of the 12 fragments securely dated to prehistory, 5 came from the transect collections (maps V_1 and 2). These are all individual finds spaced at least 250 meters apart. Two were collected from the upper portions of the eastern ridge, from field blocks 38b and 22b and another pair was collected from the western ridge, from field blocks 167 and 208b. The fifth fragment came from field block 377a from the southeast corner of the survey area, near the floor of the basin. The grid survey covering field block 38b confirmed the presence of small quantities of prehistoric material: two fragments came from grid units within the limits of this field and another one in the northern end of the grid, about 70 meters away (map V_3). 350 to 500 meters to the south of grid 1, below the tip of the eastern ridge, the total grid survey revealed another dispersed group of late prehistoric fragments. Two isolated finds came from grids 11 and 4, spaced about 110 meters apart. Finally, a pair of prehistoric shards was discovered among the total collection from the central unit on grid 6, which gave the highest density of 1.33 fragments per 100 sq meters. At present it is impossible but to speculate about the real significance of these remains. But following the interpretation offered for similar phenomena discovered in the Sopot survey and in other intensive survey projects, one could conclude that these were probably the traces of smaller and not very long-lived installations380. The facts that these finds were discovered at considerable distances from each other and that the find-spots differ little in terms of density or the degree of preservation offers a further support for this thesis. Nucleated prehistoric settlements are likely to leave more substantial surface clusters; as we know from the experience of the Sopot survey where even hamlets not larger than half a hectare produced full domestic assemblages consisting of hundreds of pottery fragments. On the other hand, the chances of survival of prehistoric material originally discarded on the surface are infinitesimally small 381. In fact it is quiet possible that these prehistoric sites were of a similar size to the later Roman farmsteads, superimposing some of the prehistoric remains. If this observation is only partly correct, it will imply that the precedent for the settlement pattern during the Roman Period with its focus on the summit of the eastern ridge was set many centuries, perhaps even millennia We’ll see that not too many prehistoric sites are known from the Skopje Basin. D. Mitrevski, 449-457, ed. B. Hänsel, 1998; D. Mitrevski, 1997; I. Mikulčić, 17-35, 1982. 380 J.L. Bintliff, P. Howard, A.M. Snodgrass, 139-168, 1999. 381 E. Neustupný, 49 ed. E. Neustupný, 1998. 379 185 earlier. However one cannot assign too much weight to the observed overlapping distribution patterns. The coincidence of the scarce late prehistoric finds with sites of later historic periods could result from the favorable taphonomic factors rather than reflecting long-term continuity in the focus of local settlement382. As discussed in appendix IV, the occurrence of small quantities of prehistoric material on later historical sites is also a consequence of the increased survey intensity on these locations383. But the total absence of this material from grid units covering field blocks in the central sections of the survey area and on the western ridge clearly indicates that the distribution of this material isn’t isotopic, nor is it necessarily linked with the Roman sites. The pair of late prehistoric shards collected from the western ridge came from individual transect collections and were spatially unrelated to sites from later periods, as were the collections from grids 1 and 4 on the eastern ridge. Plausible as it seems, the thesis that these scattered finds are the vestigial remains of prehistoric farms or huts is difficult to support without repetitive collections from these find-spots or perhaps, sub-surface research. One shouldn’t exclude the possibility that these scatters were derived from non-residential site categories, such as burials, votive pits or other types of sub-surface deposits384. At the moment, it remains unknown if these fragments are datable to a single prehistoric period. Even if we allow that at least some of these find-spots are the remains of farmsteads, their chronology remains problematic. Scholars that have studied small dispersed prehistoric settlements believe that even if all such settlements are datable within single prehistoric period, they weren’t particularly long-lived and it is very unlikely that more than a few were truly contemporary385. The second group of finds with a possible prehistoric date forms a coherent fabric category. Although they are slightly more numerous amounting to a total of 18 fragments, they were less dispersed than the heterogeneous group of securely dated prehistoric material (cf. tables 1 and 2, Appendix IV, map V_3). 16 fragments came from two locations near the summit of the eastern ridge, covered by grids 11 and 6. The larger cluster was discovered on grid 11, where we recorded the maximum density of 3.3 fragments per 100 sq meters. The second smaller cluster was discovered in the central parts of grid 6, less than 200 meters to the south and on slightly lower ground. In both cases the core of the clusters is limited to a single grid unit (about 150 sq meters) with individual fragments appearing on units contingent to the core. The other two fragments belonging to this fabric class came from grids 8, 100 meters to the north of the cluster on grid 6 and from the northern end of grid 2, about 280 meters northwest of the cluster on grid 11. Examples of this fabric never appear outside the higher portions of the eastern ridge and not a single shard came from the individual transect collections. Clearly these are more substantial clusters and the possibility that they are vestigial traces of prehistoric farmsteads seems even likelier. However there remains the problem of their chronology, along with the chances that they were a component of some later assemblages. Judging by the fabric features and the primitive hand-modeling this pottery is also prehistoric. But suspicions regarding its dating were stirred by its pattern of distribution. Both grids 11 and 6 cover Roman Period sites and the isolated examples from grids 2 and 8 also came from amidst or 382 J.L. Bintliff, P. Howard, A.M. Snodgrass, 156, 1999; M. Kuna, 29-44, eds. J. Bintliff, M. Kuna, N. Venclová, 2000; for criticism see, J. L. Davis, Are the landscapes of Greek prehistory hidden? 22-35, eds. S.E. Alcock, J.F. Cherry, 2004. 383 J.L. Bintliff, P. Howard, A.M. Snodgrass, 145, 149, 1999. 384 See examples in M. Godja ed. 2004. 385 J.L. Bintliff, E. Farinetti, et al. 665-674, 2006. 186 near dense clusters of Roman material. At the same time however, there is a considerable overlap with material securely dated to prehistory, while the Roman clusters on the lower terraces of the eastern ridge and near the valley floor never included fragments of the possible late prehistoric fabric, although the assemblages differed little from those that constituted the Roman clusters on the eastern ridge. The find-spot of the highest artifact density for the group securely dated to prehistory on grid 6 is also the core of the second, smaller cluster of the possible prehistoric fabric group. There are further overlaps on grids 8 (field block 22b) and 11. Only on grids 4 and 1 do the securely dated prehistoric shards appear unaccompanied by examples of the discussed fabric group, although a fragment of the latter group came from the northern end of grid 2, at a distance of only 65 meters from the scatter of securely dated prehistoric fragments from the eastern half of grid 1 (field block 38b). Thus the pattern already exhibited by the group of finds securely dated to later prehistory is partly repeated and enhanced. There is a clear focus on grids 11 and 6 (point 501 on map V_2), which is a location analogous to the one occupied by the fortification described in Appendix III (point 540). Finds securely dated to prehistory were found dispersed over a somewhat larger area, with a few fragments coming from the upper terrace, above the contour line of 520 meters. Understandably there is very little ground to speculate about the chronological relation between the two groups of fragments, especially because the securely dated finds do not form a homogenous group in terms of fabric categories. In this particular case we believe that regardless of the considerable overlap between the two groups, they don’t appear to form an integral contemporary assemblage. There are considerable difference in the fabric characteristics between some of the fragments securely dated to prehistory and the possible prehistoric fabric group. The complete absence of examples of the latter group on grid 1, one of the find-spots featuring slightly greater concentration of securely dated prehistoric material cannot be overlooked, as much as its fairly sporadic appearances among the possible prehistoric clusters on grids 6 and 11. Either these two groups represent different prehistoric strata or the fabric for which we suggested a possible prehistoric date belongs to the Roman assemblage. As demonstrated in appendix IV, both interpretations agree well with the distribution maps. It was in any case deemed important to briefly consider the implications of there being a separate prehistoric fabric group, apart from the handful of finds with a secure prehistoric date. Even though remaining unanswered the issue has no impact on the long-term settlement dynamic in the second survey. If we adopt the view that these scatters are the vestigial remains of prehistoric farms, the similarity with the distribution of the settlements from the Roman Period is indeed remarkable. As will be shown when discussing remains of later periods, most settlements in the survey area will retain the rank of farmsteads, whether isolated or forming dispersed networks. The preference for locations close to, or at the very summit of the eastern ridge could be related to a number of factors, primarily agricultural exploitation. It is notable that this type of locations is analogous to the micro-location of the modern village of Mirkovci situated less than 2 kilometers to the west of the surveyed ridge. These flattened ridges offer both relative safety from eventual floods and an immediate access to the agricultural land on the lower slopes and on the basin’s floor. At the same time, from these micro-locations it is possible to observe a large portion of Montenegro’s foothills. The local north-south roads linking the plain with the mountain usually follow the crests of these ridges, like the modern asphalt road running between Skopje and Pobuže. This natural corridor passes immediately next to the cluster on grid 11 and less than 100 meters east of the cluster on grid 6. Further north, it probably passed by the small 187 scatter of later prehistoric finds on grid 1 and the Late Roman fortification that marks the northeast corner of the surveyed basin. The couple of securely dated prehistoric fragments from the summit of the western ridge are in all aspects unrelated to the prehistoric material from the eastern ridge (map V_3a). These finds are made in different fabrics and show little or no similarity to the prehistoric shards from the eastern ridge. In all probability they belong to a different prehistoric period. This is consistent with later developments for as will be shown, the two ridges were never occupied within the same time-period. Although seemingly there is very little difference between the neighbouring ridges, there are certain peculiarities regarding access to communication and vegetation. In this respect the western ridge is slightly disadvantageous, being more isolated from the local roadnetwork and drier. V.3.1 The Roman Period: overall and on-site distribution(2nd-4th c AD? Tables 3-22, graph 1-21, Appendix IV) This is the earliest period for which there is a substantial surface evidence for human occupation in the survey area. We will see that such a discontinuous sequence of settlement is a typical feature of the Skopje Basin. As in the first survey area, it isn’t possible to work with a more refined chronology. The upper and the lower limits of the period remain vaguely defined. This is expected considering the fact that so far, very little is known about Early and Mid Imperial Scupi. Thankfully the ongoing salvage excavations on the city’s eastern necropolis have unearthed hundreds of burials dated between the early 2nd and the early decades of the 4th century AD. The excavated material mostly comes from enclosed, undisturbed deposits and was a precious source for comparison386. Close direct parallels were found for most of the fabric groups broadly dated to the Roman Period. In fact it is not impossible that the material from both locations was made in the same fabrics. However this hardly determines the actual dating of the survey finds. On a few locations along with the bulk of the plain, soft fabrics, there are rare fragments of fine Late Roman sigillata, also found in late 4th century deposits in Scupi387. At the same time there lacks among these collections, the local pottery production that typifies the 5th and the 6th centuries AD. This is a better known and more recognizable material. The Late Roman pottery from Scupi was published several years ago and there are parallels on a number of other sites in the Skopje Basin dated through architectural survey and excavations 388. As will be shown there are finds datable to the last two centuries of Antiquity in the surface record, but they are unrelated to the main concentrations of finds broadly dated to the Roman Period. Hence the proposed upper chronological limit rests solely on negative evidence and consequently, it should be accepted with caution. Material datable to the Roman Period is the second most dominant chronological group in the surface record of the second survey area. Over 3670 fragments were dated to the Roman Period and it is certain that their number is much greater (table 3, Appendix IV). Nearly 50% of the material from certain collection units represented fragments worn beyond recognition, which 386 This material is yet to be published; courtesy to L. Jovanova, a custodian at the Museum of Skopje in charge of the excavations, for allowing me to look at some of the material. So far only the Late Roman pottery from Scupi has received a separate study, M. Ončevska-Todorovska, 2004. 387 M. Ončevska-Todorovska, 45-48, Tab. 1-6, 2004. 388 I. Mikulčić, Docnorimski grobovi od Skupi, 109-143, Godišen Zbornik na Filozofskiot Fakultet 26, 1974; I. Mikulčić, N. Nikuljska, Markovi Kuli-Vodno, kaj Skopje 65-74, Macedoniae Acta Archaeologica 5, 1979; N. Čausidis, 183-197, 1989. 188 we decided to discard after weighing and counting. Later it was discovered that the majority of these tiny and defaced shards were in fact small bits of Roman brick and tile. Was this material kept and included in the analysis, the number of Roman finds would certainly grew for up to 50%. Nevertheless because the number of thrown fragments of Late Ottoman-Early Modern tile is many times greater, in reality Roman material comprises less than 30% of the total surface record. In absolute terms the Roman collection from Skopian Montenegro is at least thrice as large as the collection from Sopot. This neatly coincides with the three times larger area of the Roman settlement in the second survey. As in the first survey area, Roman finds are to be found over a large portion of the surveyed basin. Almost 40% of the field blocks featured at least one fragment datable to the Roman Period, which is again remarkably similar to the percentage of field blocks with Roman material in the first survey. There is however an important difference. While in the first survey Roman finds were found in virtually all survey sections, in Skopian Montenegro over 98.5% of the material was collected from field blocks to the east of the Skopje-Kučevište asphalt road. Thus the main focus of residential and other activities was obviously on the eastern ridge and in the central survey sections. Consequently the mean overall density recorded by the transect survey is extremely skewed, with only 1.8 fragments per 1000 sq meters. In the case of the Roman material in the second survey, the mean district values are far more reliable references. The mean density of Roman material on field blocks east of the Skopje- Kučevište road is nearly 3 fragments per 1000 sq meters, while on field blocks to the west, on average the transect survey recorded 2.2 fragments per 1 hectare. Moreover as collections by individual field walking transects were not particularly thorough on a number of site locations, it is likely that even these values are a slight underestimate, especially for field blocks in the eastern survey half. The analysis of the statistical distribution of the Roman finds by field blocks defined three basic zones of artifact density. The great majority of field blocks or over 80% belong to the zone of low or very low artifact density, featuring between 0 and 2.5 fragments per 1000 sq meters. Except for the few cases of data loss or where transect collections were purposefully less intensive, these field blocks cover the off-site zone. The second zone characterized by artifact densities ranging between 2.5 and 11.3 fragments per 1000 sq meters is predictably far more limited. Less than 20% of all field units belong to this zone. This group includes field blocks that feature artifact densities close to the mean value for the eastern survey half, but also units that feature densities that are 2-3 times higher. It is therefore useful to further distinguish between the group of field blocks featuring between 2.5 and 5.1 fragments per 1000 sq meters, which roughly equals the median value for this survey area and field blocks with artifact densities ranging between 5.1 and 11.3 fragments per 1000 sq meters. The former group comprises less than 9%, the latter 11% of all field units in the second survey. On some of these field units the total grid survey revealed on-site densities, at least over portions of the field block area, while on others, scatters of average or lower than average artifact density. The zone of the highest artifact density consists of field blocks on which the transect survey recorded over 11.3 fragments per 1000 sq meters. They comprise only 3% of all field walking units and unless there were radical changes in the surface record during the field survey, they signaled the presence of Roman sites. This is in tune with the findings of the Sopot survey where the on-site densities were at least two to three times the mean district values and on the great majority of sites they were many times higher. There is a more or less clear pattern in the spatial distribution of the various zones of artifact density (map V_4a). Field blocks featuring between 5.1 and 11.3 and those featuring over 11.3 fragments per 1000 sq meters don’t appear to the west of the Skopje-Kučevište road. 189 They are limited to the eastern ridge and the northern half of the central survey section. We see that field blocks that belong to these two zones almost always appear together, forming fairly extensive patches of high artifact densities. Whenever the collections by individual field walking transects included at least one third of the material counted, there emerged roughly concentric patterns with one or two field blocks featuring over 11.3 fragments per 1000 sq meters, surrounded by a few field blocks that featured between 5.1 and 11.3 fragments per 1000 sq meters. These clusters of field blocks usually occupy an area of between 1 and 2 hectares. In some cases as on the top of the eastern ridge, they are contingent, forming large continuous belts of high or very high artifact density. As a result it is difficult to count their exact number, although it is evident that there were at least 7 or 8 such clusters in the eastern survey section. There is an apparent concentration along the top and the upper portion of the eastern ridge, where one can observe at least 6 such clusters, centered on field blocks 351, 336, 320, 22a-b, 1-2 and the northernmost on field block 37. Only two clusters were revealed by the transect survey collections along the lower terraces of the eastern ridge: one centered on field blocks 47a, 49-50, the other on field blocks 289a-290a, 250 meters to the south. In the central survey section, there is one potential find-spot on field block 137, near the northern survey limit. A larger concentration of field blocks featuring between 5.1 and 11.3 fragments per 1000 sq meters was revealed on field blocks 84-85 and 89, about 300 meters to the east and also located by the northern limit of the survey area. Field blocks featuring over 11.3 or between 5.1 and 11.3 fragments per 1000 sq meters rarely appear isolated. Unless related to the small size of the field block artificially enhancing the artifact density or to the inconsistent collections by individual transect units, these isolated peaks could signal the presence of a different site category. But in the majority of cases, the roughly concentric pattern is further emphasized by the distribution of field blocks featuring between 2.5 and 5.1 fragments per 1000 sq meters. They too tend to cluster around field blocks with artifact densities higher than the mean district value and are nearly absent from the western survey half. Indeed one notes that this group of field units often forms perfect outer rings around the zones of high or very high artifact densities. They are mostly concentrated on field blocks to the west of the high-density clusters along the top of the eastern ridge and on the stretches that separate the high-density clusters on field blocks 47a, 49-50, 66, 289a-290a and 137. Finally, field blocks featuring less than 2.5 fragments per 1000 sq meters of transect survey cover over 95% of the western ridge, but they also cover the greater portion of the central survey section and considerable stretches on the eastern ridge. On the eastern ridge they often form larger compact stretches separating the zones of average or higher than average artifact density. These are exemplified by the large group of field blocks in the northeast corner of the survey, on the very top of the eastern ridge and by the nearly sterile stretch separating the high-density clusters on the upper portions of the ridge from those on its lower terraces. Unfortunately this overall pattern of distribution came to light only after the study of the collections by individual transects. By that time, the total grid surveys were already being carried out and the only guidance we had for determining the focus and the limit of the total collections was the overall artifact density. As anticipated during this stage of fieldwork, the zones of high overall artifact density don’t necessarily overlap with the zones of high density of Roman material. This is most clearly reflected on the western ridge where not a single Roman shard was found among the transect collections from field units featuring very high overall density. But there are also mismatches on certain field blocks in the central survey section and on the eastern ridge, such as field block 263 by the Skopje-Kučevište asphalt road or field blocks 38a/b-40a/b 190 in the northeast corner of the survey area. The chronological composition of the surface record on each of these locations is given in Appendix IV. Here it suffices to mention that the distribution of the total surface record in the second survey in general was chiefly determined by the distribution of the predominant Late Ottoman-Early Modern material. Nevertheless when present in larger quantities, theRoman material did have an effect on the amount of the total surface record, even on field blocks where material dated to the last couple of centuries was absent or present in modest quantities. Thus the zones of high density of Roman material almost always coincided with the zones of high overall density, but the opposite wasn’t always the case. This circumstance explains the fairly satisfactory coverage of the zone of higher density of Roman material by the regular grid survey (map V_4b). More than 50% of the field blocks that feature over 11.3 fragments per 1000 sq meters were included in the total survey. Most of the field blocks that belong to the high density zone and were left out of the total grid survey presented cases where upon return, it was simply impossible to locate the large quantities of surface finds counted during the transect survey. These include two groups of field blocks situated along the lower terraces of the eastern ridge, field blocks 47a-b, 49, 50 and 66 and 281a-290a and field block 328b along the eastern limit of the survey area. It should be noted that although nearly all of these field blocks featured higher than average overall artifact densities, only one or two belonged to the zone of the highest overall density. As explained in greater details earlier in this chapter and in Appendix IV, the changing surface conditions presented a serious problem in the second survey area. Equally disturbing was the accidental discovery of a medium-sized Roman site on field blocks 102-104a/b where the transect collections recorded an almost complete absence of Roman finds. Other examples of field blocks where the high density of Roman material wasn’t reflected in the overall artifact density recorded by the transect survey were field blocks 351, 348 and 342. Although these three field blocks feature overall artifact density higher than the average, the perceived quantities of surface material were simply deemed too low to merit detailed attention. In this case however, after realizing the prominence of Roman material among the transect collections we were able to locate a denser cluster and regular grid survey was carried out over both field blocks 348 and 351. In addition to field blocks featuring higher density of Roman finds, inadvertently the total grid survey included field blocks with average or lower than average density of Roman material. In fact nearly 40% of the gridded area falls outside the zone of very high density of Roman material, excluding field blocks where the transect survey indicated the presence of other periods or where collections by individual transects were less intensive. This total coverage of the offsite zone opened an important insight into the distribution of the finds beyond the site limits and it also enabled us to roughly assess the true density of Roman material on the basis of the transect survey record. The total grid survey revealed at least one very significant weakness in the data produced by the transect survey. We encountered the same problem in the first survey area. It is very difficult to predict the true artifact density and nearly impossible to draw the limits of the sites on the basis of the transect survey record. This is particularly problematic in the second survey because sites were often smaller than 1000 sq meters. The large and irregular field block can hardly provide an adequate spatial frame for documenting surface clusters of such size. It is merely possible to point to the presence and the approximate locations on the basis of the collections by individual transect units. Nevertheless these proved to be an indispensable tool for delimiting the site limit on locations left out of the regular grid survey, as well as for delimiting 191 the extent of the site halos. Although highly focused, the total grid survey with its coverage of only 10% of the survey area could only hope to document a portion of the true number of sites and satellite clusters. Even when a site location was recognized during the course of fieldwork and total collections were carried out, it was often very difficult to determine the focus and the limits of the regular grids. This was true regardless of the fact that in comparison to the first survey area, the maximum densities recorded on the Roman sites were higher, with a considerable difference between the minimum and the maximum value. As explained in Appendix IV, two main factors caused these difficulties: the fairly low quality and obtrusiveness of the material datable to the Roman Period and the irregular, focalized patterns of distribution, both on-site and in the immediate surroundings of the sites. This latter circumstance proved particularly problematic not only during fieldwork, but also during the analysis of the grid survey records. Only on a minority of examples was it possible to observe a concentric pattern of distribution across the site areas. The great majority of the sites revealed by the total grid survey consisted of at least two cores separated by wider stretches of lower artifact density, in some cases dropping below the site threshold. These considerable fluctuations in the artifact density over relatively short distances continue outside the site areas, with low peaks regularly appearing in the halo zone. Predictably the total grid survey rarely managed to capture these satellite clusters of intermediary density, but their presence was often picked up by the transect survey. In order to obtain a fuller understanding of the extent of the site area and the distribution in the halo zone, it was necessary to combine the results of the transect and the grid survey. The unequal intensity was another significant challenge in using the collections by individual field walking transects as records of artifact density389. In the second survey area, on nearly 70% of the field units the transect collections included over 33% of the material counted. Collections including less than 30% of the material counted were mostly related to data loss or to a deliberate decision to save the surface record for the total grid collections. Even smaller samples would have recorded all major periods in the surface record, but comparing the density of certain categories of material on field blocks where the transect collections included 40 and 60% of the counted material is understandably misleading. This factor must be taken into account when trying to draw the site limits on locations not included in the regular grid survey. But as elaborated in a greater detail in Appendix IV, adjusting the individual transect collections so that they represent 100% of the material counted results only in minor and localized corrections (map V_4c). Overall there is a slight increase in the number of field units featuring very high or higher than average density of Roman material. About 4% of the field units belong to the very high density zone, while 18% feature average or higher than average artifact density (cf. graphs 20 and 21, Appendix IV). There aren’t any major changes in the pattern of distribution that was described in the preceding paragraphs. Very high concentrations emerge on field blocks 277b and 10, unmasking the small size of the transect collections. Some of the field units for which we suspected on-site densities form better defined clusters (field blocks 125-129, field block 342), while in a few cases the analysis blurred the limits indicated by the unadjusted record of the transect survey (field blocks 336, 289a-290a). Even though the individual transect collections represented over 30% of the material counted, on almost ¾ of the field units, the predicted densities are considerably higher. The lower threshold of the average density zone increases from 2.5 to 7, while the lower threshold of the very high density zone from 11.3 to over 38 fragments per 1000 sq meters. While the difference between the density figures recorded by the regular grid survey and the unadjusted transect collections were often 389 J.L. Bintliff, P. Howard, A.M. Snodgrass, et al. 19-20, 2007. 192 tenfold, the adjusted transect collections produce densities that are 2 to 4 times lower than those recorded by the total collections. The total collections on grid 1 covering an area of nearly 1.5 hectares and including field blocks 38a/b-40a/b, 59, 61a-b were carried out prior to the study of the material collected by the individual field walking transects (maps V_5a-c). These field blocks were selected for a total grid coverage because of the very high overall artifact density, but both the transect and the grid surveys confirmed that the great majority of the finds dated to the last couple of centuries, with a very small percentage of Late Byzantine-Early Ottoman and prehistoric finds (tables 4-5 graph1, Appendix IV). This was the first clear signal that the off-site debris from the last couple of centuries was not only present in considerable quantities in the survey area, but its density also varied considerably from field to field. Nevertheless the small collection of Roman finds comprising less than 15% of the total surface record on this location formed a tiny cluster with on-site densities in the southeast corner of the gridded area (map V_5b). It is defined by artifact densities higher than 6.5 fragments and a maximum of 11.5 fragments per 100 sq meters at the site core. Excluding the site halo, the average density is 8 fragments per 100 sq meters, slightly over twice the district average. This is a compact cluster measuring less than 500 sq meters. It barely emerges from the off-site segment revealed on the rest of the gridded area where artifact densities consistently measure about 0.65 and on certain location rise to over 3 fragments per 100 sq meters. It is questionable if the small and thin cluster of Roman finds on grid 1 represents a separate establishment or it is a part of the much larger cluster uncovered by grid 2, about 50-60 meters to the south-southeast (maps V_5a-c, tables 4-5, Appendix IV). As explained in Appendix III, the transect survey recorded an increased amount of the overall surface record on the field blocks immediately to the south of grid 1 stretching partly over the same, partly over the lower terrace. Unlike on the field blocks covered by grid 1, the individual transect collections from field block 37 indicated the presence of Roman material, giving a density of nearly 12 fragments per 1000 sq meters. To some degree this was confirmed by the results of the total grid survey, which apart from field block 37 covered partly or entirely field blocks 33-35. This was a many times larger site, with a higher maximum density and a slightly higher on-site threshold. Its limits coincide with a narrow belt with artifact densities ranging between 6.5 and 8.3 fragments per 100 sq meters. On the central portion of the site in the southern half of the gridded area, artifact densities reached to over 20 fragments per 100 sq meters. Thus over ¾ of the site area are located on the lower terraces on field block 34 and only the northern end of the site stretches over to field block 37. This imprecision of the transect survey data has to do with the inconsistent intensity of the collections by individual transects. Once the individual transect collections are adjusted to represent 100% of the material counted, field block 34 joins the zone of higher than average artifact density, though it is still ranked lower than field block 37. On field block 32 covering a segment of the site halo, the density predicted are lower and this unit is shifted to the zone of average artifact density (map V_5c) Further in the text we will see that in comparison to other Roman sites in this survey area, the cluster on grid 2 is fairly compact. Even so there stretches across the centre of the cluster a narrow strip of intermediary density, nearly splitting the site area into two halves. This contributes to the relatively low average density on this site of only slightly over 8 fragments per 100 sq meters. It must be noted however that this figure doesn’t include the discarded fragments, which in the case of grid 2 mostly comprised worn bits of Roman tile. Similar narrow strips of artifact density higher than 8.3 fragments per 100 sq meters cross over to the upper terrace on 193 field block 37. Including only grid units featuring densities higher than 8.3 fragments per 100 sq meters, the cluster on grid 2 occupies almost 5 500 sq meters. Turning back to the issue of the relation between the neighbouring clusters on grids 1 and 2, it was thought helpful to analyze the distribution of the Roman material by basic fabric categories (graph 2, Appendix IV). In general as in the Sopot survey, it was easy to make a distinction between fine and coarse fabric groups, the latter often featuring a number of different inclusions in the paste. Apart from cooking, some of these fabric groups could have been used for storage and transport. Compared to the Roman pottery from the Sopot survey, there are a considerable number of coarse fabric groups. It has to be stressed however that because of the character of the local material, the category of fine ware actually consists of plain pottery or pottery where the slip was worn. Truly fine, decorated or glazed pottery was almost completely absent390. It is equally possible that a portion of the plain ware was also used for transport and storage, but these were evidently fragments of smaller vessels, possibly serving as smaller jars or table jugs. In this respect, the classifications presented are not comparable to the similar analyses in Greece and the Aegean, where pottery experts can distinguish between table ware, cooking or processing vessels and transport and storage vessels391. In principle the analyses of the assemblages from both survey areas distinguish between plain and coarse fabric groups, broad categories that shouldn’t be directly related to function. In addition we were able to give more precise estimates about the true quantities of architectural ceramics. As discussed in the appendix, the two assemblages have a similar composition. In both collections architectural ceramics is the most predominant category, comprising between 60 and 70% of the collections of Roman material. Plain pottery is slightly more numerous on the cluster on grid 2 representing about 20%, while coarse fabrics represent around 13% of the collected material. On the cluster on grid 1 both categories are represented by about 17% of the collections. It is noteworthy that this composition is very similar to the composition of the Roman-Late Roman assemblages from Sopot, especially those from site 5a-b. This relative consistency in the shares of the basic ceramic categories on the clusters on grid 1 and 2 is rather exceptional for the Roman sites in the second survey. The collections from most other neighbouring clusters showed considerable variations. The fact that the Roman collections from grids 1 and 2 were made up of identical fabrics suggests that the two were most probably contemporary. At the same time, the similar composition of the assemblages from both clusters can be interpreted as a sign that both sites had a similar function and existed independently. However their very close proximity to each other, the incomparably smaller size of the cluster on grid 1 and above all, the fact that this pattern of closely spaced pairs consisting of one small and one many times larger cluster is repeated on a number of other locations in the survey area indicate that the two probably functioned as components of a single establishment. Being ten times smaller and situated only 50 meters away, the cluster on grid 1 could be an outbuilding of the main settlement unit on grid 2. This is further suggested by the fairly extensive zone of intermediary density spreading over most of the area covered by grid 1. It is unlikely that this off-site material was produced by the small cluster on grid 1. This zone of intermediary density must be the halo of the site on grid 2, spreading over a distance of 140 meters to the north of the site. It is nearly certain that similar densities would On the special value assigned to this pottery class in the Roman World see J. T. Peňa, 20-21, 2007. J.L. Bintliff, P. Howard, A.M. Snodgrass, tab. 5.3, 2007; T. Whitelaw, 227-243, eds. R. Francovich, H. Patterson, 2000; M. van der Enden, Koroneia 2009: Surface survey sampling experiments, 33-41, J. Bintliff, B. Slapšak, The Leiden-Ljubljana Ancient Cities of Boeotia 2009 seasons, 1-58, Pharos 17(2), 2012. 390 391 194 have been revealed was grid 2 extended over the neighbouring fields to the east and west. According to the collections by individual transects, the density of Roman material stays above 5.1 fragments per 1000 sq meters on most of these field units, although the extent of the halo in these directions is limited by the proximity of the neighbouring sites on field blocks 47a-b, 4950, 66 and on grids 5-11. On field blocks to the east and west of the cluster on grid 2, the site halo spread for not more than 50 meters. The site on grids 1 and 2 occupies the upper portion of the eastern ridge, the terrace between the contour lines of 510 and 520 meters above sea level. It is located immediately below the summit of the ridge, on the gentler and sunnier western slope. Topographically this microlocation is hardly prominent. It is situated above the most fertile stretches of land in the surveyed basin and has an excellent visual control over the wider surroundings. Like most of the find-spots of prehistoric material, it has an immediate access to the main local road that leads along the summit of the ridge. The small fortification is situated about 350 meters to the north, overlooking the slopes occupied by the site on grids 1 and 2. About 130 meters to the east-northeast of the edge of the cluster on grid 2, we observed larger quantities of Roman material on field blocks 10 and 11 (grid 3 on map V_5b). This pair of contiguous field units was immediately selected for total grid collections and the transect collections included but a few fragments representing a tiny fraction of the material counted. Assuming that all material counted was included in the transect collections will only partly compensate for this bias, elevating the density on field block 10 above the site threshold (map V_5c). Only after the total collections on grid 3 did we clarified the exact size and location of this site (map V_5b, table 5, Appendix IV). It was confined to a pair of grid units on the southern periphery of the gridded area that featured around 15 and 20 fragments per 100 sq meters. Judging by the apparent traces of secondary dislocations along the eastern edge of the field, originally this site was larger and perhaps slightly denser. It is in any case evident that it didn’t occupy an area larger than 300 sq meters. Like the clusters on grids 1 and 2, it is surrounded by a narrow belt of artifact densities ranging between 4 and 6 fragments per 100 sq meters. Excluding this zone from the site area, the cluster on grid 3 has one of the highest average densities in this survey area, but this merely reflects its small size. The total collections from the rest of the units on grid 3 revealed artifact densities ranging between 1 and 4 fragments per 100 sq meters. As argued in Appendix IV this zone coincides with the site halo, which in the case of the cluster on grid 3 cannot be followed outside the gridded area covering field blocks 10 and 11. The neighbouring field blocks to the west were thickly overgrown and Roman finds were nearly absent among the transect collections from the neighbouring fields to the north and south. Seemingly this cluster is of the same rank as the cluster on grid 1. There are however a few characteristics that set it apart from the rest of the small Roman sites in the second survey. The collection of Roman finds from grid 3 is mostly made up of the same local fabric groups that constitute most other Roman clusters and the composition of the assemblage is very close to the one collected from grid 2 (graph 2, Appendix IV). But the collection from grid 3 is the only one in the second survey that includes a small amount of fine table ware. It consists of about a couple of dozen fragments covered with a high-quality red slip and in one case, a finely executed stamped decoration. Among the material collected one can recognize rim fragments from cups, a ring foot from a dish or a plate and bowl fragments. There is also a small group of tile fragments made in a fabric that doesn’t appear on any other location in this survey area. Initially it was thought that these finds could represent a later phase on site 3, but as they are datable to the 4 th century AD, there are no firm grounds to chronologically separate them from the rest of the 195 Roman material392. Even if there was a later phase on this site, the character of the collected finds doesn’t indicate a full residential site. The local coarse ware and the architectural ceramics typical for the 5th and the 6th century are absent from grid 3, as from the rest of the survey area. The cluster on grid 3 is also the only Roman site on which definite traces of stone masonry were found. These are roughly hewn blocks made from a local stone, very similar to those found along the wall of the small fortification discovered outside the northeast corner of the survey area. Like much of the surface material on site 3, they were found dislocated along the edges of the fields, often inserted into the terrace walls that delimit field blocks 10 and 11 from the north and south. Unlike the similarly sized cluster on grid 1 and other small clusters revealed by this survey, site 3 stands fairly isolated. The nearest larger sites are located on grids 2 and 5-11, at distances of 120 and 140 meters. Its micro-location is also far more prominent in comparison to the rest of the sites. The site on grid 3 is the only one situated above the contour line of 520 meters above sea level. It stands lower only to the small fortification discovered 450 meters to its northwest, closer to the apex of the surveyed basin. Although there is little evidence to elucidate the chronological relation between the two, it is very possible that they were at least partly contemporary. As discussed in Appendix III, currently the fortified area has been turned into a grove and was left out of the survey area. During the recording of its plan, it became clear that there is very little surface material, but it was also observed that the craftsmanship of the stonework was similar to that discovered on grid 3. These two sites are both situated near the edge of the hypothetical terrace carved by the receding Central Balkan Lake. The site on grid 3 literally sits at the very edge of this terrace, bound at the south by the contour line of 520 meters above the sea. Both sites occupy locations at the summit of the eastern ridge. The crest of this ridge has long since been turned into agricultural fields. The asphalt road linking the village Pobužje with Skopje was cut along a lower line, on the eastern slope of the ridge. But the easiest natural line of communication is the one following the very summit, passing by site 3 and the fortification and leading directly to a point where the roads leading out of Pobužje and Kučevište meet. The clusters on grid 1and 2 occupy a slightly lower ground, between the contour lines of 510 and 520 meters above the sea. They are positioned on a gently sloping but more spacious and also probably, more fertile terrain. The rest of the clusters of Roman material occupy the lower terraces; the closer to the central axis of the basin, the further away from the natural line of communication connecting the plain and the mountain. As in Appendix IV they will be presented by terraces in a descending order. 350 meters west-northwest of the cluster on grid 1, at the southwest foot of the small fortification, the transect survey revealed a small group of 4 contingent field blocks featuring above 5.1 fragments per 1000 sq meters (maps V_6a-c, grid 14). As with the site on grid 3, this concentration of material was discovered at an early stage and it was decided to collect only a small sample of finds by individual transect units from field block 86, where most of the material was obviously concentrated. When the transect collections are adjusted to represent 100% of the material counted, both field blocks 86 and 84a come to prominence with nearly 50 fragments per 1000 sq meters. The total collections on grid 14 confirmed what was anticipated in the course of fieldwork, although the quantity of the gathered finds was lower than expected (tables 6-7, Appendix IV). This site was revealed on the border between field blocks 86 and 84a, most of the site area encompassed by the former unit. The increased density of Roman material on the surrounding fields indicated by the fairly thorough individual transect collections probably marks 392 The dating of this material was suggested by D-r P. Bes from the CatholicUniversity in Leuven, who kindly accepted to look at some of the photographs. 196 the extent of the site halo, mostly spreading to the north and west of the site. Its maximum radius measured not more than 70 meters in a northwest direction. Like the cluster on grids 1-2, this site is located at the southern edge of its impact zone, occupying a lower ground. This is a much smaller site occupying an area of about 600 sq meters and featuring maximum artifact densities of about 15 fragments per 100 sq meters. There are no traces of secondary dislocations and it is possible to observe an almost regular concentric pattern. The site area is compact, limited to 3 or 4 contingent grid units in the centre of the grid. It is surrounded by a narrow belt featuring between 4 and 6.5 fragments per 100 sq meters. The rather low average density on this site is again related to the decision to discard a considerable portion of the material gathered, especially defaced fragments of architectural ceramics (table 7, Appendix IV). In the eastern third of the gridded area, the density of Roman finds never exceeds 2 fragments per 100 sq meters. This confirms the findings of the transect survey according to which the site halo mostly spread to the north and northwest of the site on grid 14. The composition of the assemblage collected from grid 14 is unlike the assemblages from the sites discussed previously (graph 4, Appendix IV). At least 80% of the collected material was fragments of brick and tile and while coarse ware was represented by slightly over 10% of the collection, plain fabric comprised less than 5%. Almost identically composed assemblages were collected from the first survey area, from sites 13a-b, 14 and 15. In fact clusters predominantly made up of architectural ceramics and coarse ware and pithos fragments in particular are far from unusual when rural sites from the Roman Period are in question. Very similar clusters are known from intensive survey projects in various parts of Greece and Italy393. During the analysis of the results from the first survey in chapter IV, we mentioned the possibility that these are special-purpose sites lacking a full domestic assemblage. However one also has to allow for the possibility that this relative scarcity of pottery was chiefly determined by site-specific taphonomic factors. It can be argued that under certain circumstances, the collapsed roofs of buildings can effectively seal off the deposits lying on the floors allowing only a small fraction of the material to enter the surface record394. In the paragraphs that follow, we’ll see that even on sites with full domestic assemblages there are components predominantly made up of brick and tile. The cluster on grid 14 is also made distinct by its micro-location. Like the sites on grids 1 and 2, it is situated between the contour lines of 510 and 520 meters above sea level. But unlike the latter it belongs to the central sections of the survey basin, at a considerable distance from the main natural line of communication running along the top of the eastern ridge. To the north, the slopes leading up to the mountainside and the location of the modern villages are relatively steep and uncomfortable for communication. Although lying at the same height as the sites on grids 1 and 2, at this point the summit of the eastern ridge is higher and there is no immediate access to the main local road from site 14. In general sites occupying the central sections of the survey area are more isolated than those standing on the top of the eastern ridge or along its upper portions. Positioned unfavorably in the context of the local road network and lacking visual control over its surroundings, the cluster on grid 14 is surrounded by gentle and spacious fertile stretches. It occupies the southwest foot of the small fortification standing over 20 meters higher. This is the head of the surveyed basin, the point where the small valley begins to take shape. 393 J.L. Bintliff, P. Howard, A.M. Snodgrass, et al. fig. 5.1, 2007; W. Cavanagh, C. Mee and P. James, et al. 2005; C. Mee, H. Forbes, et al. 1997; T.W. Potter, The Changing Landscape of South Etruria, New York 1979. 394 J. Bintliff et al. 261, 2002. 197 Recall that because of the relative spaciousness and fertility, we predicted this and similar locations to be one of the main settlement loci in this survey area. One of the main targets of the regular grid survey on the eastern ridge was a group of field blocks along the eastern limit of the survey. A number of closely spaced field units featured high or very high overall artifact density and in this case, we were lucky that at least on half of them, the Roman material was present in on-site quantities. On the northernmost group of field blocks 1-6, the individual transect collections indicated that the zone of very high density of Roman material was even more extensive than the zone of high overall density (map V_7a-7c, tables 6-7, Appendix IV). Unfortunately at the time of the grid survey on these locations, the transect collections were yet to be analyzed and the badly worn, defaced fragments left an impression that we were dealing with a typical off-site scatter. Although far more numerous, the volume of the collection was so small and unpromising, it was decided to limit the total grid survey to field blocks 2 and 6 and part of field block 1. Some of the highest on-site densities of Roman material in this survey were recorded by the total collections on grid 11, covering field blocks 1 and 2. It reaches nearly 40 fragments per 100 sq meters and were all finds included in the analysis, it would have surely risen to over 66 fragments per 100 sq meters. Not surprisingly the average on-site density is also very high, reaching nearly 15 fragments per 100 sq meters, even when the discarded finds are excluded from the analysis. The cluster on grid 11 is more compact and considerably larger than the clusters on grids 1, 3 or 14. The revealed portion measures about 2200 sq meters, but given the fact that densities of about 10 fragments per 100 sq meters were recorded on units along the western edge of the grid, it is likely that the site spread further in this direction, over field block 1. An additional row of grid units to the north and west would have probably revealed this site in its entirety. Only about 30 meters from the northwest edge of the cluster on grid 11, on field block 6, both the transect and the grid collections revealed a larger concentration of Roman material. Although on all field blocks of this group the density of the Roman finds was above the survey’s average, there is an apparent increase on field block 6. Artifact densities close to or above the threshold of 8.3 fragments per 100 sq meters were recorded on the northern row of units on grid 5. They spread over an area of 750 sq meters. Regarding size and artifact density, the cluster on grid 5 is slightly larger than the clusters on grids 1, 3 and 14, but it lacks one of the crucial features of on-site clusters (table 7, Appendix IV). The total grid survey revealed that this cluster didn’t have a focus or a gradual transition towards the off-site. The density of Roman finds fluctuated at random on the northern row of grid units, suddenly declining to an off-site level in the southern half of the field and probably, outside its limits. As a result there emerged the possibility that the cluster on grid 5 is merely a continuation of the peripheral zone recorded on grid 11, with considerable consequences regarding the size of the cluster on grid 11. This would imply that site 11 stretched over most of field blocks 1-6 occupying an area of 5-6000 sq meters. However the detailed analysis of the transect survey results showed that these are two physically distinct clusters, separated by a zone of average artifact density that spread over field blocks 4 and 5 (map V_7c). To a certain degree this observation was confirmed by the total grid survey, though as it often happened in the Sopot survey we failed to clearly demonstrate it by merging the two grids. Finally, we should take into account the possibility that the truncated cluster on grid 5 represents the remains of a disturbed site situated on the edge between field block 6 and the neighbouring field unit to the north, 7a. As described in appendices III and IV, field units 5 and 6 198 are delimited by a very tall escarpment on the north. Faint traces of rubble on the exposed crosssection point to the possibility that the original site was cut and leveled when the escarpment was built. But this possibility hardly changes its rank and relation to the site on grid 11. Adopting the view that the clusters of Roman pottery revealed on grids 5 and 11 are physically separate, their similarity to the pair of sites discovered on grids 1 and 2 cannot escape notice. In both cases a larger and denser cluster is accompanied by much smaller and thinner satellite, situated less than 70-80 meters from the cores of the larger clusters. On the surface, the pair of clusters on grids 5 and 11 appears as a smaller, but much denser replica. As was shown however, there are certain differences between the two. The maximum artifact density on this site is two to three times the maximum density on grid 2, so far the largest Roman site in the surveyed basin. It remains unclear if this is related to the extreme fragmentation of the surface material on grid 11 or it truly reflects prolonged human occupation. The former seems likelier because even the most numerous collections weighed barely one kilogram. The maximum density of over 66 fragments per 100 sq meters was essentially produced by a hundred tiny and defaced bits of ceramics, in all likelihood fragments of brick and tile. As outside the limits of field blocks 1 and 2, both the quality and the size of the ceramic fragments increase, it was suggested that the increased density on grid 11 should be related to the fine harrowing of these fields. This extremely poor state of preservation was the chief reasons why this site was only partly documented, despite the fact that Roman finds represented over 90% of the total surface record on grid 11 (graph 3, Appendix IV). Seemingly far more notable is the difference between the fabric compositions of the two assemblages from grids 1-2 and 5-11 (cf. graphs 2 and 4, Appendix IV). In the collections from grid 11, plain fabrics represent well over 30%, while coarse fabrics are represented by less than 10% of the finds. However when the collections from grids 5 and 11 are joined into a single assemblage, its composition is very similar to the composition of the Roman assemblage from the site on grids 1 and 2. Both coarse ware and architectural ceramics figure far more prominently in the collection from grid 5. This spatial differentiation between the basic ceramic categories almost became a rule on all Roman sites in the second survey. But even when the collections from grids 5 and 11 are joined into a single assemblage, the prominence of the plain fabrics on this site cannot be denied. It was the only site in the survey area that produced a fragment of a lamp and there were a considerable number of fragments covered with a poorly preserved slip. The total grid survey failed to reveal even a small portion of the halo of site 5-11. It is nevertheless certain that it entirely covered field blocks 3-5 and spread over the neighbouring field block 7a to the north, on a higher ground (map 7a-c). Measured from the northern edge of the cluster on grid 11, it extended for over 70 meters. The collections by individual transect collections showed that the site halo also spread to the west, where it reached its maximum extent of over 100 meters and to the south where it is limited by the proximity of the cluster on grid 8. Understandably it is quite possible that the halo also spread to the east of the site area, outside the survey limits. The sites on grids 1-2 and 5-11 occupy the same type of micro-locations. They are both located on the gently sloping terraces immediately below the crest of the eastern ridge. The characteristics of the location of the site on grids 1 and 2 also apply to the locus of the site on grids 11-5. The latter is situated even closer to the top of the ridge and the main local corridor. Consequently it is less sheltered and despite its lower altitude, it has a good visual control not only over the surveyed basin, but also over the neighbouring basin to the east. 199 We mentioned that the total grid survey failed to cover all field blocks where the transect survey recorded larger concentrations of Roman material. This is particularly true for the lower terraces of the eastern ridge, where either the presence of the Roman material was realized only after the processing of the finds or it simply proved impossible to locate on-site densities when the fields were revisited for a total collection by regular grids. In these cases the collections by individual transect units present the only record we have. In trying to extract the maximum information from these data, it is important to bear in mind the inherent deficiencies of the transect survey records. The individual transect collections recorded one of the most extensive zones of very high density of Roman material. It includes field blocks 47a/b, 49-50 and 66, situated only about 7080 meters to the west and southwest of the larger cluster on grid 2 (maps V_19a-c, table 8, Appendix IV). On two of these field units, 47a and 49, the transect survey recorded artifact densities of nearly 15 fragments per 1000 sq meters. The majority of the field units where the total grid survey recorded site remains featured similar artifact densities and some (such as field block 34) featured even smaller amounts of Roman material. Only fifty meters northeast of this group of field blocks, on field block 66, the transect survey recorded an even greater concentration of Roman material. The overall artifact density on this field was average, but the density of the Roman finds was over 18 fragments per 1000 sq meters. This pattern is further enhanced if we assume that the individual transect collections included 100% of the material counted, with the difference that the concentration on field block 66 now appears thinner and more isolated. This concentration is not very far from the western edge of the site on grid 2, sharing the same terrace. Situated at about 70 meters from the latter, it is possible that it marks the remains of yet another satellite of site 1-2. Failing to locate and document these clusters using the more intensive grid survey, we can say very little about their exact size, character and location. Likewise one can only guess if they were single clusters, a combination of a larger and a smaller cluster or a series of smaller clusters. We will see that all three combinations were encountered during the grid survey. But judging by the individual transect records, this site certainly consisted of more than one separate cluster (map V_19b). On the southern group of the field blocks (47a-b) the greatest concentration of Roman finds was recorded in the southern half of field block 47a, though in all probability the cluster spread over much of this field block, the site halo spreading over field block 47b. It is also certain that the smaller concentration recorded along the boundary between field blocks 49 and 50, about 60 to 70 meters on the north formed a separate and smaller cluster. As explained before, the status of the very high concentration on field block 66 situated approximately 50 meters to the north-northwest of field block 50 remains uncertain. We are actually more inclined towards viewing it as a satellite of the site on grids 1-2. Predictably it is impossible to decide which of these were the central clusters and which were their satellites. The individual transect collections suggest that the southernmost, the one on field block 47a was slightly larger and denser than the rest. The transect survey records also roughly point to the extent of the halo surrounding the clusters on field blocks 47a-b, 49-50 and 66 (map V_19a, 19c). If the latter cluster truly belongs to the impact zone of the site on grids 1-2, then the halo of the other two clusters is mostly limited to the field blocks on their south. It extends for over 80 meters measured from the southern limit of field block 47a. Here it borders and possibly intersects with the halo surrounding the site on field blocks 289a-291a, situated at a distance of about 200 meters. To the east the zone of intermediary density extends for almost 70 meters from the edge of field block 200 47a. The low density on the neighbouring field blocks to the west and north indicates that the halo zone didn’t spread symmetrically around the site. Understandably if we interpret the concentration on field block 66 as a component of this site, on the north its impact zone will extend for at least 140 meters measured from the edge of field block 50. In such a case this will become one of the most extensive site halos in this survey area. Although not impossible this conjecture is not very likely. The fact that the remains on these field blocks practically disappeared from the surface record in the latter part of the first year’s campaign made us believe that they were small and not very substantial clusters. Taken together around 65% of the Roman finds collected by individual transects from these fields comprised brick and tile fragments (graphs 5-6, Appendix IV). There is an almost equal percentage of fine/plain and coarse pottery. Excluding the collection from field block 66 wouldn’t cause dramatic changes in the composition, though the percentage of architectural ceramics and coarse fabrics will certainly decrease. Looking at the collections from each of these field blocks separately, we observed some striking variations in the presence of the basic ceramic categories. The cluster on field blocks 49-50 was mostly made up of architectural ceramics, with pottery represented by less than 20% of the collected fragments. On the other hand, on field blocks 47a-b pottery comprises over 50% of the finds and plain fabric groups are particularly prominent. Thus as on the site on grids 5-11 the basic categories of ceramic artifacts exhibit divergent patterns of distribution. Apart from the work of certain taphonomic processes, this could reflect an original discard behavior or that the site components were foci of different types of activities. The site on field blocks 47a-b and 49-50 occupies a location similar to that of site 14. It is situated even lower on the slopes of the eastern ridge, at a considerable distance from the main local road. Measuring from the eastern limit of field block 47a, the crest of the eastern ridge lies at a distance of over 270 meters, while the difference in height between the two points is over 30 meters. In return for this relatively unfavorable positioning concerning proximity to communication and visual control, the site on field blocks 47a-b and 49-50 is in the centre of a spacious and fertile stretch. The surrounding terrain is gentle and space is not limited by escarpments as on the upper portions of the ridge. Towards the end of the second year’s campaign we accidentally discovered another Roman site on field blocks 102a/b-104a/b, by the northern survey limit. As explained in Appendix IV, at the time of the transect survey this group of fields was fallow and both the overall and the density of Roman material were very low (table 9, map V_8a, Appendix IV). However after being ploughed sometime during the second year, there emerged a substantial cluster of Roman material that was accidentally spotted during the total collections on the neighbouring field blocks to the west. This was confirmed by the total survey on grid 27, completely covering all six fields. It revealed a medium-sized cluster occupying an area of almost 1800 sq meters (map V_8b). The site is located in the western half of the grid and consists of two cores separated by a narrow strip where artifact density drops to slightly over 3 fragments per 100 sq meters. It divides the site into a northern, denser and larger core and a smaller component in the southern half of the grid. The maximum artifact density recorded on the northern component is second only to the very dense core on site 5-11. Counting the discarded material, it nearly reaches 30 fragments per 100 sq meters. The southern cluster is considerably thinner and the maximum artifact density barely exceeds 15 fragments per 100 sq meters. Because of the density fluctuations within the site limits, the average on-site density is somewhat lower or almost 9 fragments per 100 sq meters (table 10, Appendix IV). None the less 201 this is still about 2.5 times the survey average recorded by the grid survey. The fact that this relatively substantial cluster emerged on a location that appeared nearly sterile during the quantification campaign, again stresses the dynamic nature of the surface record in the second survey area. It warns us that even medium-sized clusters could have passed completely unnoticed during the quantifications campaign. Interestingly while the cluster on grid 27 remained hidden during the first year’s campaign, the thin scatter of Roman material spreading over field blocks to its south was recorded by the transect survey (maps V_8a, 8c). This scatter almost certainly represents the halo of site 27, extending over a distance of 60-70 meters measured from the southern edge of the site. The grid survey confirmed that this zone of intermediary density spreads to the east of the site for at least 50 meters, covering the entire eastern half of the gridded area. On field blocks to the north, the collections by individual transect units picked up a sparser off-site carpet. The fact that on one or two field units artifact density approached the mean district value merely reflects the relatively small size of these field units. Immediately to the west of site 27, the regular grid survey revealed an even larger site. It is thus evident that the site halo mostly spread to the east and south of the site area, over a lower ground. The composition of the assemblage from this site is very similar to the assemblage from the site on grids 1-2 or to the combined collections from field blocks 47a-b and 49-50 (graph 8, Appendix IV). Architectural ceramics represents nearly 60% of the all finds collected, while pottery fragments about 40%. Unlike the assemblage from grids 5-11, coarse ware is slightly more numerous than plain ware. Looking at the on-site distribution of the basic ceramic categories, we see once again a clear spatial differentiation (map V_9). Most of the brick and tile alongside a portion of the coarse fabric groups came from the southern component, while on the northern core plain fabrics were predominant and architectural ceramics comprised a minority in the total collections. About 60-70 meters to the west of site 27, the transect survey recorded a very high overall density on four closely spaced field units. The collections by transect units revealed that the Late Ottoman-Early Modern Period was the most numerous chronological group in the surface record on this location, although finds datable to the Roman Period also contributed to the high overall density, especially on field blocks 125, 129 and 137 (map V_8a, table 9, Appendix IV). In fact adjusting the transect collections so that they represent 100% of the material counted, there emerges a fairly compact zone of higher than average artifact density, spreading over field blocks 125, 129, 134-137 (map V_8c). Nevertheless until the total collections on grids 15-18 were carried out, we believed that the cluster of Roman material was located further west, on field block 137. The total grid survey proved otherwise (map V_8b). In fact the density of Roman finds was so low on this field, it was decided not to extend grid 18 over its entire area. The total grid survey discovered three separate clusters; two on field blocks 125, 129 and a third one on the border between field blocks 130 and 136. The southernmost component is the largest. It was located in the western half of field block 125, covered by grid 15. The maximum artifact density of about 30 fragments per 100 sq meters was recorded on 3 neighbouring grid units. The second core is situated about 25-30 meters to the northeast. It was discovered in the eastern half of field block 129, covered by grid 17. As on grid 27, this core is thinner featuring less than 20 fragments per 100 sq meters. It is also much smaller, limited to a single grid unit. Finally, the third and northernmost component is situated about 45 meters to the north-northwest, along the eastern edge of grid 18. Similarly to the central component, it featured about 20 202 fragments per 100 sq meters, but it is larger spreading over at least three neighbouring grid units. The very low average on-site densities, barely twice the survey average, reflect both the fact that the site cores were limited to very small segments of the gridded areas and that a considerable portion of the discarded finds consisted of worn Roman tile (table 10, Appendix IV). It should be noted that unlike the sites on grids 1-2 or those on field blocks 47a-b and 49-50, the three components of this site are united by a discontinuous zone of on-site density, mostly higher than 6-7 fragments per 100 sq meters. Only rarely does the artifact density drop below the threshold of 8.3 fragments per 100 sq meters within the limits of the site. Including this zone in the site area, the site on grids 15-18 equals the site on grids 1 and 2, measuring approximately 5500 sq meters (map V_10). As on grid 1, the total collections on grids 16 and 18 covered a considerable portion of the site halo. North of the site area on grid 18, it is possible to clearly follow the transition between the on-site and the off-site. The cluster situated along the eastern edge of this grid is surrounded by a narrow belt with artifact densities ranging between 2.5 and 6 fragments per 100 sq meters. After about 15-20 meters, the artifact density decreases to about 1.5-2 fragments per 100 sq meters and it stays at this level throughout the northern half of the grid. However in the northwest corner of the gridded area there is another increase, possibly indicating the presence of a satellite cluster on field blocks 137 or 140. Including the latter field unit and measuring from the northern component on grid 18, the maximum radius of the northern halo is at least 85 meters. The total collections on grid 16 showed that the western halo is far less extensive, with artifact densities dropping below 2 fragments per 100 sq meters, after 50 meters from the site edge. The transect survey collections suggest that the site halo also extends over the fields east of the site area. This group of field blocks could equally belong to the impact zone of the site on grid 27, but even if it is notionally divided between the two sites, the southern half belonging to the site on grids 15-18, the halo radius will measure over 120 meters in this direction. South of the site area there lacks a continuous carpet of intermediary density, but note the isolated peak on field block 119 featuring nearly 6 fragments per 1000 sq meters. This could very well present the remains of another small satellite, probably situated in the eastern half of the field block. The composition of the Roman assemblage from grids 15-18 repeats the pattern revealed on grids 5-11 and 27 (graph 8, map V_9). In this case the southern and central component are made up of equal quantities of architectural ceramics and plain pottery, both groups representing about 35% of the collection. However coarse fabrics are likewise present in considerable quantities, comprising about 25% of the finds. The percentage of pottery would probably decrease by a small margin were all finds included in the analysis. But in the case of this site, the “brick and tile component” was evidently on the northern core on grid 18. Here architectural ceramics comprises almost 70% of the total collections, though unlike in the collections from grid 14, both plain and coarse fabrics are present in considerable numbers. The sites on grids 15-18 and 27 occupy the head of the surveyed basin. They are situated immediately below the contour line of 500 meters above the sea. The surrounding terrain is gentler and far more spacious than on the upper portions of the eastern ridge. Although there is no running water on the surface, the high water-table allows the cultivation of garden cultures and fruits. The basin’s floor probably has the most fertile soils in the survey area. It was therefore surprising to discover the faint remains of prehistoric settlement on the eastern ridge and not on the floor of the basin. Regarding proximity to natural lines of communication and visual control over the surroundings, the locations of both sites are less favorable than the locations of their neighbours on the eastern ridge. The lower portions of the basin are still within 203 sight, but it is impossible to see beyond the summits of the ridges. Presently the SkopjeKučevište asphalt road passes only about 150 meters west of the site on grids 15-18. However this modern road is far from following the easiest natural line of communication. It climbs the relatively steep slopes of the western ridge avoiding the even steeper, northern side of the surveyed basin. As explained in the chapter describing the survey area and the broader region of Skopian Montenegro, all major local and regional roads follow the summits of the ridges rather than the narrow valley floors. At this same altitude but south of the site on grids 5-11 on the eastern ridge, the transect survey revealed an extensive zone of higher overall density spreading for nearly 200 meters along the upper portions of the ridge. As a total coverage of this entire zone would have been impractical, we aimed at covering only those field blocks with very high overall density. However the study of the individual transect collections showed that this increment in the overall artifact density was largely contributed by the off-site debris from the last couple of centuries, especially on field blocks in the eastern half of this zone. Roman material was also present, especially on field blocks in the western half, on the other side of the dirt road that connects the fields in this area with the Skopje-Pobužje asphalt road. On certain field blocks, such as field block 328b, we simply couldn’t locate any material from the Roman Period despite the fact that these finds were present among the individual transect collections (maps V_11a – 11c, tables 1112, Appendix IV). The northern half of this group of field blocks was covered by grids 8 and 4 (map V_11b). On grid 4 we revealed a larger off-site segment, similar to those recorded on grids 1 and 18. On-site densities were recorded on grid 8, covering the narrow fields on the other side of the dirt road. We believe that the grid survey revealed only a portion of this site or rather, on of its components. It was discovered on the boundary between field blocks 22a and 22b, with an evident focus on a single unit from the northern row. On this core the regular grid survey recorded almost 20 fragments per 100 sq meters, including the discarded material. On the surrounding grid units the density of Roman material suddenly drops to about 6.5 fragments per 100 sq meters, marking the edge of the site. The revealed portion of this site measures about 750 sq meters and features an average density of 9 fragments per 100 sq meters or about 2.5 times the survey average. According to the collections by individual transect units there was a similarly sized core on the neighbouring field to the north (map V_11c). It was probably situated in the northern half of field block 23a, at a distance of only 20 meters from the core on grid 8. Thus in all likelihood we are encountering a similar situation to those revealed on grids 15-18 or on field blocks 47a-b, 49-50. Naturally in the case of the site on grid 8, it remains unclear if on-site densities spread across both field blocks 22a-b and 23a-b. If the situation on field block 23a-b is comparable to that revealed on its southern neighbour, it is likely that the on-site density was limited to a smaller area and that these were two separate clusters. Analyzing the composition of the collected material from grid 8, we observed a nearly identical cluster to the one on grid 14 (graph 10, Appendix IV). It is predominantly made up of architectural ceramics, fragments of brick and tile made almost exclusively in one fabric. They constitute over 80%, while pottery fragments not more than 15% of the collections. As on grid 14, coarse pottery is more numerous than plain fabrics. In the case of the assemblage from grid 8, coarse ware is almost twice as numerous as the plain fabrics. Given the possibility that this site had another component on field block 23a, it was decided to separately analyze the transect collections from this and other field blocks surrounding the site. On all field units belonging to this group, except for field block 25a situated between 50 and 70 meters from the core on grid 8, 204 the ratio of tile to pottery was 4 to 1, identical to their ratio in the total collections. Thus even if the site extended on the neighbouring terrace to the north, it is unlikely that the composition of the assemblage would have changed dramatically. Initially the scarcity of pottery on this site was related to the possibility that this was a non-residential site, but this view wasn’t supported by the fairly large extent of the site halo (map V_11a, 11c). If the site on grid 8 was truly a non-residential site one would expect to reveal only a very narrow belt of intermediary density, spreading for not more than a few dozen meters from the site edge. But in the case of site 8, after this narrow intermediary zone marking the site periphery there spreads an extensive area with artifact density ranging between 0.65 and 1.6 fragments per 100 sq meters. This segment of the off-site was captured by the total survey on grid 4. With slight declines and peaks, it stretches for a maximum of 90 meters measured from the eastern edge of the site. The fact that it partly intersects with the halo of its southern neighbour on grids 6-7 doesn’t reduce its extent. That the site halo spreads in almost all directions from the site area is documented by the transect survey collections. It mostly spread on the field blocks to the west of the site, where the maximum halo radius was measured at about 80 meters from the site edge. To the north and south, the site on grid 8 is sandwiched between the sites on grid 5-11 and 6-7. The transect survey revealed densities higher than the district average on most of the intervening field units, but it is certain that these field blocks at least partly belong to the impact zone of the neighbouring sites. Hence it appears that the halo of site 8 mostly extends to the east and west of the site area, along the same terrace occupied by the site. Only about 60 meters south of the site on grid 8, the transect collections indicated very high density of Roman material on field block 320 (maps V_11a, 11c, tables 12-13, Appendix IV). Indeed when the transect collections are adjusted to represent 100% of the material counted, field block 320 becomes the unit with the highest density of Roman material in the second survey or over 118 fragments per 1000 sq meters. This field block also ranked very high by the overall artifact density, which was equally contributed by the material datable to the Late Ottoman-Early Modern and the Roman Period. The total survey on grid 6 documented almost the entire site area. This was a compact cluster featuring a typical concentric pattern of distribution. Artifact density gradually declines in all directions from the site core, occupying the centre of the gridded area. As usual the maximum density was limited to a single grid unit, featuring almost 23 fragments per 100 sq meters. On grid units along the site periphery, artifact density sharply decreases to between 3-4 fragments per 100 sq meters. A small portion of the site probably extended beyond the western limit of the grid, on field block 321. Allowing for wider margins along the western edge, the site on grid 6 is only slightly larger than its neighbour on grid 8, occupying not more than 800 sq meters. According to the composition of the material collected, the site on grid 6 is nearly identical to the clusters on grid 11 or on grids 15 and 17 (graph 10, Appendix IV). In the case of site 6 architectural ceramics are still the most prevalent group representing 45% of the collections, but plain pottery is only slightly less numerous with nearly 40% of all Roman finds on this grid. As on grid 11 coarse pottery is very scarce, comprising less than 10% of the collection. Concerning their composition, the assemblages collected from grids 6 and 8 are “complementary”, perhaps indicating that these two clusters were components of the same site. Admittedly the distance between the two is slightly larger than on the rest of the sites and they are separated by a wider stretch of average artifact density. Similar distances were measured between the clusters on grids 1-2 and 5-11, but in these examples, one of the clusters was many times smaller. The clusters on grids 6 and 8 on the other hand were of a roughly equal size. Even 205 if we allow that the site on grid 8 spread further north, it wouldn’t have been much larger than its neighbour on grid 6. Further confusing the interpretation, the total survey on grid 7 covering field blocks on the eastern side of the dirt road revealed an even smaller cluster, situated only about 25 meters from the eastern edge of site 6 (mapV_11b, table 12, Appendix IV). This was a tiny site, with artifact densities above the site threshold limited to three grid units. We believe that this circumstance explains why it went unnoticed during the transect survey; even the corrected transect collections don’t reveal on-site densities on the field blocks covered by grid 7 (maps V_11a, 11c, table 13, Appendix IV). As two of these units barely featured 6.5 fragments per 100 sq meters, it is likely that the site area didn’t exceed 400 sq meters. Interestingly the maximum density was slightly higher than on grid 6, featuring about 25 fragments per 100 sq meters. It is highly probable that a similarly sized cluster elevated the artifact density on the neighbouring field to the east, on field block 328b. On this field unit the collections by individual transects recorded over 20 fragments per 1000 sq meters, well above the threshold of 11.3, but upon return we discovered the field nearly sterile and it wasn’t included in the gridded area. The cluster on grid 7 was like its neighbour on grid 6 predominantly made up of pottery fragments (graph 10, map V_12, Appendix IV). Brick and tile comprised only slightly over 30% of the total collections. Representing nearly 45% of the assemblage, plain pottery is by far the most numerous group, although there are considerable quantities of coarse fabrics. It is noteworthy that in comparison to the material from grid 6, these finds were rather poorly preserved. In this aspect, they resemble the material collected from grids 11, 15 and 17. It is possible that the fairly high maximum density on grid 7 was merely reflecting the extreme fragmentation of the material. Regarding their very close proximity to each other, it is likely that the clusters on grids 6 and 7 were the two components of the same site. What remains uncertain is the status of the cluster on grid 8 and its relation to the latter pair of sites. Judging by the composition of the Roman collection from this grid, it would have neatly complemented the “pottery-based components” on grids 6 and 7. We saw a very similar pattern on a number of other sites in this survey area. The problem with this group arises from the considerable dispersion of the three clusters. The small cluster on grid 7 is almost 100 meters away from the site on grid 8. In most other examples the various components formed more compact groups, the distance between neighboring components rarely exceeding 50 meters and the intervening stretches often featuring densities higher than the threshold of 6.5 fragments per 100 sq meters. If for the moment we adopt the view that the clusters on grids 6 and 7 formed a single site that existed independently of the cluster on grid 8, the extent of its site halo is considerably limited by the proximity of the neighbouring sites. It certainly extended for between 75 and 85 meters to the north, across the field units covered by grid 4. In this direction it overlapped with the halo emanating from the site on grid 8. The extent of this zone to the east of the site is unclear, because of the problematic status of the concentration on field block 328b. If this was another component of the same site, the maximum halo radius will measure about 110 meters from the eastern edge of the cluster on grid 6 or 75 meters from the cluster on grid 7. But this field unit is situated by the eastern survey limit and it is possible that it is a component of yet another site, situated outside the survey area. The zone of intermediary density also spreads to the south of the cluster on grid 7, but in this direction it is evident that it intersected with the halo of the larger site on grid 10. The transect survey recorded artifact density close to or slightly above the district average on the field blocks to the west of the cluster on grid 6, but here too, its 206 halo zone probably overlapped with the halos of the sites on grids 8 and 10. Dividing this zone into two equal halves and assuming that the halo of site 10 spread exclusively to the south, the maximum radius of the western halo was about 55 meters long, measured from the western edge of the cluster on grid 6. Understandably if the clusters on grids 6-8 are joined into a single site, the extent of the halo zone will grow considerably. In such a case however there emerges another problem. As with the clusters on field blocks 47a-b, 49 and 50, it is impossible to decide which of the three components was the centre of this group. Measuring from the centrally positioned cluster on grid 6, the halo radius will extend for over 120 meters to the north, bordering with the halo of site 511. To the west, measured from the edge of the cluster on grid 6, the maximum halo radius will extend for nearly 160 meters, making it one of the most extensive site halos in the second survey. Given the fact that the combined areas of the three clusters barely exceed 2000 sq meters, it is unlikely that their halo was more extensive than the halos of the sites on grids 1-2 or 15-18. Recall that these two sites were more than two times larger. Predictably taking the cluster on grid 8 as the centre of this group will extend the halo radius even further. Its maximum extent in a southeast direction measured from the edge of the cluster on grid 8 will reach almost 200 meters. Mainly because of the extensive zone of intermediary density surrounding the cluster on grid 8, we maintain the view that this was a separate residential site. These three clusters of Roman material, each presenting a differently composed assemblage are situated around the contour line of 490 meters above sea level. They occupy slightly lower ground than the cluster on grids 15-18, but essentially sit on the same terrace. Naturally the sites uncovered on grids 6-8 are much closer to the crest of the eastern ridge, lying at a distance of only 120 to 140 meters. Their locations are analogous to the locations occupied by the Roman sites on grids 1 and 2, 5 and 11. These are the upper terraces of the eastern ridge, gently sloping in a southwest direction. In fact looking more carefully at the topographic map, one notes that this group of sites is located on the top of a low off-shoot of the eastern ridge, enclosing the surveyed basin from the southeast. It seems as if the low depression in the central part of the eastern ridge was deliberately avoided, although we saw that there are Roman sites in this section that the total grid survey failed to locate. As explained earlier the advantage of the upper portions of the ridges as settlement locations mainly lies in their proximity to the natural lines of communication. The tendency to occupy the upper portions of the basin could have also been guided by the desire to occupy less of the most fertile fields in the surroundings395. As explained in appendices III and IV, the zone of high overall density extends for a short distance to the south along the low off-shoot of the eastern ridge and then continues for over 400 meters in a westward direction, across the slopes of the eastern ridge. After the study of the individual transect collections, it became clear that much of the material from this part of the survey area consisted of Late Ottoman-Early Modern off-site debris, although on a number of field units there were evident concentrations of Roman material. It has to be stressed that even on these field units, the artifact density recorded by the transect survey barely exceeded the thresholds of 11.3 fragments per 1000 sq meters or 38 fragments per 1000 for the adjusted transect collections (tables 14-16, Appendix IV). Unfortunately there wasn’t much overlap between the layers of the Late Ottoman-Early Modern and the Roman material and the total grid survey included only a portion of the field units featuring higher densities of Roman material (cf. maps V_13a/13c and 13b). 395 Cf. R. Shiel, A. Stewart, 95-109, J. Bintliff, P. Howard, A. Snodgrass et al, 2007. 207 In fact definitive traces of Roman occupation were only discovered on field blocks 332 and 333a covered by grid 10 (table 15, Appendix IV, map V_13b). On these fields the total collections by regular grids revealed a substantial and a fairly compact cluster. The core of the site with maximum artifact densities of over 30 fragments per 100 sq meters was located in the western half of field block 333a. It is limited to a single grid unit. On the rest of the site area, we recorded lower artifact densities. A thinner carpet of Roman material featuring about 10 fragments per 100 sq meters extends for about 60 meters from the site core, mostly on its west over field block 333a. On the opposite eastern side, the density of Roman material sharply decreases to less than 7 fragments per 100 sq meters. The eastern edge of the site was probably left out of the gridded area. The cluster on grid 10 measures about 2000 sq meters. It has a pronouncedly elongated shape and unusually for its size, it is very compact with average artifact densities of nearly 10 fragments per 100 sq meters. The total survey on grid 9 located immediately to the west of grid 10 confirmed the western extent of this site (map V_13b). At a distance of 40 meters from the site’s western edge, the Roman material almost completely disappears from the surface record (graph 11, Appendix IV). As the cluster on grid 6 is situated only about 50 meters to the north of the edge of site 10, it is evident that the site halo spread mostly to the east and south of the site area. According to the individual transect collections there is a wider zone of intermediary density covering the lower terraces, south of the site. It extends for over 250 meters along the longer west-east axis, but measured from the southern edge of the site area, the maximum halo radius is about 80 meters long. As on a number of other medium or large-sized sites in this survey area, there is at least one possible satellite cluster situated towards the edge of the halo zone. In this case, its presence is probably indicated by the elevated density on field block 342 situated about 50 meters to the south of the site limit. However when the individual transect collections are adjusted to represent 100% of the material counted, we see a considerable increase on all field units covering the terrace to the south of site 10, while field block 342 is shifted to the rank of very high artifact density, characteristic for field blocks covering genuine site areas (map V_13c). It is thus quite probable that we are dealing with another small to medium-sized residential site on this field unit, tying onto the chain of sites on grids 6-8 and 10. In such a case, the southern halo of site 10 is less extensive, probably limited to the first pair of field units to the south. The eastern halo of this site is larger, including a possible satellite on its periphery, on field block 336, at a distance of over 100 meters from the eastern edge of site 10 (maps V_13a, 13c). Here the collections by individual transect units recorded densities close to the site threshold, though lower than on field blocks 320 or 342. Like the concentration on field block 328b, this possible cluster is situated near the survey’s eastern limit and at a considerable distance from the site on grid 10. Therefore one cannot exclude the possibility that it gravitated towards an unknown site outside the survey’s eastern limit. But in view of the low artifact density recorded on the intervening stretch between field blocks 336 and the eastern survey limit, this isn’t a very likely interpretation. We believe that the concentration on field block 336 is a satellite of the site on grid 10. Thus the halo of this site reaches its maximum radius in an eastern direction, measuring almost 130 meters from the edge of the site area. Like its northern neighbours on grids 6-8, the greater portion of the site halo is spread over the same terrace occupied by the site. Concerning the presence and the distribution of the basic categories of ceramic finds, the collection from grid 10 is similar to the majority of on-site collections in the second survey (graph 12, Appendix IV). Nevertheless there are slight variations and in more specific terms, the collection from grid 10 stands in-between the collections from grids 1-2 or 27 and those from 208 grids 6, 7 or 15. Architectural ceramics is the predominant category, but it barely represents 50% of the assemblage including the discarded material. Typically for the majority of the sites along the upper portions of the ridge, plain pottery comprises over 30% of the material, but coarse fabrics are also well represented by nearly 20% of the finds collected. As on most of its neighbours, the pottery and the architectural ceramics exhibit divergent patterns of distribution. In the case of site 10, the “brick and tile component” was situated in the western half of the grid, while most of the plain pottery came from the eastern half and specifically from the site core (map V_12). The cluster on grid 10 is part of the chain of Roman sites occupying the low off-shoot of the eastern ridge that marks the southeast limit of the survey area. These are the small clusters uncovered on grids 6-8. The cluster on grid 10 occupies a slightly lower terrace, but this is essentially the same micro-location. The contour line of 480 meters above sea level marks the southern limit of the site area. This location still offers a good visual control over the lower portions of the basin and easy communication with the fields and the main local road on the crest of the eastern ridge. Going towards the central axis of the basin, it is evident that Roman sites become scarcer, at least along the central and southern sections of the survey area. Further west along the contour lines of 470 and 480 meters above the sea, the total grid surveys didn’t reveal definite traces of Roman sites. Targeting the field blocks that featured very high overall artifact density, the total survey on grids 12 and 13 missed the major concentration of Roman material in this part of the survey area (maps V_13b, 13d). After the study of the collections by individual transect units, it became evident that larger quantities of Roman material were to be found on field blocks 289a-291a, immediately north of grid 12. Field blocks 289a and 290a situated almost 300 meters west-northwest of the cluster on grid 10 and 190 meters south of the site on field blocks 47a-b and 49-50, both featured slightly over 12 fragments per 1000 sq meters (maps V_13a, 13d, 20a, table 17, Appendix IV). This cluster almost certainly spreads further north and west over parts of field block 291a and 303b, where the collections by transect units were less thorough. Analyzing the distribution of the Roman material by individual transect collections, it becomes evident that the cluster consists of at least two, possibly three separate components. One is located in the northern half of field block 289a and possibly spreading over field block 303b, with a second smaller cluster in the northern half of the neighbouring field block 290a, about 30 meters to the north. If there truly was a third northernmost component, it was probably located in the central part of field block 291a, approximately 25 meters to the northeast of field block 290a. As on its northern neighbour on field blocks 47a-b and 49-50, the collections from each of the components present a differently composed assemblage (graph 14, Appendix IV). The percentage of brick and tile gradually decreases from nearly 80% in the collections from field block 289a, to less than 50% on field block 291a. Apparently the “brick and tile component” was located in the southern half of the site area. The category of plain fabrics exhibits a nearly inverse distribution. It is virtually absent on the southern component on field block 289a, its share increasing to 10% on field block 290a in the centre of the site area and to over 20% on the northernmost component. The percentage of coarse ware is much more stable, barely increasing from less than 20% on field block 289a to 23% of the collections from field block 291a. Taken together, the collections from these three field blocks constitute an assemblage not much different than those collected from the majority of the Roman sites in this survey area. The only point of significant difference is the ratio of coarse to plain pottery. In the case of the elusive site on field blocks 289a-290a, coarse fabrics are more than twice as numerous as the plain fabric 209 groups. On most other sites of the Roman Period in the second survey plain pottery was predominant, often comprising over 30% of the collections. The presence of a full residential site on field blocks 289a-291a explains the slightly ambiguous results of the total survey on grids 12 and 13. These two grids were located within or just outside the limits of the site halo and the average artifact densities never exceeded the limit of 1 fragment per 100 sq meters (table 15, Appendix IV). The small but sudden increase in the density of Roman material on grid 12 has to be related to the halo zone generated by the site on field blocks 289a-291a. Grid 12 is located less than 20 meters to the south of the site and partly covers the 70-80 meters wide carpet of intermediary density, possibly stretching for over 110 meters to the south of field block 289a (maps V_13d and 20b). The collections by individual transect units indicate a possible satellite cluster on field block 284, immediately to the south of grid 12. East of the site area, it seems that the site halo is limited to the neighbouring pair of field blocks 294 and 303b, where the transect survey recorded artifact densities only slightly higher than the sector’s average. But when the transect collections are adjusted to represent 100% of the material counted, there emerges a more extensive zone of average and higher than average artifact density. It spreads over field blocks 294, 295, 295’ and 296, but not beyond the eastern limit of field block 303b. This is also confirmed by the results of the total survey on grid 13, 4050 meters southeast of field block 289a. Here Roman material was absent on all but a single unit in the centre of the gridded area. We believe that similarly sized, satellite clusters produced the increased artifact density on field block 284, situated 70 to 80 meters to the south of the site area. Note that these tiny clusters are always located at the edge of the halo zone. Further to the south and east there spreads a larger zone of low artifact density. Although the artifact densities on these satellites don’t exceed the site threshold, the increase from the neighbouring units is considerable. This is most clearly illustrated on grid 13, where all of the sudden, the density of Roman material increases to 6.5 fragments per 100 sq meters, nearly reaching on-site densities. Was this increment a part of a wider zone of intermediary or low artifact density, it would have been interpreted as an anomalous fluctuation in the off-site record. But in this particular case, the virtual absence of Roman finds on the surrounding grid units points to the possibility that this is a special purpose site located outside the zone of intense manure. This seems to be a plausible interpretation, especially in the light of similar phenomena observed in the hinterland of Thespiai and elsewhere in Boeotia396. However one needs to be cautious when interpreting this scanty evidence, for the collections from grid 13 remains an isolated example. In addition pottery from Roman necropoleis doesn’t exhibit features that distinguish it from the material found on settlement sites397. Given that this reading of the transect survey record is correct, the site on field blocks 289a-291a has a fairly extensive halo, stretching for about 80 meters to the south and over 100 meters to the east of the site area. According to the individual transect collections, north of the site area, on the terraces above the site, the halo of this site was only slightly more contracted (maps V_13d, 20b). In this direction there is a continuous carpet of average density of Roman material stretching between the sites on field blocks 289a-291a and 47a-b, 49-50. Assuming that these two neighbouring sites were of equal rank, the maximum halo radius was not longer than 90 meters, measured from the northern edge of field block 291a. The transect survey clearly demonstrates that the halo zone didn’t spread to the west of the site area or was limited to a very narrow belt along the site periphery. 396 397 J.L. Bintliff, P. Howard, A.M. Snodgrass et al. 2007, LSE4, THS1, THS11. J. Bintliff, 44-52, eds. F. Vermeulen, G-J. Burgers, S. Keay, C. Corsi, 2012. 210 Judging by the considerable extent of the halo zone, the concentration of Roman material on field blocks 289a-291a signals the presence of at least a medium-sized cluster. It occupies the foot of the eastern ridge and it is closer to the Skopje-Kučevište asphalt road than to the summit of the eastern ridge. Concerning proximity to the local road network and visual control over the surroundings, it is in a less favorable position than its northern neighbour on field blocks 47a-b, 49-50 or the sites on grids 15-18 and 27. We’ll repeat that the only advantages of this type of locations are the absence of physical barriers and the immediate access to fertile soils. Roman finds were completely absent among the transect collections from field block 263 (maps V_14a, 14c, table 18, Appendix IV). The very high overall density on this field unit can wholly be attributed to the sudden increase in the amount of the Late Ottoman-Early Modern material. However the total survey on grid 19 covering field block 263 in its entirety yielded a small collection of badly worn fragments datable to the Roman Period (map V_14b, table 19, Appendix IV). After the processing of the finds it became clear that this was a slightly larger collection with a maximum density of 8.3 fragments per 100 sq meters. Recall that similar artifact densities were recorded on the periphery of site areas. Moreover this small cluster exhibited a perfectly concentric pattern, typically seen on small compact sites, such as those on grids 1 or 7. As discussed in Appendix IV this small collection wasn’t assigned an on-site status, because of the relatively low artifact density and the extreme fragmentation of the collected finds. The average density on this grid is well below the survey average and equals the density recorded on grid 12. In this context, one wonders if the increased artifact density on this field isn’t a direct result of the poor state of preservation of the material. But relegating the collection from grid 19 to an off-site level doesn’t entirely solve the problem of its sudden appearance in this survey section. Even if we agree that this was a part of the off-site carpet, it is strange that there are no major sites in its vicinity. The nearest known sites are situated at distances of almost 300 meters. As we saw densities higher than 2-3 fragments per 100 sq meters were normally recorded in the halo zone, close to residential sites. The last Roman site in the central survey section was discovered during the transect survey on field block 277b (map V_14a, table 18, Appendix IV). Being immediately recognized as a discrete site, the collections by individual field walking transects were less intensive, but we were able to locate and record the full extent of this site during the regular grid survey. Once the individual transect collections are adjusted to represent 100% of the material counted, this bias is unmasked and field block 277b joins the zone of higher than average artifact density (map V_14c). Similar or even higher densities are predicted for field blocks 270a-271, about 370 meters to the north and for field block 394, situated about 30 meters to the south. Because these field units are surrounded by extensive stretches of low artifact density (i.e. there lack traces of the site halos), we believe that these are small concentrations, possibly representing nonresidential remains. In terms of size and character they are probably comparable to the site revealed on grid 20. This site was located in the southwest corner of field block 277b, partly spreading to the north over the neighbouring field block 276b (map V_14b). According to the total survey on grid 20 this was one of the smallest, but at the same time one of the densest clusters of Roman material in the second survey (table 19, Appendix IV). Including the discarded finds, the grid survey recorded a maximum density of nearly 80 fragments per 100 sq meters in the southwest corner of the grid. Including this material, the average on-site density could increase to over 20 fragments per 100 sq meters. Outside the site core to the north and east, the density of Roman finds sharply decreases to an off-site level. On the eastern side the decrease is slightly gentler 211 and it is possible that the site area extended slightly beyond the grid limit. It is also possible that parts of the site remain hidden beneath the field hedges or the Skopje-Kučevište road. But in all likelihood the actual site area wasn’t much larger than the portion revealed by the grid survey, measuring roughly 600 sq meters. Regarding the composition of the collected finds, the cluster on grid 20 belongs to the group of clusters predominantly made up of brick and tile (graph 16, Appendix IV). This category comprises over 90% of the assemblage, while plain pottery is only slightly more numerous than coarse ware. In the case of this site, the predominance of architectural ceramics is even more pronounced than on the sites on grids 8 or 14. Predictably because of the small size and fairly compact character of the site area, there are no obvious divergences in the distribution of tile and pottery (map V_15). The cluster on grid 20 along with the one on grid 3 and the hypothetical clusters on field blocks 270a-271 and 394 were the only Roman sites in this survey area that weren’t surrounded by a more extensive zone of intermediary artifact density. The collections by transect units on the surrounding field blocks included at least 40% of the material counted, but except for one field unit the density of Roman material never exceeded 2.5 fragments per 1000 sq meters. This sudden decrease is also indicated by the results of the regular grid survey. Artifact density drops below 1 fragment per 100 sq meters on the northernmost grid unit, a bare 30 meters from the site core. It has to be pointed out though that grid 20 failed to cover the very edge of the site area on the eastern side. The existence of a more extensive halo to the south of the site shouldn’t be excluded, although this is not indicated by the individual transect collections. The cluster of Roman finds on grid 20 occupies the lowest location among the Roman sites in the survey area. It is situated just below the contour line of 460 meters above the sea, in the central part of the surveyed basin. At this altitude the small stream begins to take shape and the first longer waterlogged stretches appear. Like the clusters uncovered on grids 15-18 or 27, the site on grid 20 has access to the most fertile portion of the basin. At present most of this zone is under gardens and orchards. Understandably there is no visual communication with the upper slopes and the crests of the ridges. The gentle relief of the survey area makes this distinction less significant, but it is still much more difficult to travel across the slopes than to follow the crests of the ridges. This feature of the studied landscape is presently obscured by the Skopje-Kučevište asphalt road following the central axis of the basin. Observed in relation to the rest of the clusters of Roman material, the site on grid 20 is truly one of the most isolated. The nearest possible Roman site with a residential character is situated 300 meters to the northeast, on field blocks 289a-291a north of grid 12. The next clusters in terms of proximity are the one on grids 15-18, 600 meters to the north and the cluster on grid 26, about 500 meters to the east by the southeast boundary of the survey area. In comparison, the rest of the Roman clusters had their nearest neighbours at distances not greater than 150-200 meters. The southernmost of the series of sites that occupy the eastern ridge or rather its low offshoot was discovered on field blocks 348 and 351 by the eastern survey limit, 150 meters to the south of grid 10. Because of the low obtrusiveness of the material and the relatively low artifact density, this cluster came to light only after the processing of the transect collections (maps V_16a, 16c, table 20, Appendix IV). The scarcity of the material datable to the last two centuries also contributed to the somewhat lower overall densities. We nevertheless managed to include field blocks 348 and 351 in the total grid survey, just before the end of the second year’s campaign. Grid 26 covered most of the site area, although the southern and western peripheries of the site remained outside the gridded area (map V_16b). On the east, Roman material 212 disappeared from the surface record on the fields on the other side of the dirt road and beyond the eastern survey limit. The total grid survey showed that on this side too, on-site densities extended slightly beyond the survey area. Artifact densities are relatively low, typical for the small satellite clusters. The average on-site density is fairly modest, with slightly over 7 fragments per 100 sq meters, though the true density is probably over 10 fragments per 100 sq meters (table 19, Appendix IV) Across the site area, artifact densities range from 10 and 15 fragments per 100 sq meters in the centre of the grid, to between 5.5 and 10 fragments per 100 sq meters on the peripheral units south and west of the core. It is therefore unlikely that the site spread for a very long distance beyond the grid limits. Allowing for wider margins along the southern and the western sides and excluding the possible eastern margin, the site on grid 26 measures almost 2200 sq meters. On the northern row of grid units artifact density drops below 3 fragments per 100 sq meters, clearly marking the site limit on this side. It also indicates that the site halo didn’t spread to the north of the site area. This was confirmed by the individual transect collections on the field blocks to the north of the site. A 30 to 40 meters wide belt of low artifact density separates the halo of site 26 from that of site 10. Typically for most on-site collections from the southern half of the eastern ridge, pottery is more numerous than architectural ceramics. However if the discarded material is included in the analysis, brick and tile will represent nearly 55% of the collections, while the share of plain pottery will drop to about 30% (graph 16, Appendix IV). Coarse fabric groups are represented by about 20% of the finds and 15% when all material is included. This is nearly identical to the composition of the assemblage from the neighbouring site on grid 10. Despite the fact that a considerable portion of the site area was revealed, there is no spatial differentiation between the basic categories of ceramic artifacts (map V_17). As already explained, the cluster of Roman finds uncovered on grid 26 is the southernmost of the chain that covers the low off-shoot of the eastern ridge. It begins with the group of sites uncovered on grids 6-8, approximately 300 meters northeast of grid 26. They all occupy very similar locations to those of the sites uncovered on grids 1-3, 5 and 11, the latter group being situated on the upper slopes or on the very top of the eastern ridge. Although positioned at only a slightly higher altitude than the cluster on grid 20, the Roman site on grid 26 has a much better visual control over the lower sections of the basin and offers a much easier access to the main local roads. According to the individual transect collections, the halo of the site on grid 26 spread mostly to the south and west of the site area (maps V_16a and 16c, table 20, Appendix IV). As these are rather low densities, one cannot be sure if this zone spread east of the site, beyond the survey limits. We saw in the preceding paragraphs that the halo zones rarely spread on more than two sides of the site areas. The halo of site 26 is especially extensive on field blocks west of the site and on the same terrace delimited by the lines of 460 and 470 meters above sea level. Measured from the southwest corner of the site, the maximum halo radius reaches almost 160 meters, extending to the western edge of field block 357. In fact the zone of between 2.5 and 5.1 fragments per 1000 sq meters patchily extends for hundreds of meters further west and on field blocks 314-315, artifact density increases to over 8 fragments per 1000 sq meters. It seems unlikely that this entire zone, spreading over an area of several hectares belongs to the halo of site 26 or to its neighbour on grid 10. Because of the observed disproportion between the areas of the sites and their halos, we suspect that there exists another residential site on the fields south of grid 10 and west of grid 26. The adjusted record of the transect collections points to field block 342 as a likely site location (map V_16c). The density predicted for this 213 field block exceeds 40 fragments per 1000 sq meters, while on all neighbouring field blocks it ranges between 7 and 15 fragments per 1000 sq meters. Thus we have the recognizable concentric pattern of one or two contingent field units with very high artifact density, surrounded by a more extensive zone of intermediary density. This same pattern is visible in the record of the unadjusted transect collections, but now the picture is crystallized and we see a definite concentration on field block 342. The location of this hypothetical site is a near replica of its neighbour on grid 10. It occupies a lower terrace on the eastern ridge and it is connected with the local road-network via the summit of the low ridge that delimits the survey area from the southeast. Even the dispersal of its halo imitate the halo of site 10, chiefly extending to the east and south of the site area. One can only speculate about the size of the area occupied by this site. None the less it is possible to infer a few reasonable conclusions. It is evident that this was a single-core site. On sites with multiple-cores, the high concentration regularly spread over more than one field unit. In fact only the very small, single-core sites, like those uncovered on grids 1, 3 or 20 were encompassed within the limits of a single field block. This implies that the possible site on field block 342 couldn’t have been much larger. The analysis of the distribution of the finds collected by individual field walking transects also showed that the main concentration was recorded on the field block’s central sections. On the basis of this record we can argue that this site measured less than 1000 sq meters. In this view its site halo is fairly large, for it could extend for over 100 meters to the east and south of the site area. As argued in the appendix, this still leaves a considerable room for the halos of sites 26 and 10. The lengths of their maximum radii merely become more proportional to their rank and size. V.3.3 The off-site zone According to the collections by individual transect units (on average, more thoroughly executed than the transect collections east of the Skopje-Kučevište road), there are no traces of Roman settlement on the western ridge. Only on a small group of field blocks in the northern end of the ridge and opposite the site on grids 15-18 does the artifact density exceed the limit of 2.5 fragments per 1000 sq meters (map V_18a, table 22a, Appendix IV). On the rest of the field blocks the Roman material was completely absent from the transect collections. In fact if the transect collections are adjusted to represent 100% of the material counted, even these group of field blocks join the zone of lower than average artifact density, further enhancing the contrast between the western and the eastern ridge (map V_18c). It is unfortunate that all but one of the grid surveys on the western ridge covered field units where the transect survey records showed near or complete absence of Roman finds (map V_18b). Nevertheless in nearly all grid collections there were small amounts of Roman material (table 22b, Appendix IV). This implies that the off-site carpet of Roman finds continues to spread over the western ridge of the survey area. The maximum density recorded on grid units on the western ridge is around 2 fragments per 100 sq meters, while the average densities on gridded areas range between less than 1 and 3.6 fragments per 1000 sq meters. These tiny collections were usually gathered from smaller groups of contingent grid units. Roman material was absent from the greater portion of the gridded areas. In this respect the off-site carpet on the western ridge strongly resembles the distribution in the off-site zone revealed in the Sopot survey. Being found alongside much larger quantities of Late Ottoman-Early Modern or 214 prehistoric material, it’s hardly a surprise that these scatters came to light only after the processing of the material. Looking at the distribution of the Roman finds on grids 21 and 23, one sees a handful of finds dispersed along one or two neighbouring rows of units and forming ultra-thin carpets of less than 1 fragment per 100 sq meters (maps V_18b, 18d). The only exception was grid 25, covering field blocks 170 and 171 in the northern end of the western ridge. Here the thin carpet of Roman finds covers the entire grid giving an average density of 3.6 fragments per 1000 sq meters. This confirms the results of the transect survey, which recorded slightly higher artifact densities on these two field units. This record also demonstrates that even very small amounts of Roman material could increase the densities recorded by the transect survey to over 4 fragments per 1000 sq meters. This is especially the case on field units with smaller areas and on which the collections by transect units were carried out more thoroughly. It warns us against overestimating the true amounts of Roman material on some of the field units that were left out of the grid survey and on which the transect survey recorded densities of 5-6 Roman shards per 1000 sq meters. Nevertheless the total survey on grid 25 confirmed an increase in the amount of Roman material along a south-north access, albeit a rather slight one. This could be related to the denser off-site carpet on field blocks surrounding the site on grids 15-18, but it could equally spread from a site situated beyond the survey’s northern limit. Note that the majority of the field units with densities of Roman material approaching the survey’s average are located along the lower terraces of the western ridge. At present the western ridge is artificially separated from the eastern ridge by a tall escarpment cut during the construction of the Skopje-Kučevište asphalt road. Originally the slopes of the two ridges must have merged more gently around the head of the shallow basin. The percentage of Roman material in the transect and grid collections from the western ridge is tiny (graph 17, Appendix IV). It is far lower even in comparison to the grid collections covering the off-site zone on the eastern ridge. Finds datable to the Roman Period never exceed 4% of the collections excluding discarded material, which in the case of the western ridge consisted almost exclusively of the Late Ottoman-Early Modern debris. This is in itself a clear indicator that there are no on-site densities on the western ridge, including its northern end. As explained in the section dealing with the remains from the prehistoric periods, despite the evident similarities concerning micro-topography and proximity to agricultural resources, the western ridge is a less favorable settlement location. In particular its eastern slopes are relatively steep and hardly inhabitable. The summit of the ridge and the western slope communicate only with the floor of the neighbouring basin to the west and with the village Kučevište. Because of this circumstance the western ridge doesn’t have a direct access to the main local road connecting the foothills of Mt. Montenegro with the Skopje Basin. In addition it seems that the western half of the survey is drier and at slightly greater distances from the nearest known freshwater springs. The analysis of the Roman assemblages from the Sopot survey showed fairly consistent differences in the composition of on-site and off-site collections. This wasn’t the case with the Roman material from the second survey, at least in its eastern half. Further confusing the distinction between the site and the off-site, the composition of small off-site collections was often rather similar to the typical domestic assemblages. In fact the problem arises from the considerable differences between the off-site collections from various sections of the eastern ridge (graph V_2). In some collections such as those from grids 4 and 9, architectural ceramics was by far the most predominant category, comprising almost 70% of the assemblage. In others, 215 such as the collections from grids 12, 16 and 19, the presence of brick and tile is far less prominent and in contrast to the collections from grids 4 and 9 there is a significant percentage of coarse ware. As discussed in Appendix IV, these variations in the composition of the off-site cannot be related to the micro-location or to the proximity of on-site clusters. Although grid 9 was located in the immediate vicinity of a site, it yielded a very different collection from those that came from the identically positioned grids 12 and 16. Similar variations were observed in the off-site collections from field blocks 294 and 303b, both situated to the west of the site on field blocks 289a-291a (graph 14, Appendix IV). Graph V_2: Composition of the total collections from the off-site and the halo zone 100% 80% Unid 60% Plain 40% Coarse 20% Tile 0% Grid 9 Grid 4 Grid 12 Grid 19 Grid 16 We have a nearly identical situation on the western ridge. Because of the small number of finds from this survey section, we analyzed the integral collections by grid and transect units. As shown on graph 18 in Appendix IV, although covering the same portions of the western ridge, there were considerable differences between the compositions of the grid and the transect collections. Thus the composition of the off-site material changes not only across various portions of the survey area, but even with changing collection strategies. Surprisingly enough the transect collections included a higher percentage of pottery and less architectural ceramics. This is contrary to what we expected assuming that brick and tile are more obtrusive than pottery fragments. Analyzing in greater details the transect collections from the site on field blocks 289a-291a and its surroundings, we predicted that architectural ceramics will always be overrepresented in the transect collections. It was thought that the more sensitive grid collections will include a larger number of pottery fragments. But the comparison of the transect and the grid collections from the western survey section showed that this wasn’t necessarily true. In the appendix we also examined the possibility that the inconsistencies in the composition of the off-site record could result from the fact that the analyzed grid collections were often small and limited to very small segments of the off-site. However the analysis of the collections by individual field walking transects from the southern and eastern halves of the eastern ridge (excluding field blocks covering certain or possible site areas) showed that the variations persist regardless of the representatives of the sample. For reasons we still cannot understand, there is a dichotomy between the composition of the transect collections from the southern and northern half of the eastern ridge. In the transect collections from the southern half, pottery is more numerous than tile and plain fabrics are more prominent than coarse ware; in collections from the northern half tile is far more numerous than plain pottery, while the percentage of coarse ware remains stable. Apparently the only more or less consistent feature of the off-site collections in the second survey is the stronger presence of coarse pottery, which is in a complete contrast to the 216 composition of the off-site collections from Sopot. Larger quantities of coarse ware were discovered in the majority of the collections regardless of their location and the collection strategy. As explained earlier, among the Roman material from Skopian Montenegro it was possible to define a number of different coarse fabric groups. On the other hand, only a few fabric groups were recognized as coarse ware among the Roman material from Sopot. It is possible that some of the fabric groups from the Sopot material were erroneously classed as plain pottery, while plain fabric groups from Skopian Montenegro were treated as coarse ware. However we believe that the classification of the material in both survey areas was fairly consistent and that the increased presence of coarse fabrics in the second survey area has to do with the local traditions of pottery production and the different character of the local raw materials. As explained in an earlier paragraph, the distinction between coarse and plain pottery was made on the basis of the formal fabric features and it doesn’t imply a functional distinction. Comparing the off-site records on the western and eastern ridge, the differences are more than apparent. On the eastern ridge the off-site carpet is far more compact covering most of the survey area east of the Skopje-Kučevište road. Sterile stretches are fairly limited. Moreover the eastern ridge features a much denser off-site carpet. As shown on grids 1 or 18, artifact densities of over 1 fragment per 100 sq meters can continuously cover fairly extensive sections. Indeed on certain grids covering the off-site zone close to sites (grid 1 or 12), the maximum artifact density could reach up to 6-7 fragments per 100 sq meters, approaching the site threshold. This gradual dissipation of the surface material outwards from the site area explains the difficulties in determining the site limits. Small but sudden peaks in the nearer off-site zone are sometimes difficult to distinguish from separate sites of a minor size or the so called satellites. The western ridge on the other hand lacks a continuous carpet of Roman material. As in Sopot there are small and very thin scatters separated by large zones where the Roman finds are completely absent from the surface record. However looking at the integral distribution of the Roman material in the second survey area, it is clear that even on the eastern ridge it doesn’t form a continuous carpet with even artifact densities. Most of the clusters featuring on-site densities are surrounded by roughly concentric zones of decreasing artifact densities. It is as if the on-site distribution pattern is repeated on a larger scale. Most of the Roman sites and particularly the clusters featuring full domestic ceramic assemblages have generated their own off-site zone, sometimes intersecting with the off-site zones of their neighbours, sometimes separated by sterile or nearly sterile stretches398. The latter are typified by the findings on grids 9 and 13 and on the grid surveys on the western ridge. On the other hand, the off-site zone surrounding the site areas was partly captured on grids 1, 4, 12, 18 and 27. They are completely covered with a thin carpet of Roman material featuring between 0.6 and 2.3 fragments per 100 sq meters, but on certain locations artifact densities could suddenly increase to 6-7 fragments per 100 sq meters. Such fluctuations are unparalleled in the farther off-site zone, where the average artifact densities barely exceed 2-3 fragments per 1000 sq meters. This denser off-site zone marks the immediate surroundings of the site, its impact zone or site halo. It could be generated by the regular bringing of manure on the most intensively cultivated fields, but also by remains of less intensive, non-residential activities. As was shown on grids 1-2 or on grid 18, larger sites are sometimes accompanied by smaller satellite clusters, barely emerging from the off-site carpet surrounding the site. Because of their low prominence in the surface record they rarely came to light prior to the processing of the finds. As they usually appear at distances not greater than 100 398 Cf. J. Bintliff, 29-38, 2005. 217 meters from the sites within the nearer off-site, it is possible that they represent outbuildings or remains of other non-residential activities. How does this distribution in the off-site zone compare to the situation revealed in the Sopot survey? Outside the site area on almost all gridded sites, it is possible to observe a narrow belt of intermediary density, extending for not more than a dozen to 20 meters from the edge of the site. This phenomenon was observed in both survey areas and regardless of the time-period. More extensive carpets of average or low artifact density were only typical for a few historic periods. But although in general there lacks the evidence of continuous carpets of surface artifacts for the prehistoric periods in the Sopot survey, we often documented small concentrations of finds usually limited to the same topographic units occupied by the settlement. We believe that these scatters extending for up to 300-400 meters from the central site have their analogy among the halos of Roman sites in Skopian Montenegro. The fact that they are less extensive than the impact zone of prehistoric sites in the first survey is in accord with their smaller size399. In the case of Skopian Montenegro in the Roman period, there clearly lacks a continuous off-site carpet emanating from a single residential centre. Rather each of the discovered sites has generated its own halo spread over several field blocks (1-2 hectares), not necessarily symmetric to the site. But the survey also identified a very thin, discontinuous carpet spread over the entire survey area that couldn’t be related to any particular site. This was sometimes termed the “farther off-site”, although it would be less confusing to simply term this phenomenon the off-site and use the term halo for the zone of intermediary density spreading for over 100 meters from the site edges. Seemingly the distribution of Roman material in the off-site zone of the first survey doesn’t exhibit a finer structuring. But in retrospective in Sopot too, it is possible to observe an increased artifact density and satellite clusters around the central settlement on site 5a-5b and an ultra-thin off-site carpet, occupying the entire western half of the survey area. V.3.3 Analysis of the integral network of Roman sites and land-use Because of the peculiar distribution patterns in both the on-site and the off-site zones, it is difficult to answer even the seemingly simple question of how many Roman sites are there in the second survey. The grid surveys fully or partly uncovered 14 certain clusters of Roman material, while the transect survey indicated at least three other potential clusters, which the total grid survey failed to locate. In addition there were a number of lower density peaks that were interpreted as satellite clusters. These were all spatially distinct clusters, separated by zones of low or average artifact densities. They stood apart from their surroundings only thanks to the increased quantities of surface material. However the differences could vary considerably, which also proved problematic for the interpretation of the survey findings. On some clusters, such as those uncovered on grids 11 and 20, the difference between the artifact densities recorded on the site cores and on the site periphery was often greater than tenfold. These two clusters featured the maximum densities of Roman material in the second survey, around 65 fragments on site 11 and 80 fragments per 100 sq meters on site 20. Much gentler differences were observed on the clusters uncovered on grids 27, 10 and 2, but still the on-site densities were several times higher than on the site’s immediate surroundings, with maximal densities of about 30 fragments per 100 sq meters. Understandably most problematic are the tiny clusters, such as those on grids 1, 14 399 J. Bintliff, A. Snodgrass, 506-513, 1988; T. J. Wilkinson, 1989, especially tab 1 where we see that even sites smaller than 1.5 ha can produce off-site carpets with radii of 200-400 meters. 218 and the dubious cluster on grid 19, all three featuring a maximum density lower than 20 fragments per 100 sq meters. On these grid units the difference between the densities recorded on the site cores and the densities recorded on peripheral units was barely two or threefold. Variations of such a scale are rare, but not untypical for the halo zone. Hence it is sometimes very difficult to decide if these are only anomalous peaks in the off-site zone or faint traces of non-residential activities. Concerning the size of the clusters, they can be grouped into four basic categories. As was shown the grid survey didn’t always manage to determine the exact limits of the sites. Nevertheless taking into account the area revealed by the grid survey, each of the discovered clusters can roughly be classed as a small, small to medium, medium and large-sized cluster400. The group of small-sized clusters includes the clusters revealed on grids 1, 3, 7, 14, 20 and possibly site 5. They all occupy areas not larger than or around 500 sq meters and with the exception of the site on grid 20, they are usually thinner than the other three categories. On these clusters the artifact density rarely exceeds 20 shards per 100 sq meters. The next group of small to medium-sized clusters includes slightly larger and denser clusters. These are basically the sites uncovered on grids 6 and 8, both problematic regarding their exact extents. They measure around 800 sq meters, with maximum densities ranging between 20 and 30 fragments per 100 sq meters. The category of medium-sized sites includes the clusters uncovered on grids 10, 26, 27 and possibly the one on grid 11. These clusters occupy areas measuring about 2 000 sq meters and are usually well-defined against the background scatters, though the maximum artifact density rarely exceeds 30 fragments per 100 sq meters. The cluster on grid 11 remains problematic, although the revealed portion measures slightly over 2 600 sq meters. There remained the problem of its relation with the concentration of Roman finds on grid 5, a couple of dozen meters from its northern edge. If this was a continuation of the same cluster, then the site on grids 5 and 11 would belong to the category of large-sized clusters. In case they were separate, the cluster on grid 11 would fall into the category of medium-sized clusters, while the one on grid 5 to the category of small-sized clusters. Site 11 is also characterized by one of the highest maximum densities of over 66 shards per 100 sq meters. Finally, the sites uncovered on grids 2 and 15-18 certainly belong to the category of large-sized clusters. They are far larger than the rest of the clusters measuring over 5 000 sq meters, though the maximal artifact densities equalled the densities recorded on smaller clusters. Earlier in the discussion we also mentioned that the tiny concentration on grid 13 could present a separate class of non-residential sites. More data are needed to test the character of this cluster. It is an isolated example in this survey area and as in the case of grid 19, the low artifact density provides no secure basis to separate it from the offsite zone. The size of the sites on field blocks 289a-291a and 47a-b, 49-50 and 66 can barely be guessed on the basis of the transect survey record. We’re nonetheless rather confident that these were the remains of multiple-core, small to medium or medium-sized sites. The collections by individual transect units suggest that the increased artifact density was limited to smaller portions of the field blocks’ areas. These concentrations probably repeat the patterns revealed on grids 1518 or 6-7, where we see two or more closely spaced small-sized clusters. It is very unlikely that on-site densities spread over the intervening stretches between the separate site components, as on site 15-18. If these two sites formed extensive and compact clusters, it is difficult to accept 400 A similar categorization has been proposed for the Roman rural sites discovered in the South Etruria Survey, T.W. Potter, 12, 1979; where apart from site area the author takes into account the character of the material to distinguish between huts, farms and villas. 219 their complete disappearance from the surface record within a period of just a few months. Finally we have the group of possible sites on field blocks 270a-271, 342 and 394. They all came to light only after the individual transect collections were adjusted to include 100% of the material counted. A common feature for these concentrations is that on-site densities are limited to one or two field units and sometimes even further confined to a certain section of the field block. It is certain that these are single-core sites, not larger than 800 sq meters. However there is the important difference between the hypothetical site on field block 342 and those on field blocks 270a-271 and 394. Like the rest of the residential sites, the former has generated a fairly extensive carpet of intermediary density, while the latter two feature only very narrow belts of intermediary density, similar to those recorded on grids 3 and 20. Tab V_1: The area of Roman clusters documented by the grid survey Cluster Number Cluster on grid 1 Grid 3 Grid 7 Grid 14 Grid 20 Grid 6 Grid 8 Maximum area ca 350 sq m ca 300 sq m ca 450 sq m ca 650 sq m > 600 sq m ca 800 sq m > 750 sq m Cluster Number Grid 5? Grid 10 Grid 26 Grid 27 Grid 11 Grid 2 Grids 15, 17, 18 Maximum area ca. 750 sq m ca 1900 sq m ca 2200 sq m ca 1800 sq m > 2 600 sq m ca 5000 sq m ca 5500 sq m Because of the peculiar context in which they were discovered, it makes little sense to assign ranks to the various site-size categories. First, we have to consider the possibility that at least in some cases a group of smaller neighbouring sites functioned as a single farming/dwelling unit or even that the entire group functioned as a single, dispersed settlement. As was shown in the preceding paragraphs and later in this discussion, it is indeed rather inconvenient to interpret each physically separate cluster as a separate dwelling unit. At the same time, the collected finds show a great uniformity regarding the absence of luxurious material. Only the cluster on grid 3 stood apart because of the small quantity of fine pottery, stone rubble and primarily because of its micro-location. There is thus no basis to differentiate between the clusters in this regard. The absence of fine ware and architectural remains is certainly significant for determining the social rank of the entire network and will be discussed below. 220 Graph V_3: Revealed site area and max artifact density (discards included) Site area and max density 90 Site 20 80 Max density per 100 sq m 70 60 Site 11 50 40 Site 5 30 Site10 Site 27 Site 2 Site 6 Site 3 20 Site 7 10 Site 15-18 Site 8 Site 14 Site 1 Site 26 0 0.00 1000.00 2000.00 3000.00 4000.00 5000.00 6000.00 Revealed site area sq m Discussing the distribution and the limits of individual clusters we noted that some of the sites differ in terms of the maximum density and the quality of the material (graph V_3). As has been demonstrated by earlier research, there is in general very little correlation between site area and on-site density401. To illustrate this it suffices to mention that the small or small to mediumsized site on grid 20 featured the highest artifact density with about 80 fragments per 100 sq meters and one of the heaviest collections in the survey area. Although ten times larger, the maximum artifact density on the clusters on grids 2 and 15-18 is not higher than 30 shards per 100 sq meters. Similarly the medium-sized site on grid 26 featured a maximum density of only 15 fragments per 100 sq meters, less than the maximum density recorded on some of the smallsized sites. There is an even weaker correlation when the average on-site densities are considered, in which case the very small sites on grids 20 and 3 are the densest. On-site density can be determined by the longevity and intensity of site use, but in this case primacy must be given to post-depositional factors. We may recall the cases of the clusters on grids 11 or 7, where the high artifact density could simply reflect the extreme fragmentation of the material. In fact we saw that entire clusters can disappear and reappear on the surface within a period of a few months. In general sites in the second survey rarely featured maximum densities higher than 35 fragments per 100 sq meters and the average density usually ranged between 7 and 15 fragments per 1000 sq meters. Note that 9 out of 14 gridded sites have maximum artifact densities ranging between 10 and 35 fragments per 1000 sq meters. Sites in the first survey area formed denser clusters, although there were considerable variations from site to site and from period to period. In fact sites dated to the Roman-Late Roman Period in the Sopot survey were likewise fairly thin, with maximum artifact densities lower than 30 fragments per 100 sq meters. 401 J.L. Bintliff, A.M. Snodgrass, 133-134, 1985. 221 Graph V_4: Revealed site areas (x axis, sq m) and maximum halo radii (y axis, ha) Site areas and max halo radii 160.00 Site 2 140.00 Site 26 Max halo radius Site 15-18 Site 11 120.00 100.00 80.00 Site 6 Site 8 Site 14 Site 10 Site 27 60.00 Site 3 40.00 Site 20 20.00 0.00 0.00 1000.00 2000.00 3000.00 4000.00 5000.00 6000.00 Revealed site area Testing the relation between the site area and the size of the halo or the nearer off-site zone seems like a far more useful pursuit. The narrow intermediary zone separating the site from the off-site is on most clusters limited to strips not wider than 15-20 meters. But the more extensive and thinner scatters forming what was called the halo zone or the nearer off-site spread over larger areas that clearly vary from site to site. Adopting the interpretation that these scatters were produced by subsidiary buildings, intense manure and rubbish disposal implies that they roughly coincide with the sites’ inner territories402. In this respect they are analogous to the satellite clusters from the first survey area, usually found limited on the same micro-topographic unit where the main cluster was situated. However there are a number of practical difficulties than need to be elaborated prior to the analysis. We saw that even determining the site area can be quite difficult because of the relatively small differences between the on-site and the off-site, as well as the peculiar distribution of the on-site clusters. The thin carpet surrounding the sites is understandably far more elusive and the peculiar on-site distribution inevitably affects the distribution in the off-site. The total grid survey demonstrated that concentric on-site patterns are the exception and not the rule in this survey area. On-site concentrations are interrupted by low density strips, while small on-site densities suddenly appear in the off-site zone. But above all it is the very low artifact density that makes the recognition of this zone impossible in practice. The couple of instances where we managed to cover larger sections of the site halos were by chance, as on grids 1 and 4 where we suspected genuine archaeological sites. In consequence the limits of this zone can hardly be determined with a greater precision. In most of the cases the extent of the halos can only be roughly estimated on the basis of the transect survey collections. As we saw these are rather difficult to interpret, especially for the purpose of estimating the extent of continuous clusters. For the present purpose, we took the lower threshold of the median density of at least 2.5 fragments per 1000 sq meters. This will most probably overestimate their extents 402 J. Bintliff, P. Howard, A. Snodgrass, 23-23, 2007; S.E. Alcock, J.F. Cherry, J.L. Davis, 142, ed I. Morris, 1994; T. J. Wilkinson, tab. 1, 1989. 222 because even very sparse off-site carpets such as those on the western ridge can increase the artifact density to over 3 fragments per 1000 sq meters of transect survey. Equally detrimental to the analysis, the transect survey can hardly distinguish between extensive, continuous and localized but dense concentrations. Further complicating the matter, both the transect and the grid survey results indicate that the halos were rarely forming symmetric rings around the sites403. For example, on grids 15-18 where the total survey covered a larger continuous block of land, on the western periphery the site halo is not wider than 50 meters, while to the north it stretches for at least 90 meters. Similar conclusions can be made on the basis of the transect survey results. The cluster on grid 10 is bounded by field units featuring higher artifact density to the south and east, but not to the west where the small radius of the impact zone is confirmed by the total grid survey. This is also characteristic for the probable clusters on field blocks 47a-b, 49-50 and 289a-291a, where the higher artifact density extends only on one or two sides of the site area and is most probably asymmetric to the main cluster. It is significant that these zones are not necessarily spreading on ground levels lower than the central clusters, as illustrated by the sites on grids 1, 14 and 15-18. The halo zones of these sites spread mostly on the terraces above the site areas. Because of the relatively low survey resolution and the irregular shapes of the scatters, we considered the maximum radius of the halo measured from the edge of the site (graph V_4). This doesn’t eliminate all problems surrounding this exercise as in a number of cases it is impossible to distinguish between the halos of neighbouring sites. One of the more problematic examples was the group of sites on grids 6, 7 and 8, all situated within a radius of 70-80 meters and “sharing” the halo zone partly revealed by grid 4 and on the surrounding field blocks. The same difficulty surrounds the drawing of the limit that separates the halos of site 2 and the cluster on field units 47a-b, 49, 50 and 66. The status of the concentration on field block 66 remains vague and although this doesn’t affect the halo radius of site 1-2, it does have a considerable effect on the maximum halo radius of the site on field blocks 47a-b, 49-50. In these and similar cases (for example, the sites on grids 15-18 and 27 or on grids 1-2 and 5-11), it was necessary to arbitrarily divide the shared portions of the halo zone into two equal halves assuming that the neighbouring sites were of an equal rank. But this is not always possible because in some cases (such as the previously mentioned example of the sites on field blocks 47a-b, 49-50 and 66), we lack information about the exact size and location of the site. As these basic parameters are missing, all three sites revealed by the transect survey on field blocks 47a-b, 49-50; 289a-291a and 342 had to be excluded from the analysis. But essentially the same problem is posed by the smaller, satellite clusters such as those on grids 1 or 5, situated in the centre of the impact zones of the larger sites on grids 2 and 11. In these cases as in the case of site 7, the satellite clusters were treated as parts of the halo zone of their larger neighbours. The suspected satellite clusters indicated by the transect survey on the halo peripheries of nearly half of the sites in this survey area were likewise treated as a segment of the halo zone. An exception was made for the few ambiguously located satellites, such as the one on field block 328b situated by the eastern survey limit or the one on field block 66 situated at an equal distance between the sites on grids 1-2 and on field blocks 47a-b, 49-50. Like the residential sites on field blocks 47a-b, 49-50 and 289a291a, they had to be excluded from the analysis. It is thus important to remember that graph V_4 correlates only the revealed site areas and the maximum halo radii. The latter parameter gives a rather imprecise estimate of the extent of the halo and in cases where the location of the central cluster is uncertain or when it lies at the 403 Cf. J. Bintliff, P. Howard, A. Snodgrass, 2007, a number of examples in Appendix A, 183-312. 223 edge of the halo zone, it can even be misleading. Therefore we also tried to provisionally determine the actual halo areas by adding together all field blocks contingent to the sites that feature artifact densities higher than 2.5-3 fragments per 1000 sq meters (graph V_5). The exact location of the site had no effect on the approximate extent of the halo zone and the decision to include or exclude one or two field units from the site halo made little difference for the maximum halo areas. When the maximum radius is measured, the inclusion of a single field unit can in certain cases extend the halo zone for nearly 100%. But measuring the approximate halo areas also has its own disadvantages. This approach is particularly problematic for sites that were revealed close to or on the very edge of the survey area. These include the sites on grids 5-11, 26 and 15-18. Particularly for the first two cases, the halo radii are a better index than the halo areas, because we don’t know if and how far they extended beyond the survey limits. Graph V_5: Revealed site areas and maximum halo areas. Site and halo areas 3.50 Site 10 Site 15-18 3.00 Site 26 Site 1-2 Site 8 Maximum halo areas 2.50 Site 5-11 2.00 Site 6-7 Site 27 1.50 1.00 Site 14 0.50 Site 3 Site 20 0.00 0.00 1000.00 2000.00 3000.00 4000.00 5000.00 6000.00 Revealed site areas Both charts demonstrate that there isn’t a particularly strong correlation between the revealed site areas and the extent of the halo zone, especially when the latter is expressed as the maximum halo area. Large clusters like those on grids 1-2 and 15-18 tend to have larger halo radii and larger halo areas, but medium-sized sites can feature equally large or even larger halos. The two largest sites on grids 1-2 and 15-18 feature maximum halo radii ranging between 120 and 150 meters, while in terms of area they measure between 2.7 and 3.1 hectares. This is very close to the extent of the halos of the medium-sized sites on grids 5-11 and 26 and if the maximum halo area is considered, the medium-sized site on grid 10 is ranked second, its halo measuring about 3 hectares. We believe however that the halo areas of both sites 10 and 26 are smaller than 3 hectares, as the periphery of their halo zones could belong to the potential site on field block 342. Note that when the maximum area is estimated, the site on grids 5-11 is ranked lower than some medium or small to medium sized sites. This is certainly related to the fact that this site was situated by the eastern limit of the survey area and it is possible that we have only revealed a portion of its halo. Predictably sites located closer to the centre of the halo zone are 224 ranked higher when the halo areas are compared. The most notable example is the site on grid 8. Classed as a small to medium-sized site, with a revealed site area of at least 750 sq meters, it has a halo with an average radius of about 80-85 meters, but in terms of area it has the fifth largest halo in the survey stretching over 2.60 ha. Admittedly the cluster on grid 8 is not the best example as it is very probable that the site occupied a larger area, a probability that seems to be confirmed by this analysis. As can be seen on the charts, there is a considerable overlap between the categories of small to medium and medium-sized sites. Some medium-sized sites like the one on grid 27 can produce rather small site halos. The maximum radius of this site’s halo is less than 75 meters long and its estimated area is barely over 1.5 hectares. The cluster on grid 6, although twice as smaller has a maximum halo radius of about 80 meters and an estimated halo area of roughly 1.75 hectares. But in the case of the site on grid 6, we have to consider the fact that the neighbouring cluster on grid 7 was eliminated from the analysis and included in the halo area of site 6. The same was applied to the rest of the satellite clusters, but in the case of the cluster on grid 7 it is possible that we are dealing with a separate core of the site on grid 6. If these two clusters are joined into a single site, it would rank as a medium-sized site and the extent of its halo would be more proportional to its site area. Naturally in such a case, we would also have to revise the eastward extent of the site halo measuring it from the edge of the cluster on grid 7 rather than from grid 6, but this wouldn’t affect the size of the maximum halo area. The other problematic cluster on grid 8 also has a possibly larger site area. Thus the two sites that mostly alter the predicted positive correlation between the site and the halo area need to be shifted slightly to the right on the charts, leaving a visible concentration of small-sized sites in the lower right corner of the charts. This doesn’t cancel the fact that there aren’t particularly strong differences between medium and large-sized sites, but it rectifies the almost random fluctuations in the extent of the halos of the small and small to medium-sized clusters. Basically all sites smaller than 1000 sq meters produced halo areas smaller than 1 hectare or a maximum radius measuring less than 80 meters. The variations between the sites, both in terms of halo radii and halo areas are relatively large. The halo radii can range from less than 30 meters on site 20 to nearly 150 meters on site 2, while the halo areas can range from 0.2 and 0.4 hectares on sites 20 and 3 to nearly 3.15 hectares on site 15-18. There is very little clustering on the charts regarding both parameters. However three sites are set apart from the rest by their very small halo areas. These are the clusters on grids 3, 14 and 20. Sites 3 and 20 have radii shorter than 50 meters and all three feature halo areas smaller than 1 hectare. In fact site 20 with a halo spreading in a radius shorter than 30 meters or over an area of 2000 sq meters can justly be treated as a site lacking a halo or a nearer off-site zone. Site 3 and 14 have slightly more extensive halo zones, but they are negligible in comparison to their larger neighbours on grids 2 and 5-11. It is perhaps no coincidence that these clusters were also made distinct by the composition of their assemblages or their micro-locations. The clusters on grids 14 and 20 were almost exclusively made of tile fragments, while the cluster on grid 3 featured rare fragments of fine ware and occupied the most prominent location in the surveyed landscape. The numerous difficulties surrounding the determination of the extent of the site halos as well as the small number of examples prevent us from observing a clear correlation between site size or character and the extent of the site halo. We can barely conclude that larger sites do tend to produce more extensive site halos, although medium and even small to medium-sized sites can sometimes produce equally large halo zones. This doesn’t have to be related to the lack of higher 225 resolution data. It has to be stressed that the extent and the prominence of the site halo is affected by the same taphonomic processes that affect the on-site density and distribution. The fact that the differences in the site areas are relatively small and the likelihood that they were of a similar socio-economic rank also must contribute to the absence of marked and consistent differences in the size of this zone. Graph V_6: Composition of the on-site assemblages. 100% 80% Other 60% Coarse 40% Plain 20% Tile/brick 0% Site 2 Site 10 Site 3 Site 15/17 Site 7 Site 14 During the interpretations of the individual clusters’ ceramic collections, there was a feeling of dismay over the significant variations often observed in their composition. The percentage of architectural ceramics ranged between 30 and 90%, while the category of plain ware could comprise anywhere between 5 and 45% of the collections. There was less fluctuation in the amount of coarse ware in the on-site assemblages, but the ratio of coarse to fine/plain ware often changed significantly from cluster to cluster. In principle however, one can distinguish between two basic types of sites on the basis of the presence of the basic categories in the ceramic assemblages (graph V_6). One is the group featuring a more balanced composition of the finds, with brick and tile representing between 35 and 65% of the assemblage and fine/plain ware being more prevalent than coarse ware. This group practically includes all on-site collections, except those collected from grids 8, 14 and 20 (the last three bars on the right). The most extreme examples of this group are the assemblages from grids 7 and 11; the former is comprised of only about 30% architectural ceramic, while in the latter the coarse ware category comprises less than 10% of the Roman collection. In the majority of the on-site assemblages brick and tile are more numerous, while coarse ware represents at least 10% of the finds404. Nevertheless the variations exhibited by these two assemblages are relatively insignificant when compared to the composition of the second group of assemblages. This group includes the clusters on grids 8, 14 and 20. They are all characterized by a very pronounced predominance of architectural ceramics, comprising between 80 and 90% of the material. Pottery fragments comprise less than 20% of the assemblages and coarse ware is usually slightly more numerous than fine/plain ware. Similarly composed assemblages were collected from the peripheral parts of the first survey area, where they were treated as non-residential sites. In the second survey these sites occupied different types of locations and they cannot be readily equatted with the sites with similarly composed assemblages from the Sopot survey. The size of the cluster is seemingly unrelated to the composition of the ceramic assemblages. Sites featuring full domestic assemblages in practice appear in all sizes, from the 404 This general scarcity of coarse ware could be related to its longer use-life in comparison to plain utilitarian vessels, as well as the higher price of certain coarse ware categories such as dolia, J.T. Peňa, 27-31, 56-7; 2007. 226 smallest (the cluster on grid 7), to the largest (the clusters on grids 2 and 15-18). However the clusters predominantly made up of brick and tile fragments either belong to the category of small sites (the clusters on grids 14 and 20) or to the category of small to medium-sized sites (the cluster on grid 8). Although we failed to demonstrate this clearly, it seems likely that in the Sopot survey too, this type of sites occupied smaller areas than the majority of sites featuring full domestic assemblages. But despite the fact that this type of sites was observed in both survey areas and is documented by other intensive survey projects405, one needs to be cautious when proposing functional interpretation solely on the basis of the composition of the ceramic assemblages. In this study it was suggested that these were possibly non-residential sites because they lacked a complete set of domestic pottery and featured a very small number of pottery fragments in general406. However the fact that the ceramic assemblages from these sites are made almost exclusively of brick and tile could very well be the result of certain taphonomic factors rather than reflecting the composition of the original assemblage. Most typically it’s been observed that after initial collapse, the heavy roof constructions can effectively seal off deposits on the floor. As a result only a few artifacts find their way into the surface record. Sites that became part of the archaeological record following this model are likely to produce little else apart from tile on the surface, especially if recently disturbed407. Another difficulty surrounding the definition of the site limits and their function was the clear tendency of spatial differentiation of the basic functional categories observed on domestic sites. In a number of cases, most prominently the clusters on grids 10, 15-18, 27 and on field blocks 47a-b, 49-50, 66 and 289a-291a, architectural ceramics and pottery were concentrated on two different portions of the site area. In nearly all examples the two portions of the site formed an integral and continuous cluster, though one cannot exclude the possibility that in some cases the two portions formed discrete clusters. We suspected that this was the case with the site on field blocks 47a-b, 49-50 and 66, where the transect collections from each of the field blocks presented a differently composed assemblage. This on-site patterning can reflect the foci of different types of settlement activities and they can be related to repeated practices of cleansing the interior of the buildings from debris and its discard in pits or on the fields surrounding the building408. This may be an intriguing revelation, but it created problems during the interpretation of the integrity of neighbouring clusters. The site areas were not always uncovered in their entirety, opening a room for doubts over the completeness of the collected assemblages. The most illustrative example is the cluster uncovered on grid 11. Analyzed separately, the Roman assemblage collected from this grid exhibits a slightly unusual composition. It was characterized by a high presence of fine and plain ware (though probably lower than 40% when all finds are included) and far more erratically, very small amounts of coarse ware. But when the combined assemblages from grids 5 and 11 are considered jointly, their composition becomes similar to that on most other clusters (graph V_7). The percentage of architectural ceramics increases to nearly 50% and that of coarse ware to over 10% of the assemblage. Although still higher than usual, the percentage of fine/plain ware is lower than 40%. The cluster on grids 1517 and 18 presented a similar example, though in this case the percentage of coarse ware was 405 J.L. Bintliff, P. Howard, A.M. Snodgrass, sites LSE7, THS3, THS12, 2007; W. Cavanagh, C. Mee and P. James, et al. 2005; C. Mee, H. Forbes, eds. 1997. 406 T. Whitelaw, 227-228, eds. R. Francovich, H. Patterson, 2000. 407 D. Pettegrew, 189- 209, 2001; J. Bintliff et al. 259-265, 2002. 408 D. Pettegrew, 200-201, 2001. 227 higher in the collections from both components of the site. There were also doubts surrounding the completeness of the assemblages from grids 6, 7, 8 and 26, where the grid survey also failed to cover the entire site area, including the site halo. In these cases the analysis of the transect collections from field units where the extension of the site area is suspected showed that the composition of the assemblage wouldn’t change dramatically. Nevertheless one needs to be reserved, for we saw that the composition of the collections can change dramatically even when different collection methods are applied. Graph V_7: Composition of the joint assemblages from the central and the satellite clusters. 100% 80% 60% Other 40% Coarse Plain 20% Tile/brick 0% Site 1-2 Site 5-11 Site 6-8 Grid 10 Grid 27 Site 1518 Grid 26 Bearing this in mind, it can be argued that the assemblages predominantly made up of architectural ceramics represent but single components of larger residential sites. Indeed at least one of these clusters wasn’t completely uncovered by the grid survey. Examining this possibility in the case of the cluster on grid 8, it was considered very unlikely that this cluster would produce a full domestic assemblage had the grid survey been extended over the neighboring fields to the north. Not only because the compositions of the transect collections differed little from the total collections, but also because such a scarcity of pottery fragments was unparalleled even on portions of residential sites where the bulk of the architectural ceramics was concentrated. However analyzing the possible halo extents, we remarked that this site has one of the largest halos in this survey area, given that the increased artifact density on the neighboring fields is truly a result of a denser off-site carpet rather than a series of small, isolated clusters. In contrast to site 8, the other two sites which lacked full domestic assemblages (on grids 14 and 20) stood fairly isolated and didn’t produce very extensive site halos. It was therefore suggested that either this was a partly revealed residential site or it formed an integral but dispersed site, including the smaller neighbours on grids 6 and 7. Indeed joining the assemblages collected from grids 6-8 will result in a full domestic assemblage, not much different than the assemblages on the majority of residential sites in the second survey area (graph V_7). By the time of the total collections by grid units, it became clear that we were dealing with an extensive network of roughly contemporary buildings. The total grid survey revealed two concentrations of sites. One much larger group formed a nearly continuous chain along the upper portions of the eastern ridge and its low off-shoot, delimiting the survey area from the southeast. The clusters uncovered on grids 1-3, 5 and 11 form the northern half of this chain, occupying the upper portions and the top of the eastern ridge. The southern half of the chain includes the sites uncovered on grids 6-8, 10, 26 and the site on field block 342, all situated along the ridge of the low off-shoot. The second smaller group of sites includes the clusters on grids 15-18, 27 and 14. These sites occupy the apex of the surveyed basin, the northern end of the survey area. Although situated in the central parts of the basin, they lie at roughly the same altitude as the sites from the 228 first group. Only the cluster on grid 20 stands isolated on the floor of the basin, approximately 500 meters from the nearest site on grid 26. There remained the problem of the possible sites overlooked by the transect survey and partly or entirely uncovered by the grid survey. The only examples are the scatters on grids 19 and 13. These two scatters exhibit different distribution patterns, but in both cases the problem arises from the low artifact densities. Although on both grids there are visible concentrations of Roman material, they are too tiny even in comparison with the category of small-sized clusters. As discussed, it is very difficult to distinguish them from similar peaks in the nearer off-site zone. For example, the scatter revealed on grid 19 differs little from the halo surrounding the site on grids 15-18. The poorly preserved fragments and the absence of larger concentrations of Roman material on the neighbouring field units cast further doubts over the character of the collection from grid 19. The case of the potential cluster on grid 13 presents a slightly different case that was already discussed. Even if this is truly a separate site, it was of a very different character than the rest of the Roman sites discovered in this survey. Clusters of such small dimensions and volume are impossible to locate and define, unless the entire area is covered by total grid survey. If the tiny cluster on grid 13 is an authentic site, then it is almost certain that there are a number of similar sites that remained unnoticed. But even in such a case, they wouldn’t affect the network of residential sites on the eastern ridge. The combined findings of the transect and the grid survey present undeniable evidence that a larger portion of the eastern ridge was occupied by installations very similar to those discovered along its ridge or in the northern end of the surveyed basin (map V_21). The two possible sites (or groups of sites) left out of the grid survey partly fill in the empty stretch in the central portions of the eastern ridge. The clusters on field blocks 47a/b, 49-50 are situated only about 100 meters west of the site on grid 2, on the same terrace as the sites on grids 5, 11 and 27. Approximately 250 meters to the south, on the same terrace as the cluster on grid 10, lies the potential site on field blocks 289a-290a. The network of Roman sites on the eastern ridge was evidently denser, though the group of sites in the northern end of the survey remains slightly more isolated. Recognizing the existence of these two sites doesn’t change the fact that the focus of the extensive network of farms and houses was on the upper portions of the eastern ridge. Distances between neighbouring clusters can range anywhere from 50-60 to 250 meters. Clearly the greatest concentration of sites is along the eastern ridge, especially in its southern part. Here the sites on grids 6-8 and 10 and the one on field block 342 are situated at intervals of 60 to 70 meters. However it has to be emphasized that all of these sites, except the one on grid 10 belong to the categories of small or small to medium-sized clusters and could form an integral, dispersed establishment. Even smaller are the distances between the clusters on grids 1 and 2 and on grids 5 and 11, where a large-sized cluster was accompanied by a smaller satellite, situated less than 40 meters from the edge of the larger cluster. Although these are physically separate clusters, they are positioned so close to each other, in all probability they functioned as parts of a single unit. They consist of the same fabric groups and the satellite clusters are too small and thin in comparison to their larger neighbours. It is therefore difficult to imagine them functioning as separate settlement units. Indeed focusing only on the distances between medium and large-sized sites, a slightly more regular pattern is revealed (map V_22). In such a case, the distances between most neighbouring groups of sites measure between 120 and 250 meters, although the clustering of sites along the top of the eastern ridge and in the northern end of the central survey section persists. 229 These groups of clusters are arranged theatrically along the gently sloping terraces of the eastern ridge. To a certain degree the possible non-residential clusters on grids 14 and 20, as well as those on field blocks 394 and 270a-271 contradict this pattern. They appear rather isolated and unrelated to any of the clusters that form the network. In this respect they are different from the similarly composed cluster on grid 8 positioned in the immediate vicinity of the clusters on grids 5-11 and 6-7-10, although this site too was ambiguously related to its neighbours. As explained in the discussion of the individual sites, one should allow that the described network of residential sites was completed by an unknown number of minor satellites, such as those on field blocks 284, 336 or 119. The cluster on grid 3 needs to be separated from the rest of the Roman sites in this survey area. In terms of size and the on-site distribution of the finds, it differs little from the rest of the small-sized clusters. However it occupies a special location in this landscape and it is characterized by pottery finds of an evidently higher quality than on the rest of the sites. This site is situated on the watershed line that separates the drainages of the surveyed basin and the small stream to the east. It overlooks both basins and stands by the main local road that links the mountain with the foothills. As can be noticed the distances between neighbouring sites are surprisingly small. Even when joining pairs of small and medium-sized or large clusters into single installations, the distances between neighboring sites rarely exceed 250 meters. And if the cluster on grid 8 is treated as a residential site with a site area larger than that revealed by the grid survey, the distances between neighbouring sites become even shorter, especially on the upper portions of the eastern ridge. Comparing these distances with those recorded by larger regional projects they are indeed unusually small, although there are examples of similarly spaced rural sites 409. Large rural villas are also known to be accompanied by less substantial outbuildings, situated not more than 250-300 meters from the main residential complex, but in the case of Skopian Montenegro we lack such a complex in the immediate surroundings410. We’ll return to the possible sociohistorical interpretations of this network in the concluding section to this chapter. It was first necessary to offer a coherent interpretation of the extent and the structure of individual sites. Understandably it is quite possible that this extensive network spreads beyond the limits of the survey area. More clusters can be expected, especially on the slopes southeast of the clusters on grids 10 and 26, as well as on the slopes north of the sites on grids 14 and 27. It is less likely that the settlement extended over to the much steeper eastern slopes, into the drainage of the neighbouring stream on the east. In all probability the small fortification discovered 170 meters northeast of the site on grid 14 was at least partly contemporary with this network of farmsteads and hamlets. If this was truly the case, it could very well mark the northeast limit of the settlement. As in the first survey area and in many other regions in the country, fortifications are inseparable components of the countryside in the Roman Period. Apart from their obvious relation with local and regional roads, we saw that they also often mark the borders of microregions. The absence of material on the surface of the fortified area prevents us from chronologically relating the fort with the rest of the Roman sites in the survey area. Yet the very 409 Most prominently in South Etruria where small farms can appear at distances of less than 300 meters, T. Potter, 125, 1979; H. Goodchild, Modeling Roman Agricultural Production in the Middle Tiber Valley, Central Italy, 2007 unpublished PhD thesis, especially tab 3.7 for distances between Early Imperial sites recorded by the South Etruria survey. 410 There are close examples from a rural survey in north Bulgaria, in the hinterland of Nicopolis ad Istrum, A.G. Poulter, Site-specific field survey: the Methodology, 583-595, ed. A.G. Poulter, 2007. 230 position of this fort and its obvious topographic relatedness to the slopes of the eastern ridge suggests that it formed a constituent part of this complex, dispersed settlement. Although lacking in strong defensive qualities, it could still act as a place of refuge for the local community and its livestock. At the same time it could’ve been used as a small station along the main local road that follows the crest of the eastern ridge. Both the fortification and the site on grid 3 are aligned along this natural route. Because of the incomplete data, but mostly because of the dispersed character of the settlement, it is rather difficult to estimate the total area occupied by domestic sites. Summing up the areas of all domestic clusters revealed by the grid survey, it turns out that the integral settlement measured at least 2.2 hectares. This figure doesn’t include the three possible clusters on field blocks 47a-b, 49-50, 66; 289a-291a and 394, as well as the problematic site on grid 8. Assuming that all three sites were made up of domestic assemblages and that they rank as small to medium or medium-sized clusters, the size of the occupied area will increase to over 3 hectares. Settlements of such size are usually ranked as small to medium-size villages, consisting of between 30 and 50 households. This is not disproportional to the size and the character of the surveyed terrain. In terms of spaciousness and fertile land, it certainly offers a greater potential than the rugged environs of Sopot. But there are at least two uncertainties surrounding this straightforward estimate. First, we don’t know if this group of sites forms an integral network, a single dispersed settlement or if they are just a segment of a much more extensive network of individual farms, dispersed across the entire plain. Knowing that there are no traces of occupation from the Roman Period in the western half of the survey, the former seems more likely, though it is possible that the network extended beyond the northern and the southeastern limits of the survey. The presence of the small fort at the top of the eastern ridge overseeing the entire network of farms, further unites the scattered estates into a single, integral settlement. The second uncertainty is related to the exact character of the various clusters that comprise this network. On all except three of the discovered clusters, the material collected formed domestic assemblages, but it is uncertain if the medium and large-sized clusters present the remains of larger, individual farms or if they are agglomerations of several or a dozen dwellings. In other words, it is difficult to decide which of the variously ranked clusters present basic settlement units and what they actually represent in socio-historical terms. Assuming that only the medium and large-sized clusters represent separate estates and that the small and small to medium sites are the remains of subsidiary buildings wouldn’t decrease significantly the total settlement area, but it would obviously decrease the number of households. This implies that the entire complex was comprised of individual or pairs of farms, surrounded by subsidiary buildings and their agricultural fields. The variations in the size and the structure of the revealed clusters could actually be related to the size of the group inhabiting a single estate or more likely, it could simply reflect the wealth of individual households and the number of subsidiary buildings. In such a case the entire network consists of not more than a dozen farmsteads, including the two possible clusters revealed by the transect survey and the potential few clusters outside the survey limits. Judging by the on-site distribution patterns, this is a more likely interpretation than the one assuming that each cluster represents a smaller agglomeration of individual dwellings. The majority of the sites had a single core; multiple cores were only recorded on grids 2, 15 and 17. Moreover these cores were in fact larger concentrations of architectural ceramics, possibly indicating the location of roofed structures 411. 411 J. Bintliff, P. Howard, A. Snodgrass, 2007; interpretation of individual sites in chapter 6 and in a greater detail, in Appendix A. 231 On the majority of the single-core sites, there was a clear spatial differentiation between the fragments of pottery and architectural ceramics. If the discovered surface clusters were agglomerations of at least several separate dwellings, one would expect to see a more even distribution of the various functional categories and multiple site cores. The electric resistivity survey on similar sites in the hinterland of Nicopolis ad Istrum in northern Bulgaria has shown that they mostly consist of one central and larger building surrounded by less substantial and smaller outbuildings412. There are two problems with this reading of the surface record. Some of the small or small to medium-sized clusters, such as those on grids 6 and 7 are characterized by full domestic assemblages and very high concentrations of surface material, although architectural ceramics is present in smaller amounts than on other clusters. But in general on these bases, they are undistinguishable from their medium or large-sized neighbours. If these are the remains of subsidiary, non-residential outbuildings, it becomes difficult to explain the strong presence of plain domestic fabrics and the fragments of cooking pots. But at the same time being so much smaller than the large-sized clusters, it is hardly tenable to claim that they are simply smaller and humbler versions of the supposed larger estates represented by medium and large-sized clusters. Either they are subsidiary structures of an unknown function or the larger and medium-sized clusters are actually agglomerations of several separate dwellings, each measuring between 500 and 1000 sq meters. It is possible that they are not the remains of building structures but the remains of middens or refuse pits, though in such a case the size and the density of the clusters is rather confusing. One should finally allow the possibility that the discovered clusters are not at all farming estates, but the winter camps of pastoralists413. As discussed in chapter IV, in later periods herding was an important component in the local economy. To be sure like in most other regions of the country, the local economies were mixed. We’ll see that the later agro-pastoralist communities carefully avoided locating the winter-camps across prime cultivable land. But in the exceptional conditions created by the founding of Scupi, the relatively large market it provided and the peculiar agrarian arrangements, it isn’t impossible that there developed settlements or estates with highly specialized production414. Again judging by the results of ethnoarchaeological research, pastoral communities leave very little ceramic material or are completely aceramic415. It is highly unlikely that purely pastoralist sites would produce such amounts of architectural ceramics and fairly extensive site halos. V.4 The Late Roman Period, 4th-6th century AD (tables 23-24, graph 22, Appendix 4) We have already pointed to the possibility that a portion of the fabric groups broadly determined as Roman or possibly even the entire assemblages belongs to the first half of the 412 A. G. Poulter, fig 2, ed. A.G. Poulter, 2007; M.J. Boyd, Geophysical survey and rural settlement architecture on the Lower Danube at the transition to Late Antiquity, 597-609, the same volume; a pattern that seemingly matches the findings of the Skopian Montenegro survey; although as we’ll see there are some notable differences. 413 A suggestion made by Prof. M. Pasquinucci, for which I’m most thankful. In both survey areas the interpretation of the results was largely shaped from a farmer’s perspective. 414 J. L. Bintliff, Going to market in Antiquity, 209-250, eds. E. Olshausen, H. Sonnabend, Zu Wasser und zu Land, Stuttgarter Kolloquium 7, Stuttgart 2002. 415 C. Chang, P.A. Tourtellote, Ethnoarchaeological survey of pastoral transhumance sites in the Grevena Region, Greece, 249-264, Journal of Field Archaeology 20-3, 1993. 232 4thcentury AD. It is nevertheless clear that the pottery production characteristic for the period of the late 4th-late 6th century is absent from the collected assemblages. This is a far more familiar material, known from a number of excavated sites and most relevant for the survey area, from Scupi416. The typical shapes and decorative techniques found in the Late Antique layers of the colony are almost completely absent from the surface collections. This is in accord with the chronology known from the great majority of open, agricultural estates in the central Balkans417. The few excavated farms or villas in the southern regions of the country were abandoned by the late 4th century AD, at the latest418. Most of the excavated villa complexes in neighboring Bulgaria, if not transformed into a nucleated settlement were also deserted by the late 4th century419. According to the survey record, there were radical changes in the surveyed basin even prior to the decline of Scupi. Excluding finds possibly datable to the 4th century AD, almost entirely limited to the cluster on grid 3, Late Roman material comprises only about 1.20% of all gathered finds (table 23, Appendix IV). The small collection datable to the Late Roman Period numbered 138 fragments. Because the great majority of the discarded finds either dated to the Roman or to the Late Ottoman-Early Modern periods, their percentage in the total surface record is even lower. This small amount of surface material was collected from about 50 field blocks dispersed across all survey sections. There is nevertheless a visible concentration of field blocks featuring at least 0.5 fragments per 1000 sq meters in the northern half of the western ridge and in the central survey section, along the Skopje-Kučevište road (map V_23a). But unlike other historical periods represented with settlement remains in the surface record, the difference between field units featuring maximal and minimal densities is negligible. The maximum densities of about 4 fragments per 1000 sq meters recorded towards the northern periphery of the survey area and on the top of the western ridge are repeated on field blocks on the eastern ridge and in the southern half of the survey. In the northwest quarter of the survey area field blocks featuring over 3 fragments per 1000 sq meters are usually accompanied by at least one field unit featuring between 1 and 3 fragments per 1000 sq meters. In the southern survey half or on the eastern ridge, field blocks featuring around 3 fragments per 1000 sq meters often stand isolated. This is the main difference between the various survey sections. In other words, the thin carpet of Late Roman finds is becoming slightly denser on the central parts of the western ridge and in the northern half of the basin floor, while the maximum artifact density remains stable across the survey area. In fact because of the low numbers, it was argued that even these slight differences are artificially enhanced by the variable size of the field blocks. Focusing on the number of collected finds per field blocks, the differences are even less significant. In principle the transect collections included not more than 2 Late Roman shards per field block (map V_23b). As explained in Appendix IV, more extensive zones of elevated density of Late Roman finds were discovered on field blocks 212a-218a in the central part of the western ridge and on two pairs of field units further north, along the top of the ridge. In the northern end of the central survey sections, larger quantities of Late Roman finds were discovered on field blocks 126, 129, M. Ončevska-Todorovska, 2004. L. Mulvin, Late Roman villa plans: the Danube-Balkan Region 377-413, eds.W. Bowden, L. Lavan, C. Machado, Recent research on the Late Antique countryside, Leiden 2004; A.G. Poulter,Cataclysm on the Lower Danube: The destruction of a complex Roman landscape, 223-254, ed. N. Christie, Landscapes of change: Rural evolutions in Late Antiquity and the Early Middle Ages, Ashgate 2004. 418 I. Mikulčić, Edna anticka villa rustica vo Tikveš 269-285, Godišen Zbornik na Filozofski Fakultet 23, 1971; K. Kepeski, Villa Rustica vo Pešterica kaj Prilep 143-156, Macedoniae Acta Archaeologica 2, 1976 419 V. Dinčev, Rimskite vili v dneshnata b’lgarska teritoria, Sofia 1997, 18-20. 416 417 233 137 and 274a. These are not contingent field units, but they form closely spaced groups, stretching over the entire northern half of the central survey section. On the eastern ridge and into the southern survey half, field blocks featuring over 1 fragment per 1000 sq meters are much more isolated. Evidently the extensive network of farms in the eastern survey sector was abandoned after the 4th century AD. On the majority of the field blocks where we recorded larger concentration of Roman material, the Late Roman finds were either completely absent or present in very small quantities. The only exception is the group of field blocks 123-124, 137 partly covered by grids 15-18, on which we discovered one of the largest Roman sites in the second survey. There aren’t any major changes in the overall pattern of distribution when the transect collections are adjusted to represent 100% of the material counted (map V_23c). In fact this operation only deepens the pattern described in the preceding paragraph. The zones of increased artifact density in the central top section of the western ridge and in the northern half of the central survey sections become slightly more compact and better defined against their backgrounds. At the same time, the isolated peaks in the southern half of the central survey section and on the eastern ridge now appear thinner, further emphasizing the increase in the northwest quarter of the survey area. To a certain degree these results were confirmed by the total collections, though only in a few cases did the regular grid survey accidentally include field units featuring higher density of Late Roman finds. On field blocks 129 and 137 covered by grids 17 and 18, the total collections by regular grids truly revealed more compact and extensive scatters of Late Roman material (map V_24). A thin, patchy carpet featuring slightly over 1 fragment per 100 sq meters covers larger portions of the gridded areas, although the very low density and the lack of focus clearly suggested that this is a segment of an off-site carpet. On field block 302 on the lower terraces of the eastern ridge and featuring 3 fragments per 1000 sq meters of transect survey, the total collections included but a pair of Late Roman shards. Similarly on the top of the western ridge, the total survey on grid 22 partly covering the large group of field blocks 212a-218a collected only a few fragments, randomly dispersed across the gridded area. In fact the situation recorded on grid 22 is almost identical to the total survey records on grids 21 and 23-24, covering field blocks where the transect collections didn’t include Late Roman finds. Small quantities of pottery were collected from the majority of the grids in the survey area, even on field blocks where the transect survey didn’t record Late Roman material. In fact the maximum density of 2.65 fragments was collected from grid 12, from a field block where the transect collections indicated a complete absence of the Late Roman material. The Late Roman finds were totally absent only among the collections from grids 5-11, 14, 20 and 26. In this respect the transect survey record was confirmed, because these sterile stretches are entirely limited to the eastern ridge (grids 14, 5-11 and 26) and in the southern half of the central survey section (grid 20). On the rest of the gridded areas, this material comprised between 1 and 5% of the total collections. As on grids 17 and 18 a handful of Late Roman finds were usually found dispersed in thin scatters across the gridded area. Only on grids 12 and 17 did the maximum density exceed the limit of 1.3 fragments per 100 sq meters. Most commonly the artifact density ranged between 0 and 1.3 fragments per 100 sq meters or in absolute terms, between 0 and 2 fragments per grid unit. Even on grid 4 where we suspected the presence of a more significant concentration, the artifact densities were below 1.3 fragments per 100 sq meters. In principle the pattern revealed by the regular grid survey consists of tiny scatters or even single finds separated by fairly large sterile intervals. This sparse carpet of Late Roman material is strikingly similar to 234 the distribution of the Roman finds in the “farther” off-site or to the Roman-Late Roman off-site in the first survey area. It is unlikely that the tiny concentrations revealed on grids 12 or 17 represent even less intensive non-residential activities, although one can never exclude that some of these scatters are derived from ephemeral phenomena such as pits, isolated burials or even agricultural huts lacking tiled roofs. As was shown, such fluctuations are not unparalleled among the off-site scatters of Roman material and they are also characteristic for the material from later historical periods. Not only the very low artifact densities, but also the character of the gathered material indicates that there are no residential sites datable to the Late Roman Period in the second survey area. The most telling signs are the very small percentage of architectural ceramics datable to this period and the actual absence of an integral ceramic assemblage. Over 80% of the collected material consisted of a single fabric group. Tile fabrics different from those accompanying the rest of the Roman material appear only on one location, on grid 4. They have thinner crosssections and simple decorative patterns typical for the Late Roman Period. As they appear concentrated on two grid units in the northern half of grid 4, it is indeed possible that this scatter of tiles is the remains of a small, non-residential structure. However they weren’t accompanied by pottery finds and the few fragments of tile barely gave a maximum density of 1 fragment per 100 sq meters. Apart from fragments of transport and storage vessels, there were no fragments of coarse ware datable to this period. It is possible that some of the fabric groups that were part of the Roman assemblages continued to be used into the Late Roman Period and the same can be argued for some of the brick and tile fabrics. This will probably increase the overall amount of the Late Roman material resulting in a denser off-site carpet, but it wouldn’t change the fact that the old network of farms was completely deserted. This implies however that the sparse carpet of Late Roman material must have originated from elsewhere, in all likelihood not far away from the survey limits. Analyzing the distribution of the Late Roman finds by field blocks, we observed an apparent increase towards the northern end of the central section and on the central top portion of the western ridge. Initially this suggested that the density of the Late Roman material gradually increased both along the southnorth and the east-west axis. But comparing the average densities recorded by grids covering various parts of the survey area, this linear trend was confirmed only for the central survey section. Here both the transect and the grid survey indicate a visible increase towards the northern survey limit; as mentioned the most extensive and compact scatters of Late Roman finds were collected from grids 17 and 18. On the western ridge the increase in the density of the Late Roman finds was limited to the central parts of the section and Late Roman finds were only slightly more numerous than on the eastern ridge. It is possible that the failure to document larger quantities of Late Roman finds in the northern half of the western ridge is related to the small coverage of the grid survey and the fact that most of the gridded area included field blocks on which the transect survey didn’t record Late Roman material. It is in any instance evident that there lacked a continuous carpet as in the northern half of the central section. If these records are accurate, it is most logical to locate the possible settlement that generated this off-site carpet not far beyond the northern limits of the survey area, on the slopes leading towards Kučevište. In fact the chapel of St. Paraskeva, just outside the northern limits of the survey could even mark the approximate location of the Late Roman settlement (map V_24). It is situated only about 50 meters away from the northernmost pair of field blocks in the central survey section. Indeed it isn’t unusual to find Christian chapels erected over Late Roman settlements, both during the Middle Age and later historical periods. In such a case however, the 235 absence of Late Roman finds from blocks 140 and 155 immediately to the south of the chapel is somewhat problematic. Field block 155 was discovered fallow and sterile, but on field block 140 not a single Late Roman shard was collected, although almost 50% of the counted material was gathered. Judging by the extent of the off-site, this hypothetical Late Roman settlement must have been larger than the farmsteads of the previous period. Focusing only on the northwest quarter of the survey area where the transect collections indicated an increased artifact density, the maximum extent of this zone is about 680 meters, measured from the small chapel to field block 218a on the central portion the western ridge. Along the Skopje-Kučevište asphalt road it extends for over 500 meters. According to some studies off-site carpets of such extents could be produced by hamlets or small villages. Indeed going back to the first survey area, we may recall that the small hamlets were usually surrounded by impact zones with similar radii. V.5 The Middle and Late Byzantine Period (early 10th-early 14th century; tables 25-26, graph 23, Appendix IV) The finds possibly datable to the Middle Ages form a very small collection, including but a few feature shards. They all consist of simple strap handles, flat bases and fragments with a low ridge running along the line of maximum diameter. Lacking more diagnostic examples, the collection can only be roughly dated between the 10th and the early 14th century. It comprised a distinct group of fabrics, limited exclusively to a small number of field units. It is significant that it doesn’t accompany finds datable to the late 14th or the Early Ottoman Period, which were found dispersed over a larger portion of the surveyed terrain. Because of the absence of clearly recognizable shapes and decorative techniques, the proposed dating should be accepted with reserves. In principle these simple shapes could date anywhere between the Late Roman and the Early Ottoman Period. The only reason for analyzing them separately is the fact that they appear limited to a single location and never accompany finds datable to the Late Roman or the Early Ottoman Period. In total only 85 fragments constitute this ceramic assemblage. They comprise 0.75% of all material studied and their share in the total surface record is certainly lower than 0.5% (table 25, Appendix IV). They were collected from 7 field units, all situated within a radius of 150 meters in the northwest corner of the survey, on the upper portions and the top of the western ridge. The pattern revealed by the transect survey is clear and it isn’t affected by the inconsistencies in the size of the field blocks’ areas and in the collections by transect units (maps V_25a and 25b). The greatest and the most extensive concentration of finds datable to the Middle Ages was revealed on a group of 4 contingent field blocks situated on the top of the western ridge, about 320 meters south-southwest from the monastic chapel dedicated to St. Paraskeva. From north to south these include field block 171, 170, 186 and 188 (table 26, Appendix IV). By far the highest artifact density was recorded on field block 171. The core of this cluster was evidently on this field unit, featuring nearly 18 fragments per 1000 sq meters of transect survey. To the south there is a sharp decline on field blocks 170 and 186 with about 3 fragments per 1000 sq meters. After about 50 meters from the southern edge of field block 170, there is another peak on field block 188, where artifact density increases to almost 6 fragments per 1000 sq meters. Finds datable to the Middle Ages disappear from the surface record on the surrounding field units. This material was present only on field block 167 situated immediately northwest of field block 171 and on field blocks 183 and 158, 100 to 150 meters to the west and 236 north of the main group. With low artifact densities ranging between 1.8 and 3 fragments per 1000 sq meters, they mark the impact zone of the possible Medieval site. Understandably adjusting the transect collections to represent 100% of the material counted the density increases on all of these field units, but the increment is fairly proportional and doesn’t affect the pattern revealed by the “raw” record of the transect survey. The total survey by regular grid units confirmed the transect survey record (map V_26). It managed to define more precisely the northern and eastern limits of the site, although it failed to reveal its very edge on the south and west. This is a compact, though irregularly shaped cluster with a possible extension on the western side. It is defined by artifact densities of about 2.5-3 fragments per 100 sq meters, recorded both along the northern and southern periphery. To the west and northeast, the decline is slightly sharper. On the site core the maximum density is relatively low. The total grid survey recorded slightly over 7.5 fragments per 100 sq meters in the centre of grid 25. This low contrast between the maximum and the minimum density, along with the low quality of the material explains why it proved difficult to document the entire site area by a regular grid survey. We encountered the same problem on certain sites from the Roman Period, most prominently the clusters on grids 5-11 and 26. It is impossible to decide where to limit the grid survey during fieldwork, by simply looking at the surface. Allowing for wider southern and western margins, the possible Medieval site measures 1500 sq meters. Based on the detailed analysis of the transect collections we argued that there was another, smaller core on field block 188, although it remains uncertain if it was a physically separate cluster or if it was related to the cluster on grid 25 through a zone of lower artifact density. This peripheral zone is characterized by an artifact density ranging between 1 and 2 shards per 100 sq meters. It was revealed along the northern and eastern site limits and it is very likely that similar total amounts contribute to the increased density on field blocks 183 and 168, situated north and west of the main cluster. Thus the collection of finds possibly datable to the Middle Age forms a cluster that features a roughly concentric distribution pattern, with a high on-site concentration surrounded by an intermediary zone of lower artifact density separating the on-site from the off-site. As we saw during the description of the Roman sites, the pattern is almost never perfectly concentric; rather the main cluster was surrounded by a number of smaller and thinner satellite clusters, separated by zones of very low or zero artifact density. Taken together, the two clusters measure between 2000 and 2500 sq meters. Similarly sized clusters of Roman material were interpreted as the remains of individual or pairs of farms. Lacking archaeological data for the rural settlements from this period, one can only speculate about the actual character of the discovered site. There is the same dilemma of whether we’re dealing with isolated agricultural estates or a cluster of smaller dwellings. In any case the very size of the possible Medieval cluster doesn’t allow for more than a few small or a single extended family. Not insignificantly it repeats the size of the medium-sized Roman sites. This indicates that during certain time-periods there was a preference for small settlement units, existing isolated or as a part of an extensive network spread across the foothills. It is almost the antipode of the Late Ottoman and Early Modern settlement pattern, with relatively large nucleated settlements, situated kilometers away from their agricultural fields The small site on grid 25 occupies a rather exposed location, much more similar to the locations of the Roman farms on the eastern ridge. The Medieval farm is situated on the top of the slightly lower, western ridge, opposite the Roman site uncovered on grids 1 and 2 and roughly at the same height of about 510 meters above the sea. The nearest active freshwater spring is on the floor of the neighbouring valley on the west, about 600 meters away. Compared 237 to its eastern neighbour, the western ridge has a drier appearance. There lacks the small groves or the lush stretches that separate the agricultural fields in the eastern half of the survey. The western ridge has a steeper eastern slope. Communication between its crest and the floor of the surveyed basin is difficult and at certain places impossible. In fact the Medieval farm on grid 25 is more orientated towards the floor of the neighbouring basin on the west than towards the surveyed area. The western slopes of the western ridge are much gentler and in this respect the two ridges covered by the survey area are very similarly shaped. Being drier and less communicable, the western ridge is slightly disadvantaged as a settlement location. To the north it leads directly to Kučevište, but to the south it ties onto a chain of low hills, which terminates in a dead-end, about 2 kilometers south of the surveyed area with the isolated hillock Čavrnjak. This circumstance explains why the modern asphalt road between Kučevište and Skopje follows the floor of the basin, rather than the top of the ridge. In the past Kučevište was accessed via the road along the eastern ridge of the surveyed basin or via the road that comes from Mirkovci. The implication is that the small Medieval farm stood isolated, hundreds of meters away from the main local roads. It communicates with the rest of the wider study region only through the head of the western ridge, through the location occupied by the modern village Kučevište. Perhaps the large area occupied by the village houses truly hides the remains of a larger nucleated settlement, contemporary with the discovered farm. In this same context of viewing the small Medieval site as a satellite of a hypothetical nucleated settlement, it is possible to argue that these are the remains of a non-residential site. Indeed in the immediate surrounding of the Late Ottoman-Early Modern village, one can still see the fading remains of sheepfolds. Until several decades ago they were used as winter-camps for the herds. However these are rather humble constructions, usually positioned on the barren mountain slopes not in the midst of the arable zone. These were simple animal sheds not intended for human habitation. The possible Medieval collection on the other hand comprised a small, but full domestic assemblage. Moreover like the majority of the Roman sites in the second survey area, it generated a relatively extensive halo, spreading for almost 50 meters to the south of the site. If the isolated concentrations on field blocks to the north and west are included, the maximum halo radius would extend for nearly 150 meters measured from the northern edge of the site. In terms of hectares the halo zone occupies between 0.5 and 1 hectare. This is considerably smaller than the halos of similarly sized Roman sites, though knowing the small size of the ceramic assemblages typically encountered on Medieval sites, this is hardly surprising420. The presence of a full domestic assemblage and of a fairly extensive halo unambiguously point to the residential character of this site. V.6 The Late Byzantine-Early Ottoman Period (late 14th- early 18th century; table 27, graph 24, Appendix 4) Although better known than earlier Medieval pottery, the material from these centuries has received inadequate attention by archaeologists working in this region421. Pottery production from this period is closely related to earlier traditions, but it also features a number of characteristics that set it apart from earlier Byzantine pottery, as well as from the ceramics dated to the 18th and 19th centuries. Lead glaze applied over a poorly fused white slip or engobe and the 420 421 A. K. Vionis, 571-572, J.L. Bintliff et al. 541-606, 2004-2005. B. Babić, 163-164, 1967. 238 characteristic wet-wash finish are among the most easily recognizable features422. Again the closest parallels come from the material excavated during the past several years from the Skopje citadel. That there lacked settlement traces dating to the Late Byzantine-Early Ottoman Period is indicated by both the statistical and the spatial distribution of the material. Finds datable to this period comprise slightly over 2% of the collection studied and only about 1% of the total surface record. The collection consists of about 245 fragments distributed over 85 field blocks, across various portions of the survey area (table 27, Appendix IV). Because of the fairly large dispersal area, even the district average values present insignificantly low thresholds. Being more numerous, the Late Byzantine-Early Ottoman assemblage exhibits a greater variation in artifact density between the field units on which this period was represented in the transect collections. But this fact doesn’t diminishes the relatively small differences between the minimum and maximum densities recorded by the transect or the grid survey. The maximums of 10 fragments per 1000 sq meters of transect and 3.1 fragments per 100 sq meters of grid survey are simply too low to represent anything but off-site or heavily truncated, vestigial remains. Even when the transect collections are adjusted to represent 100% of the material counted, the maximum density predicted barely exceeds 26 fragments per 1000 sq meters. In this respect the collection of the Late Byzantine-Early Ottoman finds resembles the Late Roman collection, the basic difference being that it is twice as large, forming a slightly denser carpet. The spatial distribution of this material is also strikingly similar to the distribution of the Late Roman finds (map V_27a). Over 50% of the field units on which the transect survey recorded finds datable to the Late Byzantine and Early Ottoman Period are concentrated in the northwest corner of the survey area. Into the southern half of the survey area and on the slopes of the eastern ridge, the carpet of Late Byzantine and Early Ottoman finds becomes evidently sparser. Along the eastern and southern limits of the survey, there are but isolated field units featuring artifact densities not higher than 1.5 shards per 1000 sq meters. For example field blocks 385, 368a and 327 feature between 2.4 and 4.4 fragments per 1000 sq meters, but they are separated by large empty stretches. On the other hand in the northwest corner of the survey, we see groups of four or five contingent field walking units featuring not much higher density of Late Byzantine-Early Ottoman pottery, although on certain isolated units the transect survey recorded over 10 fragments per 1000 sq meters, such as field block 168. In absolute terms these are evidently low values, but the figure itself isn’t decisive in determining the off-site character of the Late Byzantine-Early Ottoman scatter. As always this is determined by the relatively small difference between the zones of high and low artifact density, which contributes to the absence of a clear focus or foci. Thus the larger concentration on field block 168 and on the four neighboruing field blocks to the east is repeated on a number of other field blocks in the northern half of the western ridge and on the basin floor. Admittedly the maximum of slightly over 10 fragments per 1000 sq meters isn’t reached, but the artifact density is still above the threshold of 2 fragments per 1000 sq meters. If this is taken as a provisional on-site threshold, it follows that there was a network of smaller sites not unlike the network of Roman farms, but with an evident focus on the northwest quarter of the survey area. Technically the only difference between the distribution patterns of the material from these two periods is one of scale. On the hypothetical cores of the clusters of Late Byzantine-Early Ottoman material the artifact density is rarely higher than 2 to 3 times the average value. In the case of the Roman collection by transect units these differences were at 422 J. Vroom, Byzantine to Modern Pottery in the Aegean: An Introduction and Field Guide, Utrecht 2005. 239 least two or threefold. Furthermore while the carpet of Roman finds was organized in concentric zones around a dozen distinct cores, the clusters of field blocks on which the Late ByzantineEarly Ottoman finds were present appeared suddenly after smaller or larger stretches of sterile surface. As with the Late Roman collection, the carpet was becoming denser towards the northwest survey quarter, while the district maximum and average values remain fairly stable or only slightly increased. The carpet of Late Byzantine-Early Ottoman finds lacks one of the basic features of settlement sites and this is their more or less concentric pattern of distribution. Two potential biases distort the transect survey record: the variable size of the field block as a basic quantitative unit and the inconsistencies in the individual transect collections. As shown in Appendix IV, these two factors are particularly potent in the northwest survey quarter; the survey sections which according to the transect collections feature the highest density of Late Byzantine-Early Ottoman finds. The field blocks in the northern half of the western ridge are on average smaller than the field blocks in other survey section, which directly enhances their artifact density. When field blocks are compared by the number of collected finds, the differences are far less significant, as in principle each field block gave between 1 and 5 fragments. Even when featuring a single fragment datable to the Late Byzantine-Early Ottoman Period, a smaller field unit will be ranked higher than a larger field unit on which the transect collections included 2 Late Byzantine-Early Ottoman shards. In addition the field blocks in the northern half of the western ridge were more thoroughly transected, the transect collections regularly including over 50% of the material counted. In other survey sections especially the northern halves of the central section and the eastern ridge, the transect collections were less intensive and often included less than 40% of the counted finds. Assuming that all counted material was collected, the northern half of the western ridge loses its prominence (map V_27b). On certain field units in the northern half of the central survey section such as field blocks 97b and 134, the artifact density could theoretically increase to over 20 fragments per 1000 sq meters. In fact the maximum density recorded by the transect survey moves from field block 168 to field block 134 in the central survey section, with over 26 fragments per 1000 sq meters. To be sure, the artifact density on the field blocks in the northern half of the western ridge also increases. Although the increase is more gradual than on certain field blocks in the central survey section, these field units continue to rank high in terms of artifact density. Most significantly the assumed cluster of field blocks centered on field block 168 loses its significance as larger and denser “clusters” appear in the central parts of the ridge or in the northern half of the central survey section. On the field blocks from the rest of the survey sections, the effects of these factors are far gentler and the overall pattern described in the preceding paragraph persists. What this analysis corrected was the apparent concentration of finds in the northern half of the western ridge. It now becomes clear that there were roughly equal or perhaps even larger quantities of Late ByzantineEarly Ottoman material in the northern half of the central survey section. As we will see this was confirmed by the total grid surveys. We were lucky that a number of field blocks featuring artifact densities higher than the threshold of 2 fragments per 1000 sq meters of transect survey were inadvertently included in the regular grid survey (map V_28a). After all, a total grid survey is the most straightforward way of deciding if an increased density represents genuine archaeological remains or a patchy off-site carpet. The total collections by regular grid units covered the greater portion of the hypothetical cluster of field blocks 168-172, though it missed the core on field block 168 (map V_28b). It further included a number of field blocks where artifact density could theoretically increase to 240 over 10 fragments per 1000 sq meters. These include field blocks 129 and 134 (grids 17 and 18) and field block 212a, covered by grid 22. After the study of the total collections, the absence of settlement remains or even the remains of less intensive, non-residential activities was confirmed. Even on grids covering the field blocks ranked higher than the average on the basis of the transect survey records, the maximal artifact density never exceeded the limit of 2 fragments per 100 sq meters. In fact on certain grids such as grid 22 or 1, the maximum densities were lower than those predicted on the basis of the transect survey record and not much higher than the selective transect collections. Recall that this was also the case with the Late Roman collection but also with the Roman material in the off-site zone. Outside the zones of on-site densities the discrepancies between the transect and the grid survey records are considerably diminished. Apart from grid 22, the only other more significant discrepancy between the transect and the regular grid survey was on grid 19 covering field block 263. Here although the transect collections didn’t include Late Byzantine-Early Ottoman finds, the total grid collections recorded the highest artifact density of 3.1 shards per 100 sq meters. In the case of this field unit, there were also considerable discrepancies concerning the presence of Roman finds, which must be related to the less thorough transect collections but also to the very low quality and obtrusiveness of the finds. Nevertheless even the maximum densities of 3.1 fragments per 100 sq meters recorded on two contingent grid units in the centre of the grid are not sufficient to elevate this collection to an on-site status. The range between the minimum and maximum density remains extremely low, while on a micro-level the “clusters” revealed on grids 19, 16-17 or 25 completely lack an inner structure. They are randomly scattered across the gridded area, the artifact density fluctuating between 0 and 2 fragments per 100 sq meters (map V_28a). In fact the pattern revealed by the transect survey is repeated on a micro-level: the carpet becomes slightly denser on certain grids while the maximum density remains stable. When dealing with carpets of surface material that lack a visible focus, the average density recorded per gridded area is a far more sensitive index of the changes across larger survey sections. However these data must be seen as a complementary to the record of the transect survey and not as its ultimate test. As argued in the appendix and in a previous section, the gridded areas are not necessarily representative of the entire survey sections. On the surface, the overall distribution of the Late Byzantine-Early Ottoman finds by grids is quite similar to the distribution pattern of the Late Roman finds (cf. graphs 22 and 24, Appendix IV). In both cases we see two significant disagreements with the transect survey record: field blocks in the northern half of the central survey section feature artifact densities much higher than fields on the western ridge and the difference between the densities on grids covering portions of the eastern and the western ridge are much lower than indicated by the transect survey. On grids 17 and 19, the average density reaches over 5 and nearly 10 fragments per 1000 sq meters. Only on grid 25 does the average density come close with about 2.5 fragments per 1000 sq meters, but in this case the increase is uncertain because the Late Byzantine-Early Ottoman collections includes a fabric group that could equally belong the earlier Medieval assemblage. Excluding this category would lower the artifact density on grid 25 to only slightly over 2 fragments per 1000 sq meters. On the rest of the grids including those covering parts of the central survey section (grid 18, 20), the average density never exceeds 2 fragments per 1000 sq meters. As with the Late Roman collection, it is difficult to observe a clear linear tendency. In the central survey section the main concentration is not located by the northern survey limit, but on grid 19 towards the centre of the survey area and on both the 241 eastern and the western ridge there is a subtle increase along a south-north axis. Finally, the total grid collections confirmed the higher density on the western ridge, although the increase from the eastern ridge is much lower than expected. Thus the grid survey record draws a slightly different pattern of distribution from the one indicated by the transect survey. The density of Late Byzantine-Early Ottoman finds in the northern half of the central survey section is not merely equal to that recorded on the northern half of the western ridge, but on certain field blocks it is much higher. On the other hand the differences in artifact densities between the eastern and the western ridge are hardly as prominent as suggested by the transect survey. We believe however that this comparison doesn’t cancel the observations made on the basis of the transect survey records. Understandably the average densities by individual grids chiefly reflect local conditions and can hardly depict an overall linear tendency. Indeed if we estimate the average density on the total of gridded areas by various survey sections, the differences are gentler, although the overall pattern remains unchanged. There is a higher concentration of Late Byzantine-Early Ottoman finds in the central survey section and this doesn’t have to be related to the fragmented character of the finds and ultimately to the fact that the fields in this part of the survey area were often finely harrowed. We suggested this interpretation because the grid surveys in the central survey sections recorded higher off-site values not only for the Late Byzantine-Early Ottoman and the Late Roman periods, but also for the off-site carpet of Roman finds. However this same argument can be used in favor of the view that the off-site carpet on the fields on the floor of the surveyed basin was truly denser and accumulated over at least three different time-periods. This shouldn’t come as a surprise knowing that this is probably the most fertile and intensely cultivated part of the survey area. This is actually indicated by the present-day land use, with the numerous gardens and orchards aligned by the Skopje-Kučevište asphalt road. The off-site character of the Late Byzantine and Early Ottoman scatters is to a certain degree reflected by the lack of a complete ceramic assemblage. The great majority of the finds belonged to fabric groups classed as fine pottery, but it is possible that some of these fabrics were used for storage or transport. Unlike contemporary pottery from the Sopot survey, there were at least three fabric groups that could be related with cooking and food preparation including the characteristic type of bread-baking pans or tzrepna. Needless to say, the dating of this pottery is far from certain423. The most numerous of these fabric groups finds parallels among the material excavated from the Late Ottoman-Early Modern deposits on the Skopje citadel. This group alone consists of over 550 shards, more than twice the entire collection of finds dated to the Late Byzantine and Early Ottoman Period from Skopian Montenegro. Only about a dozen fragments belonged to a fabric group resembling cooking pot fabrics from earlier centuries. For this group direct parallels were found among the material from one of the abandoned villages mentioned in the Early Ottoman censuses, its location still being remembered by the local inhabitants. It is also possible that at least a portion of the tzrepna fragments date prior to the 18th century. Interestingly not a single fragment of this characteristic pottery was collected from grid 25, from the Medieval farmstead. As we discovered during the analysis of the contemporary material from the Sopot survey, there are no fragments of architectural ceramics that could be related to this period. The predominant tile fabrics are found dispersed across the entire survey area, accompanied by other Late Ottoman and Early Modern fabric B. Ristevski 1044, 2006. B. Babić, 298-306, 1986; R. Carlton, An ethnoarchaeological study of pottery production on the Dalmatian island of Iž, 101-124, eds. J. Chapman et al. Recent developments in Yugoslav Archaeology, Oxford 1988. 423 242 groups. Like the cooking pot fabrics, each of these groups is at least twice as numerous as the entire collection dated to the Late Byzantine and Early Ottoman Period. Because of the problematic chronology some of these fabric groups were left out of the analysis, although it was possible to notionally partition these classes between the Late Byzantine-Early Ottoman and the Late Ottoman-Early Modern periods424. This could result in fuller assemblages for the earlier period, but it wouldn’t change the overall distribution pattern and the interpretation proposed. The increase in the amount of the Late Byzantine-Early Ottoman material in the northwest quarter of the survey and in the central survey section in particular points to the nearest possible settlement that generated this off-site carpet. As will be shown, the distribution pattern revealed foreshadows the distribution of the off-site debris discarded during the past couple of centuries. The carpet of Late Byzantine-Early Ottoman finds appears as its more contracted and sparser predecessor. At the same time it deepens the pattern exhibited by the finds dated to the Late Roman Period, which were mostly concentrated along the survey’s northern limit. While it isn’t impossible that a certain percentage of the Late Byzantine-Early Ottoman material comes from non-residential activities, it is certain that these are not settlement remains but off-site material related to a larger settlement in the nearby vicinity. The most likely candidate is Late Byzantine and Early Ottoman Kučevište, either situated beneath its Late Ottoman successor or in its immediate vicinity. This is indicated by the location of the main village church, built sometime in the first half of the 14th century and renewed towards the middle of the 17th century425. Kučevište is situated at a roughly equal distance from the survey area as the neighbouring Mirkovci to the west. Both villages lie at a distance of about 1.3 kilometers, but while Kučevište has a direct access via a downwards road, Mirkovci is situated on a slightly lower ground, across the valley of the Kučeviški Stream. Regarding logistics it is clear that the discovered off-site debris originated from Kučevište and not from Mirkovci426. Although there is undeniable evidence for settlement displacements in the region during the 18th and the early 19th century, it is evident that the basis for the Late Ottoman and Early Modern settlement pattern was established as early as the 14th century, in all probability prior to the arrival of the Ottomans. The survey actually captured the very periphery of the inner village territory of Kučevište. The gradual faltering of the off-site carpet in the southern half of the survey area roughly draws the limits of the intensely cultivated land. To our surprise individual fragments from this period were collected from field blocks lying at a distance of nearly 2.5 kilometers, measured from the church of the Holy Savior, the core of the Late Medieval and Ottoman Period community. It is very likely that rare, isolated finds from this period can also be expected even on fields beyond the survey’s southern limit, but the more intensely cultivated area was limited within a radius of 1.6 kilometers from the centre of the Late Ottoman-Early Modern settlement. According to some researchers, off-site carpets of such an extent appear around major villages or even small towns. This qualification is not necessarily exaggerated regarding the population of Late Byzantine-Early Ottoman Kučevište427. It was possibly the largest village in the region of Skopje and one of the largest villages in the country. 424 Cf. J. Bintliff, P. Howard, A. Snodgrass, 13, 18, 2007; J. F. Cherry, et al. 331-334, 1991. P. Miljković-Pepek, Crkvata Sv. Spas vo Kučevište, 417-421, eds. V. Mošin et al. 1975. 426 P. Howard, 111-128, eds. J. Bintliff, P. Howard, A. Snodgrass, 2007. 427 T.J. Wilkinson, tab.1, 1989. 425 243 V.7 The Late Ottoman-Early Modern Period 1800-1950 (table 28, graphs 25-27, Appendix IV) As explained in the previous chapter, while determining the limits of the survey area we’re attempting to avoid the narrower village territories, regularly covered with large amounts of material discarded during the past couple of centuries. But the study of the collected finds proved that the extent of this zone was greatly underestimated. In fact as the analysis in the previous section showed, a larger portion of the surveyed terrain was already part of the narrower territory of Early Ottoman Kučevište. Predictably in comparison to the finds dating to the Late Byzantine-Early Ottoman Period, the volume of Late Ottoman-Early Modern material was many times greater. This was realized by the time of the second year’s campaign, but because of the sheer amount of material, the great variety of fabric groups and the seemingly erratic distribution pattern, the fact simply couldn’t be accepted without a certain dosage of suspicion. Yet the prolonged study of the fabrics, comparison with the Late Ottoman and Early Modern finds from the Sopot survey and the material from recent excavations on the Skopje citadel removed even the slightest doubt about the dating of these finds. The pottery and the architectural ceramics from this period have never been a research subject in regional archaeology, but they exhibit certain characteristics that distinguish them from earlier ceramic production. The high firing temperatures reflected in the great solidity of the paste and the very frequent use of a poor vitreous glaze applied directly on the surface are among the most apparent features 428. Almost 6350 fragments or 55% of the studied material could be dated to the Late Ottoman-Early Modern Period (table 28, Appendix IV). If all finds collected by regular grid units are included in the analysis, their number will be nearly doubled and they would have represented approximately 75% of the total surface record. Even this is an underestimate, because it is almost certain that much of the material that was counted but wasn’t included in the transect collections also dates to this period. Given the considerable distance from the centre of the traditional village, this was a surprising discovery. Equally baffling were the very high maximum and average densities recorded by both the transect and the grid survey. Comparing these records with those for the material datable to the Roman Period there are hardly any differences, apart from the small ratio between the average densities recorded by the grid and the transect survey. Typically for the more evenly widespread collections, the grid survey average is about 2.5 times the average recorded by the transect survey, corrected for the lesser degree of survey intensity. But at the same time the range between the maximum and the minimum values is slightly higher for the Late Ottoman-Early Modern material when all collected finds are taken into account. Despite the very large dispersal, variations across field blocks can often be quite significant. Finds datable to the last couple of centuries were discovered on the great majority of the field walking units or more precisely on over 85% of the survey area. The Late Ottoman-Early Modern finds were completely absent only from the surface of field blocks where no surface material was recorded or for which data were missing. Rather intriguingly this material was also completely absent in the transect collections from field blocks 1, 3, 4, 6 and 351 (map V_29a). These field blocks, all situated by the survey’s eastern limit were covered by grids 5-11 and 26, on both of which Roman sites were discovered. As was shown earlier in the text, the transect survey records were largely confirmed by the study of the total grid collections. On both locations Roman material represents over 80% of the total surface record. The total collections 428 J. Vroom, fig. 5.1-5.5, 6.1-6.2, 2005. 244 from grids 14 and 20 featured very similar chronological profiles. On the fields covered by these grids the carpet of Late Ottoman-Early Modern finds suddenly becomes thinner, although the share of this material in the total collections is slightly higher, representing 20 and 14% of the material studied. In the rest of the transect and grid collections, material dating to the last couple of centuries represented at least 50% of the collected finds. This implies that the distribution of the total surface record was to a large extent determined by the distribution of the Late OttomanEarly Modern finds. The zones of higher artifact densities recorded by the transect survey were actually zones of higher concentrations of Late Ottoman and Early Modern material. Thanks to this circumstance a large percentage of the field blocks with high artifact density were included in the regular grid survey. Despite the very large amount, there are absolutely no doubts about the off-site character of the surface material datable to the last couple of centuries. The possible existence of more intense, non-residential activities in the survey area would’ve hardly gone unnoticed in the earliest ethnographic and travelers’ accounts. It is highly unlikely that even non-residential sites such as tombs or chapels, built in the last couple of centuries would simply disappear in the plough-zone, leaving no memory in the local toponomy and oral tradition. The difficulty arises from the fact that there aren’t too many other indicators of the off-site character of the Late Ottoman and Early Modern material, inherent to its distribution pattern. While discussing the distribution of the Late Byzantine and Early Ottoman finds, it was remarked that although they are three times more numerous than the material possibly datable to the Middle Age, unlike the latter group they don’t represent the remains of intensive, residential activities. This was reflected in the rather small differences between the zones of high and low artifact densities. A similar observation was made on the distribution of the finds datable to the Late Roman Period. In both cases the difference between field blocks featuring very high and very low artifact density was rarely greater than threefold. In the case of the Late Ottoman-Early Modern collections, the difference between field blocks featuring very high and very low artifact densities was in some cases tenfold. Looking at the results of the selective transect collections, field units belonging to the zone of very high artifact density can feature between 17 and 51 fragments per 1000 sq meters, while field units with low artifact density featured between 0.7 and 4 fragments per 1000 sq meters. These contrasts are far greater than the differences in the density of Roman finds recorded in the on-site and off-site zones. Bearing this in mind, it is no wonder that so much time and effort was spent on total grid collections on field blocks where the surface record almost entirely consisted of material datable to the Late Ottoman-Early Modern Period. The sheer quantity of the surface material on field blocks 231-33, 263 or 495 covered by grids 19, 21 and 23 was taken as a clear sign that there were genuine archaeological sites on these fields. But the study of the material confirmed the suspicions already incited at the fieldwork stage. Not only on grids 21-24 covering the southern half of the western ridge, but also on a number of other grids in this survey area the bulk of the surface material dated to the last couple of centuries. According to the transect survey collections, the highest concentration of this material was in the northwest quarter of the survey area. In this respect the distribution of the finds datable to the last couple of centuries follows the distribution of the off-site carpet from the preceding Late Byzantine-Early Ottoman and the Late Roman periods. There is a gentle decrease on the field blocks in the northern half of the survey section and on the field blocks on the eastern ridge. There are no considerable variations between the northern and the southern half of the eastern ridge, the lower and the upper terraces. It is evident however that the field blocks on 245 which the transect collections didn’t include finds datable to the Late Ottoman-Early Modern Period tend to cluster along the eastern survey limit and on the top of the ridge. In the southern half of the central survey section artifact density remains at the same level. It is actually higher when compared to certain portions of the eastern ridge. But this off-site carpet is most visibly diminished on the southern half of the western ridge, where we see the largest number of sterile field units. In this survey section the decrease in the amount of Late Ottoman-Early Modern finds has obviously caused the thinning out of the total surface record. In terms of mean densities by survey sectors, the northern half of the western ridge features slightly over 10 fragments per 1000 sq meters (graph 25, Appendix IV). In the northern half of the central survey section the transect collections recorded 8.5 and on the northern half of the eastern ridge 7.5 fragments per 1000 sq meters. The density of Late Ottoman-Early Modern finds remains stable and in fact it is slightly higher on the southern half of the eastern ridge and in the southern half of the central survey section. Only on the southern half of the western ridge do we see a sharper decrease, with a mean sector’s value of less than 6 fragments per 1000 sq meters. The overall pattern doesn’t change significantly assuming that total collections by transect units were made on all field blocks (map V_29b). The northern halves of the western ridge and the central survey section feature roughly equal mean densities, while on the eastern ridge the decline is sharper than indicated by the transect collections, especially in the northern half. Understandably if all counted finds were collected the contrast between the southern half of the western ridge and the rest of the survey sections also becomes sharper. The overall pattern is very much clear. Artifact density decreases from north to south and more gently from west to east. Regarding the locations of the present-day villages, this distribution is far from surprising. It is but an enhanced version of the Late Byzantine-Early Ottoman off-site carpet. The difficulties of interpretation begin to arise once the distribution by field blocks is looked at in a greater detail. We are faced not only with considerable variations within the limits of single survey sections, but also with roughly concentric patterns of distribution. This is especially evident in the sections featuring average or lower than average artifact densities, such as the eastern ridge or the southern half of the western ridge. The density of Late Ottoman-Early Modern finds suddenly increases on groups of contingent field blocks, covering areas of over 1 hectare. It is often possible to observe a typical on-site patterning, with field blocks featuring very high artifact densities surrounded by field units featuring average or higher than average densities. The groups of field blocks 38a/b-40a/b and 495-497 on the upper terraces of the eastern ridge and in the southern half of the western ridge are particularly good examples. They are in principle undistinguishable from the clusters of field blocks featuring onsite densities of Roman material. It has to be emphasized though that this patterning is to a certain degree, the artificial product of the inconsistencies in the individual transect collections. If total collections by individual transects are assumed, the density of Late Ottoman-Early Modern finds further increases on these groups of field blocks, while on the surrounding field units the increase is slight. Consequently the roughly concentric pattern disappears and these groups of field blocks emerge as isolated density peaks. In most instances the total survey by regular grids confirmed the results of the transect survey and considering the average density per grid, they are very close to the values predicted on the basis of the transect survey (graph 27, Appendix IV, map V_30). This was the case not only on those field blocks where the transect survey recorded very low artifact densities (covered by grids 5-11 or 20), but also on the field blocks featuring very high densities (grids 19, 15-17). On certain grids however, the total collections didn’t produce the expected maximums. This was 246 especially pronounced on grids 12, 22 and 25. In all three cases we expected fairly high artifact densities, close to those recorded on grids in the central survey section. These discrepancies between the transect and the grid survey records could be related to changes in the surface layer, though we observed that in both survey areas finds datable to the last two centuries formed the most stable fraction of the surface record. We also thought of the possibility that the material on these units is more obtrusive (indeed tile is fairly numerous on grid 12).But the problem arose not so much from overestimating the true densities on field blocks covered by grids 12 or 22 as from the underestimating of artifact densities on the field blocks in the central survey section. The true amount of surface material in this survey section came to light only after the regular grid survey, although for some field units the transect survey also indicated high artifact density. The patterns revealed by the total grid survey look much more like the typical off-site distributions. On the majority of the grids regardless of the artifact density, the carpet of Late Ottoman-Early Modern finds lacks a clear focus (map V_31, 32). The material is evenly spread over much of the gridded areas and the recorded density rarely exceeded 10 fragments per 100 sq meters. As we saw on certain grids (5-11, 14 and 20) larger portions of the gridded areas were sterile with low isolated peaks. Essentially the same pattern was revealed on grids 15-17 and 19, the difference being that on these grids, the density of Late Ottoman-Early Modern finds was considerably higher. The great majority of the grid units featured over 15 fragments per 100 sq meters and on grid 19, the total survey recorded over 30 fragments per 100 sq meters on three contingent units in the centre of the gridded area. It should be noted that the true maximums are much higher. On certain grids they could increase for almost 100% were all collected finds included in the analysis. However on a few grids the concentric pattern characteristic for the on-site distributions returns. This is particularly apparent on grids 1 and 23-24, where we see both wider ranges between the maximum and the minimum densities and clustering of grid units with high artifact densities in certain parts of the gridded areas. Finer variations can also be observed on grids 21 and 22, both covering the central parts of the western ridge. As explained in the appendix, it is not by chance that these grids spread over more than one field unit. This is not a strict rule, as grids 2 or 4 although covering at least two or three field units revealed thin and even carpets across the gridded area. On the other hand, the smallest variations were observed on smaller grids, limited to one or two neighbouring fields. In other words, there were no significant fluctuations on a field block level, within the limits of individual agricultural parcels. This is most apparent on grids covering larger continuous segments of the survey area, such as grids 1-2 or 15-18. Here grid units with higher artifact densities are usually limited to certain agricultural fields. The large amount of Late Ottoman-Early Modern material collected from grids 1 and 17 cannot be traced across the grid limits and into the neighbouring fields covered by grids 2 and 18 (map V_31). That there were significant variations in artifact densities across the grids is finally indicated by the comparison of the average densities recorded on the gridded areas (graph 27, Appendix IV). This revealed very dramatic fluctuations over rather short distances, which we pointed out during the discussion of the chronological composition of the total collections from the Roman sites on the eastern ridge. One may recall the case of grids 19 and 20; although spaced only 200 meters apart in the central parts of the survey area, the difference in artifact density is more than tenfold. Over 110 fragments per 1000 sq meters were recorded on grid 19 and only about 10 fragments per 1000 sq meters on grid 20. Similar drastic fluctuations were revealed on grids 1 and 2, situated next to each other on the upper terraces of the eastern ridge. 247 The linear trends observed in the distribution of the Late Ottoman-Early Modern finds by field blocks disappear even when comparing the average densities recorded on grids that cover parts of different survey sections. This peculiar distribution simply reflects the fact that each of the agricultural fields present a discrete locus of material deposition. It has to be stressed that at least some of the agricultural parcels which we used as field walking and quantification units could be a couple of centuries old. The sudden and often sharp differences in the amounts of off-site debris are actually revealing the distinct histories of material deposition on each separate field. After all one has to assume that the off-site debris was deposited by individual or possibly, by groups of contingent fields and not in large sways over the entire agricultural zone. Apart from the accessibility of the fields, other factors that could influence the variable distribution of off-site material are the soil properties and the type of cultures grown 429. One also has to take into account the fact that manure was not equally available to every family and even if it was, it will hardly be distributed evenly across all fields, regardless of the location, soil type or culture. It has been pointed out that concentrating manure on certain fields is primarily a matter of investment strategies of individual farmers430. To a large degree this helps us better understand the very dense concentrations of Late Ottoman-Early Modern off-site debris revealed by the regular grid survey on certain fields in the central survey section. As with the much thinner carpets of Late Byzantine-Early Ottoman and Late Roman finds, the average densities recorded on grids 17 and 19 are considerably higher than on the rest of the gridded areas. This suggests that at least some of the gardens along the Skopje-Kučevište asphalt road have been a focus of intensive agricultural exploitation during three different time periods; four, if we consider the dense off-site of Roman finds on grid 19. During the description of the survey area, we mentioned that like the wider study region it presents a mosaic of various soil types. Ethnographers have observed that the local inhabitants had developed an elaborate classification system for the local soils431. Unfortunately there are no detailed soil maps of this region and it is impossible to relate this variable to the fluctuations in the off-site record. It is equally difficult to point to a straightforward relation between the types of cultures grown and the presence of off-site debris. The most illuminating examples are the densities recorded on field blocks 2 and 320, on both of which the transect survey records were confirmed by the regular grid surveys. Both fields were vineyards, situated less than 200 meters apart and easily accessible from the local road-network. Yet while field block 320 featured over 20 Late Ottoman-Early Modern fragments per 1000 sq meters, this material was almost completely absent from field block 2. Nevertheless we can observe that over 90% of the field units featuring very high density of Late Ottoman-Early Modern off-site material were planted with cereals. This group included only a small number of gardens and vineyards, mostly the much-discussed examples along the Skopje-Kučevište road. Hardly surprising off-site material was mostly absent on fallow or abandoned fields, although Late Ottoman-Early Modern artifacts were sometimes scarce even on fields planted with cereals and on gardens. Despite the often significant localized variations, the survey results confirmed that the spreading of this off-site carpet is largely related to the proximity of the contemporary settlements. There is an undeniable decrease in the quantity of the off-site material along the north-south and west-east axis, although the proximity to local roads apparently wasn’t a 429 R. Shiel, A. Stewart, 95-100, J. Bintliff, P. Howards, A. Snodgrass, et al. 2007. H. Forbes, 159-172, ed. R. Jones, 2012. 431 S. Tomić, 455-464, 1905. 430 248 particularly strong factor. These tendencies were actually slightly obscured by the decision to collect less material from a number of field blocks in the northern halves of the central survey section and the eastern ridge. Naturally this mostly affects field blocks featuring very high overall artifact densities, which automatically decreases the differences in artifact densities between the survey sections. We saw that if all counted material is included in the analysis, the concentrations of the Late Ottoman-Early Modern material in the northwest survey quarter and especially along the Skopje-Kučevište road would have been much more pronounced. But the Montenegro survey also shows that there are other, less predictable factors related to the varying agricultural practices of individual farmers or even to isolated events of debris discard. V.8 Conclusions: the history of settlement in the second survey and the wider study region The second survey area had a fairly short and patchy history of human occupation. In contrast to the settlement history of the first survey area, there is very little substance to relate with the developments on a broader regional level. In fact this applies for the integral region of the Skopje Basin; much of its settlement history is probably lost forever under the strong waves of urbanization in the past several decades. It is no accident that the great majority of the known archaeological sites come from the periphery of the basin. But even in these micro-regions, only certain periods of the past are represented in the archaeological record. In this respect the sequence of settlement revealed in the survey area largely follows the known regional developments, although there are certain differences. Before summarizing the history of human habitation in the second survey area, it is important to consider some of the factors possibly related to the presence of such a small number of periods in the surface record. After all compared to the first survey area, this is a larger territory, characterized by gentler relief and richer agricultural resources. Only the intensively surveyed area extends over 1.1 kilometers of almost 100% arable land. If we assume that the productivity of these soils is comparable to that of the thin soils along the Middle Vardar and allow for 5.5 hectares large individual estates, the farmland available only in the limits of the survey area could comfortably sustain a community of almost 20 families.Allowing for a more intensive agricultural regime with individual estates occupying about 3.6 hectares, the maximum number of families could rise to almost 30432. Understandably the catchments of the possible settlements in the second survey can extend far beyond the survey limits, encompassing most of the central portion of the foothills. As the small valleys that dissect the plain are not very steep, even the exploitation of the land along the banks of the neighbouring Kučeviški and the Pobuški Stream could still be feasible; although one could predict that the maximal theoretical area would have spread further south, within the limits of the surveyed basin. Setting the limits of this theoretical territory at a distance of 2-3 kilometers from the hypothetical settlement centre433, it will measure between 15 and 20 sq kilometers. Except for the northern end including a small section of the mountainside, the rest of this land is cultivable. In fact because of the local topography, it is likely that the territory of this hypothetical settlement will mostly spread to the south, with arable land comprising almost 100%. Such an extensive farmland can support a community of over 300 households, exceeding even the largest of the present-day villages by 50%. It is thus more than apparent that the carrying capacity of the survey area and the wider 432 For the minimum and average size of individual farming estates see the discussions in M.H. Jameson et al, 400-415, 1994; J. Bintliff, P. Howard, A. Snodgrass, et al. 158-159, 2007. 433 J. Bintliff, 510-511, ed. G. Barker, 1999. 249 study region couldn’t have been a constraining factor in the local settlement history. Even the area of slightly over 1 sq kilometer covered by the intensive survey can sustain a hamlet or even a small village. Other factors must have contributed to the discontinuous history of habitation in the second survey area. All in all there are three main phases of human settlement in the second survey area. The earliest, dating to an unknown later prehistoric period is represented by a couple of dozen fragments. They were found scattered on several field blocks along the top and the upper slopes of the eastern and western ridges. Their concentration is far greater on the eastern ridge, which isn’t necessarily related to the fact that the more intensive regular grid survey covered a larger portion of this survey section. Both on the eastern and western ridges this material was confined to the upper slopes and the crests, never appearing on the lower sections and in the central parts of the basin. Although the collection is very small and not fully homogeneous, the distribution of the finds provides a hint for the earliest chapter of the settlement history in the second survey area. Most of the finds were found isolated, accompanied by larger amounts of Late OttomanEarly Modern debris or more often, by loads of unearthed Roman material. Only rarely did they form tiny clusters of several fragments, always on later Roman sites. These individual or small groups of finds are spaced a few hundred meters apart, heralding the distribution of the Roman farms. It is difficult to come up with an interpretation different than that adopted for the Bronze Age settlement in the Sopot survey. The handful of prehistoric fragments from the Montenegro survey are the vestigial remains of small, isolated farmsteads (map V_33). In all probability they didn’t differ essentially from their Roman successors. The thin and highly dispersed scatter could hardly represent the remains of a larger nucleated settlement. At the same time it is also evident that this is not a mere section of a more extensive carpet of prehistoric finds. It was emphasized that they were exclusively confined to the crests and the upper slopes of the ridges, with a possible greater concentration on the eastern ridge. Thus as in the Roman Period, farming was probably concentrated on the slopes and on the valley floors. The small size of these establishments, their likely short life-spans, coupled by a centuries-long history of intensive agricultural exploitation have ensured that only a tiny fraction of the original ceramic assemblage survived or made its way into the surface record434. We repeat the fact that the small collection of prehistoric finds doesn’t form a coherent ceramic assemblage. It actually remains uncertain if the entire collection dates to a single or more than one prehistoric period. Judging by the basic fabric features, none of these finds dates earlier than the Late Bronze Age. In the entire region of Skopje, only 25 known sites are datable between the end of the 13th and the end of the 4th century BC (map V_34)435. In the Skopje Basin itself, there are less than a dozen late prehistoric sites, all but one of which are dated between the end of the 8th and the end of the 4th century BC. These consist almost exclusively of fortified hill-tops and accidentally discovered necropoleis. In fact only two of these sites have been characterized as open settlements, although it is clear that at least some of the necropoleis also belonged to open settlements. Mound burials, the hallmark of the Iron Age landscape in the southern regions of the country are unknown in the Skopje Basin436. In contrast to the settlement distribution during the Neolithic, almost all of the known late prehistoric sites are situated along the periphery of the 434 1998. M. Kuna, 29-44, eds. J. Bintliff, M. Kuna, N. Venclová, 2000; E. Neustupný, 49, ed. E. Neustupný, Arheološka karta na Republika Makedonija, vol. II, 609-616, Skopje 1996. D. Mitrevski, 1997. 435 436 250 basin. This is the hilly zone separating the basin’s floor from the surrounding mountainside, described in chapter IV437. The micro-regions that belong to this geo-pedologic zone are one of the oldest settlement niches in the Skopje Basin. Most of the known Iron Age and Early Antique sites are situated near villages from later historical periods. The survey area and the wider study region are also part of this zone and it is very possible that the “oppidum” near modern Brazda was not the only late prehistoric settlement in this large and relatively fertile region 438. Indeed ethnographers have made a number of vague references to “prehistoric” tombs and lithics accidentally unearthed near the villages Mirkovci, Kuševište and Pobužje, but this is impossible to confirm at present439. Despite the low degree of systematic research, it is undeniable that the Skopje Basin like many other regions in the country experienced a growth during the periods of the Iron Age and Early Antiquity. The appearance of the small amount of late prehistoric material in the surface record of the second survey area is therefore hardly a surprise. If these are truly the remains of isolated or small group of farmsteads, it implies that the late prehistoric settlement had set the precedent for all later settlement in the surveyed basin. As explained in the conclusion to chapter III, the fortified Iron Age and Early Antique settlements came to an abrupt end by the late 4th century BC440. During the next three centuries up until the Roman conquest of Moesia, the region of Skopje enters a “Dark Age” phase. According to the archaeological atlas, only one accidental find in the entire region of Skopje can be dated to the Hellenistic Period. Professor I. Mikulčić mentions several burials on the later urban territory of Scupi, relating them with the small predecessor of the Roman colony, but these findings have not been confirmed by later excavations441. While not denying the evident decline and contraction of settlement during this period, it is difficult to accept that the relatively large region of the Skopje Basin was simply deserted during the Hellenistic Period. Apparently older settlements have either contracted or were shifted to other less prominent sites, still awaiting their discovery and proper documentation. The known historical events pertaining to the Skopje Basin, the advance of the Dardanian tribes and the constant wars with Macedon can only explain the abrupt breach with earlier tradition, but not the complete lack of information about the archaeology of this region during the Hellenistic Period. As will be shown, a very similar episode follows the end of Antiquity lasting through most of the Middle Ages. The most substantial settlement remains from the second survey are broadly dated to the Roman Period, not later than the first half of the 4th century AD. As it often happens there are not too many elements to propose a more precise dating. Archaeologists have only recently started to uncover the remains of Early and Mid-Imperial Scupi; most of the data published over the past decades come from excavations on the city necropolis442. In fact the dating proposed is mostly based on negative evidence, on the absence of pottery characteristic for the Late Roman Period 437 J. Trifunovski, 347-349, 1955. D. Koračević, Praistoriski naod od Brazda, 57-65, Macedoniae Acta Archaeologica 3, 1977; I. Mikulčić, 21-23, 1982, I. Mikulčić, V. Sokolovska, 79-93, 1990. 439 J. Trifunovski, 7, 1971; the author mentions a mound with urn burials destroyed during the construction of a road near Mirkovci. Unfortunately he doesn’t give a more precise location. As explained in an earlier chapter, mound burials typically belong to the first half of the first millennium BC. 440 I. Mikulčić, 20-21, 1982; although the most recent excavations on the Skopje citadel didn’t confirm the existence of Iron Age layers, the general observations concerning the chronology of settlement in the Skopje Basin are still valid. 441 I. Mikulčić, Neilirski elementi medju skupskim nalazima na prelazu iz predrimskog u Rimsko Doba, 245-258, Sahranjivanje kod Ilira (a symposium), Beograd 1979; L. Jovanova, 197-221, 2008. 442 I. Mikulčić, 89-102, T. I-XII, 1973-1974; L. Jovanova, D. Mihailova, Skupi-Istočna nekropola, 203-248, Macedoniae Acta Archaeologica 15, 1999. 438 251 among the discovered assemblages. The material finds close parallels among the pottery excavated from Scupi’s eastern necropolis, from tombs dated between the early second and the middle of the 4th century AD. We believe that there were at least 9 farmsteads in the second survey area during the Roman Period (map V_35). They are irregularly distributed into three groups: one larger, occupying the upper terraces of the eastern ridge and its low off-shoot that encloses the survey area from the southeast and two smaller in the central parts of the ridge and in the northern end of the survey area. According to the distribution of the halo zone, in total occupying about 20 hectares and mostly concentrated on the terraces, in-between the central sites, the most intensive farming was focused on the slopes of the eastern ridge and only on certain sections of the valley floor. It is evident that these 20 hectares represent only the most intensively cultivated parts of the landscape and that the agricultural territory of these farms spread over most of the eastern ridge and possibly over parts of the western ridge. This pattern of settlement and agricultural exploitation consisting of an inner ring of fields planted with labour-demanding cultures and an outer ring of fields with crops that require less care and no irrigation is attested both in historical sources and archaeologically443 and we also observed in the first survey area. The network of Roman farmsteads was completed by an isolated, small-sized cluster and a small fortification, both occupying the top of the eastern ridge. Like the forts discovered in the Sopot survey, the surface of the fortified area was almost sterile. Consequently there is no clear positive evidence to chronologically relate this fortification with the network of agricultural estates spread on the lower terraces. Only the close topographical relatedness between the fortified hill-top and the western slope of the ridge indicate that the fort and the network of farmsteads were at least partly contemporary forming an integral settlement complex (map V_35). An alternative explanation would be that the fort and perhaps even the small site on grid 3 were established after the network of dispersed farms was abandoned. Because of the rough chronological resolution, this thesis is equally difficult to support with positive evidence. Needless to stress there is very little information about the types of rural settlements during the Roman Period, especially for the regions of the Balkan interior. In regional archaeology, surface scatters of Roman material found in the countryside are often indiscriminately called villae or vici, or simply “settlements”444. These terms particularly the villa, often carry only vague connotations; sometimes designating independent rural settlement units, sometimes referring to larger estates that are part of larger nucleated settlements. The term vicus designates nucleated rural settlements, usually but not necessarily, protected by fortifications. Theoretically the distinction is very much clear-cut: villas are larger agricultural estates belonging to medium or larger landowners, while vici are nucleated, rural settlements, roughly the equivalent of modern villages and hamlets. The latter are often associated with preRoman settlements that didn’t develop into truly urban centres after the Roman conquest445. However the archaeological identification of these two categories of rural settlements is far more troublesome. In total only two Roman villas have been excavated and published on the territory of the Republic of Macedonia, both discovered in the southern parts of the country446. Much more data 443 .J.L. Bintliff, P. Howard, A.M. Snodgrass, et al. fig 4.3, 2007; traditional peasant-farmers seem to make an analogous distinction between “wet” and “dry” cultures, J. Trifunovski, fig. 3, 1955. 444 Arheološka Karta na Republika Makedonija, vol II, 1996, and in most reconnaissance reports. 445 I. Mikulčić, 64, 1982. 446 I. Mikulčić, 269-285, 1971; K. Kepeski, 143-156, 1976. 252 come from neighbouring regions in the Balkan interior, especially from the provinces along the Danube447. In the majority of cases these are large compounds, including elaborate residential buildings, spacious yards and various production facilities. But in some cases, Roman villas lack elaborate residential components and are limited to the storage and production facilities, the so called pars rustica448. Smaller landowners would naturally build smaller and simpler estates and one also has to take into account the possible socio-economic and regional differences. On the other hand, almost nothing is known about the size and the character of the vicus as a form of rural settlement, in most countries of the Balkan Peninsula449. When applied the term simply refers to rural sites, lacking in urban planning or traces of monumental architecture. In the regional archaeological literature there are no examples of Roman vici documented through surface survey or excavations450. Most commonly the distinction between Roman villae and vici is purely architectural, despite the fact that theoretically the two terms designate different types of settlement units. Some authors have proposed that strictly speaking a Roman rural villa shouldn’t be related to a nearby urban or rural agglomeration451. According to this perspective, a villarustica in the real sense of the word is an autonomous settlement unit, with its own territory and consequently farms or other agricultural estates in the near vicinity of nucleated settlements don’t belong to this category, despite the similarities in the architecture and planning. Obviously the aim of this fairly narrow definition is to avoid distinctions solely based on formal and architectural characteristics. But there are problems associated with this approach. Understandably it shouldn’t be difficult to maintain this criterion when distinguishing between isolated villas and complexes in the agricultural territories of larger urban centres. But this is hardly attainable when one needs to make a distinction between an autonomous villa and an agricultural estate gravitating towards a minor, nucleated settlement. This is practically impossible in the absence of detailed survey or excavation data. Even assuming that the location and the character of the nucleated settlements are known, there still remains the delicate problem of drawing the limits of its agricultural territory452. It is evident that this definition encompasses only the very large estates, the Roman latifundia, which according to ancient authors owned at least half a square kilometer of agricultural land, pastures and groves453. The size and the distribution of the clusters of Roman material found in the Montenegro survey bear an undeniable resemblance to the plans of excavated Roman villas. Studying the villas discovered in other regions of the province of Upper Moesia, mostly on the territory of modern-day Bulgaria, it is evident that the size of the majority of the fully excavated examples A. Mócsy, 1972; L. Mulvin, 377-413, eds.W. Bowden, L. Lavan, C. Machado, 2004; V. Dinčev, 1997; A.G. Poulter, The Roman to Byzantine transition in the Balkans: preliminary report on Nicopolis and its hinterland, 347-358, Journal of Roman Studies 13, 1999; A.G. Poulter, 51-98, ed. A.G. Poulter, 2007. 448 V. Dinčev, 17-18, 1997; L. Mulvin 379, 2004; for Roman villas in general, J. T. Smith, Roman villas: a study in social structure, New York 1997. 449 A.G. Poulter, 223-254, ed. N. Christie, 2004. Vici are better known from the western Roman provinces, M. Rorison, Vici in Roman Gaul, Oxford 2001. 450 I. Mikulčić 64, 1982 has identified three vici among the Roman fortifications in the SkopjeBasin, but the author solely relies on the meager construction of the walls and the simple planning. 451 V. Dinčev 9-24, 1997; see the criticism by A.G. Poulter 229-230, ed. N. Christie, 2004. 452 V. Dinčev 18, 1997. In the hinterland of Nicopolis ad Istrum, Roman villas were discovered at every 2.5 kilometers, dividing the landscape into very large estates, roughly measuring 500 ha. A. G. Poulter, 239, ed. N. Christie, 2004. 453 K.D. White, Roman farming, London 1970. 447 253 ranges between 1-2000 and 5-6000 sq meters454. Only a few exceptional examples occupied areas greater than one hectare. These are usually estates of very rich landowners or higher ranking members of the provincial administration. Thus the areas occupied by the average villas are nearly identical to the areas over which our medium and large-sized clusters are spread. Moreover in some cases very great similarities were noted between the plans of excavated villa complexes and the shapes and on-site distributions of surface material on the sites discovered in the second survey. Usually the various buildings of the villa compound are arranged around an inner court-yard, but there are also many examples with plans in the shape of the Greek letters Γ or Π. The elongated shapes of some of the discovered clusters look very much like the so called corridor-shaped units, often enlarged in later phases by adding buildings at the ends of the corridor. More importantly some of the excavated villas lacked an inner courtyard and the various components of the complex were dispersed across larger areas. In these cases the major component (presumably the residential part) can be accompanied by several, many times smaller outbuildings, sometimes situated at distances of a few hundred meters. The similarities with some of the discovered clusters are difficult to ignore, especially the group of clusters on grids 68 and field blocks 47a-b, 49-50. Even the size of the smaller subsidiary structures is strikingly similar to our small-sized clusters, rarely exceeding 1000 sq meters. Regarding size and inner planning, it seems there is little room for doubts over the character of the sites discovered in the Montenegro survey. However turning to the other aspects that characterize Roman villas in the inner Balkan provinces, the evidence is far from positive. First of all, the context in which they were found doesn’t meet the criterion of the villa as an isolated, independent settlement unit. Even allowing for an extreme dispersal of the various components of the villa complex, it is obvious that there were at least three separate complexes in the survey area. While it is possible to join the sites situated between grids 5-11 and 10 into a single very large complex, it is evident that the clusters on grids 15-18 and 27 and the clusters on grids 1 and 2 formed separate estates455. Such close co-existence of at least three separate villa complexes cannot be cited in the literature available. The greatest distance between two neighboring medium and large-sized cluster never exceeds 250-300 meters. Assuming that the surrounding agricultural land was evenly divided between these farms, each was allotted with agricultural areas not larger than several hectares. This is still above the limit of 10 iugera or about 2.5 hectares, according to Roman authors, the minimum size of a small estate 456. But they are obviously far from the size of a large or a medium-sized estate and in this respect they don’t qualify as typical rural villas457. In this context it is noteworthy that the potential agricultural territories of the discovered sites are surprisingly close to the estimated halo areas. On the majority of residential sites they ranged between 1.5 and 3 hectares. Understandably it shouldn’t be excluded that these farms owned land outside their immediate catchments, but the most V. Dinčev 1997; regarding planning and size, some of the villas are near replicas of the Roman sites in the second survey, especially pl. 3, 15, 41. There are also close similarities with the villa rustica near Prilep, K. Kepeski pl. 1, 1976. See also examples in J.T. Smith, 199-217, 1997; L. Mulvin, 385-394, eds.W. Bowden, L. Lavan, C. Machado, 2004; M. J, Boyd, fig 2, 3, ed. A. G. Poulter, 2007. 455 The greatest distance between two outbuildings belonging to the same villa-complex rarely exceeds 200250 meters, V. Dinčev 15, 1997; A.G. Poulter, 235, ed. N. Christie, 2004. 456 K.D. White 387-388, 1970; note however, that the author is citing older sources, from the time of the LateRepublic. 457 However, if we look at the evidence of the South Etruria survey we will see that the majority of the small rural sites had similar territories, T. Potter, 122, 125, 1979; the author stressing the fact that the physical environment in the survey region would make large landholdings impractical; H. Goodchild, 110-118, 2007. 454 254 intensively exploited land would have been limited to the fields in the immediate vicinity of the sites, roughly coinciding with the site halos. The other negative evidence comes from the character of the collected finds. Roman villas are commonly associated with solid architecture, often with visible building remains on the surface458. The studied villa complexes from Bulgaria often include large dining halls, baths, sanctuaries and in some cases even defensive walls. Fragments of columns and architectural pieces are common, even as surface finds. Luxurious table ware, glass and mosaic tesserae are regular inventories of the archaeological record on villa-sites. Apart from a handful of fine pottery fragments of a higher quality, none of this was present in the surface record of the second survey area. Even stone rubble was very scarce, despite the large quantities of architectural ceramics. In fact fragments of fine sigillata and roughly hewn stone-blocks were only found on the cluster on grid 3, underlining its special character. On the rest of the clusters the domestic pottery was either plain or had faintly preserved traces of poor, lean slip. With the evidently humble character of the finds and the lack of visible architectural remains on the surface, these small agricultural estates are incomparable to the elaborate villa complexes in the Balkan provinces. Although seemingly planned along very similar lines, their walls were probably built of mud-brick or roughly cut stone joined by mud. The surface material gathered indicates that they were inhabited by a poorer community of farmers or herdsmen. Despite the close proximity to Scupi, almost none of the luxury items produced in the colony found their way into the surface record of the survey area. Admittedly the absence of solid architectural remains and luxurious artifacts doesn’t necessarily exclude the possibility that these are individual agricultural estates of rich landowners459. Examples of Roman villas without residential components are known from other regions of Upper Moesia, in modern northwest Bulgaria. These are essentially the properties of rich town-dwellers whose estates were run by slaves, tenant farmers or sharecroppers. The examples published are indeed much humbler in comparison to residential villas460. They are characterized by very simple plans, poor masonry and earthen floors. In one or two cases the researchers have noted the scarcity of building stone, concluding that the upper parts of the walls were entirely constructed of wood and mud-brick461. Not surprisingly however, only a few examples of villas without residential components are known and in nearly all of these cases, the complexes were only partly uncovered. To further complicate matters, dwellings that are parts of rural nucleated complexes sometimes repeat the architectural planning of villa complexes. In fact as noted by earlier authors, by the time of the Late Republic and the Early Empire the typical villa plan with an inner courtyard, surrounded by the various functional components is obviously inspired by the urban palaces with peristyle yards462. Hence it is impossible to determine the social rank of the estate solely on the basis of the plan and size of the buildings. Detailed survey data from the surroundings is necessary and this is rarely available in the published studies. There is one final possibility that needs to be briefly considered. The rural sites survey in the hinterland of ancient Nicopolis ad Istrum revealed a peculiar pattern of a large, residential villa complexes accompanied by outbuildings and the humble remains of what the author interpreted as a village or a hamlet, situated several hundred meters away from the central I. Mikulčić, 1971; V. Dinčev, 1997; L. Mulvin, 381, eds.W. Bowden, L. Lavan, C. Machado, 2004. V. Dinčev, 17-18, 1997.A.G. Poulter, 229-230, ed. N. Christie, 2004. 460 In total, only five examples are cited by V. Dinčev 68-73, 1997. 461 M.J. Boyd, 601, ed. A.G. Poulter, 2007. 462 A.G. McKay, Houses, villas and palaces in the Roman World, London 1975. 458 459 255 residence463. It is possible that in our case, we managed to document only a part of the original arrangement, the main residence remaining hidden beyond the limits of the survey. This is an option that needs to be taken seriously, although there are two chief objections against this view. First, unlike in the cases of Nicopolis ad Istrum and elsewhere464, the supposed community of dependent farmers didn’t live in a nucleated village or hamlet but in a highly dispersed settlement, with clear evidence that the land lying in-between the settlement units was intensely cultivated. We already agreed that the discovered sites were not the remains of agglomerated dwellings but farmstead, albeit of a much lower rank than the rich villas commonly found in the literature. Secondly, as we’ll see in the following paragraphs, although there is definite evidence that high ranking officials from the nearby colony of Scupi had properties in Skopian Montenegro, none of the accidentally discovered tombstones or sarcophagi can be related to the area covered by the survey. The remains of a large residential villa complex can hardly remain completely hidden. Regardless of the exact socio-historical character of the revealed network of Roman sites, for the moment it is more likely that these are the remains of individual or pairs of agricultural estates, rather than a nucleated settlement organized into separate dwelling quarters. For the latter there are simply no known parallels from the central regions of the Balkan Peninsula. But at the same time, it must be underlined that while villa estates are relatively well researched in certain parts of the Empire, other forms of rural settlement remain a complete mystery, especially in the central Balkan provinces. We still lack even the slightest idea of what the traditional rural houses looked like during this period, how big they were or whether they formed nucleated agglomerations or existed as isolated estates465. Briefly turning to the situation in Sopot during the Roman Period, one also notes a certain degree of settlement dispersal, although in this survey area there were much fewer separate clusters, set apart at greater distances. The network actually consisted of one larger and one medium-sized cluster, spaced nearly a kilometer apart and at least three other, possible non-residential sites, a kilometer and a half from the main cluster. The findings of the Montenegro survey beget a slight rethinking of the interpretations of the Sopot survey results. In the first survey area, there were at least 6 separate clusters of Roman finds, three of which were identified as fully residential sites. Of the latter, two formed a closely related pair and were merged into a single site 5a-5b. The third, much smaller cluster comprised a full domestic assemblage and was situated at a considerable distance of nearly 1 kilometer, in a different survey sector. Still further away, at the other end of the survey area were the three clusters featuring assemblages predominantly made up of architectural ceramics. They too formed a closely related pair and an isolated cluster, situated nearly a kilometer away from the latter. This arrangement was interpreted as a combination of a small nucleated settlement and a farm accompanied by a group of non-residential sites or alternatively, as two nucleated settlements accompanied by satellite farms. The pattern is evidently different than that in the second survey area, with individual clusters spaced at greater distances and with a thinner offsite carpet. But this difference is largely dictated by the local topography. The fragmented character of the terrain in the first survey precluded clustering of more than one residential site 463 A.G. Poulter, fig 2, in ed. A.G. Poulter, 2007. A very similar example is presented by V. Dinčev, 60-67, fig.59, 1997, from central Bulgaria. 465 A.G. Poulter, 347-348, 1999; out of 15 sites documented by intensive survey and electric resistivity in the area of Nicopolis, 14 were determined as villas, although there is epigraphic evidence for rural settlements from the same area. 464 256 per topographic unit. Each of the topographic units that constituted the surveyed valley presented a physically separate settlement location. In the second survey, such micro-topographic divisions are non-existent. The surveyed basin forms an almost continuous stretch of gently sloping land. Although there are differences between the various points of the terrain concerning access to local roads and visual control, there is only a very vague physical fragmentation. In these conditions it is expected and indeed logical to see evenly spaced settlement units across a larger stretch of gentle and fertile territory. But even in the second survey there was an apparent preference for the eastern ridge and its upper slopes in particular. Site 5a-b in the Sopot survey, interpreted as a small nucleated settlement measured between 0.8 and 1 hectare. In absolute terms it is not much larger than the large-size clusters from the second survey, although in comparison its area is almost twice as large. But is this difference sufficient to see this site as a nucleated settlement, rather than as a larger agricultural estate?466 Examples of villa estates occupying areas of over 1 hectare are not uncommon for the central Balkans. However these are normally very luxurious establishments, with visible architectural remains on the surface467. In this respect the material that was encountered on the surface of site 5a-b differed little from the surface remains on the Roman sites in the second survey. Architectural material consisted exclusively of brick and tile fragments, while stone rubble was completely missing. Interestingly in both survey areas, there was a spatial differentiation between architectural ceramics and pottery suggesting that this is most probably the result of discard behaviour and taphonomic processes. There are however a number of differences between the major Roman site in the Sopot survey and the large-sized clusters discovered in the Montenegro survey. First of all, there is the chronological difference. The great majority of the finds collected from site 5a-b in Sopot were dated to the Late Roman Period, while the Roman sites in Skopian Montenegro were dated not later than the late 4th century. Knowing that nearly none of the securely dated Roman villas in the region post-date the late 4th century, it is very unlikely that the region of Sopot, situated near the busy Vardar Valley corridor was an exception. But this doesn’t mean that individual farming estates were unknown in the period of Late Antiquity. In fact the small Late Roman cluster on site 8 in the first survey, measuring around 3000 sq meters can only be interpreted as the remains of an individual farmstead. There are also slight differences regarding the on-site distribution patterns. In Sopot we also observed a tendency to linear clustering of the on-site material, but the zones of higher concentration were more widespread, while multiple cores were common. On the other hand, the Roman sites in the Montenegro survey usually had the core limited to one or two grid units, featuring far greater artifact density than the rest of the site area. This is potentially related to the social rank and the architectural organization of the settlement, though it could be also determined by post-depositional factors468. The distribution pattern on the small cluster on site 8 looked much more like the on-site distribution in the Montenegro survey, with the bulk of the finds concentrated on a pair of grid units and gradually decreasing artifact density along a single row of grid units. 466 In regional survey projects in Greece sites of similar size are usually identified as small hamlets or large farms; J.L. Bintliff, P. Howard, A. Snodgrass, 185, 2007; W. Cavanagh W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al. 2002; C. Mee, H. Forbes, et al. 1997. 467 V. Dinčev, pl.12, 18, 1997. 468 Cf. J. Bintliff, P. Howard, A. Snodgrass, sites THS 12 and THS 14, 2007. 257 Finally, there is an important difference in the general property of the ceramics. The Roman material from the Sopot survey was part of a centuries-long tradition of local ceramic production. The shapes and techniques differed very little from earlier, prehistoric pottery production, while lacking clear parallels among the Roman pottery typically found in larger settlements. The relatively large number of misfired tile and pottery is also suggesting local production. In comparison, misfired fragments and wasters were rare among the Roman finds from Skopian Montenegro. The second survey area obviously lacked a strong pre-Roman tradition and the tile and pottery collected find very close parallels among the material recently excavated from Scupi’s eastern necropolis. Doubtless it is near impossible to clearly determine the architectural organization of the discovered sites solely on the basis of the distribution of surface finds. The differences observed in the on-site distributions in Sopot and Skopian Montenegro can equally reflect divergent settlement type and organization and different post-depositional processes. Nevertheless the size of the discovered clusters, as well as the character of the surface material,clearly determine the social rank of these settlements. These were either the remains of small hamlets or of individual farmsteads belonging to smaller landholders. For reasons already discussed, it is likely that the only true nucleated settlement was discovered on site 5a-b in the Sopot survey. The rest of the fully residential sites in the two survey areas were remains of individual farms. In all probability these are the lowest ranking settlements that one can expect in the Roman countryside. The small estates in Skopian Montenegro could produce enough surplus for trade, but there is very little or no positive evidence. The intensive collection and study of the surface material revealed an almost total absence of luxury items produced in larger settlement centres. Hence trade with the neighbouring towns was probably limited to a range of most basic products. One notable feature of the Roman and Late Roman sites discovered in the two surveys is the relatively low on-site artifact density. In this respect, the clusters revealed in the second survey area featured somewhat higher artifact densities, but they were still much thinner in comparison to the high concentrations of artifacts encountered on the surface of prehistoric sites in the Sopot survey, with maximum densities often exceeding 50 fragments per 100 sq meters. This relative sparseness of the Roman on-site scatters becomes even more pronounced knowing that more than 50% of the Roman surface finds consisted of architectural ceramics, an artifact category that wasn’t represented in pre-Roman assemblages. Moreover being deposited much later than prehistoric artifacts and generally being fired at higher, more stable temperatures, Roman ceramics is more likely to survive in the surface record 469. How then to explain this apparent disagreement between expected and actual results? Other factors that can potentially influence the density of surface artifacts are the intensity and the character of occupation, its longevity and the quantity of ceramic artifacts produced and utilized by different cultures and societies470. Concerning the latter aspect it is certain that at least in some regions of the country, the Roman Period expanded the range of ceramic artifacts used by the local societies. The most apparent example is the introduction of the various forms of architectural ceramics, but there is also a wider range of pottery categories. The Sopot survey also showed that the longevity of occupation wasn’t necessarily decisive, as some of the very dense prehistoric clusters could be 469 J.L. Bintliff, P. Howard and A.M. Snodgrass, 145-149, 1999; M. Kuna, 33-34, eds. J. Bintliff, M. Kuna, N. Venclová, 2000. 470 M. Kuna, 33, eds. J. Bintliff, M. Kuna, N. Venclová, 2000; M. Millett, Dating, quantifying and utilizing pottery assemblages from surface survey, 53-59, eds. R. Franovich, H. Patterson, 2000; E. Neustupný, 51, ed. E. Neustupný, 1998. 258 dated within a period of a couple of centuries471. The remaining cultural factor, the intensity and character of occupation is the most elusive one for surface archaeology. Analyzing the composition of the various Roman assemblages in the two survey areas, it was indeed concluded that certain assemblages were “incomplete”. Certain categories of pottery normally found in domestic assemblages were almost completely missing. But full domestic assemblages didn’t necessarily appear in the denser clusters. One wonders then, if the relatively low artifact density on residential sites from the Roman Period isn’t related to the organization of domestic space in the countryside. There seem to be no apparent fault in assuming that a greater number of households per settlement will produce a higher density of surface debris. All other conditions being equal, individual agricultural estates including inner courtyards, animal sheds and other subsidiary buildings would inevitably leave thinner on-site scatters than a nucleated settlement consisting of a number of separate households and occupying an equally large area472. In other words, the larger vernacular complexes meant that there were less people per hectare of occupied area. Understandably these features are period and region-specific and more research is needed in order to infer the possible number of inhabitants on the basis of settlement areas. Finally, we have to allow for the agency of certain taphonomic factors in the determining the on-site artifact density. In the case of the sites from the Roman and Late Roman periods, although the introduction of tile and brick expands the range of ceramic artifacts associated with residential sites, the presence of heavy, tiled roofs can sometimes have the opposite effect 473. This could offer a possible explanation for the “brick and tile clusters” discovered in both survey areas. The sudden emergence of an extensive network of farmsteads in the second survey area mirrors the developments in the wider study region and in the Skopje Basin in general during the Roman Period. In contrast to the Hellenistic Period represented by only two sites for the entire region of Skopje, there are 111 sites dated to the Roman Period in the country’s archaeological atlas. This figure probably including a number of Late Roman sites is far from definitive. In fact 25 years prior to the publication of the archaeological atlas, in the Tabula Imperii Romani series, 160 Roman sites were entered for the same geographical and administrative region 474. The character of these sites (dislocated funerary steles, votive altars) indicates that despite the evident increase from the Hellenistic Period, the true number of Roman sites is probably much greater. Less than 30% of the Roman sites registered in the T.I.R. series and less than 20% of the Roman sites in the archaeological atlas were identified as open settlements. In both publications, more than 50% of the discovered sites consisted of “isolated” epigraphic monuments or forts. The majority of these sites were discovered accidentally. They actually represent the portion of sites with monumental architectural remains. Not surprisingly there usually lacks information about the precise location of these finds and about the surrounding context. Although rather patchy and truncated, the data available are to a certain degree instructive of the settlement pattern in the Skopje Basin during the Roman Period. It definitely points to a strong expansion compared to earlier periods, though it has to be remembered that the character of the Roman sites is for the greater part dubious. The great majority of the accidentally discovered sepulchral monuments come from small, family mausoleums, in all probability, but 471 J.L. Bintliff, A.M. Snodgrass, 131, 1985. Low artifact density and a thin habitation layer has also been observed on excavations of villa complexes, K. Kepeski, 146, 1976; the author has concluded that the villa was inhabited seasonally, during the harvest season. 473 D. Pettegrew, 201, 2001. 474 I. Mikulčić, Sectio Naissus-Dyrrachion-Scupis-Thessalonike, Tabula Imerii Romani, K-34, Ljubljana 1976. 472 259 not necessarily erected near the agricultural estates of their owners. This indicates that individual farms and villas were a significant component in the settlement pattern of the Early and Middle Imperial periods in the Skopje Basin. The issue is inextricably related to the foundation of the Roman colony and the actual extent of colonization475. As explained earlier, there is very little archaeological evidence apart from the corpus of epigraphic monuments and the dislocated remains of funerary architecture. They contain undeniable evidence relating to the settlement of army veterans in the region and a number of examples have actually come from Skopian Montenegro. As one might expect, the majority of the funerary steles and votive monuments come from the lower portions of the basin, closer to the more fertile stretches along the Vardar Valley, Scupi and the main interregional road476. Some authors have even speculated that the peripheral zones of the region were left to the autochthonous population477. But these zones, the mountainside and the peripheral hilly regions can also offer important natural resources, such as pastures, timber or minerals. These regions of the Skopje Basin were an integral part of the territory of Roman Scupi and the distribution of epigraphic monuments erected over the tombs of army veterans confirms that they owned properties in the peripheral parts of the Skopje Basin. Looking at the spatial distribution and the chronology of these monuments, it is evident that they are not exclusively limited to the inner Skopje Basin, although they become scarcer in regions near the eastern and western limits of Scupi’s territory (maps V_36 and 37)478. The funerary steles of at least eight colonists have been found in the region of Skopian Montenegro, mostly dislocated or from uncertain locations (map V_38). One L. Marcianus, a veteran legionary of the VII Claudiae, later, a quaestor and duumvir in Scupi was buried at his estate, about 3 kilometers to the south of the second survey area479. He died at the end of the 1st or the early 2nd century AD and was certainly one of the earliest Roman colonists in this region. Another stele from the same period comes from the territory of modern Mirkovci, 1.5 to 2 kilometers west of the survey area. It was erected over the tomb of a veteran from the I Italica legion, though there is no evidence that he held offices in Scupi. Two other steles from the environs of this village belonged to veteran legionaries and city officials, one of which also honors a veteran of the VII Caludiae legion and consequently cannot be later than the early 2 nd century. Veterans from the VII Claudiae were also buried near the villages Banjane and Kučevište. The majority of modern researchers accept that veterans of this legion were among the first Roman colonizers of the Skopje Plain and the surrounding valleys 480. Their funerary monuments appear not only in the region of Skopian Montenegro, but across the entire administrative area of modern and Roman Skopje. Examples are known from as far as the northern entrance in the Taor Canyon, not far from the first survey area and from the valley of the Markova Reka, tens of kilometers away from Scupi. The wider study region, situated 7-8 B. Josifovska-Dragojević, 25-30, 1982; cf. T.W. Potter, 1979; P.A.J. Attema, G-J. L.M. Burgers, P.M. van Leusen, Regional Pathways to Complexity: settlement and land-use dynamics in early Italy, from the Bronze Age to the Republican Period, Amsterdam 2010. 476 B. Josifovska-Dragojević, 28, 1982; I. Mikulčić, Teritorija Skupa 463-480, Živa Antika 21, 1971; cf. T.W. Potter, fig 27, 1979. 477 B. Josifovska-Dragojević, 32, 1982. 478 Mostly after I. Mikulčić, map 1, 2, 1971; I. Mikulčić, 1982; Arheološka Karta na Republika Makedonija, vol. II, 362-387, 1996. 479 B. Josifovska-Dragojević, n.46, 37, 52, 68, 44, 54, 62 1982; L. Jovanova, Two Latin inscriptions from the village Kučeviška Bara, near Skopje 69-84, Macedonian Heritage 9-24, 2005. 480 D. Koračević, 2002, B. Josifovska-Dragojević, 25, 1982. 475 260 kilometers north of Scupi was at least partly occupied during this main wave of colonization, in the later part of the 1st century AD. Of particular significance is the fact that a great number of the monuments are dedicated to or by high-ranking city officials. We hear of praetors or aedils of Scupi who owned estates in some of the most peripheral parts of the Skopje Basin481. In the wider study region in particular, among the spolia built into the main parish church of Kučevište, there is a funerary stele commemorating an aedile and a member of the city council in Scupi. One of the steles from Mirkovci, dated to the mid-Imperial Period marked the tomb of another city council member and a quaestor. There is finally a votive altar, built into the main church in Ljuboten dedicated by an ex-slave and a member of the college of priests responsible for the Imperial cult in Scupi, the augustales. Apart from indicating that a wealthier class of citizens owned land in the wider study region, the epigraphic data hint at something far more significant regarding the character of the agricultural estates during the Early and Mid-Imperial periods. The abovementioned local functionaries had to be based in Scupi, at least during their time in office. This means that their agricultural estates were either run by middle-men or were only seasonally occupied. Doesn’t this explain then the humble character of the finds and the low artifact density on the clusters discovered in the second survey area? To be sure not a single epigraphic monument has been reported from the surveyed basin, although this could very well be a mere coincidence. After all the majority of the steles used as spolia in the village churches have an unknown provenance. Naturally there remains the problem of the indigenous, pre-Roman population of the Skopje Basin. One of the earlier researchers of the topography of Scupi and its necropoleis claimed that there existed a small community on Scupi’s acropolis immediately prior to the arrival of the Roman legions482. This claim has not been supported by later research, although there are no firm grounds to fully reject it. Personal names of indigenous people abound in later epigraphic sources, along with evidence for the existence of local religious cults483. Specifically for the wider study region, we have one votive altar dedicated to the deity Zbelturd, a local interpretation of Jupiter. It was found built into a ruined church between the villages Ljubanci and Pobužje, situated about 1.7 kilometers from the survey area484. Many of the inscriptions from Skopian Montenegro bear typical Roman cognomens, the Cornelii and the Rufii being among the most prominent. But there are a number of names unattested in other parts of the Empire, such as one Dardanus, a son of the veteran who owned land near modern Mirkovci or Solius Sur, mentioned on an epitaph from a site situated only 1 kilometer from the survey’s eastern limit. These men were part of the Romanized local population. They attained the status of Roman citizens either through military service or through inheritance. It is beyond any doubt that the autochthonous pre-Roman population existed side by side with the Roman colonizers and migrants from the Greek-speaking and Oriental provinces. What remains unclear is the type of settlements in which this community was organized. Quite possibly their presence in the Skopje Basin prior to the arrival of the Roman legions was not as strong as in the neighbouring region of the southern half of Upper Moesia. This partly explains the intensive colonization of the Skopje Basin and the surrounding valleys, as well as the 481 M. Basotova, A new veteran of the legion VII Claudia from the Colonia Flavia Scupi 405-409, Arheološki Vestnik 58, 2007; for steles of city-functionaries in the wider study region, B. Josifovska-Dragojević, n. 62, 68, 11, 1982 482 I. Mikulčić, 249, 1979. 483 B. Josifovska-Dragojević, 32-37, 1982; F. Papazoglu, 1969, especially 168-189. 484 B. Josifovska-Dragojević, n. 18, 1982. 261 relatively swift Romanization of the area. In comparison in the neighbouring region to the east of Skopje, the region of modern Kumanovo, typically Roman aspects of material culture begin to appear only in the latter part of the 2nd and the early 3rd century AD. Votive dedications and funerary steles, architectural sculpture, coinage and even typical Roman pottery are still very rare prior to the middle of the 2nd century AD485! Unfortunately we lack information about the exact locations on which the funerary steles of the earliest colonists in the study region were found. The majority of the funerary steles from Mirkovci are known to have come from fields along the valley of the Kučeviški, east of the village. The group of steles near the village Radišani, 2-3 kilometers south of the survey area were reportedly found west of the village, on a location very similar to the surveyed terrain486. Another group of Early to Middle Imperial funerary steles was recently discovered very near the latter site, in the lower parts of the surveyed basin, over 3 kilometers to the south487. The rest of the monuments found in the village churches have uncertain provenance. There is reliable evidence revealing that the architectural sculpture and funerary monuments were sometimes brought from distances of several kilometers. But the monuments found in-situ clearly demonstrate that there was a roughly even dispersal of isolated or groups of farms across the foothills. Apart from dislocated or monuments used as spolia, there is very little evidence from the zone of modern housing along the mountain foot. It remains uncertain if this is related to chances of preservation or if it reflects a genuine preference to settle the central and lower portions of the plain, the “lower shelves”. Apart from tombs and dislocated funerary monuments, on two sites in the wider study region there were remains of what was interpreted as settlement buildings. In fact one of the two sites is entered as a villa rustica in the archaeological atlas488. It is situated near the village Banjane by a copious fresh-water spring, 2.5 kilometers west-southwest of the survey area. The results of the rescue excavations were never published. It is only mentioned that the excavation revealed the foundations of a building, featuring several separate rooms arranged in the shape of the Greek letter Г. On the northern end it terminated with an apsidal room with a colonnaded porch. These elements of the building plan are indeed typical for villa complexes. The large amounts of re-used architectural sculpture perhaps point to a later date for this estate, the late 3rd4th century AD. It would be particularly interesting to learn if the earliest colonial farms were independent units of settlement, formed extended networks or were joined to existing rural communities 489. It wouldn’t be surprising if all three modes were applied, including close cohabitation between agricultural estates and suburban villas. During the construction of a railway line, only about one kilometer northeast of Scupi, an accidental discovery brought to light a funerary stele dedicated by the inhabitants of vicus Cavadinus to a member of the local community490. In the immediate vicinity of this find-spot, archaeologists have observed large concentration of surface finds datable to the Roman Period, identifying the site with the vicus mentioned in the inscription. This settlement was situated in the narrower ager of the colony. One of the largest concentrations of I. Mikulčić, 469-473, 1971b. B. Josifovska-Dragojević, 241-249, 1974. 487 L. Jovanova, 69-70, 2005. 488 Arheološka Karta na Republika Makedonia, 364, 1996. 489 A combination of a villa and a dependent hamlet spaced a few hundred meters apart are known from northern Bulgaria and central Greece, though in the Late Roman Period, V. Dinčev, fig. 59, 1997; J. Bintliff, P. Howard, A. Snodgrass, sites THS 2 and THS 16, 2007. 490 B. Josifovska-Dragojević, n. 86, 1982. 485 486 262 Early Roman funerary monuments honoring the citizens of Scupi comes from an area situated only 1.5 kilometers to its east. The fact that early Roman colonists were allotted land properties in various parts of the Skopje Basin perhaps suggests that the autochthonous communities were organized into a few small nucleated settlements, leaving larger swathes of unoccupied territory. This considerable dispersal of colonists’ estates could both aim at optimal exploitation of the newly conquered territory and at a more efficient Romanization and pacification of the local population. Apart from vicus Cavadinus, there are only two other open settlements in the narrower Skopje Basin; the rest of the known archaeological sites consist of isolated objects (villas?) and isolated or groups of tombstones (map V_37). Considering the accidental character of these discoveries and the lack of systematic research, it is very probable that the true number of open, nucleated settlements is many times greater. Yet the same approach has produced rather different results in regions in the western and eastern periphery of the Skopje Basin, near the limits of the colony’s territory. Here one still finds isolated funerary steles or votive inscriptions, but there are an equal or greater numbers of sites entered as open, rural settlements in the archaeological atlas. Judging by the evidence available in the published literature, the main focus of settlement for the Roman colonizers was the narrower Skopje Basin, including the wider study region and the neighbouring valley of the Markova Reka to the south. This brief overview of the settlement pattern in the region of Skopian Montenegro and in the wider region of Skopje during the Early and Mid-Imperial Period sheds further light over the possible character of the sites discovered in the second survey. It is evident that the group of agricultural estates revealed in the second survey was a part of an extensive network of villas and farmsteads, spread across the entire Skopje Basin by the beginning of the 2nd century AD. Even if they belonged to the local native families, they were deeply Romanized by the early second century. The poor character of the material and the low on-site artifact density probably indicates the lower social status of their owners, as well as the fact that apart from the residence there were a number of non-residential, subsidiary buildings. In fact they could have still belonged to richer, city based landholders, but inhabited and maintained by middle-men and workers. It is nearly impossible to arrive at a certain conclusion without more comparative data. Surrounded by estates of colonizers and city officials, it would be rather unusual if the survey area was an enclave inhabited by local farmers. However the case of vicus Cavadinus demonstrates that such close coexistence was not uncommon. At the same time, the fact that the revealed farms form an integral network capped by a fortification and possibly a special-purpose site runs against the common perception of isolated villa-complexes. One wonders how much of this perception is actually shaped by our poor knowledge of settlement types other than the larger villa sites. The extant archaeological evidence from the Skopje Basin and the wider study region in particular demonstrate that this pattern of individual or groups of agricultural estates and villas survived well into the 4th century. For this period there is even less evidence to rely on, as inscribed steles or altars are extremely rare in most parts of the Balkan Peninsula after the late 3rd century AD. Nevertheless on the few excavated villa complexes in the country, there is a clear evidence for 4th century occupation491. The partly excavated villa near the village Banjani probably dates within the period between the late 3rd and the late 4th century. The much larger corpus of excavated evidence from Bulgaria confirms that by the late 4th century, life definitely Arheološka Karta na Republika Makedonija vol.2, 364, 1996; K. Kepeski, 155-156, 1976. 491 263 ended on the great majority of villas in the countryside492. They were either completely deserted or literally squatted on and turned into small hamlets. This chronology is in accord with the dating proposed for the collected pottery finds and the date suggested for the fortification. The patterns established in the Early Imperial period survived uninterrupted until the last quarter of the 4th century and at least for some regions in the country, the late 3rd-4th century is a period of expansion. However by this time a profound transformation was already starting to take place. In the landscape of the countryside the changes were announced by the reintroduction of low fortified hilltops493. In many cases the old pre-Roman fortifications, such as the one near Brazda were refortified, but a certain number were built anew. Unlike the fortifications built in later centuries, this first wave of fortification was mostly limited to the transitional hilly zone, close to the arable land and near the main roads494. In many examples the simple planning and the humble masonry technique have prompted researchers to conclude that a number of these forts were built on the initiative of the local communities, chiefly for reasons of security495. Indeed the near complete absence of surface remains on many of these forts indicates that the focus of everyday life was elsewhere. Nevertheless their appearance paves the way for the future developments. Many fortifications will be renewed and expanded during the following two centuries and on a certain portion of them, there is undeniable evidence for permanent occupation. The small fort discovered in the second survey area is most probably part of this first wave of fortification. It isn’t an isolated example in the region of Skopian Montenegro (map V_39). Apart from the abovementioned fort near Brazda, another fort was discovered about 2 kilometers east of the survey area, occupying an almost identical location. Another two forts from this period were documented near the villages Bulačani and Viniče, about 7-8 kilometers to the east. In these latter cases the forts have already retreated to the mountainside, although they are still relatively close to the arable zone. Similarly planned hill-forts have been discovered in other parts of the Skopje Basin. Another novelty introduced during the reign of the Tetrarchs and Constantine’s dynasty involved a change in the burial customs. Skeletal burial replaces cremation, inevitably initiating changes in the funerary architecture. These changes were first observed on the large city necropoleis of Stobi and Scupi, but monumental tombs and sarcophagi also appeared isolated in the countryside496. At least three monumental sarcophagi dated to the mid-4th century come from the wider study region497. They were already mentioned while discussing the epigraphic evidence. In all three cases, the sarcophagi were constructed of reused funerary steles, often cut or broken into pieces and builtin the sarcophagi. By the end of the 3rd century AD the habit of V. Dinčev, 18-20, 1997; A. G. Poulter, 241-242, ed. N. Christie, 2004; A. G. Poulter, 51-98, ed. A.G. Poulter,2007. 493 I. Mikulčić, 2002; M. Milinkovitch, 159-192, ed. J. Henning, 2007; J.H.W.G. Liebeschuetz, 101-134, ed. A.G. Poulter, 2007. 494 It must be admitted though that the true extent of this first wave of fortification is to a large degree unclear, as the dating mostly pertains to the most conspicious, usually later phases. 495 I. Mikulčić, 64-65, 1982; J.H.W.G. Liebeschuetz, 107, ed. A.G. Poulter, 2007. 496 This subject has not yet received an adequate, synthetic study; for Scupi, I. Mikulčić, 109-143, 1974; (Stobi) A. B. Wessolowsky, 97-137, eds., J. Wiseman,Dj. Mano-Zissi 1973; (“isolated” examples) V. Lilčić, Ranohristijanska zasvedena grobnica kaj s. Orman, 143-146, Macedoniae Acta Archaeologica 7-8, 1987; Ð. Petački, 79-88, 1997. 497 A. Keramitčiev, Edna docnorimska grobnica od s. Pobužje, skopsko 143-148, Živa Antika 13/14, 1964; B. Josifovska-Dragojević, 241-249, 1974; L. Jovanova, 69-84, 2005. 492 264 erecting funerary or votive inscriptions dies out in the central Balkan provinces 498. The fact that they used the tomb-stones of their predecessors, effectively erasing their memory, certainly suggests a definitive breach with earlier tradition. Although lacking precise information about the location of the finds, it is clear that at least one of the 4th century sarcophagi was discovered in the same locality where earlier inscribed steles were reported. The 4th century landowners in the region retained the old burial places, though they were obviously unrelated to the earlier landowners. The tombstones of the old landowning elite were merely seen as an easily accessible source of building material. Neither the Early and Mid-Imperial funerary inscriptions nor the three fourth century sarcophagi can be related to the network of farmsteads revealed in the second survey area. The closest one of the sarcophagi was discovered at a distance of about 1 kilometer from the northeast corner of the survey. Obviously the find-spot is too distant and it could hardly belong to some of our sites, although it is quite possible that they were contemporary. Because of the low chronological resolution of the finds, it is impossible to assess the impact of the social and cultural transformations on the settlement pattern in the wider study region. But looking at the distribution of the known 4th century sarcophagi in the Skopje Basin, it is evident that the network is much sparser in comparison to the network of earlier funerary steles. This could be largely related to the character of the finds, sarcophagi fragments being more difficult to recognize than inscribed tombstones. However one shouldn’t exclude the possibility that this reflects an actual change in the structure of landownership and local economy499. The last quarter of the 4th century marks the end of the old settlement pattern. According to the archaeological atlas and other sources, with the exception of Scupi, the inner part of the Skopje Basin was almost completely abandoned during the 5th and the 6th centuries500. The latest category of finds from this zone,are the abovementioned 4th century sarcophagi. As discussed in previous chapters, this seemingly radical breach with earlier settlement applies not only to the region of Skopje but to most central Balkan provinces501. The last two centuries of Antiquity are marked by an apparent decline and contraction of the old urban centres and an intensive campaign of fortification and refortification in the countryside. Around 50 forts have been discovered and documented, solely in the region of Skopje502. Their actual number is probably even greater, although not all were built anew in the 5th and 6th century AD. As one might expect, there is a considerable variety regarding size, position, planning and construction. In contrast to the forts of the late 3rd and 4th century, the new fortifications were built in mountainous regions occupying barely accessible locations, away from the major plains and valleys. And yet a number of these new fortified centres exhibit a surprisingly high level of sophistication in urban planning and construction. These were by all standards of the time urban settlements, with secured water-supply, public buildings and high-quality construction. At the same time it seems that the old network of villas and farms was completely abandoned. For the administrative region of modern Skopje, the ratio between open settlements dated to the Roman and Late Roman Period in the archaeological atlas is 10 to 1. It should be stressed that this figure doesn’t include fortifications or funerary monuments. In principle surface 498 A. G. Poulter, 9-11, ed. A.G. Poulter, 2007. Similar tendencies have been observed in northern Bulgaria, A. Poulter, 242-243, ed. N. Christie, 2004. 500 Excluding Scupi, only 3 open settlements are dated to Late Antiquity, none of which is located in the central parts of the basin, Arheološka Karta na Republika Makedonija vol. 2, 609-615, 1996. 501 I. Mikulčić, 2002; A. Poulter, ed. 2007; M. Milinković, 159-161, ed. J. Henning, 2007. 502 I. Mikulčić, 1982. 499 265 clusters dating to the late 4th through the 6th century shouldn’t be less visible than surface clusters of pre-4th century date. On the contrary, it’s been justly argued that the ceramic material of the Late Roman Period is far more obtrusive and recognizable than pottery from the Early and MidImperial phases503. It is equally possible that some of these sites, solely known from the brief entries in the archaeological atlas were only broadly dated to the Roman Period and that they also had a Late Imperial phase. It is almost certain that in reality, the number of open rural settlements from both periods is greater, but this doesn’t demean the significance of the mentioned ratio. The few open settlements dating to Late Antiquity come from the western periphery of the region, the same areas in which open settlements from the Early and MidImperial periods were mostly concentrated. The second survey area and also probably the wider study region didn’t escape this tendency of nucleation and retreat. The survey showed that by the late 4th century, the network of farms occupying the entire eastern half of the survey was completely abandoned. There are no remains of intensive, residential activities dating to the 5th and 6th century AD. Judging by the number of known monumental archaeological remains, the wider region of Skopian Montenegro experienced a significant shift in the focus of settlement. On the plain at the foot of Mt. Montenegro, not more than four sites can be dated to the Late Imperial period, none of which has received a proper publication504. One of these sites is an Early Christian basilica, vaguely located in the southwest periphery of the region; while the other three are necropoleis dated on the basis of the characteristic cist burials. Again there exist only vague descriptions of their exact locations and the surrounding context. According to a brief entry in the archaeological atlas, one of these necropoleis is situated 1 kilometer to the north of Mirkovci, about 1.8 kilometers west of the survey area. This is very close to the location of the abovementioned villa rustica near Banjane and it is possible that this necropolis belonged to a rural settlement that succeeded the earlier villa. The second necropolis was accidentally discovered during a road construction in modern Pobužje, on the south periphery of the traditional core of the village and about 1.5 kilometers north of the survey area. Finally, the third necropolis is situated only 1.2 kilometers east of the latter, between the villages Pobuže and Ljubanci close to the mountain foot. According to the information available in the literature, on all three sites there were a larger numbers of burials suggesting that these were communal necropoleis. Compared to the number of accidentally discovered monuments from the previous centuries, there is an obvious reduction. But at the same time, one has to take into account the fact that while earlier Roman steles were determined as isolated or small groups of burials, the Late Roman cists were interpreted as parts of communal necropoleis. One should also point to the new change in burial traditions, introduced with the spread of Christianity in the countryside. Being less lavish and accompanied by rare finds, the burials from this period are certainly less conspicuous than the funerary monuments from the Early and Mid Imperial Period. It is also evident that there is little continuity with earlier burial sites. If there were inscribed steles in the immediate vicinity of the Late Roman necropoleis, it is very unlikely that they would’ve gone unnoticed, although it is possible that earlier steles were used in the construction of the cists. All of the three necropoleis were situated close to the mountain foot, near the sites of the Late Medieval and Ottoman villages. In fact the necropolis between Pobužje and Ljubanci occupies an empty settlement niche, identical to those occupied by the rest of the villages in the region. 503 D.K. Pettegrew, 743-784, 2007. Arheološka Karta na Republika Makedonija, vol. II, 375, 378, 380; I. Mikulčić, Tragi od ranohristijanski baziliki okolu Skopje, 97-123, Godišen Zbornik na Filozofskiot Fakultet. 33, 1981. 504 266 The central portions of the foothills, including the intensively surveyed area were apparently abandoned. Either because of security or changes in the local economy, settlement moved to the less exposed mountain foot, closer to the mountainside. This indicates that the roots of the Late Ottoman and Early Modern settlement pattern can perhaps be traced back to the period of Late Antiquity. It is unfortunate that this thesis cannot be confirmed by a systematic research, as the entire area along the foot of Mt Montenegro is under modern housing. But a clue to the possible location of the nearest Late Roman settlement is perhaps present in the surface record of our survey area. Although there definitely lacked settlement remains from this period, there was a very thin off-site carpet, gradually becoming denser towards the northwest corner of the surveyed basin. In this respect too, it strikingly resembles the later carpets of Ottoman and Early Modern off-site debris. This thin cover of Late Roman material must have originated from a settlement situated somewhere in the direction of Kučevište. We mentioned the small chapels of St. Paraskeva, immediately to the north of the survey area and the chapel of St. Athanasius, about 600 meters from the survey’s northwest corner, as potential settlement location. Knowing that the Late Byzantine-Early Ottoman off-site was generated by a settlement most probably situated beneath the present-day village, one shouldn’t exclude the possibility that the Late Roman settlement was also situated at a greater distance from the survey area. This however would imply that like its Late Medieval and Ottoman successors, it was a larger nucleated settlement. It is in any case difficult to imagine a different source for this relatively widespread and even carpet of surface material. Recall that a very similar phenomenon was observed in the Sopot survey, where even a settlement smaller than one hectare produced an off-site carpet that spread over a distance of several hundred meters. Although far from certain, the fragmented evidence points to a possible nucleation of settlement and retreat from the central portions of the foothills. Traces of 6th century occupation were discovered at the fort near Brazda, though the extent of this phase is unknown 505. That the focus was increasingly shifting towards the mountainside is confirmed by the building of at least three new fortifications, up to several kilometers into the massif of Skopian Montenegro 506. Two were discovered to the north of Kučevište, along a road that leads across the mountain and a third one, about 4 kilometers northeast of Ljubanci. Apart from architectural remains, there is very little surface evidence on these sites. At present they are completely covered by forest and ceramic fragments are extremely rare on the surface. The few collected fragments of tile and pottery and above all the masonry, unambiguously determine them as Late Roman. The one near Ljubanci in particular cannot be dated earlier than the 6th century507. These forts occupy topographic units that are strikingly similar to the ridges included in the survey area. As discussed in chapter IV, the relief of the mountain interior is basically a dramatic replica of the relief in the foothills. There is however one essential difference and this is the scarcity of arable land in the mountain interior. This implies that either the retreat to the mountain interior was related to a change in the local economy or the fortified centres weren’t permanently inhabited. In this context, it is noteworthy that settlement in the mountainside is not uncommon in later periods. I. Mikulčić, 72-73, 1982. N. Čausidis, , 183-197, 1989. 507 N. Čausidis, 187-189, t. 2a, 2b, 1989; this is clearly indicated by the masonry and the appearance of pentagonal towers. 505 506 267 Both archaeologists and historians that work in the central Balkan regions agree that the dense network of Late Roman fortifications came to an abrupt end by the end of the 6th and the beginning of the 7th century AD508. However research in the past decade on a number of sites in the Aegean and central Greece has revealed definite traces from this time-period, suggesting that on certain sites life continued well into the 7th and even the 8th century509. Although one has to allow for regional differences, the fact that the ceramic assemblages associated with this period are a peculiar mixture of pottery with low diagnosticity and certain forms that survive from the period of Late Antiquity warns us that the absence of evidence from these centuries could easily reflect the difficulties in recognizing this material. The complex sets of historical events and processes that brought about the sudden end of this era are well known and there is little use in repeating them. During the several centuries after the end of Antiquity, the Skopje Basin shares the fortune of the rest of the central Balkan Peninsula510. Archaeological and textual evidence is too scarce, even when compared to certain prehistoric periods. In the entire administrative region of modern Skopje, only two sites have been dated to the period between the early 7th and late 9th century. In both cases the discoveries were made during excavations of earlier sites and very little is known about the extent and the exact character of the Early Medieval remains511. One of these sites is Roman Scupi, raising the abovementioned issue of the possible continuity of occupation on certain Late Antique settlements512. Near the city-baths, brief rescue excavations revealed a small concentration of pottery related to a post-Antique occupation layer. Based on stratigraphic observations, the material exhibiting close resemblance to the local Late Roman pottery production was dated to the 7th and the 8th century. The second site with remains datable to this period is situated in the central parts of the Skopje Basin, approximately 12 kilometers east of Scupi. The results of this excavation have not been fully published. Only the pottery was analyzed as a part of the small corpus of Early Medieval pottery from the region of Skopje513. The site was accidentally discovered in the near vicinity of a prehistoric necropolis. Unlike the material from Scupi, the pottery exhibits more significant deviations from the Late Roman traditions, although the basic forms are repeated. It is interesting to note that both sites are located not far from the Vardar Valley, in the inner Skopje Basin. They occupy gentle and fertile stretches, covered with Quaternary sediments. Based on the character of the pottery discovered, it has been suggested that these were small rural communities, but their true extent remains unknown. There are no material remains from this period on the surface of the second survey area. As in the first survey, the period between the early 7th and the beginning of the 10th century remained elusive. Knowing the humble character of the accidentally discovered traces in Scupi, it becomes clear that more sites from this period can hardly be expected from the traditional method of field survey. Add to this the fact that even pottery experts have difficulties in V. Popović, 230-257, 1980; I. Mikulčić, 2002; A. Poulter, 223-254, ed. N. Christie, 2004. A.K. Vionis, J. Poblome, M. Waelkens, 147-165, 2009; P. Armstrong, Trade in the Eastern Mediterranean in the 8th century, 159-178, ed. M. Mango, Byzantine Trade, 4th-12th century, Ashgate Publishing 2009; A. K. Vionis, 570-578, J.L. Bintliff et al. 541-606, 2004-2005. 510 J.H.W.G. Liebeschuetz, 130-132, ed. A.G. Poulter, 2007. 511 B. Ristevski, Ranosrednovekovna sadova keramika od Republika Makedonija, Vizantiiskii Vremenik 89 (65), 2006. 512 D. Koračević, 165, 2002; B. Ristevski, Early Medieval finds from Scupi, 71-81, Macedonian Heritage 20, 2002; cf. J. Bintliff, 384-387, 2012. 513 D. Mitrevski, Praistoriska nekropola Klučka-Hipodrom kaj Skopje, 61-89, Zbornik na Arheološki Muzej 1, 1995; B. Ristevski, 51-52, 2004. 508 509 268 recognizing the material from this period, it is no wonder only a dozen sites from the entire country are dated to the Early Middle Age514. As the scanty findings from this period have only been recently synthesized and studied in greater details, it is possible that there is a greater number of Late Roman sites with unrecognized Early Medieval phase515. The number of the sites securely dated to the Middle and Late Byzantine Period in the region of Skopje indicates only a very weak and slow recovery (map V_40). Up until the 14th century, the map of archaeological remains in the Skopje Basin, including coin hoards and sacral monuments appears grossly incomplete. For the entire administrative region of modern Skopje, the archaeological atlas lists not more than 13 sites including coin hoards and vaguely dated surface remains, which could very well belong to the Early Ottoman Period. Securely dated settlement remains have been discovered only on 6 locations, four of which are fortified centres516. This situation doesn’t correspond to the information in the written sources. By the mid 13th century, we hear not only of Skopje and the surrounding forts, but also of a number of rural settlements and monasteries. In fact according to the written documents, some of these settlements were established as early as the 11th century. After the Byzantine re-conquest of this region, in the early decades of the 11th century, Skopje became the main centre of Byzantine power in the central Balkans. The town and the wider region were also the power base of later separatist kingdoms and often changed hands between regional rulers517. Bearing in mind the geo-strategic importance of the region, its relative fertility and favorable climatic conditions, it is difficult to accept that the countryside in the region of Skopje was so sparsely populated prior to the 14th century. Excluding the fortified centres and sacral monuments, there are only two sites datable between the 11th and the 14th century in the entire Skopje Basin. Thus it turns out that there is hardly any increase from the Early Middle Age. One of the sites where remains from this period have been discovered is actually Scupi, though it isn’t clear if there is a direct continuity with the Early Medieval phase. The Middle Byzantine remains are more substantial, including an earthen floor, a refuse pit and a number of pit-burials, dated between the late 9th and the 11th century. They were discovered on a different micro-location, but still within the urban core of ancient Scupi518. The second open settlement dated to this period is situated even closer to the survey area, in the western periphery of the wider study region 519. The site is situated 800 meters northeast of the village Gluvo and 2.4 kilometers west of the survey area. The salvage excavations were limited to a group of three pits, probably dug for industrial purposes. Approximately 200 meters from the group of pits, the author mentions an agglomeration of building remains. Giving little further information, the two sites are related and identified with a village mentioned in later historical sources under the name Marušane. The researcher dated the pits to the period of the 11th and 12th century on the basis of the discovered pottery and tile. The report contained brief remarks about the excavated ceramics. Fine glazed examples were discovered alongside tile Z. Beldedovski, 45-49, Zbornik na Štipskiot Naroden Muzej VI, 1990; Arheološka Karta na Republika Makedonija, vol. II, 1996. 515 B. Ristevski, 1035-1059, 2006. 516 B. Babić, Wann sind die ältesten Teile der mittelalterlichen Burg Kale in Skopje gebaut, 359-364; Actes du VIIIe Congrès international des sciences pré et protohistorique, Beograd 1971; I. Mikulčić, 118-135, 1982; V. Lilčić, Markov Grad, Matka; ukrepen srednovekoven manastir, 268-280, Istorija 19-2, 1983. 517 T. Tomoski, Skopje od XI do XIV vek, 57-74, eds. V. Mošin et al, 1975. 518 D. Koračević, 165-173, 2002. 519 M. Bilbija, Novi arheološki naodi od Skopje i okolinata, 134 -135, Macedoniae Acta Archaeologica 4, 1978. 514 269 fragments, both finding parallels among the material from the Skopje citadel. It is noteworthy that the material found in Scupi and in the second survey area was very different, almost entirely consisting of plain utilitarian and coarse ware. There is no detailed information about the size and the exact micro-location of the excavated site. It is only certain that it was situated by a local stream, close to the mountain foot. The surrounding terrain is much gentler than in the survey area. Further evidence for pre-14th century rural settlements in the wider study region comes from the literary sources520. Of particular importance for the region of Skopje and Skopian Montenegro in particular are two imperial edicts dated to the middle and the end of the 13th century. These two documents describe in great details the rights and the properties of the main monastery in Medieval Skopje, St. George-Gorgos or Nikephoros. Apart from being an invaluable historical source for the agrarian relations, the social hierarchy and legal regulations, they reveal a great number of toponyms in the Skopje Basin and the surrounding regions. Both documents confirm the rights and land granted to the monastery by earlier Emperors, beginning from the founder and the first donor, the Byzantine Emperor Roman III, who reigned in the second quarter of the 11th century. Thus indirectly they inform us about the agrarian arrangements in the region of Skopje, for the period immediately after the Byzantine conquest of Skopje in 1004. The texts reveal a populated, agriculturally exploited countryside, parceled into fields, vineyards, summer and winter pastures, meadows, watermills, gardens and forests. Except for the land and villages belonging to the monastery, it is evident that a considerable portion of the land was given to the military aristocracy. This is hardly surprising knowing that the region of Skopje was on the northern periphery of the Empire, during most of the 11th and 12th century. Most important for the present study is the direct mention of the village Pobužje and of the region of Montenegro, which presents an undeniable proof that the area was inhabited and recognized as a minor administrative unit by the middle of the 11th century. The surface archaeological evidence also indicates that the surveyed area was reoccupied sometime in the period between the 10th and the 14th century. The new settlement was many times smaller than the dispersed network of Roman farms. Its remains consisted of a tiny collection of pottery fragments, probably distributed in one larger and one smaller, ill-defined cluster. Their combined areas totaled not more than 2500 sq meters, equaling the size of the medium-sized clusters of Roman material. This was in other words a single farmstead or a small group of households. According to the surface remains, the focus of agricultural production was on the slopes and on the top of the western ridge (map V_41). These were lightly constructed dwellings, for there were no traces of building ceramics or stone rubble. We’re again dealing with an ultra-thin carpet of surface material, perhaps indicating a short-lived occupation. In this context however, attention must be turned to the fact that because of the peculiarities of the cuisine and dinning practices of the time-period and the increased usage of metal vessels, Medieval and post-Medieval sites tend to produce smaller ceramic assemblages than sites from Antiquity or certain prehistoric periods521. One may recall the small collection of pottery fragments, which we associated with Late Byzantine-Early Ottoman Sopot. Although this settlement was five times larger than the Medieval settlement in the second survey, it produced a fairly small volume of finds. 520 521 V. Mošin, 97-177, eds. V. Mošin et al, vol. 1, 1975. A. Vionis, 177-197, eds. J. Bintliff, H. Stoeger, 2009. 270 The extant literary sources provide a potential clue about the socio-historical character of this settlement522. The above-mentioned imperial edicts are very precise when describing the villages and their territories given in possession of the monastery of St. George in Skopje. In a number of instances while defining the limits of the monastic land, they talk in great details about the various types of assets that came along with certain villages (vineyards, watermills, fishing ponds, or hunting reserves) including minor, satellite hamlets. While the names of larger central villages have often survived until the present-day, the memory of these smaller satellites has almost been completely erased. Only in a few, rare cases were their names preserved in the local toponomy. One should also point to the cases where the ruler gives the specific rights to a landowner to settle small communities of dependent farmers on his own estate. Such small communities could quietly disappear, leaving only faint remains in the surface records. Regardless of whether the discovered cluster of Medieval pottery came from a separate settlement unit or a larger settlement’s satellite, from a long-term perspective it seems to herald a return to the dispersed settlement scheme, characteristic for the Early and Mid-Imperial Period. Apart from the brief report on the rescue excavations near Gluvo, there is very little archaeological evidence to support this thesis. The approximate location of a third village situated in the foothills is revealed by local toponyms, though we don’t know if it predated the 14th century. It was situated somewhere along the lower course of the Jazirski Stream, at a distance of over two kilometers from the mountain foot and the modern village Ljuboten. The location is very similar to the Middle Byzantine remains near Gluvo: both sites occupy locations on the valley floors, near the confluence of two streams. On the basis of the extant historical and topographic data, it is nearly certain that by the early 14th century the majority of the villages in the region were firmly established at their present-day locations. Seven of the ten presently extant villages are mentioned under the same or very similar names in the historical sources of the time523. The beginning of the 14th century marks a new era in the demographic history of the country. Historically this is the period of expansion of the Serbian Kingdom524. Already by 1282 Skopje and the surrounding regions were conquered by the Serbian King Milutin. By the end of the 13th century, the army of the Serbian King reached Thessaly and the plain of Thessalonica. After long and uncertain negotiations, the Byzantine Emperor transferred the land along the Upper Vardar as a dowry to King Milutin. This historical event is of a great importance for Skopje and the wider study region in particular, because after the treaty with Byzantium in 1299, the Serbian King will gradually begin to move the royal court to Skopje, closer to his new ally. Discovering the newly conquered land completely ravaged by the constant warfare in the last decades of the 13 th century, we are told that the king started an intensive building campaign, particularly focused on the city of Skopje and on the surrounding regions525. According to his biography, over 40 churches were either restored or built anew during his reign and particularly in the period between 1300 and 1321. Although the actual extent of this building campaign remains dubious, through the study of inscriptions, historical documents, studies of architecture and fresco-decoration, it’s been V. Mošin, 124-146, eds. V. Mošin et al, vol 1, 1975. V. Mošin, 128-129, 263-324; eds. V. Mošin et al, vol 1, 1975; these references in the historical documents are summarized by J. Trifunovski, 31-34, 1971. 524 T. Tomoski, 60-63, eds. V. Mošin et al, vol. 1, 1975; G. Ostrogorsky, History of the Byzantine State, 1957. 525 V. Mošin, 309-337, eds. V. Mošin et al, vol. 2, 1977. 522 523 271 ascertained that at least a dozen churches were built in this period, both in Skopje and in its surroundings. Two, possibly three of these churches were built in the wider study region. These are the parish church in Čučer, the monastic church St. Nikita, near Banjane and possibly a small chapel near Kučevište526. The church of St. Nikita near Banjane is particularly important, because it is mentioned in an extant edict of the Byzantine Emperor Michael IX, where the Emperor confirms the rights and land given to the monastery by his brother in law, King Milutin527. This authentic document presents a precious historical source, pertaining directly to the region of Skopian Montenegro. It is here that we for the first time hear of the villages Banjane, Kučevište and its unnamed summer camp and the abovementioned village Lopušane, near the Jazirski Stream. All these villages, along with the people and the land were given in the possession of the restored monastery of St. Nikita, later, all together transferred to Chilander on Mt. Athos. Thus in essence almost the entire region of Skopian Montenegro was a monastic land, shared between the monasteries of St. George in Skopje (a monastery that had earlier possessions in this region, including a portion of the village Pobužje), St. Nikita near Banjane (owning the greatest portion of the land) and the newly built monastery of the Holy Archangels in Prizren, Kosovo (built slightly later and given one half of Kučevište and its summer camp). Only the villages Ljubanci and Ljuboten, the easternmost pair of villages in the region were definitely in the hands of the military aristocracy. It is equally important to note that the edict had to be issued no later than the first decades of the 14th century, which implies that these settlements were established by the beginning of the century. Whether they already existed prior to the Serbian conquest or were newly colonized communities of serfs is impossible to know without archaeological research. In the historical sources there are no direct mentions of colonizing activity528. The only indisputable fact is that a significant portion of the military aristocracy and the monastic communities came from the old core of the Serbian Kingdom, northern Kosovo, the western parts of present-day Serbia and the east of Bosnia and Herzegovina. More archaeological data are needed in order to determine if the beginning of the 14th century saw the establishment of a new settlement pattern in the region or it merely revitalized the existing network of hamlets and villages. The fact that so many of the country’s pre-Ottoman monuments date to the 14th century could simply reflect the fact that this was the last phase in which fortifications and lavish monastic complexes were built. Nevertheless the considerable difference in the number of archaeological sites dated prior to and after 1300 cannot be wholly attributed to external factors. The archaeological evidence available from the wider study region, including the survey results demonstrate that the old Middle-Byzantine network of settlements wasn’t completely renewed during the 14th century. Excavated material from the site Marušane near Gluvo, on the western periphery of the foothills is dated no later than the second half of the 12th century. In the survey area itself, the finds that can be securely dated to the end of the Byzantine and the Early Ottoman periods rarely accompany the assemblage broadly dated to the Middle Age. This S. Nikolovska, Izgradeni i obnoveni manastiri i crkvi od kralot Milutin, 509-519, eds. V. Mošin et al, vol. 2, 1977. 527 V. Mošin, 317-324, eds. V. Mošin et al, vol. 1, 1975. 528 This is implied, though never explicitly stated in modern historical studies. As pointed out in one of these studies, the campaign of founding or renewing abandoned monasteries could have been one of the crucial instruments of installing new monastic centres and legally expropriate land captured by arms, V. Mošin, 168-169, eds. V. Mošin, et al, vol. 1, 1975. 526 272 circumstance largely influenced the dating of the Medieval assemblage; fragments that belonged to some of its constituent fabric groups never appeared alongside the Late Byzantine-Early Ottoman off-site collections. It presents the base for the assumption that by the time of the 14 th century, the surveyed terrain definitely became a part of the settlement territories of Kučevište and Pobužje and that the hypothetical satellites pre-date the 14th century and King Milutin’s building campaign. It is in any instance clear that the small Medieval farm or a hamlet presented the last settlement phase in the survey area. Although the exact date remains uncertain, it was in all probability a satellite of some larger settlement, which implies that the survey area was part of a larger settlement’s territory even prior to the 14th century. Judging by the locations of the oldest village churches, the major settlements in Skopian Montenegro didn’t suffer considerable displacements after being incorporated into the Medieval Serbian Kingdom in the late 13th century (map V_42). As already explained the network of fortified towns, monastic centres and villages established during the 14th century will survive the turbulent decades of the Ottoman conquest and prosper during the first three centuries of Ottoman rule. To be sure there were significant settlement and population displacements in certain regions of the country, but specifically for the region of Skopian Montenegro, the Ottoman conquest of Skopje didn’t introduce radical changes regarding the continuity of population, settlement pattern or agrarian relations. The rights of ownership were merely transferred from the monasteries of St. Nikita and St. George to the newly erected mosques in Skopje. In fact according to the historical evidence available, it is very possible that the monastery of St. George retained its autonomy until the mid-decades of the 15th century, 40 years after the conquest of Skopje. The most direct testimony to the wealth and size of the villages in Skopian Montenegro during the Early Ottoman Period comes from the official, exhaustive censuses of the 15 th and 16th century529. According to the earliest census available, dated to the middle decades of the 15th century, the majority of the villages in the region featured between 30 and 60 households, ranking them as medium-sized villages. Only Banjane stands apart with over 100 houses, though in this case it seems that the population of the neighbouring villages Čučer and Gornjane were subsumed under Banjane. These two villages are not mentioned separately in this census, although it is clear that Čučer was established prior to the Ottoman conquest. In the next exhaustive census for the years 1467/1468, there is an evident population increase in nearly all villages in the region. Pobužje almost reaches the maximum of around 60 households, while Banjane, along with Čučer and Gornjane counted 158 families. Other larger, presently active villages, such as Brazda and Mirkovci are mentioned for the first time in the 15th century. Knowing the size of the latter, it is likely that these two villages existed prior to the Ottoman conquest, but weren’t mentioned in the imperial edicts, because they belonged to the military aristocracy. But the real population explosion happened in the late 15th and the early 16th century, when the number of houses in some of the villages more than tripled530. It is unfortunate that the exhaustive censuses from this century are yet to be fully translated and published. According to the brief remarks in the publications of the earlier 15th century censuses, Kučevište and Mirkovci had over 100 families by the first quarter of the 16th century. From the information available in 529 M. Sokolovski, A. Stojanovski,1973; M. Sokolovski, 534, 537-538,1971;M. Sokolovski, Vakafi i vakafski imoti vo Skopje i skopsko, vo XV i XVI vek, 42, 51 Prilozi VIII, Makedonska Akademija na Nauki i Umetnosti, 1977. 530 Cf. J.L. Bintliff, 193-224; eds. A. Hurst, A. Schachter, 1996; J. Bintliff, 437-441, tab. 20.1, 2012. 273 the published studies, one cannot be certain if this number refers solely to these villages or it includes the population in the neighbouring dependent hamlets. From the Early Ottoman censuses we also learn that the settlement network in the wider study region was more extensive. Apart from the major villages, situated along the foot of the mountain, there were an unknown number of smaller, long-since abandoned villages. Mostly thanks to the ethnographic studies carried out in the early part of the last century, it is now clear that the majority of these settlements were situated in the mountainside531. Faint traces are preserved in the local toponomy and in the memory of older inhabitants. These villages were situated on the small mountain plateaus or on the gentler ridges, usually near natural springs. The closest example is the village Brodec, situated 3 kilometers northeast of Pobužje, deep into the mountain massif. The ruins of another deserted village, known from the historical sources and oral tradition as Zgurovce can still be seen on the ridge that rises above Ljubanci. It is situated less than 2 kilometers from the latter, but it is mentioned separately in the 15th century Ottoman censuses, counting around 20 houses. Toponyms hiding the names of old villages can also be found above Kučevište and Banjane and further into the mountainside. It is certain that some of these settlements existed prior to the Ottoman conquest. The fact that they are not mentioned as separate settlements in the exhaustive censuses, perhaps suggests that they were treated as dependent hamlets of the larger villages at the mountain foot, Kučevište and Ljubanci. It is noteworthy that the micro-locations of these mountain hamlets are very similar to those occupied by the villages in the foothills. They are regularly positioned on the top or the upper portions of the ridges, usually facing southwest. The Late Roman forts also exhibited similar orientation, though they were obviously located on less accessible, narrower micro-topographic units. Only one village mentioned in the 15th century censuses can be located on the plain at the foot of the mountain. This is the abovementioned village by the Jazirski Stream, situated about 2 kilometers southwest of Ljubanci and Ljuboten. According to the local toponomy there is a place called Lopušani near the confluence of the Jazirski and the Pobuški Stream. A village with this name is indeed mentioned in the 14th century sources, but it never appears in the Ottoman censuses. Instead there is a village called Jazirce, obviously situated somewhere along the Jazirski Stream. Possibly Lopušani changed its name into Jazirce after the Ottoman conquest, as the valley is too small to host two separate communities. On the other hand the village known as Marušane in the local oral tradition, near Gluvo, doesn’t appear in the Ottoman censuses. It could have survived as a dependent hamlet of its neighbour, though as discussed it was partly excavated and dated a couple of centuries prior to the Ottoman conquest. The list of taxes paid in cash or kind indicates a varied agricultural economy532. Apart from the usual taxes paid for cultivating cereals, garden cultures and fruits, taxes were paid for the raising of sheep, pigs and silkworms. Taxes were also paid for at least 5 water-mills and for wine production. A mid-16th century document informs us that the inhabitants of Kučevište were skilled builders and craftsmen. They were largely responsible for the building of an aqueduct, which not only fed Skopje and its baths with running water, but also irrigated the fields, vineyards and gardens surrounding the town. In fact in another document from 1568, the highest juridical official for the region of Skopje made the inhabitants of Kučevište permanently S. Tomić, 420-429, 1905; J. Trifunovski, Tri isčezla Srednjovekovna sela, 156-157, Glasnik Srpskog Geografskog Društva I/2, 1970. 532 M.Sokolovski,A Stojanovski, 281-282, 1973; M.Sokolovski, 534, 538, 1971. 531 274 responsible for the water-supply and irrigation of Skopje and its surroundings. In return they were relieved from a number of taxes and obligations533. This brief overview of the settlements of Skopian Montenegro in the first couple of centuries of Ottoman rule clearly demonstrates that the pattern established in the 14th century experienced further development and possible extensions. The bulk of the population was concentrated in the villages, forming a chain along the mountain foot. Only one settlement can be located in the central parts of the foothills, at its southern periphery. Finally, a smaller portion of the population lived in small hamlets, situated deep into the mountain massif. This second known reoccupation of the mountain after the Late Roman Period was certainly related to the local pastoralist economy. The rugged terrain surrounding the mountain hamlets offers very little arable land and this circumstance supports the scanty literary evidence for local semi-nomadic herding. These hamlets must have developed from temporary, spring and summer camps. They are regularly positioned by the local mountain roads and near mountain-springs. Immediate access to water would have been of utmost importance for the stationing of larger herds534. The complexity and the integrity of the Late Byzantine-Early Ottoman settlement pattern in the wider study region are apparent. This network of settlements encompassed a fairly large territory, featuring diverse relief and natural resources. In contrast to the hamlets from the first survey area, these communities were bound to share a larger territory with few physical boundaries. In this light it is easier to understand why the communities inhabiting Skopian Montenegro formed one of the earliest administrative units in the country, nearly becoming a separate ethnic group. It was surprising to discover that by the middle of the 16th century, the majority of the settlements of Skopian Montenegro were by all standards medium and large-sized villages, some approaching the size of Classical Greek poleis. Banjane, Mirkovci and Kučevište all had near or over 100 households by the early decades of the 16th century. In fact some of these settlements like Pobužje were more populous in the 16th than in the early 20th century. Despite the richness in agricultural and other resources of the wider study region, it was unexpected to find such a large concentration of rural population. Unlike the settlements in the first survey area, these are communities of a higher rank, sufficiently large to allow for intermarriage between members of the same community. Social cohesion and order were partly achieved through horizontal subdivisions, reflected in the existence of satellite hamlets and separate quarters or mahalas within the limits of the central settlements535. These large communities of farmers and herdsmen left their mark outside the narrower settlement areas, in the surface record of the fields at the foot of the mountain. Among the vast amounts of off-site debris spread across the entire survey area, there was a small percentage of finds that could be dated prior to the 18th century. They too were found dispersed across most of the survey area, but the carpet of finds became visibly denser in the northwest survey quarter and especially on the basin’s floor, on the fields by the Skopje-Kučevište asphalt road. As with the Late Roman finds, its distribution pattern clearly pointed to the settlement that generated this material. This was the village Kučevište. Consisting of over 100 houses already towards the end of the 15th century, it was a true giant even in comparison to many modern villages. The size and the position of this village make it the obvious source of the off-site carpet discovered in the second survey. It is noteworthy that Pobužje, the smaller eastern neighbour of Kučevište left a 533 M. Sokolovski, 50-51, 1977. J. Trifunovski, 34, 1971; C. Chang, P.A. Tourtellote, 249-264, 1993. 535 J.L.Bintliff, 511, ed. G. Barker, 1999. 534 275 humbler mark in the surface record on the eastern ridge. Bearing in mind the very large population concentrated in these villages, the discovery of a thin off-site carpet dating to the Late Byzantine-Early Ottoman Period is not totally out of place. Somewhat paradoxically, beginning from the 17th century there are very few published historical accounts pertaining to the wider study region. The corpus of published Ottoman censuses belongs almost entirely to the period of the 15th and 16th century. It is particularly unfortunate that we know so little about demographic developments in the period of the last decades of the 17th and the early decades of the 18th century. According to later historiography and the oral historical traditions recorded by ethnographers at the very beginning of the 20th century, this was a period of radical demographic changes, commonly related to the aftermath of the Austro-Ottoman Wars of the late 17th century536. The changes were brought about by two complementary migratory movements. The first was the exodus of an unknown portion of the local population in the years following the end of the Austro-Ottoman war in 1689. Fearing reprisals, families that allied with the Austrians against the Ottoman authorities retreated with the Austrian army beyond the Danube. Although this is confirmed by written evidence, the real effects of the migration can only be guessed at without the precise information of the official censuses or archaeological data. The other migratory process allegedly started about a century later and lasted throughout the first half of the 19th century. It’s been maintained that the vacuum left by the earlier exodus was filled-in by a wave of settlers from northern Albania and Kosovo. Being largely a transhumant population, the new colonists were particularly opting for the lush mountain pastures. By the early decades of the 19th century, they reached the mountains west of Veles, significantly changing the ethnic and religious composition of the population in many regions of the country, especially in the northwest. In Skopian Montenegro, most affected were the small communities that inhabited the mountainside, such as Zgurovce or Brodec. But in principle the migration shook one of the bases of the regional economy and affected all communities in Skopian Montenegro. Maintaining the large flocks of sheep was impossible without free access to mountain pastures. Pressed by the new migrants, they were forced to abandon the high mountain pastures along the main mountain ridge, while the old population concentrated in the villages at the foot of the mountain. By the beginning of the 20th century, the importance of sheep and goat herding had visibly lost its significance in the local economy, although the end of this tradition came only with the industrialization of the country and the radical socio-economic reforms in the middle decades of the last century. One cannot fully discredit the local oral traditions as it is undeniable that by the second half of the 19th century, the villages in the central part of the mountain massif were 100% Albanian and the small villages on the southern slopes of the mountain, mentioned in Late Byzantine and Early Ottoman sources were completely abandoned by the beginning of the 19th century537. In all probability the semi-nomadic herding retained some role in the local economy, but because of the tensions with the competing transhumant population of Albanian settlers, the focus was slowly shifting towards the foothills and the Skopje Basin538. But in spite of the declining political and economic situation during the last centuries of Ottoman rule, the villages of Skopian Montenegro were teeming with inhabitants towards the end of the 19th century. With nearly 200 houses in 1905, Kučevište was probably the largest J. Trifunovski, 22, 1971; S. Tomić, 453, 1905. S. Tomić, 450-454, 1905; J. Trifunovski, 34-46, 1971. 538 J. Trifunovski, 51, 1971. 536 537 276 village in the Skopje Basin and one of the largest in the country539. According to the same census, although situated only 2 and 3 kilometers away, both Mirkovci and Ljubanci had over 100 households. In fact population growth in the region will continue unchecked until the second half of the 20th century. The two Balkan and World Wars apparently had no long-term negative effects. As in most other parts of the country, the decline of these old rural communities begins only in the second half of the last century. However because of their long history and rich tradition and because of the proximity to Skopje, the villages of Skopian Montenegro didn’t share the fortune of the majority of villages in the country. They are still one of the largest rural settlements in the Skopje Basin, although population growth has evidently come to a halt in the past several decades. These large communities increasingly relying on the agricultural exploitation of the farmland in the foothills produced a very dense off-site carpet. The density of artifacts datable to the last couple of centuries was further enhanced by the positioning of the villages above the plough-zone, a circumstance that facilitated the transportation of manure to the overexploited fields540. The Late Ottoman-Early Modern off-site carpet confuses with its sheer amount, density variations across the fields and finally, with the variety and quality of the finds. Almost 12 000 fragments of pottery and tile were recorded over an area of 1 sq kilometer, lying at a distance of up to 2.5 kilometers from the centre of the village. Such vast quantities of discarded material were unexpected, but they truly reflect the very large size, the convenient positioning and the wealth of the villages in the wider study region. In all likelihood, they had both developed their own ceramic production and enjoyed a ready access to ceramic goods produced in Skopje. The landscape that presently characterizes the flat portions of Skopian Montenegro, with its dense field divisions, tall hedges and terrace walls could very well be the product of this last phase of occupation. The historic and ethnographic evidence, the appearance of the modern landscape and the portion of the surface record revealed through surface artifact survey are all in unison about the developments in the wider study region during the period of the last two centuries. That this last period of occupation in the wider study region was indeed marked by an increased pressure on the natural resources is finally suggested by a rough estimate of the region’s carrying capacity. Looking at the agricultural land available in the foothills of Skopian Montenegro, one has to conclude that the agricultural resources were truly being stretched to their limits. With not more than 35 sq kilometers of arable land, Skopian Montenegro could sustain about 750 households, assuming that the size of the average estate was in the region of 55.5 hectares and that on average one household included about 7 individuals. On the other hand, the official censuses list over 1000 households in this region at the end of the 19 th century. It is thus evident that arable land was becoming a precious resource and that the average size of individual estates was closer to 4 hectares. In addition one has to account for the fact that 539 J. Trifunovski, tab. 4, 1971. There is an explicit testimony to the importance attached to manure by the local farming communities in the ethnographic work of S. Tomić, 433, 1905; explaining the logic behind the villages’ locations at the very edge that separates the foothills from the mountainside, the author writes: “This is because of purely economic reasons. The peasant who apart from farming engages in herding avoids the plains, i.e. the central parts of the agricultural land and settles the fields’ edges, at the mountain foot, so that both the fields and the mountain are at an equal distance… In these villages it is also easier to bring the manure to the fields. The village is above the line of the fields and then the manure is simply smeared downhill rather than being carried uphill. And the rain itself washes away the manure that accumulates around the houses spreading it over the fields and meadows.” This last detail about the “manure that accumulates around the houses” clearly implies that household rubbish could easily be included in the manure. Cf. H. Forbes, 163-64, ed. R. Jones, 2012. 540 277 throughout the Ottoman Period, a considerable percentage of the agricultural yields from this region went to the secular and religious feudal lords in Skopje. Thus the large communities in the study region had to live off an even narrower subsistence base. During the Early Ottoman Period this scarcity of land was compensated by the exploitation of extensive mountain pastures and other resources in the mountainside, but in the last couple of centuries the local communities had to find alternative sources of existence, including increased specialization in crafts and trade and temporary migration to the urban centres in the region and abroad 541. In the light of this evidence it is much easier to understand the appearance of such a dense carpet of discarded ceramic artifacts from the last couple of centuries. Enhancing the fertility of the soils was essential for securing at least an average yield. This practice continues to the present day, with pottery gradually disappearing from everyday use and the rubbish heaps and synthetic fertilizers partly replacing organic manure. With so few periods represented in the surface record, it is obviously difficult to observe a clear long-term pattern in the history of settlement in the second survey area. Even more so than in the first survey area, in this specific geographic setting it is nearly impossible to understand the developments in one section of the region without having at least indirect information about the developments in other parts. Evidently one could say that the settlement history of the second survey area consisted of brief episodes of ephemeral occupation, alternating with long intervals during which the basin was abandoned or became the agricultural territory of a larger, nucleated settlement. Remaining unclear is how this dynamics relates to the developments in the wider study region, which are very vaguely known. In the majority of cases it is impossible to say if the settlement remains discovered are independent units of a wider, dispersed network or satellites of larger settlements. Even for the Roman Period for which we have the largest corpus of archaeological evidence from both the survey area and the wider region of Skopian Montenegro, a certain number of issues were left undecided. Are we seeing a peculiar form of dispersed settlement or a concentration of individual farmsteads representing a segment of a wider network? What was the community’s relation with the citizens of Scupi, some of which we know held property in the wider area? On the other hand, only for two, possibly three periods does the surface evidence indicate that the episodes of abandonment were in fact periods when the survey area became the agricultural territory of a larger nucleated settlement. This obviously happened during the Late Ottoman-Early Modern Period, but also during the Late Byzantine-Early Ottoman and also possibly during the Late Roman Period. There were no remains of permanent occupation dating to these three periods in the surface record, but the appearance of more or less evenly dispersed off-site carpets suggested that there had to exist a larger nucleated settlement, situated not far away from the survey area. Whether the survey area was occupied or turned into an agricultural land, data from the surrounding regionsare crucial for understanding the developments in the intensively surveyed area. Much time and energy was spent in trying to synthesize the scarce archaeological data from the wider study region and the Skopje Basin in order to relate the developments at different regional scales. It is possible to infer some very general conclusions, but without more detailed evidence from the wider study region, they remain in the realm of working hypotheses and speculations. One thing in common for all three periods of settlement is the size and character of the basic settlement unit. Whether it stood isolated or as a part of a wider network, the basic settlement unit never exceeded the rank of a farm or a small hamlet. As discussed earlier, 541 S. Tomić, 462-464, 1905. 278 throughout its entire settlement history, the survey area never achieved the status of a stable settlement locus. It almost did during the Roman Period, but even then the area was inhabited by less than a dozen families. There never developed a larger, nucleated settlement. As we’ll see in the concluding chapter, this was to a great degree predetermined by the specifics of the regional geography rather than by the lack of certain resources or the limited carrying capacity of the region. The second survey area was certainly not deficient in space or agricultural resources, but its place in the wider regional context made it a settlement locus of a lower rank. The surrounding settlement loci were preferred for a number of reasons and because of their proximity and the lack of significant topographic barriers, the survey area often became part of the territory of the larger neighbouring settlements. Interestingly even in the first survey area, which retained a level of stable settlement niche throughout most of the periods during the last 7000 years, we observed a surprising level of continuity regarding settlement size. The basic cycle of development in the second survey area essentially consisted of transformations from vacant territory/agricultural land, to dependent farms or hamlets and to a tight network of (independent?) farms. The latter level wasn’t exceeded and it is possible that it represented a unique episode in the local settlement history, rather similar to the Late Iron Age expansion in the first survey area. The surface evidence clearly demonstrated that for the greater part of its history, the second survey area was either a vacant land or a part of another settlement’s territory. The brief episodes of occupation always coincided with major socioeconomic and demographic transformations in the wider study region and in the Skopje Basin. Knowing the very high level of integration of Skopian Montenegro and its proximity to a larger civic centre, this is hardly a surprise. As pointed out, a number of issues remain completely unanswered or only vaguely determined. More evidence is needed in order to securely determine the character of the occupation during the Middle or the Late Byzantine periods. It is impossible to decide if the small hamlet belongs to a time when the chain of nucleated settlements at the mountain foot was not fully established or if it was a short-lived satellite of Kučevište’s predecessor. A more precise dating of the finds will certainly help settle this issue. Judging solely by the location of the site, its size and character, it seems that this wasn’t a return to the Roman network of dispersed farms, but rather an early phase of the Late Medieval and Ottoman network of villages and hamlets. It is also unclear if this pattern had a predecessor during the Late Roman Period; the survey data and the scanty information from the wider study region suggest that this could very well be the case. Finally, apart from the issues surrounding the socio-historical character of the group of Roman farms, we don’t know if this was a unique episode or if it had a distant precedent in prehistory. The faint traces of prehistoric occupation merely confirm that the second survey area was never occupied by a stable, nucleated settlement. These and other questions pave the way for future research, which unfortunately has to be focused on one of the presently occupied settlement niches. We suspect that it is these locations that attracted human settlers in this region during most periods of its settlement history. If this hypothesis proves correct, it will present a striking testimony to the stability of settlement locations in the regions along the Vardar Valley. 279 Chapter VI: General conclusions VI.1 Patterns of habitation and land-use in the two survey areas In most periods of the past when the survey areas were inhabited, the bulk of the surface material was found concentrated on a single location. These locations were usually referred to as the central sites. It has been generally accepted that major surface clusters featuring high artifact density and varied ceramic assemblages are the remains of past settlements542. Although there are no generally applicable rules, in most instances there are clear indicators, period and region specific, that one is dealing with remains of intense occupation: large quantities of architectural ceramics (though not necessarily), carbonized pieces of wattle and daub, fragments of cooking stoves etc. For the survey areas presented in this study, all major surface clusters were interpreted as settlement remains. They all usually featured balanced ceramic assemblages (roughly equal percentage of coarse – cooking, storage and transport vessels – and fine, table and utilitarian ware) and there were specific artifact categories that indicated domestic occupation, such as the abovementioned (brick and tile, pithos, braziers). On some of these sites we also found evidence for other types of activities, industrial and cultic. The sheer density and the extent of the central clusters preclude alternative explanations. Needless to explain, the settlements are the central element of the inhabited landscapes. Their size and spatial arrangements directly reflect the size and structure of the local community, while their location indicates the focus of local production and attitudes to other types of resources, including security and communication. However throughout this study we saw that settlements or the central sites were not the only anthropogenic features in the landscapes. Apart from the settlement sites, the hyper-intensive surveys revealed a whole range of different phenomena in both survey areas. These include the intermediary and low density extensive scatters, the site halos and the off-site, as well as the so called satellite clusters, characterized by intermediary to high artifact density and areas much smaller than the central sites. What follows is a brief summary of the phenomena revealed by the surface artifact surveys in the two areas. It seems that post-depositional processes at least partly contribute to the phenomenon of site edge. These narrow belts of intermediary density running along the site periphery have become differentiated by the smearing of freshly unearthed archaeological material or site erosion. The processes are initiated by natural forces, such as soil erosion, the activity of animals and cultivation. It partly explains the low state of preservation of the finds collected from the site periphery, as well as the fact that they can often be observed even on small, possible nonresidential sites.They rarely measured more than 10-20 meters in width and they ought to be differentiated from the more extensive halos and off-site scatters that could very well be the product of original discard behavior. However, in certain cases when this intermediary belt is wider (the Late Iron Age in the first survey area, the Roman sites from the second survey), it is impossible to distinguish between natural and cultural agencies. Both the natural weathering of sites and the more intensive cultivation of the fields that are nearest to the settlement work to produce an extensive scatter of intermediary density or a site halo543. Perhaps the most controversial of all the phenomena discussed is the interpretation of the extensive low density scatters or the off-site. Basically the debate revolves around the following 542 For various definitions see J.F. Cherry, 394-397, eds. D. R. Keller, D.W. Rupp, 1983; T.W. Gallant, 408-409, 1986; J. Bintliff, P. Howard, A. Snodgrass et al, 16-18, 2007. 543 S.E. Alcock et al, 159-160, ed. I. Morris, 1994; J.L. Bintliff, E. Farinetti, et al, 665-674, 2006. 280 question: are the extensive low density scatters mainly the result of natural, post-depositional process of site weathering or are they the result of past anthropogenic factors or the remains of intense spreading of manure in the past544? Some of the findings of the surveys presented here provide clear support for the latter view. Most significantly there is ample ethnographic evidence pertaining specifically to the second survey area for the practice of spreading manure and other debris on the fields prior to the autumn and spring ploughing545. It is important to stress that these accounts date to the very beginning of the 20th century, the period to which the bulk of the off-site debris collected in the two survey areas was dated. One problem with adopting the ancient manure hypothesis specific to these two smallscale surveys is that extensive zones of ancient manure are usually associated with larger urban centres, many times the size of the small, rural sites that were the subject of this study. To be sure, off-site scatters are not unique for large urban or semi-urban centres. Research has shown that the extent of this zone varies proportionally to the size of the settlement that has generated it546. Smaller sites will naturally leave smaller impact areas. But in the first survey area, (due to the low resolution of the large block survey) it sometimes proved rather difficult to determine if one is dealing with an ultra-thin off-site carpet limited to the same topographic unit as the settlement or with small, ephemeral satellite clusters. It seems that only for the Late OttomanEarly Modern, the Roman-Late Roman and possibly for the Late Iron Age and the Late Byzantine-Early Ottoman periods is there clear evidence for intensive field manure. At this point we need to make a distinction between the situations recorded in the first and the second survey. In the former a thin carpet of mostly Late Roman finds spread continually over the entire western half of the survey area, featuring maximum artifact density not greater than 1 fragment per 100 sq meters. In the eastern survey sectors, across the central valley, this thin carpet of surface material disappears and the Roman finds were found concentrated in small density clusters. We discovered an equally nuanced picture in the second survey area. On the basis of the artifact densities recorded by the transect and the grid survey, it was possible to distinguish between two off-site zones. The one nearer to the farm-sites and spreading over much of the survey’s eastern half featured maximal artifact density of up to 5-6 fragments per 100 sq meters. This is the site halo, a zone characterized not only by higher artifact density, but also by more dramatic fluctuations than in the rest of the off-site547. The off-site segment further away from the sites and spreading over the western survey half was characterized by artifact densities consistently lower than 1 fragment per 100 sq meters. In terms of artifact density and patterns of distribution, this segment of the off-site is similar to the Roman-Late Roman off-site in the first survey area. At a first sight it appeared that site halos were absent around Roman-Late Roman sites from the first survey, although it is equally possible that we failed to document them. The micro-topography of the terrain in the first survey, as well as the peculiar locations of the Roman-Late Roman sites could further contribute to the weak prominence of this phenomenon. The largest site 5a-b was delimited on two sides by steep ravines and we weren’t allowed to collect finds from the fields situated immediately to the east of the site. Similarly the Late Roman farm on site 8 was surrounded by overgrown stretches on all sides. However in both of these cases the off-site carpet did become slightly denser on the fields closer to the sites and in 544 S.E. Alcock, et al, 160, 164-165, ed. I. Morris, 1994; A. Snodgrass, 197-200, the same volume; J.L. Bintliff et al, note 1, 2006. 545 S. Tomić, 433, 1905. 546 T.J. Wilkinson, tab. 1, 1989; J. Bintliff, P. Howard, figs. 2 and 3, 1999. 547 Cf. J.L. Bintliff, 29-38, Pharos XIII, 2006; J. Bintliff, P. Howard, A. Snodgrass et al, 23-26, 2007. 281 the immediate surroundings of site 5a-5b it was possible to observe small, low peaks comparable to those recorded on the halos of the Roman sites in the first survey. Low density scatters were recorded even on the fields surrounding sites 13a-13b and 14, on the otherwise sterile eastern ridge, as well as around site 8 in the second survey, predominantly made up of architectural ceramics. Adopting the view that this increased density in the off-site is the result of intensive manure and discard of rubbish, the very presence and the extent of the halo zone is instructive of the residential nature of these sites. In principle the Roman settlement in the second survey with its fairly structured off-site zone is only superficially unique, though it could signal different discard behavior and different agricultural practices, as well as different taphonomic conditions. But in essence the pattern is not much different than that recorded for the settlements from various other periods in the first survey. In both micro-regions, the total dispersal area of certain chronological category of finds was usually found limited to the topographic unit on which the contemporary settlement was located. During the analysis of the results, the total dispersal areas of the chronological categories represented in the surface record was regarded as an indicator of the settlement’s impact area or its inner territory, regardless of the mechanisms that generated it. This doesn’t refer exclusively to the land under intensive agriculture, nor is it necessarily defined by a continuous carpet of ultra-low density. We saw that the latter was characteristic only for a few periods in the past. Most of the pre-Roman periods discovered in the first survey, as well as the Medieval occupation in the second survey lacked continuously spreading off-site carpets. The main settlement was rather accompanied by a few or several clusters of intermediary density and a much smaller size. It is the distribution of these satellite clusters that was taken as indicator of the settlement’s inner area. As explained earlier, these small and elusive clusters were usually discovered within the same micro-topographic units, at a distance of not more than a few hundred meters from the settlement. Their interpretation remains highly problematic, not least because they were often recognized only after the processing of the finds. Moreover it is obviously related to the interpretation of the settlement in social and economical terms and to the specific period in question. These phenomena can represent the remains of a number of landscape features normally associated with settled, agro-pastoral communities: animal sheds, industrial facilities, refuse pits, votive offerings and burials. Without comparative evidence from the surrounding regions, one can but speculate. At this point, one wonders if the settling of this issue remains beyond the limits of intensive surface survey. For most of these scatters, it is actually questionable if even excavations or geo-physical prospection can offer a solution to the problem548. For the purposes of the present study, it is important to acknowledge that their distribution roughly coincides with the limits of the same topographic entity occupied by the central cluster, the settlement. As one might suspect, this fairly simple scheme of a settlement marking its inner territory by a series of satellite features or an extensive off-site carpet, usually limited to a single, microtopographic entity wasn’t necessarily the norm. In certain periods of the past in both survey areas, the human landscape was far more complex, extending over several topographic units. However the basic principles are not necessarily changed, merely the scale is different. This fact is best exemplified by the Late Iron Age settlement in the first survey area, where the scheme of one central and a number of satellite clusters is repeated at a micro-regional level. 548 A.M. Snodgrass, Survey archaeology and the rural landscape of the Greek city, 113-136, eds. O. Murray, S. Price, The Greek city from Homer to Alexander. Oxford 1990. Also see the often ambiguous results of the Laconia Rural Sites Projects, W. Cavanagh, C. Mee and P. James, et al, 2005. 282 The Late Roman Period in the same survey area offers a seemingly similar picture. The entire basin, in this case, virtually its every corner was occupied or at least covered with a sparse layer of ceramic finds. In reality however, the situation is far more complicated. It is again possible to recognize a central cluster (site 5a) defined by its larger size and varied ceramic assemblage, but now it is accompanied by a larger number of satellite clusters, some of which are not much different from the central cluster. But there are also considerable differences between these secondary or satellite clusters. As mentioned, some are very similar to the central cluster in terms of artifact density, intra-site distribution or the composition of the assemblage, the only difference being the size (site 5b, the Late Roman phase on site 8). Others are characterized by the very small amounts of fine table ware and tile, often found in closely spaced, tiny clusters (such were the rare finds collected from field units situated south and southwest of site 5a and possibly the clusters from sector I). A third group of clusters, whose size remains ill-defined but is probably much smaller than site 5a are characterized by the composition of the ceramic assemblage, consisting almost exclusively of architectural ceramics and rare fragments of storage vessels. In addition there are what appear to be tiny isolated concentrations of architectural ceramics, such as those discovered on the southern tip of sector IX, near site 11 or on the two newly discovered forts, sites 9 and 10.All of this point to a highly developed and functionally stratified landscape, each of its corners being optimally exploited. Also differing from the model of a central cluster plus satellites bound within certain topographic limits is the Middle Neolithic settlement. But in contrast to the far more extended and developed scheme characterizing the Late Iron Age and the Late Roman periods, during this phase the total dispersal area of the surface finds practically overlapped with the limits of the settlement. Not a single shard dating to the Mid-Neolithic was discovered outside the central cluster. These findings are in accord with what has been learned so far about the local Early and Mid-Neolithic cultures. Excavations have shown that at least some of the activities associated with locations outside the settlements in later periods, such as burial or refuse disposal, were carried out within the living space of the Neolithic communities549. The settlements from the Roman Period in the second survey and to a certain degree, the Late Bronze-Early Iron Age settlement in the first survey area, offer yet another different pattern. To be sure there are considerable differences between the organization of settlement during these two periods. A common characteristic for both phases is the existence of at least several clusters of roughly equal size, spaced at regular intervals across a single topographic unit. This scheme is most pronounced in the second survey area during the Roman Period, where it is impossible to single out one of the dozen clusters as the main focus of settlement. Rather they seem to be arranged in relation to the small fortification occupying the top of the ridge and lacking a substantial surface record. In the case of Late Bronze-Early Iron Age Sopot, it is possible to observe a central and a slightly larger cluster, surrounded by minor satellites spaced at short distances from each other. The further away from the central cluster, the sparser the network of satellite clusters. In both of these cases, it was suggested that at least some of the satellite clusters were the remains of domestic occupation. Apart from the artifact density there are no other indicators of their residential or non-residential character. It is quite possible that these are examples of dispersed rural settlements, a settlement type for which there are hardly any parallels in the archaeological literature from the region, but which are known from later historic M. Garašanin, 82-106, ed. A. Benac, 1979;M. Garašanin, 89, 1983; J. Chapman, “Rubbish dumps” or “Places of deposition”: Neolithic and Copper Age settlements in Central and Eastern Europe, 347-362, ed. A. Ritchie, Neolithic Orkney in its European Context, Cambridge 2000; P. Halstead, ed. 1999. 549 283 periods, the Ottoman and Early Modern periods550. Admittedly this type of rural settlements, known as dispersed villages in the ethnographic and geographic literature, are usually associated with the rugged, mountainous regions, where animal husbandry plays an important role in the local economy. The two survey areas present a very different environmental setting in which the nucleated village was the norm, at least in later historic periods. Most of the settlement sites discovered in the two surveys measured between 1-2000 sq meters and 1 hectare. This is not an untypical finding for intensive surveys, especially focused on the rural sectors551. A smaller group of domestic assemblages occupied less than 1000 sq meters, but these were either satellite clusters (a few of the Roman sites in the second survey) or vestigial remains of earlier prehistoric settlements of the former group (the Bronze Age settlement in the first survey or the ultra-thin, late prehistoric scatter in the second survey). It is quite probable that the smallest independent settlement unit measured not less than 1000 sq meters. Such small settlement sites could only represent individual farmsteads, consisting of a single building and probably, an adjacent yard552. Examples come from both survey areas and from different timeperiods.These include the Bronze Age, the Roman phase on site 12 and possibly the the Hellenistic Period in the first survey, the Roman Period sites 6, 7, 10, 26 and 27 and the Medieval site 25 from the second survey. Table VI_1: Size and possible rank of the settlement sites discovered in the two survey areas First survey Site Area in sq num/dating m Sites 3/BA 825 Site 4/LB-Eot 10 000 Sites 5a+5b/R 10 000 Site 5a/EI >2-3000 Site 6/EI 4000-5000 Site 7/LBA-EI >5000 Site 8/LIA 36 000 Site 8/LRom Site 11/MNeo Site 11/Undat Site 12/Hell Site 12/Rom 3500 8500 9000 1800 1500 Rank Site num. Second survey Area in sq m Rank Farm Hamlet Hamlet? Hamlet? Hamlet? Hamlet? Small village Farm Hamlet Hamlet Farm Farm Sites 1-2/R Site 3/R Site 5-11/R Site 6/R Site 7/R Site 10/R Site 15-18/R 5-6000 < 500 >3500 1200 500 2000 5500 Farm? Farm? Farm Farm Farm? Farm Farm Site 26/R Site 27/R Site 25/Med 2500 1850 1750 Farm Farm Farm When it comes to the Roman and the Late Roman periods, there is no reason to put the sites measuring up to several thousand sq meters into a different rank. Indeed there are examples of larger Roman villas from Serbia and Bulgaria occupying areas of over 1 hectare, while in Greece it’s been ascertained that both Hellenistic and Roman farms tend to be larger, often J. Cvijić, 1918. T.W. Potter, 1979; J.F. Cherry, Appendix A: register of archaeological sites on Melos, eds. C. Renfrew, M. Wagstaff, 1982; M.H. Jameson, et al, tab. 4.7, 1994; R.W.V. Catling, 162-163, eds. W. Cavanagh et al, 2002. 552 R.W.V. Catling, note 22, eds. W. Cavanagh et al, 2002; J. Bintliff, P. Howard, A. Snodgrass, et al. 134135, 2007. 550 551 284 exceeding 0.5 ha553. But it is somewhat more problematic to determine the character of similarly sized settlement sites from earlier periods, such as sites 6 and 7 from the first survey. There are hardly any parallels in the wider region from the period of transition between the Late Bronze and the Early Iron Age or from the first couple of centuries of the 1st millennium BC. The few settlement sites from this period are almost exclusively hill-forts known from extensive surveys. In principle, the only difference between the prehistoric settlements on sites 6 and 7 in the first survey and sites 1-2, or 15-18 from the second survey is the lower maximal artifact density recorded on the latter.On the majority of sites from the Roman and Late Roman periods, the maximum artifact density rarely exceeded the threshold of 25-30 fragments per 100 sq meters. In comparison, the Early Iron Age settlement on site 6 in the first survey featured up to 45 fragments per 100 sq meters and its predecessor on site 7, over 50 fragments per 100 sq meters. This is not necessarily related to the longevity of occupation, as the prehistoric assemblages collected from sites 6 and 7 were dated within periods not longer than three centuries. However it could very well reflect a higher intensity of occupation during the prehistoric periods, changed living standards and different organization of the living space554. Basically the question is whether the small prehistoric sites from the first survey are individual farmsteads or very small, nucleated settlements consisting of not more than a few families or a single clan. It is obviously difficult if not impossible to answer this dilemma solely on the basis of evidence from surface artifact survey. On-site artifact density and distribution can merely offer ambiguous hints, as these variables are determined by a wider range of site-specific factors. With the data presently available, we can only conclude that in a number of periods in the past, rural settlement was of the rank of individual farmsteads or very small hamlets. In fact, the settlement rank most commonly associated with the present-day countryside, the Late Ottoman-Early Modern village was only achieved in the first survey area and only in certain time-periods. These include the Mid-Neolithic settlement and its successor on site 11, the Late Iron Age settlement on site 8, the Late Byzantine-Early Ottoman village (site 4) and possibly, the Roman-Late Roman settlement on site 5a-5b. 553 D. Srejović, et al. Gamzigrad: an imperial palace of late classical times, Belgrade 1983;V. Dinčev, 1997; L. Mulvin, 377-413, eds. W. Bowden, L. Lavan, C. Machado, 2004; S.E. Alcock, 87-135, 1989; M.H. Jameson et al, 255-256, 1994; H. Bowden, D. Gill, Roman Methana, 77-83, eds. C. Mee, H. Forbes, 1997; 554 J. Bintliff, K. Sbonias, Demographic and ceramic analysis in regional survey, 244-258, eds. R. Francovich, H. Patterson, Extracting Meaning from Ploughsoil Assemblages. Oxford 2000. 285 Graph VI_1: Maximum radii for the site halos and impact zones and estimated site areas Revealed site areas and max halo radii 1000.00 Site 8 I 900.00 800.00 Maximum halo radius 700.00 Site 11a I 600.00 500.00 Site 7 I 400.00 Site 8a I 300.00 Site 12 I 200.00 Site 6 I Site 4 I Site 25 II 100.00 Site 26 II Site 3 I Site 27 II Site 11 I 0.00 0.00 Site 5a-5b I 5000.00 10000.00 15000.00 20000.00 25000.00 30000.00 35000.00 40000.00 Revealed site areas In order to have a fuller understanding of the land-use patterns during the different periods of settlement, we also had to turn to the other forms of surface artifact phenomena. Measuring the extent of the halos that surround the Roman sites in the second survey, we encountered a series of problems, despite the fact that we were dealing with more or less continuous artifact scatters. In the first survey, the task is made even more difficult by the absence of a continuous site halo. To be sure, we can easily measure the distance between the edge of the site and the furthest occurrence of an artifact from the same period, but one cannot be sure if this impact zone is fully identical to the halos around Roman sites in the second survey. Furthermore the comparability of the extent of the impact zones of the sites in the first survey is problematic in itself, as these sites date to different epochs, with different technological capacities and perceptions of the physical surroundings. Nevertheless the size of the settlement was certainly one of the decisive factors concerning the extent of the impact zones. The larger nucleated settlements from the Sopot survey predictably featured impact zones much larger than the halos surrounding Roman sites from Skopian Montenegro. On graph VI_1 in addition to the sites from all known periods with clearly established site areas, we added two medium-sized Roman farms from the second survey, sites 26 and 27 and the Medieval farmstead on site 25. These three sites alongside the small Bronze Age farm from the first survey occupy the lower end of the scale. They would have been joined even by the largest Roman farms from the second survey area, as these too have radii shorter than 200 meters. In this respect the prehistoric and the Roman-Late Roman hamlets from the first survey area are a scale higher, with impact zones stretching over distances between 200 and 400 meters. In addition, the higher rank of the Late Iron Age settlement in the first survey is reflected in the very large radius of the impact zone, reaching nearly 1 kilometer. The close correspondence with the extent of impact zones around sites of different ranks and from different time periods in the 286 Levant is indeed striking555. There too settlements of the rank of farms or hamlets (lumped into a single category of settlements occupying less than 1.5 hectares) have halo radii between 200 and 400 meters, while villages (occupying between 2 and 9 hectares) have halo radii measuring between 600 and 1000 meters. But as in the second survey the correlation between settlement size and the extent of the impact zone is not particularly strong and in a few instances there are considerable deviations. Apart from the Middle Neolithic hamlet leaving no visible impact on the surroundings but the narrow peripheral belt of intermediary density enveloping the site, these include the Hellenistic farmstead and the possible Late Neolithic hamlet on site 11. These two settlements have impact zones slightly larger than expected. The Hellenistic farmstead with a site area estimated at 1800 sq meters has an impact zone with a maximum radius of about 300 meters, while isolated finds possibly dating to the Late Neolithic were collected at distances of over 600 meters from the site’s northern edge. It is possible that in both examples, at least some of the satellite clusters were traces of dispersed settlement units, which extended the radii of the impact zones. In fact the halo radius of the Late Bronze-Early Iron Age hamlet would also reach over 600 meters, if we didn’t assign a settlement status to some of the clusters north of the central site. The halo radii of the Roman to Late Roman hamlet on site 5a-5b and of the Late Byzantine-Early Ottoman hamlet on site 4 roughly equal the extent of the halos surrounding the smaller prehistoric sites, but they are still within the 200 to 400 meters range. The relatively small halo, spreading for about 220 meters to the north of the Late Byzantine-Early Ottoman site could be related to the low overall density of this material. In the case of the Roman to Late Roman settlements, it is symptomatic that the hamlet on site 5a-5b has a maximum halo radius only slightly larger than that of the Late Roman farm on site 8. It should be stressed that unlike the rest of the periods represented by settlement remains in the surface record, the impact zones of the Roman to Late Roman settlements in both survey areas were limited to the site halo and didn’t include the farther off-site. If this thinner off-site segment is estimated in the impact zone, the maximum halo radius of the hamlet on site 5a-5b would increase to over 800 meters spreading over the entire southern foothills of Prisoj. The extent of the halo of the Late Roman farm on site 8 would remain unchanged, but still measuring considerable 370 meters from the site’s northern edge. In this particular case however, it remains unclear if the slightly increased density at the foot of site 9 represents intense cultivation or a focus of separate activities. The same problem surrounds the low density scatters in the rest of the survey sectors in the first survey area, especially in sectors IX and XI. Regardless of whether one calculates the farther off-site when measuring the site halo, the Roman-Late Roman sites from the first survey area feature considerably larger impact zones than the farms in the second survey area. The latter were spaced at distances not greater than 300 meters, with satellite clusters occurring at about 80-100 meters from the site edge. In this respect, they appear as a more condensed version of the pattern revealed in the first survey area. This disparity between the maximum radii of the halos around Roman settlements in the two survey areas further underlines the differences in settlement and landscape organization. It is possible that the two communities practiced different agricultural regimes and invested in different types of cultures. It was argued that the network of farms in Skopian Montenegro probably represent a different settlement category from the hamlet on site 5a-5b in Sopot. This would in turn imply differences in the patterns of ownership and agricultural exploitation. 555 T.J. Wilkinson, tab. 1, 1989. 287 It is noteworthy that both the halos around the Roman sites in the second survey and the scatters that constitute the impact zones of the sites from various periods in the first survey are rarely spreading symmetrically around the settlement sites. In fact for the majority of periods represented by settlement remains, the settlement is located at the very edge of the impact zone. Such was the case for the possible Late Neolithic site, the Late Bronze-Early Iron Age, the Roman to Late Roman hamlet on site 5a-5b, the Late Byzantine-Early Ottoman hamlet on site 4 and for a number of Roman farmsteads in the second survey. Probably the most striking is the example of the Late Bronze-Early Iron Age assemblage, which was found exclusively limited to the east of the Sopot-Vetersko dirt road, despite the lack of topographic barriers and the fact that the settlement was positioned by this same road. We saw that the dispersal areas of the various chronological categories were usually limited to single micro-topographic units. This could very well reflect a genuine preference for certain sections of the survey areas regarding the agricultural exploitation and other non-residential activities.Unfortunately this cannot be correlated with the local pedology or the presence of other natural resources. The detailed analysis of the survey results demonstrated that the remains of nonresidential activities or the so called special-purpose sites are nearly impossible to identify solely on the basis of the surface record. In certain cases this was suggested by the location (site 5b or 14 in the first survey, site 3 in the second survey) or the absence of a site-halo (sites 3 and 20 in the second survey). In the case of the Roman and the Late Roman Period, we also pointed to a category of sites that almost exclusively consisted of brick, tile and small amounts of coarse ware (sites 2, 13a-13b, 14 in the first survey, 14 and 20 in the second survey). But as explained, this peculiar composition of the ceramic assemblage could very well reflect the local postdepositionalhistory rather than the original discard behavior. It is quite possible that at least some of these sites were locations of industrial, religious or other forms of non-residential activities, but we simply lack positive evidence. The only site categories that can be related to non-residential activities with certainty are the fortifications, the Ottoman tower and the Late Iron Age mound necropolis. A common feature for all of these sites was the very low artifact density, the architectural remains being the only traces of anthropogenic activity. Not surprisingly these site categories invariably date to periods of population growth, such as the Late Iron Age, the Roman to Late Roman or the Ottoman Period. It is reasonable to allow that similar categories existed during other periods of settlement, but these were either humble buildings or slightly adapted natural features. One cannot ignore the fact that they always appear on the very periphery or outside the microregional units. The locations of most of these monuments aren’t ideally suited for the exploitation of the surrounding land. In certain cases there is a very close topographical connection with the surrounding basins (forts 9 and 10, the mound necropolis in the first survey, the fort in the second survey), while in others these monuments belong to different microregional units and lack immediate access to the basin (the fort over the monastery of St. George or the isolated tower on the Vardar). It is possible that this distinction reflects the fact that the building of the latter two monuments wasn’t initiated by the local community. Considering the size and the elaborate layout of the fort on site 10, it is likewise possible that we’re dealing with a state-sponsored building. But the majority of these monuments were certainly built on the initiative and for the purposes of the local communities. Even these small agro-pastoralist groups could in certain periods produce architecture or earthworks of a monumental scale, showing a deep understanding of their physical surrounding and its place in the local geography. 288 VI.2 The relation between local, micro-regional and broader regional dynamics; the importance of the geographic setting One of the general aims of this study was to compare the long-term developments in two micro-regions, featuring different environments and situated in contrasting geo-political contexts. Often dealing with difficult ground conditions and fairly rich data sets, this general goal rarely had the chance to come into the focus of discussion. It was necessary to process and correctly interpret the recorded field data and analyze the material collected, before we could even begin thinking about the relations between known, regional and the micro-regional developments. During this long and delicate process of data recording and analysis, there emerged a number of unpredicted, but not less important issues. We had to devote a great deal of time and efforts in explicating the adopted method of fieldwork and the reasoning behind the interpretations proposed, not only because they form the basis for further, more general analysis, but also because they present important research subjects in their own right. Actually in the end, it can turn out that some of the most important contributions of these two micro-regional surveys were precisely on the subjects that we failed to mention among the basic goals of the research, most prominently, the definition of distinct ceramic assemblages, but also a certain number of methodological and interpretative issues. Nonetheless a considerable portion of this study was devoted to the comparison of the long-term developments in the two survey areas and to their relations with the developments in the wider region of the Vardar Valley. In addition to paving the road for future research, it was possible to make a few important observations that deserve a brief summary. Over the past decade and a half it has been argued that one of the major set-backs of the very intensive, regional surveys in the Mediterranean is their narrowness of perspective, especially when it comes to interpreting the local settlement dynamics556. More precisely, it has been suggested that by focusing on ever smaller geographic regions, survey archaeologists have given up the possibility to relate the local to the broader inter-regional developments. The accumulation of datasets of unprecedented detail and richness meant that the wider perspectives had to be sacrificed. Focused on ever smaller regions and often on the rural sectors of the landscape, contemporary regional survey projects can hardly hope to address issues such as interregional dynamics, settlement hierarchy or the impact of imperialism. To a certain degree this argument seems reasonable, but the call for a return to the less intensive, site-based approach is impetuous. Over the course of this study time and again it was stressed that at best, only a tiny fraction of the sites present in the surface record would have been detected using a less intensive survey strategy. In the particular case of the study areas and the broader region of the Vardar Valley, a more traditional extensive survey would have only confirmed and perhaps, expanded on the pattern already known from earlier research. As discussed in chapter I, the major archaeological sites in the region (mostly fortified hill-tops dating to the Late Roman and Medieval periods and the larger settlements and necropoleis from certain prehistoric periods) have been discovered and documented during the reconnaissance campaigns carried out over the past several decades. What is not known, the type of settlement during a number of prehistoric and historic periods, the rural settlement of the Roman and Medieval Periods, the size and the inner organization of settlements, these and similar goals can only be achieved by the means of 556 E. Fentress, What are we counting for? 44-52, eds. R. Francovich, H. Patterson, 2000; R.E. Blanton, 327-329, Antiquity 75, 2001; and the criticism in M. Given, Mapping and manuring: Can we compare shards density figures? 13-21, eds. S.E. Alcock, J.F. Cherry, 2004; W.R. Caraher, D. Nakassis, D.K. Pettegrew, 7-10, 2006. 289 intensive and systematic survey inevitably limited to smaller geographic units. As a possible way out of this methodological dead-end, some scholars have justly stressed the importance of comparative regional studies, especially where intensive survey data are available 557. This path however is not without its own problems, as sometimes even the comparison between the results of surveys in two geographically close regions requires that a number of conditions are met (comparability of recording and collecting methods, density estimates, site definition etc). In general the debate seems to address the issue of compatibility between the means and the aims of the research. Obviously if one is interested in studying inter-regional relations from a core-periphery perspective, the distribution, size and positioning of small, isolated farmsteads or hamlets can be of little use, even when regions of 60-70 sq km are in question, let alone surveys on a “microscopic” scale, such as the ones presented in this study. However, if the basic goals of the survey are directed towards unraveling the type and the size of rural settlements and their locational preferences, it is difficult to see a fault in the approach adopted in these studies. Admittedly one can argue (and this was explicitly acknowledged during the analysis of the settlement histories in both survey areas) that it is difficult, if not impossible, to understand the local settlement dynamics having little or no information about the developments in the neighbouring micro-regions. Comparable data is certainly lacking, but the surveyed areas weren’t blindly located in regions that were an archaeological and historical terra incognita. When choosing the survey areas, we had a good idea not only about the historical geography of the broader regional context, but also about the archaeology and the main historical developments in the region. As was hopefully demonstrated in the preceding chapters, much can be made of the existent historical and archaeological data, despite its raw character and the consequent problems of comparability558. In the first survey area with its millennia-long history of habitation, it was possible to put most of the discovered settlement phases within the wider network of the known, contemporary settlements in the region of the Middle Vardar and beyond. Thus the Middle Neolithic settlement fitted surprisingly well into the network of known Middle and Late Neolithic settlements occupying the extensive geo-pedologic zone of Neogene sediments that covers most of the MidVardar Valley and the basins to the east. Moreover it filled-in an apparent gap separating the Middle Neolithic settlements on the Ovče Pole Plateau and those in the Skopje Basin, maintaining the 10-15 km interval between neighbouring settlements. It is also very probable that the Late Iron Age centre with its mound necropolis was a part of a similar wide network of contemporary settlements, occupying nearly identical geographic locations: the small, marginal lateral valleys that drain the Tertiary basins east of the Vardar. We even predicted the existence of a similarly sized, contemporary settlement in a valley, 7-8 km northeast of the survey area, where earlier surveys have reported funerary mounds that closely resemble those discovered near Sopot. The dispersed Late Roman settlement, with its combination of small hamlets or villas and fortifications finds numerous parallels, not only in the neighbouring micro-regions, but also in the more distant regions, featuring similar topography and resources 559. On the other hand, for certain periods such as the Bronze Age, the Hellenistic or the Late Byzantine-Early Ottoman, finding parallels even in more distant areas proved far more difficult. For the first two periods, it 557 S. E. Alcock, J.F. Cherry, Introduction, 1-9, eds. S.E. Alcock, J.F. Cherry, 2004; P.A.J. Attema, G-J. L.M. Burgers, P.M. van Leusen, 2010. 558 Cf. E. Farinetti, Boeotian landscapes: A GIS-based study for the reconstruction and interpretation of the archaeological datasets of ancient Boeotia, Oxford 2011. 559 A.G. Poulter, 51-97, ed. A.G. Poulter, 2007. 290 can be argued that the general scarcity of data is related to the character of the settlements: small, dispersed farmsteads, measuring not more than 0.1-0.2 ha can hardly be detected by the traditional method of extensive, site-oriented surveys. As for the Late Byzantine-Early Ottoman Period, the reasons may very well stem from the general lack of interest among scholars working in this region, with the exception of art historians and the historians of the early centuries of Ottoman rule. Finally, for certain periods such as the Eneolithic, the end of the Bronze Age or the Early Middle Ages, the very absence of data is paralleled not only in the region of the MidVardar Valley, but in the central Balkans in general. This fact can very well reflect overall population decline or deteriorating living conditions, though it is equally possible that the material culture from these periods has “low visibility” and that we are barely beginning to recognize it560. In the case of the first survey area, one may conclude that the local developments were largely in accord with the broader regional dynamics, known from decades of extensive surveys and excavations. But in addition to simply putting the revealed local settlement history into a wider geographical context, we also tried to understand the local settlement dynamics from “within”, on the basis of the local distribution of resources. In this context, turning back to the previously discussed criticism of the limited nature of intensive survey data, it is difficult to understand why should one give priority to broader regional or interregional data over local, micro-regional data? Following this line of reasoning, should we treat archaeological data pertaining to single, multi-period sites as of a lesser order and importance? Again it all depends on the particular research interests and goals: an intensive survey of the hinterland of a single rural settlement is certainly not the appropriate approach if one whishes to learn something about polity interactions or the emergence of social complexity. But this certainly doesn’t imply that the study of micro-regional histories is an inappropriate research goal or that it can only be legitimately approached after the broader regional and interregional developments have been clarified. When dealing with micro-regions or more precisely with the hinterlands of individual settlements, the settlement dynamics consists of two basic facets: diachronic changes in the size and rank and changes in the settlement locations and habitation strategies. The basic difference between micro-regional and regional or interregional analysis is that the former lacks the third aspect of settlement dynamics and this is the changing settlement hierarchy. The very size of the survey areas allows for the existence of not more than one community per period and in this respect, the surveyed areas were either inhabited or abandoned/absorbed into the territory of a neighbouring settlement. It is mostly this third aspect that naturally remains beyond the scope of the micro-regional analysis. Unless we have access to data from the neighbouring micro-regions or from the broader region, it is impossible to know if abandonment is related to a nucleation or to a general regional decline561. The same is to a large degree true about the first aspect, the changes in the rank and size: a contraction of the studied settlement could equally reflect its subordinate status in relation to a neighbouring settlement or an overall population decline. Concerning these aspects, the micro-regional analysis is inevitably limited to the charting of the local cycles of growth and contraction. It can hardly offer an unambiguous explanation, although nothing guarantees that a broader regional analysis will be more successful in this respect. But 560 A.K. Vionis, J. Poblome, M. Waelkens, 147-165, 2009; J.B. Rutter, 137-142, eds. D. Rupp, D. Keller, 1983. 561 N. Terrenato, Sample Size Matters! The Paradox of global trends and local surveys, 36-48; eds. S.E. Alcock, J.F. Cherry, 2004. 291 when it comes to the second aspect of settlement dynamics, the displacement of settlement, its nucleation and dispersal, there are no grounds to doubt the potential of an “internal”, microregional analysis. Settlement location can readily be related to natural resources, (such as certain types of soils, proximity to freshwater springs and surface water) and communications. To be sure in this aspect too, exterior forces can be as influential as the interests of the local community, but the interplay between settlement and its physical environment are obvious and they cannot be justly ignored562. When trying to understand the long-term settlement dynamic, we’re particularly hampered by our poor understanding of the chronology of the different pottery groups. This proved particularly problematic in the first survey where we had an obvious succession of assemblages comprising similar fabric groups. Because of the low chronological resolution, we don’t know if the local settlement experienced continuous transformations or if there were gaps between two subsequent periods. In other words, we don’t know for certain if the periods when the local settlements was of the rank of a farmstead represent isolated episodes or early stages in the medium-term cycles, culminating with the emergence of a hamlet. At least for the Hellenistic to Late Roman period, the latter seems to be the likelier scenario, although one has to allow for the possibility that in certain political and socio-economic circumstances, farmsteads (standing isolated or forming extensive networks) were the preferred settlement type. In other words, the appearance of individual or networks of farms shouldn’t necessarily be seen as an episode of demographic contraction or the early stage of the development of a nucleated settlement. We already stressed the strong continuity concerning settlement size in both survey areas. In terms of population size, the rural settlements in the first survey ranged between a single extended or a few individual families to a clan consisting of up to 30 families. Excluding the Late Iron Age, this upper limit was never exceeded. In general, this long term tendency can be determined by two separate constraining factors: the carrying capacity of the settlements’ catchments and the underlining social structure of the local communities563. Analyzing the agricultural potential of the basin of Sopot and of portions of the neighbouring micro-regions, it was concluded that most of the local settlements could comfortably secure their subsistence by exploiting the natural resources of the area and even grow further. It seems that only the Late Iron Age settlement had stretched the agricultural potential of the area to its limits. This means that the chief factor limiting the size of the local communities was inherent to the nature of these societies. Featuring not more than 150 individuals, social order and cohesion in these groups could have been maintained through face-to-face relations or direct negotiations between families and individuals. Once the community exceeds the threshold of 150 individuals, this primordial regulating mechanism cannot be sustained, simply because there is a natural limit to the number of individuals with which a human can maintain face-to-face relations. This observation is based on the study of both groups of primates and traditional human societies, the fact that a subject can interact only with a limited number of individuals being predetermined by the size of the human brain564. Communities including between 80 and 150 individuals occur regularly across very different cultures and time-periods and this is surely an index of their strong stability. The fact that the fabric of society is woven through face-to-face communication eliminates the need for a permanent social hierarchy or horizontal subdivisions, threatening to 562 Cf. J. Chapman, R. Shiel, Settlements, Soils and Societies in Dalmatia, 62-76, eds. G. Barker, J. Lloyd, 1991; J. Bintliff, P. Howard, A. Snodgrass, et al, 2007. 563 J.L. Bintliff, 526-532, ed. G. Barker, 1999. 564 R.I.M. Dunbar, 4-6, 1993 292 cause inter-societal tensions and eventually, settlement fission. There are however certain problems and perhaps the most significant is that in these societies intermarriage is not a viable solution. As a result, communities of this type are always exogamous and forced to maintain at least some kind of formal relations with the neighbouring settlements. Closely related to the practice of exogamy is the inevitable dispersal of properties transacted between intermarrying families from different settlements.565 Indeed apart from the chances of expanding their territories and wealth, avoiding this problem could be one of the chief incentives behind the settlement’s tendency to grow beyond the threshold of 30 households. However in order to achieve a higher rank and become endogamous, a community needs to have at least 500 inhabitants to secure a sufficiently large genetic pool that would enable intermarriage between members of the same community and that would allow for the individual properties of the community members to remain concentrated within the limits of the settlement’s territory566. This implies that there is a theoretical transitional phase characterized by populations higher than 150 but lower than 500; a phase when the population level is not high enough to initiate settlement fission or transformation into a so called corporate community567. At its peak Late Iron Age Sopot was probably in a similar state: it was considerably larger than all of its predecessors and successors, but it never reached the level of population that would enable it to grow into a corporate, endogamous community. Although both the size of the settlement revealed through surface artifact survey and the mortuary evidence suggest that it could have almost approached this status, estimating the agricultural potential of the survey area and its surroundings it was concluded that the Late Iron Age settlement probably never had more than 60 households. As indicated by the extent of the mound necropolis, Late Iron Age Sopot was probably qualitatively different than the settlements from other periods in the first survey area, the appearance of groups of smaller mounds probably indicating horizontal subdivisions into several clans. However we believe that this settlement failed to achieve the status of a fully autonomous polity and even if it came close to becoming a corporate community, it was only for a very brief period of time. The limited agricultural resources ensured that like its predecessors and successors, it remained a part of a wider network of settlements of a similar rank and size, occupying the small lateral valleys along the Mid-Vardar. Again comparable data from the neighbouring settlement niches is crucial in addressing this issue. In the first survey area, for its size featuring a considerable variety regarding the distribution of natural resources, it was possible to follow the displacement of settlement from the eastern sectors, covered with lighter but less fertile Tertiary deposits, to the western sectors, covered with Quaternary, stony but more fertile soils. As might be expected, the earliest MidNeolithic settlement and its successor chose the former soils, which although less fertile were more suitable for primitive hand cultivation. In later prehistoric and historic periods, settlement was nearly always located on the Quaternary sediments on the western bank or on the flysch, covering the Vardar Valley floor. The only exception was the Hellenistic settlement, which for some unknown reason chose to return to the eastern survey sectors and was not only far away from the most fertile part of the landscape, but also lacked a freshwater source in its immediate vicinity. Obviously in this case, apart from the proximity to the basic natural resources other 565 J. L. Bintliff, 532-533, ed. G. Barker, 1999. J. L. Bintliff, 532, ed. G. Barker, 1999. 567 As observed by R.I.M. Dunbar, 3-4, 1993; “unlike bird flocks which can shed individuals through trickle migration as soon as the group exceeds the optimal number of individuals, human communities have to wait until they are sufficiently large to permit fission into daughter settlements.” 566 293 factors were also at play. This is equally true for the settlements situated in the western survey sectors, which although exploited the same type of soils weren’t always located on the same location. Thus unlike the settlements from earlier periods located closer to the foot of Prisoj, the Late Ottoman-Early Modern village and its Late Byzantine-Early Ottoman predecessor occupied a location very close to the small stream, near its confluence with the Vardar. It was suggested that for this community, gardening may have become a more important component in the local economy, along with the possibility of exploiting the power of running water. But in order to understand the constant shifts of settlement location within the western survey sectors and the seemingly inexplicable withdrawal of the Hellenistic settlement, it was necessary to introduce yet another locational factor and this was the main line of natural communication in the survey area. Knowing that this was very likely an active section of the interregional road known as the Via Axia in the Roman Period, it was possible to examine the location of the settlements in relation to the main road roughly following the east-west axis of the surveyed basin. Finally, the size of the settlement can also be viewed as a separate locational factor. Except for the Late Iron Age, the rest of the settlements in the first survey area were of the rank of small hamlets or farms. As such, their inner territory was most probably limited to certain portions of the surveyed basin or at least, this is what the distribution of the off-site carpet and the satellite scatters suggests. We arrived at a similar conclusion after analyzing the carrying capacity of the valley of Sopot: during most periods of settlement, the size of the local community didn’t exceed the agricultural potential of their immediate surroundings. Hence none of these settlements occupied the very centre of the basin, but rather tended to concentrate on certain micro-topographic units, usually on the lower western bank. In contrast the Late Iron Age settlement, the only one that came close to achieving the rank of a village and possibly exploiting the full potential of the surveyed basin, occupied the geometric centre of the integral area, assuming an equal access to both banks of the valley, although the focus was evidently on the western bank. For this settlement too, the proximity to the main line of communication doesn’t seem to be of a particular importance. By its location, it belongs to the “sheltered” group of settlements, but the positioning of the mound necropolis on the top of the Jakupica Ridge, the eastern limit of the basin, clearly indicated that security wasn’t a major concern for the Late Iron Age inhabitants. It is thus evident that when determining the importance of a certain resource (and communications in particular) as a locational factor, focusing solely on the location of the central settlement can often lead us into bringing incorrect interpretations. If the goal is to study the relation of the local communities to their physical surroundings, the integral surface archaeological record dating to a certain time-period has to be considered. This type of analysis is reminiscent and partly inspired by the studies of ethnographers working within the framework of the early Anthropo-geographic School568. According to one of their central theories, given that all conditions are optimal, the settlement’s location should reflect the consideration of a number of factors, including sufficient living space, access to good arable land and pastures, access to water, access to natural lines of communication, preferable exposure to the elements etc. If one or more of these factors is disregarded and the settlement location deviates from the optimum, then its location must be influenced by other, nongeographical factors. These may include particular historical developments, insecurity and 568 J.L. Bintliff, 147-164, eds. R.A. Bentley, H.D. Maschner, 2008; A. Holt-Jensen, Geography: History and Concepts, London 1999; for an example of an early application of this approach, H. Lehmann, 212-238, , 1939, after M. Gkiasta, 72-75, 2008; this tradition has also been very influential in the old Yugoslav school of Human Geography, J. Cvijić, 1922; J. Trifunovski, 345-517, 1955. 294 demographic pressure, but also the work of external and internal political, ideological and natural forces. It is evident that this theoretical position is carefully formulated, so that it doesn’t descend into a rigid geographic determinism. But despite its breadth and potential, especially for the analysis of the location of the Late Ottoman-Early Modern rural settlements, it was never applied systematically. On the other hand, its application to rural sites from the more distant past could be somewhat more problematic, simply because we lack information about the local historical conditions and developments, the local economic and environmental conditions. To take the most obvious example, the remains of ancient settlements are often found in presently barren and inhospitable environments, which wasn’t necessarily the case in the more distant past when these settlements were active. At the same time working on a micro-regional level, one has few other choices but to relate the settlement locations with factors such as access to natural resources and communications. Giving up this or similar perspectives, it becomes impossible to make any sense of the constant displacement of the main settlement within the narrow frames of a single parish. But it is important to recognize the main disadvantages of the archaeologist when attempting to apply this or similar geographic approaches to intensive survey data. Unlike geographers or ethnographers, archaeologists can rarely identify the recorded surface phenomena with the known habitational components with certainty. In fact an archaeologist can barely guess what proportion of the original artificial features in the studied landscape has survived in the surface record569. As we learned from the experience of these two small-scale surveys and from the large regional projects carried out over the past few decades, the settlement is but a single component of the inhabited landscapes. Clearly settlements are the central elements of human habitation and their locations are certainly instructive of the living standards, the economy and social conditions and perhaps even of the ways in which the local communities perceived their physical surroundings. However during most periods of the past there were a number of other features through which humans exploited and organized their environments (various agricultural and industrial facilities, field huts and animal sheds, refugia and cultic locations) and the logic behind their location is often totally opposed to the logic behind the locations of settlements. We saw this through the example of the Late Iron Age settlement in the first survey area. Equally illuminating was the Roman-Late Roman settlement in the same survey area, when there existed two parallel schemes reflecting two contrasting relations to the environment. When analyzing the factors that influenced the settlement location, one has to approach each of the settlements separately, taking into account other habitational components and acknowledging the possibility that they are simply not preserved in the surface archaeological record. One last difficulty in adopting a purely geographic perspective when trying to understand the local settlement dynamics stems from the fact that this approach was primarily devised with the aim of analyzing the locations of contemporary settlements, possessing more or less equal technological capacities, similar economies and social organization570. This is hardly the case for an archaeological research whose subject of study is the long-term settlement dynamics from the Neolithic to the present-day. Needless to stress, over the course of the last 8 millennia there happened profound transformation of the technologies, the social and economic organization. Therefore while it isn’t necessarily erroneous to define an optimal location for settlements dating to the same or historically close epochs, this is obviously unviable for the purposes of a diachronic analysis. The problem is that while geographers try to understand the logic (or its 569 570 J. Bintliff, P. Howard, A. Snodgrass, 139-168, 1999 T.K. Earle, R.W. Preucel, 501-513, 1987; with numerous comments and a reply by the authors.. 295 absence) behind the settlement location in a known social and economic context, the goal of the archaeologist is to catch a glimpse of precisely these contexts on the basis of the sites’ location and character. In other words, although seemingly striving towards similar goals, the path undertaken by the landscape archaeologist and the geographer cannot be fully convergent. Because of the differences in the starting points and the specific study subjects, but also because we had no access to past environmental data, the discussion of the “inner” settlement dynamics remained chiefly descriptive, although we attempted to establish a rough topology for both survey areas. The case of the second survey area and the integral region of Skopian Montenegro nicely illustrate just how powerful the geographic factors can be in determining the location of settlements. For this region the accounts of the early 20th century ethnographers are particularly helpful, because they explicitly state the logic behind the positioning of the Late Ottoman-Early Modern villages571. The second survey area was carefully situated in the very centre of the fertile, gently rolling foothills of Mt. Montenegro, at an equal distance from the nearest contemporary settlements. In comparison to the first survey area, it looked much more promising and richer in natural resources. Also being slightly larger, we expected to find at least an equal number of periods represented in the surface record as in the first survey area. Indeed briefly estimating the carrying capacity of the intensively surveyed area, it was concluded that it could comfortably sustain a settlement of a similar rank to those discovered in the first survey. Understandably the carrying capacity of the theoretical catchments is much greater, allowing for the emergence of larger, town-like settlements. Furthermore in order to ensure the discovery of settlement remains, we carefully positioned the survey over the same type of topographic units occupied by the Late Ottoman-Early Modern villages. But as we saw in the preceding chapter, these expectations came to nothing. Although the survey did reveal traces of settlement from at least three periods of the past, the surface record was nothing like that in the first survey area, the prehistoric periods being particularly underrepresented. None of the settlements revealed in the second survey were of a nucleated type and there lacked a distinct, local ceramic production. Leaving aside the potential post-depositional factors, above all the possibility that earlier surface remains are buried beneath deep colluvial sediments, it was suggested that the relative scarcity of settlement traces in the surface record reflects genuine absence of settlements during most periods of the past. In fact we are still rather confident that if conditions allowed and if the survey was carried out in the immediate vicinity of one of the Late Ottoman-Early Modern villages, the results would have been similar to those obtained from the first survey. When trying to understand the place of the secondsurvey area in the wider study region and the absence of a long history of settlement, it is important to take into account the wider geographical setting. The rugged plain at the foot of Mt. Montenegro is a larger and compact regional unit, measuring nearly 30 sq kilometers. There are no clear topographic divisions; the terrain consists of a series of narrow valleys, alternating with low, gentle ridges. These vague vertical divisions are complemented by a series of concentric terraces, dividing the ridges along the horizontal axes. Thus instead of series of small, physically separate valleys, the region of Skopian Montenegro is a mosaic of old lake terraces, broken up into separate shelves by the mountain streams. In this geographic setting the small valleys are too narrow to accommodate the settlement with its fields. Therefore the focus of human settlement was on the low ridges, with their gently sloping sides. Each of these “shelves” could accommodate a settlement with its inner territory. Prior to the field survey, we hoped that there was a greater dynamism in the S. Tomić, 433, 1905. 571 296 settlement history of the wider study region, with settlements shifting more frequently across the terraced landscape. But the survey results showed that this didn’t happen particularly often, at least not in the intensively surveyed portion of the region. Why was the central portion of the foothills, partly covered by the second survey, so resiliently avoided during most periods of the past? It offered nearly ideal conditions for the development of a small to medium-sized, if not larger agrarian community: a plenty of living space and fertile soils, access to water and communication. Maintaining the locational perspective which we briefly elaborated upon, we can repeat that the locations of the Late Ottoman-Early Modern villages are advantageous because they integrate one additional factor in their positioning and this is the access to the resources of the mountainside. Being located at the very foot of the mountain has the advantage of offering equal access both to the fields in the foothills and to the mountain resources. This positioning eliminates the major logistical problem of all agro-pastoral communities: the transport of the flocks from the winter to the summer pastures, especially during the early spring months, by which time most of the local cultures normally begin to sprout572. Locating the settlement in the midst of the plough-zone, a considerable portion of the agricultural land falling within the settlement’s catchment has to be given up to houses and outbuildings, shelters for animals, the communal cemeteries etc. For the large communities that inhabited the wider study region over the past 5-6 centuries, arable land was simply too precious to afford such an arrangement. In addition to these economical and logistical considerations, the early 20th century ethnographic record reveals another important factor that influenced the location of settlement, especially during later historic periods573. For the local inhabitants, the presence of the Medieval churches and monasteries in the immediate vicinity of the villages was of equal, if not of a greater importance. The saints to which these churches were dedicated were seen as patrons and protectors of the entire communities whose very large size was itself providing a sense of security. Recall that by the beginning of the 18th century, life in the small satellite hamlets in the mountainside has probably become perilous. Thus in this case, we see the purely economic factors being reinforced by symbolic or ideological means. One can imagine that the latter factors were particularly important, especially during the turbulent Late Ottoman Period, when they obviously played an important role in the preservation of the local Christian identity. The surface archaeological record in the survey area and in the wider study region indicates that similar considerations influenced the types of settlement locations in the more distant past. The distribution of the thin off-site scatters of the Late Roman and the Late Byzantine-Early Ottoman finds points to the direction of the Late Ottoman-Early Modern villages as the location of these periods’ settlements. This is further supported by the location of the monastic churches, mostly founded in the first half of the 14th century and also by the accidental discoveries of agglomerated Late Antique cist burials, both situated within the borders of the Late Ottoman-Early Modern villages and in their immediate vicinity. Obviously for these periods, we lack the nuanced ethnographic narratives and it would be too simplistic to project the economic and ideological perspectives of the Late Ottoman-Early Modern communities to their distant predecessors. Nevertheless the little evidence that we have, indicates that during these 572 For a discussion of this issue from the point of view of the possibility of transhumance in antiquity, S.E. Alcock et al, 148-149, ed. I. Morris, 1994; P. Halstead, Traditional and ancient rural economy in Mediterranean Europe: plus ça change? 77-87, Journal of Hellenic Studies 107, 1987. 573 S. Tomić, 453-54, 1905; J. Trifunovski, 1971. 297 two periods the local settlement pattern closely resembled the one that has survived until the present-day. But the intensive survey of the central portion of the foothills, along with accidental discoveries in other parts of the wider study region suggest that this seemingly optimal pattern of settlement was fully or partly abandoned during at least three periods in the past. Does this indicate a change in the local economy, with the pastoral component loosing its importance? Understandably with the means presently at our disposal we can never be sure, but for a small and predominantly agrarian community, the locations of the Late Ottoman-Early Modern villages are certainly not the most convenient. Located at the very foot of the mountain, half of the villages’ catchments will fall to the mountainside, mostly consisting of steep ridges and narrow valley floors, offering little cultivable land. From a purely agrarian perspective, the low ridges in the central parts of the foothills presented a more advantageous settlement location, offering immediate access to cultivable land on all sides. We may recall the locations of some of the prehistoric farms in the first survey area, boldly located in the midst of the modern ploughzone and near the central axis of the region. At the same time, one shouldn’t forget that such locations are optimal only for settlements of a minor rank. A medium or large-sized village would have not only consumed a considerable portion of its arable land, but would also disrupt the existing pattern of villages with territories spreading into narrow elongated strips that cut across both the foothills and the mountainside. Thus one can argue that the present-day pattern with villages located along the mountain foot and the one characterized by farms and hamlets dispersed across the plain were incompatible, unless the smaller establishments in the plain are seen as satellites of the main settlements at the mountain foot. In comparison to the first survey area, the second survey, as well as the wider study region of Skopian Montenegro is much more uniform concerning the geo-pedological substrate. More than 95% of the foothills are covered with moderately eroded, Tertiary deposits. In such conditions it is obviously impossible to include this factor in the analysis of the local settlement dynamics, although it has to be emphasized that there are much finer, local varieties of soil types. Unfortunately we only have a vague, general idea of their distribution. Thus when examining the distribution of the settlements revealed in the second survey area, we had to operate with two basic parameters: micro-topography and relation to the local road-network. The results were nevertheless satisfactory, because the analysis helped us explain the clear preference for settling on the upper portions of the eastern ridge. This was observed both for the later prehistoric settlement and for the agglomeration of Roman farmsteads. These locations close to the top of the ridge offered access to arable land and to the main road artery in the region. There were no freshwater springs in the immediate vicinity, but at the eastern foot of the ridge, on a narrow valley floor, which on the other hand didn’t offer sufficient living space. The very top of the ridge was not occupied, probably in order to avoid the northerly winds that blow constantly from the direction of the mountainside. The only exception is the small site 3, a circumstance which along with the character of the ceramic assemblage was instructive of its special-purpose character. While the Roman and the later prehistoric settlements occupy roughly identical locations, on two occasions during prehistory and the Middle Age, the opposite western ridge was briefly occupied. Although looking as an identical replica of the eastern ridge, this topographic unit presented a less favourable settlement location. It was not only drier and situated at a greater distance from the freshwater sources that issue from the foot of the eastern ridge, but it also lacked direct access to the main road-network. More precisely, it was connected to the local 298 road-network only via the site of the later village Kučevište. Access to both the outside world and to the mountain resources was only possible through the site of the larger settlement at the mountain foot. This circumstance reflects the difference in status between the settlements that occupied the western and eastern ridges. The networks of Roman and possibly, late prehistoric farms that occupied the eastern ridge came closer to becoming a separate settlement. The distribution of Roman farms showed traces of spatial planning and perhaps even a settlement focus on the top of the ridge. Recall that in total, they occupied an area of nearly 3 hectares, sufficiently large to accommodate a small-sized village, though most probably the community consisted of about 10 families. In contrast the Medieval settlement on the western ridge was at least ten times smaller, leaving extremely faint traces in the surface record. This interpretation implied that the highly integrated and stable settlement pattern that characterized the wider study region during the past several centuries and possibly during Late Antiquity could predate the 14th century. Only during the Roman and the later prehistoric period was this scheme abandoned in favor of a more dispersed, (purely?) agrarian based pattern. If we look at the broader context, we’ll see that in general the settlement history of the survey area and the wider study region is closely related to the major developments in the Skopje Basin. Although difficult to interpret and lacking a geographically close parallel, the scant traces of Late Bronze or Iron Age activity in the second survey area are hardly surprising. As in the rest of the lateral valleys of the Skopje Basin, the first stable settlements date no earlier than the first millennium BC. There is very little or no evidence of Bronze Age or Neolithic activity in these parts of the plain. The group of Roman farms was also a part of a broader, regional network that extended across the entire region of modern Skopje. It was closely related to the foundation of Scupi and the colonization of the basin and the lateral valleys. During the period between the late 1st and the late 4th century AD, the entire Skopje Basin formed a part of the colony’s agricultural territory. This extreme settlement pattern, with one very large metropolis and an unknown number of various agricultural estates and but a few hypothetical villages will never be repeated in the Skopje Basin and it is most probably unique for the entire country. The abandonment of this pattern during the period of Late Antiquity, accompanied by the retreat in the more mountainous regions and a possible nucleation was inevitably reflected in the survey area and in the wider study region. It is quite possible that for the first time in this period, the survey area became a part of the agricultural territory of a larger nucleated settlement. After a period of a few centuries of decline and possible complete abandonment of the wider study region, settlement returns in the area by the Mid-Byzantine Period. Again this corresponds with the establishment of Medieval Skopje as the main administrative and economic centre in the wider region. Unlike the Early and Middle Roman Periods, there were a number of other forts and nucleated rural settlements, especially after the 14th century. Nevertheless the ancient agrarian relations were basically restored, because a large portion of Skopian Montenegro, along with its inhabitants belonged to a major landowner based in Skopje, the monastery of St. George Nikephoros. This relationship will survive the Ottoman conquest and it will be maintained throughout the entire Ottoman Period. But despite of the observed continuity in agrarian relation, on the local level the settlement pattern had changed and the Roman villas and farms were replaced by nucleated communities of dependent peasants. In such constellations, the survey area became but a part of the agricultural territories of these newly developed nucleated settlements. As for the first survey area it is possible to arrive at tentative, but well argued interpretations of the local settlement dynamics, despite the scanty archaeological and literary 299 evidence. But in order to approach the problem, it was necessary to take into account both the “inner” and the external factors, the relationship between the settlements and their physical environment and the wider, regional context. Obviously small-scale, micro-regional studies can hardly be informative about trends and developments outside the narrow limits of the survey areas, but they can be more than a mere methodological exercise. As exemplified by both casestudies and especially the second survey, micro-regions are certainly not enclosed microuniverses and it is impossible to understand the local developments, without at least a minimum insight into the broader context. At the same time however, they do exhibit peculiar inner dynamics, which can only be understood through careful study of the relationship between settlements and other habitational components and their physical surroundings. We still need to address the issue of the apparent differences between the two survey areas in terms of their respective settlement histories. During the early stages of the research, it was deliberately decided to survey and compare two micro-regional entities that featured contrasting environmental conditions and that belonged to regions with different geo-political and historical backgrounds. This would obviously make direct comparison more difficult, but at the time it seemed more important to record the amount and distribution of surface material in various environments and open an insight into issues such as the size and types of rural sites through various periods of the past and across different natural settings. The first survey area roughly corresponds to one of the dozens small valleys that drain the banks of the Middle Vardar. Geographically this is a well-defined territorial unit, separated from the neighbouring valleys by low, but extensive and barren hills. At present this is a marginal, dry land with little fertile soils and no running water on the surface. Basically its only resource is its strategic location in the broader geographic context, as it occupies the point where one can most easily leave the Taor Canyon and continue southwards, towards Thessalonica and the Aegean. This micro-region is marginal not only in terms of agricultural and other natural resources, but also in a geologic, cultural and political aspect. In a number of historical periods, the area found itself at the very edges of the political entities that dominated the lands along the Middle and the Lower Vader. In contrast the second survey area and the wider region of Skopian Montenegro, throughout all of its known history lied in the heartland of the polities that dominated Skopje and the Skopje Basin. Ever since the Iron Age, this region was within a day-walk from the main regional centres, pre-Roman and Roman Scupi and Medieval Skopje. In terms of agricultural and other natural resources, it also offers far more favourable conditions than the barren, rocky landscape that surrounds modern Sopot. Apart from fertile and thicker soils, this region is rich in water and freshwater springs (in the past, Roman and Ottoman Skopje and their fields and gardens were fed from these springs), timber and pastures. To illustrate this contrast in the productivity and wealth of resources, it suffices to compare the modern population figures for the wider regions of the two survey areas: at present, roughly 8000 people inhabit the region of Skopian Montenegro, less than 200, the villages in the region of the southern end of the Taor Gorge. All environmental, cultural and historical factors are more favourably inclined towards the second survey area and the region of Skopian Montenegro and yet the results of the intensive surveys proved counterintuitive. While the first survey area was inhabited in most periods during the last 8 millennia and it is still occupied by a small (albeit disappearing) village, the second survey area was thinly inhabited only in three unrelated periods in the past and never grew into a stable settlement niche that sustained a nucleated settlement. The specific mechanisms that brought about this seemingly paradoxical situation were analyzed separately in the preceding 300 paragraphs and at this point we can briefly summarize them. To a certain degree, the second survey area was unfortunately chosen, because it is located in the neighbourhood of a much more favourably positioned settlement niche and it remained but a part of the hinterland of the large nucleated settlements that occupied its northern neighbours. The basin of Sopot on the other hand, despite its barren and inhospitable appearance is a clearly delimited micro-geographic entity, a true settlement chamber that was inhabited in all, but the most precarious periods of the past574. The very fact that it was surrounded by an extensive area of dry and barren rock preconditioned restricted settlement mobility. The opposite is the case of Skopian Montenegro, where seemingly every corner of the 30 sq km large foothills offers a suitable settlement location. As was shown however, not all of its parts were equally suitable for the large agropastoral communities that inhabited the region over the past several centuries. They established a highly integrated pattern of settlement, where even small, localized displacements could affect the entire network. We believe that settlement in this region followed the same or similar patterns in most other periods of the past. Finally, unlike the hinterland of Sopot and to a certain degree the locations of the Late Ottoman-Early Modern villages of Skopian Montenegro, the second survey area lacked a visible physical integrity. Micro-topographically it is barely distinct from the surrounding basins and ridges, but for the people inhabiting this landscape it was but a section of a wider terrain, without clear borders or micro-topographic specifics that would’ve formed the basis for the development of a local identity. Again we see a synergy between the forces of nature and the cultured perceptions of humans, strongly influencing the choice of human habitat. Perhaps the most striking implication of these findings is the incredible stability of the settlement niches in the regions along the Vardar Valley. Once occupied by a nucleated settlement even of a minor size, the prevailing tendency was that they remained inhabited, often until the present-day. In fact, it can be argued that in all micro-regions where settlement has survived to this day (even if presently lying in ruins), one can confidently expect to find settlement remains from a number of other historic and prehistoric periods. It took dramatic and extreme historic episodes to interrupt or relocate the established pattern of settlement, such as the foundation of Roman Scupi, the near collapse of society at the end of Antiquity or the radical modernization of the country after World War II. But even after such dramatic events, the chances were that once the old conditions returned, settlement will also return to the old niches rather than occupy alternative locations and completely replace the old settlement pattern. Having studied but a few potential niches, it is certainly too early to generalize on the basis of such a thin corpus of evidence. Indeed studying the historical toponomy in the region of the Middle Vardar Valley, we found a number of examples of villages relocated over distances of several kilometers and occupying completely different drainages. In all likelihood however, these are the exceptional cases. The fact that toponyms associated with the old village often survive in the landscape long after it’s been abandoned, most plainly illustrates the profound connection between the community and its physical surroundings. 574 Cf. J.L. Bintliff, 193-224, eds. A. Hurst, A Schachter, 1996. 301 Referrences: Alcock, S.E. Archaeology and imperialism: Roman Expansion and the Greek City, 87135, Journal of Mediterranean Archaeology 2, 1989 Alcock, S.E. Urban survey and the city of Phlius, 421-463, Hesperia 60-4, 1991 Alcock, S.E. Graecia Capta: The Landscapes of Roman Greece, Cambridge 1993 Alcock, S. E. Cherry, J.F. Davis, J.L. Intensive survey, Agricultural practice, and the Classical Landscape of Greece, 135-168, ed. I. Morris, Classical Greece: Ancient Histories and modern Archaeologies, Cambridge 1994 Alcock, S.E. Osborne, R. eds. Placing the Gods. Sanctuaries and sacred space in ancient Greece, Oxford 1994 Alcock,S. E. Cherry, J.F. Introduction, 1-9, eds. S.E. Alcock, J.F. Cherry, Side-by-side survey: comparative regional studies in the Mediterranean World, Oxford 2004 Alcock, S.E. Cherry, J.F. eds. Side-by-side survey: comparative regional studies in the Mediterranean World, Oxford 2004 Alcock S.E., Rempel, J.E. The More Unusual dots on the map: Special purpose sites and the texture of landscape, 27-46, eds. P. Guldager Bilde, V.F. Stolba, Surveying the Greek Chora: The Black Sea Region in a Comparative Perspective, Aarhus 2006 Aleksova, B. Mango, C. Bargala: a preliminary report, 269-281, Dumbarton Oaks Papers 25, 1971 Aleksova, B. Bargala i sredniot tek na Bregalnica, 61-71, Glasnik na Institutot za Nacionalna Istorija 3, 1983 Aleksova, B. Loca sanctorum macedoniae, Skopje 1995 Antoljak, S. Srednovekovna Makedonija, vol. I, Skopje 1985 Attema, P. Two challenges for Landscape Archaeology, 18-26, eds. P. Attema et al, New developments in Italian Landscape Archaeology, Oxford 2002 Attema,P.A.J. Burgers, G-J. L.M. van Leusen, P.M. Regional Pathways to Complexity: settlement and land-use dynamics in early Italy, from the Bronze Age to the Republican Period, Amsterdam 2010 Arheološka Karta na Republika Makedonija, volume 2, Skopje 1996 Armstrong,P. Trade in the Eastern Mediterranean in the 8th century, 159-178, ed. M. Mango, Byzantine Trade, 4th-12th century, Ashgate Publishing 2009 Babić, B. Mikulčić, I. Skopje-a prehistoric and Medieval fortress, 162-163, Arheološki Pregled 9, 1967 Babić, B. Suli-an (Late Middle Ages-Ottoman Period), 163-164, Arheološki Pregled 9, 1967 Babić, B. Trouvaille scellée de poterie de table Byzantine à Skopsko Kale 45-54, Archaeologica Iugoslavica 13, 1971 Babić, B. Wann sind die ältesten Teile der mittelalterlichen Burg Kale in Skopje gebaut, 359-364; Actes du VIIIe Congrès international des sciences pré et protohistorique, Beograd 1971 Babić, B. Materijalnata kultura na makedonskite Sloveni, vo svetlina na arheoloshkite iskopuvanja vo Prilep, Prilep 1986 Banaji,J. Agrarian Change in Late Antiquity: Gold, Labour and Aristocratic Dominance, Oxford 2001 302 Barker, G. Approaches to Mediterranean Landscape History, 1-16; ed. G. Barker, A Mediterranean Valley: Landscape archaeology and Annales History in the Biferno Valley, Leicester 1995 Basotova, M. A new veteran of the legion VII Claudia from the Colonia Flavia Scupi, 405-409, Arheološki Vestnik 58, 2007 Bavant, B. Caričin Grad and the changes in the nature of urbanism in the central Balkans in the sixth century, 337-374, ed. A.G. Poulter, The Transition to Late Antiquity on the Danube and Beyond, Oxford 2007 Beldedovski, Z. Bregalničkiot Bazen vo Rimskiot i Ranosrednovekoven Period, 9-41, Zbornik na Štipskiot Naroden Muzej VI, 1990 Benac. A ed. Praistorija Jugoslovenskih Zemalja, vol.1-5, Sarajevo 1977-1989 Bender, B. ed. Landscapes, politics and perspectives, Berg 1993 Benesh, J. Zvelebil, M. A historical interactive landscape in the heart of Europe: the case of Bohemia, 73-93, eds. P. Ucko, R. Layton, The archaeology and anthropology of landscape, New York1994 Bes, P. Poblome, J. Bintliff J, Puzzling over pottery. Thespiae, Tanagra and methodological approaches towards surface pottery, 339-345, eds. D. Malfitana, J. Poblome, J. Lund, Old pottery in a new century: Innovating perspectives on Roman Pottery Studies, Ibam, National Museum of Denmark, Leuven, Icrates 2006 Biagi, P. Spataro, M. New observations on the Radiocarbon Chronology of the Starčevo-Criş and Kōrōş Cultures, 35-50, eds. L. Nikolova, J. Fritz, J. Higgins, Prehistoric Archaeology & Anthropological Theory and Education. RPRP 6-7, 2005 Bilbija, M. Novi arheološki naodi od Skopje i okolinata, 134 -135, Macedoniae Acta Archaeologica 4, 1978 Binford, L.R. Archaeology as Anthropology, 217-225, American Antiquity 28-2, 1962 Binford, L.R. A Consideration of Archaeological Research Design, 425-441,American Antiquity 29-4, 1964 Bintliff, J.L. Snodgrass A.M., The Cambridge/Bradford Boeotian Expedition: The first four years, 123-161, Journal of Field Archaeology 12, 1985 Bintliff, J.L. Snodgrass, A.M. Mediterranean survey and the city, 57-71, Antiquity 62, 1988 Bintliff, J. Snodgrass, A.M. Off-site pottery distributions: A regional and interregional perspective, 506-513, Current Anthropology 29-3, 1988 Bintliff, J.L. ed. The Annales School and Archaeology, Leicester 1991 Bintliff, J. Davis, B. et al, Trace metal accumulations in soils on and around ancient settlements in Greece, 9-24, ed. P. Spoerry, Geoprospection in the Archaeological Landscape, Oxford 1992 Bintliff, J. Territorial behavior and the natural history of the Greek polis, 207-249, eds. E. Olshausen, H. Sonnabend, Stuttgarter Kolloquium zur Historischen Geographie des Altertums, Amsterdam 1994 Bintliff, J.L. Archaeological survey of the Valley of the Muses and its significance for Boeotian history, 193-224, eds. A. Hurst, A Schachter, La montagne des Muses, Geneva 1996 Bintliff, J. L. Regional Survey, Demography, and the Rise of Complex Societies in the Ancient Aegean: Core-Periphery, Neo-Malthusian, and other Interpretive Models, 1-38, Journal of Field Archaeology 24, 1997 Bintliff, J.L. Further considerations on the population of ancient Boeotia, 231-252, ed. J. Bintliff, Recent Developments in the History and Archaeology of Central Greece, Oxford 1997 303 Bintliff, J.L. Howard, P. Studying needles in haystacks-Surface Survey and Rural Landscapes of Central Greece in Roman Times, 51-91, Pharos 1999 Bintliff, J. Howard, P. Snodgrass, A. The hidden landscape of prehistoric Greece, 139168, Journal of Mediterranean Archaeology 12-2, 1999 Bintliff, J.L. The Ottoman era in the context of long-term settlement history. A casestudy: the archaeological survey of the Valley of the Muses, Boeotia, Central Greece, 203-229, Arab Historical Review for Ottoman Studies 19-20, 1999 Bintliff, J. Settlement and territory, 505-545, ed. G. Barker, Routledge Companion Encyclopedia of Archaeology, London 1999 Bintliff, J.L. et al, Deconstructing “the sense of place”? Settlement systems, field survey and the historic record: a case study from Central Greece, 123-149, Proceedings of the Prehistoric Society 66, 2000 Bintliff, J. L. The concepts of 'site' and 'offsite' archaeology in surface artifacts survey, 200-215, eds. M. Pasquinucci, F. Trément Non-Destructive Techniques Applied to Landscape Archaeology, Oxford 2000 Bintliff,J. Sbonias,K. Demographic and ceramic analysis in regional survey, 244-258, eds. R. Francovich, H. Patterson, Extracting Meaning from Ploughsoil Assemblages. Oxford 2000 Bintliff, J. L. Going to market in Antiquity, 209-250, eds. E. Olshausen, H. Sonnabend, Zu Wasser und zu Land, Stuttgarter Kolloquium 7, Stuttgart 2002 Bintliff, J. L. et al. Classical farms, hidden prehistoric landscapes and Greek rural survey: a response and an update, 259-265, Journal of Mediterranean Archaeology 15, 2002 Bintliff, J. L. et al. The Tanagra project: investigations at an ancient city and its countryside (2000-2002), 541-606, Bulletin de Correspondence Hellènique128-129, 2004-5 Bintliff, J.L. Farinetti, E. Landscape and Early Farming: Settlement Dynamics in Central Greece, 665-674, Geoarchaeology 21, 2006 Bintliff, J. The Leiden University Ancient Cities of Boeotia Project: 2005 season at Tanagra, 29-38, Pharos 13, 2006 Bintliff, J.L. Slapšak, B. The Leiden-Ljubljana Ancient Cities of Boeotia Project: season 2006, 15-27, Pharos 14, 2007 Bintliff, J. L. The contribution of regional survey to the Late Antiquity debate: Greece in its Mediterranean context, 649-678, ed. A.G. Poulter, The Transition to Late Antiquity on the Danube and Beyond, Oxford 2007 Bintliff, J. Howard, P. Snodgrass, A. et al, Testing the hinterland: the work of the Boeotia survey (1989-1991) in the southern approaches to the city of Thespiai, Cambridge 2007 Bintliff, J.L. History and Continental Approaches, 147-164, eds. R. A. Bentley, H. D. Maschner, Handbook of Archaeological Theories, Altamira Press 2008 Bintliff, J. Contemporary issues in surveying complex urban sites in the Mediterranean region: the example of the city of Thespiai (Boeotia, Central Greece), 44-52, eds. F. Vermeulen et al, Urban Landscape Survey in Italy and the Mediterranean, Oxford 2012 Bintliff, J. The Complete Archaeology of Greece: From Hunter-gatherers to the 20th century AD, Wiley-Blackwell 2012 Blanton, R.E. Mediterranean myopia, 627-629, Antiquity 75, 2001 Bogaard, A. The Nature of Early Farming in Central and Southeast Europe, 49-58, Documenta Praehistorica XXXI 2004 304 Bowden, H. Gill, D. Late Roman Methana, 84-91, eds. C. Mee, H. Forbes, A rough and rocky place: The landscape and settlement history of the Methana Peninsula, Greece, Liverpool 1997 Boyd, M.J. Geophysical survey and rural settlement architecture on the Lower Danube at the transition to Late Antiquity, 597-609, ed. A.G. Poulter, The transition to Late Antiquity on the Lower Danube and Beyond, Oxford 2007 Branigan, K. Late Neolithic Colonization of the Uplands of Eastern Crete, 57-65, ed. P. Halstead, Neolithic society in Greece, Sheffield 1999 Cavanagh, W.G. Hirst, S. Litton, C. Soil Phosphate, Site Boundaries and Change Point Analysis, 67-83, Journal of Field Archaeology 15, 1988 Cavanagh, W. Crouwel, J. The Survey Area in the Prehistoric Periods, 121-150, eds. W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al. The Laconia Survey: Continuity and Change in a Greek Rural Landscape, Volume I, London 2002 Cavanagh, W. Mee, C. and James, P. et al, The Laconia Rural Sites Project, Athens 2005 Caraher, W.R. Nakassis, D. Pettegrew, K.D. Siteless Survey and Intensive Data Collection in an Artifact-rich Environment: Case Studies from the Eastern Corinthia, Greece, 743, Journal of Mediterranean Archaeology 19-1, 2006 Carlton,R. An ethnoarchaeological study of pottery production on the Dalmatian island of Iž, 101-124,eds. J. Chapman et al. Recent developments in Yugoslav Archaeology, Oxford 1988 Catling, R.W.V. The Survey Area from the Early Iron Age to the Classical Period (1050300), 151-256, W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al. The Laconia Survey: Continuity and Change in a Greek Rural Landscape, Volume I, London 2002 Chang,C. Tourtellote,P.A. Ethnoarchaeological survey of pastoral transhumance sites in the Grevena Region, Greece, 249-264, Journal of Field Archaeology 20-3, 1993 Chapman, J.C. The early Balkan village, 33-53, Varia Archaeologica Hungarica 2, 1989 Chapman, J. Shiel, R. Settlements, Soils and Societies in Dalmatia, 62-76, eds. G. Barker, J. Lloyd, Roman Landscapes: Archaeological surveys in the Mediterranean Region, London 1991 Chapman, J. “Rubbish dumps” or “Places of deposition”: Neolithic and Copper Age settlements in Central and Eastern Europe, 347-362, ed. A. Ritchie, Neolithic Orkney in its European Context, Cambridge 2000 Chase-Dunn, C.K. Hall, T.D. Rise and Demise: Comparing World – Systems, Boulder 1997 Cherry, J. F. A Preliminary definition on site distribution on Melos, 16-23, eds. C. Renfrew, M. Wagstaff, An Island Polity: the archaeology of exploitation on Melos, Cambridge 1982 Cherry, J.F. Appendix A: register of archaeological sites on Melos, eds. C. Renfrew, M. Wagstaff, An Island Polity: the archaeology of exploitation on Melos, Cambridge 1982 Cherry, J.F. Frogs around the pond: Perspectives on current archaeological survey projects in the Mediterranean region, 383-417, eds. D. R. Keller, D.W. Rupp, Archaeological Survey in the Meiterranean Area, Oxford 1983 Cherry, J.F. et al, Archaeological survey in an artifact rich landscape: a Middle Neolithic example from Nemea, 159-176, American Journal of Archaeology 92-2, 1988 Cherry, J. F. et al, Landscape archaeology as long-term history: northern Keos in the Cycladic Islands from earliest settlement until modern times, Los Angeles 1991 305 Curta,F. The Making of the Slavs: History and Archaeology of the Lower Danube Region (c. 500-700), Cambridge 2001 Cvijić, J. Osnovi za Geologiju i Geografiju Južne Makedonije i Stare Srbije, Beograd 1906 Cvijić, J. Balkansko Polustrovo i Južnoslovenske Zemlje, I, Beograd 1918 Čausidis, N. Novootkrieni docnoantički tvrdini na Skopska Crna Gora, 183-196, Macedoniae Acta Archaeologica 10, 1989 Davies, S. Davis, J.L. eds. Between Venice and Istanbul: Colonial Landscapes in Early Modern Greece, Hesperia Supplement 40, 2007 Davis, J. L. Are the landscapes of Greek prehistory hidden? 22-35, eds. S.E. Alcock, J.F. Cherry, Side-by-side survey: comparative regional studies in the Mediterranean World, Oxford 2004 Dickinson, O. The Aegean from Bronze Age to Iron Age, continuity and change between th the 12 and 8th century BC, London-New York 2006 Dinčev, V. Rimskite vili v dneshnata b’lgarska teritoria, Sofia 1997 Dinčev,V. Ranovizantiite kreposti v B’lgaria i s’sednite zemi, Razkopki i proučvania XXXV, 2006 Dimitrioska,G. Grupa Docnorimski grobovi od Kamnik, 131-138, Macedoniae Acta Archaeologica 5, 1979 Donev, D. Some observations on the northern frontier of the ancient Macedonian Kingdom, 29-63, Macedonian Heritage 26, 2005 Dunn, A. The transition from polis to kastron in the Balkans (III-VII c): General and regional perspectives, 60-80, Byzantine and modern Greek studies 18, 1994 Dunbar, R.I.M. Coevolution of Neo-cortex size, group size and language in humans, 681735,Behavioral and Brain Sciences 16 (4), 1993 Earle, T.K. Preucel, R. Processual Archaeology and the radical critique, 501-512, Current Anthropology 28-4 1987 Erdoğu, B. The Prehistoric Settlement of Eastern Thrace, Oxford 2005 Farinetti, E. Boeotian landscapes: A GIS-based study for the reconstruction and interpretation of the archaeological datasets of ancient Boeotia, Oxford 2011 Fentress,E. What are we counting for? 44-52, eds. R. Francovich, H. PattersonExtracting meaning from ploughsoil assemblages, The Archaeology of Mediterranean Landscapes 5, Oxford 2000 Filipović, M. Severna Veleška Sela, 489-573, Srpski Etnografski Zbornik XXIX, 1935 Flannery, K. Empirical determination of Site Catchments in Oaxaca and Tehuacán, 103116, ed. K. Flannery, The Early Mesoamerican village, New York 1976 Flannery, K. ed. The Early Mesoamerican village, New York 1976 Forbes, H. Turkish and modern Methana, 101-117, eds. C. Mee, H. Forbes, A rough and rocky place: The landscape and settlement history of the Methana Peninsula, Greece, Liverpool 1997 Forbes, H. Lost souls: Ethnographic observations on manuring practices in a Mediterranean community, 159-172, ed. R. Jones, Manure Matters: Historical, Archaeological and Ethnographic Perspectives, Ashgate 2012 306 Forsen, B. Karivieri,A. The Roman-Early Modern Periods: conclusions, 307-331, eds. J. Forsen, B. Forsen, et al. The Asea Valley Survey: An Arcadian Mountain Valley from the Paleolithic Period until Modern Times, Stockholm 2003 Gallant, T.W. “Background noise” and site definition: A contribution to survey methodology, 403-418, Journal of Field Archaeology 13-4, 1986 Garašanin, M. Garašanin, D. Arheološka iskopavanja u selu Radanju, kod lokaliteta “Krivi Dol” 9-61, Zbornik na Štipskiot Naroden Muzej I, 1959 Garašanin, M. Contributions à la chronologie de l’àge du fer en Macédoine, 173-186, Živa Antika X 1-2, 1960 Garašanin, M. Garašanin, D. Arheološke beleške sa rekognosciranja u Istočnoj Makedonji, 69-95, Zbornik na Štipskiot Naroden Muzej II, 1961 Garašanin, M. Garašanin, D. Iskopuvanja na mogilata Krst, kaj selo Tarinci, 65-69, Zbornik na Štipskiot Naroden Muzej II,1961 Garašanin, M. Spasovska, G. Novi Iskopuvanja vo Zelenikovo kaj Skopje, 85-117, Macedoniae Acta Archaeologica 1, 1976 Garašanin, M. Centralnobalkanska zona, 82-106, ed. A. Benac Praistorija Jugoslovenskih Zemalja vol. II, Sarajevo 1979 Garašanin, M. The Stone Age in the Central Balkan Area, 75-186, Cambridge Ancient History, volume III, Cambridge 1983 Georgiev, D. Šerif, A. Turski dokumenti za istorijata na makedonskiot narod, Skopje 1994 Georgiev, Z. Karakterot i značenjeto na vongrobnite naodi vo južnopovardarskite nekropoli od Železno Vreme, 37-53, Macedoniae Acta Archaeologica 6, 1980 Georgiev, Z. Praistoriski naodi od Kočani i prašanje na ramnite nekropoli po Bregalnica, 65-79, Godišen Zbornik na Filozofski Fakultet 7, 1981 Georgiev, Z. Tri pred-rimski naselbi kraj Pčinja, 91-101, Godišen Zbornik na Filozofski Fakultet 17-18, 1990-1991 Georgiev, Z. Basarabi stilot i Makedonija, 81-96, Macedoniae Acta Archaeologica 13, 1992 Georgievski, B. The ancient road Scupi-Stobi after the Tabula Peuntigerianna, 117-134, Macedoniae Acta Archaeologica 12 1991 Georgievski, B. Praistoriski gradišta od kumanovsko, 51-68, Macedoniae Acta Archaeologica 13, 1992 Gimbutas, M. et al. Neolithic Macedonia, volume I, Los Angeles 1976 Given, M. Mapping and manuring: Can we compare shards density figures? 13-21, eds. S.E. Alcock, J.F. Cherry, Side-by-side survey: comparative regional studies in the Mediterranean World, Oxford 2004 Gkiasta, M. The Historiography of Landscape Research on Crete, Leiden 2008 Godja, M. ed. Archaeology, Ancient Landscape, Settlement dynamics and Nondestructive Archaeology, 2004 Goodchild,H. Modeling Roman Agricultural Production in the Middle Tiber Valley, Central Italy, unpublished PhD thesis, 2007 Grammenos, D. ed. Recent research in the prehistory of the Balkans, ThessalonicaAthens 2003 307 Gregory, T.E. Less is better: the quality of ceramic evidence from archaeological survey and practical proposals for low-impact survey in a Mediterranean context, 15-35, eds. E.F. Athanassopoulos, L-A. Wandsnider, Mediterranean Archaeological Landscapes: current issues, Philadelphia 2004 Grozdanova, V.B. Spomenici od Helenističkiot Period vo S.R. Makedonija, Skopje 1987 Grujić, R. Gusari na Svetoj Gori i Hilandarski pirg Hrusija od XII-XIV veka, 1-32, Glasnik Skopskoga Naučnoga Društva XIV, 1934 Grujić, R. Pološko-tetovska eparhija i grad Lješek, Glasnik Skopskoga Naučnoga Društva d.n. XII, 1933 Ðidrova, L. Heraklea Lyncestis; problem na hronologija i interpretacija, 269-291, Macedonia Acta Arhaeologica 15, 1999 Jameson, M.H. et al, A Greek Countryside: The southern Argolide from Prehistory to present day, Stanford 1994 Hađi-Pešić, M. B. Keramika u Srednovekovnoj Srbiji, Belgrade 1981 Halstead, P. Traditional and ancient rural economy in Mediterranean Europe: plus ça change? 77-87, Journal of Hellenic Studies 107, 1987 Halstead, P. ed. Neolithic society in Greece, Sheffield 1999 Hammond, N.G. Griffith, T.G. A history of Macedonia, vol. II, Oxford 1979 Hänsel, B. Kastanas. Die Grabung der Baubefund, Prähistorische Archäologie im Südosteuropa, Band 7-2, Berlin 1989 Harding, A.F. European Societies in the Bronze Age, Cambridge 2000 Heurtely, W. Prehistoric Macedonia, Oxford 1939 Howard, P. Spatial analysis of Boeotia field-walking survey data, 111-128, eds. J. Bintliff, P. Howard, A. Snodgrass, Testing the hinterland: the work of the Boeotia survey (19891991) in the southern approaches to the city of Thespiai, Cambridge 2007 Hochstetter, A. Kastanas: die Handgemachte Keramik, Prähistorische Archäologie in Südosteuropa, Band 3, Berlin 1984 Holt-Jensen,A. Geography: History and Concepts, London 1999 Hope-Simpson, R. Mycenaean Greece, Noyes Publishers 1981 Horden, P. Purcell,N. The Corrupting Sea: A study of Mediterranean history, Oxford 2000 Janakievski,T. Docnoantička mikrostanbena celina nad teatarot vo Heraclea Lyncestis, Bitola 2001 Josifovska, B. Prilog lokalizovanja grada Argosa u Peoniji,117-136, Živa Antika XV, 1964 Josifovska-Dragojević, B. Izveštaj za arheološkoto rekognosciranje po dolinata na Sreden Vader, 106-127, Zbornik na Arheološkiot Muzej IV-V, 1966 Josifovska-Dragojević, B. Inscriptions de la Mésie supérieure, Regions de Skopje et Kumanovo, vol. VI, Belgrade 1982 Jovanova, L. Mihailova,D. Skupi-Istočna nekropola, 203-248, Macedoniae Acta Archaeologica 15, 1999 308 Jovanova, L. Two Latin inscriptions from the village Kučeviška Bara, near Skopje 69-84, Macedonian Heritage 9-24, 2005 Jovanova, L. Skupi-Colonia Flavia Scupinorum: topografija i urbanistička struktura na gradot od I do III vek, 197-222, Macedonia Acta Arhaeologica 18, 2008 Jovanović, P.S. Reljef Skopske Kotline, 62-116, Glasnik Skopskoga Naučnoga Društva X, Skopje 1931 Kalamagdeska-Mihailova, D. Late AntiqueAREA MACERIA CINCTA from Novo SeloZelenikovo23-28, Macedonian Heritage 20, 2002 Karpuzova, S. Pecovska, N.P. Iskopuvanje vo selo Korešnica-Demir Kapija, 77-86, Macedoniae Acta Archaeologica 14, 1996 Keller, D. R. Rupp D.W. eds. Archaeological Survey in the Meiterranean Area, Oxford 1983 Kepeski, K. Villa Rustica vo Pešterica kaj Prilep, 143-156, Macedoniae Acta Archaeologica 2, 1976 Keramitčiev, A. Beleški od rekognosciranjata vo strumičkiot basen i po dolinata na Vardar, 115, Zbornik na Arheoloshki Muzej III, 1961 Keramitčiev,A. Edna docnorimska grobnica od s. Pobužje, skopsko 143-148, Živa Antika 13/14, 1964 Keramitčiev, A. Za staroto rudarstvo vo Mariovo I, 89-104, Macedoniae Acta Archaeologica 2, 1976 Keramitčiev. A. Za staroto rudarstvo vo Mariovo II, 67-81, Macedoniae Acta Archaeologica 4, 1978 Knapp, A.B. ed. Archaeology, Annales and Ethnohistory, Cambridge 1992 Kondijanov, J. ed. Numismatic collection of the National Bank of the Republic of Macedonia, Skopje 1999 Koračević, D. Praistoriski naod od Brazda, 57-65, Macedoniae Acta Archaeologica 3, 1977 Koračević, D. Skupi- gradska teritorija, Skopje 2004 Kosso, C. The archaeology of public policy in Late Roman Greece, Oxford 2003 Kotsakis, K. What tells can tell – social space and settlement in the Greek Neolithic, 6676, ed. P. Halstead, Neolithic society in Greece, Sheffield 1999 Kuna, M. Zvelebil, M. Foster, P. Dreslerová, D. Field survey and landscape archaeology research design: methodology of a regional field survey in Bohemia, 110-130, Památky Archeologické LXXXIV-2, 1995 Kuna, M. Method of surface artifact survey, 77-83, ed. E. Neustupný, Space in Prehistoric Bohemia, Prague 1998 Kuna, M. Surface artifact studies in the Czech Republic 29-44, eds. J. Bintliff, M. Kuna, N. Venclová, The future of archaeological field survey in Europe, Sheffield2000 Lavan, L. ed. Recent research in Late Antique urbanism, Journal of Roman Archaeology supplement 42, 2001 Lehmann, H. Die Siedlungsraeume Ostkretas, 212-238, Geographische Zeitschrift 45, 1939 309 Liebeschuetz, J.H.W.G. The Lower Danube under pressure: from Valens to Heraclius, 101-134, ed. A.G. Poulter, The Transition to Late Antiquity on the Danube and Beyond, Oxford 2007 Lilčić, V. Markov Grad, Matka; ukrepen srednovekoven manastir, 268-280, Istorija 19-2, 1983 Lilčić, V. Ranohristijanska zasvedena grobnica kaj s. Orman, 143-146, Macedoniae Acta Archaeologica 7-8, 1987 Lilčić, V. Docnežnoantičkite tvrdini vo Tikveš i Vitačevo, 115-136, Godišen Zbornik na Filozofskiot Fakultet 41-42, 1988-89 Lilčić, V. Antička gradežna keramika vo R. Makedonija, Skopje 1996 Lilčić, V. The canyon of Matka, 47-59, Macedonian Heritage 3, 1997 Lilčić, V. Fauces Pelagoniae, 4-50, Macedonian Heritage IV, 1997 Lilčić, V. Late Antique communications and fortification in the Mid-Vardar Region, 1353, Macedonian Heritage 7, 1998 Lilčić, V. Makedonskiot kamen za bogovite, Hristijanite i za život po životot, volumes I and II, Skopje 2001 Livy, The dawn of the Roman Empire, book 31-40, translated by J.C. Yardley, Oxford 2000 Lloyd, J. Barker, G. Rural settlement in ancient Molise: problems of archaeological survey, 289-304, eds. G.W. Barker, R.A. Hodges, Archaeology and Italian society: Prehistoric, Roman and Medieval studies, Oxford 1981 Luković, M. Tektonika i geološka istorija skopskog bazena, 4-55, Glasnik Skopskoga Naučnoga Društva X, 1931 Maneva, E. Nekolku Tipovi i Formi na Elinistička i Rimska keramika od Herakleja, Bitola 1979 Maneva, E. Srednovekoven Nakit vo Makedonija, Skopje 1992 McInerney,J. Sacred land and the margins of the community, 33-59; eds. R. Rosen, I. Sluiter, City, countryside, and the spatial organization of value in Classical Antiquity, LeidenBoston, 2006 McKay, A.G. Houses, villas and palaces in the Roman World, London 1975 Mee, C. Forbes, H. Survey Methodology, 33-41, eds. C. Mee, H. Forbes, A rough and rocky place: The landscape and settlement history of the Methana Peninsula, Greece, Liverpool 1997 Mee, C. Forbes, H. eds. A rough and rocky place: the landscape and settlement history of Methana Peninsula, Greece, Liverpool 1997 Metcalf, M.D. Avar and Slav invasions into the Balkan Peninsula (c. 575-625): The nature of numismatic evidence, 140-148, Journal of Roman Archaeology 4, 1991 Mikulčić, I. Izveštaj sa probnog iskopavanja halštatske nekropole kod s. Star Karaorman 95-112, Zbornik na Štipskiot Naroden Muzej I, 1959 Mikulčić, I. Iliro-Arhajski grobovi is okoline Štipa 197-209, Starinar XIII-XIV, 1965 Mikulčić, I.Pelagonija u svetlosti arheoloških nalaza, Beograd-Skopje 1966 Mikulčić, I. Jovanović, M. Helenistički opidum u Krševici, kod Vranja, 355-375, Vranjski Glasnik IV, 1968 Mikulčić, I.Nepoznat antički grad kaj s. Mojno, Godišen Zbornik na Filozofskiot Fakultet 22, 1970 310 Mikulčić, I. Edna antička villa rustica vo Tikveš, 269-285, Godišen Zbornik na Filozofski Fakultet 23, 1971 Mikulčić, I. Teritorija Skupa, 463-480, Živa Antika 21, 1971 Mikulčić, I. From the topography of Scupi, 29-35, Archaeologica Iugoslavica XIV, 1973 Mikulčić,I. Docnorimski grobovi od Skupi, 109-143, Godišen Zbornik na Filozofskiot Fakultet 26, 1974 Mikulčić, I. Topografija na Eudarist, 173-197, Macedoniae Acta Archaeologica 1, 1975 Mikulčić, I. Ranorimski skeletni grobovi iz Skupa, 89-102, Starinar XXIV-XXV, 1975 Mikulčić, I. Ubikacija na Bila Zora, 147-165, Godišen Zbornik na Filozofkski Fakultet 2, 1976 Mikulčić, I. Sectio Naissus-Dyrrachion-Scupis-Thessalonike, Tabula Imerii Romani, K34, Ljubljana 1976 Mikulčić, I. Nikuljska, N. Markovi Kuli-Vodno, kaj Skopje, 65-74, Macedoniae Acta Archaeologica 5, 1979 Mikulčić,I. Neilirski elementi medju skupskim nalazima na prelazu iz predrimskog u Rimsko Doba, 245-258, Sahranjivanje kod Ilira (a symposium), Beograd 1979 Mikulčić,I. Tragi od ranohristijanski baziliki okolu Skopje, 97-123, Godišen Zbornik na Filozofskiot Fakultet, 33, 1981 Mikulčić, I. Some new factors in the history of Stobi, 205-229, eds. J. Wiseman, B. Aleksova, Studies in the Antiquities of Stobi, vol. III, T. Veles, 1981 Mikulčić, I. Bilbia, M. Markovi Kuli, Vodno, Skopje, 1977-1980, 205-220, Macedoniae Acta Archaeologica 7-8, 1981-1982 Mikulčić, I. Staro Skopje so okolnite tvrdini, Skopje 1982 Mikulčić, I. Problemot na Antigonea, 111-135, Godišen Zbornik na Filozofkski Fakultet 11, 1983 Mikulčić, I. Stobi vo pred-rimsko vreme, 101-117, Godišen Zbornik na Filozofkski Fakultet 13, 1985 Mikulčić, I. Frühchristlicher Kirchenbau in der S.R. Makedonien, 221-251, Corso di cultura sull’arte ravennate et bizantina XXXIII, Ravenna 1986; Mikulčić, I. Topografija na Stenae-Prosek-Demir Kapija, 65-78, Godišen Zbornik na Filozofkski Fakultet 15, 1988-1989 Mikulčić, I. Sokolovska, V. Grobnicata vo Brazda, kaj Skopje, 79-92, Macedoniae Acta Archaeologica 11, 1990 Mikulčić, I. Srednovekovni Gradovi vo Makedonija, Skopje 1996 Mikulčić, I. Antički Gradovi vo Makedonija, Skopje 1999 Mikulčić, I. Spätantike und frühbyzantinische Befestigungen inNordmakedonien, Munich 2002 Mikulčić, I. Stobi, an ancient city, Skopje 2003 Milinković, M. Stadt oder “Stadt”: Frühbyzantinische Siedlungsstukturen im nördlichen Illyricum, 159-192, ed. J. Henning, Post-Roman towns, trade and settlement in Europe and Byzantium vol. 2, Berlin 2007 Miljković-Pepek, P. Crkvata Sv. Spas vo Kučevište, 417-421, eds. V. Mošin, Spomenici za srednovekovnata i ponovata istorija na Makedonija, volume 1, Skopje 1975 Millett, M. Pottery: population or supply pattern? The Ager Tarraconensis approach, 1826, eds. G. Barker, J. Lloyd, Roman Landscapes: Archaeological surveys in the Mediterranean Region, London 1991 311 Millett, M. Dating, quantifying and utilizing pottery assemblages from surface survey, 53-59, eds. R. Franovich, H. Patterson, Extracting meaning from ploughsoil assemblages, The Archaeology of Mediterranean Landscapes 5, Oxford 2000 Minić, D. Prilog datovanju srednjovekovne keramike iz Kruševca, 43-49, Starinar XXX, 1980 Mitrevski, D. Prilog kon poznavanjeto na Donovardarskata-Paionska grupa na Železnoto Vreme, 145-159, Macedoniae Acta Arhaeologica 12, 1990-1991 Mitrevski,D. Praistoriska nekropola Klučka-Hipodrom kaj Skopje, 61-89, Zbornik na Arheološki Muzej 1, 1995 Mitrevski, D. On the Hellenization process of the Iron Age culture in Macedonia, or on the early contacts between Macedonia and the Aegean regions, 191-203, Thracia 11, 1995 Mitrevski D. Protoistoriskite Zaednici vo Makedonija preku pogrebuvanjeto i pogrebnite običai, Skopje 1997 Mitrevski, D. New aspects on the Bronze Age sites on the northern periphery of the Mycenaean World, 449-457, ed. B. Hänsel, Man and environment in the European Bronze Age, Kiel 1998 Mitrevski, D. Prehistory in FYRO Macedonia, 13-73, ed. D.V. Grammenos, Recent research in the prehistory of the Balkans, Thessalonica-Athens 2003 Mitrevski D. et al, Vardarski Rid vol.I, Skopje 2004 Mócsy, A. Pannonia and Upper Moesia, London 1972 Morris, I. Burial and Ancient Society: the rise of the Greek city-state, Cambridge 1987 Morris, I. Archaeologies of Greece, 8-48, ed. I. Morris, Classical Greece: Ancient Histories and Modern Archaeologies, Cambridge 1994 Mošin, V. et al. Gramoti na manastirot Sv. Gorgi-Gorg Skopski, 181-209, eds. V. Mošin et al. Spomenici za srednovekovnata i ponovata istorija na Makedonija, vol 1, Skopje 1975 Mošin, V. et al. Spomenici za srednovekovnata i ponovata istorija na Makedonija, volumes I and II, Skopje 1975, 1977 Mulvin,L. Late Roman villa plans: the Danube-Balkan Region, 377-413, eds.W. Bowden, L. Lavan, C. Machado, Recent research on the Late Antique countryside, Leiden 2004 Neustupný, E. The transformation of community areas into settlement areas, 45-61, ed. E. Neustupný, Space in Prehistoric Bohemia, Prague 1998 Neustupný, E. ed. Space in Prehistoric Bohemia, Prague 1998 Nikolova, E. Stobi-zapadna nekropola, 247-270, Macedonia Acta Arhaeologica 18, 2008 Nikolovska, S. Izgradeni i obnoveni manastiri i crkvi od kralot Milutin, 509-519, eds. V. Mošin et al, Spomenici za srednovekovnata i ponovata istorija na Makedonija, vol. 2, 1977 Novaković, P. Detecting territoriality and social structure in the Bronze and Iron Ages. GIS and the hillforts in the Kras region, 101-115, ed. B. Slapšak, On the Good Use of Geographic Information Systems in Archaeological LandscapeStudies, COST Action G2, Ancient Landscapes and Rural Structures, 2001 Orton, C. Hodder, I. Spatialanalysis in archaeology, Cambridge 1976 Orton, D. Herding, Settlement and Chronology in the Balkan Neolithic, 5-40, European Journal of Archaeology 15-1 2012 Osborne, R. Demography and survey, 63-72, eds. J.F. Cherry, S.E. Alcock, Side – by side survey: comparative regional studies in the Mediterranean World,Oxford 2004 312 Osnovna Geološka Karta S.F.R. Jugoslavije, 1:100 000, sekcija K34-55, Uroševac, Belgrade 1984 Ostrogorsky, G. History of the Byzantine State, 1957 Papazoglu, F. Srednjobalkanska plemena u pred-Rimsko doba, Sarajevo 1969 Papazoglu, F. Istoriski priliki: Lihnid vo vremeto na ranoto Carstvo, 101-113, Ohrid i ohridsko niz istorijata, Skopje 1985 Papazoglu, F. Les villes de Macédoine à l'époque romaine, Bulletin de Correspondence HellèniqueSupplement XVI, 1988 Papazovska-Sanev, A. Keramika od Železnoto Vreme po dolinata na Vardar, unpublished MA thesis, 2006 Pašić, R. Gradište kao tip naselja u Skopskoj Kotlini, 145-153, Materijali: Naseljavanje i naselja u praistoriji, X kongres arheologa Jugoslavije, Prilep 1976 Peňa, J.T. Roman Pottery in the Archaeological Record, Berkeley 2007 Perlès, C. The distribution of Magoules in Eastern Thessaly, 42-56, ed. P. Halstead, Neolithic society in Greece, Sheffield 1999 Perlès, C. The early Neolithic in Greece: the first farming communities in Europe, Cambridge 2001 Petački, D. Dabici-nekropola od postaro Železno Vreme, 71-73, Arheološki Pregled 26, 1986 Petački,Ð. A complex of Late Antique vaulted tomb chambers from Veles, 7988,Macedonian Heritage5, 1997 Petrova, E. Paeonia, Skopje 1999 Pettegrew, D.K. Chasing the Classical farmstead: assessing the formation and signature of rural settlement in Greek landscape archaeology, 189-209, Journal of Mediterranean Archaeology 14, 2001 Pettegrew, D.K. The Busy Countryside of Late Roman Corinth, 743-784, Hesperia 76, 2007 Popović, V. Aux origins de la slavisation des Balkans, 230-257, Comptes rendu des sèances de Acadèmie des inscriptiones et belles lettres 1, 1980 Potter,T.W. The Changing Landscape of South Etruria, New York 1979 Poulter, A.G. The Roman to Byzantine transition in the Balkans: preliminary report on Nicopolis and its hinterland, 347-358, Journal of Roman Studies 13, 1999 Poulter,A.G. Cataclysm on the Lower Danube: The destruction of a complex Roman landscape, 223-254, ed. N. Christie, Landscapes of change: Rural evolutions in Late Antiquity and the Early Middle Ages, Ashgate 2004 Poulter, A. G. The transition to Late Antiquity, 1-50, ed. A.G. Poulter, The Transition to Late Antiquity on the Danube and Beyond, Oxford 2007 Poulter, A.G. The transition to Late antiquity on the Lower Danube: the city, a fort, and the countryside, 51-97, ed. A.G. Poulter, The Transition to Late Antiquity on the Danube and Beyond, Oxford 2007 Poulter,A.G. Site-specific field survey: the Methodology, 583-595, ed. A.G. Poulter, The transition to Late Antiquity on the Lower Danube and Beyond, Oxford 2007 “Prirodni i socio-geografski karakteristiki na zonite na nomadsko i polu-nomadsko stočarstvo vo Makedonija” (a symposium), Skopje 1981. 313 Rackham,O. Observations on the historical ecology of Laconia, 73-119,eds. W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley,The Laconia Survey: Continuity and Change in a Greek Rural Landscape, Volume I, London 2002 Radojčić, Sv.Menada iz Tetova, Glasnik Skopskoga Naučnoga Društva d.n. XII, 1933 Rapp, G.R. McDonald, A.W. eds. The Minnesota Messenia Expedition: Reconstructing a Bronze Age Regional Environment, Minneapolis 1972 Renfrew, C. Wagstaff, M. eds. An Island Polity: The Archaeology of Exploitation on Melos, Cambridge 1982 Ristevski, B. Early Medieval finds from Scupi, 71-81, Macedonian Heritage 20, 2002 Ristevski, B. Medieval ceramics from the Skopje area, 49-71, Macedonian Heritage 21, 2004 Ristevski, B. Ranosrednovekovna sadova keramika od Republika Makedonija, Vizantiiskii Vremenik 89 (65), 2006 Rorison, M. Vici in Roman Gaul, Oxford 2001 Runnels, C.N. van Andel, T.H. The Early Stone Age of the nomos of Preveza: Landscape and Settlement, 47-133; eds. J. Wiseman, K. Zachos, Landscape archaeology in southern Epirus, Greece I, Athens 2003 Sarris A. et al, The Nicopolis Project – the integration of geophysical prospection, satellite remote sensing, and GIS techniques in the study of Epirus, Greece, International Symposium of Archaeometry,Urbana, Illinois U.S.A.1996 Rutter, J.B. Some thoughts on the analysis of ceramic data generated by site surveys, 137-142, eds. D. Rupp and D. Keller, Archaeological Survey in the Meiterranean Area, Oxford 1983 Sanev, V. et al. Urgeschichte Makedoniens, Skopje 1976 Sanev, V. Neolitskata naselba Rug-Bair kaj s. Gorobinci, 203-247, Zbornik Naroden Muzej Štip IV-V, 1976 Sanev, V. Mogilite of Kunovo Čuki kaj s. Orizari-kočansko, 7-27, Zbornik na Arheološkiot Muzej VIII-IX, 1978 Serafimova,A. The Middle Ages, 93-174, Macedonia; cultural heritage, Skopje 1995 Schiffer, M.B. Sullivan, A.P. Klinger, T.C. The design of archaeological surveys, 1-28, World Archaeology, 10-1, 1978 Sherratt, A. Water, soil and seasonality in early cereal cultivation, 313-330, World Archaeology 11, 1980 Sherratt, A. Plough and pastoralism: Aspects of the secondary products revolution, 261– 305, eds. I. Hodder, G. Isaac, N. Hammond. Patterns of the past. Studies in honor of David Clarke, Cambridge 1981 Shiel, R. Stewart,A. The soils and agricultural potential of the Thespiai area, 95-109, J. Bintliff, P. Howard, A. Snodgrass et al, Testing the hinterland: the work of the Boeotia survey (1989-1991) in the southern approaches to the city of Thespiai, Cambridge 2007 Shortman, E.M. Urban, A. Resources, power and interregional interaction, Springer1992 Slapšak, B. The 1982-1986 Ager Pharensis survey. Potentials and limitations of “wall survey” in karstic environments, 145-149, eds. J.C. Chapman, et al. Recent developments in Yugoslav archaeology, Oxford 1988 Smith, J. T. Roman villas: a study in social structure, New York 1997 314 Snively,C. Golemo Gradište at Konjuh: Reports on the excavations in 2000, 293-302, Dumbarton Oaks Papers 56, 2002 Snodgrass, A.M. The Dark Age of Greece, Edinburgh 1971 Snodgrass, A.M. Cherry, J.F. On not digging up the past, 9-13, The Cambridge Review 109, 1988 Snodgrass,A.M. Survey archaeology and the rural landscape of the Greek city, 113-136, eds. O. Murray, S. Price, The Greek city from Homer to Alexander. Oxford 1990 Snodgrass, A.M. The Archaeological aspect, 197-200, ed. I. Morris, Classical Greece: Ancient Histories and Modern Archaeologies, Cambridge 1994 Sodini, J-P. The transformation of cities in Late Antiquity within the provinces of Macedonia and Epirus, 311-336, ed. A.G. Poulter, The Transition to Late Antiquity on the Danube and Beyond, Oxford 2007 Sokolovska, V. Stadion stone from Isar-Marvinci, Archaeologica Iugoslavica XXIIXXIII, 1982-1983 Sokolovska, V. Isar Marvinci i Povardarieto vo Antičko vreme, Skopje 1986 Sokolovska, V. Pajonskoto pleme Agrijani i vrskite so Damastion, 9-32, Macedoniae Acta Archaeologica 11, 1990 Sokolovski, M. Turski Dokumenti; Opširen Popisen Defter, 4 (1467-1468), Skopje 1971 Sokolovski, M. Stojanovski, A. Turski dokumenti za istorijata na makedonskiot narod, vol. II, Skopje 1973 Sokolovski, M. Vakafi i vakafski imoti vo Skopje i skopsko, vo XV i XVI vek, 42-51 Prilozi VIII, Makedonska Akademija na Nauki i Umetnosti, 1977 Srejović, D. et al. Gamzigrad: an imperial palace of late classical times, Belgrade 1983 Stankovska, Lj. Makedonska Onomastika, Skopje 2002 Stankovski, J. Tumulot I od tumularnata nekropola kaj s. Strnovac, opština Staro Nagoričino, 135-153, Macedoniae Acta Archaeologica 18, 2008 Stojanović-Anderson, V. Stobi, the Hellensitic and Roman Pottery, Princeton 1992 Stojanovski, A. Dervenđistvoto vo Makedonija, Skopje 1974 Stojanovski, A.Makedonija vo turskoto srednovekovie (XIV-XVIII vek), Skopje 1989 Sunders,G.D.R. Recent developments in the chronology of Byzantine Corinth, 385-399, eds. C.K. Williams II, N. Bookidis, Corinth: the Centenary, 1896-1996, Athens 2003 Sutton, S. B. Settlement patterns, settlement perceptions: Rethinking the Greek village, 313-336, ed. P.N. Kardulias, Beyond the Site: Regional Studies in the Aegean, 1994 Tanev, Gj. Počvite pomegju Pčinja i Vardar, sekcija Skopje-Veles II, Skopje 1969 Terrenato,N. Sample Size Matters! The Paradox of global trends and local surveys, 3648; eds. S.E. Alcock, J.F. Cherry, Side-by-side survey: comparative regional studies in the Mediterranean World, Oxford 2004 Todorović, D.B. Pedološka ispitivanja u skopskog bazena, 242-268, Glasnik Skopskoga Naučnoga Društva X 1931 Tomoski, T. Dali postoel grad Polog,Glasnik na Institutot za Nacionalna Istorija I-1, 1959 Tomoski, T. Beitrage zu einer Rekonstruktion der Strecke Scupi-Stobi auf der Tabula Peutingeriana, 113-125, Ziva Antika VI-1 1961 Tomoski, T. Skopje od XI do XIV vek, 57-74, eds. V. Mošin et al, Spomenici, za Srednovekovnata i ponovata istorija na Makedonija, volume 1, Skopje 1975 315 Tomoski, T. Ovče Polje vo Sredniot Vek, 243-265, Godišen Zbornik na Filozofskiot Fakultet 30, 1978 Tilley, C. A phenomenology of Landscape, Oxford 1994 Tomoski,T. Skopskata oblast of XI do XIV vek, 57-75, ed. V. Mošin, Spomenici za srednovekovnata i ponovata istorija na Makedonija, vol.1 Skopje 1975 Tomić, S. Skopska Crna Gora, 409-509, Srpski Etnografski Zbornik III 1905 Trifunovski, J. Seoska Naselja Skopskog Polja, 345-517, Srpski Etnografski Zbornik LXIX, 1955 Trifunovski, J. Isčeznati naselbi na Skopska Crna Gora, 145-149, Godišen Zbronik na Prirodno-Matematički Fakultet 18, 1970 Trifunovski, J. Tri isčezla Srednjovekovna sela, 156-157, Glasnik Srpskog Geografskog Društva I/2, 1970 Trifunovski, J. Skopska Crna Gora-anthropogeographische Forschungen, Skopje 1971 Trifunovski, J. Les arbres culturelles dans la region de Skopska Crna Gora, 129-137, Bulletin de l’institut ethnographique de l’Academie Serbe XXIV, 1975 Trifunovski, J. Manastirski konaci u Skopskoj Crnoj Gori, 113-124 Glasnik Etnografskog Instituta XXIX 1980 Trigger, B. A History of Archaeological Thought, Cambridge 1989 van Berghem, J-W Fiselier, J. Soils and Land Use Potential, 57-71,eds. W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al. The Laconia Survey: Continuity and Change in a Greek Rural Landscape, Volume I,London 2002 van der Enden, M. Koroneia 2009: Surface survey sampling experiments, 33-41, J. Bintliff, B. Slapšak, The Leiden-Ljubljana Ancient Cities of Boeotia 2009 seasons, 1-58,Pharos 17(2), 2012 Vasić, R. Kulturne grupe starijeg gvozdenog doba u Jugoslavii, Belgrade 1973 Vasić, R. Gevgeliska grupa starijeg Gvozdenog Doba, 637-38, ed. A. Benac, Praistorija Jugoslovenskih Zemalja V, Sarajevo 1987 Videvski, Z. Bronzenoto Vreme na Vardarski Rid, 91-113, ed. D. Mitrevski, Vardarski Rid-Staromakedonski grad, vol I, Skopje 2005 Vita-Finzi C. Higgs, E.S. Prehistoric economy in the Mount Carmel area of Palestine: Site catchment analysis, 1-37, Proceedings of the Prehistoric Society 36, 1970 Vionis,A.K. The Medieval and post-Medieval pottery and Tanagra village history, 570578, J.L. Bintliff, et al. The Tanagra Project: Investigations at an ancient Boeotian city and in its countryside, 541-606, Bulletin de Correspondence Hellènique 128-129, 2004-2005 Vionis, A. “Crusader” and “Ottoman”Material Life: The Archaeology of Built Environment and Domestic Material Culture in the Medieval and post-Medieval Cyclades, Greece, Leiden 2006 Vionis, A. Material culture studies: the case of the Medieval and post-Medieval Cyclades, 177-197, eds. J. Bintliff, H. Stoeger, Medieval and post-Medieval Greece, the Corfu papers, Oxford 2009 Vionis, A.K. Poblome, J. Waelkens, M. The hidden material culture of the Dark Ages. Early Medieval ceramics from Sagalassos (Turkey): new evidence (ca AD 650-800), 147-165, Anatolian Studies 29, 2009 Vroom, J. After Antiquity. Ceramics and Society in the Aegean from the 7th to the 20th centuries A.C. A Case Study from Boeotia, Central Greece, Leiden, 2003 316 Vroom, J. Byzantine to Modern Pottery in the Aegean: An Introduction and Field Guide, Utrecht 2005 Vujević, P. Klima skopskog bazena, 125-240, Glasnik Skopskoga Naučnoga Društva X, 1931 Vulić, N. Teritorija rimskog Skoplja, Glasnik Skopskoga Naučnoga Društva d.n I/1, 1924 Vulić, N. Antički spomenici naše zemlje, Spomenik Srpske Kraljevske Akademije LXXI, Belgrade 1931 Vulić, N. Geografija Južne Srbije u antičko doba; Glasnik Skopskoga Naučnoga Društva d.n. XIX, 1938 Vučković-Todorović, D. Antička Demir Kapija, 229-267, Starinar XII, 1961 Wagstaff, M. The formation of the modern landscape of the survey area, 403-420, eds. W. Cavanagh, J. Crouwel, R.W.V. Catling, G. Shipley et al.The Laconia Survey: Continuity and Change in a Greek Rural Landscape, Volume I,London 2002 Wandsnider, L. Camilli, E.L. The character of surface archaeological deposits and its influence on survey accuracy, 169-188, Journal of Field Archaeology 19-2, 1992 Wessolowski, A. Burial customs in the west cemetery 97-141, eds.J. Wiseman, Dj. Mano-Zissi, et al. Studies in the Antiquities of Stobi I, T. Veles 1973 Whallon, R. Methods of Controlled Surface Collections in Archaeological Surveys, 7383, eds. D. Rupp and D. Keller, Archaeological Survey in the Meiterranean Area, Oxford 1983 White, K.D. Roman farming, London 1970 Whitelaw, T. Reconstructing a Classical Landscape with Figures: some interpretative explorations in NW Keos, 227-243, eds. R. Francovich, H. Patterson, Extracting meaning from ploughsoil assemblages, The Archaeology of Mediterranean Landscapes 5, Oxford 2000 Wilkinson, T.J. The definition of ancient manured zones by means of extensive shardsampling techniques, 323-333, Journal of Field Archaeology 9, 1982 Wilkinson, T.J. Extensive shard scatters and land-use intensity: some recent results, 3146 Journal of Field Archaeology 16-1, 1989 Wiseman J. Mano-Zissi, Ð. eds. Studies in the Antiquities of Stobi, vol. I-III, Skopje-T. Veles 1973, 1975, 1979 Wiseman, J. Zachos, K. eds. Landscape Archaeology in southern Epirus, Greece, vol. I, Hesperia Supplements, 2003 Whittaker,C.R. Late Roman trade and traders, 163-180, eds. P. Garnsey, K. Hopkins, C.R. Whittaker, Trade in the Ancient Economy, Berkeley-Los Angeles 1983 Wright J. et al, The Nemea Valley Archaeological Project, a preliminary report, 579-659, Hesperia 59-3, 1990 Zangger E. et al, The Pylos Regional Archaeological Project, part II: Landscape Evolution and Site Preservation, 549-641, Hesperia 66-4, 1997 Zdravkovski, D. Sreden Neolit vo Gorno-Vardarski Region, 55-63, Macedonia Acta Arheologica 18, 2008 Šeldarov, N. Macedonia and Paeonia, Skopje 2003 Šurbanovska, M. Rimska keramika od Stobi 151-162, Macedonia Acta Arhaeologica 17, 2006 317 Šurbanovska M. et al. Arheološki istrazuvanja-Stobi 2001, 271-282, Macedonia Acta Arhaeologica 18, 2008 www.mav.mk www.tafhr.org List of maps Map I_1:The central parts of the Balkan Peninsula Map I_2:The Republic of Macedonia with the main interregional corridors 318 Map I_3: Section of the mid-Vardar Valley showing borders of topographic entities, red line and 1.5 km buffers, black circles, around existing settlements Map I_4: 1: 2 500 vertical map of the first survey area with the field walking units Map I_5: Field blocks and pottery counts per individual transects Map I_6: Regular grids one and two over field blocks in the first survey List of tables TableI_1: The field walking forms Table III_1: Location of the central settlements in relation to the main east-west axis TableIV_1: List of the villages in Skopjan Montenegro and the number of houses and inhabitants at the end of the 19th century (after S. Tomić, 1905, 507-508) TableV_1: The area of Roman clusters documented by the grid survey Table VI_1: Size and possible rank of the settlement sites discovered in the two survey areas List of graphs Graph III_1: Distribution of the Late Iron Age assemblage by fabrics Graph III_2: Total site areas by period, in hectares (the area of the castra excluded) Graph V_1: Chronological profiles of sample and total collections from various grid units Graph V_2: Composition of the total collections from the off-site and the halo zone Graph V_3: Revealed site area and max artifact density (discards included) Graph V_4: Revealed site areas (x axis, sq m) and maximum halo radii (y axis, ha)Graph V_5: Revealed site areas and maximum halo areas Graph V_6: Composition of the on-site assemblages Graph V_7: Composition of the joint assemblages from the central and the satellite clusters Graph VI_1: Maximum radii for the site halos and impact zones and estimated site areas List of maps (on CD) Map II_1: The Republic of Macedonia, the black square marks the survey area Map II_2: Section of the Taor Canyon with the natural limits of the valley of Sopot Map II_2a: Lower course of the central valley Map II_2b: The upper course and the headwaters of the central valley Map II_3: Geological framework of the first survey area Map II_2c: Rectified aerial photograph of Sopot’s administrative area Map II_4: Division of the survey area into 11 sectors Map II_5: Ground visibility in sector I Map II_6: Ground visibility in sectors II through V Map II_7: Ground visibility in sectors IV through X Map II_9: Sector XI, ground visibility conditions Map III_1a: Distribution of the Middle Neolithic material by field units (Appendix 2) Map III_1b: Density of the Middle Neolithic finds per 1000 sq meters 319 Map III_2: Distribution of the Middle Neolithic finds by grid units Map III_2a: Density of the Middle Neolithic finds in 1000 sq meters (Appendix 2) Map III_2b: Density of the MN material in 1000 sq m, adjusted collections (Appendix 2) Map III_3: Distribution of the assemblage possibly dating to the Late Neolithic Map III_4a: Distribution of the possible Late Neolithic finds in sector IX Map III_4c: Density of the Late Neolithic finds, adjusted collections Map III_5: Distribution of the Bronze Age finds in the survey area Map III_6: Distribution of the Bronze Age finds on site 3 Map III_7: The Bronze Age settlement in the survey area Map III_8a: Distribution of the Late Bronze Age finds, adjusted collections Map III_8b: Distribution of the Late Bronze Age finds by transect units, adjusted collections Map III_9: Distribution of the Late Bronze Age finds on site 3 Map III_10: Distribution of the Late Bronze Age finds by grid units Map III_10a: Distribution of finds datable to the Late Bronze Age and the Late Bronze Age-Early Iron Age Map III_11a: Distribution of the Late Bronze Age-Early Iron Age finds. Map III_11b: Distribution of the Late Bronze Age-Early Iron Age finds, adjusted collections Map III_12: Distribution of the Late Bronze Age-Early Iron Age finds on site 7 Map III_13: Distribution of the Late Bronze Age-Early Iron Age finds at the foot of site 9 Map III_15a: Distribution of the Early Iron Age assemblage by transect units Map III_15b: Distribution of the Early Iron Age finds, adjusted transect collections Map III_16: Distribution of the Early Iron Age finds on site 6 Map III_17: Distribution of the Early Iron Age finds on site 5a-5b Map III_18a: Distribution of the Late Iron Age finds by transect units Map III_18b: Distribution of the Late Iron Age finds, adjusted collections Map III_19: Distribution of the Late Iron Age finds on site 8 Map III_20: Distribution of the Late Iron Age fabric groups Map III_21a: Distribution of the Hellenistic material Map III_21b: Distribution of the Hellenistic material, adjusted collections Map III_24: Distribution of the Hellenistic finds on site 12 Map III_25a: Distribution of the Roman to Late Roman finds in the survey area. Map III_25b: Distribution of the Roman to Late Roman finds in the survey area, adjusted collections Map III_26a: Distribution of the Roman to Late Roman finds in sector I Map III_26b: Density of the Roman to Late Roman finds in sector I Map III_26c: Density of the Roman to Late Roman finds in sector I, adjusted collections Map III_27a: Distribution of the Roman to Late Roman finds in sectors II and III Map III_27b: Density distribution of the Roman to Late Roman finds in sectors II-V, adjusted collections Map III_28: Map 28: Distribution of the R-LR finds on site 5a-5b and the surrounding fields and the number of tile fragments per grid unit Map III_30: Map III_30: Distribution of the fabric groups possibly dating to the pre-4th century Map III_31a: Distribution of the Roman to Late Roman finds in sectors V and VII 320 Map III_31b: Density of the Roman to Late Roman finds in sectors V-IX, adjusted collections Map III_32: Distribution of the Roman to Late Roman finds in sector VII Map III_32a: Density of the Roman to Late Roman finds in sectors, VI-VII, IX-X Map III_32b: Density of the Roman to Late Roman finds in sectors, VI-VII, IX-X, adjusted collections Map III_33: Distribution of the Roman to Late Romanfinds on site 11 and sector IX Map III_34: Distribution of the Roman finds on site 12 Map III_35: Distribution of theRoman to Late Romanfabric groups on site 12 Map III_36a: Distribution of the Roman to Late Roman finds in sector X Map III_36b: Distribution of the Roman to Late Romanfinds in sector X, adjusted collections Map III_37a: Distribution of the Roman to Late Romanfinds on sites 13a and 13b Map III_37b: Distribution of the Roman to Late Roman finds on sites 13a-13b, adjusted collections Map III_38a: Distribution of the Roman to Late Roman finds on sites 14-15 Map III_38b: Distribution of the Roman to Late Roman finds on sites 14-15, adjusted collections Map III_38c: Density of the Roman to Late Roman finds in sector VIII, in 1000 sq meters Map III_38d: Distribution of the Roman to Late Roman finds in sector VIII, adjusted collections Map III_39a: Distribution of the R-LR finds in sector XI, in 1000 sq meters Map III_39b: Distribution of the Roman to Late Roman finds in sector XI, adjusted collections Map III_42: Approximate sketch of the architectural surface remains of Late MedievalEarly Ottoman Sopot Map III_43: Distribution of finds by period in sector II, grid and transect units Map III_44: Distribution of the Late Byzantine_Early Ottoman finds in sector II and grid coverage Map III_44a: Density of the Late Byzantine_Early Ottomanfinds in sector II Map III_44b: Density of the Late Byzantine_Early Ottoman finds, adjusted collections Map III_45: Distribution of the LO-EM finds in the survey area, grid and transects Map III_46a: Distribution of the Late Ottoman_Early Modern finds in the western survey half Map III_46c: Distribution of the Late Ottoman_Early Modern finds in the western survey half, adjusted collections Map III_46b: Distribution of the Late Ottoman_Early Modern finds in the eastern survey half Map III_46d: Distribution of the Late Ottoman_Early Modern finds in the eastern survey half, adjusted collections Map III_46e: Distribution of the Late Ottoman_Early Modern finds in sector XI, adjusted collections Map III_46f: Distribution of the Late Ottoman_Early Modern finds in the western survey sectors, grid and transect 321 Map III_46h: Distribution of the Late Ottoman_Early Modern finds in the eastern survey sector, grid and transect Map III_46g: Distribution of the Late Ottoman_Early Modern finds in sectors V and VII, grid and transects Map III_47: Settlement and land-use during the Mid-Neolithic Map III_48: Distribution of the Neolithic settlements in the regions of Skopje, Sveti Nikole and Veles, after the archaeological atlas; red stars-attested EN layer, black stars-broad Neolithic dating Map III_49: Settlement and land-use during the Late Neolithic Map III_50: Settlements broadly dated to the Bronze Age , black stars and to the “Dark Ages”, red stars Map III_51: Settlement and land-use in the survey area during the Bronze Age Map III_52: Settlement and land-use during the Late Bronze Age Map III_53: Settlement and land-use during the Late Bronze Age_Early Iron Age Map III_54: Settlement and land-use during the Early Iron Age Map III_55: Distribution of the Late Iron Age sites, black stars-necropolis, red starssettlements Map III_56: Settlement and land-use during the Late Iron Age in the survey area Map III_57: Distribution of settlements during the Hellenistic Period; black stars-older centres with Hellenistic phase, red stars-newly founded centres; red circles-flat Hellenisticsettlements Map III_58: Settlement and land-use during the Hellenistic Period Map III_59: Distribution of Early and Mid Roman sites, circles-mounds Map III_61: Settlement and land-use during the Late Roman Period Map III_62: Distribution of sites broadly dated to the Middle Ages, star-towns, crossescemeteries, squares-monasteries, black circles-small forts, red circles-flat settlements Map III_63: Medieval to Modern Veles and its environs Map III_64: A Middle to Late Byzantine fort and monastery, 6 kilometers north of Sopot Map III_65: Settlement and land-use during the Late Byzantine_Early OttomanPeriod Map III_66: Settlement and land-use during the Late Ottoman_Early ModernPeriod Map III_67: Distribution of archaeological sites in the survey area; red line represents the modern highways, the main east-west axis; the central stream is the north-south axis Map III_68: Theoretical territories of sites 5a-5b and 4; 2 km radii black line; 500 m red line Map III_69: Modified theoretical catchments of the three settlement loci in the survey area Map IV_1a: The central and western Balkans Map IV_1b: Skopian Montenegro and the Skopje Basin Map IV_2: The region of Skopian Montenegro Map IV_3: The area of the foothills Map IV_4: Hydrologic divisions in the wider study region Map IV_5: Geology of the wider study region Map IV_6: Thiessen polygons drawn around modern villages Map IV_7: The western half of the foothills Map IV_8: Borders of the second survey area 322 Map V_1: Distribution of the prehistoric finds in the second survey area Map V_2: Distribution of certainly dated prehistoric shards Map V_3: Distribution of all finds possibly dating to prehistory on the eastern ridge Map V_3a: Distribution of the prehistoric finds on the western ridge Map V_4a: Distribution of the Roman material in the second survey area Map V_4b: Distribution of the Roman material and grid coverage Map V_4c: Distribution of the Roman finds, adjusted transect collections Map V_5a: Distribution of the Roman finds across sites 1-3 Map V_5b: Distribution of the Roman finds by grid units on sites 1-3 Map V_5c: Distribution of the Roman finds across sites 1-3, adjusted collections Map V_6a: Distribution of the Roman finds across site 14 Map V_6b: Distribution of the Roman material by grid units on site 14 Map V_6c: Distribution of the Roman finds across site 14, adjusted collections Map V_7a: Distribution of the Roman material across site 11-5 Map V_7b: Distribution of the Roman finds by grid units on sites 11-5 Map V_7c: Distribution of the Roman finds across site 11-5, adjusted collections Map V_19a: Distribution of the Roman material across field blocks 47a-b, 49, 50, 61 Map V_19b: Approximate extent of the cluster on field blocks 47a-b, 49, 50 and 61 Map V_19c: Distribution of the Roman finds across field blocks 47a-b, 49, 50 and 61 Map V_8a: Distribution of the Roman finds in the northern central parts of the survey area Map V_8b: Distribution of the Roman finds on the sites on grids 15-18 and 27 Map V_8c: Distribution of the Roman finds in the northern central parts of the basin, adjusted collections Map V_9: Distribution of fabric categories on grids 15-18 and 27 Map V_10: Approximate extent of the site on grids 15-18 Map V_11a: Distribution of the Roman material on field blocks along the eastern survey limit Map V_11b: Distribution of the Roman finds on grids 4, 6-8 and 10 Map V_11c: Distribution of the Roman finds on field blocks along the eastern survey limits, adjusted collections Map V_12: Distribution of fabric categories on grids 4, 6-8 and 10 Map V_13a: Distribution of the Roman finds on field blocks in the southern half of the eastern ridge Map V_13b: Distribution of the Roman material on grids 9, 10 and 12 Map V_13c: Distribution of the Roman finds in the southern half of the eastern ridge, adjusted collections Map V_13d: Distribution of the Roman finds across the site on field blocks 289-291, adjusted collections Map V_20a: Approximate extent of the site on field blocks 289a-290a Map V_20b: Distribution of the Roman finds on field blocks 289a-291a Map V_14a: Distribution of the Roman finds in the central section of the survey area Map V_14b: Distribution of the Roman finds on grids 12-13, 19 and 20 Map V_14c: Distribution of the Roman finds in the central survey section, adjusted collections 323 Map V_15: Distribution of fabric groups on grids 12, 19 and 20 Map V_16a: Distribution of the Roman finds on site 26 and the surroundings Map V_16b: Distribution of the Roman finds on grid 26 Map V_16c: Distribution of the Roman finds on site 26 and its surroundings, adjusted collections Map V_17: Distribution of fabric groups on site 26 and its surroundings Map V_18a: Distribution of the Roman finds on the western ridge Map V_18b: Distribution of the Roman finds on grids 21-25, the western ridge Map V_18c: Distribution of the Roman finds on the western ridge, adjusted collections Map V_18d: Distribution of the Roman finds on grids 21 and 23-24 Map V_21: Distribution of the clusters with on-site densities Map V_22: Interpretational map of the integral network of Roman sites Map V_23a: Distribution of the Late Roman finds in the second survey area Map V_23b: Distribution of the Late Roman finds/number of finds per transect Map V_23c: Distribution of the Late Roman finds, adjusted transect collections Map V_24: Distribution of the Late Roman finds collected by grid and transect units Map V_25a: Distribution of finds broadly dated to the Middle Age Map V_25b: Distribution of finds possibly dated to the Middle Age, adjusted collections Map V_26: Distribution of finds possibly dated to the Middle Age on grid 25 Map V_27a: Distribution of the Late Byzantine_Early Ottoman finds Map V_27b: Distribution of the Late Byzantine_Early Ottoman finds, adjusted transect collections Map V_28a: Distribution of the Late Byzantine_Early Ottoman finds by grid units Map V_28b: Distribution of the Late Byzantine_Early Ottoman finds on the site on grid 25 Map V_29a: Distribution of the Late Ottoman_Early Modern finds Map V_29b: Distribution of the Late Ottoman_Early Modernfinds, adjusted transect collections Map V_30: Overall distribution of the Late Ottoman_Early Modernfinds, grid and transect units Map V_31: Distribution of the Late Ottoman_Early Modernfinds in the northeast survey quarter Map V_32: Distribution of the Late Ottoman_Early Modernfinds in the central and southwestern survey sections Map V_33: Interpretative map of the prehistoric surface finds Map V_34: Distribution of the pre-Roman sites from the Skopje Basin Map V_35: Interpretative map of the surface remains from the Roman Period Map V_36: Distribution of the Roman sites in the Skopje Basin (end of the 4th century) Map V_37: Distribution of the Roman sites (end of the 4th century) by site categories Map V_38: Distribution of the Roman epigraphic monuments from the wider study region Map V_39: Distribution of the Late Roman sites in Skopian Montenegro by categories Map V_40: Distribution of the post-Antique sites in the Skopje Basin Map V_41: Interpretative map of the surface remains from the Medieval Period Map V_42: Distribution of the post-Antique settlements in the wider study region 324 List of photos (on CD) Photo II_1: Section of the Taor Canyon Photo II_2: The southern slopes of Prisoj Photo II_3: The western slopes of Radičica Photo II_4: The hillock at the southeast corner of the survey area Photo II_5: A view of the Taor Canyon Photo II_7: Sectors VI and VII Photo II_11: The lower half of the survey basin Photo II_8: Sectors VII and IX Photo II_9: Sectors IX and X Photo II_10: The upper course of the valley Photo III_1: A view on the western half of site 4, Late Byzantine-Early Ottoman Sopot Photo IV_1: Agricultural divisions in the second survey area Photo IV_2: A view at the second survey area Contents of the appendices (on CD) Contents of appendix I: Analysis of the survey results; the distribution of the total surface record in the first survey area App. I.1: The analyzing procedures App. I.2: The overall distribution of surface finds App. I.3.1: Analysis of the distribution of surface finds by individual sectors App. I.3.2: Sector I, the terraces on the Vardar’s left bank App. I.3.3: Sector II, Sopot SW App. I.3.4: Sector III, Sopot NW App. I.3.5: Sector IV, Sopot SE App. I.3.6: Sector V, Sopot NE App. I.3.7: Sector VI, the central valley App. I.3.8.1:Sector VII, the eastern foothills of Prisoj App. I.3.8.2: “Site number 8” 25 App. I.3.8.3 Site 9, grids 14 and 15 App. I.3.9: Sector VIII, the northern foot of Gaber App. I.3.10: Sector IX, Ramnište App. I.3.11: Sector X, Jakupica App. I.3.12: Sector XI, the upper course of the central valley App. I.4: Conclusion List of tables in Appendix 1 Table 1: Field blocks west and east of the central valley Table 2: Statistical distribution of the overall surface record in sector 1 325 1 2 6 8 11 14 18 19 23 24 29 31 33 37 41 42 Table 3: Statistical distribution of the overall surface record on grid 16a Table 4: Statistical distribution of the overall surface record on grid 16b Table 5: Statistical distribution of the overall surface record in Sopot SW Table 6: Statistical distribution of the overall surface record in Sopot NW Table 7: Statistical distribution of the overall surface record on grid 5a Table 8: Statistical distribution of the overall surface record on grid 5b Table 9:Statistical distribution of the overall surface record in sector IV Table 10: Statistical distribution of the overall surface record in sector V or Sopot NE Table 11: Statistical distribution of the overall surface record on grid 6 Table 12: Statistical distribution of the overall surface record on grid 7 Table 13: Statistical distribution of the overall surface record on grid 8 Table 14: Statistical distribution of the overall surface record in sector VI Table 15: Statistical distribution of the overall surface record in sector VII Table 16: Statistical distribution of the overall surface record on grid 9 Table 17: Statistical distribution of the overall surface record on grid 10 Table 18: Statistical distribution of the overall surface record on grid 12 Table 19: Statistical distribution of the overall surface record on grid 13 Table 20: Statistical distribution of the overall surface record on grid 14 Table 21: Statistical distribution of the overall surface record on grid 15 Table 22: Statistical distribution of the overall surface record in sector VIII Table 23: Statistical distribution of the overall surface record in sector IX Table 24: Statistical distribution of the overall surface record on grid 11 Table 25: Statistical distribution of the overall surface record on grid 17 Table 26: Statistical distribution of the overall surface record in sector X Table 27: Distribution of the overall surface record on sites 14 and 15 Table 28:Distribution of the overall surface record on sites 13 and 13a Table 29: Statistical distribution of the overall surface record in sector XI Table 30: size of the documented sites and density of finds on the surface List of graphs in Appendix I Graph 1: Number of counted ceramic fragments per sectors Graph 2: Overall mean densities in some regions of the Mediterranean and the Near East in 1000 sq meters List of photos in Appendix I Photo 1: The rectangular tower in sector I Photo 2: Site 4, the upper eastern terrace Photo 3: Site 10, a view from the northwest Photo 4: Sopot and the southern foot of Prisoj, a view from site 10 List of maps in Appendix I Map 1: Division of the survey area into field blocks and ground visibility Map 2: Distribution of the overall surface record in the first survey area 326 Map 3: Distribution of bone and modern rubbish in the first survey area Map 4: Distribution of architectural remains in the first survey area Map 25: Distribution of the known and newly discovered forts in the first survey area Map 5: Distribution of the overall surface record in sector I Map 6: Distribution of the overall surface record in sector I, quantity and weight (in gr) Map 7a: Distribution of the overall surface record in sector II-Sopot SW Map 7b: Number of ceramic fragments by individual field walking transects in sector II Map 7c: Number of finds counted by individual transects and grid units in sector II Map 8: A simplified plan of the architectural remains on site 4 Map 9a: Distribution of the overall surface record in sector 3-Sopot NW, grid and transect Map 9b: Distribution of the overall surface record on site 5a-5b Map 10: Distribution of the overall surface record in sector IV Map 10a: Distribution of the overall surface record in sector IV-Sopot SE, by individual transects Map 11: Distribution of the overall surface record in sector V-Sopot NE, by individual transects Map 12: Distribution of the overall surface record on grids 6 and 7 Map 13: Distribution of the overall surface record on grids 6 and 7, combined grid and transect Map 14: Weight distribution of the overall surface record on grid 7 Map 15: Distribution of the overall surface record on grid 8 Map 16: Weight distribution of the overall surface record on grid 8 Map 17: Distribution of the overall surface record in sectors VI-X Map 18: Distribution of the overall surface record in sector VII, grid and transect units Map 19a: Distribution of the overall surface record on site 8 Map 19b: Weight distribution of the total collections on site 8 Map 20: Distribution of the overall surface record on grids 12-14 Map 21: Weight distribution of the total collections from grids 12-14 Map 22: Distribution of the overall surface record in the northern end of sector VII Map 23: Weight distribution of the total collections from grids 14 and 15 Map 24: Distribution of the overall surface record in sector VIII Map 27: Distribution of the overall surface record on grid 11 and the surrounding field blocks Map 28: Weight distribution of the total collections from grid 11 Map 29: Distribution of the overall surface record on grid 17 (site 12) Map 30: Weight distribution of the total collections from grid 17 Map 31a: Distribution of the overall surface record in the southern half of sector X Map 31b: Distribution of the overall surface record in the northern half of sector X Map 32: Distribution of the overall surface record on sites 14 and 15 Map 33: Distribution of the overall surface record on sites 13a and 13b Map 34: Distribution of the overall surface record and the burial mounds in the eastern survey sectors Map 35: Distribution of the overall surface record in sector XI Map 36: Distribution of the overall surface record in sectors I-X, grid and transect units 327 Contents of appendix II:Surface material distribution by periods from survey area 1 App. II.1:The Middle Neolithic 1 App. II.2: The Late Neolithic (5th and first half of the 4th millennium?) 4 rd nd App. II.3: The Bronze Age (3 and 2 millennium BC, until c.a. 1200 BC) 9 App. II.4: The Late Bronze Age (1600-1200 BC) 12 App. II.5: The Late Bronze-Early Iron Age? (1200-1000 BC) 15 App. II.6: The Early Iron Age? (1000-800BC) 22 App. II.7: The Late Iron Age (7th and 6th century BC) 28 App. II.8: The Hellenistic Period (late 4th – late 1st century BC) 38 App. II.9.1: The Roman to Late Roman Period (AD 1st-end of 6th century) 44 App. II.9.2: “Site 2” and the distribution of the Roman-Late Roman finds in sector I 49 App. II.9.3: Site 5a-b and surrounding clusters 50 App. II.9.4: The off-site zone, sectors IV and V 57 App. II.9.5: Sector VII, the Late Roman phase on site 8 58 App. II.9.6: The field blocks surrounding site 9 61 App. II.9.7: Sector IX, Ramnište and site 12 64 App. II.9.8: Sector X, sites 13a-13b, 14 68 App. II.9.9: Sectors VIII and XI 74 th th App. II.10: The Late Byzantine and Early Ottoman Period (14 -18 century) 76 App. II.11: The Late Ottoman and Early Modern Period (1800-1950) 82 List of tables in Appendix II Table 1: Amount and distribution of the Middle-Neolithic finds Table 2: Mean district density and density on field blocks covering the site area in 1000 sq meters Table 3: Amounts and distribution of the assemblage possibly dating to the Late Neolithic Table 4: Mean district density and density on field units covering possible site areas, in 1000 sq meters Table 5: Amount and distribution of the Bronze Age finds Table 6: Mean district (grid and transect) and on-site density of the Bronze Age finds Table7: Amount and distribution of the Late Bronze Age finds Table 8: Mean district density and density on field blocks covering find-spots of Late Bronze Age material, in 1000 sq meters Table 9: Mean district and on-site density recorded by the grid survey, in100 sq meters Table 10: Amount and distribution of the assemblage dated to the Dark Ages Table 11: Mean district density and density on field blocks covering potential site locations, in 1000 sq meters Table 12: Mean district density and density on find-spots recorded by the grid survey, in 100 sq meters Table 13: Amount and distribution of finds datable to the Early Iron Age Table 14: Mean district density and density on field blocks featuring EIA material, in 1000 sq meters Table 15: Amount and distribution of finds datable to the Late Iron Age 328 Table 16: Mean density in the western half of the survey area and density on possible site locations, in 1000 sq meters Table 17: Amount and distribution of finds datable to the Hellenistic Period Table 18: Mean district densities and densities on field blocks with Hellenistic material recorded by transect units, in 1000 sq meters Table 19: Amount and distribution of the finds datable to the Roman Period Table 20: Mean overall density and density on field blocks covering probable site locations, in 1000 sq meters Table 21: Mean overall density and density on field blocks with possible Roman to Late Roman sites, in 1000 sq meters Table 22 Mean district density and average density recorded by regular grid units, in 100 sq meters Table 23: Mean district density recorded by the grid survey and average densities per grid, in 100 sq meters Table 24: Mean overall density and density on field blocks with Roman material in sector IX, in 1000 sq meters Table 25: Mean district density and average density per grid in sector IX, in 100 sq meters Table 26: Mean overall and density on field blocks covering possible site locations in sector X, in 1000 sq meters Table 27: Mean overall density of R-LR finds and the maximal densities recorded by the transect survey in sectors VIII and XI, in 1000 sq meters Table: 28 Amount and distribution of finds datable to the Late Byzantine and Early Ottoman Period Table 29: Mean district density and density on field blocks covering the area of site 4, in 1000 sq meters Table 30: Amount and distribution of finds datable to the Late Ottoman and Early Modern Period Table 31: Mean district values for sectors II and III and various levels of density by field blocks, in 1000 sq meters List of graphs in Appendix II Graph 1: Distribution of finds by periods on site 11 Graph 2: Distribution of finds by period on site 3 and surrounding field units Graph 3: Distribution of finds by period on “site 7” and adjacent fields to the south Graph 4: Distribution of finds by periods on site 9 and adjacent fields Graph 5: Distribution of finds by period on site 6 and the surrounding fields Graph 6: Distribution of finds by period on sites 5a and 5b Graph 7: Chronology of the material gathered by transect units from sectors II through VII Graph 8: Distribution of finds by period on different parts of “site 8” and adjacent fields Graph 9: Distribution by periods of the material from field blocks 31, 80, 87 and 107 Graph 10: Distribution of finds by period on “site 12” and the surrounding fields Graph 11: Distribution of the Roman finds collected by field walking units, by sectors (in 1000 sq meters, visibility and survey intensity corrected) 329 Graph 12: Chronological distribution of surface finds on “sites” 2 and 3 and in sector I Graph 13: Distribution of Roman finds by classes on “site 5a-5b” and in sectors I-III Graph 14: Distribution of the basic classes of Roman material collected by grid units Graph 15: Distribution of the basic classes of Roman material on grids 12, 13-14 and 15 Graph 16: Distribution of finds by period at the foot of site 9 Graph 17: Distribution of the Roman finds by classes from sector IX Graph 18: Distribution of finds by period in sector X Graph 19: Distribution of the Roman finds by basic classes on sites 13a, 13b and on field blocks 152-155 Graph 20: Chronological composition of the transect collections from sector VIII and XI Graph 21: Distribution of finds by period from sector II Graph 22: Distribution of finds by period in the southern half of sector II Graph 23: Density of LO-EM finds collected by transects, by sector (per 1000 sq m) Graph 24: Distribution of Late Ottoman-Early Modern finds by basic functional classes Contents of Appendix III: Analysis of the survey results; the distribution of the total surface record in the second survey area App III.1: The field survey and the analyzing procedures App III.2: Overall distribution of the total surface record by field blocks App III.3: Other categories of surface finds App III.4: The regular grid survey results App III.5: Conclusion 1 5 14 18 43 List of tables in Appendix III Table 1: The field walking forms for individual transects used in the second survey area Table 2: The field walking forms for grid collections used in the second survey area Table 3: Comparison of the size and the artifact density in the two survey areas List of graphs in Appendix III Graph 1: Distribution of the field units across the basic density ranges in the survey areas Graph 2: Distribution of the field units across the density ranges recorded in the first survey area List of photos in Appendix III Photo 1: An isolated cross on the eastern survey limit Photo 2: A view of the “sacred grove” in the northeast corner of the survey Photo 3: The western wall of the fortification List of maps in Appendix III Map 1: Division of the survey area into field blocks Map 2: Ground visibility conditions in the second survey area 330 Map 3: Distribution of the overall surface record in the second survey area Map 4: Distribution of the overall surface record, density ranges and quantity Map 5: Distribution of the overall surface record on the northern half of the eastern ridge Map 6: Distribution of the overall surface record in the northwest survey section Map 6a: Distribution of the overall surface record in the southwest survey section Map 7: Distribution of the overall surface record by individual transect units Map 8: The southwest corner of the survey area, number of shards by individual field walking transects Map 9: The northeast corner of the survey area, number of shards by individual field walking transects Map 10: Distribution of the piles of modern rubbish Map 11: Plan of the fort in the northeast corner of the survey area Map 12: The location of the two forts in the wider study region Map 13: Distribution of the overall surface record on grids I-III, grid and transect units Map 14: Distribution of the overall surface record on grids I-III, quantity ranges Map 15: Weight distribution of the total collections from grids I-III Map 16: Distribution of the overall surface record on grids 4-11 and the surrounding fields Map 17: Distribution of the overall surface records on grids 4-11, quantity ranges Map 18: Weight distribution of the total collections from grids 4-11 Map 19: Distribution of the overall surface record on grids 12, 13 and 19 and on the surrounding fields Map 20: Distribution of the overall surface record on grids 12-13 and 19, quantity ranges Map 21: Weight distribution of the total collections from grids 12, 13 and 19 Map 22: Distribution of the overall surface record on grid 14 and the surrounding fields Map 23: Distribution of the overall surface record on grid 14, quantity ranges Map 24: Weight distribution of the total collections from grid 14 Map 25: Distribution of the overall surface record in the northeast and central parts of the survey area Map 26: Quantity of the overall surface record in the central north survey section, grid and transect Map 27: Distribution of the overall surface record, quantity ranges for grids 15-18 and 27 Map 28: Weight distribution of the total collections from grids 15-18 and 27 Map 29: Distribution of the overall surface record in the central and western survey sections, grid and transect Map 30: Distribution of the overall surface record on grids 19 and 20, quantity ranges Map 31: Weight distribution of the total collections from grids 19 and 20 Map 32: Distribution of the overall surface record on the western ridge Map 33: Distribution of the overall surface record on grids 21-24 and the surrounding fields Map 34: Weight distribution of the total collections from grids 21-24 Map 35: Distribution of the overall surface record on grid 26 and the surrounding fields Map 36: Weight distribution of the total collections from grid 26 Map 37: Distribution of the overall surface record on grid 25 and the surrounding fields Map 38: Weight distribution of the total collections from grid 25 Map 39: Distribution of the artifact clusters in the second survey area 331 Contents of Appendix IV: Surface material distribution by periods from survey area 2 App IV.1.1: The prehistoric assemblage App IV.1.2: The Roman Period (2nd-4th c AD?) App IV.1.2.1: Statistical and overall distribution App IV.1.2.2: Grids 1, 2 and 3 App IV.1.2.3: Grids 14, 11 and 5 App IV.1.2.4: The site on field blocks 47a-b, 49-50 and 66 App IV.1.2.5: Grids 27, 15-18 App IV.1.2.6: Grids 4, 6-8 App IV.1.2.7: Grids 9-10, 12-13 App IV.1.2.8: The site on field blocks 289a-291a App IV.1.2.9: Grids 19, 20 and 26 App IV.1.2.10: The off-site and the site halos App IV.1.3: The Late Roman Period App IV.1.4: The Middle and Late Byzantine Period App IV.1.5: The Late Byzantine-Early Ottoman Period (late 14th- early 18th 105 App IV.1.6: The Late Ottoman-Early Modern Period (18th-20th century 1 5 5 9 15 21 25 33 44 53 58 70 94 101 century) 112 List of tables in Appendix IV Table 1: Amount and distribution of the finds securely datable to a late prehistoric period Table 2: Amount and distribution of all finds possibly datable to later prehistory Table 3 Amount and distribution of the finds datable to the Roman Period Table 4: Mean district density and density on possible site locations in the northeast survey quarter, in 1000 sq meters Table 5: Mean overall and on-site density recorded by the total grid survey, in 100 sq meters Table 6: Mean overall density and density on field blocks covering potential site locations, in 1000 sq meters Table 7: Mean overall and on-site density recorded by the regular grid survey, in 100 sq meters Table 8: Mean district density and density on field blocks covering potential sites, in 1000 sq meters Table 9: Mean district and density on field blocks covering potential sites, in 1000 sq meters Table 10: Mean overall and average densities for parts of grids 15-18, in 100 sq meters Table 11: Mean district and density on field blocks covering possible site locations, in 1000 sq meters Table 12: Mean overall densities and densities recorded on grids 4, 6-8, in 100 sq meters Table 13: Mean district and densities on field blocks covering possible site locations, in 1000 sq meters Table 14: Mean district density and density on field units covering possible site locations, in 1000 sq meters 332 Table 15: Mean overall density and mean density on grids 9, 10, 12 and 13, in 100 sq meters Table 16: Mean district density and density on field blocks covered by grids 9, 12 and 13, in 1000 sq meters Table 17: Mean district and density on field blocks 289a-291 and the immediate surroundings, in 1000 sq meters Table 18: Mean district density and density on field blocks 270-271, 277b and 394, in 1000 sq meters Table 19: Mean overall density and mean density on grids 19, 20 and 26, in 100 sq meters Table 20: Mean district density and density recorded on field blocks covering site 26 and its immediate surroundings, in 1000 sq meters Table 21: Mean district density and density recorded on field blocks covering the outer territory of the site on grid 26, in 1000 sq meters Table 21: Mean district density and density on field blocks in the western survey half in 1000 sq meters Table 22: Mean district density and average density recorded on grids in the western survey half, in 100 sq meters Table 23: Amount and distribution of Late Roman finds Table 24: Mean district density and density on individual field blocks in the northwest survey section, in 1000 sq meters Table 25: Amount and distribution of finds datable to the Middle and Late Byzantine Period Table 26: Mean district and density on field blocks featuring finds broadly dated to the Middle Age, in 1000 sq meters Table 27: Amount and distribution of finds datable to the Late Byzantine-Early Ottoman Period Table 28: Amount and distribution of finds datable to the Late Ottoman-Early Modern Period List of graphs in Appendix IV Graph 1: Chronological composition of the finds on grids 1-3 Graph 2: Distribution of Roman material by classes on grids 1-3 Graph 3: Distribution of finds by periods on grids 14, 5 and 11 Graph 4: Distribution of Roman pottery by basic categories, grids 14, 5 and 11 Graph 5: Chronological profile of the individual transect collections from field blocks 47a-b, 49, 50 and 66 Graph 6: Distribution of the collections by basic functional categories Graph 7: Distribution of finds by periods from grids 27, 18, 15-17 Graph 8: Composition of the Roman assemblages from grids 27, 18, 15-17 Graph 9: Distribution of finds by period on grids 4, 6-8 Graph 10: Composition of the Roman assemblages from grids 4, 6-8 Graph 11: Distribution of finds by periods on grids 9, 10, 12 and 13 Graph 12: Composition of the Roman assemblages from grids 9-10, 12-13 333 Graph 13: Distribution of finds by periods from field blocks 289-291a, 294, 303b, 295 and 296 Graph 14: Composition of the Roman assemblages by basic functional categories Graph 15: Distribution of finds by period on grids 19, 20 and 26 Graph 16: Composition of the Roman assemblages from grids 19, 20 and 26 Graph 17: Chronological composition of the material from the western ridge Graph 18: Composition of transect collections of Roman material from the eastern ridge and of the total Roman collection from the western ridge Graph 19: Average densities of Roman material in the off-site recorded by grid areas in 1000 sq m Graph 20: The extent of the three basic density zones in the second survey area in percentage Graph 21: The extent of the three basic density zones according to the record of the adjusted transect collections, in percentage Graph 22: Average density of Late Roman finds by gridded areas (in 1000 sq m) Graph 23: Distribution by periods of the material collected from grid 25 and the surrounding fields Graph 24: Average density of LB-EO finds by gridded areas (in 1000 sq meters) Graph 25: Average density of LO-EM finds across different survey sections, in 1000 sq meters Graph 26: Distribution by periods of the total collections from grids 21, 22 and 23-24 Graph 27: Average densities of LO-EM finds by grids (in 1000 sq meters) Abbreviations: MN, M. Neo = Middle Neoltihic LN, L. Neo = Late Neolithic BA = Bronze Age LBA = Late Bronze Age LBA-EIA = Late Bronze Age-Early Iron Age EIA = Early Iron Age LIA = Late Iron Age Hl, Hell = Hellenistic Period R-LR, R-LRom = Roman-Late Roman LB-EO = Late Byzantine-Early Ottoman LO-EM = Late Ottoman-Early Modern 334 Curriculum Vitae Damjan Donev was born on the 29th of June 1979, in Skopje. After studying for three years at the Art History and Archaeology department at the State University of Sts. Cyril and Methodius in Skopje, he completed his undergraduate studies at Davis and Elkins College, Elkins, WV, earning a degree in Cultural Anthropology. He completed his Master’s degree at the department of Archaeology and Art History, at Bilkent University, Ankara. Over the past decade and a half Damjan Donev has participated in excavation and survey projects in the Republic of Macedonia and Greece. His geographic focus is on the Balkan interior, while his research interest includes landscape studies, with a special emphasis on territoriality and settlement patterns, archaeological method and theory and sepulchral archaeology. emails: damjanec@yahoo.com; damjaned@gmail.com 335 Abstract The prevalent type of archaeological field surveys in the Republic of Macedonia has ensured that only a smaller percentage of the archaeological sites visible on the surface were included in the archaeological map of the country. Up until now there were no systematically gathered field data pertaining to the size, positioning and intra-site organization of nonarchitectural surface remains. The two small-scale and hyper-intensive surface artifact surveys presented in this study were the first glimpse of the type and distribution of settlement on a parish level and in a rural context, in the regions along the Vardar Valley. Not attempting to offer a representative coverage of the region as a whole or of certain types of micro-geographic entities, the surveys were rather concentrated on 1) reconstructing the long-term history of individual settlements (by means of highly intensive and systematic survey coverage and careful study of the ceramic fabrics); 2) understanding the integral set of habitation practices (by adopting a site-less approach in the interpretation of the surface artifact scatters) and 3) exploring the type of micro-topographic elements preferred by the local farming communities (the concept of settlement niche). The study and interpretation of the field data faced us with the problem of understanding the settlement dynamic on a micro-level, but it also brought up a series of interpretative and methodological problems inherent to all studies of surface archaeological material. 336 Samenvatting Het overwegend gebruikte type van archeologische veldonderzoek in de Republiek van Macedonië heeft er voor gezorgd dat een klein percentage van de aan de oppervlakte zichtbare archeologische sites opgenomen werden in de archeologische kaart van het land. Tot op heden waren er geen systematisch verzamelde data betreffende grootte, positie en intra-site organisatie van de niet-architectonische prospectie vondsten. De twee kleinschalige en hyper-intensieve prospectie veldonderzoeken naar artefacten gepresenteerd in dit onderzoek zijn een eerste blik op het type en de verspreiding van nederzettingen op district-niveau en binnen de plattelands context, in de regio van de Vardar Vallei. The onderzoeken pogen niet een representatieve dekking van de gehele regio of van bepaalde typen van micro-geografische eenheden te geven, maar zijn eerder geconcentreerd op 1) het reconstrueren van de lange term geschiedenis van de individuele nederzettingen (door middel van intensieve en systematische veldonderzoek dekkingsgraad en nauwgezet onderzoek naar de ceramische materialen); 2) het begrijpen van de integrale set van bewoningspatronen (door middel van een site-less aanpak bij de interpretatie van de spreidingspatronen van de oppervlakte artefacten) en 3) het onderzoeken van de voorkeuren voor micro-topografische elementen door de lokale plattelandsgemeenschappen (het concept van settlement-niche). Het onderzoek en de interpretatie van de veld data presenteerde ons het probleem van het begrijpen van de dynamiek van nederzettingen op het microniveau, maar het bracht ook een reeks van interpretatieve en methodologische problemen behorend tot alle onderzoeken naar materiaal gevonden via prospectie archeologie. 337