(Studies in Generative Grammar 83) Katalin E. Kiss-Universal Grammar in The Reconstruction of Ancient Languages-Mouton de Gruyter (2005)
(Studies in Generative Grammar 83) Katalin E. Kiss-Universal Grammar in The Reconstruction of Ancient Languages-Mouton de Gruyter (2005)
(Studies in Generative Grammar 83) Katalin E. Kiss-Universal Grammar in The Reconstruction of Ancient Languages-Mouton de Gruyter (2005)
WDE
G
Studies in Generative Grammar 83
Editors
Henk van Riemsdijk
Jan Köster
Harry van der Hulst
Mouton de Gruyter
Berlin · New York
Universal Grammar in
the Reconstruction
of Ancient Languages
edited by
r
Katalin E. Kiss
Mouton de Gruyter
Berlin · New York
Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter G m b H & Co. KG, Berlin.
ISBN-13: 978-3-11-018550-8
ISBN-10: 3-11-018550-4
Introduction 1
Katalin E. Kiss
Some firm points on Latin word order: The left periphery 429
Giampaolo Salvi
Index 519
Katalin Ε. Kiss
2. Predecessors
The view that ancient languages can be modelled successfully in the gen-
erative framework despite the lack of native speakers has been shared by a
number of linguists ever since the ninteen sixties. Some scholars assumed
that the lack of native speakers can be made up for by a corpus of adequate
size and variation (Ehlich 1981), while others claimed that they had devel-
oped a native-like intuition in the language studied by them (Pillinger 1980).
4 Katalin Ε. Kiss
It has been mainly Latin whose grammatical phenomena have been sub-
jected to generative analyses. The first major generative study of Latin
syntax was Robin L a k o f f s book on Latin complementation (1968), dis-
cussing various types of finite and non-finite complement clauses, and the
licensing of various moods. In the seventies and early eighties, the period
of the emergence of Government and Binding Theory, problems of Latin
infinitival phrases aroused interest, which has persisted up until today.
Pepicello (1977) analyzed accusative with infinitive as Subject-to-Object
Raising, and his views elicited alternative proposals by Bolkenstein (1979),
Pillinger (1980), and Wales (1982). Questions of accusative with infinitive
constructions, for example, the source of the accusative case, kept recurring
also in the proceedings of a series of International Colloquia on Latin Lin-
guistics, edited by Pinkster (1983), Touratier (1985), Calboli (1989), and
Herman (1994). Maraldi (1983), for example, identifyied Acl as Excep-
tional Case Marking, licensed by S' deletion; however, she also noticed
instances of Acl occurring in the context of a passive matrix verb. The
same problem was also addressed by Calboli (1983, 1989), and recently by
Cecchetto and Oniga (2002).
Areas of Latin syntax examined in the generative framework include the
case system - see Binkert (1970) and Murru (1977). Maurel (1983, 1989)
discussed problems of Latin relativization. Bertocchi and Casadio (1983),
and Bertocchi (1989) dealt with questions of binding, particularly with the
Latin se and suus, which are anaphors that can be bound not only locally
but also at a long distance. Salvi (1996) and Giusti (2001) examined the
Latin pronominal system from a historical perspective, as the predecessor
of the Romance systems of strong, weak, and clitic pronouns and the Ro-
mance article. Basic questions of Latin syntax concerning sentence struc-
ture, word order variation, its derivation, and its discourse functions were
addressed by Ostafin (1986), Salvi (1999/forthcoming), and Polo (2003).
Benucci (1996) analyzed Umber, a close relative of Latin. Recently issues
of Latin morphosyntax have gained significance, particularly in the frame-
work of Distributed Morphology - see Embick's derivation of the synthetic
and analytic forms of the perfect (2000), and Embick and Halle's analysis
of the Latin conjugation (1999). Oniga (2004) provides a generative analy-
sis of the major morphological and syntactic structures of Latin.
Studies of Ancient Greek in the generative literature had been sporadic
up until recently. The analysis of agreeing predicative adjectives in non-finite
clauses by Andrews (1971) provided important evidence for the presence of
a covert subject in infinitives, and contributed to the shaping of control and
Introduction 5
3.1. Accusative with infinitive without ECM and without object control
The fact that the possessor can appear either as a determiner of the posses-
sion, or as an extracted complement, or an adjunct external to the projection
of the possession is well-known from a number of languages. Since the
case-ending or the preposition of an external possessor is often different
from that of an internal one, the recognition of an external possession con-
struction can be a very difficult descriptive problem. External possessors
also raise theoretical questions such as what triggers the external ization of a
possessor, and how an external possessor construction is derived.
Two papers of the volume are devoted to problems of the possessive con-
struction. Gabor Zolyomi's paper paper entitled Left-dislocated possessors
in Sumerian discusses two different Sumerian constructions with a topical-
ized external possessor. In one of them, called the anticipatory genitive
construction, the left-dislocated possessor bears the genitive case, and is
coindexed with a possessive pronoun internal to the projection of the pos-
session. The possessor and the possessive pronoun coindexed with it need
not be subjacent; for instance, in one of the examples discussed, the posses-
sor in the left periphery of the matrix clause is coindexed with a possessive
pronoun in a relative clause. The lack of subjacency and the resumptive
pronoun strategy are interpreted as evidence that the anticipatory genitive
construction is a base-generated construction, involving no movement.
In the other possessive construction with a left-dislocated possessor, the
so-called external possession construction, the left-dislocated possessor,
though copied by a possessive pronoun in the projection of the possession,
is not in the genitive case but bears the same syntactic case as the posses-
sion itself. This construction can be used in the case of an inalienable pos-
session. Interestingly, the verb agrees with the external possessor instead of
the head of the possessive construction.
10 Katalin Ε. Kiss
(3)
represent either a weak or a strong form. The ea in (4) must be strong, be-
cause it supports the enclitic vero.
In view of these, the structure in (4) does not contradict the hypothe-
sized sentence structure any more; the pronoun doubling the topic constitu-
ent is a strong pronoun occupying Spec,IP, hence it is not subject to the
constraints on weak pronoun placement.
The left periphery of embedded clauses is somewhat more complex than
the [ΤΟΡΡ· · · [ F P - · · [IP [VP··· structure identified in main sentences; an embed-
ded clause also contains a clause-initial relative WH-element, or a pre- or
post-topic complementizer. In order to provide place for these constituents,
Salvi supplements his left periphery-model along the lines proposed by
Rizzi (1997). In Rizzi's theory, the complementizer domain of the sentence
has three layers, to be occupied by a relative WH-element, a topic, and a
focus, respectively. The complementizer can appear in the head position of
any of these projections. (In fact, in Latin it has to be ensured that the com-
plementizer precede the filler of Spec,FP, i.e., it occupy the head of the
projection harboring the relative WH-phrase or that harboring the topic.) In
embedded clauses the complementizer also provides an additional target for
weak pronoun placement.
Lisa Lai-Shen Cheng, Aniko Liptäk, and Chris Reintges's paper entitled
The nominal cleft construction in Coptic Egyptian analyzes a focussing
device of Coptic which is also known from many present-day languages of
the world. Its discourse function being clear, the investigation aims to es-
tablish how the parameters involved in the formation of the construction are
to be set so that the restrictions attested fall out. These are the following :
(i) The cleft constituent is a DP in the left periphery of the sentence, (ii)
The backgrounded proposition is represented by a relative clause whose
relative pronominal element is coindexed with the cleft constituent. In sub-
ject relatives the relative pronoun is empty, otherwise it is a resumptive
pronoun in situ, (iii) The element linking the cleft constituent and the back-
grounded proposition is a deictic pronoun agreeing in number and gender
with the cleft constituent.
The authors claim that the Coptic nominal cleft construction is to be
derived from a small clause, with the cleft constituent functioning as its
subject, and the relative clause functioning as its predicate. This explains
why the cleft constituent cannot be anything but a DP. The relative clause
has a relative operator in situ, represented by a resumptive pronoun. The
authors adopt an analysis in which the pronominal operator undergoes
movement to Spec,CP in narrow syntax; at PF, however, the lower copy of
16 Katalin Ε. Kiss
Two papers of the volume, The correlation between word order alterna-
tions, grammatical agreement and event semantics in Older Egyptian by
Chris H. Reintges, and VSO and Left-conjunct Agreement: Biblical Hebrew
versus Modern Hebrew by Edit Doron, deal with VSO-SVO word order
variations. Both of the studies derive the variation from a lexical idiosyn-
cracy: the lack or presence of an EPP feature on a functional category ex-
tending the verb phrase. Both of them also relate the variation to subject-
verb agreement, linking VSO to the lack of agreement, and SOV to the
presence of it. However, whereas in Reintges's theory, it is agreement that
determines SVO order (SVO being a consequence of the EPP feature of
Agreement), in Doron's approach it is the other way round; it is the SVO
order (triggered by an optional EPP feature of Tense) that elicits agreement.
Reintges argues that in Older Egyptian, the VSO-SVO word order variation
is not random but is linked to the type of the eventuality; sentences describ-
ing an event have a VSO word order, whereas those expressing a state are
SVO. This distribution of preverbal and postverbal subjects might recall a
theory elaborated by Kratzer (1995), Maleczki (1999), and others, accord-
ing to which sentences with an event variable can have a spatiotemporal
expression (whether spelled out or unarticulated) in their external argument
position, which allows the subject to remain in the VP. Sentences describ-
ing a state, which lack an event variable, on the other hand, can only have
Introduction 17
movement, and subject movement to Spec,AgrSP. That is, the word order
difference between eventive and stative sentences is eventually reduced to
the presence of an AgrP projection with an EPP feature in the latter.
Edit Doron demonstrates in her paper that the seemingly identical VSO
sentences of Biblical Hebrew and Modern Hebrew represent different struc-
tures; Biblical Hebrew sentences have the structure in (5a), whereas Mod-
ern Hebrew sentences display the structure in (5b). This difference is de-
rived from the assumption that in Biblical Hebrew, the Tense head does not
have an intrinsic EPP feature (although it can be supplied with an EPP fea-
ture optionally, which yields an SVO sentence). In Modern Hebrew, on the
other hand, EPP is a lexical property of Tense.
In Biblical Hebrew, the lack of the attraction of the subject to Spec,TP
also affects the Agree relation between the subject and the verb. This be-
comes evident in the case of a conjoined subject of the following type:
If Τ has an EPP feature (as is always the case in Modern Hebrew and can
also happen in Biblical Hebrew), it is the whole conjoined noun phrase (the
highlighted DP) that is subjected to movement to Spec,TP and agreement
with V+T - owing to constraints on movement. If, on the other hand, Τ
does not have an EPP feature (as is usually the case in Biblical Hebrew), Τ
can agree with the minimal D constituent closest to it which allows the
derivation to converge. That is, the VSO sentences of Biblical Hebrew dis-
play left-conjunct agreement; for example, in the equivalent of 'And lifted
David and the people with him their voice', the verb is in the singular.
In a few V S O sentences of the Bible, we attest full agreement between
the verb and a conjoined subject noun phrase following it. As Doron demon-
strates, these sentences all represent structure (5b); the verb has been raised
to a functional projection above TP, while the postverbal subject still occu-
pies Spec,TP.
Introduction 19
Surprisingly, the conjunction may also appear cliticized after the first con-
junct - e.g. in the following Vedic Sanskrit example:
This example seemingly contradicts not only the Latin examples cited above,
but also what we attest in present-day Indo-European languages. A con-
junction cliticized to the first conjunct is sharply ungrammatical - as illus-
trated by the English example *He left and. He didn't even say good-bye
(cf. Munn 1992).
On the basis of this seemingly contradictory array of facts, a traditional,
inductive description cannot but conclude that anything goes; the conjunc-
tion^) can stand anywhere in the coordinated phrase except in initial posi-
tion - which is not true. (For example, no 'XP and XP XP' order is attested.)
If, on the other hand, the binary branching conjunction theory following
from the principles of Universal Grammar, formulated in somewhat differ-
ent versions by Munn (1992) and Kayne (1994), is used as the framework
of the analysis, all the facts attested can be derived, and no unattested pos-
sibilities are predicted. In this framework, coordinated phrases are assigned
the following universal structure:
20 Katalin Ε. Kiss
(9) &P
& YP
(10) &2P
&1 XP &2 YP
X Y
In this structure both X and Y undergo head movement to &1 and &2, re-
spectively - unless e.g. &2 is not enclitic, as happens in (7). In this frame-
work, the source of the seemingly idiosyncratic example in (8) is a version
of structure (10), in which the 2nd conjunction (&2) is phonetically null.
The benefits of deriving the coordination possibilities of ancient Indo-
European languages from the binary branching hypotheses of Universal
Grammar are mutual: not only seemingly idiosyncratic facts of a number of
languages receive a principled explanation, but also a speculative hypothe-
sis of Universal Grammar receives empirical support; ancient Indo-
European languages realize possibilities that are derivable from the Munn-
Kayne theory but have not been attested so far.
must be that the N u m head involved in D-stems can only merge with a vP.
Thus the interpretation of the D-stem of the Akkadian verb as a N u m P pro-
jection subsuming a vP accounts for both the seemingly idiosyncratic effect
of the D-stem on argument structure, and the sometimes delicate difference
between the interpretations of the G-stem and D-stem of one and the same
transitive or unergative verb. Huber's analyses can also be extended to
similar facts of related languages. For example, the corresponding stems of
Arabic and Hebrew verbs can have an 'intensive' interpretation. Huber
analyzes such sentences (meaning, for example, 'x bit y (fiercely or repeat-
edly)') as expressing cumulative affectedness, with either a series of transi-
tions predicated of the same entity, or the output of one transition serving
as input for the next transition.
The problem that Bartolotta aims to analyze is the following: even though
the Greek perfect (w)oida Ί see (with the mind's eye)' and the aorist
(w)iidon Ί saw' are verb forms derived from the same root *weid- 'to see',
associated with the same theta-grid consisting of an experiencer and a the-
me, they select different cases for their theme argument. The aorist form
takes an accusative object, whereas the perfect form occurs both with an ac-
cusative and a genitive object. The genitive appears to be the older variant; in
Iliad there are 23 occurrences of it, with only three in Odyssey. Bartolotta
wants to answer two questions: why one and the same verb had its theme
argument marked with different cases in different tenses at one stage of the
language; and why the genitive marking of the theme of the perfect verb
form eventually disappeared.
As for the first question, Bartolotta demonstrates that the Indo-European
*weid- stem underwent a meaning-split; perceptive vision, i.e., simple eye-
sight, came to be differentiated from intentional vision, i.e., internal acqui-
sition, thinking. The different tense forms were likely to elicit one or the
other of the two meanings of the verb; the [-stative] aorist form elicited the
'perceptive sight' interpretation, whereas the [+stative] perfect form evoked
the 'intentional vision' meaning.
In the early Greek represented by Iliad, the aorist (w)eidon, denoting per-
ception, always takes an accusative object because in that language variant
verbs expressing immediate contact with their object take an accusative
object. The object of the perfect (w)oida, meaning 'see with the mind', on the
other hand, is in the genitive because in that period Greek verbs denoting
mediated contact with their object take a genitive object.
The process in the course of which the genetive marking of the object of
(w)oida 'see with the mind' gradually gave way to accusative marking is
claimed by Bartolotta to be a manifestation of Greek developing from the
active-stative system of Pre-Indo-European to a nominative-accusative sys-
tem. This typology, developed in Indo-European linguistics (cf. Lehmann
1993), is recast by Bartolotta in a generative terminology. In her formula-
tion, the change from the active-stative system to the nominative-accusative
system brought about the following major changes: The semantic-lexical
relationship between the verb and its complements typical of the active-
stative type, with the N P matching the verb in its [+/-stative] feature, be-
came a structural relationship of the nominative-accusative type. Inherent
case assignment coupled by theta-role assignment gave way to structural
case assignment with no theta-role assignment involved. In the active-
stative system, the object is assigned a theta-role and case in situ; in the
24 Katalin Ε. Kiss
Acknowledgments
References
Andrews, Avery D.
1971 Case agreement of predicate modifiers in Ancient Greek. Linguistic
Inquiry 2: 127-151.
Benucci, F.
1996 Studi di sintassi umbra. II verbo nelle Tavole Iguvine e nelle iscrizioni
minori. Padova: Libraria Padovana.
Bertocchi, A.
1989 The role of antecedents of Latin anaphors. In Subordination and
Other Topics in Latin, Gualtiero Calboli (ed.), 441^462. Amsterdam:
John Benjamins.
Bertocchi, A. and C. Casadio
1983 Anaphoric relations, pronouns and Latin complementation. In Latin
Linguistics and Linguistic Theory, Harm Pinkster (ed.), 27-40. Am-
sterdam: John Benjamins.
Introduction 25
Binkert, P. J.
1970 Case and prepositional constructions in a transformational grammar
of Classical Latin. [Unpublished Ph.D. diss., University of Michigan.]
Bolkenstein, Machtelt
1979 Subject-to-object raising in Latin, Lingua 49: 15-34.
Bresnan, Joan
1982 Control and complementation. Linguistic Inquiry 13: 343-433.
Calboli, Gualtiero
1983 The development of Latin (Cases and infinitive). In Latin Linguistics
and Linguistic Theory, Harm Pinkster (ed.), 42-58. Amsterdam: John
Benjamins.
Calboli, Gualtiero (ed.)
1989 Subordination and Other Topics in Latin. Amsterdam: John Benjamins.
Cardinaletti, Anna and Michal Starke
1999 The typology of structural deficiency: a case study of three classes of
pronouns. In Clitics in the languages of Europe, Henk van Riemsdijk
(ed.), 145-233. Berlin: Mouton de Gruyter.
Cecchetto, Carlo and Renato Oniga
2002 Consequences of the analysis of Latin infinitival clauses for the
theory of case and control. Lingue e linguaggio 1: 151-189.
DeCaen, Vincent J. J.
1995 On the placement and interpretation of the verb in standard Biblical
Hebrew prose. [Unpublished Ph.D. diss., University of Toronto.]
Doron, Edit
2000 V S O and left-conjunct agreement: Biblical Hebrew vs. Modern He-
brew. In The syntax of verb-initial languages, Andrew Carnie and
Eithne Guilfoyle (eds.), 75-96. Oxford: Oxford University Press.
Eh 1 ich, Κ.
1981 Native Speaker's heritage. On philology of 'dead' languages. In A
Festschrift for Native Speaker, F. Coulmas (ed.), 153-165. The Hague:
Mouton.
Embick, David
2000 Features, syntax, and categories in the Latin perfect. Linguistic In-
quiry 31: 185-230.
Embick, David and Morris Halle
1999 The Latin conjugation. [Unpublished Ms., MIT.]
Garrett, A.
1990 The Syntax of Anatolian Pronominal Clitics. [Unpublished Ph.D.
dissertation, Harvard University.]
Giusti, Giuliana
2001 The birth of a functional category: From Latin ILLE to the Romance
article and personal pronoun. In Current studies in Italian syntax,
Essays offered to Lorenzo Renzi, Guglielmo Cinque and Giampaolo
Salvi (eds.), 157-171. Amsterdam: Elsevier.
26 Katalin Ε. Kiss
Gragg, Gene B.
1968 The syntax of the copula in Sumerian, in: J.W.M. Verhaar (ed.), The
verb 'be' and its synonyms, III (Foundations of Language Supple-
mentary Series 8). Dordrecht, 86-196.
1973 Linguistics, method, and extinct languages: The case of Sumerian.
Orientalia N.S. 42: 78-96.
Groneberg, Brigitte R. Μ.
1987 Syntax, Morphologie und Stil der jungbabylonischen hymnischen
Literatur, I—II (Freiburger altorientalische Studien 14. I—II.) Stuttgart:
Franz Steiner Verlag.
Hale, Kenneth
1983 Warlpiri and the grammar of non-configurational languages. Natural
Language and Linguistic Theory 1: 5 - 4 7 .
Hale, Mark
1987 Studies in the comparative syntax of the oldest Indo-Iranian lan-
guages. [Unpublished Ph.D. dissertation, Harvard University.]
1990 Preliminaries to the study of the relationship between syntax and
sandhi in Rigvedic Sanskrit. Münchener Studien zur Sprachwissen-
schaft 51: 77-96.
1991 Some observations on intersentential pronominalization in the language
of the Taittirlya Samhitä. In Sense and Syntax in Vedic, J. Brereton
and S. Jamison (eds.). Leiden: Brill.
1993 Tmesis and movement in Avestan. Indo-Iranian Journal 36: 29^13.
1996 Clitic placement in the language of the Rigveda. In Approaching
Second. Second Position Clitics and Related Phenomena, Aaron L.
Halpern and Arnold M. Zwicky (eds.), 165-197. Stanford: CSLI
Publications.
Herman, Jözsef (ed.)
1994 Linguistic Studies on Latin. Selected papers from the 6th Interna-
tional Colloquium on Latin Linguistics, Budapest, 23-27 March,
1991. Amsterdam: John Benjamins.
Holmstedt, Robert
2002 The relative clause in Biblical Hebrew. [Ph.D. dissertation, University
of Wisconsin at Madison. Forthcoming as The relative clause in Bib-
lical Hebrew. Linguistic Studies in Ancient West Semitic. Winona
Lake, Ind.: Eisenbrauns.]
Huber, Christian
1989/90 Ergative Sprachen in der GB-Theorie und einige Überlegungen zum
Sumerischen. Wiener Linguistische Gazette 4 5 - 4 6 : 53-78.
1996 Some notes on transitivity. Verb types, and case with pronouns in
Sumerian. Wiener Zeitschrift für die Kunde des Morgenlandes 86
(.Festschrift Hirsch)·. 177-189.
Introduction 27
Maurel, J.-P.
1989 Subordination seconde du relatif en latin et theorie du 'COMP'. In
Subordination and Other Topics in Latin, Gualtiero Calboli (ed.),
181-196. Amsterdam: John Benjamins.
Morrell, K. S.
1989 Studies on the phrase structure of early Attic prose. [Unpublished
Ph.D. dissertation, Harvard University.]
Munn, Alan
1992 A null operator analysis of ATB gaps. The Linguistic Review 9: 1-26.
Murru, F.
1977 Studio di un modello semantico-generativo per l'insegnamento del
latino. II problemo dei casi. [Unpublished Ph.D. dissertation, Univer-
sity of Torino.]
Oniga, Renato
2004 Breve introduzione linguistica. Milano: Franco Angeli.
Ostafin, David Mark
1986 Studies in Latin word order: A transformational approach. [Unpub-
lished Ph.D. dissertation, University of Connecticut at Storrs.]
Pepicello, William
1977 Raising in Latin. Lingua 42: 209-218.
Philippaki-Warburton, I. - G. Catsimali
1997 Control in Ancient Greek. Greek linguistics '95. In Proceedings of
the 2nd International Conference on Greek Linguistics, Salzburg,
September 1995, G. Drachman, A. Malikouti-Drachman, J. Fykias
and C. Klidi (eds.), 577-588. Graz: W. Neugebauer Verlag.
Pillinger, Ο. Stephen
1980 The accusative and infinitive in Latin: A refractory complement
clause. Journal of Linguistics 16: 55-83.
Pinkster, Harm (ed.)
1983 Latin Linguistics and Linguistic Theory. Amsterdam: John Benjamins.
Polo, Chiara
2003 Latin, Italian and Slovene between Morphology and Syntax. [Ph.D.
dissertation, University of Padua, to be published by Unipress in
2004.]
Pustejovsky, J.
1991 The syntax of event structure. Cognition 41: 47-81.
Reintges, Chris H.
1996 Egyptian Root-and-Pattern Morphology. Proceedings of the Interna-
tional Conference on Egyptian Grammar. Crossroads 111, Yale 1994.
Lingua Aegyptia 4: 213-244.
1997 Passive voice in Older Egyptian. (HIL Dissertations 28.) The Hague:
Holland Academic Graphics.
Introduction 29
Taylor, Ann
1994 The change from OV to VO in Ancient Greek. Language Variation
and Change 6: 1-37.
1996 A prosodic account of clitic position in Ancient Greek. In Approach-
ing Second. Second Position Clitics and Related Phenomena, Aaron
L. Halpern and Arnold M. Zwicky (eds.), 477-503. Stanford: CSLI
Publications.
Touratier, Christian (ed.)
1985 Syntaxe et latin, Actes du lime Congres International deLinguistique
Latine, Aix-en-Provence, 28-31 Mars 1983. Aix-en-Provence, Mar-
seille: Universite de Provence/J. Lafitte.
Wales, M. L.
1982 Another look at the Latin accusative and infinitive. Lingua 56: 127-
152.
Zolyomi, Gabor
1996 Genitive constructions in Sumerian. Journal of Cuneiform Studies
48: 39^15.
The correlation between word order alternations,
grammatical agreement and event semantics in
Older Egyptian
Chris H. Reintges
1. Introduction
b. The S V O alternative
n-ntt N N pn rx(-w) rn n(j) wßi'-w (...)
since N N DEM:SM learn-3MSTAT name LINK(-SM) fowler-PM
'Since this N N (the male deceased) knows (by learning) the names
of the fowlers ( . . . ) . ' (Coffin Texts VI 2 2 o / B l B o )
The contrast between verb-initial and subject-initial word order does not
reflect discourse-configurationality, where the variable position of the sub-
ject is related to its topic or focus role. Rather, V S O and SVO structures are
associated with different aspectual viewpoints from which a given situation
is presented. The V S O clause in ( l a ) above has an event-related interpreta-
tion, describing the acquisition of some knowledge, while the SVO 'alter-
native' in ( l b ) above has a state-related interpretation, describing the pos-
session of some knowledge through learning. Apart from word order, Older
Egyptian employs two morphologically distinct finite verb conjugations,
the Eventive and the Stative, to formally distinguish event- and state-
denoting verbs that are derived from the same root. The Eventive-Stative
alternation that yields minimal pairs like j.r% 'learn about' and rx(-w)
'know (through learning)' is fully productive in various lexical classes of
transitive, unergative and unaccusative verbs.
Older Egyptian thus represents the typologically marked case of a lan-
guage where a stative-resultative verb form cannot be derivationally related
to a non-stative base form, but where the members of the opposition, stative
and eventive, are encoded by different types of inflectional paradigms
(Nedjalkov & Jaxontov 1988: 29). The aim of this study is to clarify the
complex relation between word order alternations, subject-verb agreement,
and event semantics. I will argue that the aspectual properties of verbs are
not specified at a lexical, but rather at a syntactic level. To assume an
event- or state-related interpretation, the subject and the verb must appear
in a particular hierarchical relation with one another. On the other hand, the
eventive or the stative interpretation of the main verb has a morphological
correlate in the finite verb inflection. I will show that grammatical agree-
ment in the traditional sense is only represented by the Stative verbal para-
digm, while the Eventive conjugation lacks agreement proper. In the
Stative, the presence of agreement excludes tense- and aspect morphology,
while the corresponding Eventive is compatible with the full range of Older
Egyptian tense-aspect-mood and voice marking. I will also provide a con-
figurational analysis of Eventive V S O and Stative S V O sentences. What I
want to show in particular is that V S O order does not correspond to a single
Word order correlations in Older Egyptian 33
parentheses) are therefore not part of the internal temporal structure of the
state eventuality itself (Smith 1991: 32 (30); Rothstein 2004: 14-6). (I and
F represent the initial and final end points of the relevant eventuality.)
(I) (F)
(I) (F)
This section examines the interaction between lexical semantics and aspectu-
al viewpoint in the derivation of verbal meaning. The following description
of the Eventive-Stative opposition is based on a simple taxonomy of verbal
classes, in which valency information is considered apart from argument
meanings. Following Grimshaw (1990) and related research, argument struc-
ture itself does not encode thematic roles like AGENT, PATIENT or THEME,
since it only represents the argument-licensing capacity of a predicate
without further specifying any semantic information about its arguments,
except for their relative prominence. The Eventive-Stative alternation in
Older Egyptian shows quite clearly that the eventuality of the verbal and
the thematic content of its arguments are not part of its lexical-semantic
frame, but are primarily determined by the morpho-syntax. The systematic
differences in meaning between Eventive-inflected and Stative-inflected
verb forms will be examined for different lexical classes of transitive-
active, unergative and unaccusative verbs.
(2) DP OBJECTS
a. Eventive verb form
jw rx-n(-j) fik? nb ft? n(-j) gnw
AUX learn-PERF-lS EV magic every secret LINK(-SM) residence
Ί learned about every secret magic of the residence.'
(Urkunden I 143: 2)
36 Chris Η. Reintges
(3) CP COMPLEMENTS
a. Eventive verb form
j.mr-n(-j) [CP nd5-k jrt-k m-ΐ jr
wish-PERF-1 Sev save-2SM E v eye-2SM f r o m - a r m make(-PTCP:SM)
r-k ]
against-2SM
Ί have come to wish (that) you save your eye from the one who
acts against you.' (Ancient Hymn Da/12)
b. Stative verb form
j.mr-k(j) [cp nd'-k jrt-k m-ΐ jr
wish-lSsiAT save-2SM E V eye-2SM f r o m - a r m make(-PTCP:SM)
n-k ]
for-2SM
Ί have the wish to save your eye from the one who acts for you.'
(Coffin Texts VI 220j/L2Li)
The exact classification of verbs like sit, stand and lie has raised some con-
troversy in the literature. Hoekstra & Mulder (1990) include such verbs of
spatial configuration in the class of (unaccusative) copular verbs, although
members of this class in Dutch appear in constructions displaying unergative
syntax. To make sense of the variable unergative-unaccusative behaviour,
Levin & Rappaport-Hovav (1995: 126-130) propose to distinguish between
three types of non-causative meanings associated with a specific spatial
configuration. The first two non-causative meanings are the 'maintain posi-
tion' and the 'assume position' sense, both of which are agentive. The third
meaning is non-agentive and describes the position of the subject with re-
spect to a particular location. In Older Egyptian verbs of spatial configura-
tion, the agentive 'maintain position' and the 'assume position' sense are
both encoded by the Eventive conjugation pattern, as shown in (7a), and the
non-agentive 'simple position' sense by the corresponding Stative, as
shown in (7b) below.
Verbs of inherently directed motion like j j 'to go' and jw 'to come' incor-
porate into their semantics a deictic orientation towards the speaker. The
location argument does not have to be overt, but can be semantically im-
plied, as seen in (9a-b) below.
b. Stative form
m?-n wj nw xpr-kw
see- perFev me Nun come.into.existence-lSSTAT
'(The god) N u n saw me (already) existing.'
(Coffin Texts I 3 3 4 a / B 2 L )
Word order correlations in Older Egyptian 43
sative syntax. The selection of the Eventive or the Stative conjugation sets
not only a basic parameter for the eventuality described in a particular sen-
tence, but also defines the theta-role of the most prominent argument that is
realized as the surface subject.
3.1. The paradigm structure of the Eventive and the Stative conjugation
Older Egyptian has two exclusively suffixal conjugation patterns for finite
verb forms, which are called the Eventive and the Stative on the basis of
their primary grammatical meaning. The complete inflectional paradigms
of both verb conjugations are represented in table 1. (The triconsonantal
verb sd3m 'to hear' has been chosen to illustrate a typical paradigm).
Word order correlations in Older Egyptian 45
EVENTIVE STATIVE
is 3 3
sd m(-j), sd m-j sd 3 m-k, sd 3 m-kj, sd 3 m-kw
2SM sd 3 m-k sd3m-tj, sd 3 m-c(j)
2SF sd 3 m-c, sd 3 m-cn sd3m-tj, sd 3 m-c(j)
3
3SM sd m-f sd 3 m-w, sd3m-jj,
3SF sd 3 m-s sd3m-tj, sd 3 m-c(j)
1P sd 3 m-n sd 3 m-wn, sd 3 m-wnj, sd 3 m-wjn
3
2P sd m-cn sd 3 m-tjwn, sd 3 m-tjwnj
3 PM sd 3 m-sn sd 3 m-w, sd3m-jj
3PF sd 3 m-sn sd3m-tj
2D sd 3 m-cnj, sd 3 m-cny sd 3 m-tjwnj
3
3 DM sd m-sny sd 3 m-wjj, sd 3 m-wj
3DF sd 3 m-sny sd3m-tjj, sd 3 m-tj
& Sportiche 1994; and Benmamoun 2000). I will argue, however, that we
are dealing with an agreement asymmetry of a rather different kind. The
personal inflections on Eventive verb forms do not instantiate subject-verb
agreement at all, but represent enclitic subject pronouns that correspond to
an argument position. By contrast, grammatical agreement proper is only
instantiated by the featurally less coherent Stative paradigm.
This section argues against the commonly held view in Egyptological lin-
guistics that Stative inflections represent incorporated subject pronouns (see
Allen 1984: 6 §11 and 384 §564 and Schenkel 1997: 199 for representative
views). The first argument concerns the role of locality in distinguishing
between pronoun incorporation and grammatical agreement. The second
argument relates to the limited distribution of pro-drop.
3.2.1. Locality
In view of the fact that the ECM complement does not exceed the domain
of the subject agreement phrase AGRSP, it is hard to see how the co-
occurrence of ECM subjects and Stative verb forms could be reconciled
with the traditional pronoun incorporation analysis of Stative inflections,
given that there would be two subjects for which only a single theta-role
and structural case would be available. If we were to assume that these
personal inflections are incorporated pronouns, we would expect them to
conform to the general principles of the Binding Theory of Chomsky
(1981). In other words, they should be free in their governing category. In
the above examples, the purported subject pronoun is bound within its gov-
erning category (i.e. the subject agreement phrase AGRSP) by either a refer-
ring expression or another pronoun. The grammaticality of (15a-b) above
vis-0-vis the Binding Theory is expected under an agreement analysis of the
Stative paradigm. On this analysis, Stative inflections redundantly express
the person and gender features of the preverbal subject, but do not occupy a
separate structural position (Bresnan & Mchombo 1987: 752ff). As pointed
out by the reviewer, a clitic-doubling account might save the traditional
pronoun incorporation analysis. Such a clitic-doubling analysis would,
however, be at odds with the limited distribution of null subjects, as we will
see next.
3SM e. m sw j-jj
INTERJ him come-3M S T AT
'Look, he (the deceased king) has arrived.'
(Pyramid Texts 1495a/P)
3SF f. wnt-s sr-tj n-k r-s
COMP-3SF foretell-3F S T AT to-2SM about-3SF
'(To inform you) that she (the deceased female) has foretold about
herself to you.' (Coffin Texts I 140g/B3Bo)
Word order correlations in Older Egyptian 49
IP g. pro s-d5f?-wn
CAUS-provide-1 PSTAT
' W e are provided (with food).' (Bissing, Gem-ni-kai I, pi. 18, 2 nd reg.)
h. jw-n s?-wn
AUX-lPEv satisfy-1 P S T A T
' W e are satisfied.' (stele Mehu-akhti (Reisner G 2375),
MIO 1 (1953), p. 328 B: 2, fig. 1)
2P/D i. pro ßr-tjwn r b?-j pn
be.far.away-2D/P S TAT from soul-lS DEM:SM
'You should be far away from this soul of mine.'
(Coffin Texts VI 7 6 c / B 3 L )
j. m-χί cny gm-tjwny
COMP you:D find-2P/D S T AT
'after the two of you (i.e. the two chopped-off hands of the god
Horus) have been found.' (Coffin Texts II 350a/B4L b )
3PM k. sk sn Γχ-jj s(j) η χτη-sn s(j)
COMP they understand-3MsTAT it NEG ignore-3P EV it
'While they (the demons) know it (the eye of Horus), they are not
unaware of it.' (Coffin Texts VII 111 j / SQ1OC)
The above data illustrate the limited distribution of pro-drop in the Stative
paradigm. Null subjects are licensed in the context of first and second person
singular and plural reference, where the inflectional endings are unambigu-
ously specified for [person] and [number], but excluded from third person
contexts where agreement lacks an explicit morphological representation of
the features [number] and, in the feminine, [person]. Furthermore, note that
the availability of pro-drop serves as a means for disambiguating the refer-
ence of the polysemous ending -tj, informally stated in (18).
A similar paradigmatic split between first and second person on the one
hand and third person on the other has been observed for verbal tenses in
Modern Hebrew (Borer 1986; Ritter 1995; Shlonsky 1997). If the inflec-
tional endings on Stative verb forms represented incorporated pronouns, it
would be a complete mystery why clitic-doubling is obligatory in third
50 Chris Η. Reintges
a. jw proexpi d5w-w
AUX be.bad-3MsTAT
'(It) has turned out bad.'
(Lepsius, Denkmäler II [pi. 63, 4th register])
b. jw proexpι sfp(-w) d3d NN pn
AUX be.bright-3MSTAT speak EV NN DEM:SM
jw proexp/ knß(-w) d3d NN pn
AUX be.dark-3MSTAT speak EV NN DEM:SM
'(It) dawns (when) this NN (the deceased male) speaks, (it) gets
dark (when) this NN speaks.' (Coffin Texts IV 29e/Sq6C)
Eventive verb forms come in two varieties: synthetic forms with concord-
marking person inflections and analytic forms without such inflections. The
distribution of both forms is strictly regulated: finite clauses with postver-
bal DP subjects require the analytic form, as (21a) below illustrates. The
corresponding synthetic form is selected in the context of pronominal sub-
jects, as (21b) exemplifies. Although analytic forms lack discrete agree-
ment markings, they can still be inflected for tense and aspect, witness the
presence of the Perfect marker n-.
Word order correlations in Older Egyptian 53
In Older Egyptian, there is a kind of ellipsis process that allows the subject
DP of asyndetically coordinated clauses to be deleted under identity with
the subject of the preceding clause (McCloskey 1991). The main point to be
observed here is that the Eventive verb appears in its analytic form, i.e.
without a personal affix.
All that needs to be said under the pronoun incorporation analysis is that
the subject position of the third conjunct contains a non-pronominal empty
category or gap (indicated as ' '), which is co-indexed with the preceding
DP subject (Huang 1984). Under the competing agreement analysis, one
has to stipulate that there is no agreement with vv/7-subject traces in co-
ordinate structures (see Borsley & Stephens 1989 for an analysis along
these lines for Breton subject ellipsis). On this view, one would expect
analytic forms to be consistently used in subject extraction contexts. This
prediction is not borne out by the empirical evidence, as the distribution of
analytic forms in relative clauses shows.
the referential content of a w/z-subject trace. This raises the question why the
subject gap of co-ordinate structures is never governed by an agreeing verb
form. Under the pronoun incorporation analysis, the personal affix -sn of the
embedded Eventive verb form would be a resumptive pronoun.
Reintges (2000) offers a locality explanation for the distribution of gaps and
resumptive pronouns. Since the embedded subject position is adjacent to
the relative complementizer, the wh-trace is properly head-governed by an
agreeing C O M P . That locality is, indeed, the relevant factor is evident in
Eventive subject relatives, where the initial verb intervenes between C° and
the subject position, thereby blocking government from C°. To rescue the
derivation, the gap has to be replaced by a resumptive pronominal. As we
can see from the contrast between (25a-b) and (25c-d) below, gaps and re-
sumptive pronouns may vary in direct object relatives.
It is hard to see how the availability of analytic and synthetic verb forms in
classical w/z-constructions can be reconciled with an agreement analysis of
the personal inflections on Eventive verb forms. On an agreement account,
one would have to stipulate a variety of construction-specific agreement
58 Chris Η. Reintges
3.3.5. COMP-cliticisation
(27) COMP-CLITICISATION
a. Finite subordinate clauses
r rd3-t ij-t(j) [Cp wnt-k h?-t(j% m Γηρ
t o give-INF understand-PASS2 COMP-2SM descend-2s S T AT in p e a c e
m Y?m ]
from Y a m
'to let it be known that you descended in peace from (the land) Y a m '
(Urkunden I 128: 7 - 8 )
b. Finite relative clauses
ncn pw [Cp nt(j)-k rx-t(j) siVj ]
god DEM:SM COMP REL (-SM)-2SM know-2sSTAT him
'this god whom you know'
(Coffin Texts V 11 l d / M 2 C ) (cf. ex. (25c))
plain and simple, its appearance in the preverbal subject position of an em-
bedded Stative clause can be accounted for in a natural and unconstrained
way. W e return to the syntax of pronominal enclisis in section 4.2.2.3.
This section addresses the question how the verb-initial and subject-initial
word orders of Eventive and Stative clauses are syntactically derived from
the interaction of verb raising and subject positioning. In Reintges' (1997:
2 8 7 - 2 9 8 ) analysis, the finite verb in Stative SVO clauses raises to the highest
inflectional position (but no further). By contrast, the finite verb in Eventive
V S O clauses always raises out of the inflectional domain, targeting the com-
plementizer position or a lower left-peripheral functional head. A crucial as-
sumption of this approach was that Eventive and Stative subjects appear in
the same structural slot, to wit, the Specifier position of the IP. The analysis
just outlined gives a correct description of the word order and agreement
facts of Stative S V O order, but does not present a complete picture of the
flexible syntax of Eventive V S O clauses, in particular, the availability of
more than one hierarchical position for the postverbal subject.
My point of departure is the view that V S O and S V O orders are both de-
rived f r o m the same V P shell, a plausible assumption given the broad lexi-
cal distribution of Eventive and Stative verb forms. The shared VP-internal
structure encodes valency and argument hierarchy information, but is void
of semantic information about event types and argument meaning. This
must be so, since eventuality and theta roles are not part of the lexical-
semantic template of the root itself, but are specified at the next deriva-
tional level or phase, in C h o m s k y ' s (2000, 2001) system. 1 1 I adopt the main
60 Chris Η. Reintges
(29)
Spec
DO
fik? nb V ROOTP
V/r-x/
Word order correlations in Older Egyptian 61
The skeletal structure of VPs shown in (29) is incompatible with the claims
of the so-called Split-VP Hypothesis, according to which one or more func-
tional functions, for instance, the Object Agreement Phrase AGROP, may
intervene between the subject and direct object licensing verbal heads ν and
V, respectively (see Harley 1995 and the references cited therein). The
departure of relatively well-established principles of clausal organisation
(in particular, the bifurcation between a lexical and functional layer) is,
however, contradicted by the empirical evidence from Older Egyptian ob-
ject shift, where the DP subject occupies a position below the target posi-
tion of the shifted object pronoun (see McCloskey 1997: 221-225, 2001:
179-189 for counterevidence from Irish object shift and dative subjects).
In generative analyses that derive VSO order via verb fronting, it is gener-
ally agreed upon that the subject occupies the specifier position below the
functional head in which the finite verb appears. 12 In a left-headed lan-
guage, this will always mean that the subject will appear to the right of the
finite verb. Within this broad consensus, the exact position of the subject
and the scope of verb movement are topics of controversy. One school of
thought represented by Emonds (1985), Sproat (1985), Stowell (1989) and
Aoun, Benmamoun & Sportiche (1994) holds that V S O order is derived
from SVO order through verb movement and adjunction to COMP in a
manner reminiscent of den Besten's (1983) influential analysis of the Verb
Second phenomenon in West-Germanic. Under the V-to-C analysis, the
subject occupies the inflectional subject position [Spec, IP], See the dia-
gram in (30) below for further illustration.
I I I
[cpC° [IP S U [R 1° [ V P . . . v ° ... ]]]]
an underlying SVO order, but the subject and the direct object are both
case-licensed in their base-position in a configuration of government by the
finite verb, as schematically represented in diagram (31) below.
I I
[r 1° [VP su [v v° DO ]]]
This section argues that Eventive VSO order can be arrived at by different
routes. There are two parameters of variation, viz. (i) the availability for
more than one clausal position for the licensing of the subject and (ii) dif-
ferent targets for verb movement.
The finite verb in Eventive VSO clauses displays the characteristic behaviour
of verb raising. First, the verb may appear to the left of sentential adverbs and
negation, suggesting that it has moved out of the vP shell. As a result of verb
movement, adverbial particles like (j)r-f 'indeed' in (32a) and js 'really' in
(32b) below occur in clause-second position, from where they precede the
subject DP. Evidently, there is no strict adjacency requirement in Older
Egyptian that holds between the finite verb and the nominal subject.
Word order correlations in Older Egyptian 63
Since Pollock (1989, 1997) and Cinque's (1999) seminal work on adverbial
distribution and clausal architecture, the placement of negative elements
has been used as a diagnostic tool for verb movement in various languages.
Older Egyptian has two negative markers, the clause-initial negative parti-
cle η (var. njj) and the clause-internal negation w 'not' (see Loprieno 1995;
Kammerzell 1993 for further discussion). As illustrated in (33) below, the
verb raises out of the vP to the left of the negative adverbial w.14
Third, there is in Older Egyptian an ellipsis process that targets only the
subject, but not the complement and the finite verb of asyndetically coordi-
nated clauses. The verb bears tense and aspect inflection, indicating that the
subject is elided after the verb has moved to Tense (McCloskey 2001: 161-3).
(35) S U B J E C T ELLIPSIS
The syntactic facts considered so far support a verb (X°) raising account of
verb-first sentences. It remains to be seen whether an alternative analysis in
terms of remnant verb phrase movement (with previous extraction of the
direct and indirect object) would account equally well for the complexities
of Older Egyptian VSO order. I leave this issue for future research.
Two interrelated questions arise from the previous discussion on verb raising:
(i) How far does the verb raise in Eventive VSO clauses?
(ii) Which positions are available for the postverbal subject?
(36) V S O C L A U S E S W I T H N P - I N T E R N A L S U B J E C T DPS
I I I 0
I
[τ Τ [vp D P S U [v. ν [Vp D P d o [ν' V [ROOTP ROOT ]]]]]]
As the locus of tense and event structure, the T°-node qualifies as a sub-
stantive rather than a functional category, falling together with V and Ν
(Chomsky 2001: 9). With respect to its featural content, one may plausibly
assume that its temporal, aspectual, and modal specifications instantiate
interpretable features, which remain accessible throughout the derivation.
Since there is no morphological correlate of subject-verb agreement on
Eventive-inflected verb forms, the stipulation of uninterpretable phi-features
on Tense is empirically unsubstantiated. The trigger for V—>T raising is
therefore not feature co-valuation, but rather the need for providing Tense
with an uncontroversial lexical root at some point in the derivation (Nash &
Rouveret 2003).
In Older Egyptian, Tense is morphologically expressed by means of
suffixes, which need to be attached to a verbal host. Verb raising is a pre-
condition for combining the functional features of Tense with the lexical-
semantic features of the verbal root. What evidence is there for the subject-
in-situ derivation in (36) above? The first argument for the location of DP
66 Chris Η. Reintges
wd3-n sw Pg? n- fw
order- PERFev him Open.one to- Shü
' T h e Open One has commended him (the king) to (the god) Shü.'
(Pyramid Texts 6 0 4 b / N )
Word order correlations in Older Egyptian 67
Object shift also affects the placement of dative pronoun, as shown in (39a)
below. When both objects are pronominal arguments, indirect object pre-
cedes the direct object in linear order, as seen in (39b) below.
There is good reason to believe that Older Egyptian object shift is a syntac-
tically rather than prosodically driven reordering process. To begin with, it
is possible for the negation adverb w and the emphatic particle js to inter-
vene between the verb and the shifted object pronoun, as exemplified by
(40a-b) below. It is clear, then, that shifted object pronouns do not form a
prosodic constituent with the main verb.
Leftward raising of the verb past the vP-internal subject represents one
route to the verb-initial surface order in Older Egyptian, but not the only
such route. A more complex derivational route by which VSO order can be
arrived at involves subject raising to the specifier of TP and verb movement
to a functional head F° at the left edge of the inflectional domain. It looks
as if the F°-node has information structure content, being involved in the
licensing of left-peripheral topic and focus phrases (see Aoun, Benmamoun
& Sportiche 1994: 204, fn.8; and Uriagereka 1995: 9 5 - 9 7 for related ideas).
Reintges (1997: 127ff) furthermore suggests that auxiliary verbs are merged
into this position, since such verbs play a crucial role in the temporal orga-
nisation of narrative discourse, introducing foregrounded chains of events.
The optional activation of the EPP feature on Tense gives rise to a VSO
variant with raised subjects.
ΐ I
The syntactic distribution of negation and information-structuring particles
has been used as a diagnostic tool for both verb movement and subject po-
sitioning. Granted that adverbial elements have a fixed position in the syn-
tactic structure, the surfacing of both the finite verb and the subject DP to
the left of the negation w and the emphatic particle js in examples like (43)
below clearly shows that both constituents have vacated the vP.
Further support for the vP-external location of the subject DP comes from
the stranding of postnominal adjectives. In Older Egyptian, universally
quantified subjects must undergo Quantifier Raising in the narrow syntax.
This is illustrated in (44) below, in which the subject χί nb 'everything'
precedes the object shifted dative pronoun r-f ' f r o m him'. The original
position of the raised quantifier is marked by the stranded adjective d3wt
'evil' (Borer 1995: 535). Further note that universal quantifiers are inter-
preted as free choice items in negative contexts.
a. Main clauses
DP S U > V E R B > CL S U > PP
wj? Mi s-qd-d-fi m-gnw jgpt
barque Re CAUS-row- IMPERF-3smEv in-inside cloudy.sky
' T h e barque of (the sun-god) Re, it sails inside the cloudy sky.'
(Coffin Texts IV 1 2 5 c / S l P )
b. Subordinate clauses (reported speech)
C O M P > DPsu > V E R B > CL S U
dM-n-k n- R? [ CP ntt Nt, jw-S; ]
say-PERF-2SM E V to R e COMP N e i t h come(-PROS)-3SF E V
' Y o u told (the sun-god) Re that (Queen) Neith would come.'
(Pyramid Texts/Neith 4 0 - 4 1 [pi. 7])
Word order correlations in Older Egyptian 71
I I0
I I I
[fp O p [ F F° [ T P CL S U [ T ' T [ v p ? c l [ v v [Vp D P d o [v V° [ROOTP ROOT ]]]]]]
ΐ I
The raising analysis of enclitic subject pronouns is empirically well moti-
vated. As shown by the contrast between the a- and the b-examples of ( 4 9 ) -
(50) below, subject DPs and clitics appear on opposite sites of the clause-
internal negation w and the emphatic particle js. This mirror image pattern
follows from their respective vP-internal and vP-external location.
That subject clitics occupy a high position in the syntactic tree is also evi-
dent from their appearance in front of shifted pronominal objects, as shown
in (51).
To answer the question where the clitic double has been raised to, a com-
parison between Stative and Eventive complement clauses is instructive.
The embedded subject position of Eventive and Stative clauses can be oc-
cupied by an enclitic pronoun, which leans on the preceding lexical com-
plementizer. But while there are two instances of the subject clitic in the
embedded Eventive clauses in (54a) below, the presence of independent
agreement morphology on the finite verb in the Stative counterpart in (54b)
automatically excludes such clitic doubling.
we can see from the following minimal sentence pair, passive Is and pas-
sive 2s appear as functional equivalents of one another in main clauses.
b. η ms-n-t(j)-j js ms-jj-t
NEG bear-PERF-PASS2-lS EV EMPH bear-GER-SF
Ί was really not born in a birthing.' (Coffin Texts I 3 4 4 c / S l C )
At this juncture, one might suspect that the auxiliary and the passive 2 verb
compete for the same syntactic position, viz. the C°-node. This cannot be
80 Chris Η. Reintges
the whole story, however. As we can see in (61) below, an auxiliary verb
like wn 'be' may appear in a conditional adjunct clause formed with the pre-
positional complementizer jr 'as for, i f . I take this to mean that auxiliary
verbs are not subject to the Verb Second Condition.
(61) A U X I L I A R Y V E R B S IN C O N D I T I O N A L A D J U N C T CLAUSES
The occurrence of examples like (62) above is in line with our expectations,
since the auxiliary verb and the passive 2 formative -tw are both located in
their designated landing site, which is the C°-node. While Verb Second
applies in full force in passive 2s, passive Is may, but need not, move all
the way up to C° and may eventually be positioned in a lower functional
head. The parameterisation of the V2 feature correctly predicts that pas-
sives 1 and passives 2 occur in free variation in main and embedded clauses
with a vacant C°-node. It is worth pointing out that weak V2 effects are by
no means a hallmark of the passive paradigm, but can also be observed in
other verbal-inflectional patterns, such as subjunctive or imperfective verb
forms (see Reintges 1997: 79ff). Future research has to clarify the morpho-
syntactic factors that underlie the parameterisation of the Verb Second
Condition.
Word order correlations in Older Egyptian 81
ΐ I
Despite their derivational connection, the merging of the subject into the
Specifier of AGRSP and the adjunction of the verb into the AgrS°-head
satisfy different syntactic constraints. The interpretable features on the
AGRS°-node are [+stative], [+finite], while the φ-features and the EPP fea-
ture represent uninterpretable features. In Chomsky's (2000, 2001) system,
the EPP, like other selectional features, seeks an XP to merge with the
category it heads. The uninterpretable φ-features act as a probe that seeks a
goal with the corresponding interpretable features to establish agreement.
The AGREE relation removes the uninterpretable features from the narrow
syntax, making it possible for the narrow syntactic derivation to converge at
LF. As we will see next, verb movement is triggered by the need to combine
the interpretable semantic features on the AGRS°-node with a verbal root.
category just like Eventive Tense, which must be combined with a verbal
root at some point in the derivation. I assume modulo Baker's (1988)
'Mirror Principle' that verb movement and adjunction to AGRS0 have a mor-
phological correlate in the templatic structure (VERB + SUFFIX) of Stative-
inflected verb forms.
If the clause-second position of Stative verb forms can be identified
with the AGRS°-node, the preverbal subject is in all likelihood located in
the specifier position of that node. This account predicts - c o r r e c t l y - that
the subject DP appears in front of adverbial particles and shifted pronomi-
nal objects. In (64) below, both the ECM subject fiq? Y?m 'the rule of
Yam' and the embedded Stative fm(-w) 'had departed' precede the subject-
oriented particle (j)r-f, which is left-adjoined to the vP.
In examples like (65a) below, the postverbal position of direct object pro-
noun cw 'you (man)' is by itself not indicative of object shift, given the
SVO surface order, but would still indicate that both the subject and the
verb have raised out of the vP domain. The applicability of pronominal
object shift to a Stative clause is, however, evident from the CL 1 0 > DP DO
order of indirect object pronouns in examples like (65b) below. Recall that
the canonical order of nominal object is DP DO > DPI 0 .
It is clear then that the preverbal subject and the finite verb of Stative
clauses have both raised out of the vP. That Stative subjects are, indeed,
located in the Specifier position of AGRSP and not higher up in the syntac-
tic tree is evident from their occurrence as ECM subjects as well as their
intermediate position between the auxiliary jw and the main verb, since
auxiliary verbs appear in a preclausal functional position F°.
The familiarity condition does not exclude Stative subjects from being as-
sociated with different types of semantic focus (E. Kiss 1998). In (68) be-
low, the interjection m- ' l o o k ' signals new information focus on the adja-
cent subject. Definite expressions (proper names, personal pronouns) are
admissible as presentational foci, since they correspond to discourse-new
Word order correlations in Older Egyptian 85
information, meaning that the relation its referent enters into is novel with
respect to the preceding stretch of discourse.
In (69) below, the sentence negation η takes wide scope over the indefinite
N P z? (J)tm 'a son of (Atum)', which therefore receives the semantic inter-
pretation of a downward entailing quantifier (NOT > A SON = NO SON). As
non-referential expressions, such quantifiers are excluded as topics, although
they do function as subjects (Alexiadou & Anagnostopoulou 1998: 5 0 4 -
511; Alexopoulou, Doron & Heycock 2003: 9-12).
The preverbal subject position hosts not only presentational foci and raised
quantifiers, but is equally well suited as a landing-site for w/?-phrases, as
seen in (70) below.
differs from these languages in that the preverbal subject position is re-
stricted to nominative arguments. It looks as if the EPP position of Stative
clauses is targeted as a landing site by various types of operators, indicating
that the AgrS 0 conveys some left-peripheral functional features (Topic,
Contrast, Focus). This provides an explanation for the absence of topicali-
sation and clitic-lefit-dislocation structures in Stative clauses.
In this section, I will show that there are Stative clauses that look superfi-
cially like CLLDs, but actually instantiate a sentence pattern with two
nominative subjects. In the so-called Broad Subject construction, the A-
and A'-properties of EPP-induced subject merge are distributed over two
specifier positions, as exemplified in (71a-b) below.
the higher specifier, while the latter are derived subjects that originate
within the vP. Since the broad subject may be either topic or focus, it seems
natural to classify the higher specifier as an A'-position and the lower speci-
fier as an Α-position. K u r o d a (1988) draws a connection between the mul-
tiple filling of the specifier of IP in Japanese and the impoverishment of
agreement. On this account, one would expect, - contrary to the facts - that
the multiple subject construction would be another syntactic variant of an
Eventive V S O clause that lacks subject-verb agreement altogether. It rather
looks that the possibility of filling and licensing multiple subjects depends
on the active status of the EPP feature on the relevant inflectional node
rather than the richness of subject agreement itself.
5. S u m m a r y and conclusions
Notes
the lexical-semantic template of the underlying root. Thirdly, a given root can
appear in only one reduplicative pattern. Reintges (2003) conjectures that the
non-cyclicity of reduplicative copying has its basis in some economy principle
preventing the generation of strings of two adjacent copies of the root mor-
pheme, which would leave room for too many parsing options.
10. There are some remarkable structural differences in the synthetic paradigms of
Older Egyptian and Modern Celtic languages. First, in contradistinction to
Modern Irish, separate affixes do exist for every person, gender and number
distinction in Older Egyptian verbal paradigms. Second, inflected prepositions
in Irish, unlike their Older Egyptian counterparts, exhibit a high degree of
morpho-phonological irregularity (McCloskey & Hale 1984: 506). Third, in
Breton, it is possible to construe a subject clitic with an analytic verb form, as
in (ia), or to attach a subject clitic to a co-referential synthetic verb form, as in
(ib). In line with Borsely & Stephens (1989), one might analyse (ia) as an in-
stance of a topicalisation construction in which the analytic form agrees with
the trace left behind by subject extraction. In (ib), we seem to be dealing with a
special type of clitic-doubling, where the second (clitic) pronoun is added to
the synthetic verb form for contrastive purposes. Contrary to what is said in
Stump (1984: 333-7), the absence of resumptive pronominalisation in the con-
text of direct object fronting does not provide compelling evidence against the
pronoun incorporation analysis of synthetic verbal paradigms, but rather shows
that Breton displays the standard subject-object asymmetries.
realised on the finite verb, whereas Eventive VSO clauses have no agreement
at all, but distinct tense and aspect projections. In short, the difference between
Eventive and Stative clauses is due to the functional superstructure above the
v P domain.
12. For Chamorro (Mariana Island, Western Austronesian), Chung (1998, chapter
4) presents detailed arguments to show that the pragmatically unmarked VSO
order cannot be derived from SVO order, but rather derives from VOS order
via lowering of the subject and adjunction to the right of the verbal predicate.
This generally shows that verb-initial languages are typologically as diverse as
subject-initial ones.
13. To account for the agreement mismatches between VSO and SVO clauses in
Modern Standard Arabic, Mohammad (1990) and Ouhalla (1994) propose that
the postverbal subject of VSO sentences is located in the VP-internal subject
position, while the structural subject position [Spec, IP] is occupied by a pho-
nologically null expletive pronoun that induces impoverished agreement in-
flection. At LF the null expletive is replaced by its associate, the subject DP in
the thematic subject position [Spec,VP], As pointed out by Alexiadou &
Anagnostopoulou (1998) and Doron (2000, this vol.), the /7ro exp |-VSO analysis
is intrinsically problematic, since VSO clauses in Modern Standard Arabic and
Modern Hebrew do not show the indefiniteness restriction. This criticism is
not entirely justified, however, as defmiteness effects are systematically absent
in Arabic VSO structures, even when the expletive pronoun is lexicalised. The
spell-out of the expletive pronoun hu 'it' is obligatory in the context of the
subordinating finite complementizer ?anna 'that'.
(i) OVERT EXPLETIVE CONSTRUCTIONS WITH DEFINITE DP SUBJECTS
a. öann-a [CP ?anna-hu wasal-a ?at-tulaabu ]
think-PERF.3SM that-ζ'ί arrive-PERF.3SM the-male.student.PM.NOM
b. Sann-a [c? ?anna-hu wasal-at ?at-taalib-aat
think-PERF.3SM that-it arrive(-PERF)-3sf the-female.students-PF.NOM
'He thought that the male/female students had arrived.'
In ?anna 'that' complement clauses, the finite verb agrees in person and gender
with the postverbal DP subject rather than the expletive pronoun singular mascu-
line hu 'it'. This is particularly clear in the context of feminine DPs, as in (ib)
above. According to Ouhalla (1994: 70, fn.2), the agreement of the embedded
verb with the postverbal subject is only apparent. His main argument is that the
third person singular masculine form of the finite verb can also appear before
feminine subjects. This agreement pattern is instantiated in Classical Arabic
(Fischer 1972: 165, §356), but judgements about its grammaticality status in
Modern Standard Arabic vary. Moreover, if gender agreement with the post-
verbal subject were, indeed, only apparent, it remains to be explained why the
selection of the third person singular feminine form of the verb in the context
of masculine subjects results in ungrammaticality. See Aoun, Benmamoun &
Word order correlations in Older Egyptian 93
References
Andersen, S. R.
1982 Where's morphology? Linguistic Inquiry 13: 571-612.
Adger, D.
1997 VSO and Weak Pronouns in Celtic. Canadian Journal of Linguistics
42: 9-31.
94 Chris Η. Reintges
Borer, H.
1981 Comments on Pro-Drop Phenomena. MIT Working Papers in Lin-
guistics 3: 127-149. Cambridge, Mass.
1986 I-Subjects. Linguistic Inquiry 17: 375-416.
1995 The Ups and Downs of Hebrew Verb Movement. Natural Language
and Linguistic Theory 13: 527-606.
Borer, H. and Y. Grodzinsky
1986 Syntactic Cliticization and Lexical Cliticization: The Case of Hebrew
Dative Clitics. In The Syntax of Pronominal Clitics, H. Borer (ed.),
175-217. Syntax and Semantics 19. New York: Academic Press.
Borsley, R. D. and J. Stephens
1989 Agreement and the Positions of Subjects in Breton. Natural Language
and Linguistic Theory 7: 407^127.
Bowers, J.
2002 Transitivity. Linguistic Inquiry 33: 183-224.
Bresnan, J. and S. A. Mchombo
1987 Topic, Pronoun, and Agreement in Chichewa. Language 63: 741-782.
Carnie, Α., Η. Harley and E. Pyatt
2000 VSO Order as Raising Out of IP? Some Evidence from Old Irish. In
The Syntax of Verb-initial languages, A. Carnie and E. Guilfoyle
(eds.), 39-59. Oxford Studies in Compartive Syntax. Oxford Univer-
sity Press.
Carstens, V.
2003 Rethinking Complementizer Agreement: Agree with a Case-Checked
Goal. Linguistic Inquiry 34: 393-412.
Chomsky, N.
1981 Lectures on Government and Binding. Dordrecht: Foris Publicatons.
1996 The Minimalist Program. Current Studies in Linguistics 28. Cam-
bridge, Mass.: MIT Press.
2000. Minimalist Inquiries: The Framework. In Step by Step: Essays
on Minimalist Syntax in Honour of Howard Lasnik, R. Martin, D.
Michaels and J. Uriagereka (eds.), 89-155. Cambridge, Mass.: MIT
Press.
2001 Derivation by Phase. In Ken Hale: A Life in Language, M. Kenstowicz
(ed.), 1-52. Current Studies in Linguistics 36. Cambridge: MIT Press.
Chung, S.
2003 The Syntax and Prosody of Weak Pronouns in Chamorro. Linguistic
Inquiry 34: 547-599.
Chung, S. and J. McCloskey
1987 Government, Barriers, and Small Clauses in Modern Irish. Linguistic
Inquiry 18: 173-237.
96 Chris Η. Reintges
Cinque, G.
1990 Types of A'-Dependencies. Cambridge, Mass.: MIT Press.
1999 Adverbs and Functional Heads: A Cross-Linguistic Perspective.
Oxford Studies in Comparative Syntax. Oxford/New York: Oxford
University Press.
Collins, C.
2002 Multiple Verb Movement in =j= Hoan. Linguistic Inquiry 33: 1-29.
Collins, C. and H. Thräinsson
1996 VP-Internal Structure and Object Shift in Icelandic. Linguistic In-
quiry 27: 3 9 1 ^ 4 4 .
Comrie, B.
1976 Aspect - An Introduction to the Study of Verbal Aspect and Related
Problems. Cambridge: Cambridge University Press.
Craenenbroeck, J. van and M. van Koppen
2002 The Locality of Agreement and the CP-Domain. Paper Presented at
GLOW 2002 Conference in Amsterdam, The Netherlands.
Diesing, M.
1990 Verb Movement and the Subject Position in Yiddish. Natural Lan-
guage and Linguistic Theory 8: 41-79.
Doron, E.
1988 On the Complementarity of Subject and Subject-Verb Agreement. In
Agreement in Natural Language: Approaches, Theories, Descriptions,
M. Barlow and C. A. Ferguson (eds.), 201-218. Papers Presented at
a Conference Held at Stanford University in October 1984. Stanford:
CSLI.
1996 The Predicate in Arabic. In Studies in Afroasiatic Grammar, Papers
from the Second Conference on Afroasiatic Languages Sophia Anti-
polis 1994, J. Lecarme, J. Lowenstamm and U. Shlonsky (eds.), 7 7 -
104. The Hague: Holland Academic Graphics.
2000 VSO and Left-conjunct Agreement: Biblical Hebrew vs. Modern
Hebrew. In The Syntax of Verb-initial languages, A. Carnie and E.
Guilfoyle (eds.), 75-96. Oxford Studies in Compartive Syntax. Ox-
ford University Press. [Reprinted in this volume.]
2003 Agency and Voice: The Semantics of the Semitic Template. Natural
Language Semantics 11: 1-67.
Doron, E. and C. Hey cock
1999 Filling and Licensing Multiple Specifiers. In Specifiers - Minimal-
ist Approaches, D. Adger, S. Pintzuk, B. Plunkett, and G. Tsoulas
(eds.), 69-89. Oxford: Oxford University Press.
Dubinsky, S. and S. R. Simango
1996 Passive and Stative in Chichewa: Evidence for Modular Distinctions
in Grammar. Language 72: 749-781.
Word order correlations in Older Egyptian 97
Edel, Ε.
1955/64 Altägyptische Grammatik. Analecta Orientalia 34/39, Rome: Ponti-
ficial Biblical Institute.
1959 Beiträge zur ägyptischen Grammatik 1. Zum angeblich geminierenden
Pseudopartizip der Verben III. infirmae. Zeitschrift für Ägyptische
Sprache und Altertumskunde 84: 105-108.
Emonds, J. E.
1985 A Unified Theory of Syntactic Categories. Studies in Generative
Grammar 19. Dordrecht: Foris Publications.
Fassi Fehri, A.
1988 Agreement in Arabic, Binding and Coherence. In Agreement in Natural
Language: Approaches, Theories, Descriptions, M. Barlow and C. A.
Ferguson (eds.), 107-158. Papers Presented at a Conference held at
Stanford University in October 1984, Stanford: CSLI.
1993 Issues in the Structure of Arabic Clauses and Words. Dordrecht:
Kluwer.
Fischer, W.
1972 Grammatik des Klassischen Arabisch. Porta Linguarum Orientalium
NS 11. Wiesbaden: Harrasowitz.
Franks, S.
1990 The Status o f N u l l Expletives. Lingua 81: 1-24.
Fukui, N. and M. Speas
1986 Specifiers and Projection. In MIT Working Papers in Linguistics 8,
Ν. Fukui, Τ. R. Rapoport and E. Sagey (eds.), 128-172.
Givon, T.
1976 Topic, Pronoun, and Grammatical Agreement. In Subject and Topic,
C.N. Li (ed.), 149-188. New York/London: Academic Press.
Greenberg, J. H.
1966 Universals of Language. Cambridge, Mass.: MIT Press.
Grimshaw, J.
1990 Argument Structure. Linguistic Inquiry Monograph 18. Cambridge,
Massachusetts: MIT Press.
Guilfoyle, Ε., H. Hung and L. Travis
1992 Spec of IP and Spec of VP: Two Subjects in Austronesian Lan-
guages. Natural Language and Linguistic Theory 10: 375-414.
Hale, K. and S. J. Keyser
1993 On Argument Structure and the Lexical Expression of Syntactic
Relations. In The View from Building 20: Essays in Linguistics in
Honour of Sylvain Bromberger, K. Hale and S. J. Keyser (eds.), 5 3 -
109. Current Studies in Linguistics 24. Cambridge, Mass.: MIT Press.
1997 On the Complex Nature of Simle Predicators. In Complex Predicates,
A. Alsina, Joan Bresnan, and Peter Sells (eds.), 29-65. CSLI Lecture
Notes 64. CSLI Publications: Stanford.
98 Chris Η. Reintges
Larson, R.
1988 On the Double Object Construction. Linguistic Inquiry 19: 335-391.
Lehmann, C.
1988 On the Function of Agreement. In Agreement in Natural Language:
Approaches, Theories, Descriptions, M. Barlow and C. A. Ferguson
(eds.), 55-65. Papers Presented at a Conference held at Stanford
University in October 1984, Stanford: CSLI.
Levin, B. and M. Rappaport-Hovav
1986 The Formation of Adjectival Passives. Linguistic Inquiry 17, 623-661.
1995 Unaccusativity At the Syntax-Lexical Semantics Interface. Linguistic
Inquiry Monographs 26. Cambridge, Mass.: MIT Press.
Loprieno, A.
1995 Ancient Egyptian - A Linguistic Introduction. Cambridge: Cambridge
University Press.
Marantz, A.
1997 No Escape From Syntax: Don't Try Morphological Analysis in the
Privacy of Your Own Lexicon. In Proceedings of the 2Γ' Annual
Penn Linguistics Colloquium, University of Pennsylvania Working
Papers in Linguistics 4.2, A. Dimitriadis, L. Siegel, C. Surek-Clark
and A. Williams (eds.), 201-225. University of Pennsylvania, Phila-
delphia.
2000 The Universality of Root and Pattern Morphology. Paper Presented
at the 5th Conference on Afroasiatic Linguistics (June 2000), Paris.
McCloskey, J.
1991 Clause Structure, Ellipsis and Proper Government in Irish. Lingua
85: 259-302.
1996 On the Scope of Verb Movement in Irish. Natural Language and
Linguistic Theory 14: 47-104.
1997 Subjecthood and Subject Positions. In Elements of Grammar,
L. Haegeman (ed.), 197-235. Dordrecht: Kluwer.
2001 The Distribution of Subject Properties in Irish. In Objects and Others
Subjects: Grammatical Functions, Functional Categories, and Confi-
gurationality, W. D. Stanley and St. Dubinsky (eds.), 157-192. Studies
in Natural Language and Linguistic Theory 52. Dordrecht: Kluwer.
McCloskey, J. and K. Hale
1984 On the Syntax of Person-Number Inflection in Modern Irish. Natural
Language and Linguistic Theory 1: 487-533.
Mchombo, S. S.
1993 A Formal Analysis of the Stative Construction in Bantu. Journal of
African Languages and Linguistics 14: 5-28.
100 Chris Η. Reintges
Mohammad, Μ. Α.
1990 The Problem of Subject-Verb Agreement in Arabic: Towards a Solu-
tion. In Perspectives on Arabic Linguistics I, Papers from the First
Annual Symposium on Arabic Linguistics, M. Eid. Current (ed.),
95-125. Issues in Linguistic Theory 63. Amsterdam/Philadelphia:
Benjamins.
Mourelatos, A. P. D.
1981 Events, Processes and States. In Tense and Aspect, P. Tedechi and A.
Zaenen, 191-212. Syntax and Semantics 14. New York: Academic
Press.
Nash, L. and A. Rouveret
2003 Cliticization as Unselective Attract. Ms., Universite Paris-7.
Nedjalkov, V. P. and S. J. Jaxontov
1988 The Typology of Resultative Constructions. In Typology of Resulta-
tive Constructions, V. P. Nedjalkov (ed.), 3-62. Typological Studies
in Language 12, 1988. Amsterdam/ Philadelphia: Benjamins.
Ouhalla, J.
1994 Verb Movement and Word Order in Arabic. In Verb Movement,
D. Lightfoot and N. Hornstein (eds.), 41-72. Cambridge: Cambridge
University Press.
Parsons, T.
1990 Events in the Semantics of English: A Study in Subatomic Semantics.
Current Studies in Linguistics 19. Cambridge, Mass.: MIT Press.
Perlmutter, D. M.
1978 Impersonal Passives and the Unaccusative Hypothesis. In Proceed-
ings of the Fourth Annual Meeting of the Berkeley Linguistic Society,
J. Jaeger (ed.). Berkeley Linguistic Society: University of California,
Berkeley.
1983 Personal and Impersonal Constructions. Natural Language and Lin-
guistic Theory 1: 141-200.
Perlmutter, D. M. and P. M. Postal
1983 Toward a Universal Characterisation of Passivisation. In Studies in
Relational Grammar 1, D. M. Perlmutter (ed.), 3-29. Chicago: The
University of Chicago Press.
Pollock, J.-Y.
1989 Verb Movement, Universal Grammar, and the Structure of IP. Lin-
guistic Inquiry 20: 365-424.
Reintges, C. H.
1996 Egyptian Root-and-Pattern Morphology. Lingua Aegyptia 4. Proceed-
ings of the International Conference on Egyptian Grammar. Cross-
roads III (Yale 1994), 213-244.
1997 Passive Voice in Older Egyptian: A Morpho-Syntactic Study. HIL
Dissertations 28. The Hague: Holland Academic Graphics.
Word order correlations in Older Egyptian 101
Shlonsky, U.
1990 Pro in Hebrew Subject Inversion. Linguistic Inquiry 21: 263-275.
1997 Clause Structure and Word Order in Hebrew and Arabic: An Essay
in Comparative Semitic Syntax. Oxford Studies in Comparative Syn-
tax 11. Oxford/New York: Oxford University Press.
Smith, C. S.
1991 The Parameter of Aspect. Studies in Linguistics and Philosophy 43.
Dordrecht: Kluwer.
Sproat, R.
19B5 Welsh Syntax and VSO Structure. Natural Language and Linguistic
Theory 3: 173-216.
Stowell, T.
1989 Raising in Irish and the Projection Principle. Natural Language and
Linguistic Theory 7: 317-359.
Stump, G. T.
1984 Agreement vs. Incorporation in Breton. Natural Language and Lin-
guistic Theory!·. 289-348.
1989 Further Remarks on Breton Agreement. Natural Language and Lin-
guistic Theory 7: 429-471.
2001 Inflectional Morphology: A Theory of Paradigm Structure. Cam-
bridge Studies in Linguistics 93: 57-149.
Talmy, L.
1985 Lexicalization Patterns: Semantic Structure in Lexical Forms. In
Language Typology and Syntactic Description, vol. III. Grammatical
Categories and the Lexicon, T. Shopen (ed.), 57-149, Cambridge:
Cambridge University Press.
Tenny, C. L.
1994 Aspectual Roles and the Syntax-Semantics Interface. Studies in Lin-
guistic and Philosophy 52. Dordrecht.
Uriagereka, J.
1995 Aspects of the Syntax of Clitic Placement in Western Romance.
Linguistic Inquiry 26: 79-123.
Vendler, Z.
1967 Linguistics in Philosophy. Ithaca (New York): Cornell University
Press.
Vikner, S.
1994 Finite Verb Movement in Scandinavian Embedded Clauses. In Verb
Movement, D. Lightfoot and N. Hornstein (eds.), 117-147. Cam-
bridge: Cambridge University Press.
1995 Verb Movement and Expletive Subjects in the Germanic Languages.
Oxford Studies in Comparative Syntax. Oxford: Oxford University
Press.
Word order correlations in Older Egyptian 103
Wasow, Th.
1977 Transformation and the Lexicon. In Formal Syntax, P. Culicover, Th.
Wasow and J. Bresnan (eds.), 327-360. New York: Academic Press.
Zagona, K.
2002 The Syntax of Spanish. Cambridge Syntax Guides. Cambridge: Cam-
bridge University Press.
Zubizaretta, M. L.
1998 Prosody, Focus, and Word Order. Linguistic Inquiry Monographs
33. Cambridge, Massachusetts: MIT Press.
The nominal cleft construction in Coptic Egyptian
1. Introduction
Turning to the topic of our paper, cleft sentences are very common in the
syntactic patterns of Sahidic Coptic to indicate the focal status of a nominal
argument. As with English clefts, Coptic cleft constructions have a bi-
clausal syntax, yet express a single proposition. The focused DP p-tfoejs
'the Lord' in (2a), the ννΛ-subject nim 'who' in (2b), as well as the contras-
tively stressed pronoun ntof 'HE' in (2c) are all associated with an out-of-
focus relative clause that is introduced by the relative complementiser et
'that'. (Relative gaps are indicated as ' ').3
(2) a. DP-clefts
p-tfoejs gar p(e) [Cp et ο m-metre
DEF:SM-lord PCL PRON:SM CREL(PRES) be as-witness
n-ta-senedis ].
to-DEF:SM: 1 S-conscience
'For (it is) the Lord who is witness to my conscience.'
(V. Pach. 89: 1-2)
b. Wh-clefts
nim tenu p(e) [CP et sorm m-p-meefe ]?
who A D V P R O N : S M CREL ( P R E S ) misguide DO-DEF:SM-crowd
'Who (is it) now that is misleading the crowd?'
(Acts of Andrew and Paul 212: 231)
c. Pronoun clefts
ntof gar pe [cp et shai gn-fatfe nim
HE PCL PRON:SM CREL (PRES)write DO-word every
[e-nere Jesus tfo ommo-u ]]
REL-PRET Jesus say D O - 3 P
'Since (it is) him who wrote down every word that Jesus had said.'
(Pistis Sophia 71: 5-6)
Different from English //-clefts, there is no copular verb in the Coptic coun-
terpart that connects the cleft constituent to the subordinate relative clause.
Instead, a deictic pronoun (glossed as P R O N ) is employed to perform this
linkage function. The deictic pronoun recurs in predicational and identifica-
tional nominal sentences, as shown by the contrast between (3 a) and (3 b)
below.4
The nominal cleft construction in Coptic Egyptian 107
Coptic cleft sentences are typically used for contrastive emphasis with
various degrees of strength. Following Rooth (1992) and subsequent re-
search, we assume that contrastive focus specifies a subset of the given set
of discourse entities for which the backgrounded proposition holds true, but
at the same time implies a set of alternatives for which that proposition
does not hold true. The set of alternatives brought into play by contrastive
focus may be overtly given, as in (4). The most natural interpretation of this
example is one in which the contrast set on which focus operates consists of
the clefted DP A ugustos Kostantinos and the DP complement Dioklctianos
of the complex preposition e-p-ma 'instead' (lit. in the place of).
In the following set of data, the contrastive focus reading is evident from
the use of dedicated focus particles. The emphatic reflexives mawaa-n
'ourselves' in (6a) and ho 'myself in (6b) below impose an exclusive inter-
pretation on the cleft constituent. It is asserted that none of the alternatives
could provide a value for the open predication contained in the restrictive
relative clause.
Consider next the cleft construction in (7), in which the long discourse
topic DP p-nute nt-a-f-fope mn 'God who has been with (DP)' functions as
the antecedent of the clefted pronoun ntof' he'. The interpretation of this
110 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng
example is clearly not contrastive, as can be seen from the presence of the
additive focus particle on 'also, too'.
In summary, Coptic nominal clefts are not a purely contrastive focus device,
but allow for a much broader range of semantic focus interpretations (cf.
Doetjes, Rebuschi & Rialand 2004 for similar observation on the French
c 'estXP que/qui 'it/that is XP who' construction).6
This section reviews the main structural properties of Coptic nominal clefts.
We will show that clefts minimally involve two clausal projections: the
matrix clause is a nominal predicational sentence into which a relative clause
is embedded: [CPi DP PRON [CP2 RELATIVE CLAUSE]]. In this structure, the
relative clause lacks nominal properties, i.e. it is not a free relative clause.
We will also look at the categorial restriction on the cleft constituent and
the agreement behaviour of the deictic pronoun.
inserted after the focus phrase, which copies its person and number feature.
Byrne (1990) provides a single-clause analysis for that construction, in
which the focused constituent (NP) is base-generated in an adjoined posi-
tion to the CP. According to this analysis, the optionally present pronoun
represents a pronominal operator that has moved from the gap position, as
schematically represented in (9b) (labels are from Byrne (1990). 7
c. Wh-in-situ questions
ant-a u f°Pe mmo-k pe.n-tfoejs pa-no ?
REL-PERF what happen to-2SM DEF:SM.lP-lord DEF:SM-king
' W h a t happened to you, our lord and king?' (Eudoxia 36: 24)
The nominal cleft construction in Coptic Egyptian 113
d. Focus-in-situ declaratives
ont-a-f-fiße gar an nkyi p-nute han ne.f-hap
REL-PERF-3SM-change PCL NOT FM DEF:SM-god in DEF:P.3SM-law
'God has not changed in his laws. (Shenoute De Iudicio, 31: 9 - 1 1 )
In Reintges (2004a: ch. 4) and Reintges, LeSourd & Chung (2005), the
relative marking of the tense-aspect word is analysed as w/z-agreement
morphology, which flags classical w/z-constructions (relative clauses, wh-
questions, and focusing constructions) and sets them apart from pragmati-
cally neutral declaratives. Notice, however, that operator-variable construc-
tions are not automatically flagged by special inflectional morphology: the
relevant operator must also be in the appropriate configuration. As shown in
(1 la), relative marking is obligatory when the w/z-phrase appears in-situ in a
clause-internal argument or adverb position. By contrast, no such special in-
flection is resorted to when the w/z-phrase appears ex-situ in a left-peripheral
focus projection to the left of the perfect marker a-, as shown in (1 lb) below.
In the Coptic counterpart, the predicate has no such DP layer, which would
give it the syntactic distribution and referential properties of free relative
clauses. In the examples considered so far, we have treated the deictic pro-
noun and the relative clause as independent sentence constituents. The very
The nominal cleft construction in Coptic Egyptian 115
same surface string can also be found in free relative clauses, however.
Compare pronominal pe in the cleft construction in (14a) with the formally
identical definite article pe- 'the' that introduces the free relative clause
pe-nt-a-i-meewe 'what I had thought' in (14b) below.
Third, as we can see from (17a-b) below, the deictic pronoun may be
dropped when the clefted constituent is an independent pronoun. By con-
trast, it is never possible for free relative clauses to omit the definite article
or the demonstrative pronoun.
Clefts in Coptic are more restricted than the English ones in that the clefted
constituent cannot be of any other category than a DP. This categorial re-
striction is illustrated in the data in (19). Notice that Coptic has both argu-
ment (subject, direct, indirect and prepositional object) and adjunct clefts,
but due to the categorial restriction it is never possible to cleft the entire
prepositional or adverbial phrase, only the DP part of it:
c. Prepositional object
u-mnt-at-noße tie) [Cp et-u-fine nso-s
INDEF:S-NOM-sinless PRON:SF C REL -PRES-3P-seek for-3SF
n-toot-n ]
from-hand-lP
'(It is) a sinless (life) which is requested (lit. which they request)
from us.' (Testament of Isaac 233: 21)
d. Locative adverb
p-kah on pe [CP nt-a-u-kto-u ero-f ]
DEF:SM-earth PCL PRON:SM REL-PERF-3P-turn-3P to-3SM
'It (is) again the earth that they (i.e. the birds and the fish) turn
themselves to.' (Zenobius 202: 15-16)
e. Cause/reason adverb
awo nai n-tei-he mn [DP ne- [CP et eine
and DEM:P of-DEM:SM-kind with DEF:P-C REL (PRES) resemble
mmo-u ]] n(e) [cp etere p-Jatfe
DO-3P PRON:P REL(-PRES) DEF:SM-word
[Cp et seh ] tfo mmo-s etßeet-u (...)]
CREL (PRES) be.written say DO-3SF because.of-3P
'(It is) such kind of people and those who resemble them that the
Scripture word (lit. the word that is written down) says about them
(...)' (Shenoute III 151: 2 6 - 2 7 )
The deictic pronoun agrees with the clefted DP in number and gender, but
not in person. Consider the data in (20) below, all of which involve clefted
DPs.
The nominal cleft construction in Coptic Egyptian 119
(20) NP clefts with gender and number agreement of the pronominal copula
a. SINGULAR MASCULINE pe
arcu p-haqios A pa Mena pe
perhaps DEF:SM-holy Apa Mena PRON:SM
[Cp nt-a-f-kyont ero-j etbe p-eret
REL-PERF-3SM-be.wrathful about-lS for DEF:SM-vow
Lcp nt-a-i-eret mmo-f na-f ]]
REL-PERF-1S-VOW DO-3SM for-3SM
'Perhaps (it is) the holy Apa Mena who has become wrathful about
me because of the vow that I made to him.'
(Mena, Mir. l i b : 25-29)
b. SINGULAR FEMININE te
te.k-hmhal Eudoxia t[e)_ [Cp et wof e-ei
DEF:SF.2SM-servant Eudoxia PRON:SF CREL (PRES) want to-go
ehun e-pe.k-aspasmos ]
PCL to-DEF:SM-2SM-greeting
'(It is) your maid servant Eudoxia who wishes to enter to greet you.'
(Eudoxia 56: 1-2)
c. PLURAL ne
ftow n-oeik n(e) [Cp et tef na-n mmeene]
four of-bread P R O N : P C R E L ( P R E S ) be.portioned for-LP daily
Clefted third person pronouns show the same agreement behavior as clefted
DPs, as seen in (21):
(21) Third person pronouns clefts with gender and number agreement of the
pronominal copula
a. SINGULAR MASCULINE pe
ne ntof mawaa-f pe [cp e-f-anakhorej
PRET HE alone-3SM PRON:SM REL(-PRES)-3SM-retreat
hm p-ma etmmau ]
in DEF:SM-place that
'(It) was him alone who was living as a hermit in that place.'
(AP Chaine no. 181, 43: 21 - 2 2 )
120 Chris Η. Reintges, Anikö Liptäk and Lisa Lai Shen Cheng
b. PLURAL ne
[DP ne- FCp nt-a-u-mu mn pe-Khristos ]]
DEF:P REL-PERF-3P-die with DEF:SM-Christ
ntou n(e) [cp et na-onh mn pe-Khristos ]
THEY PRON:P CREU FUT-live with DEF:SM-Christ
' ( A s for) those who died with Christ, (it is) them w h o will also
live with Christ.' (Shenoute IV 4: 1 8 - 1 9 )
(22) First and second person pronoun clefts with gender and number, but
not person agreement of the pronominal copula
a. SINGULAR MASCULINE pe
ontok pfe) [cpef neu ]
YOU(-SM) PRON:SM CREL (PRES) come
'(Is it) you w h o will c o m e ? ' (Luke 7: 20)
a.' *antok antak [Cp et neu ]
YOU(-SM) PRON:2SM C r e l (PRES) come
b. PLURAL n e
onto ton nie) [Cp et tmayo ammo-tan ]
YOU(-P) PRON:P CREL (PRES) justify DO-2P
'(It is) you that justify themselves.' (Luke 16: 15)
b.' * an to ton on to ton [Cp et tmayo ammo-ton ]
YOU(-P) PRON:2P C r e l (PRES) justify DO-2P
Finally, pronoun clefts allow for the deletion of the pronominal copula
when the clefted pronoun is the antecedent of a subject relative clause. 8
DP yes no no
small clause (SC) structure (see Stowell 1981; Moro 1997 on small clause
structures in general, and Rothstein 1995; Doron 1986 on the role of pro-
nominal copulas therein).
(25) SC
b. Q-PCL » CLEFTED W H
eye u p(e) [Cp et na-fope ham p-et-fufu ]?
Q what? PRON:SM CREL FUT-happen to DEF:SM-CREU-dry
'What (is) it that will happen to the dried out one (the tree)?'
(Luke 23: 31)
However, as we can see in (29) below, topic phrases can also occur to the
right of the clefited constituent. This lower topic position hosts various types
of elements: left-dislocated personal pronouns (29a), emphatic reflexives
(29b), topicalised time adverbials (29c).
We thus propose that the left peripheral position of the cleft constituent is the
result of w/z/focus-movement in the overt syntax. See diagram (30) for fur-
ther illustration.
(30) [CP[τορΡ [FOCP DPi [FOC· [τορΡ [sc ti [sc· {pe, te, n e } [ R C ... ]]]]]]]]
ΐ I
The proposed focus fronting process is cross-linguistically well attested as
a general strategy in languages like Hungarian (E.Kiss 1987), Basque
(Ortiz de Urbina 1989), Greek (Tsimpli 1995), and Hausa (Green & Jaggar
2003). Focus fronting is also available as a marked alternative to in-situ
focus in Coptic Egyptian (Reintges 2003, 2004a). 9
126 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng
In this final section, we spell out how the cleft constituent is referentially
linked to a gap or a resumptive pronoun in the associated relative clause.
The broad syntactic distribution of resumptive pronouns in relative clauses
is illustrated in (19b-e) above. Coptic resumptive pronouns behave differ-
ently from the ones found in English type languages in that they are not
restricted to the context of island violations (see, among various others,
Demirdache 1991; Tellier 1991; Shlonsky 1992). 10
Following Demirdache (1991, 1997) we analyze resumptive pronominals
as in-situ relative operators on a par with w/z-in-situ phrases. The relative
operator undergoes w/z-movement to the specifier of CP thereby creating
the relevant operator-variable dependency. On this view the resumptive
pronoun represents the spell-out of the lower copy of the operator. The
movement configuration of resumptive pronominalization is represented in
diagram (31). See Browning (1997) for an early account of relative opera-
tors as null pronominals (OPPR0N).
. . . OPpRoN · · ·
That locality is indeed the relevant licensing condition for the in-situ
placement of the relative operator is evident from subject relative clauses in
which intervening material disrupts the adjacency between the relative
complementizer and the subject position. Consider the stacked relative
clause in (33), where the possessive auxiliary verb wanta ' H A V E ' appears
sandwiched between the complementizer allomorph ete- and the embedded
subject position. Notice that in this context a resumptive pronoun appears
in the embedded subject position.
(33) Stacked relative clause with both subject gap and resumptive pronoun
pei-rome, [CP1 et , waaß [CP2 ete-wsnta-f, mmau
DEM:SM-man CREL (PRES) be.holy REL-HAVE-3SM there
rn-pei-tncefe n-arete j]
DO-DEM:SM-multitude of-virtue
'this holy man who possessed such a multitude of virtues'
(Budge, Horn. 2: 1-2)
5. Concluding remarks
This paper dealt with the syntactic and semantic aspects of the nominal
cleft construction in Coptic Egyptian. Nominal clefts in this language were
shown to belong to a larger family of focusing constructions, where the
fronted cleft constituent corresponds to contrastive (identificational) focus
in the majority of cases, although non-contrastive, presentational focus
readings are also available. As we have shown, Coptic clefts partake in the
grammar of copular constructions. The deictic copula is merged into the
128 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng
head position of a small clause. It links the small clause subject, the clefted
NP, to its predicate, the associated relative clause. The clefted DP cannot
remain in-situ in the subject position of the small clause, but has to undergo
focus fronting to the specifier position of a designated focus phrase in the
left periphery. The underlying small clause configurationality of Coptic
nominal clefts provides a principled explanation for the categorial restric-
tion on electable elements, which can only be DPs.
Acknowledgements
This paper has benefited from detailed comments by Marcel den Dikken,
Barbara Egedi and an anonymous reviewer. The work of the first two authors
was supported by the Netherlands Organisation for Scientific Research
(NWO). All remaining errors are ours.
Notes
1. Since the Arab conquest of Egypt in 641 CE, Coptic was gradually replaced by
Arabic for most practical purposes and reduced to a mainly ecclesiastical lan-
guage, cultivated only by educated members of the Christian minority. The
appearance of Coptic grammars, vocabularies, and textual editions written in
Arabic in the 13th century CE signal a revived interest in Coptic philology by
Egyptian Christian scholars, but also the disappearance of Coptic as a spoken
language.
2. Coptic Egyptian is the linguistic outcome of widespread bilingualism within a
speech community, with Greek as the politically and culturally predominant
language. Greek was not only the language of the literate elite, but also the
language of the Holy Scriptures and the new religion and therefore a language
of great cultural importance. Although no clear statistics are available at pre-
sent, it is estimated that approximately forty percent of the Coptic vocabulary
consists of Greek loan words. The transfer of Greek lexical material was not
confined to lexical items, but also involved a considerable amount of function
words, such as sentence conjunctions, discourse markers, manner and time ad-
verbiale, and even some prepositions. Language contact phenomena at all
grammatical levels (lexicon, syntax, discourse structure) show that Coptic
should be classified as a bilingual language variety with two parent languages,
Egyptian and Greek (see Reintges 2004b for a more detailed discussion).
The nominal cleft construction in Coptic Egyptian 129
3. The following abbreviations are used in the glosses. (Glosses are given in paren-
theses for morphemes that have no surface-segmental shape): 1 'first person';
2 'second person'; 3 'third person'; ADV 'adverb', C 'subordinating com-
plementiser'; CRQ 'relative c o m p l e m e n t e r ' ; CAUS.INF 'causative infinitive';
CONJ 'conjunctive'; DEF 'definite article', DEM 'demonstrative article'; DO
'direct object marker'; F 'feminine'; FM 'focus marker'; FUT 'future'; HAB
'habitual aspect'; INDEF 'indefinite article'; INF 'infinitive'; Μ 'masculine';
NEG 'negative scope marker'; NEG.PERF 'negative perfect'; NOM 'nominalising
prefix'; PCL 'particle'; PERF 'perfect'; Ρ 'plural'; PRES 'present tense'; PRET
'preterit'; PRON 'deictic pronoun/pronominal copula'; Q 'question particle';
REL 'relative marker'; s 'singular'; TEMP 'temporal conjugation'. We distin-
guish relative markers from relative complementizers, since the former but not
the latter can also appear in main clauses. See Reintges (2004a) for the text
editions used in this article.
4. Note that the pronominal copula in nominal clefts is phonologically reduced
(as indicated by the parentheses), but not in nominal sentences. This is because
the deictic copula of clefts is a proclitic element that is attached phonologically
to the adjacent relative clause, while its counterpart in equatives is a clause-
second enclitic. The phonological reduction of the deictic copula stems from
an optional process of vowel elison (cf. (2a,c)): pe, te, ne —> p, t, η / CR|:1
et, ete, e, ere (cf. Polotsky 1962: 414 and Layton 2000: 371, §464).
5. As we can see from the grammaticality contrast between the (a) and the (b)
examples of (i) and (ii), there is a definiteness restriction in the corresponding
cleft constructions in Morrocan Arabic and Modern Hebrew.
(i) Definite restriction in Moroccan Arabic
a. L-WLAD huma lli sarrd-at (-hum) Nadia.
the-children PRON.they RM sent-she (-them) Nadia
' It was the CHILDREN that Nadia sent.' (Ouhalla 1999: 341 (5b))
b. *WLAD huma lli sarrd-at (-hum) Nadia.
children PRON.they RM sent-she (-them) Nadia
*'It was CHILDREN that Nadia sent.'
(ii) Definite restriction in Modern Hebrew
a. Dani hu Se 'azar le Dina
Dani he that helped to Dina
' It is Dani who helped Dina.' (Doron & Heycock 1999: 77 (21))
b. ??veled hu Se 'azar le Dina
boy he that helped to Dina
'It is a boy/one boy who helped Dina.' (Edit Doron, p.c.)
Ouhalla (1999: 341) hypothesizes that "the restrictive scope of the focus position
in Arabic clefts may well have to do with the involvement of the pronominal
copula", a plausible assumption given the mismatch in definiteness that would
130 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng
References
Adger, David, Susan Pintzuk, Bernadette Plunkett and George Tsoulas (eds.)
1999 Specifiers: Minimalist approaches. New York: Oxford University
Press.
Agbayani, Brian
2000 if/z-subjects in English and the vacuous movement hypothesis. Lin-
guistic Inquiry 31: 703-713.
Anagnostopoulou, Elena, Henk van Riemsdijk and Frans Zwarts (eds.)
1997 Materials on left-dislocation. Linguistik Aktuell / Linguistics Today
14. Amsterdam/Philadelphia: Benjamins.
Barbiers, Sjef, Johan Rooryck and Jeroen van de Weijer (eds.)
1998 Small words in the big picture: Squibs for Hans Bennis. HIL Occa-
sional Papers 2, Leiden: Holland Institute of Generative Linguistics.
132 Chris Η. Reintges, Anikö Liptäk and Lisa Lai Shen Cheng
Borer, Hagit
1984 Restrictive relatives in Modern Hebrew. Natural Language and
Linguistic Theory 2: 219-260.
1986 The syntax of pronominal clitics. Syntax and Semantics 19: 313-332.
Orlando: Academic Press.
Browning, Maggie
1987 Null operator constructions. Ph.D. dissertation, Massachusetts Insti-
tute of Technology.
Byrne, Francis
1990 Toward an account of preclausal focus in some Creole languages.
Linguistics 28: 661-688.
Cheng, Lisa Lai-Shen and Norbert Corver (eds.)
2005 Wh-Movement moving on. Cambridge, Mass.: MIT Press.
Chomsky, Noam
2001 Derivation by phase. In Ken Hale: A life in language, M. Kenstowicz
(ed.), 1-52. Cambridge, Mass.: MIT Press.
Demirdache, Hamida
1991 Resumptive chains in restrictive relatives, appositives and dislocation
structures. Ph.D. dissertation, Massachusetts Institute of Technology.
1997 Dislocation, resumption and weakest crossover. In Materials on left-
dislocation,, Elena Anagnostopoulou, Henk van Riemsdijk and Frans
Zwarts (eds.), 193-231. Linguistik Aktuell / Linguistics Today 14.
Amsterdam/Philadelphia: Benjamins.
Doetjes, Jenny, Georges Rebuschi and Annie Rialand
2004 Cleft sentences. In Handbook of French semantics, Henriette de
Swart and Francis Corblin (eds.). Stanford: CSLI.
Doron, Edit
1986 The pronominal copula as agreement clitic. In The syntax of pro-
nominal clitics, Hagit Borer (ed.), 313-332. Syntax and Semantics
19. Orlando: Academic Press.
Doron, Edit and Caroline Heycock
1999 Filling and licensing multiple specifiers. In Specifiers'. Minimalist
approaches, David Adger, Susan Pintzuk, Bernadette Plunkett and
George Tsoulas (eds.), 69-89. New York: Oxford University Press.
Green, Melanie and Philip Jaggar
2003 Ex-situ and in-situ focus in Hausa. In Research in Afroasiatic grammar
II, Papers from the fifth conference on Afroasiatic languages (Paris,
June 2000), Jacqueline Lecarme (ed.), 189-214. Current Issues in
Linguistic Theory 241. Amsterdam/Philadelphia: John Benjamins.
Green, Melanie and Chris H. Reintges
2004 Syntactic anchoring in Hausa and Coptic w/?-constructions. In Papers
of the Berkeley Linguistic Society, Special Session on Afro-Asiatic Lan-
guages, Andrew Simpson (ed.), 61-72. Berkeley: Linguistic Society.
The nominal cleft construction in Coptic Egyptian 133
Heggie, Lorie
1993 The range of null operators. Natural Language and Linguistic Theory
11: 45-84.
Huddleston, Rodney and Geoffrey K. Pullum
2002 The Cambridge grammar of the English language. Cambridge:
Cambridge University Press.
Kenstowicz, Michael (ed.)
2001 Ken Hale: A life in language. Current Studies in Linguistics 36.
Cambridge, Mass.: The MIT Press.
Kiss, Katalin E.
1987 Configurationality in Hungarian. Budapest: Akademiai Kiado.
1998 Identificational focus versus information focus. Language 74: 2 4 5 -
273.
Kiss, Katalin E. (ed.)
1995 Discourse configurational languages. New York: Oxford University
Press.
Koopman, Hilda
1982 Les questions. In Syntaxe de l'Haitien, Claire Lefebvre, Helene
Magloire-Holly and Nannie Piou (eds.). Ann Arbor, Michigan:
Karoma.
Lambrecht, Knud
2001 A framework for the analysis of cleft constructions. Linguistics 39:
463-516.
Layton, Bentley
2000 A Coptic grammar with chrestomathy and glossary. Porta Lingua-
rum Orientalium N.S. 20. Wiesbaden: Harrassowitz.
Lecarme, Jacqueline (ed.)
2003 Research in Afroasiatic grammar II. Papers from the fifth confer-
ence on Afroasiatic languages (Paris, June 2000). Current Issues in
Linguistic Theory 241. Amsterdam/Philadelphia: John Benjamins.
Lefebvre, Claire, Helene Magloire-Holly and Nannie Piou (eds.)
1982 Syntaxe de l'Haitien. Ann Arbor, Michigan: Karoma.
Lumsden, John S.
1990 The biclausal structure of Haitian clefts. Linguistics 28: 741-759.
McCloskey, Jim
1990 Resumptive pronouns, A'-Binding, and levels of representation in
Irish. Syntax and Semantics 23.
Moro, Andrea
1997 The raising of predicates: Predicative noun phrases and the theory of
clause structure. Cambridge: Cambridge University Press.
Ortiz de Urbina, Jon
1989 Parameters in the grammar of Basque. Dordrecht: Foris.
134 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng
Ouhalla, Jamal
1999 Focus and Arabic clefts. In The grammar of focus, Georges Rebuschi
and Lauri Tuller (eds.), 335-359. Amsterdam/Philadelphia: Benjamins.
Polotsky, Hans Jacob
1962 Nominalsatz und Cleft Sentence im Koptischen. Orientalia 3 1 : 4 1 3 -
430.
Rebuschi, Georges and Laurie Tuller (eds.),
1999 The grammar of focus. Amsterdam/Philadelphia: Benjamins.
Reintges, Chris H.
1998 The 'Highest Subject Restriction' in Coptic relative clauses. In Small
words in the big picture: Squibs for Hans Bennis, Sjef Barbiers,
Johan Rooryck, and Jeroen van de Weijer (eds), 83-89. HIL Occa-
sional Papers 2, Leiden: Holland Institute of Generative Linguistics.
2003 Syntactic conditions on special inflection in Coptic interrogatives. In
Research in Afroasiatic grammar II, Papers from the fifth conference
on Afroasiatic languages (Paris, June 2000), Jacqueline Lecarme (ed.),
363^108. Amsterdam/Philadelphia: John Benjamins.
2004a Coptic Egyptian (Sahidic dialect): A learner's grammar. Afrikawis-
senschaftliche Lehrbücher 15. Cologne: Rüdiger Köppe Verlag.
2004b Coptic Egyptian as a bilingual language variety. In Lenguas en con-
tacto de la Antigüedad a la Edad Media [Languages in contact in
Antiquity and the Middle Ages], S. Torallas Tovar (ed.). Madrid:
Consejo Superior de Investigaciones Cientificas.
Reintges, Chris H., Philip LeSourd and Sandra Chung
2005 Movement, Wh-Agreement, and apparent Wh-m-s\t\i. In Wh-Move-
ment moving on, Lisa Lai Shen Cheng and Norbert Corver (eds.).
Cambridge, Mass.: MIT Press.
Rooth, Mats
1992 A theory of focus interpretation. Natural Language Semantics 1: 7 5 -
116.
Rothstein, Susan
1995 Small clauses and copular constructions. Syntax and semantics 28:
Small clauses, 27-48. New York: Academic Press.
Shlonsky, Ur
1992 Resumptive pronouns as a last resort. Linguistic Inquiry 23: 443-468.
Simpson, Andrew (ed.)
2004 Proceedings of the twenty-seventh annual meeting of the Berkeley
Linguistics Society, March 22-25, 2001, Special Session on Afroasi-
atic Languages. Berkeley: Linguistic Society.
Stowell, Tim
1981 Origins of phrase structure. Ph.D. dissertation. Massachusetts Insti-
tute of Technology.
The nominal cleft construction in Coptic Egyptian 135
Barbara Egedi
1. Introduction
structions: one with simple definite possessions and one for indefinite, modi-
fied or deictically marked head nouns. After establishing the distribution and
the use of the two genitive constructions, I will point out the differences
between their internal structures.
The paper is organised as follows: section 2 expounds the theoretical
assumptions regarding the noun phrase research and the minimalist pro-
gram; section 3 examines the Coptic data, the distributional conditions of
noun phrases and the internal structure of the two genitive constructions
with particular attention to the nature of linking elements between the pos-
session and the possessor. In the last section, I analyse the derivation of
simple NPs and that of genitive constructions in the framework of a modi-
fied checking theory.
2. Theoretical background
The research of the noun phrase does not have as old a tradition in generative
theories as the research of the verbal constructions and the sentence as a
whole. (On the history of this research see Giorgi-Longobardi 1991.) How-
ever, since the end of the eighties several suggestions have appeared for an
analysis of noun phrases similar to that of sentences, supposing functional
categories (whose role is essentially grammatical) above lexical nouns and
their phrases. In other words, the noun phrase is actually a DP projected
from a Det(erminer), whose complement is an NP. The DP hypothesis tra-
ditionally has been ascribed to Abney (1987) but the comparison of NPs
with VPs can be traced back to Chomsky's earlier works. 2
Once this analysis was widely accepted, functional heads appeared in a
greater number in the noun phrase corresponding to the multiplied number
of functional projections in the clause. Bernstein (1991), Ritter (1991) and
others suggested that this functional FP between the lexical NP and the
functional DP should be a NumP, a projection responsible for number
specifications, and this view is going to be adopted in this paper as well.
The main motivation for supposing an intermediate FP was the require-
ment to explain the word order variations within the noun phrase across
languages and to provide a landing site for N-movement. As the approach
presented here intends to be a minimalist one, the notion of movement needs
a more accurately formulated motivation 3 in terms of checking theory. In
Genitive constructions in Coptic 139
the next section I give a short summary of the principles and methodology
of the minimalist program.
3. The Coptic NP
Traditional grammars (e. g., Till 1986: §§44-122; Lambdin 1986: passim·,
Vergote 1950) confine themselves to list exhaustively the possible orders of
constituents, the types of articles and other determiners, without trying to
explain the relationship between certain phenomena or establishing the
140 Barbara Egedi
basic phrase structure of the NP. Two exceptions to this generalisation are
Shisha-Halevy (1986, §5.1.1. and passim) - however, his view is entirely
different from my own - and Layton (2000), who is concerned with article
phrases and specifier phrases. 5
According to the DP hypothesis, argument noun phrases are DPs,
maximal projections of the functional category D (determiner). What is more,
the complement of the D is another functional category, an intermediate
projection, often associated with the number specification of the lexical NP.
However, only necessary functional projections must be built (Grimshaw
1991).
This supposition of different NP types (NPs, NumPs, DPs) is not an
arbitrary classification, it is also motivated by the syntactic distribution of
noun phrases: the various noun phrase projections distribute differently
across sentence positions (E. Kiss 2000: 124) as the evidence in 3.2.2 will
show it.
The definite article and the demonstrative have three distinct forms: in the
singular there is a masculine as well as a feminine form, while in the plural
no such morphological distinction can be found. The indefinite article has a
singular and a plural variant. The possessive article does not only mark the
number and gender of the possessed noun like the article but also the per-
son, the number and - in 2. and 3. sg. - the gender of the possessor. 6
3.2.2. Distribution
a.) in the so-called 'first present' type sentences, the subject cannot be in-
definite, more precisely, an indefinite subject must be preceded by an
existential particle:
c.) Coptic has two types of genitive constructions, in one of which definite-
ness is obligatory. In fact, the situation is more complicated, since in
this pattern the possessed noun must have a definite article but cannot
have any other determiner or modifier.
142 Barbara Egedi
Let us call the two genitive constructions pattern A and pattern B, exempli-
fied in ( 1 0 M 1 3 ) :
Pattern A
e
(10) p-sere m-p-röme
def:sg.m.-son o/-def:sg.m.-man
'the man's son'
Pattern Β
e
(11) ou-sere nte-p-röme
indef:sg.-son o/-def:sg.m.-man
'a son of the man'
e
(12) p-sere "n-sabe nte-p-röme
def:sg.m.-son adj.prt.-clever o/-def:sg.m.-man
'the clever son of the man'
c e
(15a) oua n-nen-eiote (15b)owa mmo-ou
one of(part)-our-fathers one of(part)-them
' o n e of our fathers' 'one of them'
e C
(17a) p-höb n-nek-cig (17b) hen-höb n-cig
the-work of-your(poss.art)-hand indef:pl.-work adj.prt.-hand
'the work of your hands' 'handiworks'
e
(17c) pa-höb n-cig
poss.art.-work adj.prt.-hand
' m y handiwork'
What is it that shows us that the in (17b) is not a genitive marker? First,
bare nouns have a very restricted use in Coptic, some kind of an article
usually appears with them. Adjectives, on the other hand, do not need any
article or determiner, apart from this linking element. 13 Loprieno considers
it a determinative pronoun, comparing it with that in Hebrew attributive
constructions (Loprieno 1995: 56), but let me call the reader's attention to
the fact that while after a definite noun there appears another definite article
before the adjective in Hebrew (hä- 'is hag-gädöl the-man the-great i.e. 'the
great man'), after an indefinite noun the article (or determinative pronoun)
is absent before the adjective, too ('is gädöl man great i.e. 'a great man'). In
my opinion this phenomenon can be explained as an agreement in definite-
ness, and it is far from being similar to the Coptic adjective particle, whose
appearance is indifferent to the definiteness of its noun, as can be seen in
(17b). (17a) would not be grammatical with an indefinite article. (17c) pro-
vides a clearer case where the presence of the possessive article evidently
excludes the genitive interpretation of the adjective linking particle.
144 Barbara Egedi
Perhaps this ambiguous status is responsible for the fact that some Coptic
grammars refer to both en- morphemes with the same rather neutral names
as mark of relationship (Layton 2000: §203) or nota relationis (Shisha-
Halevy 1986: 20). I prefer calling it a genitive particle (cf. Till 1986: §113:
"Genitivpartikel") or rather marker, thus indicating that it functions as a
case-marker or much alike.
It is worth mentioning that the notion of "case" in Coptic does not go
without saying. The only obvious structural case is the nominative (for its
position see Reintges 2001: 100). The object of the verb can be expressed
either by a prepositional phrase or by forming a construct state with the
verb. The latter is a procedure, which is not possible in every conjugation
type. Apart from these cases all of the relations within the sentence are
expressed by prepositions. That is why the case-marker status of the e n -
morpheme is not self-evident for me.
The "nte-, enta- of pattern Β is a better candidate for being a preposi-
tion: it has two allomorphs, the first of which was exemplified in (11)-(13).
The second form serves for introducing pronominal possessors.
e
(18) ou-sere nta-f
indef:sg.-son o/proNiB.sg.m.
'a son of his'
This view is shared by many authors (Till 1986: §113; Lambdin 1983).
Nevertheless, Layton (2000: §204) finds it questionable, arguing that it
never modifies a preceding verb or verbal clause. On the one hand there is
no doubt about its prepositional origin (Loprieno 1995: 71), on the other the
above criterion does not seem to be relevant in defining the preposition as a
category.
In the Coptic data, there is a well-observable complementary distribu-
tion between the two patterns. Although earlier grammars pointed out the
syntactic conditions of this distribution, Layton explains it on semantic
grounds. He speaks about a restrictive expansion, a construction of two
entity terms such that one restricts the meaning of the other by limiting the
number of referents to which it applies (Layton 2000: §146). In his termi-
nology, restrictive expansion has two types: the general (possessive) rela-
tionship and the appurtenance construction, but they do not correspond
exactly to our pattern A and pattern Β division. Shisha-Halevy argues that
the original opposition was essential possession vs. incidental possession or
appurtenance but it is usually neutralised and maintained in isolated cases
only (Shisha-Halevy 1986: 21). I maintain that the factors determining the
Genitive constructions in Coptic 145
choice between them are purely syntactic: the choice depends on what kind
of other N P modifiers are present. In 3.2.3., it has been shown that pattern
A is used with simple definite possessions only, while pattern Β appears
practically in every other case.
As a matter of fact, the distribution seems unusual at first sight, but we
find a similar phenomenon in the English preposed/postposed genitive alter-
nation. The preposed possessive construction {John's book, a teacher's work)
is always understood as definite. Lyons explains it by the fact that the geni-
tive phrase is in Det position, and the filling of this position forces a definite
interpretation (Lyons 1986: 138-140). In the case of an indefinite head noun,
another construction has to be used (a friend of mine), typically a PP com-
plement. This PP construction occurs when the head Ν is modified by a
demonstrative (this friend of mine). As a general rule, it is available to
make possible anything other than a simple definite possessive, which is all
the preposed construction can express. 14
The resemblance (of the English constructions) to the distribution of
Coptic genitive constructions is striking. Though Coptic has no preposed
constructions (except with a pronominal possessor), pattern A implies the
same simple definite interpretation. While in English the preposed genitive
and the definite article cannot co-occur (*the John's book) - presumably
because they occupy the same structural position, 15 the Coptic genitive
phrase in pattern A is in a postposed position, and the noun phrase has an
overt definite article.
I claim that there is a correlation between the English and Coptic data,
and in the next section I intend to provide an adequate account of this fact
and determine the exact structure of Coptic possessive patterns.
4. The structure
In 2.2. we have asserted that for being checked, formal features must be in
some local structural relation and it is this requirement that motivates
movement. First, Chomsky (1993) formulated the principle of Greed: Move
a applies to an element a only if morphological properties of a itself are
not otherwise satisfied. In other words, movement is triggered by the mor-
phological requirements of the moved element. Lasnik (1999) suggests a
revision of Greed in terms of Enlightened Self Interest (ESI): The morpho-
146 Barbara Egedi
logical requirement can be either one of the moving element (as with Greed)
or one of the position it is moving to.
At the same time Chomsky (1995) modifies his principle apparently to
the contrary, his "Suicidal Greed" insists that only the target is relevant:
Movement happens only when a category with an uninterpretable feature
attracts some feature into its checking domain, (cf. The principle Attract:
Chomsky 1995: 297)
Carstens, giving a version of Lasnik's ESI, replaces Chomsky's Attract
with the principle Move (Carstens 2000: 324), which essentially suggests
the same: the features of the target have no special status in the motivation
of movement; if a category has uninterpretable features, it may either raise
itself or "attract" raising for checking purposes. The features may be
"strong", requiring immediate checking via overt movement, or "weak", in
which case checking takes place only in covert syntax - obeying the econ-
omy principle of Procrastinate.
This extended version of the checking theory will fit our analysis of
genitive constructions, but first of all I intend to show the base structure of
the simple noun phrase in Coptic.
(19) (20)
NumP Num°
Num NP N° Num°
[β number] [a gender] [β number]
Ν
[a gender]
Genitive constructions in Coptic 147
(21) en-sney e
n-saben (22) c
n-sabeeye]9
def:pl.-brother:pl. wise def:pl.-wise.pl.
'the wise brothers' 'the wise ( m e n ) '
The indefinite articles (see the forms in (5)) are merged probably in Spec,NP,
but they have to check their uninterpretable number features, hence they
raise to the specifier of NumP, where they are in appropriate spec-head
checking relation with the N+Num head.
The definite articles are heads of the DP determiner phrase or more pre-
cisely - adopting Lyons's theory (1999) - of the definiteness phrase. As we
could see in (2), the Coptic definite article shows agreement in gender and
number features with the noun. As a matter of fact, agreement (i.e. concord)
takes place between D°, whose uninterpretable gender and number features
motivate raising to check them, and the features of N+Num head. The raising
of features takes place in covert syntax as no overt movement can be ob-
served. 20
(23) DP
148 Barbara Egedi
The pattern A exemplified in (10) provides a real challenge for our investi-
gation. We have to face several questions: what happens during the deriva-
tion ensuring that the possessed noun always be definite and at the same
time excluding any other determiner from the phrase? Where is the genitive
phrase base-generated?
Derived (deverbal) nouns (especially those expressing a process or an
event) inherit the theta-grid of the verb. The arguments of the noun should
be generated in the corresponding external/internal argument positions
(Ritter 1991; Siloni 1996). For example, the agent argument is assigned
nominative case in verbal constructions and genitive case in nominal con-
structions. The theme argument is assigned accusative case in both con-
structions. This is the case in Hebrew (26) and in Coptic, as well: 21
e
(24) p-ei ebol m-p-israel hen-keme22
the-come:inf. forth of-the-Israel from-Egypt
'the coming forth of Israel from Egypt'
If we suppose that the base position of the phrase expressing the possessor
is in the SpecNP, the combination of a genitive construction with an adjec-
tival modifier (possibly adjoined to NP 2 5 ) would yield the Possession +
Adj(s) + Possessor order:
Genitive constructions in Coptic 149
e e
(27a) *p-sere n-cabe m-p-röme
def:sg.m.-son clever o/-def:sg.m.-man
'the man's clever son'
(27b) *DP
D NumP
P-
Num NP
sere.
AdjP NP
L
n-cabe
Spec N'
e
m-p-röme
Ν
ti
However, in Coptic the possession and the possessor require direct adja-
cency in pattern A, therefore adjectives must follow both of them (which
otherwise results in structural ambiguity, since the adjective may be the
modifier of either of the nouns):
e e
(28) p-sere m-p-r5me n-cabe
def:sg.m.-son o^defisg.m.-man clever
a) 'the m a n ' s clever son'
b) 'the clever m a n ' s son'
To ensure the direct adjacency and the licensing of genitive case we have to
assume a PossP projection with an abstract possessive morpheme 26 in its
head, and with the possessor phrase in its specifier. It would explain the ex-
ceptional behaviour of the genitive marker en-, em-, which can be regarded
as a case marker on the possessor.
But we have not given account of the fact that the head noun in pattern
A cannot be indefinite or modified. It seems reasonable to assume that the
possessor phrase works as a determinant, it has a [definite] feature. There is
a similar assumption for Hungarian caseless possessor phrases, which are
claimed to have a [determininer] feature and move to the DP domain to
have it checked. 27
In Coptic, the possessor itself does not raise. Its [def] feature is weak, and
there is no overt movement - as is attested in Hungarian or English {a boy's
book). Nevertheless, the [definite] feature of the possessor has to be raised
in order to be checked by the D head. The definite article - as a default
150 Barbara Egedi
(29) DP
D NumP
def.art.
Num PossP
— Spec Poss'
possessor
[def] Poss ΝΡ
' 1
T ^
Ν
possession
The uninterpretable gender and number features of the definite article will
be checked as illustrated in (23).
Returning to the original assumption that the possessor phrase originates
in Spec,NP we could also exclude the appearance of an indefinite possession
in pattern A. I supposed earlier that Spec,NP is the starting-point also for
the indefinite articles, so the indefinite noun and the possessor of type A
mutually exclude each other. However, the possessor merged in Spec,NP
has a further [gen] case feature (beside his [def] feature), which can be
checked in Spec,PossP. The possessor phrase is raised to the PossP domain
overtly, the proof of which is given by the possession-possessor-modifier(s)
word order discussed above (see (27) and (28)).
Genitive constructions in Coptic 151
(30) DP
D NumP
def.art.
Num PossP
— Spec Poss'
possessor
[defj Poss NP
AdjP NP
modifier
Spec Ν'
possessor
[gen],[def] Ν
- possession
The possessive article which encodes the pronominal possessor has an in-
terpretable [definite] feature (see (4) and (14)) and its derivation is parallel
in manner with that of demonstratives. They are in the D-head and combine
with the N P by merge.
However, the definite feature assignment to the possessor is not so self-
evident if we take into consideration languages classified as adjectival-
genitive by Lyons (1999). From (31) it can be seen, that in Italian the pro-
nominal possessor can co-occur with a definite/indefinite article.
Two conclusions can be drawn from (31): the pronominal possessor does
not occupy the D° position since it seems to be reserved for the definite
article il, moreover, the possessiveness itself does not involve defmiteness
as a rule. However, this divergence can be due to parametric variation among
languages.
152 Barbara Egedi
Returning to genitive constructions, we can state that the cnte- phrase follows
the possession, although it can be separated from it even by a short embedded
relative clause (33).
e
(33) p-moou et-onh nte-p-ouoin29
def:sg.m.-water rel.conv.-0-alive of- def:sg.m.-light
'the living water of the light'
NumP PP DP PP
e e
nte-p-röme nte-p-röme
ou-sere pei-sere/
p-sere e n-sabe
(36) NumP
Num NP
sere
AdjP NP
e
n-sabe
PP NP
nte-p-röme
5. Conclusion
Acknowledgements
Notes
1. e.g. rules like the pronunciation of the ou sequence that has two values: a con-
sonantal [v] and a vocal [u] depending on its position within a syllable. The
transcription is based on Lambdin (1983: χ.) I initiated only one more diacritic
sign: e , which corresponds to the Coptic supralinear stroke. It is used to indi-
cate either a syllabic consonant or a short schwa before the letter in question.
The abbreviations used for grammatical expressions: acc. = accusative; circ. =
circumstantial; conv. = converter; def.= definite; dem. = demonstrative article;
foe. = focus-marker; gen. = genitive; indef. = indefinite; poss. = possessive
article; proN = pronoun; prt. = particle; ps. = present; rel. = relative.
Genitive constructions in Coptic 15 5
16. Adjectives can precede their nouns but its conditions and problems do not con-
cern us here. I suppose for now that the Ν Adj sequence is the unmarked one.
17. About one hundred nouns have a distinct plural form as well (according to
Layton to express the category of individual concrete plurality, Layton 2000:
§ 108b).
18. From son 'brother' in singular
19. From sabe 'wise' in singular (Lambdin 1983: 58)
20. This analysis follows that of the Italian determiner le in Carstens (2000: 329)
21. In reality, Coptic uses the infinitive form of the verb in both constructions. An
alternative analysis for DPs supposes an nP "shell" above the core NP - fol-
lowing the VP-shell hypothesis - and its specifier provides the position where
the possessor or the agent of derived nouns can be generated (Carstens 2000;
Radford 2000). This analysis seems to be redundant for Coptic because of the
above-mentioned infinitival character of the nominal constructions. One hardly
finds any deverbal derived noun with its full argument structure.
22. Ps 113(114): 1. (Layton 2000: §105)
23. Matt 26:12. (Layton 2000: §105) The personal pronoun is in canonical object
position. Its realization is clearly different from that of pronominal possessors
of simple nouns lacking arguments. (25b) would be grammatical only with this
latter type.
24. Ritter 1991, 39 (2b)
25. There is a general uncertainty concerning the generation of adjectives: are they
adjunctions or generated in SpecXP? For the moment I accept the common as-
sumption that they are adjoined to a maximal projection. About this problem
and a different point of view: Cinque (1995)
26. Radford (2000); this view is supported by the fact that in Hungarian (and in
Turkish) there is a visible agreement inflection on the possessed noun. Cf. E.
Kiss (2000)
27. E. Kiss (2000: 134). The Hungarian noun phrase has a relatively complex struc-
ture with real agreeing properties in genitive constructions. Its demonstration
would need a more detailed analysis than seems relevant to our understanding
of Coptic NPs. Beyond a desultory comparison, there seem to be more differ-
ences than similarities between the constructions of the two languages. (For
references, see: Szabolcsi 1994; E. Kiss 2000, 2002: Ch. 7. among others.)
28. Polotsky (1960: §33) exhaustively enumerates the possible subject types, but
these cases are practically all indefinite.
29. From p.Beroliensis 8502, 26, 20. after Till 1986, § 113
30. After Reintges (2002, 350. (9))
31. And it freely receives a default possessor interpretation (Ritter 1988: 921). In
Ritter (1991) she modifies her theory, and supposes two sources of sei: either a
realization of case assigned by Ν to an argument inside NP or the head of a DP
adjunct. (Ritter 1991: 48) The Coptic "nte-phrase does not seem to appear for
expressing an argument of the infinitival nominal constructions.
Genitive constructions in Coptic 1 57
32. In the theory of Cinque (1995) every adjective and attributive modifier in the
phrase has its own projection with the adjectives in the specifier position and
these projections can be ordered hierarchically. If there were also a so-called
c
«/e-projection, it could be positioned directly over the NP.
References
Abney, Steven
1987 The English noun phrase and its sentential aspect. Ph. D. diss., Cam-
bridge, Mass.: MIT.
Bernstein, Judy
1991 DPs in French and Walloon: evidence for parametric variation in
nominal head movement. Probus 3(2): 101-126.
Carstens, Vicki
2000 Concord in Minimalist Theory. Linguistic Inquiry 31:319-355
Chomsky, Noam
1993 A minimalist program for linguistic theory. In The Minimalist Program,
1995, Noam Chomsky, 167-217. Cambridge, MA: MIT Press.
1995 Categories and transformations. In The Minimalist Program, 1995,
Noam Chomsky, 219-394. Cambridge, MA: MIT Press.
Cinque, Guglielmo
1995 On the evidence for partial N-movement in the Romance DP. In Ital-
ian Syntax and Universal Grammar, Guglielmo Cinque (ed.), 2 8 7 -
309. Cambridge: Cambridge University Press.
E. Kiss, Katalin
2000 The Hungarian noun phrase is like the English noun phrase. In Papers
from the Pics Conference, Approaches to Hungarian 7, Alberti Gabor
and Kenesei Istvän (eds.), 121-149. Szeged: JATE Press.
2002 Hungarian Syntax. Cambridge: Cambridge University Press.
Giorgi, Alessandra and Giuseppe Longobardi
1991 The Syntax of Noun Phrases: Configuration, Prameters, and Empty
Categories. Cambridge: Cambridge University Press.
Grimshaw, J.
1991 Extended projection. Ms., Brandeis University.
Lambdin, Thomas O.
1983 Introduction to Sahidic Coptic. Macon: Mercer University Press.
Lasnik, Howard
1999 Minimalist Analysis. Oxford: Blackwell.
Layton, Bentley
1990 The Coptic Determinator Syntagm and its Constituents. Journal of
Coptic Studies 1: 79-97.
158 Barbara Egedi
Layton, Bentley
2000 A Coptic Grammar with Chrestomathy and Glossary. Sahidic Dialect.
Wiesbaden: Harrassowitz Verlag.
Loprieno, Antonio
1995 Ancient Egyptian. A Linguistic Introduction. Cambridge: Cambridge
University Press.
Lyons, Christopher
1986 The syntax of English genitive constructions. Journal of Linguistics
22: 123-143.
1999 Definiteness. Cambridge: Cambridge University Press.
Polotsky, Hans Jacob
1960 The Coptic conjugation system. Orientalia 29: 392-422.
1987 Grundlagen des koptischen Satzbaus. Erste Hälfte. American Studies
in Papyrology 28. Decatur, GA: Scholars Press.
Radford, Andrew
2000 NP shells. Essex Research Reports in Linguistics 33: 2-20.
Reintges, Chris H.
2001 Agreement marking, case assignment and the composition of the Coptic
clause. Göttinger Miszellen 180: 97-102.
2002 A configurational approach to Coptic second tenses. Lingua Aegyptia
10: 343-388.
Ritter, Elisabeth
1988 A head-movement approach to construct-state noun phrases. Linguistics
26: 909-929.
1991 Two functional categories in noun phrases: Evidence from modern
Hebrew. In Perspectives on phrase structure: heads and licensing,
Susan Rothstein (ed.), 37-62. (Syntax and semantics 25.) San Diego:
Academic Press.
Shisha-Halevy, Ariel
1986 Coptic Grammatical Categories. Structural Studies in the Syntax of
Shenoutean Sahidic. (Analecta Orientalia 53.) Roma: Pontificium In-
stitutum Biblicum.
Siloni, Tal
1991 Noun raising and the structure of NPs. MIT Working Papers in Lin-
guistics 14: 255-270.
1996 Hebrew noun phrases. In Parameters and Functional Heads: Essays in
Comparative Syntax, Belletti, Adriana and Luigi Rizzi (eds.), 239-267.
Oxford University Press.
Genitive constructions in Coptic 159
Szabolcsi, Anna
1981 The possessive construction in Hungarian: a configurational category
in a non-configurational language. Acta Linguistica Academiae Scien-
tiarum Hungaricae 31: 261-289.
1994 The noun phrase. In The Syntactic Structure of Hungarian. Kiefer,
Ferenc and E. Kiss Katalin (eds.), 179-274. (Syntax and Semantics
27.) San Diego: Academic Press.
Till, Walter C.
1986 Reprint. Koptische Grammatik (Saidischer Dialekt). 2. Aufl. 1961.
Leipzig: VEB Enzyklopädie Verlag.
Vergote, J.
1950 La phrase nominale en copte. In Coptic Studies in Honor of Walter
Ewing Crum, 229-242. (The Bulletin of the Byzantine Institute 2.).
Boston: The Byzantine Institute.
Left-dislocated possessors in Sumerian
Gabor Zolyomi
1. Introduction
2. Sumerian in a nutshell
As the elements in P4, P5, and the possessive pronominal in P3 are encli-
tics being attached to the final word-level constituent of the NP, all these
elements cumulate at the right end of the phrase in double and triple geni-
tive constructions like (1) and (2).
The Sumerian noun phrase projects the same functional projections in
the same order as has been attested in various present-day languages, e.g. in
Hungarian (cf. Bartos 2000), in such a way that complements preceede
their head, while specifiers follow it. The N P in (1) can be represented by
the tree in Figure l : 9
164 Gabor Zolyomi
K(ase)P
Κ
PossP
fDATl
Poss' Spec
NP Poss
K(ase)P
PossP Κ
IDATI
The last surface position of the Sumerian noun phrase accommodates the
case-markers. Ten case-markers can be distinguished in Sumerian: -/o/, -/e/,
-/'a/, -/ra/, -/ta/, -/da/, -/se/, -/ak/, and -/gin/. The case-markers are enclitics
that function to distinguish cases. In Sumerian cases are distinguished,
however, not solely by nominal case-markers, the verbal affixes also play
an essential role in the identification of cases.10 Three of the nominal case-
markers (-/ra/, -/'a/, and -Id), and one of the verbal affixes (/i/ in S9) are
166 Gabor Zölyomi
human non-human
ERGATIVE -Id -Id final pronominal prefix
(S10)11 and pronominal
suffix (SI3)
ABSOLUTIVE -lei -lei final pronominal prefix (S10)
and pronominal suffix (SI3)
ADVERBIAL CASES
The eleven cases can be classified into three groups: i) ergative and absolu-
tive, encoders of the subject and the object, the primary syntactic functions;
ii) adverbial cases; iii) adnominal cases which relate to no corresponding
verbal affixes.
S i l ) ; and there are three slots after the stem. Unlike in German or English
where among the participants of a verb only the subject is cross-referenced
with an affix on the verbal form, Sumerian verbal forms may cross-
reference up to four participants of the verb.
The Agent, the Subject and the Patient of a verb are cross-referenced with
pronominal affixes in S10 and SI 3. The syntactic function of their referents
is indicated mainly by their position, so, e.g., in the present-future conjuga-
tion the pronominal suffix in S13 cross-references A and S, while the final
pronominal prefix in S10 cross-references P. In the case of the other par-
ticipants, the syntactic function is indicated with one of the prefixes in the
ADVERBIAL slots (S6-S9), while the person, number, and gender of the par-
ticipant may be marked with an initial pronominal prefix in S5. The set of
pronominal prefixes in S5 contains no morpheme cross-referencing the 3rd
ps. sg. non-human participant. The pronominal prefix referring to 3rd ps.
sg. non-human participants developed from the middle-marker in S4. The
cislocative prefix in S3 functions as the 1 st ps. sg. pronoun before a dative
(see, e.g. [18] below), an oblique-locative, or an oblique-directive-prefix.
16 8 Gabor Zölyom i
If there is more than one adverbial prefix in a verbal form, then the pro-
nominal prefix in S5 or S4 specifies the person, number, and gender of the
one which is the nearest to it. The other adverbial prefixes refer to a
3rd.ps.sg. non-human participant by default. The verbal affixes cross-
referencing verbal participants function similarly to the pronouns of other
languages, so verbal participants introduced earlier into the discourse may
be encoded only with the verbal affixes in Sumerian.
The string of 14 slots identified by Sumerian philology, represented in
Table 3, suggests a left-branching, head-final sentence structure involving a
CP, an AgrSp, a TenseP, and a VP projection (and probably many more). 12
The subordination suffix in S14 functions as the head of the CP (see below
(8) for an example with this morpheme). The verbal complex is preceded
by focus and topic positions, presumably left-adjoined to CP.
CP
AgrSP C
TenseP AgrS
In (3) the possessive pronominal enlitic -/bi/ agrees in person, gender, and
number with the left dislocated non-human PR "Namnunda-kigara", which
is marked with the genitive case-marker -/ak/. In (4) the possessive pro-
nominal enlitic -/ani/ agrees in person, gender, and number with the left
dislocated human PR "Ur-tukula", which is marked with the genitive case-
marker -/ak/.
(9) G u d e a C y l A 29: 1 4 - 1 7 ( 2 . 1 . 7 )
e2-a ni2 gal-bi,
P1 e= P5 ak P1 ni P2 gal= P3 bi=p 5 0
house=GEN fear great=3NH.POSS=ABS
kalam-ma mu-ri,
kalam=' a S3 mu- S9 n- S11 ri- S13 0
In (9) the first clause contains an A G C introducing the " h o u s e " as the topic.
In the following clause the same participant remains the topic, but being
cognitively active it is expressed only with a pronominal enclitic on the PM.
The announcement of a new topic as the function of the A G C is espe-
cially clear in (4) and (6). These examples come f r o m an administrative
letter and a legal document respectively, which represent a simpler dis-
course situation than the royal inscriptions and the literary texts. (4) is the
first sentence in the message part of a letter, preceded only by the address
formula. The participant functioning as the PR, "Ur-tukula", must be there-
fore in a cognitively inactive state. The sender of the letter, however, must
have assumed that the receiver can identify it, so it should be cognitively
accessible. The same applies to (6), the very first sentence of a legal docu-
ment about Ur-Sulpae.
Syntactically, the left-dislocated PR is in one of the topic positions situ-
ated at the left periphery of the sentence. 1 5 The fact that the left-dislocated
PR is associated with a resumptive pronoun in the projection of the posses-
sor suggests that it is generated in its adjoined position, involving no
movement. This assumption is corroborated by (8) above, where the re-
sumptive pronoun is inside a relative clause showing that the topicalized
P R does not have to observe subjacency in AGCs.
172 Gabor Zolyomi
Zolyomi (1999: 231-237) showed that EPCs are also present in Sumerian.
Consider the following example (the external PR, the co-indexed resump-
tive pronominal enclitic, and the cross-referencing verbal pronominal and
adverbial prefix are underlined in the glosses):
Left-dislocated possessors in Sumerian 173
In (11) the P R of the word su " h a n d " is the god Ningirsu. The P R is, how-
ever, not in P3 of the N P whose head is the word "hand", but in a sentence
initial position. P3 is filled with a pronominal enclitic co-indexed and
therefore agreeing in gender, person, and number with the PR. The PR is
cross-referenced with the verbal prefix l\l of S9 glossed as oblique-locative
here. The pronominal prefix preceding the oblique-locative prefix in S5 is
co-indexed with the PR, but not with the PM. The PR and the PM are case-
marked with the case-markers -/ra/ and -/'a/ respectively which correspond
to the human and non-human markers of the oblique-locative case.
The same verbal expression is used in (12), but here the word " h a n d "
has no possessor. Accordingly the pronominal prefix in S4 agrees in gen-
der, person, and number with the word "hand".
(12) A W L 81 5: 1 - 4
en-ig-gal, nu-banda3, e2 zag iri-ka-ka,
eniggal nubanda=e e zag iri=ak=ak= > a
PN overseer=ERG house border city=GEN=GEN=LOC
su-a bi2-gi4
su= 5 a s4b-s9i-sion-sngi-si30
hand=OL.NH 3NH-OL-A.3SG-return-P.3NH
'Eniggal, the overseer has delivered (lit. 'returned to h a n d ' ) them
(= various sorts of wood) in the house at the border of the city.'
the PR but not with the PM as is usual. This type of EPC will be referred to
as Type A EPC.
(13) Ean. 1 4 : 2 4 - 2 6
Min-hur-sag-ra, dug3 zid-da-na,
ninhursag=ra dug zid-a=ani= 'a
DN=OL.H knee right-PT=POSS.3SG=OL.NH
mu-ni-tus
s3mu-s5nn-s9i-sion-siitus-si30
CISL-3SG-OL-A3.SG-sit-P.3SG
'She (= Inana) has made him (= Eanatum) sit on the right knee of the
goddess Ninhursag.'
( 1 6 ) En. 1 33 2: 1 3 - 3 : 6
kur-kur su-ni-se3,
kur~kur=0 su=ani=se
country~PL=ABS hand-POSS.3SG-TERM
mu-se3-gar-ra-a
S3 mu- S 5n- S8 si- S9 n- S 1() gar- s 13 0- s 14 ' a-' a
CISL-3SG-TERM-A.3SG-put-P.3NH-SUB-LOC
' W h e n h e ( = L u g a l - U R U x K Ä R ) has given all the
foreign lands into his ( = Enannatum's) hands.'
(17) Luzag. 1 1: 4 4 - 4 5
kur-kur giri3-na,
kur~kur=0 giri=ani='a
country~PL=ABS foot-POSS.3SG-OL.H
e-ni-se3-ga-a
S5 enn- S9 i- S1() n- S11 seg- S13 0- S i4 a- a
3SG-OL-A.3SG-throw-P.3SG -SUB-LOC
' ( w h e n ) h e ( = Enlil) made all the countries serve him
( = L u g a l z a g e s i ) (lit. threw all the countries to his f e e t ) . '
lations are provided with (18)-(19) to demonstrate how easily these exam-
ples translate with German EPCs. This type of EPC will be referred to as
TypeBEPC.19
Type A and Β EPCs are similar in that their predicate is a verb denoting a
motion event with a spatial endpoint. In their literal meaning these verbs
imply a physical contact. The PM of these EPCs is marked with one of the
adverbial cases whose meaning implies a motion towards an entity.
Type C and D EPCs differ from the Type A and Β EPCs in that their predi-
cate does not imply any tangible affectedness. All occurrences of these two
type come from literary texts or incantations in which a deity, a city, a tem-
ple, or a deified object is praised and is addressed directly in 2nd ps.
Left-dislocated possessors in Sumerian 179
A further distinction can be made among the EPCs in terms of the presence
or the absence of a left-dislocated lexical PR. EPCs where the external PR
is overtly present as a lexical N P will be referred to as lexical EPCs, while
180 Gäbor Zölyomi
NORMAL - + -
AGC + + -
LEXICAL EPC + - +
7. Conclusion
The present paper attempted to show that Sumerian has two different con-
structions with a left-dislocated possessor. In both of them the possessor is
positioned before the head of the possessum in one of the topic positions of
182 Gäbor Zolyomi
Acknowlegements
Notes
22. The enclitic copula after the words h e 2 - g a l 2 and g i r i z a l x are considered
to function here as focus marker. The cleft structure of the English translation
attempts to render this.
23. Another possible example of this type is Gudea Cyl. A (2.1.7) 8: 2 3 - 9 : 4,
which uses the verb ζ u "to know", and the external possessor is marked with
the comitative. See Zolyomi 1999 (a): 183 for the interpretation of this passage.
24. He uses in fact the term 'possessor raising'.
25. Constructions similar to Sumerian lexical EPCs are described by Heine (1997:
158-161) as the result of possessor specification involving the Topic Schema.
Heine states that these "constructions are occasionally described as instances of
possessor-possessee apposition. ... in languages having a system of overt case
marking, such constructions are likely to be characterized by case agreement in
that the possessee (= the specifier) receives the same case marking as the posses-
sor (= the specified). Such a situation appears to obtain, for example in many
Australian languages (cf. Dixon 1980: 293), even if possessor specification of
this type tends to be confined to 'inalienable' possession." (Heine 1997: 159)
References
Attinger, Pascal
1993 Elements de linguistique sumerienne. La construction de dui/e/di
'dire'. (Orbis Biblicus et Orientalis. Sonderband.) Fribourg, Suisse/
Göttingen: Editions Universitaires/Vandenhoeck & Ruprecht.
Bauer, Josef
1972 Altsumerische Wirtschaftstexte aus Lagasch. (Studia Pohl, 9.) Rome:
Biblical Institute Press.
Bartos, Huba
2000 Az inflexiös jelensegek szintaktikai härtere. In Strukturälis magyar
nyelvtan 3: Morfologia [Structural Hungarian Grammar, v. 3:
Morphology], Ferenc Kiefer (ed.), 653-762. Budapest: Akademiai
Kiado.
Black, Jeremy, Graham Cunningham, Jarle Ebeling, Esther Fltickiger-Hawker,
Eleanor Robson, Jon Taylor and Gabor Zolyomi
1998- The Electronic Text Corpus of Sumerian Literature. Oxford,
(www-etcsl.orient.ox.ac.uk)
Chappell, Hilary and William McGregor
1996 Prolegomena to a Theory of Inalienability. In The Grammar of In-
alienability. A Typological Perspective on Body Part Terms and the
Part-Whole Relation, Hilary Chappel and William McGregor (eds.),
3 - 3 0 . (Empirical Approaches to Language Typology, 14.) Berlin/
New York: Mouton de Gruyter.
186 Gäbor Zölyomi
Coulmas, Florian
2002 Writing Systems. An Introduction to Their Lingusitic Analysis.
(Cambridge Textbooks in Linguistics.) Cambridge, etc.: Cambridge
University Press.
Dixon, R.M.W.
1980 The Languages of Australia. Cambridge: Cambridge University Press.
Edzard, Dietz Otto
2003 Sumerian Grammar. (Handbook of Oriental Studies, Sect. I, 71.)
Leiden/Boston: Brill.
Farber, Gertrud and Walter Färber
2003 Vor einem, der auszog, ein g u d u 4 zu werden. In Sallaberger, Volk,
and Zgoll (eds.)2003: 99-114.
Geller, Mark
1997 The Last Wedge. Zeitschrift für Assyriologie 87: 43-95.
Haspelmath, Martin
1999 External Possession in a European Areal Perspective. In Payne and
Barshi (eds.) 1999: 109-135.
Hayes, John
1991 Some Thoughts on the Sumerian Genitive. Acta Sumerologica 13:
185-194.
2000 A Manual of Sumerian Grammar and Texts. (ARTANES 5.) 2nd
revised and expanded ed. Malibu: Undena Publications.
Heine, Bernd
1997 Possession. Cognitive Sources, Forces, and Grammaticalization.
(Cambridge Studies in Linguistics, 83.) Cambridge, etc.: Cambridge
University Press.
Klein, Jacob
1990 Sulgi and Ismedagan: Originality and Dependence in Sumerian
Royal Hymnology. In Bar-Ilan Studies in Assyriology Dedicated to
Pinhas Artzi, Jacob Klein und Aaaron Skaist (eds.), 65-136. (Bar-
Ilan Studies in Near Eastern Languages and Culture.) Ramat Gan:
Bar-Ilan University Press.
Krebernik, Manfred and Hans J. Nissen
1994 Die sumerisch-akkadische Keilschrift. In Schrift und Schriftlichkeit /
Writing and Its Use, 1 (1), H. Günther and O. Ludwig, (eds.), 2 7 4 -
288. Berlin/New York: Walter de Gruyter.
Krecher, Joachim
1985 Die /m/-Präfixe des sumerischen Verbums. Orientalia NS 54: 133-181.
Lambrecht, Knud
1994 Information Structure and Sentence Form. Topic, Focus, and the
Mental Representations of Discourse Referents. (Cambridge Studies
in Linguistics, 71.) Cambridge etc.: Cambridge University Press.
Left-dislocated possessors in Sumerian 187
Ludwig, Marie-Christine
1990 Untersuchungen zu den Hymnen des läme-Dagan von Isin. (SANTAG,
2) Wiesbaden: O. Harrassowitz.
Malul, Meir
1989 An Ur III Legal Document in the Possession of the Museum of the
Kibbutz of Bar-Am, Israel. Acta Sumerologica 11: 145-154.
McGregor, William
1999 External Possession Constructions in Nyulnyulan Languages. In
External Possession, Doris L. Payne and Immanuel Barshi (eds.),
4 2 9 ^ 4 8 . (Typological Studies in Language, 39.) Amsterdam/Phila-
delphia: John Benjamins.
Michalowski, Piotr
1980 Sumerian as an Ergative Language. Journal of Cuneiform Studies
32: 86-103.
Payne, Doris L. and Immanuel Barshi (eds.)
1999 External Possession. (Typological Studies in Language, 39.) Amster-
dam/Philadelphia: John Benjamins.
1999 (a) External Possession. What, Where, How, and Why. In Payne and
Barshi (eds.) 1999: 3-29.
Poebel, Arno
1923 Grundzüge der sumerischen Grammatik. (Rostocker orientalistische
Studien, 1.) Rostock: Selbstverlag des Verfassers.
Sallaberger, Walther, Konrad Volk and Anette Zgoll (eds.)
2003 Literatur, Politik und Recht in Mesopotamien. Festschrift für Claus
Wilcke. (Orientalia Biblica et Christiana, 14) Wiesbaden: Harrassowitz
Verlag.
Sjöberg, Äke
1963) giri x (= KA)-zal. Zeitschrift für Assyriologie 55: 1-10.
Sollberger, Edmond
1966 The Business and Administrative Correpondence under the Kings of
Ur. (Texts from Cuneiform Source, 1.) Locust Valley, NY: J. J.
Augustin Publisher.
Steible, Horst
1982 Die altsumerischen Bau- und Weihinschriften. 2 vols. (Freiburger
altorientalische Studien, 5.) Wiesbaden: Franz Steiner Verlag.
Thomsen, Marie-Louise
1984 The Sumerian Language. An Introduction to Its History and Gram-
matical Structure. (Mesopotamia, 10.) 3rd ed. Copenhagen: Akade-
misk Forlag.
Wilcke, Claus
1994 Die Keilschriftkulturen des Vorderen Orients. In Schrift und Schrift-
lichkeit /Writing and Its Use, 1, 1, H. Günther and O. Ludwig, (eds.),
491-503. Berlin/New York: Walter de Gruyter.
188 Gabor Zölyomi
Zolyomi, Gabor
1996 Genitive Constructions in Sumerian. Journal of Cuneiform Studies
48: 39^15.
1999 Directive Infix and Oblique Object in Sumerian: An Account of the
History of their Relationship. Orientalia N S 68: 215-253.
1999 (a) Review of D. O. Edzard, Gudea and His Dynasty. Orientalische
Literaturzeitung 94: 178-189.
2005 Sumerisch. In Schriften und Sprachen des Alten Orients, Michael
Streck (ed.), 11-43. Darmstadt: Wissenschaftliche Buchgesellschaft.
Complex predicate structure and pluralised events
in Akkadian
Christian Huber
Introduction
further byforms could be derived, among them iterative stems which are not
found in Arabic, Hebrew, or Ethiopian Semitic. The present paper concen-
trates on the G-stem, D-stem and S-stem, whose morphological properties
are briefly summarised in Table 1, the radicals (i.e. root consonants) are
schematically represented by CiC 2 C 3 (final -u in the infinitive forms is the
nominative case marker).
Table 1.
The D-stem displays an erratic behaviour with different verb types which had
always made it a troublesome task to assign it a uniform function. For ex-
ample, it brings about a transitivity or valency asymmetry, illustrated in (1).
The fact that from a great number of G-stem verbs (or their roots) a D-stem
can be derived systematically according to ( l a ) or ( l b ) (see below footnote
4) suggests that the D-stem does not come about by mere coincidence in
these cases but that it is an important derivational tool of the language.
Synchronically, thus, it would not seem unreasonable to assume that there
are not two distinct morphosyntactic processes at work in ( l a ) and ( l b )
which only happen to share the same morphological guise but that it is the
same process that applies in ( l a ) and (lb).
In the present paper I will mainly be concerned with the asymmetry in
(1) and the properties of predicates that participate in it either in the way of
( l a ) or as in (lb). As I hope to show, general assumptions in generative
grammar about the syntactic structure underlying different verb types to-
gether with certain assumptions about event structure allow to formulate
the properties of the D-stem in a way that accounts for both ( l a ) and ( l b )
while still deriving both ( l a ) and ( l b ) from the same set of properties.
In section 1 I present some data that exemplify the basic syntactic be-
haviour of unaccusative, transitive and unergative verbs in the G-stem, D-
stem and S-stem. On the basis of the observations made in section 1 I make
a proposal in section 2 as to the structural configurations underlying the D-
stem and S-stem and how they differ from each other and from the G-stem.
I suggest, in particular, that the D-stem involves a v-VP configuration (Hale
& Keyser 1993; Chomsky 1995) together with an additional property,
which later will turn out to be verbal plural. Section 3 deals with a number of
verbs or verb classes that undergo a G-stem/D-stem transitivity alternation,
with particular attention paid to the behaviour of so-called psychological
predicates. In section 4 I investigate the properties of D-stem verbs derived
from basic transitives that do not undergo a transitivity alternation and il-
lustrate the phenomenon of verbal plural as recognised e.g. by Poebel
(1939), Greenberg (1991), and Kouwenberg (1997). Section 5 is concerned
with verbal plural and its properties with different types of predicates. I
attempt to make the notion of verbal plural more precise and suggest that
verbal plural should be analysed as eventual plural. I propose an approach
to pluralisation of events inspired by Pustejovsky (1991) that recognises
different kinds of event pluralisation and pluralised events and thereby allows
us to account for the discussed instances of verbal plural in Akkadian as well
as to articulate a difference between verbal plural in the Akkadian D-stem
and in the Arabic and Hebrew intensive stems. The D-stem phenomena
discussed, among them the transitivity asymmetry sketched in (1), are ar-
gued to reduce to the type of event involved, the way of pluralising it and
192 Christian Huber
the fact that the D-stem involves a v-VP configuation. In section 6, finally,
I briefly treat some questions pertaining to the morphological make-up of
the D- and S-stems.
A few remarks need to be added here. The present article is mainly con-
cerned with the G- and D-stem and, to a lesser degree, the S-stem. I will not
deal with other stems. Likewise, I will not deal with the corresponding
stems in other Semitic languages apart from occasional remarks. It must
also be noted that in the stages of the language that are accessible to us
through the texts, the D-stem is not a fully productive grammatical cate-
gory. D-stem forms are sometimes lexicalised to a varying degree and have
acquired specialised meanings. 3 From certain verbs apparently no D-stem
could be derived at all, whereas a few other verbs seem to be attested only
in the D-stem but not in the G-stem. Lexicalised or otherwise exceptional
D-stem instances I largely ignore. 4 Nevertheless I can discuss only a lim-
ited choice of the relevant data. Due to space limitations data will some-
times be presented only schematically. The present article is thus by no
means an exhaustive account of all phenomena relevant in connection with
D-stem and S-stem.
Another point that should be kept in mind is that working with data from
a dead language is always problematic. As native speakers are no longer
available to provide judgements or comments and all evidence comes from
the surviving (and, discovered, evaluated, and accessible) written texts, a
good deal of the finer (or even coarser) grained nuances will almost neces-
sarily escape us and many a question regarding interpretation, productivity,
grammaticality or the like will remain open to speculation.
Finally, I will not speculate on the historical development of the D- and
S-stems.
In this section I will briefly illustrate some basic syntactic properties of the
D- and S-stems when applied to intransitive and transitive G-stem verbs.
With intransitives I will make the familiar distinction between unaccusa-
tives and unergatives. Unaccusatives are intransitive verbs whose surface
subjects are underlyingly objects. Unergatives are intransitive verbs whose
surface subjects are also underlyingly subjects. The D-stem is sensitive to
this distinction in that it brings about a transitivity alternation with unaccu-
satives that is not met with transitives and unergatives.
Complex predicate structure andpluralised events in Akkadian 193
The adjectival roots in (2) surface in the G-stem prefix conjugations as in-
transitive predicates denoting the process that leads to, or increases the de-
gree of, the state described by the respective adjective. Both the D-stem and
the S-stem render 2-participant forms. The D-stem gives rise to transitive
predicates. The S-stem, employing the causative morpheme s, yields causa-
tive predicates.
Vspl sapälu 'be(come) deep, suppulu 'deepen, l o w e r * ' suspulu Ί./c. χ (to) be(come)
low' deep, low'
Vsrh (sarähu) (Stative o n l y : surruhu ' m a k e χ glorious, susruhu Ί./c. Λ: (to) be(come)
be glorious, magnificent' glorious, magni-
magnificent) ficent'
From the roots in (4) both the G-stem and the D-stem yield transitive predi-
cates featuring a subject and a direct object, whereas the S-stem yields dou-
ble-transitive, causative predicates with the causer surfacing as the subject
in nominative case and the causee and the object surfacing with accusative
case.
The roots in (7) give rise to verbs denoting body functions or utterances of
sound, thus providing good candidates for unergatives. These verbs pattern
with transitives in that there is no increase of valency in the D-stem but
only in the S-stem. 6
Complex predicate structure andpluralised events in Akkadian 195
VsM sa 'älu "cough' su ' 'ulu 'cough' (St: 'be caused to cough')7
Vsrt sarätu 'fart" surrutu 'fart' —
Among the unergatives listed in (7) some appear in both the G-stem and the
D-stem, whereas others are attested only in one or the other. In those cases
where the verb can appear in both stems, however, the D-stem never gives
rise to causativised versions of the G-stem, which can only be derived by
the S-stem.
The comparison of (2) and (4)/(7) shows that the D-stem turns the G-
stem unaccusatives into transitives, while it does not extend the valency of
unergative verbs or verbs which are already transitive in the G-stem. This
sets it apart from the S-stem, which derives causative forms from G-stem
unaccusatives as well as G-stem transitives and unergatives. The resulting
surface patterns are summarised in Table 2.
Table 2.
Considering the data in section 1 it can be observed that the D-stem is pre-
sent only with an external argument. Since unaccusatives become transitive
in the D-stem whereas the valency of transitives and unergatives is not ex-
tended it can be concluded that the D-stem adds something to the structure
of unaccusatives which allows the introduction of an external argument and
which is already present with basic transitives and unergatives. However,
the fact that the D-stem can also be applied to verbs that already have an
external argument in their basic form suggests that it provides not only
some element that allows for an external argument but that the D-stem has
some additional property.
In sum this indicates that the D-stem is a complex phenomenon that
involves some element which is only present with an external argument but
also involves yet another element or feature. I will consider both issues in
turn. In the remainder of this section I will discuss the configurational as-
pects of the D-stem and the S-stem. What sets apart G-stem transitives
from their D-stem counterparts will be discussed in section 4.
VP vP
EXT
VP
(internal structure of VP omitted)
I will therefore propose that in Akkadian, the D-stem always involves a v-VP
configuration. It transfers G-stem unaccusatives into a v-VP configuration,
thereby making them into transitives with an external argument. G-stem
transitives and unergatives, in contrast, already involve a v-VP configura-
tion and an external argument in their basic form. Therefore their configu-
ration remains unchanged in the D-stem and valency is not increased.
The fact that unaccusatives are turned into transitives in the D-stem but the
valency of transitives remains unchanged suggests that in the latter case some
additional property of the D-stem (to be dealt with in section 4) visibly takes
effect, hinting at its complex nature. Configurationally, this complex nature
of the D-stem can be captured in two ways: (i) as involving another func-
tional head in addition to little ν that is responsible for some additional
property of the D-stem, or (ii) as a little ν with some special feature.
There are two options for a v-VP configuration in combination with
another head, depending on whether all non-theta-related functional projec-
tions are required to be above vP (Chomsky) or whether a non-theta-related
functional projection is allowed to intervene between the two portions of a
double VP (e.g. Aspect in Baker 1996, among others).
In (10a) a functional head F selects a vP complement, whose head ν in
turn takes VP as a complement. In (10b) ν selects FP and F selects VP so
that F is sandwiched between the upper and the lower portion of the v-VP
configuration. However, the approach (10c) would also allow for F to take
a VP complement without having been selected by v, resulting in an unac-
cusative VP that is complement of F. This raises the questions of what the
appearance and properties of such a construction would be. The issue will
be returned to in section 4.
198 Christian Huber
vP vDP
EXT EXT vD
VP VP vD VP
Taking the D-stem to involve a v-VP configuration also accounts for the
difference between the D- and S-stems with respect to their causativising
effect. Little ν is a theta-relevant head that allows the addition of an exter-
nal argument. Causation via ν is relevant only in that part of the structure
that affects the thematic properties of a predicate. In contrast, causation via
the S-stem need not interact with the thematic properties of a predicate. It
may thus embed predicates that appear thematically "closed" in the G-stem
under a causative event and introduce a causer argument, irrespective of
whether or not the embedded predicate contains an external argument. With
Complex predicate structure andpluralised events in Akkadian 199
b. sa saqqulim usaqqil-ma
R E L pay(D).inf.GEN pay(D).pret.3sgm-MA
' H e paid what (there was) to pay.' ( O I P 27: 57 2 5 )
(12a/b) and (13a/b) illustrate once again the G-stem/D-stem valency asym-
metry, (12c) and (13c) show that the predicate embedded under the S-stem
causative may be transitive (in (12c)) as well as unaccusative (in (13c)).
Unergatives, for which I omit examples here (see (7) above), pattern with
the transitives in (12). Since the predicate embedded under the S-stem
200 Christian Huber
(14) sP
contexts the D-stem could not be used as a matrix causative for verbs that
are transitive in the G-stem, which again indicates that the D-stem is re-
stricted to a thematic configuration with only a V P complement, whereas
the S-stem is not restricted to a thematic configuration in that it may also
take complements larger than VP, such as vP.
The fact that the D- and S-stems seemingly could be interchanged with a
number of G-stem unaccusatives (beside the fact that some take only one or
the other) also indicates that with unaccusatives, the S-stem apparently did
not bar an agentive interpretation of the causer argument and that in such
cases a s - V P configuration could be interpreted thematically equivalent to a
v-VP configuration, both introducing an agent/causer argument where there
is none in the G-stem.
With verbs that appear to be ambiguous between a non-agentive and an
agentive interpretation in the G-stem such as elü 'rise, go up, be(come)
high(er)' in (17), causativisation seems to differentiate between the readings
in so far as the agentive version, which arguably contains an external argu-
ment in Spec,vP, is targetted by the S-stem and not by the D-stem. To convey
causative meanings of the verb in (17a) like ' m a k e someone go up (to a
higher location)', ' m a k e a person come forth (as a witness in a law suit)',
etc., apparently only the S-stem could be used. D-stem forms as in (17b)
rather seem to pattern with D-stems of unaccusative deadjectivals.
b. resT-su lü ulli
head.ACC.cstr-possCL3sgm indeed raise(=D).pret.lsg
Ί raised its (=wall) s u m m i t . ' ( L I H 57 i 17)
That is, the D-stem version rubbu 'enlarge x' of the G-stem rabü 'be(come)
big(ger)', for example, or the D-stem ullü 'raise x' of the G-stem elü 'rise,
go up' apparently could not be morphologically further causativised,
making it impossible in Akkadian to express something like John
had/let/made Mary enlarge the springs or John let/made/had Mary raise
the wall's summit by means of verbal morphology. 13 The issue will be re-
turned to in section 6.14
Only few motion predicates appear to have a D-stem. If so, they are turned
into transitives with an agent argument. One example is given in ( 2 0 ) . 1 5
The stative predicate edü/wadü 'know' does not allow for a double object
construction in the D-stem in most dialects. Again the D-stem adds an
agentive argument but the resulting verb's internal domain may contain
only one of the two arguments found in the G-stem (predominantly the
theme).
The verb hasäsu 'remember' allows for two different D-stem constructions.
One appears to be an extension of the G-stem version, to which an agent/
causer argument is added (see (25 i)). The D-stem verb here takes as its object
the entity whose memory is being affected (and occasionally the target of
remembering as a second object). The other D-stem construction, (25 ii), still
involves an agentive subject but rather denotes a conscious mental activity
directed at some target.
In all three cases, lamädu 'learn', edu 'know' and hasäsu 'remember', the
G-stem versions are non-agentive verbs whose valency seems increased in
the D-stem by addition of an agent/causer argument with the option of re-
taining the original arguments of the G-stem verbs in at least two cases.
This might suggest that in the G-stem versions none of the arguments oc-
cupies a VP-external position, be it a bare VP or a v-VP structure lacking
an argument in the specifier of vP. The case of edü ' k n o w ' is rather remi-
niscent of the psych-verbs in (21) in that in the general case only one of the
two arguments of the G-stem version may be retained in the D-stem. Ques-
tions arise also with respect to the second D-stem construction of hasäsu
'remember' in (25ii). All mentioned cases need closer investigation. Both
the psych-verbs in the D-stems and the D-stem versions of the verbs of
knowledge, however, clearly come in transitive structures with an agent or
causer argument as the subject that is not present in the G-stem.
The question of the actual G-stem structure(s) also arises with verbs
such as those in (26) that seem to oscillate between a reflexive agentive and
a non-agentive or stative reading. In (26) the D-stem targets only the
stative/non-agentive reading of the corresponding G-stem verb, to which it
adds an agent or causer argument, allowing D-stem forms with a complex
internal domain. 1 6
G: labäsu D: lubbusu
(i) ' w e a r y (=ACC)' 'dress * (=ACC) (with y (=ACC))'
(ii) 'put o n y ( = A C C ) \ (* ' m a k e * put o n / )
'dress oneself ( w i t h y (=ACC))' (* ' m a k e * dressx-self ( w i t h y ) ' )
Complex predicate structure andpluralised events in Akkadian 205
Let us now return to the question of what sets D-stem transitives like those
in (4), partly repeated in (27), apart from their G-stem counterparts. Recall
that G-stem transitives such as those in (4)/(27) are not causativised in the
D-stem but that their valency remains unchanged.
Kouwenberg observes that the D-stem is frequently used with plural (or
mass-noun) direct objects, but that it occurs also with plural indirect objects
or adjuncts, occasionally also with plural subjects or in the case of durative
or repeated action. Yet, the use of the D-stem is not obligatory in these
cases. The G-stem is free to occur with a plural subject, object or adjunct.
However, the D-stem is always used in case of a contrast of singular vs.
plural. Kouwenberg gives examples like those in (28) to (34).
In (28) the contrast of the use of the G-stem with a singular direct object
(28a) and the use of the D-stem with a plural direct object (28b) is illustrated.
206 Christian Huber
b. närätim upetti
river.plf.ACC open(D).pret. lsg
Ί opened canals.' (RIME 4, 603: 47f.; Kouwenberg p. 120)
The same contrast is also found under passivisation, where after NP-
movement an underlying object NP surfaces as the subject in nominative
case. In (29a) the root Vbtq 'cut (off/through)' occurs in the N-stem, which
serves as a passive of the G-stem, whereas in (29b) it occurs in the Dt-stem,
which serves as a passive of the D-stem.
b. butuqätu ubtattaqä
sluice channel(?).plf.NOM cut(Dt).pres.3plf
'the sluice channels(?) will be cut t h r o u g h "
(YOS 10, 26 iii 29; Kouwenberg p. 123)
There are only few attestations of the D-stem with a plural subject and a
singular direct object, as in (33) (see also (34) below).
c. sarräku
king.statlsg
Ί am k i n g ' (c. sarru ' k i n g ' , anäku Τ)
On the surface, at least, this suggests that in contrast to (38a) the "holder"
argument in (38b) is no external argument but originates internally or per-
haps expresses a different theta role. Note that this "active" use of the stative
conjugation is not possible with a verb like write.
Under the assumption that the D-stem creates basically agentive, transitive
forms, no "active" D-stem statives are expected to occur, D-stem statives
are expected to appear in the "regular" stative construction (then resembling
passives in that the patient/theme cannot remain in the object position).
That is, if (38b) were put in the D-stem one would expect a case alternation
to the effect that the holdee appears as the surface subject in nominative case
and the holder would be allowed to surface only as a όν-phrase-type adjunct.
This state of affairs is indeed found. The G-stem example (41a) is an in-
stance of the "active" stative construction presented schematically in (38b).
In (41a) the "holder", filament (singular), appears in the nominative and the
"holdee", epigastrium, in the accusative case. In (41b) however, which
differs from (41a) only in that the "holder" is plural and accordingly the
210 Christian Huber
verb is in the D-stem, we find the reverse situation. N o w the "holdee", epi-
gastrium, surfaces in the nominative case whereas the "holder", filaments
(plural), appears as a fry-phrase type adjunct in the instrumental accusative,
as schematically illustrated in (40a). 2 '
The data in (41) merit some comments. In (41a) the stative verb agrees with
qü ' f i l a m e n t ' . In (41b) the stative verb agrees with res libbi 'epigastrium',
which follows from the fact that it is inflected for third person masculine
singular and not plural, as would be required for agreement with a plural
'filaments'. There is no morphological marker of nominative or accusative
visible on res libbi 'epigastrium (lit. 'head of the inside')' because the noun
on which it would appear (resu ' h e a d ' ) is in the construct state. Instru-
mental accusative is morphologically identical to the 'regular' accusative as
the object case, which gives the construction a strange flavour at first sight,
together with the fact that word order remains unaffected. The case (and
number) alternation, however, is visible on qu. 'filament'. Word order re-
mains unaffected by the case alternation because res libbi 'epigastrium' is
topicalised in both (41a) and (41b), preceding the focus portion of the
clause (minimally, qu/qe 'filament(s)').
The change of construction in (41) seems to be due to the fact that in the
D-stem version of the verb the "holder" argument must originally occupy
the Spec,vP position (whereas the subject-NP of "active" statives perhaps
represents an argument that cannot occupy Spec,vP). Due to the passive-
like properties of the stative construction the "holder" can appear only as an
adjunct whereas the object, the "holdee", as the internal argument of the
corresponding verb becomes the surface subject.
Complex predicate structure andpluralised events in Akkadian 211
G: aräku D: urruku
' b e c o m e long' 'lengthen x'
(not: ' b e c o m i n g long of many
'become extremely long', etc.)
c. G: kasädu D: kussudu
'arrive, reach y ( = A C C ) ' 'chase (away) JC' (lexicalised) 2 2
(not: 'arriving of many x\ 'arrive many
times/at many places', etc.)
The fact that there is no principled reason apparent why verbal plural
should be excluded with unaccusatives like the verbs in (42) (say, with a
plural surface subject) supports the assumption that the verbal plural prop-
erty of the D-stem is linked to a v-VP structure. 23
At this point we are able to address a question raised in section 2. In
section 2 it was argued that the D-stem is a complex phenomenon that in-
volves a v-VP configuration and yet another property, which in the present
section was identified as verbal plural. Two ways were sketched regarding
how the complex nature of the D-stem could be captured configurationally:
(i) as involving another (non-theta-relevant) functional head F in addition
to little v, or (ii) as a little ν with some special feature, indicated as v D , F or
VD now assumed to be responsible for "verbal plural". While v D is perfectly
compatible with the claim that a v-VP configuration is an integral part of
the D - s t e m ' s outfit there remains the question of the structural position of a
potential head F. Two options were outlined, repeated here for conven-
ience: merger of F above vP (10a), or a head F sandwiched between the
212 Christian Huber
two portions of the ν-VP configuration (10b). It was noted that option (10b)
would also allow for F to take a V P complement without having been se-
lected by v, resulting in an unaccusative V P that is complement of F, rais-
ing the question what the properties of such a construction would be.
Another issue that becomes relevant at this point is the fact that beside the
D-stem Akkadian has another set of candidates for some verbal plural-like
property, namely the so-called iterative stems that can be derived from any
of the four main stems G, D, S and N. The properties of iterative stems
differ somewhat from those of the D-stem. For example, in contrast to D-
stem versions of G-stem unaccusatives, iterative stem versions of G-stem
unaccusatives remain unaccusative, see e.g. the iterative versions of abälu
'become dry' and maräsu 'become ill' in (43), predicating of some entity
repeated becoming dry or becoming ill, respectively.
However, the relation between the D-stem and the iterative stems as well as
many issues regarding their properties and lexical distribution still remain
to be investigated in more detail. I will therefore not further deal with issues
such as more precise positions of functional projections, potential inter-
actions, or feature checking or agreement mechanisms.
Before proceeding another fact needs to be mentioned: while in the gen-
eral case the appearance of the D-stem of transitive verbs is visibly connected
to the presence of some pluralic element, say, a plural object, there is no
such requirement for the D-stem versions of unaccusatives, as illustrated in
(44), featuring rubbü 'make big, raise (a child)', which is the D-stem of
rabü 'become big', and urruku 'lengthen', which is the D-stem of aräku
'become long'. In both cases the verb appears with a singular object (and
subject).
The question why there is no plural requirement for the object of a D-stem-
transitivised former G-stem unaccusative will be addressed in section 5.
214 Christian Huber
The crucial question now is: how can (45i) and (45ii) be related? The answer
must have to do with the mechanisms behind "verbal plural" and its effects
with different types of predicates.
Following Greenberg 1991 and Kouwenberg 1997 I took the D-stem to
be associated with "verbal plural". In section 2 it was proposed that the D-
stem reflects a functional head that co-occurs with an external argument,
either some functional head F that is associated with ν in a v-VP configura-
tion or a special type of v, indicated as vD. Under this assumption "verbal
plural" as met in the Akkadian D-stem must therefore be encoded in vD or
F, which in turn must associated with the Number system, perhaps also
associated or interacting with the aspectual system.
In the present section I will try to make the notion of verbal plural more
precise and propose an approach that allows us to account for the properties
of D-stems of transitive and unaccusative verbs as well as for some differ-
ences in the properties of the Akkadian D-stem and its counterparts in Arabic
and Hebrew. For the moment I will largely limit my attention to transitive
verbs as discussed in section 4 and unaccusatives as presented in section 1
(e.g. grow, lengthen, etc.), admitting that widening the scope of investiga-
tion will certainly refine the picture. I will also point out some differences
between Akkadian on the one hand and Arabic and Hebrew on the other
hand.
Verbal plural will then be interpreted as "many events of the same type". I
will therefore propose that the D-stem pluralises (sub-)events. Morphologi-
cally the pluralic value of the D-stem is iconically encoded via gemination
of C 2 . Gemination may be regarded as an instance of reduplication, i.e., in
the simplest case, repetition.
Recall from section 4 that the basic G-stem may occur freely with a plural
subject, object or adjunct. That is, it is unspecified with respect to "verbal
number". The difference between the G-stem and the D-stem can thus be
characterised in the following way:
(46) a. S b. Τ
I A
e -'e e
For present purposes let me propose that the D-stem pluralises transitions, i.e.
it creates a series of transitions of the same type. In addition, I will assume
that transitions may differ with respect to the (type of) state in which they
result so that there may be several types of simple transitions. For exposi-
tion, let us consider three types of predicates.
First, consider predicates such as (i) build, break(-in-two) or (ii) open,
close. These predicates express a binary opposition. The transition brings
about a terminal state predicated of some entity that is evaluated relative to
its opposition and cannot be further subdivided. The resulting state has to be
reversed before it can be brought about anew with respect to the same entity.
216 Christian Huber
-s
ι affected affected
That is, some entity may be affected by, say, hitting, and its having been
affected by hitting will remain part of its further history, but nevertheless
that entity can be affected by hitting again, which will again remain a part
of that entity's further history, and so on. Thus, it is possible to say that
John beat the dog three times whereas it is not possible to say that *John
killed the dog three times, assuming the dog stands for the same individual
in all three killing instances.
Finally, consider predicates such as grow and lengthen. These predicates
do not express a binary opposition (in absolute terms). The transition does
not result in a state that can be evaluated relative to its opposition, but
rather results in a (change of) degree which is evaluated relative to another
degree on a scale, predicating of an entity a (change of) degree (i.e. progress
Complex predicate structure andpluralised events in Akkadian 217
on a scale), whereby the interval between the degrees can be infinitely sub-
divided.
Informally put, again, growing of, or lengthening, some object does not
necessarily imply that the object has been short before, it may still have
been longer than any other object in a given comparison set. However, its
degree of longitudinal extension increased (or keeps increasing), resulting
in an interval, which, if mapped onto a scale at some time of inspection,
will yield a distance between two points respresenting the degrees of longi-
tudinal extension at the beginning and at the end of the inspection interval
such that degree a < degree b whereby the interval between the two degrees
can be infinitely subdivided. Thus, no absolute state (like e.g. existence) is
evaluated relative to a corresponding non-state (like e.g. non-existence) but
rather one degree of longitudinal extension relative to another degree of
longitudinal extension.
In Akkadian, verbs of types (47) and (48) are generally realised as tran-
sitives in the basic stem, verbs of type (49) are always realised as unaccusa-
tives in the basic stem. 25
Table 3.
Let us now turn to the question how transitions may be pluralised. In the
simplest case, pluralising transitions has the effect that the number of rele-
vant transitions be not restricted to one. We may thus think of a series of
transitions of the same type. 26 However, there may be series with different
properties. Consider the following three types of transition series. Type I,
illustrated in (50), is a series of independent transitions, the output of one
transition does not serve as input for the next transition.
218 Christian Huber
Type II, illustrated in (51), is a recursive series, the output of one transition
serves as input for the next transition.
Verbs may of course be assigned to more than one event type. For example,
a verb like punch may be treated according to (48) predicating affectedness
by punching or as "punch-do, give a punch" etc, then corresponding to (47).
With (overt or covert) three-place predicates, thus, a number of options
come into play. Through interaction with the particular way of pluralising
the relevant (sub-)event various interpretative possibilities arise.
Instances of verbal plural as illustrated in section 4 result from which
type of transition is pluralised by means of which type of series. However,
not all possible combinations occur in Akkadian. Semitic languages appar-
ently differ as to what combinations they allow. As I will argue later, this
accounts for the fact that "intensive" D is largely absent in Akkadian.
Complex predicate structure andpluralised events in Akkadian 219
(32) ( . . . ) at (a measure of) 2 PI each I marked out a plot of land for the
citizens of Nineveh.
(34) the wall which R(oyal) N(ame), RN 2 (...) (and) RN 5 (...) had built (...)
(42) a. enlarge χ
b. lengthen χ
etc.
Consider now eventual plural with affect-type verbs. The state resulting
from an affect-type predicate does not have to be reversed before it can be
brought about anew with respect to some entity. In Akkadian eventual plural
mostly brings about a type-I-series here that is interpretable only with plural
count nouns (or mass nouns), to the effect that the affected state is predi-
cated of a multitude of entities as the affected state resulting from each
transition must be predicated of a different entity. The option of multiply
predicating an affected state of the same entity seems to be realised only
marginally in Akkadian. One possible instance is (33), which is a case of
the D-stem with a plural subject and a singular object. Another potential
instance is (35), if to be understood as something like "look at χ many
times".
For unergatives several analyses are available. In Hale and Keyser's (1993)
analysis of unergatives as concealed transitives involving an incorporated
bare N, unergative verbs such as work or dance are underlyingly composed
of something like 'work-do' or 'dance-dorespectively. Thus, a terminal
change of state may be predicated of the underlying nominal component of
an unergative, which however has mass noun-like properties in that it re-
sists counting. This might explain why discernible differences between the
G- and D-stem of unergatives are largely absent, and may also be the rea-
son why many unergatives only occur either in the G-stem or the D-stem.
Alternatively, an unergative verb may be treated as involving a terminal
change-of-state transition predicated of a single entity such as 'act of swal-
lowing', 'act of barking', etc., whose coming about is denoted by the verb.
In this case pluralising the event will yield a repetitive interpretation ('swal-
low, bark, etc. several times'). Potential instances are (36b) (bark) and (30)
(swallow).
Complex predicate structure andpluralised events in Akkadian 221
Above it was said that in Akkadian, the D-stem generally does not yield an
"intensive" interpretation as is often found with the corresponding stems in
Arabic and Hebrew. The approach just outlined allows us to make the
somewhat intuitive notion "intensive" more precise and at the same time
determine in which respects Arabic and Hebrew differ from Akkadian. I
will consider two cases. First consider the Arabic data in (53), where x
stands for a singular count noun object.
In (53b) the "intensive" interpretation may come about in two ways: (i) by
predicating a series of independent "affected" states ("affected by biting/
tweaking") of a singular count noun entity, which yields an interpretation of
repeated affectedness by biting or tweaking as opposed to a single affected-
ness by biting or tweaking, or (ii) the output of one transition serves as
input for the next transition so that the resulting states (or rather, degrees)
add up, yielding an interpretation of increased affectedness by biting or
tweaking (ie. an increased degree of affectedness).
Next, consider the Arabic and Hebrew data in (54).
Under the approach taken here, the "intensive" interpretation in (54b) re-
sults from a series of binary terminal state transitions predicated of a singu-
lar count noun entity, whereby (i) each resulting state is predicated of the
same singular count noun entity, or (ii) the output of one transition serves
as input for the next transition (i.e. in both cases the singular count noun
entity is interpreted in a mass noun fashion). 3 1
Akkadian seems to lack options (53bii) and (54b) completely and to
allow for option (53bi) only marginally. The difference between Akkadian
and Arabic/Hebrew can therefore roughly be stated as follows. Akkadian
does not allow the predication to a singular count noun entity of the state(s)
222 Christian Huber
Let us now return to the question how the D-stem effects, for convenience
repeated in (55), can be related.
Table 4.
intransitive transitive
grow-type verbs G-stem D-stem no transitive graw-type
verbs in the G-stem
Despite the fact that the SD-stem combines the morphological characteristics
of both the D- and S-stems, however, the respective forms never yield an
interpretation of a pluralic transitive verb embedded under a matrix causa-
tive, nor do they yield causativised versions of transitivised unaccusatives.
Rather, the attested instances merely correspond to the D- or S-stems of the
respective verbs (see Kouwenberg 1997: 336ff for a detailed discussion).
To the extent that the issue is real, I will submit tentatively that the reasons
for this might have to do with morphology and the interpretability of mor-
phological encoding. Let us therefore take a brief look at the morphological
composition of the D- and S-stems.
The fact that the D-stem combines two seemingly unrelated properties -
transitivity and eventual plural - strongly suggests that it results from the
conflation of two originally distinct categories. This view receives support
from the D-stem's morphological make-up, although I will not speculate
here about what historical development(s) may have led to this situation.
Let us consider the morphological characteristics of the D-stem: (i) it
displays a specific "^-coloured" vowel pattern, and (ii) it displays gemina-
tion of the middle radical, which occurs in all tenses. However, neither of
these characteristics is restricted to the D-stem. As illustrated in Table 5,
gemination of the middle radical is found not only in the D-stem but also in
"present tense" of G- and N-stems. 33 The specific "w-coloured" vowel pat-
tern is common to both D- and S-stems.
Table 5.
gemination of C2
(G present) iPaRRa/i/uS
u-coloured (D, S present) uPaRRaS usaPRaS
vowel pattern (D, S preterite) uPaRRiS usaPRiS
s prefix
Complex predicate structure andpluralised events in Akkadian 225
From this it follows that (i) gemination of Cj does not automatically trigger
the special «-coloured vowel pattern, and (ii) the special «-coloured vowel
pattern does not automatically trigger gemination of C 2 . That is, «-vocali-
sation and gemination of C2 do not depend on each other.
Also, each seems to represent a different property. The fact that gemina-
tion of C 2 is availabe for present tense also of unaccusatives and other types
of non-agentive and stative predicates suggests that gemination is not a
morphological indicator of a transitive or causative relation or the presence
of an external argument or some other causer, despite being compatible with
its presence. Similarly, the fact that the «-vocalisation pattern is also a mor-
phological feature of the S-stem, which serves as a causative of transitives,
unergatives and unaccusatives and does not display verbal plural properties
as those witnessed in section 4, suggests that the «-vocalisation is not a
morphological indicator of verbal plural, despite being compatible with it.
However, it can be seen that the «-vocalisation occurs when some notion of
causativity is involved, as it is found in both the D- and S-stems.
In sum all this indicates that the two morphological characteristics of the
D-stem do not form a unit in only together expressing a single, unique
property but rather hint at the D-stem being a conflation of two originally
distinct categories. It may therefore be assumed that the two effects of the
D-stem - transitive/causative, eventual plural - correspond to the two mor-
phological characteristics under discussion, namely gemination of the middle
radical and the specific vowel pattern, as indicated in (59).
and usapras from the preterite forms uparris and usapris, respectively. In
(60) eventual plural is morphologically encoded by gemination of the middle
radical, the transitive/causative value is encoded by the vowel pattern, more
precisely the prefix vowel [u]. In (61), where no gemination is present, 35
the causative value is doubly encoded: once by the vowel pattern (again,
more precisely the prefix vowel [u]) and once by the causative prefix s.
vprs
••' valency-sensitive \
apophonic vowel: tense"-sensitive
a=>u: increase of apoph. vowel: )
valency i=>a: preterite => present ./'
(causative) ...·•''
from G-stem unaccusatives on the one hand and verbal plurals derived from
G-stem transitives on the other hand) might have to do with the fact that the
causative value of the S-stem is morphologically doubly encoded. Suppose
the D-stem requires u of the vowel pattern to be interpreted as internal
causative (i.e. transitive), whereas the S-stem requires u of the vowel pat-
tern to be interpreted as one component of the external (i.e. matrix) causa-
tive, the other component being the prefix s. Interpreting a SD form as a
causativised D-stem form would then require to interpret u at the same time
as indicating internal causative (i.e. transitive) as well as external (i.e. ma-
trix) causative. Assuming that the morphological marker u (or the operation
of apophony behind it) could not receive such simultaneous double inter-
pretation (i.e. be employed to signal two different derivational operations)
would explain for the apparent absence of SD forms being interpreted as a
causativised version of the corresponding D-stem form. In addition, it fol-
lows that in Akkadian any morphologically indicated causativising opera-
tion could be applied only once.
In this paper I have proposed an analysis of the Akkadian D-stem that allows
to account for its effects when applied both to unaccusative and transitive
verbs. I suggested that the D-stem represent a transitive (v-VP) configuration
with some additional head or feature that pluralises events, other effects fol-
lowing f r o m the type of event involved and how it is pluralised. Mapping
transitives onto that configuration therefore does not increase their valency
but endows them with the verbal, or rather eventual, plural property as il-
lustrated in section 4. Mapping unaccusatives onto that configuration makes
them transitive but the verbal plural property is concealed by the fact that
pluralising the event does not result in a multitude of states here that need
to be predicated of different entities. As only a limited set of data and rele-
vant questions could be treated here, a wider scope of investigation will
certainly yield a richer picture. The issue of verbal plural of the type en-
countered in Akkadian does not seem to have received much attention in
the generative literature and clearly merits further research.
In the majority of cases, there obtains a predictable and systematic relation-
ship between a v e r b ' s G- and D-stem. As is the case with the corresponding
stems in other Semitic languages, however, also the Akkadian D-stem is not
a fully productive category but may be subject to lexicalisation in varying
228 Christian Huber
degrees, beside some dialectal variation. Not all Akkadian verbs may occur
in the D-stem, the D-stem may also be restricted to special meanings or
usages. One does thus not expect a fully coherent picture. To provide just
one example, consider (63), which is a case of a verb that has no G-stem
but occurs in the D-stem only and allows for a reading that is neither causa-
tive nor displays any verbal plural properties (see (i)). In addition, it allows
for a valency alternation without a change of stem.
Deviations from the regular pattern are also found in the corresponding stems
of other Semitic languages, e.g. in Hebrew, where some instances of unac-
cusative verbs can be found in the corresponding stem (data from Joosten
1998: 223):
Acknowledgements
I would like to thank the people w h o over the years have discussed with me
issues related to this paper, notably Henry Davis, Jean L o w e n s t a m m , David
Pesetsky, D o m i n i q u e Sportiche, Tim Stowell, and especially Friedrich
Neubarth. Parts or versions of this paper w e r e presented at the 35th
C o l l o q u i u m of Linguistics (2000, University of Innsbruck); at Ö L T 2000
(University of Graz); at the 2001 Conference on the Syntax and Semantics
of Semitic L a n g u a g e s at the University of Southern California, Los
Angeles; and at Ö L T 2 0 0 2 (University of Innsbruck). Let m e again express
my thanks to the organisers of C S S S L for reimbursing m y travel expenses
and providing m e a c c o m m o d a t i o n in Los Angeles. I am also grateful to the
two reviewers for their c o m m e n t s and to Gillian R a m c h a n d for additional
comments.
Notes
1. For details on the language see the grammars of Soden (1995), Buccellati (1996),
and Huehnergard (1997). (Note however that Buccellati and Huehnergard do
not always indicate attestations for the data they present.) I give data in bound
transcription, using standard semitistic transliteration symbols. By convention,
s = [J], h = [x], 5 = [?]; s and t are emphatics; ä, e, Τ, 0 represent long vowels;
ä, e, i, ü represent long vowels resulting from contraction. For some sugges-
tions as to the actual realisation of the corresponding Akkadian graphs see e.g.
Diakonoff 1992 and Soden 1995. Further symbols are: d (emphatic), 1 = [Ϊ], h
= [h]. As a citation form I use the infinitive, which also was the citation form
in ancient Mesopotamian lexical and grammatical works. Abbreviations used
in glosses and tables are as follows. V = root; C, V = consonant, vowel; EXT =
external argument; DN = devine name, RN = royal name; cases: NOM = nomi-
natve, GEN = genitive, ACC = accusative, ACC.INST = instrumental accusative;
DAT = dative, OBL = oblique; PP = preposition phrase; abs = absolute state,
cstr = construct state; verb stems: G = G-stem, D = D-stem, S = S-stem, Ν =
N-stem; trans. = transitive, intrans. = intransitive; REFL = reflexive; inf = in-
finitive, sub = subjunctive, prec = precative; gender: f = feminine, m = mascu-
line, c = communis; number: sg = singular, pi = plural (with nouns only plural
will be indicated in glosses); CL = clitic(ised pronoun), possCL = possessive
pronominal clitic ('your', 'his', etc. case of clitics will be indicated explicitly
only with dative clitics); persons: 1, 2, 3 = 1st, 2nd, 3rd person; (thus,
'possCL3sgm' = 'possessive pronominal clitic 3rd person singular masculine');
vent = ventive; tenses: pres = present, pret = preterite, perf = perfect, stat =
Stative; NEG = n e g a t i o n , REL = relative particle; MA: t h e enclitic particle -ma,
230 Christian Huber
which I will not discuss, is glossed as MA. Although Akkadian data are given
in bound trascription, a clitic will be separated from its host by a hyphen in or-
der to make glosses more transparent. In glosses to verb forms, the basic stem
meaning is followed by the indications of the stem, tense, and per-
son/number/gender, separated by a hyphen from cliticised pronouns (i.e. pro-
nominal affixes) or other clitics. The form umahhisüsuma 'they beat him' will
thus be spelled out as umahhisü-su-ma and glossed as beat(D).pret.3plm-
CL3sgm-MA, indicating that the verb is a D-stem form in the preterite, 3rd per-
son plural masculine, followed by the 3rd person singular masculine pronomi-
nal affix (as its cliticised object), which in turn is followed by the enclitic par-
ticle -ma. Example sentences are accompanied by indications of the textual
source, given in standard Assyriological abbreviations. For a key to these ab-
breviations as well as for general quotations of Akkadian vocabulary items
outside a context (mainly verbs quoted merely as infinitives) see the dictionar-
ies AHw and CAD.
2. In contrast to Arabic and Hebrew the D-stem template cannot host quadriliteral
(or reduplicated biliteral) roots in Akkadian. Quadriliteral roots appear with a
prefix η or s instead, e.g. Vblkt N: nabalkutu 'to cross, pass over', S: subalkutu
'to let/make cross, pass over' (*bulkutu).
3. That verbs may acquire special meanings is not uncommon among languages.
In some instances relevant here it is also doubtful whether the G-stem and the
D-stem belong to the same root. However, the possibility of a D-stem does not
depend on the existence of a corresponding G-stem verb, as can be seen from
the fact that the D-stem may also be employed in denominal verbs, and that a
number of verbs are not found in the G-stem but only in the D-stem (see foot-
note 4). This might indicate that the D-stem (and perhaps also other stems)
could be derived directly from the root.
4. See Kouwenberg (1997) for an excellent survey of actually attested D-stem
forms and their distribution among the dialects/language stages and text gen-
res, also in relation to other verb stems.
Of course, no exact numbers can be given of attested D-stem forms and their
distribution. For mere orientation purposes, the following - approximate -
figures may be gleaned from Kouwenberg's survey (note, however, that
Kouwenberg himself hesitates to give any figures): ca. 165 G-stem verbs (ex-
cluding unergatives) do not undergo a valency alternation in the D-stem but
display verbal plurality instead. Among these are ca. 9 verbs whose D-stems
can be used not only to underline plurality but, in another use, may also ex-
press a specialised or somewhat different meaning (in some cases indicating a
higher degree of agentivity). In ca. 13 further verbs that do not undergo a va-
lency alternation, the semantic relationship between the G- and D-stems is gen-
erally not, or not fully, predictable. The lexicalised forms are usually restricted
to a specific dialect. With a handful of G-stem verbs, the D-stem primarily
Complex predicate structure andpluralised events in Akkadian 231
11. The reverse case is found e.g. in maräsu 'be(come) ill, distressed', where
causativisation appears to be via the S-stem in the regular case and only excep-
tionally via the D-stem.
12. It is also unclear why for example the transitivised/causativised version of
mädu 'be(come) (more) numerous, much/more' appears in the S-stem only and
no D-stem seems to have been possible.
13. Note that a construction such as John had/let/made Mary enlarge the springs
could not be yielded by applying the S-stem form of rabü because the S-stem
simply causativises the respective G-stem verb and unaccusatives such as rabü
'be(come) big(ger)' cannot undergo an "ergative" alternation, i.e. do not allow
for a transitive alternant in the G-stem which the S-stem could causativise.
Akkadian seemingly had no general periphrastic causative construction com-
parable to e.g. the English ma£e-causative.
Also note, in this connection, the proposal in Streck (1994: 177f.), where it
is suggested that the St2-stem of the verb kämt 'be(come) firm' correspond to a
causative of that verb's Dt-stem ('be made firm'=) 'be checked', yielding the
meaning 'have χ checked', 'let jc be checked' ("veranlassen, daß [χ] überprüft
wird"). AHw renders the meaning of the form in question as "überprüfen" ("to
check"). If Streck's assumption is on the right track, the St 2 -stem of the verb
känu constitutes an exceptional case in Akkadian. Perhaps restructuring is an
issue here as well.
14. An anonymous reviewer points out to me Doron's (2003) work on verbal tem-
plates in (Modern) Hebrew, of which I was not aware. Doron's analysis, which
cannot be discussed here in any greater detail, assumes that two additional
kinds of functional heads beside little ν are involved in the syntactic construc-
tion of verbs: agency heads and voice heads. In that approach, the Hebrew in-
tensive template (i.e. the Hebrew counterpart of the Akkadian D-stem) in-
volves an agency head that is a modifier of the root. It classifies the event as
action, licenses ν and determines that the role of the external argument be Ac-
tor (c. also Kouwenberg's (1997) "high transitivity" approach). The characteri-
sation of the external argument as Actor (in the weak concept of action of
Doron 2003: 19) seems by and large compatible also with the Akkadian D-
stem evidence. The Hebrew causative template (i.e. the Hebrew counterpart of
the Akkadian S-stem) involves an agency head that merges with a fully con-
structed verb and introduces its own argument.
However, Doron's approach cannot be straightforwardly applied to Ak-
kadian. The properties of the Akkadian template system are somewhat differ-
ent from the Hebrew system of templates (in the sense of what is traditionally
referred to as stems). For example, iterative templates such as the Akkadian it-
erative stems are altogether absent from Hebrew. Also the verbal plural prop-
erty of the Akkadian D-stem does not seem to have a counterpart in the Modern
Hebrew intensive template and therefore plays no role in Doron's approach.
Complex predicate structure andpluralised events in Akkadian 233
and other constructions with passive-like properties that display verbal plural.
This is compatible with Chomsky's (1999: 6f.) view that also passives involve
a v-VP configuration, however with ν being phi-incomplete (as opposed to phi-
complete ν in the transitive/active case). See also Chomsky 1999: 14 and fn. 35
and 36 on participles.
24. For some - exceptional - pluralic unaccusative D-stem cases in Biblical Hebrew
see below footnote 31 and ex. (64).
25. Exceptions to this rule are rare and often occur only in technical language (for
example, an intransitive version of petü 'open' may denote the becoming dis-
tant of stars). I will ignore them here.
26. Since the D-stem is not barred by countability (see e.g. (31) above in section 4)
it must be assumed that such a number or series of transition may be explicitly
limited by a numeral.
27. Taking place of proto-events/proto-transitions without constituting a transition
will yield an atelic reading (e.g. conative {try, ...), etc.); possibly effected by
rendering the series incomplete). Note that (52) could also be seen as a transi-
tion series applying to proto-entities which in their totality constitute an entity
(e.g. an object).
28. Another relevant case here would be a series of (distributive) benefactive events
involving a mass noun(-like) direct object.
29. Another scenario might yield a change of degree with respect to progress. In a
type I-series progress in degree would be predicated of different entities, an
option not found in Akkadian.
30. Arabic form I/Hebrew Qal = G-stem; Ar. form II/Hebr. Pi'el = D-stem; cited
forms: perf 3sg masc.
31. Consider also the possibly exceptional case in (ib) from Hebrew, involving a
mass noun. If Joosten's (1998) rendering of (ib) is correct, this could be a case
of a type II series of grow-type events whereby multiple embedding is inter-
preted as a larger-than-usual degree.
(i) Biblical Hebrew: a. Qal: samah 'grow, sprout'
b. Pi c el: simmeh 'grow abundantly'(?) (said of hair)
(cf. Joosten 1998: 223)
32. See also above footnote 25.
33. Gemination of the middle radical might be seen as akin to or an instance of
reduplication. For the development of reduplicated roots from a 'repetition'
meaning towards 'imperfective/present' see e.g. Bybee, Perkins and Pagliuca
1994 ch 5.1 will not discuss here the morphology of iterative stems.
34. The prefixes in the G-stem are a-, ta-, i- {< yi- < ya-) and ni- (< na-). By apo-
phonic vowel change they become u-, tu-, u-, and nu-. In the case of i- and ni-
I take the (at least historically) underlying forms ya- and na- to serve as input
for the apophonic step a=>u.
Complex predicate structure andpluralised events in Akkadian 235
35. Note, however, that there still seems to be an underlying geminate in the pre-
sent tense form of the S-stem which cannot surface with sound verbs only for
templatic reasons. Evidence for this comes from verbs with a so-called "weak"
first radical. Consider e.g. the S-stem forms usähiz (preterite) and usahhaz
(present) from the root V 'hz 'take' with the weak first radical ' (= [?]). The
weak first radical is deleted, allowing in the preterite form the vowel [a] to
spread to the empty position, resulting in a long vowel [aa] (= ä) (usa > hiz =>
usaahiz). In the present form, however, deletion of the weak first radical al-
lows to restore the underlying geminate. Here, [h] is attracted to the empty
slot, resulting in the geminate [hh], and [a] remains short (usa'haz =>
usahhaz). I assume this hidden geminate to be due to a morphological site in
the template of Akkadian verbs originally identified by Guerssel & Lowen-
stamm (1993) for Arabic which they call "Derivational Syllable". However, is-
sues of Akkadian templatic morphology cannot be discussed here any further.
References
Baker, Mark
1996 The Polysynthesis Parameter. New York/Oxford: Oxford University
Press.
1997 Thematic roles and syntactic structure. In Elements of Grammar,
L. Haegeman (ed.), 73-137. Dordrecht: Kluwer.
Belletti, Adriana, and Luigi Rizzi
1988 Psych-verbs and θ-theory. Natural Language and Linguistic Theory
6: 291-352.
Brockelmann, Carl
1908 Grundriß der vergleichenden Grammatik der semitischen Sprachen.
1. Band: Lautlehre und Formenlehre. Berlin.
Buccellati, Giorgio
1996 A Structural Grammar of Babylonian. Wiesbaden: Harrassowitz.
Bybee, Joan, Revere Perkins and William Pagliuca
1994 The Evolution of Grammar. University of Chicago Press.
Chomsky, Noam
1995 The Minimalist Program. Cambridge, Mass.: MIT Press.
1999 Derivation by Phase. MIT Occasional Papers in Linguistics 18.
Diakonoff, Igor
1992 Proto-Afrasian and Old Akkadian. A study in historical phonetics.
Journal of Afroasiatic Languages 4: 1-133.
Doron, Edit
2003 Agency and Voice: The Semantics of the Semitic Templates. Natural
Language Semantics 11: 1-67.
236 Christian Huber
Greenberg, Joseph
1991 The Semitic "Intensive" as Verbal Plurality. In Semitic Studies in
honor of Wolf Leslau, Alan S. Kaye (ed.), 576-587. Wiesbaden: Har-
rassowitz.
Guerssel, Mohand and Jean Lowenstamm
1993 Apophony in Classical Arabic. Unpublished manuscript.
1996 Ablaut in Classical Arabic Measure I Active Verbal Forms. In Stud-
ies in Afroasiatic Grammar, Jacqueline Lecarme, Jean Lowenstamm
and Ur Shlonsky (eds.), 123-134. The Hague: Holland Academic
Graphics.
Hale, Kenneth L. and Samuel J. Keyser
1993 On Argument Structure and the Lexical Expression of Syntactic
Relations. In The View From Building 20; Essays in Linguistics in
Honor of Sylvain Bromberger, K. Hale and S. J. Keyser (eds.), 5 3 -
109. Cambridge, Mass.: MIT Press.
Huehnergard, John
1997 A Grammar of Akkadian. Atlanta, Georgia: Scolars Press.
Joosten, Jan
1998 The Functions of the Semitic D-stem: Biblical Hebrew Materials for
a Comparative-Historical Approach. Orientalia N.S. 67: 202-230.
Kaye, Alan and Judith Rosenhouse
1997 Arabic Dialects and Maltese. In The Semitic Languages, Robert
Hetzron (ed.), 263-311. London/New York: Routledge.
Kouwenberg, N. J. C.
1997 Gemination in the Akkadian Verb. Assen, The Netherlands: Van
Gorcum.
McCarthy, John
1981 A prosodic theory of nonconcatenative morphology. Linguistic In-
quiry 12: 373-418.
Pesetsky, David
1994 Zero Syntax. Experiencers and Cascades. Cambridge, Mass.: MIT
Press.
Poebel, Arno
1939 Studies in Akkadian Grammar. [Assyriological Studies 9] Chicago,
111.: The University of Chicago Press.
Rowton, M.B.
1962 The Use of the Permansive in Classical Babylonian. Journal of Near
Eastern Studies 21: 233-303.
Pustejovsky, James
1991 The syntax of event structure. Cognition 41: 47-81.
Soden, Wolfram v.
1995 Grundriß der akkadischen Grammatik, 3. Auflage. Wiesbaden: Har-
rassowitz.
Complex predicate structure andpluralised events in Akkadian 237
Streck, Michael
1994 Funktionsanalyse des akkadischen St2-Stamms. Zeitschrift für Assyrio-
logie 84: 161-197.
Szabolcsi, Anna
1994 The Noun Phrase. In The Syntactic Structure of Hungarian, Syntax
and Semantics 27, Katalin E. Kiss and Ferenc Kiefer (eds.), 179-
274. San Diego and London: Academic Press.
Ungnad, Arthur
1949 Grammatik des Akkadischen. 3. Auflage. München: Biederstein
Verlag.
Bibliographical abbreviations
Edit Doron
VSO word order may on principle be the reflection of very different clause
structures. This chapter argues that for many languages where VSO order is
attested (Semitic, Celtic, Romance, Greek), it reflects a particular clause
structure, which I will call a "VSO clause," where the subject does not raise
out of the c-command domain of the tense head of the clause:
τ A
...s ...o...
A VSO clause is derived only if Τ has the following property:
This chapter shows that left-conjunct agreement follows from (i) above,
and more precisely from the weaker (ii):
I assume (following Borer 1986 and Chomsky 1998) that what triggers the
attraction of DP to Τ is the Agree relation between Τ and DP. Therefore, if
Τ attracts DP, then they must be related by Agree. But if Τ has property (ii) -
i.e., it does not attract DP - then Τ need not be related by Agree to DP.
Rather, I propose that the Agree relation be defined to hold between Τ and
the minimal D constituent closest to Τ which allows the derivation to con-
verge. In a derivation which requires D to move, the minimal D that does
not lead to a violation of the constraints on movement (such as the Coordi-
nate Structure Condition) is the full subject DP. But in a derivation where
D is not required to move, the minimal D constituent is a D head. In section
3, I show that in clauses with a conjoined subject, it is the head of the left
conjunct that is the D head closest to T.
VSO clauses share property (ii) with another very different type of
clause, those with an expletive subject. In this type of clause, Τ has the EPP
feature - i.e., it does not have property (i). The EPP feature of Τ is satisfied
by merging an expletive rather than by attracting the subject; that is, clauses
with expletive subjects have property (ii). Yet both VSO clauses (which
also have property (i)) and clauses with expletive subjects (which do not
have property (i)) exhibit left-conjunct agreement, which shows that this
kind of agreement follows from a property weaker than (i). It will be shown
in section 3 that left conjunct agreement follows from property (ii) - i.e.,
from the lack of DP-raising to Τ - either because Τ does not have the EPP
feature (property (i)), or because the EPP feature is satisfied by merging an
expletive.
Left-conjunct agreement is illustrated below for English, in a clause
with an expletive subject, and for Modern Hebrew, in a clause with a null
expletive:
In general, VSO clauses are not expletive constructions, since they do not
show the defmiteness effect found in (2) and (3). (1) above, for example, is
a VSO clause which is clearly not an expletive construction, since its sub-
ject is definite. In addition, not every VSO sequence is a VSO clause. VSO
word order is found in Modern Hebrew, but not VSO clauses. First, VSO
word order occurs in Modern Hebrew in a null expletive structure such as
(3) above. In addition, Modern Hebrew has VSO sequences following any
preverbal constituent. 3 But there is no VSO sequence in Modern Hebrew
which by itself constitutes a clause. Example (4) below shows that clause
initial VSO is ungrammatical in non-expletive constructions, irrespective of
the agreement features of the verb. (5) shows that VSO is possible only
when some other constituent precedes the verb. The contrast between (5a)
and (5b) indicates that left-conjunct agreement is disallowed. These con-
structions are therefore different in some crucial way from the expletive
construction in (3), which allows left-conjunct agreement:
b. Modern Irish
da mbeinn -se agus tusa ann
if be.COND.Is EMPH and you there
'if you and I were there' (McCloskey and Hale 1984: 31a)
c. Spanish
Estaba abierta la tienda y el mercado
was,3S open.FS the shop.F and the market.Μ
'The shop and the market were open.' (Rodrigo Gutierrez, p.c.)
d. Greek
Irthe ο Pavlos kai ο Giannis sto parti
c a m e J S the Paul and the John to-the party
'Paul and John came to the party.' (Anastasia Giannakidou, p.c.)
The prevalent word order in Biblical Hebrew is verb initial, as shown again
in (7) below. In Modern Hebrew, on the other hand, the prevalent word
order is SVO, as shown in (8a), whereas verb initial sentences are in gen-
eral ungrammatical, as shown again in (8b):
SVO is also often attested in Biblical Hebrew. Example (9), exactly as is, is
a perfect grammatical sentence of both Modern and Biblical Hebrew:
(10) V S O
way-yiqqafi mo:se ?et-matte ha:? E lo:hi:m b°- ya:do:
and-tookJA/S Moses ACC-rod (of)the.God in-hand.his
'And Moses took the rod of God in his hand.' (Exodus 4: 20)
In verb initial sentences such as (10), the complementizer ' a n d ' cliticizes to
the verb, yet it does not follow that the verb raises to C in Biblical Hebrew.
Rather, the complementizer in both (9) and (10) lowers to cliticize to the
left edge of the clause, similar to what is argued by Shlonsky (1988) for the
Modern Hebrew complementizer se 'that' and by McCloskey (1996b) for
Irish. Indeed, when sentences like (9) and (10) are preceded by adverbial
clauses (themselves introduced by complementizers), then the main-clause
complementizer follows the adverbial clause, as shown in ( l l a - b ) . Notice
that it should not be inferred from the syntax of the corresponding King
James translations that the clause following the adverbial clause is an em-
bedded clause. In Biblical Hebrew, unlike English, a clause with an overt
complementizer is possible as a main clause. In fact, the complementizer
'and' never introduces an embedded clause (cf. footnote 4):
In the examples in (13), the form of the verb is singular, yet the subject is
clearly plural. It is implausible to analyze these examples as containing a
singular subject combined with a comitative phrase, though this is the in-
terpretation often offered by traditional interpreters of the Bible, such as
Rashi (Rabbi Shlomo Yitzhaki, 1040-1105), and traditional translations
such as the King James Bible (see in particular the translation of 13c and
13e). 5 (13d), for example, contains a plural reflexive dative bound by the
subject. (13g) contains the idiom ' X lifted up X ' s voice' which is obligato-
rily reflexive: There is no lifting up anybody's voice but o n e ' s own. But
this entails, since X is marked as plural in the second part of the idiom, that
the subject is plural as well. The examples in (13) are, therefore, examples
with plural subjects.
The contrast between full agreement of the preverbal subject and partial
agreement of the postverbal subject is also found in the other languages
vith left-conjunct agreement. Postverbal left-conjunct agreement, as in the
(a) sentences below, alternates with full preverbal agreement, in the (b)
sentences:
246 Edit Doron
(15) Spanish
a. Estaba abierta la tienda y el mercado
w a s . i S open.FS the shop.F and the market.M
'The shop and the market were open.'
(16) Greek
a. Irthe ο Pavlos kai ο Giannis sto parti
c a m e J s the Paul and the John to-the party
'Paul and John came to the party.'
It has already been argued extensively that the verb in Modern Hebrew
raises out of the VP. 6 First, there is evidence (Shlonsky 1987) that the verb
can precede sentential adverbs, like bevaday 'certainly':
Second, there is evidence (Shlonsky 1991) that quantifiers are floated from
a postverbal position:
Third, VP-ellipsis strands the verb (Doron 1990, 1999), which shows that
the verb has raised out of the VP. The second reading available for (19)
shows that VP-ellipsis has applied, stranding the verb in T:
In Biblical Hebrew, it is possible to show that the verb may be found to the
left of sentential adverbs, which indicates that it raises out of VP:
In addition, at least for the absolute form (ABS), the verb can be shown to
raise out of VP, since ABS precedes not only Τ (i.e., the tensed form of the
verb), but negation (NEG) as well, as mentioned by Levin (1971) 7 :
The original position of ABS is within VP, as can be seen in untensed clauses.
The order ABS -Τ observed in (21) is not found in imperative clauses, which
are untensed. Rather, the order found in imperative clauses is T- ABS, as in
(22):
(23) FP
V+T+F TP
SUBJ TP
agreement
tv+r VP
VP
tv OBJ
(24) FP
V+T+F TP
la'iibat
playedJre S U B J TP
Mariam
tv+T VP
tsuBJ VP
tv PP
fi 1-bayt
in the-house
(25) FP
FP PP
FP FP fi 1-bayt
in the-house
V+T+F TP V+T+F TP
e
laiibat / \
playedJra SUBJ T P SUBJ χρ
Mariam Zayd
tv+T VP tv+T VP
tsUBJ V P t S U B J V P
tv PP tv PP
I I
tpp tpp
In (25), the second occurrence of the verb 'play' is ellided, whereas the PP
'in the house' is Right Node Raised (RNR). Yet this account is problematic
even for Arabic. It predicts that the RNR constituent should show singular
agreement, since this constituent supposedly originates from two singular
clauses. This prediction cannot be tested with an RNR constituent such as
'in the house' in (25), which does not exhibit agreement. Yet in examples
where the RNR constituent is a predicate with overt agreement, this predic-
tion is systematically falsified. In (26), for example, the RNR constituent
yallabaini fi l-bayt 'play.3MD in the house' is obligatorily marked as dual
(D), but if it were raised from two singular clauses, it should be marked as
singular:
Aoun and Benmamoun (1999) deny the existence of this problem by show-
ing that it is not attested in either Lebanese or Moroccan Arabic. Yet this
problem arises for Standard Arabic, as (26) shows, and moreover, it is also
found in Irish and Biblical Hebrew. The relevant example from Irish is
VSO and Left-Conjunct Agreement 251
shown in (27), where the constituent which is Right Node Raised from two
singular clauses (according to Aoun, Benmamoun, and Sportiche), 'ndr sui,
is nevertheless plural:
(27) Irish
Bhinn fein agus an seanduine 'när sui.
was IS EMPH and the old-fellow IP sitVN
'The old fellow and I used to be sitting.' (McCloskey 1986: ex. 37)
(28) TP
V+T VP
agreement
SUBJ VP
tv OBJ
Something needs to be said about how the structure in (28) satisfies the
EPP. 8 According to Chomsky (1995), the EPP is satisfied by covert raising
of the features of the subject to T. Chomsky (1998) argues against covert
feature movement, and moreover proposes to view the EPP as a feature not
just of Τ but of functional heads in general. This feature is not necessarily a
252 Edit Dor on
lexical property of functional heads, but can be added independently into the
derivation. I propose that in a strictly VSO language such as Irish, Τ is not
compatible with the EPP feature, neither as part of its lexical specification
nor as an addition by the derivation. In Biblical Hebrew, Arabic, Romance,
and Greek, on the other hand, an EPP feature may be added to Τ in some
derivations, though it is not part of the lexical specification of T. Accord-
ingly, in these languages, a V S O structure such as (28) is derived with a
numeration which does not include the EPP feature. An SVO structure such
as (29) is the outcome of a different derivation, one which includes an EPP
feature but no expletive:
(29) TP
SUBJ TP
V+T VP
agreement
VP
tv OBJ
The two derivations in (28) and (29) have different agreement patterns,
based on the same operation Agree:
In (30), if Τ has an EPP feature not satisfied by the pure merging of an ex-
pletive, it enters into the AGREE relation with the highlighted DP, since this
is the DP closest to T, and since this is the minimal constituent within the
closest DP that may move without violating the constraints on movement:
VSO and Left-Conjunct Agreement 253
(30)
D NP CONJ DP
Yet, if Τ in (30) does not have an EPP feature (or if the EPP feature can be
satisfied by a pure merging of an expletive), no movement is forced, and
the AGREE relation holds with the closest D head, which is the boldfaced D
in (30). Crucially, I assume the asymmetric structure of coordinate structures
argued for by Larson (1990). where conjunctions head coordinate struc-
tures. I also assume that conjunctions lack any formal features, from which
it follows that the category they project is that of the conjuncts themselves.
Moreover, the number specification of a conjoined DP is not a morphosyn-
tactic feature of the conjunction head, as argued hy Farkas and Zee (1995).
b. TP
tv toBJ
OVS sentences like (31a) are also attested in Modern Hebrew. As men-
tioned in the introduction, Modern Hebrew allows postverbal subjects on
condition that some other constituent - e.g., the object - precedes the verb.
Indeed, (31b) is proposed by Borer (1995) as the structure of Modern
Hebrew OVS sentences. Yet if this is the right structure for OVS sentences
in Modern Hebrew, and if left-conjunct agreement is accounted for struc-
turally, then the prediction is that (31 b) should exhibit left-conjunct agree-
ment in Modern Hebrew as well. But this is not the case, as already men-
tioned. In Modern Hebrew, only full agreement is attested:
Examples (33) and (34) above fall under clause (I) of this generalization,
whereas (35) and (36a) fall under clause (II). 10 The problem is that the two
clauses of the Moreshet Generalizaton do not seem to constitute a natural
class of syntactic environments.
Fortunately, the analysis proposed for clitics in Semitic by Doron (1996)
and Doron and Hey cock (1999) makes it possible to subsume condition (II)
of the Moreshet Generalization under condition (I). According to this
analysis, clitics may be viewed as anaphors bound by preverbal constitu-
ents, as in (37):
THE V 2 CONDITION
The verb in Biblical Hebrew does not raise beyond Τ unless preceded
by some constituent.
(39) FP
tv toBJ
(i) The verb is always preceded by some constituent, not necessarily the
subject.
(ii) There is no left-conjunct agreement when the subject follows the verb,
other than in expletive constructions.
Notes
1. All the Biblical Hebrew translations are from the King James Bible (1611),
which is generally more literal than the other translations.
2. I am very grateful to Shraga Assif for the phonetic transcription of the Biblical
Hebrew data.
3. The preverbal constituent may be null even when it is not expletive, as is
generally assumed for "narrative inversion" and for "all-focus sentences," e.g.:
(i) Modem Hebrew
hitqaser ?aba selka
called father yours
'Your father called.'
4. The main clause complementizer 'and' has an important role for text cohesion
(for recent discussion see de Caen 1995 and Hatav 1997). It is in complementary
disiribution with other complementizers - e.g., ki: 'for/that', pen 'lest', h a ' Q '
(a yes-no interrogative complementizer) - which only introduce embedded
clauses:
(i) ?al ti:r?i: ki: sa:ma? ? e lo:hi:m ?el qo:l hannaTar
1
n e g fear,2FS for heard.iMS God to voice (of)the.lad
'fear not; for God hath heard the voice of the lad.' (Genesis 21: 17)
(ii) ki: ?a:mru glisti:m ρεη yaVasu: ha:?ibri:m
for said.iMP Philistines lest will.make.3MP the.Hebrews
fisrsb ?o: fiani:t
sword or spear
'For the Philistines said, lest the Hebrews make them swords or spears.'
(1 Samuel 13: 19)
(iii) way-yo:msr h a tafiat ? e lohi:m ?a:no:ki:
and-said.iMS β. instead God I
'and he said, Am I in God's stead.' (Genesis 30: 2)
260 Edit Doron
References
Benmamoun, Elabbas
1992 Structural conditions on agreement. Proceedings ofNELS 22: 17-32.
Benmamoun, Elabbas and Shalom Lappin (eds.)
1999 Fragments: Studies in Ellipsis and Gapping. Oxford: Oxford Univer-
sity Press.
den Besten, Hans
1983 On the interaction of root transformations and lexical deletive rules.
In On the Formal Syntax of Wesgermania, Werner Abraham (ed.),
47-131. Amsterdam: John Benjamins.
Bobaljik, Jonathan D. and Andrew Carnie
1996 A minimalist approach to some problems in Irish word order. In The
Syntax of the Celtic Languages. Robert Borsley and Ian Roberts (eds.).
Cambridge: Cambridge University Press.
Borer, Hagit
1986 I-subjects. Linguistic Inquiry 17: 375-416.
1995 The ups and downs of Hebrew verb movement. Natural Language
and Linguistic Theory 13.3: 527-606.
Borsley, Robert and Ian Roberts (eds.)
1996 The Syntax of the Celtic Languages. Cambridge: Cambridge Univer-
sity Press.
Carnie, Andrew and Eithne Guiloyle (eds.)
2000 The Syntax of Verb Initial Languages. Oxford: Oxford University
Press.
Chomsky, Noam
1995 The Minimalist Program. Cambridge, Mass.: MIT Press.
1998 Minimalist inquiries: the framework. MIT Occasional Papers in
Linguistics 15.
Chung, Sandra
1990 VPs and verb movement in Chamorro. Natural Language and Lin-
guistic Theory 8: 559-620.
Chung, Sandra and James McCloskey
1987 Government, barriers and small-clauses in Modern Irish. Linguistic
Inquiry 18: 173-237.
Cinque, Guglielmo and G. Giusti (eds.)
1995 Advances in Roumanian Linguistics. Amsterdam: John Benjamins.
DeCaen, Vincent
1995 On the Placement and Interpretation of the Verb in Standard Biblical
Hebrew Prose. Ph.D. dissertation, University of Toronto.
Doron, Edit
1990 VP-ellipsis in Hebrew. MS., The Hebrew University of Jerusalem.
1996 The Predicate in Arabic. In Studies in Afroasiatic Grammar,
Jacqueline Lecarme, Jean Lowenstamm and Ur Shlonsky (eds.), 7 7 -
87. The Hague: Holland Academic Graphics.
VSO and Left-Conjunct Agreement 263
Levin, Saul
1971 The Indo-European Languages. Albany, NY: SUNY University Press.
Massam, Diane
2000 VSO and VOS: Aspects of Niuean word order. In The Syntax of Verb
Initial Languages, Andrew Carnie and Eithne Guilfoyle (eds.), 97-
116. Oxford: Oxford University Press.
McCloskey, James
1986 Inflection and conjuction in Modern Irish. Natural Language and
Linguistic Theory 4: 245-281.
1991 Clause structure, ellipsis and proper government in Irish. Lingua 85:
259-302.
1996 a Subjects and subject positions. In The Syntax of the Celtic Languages,
Robert Borsley and Ian Roberts (eds.). Cambridge: Cambridge Uni-
versity Press.
1996b On the scope of verb movement in Irish. Natural Language and
Linguistic Theory 14: 47-104.
McCloskey, James and Kenneth Hale
1984 On the syntax of person-number inflection in Modern Irish. Natural
Language and Linguistic Theory 4: 487-533.
Mohammad, Mohammad
1990 The problem of subject-verb agreement in Arabic: towards a solution.
In Perspectives on Arabic Linguistics /., Mushira Eid (ed.). Amster-
dam: John Benjamins.
Moreshet, Menachem
1967 Hanasu hakodem lishney nos'im bilshon hamikra. Leshonenu 31: 1-10.
Platzak, Christer
1983 Germanic word order and the Comp/Infl parameter. Working Papers
in Scandinavian Syntax 2. Trondheim.
Shlonsky, Ur
1987 Null and Displaced Subjects. Ph.D. Dissertation, MIT.
1988 Complementizer cliticization in Hebrew and the Empty Category
Principle. Natural Language and Linguistic Theory 6: 191-205.
1991 Quantifiers as functional heads: a study of quantifier float in Hebrew.
Lingua 84: 159-180.
Shlonsky, Ur and Edit Doron
1992 Verb-second in Hebrew. Proceedings ofWCCFL 10: 431-446.
Zwart, Jan Wouter
1993 Dutch Syntax: A Minimalist Approach. Groningen Dissertations in
Linguistics 10.
IE *weid- as a root with dual subcategorization
features in the Homeric poems
Annamaria Bartolotta
This paper is organized as follows: the first section sketches the theoretical
background involved in the case study of Old Greek eidon/oida. As is well
known, the aorist eidon takes only an accusative DP-object, while the perfect
oida can take either a genitive or an accusative DP-object. Sections 2-5 aim
to prove that the diachronic development of the root *weid- in early Greek
must be take into consideration to explain the synchronic phenomenon of
dual subcategorization features. This root proves indeed to be polysemous
and is split into two different meanings which are lexicalised by means of
different bridging contexts and different morphological developments. In
section 6 the peculiar evolution of oida from a genitive/accusative to an ac-
cusative syntax is considered from a typological point of view. The genitive/
accusative syntax is held to be a residual trace of a previous active-stative
system which had characterised the Proto-Indo-European before it adopted
a nominative-accusative system. Section 7 examines the available evidence
about the inherent status of genitive and accusative in Homeric Greek. In
section 8 the modular conception of grammar is discussed by considering
the role played by the morpho-semantic properties of the root when dealing
with a THETA-related case system. Sections 9 - 1 0 are concerned with a
minimalist representation of the syntactic structure of the early Greek verb.
Against the LATE INSERTION principle, a model endowed with three func-
tional heads (TP, AspP, vP) is put forward which requires the root features
to be visible in syntactic derivation before Spell-out. Thus, the syntactic
object case feature at that stage of the language is taken to have been inter-
pretable in LF (Object-in-situ hypothesis). Section 11 examines the Greek
sentence structure after the nominative-accusative type had been stabilized:
once case had turned to a purely formal feature, it became uninterpretable;
so, the semantic properties of the root were not able to determine the inher-
ent case assignment any longer, and tense tended to incorporate all tempo-
ral distinctions related to the verb (including aspect).
266 Annamaria Bartolotta
In the Homeric poems, the Greek perfect (w)oida Ί see with the m i n d ' s
eye, I k n o w , ' shows a peculiar behaviour as regards its morphosyntactic
structure. Although the aorist (w)eidon Ί saw', derived from the same root
*weid- 'to see', is characterized by an identical lexical entry, we can observe
a different case mapping between the two verb forms when they are bivalent
and select their internal argument (i.e. a nominal direct object). Regardless
of the same categorization features [ ] v , the same sub-categorization
features [+ N P ] Vp, the same selective restrictions as for the subject
N P [+an] , and the same theta-grid structure <Experiencer, Theme>,
the aorist is seen to regularly assign the accusative case, while the syntax of
the perfect alternates between an assignment of the accusative and genitive
cases in an apparently unpredictable way. 1 Compare, for example, the fol-
lowing passages quoted f r o m the Iliad, where oida sometimes assigns the
genitive case, as in Iliad X V , 4 1 1 - 1 2 , 2
At first glance, this outline may seem very unusual if we refer to the tradi-
tional generative thematic theory, according to which properties of verb
selection and valence are invariable for all the tenses, moods and persons of
the inflected verb. Furthermore, it should be noted that inherent case assign-
ment occurs simultaneously with internal thematic role assignment, which
is specified at the lexical entry of the verb itself (Haegeman 1996: 149).
IE *weid- as a root with DS features in the Homeric poems 267
The first step of the differentiation between *weidon and *woida has to be
found in the meaning split that the original IE root *weid-, from which both
verbs derive, undergoes. The split is motivated by the linguistic sign's
POLYSEMOUS nature: originally, the root *weid- 'to see' w a s a lexeme with
more than one related sense, denoting a kind of vision that could be both
perceptive and intentional. 6 As in most Indo-European languages, also the
Greek-derived stem *wid- / *weid- / *woid- bears more than one meaning,
with a semantic extension from 'see' to ' k n o w ' . 7 In the Homeric text, the
zero-grade stem of the aorist *wid- 'see' meant not only a kind of perception
which consisted of simple eyesight, but it also denoted the beginning of an
268 Annamaria Bartolotta
internal acquisition process after working-out data that had been seen, to-
wards a brief form of thought. This is what the analysis of eidon occurrences
in the Homeric poems tells us clearly (see below § 4). Soon, probably al-
ready at a Proto-Indo-European stage, as the comparison with the Vedic
developments of the same root lets us reconstruct (Bartolotta 2002a: 17-20),
this semantic complexity found a lexical counterpart in a new stem-formation
which entailed a vowel change (ablaut). Therefore, it is possible to distin-
guish in Greek the results of such semantic splitting:
(i) the PERCEPTIVE side of sight may be attributed to the zero-grade stem
*wid-\ the aorist eidon denotes a punctual action characterized by an
Aspect feature [-Stative]. This verb assigns the accusative case to its
internal argument, a case which semantically reflected an IMMEDIATE
contact with the object of sight;
(ii) the INTENTIONAL side of perception may be instead attributed to the
o-grade stem *woid-: the perfect oida denotes a process characterized
by an Aspect feature [+stative] (see below § 4). This verb originally
assigns the genitive case to its internal argument, a case which se-
mantically reflected a MEDIATED contact with the object of sight. 8
In the following sections I shall try to examine the way this semantic split
will condition the morphosyntactic structure of the verb.
3. Distributional contexts
Table 1. The physical organ involved in the action/state described by the aorist
eidon and the perfect oida.
Now, focusing on the context allows us to distinguish two crucial steps of the
semantic change, (1) the so-called STAGE OF OVERLAP and (2) the STAGE
OF LEXICALIZATION. According to Evans and Wilkins (2000: 549), in the
first stage, the semantic change from a meaning A to a meaning Β involves a
transitional stage of POLYSEMY where only one form takes both the meanings
(stage of overlap). 1 0 At this stage, the new meaning is not yet lexicalised.
On the basis of these claims, we would have the following evolution path:
270 Annamaria Bartolotta
1. Stage of overlap
(a) initially, one polysemous form has in nuce both the overlapping
meanings A and B: the aorist eidon denotes the two senses inherited
from the IE root *weid-, i.e. 'to see with the eyes' (MEANING A) and
the extended form 'to see with the mind' (MEANING B);
(b) the primitive meaning A is enriched by a regular contextual support
(bridging context) which enables the speakers to distinguish clearly a
new meaning B: some Homeric passages testify the presence of such
BRIDGING CONTEXTS which support the semantic extension of eidon
(see below).
2. Stage of lexicalization
The contextual sense (MEANING B) becomes then lexicalised after
creating a different morphological status of the root. This new status
is here created by deriving the Aktionsart category of Perfect through
the change of the root's vowel (*woid-). *woida is in fact character-
ized by a strong paradigmatic autonomy and an early morphological
lexicalization. Evidence for this early lexicalization comes from the
absence of reduplication (which normally marks the perfect), the
presence of proper personal endings, the presence of the vowel
change (apophony) and the presence of a new categorization form
(Aktionsart), which denotes the so-called 'resulting state', i.e. a state
resulting from a process. 11
Evidence for the first stage comes from Homeric passages that seem to be
an exception to the regular distribution discussed in §3 (table 1). As stated
above, the dative ophthalmoisi 'eyes' occurs in fact with the aoristic forms
since they properly refer to the original sense of 'seeing' related to eye-
sight; whereas datives such as thymö, phresin, etc. 'mind', are linked to the
perfect, since this latter relates to a cognitive process which takes place
only in the mind. Now, the BRIDGING CONTEXT is instead represented by
those passages in which the aorist forms occur with the syntactical expres-
sions usually proper to the perfect. Let us analyse the text. In Iliad XXI, 61,
as shown in (3),
IE *weid- as a root with DS features in the Homeric poems 271
the locative expression eni phresin 'in the mind', which occurs with the
aorist subjunctive idömai Ί see', represents a bridge between the first
meaning 'to see with the eyes', normally conveyed by eidon from *wid-,
and the extended sense 'to see with the mind', that involves a cognitive
inferential process triggered by ocular sight. A confirmation to what has
been claimed comes from the presence, in the context, of the subjunctive
daeiö (derived from the root Vda 'to know') which is closely related to the
subjunctive idömai Ί see': in this way a form of knowledge which com-
pletes the idea of a whole cognitive process is expressed. The other exam-
ple is in Odyssey VIII, 450, as shown in (4),
where the irregular coexistence of the dative thymö 'mind' with the aorist
ide 'he saw' is again the residual trace of a semantic change in progress.
Further supporting evidence for bridging contexts which function as
catalysing support for linguistic change is to be found in Homeric passages
where the aorist eidon is immediately followed by cognitive verbs such as
noeö Ί perceive', gignoskö Ί come to know', phräzomai Ί think, I per-
ceive', hormainö Ί revolve in the mind', etc. Here it is again possible to
distinguish two steps of the same process which begins with (i) the visual
stimulus and ends with (2) an internalization of knowledge (see Bartolotta
2002b: 25).
(ii) the semantic value of the so-called INHERENT case, which is assigned
by a lexical head V to the N P it governs, in conjunction with θ-role
assignment, now makes way for a new purely syntactic evaluation of
the STRUCTURAL case, which is not assigned in conjunction with Θ-
role assignment by a lexical head. 13
From a synchronic point of view, one of the most striking pieces of evi-
dence of the semantic value related to case-assignment comes from those
passages where oida assigns different cases practically to the same DP-
object. This case assignment is not random, for, as I shall argue, it is
closely connected to the semantic relationship between verb and noun.
Compare, for example, Iliad XII, 100 (= V, 11; V, 549)
In (5) the genitive singular adjective pdses, which refers to mdkhes, means
'all' taken as a whole, and not 'each' as some classical translations give. 16
It refers to a knowledge as a result of a whole complex of military opera-
IE *weid- as a root with DS features in the Homeric poems 275
tions of which a battle consists. The same explanation is indeed valid for
the genitive pases sophies in (1), where pases refers to a kind of knowledge
which is the result of a complex of techniques taken as a whole, the fruit of
an internalization and re-elaboration of data from which a sort of science or
art is inferred (mediated relationship between verb and its DP-object). More-
over, it should be noted that, morphologically speaking, the genitive is not
due to the fact that oida is inflected as a nominal form, i.e. the participle
eidote in (5), because the same construction is to be found in (1), where eide
is a subjunctive.
In (6), instead, the plural accusative mäkhas refers to a series of military
operations considered not as a whole, but as single concrete events, directly
experienced without referring to a general abstract form of knowledge
(non-mediated relationship between verb and its DP-object). The plural
forms are often preferred in similar constructions, as we can see in Iliad XI,
719, where Nestor is too young and has not learnt yet the art of war deriv-
ing from the internalization of the concrete experience of a series of mili-
tary actions:
b. eidos pugmakhies
knowing-PERF.PART.DL boxing-GEN.SG
' k n o w i n g (seeing in the mind) the art of boxing' (II. XXIII, 665)
c. aikhmes eu eidos
spear-GEN.SG well-ADV knowing-PERF.PART.SG
'knowing (seeing in the mind) well the art of the spear' (II. XV, 525)
276 Annamaria Bartolotta
d. eidote khärmes
knowing-PERF.PART.DL fight-GEN.SG
'knowing (seeing in the mind) the art of the fight' (II. V, 608)
b. eidos pantoious te
knowing-PERF.PART.SG each.kind.of-ADJ.ACC.PL and-CONJ
dölous käi medea pyktui
crafts-ACC.PL and-CONJ devices-ACC.PL cunning-ACC
'(viz. Odysseus) knowing (seeing in the mind) each kind of crafts
and cunning devices' (II. Ill, 202)
After this analysis of the status of the features involved in the oida perfect,
it is possible to clarify a synchronic account of the early Greek verb struc-
ture. The presupposition behind this is that the peculiar situation of eidon
and oida, which are tenses belonging to the same verb paradigm, is ex-
plainable only by considering the idiosyncratic lexical and morphological
properties of the root-stem. These properties play a role in syntax since
only they can account for certain future developments responsible for any
morphosyntactic structure of the verb. The morphological and lexical fea-
tures of the root have to be visible in the syntactic derivation before and not
after Spell-Out: only in this way will it be possible to give a predictive syn-
chronic explanation of all the linguistic phenomena related to the verb. That
is why in this treatment I follow Embick (2000: 187-188) when he argues
against the Distributed Morphology principles such as those of the LATE
INSERTION a n d t h e FEATURE DISJOINTNESS e x p r e s s e d in M a r a n t z ( 1 9 9 4 ,
1995). According to these principles, which defend a modular concept of
grammar, the morphological features and the arbitrary properties of vo-
cabulary items are irrelevant for syntax. They must not be present in the
syntactic derivation, with the result that the lexical Roots and the functional
morphemes have to be inserted after Spell-Out. The *wiidon / *woida case
study seems instead to highlight the role played in structuring syntactic
configurations by the idiosyncratic semantic properties of the verbal Root
with its morphological features, such as Aspect and Aktionsart (carried by
apophony). 20 This claim is valid for the inherent case-selection and, as we
shall see in the next section, for the related feature-checking process.21
Moreover, it is worth noting that dead languages such as Sanskrit, for
example, allow us to reconstruct a Proto-Indo-European language whose
lexicon was essentially based on a root-items system.
In order to describe the structure of the Homeric Greek syntax, I shall assume
a clause structure endowed with three functional heads, Tense, Aspect and
278 Annamaria Bartolotta
Light verb, within the framework of the Agr-less checking theory of the
Minimalist model. Following some suggestions coming from Embick's
analysis of Latin verb, the notation V in (10) is for the Root, «that is, the
member of the open-class vocabulary appearing in this position» (Embick
2000: 195). Moreover, according to Stroik (2001: 368), it should be noted
that 'the light verb v, which bears the [Vform] feature, is a functional cate-
gory that has both morphophonetic and semantic justification'. The [Vform]
contains all the morphological features of the root, such as Aktionsart, etc.,
which are responsible for the case selection of the verb. 22 These morpho-
logical features, that have to be visible in the syntax by virtue of the EARLY
INSERTION of the root (i.e. from the outset of the derivation), will be
checked in the head v. The structure will be then represented as follows:
(10) TP
Τ AspP
V DP
The TP bears the temporal feature, but it has been hypothesized that an
Agr(eement) node can be added in the morphological component to form a
T-Agr position (see Chomsky 1995). Here it is assumed that the verb under-
goes the checking of φ-features such as person, number, gender, including
thematicity feature [±th] (Embick 2000: 197), i.e. a morphological feature
which reveals the presence of a theme vowel as in weid-o-n. However, it
should be noticed that tense inflection and agreement inflection seem to be
two quite different systems, as pointed out, to quote just a few, by Friedmann
and Grodzinsky (1997, 2000) in their neurophysiological analysis, or by
Guasti and Rizzi (2000: 1). It could be therefore assumed that agreement is
checked in a mechanism different from that for tense. So, since agreement
may not have a node of itself, it may check in one of the other checking
points below Τ (Friedmann and Grodzinsky 2000: 99). The AspP projects
the perfective or imperfective aspectual feature, which is related to the verb
and then to v. It deals with a basic aspectual opposition, which plays a fun-
damental role in the syntax of early Greek, as already pointed out. Then,
the vP head contains all the morpholexical properties ([Vform]) which the
IE *weid- as a root with DS features in the Homeric poems 279
verb is provided with. Among them, the most important is the feature related
to the Aktionsart category, including notions such as stativity, causativity,
agentivity, punctuality, and so on. As for eidon and oida, the [Vform] fea-
tures relate respectively to a punctual action (aorist) and a resulting state
(perfect). 23 As regards the Λ/Ρ, it should be noted once again how important
it is that the verbal Root is present in the syntax from the outset of the deri-
vation. In fact, the case difference between accusative or genitive overtly
manifested by the morphology depends on the lexical values of the Root: 24
both cases had originally a semantic (inherent) value which was available
in order to make the verb semantically complete. Moreover, the presence of
the Root playing a role in Homeric Greek also finds a diachronic justifica-
tion, as results from the well known fact that in ancient languages such as
Vedic and even Proto-Indo-European (from which Homeric Greek directly
derived) the lexical items were memorized as roots. Finally, as for the verb's
arguments, I follow Lopez (2001: 714) in distinguishing the different proper-
ties of subject and object feature checking. In other words, while for the ex-
ternal argument, i.e. the subject, it has been hypothesized that, after origi-
nating from the VP shell (the Specv position), it checks its formal features
with T, a different treatment seems to characterize object feature checking.
This is what I shall try to argue in the next section.
assignment should be more emphasized. The oida case study has shown
that alternating accusative or genitive is not a mere structural fact, but a
matter of morphological-syntactic-semantic interface. That is to say, a verb
with dual subcategorization features, as is seen to occur with oida in the
early stage of Greek, points out that case assignment of internal arguments
depends on the morphosyntactic properties of the lexical verb. In structural
terms, object can also check its features in situ, against the same verbal head
that assigns it a θ-role (Lopez 2001: 699) without moving anywhere. 2 5 This
is indeed the result of the fact that accusative case is here a feature which is
interpretable at the semantic component LF. Only this hypothesis, which is
consistent with some recent suggestions about the role of semantically un-
interpretable morphological features in dislocation property (Chomsky
2000: 26), allows us to highlight the typological difference between a
nominative-accusative type, where the accusative case has only a structural
configurational value (non θ-related), and an active-stative type, where the
case still involves a semantic value which plays an important role in identi-
fying the verb's meaning. 2 6 This solution also maintains strict locality of Θ-
marking: an Object-in-situ hypothesis points out indeed the close semantic
relationship between the verb and its internal argument through maintaining
the highest level of adjacency between the two constituents. Crucially, de-
spite some widely accepted assumptions of Chomsky's theory (see Adger
et al. 1999: 5), in certain languages - for example Homeric Greek - case
becomes an interpretable feature which does not need movement for check-
ing since it survives in the mapping to the semantic component and is not ' a
formal feature that plays a role only in the syntactic derivation itself (Hale
1998: l l ) . 2 7 Under these assumptions, it could therefore be hypothesized
that a language which is typologically classified as active-stative is a [+0PC]
language. According to this THETA POSITION CHECKING Parameter, elements
can undergo feature checking at their θ-position (Ura 2000: 38), that is the
same position where they receive a θ-role. It could be hypothesized that
object enters into a checking relation with ν without moving anywhere, that
is, following Ura (2000: 220), the morphosyntactically tight relation be-
tween a verb and its object is connected to an obligatory checking relation
between them. Accordingly, it has been proposed that 'the [±0PC]-ness
lies in v's lexical properties' and that 'v's property concerning formal fea-
tures checking depends upon the aspectual-head that selects ν as its com-
plement' (Ura 2000: 219). 28 As a consequence, in the proposal made here,
the role of such a Parameter turns out to be fundamental in the analysis of
Homeric Greek, which still shows traces of an active-stative type. Hence, I
IE *weid- as a root with DS features in the Homeric poems 281
(i) the verbal system turns out to be provided with both AspP and TP
since the Indo-European active-stative type, which is directly inher-
ited in Greek, shows that the aspectual distinctions play a role more
important than the tense distinctions;
(ii) the vP contains all the morphosemantic features, such as [±stative],
proper to the verbal Root (i.e. Aktionsart), which are f u n d a m e n t a l for
case assignment (the option for genitive or accusative); this is the rea-
son w h y we have to hypothesize an Early Insertion of the Root, which
is responsible for the lexical entry of the verb and its dual subcatego-
rization features;
(iii) the verb has to be represented as a Root, since are the semantic prop-
erties of this latter which are relevant in assigning the inherent case:
w h e n the Root is *wid- [-stative] the aorist eidon assigns the accusa-
tive; w h e n the Root is *woid- [+stative] the perfect oida is able to as-
sign also the genitive;
(iv) the internal argument (object) checks its formal feature of inherent
case (genitive or accusative) in situ (without involving a movement),
i.e. where it is base-generated and receives its θ-role f r o m the related
lexical head ( * w i d - / *woid-). It is in fact semantically interpreted at
the LF. This solution fits the hypothesis that the N o u n and the Root
are listed together in a derived lexical entry;
(v) the external argument (subject) checks regularly its formal feature
(nominative case) against the functional head T.
282 Annamaria Bartolotta
On the basis of these considerations, the Homeric Greek verb structure re-
lated to eidon and oida results as follows: 3 0
(11) TP
DP, Τ
q-JJ
AspP
[aor] / [perf]
[φ-features] Asp 1
(V+v)i
Asp 0 vP
[±perfective]
t (V + v)j t. v
,0
VP
[Aktionsart]
t(^+v)i t (V Root), DP
(i) the internal argument (object) has to move in order to check its [-inter-
pretable] feature of accusative structural case, since it does not receive
a case in situ from the lexical head V (oida + accusative in Classical
Greek);
(ii) it becomes less and less important that the semantic properties of the
Root have to be visible in a pre-syntactic level, since they do not deter-
mine an inherent case assignment (oida does not show case alternation
between accusative and genitive in Classical Greek);
(iii) both AspP and vP, which bear the morphosyntactic characteristics of
the verb, play now a marginal role in respect of TP, since Tense tends
to incorporate every temporal distinctions related to the verb.
12. Conclusion
(12) The meaning of an item determines most (if not all) of its syntactic
properties.
age. Hence, unlike theories of linguistic change which suggest that mor-
phosyntactic change is abrupt and total, because it is the result of the reset-
ting of certain parameters (Timberlake 1977, Lightfoot 1991, 1998, Harris
and Campbell 1995), the line of research proposed here is more consistent
with some recent approaches, according to which parameter settings do not
change abruptly. On the contrary, change occurs as a result of competition
between alternative parameter settings during periods of syntactic variation
(Pintzuk 2003). Typology and semantics in a diachronic perspective have
had to be taken into account to put forward an explanation of peculiar case
assignments by different forms of one verbal paradigm in a synchronic
state of language. 34 My working hypothesis has involved diachrony in the
interpretation of this phenomenon. As for the implications and the motiva-
tions of the linguistic change that I have supposed, further investigation is
required.
Notes
1. From the analysis of all the occurrences of oida in the Homeric poems, the use
of genitive or accusative case due to the definite or indefinite nature of the ob-
ject NP involved does not seem likely (cf. Belletti 1988: 31). For the analysis of
the oida occurrences and such related matters, see Bartolotta (2002a: 57-70).
2. The following abbreviations are used in the paper: AN = animate; AOR = aorist;
PERJF = perfect; IND = indicative; SUBJ = subjunctive; PART = participle; INF =
6. For more discussion about this definition initially introduced by Vendryes, see
De Boel (1987: 20).
7. Compare Old Indie (yeda), Avestan (vaedä), Gothic (wait), Old Irish (ro.fetar),
Armenian (gitem), Old Slavic (vede), etc., which all reveal a deep connection
between the meaning of 'knowing' and the meaning of 'seeing', from which
the former originated. For a detailed description of the semantic extension
among different domains see Sweetser (1998).
8. In a previous work (Bartolotta 2002a: 116 ff.) I have hypothesized the existence
of a GENITIVE OF INFERENCE. See also below, note n. 15.
9. It can be easily assumed that the three nouns thymos, phren and prapides refer
to the human mind, whose seat is respectively placed in the heart, in the dia-
phragm and in the midriff.
10. It has been recently pointed out that 'the great majority of cases where different
meanings of a verb are associated with different subcategorization frames in-
volve polysemy. That is, these verbs exhibit highly related meanings, often with
a more concrete physical sense and extensions to more abstract and metaphorical
uses' (Hare, McRae and Elman 2003: 283).
11. See Di Giovine (1996, 2: 127-131) for a detailed discussion on each piece of
evidence regarding this early morphological lexicalization of the perfect oida.
12. In an active-stative type, as has been supposed for the Proto-Indo-European
language, to assign an argument depends on the semantic properties of the
lexical head, that is, the choice of a Noun and its related case strictly depends
on the [± stative] feature of its head.
13. The inherent (theta-related)-structural opposition of generative grammar corre-
sponds to the concrete-abstract opposition of historical linguistics (see Haudry
1977).
14. The genitive case was traditionally defined in historical linguistics as the case
of 'general determination' (Adrados 1992: 152), since it is shown to hold a
wide range of different meanings in respect of the other cases. For the original
functions of the genitive case, see Morani (1992: 221).
15. This terminology is taken from a previous work (Bartolotta 2002a: 113). The
INFERENCE refers to the peculiar cognitive process involved with oida where
knowledge is not immediate but inferential, that is, the result of a complex ac-
tivity of the mind including more than one step, from the visual perception to
the internal acquisition (memorization) and finally to the elaboration of the
data. The option for the genitive case would reflect the fact that, as underlined
also in Belletti (1988: 3), 'the most typical instance of an inherent Case is one
where a particular Case is associated with a particular θ-role'.
16. See, among others, Murray (1924): 'well skilled in all manner of fighting',
Mazon (1957-61): 'experts ä tous les combats', Calzecchi Onesti (1950): 'es-
perti d'ogni battaglia'. Note also that the adjective pases seems to be dislo-
cated on the right peryphery in order to emphasize by means of topicalization
the comprehensive meaning of 'all' as a whole.
IE *weid- as a root with DS features in the Homeric poems 287
17. It has to be noticed that in the Homeric poems, even when the accusative refers
to an abstract object, e.g. 'thoughts' in (9c), every mental experience is experi-
enced in a very concrete way, as results from a wide literature on this topic.
See Bartolotta (2002a: 123) and references cited there.
18. Further details and occurrences in Bartolotta (2002a: 116-124).
19. Both hypotheses are to be discarded also for other verbs of perception like
mnäomai (see Bartolotta 2003: 54).
20. On the role of Aktionsart in the verbal system see, among others, Van Valin and
La Polla (1999: 92): '[...] verbs have a basic Aktionsart type, which is how they
are represented in the lexicon', and Gennari (2003: 35), who underlines that
we need not only the tense morphemes, but also the Aktionsart morphemes in
order to determine the exact temporal interpretation of a sentence.
21. Although the dissociation between the lexicon and the computational system
has been pointed out in order to corroborate the thesis of the modularity of the
Grammar (Anderson and Lightfoot 2000: 20), the close interaction between the
lexicon and morphological features on the one hand and the computational op-
erations on the other should not be underestimated, since the latter are often
triggered by the former during the syntactic derivation (Lorenzo and Longa
2003: 651). Also White (2003: 3) points out that 'properties of items that enter
ino a computation may vary in feature composition and feature strength, with
associated syntactic consequences'.
22. With regard to the distribution of more than one morphological feature related
to one Head 'little v', it should be said that this is a first attempt to include
those features in the syntactic configuration. Subsequent thorough examinations
of such relationships should lead us to a more homogeneous representation,
which perhaps will include a one-to-one relationship between a Head and its
related features.
23. It has been shown that 'little v' morphemes with semantic contents as agentive,
Stative, inchoative, causative, etc., which can form different types of verbs from
the same root, are responsible for many syntactic effects (see Arad 2002).
24. Maling (2001: 458) shows that 'although there is no neat one-to-one correlation
between m-case and thematic role, case does serve to identify verbal arguments
on a verb-by-verb basis'.
25. The theory which distinguishes subject and object Feature checking is in line
with Alexiadou and Anagnostoupolou's generalization (2001) according to
which 'by Spell-Out VP can contain no more than one argument with an un-
checked Case feature'. Furthermore, some current issues offer new inputs
against movement for feature checking. Among others, Picallo (2002: 118-
119) assumes 'following Chomsky 1999, that Case-φ checking (i.e. the opera-
tion Agree) is a syntactic process that applies independently of movement';
Moro (2004: 5) points out that 'movement of noun phrases in languages with
fully overt Case morphology (such as Latin) is hard to explain' if maintaining
that movement is triggered by uninterpretable features.
288 Annamaria Bartolotta
26. Evidence for the relevance of the object noun phrase to the aspectual interpre-
tation of the verb with which it is closely integrated and which denotes a com-
plex activity is to be found also in non-Indo-European languages such Hungarian
(see Kiefer 1994: 185-205).
27. Ormazabal (2000: 236) has demonstrated that in constructions using certain
verbs the object movement is triggered by an animacy feature; differently it
stays in situ.
28. This parameter has been introduced by Ura in order to characterize the ergative
languages compared to 'languages like English or other European languages:
[...] whereas Checking may coincide with θ-assignment in some languages, it
may not in others. [...] Put differently, [±0PC] is the key to the distinction be-
tween nominative-accusative languages and ergative-absolutive languages, a big
topic to which numerous studies have been devoted'. Given this background,
referring to the comments of the anonymous reviewer, the [±©PC] Parameter
could be considered as a macroparameter (in the sense of Baker 1996), to the
extent that its implications have a fundamental impact on the typology of lan-
guages such as Old Greek or Vedic at earlier stages.
29. A confirmation to what has been here hypothesized comes also from the studies
of non-Indo-European languages such as Finnish, where 'aspect is not defined
in terms of a simple, one-to-one correspondence between form and meaning
but rather as a result of the contribution of one or more of the following factors:
the lexical semantics of the verb, derivational morphology, case variation of
the object [...]' (Bache, Basboll and Lindberg 1994: 11).
30. It should not be disregarded that, as some recent hypotheses might suggest (see
Lorenzo and Longa 2003: 648), not only the case but also morphological fea-
tures such as Tense, Aspect and Aktionsart could be considered [+Interpretable],
and then they would not need to be moved in order to be checked.
31. Notice that, differently from here, Tanaka does not take into account any vP,
but he simply splits the VP in two shells, inside which an AspP is assumed to
be placed.
32. With regard to this topic, see Maling (2001: 436^137), who has pointed out the
role of the differences between a direct θ-role which is associated with a direct
argument that receives structural case and one which receives lexical case in
German.
33. Following Pinker, Hare, McRae and Elman (2003: 283) point out that 'subtle
semantic distinctions between otherwise similar verbs often determine the sorts
of syntactic structure in which the verbs may appear'.
34. As stated in Hale (1998: 1), who follows a Minimalist perspective, 'the study
of change may provide valuable insights into the proper characterization of
certain syntactic phenomena within current syntactic theory'.
IE *weid- as a root with DS features in the Homeric poems 289
References
Chomsky, Noam
1995 The Minimalist Program. Cambridge, Mass.: MIT Press.
2000 Linguistics and Brain Science. In Image, Language, Brain, Alec
Marantz, Yasushi Myashita and Wayne O'Neil (eds.), 13-28. Cam-
bridge, Mass./London, England: MIT Press.
De Boel, Gent
1987 La syntaxe des verbes 'voir' chez Homere. Glotta 65/1-2: 19-32.
Di Giovine, Paolo
1996 Studio sulperfetto indoeuropeo. II. Roma: II Calamo.
Embick, David
2000 Features, Syntax, and Categories in the Latin Perfect. Linguistic
Inquiry 31/2: 185-230.
Evans, Nicholas and David Wilkins
2000 In the mind's ear: the semantic extensions of perception verbs in
Australian languages. Language 76/3: 546-592.
Friedmann, Na'ama and Yosef Grodzinsky
1997 Tense and Agreement in Agrammatic Production: Pruning the Syn-
tactic Tree. Brain and Language 56: 397-425.
2000 Split inflection in neurolinguistics. In The Acquisition of Syntax,
Marc-Ariel Friedemann and Luigi Rizzi (eds.), 84-104. Harlow,
England: Longman.
Gennari, Silvia P.
2003 Tense Meanings and Temporal Interpretation. Journal of Semantics
20/1: 35-71.
Giorgi, Alessandra and Fabio Pianesi
1997 Tense and Aspect. From Semantics to Morphosyntax. New York/
Oxford: Oxford University Press.
Guasti, Maria Teresa and Luigi Rizzi
2000 Agreement and Tense as distinct syntactic positions: evidence from
acquisition, in: http://www.ciscl.unisi.it/persone/rizzi.htm.
Haegeman, Liliane
1996 Manuale di Grammatica Generativa. Milano: Hoepli. (Introduction
to Government and Binding Theory. 2nd Edition. UK: Blackwell
Publishers).
Hale, Mark
1998 Diachronic Syntax. Syntax 1/1: 1-18.
Hare, Mary, Ken McRae and Jeffrey Elman
2003 Sense and structure: Meaning as a determinant of verb subcategoriza-
tion preferences. Journal of Memory and Language 48 (2): 281-303.
Harris, C. Alice and Lyle Campbell
1995 Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cam-
bridge University Press.
IE *weid- as a root with DS features in the Homeric poems 291
Haudry, Jean
1977 L 'emploi des cas en vedique. Introduction a l'itude des cas en indo-
europeen. Lyon: L'Hermes.
Jannaris, Antonius Ν.
1897 An Historical Greek Grammar Chiefly of the Attic Dialect. Hildes-
heim: Georg Holms.
Kiefer, Ferenc
1994 Some peculiarities of the aspectual system in Hungarian. In Tense,
Aspect and Action. Empirical and Theoretical Contributions to Lan-
guage Typology, Carl Bache, Hans Basb0ll and Karl-Eric Lindberg
(eds.), 185-205. Berlin/New York: Mouton de Gruyter.
Kurylowicz, Jerzy
1964 The inflectional categories of Indo-European. Heidelberg: Carl Winter
Universitätsverlag.
Lasso de La Vega, Jose S.
1968 Sintaxis Griega I. Madrid: Patronato Menendez y Pelayo.
Lehmann, Winfred P.
1999 La linguistica indoeuropea. Bologna: II Mulino (Theoretical Bases
of Indo-European Linguistics. 1993: London-New York: Routledge).
Lightfoot, David W.
1991 How to set Parameters: Arguments from Language Change. Cam-
bridge, Mass: MIT Press.
1998 The Development of Language: Acquisition, Change and Evolution.
Oxford: Blackwell Publishing.
2002 Syntactic Effects of Morphological Change. Oxford: Oxford Univer-
sity Press.
Lopez, Luis
2001 On the (Non)complementarity of Θ-Theory and Checking Theory.
Linguistic Inquiry 32/4: 694-716.
Lorenzo, Guillermo and Victor M. Longa
2003 Minimizing the genes for grammar. The minimalist program as a bio-
logical framework for the study of language. Lingua 113: 643-657.
Luraghi, Silvia
1996 Studi su casi e preposizioni nel greco antico. Milan: Franco Angeli.
Maling, Joan
2001 Dative: The heterogeneity of the mapping among morphological case,
grammatical functions, and thematic roles. Lingua 111: 419^164.
Marantz, Alec
1994 A late note on Late Insertion. Cambridge, Mass: MIT Press.
1995 'Cat' as a phrasal idiom: Consequences of Late Insertion in Distrib-
uted Morphology. Cambridge, Mass.: MIT Press.
292 Annamaria Bartolotta
Morani, Moreno
1992 Linee di storia della flessione nominale indoeuropea. Alessandria:
Edizioni dell'Orso.
Moro, Andrea
2004 Linear Compression as a Trigger for Movement. In Triggers, Henk
van Riemsdijk and Anne Breitbarth (eds.), 387-429. Berlin: Mouton
de Gruyter.
Ormazabal, Javier
2000 A Conspiracy Theory of Case and Agreement. In Roger Martin,
David Michaels and Juan Uriagereka (eds.), Step by Step. Essays on
Minimalist Syntax in Honor of Howard Lasnik, 235-260. Cam-
bridge, Mass./London: MIT Press.
Penke, Martina
2003 On the morphological basis of syntactic deficits. Brain and Language
87 (1): 50-51.
Pesetsky, David
1996 Zero Syntax. Experiencers and Cascades. Cambridge, Mass./London,
England: MIT Press.
Picallo, M. Carme
2002 Abstract agreement and clausal arguments. Syntax 5 (2): 116-147.
Pintzuk, Susan
2003 Variationist Approaches to Syntactic Change. In Joseph D. Brian and
Richard D. Janda (eds.), The Handbook of Historical Linguistics,
509-528. Oxford: Blackwell Publishing.
Schwyzer, Edward
1953 Griechische Grammatik I-II. München: C. Η. Beck.
Speijer, Jacob S.
1886 Sanskrit Syntax. Leiden. [Reprint 1988.] Delhi: Motilal Banarsidass.
Stroik, Thomas
2001 On the Light Verb Hypothesis. Linguistic Inquiry 32 (2): 362-369.
Sweetser, Eve
1998 From etymology to pragmatics. Metaphorical and cultural aspects of
semantic structure. 6th Edition. Cambridge: Cambridge University
Press.
Tanaka, Hidezaku
1999 Raised Objects and superiority. Linguistic Inquiry 30: 317-325.
Timberlake, Alan
1977 Reanalysis and Actualization in Syntactic Change. In Mechanisms of
syntactic change, Charles N. Li (ed.), 141-177. Austin: University of
Texas Press.
Ura, Hiroyuki
2000 Checking Theory and Grammatical Functions in Universal Gram-
mar. New York, Oxford: Oxford University Press.
IE *weid- as a root with DS features in the Homeric poems 293
Texts
Vassilios Spyropoulos
1. Introduction
There are a number of parameters associated with these principles, which de-
termine the internal structure of an infinitival clause and the constructions
in which this infinitival clause can participate. A major parameter is con-
cerned with the issue of finiteness, i.e. the ability of the functional category
INFL to assign case to the subject of the clause. The issue of finiteness is
closely associated with the feature specification of INFL, i.e. the existence
of tense and agreement features, the parameters of which are argued to be
morphosyntactically determined (Chomsky 1993, 1995). Another parameter
is concerned with the categorial status of the infinitival clauses, i.e. whether
they are INFLPs or CPs. The interaction of these parameters determines the
conditions under which an infinitival clause can have an overt DP-subject
or participate in control and raising structures.
We argue that CIGr. infinitival clauses are not finite, in the sense that
their INFL does not assign case to their subjects and, consequently, CIGr.
exhibits raising and control phenomena. We also claim that CIGr. infinitival
clauses have an articulated C-domain, the lower functional projection of
which has the option of assigning accusative case to the DP-subject of the
infinitival clause, when it is properly activated. This analysis of CIGr. in-
finitives provides an explanation for a number of idiosyncratic properties of
infinitival syntax in this language, namely the occurrence of an overt DP-
subject in exactly the same environments where raising and control phe-
nomena are also found. In addition, it offers a classification of infinitival
constructions based on the properties of the infinitival clause itself and not
on the characteristics of their surface structure, as descriptive analyses of
traditional grammars do.
The paper is organized as follows: In section 2 we present some of the
main morphological and syntactic properties of CIGr infinitives. In section
3 we examine the so-called Accusativus cum Infinitivo construction and we
argue against an analysis that suggests that CIGr. infinitival INFL, being
specified for [T], is able to assign accusative case to the DP-subject of the
infinitival clause (Philippaki-Warburton and Catsimali 1989, 1991, Sevdali
2003). Instead, in section 4, it is argued that this DP-subject is assigned
accusative case by a relevant head inside the C-domain. Section 5 outlines
how the interaction of UG principles with the language specific parameters
can account for the whole range of CIGr. infinitival clauses. Section 6 con-
cludes the paper.
The syntax of Classical Greek infinitive 297
2.1. Morphosyntax
CIGr. infinitive inflects for four of the traditional tenses (present, aorist
(past), future, (present) perfect) of the CIGr. verbal system and for three
voices (active, passive and medio-passive). The relevant forms are shown
in the following table:
Voice
Tense
active passive medio-passive
present lyein lyesthai lyesthai
future lysein lythe:sthai lysesthai
aorist lysai lythe:nai lysasthai
(present) perfect lelykenai lelysthai lelysthai
It has been noticed that tense distinctions in CIGr. infinitive mask an aspect
distinction (Jannaris 1897; Schwyzer 1950; Binnick 1991): the present in-
finitive denotes durative aspect, the aorist infinitive punctual aspect and the
perfect infinitives perfect aspect.3 The future infinitive is partly out of this
system4 for the following reason: It is considered to be a novelty in the
CIGr. verbal system that resulted from the grammaticalization of direct
speech (Schwyzer 1950). As a consequence, it has a very limited distribu-
tion and can be mainly used after the so-called verba dicendi et sentiendi,5
where it substitutes for the future indicative of the direct speech:6
On the basis of this variability, it has been suggested that CIGr. infinitive
carries morphological specification for tense, and thus its INFL is specified
as [+T] (Philippaki-Warburton and Catsimali 1989, 1991; Tantalou 2003;
Sevdali 2003). However, this variability is not always possible. As it was
mentioned above, the future infinitive has a very limited distribution and it
is mainly found after the verba dicendi/sentiendi, where a full sequence of
tenses is established. However, in other embedded clauses the future infini-
tive is not standardly used and the different infinitival forms denote only
aspect: present - durative, aorist - punctual, perfect - perfect:
b. l*eksoplisesthai /eksoplisasthai
give.arms-INF.MED.FUT /give.arms-INF.MED.AOR
/eksoplisthai
/give.arms-INF.MED.PRF
In addition, there are also embedded infinitival clauses where the infinitive
cannot vary for tense, and only one infinitive form (usually the present in-
finitive) is possible. In these cases, the infinitive and the matrix verb denote
the same event:
Leaving aside the so called 'absolute infinitive' and the infinitive in main
clauses, we move on to examine the infinitival syntax in CIGr. by focusing
on the case properties of the subject of the infinitival clause.
either the subject of the embedded infinitival clause or the DP-object of the
matrix predicate. Recall examples (14b,c) repeated as (17a,b):
The AccI sequences of the second type constitute object control constructions
and should not be confused with those of the first type. From now on, we will
use the term AccI syntax for referring to the constructions of the first type
only, i.e. for the structures in which the DP in the accusative is the structural
subject of the embedded infinitival clause.
Given that the DP in the accusative case in AccI syntax is the structural
subject of the infinitival clause, the source of its accusative case calls for an
explanation. The principles of case assignment present us with two possi-
bilities: (a) AccI syntax involves Exceptional Case Marking (ECM), i.e. the
DP-subject receives its accusative case from the matrix predicate; and (b)
the accusative case is assigned by a head inside the infinitival clause.
the exact syntactic mechanism that assigns this accusative case, 8 but there
is a consensus about the conditions such a mechanism should obey: (a) the
matrix predicate must be able to assign structural case; and (b) the embedded
infinitival clause should project only to the level of INFLP, i.e. there should
be no CP intervening between the matrix predicate and the embedded DP-
subject.
However, AccI syntax appears in constructions that do not obey these
conditions: First, the infinitival clause in AccI constructions is a CP, as co-
ordination with an o/7'-clause shows (Sevdali 2003; Tantalou 2003):
b. e:goumai [hymas pantas eidenai [cp oti ego: men lego: ortho:s]
[infCL touton de einai skaion]]
Impersonal verbs do not assign an external theta role and take a clausal
complement that realizes a 'theme' theta-role. Thus, they are not able to
assign accusative case by Burzio's Generalization (Burzio 1986). Conse-
quently, the DP-subject of the infinitival clause in these structures cannot
receive its accusative case from the matrix verb. Similarly, the AccI syntax
in an infinitival clause introduced by a verbum dicendi is maintained even
when passivization has turned the matrix verb into an 'impersonal' verb.
Consider example (14a) repeated here as (23a) and its active equivalent
(23b):
Nouns are assumed to assign inherent case (Chomsky 1981, 1986), and
thus they cannot assign case to the DP-subject of an embedded infinitival
clause, because they do not theta-mark it. This is the reason why ECM is
banned in infinitival clauses that are complements of nouns as the un-
grammaticality of (27) shows:
Thus, the grammaticality of example (26) indicates that AccI syntax does
not involve ECM.
Let us summarize so far: We presented evidence that AccI syntax can
appear in environments which do not respect the conditions of ECM. Thus,
it was shown that infinitival clauses with AccI syntax (a) are CPs and (b)
appear as complements to matrix predicates that are not able to assign
structural accusative case. All these facts indicate that AccI syntax does not
involve ECM and the source of the accusative case of the DP-subject of the
infinitival clause is clause internal. This conclusion is reinforced by the
following examples, where no case assignment relation can be established
between the matrix predicate and the DP-subject of the infinitival clause
because of the syntactic status of the infinitival clause:
The syntax of Classical Greek infinitive 307
That the predicate modifier is marked with accusative case indicates that
the empty subject of the embedded infinitival clause, which the predicate
modifier modifies, is also marked with accusative case, so that it blocks
case transmission from the controller in the matrix clause. Thus, examples
(34) and (35) involve a hidden AccI construction, in which the infinitival
clause contains a controlled null-subject pro in the accusative case:
Since the matrix predicate has already assigned genitive/dative case to its
DP-object, ECM cannot be the explanation for the accusative case of the in-
finitival empty subject.
To sum up, the structures examined above clearly show that AccI syntax
appears in environments where ECM is not possible. All these facts indi-
cate that there is a head inside the infinitival clause that is able to assign
accusative case to its subject. In what follows we will present and assess
two hypotheses: (i) the Tense-hypothesis, according to which the source of
this accusative case is the [+T] feature of infinitival INFL, (ii) the C-
hypothesis, which claims that the accusative case is assigned by a null
prepositional complementizer.
CIGr. infinitives inflect for tense and in this respect they differ from the
infinitives of languages such as English, Italian etc. Although not all the
infinitival clauses exhibit the same temporal properties, as we will see be-
low, CIGr. infinitival INFL may be considered to be specified as [+T], as a
reflex of this morphological marking. From this point of view, the [T] fea-
ture of the INFL is assumed to be morphologically defined. Philippaki-
Warburton & Catsimali (1989, 1991) and Sevdali (2003) capitalize on this
310 Vassilios Spyropoulos
tense property of the infinitival INFL and suggest that the infinitival INFL
in CIGr has case assigning properties, because it is specified as [+T]. 10 Let
us call this approach as the Tense-Hypothesis. Significantly, this account
associates the case assignment ability of INFL with the existence of a mor-
phologically defined [T] feature." Thus, an INFL head may have the fol-
lowing case assigning properties according to its specification:
In such examples the infinitival clause has both morphologically and inter-
pretatively full temporal properties. However, not all infinitival clauses
share the same temporal properties. In examples such as the ones in (39),
The syntax of Classical Greek infinitive 311
the infinitival clause has a fixed temporal interpretation (the event time of
the infinitival clause is future oriented with respect to that of the matrix
clause) and the variation among the infinitive forms simply denotes aspec-
tual variation. Crucially, the future infinitive is not normally found in such
examples: 1 3
b. /*eksoplisesthai /eksoplisasthai
give.arms-INF.MED.FUT/give.arms-INF.MED. AOR
leksoplisthai
give.arms-lNF.MED.PRF
There are also examples in which the infinitival clause can be argued to
have no temporal properties at all:
follows, we will show that we need not worry about that, since there is
enough evidence that shows that the assignment of accusative case to the
DP-subject of the infinitive is not determined by the [+T] specification of
the infinitival INFL.
c. Accl syntax
ton assyrion eis te:n kho. ran embalein aggellousi
the Assyrian-ACC in the country invade-INF.AOR anounce-3SG
'They announce that the Assyrian invaded the country'
It is obvious that the Tense-hypothesis has no account for the examples with
N o m I syntax, because in these examples it is evident that the infinitival
INFL, despite its [+T] specification, assigns no case. This conclusion gains
further support f r o m the following examples:
314 Vassilios Spyropoulos
In examples (45) and (46) the predicate modifier, which modifies the empty
subject of the embedded infinitival, is marked with nominative case. Given
that predicate modifiers always agree in case with the element they modify,
the nominative case of the predicate modifiers in (45) and (46) indicates
that no case assignment has taken place in the embedded infinitival clause
and that the ec-subject is an NP-trace and a PRO respectively.
The facts presented above show that AccI syntax is in free variation
with control and raising phenomena in FIs. Interestingly, AccI syntax is in
free variation with a control structure in DIs and after a big variety of
predicates as well, which indicates that this variation is a generalized phe-
nomenon of CIGr. complementation structures:
b. AccI syntax
kindyneuein ouk eboulonto .... autous
risk-INF.PRS NEG wanted-3PL these-ACC
'They didn't want for them to r i s k . . . '
b'. pro\ ouk eboulonto [Cp autouSj kindyneuein] (Th. IV. 15)
The verbs want and try are control verbs and take as a complement an in-
finitival clause with a P R O subject (50a) and (51a). The control pattern is
suspended in (50b) and (51b) by the insertion of the preposition for as C.
The preposition assigns accusative case to the subject of the infinitival
clause, allowing for it to be overtly realized, and thus control is suspended.
The similarity with the corresponding constructions from CIGr. seems not
to be accidental:
CIGr. uses AccI syntax in the same way as English uses ybr-infinitivals.
This similarity indicates that the accusative case in AccI syntax is deter-
mined by the C-layer. W e will call this analysis the C-Hypothesis (Tantalou
2003).
AccI syntax is also found in Latin (Bolkestein 1976; Pillinger 1980;
Cecchetto and Oniga 2002):
Cecchetto and Oniga (2002) examine the properties of AccI syntax in Latin
and conclude that it does not involve ECM. They propose that the DP-
subject of the embedded clause is assigned accusative case by an empty
prepositional complementizer, something like a null for. Tantalou (2003)
claims that CIGr. AccI syntax should be analyzed in the same way and she
argues that this analysis is supported by the CP-status of the infinitival
clause in an AccI construction. We believe that this analysis is on the right
track, mainly because it can account for the optional assignment of the ac-
cusative case. If this null complementizer is an optional element, as for is in
English, control and raising are also possible with the same infinitival
clause, when this null-C is absent. But, such an analysis requires a number
of issues to be addressed: (i) the licensing of this element and (ii) its exact
position and status.
Let us assume that the DP-subject in an infinitival clause with AccI syntax
is assigned accusative case by a null prepositional complementizer. We will
call this element 0for. 0Jor as a null element requires licensing. Empty com-
plementizers have been assumed to have an affix status and, therefore, they
are licensed by getting incorporated into the matrix verb (Stowell 1981;
Kayne 1984; Pesetsky 1991; Ormazabal 1995). This kind of incorporation
is considered to be a strictly local operation, so that it is banned when adja-
cency is not respected (Boskovic 1997; Martin 1996). This analysis ex-
plains why ίΛαί-deletion in English is blocked, when (i) the embedded
clause is in subject position (57b), (ii) an adverbial element intervenes be-
tween the main predicate and the embedded clause (57c), or (ii) the embed-
ded clause has been extraposed (57d):
In CIGr., infinitival clauses with AccI syntax can be found in all the envi-
ronments mentioned above:
318 Vassilios Spyropoulos
b. Extraposition
kindynos oun pollous apollysthai
danger-NOM then many-ACC loose-PASS.lNF.PRS
'Then, there is a danger for many people to get killed'
(X. An. V . l , 6)
c. Intervention of adverbials
synebe: te:s aute.s he:mera:s en te:i sikelia. i
happened-3SG the-GEN this-GEN day-GEN in the-DAT Sicily-DAT
gelo:na nika.n amilkan
Gelon-ACC defeat-INF.PRS Amilcas-ACC
'It happened that on the same day in Sicily Gelon defeated Amilcas'
cHdt. VII. 166)
The data in (55) show that the licensing of CIGr. & f o r is different from that
of English 0,/, a( , in that no adjacency with the matrix verb is required. This
means that the relevant mechanism for the licensing of 0/„,. is not incorpo-
ration/affixation into the matrix verb. Rather, CIGr. 0 / o r is licensed inside
its infinitival clause by a clause internal mechanism.
The answer to the licensing issue lies behind the distribution of AccI
syntax. The data presented in sections 3.2 and 3.3 shows that AccI syntax is
possible with FIs and DIs but not with Als. Als occur after the so-called
'verbs of obligatory coreference' (translation of the Greek term). These are
mainly obligatory subject control verbs such as the following:
The main characteristic of these verbs is that they take an infinitival com-
plement that does not have temporal properties, since its event point is the
same or simultaneous with that of the matrix predicate. This is because the
matrix and the embedded predicate in fact express the same event. Based
The syntax of Classical Greek infinitive 319
He claims that English gerunds have temporal properties and proposes that
(covert) V-to-C raising licenses these temporal properties. This (covert)
movement also licenses the accusative case of the gerundival DP-subject.
The existence of an overt complementizer blocks the (covert) movement
and consequently, the licensing of the accusative case. Thus, in example
(59b), a PRO is licensed as the subject of the infinitival clause. Similarly,
Cecchetto and Oniga (2002) associate covert V-to-C movement for tense
checking with AccI syntax in Latin infinitival clauses. They propose that the
infinitive covertly moves to C in order to license its temporal properties,
320 Vassilios Spyropoulos
Such examples are unexpected, if 0{nr merges with the C position. These
examples involve an infinitival clause with AccI syntax, which is intro-
duced by an overt complementizer. The availability of the AccI syntax in
such an environment indicates that 0for occupies another position than the
normal C position. The pattern is explained by assuming the Split-C-
Hypothesis proposed by Rizzi (1997). Rizzi proposes that the functional
category C is decomposed to two basic C heads: The first carries informa-
tion about the clause-typing properties of the clause and is called Cforce.
Proper complementizers are assumed to merge with this position. The sec-
ond carries information about the finiteness of the clause and is called Cfin.
The feature content of this head is therefore relevant to the grammatical
properties of INFL (mood, aspect, tense, agreement). Topic and Focus pro-
jections are also assumed to occur between these two heads:
(62) [cPl Cf orce [TopP/FocP Top/FoC [cP2 Cfi„ [iNFLP INFL ... ]]]]
We propose that 0for merges with Cfin and, thus, it is compatible with other
complementizers that merge with the Cforce position. Significantly, the posi-
tion of 0for and its licensing are now straightforwardly related. We argued
in the previous section that 0for is licensed when a [T] feature is licensed in
the C-domain. Given that the locus of this feature is the Cfin head, it follows
that this head is also the host of the 0f„r\
Thus, the existence of a [T] feature, which expresses the temporal properties
of the infinitival clause, activates the C^,, head in the C-domain. Being acti-
vated, Cfln can host 0/ o r , which is responsible for the AccI syntax. In other
words, the relation between [T] and the licensing of 0for is indirect: [T]
licenses the position that 0for occupies and not the element itself.
322 Vassilios Spyropoulos
The discussion of the properties of AccI syntax has led to an analysis that
incorporates the following assumptions:
[Agr], since CIGr. infinitives do not inflect for subject agreement. However,
they can establish a checking relation with the DP-subject of the embedded
clause. As a result of this checking relation, [Agr] is licensed and the DP-
subject of the infinitival clause acquires accusative case. CIGr. infinitival
clauses with AccI syntax are therefore proposed to have the following
structure:
(64) [CP, C force ... [CP2 Cfin[T, Agr] [INFLP DP-subject INFL[+T] ...]]]
On the basis of (i) the complex data reviewed in the previous sections and
(ii) the analysis of AccI syntax outlined above, we can now formulate a full
account of CIGr. infinitival syntax, which is based on the general principles
that define the feature constitution of the functional heads and their licensing.
More specifically, we argue that CIGr. infinitival clauses have the universal
clause structure, which consists of the following functional heads:
DI structure allows for two options: When C fin is not specified for [Agr], a
control structure is established and the subject of the infinitival clause is a
caseless PRO. Depending on the case of the controller, this structure gets
the surface form of NomI, GenI, DatI and AccI constructions (69a,b,c, and
d respectively):
When Cfi„ is specified for [Agr], then an overt DP-subject in the accusative
case is licensed as a result of the checking relation established between the
[Agr] and the DP-subject. As a consequence, control is suspended and ob-
viation is established. The structure surfaces as an AccI construction:
Example (71) differs minimally from example (72), in that the controlled in-
finitival subject surfaces as an emphatic pronominal in the accusative case,
which is controlled by the matrix subject. Example (71) shows that CIGr re-
sorts to AccI syntax whenever emphasis demands that the infinitival subject
be overt. Theoretically speaking, by being assigned accusative case the em-
bedded subject escapes the case-filter (Vergnaud 1980; Chomsky 1981) and
is licensed so as to surface as an emphatic pronominal. Consider now the
examples (73) and (74), which involve GenI and DatI syntax respectively:
The syntax of Classical Greek infinitive 327
These examples were discussed in section 3.1 and it was concluded that the
(b) examples involve a hidden AccI syntax with a pro-ACC subject in the
infinitival clause, as evident from the accusative case of the predicate modi-
fier. It is therefore suggested that [Agr] in C fin can also license a null-
subject pro-ACC in the infinitival clause, which is controlled by a matrix
argument. Noticeably, morphological manifestation of [Agr] and pro is
achieved by means of the agreement relation established between pro and
the predicate modifier. As a consequence this structure is detectable only in
the presence of a nominal modifier that modifies the empty subject of the
infinitival clause, and the structure surfaces as an AccI construction. The
relevant structures are presented in (75):
328 Vassilios Spyropoulos
Finally, FIs also allow for two options, depending on the presence vs. ab-
sence of [Agr] in Cfj„. In the absence of [Agr], the subject of the infinitival
clause is assigned no case, allowing for control and raising. 15 These struc-
tures surface normally as NomI constructions (76a-b) and occasionally as
DatI constructions (76c):
The following diagram summarizes the relation between the possible infini-
tival structures with the feature specification of the relevant functional
categories and their surface manifestation in terms of the traditional gram-
mar descriptions:
6. Conclusions
In this paper, w e investigated the properties of infinitival clauses in CIGr. It
was shown that infinitival syntax in CIGr. is subject to the general principles
of U G that define the clause structure, the licensing of empty elements, and
the licensing of the features of the functional and lexical categories. On the
other hand, certain parameters determine the feature content of the relevant
functional categories of the C- and INFL-domains. More specifically, it
was shown that CIGr. infinitival INFL is specified as [+T, - A G R ] , because
infinitives carry tense but no agreement inflection. As a consequence, CIGr
infinitival INFL does not establish a checking relation with its subject,
which is therefore allowed to establish a checking relation with another
element, or to be a PRO. When not a PRO, the infinitival subject can estab-
lish a checking relation with either an element in the matrix clause (overt/
330 Vassilios Spyropoulos
Appendix
Notes
1. See Anttila (1989: Part III), Campbell (1998: Chapter 5), Szemerenyi (1999)
and Fox (1995) for overviews of the principles of the comparative method.
2. The term Classical Greek refers to a specific period of the Ancient Greek lan-
guage that spans between the 6 th and the 3 rd centuries BC. It describes the variety
used in the literature production of this period, which is in most cases written
in the Attic and Ionic dialects. Some representative grammatical descriptions of
the syntax of CIGr. can be found in Jannaris (1897), Smyth (1918), Schwyzer
(1950). Our data are drawn from the texts of this period. However, for the ease
of exposition, in some cases we also use elementary textbook examples, either
taken from existing grammar textbooks, or invented by ourselves. In the latter
case, it has been verified that the constructions represented in these examples
occur in the texts of the period. CIGr. examples are transliterated in a broad way,
which relies on the main phonological values represented by the graphemes
and abstracts away from phonological and phonetic details, such as length and
tone, when they are not represented by separate graphemes and are not gram-
matically significant.
3. Despite its significance in determining the morphosemantics of the CIGr. verbal
system, aspect seems to have no effect on the fmiteness of a clause and subse-
quently on the case marking of the subject of the infinitival clause. We will
therefore ignore the role of aspect in the CIGr. verbal system, because it is not
related with the issues of finiteness examined in this paper.
4. Future infinitive is suggested to denote punctual aspect (Schwyzer 1950;
Binnick 1991).
5. Literally, the Latin term is translated as 'verbs of saying and feeling'. In fact,
this category of verbs includes report-verbs and judge-verbs, such as lego: 'to
say', aggello: 'to bring the news', nomizo: 'to decide, to judge', dokeo: 'to be-
lieve, to decide', etc.
6. Sometimes, a future infinitive may be found after verbs that express promise,
will, hope, desire or expectation, as a result of denoting an inherently future
event:
(i) epaggellometha Ariaio.i... eis ton thronon
promise-1 PL Ariaeus-DAT in the throne-ACC
ton basileion kathiein {X. An. 11.1,4)
the royal-ACC sit-iNF.FUT
' W e promise to Ariaeus that he will ascend to the royal throne'
7. The following abbreviations are used in the paper. A O R : aorist, COMP: comple-
mentizer, FEM: feminine, IMP: imperative, INF: infinitive, MED: mediopassive,
M S C : masculine, N E G : negation, NTR: neuter, OPT: optative, PART: participle,
P A S S : passive, PL: plural, PRF: perfect, PRS: present, PRT: particle, PST: past,
SBJ: subjunctive, SG: singular.
332 Vassilios Spyropoulos
8. In the Government & Binding framework government from the matrix predicate
is proposed to assign the accusative case to the DP-subject of the infinitival
clause (Chomsky 1981; Kayne 1984). In the Minimalist Program government
has been dispensed with as a syntactic notion and case assignment has been
subsumed under the more general checking theory (Chomsky 1993). Two dif-
ferent analyses have been proposed in this framework: (i) ECM involves covert
case assignment either by means of covert feature movement to the vP projec-
tion of the matrix verb (Chomsky 1995), or by means of operation Agree es-
tablished for agreement checking between the matrix ν and the embedded DP-
subject (Chomsky 2000). (ii) ECM involves overt movement of the DP-subject
of the infinitival clause to a specifier position in the vP-domain of the matrix
predicate (Lasnik 2001).
9. This agreement pattern is retained to Modern Greek as well (see Spyropoulos
1998, 1999, 2005)
10. To be more specific, Philippaki-Warburton and Catsimali (1991) attribute the
case assigning properties of INFL to the positive specification of a finiteness
feature [fin]. They claim that [fin] in CIGr. infinitives can be positively speci-
fied, and thus assign accusative case, because they inflect for tense. This
analysis can be considered as the INFL variant of the analysis we will propose
in sections 4 and 5.
11. Philippaki-Warburton and Catsimali (1989, 1991) and Sevdali (2003) also pro-
pose that the infinitival INFL cannot assign nominative case because it is speci-
fied as [-AGR]. This constitutes an argument for the hypothesis that nomina-
tive case is, in fact, the result of subject-agreement checking (Chomsky 2000,
2001). Evidence for the agreement-nominative association also comes from
Portuguese inflected infinitives and Modern Greek subjunctives: (a) Portuguese
infinitives may inflect for subject-agreement; when such an infinitive is used
the subject of the infinitival clause appears in nominative case (Raposo 1987).
(b) Modern Greek lacks the category of infinitive and uses the subjunctive in-
stead in complementation. MG subjunctives inflect for subject agreement and
as a result nominative case is always assigned to its subject (Philippaki-
Warburton and Catsimali 1999; Philippaki-Warburton and Spyropoulos 2002,
in prep.).
12. See Raposo (1987).
13. But, see note 6.
14. Here, we follow the widely accepted assumption that the infinitival clause in
raising constructions is deprived of its CP status and it projects only to the
INFLP level (Chomsky 1981, 2001).
15. In raising structures the infinitival clause is deprived of its CP-status (see also
note 14). That CIGr. infinitival clauses may not sometimes project to the level
of CP is also evident from examples such as the following, which involve
ECM:
The syntax of Classical Greek infinitive 333
References
Chomsky, Noam
2000 Minimalist inquiries: The framework. In Step by step: Essays on
minimalist syntax in honor of Howard Lasnik, Roger Martin, David
Michaels and Juan Uriagereka (eds.), 89-115. Cambridge Mass.: The
MIT Press.
2001 Derivation by phase. In Michael Kenstowicz (ed.), 1-52.
Cecchetto, Carlo and Renato Oniga
2002 Consequences of the analysis of Latin infinitival clauses for the the-
ory of Case and control. Lingue e Linguaggio 1: 151-189.
Davies, William D. and Stanley Dubinsky (eds.)
2001 Objects and other Subjects: Grammatical Functions, Functional
Categories and Configurationality. Dordrecht: Kluwer.
Drachman, Gaberel, Angeliki Malikouti-Drachman, John Fykias and C. Klidi (eds.)
1997 Greek Linguistics '95: Proceedings of the 2nd International Confer-
ence on Greek Linguistics, Salzburg, September 1995\ Vols. I—II.
Graz: W. Neugebauer Verlag GmbH.
Fox, Anthony
1995 Linguistic Reconstruction: An Introduction to Theory and Method.
Oxford: Oxford University Press.
Haegeman, Liliane
1992 Generative Syntax: Theory and Description. A case study of West
Flemish. Cambridge: Cambridge University Press.
1994 Introduction to Government and Binding Theory. (2nd edition.) Ox-
ford: Blackwell.
Haegeman, Liliane (ed.)
1997 Elements of Grammar: A Handbook in Generative Syntax. Dordrecht:
Kluwer.
Hale, Ken and Samuel J. Keyser (eds.)
1993 The view from Building 20. Cambridge, Mass.: The MIT Press.
Huang, Cheng-Teh James
1982 Logical relations in Chinese and the theory of Grammar. [Unpub-
lished Ph.D. dissertation, MIT.]
Jannaris, Antonius Ν.
1897 An Historical Greek Grammar, chiefly of the Attic dialect. London:
McMillan.
Joseph, Brian D., Geoffrey Horrocks and Irene Philippaki-Warburton (eds.)
1998 Themes in Greek Linguistics II. Amsterdam: John Benjamins.
Kayne, Richard
1984 Connectedness and Binary Branching. Dordrecht: Foris.
Kenstowicz, Michael (ed.)
2001 Ken Hale: A life in language. Cambridge Mass.: The MIT Press.
Landau, Idan
2004 The scale of finiteness and the calculus of control. Natural Language
and Linguistic Theory 22: 811-877.
The syntax of Classical Greek infinitive 335
Lasnik, Howard
2001 Subjects, objects and the EPP. In Objects and other Subjects: Gram-
matical Functions, Functional Categories and Configurationality,
William D. Davies and Stanley Dubinsky (eds.), 103-121. Dordrecht:
Kluwer.
Longobardi, Giuseppe
1996 The Syntax of N-Raising: A Minimalist Theory. (OTS Working Papers
in Linguistics 96-005.)
Manzini, Rita
1983 On control and control theory. Linguistic Inquiry 14: 421-^146.
Martin, Roger
1996 A Minimalist Theory of PRO and Control. Unpublished Ph.D. dis-
sertation, University of Connecticut.
2001 Null case and the distribution of PRO. Linguistic Inquiry 32: 141-166.
Martin, Roger, David Michaels and Juan Uriagereka (eds.)
2000 Step by step: Essays on minimalist syntax in honor of Howard Lasnik.
Cambridge Mass.: The MIT Press.
Morrell, K. S.
1989 Studies on the Phrase Structure of Early Attic Prose. Unpublished
Ph.D. dissertation, Harvard University.
Ormazabal, Javier
1995 The Syntax of Complementation: On the Connection between Syn-
tactic Structure and Selection. Unpublished Ph.D. dissertation, Uni-
versity of Connecticut.
Pesetsky, David
1991 Zero Syntax, vol. 2: Infinitives. Unpublished MS.
Pesetsky, David and Esther Torrego
2001 T-to-C Movement: Causes and consequences. In Ken Hale: A life in
language, Michael Kenstowicz (ed.), 335-426. Cambridge Mass.:
The MIT Press.
Pillinger, Ο. S.
1980 The accusative and infinitive in Latin: A Refractory Complement
Clause. Journal of Linguistics 16: 55-83.
Philippaki-Warburton, Irene and Georgia Catsimali
1989 'Accusativus-cum-infinitivo' sta arxea elinika (Accusativus-cum-in-
finitivo in Ancient Greek). Studies in Greek Linguistics 10: 89-108.
1991 The theory of control and raising in Ancient Greek and the status of
PRO and NP-trace: Implications for the definition of subject. Unpub-
lished MS.
1997 Control in Ancient Greek. In Greek Linguistics '95: Proceedings of
the 2nd International Conference on Greek Linguistics, Salzburg,
September 1995: Vols. I—II, Gaberel Drachman, Angeliki Malikouti-
Drachman, John Fykias and C. Klidi (eds.), Vol. II, 577-588. Graz:
W. Neugebauer Verlag GmbH.
336 Vassilios Spyropoulos
Szemerenyi, Oswald J. L.
1999 Introduction to Indo-European Linguistics. Oxford: Oxford Univer-
sity Press.
Tantalou, Niki
2003 Infinitives with overt subjects in Classical Greek. Studies in Greek
Linguistics 23: 358-365.
Latin object and subject infinitive clauses
Lucio Melazzo
The main Harvard Professor Joshua Whatmough, a
rather pompous character, got up, interrupted after
10 minutes or so: "How would you handle..." and
then he mentioned some obscure fact in Latin. I
said I didn't know and tried to go on, but we got
diverted and that's what we talked about for the
rest of the time...
(N. Chomsky, On Nature and Language)
1. The data
Should any demonstration be needed, sentences (1) - (3) show that infini-
tive clauses in Latin exhibit a subject in the accusative.1
si a virtute defeceris
if from virtue^ departPF.SUBJ.2.sG
'It is hard for friendship to remain if you have forsaken virtue.'
(CIC. Amic. 11.37)
The subject is always in the accusative, regardless of the type of the infini-
tive clauses. In conformity with traditional classification these are divided
into object clauses, e.g. those in (1); subject clauses, e.g. those in (2); and
epexegetical clauses, e.g. those in (3).
Besides infinitive clauses, there was another type of clause which may
be held equivalent to the finite ίΛαί-clause of English, as exemplified in (4).
In this type the clause introduced by quod had its verb in the indicative as
in (4a), (4b) and (4d), or in the subjunctive as in (4c) and its subject, when
phonetically realized, in the nominative. It is precisely the continuation of
342 Lucio Melazzo
this type with the so called declarative quod that is found in the Romance
languages. 2 In these languages, on the other hand, there exist no infinitive
clauses corresponding precisely to the Latin infinitive clauses of the type
exemplified in ( l ) - ( 3 ) . This situation is related to a diachronic change
which affected this area of Latin syntax: quod-clauses of the kind presented
in (4) had begun to supplant infinitive clauses in Latin since the 2 nd century
AD. 3 Bearing in mind this circumstance and considering that both infinitive
clauses and quod-clauses in Latin could occur in the same contexts intro-
duced by the same verbs, school grammarians speak of object quod-c\msQS,
subject quod-c\auses, epexegetical object quod-clauses, and epexegetical
subject quod-clauses. These types are exemplified in 4(a-d). They also de-
fine clauses like those in (4) introduced by quod as explicit (i.e. finite), and
describe the corresponding clauses in the infinitive as implicit (i.e. non-
finite).
In (5) two distinct accusative forms combined with doceo, while in (6)
aliquem was the object of doceo. Thus (7) can be explained in the light of
(6).
The pronominal form me in (7) was originally the object of dicunt and then
could be understood as a real subject of the infinitive form uenire. This, on
the other hand, was intrinsically a verbal noun but eventually came to be
equated with finite verb forms with their distinctions of tense and concord.
Bolkenstein (1976) rightly rejected the hypothesis, supported by Ernout-
Thomas (1953: 271), that Acl originated in structures where perception
verbs were involved, for these verbs always admit of two interpretations: the
one has a merely perceptive value, the other is connected with a perceptive-
cognitive purport. 8
Bolkenstein (1976) also excludes that Acl has structures with verbs like
admonere 'to advise' as its starting point. In point of fact, sentence (8) allows
of two interpretations, as (9) and (10) make these clear.
In (9) eum depends directly on the main verb, which is not the case in (10).
On analogous grounds Miller (1974) denies that the origin of Acl can be
found in structures with a verb like docere, as was implicitly suggested by
344 Lucio Melazzo
Object clauses like him to be a fool in (14) are in fact thought not to have a
left periphery, i.e. a CP-layer, and to be the complement of the upper V.
This is what is usually called ECM (EXCEPTIONAL CASE-MARKING), a no-
tion that Chomsky (1995) deals with again. As a notion, ECM is generally
accepted in generative grammar and continues to stimulate researchers to
further efforts to explain it. See e.g. Lopez's (2001) and Tanaka's (1999)
recent attempts. This hypothesis has been accepted and adapted to shed
light on Latin infinitive clauses even though Chomsky (1988: 140, note 25)
himself recognizes that the case of accusative subjects of infinitives in
Greek or Latin remains more problematic. Chomsky's theory finds an appli-
cation to Latin in Calboli (1983) and Maraldi (1983). According to Calboli
(1983), 12 (15) should have (16) as a structure.
In (17) dico is not a control verb. So it can assign the objective case to the
pronoun te in as much as the IP of the esse-clause, not being dominated by
Latin object and subject infinitive clauses 347
Further, from the contrast displayed in (19), which I have taken from Ra-
poso (1987: 87), it is clear that a similar situation is to be found in Portu-
guese:
trabalhado pouco ]
work PPPLE littleADV
Ί think /maintain that the delegates have worked little (i.e. done
little work).'
b. *Eu penso / afirmo [05 deputados terem
INOM think / maintainPRES.i.SG [theNOM delegates NOM haveINF PRES
trabalhado pouco],
workp.PPLE littleADV]
Ί think / maintain that the delegates have worked little.'
(21) a. In the absence of [+Tense], Infi (or Agr in Infi) is capable of as-
signing (the 20 ) nominative Case to a lexical subject only if it is it-
self specified for Case
b. penso [CP[C[C/T[I Agr],] [ IP os deputados [,· t, VP]]]].
tibi, once they have been moved, must have ended up in some functional
position, and the place where they are found in their respective strings must
be in the structure of split-CP which has been proposed by Rizzi (1997). I
shall not go into details about this matter now. 2 2 1 shall instead quote (24) in
support of what I have just stated.
Accordingly, (24) helps to show that object infinitive clauses in Latin can-
not be regarded as devoid of left periphery. In point of fact, scire has moved
upwards beyond its subject, which is the long anaphor se. The facts speak
for themselves, and (24) has not been chosen at random. It is an instance
taken from a prose work by such a writer as Cicero, who is esteemed an
authority of the utmost importance on Latin language and style. 23 Of course,
many passages could have been cited here from texts either in verse or in
prose, but an instance taken from prose writings has been preferred to con-
vince those who believe that poetry enjoys a great deal of freedom to
change the normal rules of language.
If I am on the right track and Latin object infinitive clauses also have a
left periphery, then credence should be given to L a k o f f s (1968: 5) view that
the structures of two constructions which can be esteemed equivalent to
each other are unlikely to be different. 24 So similar were infinitive clauses
and gwoJ-clauses that the latter was to supplant the former, with which it
had coexisted for a long time. Hence, to speak of infinitive clauses as pos-
sessing the structure of a CP as well does not in any way sound odd.
ceivable way of solving the problem would be for (3a) to have a structure
like (25).
(25) CP
ego, id, IP
t'/ respondeoi VP
te dolorem... fuisse id
ferre moderate humanitatis tuae
potuisse
non commoveri
As can be easily seen in (25), the first id in (3a) is assumed to have landed
at a place in the highest CP-layer - possibly Spec,Foe - by moving from the
specifier position of a DP which has an &P as its complement. Alternatively,
it could be assumed that it occupies the head position in DP, and the matrix
DP is focussed, with the &P extraposed. The complement of this &P, whose
head is the second nec of (3a), is the CP represented by the string fuisse id
humanitatis tuae, while its specifier position is filled by another &P with
the first nec in (3a) as a head and two CPs occupying the other two posi-
tions: te dolorem... ferre moderate is found in the position of the specifier,
whereas potuisse non commoveri occurs in that of the complement.
352 Lucio Melazzo
It is now worth considering cases like (4c), which undoubtedly permits one
to infer (26) as its structure.
reprehend is,
solere^ me w ty dicas
ι* tw de m e ipso
gloriosius praedicare
When considered carefully, (26) indicates that hoc in (4c) has undergone a
m o v e m e n t parallel to that described for the id of (3a) in (25): the starting-
point is identical and the final destination in the CP-layer might also be the
same. Consistently with both the school grammarians' view mentioned at
the end of section 1 and the observation that the use of a non-finite verb in
a Latin A c l construction is always accompanied by a null complementizer,
a finite verb being only found with an overt complementizer, (25) and (26)
suggest that both Latin object clauses (that with the verb in the infinitive
and that introduced by quod) are base-generated in the complement posi-
tion of a D P with another DP as a specifier 2 6 while, as is shown in (25), the
possibility is not excluded that a &P is inserted into this complement posi-
tion of the former DP. 2 7 Odd though it might appear, my hypothesis that
Latin Acl constructions are CPs embedded within a DP the specifier of
which is filled by another DP is not groundless. After all, this proposal pro-
vides some syntactic foundation for the semantic intuition that Acl con-
Latin object and subject infinitive clauses 353
Note that, unlike Latin, Ancient Greek had the definite article, which origi-
nally was a demonstrative pronoun. In the neuter gender, this Greek form
could precede infinitive clauses. It is all grist to the mill from my perspec-
tive, for this further evidence is corroboration of the interpretation I propose.
Indeed, to in (27) has landed at a position in the higher CP-layer as id has
in (3a). In (3a) id is preceded by ego, the topicalized subject of respondeo;
so it can be assumed to occupy the lower Spec,Top of the higher CP-layer.
In (27) to is followed by the enclitic pronoun min in compliance with
Wackernagel's (1892) law. Being the subject of the infinitive clause with
tethndmen as a predicate, the accusative form min must consequently have
moved to its final place via some position in the lower CP-layer. 29 The syn-
tactic status of Wackernagel's Law should be clarified before reaching a
decision on this matter. 30 As to the position from which to in (27) started to
move, consider (28).
354 Lucio Melazzo
Here D°, combining with CP, also has a DP as its specifier. Like id in (3 a),
to in (27) might have been base generated in the DP filling the specifier
position in (30). 31 For the time being, however, I am not able to be any
more precise.
Further clarification of my hypothesis on the structure of Latin infinitive
clauses can be provided by (31)—(33), in that they give more details about
(22)-(24), which were cited in the previous section.
Latin object and subject infinitive clauses 355
(31) CP
haec, IP
pro, meminit VP
t, t* &P
t, &P
et DP
CP
P R O victum, IP
frustra y contendere.
ty Thyrsin w t'z VP
tx t w t 2
(32) DP
plurima CP
quae, IP
t, invideanti VP
DP VP
Δ
t/ tk DP
CP
pure ; adparere v IP
tibi v
rem,,, tV t, VP
tvv t.v t y
356 Lucio Melazzo
(33) CP
negat, IP
Piso, t, VP
t, t, DP
pro CP
scire* IP
se* f , VP
ti t j quicquam
(34) kwo- ... *to- *kwo- ... */- -> *i- ... *kwo- 0 ... *kwo-
Phase 1 Phase 2 Phase 3 Phase 4
This is the pattern which Haudry (1973) and Lehmann (1974) have called
'diptyque normale' and 'correlative diptych' respectively. In this pattern, in
which the w/z-word originally preceded and eventually followed the deter-
miner, the correlation had a meaning which can be exemplified by expres-
sions like 'the man who/the thing which... this man!this thing...' corre-
sponding to the Latin 'qui!quod... is/id...'.32 The transition from the Indo-
European to the Latin phase has been convincingly stated by Haudry (1973),
who recognizes three different stages in the process of change. Here it is
worth noting that the w/2-form could be converted to a complementizer only
when, as described by Phase 3 in (34), the two members of the diptych had
already inverted their positions. 33 In all likelihood it can be maintained that
the structure of both the Latin constructions here at issue, that with the sub-
ject in the accusative and the verb in the infinitive, and that with quod and a
finite verb form, is directly related to Phase 3 or Phase 4 of (33), the neuter
pronoun, either overt or covert, and the infinitive clause or the quod-clause
being the two members of a correlative diptych.
As things stand, the structure of both the Latin object quod-clause or, as
will be said in section 7, the subject quod-clause and its or their counterpart
with covert C° and verb in the infinitive somewhat resembles that of Eng-
lish relative clauses which has been proposed by Kayne (1994) and is re-
produced here as (35).
Undoubtedly, omnes and ego in (36) and Artabanus in (37) are in the
nominative. Both (36) and (37) exhibit verbs in the present infinitive where
the imperfect indicative form is expected. Displaying root sentences, (36)
does not have any overt C°, while ubi is found in the first part of (37),
which is an adjunct with a temporal meaning. On the other hand, as shown
by (38), a subject in the accusative occurs with an exclamatory infinitive.
These facts cannot be left out of consideration in any attempt to explain the
case form in which the subject of the infinitive is found.
If the 'diachronic' interpretations put on Latin Acl and discussed above
in section 2 are left aside and the analysis is not limited to the language of
ancient Rome, then it is clear that the explanations proposed so far by
scholars either accept E C M or bring up the TP or the C P or both of them.
As has been shown above, a CP cannot be disposed of in Acl; hence the
ECM-hypothesis does not hold good, and nothing remains except for me to
search for a solution involving the CP-layer as well. My suggestion profits
from the indisputable fact that a complementizer is connected with certain
properties of the predicate of a clause. Thus, for example, a complementizer
is chosen in accordance with the finite or non-finite character of an embed-
ded clause and, in the case of Latin finite embedded clauses, quod intro-
duces those with a verb in the indicative and ut is selected for those with a
subjunctive. Consequently, a complementizer can be thought to match the
inflectional features of the verb in the embedded clause. To some extent my
hypothesis is in line with those proposals which suggest a connection of the
temporal properties of a clause with the C-domain. According to the analyses
of Stowell (1982), Chomsky (2001), Martin (2001), the temporal properties
of 1° are selected by a tense feature [T] of C° and licensed through a covert
movement that ensures their matching. Furthermore, Rizzi's (1982) and
R a p o s o ' s (1987) explanations of the Italian and Portuguese infinitive
clauses that have been reported above in section 3 go in the same direction,
and Longobardi (1996) assumes that a covert operation of V-to-C raising
licenses the accusative case of the subject in English gerunds.
In the light of so close a connection between C P and IP I shall assume
that the bundle of features which substantiates FIN° in Split-CP includes a
specification for case. This comes into play when the case of the subject
cannot be licensed in the IP-layer in accordance with what is common
knowledge in generative grammar. When this condition exists, the subject
rises overtly or covertly to the higher projection to have its case checked.
The specific case-feature will be [+Acc] if the structure containing a verb
in the infinitive form is a complement clause, and it will be [ + N o m ] other-
wise. I assume that the infinitival CP is a complement to a D head only when
the infinitival clause functions as an argument of the matrix predicate. Hence
root and adjunct infinitival clauses are generated without a subsuming DP.
That is, the FIN° head can have either of the following two sets of features:
360 Lucio Melazzo
Evidently, the accusative case of the infinitival subject in (31), (32), and
(33) is licensed by the feature specification in (39), whereas the nominative
case of the infinitival subject in (36), and (37) is licensed by (40). (38),
containing an exclamative root infinitive with an accusative subject, ap-
pears at first sight to represent a problem for (40). In fact, we could assume
that an exclamative clause is complement to an implicit speech act verb.
This assumption is supported by the fact that in many languages exclama-
tive clauses begin with the regular subordinating complementizer. Under
this proposal, the FIN° head of (38) is specified as in (39) 34 .
My explanation also receives confirmation from (41).
Now is the time to tackle the problem of subject infinitive clauses, a type
exemplified by (2a) and (2b). From (3b) it is possible to infer that the struc-
ture of these clauses is not different from that of object infinitive clauses
given in (25). If I am on the right track, in fact, in (3b) the DP containing
illud has moved into Spec,IP, whereas the infinitival complement of D has
been extraposed. 35 Moreover, the reason why the subject me carries the
Latin object and subject infinitive clauses 3 61
accusative case is the same as that which has been given above with respect
to object infinitive clauses. Further, (4d), which exhibits a subject quod-
clause like (4b), can be assumed to have a similar structure to that in (26)
once the quod-clause is admitted to have been moved upwards. Subject in-
finitive clauses also have a CP-layer. This is shown by (42).
Here abesse has in fact moved leftwards across the subject tris cohortis
(= tres cohortes) to arrive at a position, in the embedded CP-layer, lower
than that of in tanta militum paucitate, which is topicalized.
In this respect the main clause (2b) might also give further evidence. It
is generally assumed that a D selecting a clause subcategorizes a CP. As a
matter of fact, in such a null-subject language as Latin, a pro associated
with the infinitive clause is evidence of a DP projection above CP. In (2b)
this pro is specified not only as 3 rd person singular but also as neuter in
gender. This can be easily inferred from the form of the adjectival predicate
difficile, which is also neuter in gender and in the nominative case, for in
Latin copular constructions subject and predicate agree in case and, if the
predicate is not a noun, in gender as well. On the other hand, the perfect
participle nuntiatum is also neuter and is also in the nominative in (2a),
where the presence of a pro is to be supposed, and the same specifications
are exhibited by the gerundive form faciendum in (3b). Here, however, to
use traditional terms, an overt illud is to be found, which is a proleptic neuter
pronoun responsible for the epexegetical character of the subject infinitive
clause, and which has, in my opinion, the same syntactic origin as the first
id of (3 a).
Exactly like that of object infinitive clauses, the CP-layer of subject in-
finitive clauses has a FIN° distinguished by [+ Complement] 35 and [+ Acc],
This should suffice to explain the case of their subject. Once both object and
subject clauses are assumed to have one and the same structure, a solution
to a problem which scholars have long puzzled over seems to have been
found. Coherently, in subject infinitive clauses subjects take the accusative
case for the same reason why the subject of object infinitive clauses is in
the accusative.
362 Lucio Melazzo
I could consider my paper as concluded for the time being if the construction
traditionally called NOMINATIUUS CUM INFINITIUO (Ncl) were not to be ac-
counted for. Like (43), (44) displays a construction which evidently in-
volves a subject infinitive clause.
In (44)-(45) the accusative case feature of the FIN° head of the infinitive
clause has not been activated. These sentences are presumably saved in the
following way: the subject of the infinitive clause is raised into the empty
specifier position of the DP subsuming the infinitival CP. When this DP is
moved into the Spec,IP position of the matrix clause, the nominative licen-
sed by the matrix inflection, generally realized on the pronoun in the speci-
fier of the DP, is now realized on the subject of the infinitive raised to this
position. Consider the structure of (44). Assuming that the subject of the
infinitive, Caesar, is in the specifier of the DP occupying the matrix
Spec,IP position, the remnant infinitive clause must have been removed
from under the DP into the matrix CP domain.
Latin object and subject infinitive clauses 363
Naturally, it can happen also in the case of object infinitive clauses selected
by a transitive verb that the case feature of the FIN° head is not activated,
however, the subject of the infinitive can assume accusative case in the
empty specifier position of the object DP subsuming the infinitive clause.
8. Conclusion
I may have taken my readers along a tortuous road, but I believe I have had
both feet on the ground. Though certainly synchronic, the explanation I
have proposed involves consideration of diachronic features going very far
back in time. On the whole, I think that the implication of diachronic con-
siderations in synchronic analyses can also contribute to our understanding
of language and mind. Much still remains to be done. Firstly, my hypothe-
sis must be checked by taking into account the data systematically extracted
from a representative sample of Latin texts from different periods. Secondly,
the other constructions containing an infinitive must be accounted for.
Thirdly, a definite attempt should be made to obtain cross-linguistic gener-
alizations. As this is a weighty matter, I willingly leave the question open
for the time being, in the hope I may soon return to these matters with fur-
ther data and explicative solutions.
364 Lucio Melazzo
Notes
1. Whenever it could be done, I resorted to actual quotations from Latin texts and
I mentioned them in round brackets. Latin and Greek passages have been for
the most part quoted from the critical editions of both series of the Clarendon
Press (Oxon), but I also made use of Borzsäk (1992), Guillemin (1961), Lau-
rand (1965), Wuilleumier (1961). This will make clear which system of chap-
ter, paragraph, etc., division, or of pagination, is being followed. Latin and
Greek authors and works are printed in the abbreviated forms adopted in the
Oxford Latin and in the Oxford Greek Dictionary respectively.
In producing the glosses on the single words of the Latin and Greek exam-
ples I have presented as evidence I have made use of the following abbrevia-
tions: 1. = first person; 2. = second person; 3. = third person; ABL. = ablative;
ACC. = accusative; ADJ. = adjective; ADV. = adverb; AOR. = aorist; ART. = arti-
cle; COMPAR. = comparative; CONJ. = conjunction; DAT. = dative; F. = femi-
nine; FUT. = future; GDVE. = gerundive; GEN. = genitive; IMP. = imperative;
IMPF. = imperfect; IND. = indicative; INDEF. = indefinite; INF. = infinitive; IN-
TERR. = interrogative; M. = masculine; MD. = middle; N. = neuter; ΝΟΜ. =
nominative; p. = past; PASS. = passive; PF. = perfect; PL. = plural; PPLE. = par-
ticiple; PREP. = preposition; PRES. = present; PRON. = p r o n o u n ; PTCL. = parti-
cle; REL. = relative; SG. = singular; SUBJ. = subjunctive; SUPERL. = superlative.
So that readers who are not acquainted with Latin could easily identify the in-
finitive clauses and the quod-clauses found in the examples that are presented,
I thought it useful to have them printed in bold type. As far as possible, my
English translations aim to be reliable, but I am well aware that the differences
existing between English and Latin might not make things easy for some, even
most readers. These are kindly requested to read the glosses carefully. For my
part I assure them I have consulted English speakers about this matter, but I
must confess that I have not been so diligent and credulous as to look for the
Latin speakers (sic\) to whom my first anonymous reviewer referred.
2. Actually, quod was replaced by quid in Late Latin. As is found in Rohlfs
(1969: 188) the Italian clause-introducing word che derives from this Latin
word quid. In standard Italian che is a multifunctional complementizer and
thus corresponds to French que, which has the same etymological origin, to
English that and German daß. By reason of its characteristics of a generalized
complementizer, che can introduce more or less all the subordinate clauses, in-
cluding relative clauses where it can occur as either a subject or a direct object
in place of the pronoun. For a comprehensive discussion of the matter, cf.
Cinque (1988: 446 and 463-465).
3. See Cuzzolin (1994) for extensive discussion on this.
4. Lehmann (1988) mentions some parallel cases in Quechua.
5. See Miller (1974), Disterheft (1980) and Coleman (1985). It is also worth no-
ticing that, as Heftrich (1992: 223) states, this type of construction can be held
Latin object and subject infinitive clauses 365
to have existed solely in Ancient Greek and Latin, provided the denomination
Acl is confined only to infinitive clauses depending on both mental-state and
speech-act verbs ( V E R B A SENTIENDI ET DICENDI).
6. The traditional meaning of the double accusative includes constructions corre-
sponding, for example, to that of (i).
(i) Caesar Aeduos frumentum flagitabat
CaesarNOM AeduiACC.PL cornACC demandIMPF.IND.3.S0
'Caesar demanded corn from Aedui.' (CAES. Gal. 1.16.1)
7. Calboli's paper also includes the previous history of the matter here at issue
with a useful bibliography.
8. See also Pinkster (1990: 196-197) and Maraldi (1980).
9. Of course, such an explanation is rejected by Bolkenstein (1976).
10. This appears to be anything but a new idea. See Cuzzolin (1994: 38, note 18).
11. Pillinger (1980) must be counted in as well, but see Comrie's (1981) reply to
his paper.
12. The same scholar asserts that an earlier infinitive clause would have been the
nominative+infinitive construction typical of ergative languages with SOV or-
der; the Acl construction would have developed once Latin had become a
nominative-accusative language.
13. I shall not treat the syntax of sentences containing a verb of command in this
paper, for I think they still need to be analysed thoroughly.
14. As to the structures of (15) and (17), see also Maraldi (1983).
15. On the other hand, a sentence like (3a) does display id as the object of the main
verb respondeo. This makes it unlikely that this verb governs other objects.
16. With reference to the accusative as the default case, see also Calboli (1996).
17. Being assumed to have an affix status, covert complementizers are thought to
be incorporated into the matrix verb (see Stowell 1981; Kayne 1984; Pesetsky
1996; Ormazabal 1995). As a local operation, incorporation is not allowed
when adjacency is not complied with (see Boskovic 1997; Martin 1996). On
that account Latin 0 C O M p differs from English 0tha, in some respects.
18. Apart from essere 'to be' and avere 'to have', also but not necessarily em-
ployed as auxiliaries, the other verbs that can occur in the infinitive in such
subordinate clauses are dovere 'to have to, must', potere 'can, to be able to',
trattarsi 'to be a matter of, to be about', esistere 'to exist', and spettare 'to be
the concern o f . For some first information on these Italian infinitive clauses
with overt subject, cf. Skytte and Salvi (1991: 527-529).
19. As (ii) displays, os deputados in (19a) is also in the nominative.
(ii) Sera dificil [eles aprovarem a proposta].
difficult [heNOM.P1. approve^.,», the proposal]
BeFUT.IND.3..SÜ
28. Even though she does not express herself in exactly the same way, this is the
sense of Giusti's argument, which takes Renzi's (1997: 1 - 1 1 ) proposal into
account. See also Bianchi (1999: 225).
29. As will become plain in the next section, this personal pronoun did not need to
move from where I think it had been base-generated in order to check its accu-
sative case.
30. See note 4 of Lanzetta-Melazzo's paper in this miscellany.
31. If this is the case, then Homeric Greek may show signs of a development of the
demonstrative into the definite article similar to that noticeable in the transition
from Latin to Romance languages. As to does not seem to be the head which
the CP permits to project into a larger DP structure, it can be inferred that at
the age of the composition of the Homeric poems the demonstrative had not
yet completed its transition to being an article - which is what is traditionally
said, after all.
32. The pattern was also realized in Latin with pairs of forms like ' q u o ("in the
same proportion as")...eo ("by that degree")...', ' q u o m o d o ("in the manner in
w h i c h " ) . . . e o d e m modo ("in the same m a n n e r " ) . . . ' , ' q u o a d ("up to the time
that")...usque eo ("up to that point in t i m e " ) . . . ' , the wÄ-stem being as recog-
nizable in the first item of each pair as the /-stem in the second.
33. It is worth noticing that the ancestor of the English complementizer that and its
corresponding forms in the other languages of the Germanic group had *to- as
a base. This could be explained by assuming the original existence of the cor-
relative diptych.
34. This hypothesis vaguely recalls one of Harris's (1976) proposals.
35. In (3b), instead of id, illud occurs, which is the neuter form in the nominative
case of ille, ilia, illud 'that', another determiner with a demonstrative value
analogous to is, ea, id 'he, she, it'.
References
Baldi, Philip
1983 Speech Perception and Grammatical Rules in Latin. In Latin Linguis-
tics and Linguistic Theory, Proceedings of the 1st International Collo-
quium on Latin Linguistics. Amsterdam 1981, Harm Pinkster (ed.),
11-26. Amsterdam/Philadelphia: John Benjamins.
Bianchi, Valentina
1999 Consequences of Antisymmetry. Headed Relative Clauses. Berlin/
N e w York: Mouton de Gruyter.
Bolkenstein, Machtelt
1976 The Relation between Form and Meaning of Latin Subordinate
Clauses Governed by verba dicendi in Latin. Mnemosyne 29: 1 5 5 -
175,268-300.
368 Lucio Melazzo
Bolkenstein, Machtelt
1989 Parameters in the Expression of Embedded Predications. In Subordi-
nation and other Topics in Latin. Proceedings of the Third Colloquium
on Latin Linguistics, Bologna, 1-5 April 1985, Gualtiero Calboli
(ed.), 3-35. Amsterdam/Philadelphia: John Benjamins.
Borzsäk, Stefan
1992 Cornelii Taciti Ab Excessu Divi Augusti Libri I—VI. Stutgardiae et
Lipsiae: In /Edibus B. G. Teubneri.
Boskovii, Zeljko
1997 The Syntax of Non-finite Complementation. Cambridge, Mass.: The
MIT Press.
Calboli, Gualtiero
1962 Studi grammaticali. Bologna: Zanichelli.
1983 The development of Latin (Cases and Infinitive). In Latin Linguistics
and Linguistic Theory, Proceedings of the Γ1 International Collo-
quium on Latin Linguistics. Amsterdam 1981, Harm Pinkster (ed.),
41-57. Amsterdam/Philadelphia: John Benjamins.
1996 The accusative as a default case in Latin. Aspects of Latin. In Papers
from the seventh International Colloquium on Latin Linguistics, Jeru-
salem, April 1993, Hannah Rosen (ed.). Innsbruck: Sonderdruck.
Cecchetto, Carlo and Renato Oniga
2002 Consequences of the Analysis of Latin Infinitival Clauses for the
Theory of Case and Control. Lingue e linguaggio 1: 151-189.
Chomsky, Noam
1981 Lectures on Government and Binding. Dordrecht: Foris.
1988 Lectures on Government and Binding. Dordrecht/Providence: Foris.
1995 The Minimalist Program. Cambridge, Mass.: The MIT Press.
2001 Derivation by Phase. In Ken Hale: A Life in Language, Michael
Kenstowicz (ed.), 1-52. Cambridge Mass.: The MIT Press.
Christol, Alain
1989 Prolepse et syntaxe indo-europeenne. In Subordination and other
Topics in Latin. Proceedings of the Third Colloquium on Latin Lin-
guistics, Bologna, April 1985, Gualtiero Calboli (ed.), 65-89. Am-
sterdam/Philadelphia: John Benjamins.
Cinque, Guglielmo
1988 La frase relativa. In Grande grammatica italiana di consultazione. I.
La frase. I sintagmi nominale e preposizionale, Lorenzo Renzi (ed.),
443-503. Bologna: II Mulino.
1999 Adverbs and Functional Moods: A Cross-linguistic Perspective. New
York/Oxford: Oxford University Press.
Coleman, Robert
1985 The Indo-European Origins and Latin Development of the Accusative
with Infinitive Construction. In Syntaxe et Latin, Christian Touratier
(ed.), 307-342. Aix-en-Provence: Universite de Provence.
Latin object and subject infinitive clauses 369
Comrie, Bernard
1981 The Theoretical Significance of the Latin Accusative and Infinitive:
A Reply to Pillinger. Journal of Linguistics 17: 345-349.
Cuzzolin, Pierluigi
1994 Sull'origine della costruzione dicere quod: aspetti sintattici e seman-
tic!. Firenze: La Nuova Italia Editrice.
Dayal, Veneeta
1994 Scope marking: cross linguistic variation in indirect dependency. In
Wh-Scope Marking, Ulrich Lutz, Gereon Müller and Arnim von
Stechow (eds.), 157-193. Amsterdam: John Benjamins.
Disterheft, Dorothy
1980 The Syntactic Development of the Infinitive in Indo-European.
Columbus, Ohio: Slavica.
Elerick, Charles
1994 How Latin Word Order Works. In Papers on Grammar, IV, Gualtiero
Calboli (ed.), 99-117. Bologna: CLUEB.
Ernout, Alfred and F r a n c i s Thomas
1953 Syntaxe Latine. 2 e edition revue et augmentee. Paris: Klincksieck.
Giusti, Giuliana
2001 The Birth of a Functional Category: from Latin ILLE to the Ro-
mance Article and Personal Pronoun. In Current Studies in Italian
Syntax. Essays Offered to Lorenzo Renzi, Guglielmo Cinque and
Francesco Salvi (eds.), 157-171. Amsterdam/London/New York/
Oxford/Paris/Shannon/Tokyo: Elsevier.
Guillemin, Anne Marie
1961 Cornelius Nepos. CEuvres. Paris: Societe d'Editions «Les Belles
Lettres».
Hahn, Adelaide
1950 Genesis of the Infinitive with Subject-Accusative. Transactions and
Proceedings of the American Philological Association 31:117-129.
Harris, Zellig S.
1976 Notes du cours de syntaxe. Paris: Editions du Seuil.
Haudry, Jean
1973 Parataxe, hypotaxe et correlation dans la phrase latine. Bulletin de la
Societe linguistique de Paris 68: 147-186.
Heftrich, Heinrich
1992 Die Entstehung des lateinischen und griechischen Acl. Rekonstruktion
und relative Chronologie. In Akten der VIII. Fachtagung der Indo-
germanischen Gesellschaft, Leiden, 31. August-4. September 1987,
Robert Beekes, Alexander Lubotsky and Jos Weitenberg (eds.), 221—
234. Innsbruck: Institut fur Sprachwissenschaft der Universität.
Hofmann, Johann Baptist and Anton Szantyr
1965 Lateinische Syntax und Stilistik. München: Beck.
370 Lucio Melazzo
Horvath, Julia
1997 The status of 'w/z-expletives' and the partial movement construction
in Hungarian. Natural Language and Linguistic Theory 15: 509-572.
Kayne, Richard S.
1984 Connectedness and Binary Branching. Dordrecht: Foris.
1994 The Antisymmetry of Syntax. Cambridge, MA, London: The MIT
Press.
Kurzovä, Helena
1989 Subordination in Latin. Eirene 26: 23-26.
Lakoff, Robin Τ.
1968 Abstract Syntax and Latin Complementation. Cambridge, Mass./
London: The MIT Press.
Laurand, Louis
1965 Ciceron. L 'amitie. Paris: Societe d'Editions «Les Belles Lettres».
Lehmann, Christian
1988 Towards a Typology of Clause Linkage. In Clause Combining in
Grammar and Discourse, John Haiman and Sandra A. Thompson
(eds.), 181-225. Amsterdam/Philadelphia: Benjamins.
Lehmann, Winfred Paul
1974 Proto-Indo-European Syntax. Austin: University of Texas Press.
Longobardi, Giuseppe
1996 The Syntax of N-raising: A Minimalist Theory. OTS Working Papers
in Linguistics, 96-105. Utrecht: Utrecht Institut of Linguistics.
Lopez, Louis
2001 On the (Non)complementarity of Θ-Theory and Checking Theory.
Linguistic Inquiry 32: 624-716.
Maraldi, Mirka
1980 The Complement Structure of Perception Verbs in Latin. In Papers
on Grammar, I, Gualtiero Calboli (ed.), 47-79. Bologna: CLUEB.
1983 New Approaches to Accusative Subjects: Case Theory vs. Raising.
In Latin Linguistics and Linguistic Theory. Proceedings of the Is' In-
ternational Colloquium on Latin Linguistics. Amsterdam 1981, Harm
Pinkster (ed.), 167-176. Amsterdam/Philadelphia: John Benjamins.
Martin, Roger
1996 A Minimalist Theory of PRO and Control. University of Connecticut:
Ph.D. dissertation.
2001 Null Case and The Distribution of PRO. Linguistic Inquiry 32: 141-
166.
Miller, Stephen G.
1974 On the History of the Infinitive Complementation in Latin and Greek.
Journal of Indo-European Studies 2: 223-246.
Ormazabal, Javier
1995 The Syntax of Complementation: On the Connection between Syntactic
Structure and Selection. Ph.D. dissertation, University of Connecticut.
Latin object and subject infinitive clauses 371
Pepicello, William
1977 Raising in Latin. Lingua 42: 209-218.
1980 An integrated Approach to Diachronic Syntax: A Case Study from
Latin. General Linguistics 20: 71-94.
Pesetsky, David
1996 Zero Syntax. Cambridge, Mass.: The MIT Press.
Pillinger, Ο. Stephen
1980 The Accusative and Infinitive in Latin: A Refractory Complement
Clause. Journal of Linguistics 16: 5 5 - 8 3 .
Pinkster, Harm
1990 Latin Linguistics and Semantics. London: Routledge.
Raposo, Eduardo
1987 Case Theory and Infl-to-Comp: The Inflected Infinitive in European
Portuguese. Linguistic Inquiry 18: 85-109.
Rizzi, Luigi
1982 Issues in Italian Syntax. Dordrecht, Cinnaminson: Foris Publications.
Rizzi, Luigi
1997 The Fine Structure of the Left Periphery. In Elements of Grammar: A
Handbook in Generative Syntax, Liliane Haegeman (ed.), 281-337.
Dordrecht: Kluwer.
Rohlfs, Gerhard
1969 Grammatica storica della lingua italiana e dei suoi dialetti. Sintassi
e formazione delle parole. Torino: Einaudi.
Saltarelli, Mario
1976 Theoretical Implications in the Development of Accusativus cum In-
finitivo Constructions. In Current Studies in Romance Linguistics,
Marta Lujän and Fritz Hensey (eds.), 88-99. Washington: George-
town University Press.
Salvi, Giampaolo
2001 The two Sentence structures of Early Romance. In Current Studies in
Italian Syntax. Essays offered to Lorenzo Renzi, Guglielmo Cinque
and Giampaolo Salvi (eds.), 297-312. Amsterdam/London/New
York/Oxford/Paris/Shannon/Tokyo: Elsevier.
Skytte, Gunver and Giampaolo Salvi
1991 Frasi subordinate aH'infinito. In Grande grammatica italiana di con-
sultazione. 11. I sintagmi verbali, aggettivali, avverbiali. La subordi-
nazione, Lorenzo Renzi and Giampaolo Salvi (eds.), 4 8 3 - 4 8 5 , 4 9 7 -
569. Bologna: II Mulino.
Stowell, Tim
1981 Origins of Phrase Structure. Cambridge, Mass.: MIT dissertation.
1982 The Tense of Infinitives. Linguistic Inquiry 13: 561-570.
Tanaka, Hidezaku
1999 Raised Objects and Superiority. Linguistic Inquiry 30:317-325.
372 Lucio Melazzo
Tantalou, Niki
2003 Infinitives with Overt Subjects in Classical Greek. Studies in Greek
L ingu is tics 23: 358-365.
Wackernagel, Jakob
1892 Über ein Gesetz der indogermanischen Wortstellung. Indogermani-
sche Forschungen 1: 353-436.
Wales, M. L.
1982 Another Look at the Latin Accusative and Infinitive. Lingua 56:
127-152.
Wanner, Dieter
1987 The Development of Romance Clitic Pronouns. From Latin to Old
Romance. Berlin/New York: Mouton de Gruyter.
Wuilleumier, Pierre
1961 Ciceron. Caton l'Ancien (de la vieillessej. Paris: Societe d'Editions
«Les Belles Lettres».
Latin word order in generative perspective:
An explanatory proposal within the sentence
domain*
Chiara Polo
The present paper explores the explicative advantages deriving from ana-
lyzing Latin word order under a generative-inspired perspective. Relying
on a minimal comparison between Latin and Italian morpho-syntactic struc-
tures, Latin is proven to be sensitive to the same set of derivational con-
straints which have been less controversially assumed for Italian under
pragmatically and / or structurally (homogeneously) marked conditions of
production, so much so that its syntax turns out to constitute no exception
to what is predicted by Universal Grammar in terms of a basic, unmarked
word order alternating with a number of derived, marked patterns. The
commonplace of traditional, functionalistic approaches which view Latin as
a free word-order language is thus demonstrated untenable and is conse-
quently abandoned in favor of a principled account of Latin formal codifi-
cation worked out under the Principles and Parameters Theory.
Introduction
The communis opinio about Latin word order is that it is arbitrary, idiosyn-
cratic, non-systematic and immune to any type of explanation: further, the
upholders of the traditional, functionalistic approach (Meillet 1903, and
Panhuis 1982, 1984, among the others) go so far as to draw a straightfor-
ward connection between the seeming 'randomness' of Latin word order
and the fully inflectional status of the language, as if the presence of a pro-
ductive system of case-specifications stood in a cause-effect relationship
with the structural anarchy detected on the surface and the linguist could be
exempted from looking for the ratio concealed underneath it. On this view,
in fact, the paradigmatic information encoded by overt morphology is taken
to play the role of unambiguously signalling the function of the case-marked
constituents: the latter would thus escape the rigid syntactic codification
374 ChiaraPolo
Table 1. Latin
Table 2. Italian
Table 3. Latin
Table 4. Italian
15%: 9% found when Cs or PPs are involved); crucially, the four remaining
word arrangements logically conceivable - VSO, V O S , O S V , and O V S -
display a recessive frequency rate which amounts in its whole to the 7% of
the total in Latin, not so different from the 5.5% found in Italian (or, once
again, the percentage becomes 6% in both languages when the argument is
represented by C or PP). N o matter how naive and superficial these obser-
vations may appear, they suffice to cast doubt upon the commonplace of
Latin as freely admitting any option of word arrangement: on the contrary,
an employ of statistics as rigorous as possible allows to intuitively ap-
preciate the clear-cut boundary holding between high- and low-frequency
orders and to shed light on the way percentages are distributed among
them, up to the final discovery of the meaningful similarities featuring Latin
and Italian syntactic behavior. Obviously, these conclusions simplistically
380 ChiaraPolo
X'7°lp FoiT^^Y
0/V/Jue Specl^ Γ
S/(J/X que
382 ChiaraPolo
The following two summarizing schemes can help understand better the
syntactic architecture just described:
Main clause: Left Periph. | Focus [SOXV] Focus | Right Periph.
Subordinate clause: que Left Periph. | Focus [SOXV] Focus | Right Periph.
2.1. A qualitative, comparative insight into Latin and Italian word ordering:
An overview in synchronic perspective
The first step I want to take is to comment upon the statistical data drawn
from a 'qualitative' overview of Latin and Italian word orders: as a matter
of fact, the statistics below open the path to a straightforward appreciation
of the surface strings which can be taken as basic, underlying in the two
languages - owing to the fact that, in the greatest majority of cases, they do
manifest a compatibility exclusively with pragmatically unmarked readings
of their constituents (while no pairing can be posited in the given context
with a Topical, Contrastive Focus, Emphatic reading of them), and no in-
dependent evidence arises for hypothesizing an association with other
structurally-driven marked strategies of derivation (such as Heavy-NP
Shifts, Verb fronting rules, etc.). 11
384 ChiaraPolo
Table 7. Latin
Table 8. Italian
As for tables 5 and 6, they register the word order types that (information-
ally or structurally) unmarked patterns are mapped into. What they unveil is
that Italian approaches the typologically ideal condition in which neutral
imports pair with one and the same string (S)VX in the 100% of cases,
while Latin slightly departs from this desirable condition because it seems
to admit two surface strings - (S)XV and (S)VX - to carry unmarked val-
ues, even if it favors the former over the latter by a significant proportion of
± 9 0 % to ±10%. This datum seems to provide a measure of our failure to
wholly penetrate the regularity we want to reconstruct for and to ascribe
also to Latin syntactic organization. The problem consists, in fact, in the
Latin word order in generative perspective 385
the main goal of what follows consists precisely in showing how they can
be exploited to successfully deal with apparently unmotivated word switch-
ing in Latin: technically, they lie at the base of my choice to classify the
surface patterns I run into in the Latin text with justified, marked orders,
rather than with unjustified ones. The survey will progress from pragmati-
cally-driven constraints to structurally-motivated ones, conforming to the
list proposed below:
2.3. Left-dislocations
As to the first two displacement rules (listed under a), they can be roughly
likened to each other on the ground of the uniform, pragmatically marked
value the moved constituent seems to bear, even if the way in which the
movement takes place may be different: what cuts across the two construc-
tion types is in fact the discourse-based feature [+ Topic] tagging the con-
stituent, which derives from the old, known, given, and presupposed infor-
mation it delivers. More precisely, as far as Left-dislocations are concerned,
the most substantial features of the construction can be summarised as fol-
lows: first, in such languages as Italian, the left-dislocated Topic signals its
syntactic relationship with the rest of the clause either through its introduc-
tory preposition, which is moved along with it (ex. 2b), or through the re-
sumptive clitic coreferential to it, which becomes obligatory whenever the
Topic coincides with the direct object (Clitic Left Dislocation - CLLD: ex.
3b), or through both of them (ex. 2b); on the contrary, in such languages as
Latin which do not possess a system of (resumptive) clitics, no such ele-
ments ever occur, with the consequence that a diagnostic tool for left-
dislocation processes happens to be unavailable. 13 Second, no uniqueness
requirement holds on structural Topic positions, given that a clause can
contain as many left-dislocated topics as its (topicalizable) arguments and
adjuncts are, so that, for example, a left-dislocated subject may co-occur
with other left-dislocated constituents which stand in-between it and the
verb (ex. 3c): 14
Latin word order in generative perspective 387
marked order - and, on the other hand, it substantiates my point that word
order variation entails the same dynamics and can be accounted for on
similar grounds in Italian and in Latin.
Let us briefly consider the following example, quoted from Chapter LXXIV
of Cena Trimalchionis:
S
(4) a. Sed hie, (qui in pergula natus est),
but this-SG.M.NOM, (who is born in the shed outside),
DO(T) V
aedes non somniatur.
mansion-PL.F.ACC not dreams-3so
'But if you are born in the shed outside, you don't aspire to the
mansion' CT, 74, [14] 18
S
b. Ma questa gente qui, (che nasce nelle catapecchie),
but this person here, (who is born in the slums),
DO(T) V
/ palazzi non se U sogna neanche
the mansions not them dream-3PL even
'But the person who is born in the slums don't even dream the
mansion'
Under the contextual conditions found here, the uniform, preverbal position
the DO aedes (or the Italian i palazzi) occupies can be given a pragmatic
explanation: the sentence is pronounced by Trimalchio after his wife Fortu-
nata has attacked him for his shameful behavior with a handsome slave of
his. The point the speaker wants to make is to convince Fortunata along
with his guests that her railing at him was extremely inopportune and un-
justified, because it did not suit her humble origins to give herself airs with
the person who rescued her from the slave-stand and made of a flute-girl
the respectable woman she now is. In the clause under scrutiny and in the
immediately preceding linguistic context, in particular, Trimalchio is re-
minding Fortunata her past wretchedness as opposed to her present welfare
Latin word order in generative perspective 389
which probably caused the arrogant behaviour she has just displayed. The
crucial remark in his argumentation is that whoever is born in the shed out-
side (as she is) does not even dream of the mansion, while she possesses it
and should content herself with the wealth she has obtained, without stand-
ing up for further rights. Focusing now on the surface string at stake, the
DO aedes can be regarded here as a Topic, bearing old information some-
how inferable from the context, as indirectly substantiated by the fact that it
stands out as the second (easily guessable) item of a contrast which is
drawn between pergula and its opposite aedes, while the rest of the clause
can be qualified as the comment, being an open sentence predicated of the
topic which introduces new information, with the verb in the negative form
qualifying as an informational focus (non somniatur). Thus, the fact that
the context allows for a Topical reading of the DO makes possible to inter-
pret the preposed position the nominal constituent occupies in the clause as
derived through a left-dislocation process (possibly also in Latin, even if in
the latter a reliable test is missing), which finds in the presence in Italian of
a resumptive clitic coreferential with the topical element a further confir-
mation. 19
2.5. Right-dislocations
V DO(T)
(7) a. Adcognosco, inquit, Cappadocem:...
recognize-lsG - he said - Cappadocian-SGM ACC:...
Ί recognize the Cappadocian, he said' CT, 69, [2]
V DO(T)
b. (Lo) riconosco, disse, il Cappadoce:...
(him) recognize-ISG, he said, the Cappadocian:...
Ί recognize the Cappadocian, he said'
The contextual circumstances under which Trimalchio utters the sentence
under investigation yield a compatibility with a Topic reading of the DO
constituent Cappadocem (or its Italian equivalent il Cappadoce)·. the state-
ment follows Scintilla's (i.e. Habinnas' wife) reaction to Habinnas' enthu-
siastic panegyric of his own slave Massa, whose versatility and talents he
has just listed one after the other. The point for Scintilla is to denounce
Habinnas' partiality in his portrayal of his slave, as proven by the fact that
he did not count his being a pimp among his trickery. The hint she makes at
Massa as a pimp causes Trimalchio to intervene with the observation re-
gistered here, where he reassures the jealous Scintilla by claiming that he -
as well as all those present - is able to recognize the Cappadocian (i.e.
Habinnas, coming from Asia as he did) from this detail he has in common
with his slave of being a pimp and a libidinous lover, which all the rest of the
people coming from this place had a reputation for. Obviously, he takes for
granted that the sum of information carried by the constituent Cappadocem
sounds as given, old, and presupposed in the set context, because it was
available and salient in the previous discourse, where Habinnas and his
slave played a role of protagonists, and the provenance of Habinnas from
Cappadocia (namely, his status of Cappadocian) has become a piece o f
shared information among his dinner guests (the very name Habinnas has
been frequently related to his Cappadocian origins). Trimalchio's utterance
392 Chiara Polo
She further likens contrastive focus with emphasis, viewing them as two
aspects of one and the same phenomenon.27 As for the syntactic properties
of the construction, the preposed constituent must be accompanied by its
introductory preposition (in Italian), but it is never resumed by a ^referen-
tial pronoun, not even so in clitic-endowed languages like Italian:
P(CF) V
(14) a. Homines sumus, non dei.
men-PL.M.NOM are-lPL, not gods-PL.M.NOM
'We're human, not gods' CT, 75, [1]
P(CF) V
b. UOMINI siamo, non dei.
men are-lPL, not gods
'We're human, not gods'
S V DO(CF)
(16) a. (Bellumpomum), qui rideatur alios;...
(a fine specimen), wh0-3sG.N0M laughs-3sG others-PL.M.ACC
'he is a fine specimen to be laughing at others;...' CT, 57, [3]
S V DO(CF)
b. Beltomo, che deride gli ALTRI (non se stesso!);...
(a fine specimen), who laughs-3sG the others (not himself);...
'he is a fine specimen to be laughing at others, rather than at him-
self;...'
The analysis worked out by Kayne (1994) for sentences like (17) is that the
PP to Bill originates in a small clause whose specifier position hosts all his
old linguistics books and, from there, it undergoes leftward movement in-
dependently of the verb past the object to a still higher specifier position
with the result that the surface string derived has the 'heavy constituent' in
clause final position. When applied to Latin, the assumption of a Heavy NP
Shift strategy provides a straightforward explanation for such instances of
( S ) V O H as the following, where the post-verbal D O ostrea pectinesque
crucially functions as antecedent for the (non-defining) relative clause fol-
lowing in apposition quae collecta puer lance circumtulif.
A deviating, VO surface string may be derived not only when the internal
argument carries pragmatically or structurally marked values which trigger
displacement phenomena, but also when there are good reasons to assume
that it is the verb that undergoes movement: this seems to be the case for
verbs of imperative clauses which tend to surface clause-initially in Latin as
a consequence of the activation of a V-to-C constraint of a type already
well-attested in a number of SOV languages (thus deriving ViO sequences,
both when the imperative is in its proper form and when it is in the form of
a hortative subjunctive). A crucial piece of evidence for assuming a V-to-C
movement of the (imperative) verb comes from recent studies on Gothic
syntax (Ferraresi 1992 and Longobardi 1994), inspired by an observation al-
ready contained in Meillet (1908/09) in relation to a passage of the Gospel
by Mathew (8, 3). 33 The analogies detected between Gothic, on the one hand,
and Latin, on the other hand, seem to justify a uniform analysis for the in-
flected verb in imperative clauses, pointing to a V-to-C movement behind
the verb first pattern they converge on, as illustrated in the following Latin
sentence:
V DO
(19) Agite, inquit, scordalias de medio.
banish-2PL, (he said), quarrels-PL.F.ACC from here
'Come on, he said, that's enough wrangling' CT, 59, [1]
subordinate clause, which results in a [ sub XP] - V2... pattern. The tendency
to move the verb to the front position of a main clause is well documented
after an ablative absolute construction (as observed by Marouzeau 1938:
80) and a conditional or, even more frequently, a temporal subordinate
clause (Mobitz 1924: 120): this is upheld by Petronious' work, where the
construction recurs frequently after temporal, but also after conditional
subordinates (Kroll 1918: 117; Mobitz 1924: 121), to substantiate its wide-
spread employ in everyday speech. According to Bauer (1995: 96-97), the
rate of occurrence of this pattern increased during the course of Latin and
gradually spread to popular texts: in Cena Trimalchionis, the construction
is documented with a certain frequency, and the properties it manifests
seem to be reminiscent of verb second phenomena attested in Germanic
languages. 34 The trigger for the movement of the verb could be tentatively
taken to be similar in nature to the one Salvi (2004: 1.2.4) proposes for V]
clauses: 35 namely, it could be the case that the role played by an abstract
operator located in the specifier of a high functional projection deriving V)
phenomena is taken over here by temporal / conditional clauses which
would land in the specifier of the same projection to result in V2 effects: 36
V IO(N)
(20) (Quod si hoc fecerit), eripiat Norbano
(if it comes off), will-steal-3sG Norbanus-SG M.DAT
DO(N)
totum favorem.
all favour-sG.M.ACC
'If it comes off, he'll put Norbanus right out of the running'.
CT, 45, [10]
Things being so, cases of verb first in main clauses preceded by subordi-
nate clauses can be considered structurally motivated deviations from the
unmarked word order (S)OV in Latin and, as such, will be referred to the
old, OV Latin grammar. 37
2.14. Movement of the clitic verb forms of sum to the Wackernagel position
Word orders which evidence a displacement of the verb from the canonical,
sentence-final position to second sentence position can be also found when
the different forms of sum happen to be affected: as underlined by Salvi
Latin word order in generative perspective 401
(2004: 1.2.5), at least some forms of sum frequently recur in second sen-
tence position immediately after the first element (word or phrase) of the
clause, both in main and in subordinate clauses (in the so-called Wackernagel
position, as defined by Adams 1994). In this typology of structures, the
deviation from an unmarked adjustment of sentence constituents can be
derived by appealing to an (independently well-documented) clitic move-
ment of the verb to a specific 'landing site': 38
would like to spend just two words on the way VSO, VOS, OS V, and OVS
orders are derivable, using to this purpose some meaningful examples. As
to the former two patterns VSO and VOS , they converge on a verb-first
organization, which might be interpreted in terms of the activation of a verb
fronting rule, jointly or not to the post-position of the subject (VOS), which
happens to be motivated on the ground of its marked (rhematic or heavy)
status - as already assumed by Ostafin (1986: 159).40 As far as this verb-
first pattern is concerned, on the other hand, according to Salvi (2004:
1.2.4.), it is frequently found with jussive and concessive sentences, along
with clauses with assertive value, clauses instantiating a contrast, presenta-
tional clauses, to which we add narrative contexts recording a crucial pas-
sage. Technically speaking, the particular typology of clauses featuring this
construction-type might be characterized in terms of the presence of an
abstract operator in the specifier of a functional projection of a certain type -
possibly, FocP as argued for by Salvi (2004: 2.1), whose semantic activation
would follow only from the presence of lexical material in the position of
the head of the same projection. On this view, the raising of the verb could
be regarded as a possible solution for the recovery of the semantic content
of the operator:
V S
(22) Convertit ad hanc scaenam Trimalchio
turned-3sG to this Scene-SG.F.ACC Trimalchio-SG.M.NOM
DO(N)
vultum et: "Amici"...
face-SG.M.ACC and: "Friends"...
'Trimalchio surveyed this tableau, and said: "My friends",...'
CT, 33, [5]
V S DO(N+H)
(23) Accipimus nos cochlearia non minus
p i c k e d - u p - l P L we-lPL.NOM spoons-PL.N.ACC n o t less
selibras pendentia
h a l f pound-PL.F.ACC weighing-PL.N.ACC
(ovaque ex farina pingui figurata pertundimus)
(and assaulted the eggs which were made of flour baked in oil)
'We picked up our spoons weighing not less than half a pound, and as-
saulted the eggs which were made of flour baked in oil' CT, 33, [6]
Latin word order in generative perspective 403
Both the VSO surface strings illustrated above are attested in Chapter
XXXIII and, crucially, occur one after the other in the same paragraph:
they seem to carry out the function of marking the return of Trimalchio on
the scene, after the short interval dedicated to finishing his game. The
'event' seems to be worth an ironic underlining, which is entrusted to a
syntactically marked pattern with a fronted verb to highlight that Trimal-
chio has resumed his role of unrivalled director of the action and a remark-
able shift has occurred in the narrative. (Under this hypothesis, then, an
abstract, narrative operator - with the meaning of 'Then, At that point' -
occurring in the specifier of the Focus Phrase could be taken to underlie the
raising of the verb to Foc° to effect its semantic content activation). As a
matter of fact, it is on Trimalchio and on his action that the focus is placed,
when he surveys the tableau newly brought in and suggests to his guests to
test whether the eggs distributed among them are soft enough to eat; in
obeisance to his orders, the protagonists pick up their spoons and assault
the eggs and their action is given the same prominence as that performed by
Trimalchio which immediately precedes and to which it is added - as in a
list - through the same syntactically marked format VSO, intended to have
the pressing rhythm of the actions orchestrated by the master of the house
suitably emphasized.
V DO(T) S(Foc+H)
(24) a. Tegebant asellum duae lances (in quarum
covered-3PL donkey-SG.M.ACC t w o dishes-PL.F.NOM, ( o n t h e r i m s
marginibus nomen Trimalchionis inscriptum erat et argenti pondus)
of which were engraved Trimalchio's name and their weight in
silver)
'Covering the donkey were two dishes, on the rims of which were
engraved Trimalchio's name and their weight in silver'
CT, 31, [10]
V DO(T) S(Foc+H)
b. (Lo) Coprivano, I 'asinello, due piatti, (sui cui bordi erano
(it) covered-3PL, the donkey, two dishes, (on the rims of which
incisi i nomi di Trimalcione e la caratura dell 'argento),
were engraved Trimalchio's name and their weight in silver).
'Covering the donkey were two dishes, on the rims of which were
engraved Trimalchio's name and their weight in silver'
404 Chiara Polo
The PVS string occurs in Chapter XXX: it refers to the moment when a
slave stripped for flogging grovels at the protagonists' feet, and proceeds to
implore them to rescue him from punishment. His words are evoked by the
narrator through the indirect speech technique and are aimed to convince
his interlocutors that the fault for which he is being punished is not that
serious. A quick glance at the formal shaping of the clause set in its context
unveils the structural complexity of the 'late' subject peccatum suum propter
quod periclitaretur, to the effect that its postposed position turns out to be
justifiable in structural terms.
Going back to the string (S)VO, we have already seen how the tradi-
tional claim of a free variation between OV and VO based exclusively on
their occurrence rates does not find any empirical support, given that the
greatest majority of deviant, (S)VO orders occur under pragmatically or
contextually marked environments and are consequently liable of being
derived from a basic (S)OV order through one of the following licensing
strategies:
At this point, once established that Petronian Latin proves consistent with
an (S)OV grammar at least as far as the domain of sentence syntax is con-
cerned, the question arises as to the residual percentage of cases (10%)
which stand out as counterevidence to the typologically ideal situation em-
bodied, for example, by Italian, in which one and only one surface pattern
is permitted to occur under informationally and structurally neutral envi-
ronments. As already hinted at, this small set of 'idiosyncratic patterns' can
be handled under a twofold proposal:
S V
(26) Nam puer quidem ... circumibat iam dudum
as for slave-SG.M.NOM was-going-around-3SG already
DO(N) DO(N) V
pedes nostros et missionem rogabat
feet-our-PL.M.ACC and mercy-SG.F.ACC p l e a d e d - f o r - 3 s G
'As for the slave who had tumbled, he was already doing the
rounds of our feet, and was pleading for mercy'. CT, 54, [3]
Acknowledgements
Notes
* In the present paper the following abbreviations and symbols have been adopted:
ABL = ablative; ACC = accusative; C = (obliquely case-marked) complement/
complementizer; CF = contrastive focus; cl = clitic; CT = Cena Trimalchionis;
D = discontinuous; DAT = dative; DO = direct object; Ε = emphatic; f. =
feminine; FOC = focus; GEN = genitive; Η = (structurally) heavy; INST = in-
strumental; IO = indirect object; LOC = locative; m. = masculine; n. = neuter;
Ν = (informationally) neutral / unmarked; NOM = nominative; NP = noun
phrase; Ο = object; Ρ = predicative complement (of the S); pi. = plural; PP =
prepositional phrase; S = subject; sg. = singular; Τ = topic; ν = auxiliary verb;
V = verb; Vpp = past participle verb; VP = verbal phrase.
1. These hypotheses find a consistent exposition in the following passage by
Meillet (1903 [1964]: 439): the author is here mainly concerned with the syn-
tactic effects brought about diachronically by the morphological levelling many
a natural language underwent, but he also hints at the typological intertwine-
ments between morphology and syntax supposedly showing up in the syn-
chrony of the Latin language:
Les transformations du type morphologique s'accompagnent de transformations
paralleles de la phrase. La phrase indo-europeenne se composait de mots auto-
nomes, dont chacun suffisait ä exprimer un sens complet et la fonction remplie.
Quand les n o m s ont reduit, puis perdu la declinaison, la fonction des noms dans
la phrase a ete indiquee par deux precedes nouveaux: 1° Un ordre de mots ä
410 ChiaraPolo
culturally refined author like Petronius. In this scenario, Adams (1976: 83-94)
strenuously defends the former view, as shown by his claim that Cena Trimal-
chionis is an untrustworthy basis to reconstruct spoken patterns ('In word order
Cena Trimalchionis is unrepresentative of genuine Vulgar Latin ... being an
artificial concoction by a man of learning'), unlike the approximately contem-
porary letters of the soldier Claudius Terentianus which, on the contrary, are
taken to give 'a glimpse of semi-literate Latin'. Hinojo (1985: 245-246) and
Zamboni (2000: 4 4 ^ 5 ) , on the other hand, tend rather to cautiously credit the
text with a documentary value for the study of Vulgar Latin features ("... con
las debidas limitaciones, <la obra> puede considerarse como uno de los textos
caracteristicos del latin vulgar"; "Un documento citatissimo in questa prospettiva
come il Satyricon di Petronio rivela ... la presenza dell'orale nello scritto (un
orale di natura mimetica, ricostruito e non oggettivamente registrato come
farebbe oggi un ricercatore), in altre parole elementi e parametri deWimme-
diato comunicativo"). Given these alternative perspectives, I take side with
Hinojo and Zamboni and consistently regard the text under scrutiny as one of
the few choices I could make to have an insight into the familiar and colloquial
registers mastered by 1st century Latin speakers.
5. The Principles and Parameters Theory captures the intuition that the interlin-
guistic variation showing up on the surface both synchronically and diachroni-
cally can be uniformly characterized as a by-product of the specific articulation
of our language faculty (roughly coinciding with Universal Grammar - UG),
which is supposed to consist of invariable principles, along with a set of open
parameters. The latter would be open choices, ideally allowing for binary op-
tions, to be set by language learners under exposure to their primary linguistic
corpus; the value parameters are given is thus a function of the triggering expe-
rience and can accordingly vary not only from language to language, but also
from generation to generation within one and the same linguistic community.
In a pre-Kaynian framework as the one tentatively adopted here, the head-
complement parameter is one of the open choices specified by UG which chil-
dren have to fix as complement-head or head-complement depending on the
linguistic environment they are surrounded by. Hence, under our proposal, the
discrepancies detected between Latin and Italian syntax are ultimately amena-
ble to a different setting of the head-complement parameter (fixed respectively
as complement-head in Latin and head-complement in Italian). On the con-
trary, under Kayne's (1994: 5-6, 35-36) antisymmetric view of syntax, there is
only one universal order underneath surface variation - Specifier-Head-
Complement (coinciding with an S-V-0 linearization borne out by the Linear
Correspondence Axiom for all language types), with the implication that the
order (S)OV which we tentatively characterize as basic and unmarked in Latin
is held to be derived out of necessity, and the clear opposition (S)OV vs (S)VO
displayed by Latin and Italian syntaxes needs to be worked out in other terms.
412 ChiaraPolo
6. It seems finally worth pointing out that the method employed to classify the
documented strings conforms to the parameters made available by modern ty-
pological studies; it willingly adheres to the idealizations and simplifications
which are commonly adopted under this field of studies to let broad linguistic
generalizations come to the surface. However, there is a number of problems 1
had to face in my comparative survey of Latin and Italian word patterns which
I think necessary to mention very quickly so as to provide a glimpse of the
cataloguing procedure I stuck to, along with its limits and the possible ways of
compensating for them. The first difficulty I was confronted with has to do
with cases where one of the nominal constituents - most frequently the DO -
evidences in Latin a syntactic discontinuity between its adjectival modifier and
the nominal head it refers to, owing to the presence in-between of some lexical
material - most frequently the verb (DO AP - V - DO N P ) - to the effect that part
of it surfaces in front of the intervening verb, while the rest comes after it:
(i) et ipse capaciorem poposcit scyphum, ...
as well he-SG.M.NOM larger-SG.M.ACC demanded-3SG cup-SG.M.ACC
'Trimalchio too demanded a larger cup,...' CT, 65, [8]
As a matter of fact, the split status of the DO makes it difficult to decide
whether the structure should be classified with OV- or, rather, with VO-strings.
In these cases, I conformed to Ostafin's practice to take the position of the
nominal head as indicative of the ordering of the discontinuous constituent {i. e.
D O a p - V - DO N P = V-DO), as made plausible by the emphasis usually placed
on the fronted, adjectival head (often in the comparative degree), which might
argue for a strategy of mise en relief par disjunction - in Marouzeau's (1922)
terms - affecting the modifier in isolation, with the enhanced effect of turning it
into a marked constituent and, as such, into an unreliable indicator of the basic
position of the constituent. Anyway, even if my choice should sound arbitrary
or questionable, the paucity of the examples exhibiting this 'anomaly' (3.5%
of the totals: 42/1181, with just 14 affecting DOs - equal to the 1.1%, as
shown in the table below) ensures that the final statistical amount recorded for
each word order pattern is not dramatically altered by the decision 1 was forced
to take.
Tab. 1: Discontinuity
For the time being, however, I leave the question willingly open, given that its
thorough discussion goes beyond the aims of the present paper; I refer to note
Latin word order in generative perspective 413
ternal and a fronted position of the verb is systematic and matches with a de-
clarative or imperative interpretation of the clause; her examination of the rela-
tive ordering of the verb with respect to locative adverbs confirms that the in-
flected verb obligatorily moves across the adverbial material in imperative
clauses.
34. For a detailed analysis of V2 constructions in Germanic languages, I refer to
Tomaselli (1990). As far as this topic is concerned, further research is needed
to arrive at more reliable conclusions; in particular, by exploiting the kind of
evidence used by Salvi in his contribution to the present volume, it would be
necessary to establish with a higher degree of precision which is the structural
position occupied by the constituent preceding the verb, in order to exclude
that it occurs in positions meant for left-dislocated elements (this could be as-
sessed by observing the distribution of weak words, given that, if the preverbal
phrase should happen to be left-dislocated, weak words would follow the verb).
However, such a verification transcends the scope of the present article, on ac-
count of the fact that the empirical corpus I relied on for investigations (i.e.
Cena Trimalchionis) seems to be lacking in the kind of examples (exhibiting
weak pronouns); the inescapable conclusion is that the identification of Ger-
manic V2 constructions with this set of Latin structures remains a hypothesis
in need of sounder proving.
35. For a more detailed description of Salvi's (2004) proposal, see section 2.15.
36. Under this hypothesis, however, the landing site for the movement of the verb
would be the head of the Focus Phrase, on analogy with what has been sup-
posed for VI constructions, in which the abstract operator in need of semantic
activation through lexical support by the verb was taken to occur in Spec,FocP
(in the case of these V2-like structures, the temporal / conditional clause would
similarly occur in Spec,FocP and act as a trigger for verb movement - I thank
Katalin Kiss for hinting at the possibility of extending the trigger assumed for
VI clauses to V2 instances). Things being so, we would not be authorised to
speak of German-like V2 phenomena, because the structural positions in-
volved would not be the same.
37. Similarly, an analysis in terms of verb second is corroborated by indirect inter-
rogative contexts which frequently match with the typical properties of verb
second phenomena.
38. The landing site involved can be tentatively identified with the C° head or, in
the case illustrated in (21), with the Foc° head, given that the fronted constitu-
ent nulla reads as an emphatic, focalised adjective, disjoined from the noun
head sententia it agrees with, and the following, raised verb est can be consis-
tently supposed to occur in Foc°. A plausible hypothesis we can further ad-
vance is that the verb second constraint of Germanic languages represents a
generalization to all finite verbs of a movement to the Wackernagel position
which was originally reserved for a limited class of verbs. This thesis, however,
420 Chiara Polo
has been already proposed by Wackernagel (1892) for Germanic V2, by Thur-
neysen (1892) for Old French, and criticized as untenable by Salvi (2003).
39. As a matter of fact, semantic evidence remains the main source of information
on which any hypothesis on Latin sentence structure can be grounded. In the
absence of grammaticality and intonational judgements, in fact, our linguistic
arguments can be based on the scanty syntactical evidence we can gather about
Latin, along with (and more heavily so) indirect pragmatic evidence.
40. Under a Kaynian approach, VOS orders could be derived by assuming the
raising of both the light constituents V and O, with the heavy S remaining in
place. Similarly, under Zubizarreta's (1998: 135-138) proposal, VOS struc-
tures could be taken to derive as in Italian from an underlying, universal SVO
order through the leftward adjunction of VO (TP) to a Focus Phrase to ensure
that the right-hand subject ends up provided with the narrow focus interpreta-
tion she attributes to it, and the conflict which would otherwise arise between
the NSR (Nuclear Stress Rule) and the FPR (Focus Prominence Rule) is suc-
cessfully handled.
41. It actually plays the role of antecedent for the non-restrictive relative clause
following in apposition.
42. According to Berretta (1994: 80) among the others, there is a correlation holding
interlinguistically between the subject and the Theme (or, to state it in Givon's
(1983) terms, the subject is a grammaticalized topic; the object, on the con-
trary, coincides with the Rheme). When this equation is not borne out, and the
subject is rhematic and / or the object is thematic, marked word orders as the
ones being reviewed arise.
43. Besides what has been already said on low-level discontinuities in note 6, I
would add some speculative remarks on the way these constructions can be de-
rived under Kayne's (1994), van Riemsdijk's (1989) and my analysis. Possibly,
the simplest way to account for them can be obtained under Kayne's frame-
work, on condition that a small amendment to it be carried out; thus, on the as-
sumption that VO is the universal order, it should be possible to generate OV
not only through the movement of the object as a whole for feature-checking
reasons (to the specifier of an appropriate functional projection), but also
through the displacement of a subset of it {i.e. the adjective agreeing with the
nominal head) capable of checking case in its (pre-verbal) position, and to
transmit it via agreement to the nominal head left behind, in post-verbal posi-
tion. Obviously, the second option is chosen only when a sub-part of the con-
stituent - namely, the adjectival component - needs focusing, and its fronting
matching with its parallel separation from the rest of the constituent is the
marked strategy made available by the language to carry out a mise en relief
function, using Marouzeau's (1922) terminology. A derivational alternative
which seems to stand out as a further refinement of Kayne's analysis is ame-
nable to van Riemsdijk's (1989) explanation of similar German constructions,
Latin word order in generative perspective 421
References
Ferraresi, Gisella
1992 Die Stellung des gotischen Verbs im Licht eines Vergleichs mit dem
Althochdeutschen. Unpublished dissertation, University of Venice.
Givön, Talmy
1983 Topic continuity in discourse: quantitative cross-language studies.
Amsterdam Philadelphia: Benjamins.
Graffi, Giorgio
1994 Sintassi. Bologna: II Mulino.
Greenberg, Joseph H.
1963 Some universals of grammar with particular reference to the order of
meaningful elements. In Joseph H. Greenberg (ed.), 7 3 - 1 1 3 .
Greenberg, Joseph H. (ed.)
1963 Universals of language. Cambridge, Mass.: The M I T Press.
Harris, Zellig S.
1960 Structural linguistics. Chicago: University of Chicago Press.
Hinojo, Gregorio
1985 Del orden de palabras en el Satiricon. In Symbolae Lvdovico Mitx-
elena Septvagenario Oblatae, Victoriaco Vasconvm. Jose L. Melena
(ed.), 245-254.
Hofmann, Johann Baptist and Anton Szantyr
1965 Lateinische Syntax and Stilistik. München: Beck.
Hornstein, Norbert and David Lightfoot (eds.)
1981 Explanation in linguistic theory: the logical problem of language
acquisition. New York: Longman Inc.
Jackendoff, Ray S.
1974 Semantic interpretation in generative grammar. Cambridge, Mass.:
The M I T Press.
Kayne, Richard S.
1994 The antisymmetry of syntax. Cambridge, Mass.: The M I T Press.
Kroch, Anthony
1989 Reflexes of grammar in patterns of language change. Journal of
Language Variation and Change 1: 199-244.
Kroll, Wilhelm
1918 Anfangsstellung des Verbums im Lateinischen, Glotta 9: 112-123.
Kühner, Raphael and Carl Stegmann
1955 Ausführliche Grammatik der Lateinischen Sprache. II. Satzlehre.
Leverkusen: Gottschalk.
Lambrecht, Knud
1994 Information structure and sentence form. Topic, focus, and the men-
tal representations of discourse referents. Cambridge: Cambridge
University Press.
Larson, Richard K.
1988 On the double object construction. Linguistic Inquiry 19: 335-391.
Latin word order in generative perspective 425
Larson, Richard K.
1990 Double objects revisited: reply to Jackendoff. Linguistic Inquiry 21:
589-632.
Lightfoot, David
1999 The development of language. Acquisition, change, and evolution.
Oxford: Blackwell.
Longobardi, Giuseppe
1994 La posizione del verbo gotico e la sintassi comparata dei comple-
mentatori germanici: alcune riflessioni preliminari. Miscellanea di
studi linguistici in onore di Walter Belardi: 353-373.
Magni, Elisabetta
2000 L'ordine delle parole nel latino pompeiano: sulle tracce di una de-
riva. Archivio Glottologico Italiano LXXXV-I: 3-37.
Marouzeau, Jules
1922 L'ordre des wots dans la phrase latine. I. Les groupes nominaux.
Paris: Champion Editeur.
1938 L'ordre des mots dans la phrase latine. II. Le verbe. Paris: Les
Belles Lettres.
1948 Quelques vues sur l'ordre des mots en latin. Lingua I: 155-161.
Meillet, Antoine
1903 Introduction α Γ etude comparative des langues indo-europeennes.
Paris, Librairie Hachette. [Reprint 1964. Alabama: University of
Alabama Press],
1908/09 Notes sur quelques faits gotiques. MSLP 15: 73-103.
Melena, Jose L. (ed.)
1985 Symbolae Lvdovico Mitxelena Septvagenario Oblatae, Victoriaco
Vasconvm.
Mobitz, Ο.
1924 Die Stellung des Verbums in den Schriften des Apuleius. Glotta 13:
116-126.
Molinelli, Piera
1986 L'ordine delle parole in latino: studi recenti. Lingua e Stile 21-4:
485^198. Bologna: II Mulino.
Ostafin, David Mark
1986 Studies in Latin word order: a transformational approach. PhD disser-
tation, The University of Connecticut, Storrs.
Panhuis, Dirk
1982 The communicative perspective in the sentence: A study of Latin word
order. Amsterdam: Benjamins.
1984 Is Latin an SOV language? A diachronic perspective. Indogerma-
nische Forschungen. Zeitschrift für Indogermanistik und Allgemeine
Sprachwissenschaft 89: 140-159. Berlin, Germany.
426 Chiara Polo
Pintzuk, Susan
1991 Phrase structures in competition: Variation and change in Old English
word order. PhD dissertation, University of Pennsylvania, Phila-
delphia.
Polo, Chiara
2004 Latin, Italian and Slovene between morphology and syntax. Padua:
Unipress.
forthc. An analysis of low-level discontinuities in Latin. Unpublished Ms.,
University of Padua.
Quintilianus, Marcus Fabius
1980 See under: Η. E. Butler.
Renzi, Lorenzo (ed.)
1989 Grande grammatica italiana di consultazione. La frase. Vol. 1.
Bologna: II Mulino.
Riemsdijk, Henk van
1989 Movement and regeneration. In Paola Benincä (ed.).
Rizzi, Luigi
1998 The fine structure of the left periphery. In Romance syntax, Paola
Benincä and Salvi Giampaolo (eds.), 112-158. Budapest: L. Eötvös
University.
Salvi, Giampaolo
2003 Teoria sintattica e spiegazione diacronica. In Syntaxtheorien: Modelle,
Methoden, Motive, Elisabeth Stark and Ulrich Wandruszka (eds.).
Tuebingen: Narr.
2004 La formazione della struttura di frase romanza. Ordine delle parole e
clitici dal latino alle lingue romanze antiche. Tuebingen: Niemeyer.
Saussure, Ferdinand de
1916 Cours de linguistique generale: publie par Charles Bally et Albert
Sechehaye avec la collaboration de Albert Riedlinger. Paris: Payot.
Schneider, N.
1912 De verbi in lingua latina collocatione. Münster: Monasterii Guest-
falorum.
Stark, Elisabeth and Ulrich Wandruszka (eds.)
2003 Syntaxtheorien: Modelle, Methoden, Motive. Tübingen: Narr.
Steinberg, Danny D. and Leon A. Jakobovits (eds.)
1971 Semantics: an interdisciplinary reader in philosophy, linguistics and
psychology. Cambridge: Cambridge University Press.
Thurneysen, Rudolf
1982 Zur Stellung des Verbums im Altfranzösischen. Zeitschrift für roma-
nische Philologie 16: 289-307.
Tomaselli, Alessandra
1990 La sintassi del verbo finito nelle lingue germaniche. Padova: Unipress.
Latin word order in generative perspective 427
Wackernagel, Jacob
1892 Über ein Gesetz der indogermanischen Wortstellung. Indogermani-
sche Forschungen 1: 333-436 (= Wackernagel 1955: 1-104).
Walsh, P. G.
1996 The Satyricon, Translated with introduction and explanatory notes by
P. G. Walsh. Oxford: Clarendon Press.
Zamboni, Alberto
2000 Alle origini dell'italiano. Dinamiche e tipologie della transizione dal
latino. Roma: Carocci.
Zubizarreta, Maria Luisa
1998 Prosody, focus, and word order. Cambridge, Mass.: The MIT Press.
Some firm points on Latin word order:
The left periphery
Giampaolo Salvi
The aim of this chapter is to show how structural and in particular genera-
tive linguistic reasoning can be applied to the study of Latin word order.
Although the possibilities of this research are severly limited by the kind of
data we have at our disposal (sect. 1), some more or less firm conclusions
may be reached: we examine here the left periphery (i.e. the initial section)
of the main clauses (sect. 2) and of the embedded clauses (sect. 3).
The material of this chapter is drawn from ch. 2 and 4 of Salvi (2004)
(written in 1998-99), to which we refer for further details and bibliography.
The analysis presented here is, however, somewhat different and, we hope,
improves on our previous one.
The theoretical frame utilized is a simplified version of Government and
Binding Theory (for which see Graffi 1994).'
1. Methodological introduction
Pi and P2 may also be kept distinct by the position of clitic pronouns: if the
preverbal constituent is in Pi, clitics are postverbal (3a); if it is in P 2 , they
are preverbal (3b) - we exemplify with a preverbal DO, so one can notice
that the position of clitics and the presence of a resumptive pronoun both
depend on the position occupied by the preverbal DO: if it is in Pi, we have
a resumptive clitic pronoun, and this is postverbal (3a); if it is in P 2 , we
have no resumptive pronoun, and the other clitics are preverbal (3b): 3
Some firm points on Latin word order: The left periphery 431
(3) a. Nos qatre dex (DO) ont (V) ies (resumptive clitic) il (S) retenu?
our four gods have them they got
'Have they got our four gods?' (La chanson d'Aspremont 8168)
In Latin, evidence of this type is much more difficult to find and to evaluate.
Firstly, the use of resumptive pronouns is not very frequent: since zero-
anaphora is also the unmarked type of anaphora between sentences, it is
completely dominant within sentences, where overt anaphora is extremely
rare. We can find some examples, but we cannot automatically draw from
them the same conclusions as from the Romance ones discussed above.
Consider (4a), where the pronoun is doubles the Subject tuus... dolor, we
could think that the Subject occupies Pi because it is doubled, and that the
predicative AP humanus occupies P2 because the resumptive (and pre-
sumably atonic) pronoun immediately follows it (as in (2) and (3b) the
resumptive clitic immediately follows the constituent in P 2 ). But consider
now (4b), where the pronoun ea doubles the Subject urbana plebes: the
Subject is doubled, so it must be in Pi; but why does the resumptive pro-
noun immediately follow the constituent in Pi, in contrast to all our previous
examples (cf. again (2) and (3b))? But if the Subject occupied P 2 , why do
we have a resumptive pronoun (cf. (1) and (3b))? 4
These contradictory data can be elegantly explained if one assumes that the
resumptive pronouns in (4a) and (4b) differ in stress and consequently, per-
haps, in grammatical category and syntactic properties: the is of (4a) would
indeed be atonic and, as a weak pronoun, could not immediately follow Pi,
while ea in (4b) (the feminine form of is) would be stressed and of category
NP, so it could occupy P2 and immediately follow Pi. In this way, we can
mantain that, in Latin too, there exists a syntactic position P) with its own
distinguishing properties: a) it cannot be immediately followed by an
atonic pronoun, and b) only constituents in Pi may be doubled. 5
Latin pronouns could serve different pragmatic uses: they could have
the textual function of focus, of contrastive topic, of new topic, or they
could be simply anaphoric. These differences in pragmatic function could
well be associated with the prosodic difference just noted - just as in mod-
ern R o m a n c e languages, free pronouns and clitics have different pragmatic
functions (or semantics, as in Calabrese's (1980) analysis, where clitics
have expected referents and free pronouns have unexpected referents).
As for the different syntactic use of Latin personal pronouns, we will as-
sume that differences in stress, in meaning and pragmatic function, and in
syntactic behaviour go hand in hand, and that they distinguish two principal
uses of Latin personal pronouns: strong (use of) pronouns and weak (use
of) pronouns. Strong pronouns were stressed and had the function of focus,
contrastive topic or new topic (unexpected referent); weak pronouns were
unstressed and had a purely anaphoric function (expected referent). Syntac-
tically, strong pronouns could occupy the same positions as nominal NPs,
but weak pronouns were limited to a number of readily definable positions
- as in Old Romance, those described by a form of Wackernagel's Law:
they were enclitic to the first constituent of a specific syntactic domain.
This hypothesis renders the explanation offered above of the contrast in
(4) a plausible one (mainly in the light of the R o m a n c e facts), but notice
that the explanation is not supported by cogent internal evidence as is the
minimal pair in (3) for Medieval Romance. Certainly, if we denied that is
in (4a) and ea in (4b) are of different syntactic categories, we would lose
an interesting explanation of the facts in (4) and a fundamental device for
the analysis of Latin sentence structure. But when we use the position of
Latin pronouns in syntactic argumentation, we must keep in mind that we
are using a hypothesis, so our results on sentence structure below will be,
so to speak, "second degree hypotheses" about Latin syntax.
In order to once again illustrate the hypothetical degree of the analysis:
one can identify the weak forms of the personal pronouns assuming (on the
basis of the pairing form-semantics) that all pronouns that do not function
as topic or focus are purely anaphoric, in other words bona fide weak
forms. This procedure gives relatively good results: the great majority of
the supposedly weak pronous behave in conformity to Wackernagel's Law
(Salvi 1996; see also the discussion in Adams 1994a). But to get this result,
we must first and foremost exclude all the personal pronouns which stay in
sentence initial position without previously questioning their seman-
tic/pragmatic status: the question does not even arise as to whether they are
weak or strong; the fact that they are sentence initial means they can only
434 Giampaolo Salvi
Besides this syntactic fact, the arguments one can use in the study of Latin
sentence structure are mainly based on the possible pairings of distributions
and pragmatic values: we may assume that marked pragmatic values are
paired with marked word orders and, if we observe the systematic recur-
rence of a given word order with a given pragmatic value, we may suppose
that the observed word order makes use of a special structural position. But
this is not without its problems.
Notice first that in the absence of other structural information, this con-
clusion may be hazardous: nothing assures us that one pragmatic function
is tied to only one structural position or that one structural position is tied
to only one pragmatic function. We may show this with the two syntactic
positions of Medieval R o m a n c e we have identified above on purely syntac-
tic grounds: it is widely acknowledged that constituents with thematic (i.e.,
non-rhematic) value may occur both in Pi (5a) and in P2 (5b) (= (3b))
without any difference in meaning (Salvi 1993) and that constituents occur-
ring in P2 may be not only themes (5b), but also foci (5c): 5
In the light of our criteria, the theme constituent vostre terre must be in P2
because there is no resumptive pronoun; in (6), therefore, a theme and a
focus co-occur in P2 (vostre terre is in P2A, qui in P2B) - this is only possi-
ble if the focus is a Wh-phrase. 8
From the examples above we may conclude that it is a good policy to
postulate different syntactic positions for constituents with different prag-
matic functions (following E. Kiss 1987, we may think that this is all the
more probable for a language which, thanks to its case system, need not
use syntactic structure in order to mark syntactic functions); but we must
bear in mind that providing syntactic evidence in their support may be very
difficult, if not impossible, in the case of dead languages.
As for the converse problem, if a constituent with a given pragmatic
function may occupy different syntactic positions, the use of pragmatic
function as a criterion for syntactic structure may be only suggestive in the
absence of other criteria.
But there is another problem with the use of pragmatic evidence: individu-
ating the pragmatic values expressed in a sentence implies the perfect re-
construction of the author's communicative intentions, a task which is not
always easy. Even if we assume that the principles which organize the dis-
course are the same universally, there are always cases where two or more
different ways of arranging what we want to say are equally suitable to our
communicative needs. We may see this if we examine different translations
436 Giampaolo Salvi
of the same text: in the sentence reported in (7) the constituent aedes has
been interpreted as a marked topic by the Italian translator (a) (cited and
discussed by Polo 2002: 168) and as part of the rheme by the translator (b):
Both solutions are defensible on pragmatic grounds, but neither has syntac-
tic evidence that may support it.
1.6. Conclusion
In spite of all this, there is some general agreement between Latin scholars
about the main lines of interpretation of word order phenomena as mani-
festing pragmatic functions - see the standard descriptions in Kühner and
Stegmann (1955: sect. 245-249), Hofmann and Szantyr (1965: sect. 397-
410) and Pinkster (1988: ch. 9).9 So, in our analysis of Latin sentence
structure, we too will make use of these facts, in addition to those concern-
ing the position of weak pronouns.
This rather long introduction was intended to stress the limits of the
evidence available in the study of Classical Latin sentence structure and of
the conclusions we may draw from this evidence. With all this in mind, we
may pass to the analysis.
a) on the basis of the position of the enclitic weak pronouns (in non-italics)
we may separate the section of the sentence preceding the host constitu-
ent of the pronoun (marked in the following examples with a vertical
bar);
b) the constituents standing in this section of the sentence pragmatically
function as the Frame (often an adverbial clause, as in (9)) or the Topic
of the sentence; and
c) they may be introduced by the preposition de, meaning 'as for' (8):
A further structural position can be identified with the help of a) the posi-
tion of the weak pronouns, b) the pragmatic function of the constituents
that occupy that structural position, and c) their category:
a) the constituent in this position immediately precedes the weak pronouns
(in non-italics);
b) it functions pragmatically as a Focus; and
c) it may contain a Wh-word (11) or a quantifier (12):
Some firm points on Latin word order: The left periphery 439
Notice that the fact that a constituent stands before the weak pronoun is not
in itself proof that that constituent stands in the structural position identified
in this section, since weak pronouns also attach to constituents which are
not focalized:
440 Giampaolo Salvi
But while non-focussed constituents may also occupy other positions in the
sentence (e.g. ista res in (11a) appears after the weak pronoun), standing
before the weak pronouns is the only possibility for Wh-phrases. Although
the situation with other focussed constituents is not so clear-cut, we may
assume that there is a special structural position for Wh-phrases and other
focussed constituents at the beginning of the sentence (but after the posi-
tion identified in sect. 2.1). Frequently, only a part of the focussed con-
stituent appears in this position, the rest of the phrase standing inside the
sentence (15) or in a right-adjoined position (16) - a situation that is not
usual with non-focussed constituents: 10
Notice that if in (15) magnam were restitued into its normal position pre-
ceding difficultatem, the sentence would have the unmarked word order S
(haec res)-IO (Caesari)-DO ( m a g n a m difficultatem)-.. .-V ( a d f e r e b a t ) : we
may conclude from this, too, that magnam moved to occupy a special
structural position preceding that of the subject.
The verb, too, which usually stands in sentence final position (cf. e.g. (15)),
may appear in sentence initial position, in which case weak pronouns im-
mediately follow it (sentence connectives such as enim in (17a) are always
put after the first word of the sentence, so they do not count in the compu-
tation of structural positions):
It is worth noting that the anteposition of the verb never co-occurs with the
anteposition of a focussed constituent. Examples such as the following are
442 Giampaolo Salvi
no exceptions, for in (18) the verb form has not been preposed as in (17),
but it occupies the position of weak forms (Adams 1994b):
In other words, the word order focus-verb is only possible with forms of
sum, which, when weak, form a cluster with other weak forms in Wacker-
nagel's position, as we can see in (19):"
Cases such as (18) not being an exception, we may conclude that, since
focussed constituents and preposed verbs are in complementary distribu-
tion, they occupy the same structural position. But in current generative
theory, where these two types of elements belong to different categories,
this latter conclusion is not directly derivable: focussed constituents are
phrases, while verbs are heads, so they cannot occupy the same syntactic
position. Another, more abstract explanation must be found.
The theory assumes that in these cases we have two distinct structural posi-
tions and that in Italian only one of these may be occupied by phonetically
realized elements. The structure underlying (20ab) is therefore (21a), an
instance of the more general structure (21b) on which every wider syntactic
structure is based (cfr. Chomsky 1986; C = complementizer, Spec = speci-
fier, C o m p l = complement):
The correctness of this hypothesis is confirmed by the fact that there are
languages where both positions may be occupied, as in substandard Italian
(23a), to which we may assign the (partial) structure (23b):
To return to the Latin sentences (11)—(13) vs. (17), we can explain this
particular complementary distribution of focussed constituent and verb
along the same lines. In particular, following Brody's (1990) analysis of
Hungarian sentence structure, we may assume that Latin had a special
functional projection F(ocus) into whose specifier focussed constituents
were moved, in which case the head position of the projection remained
empty:
When the verb was preposed, it occupied the head position of the func-
tional projection, while the specifier position remained empty:
2.6. Conclusion
We can thus conclude that in the left periphery of the Latin sentence struc-
ture we must minimally distinguish two positions: one first position for
topic-like constituents, with the properties examined in 2.1, and one second
position for focus-like constituents, with the properties examined in 2.2.
More accurately, since we may have more than one topic-like constituent at
the beginning of the sentence (as in (27), where the constituents are num-
bered), we must assume that the first position is a recursive one:
But constituents with the function of topic/frame may also follow the in-
troductory element of the subordinate clause (33); moreover, the introduc-
tory element may be followed by a focus element (34); and topic/frame and
focus constituents may appear together (35) (=(9c)):
Since the left-dislocated constituents may both precede and follow the ele-
ments introducing the embedded clause, a univocal serialization may not
be established if the subordinators are placed in only one position. Rizzi
(1997) resolves this problem by postulating that the functional projection
responsible for subordination is composed of different layers: the specifier
of each layer can host only a specific type of constituent (relative Wh-phrase,
topic/frame constituent, interrogative Wh-phrase), while the complementizer
may appear in each one of the heads of the different layers (modulo other
independent principles of grammar which regulate the grammatical combi-
nations). If we assume the existence of three layers (Ci", Top" and C 2 "), we
get the structure (39) and the analyses in (40) for the sentences in (36)—(38):
A parallel solution may be envisaged for Latin (with F" instead of C2"): cf.
(41), an extension of structure (28), and the analyses for (32)—(35) in (42):
In the preceding discussion we did not consider the position of weak ele-
ments, a type of evidence which played an important role in our discussion
of word order in main clauses. In sect. 2 we saw that the domain inside
which weak elements occur exclude the sentence initial topic/frame con-
stituents (i.e. Top"): weak pronouns appear after the first constituent of F"
(focussed constituent (11)-(13) or preposed verb (17)) and, if F" is not
present, after the first constituent of I" (as in (14)). We therefore expect
them to appear in similar positions in embedded clauses, too.
This expectation is only partially fulfilled. When SpecF" is present, we
find the weak pronouns after the focussed constituent, as in (30a), (32a),
(34) and (35); also the examples in (29) would conform to our expectations
if we assumed that there the complementizer occupies the head F. But the
type represented by (30b), where the weak pronoun follows a relative Wh-
phrase, is not expected, as the host of the weak pronoun is outside the F"
domain (the relative Wh-phrase being in SpecC"). Likewise in the follow-
ing examples the weak pronouns precede the focussed constituent and are
hosted by a complementizer outside the domain of F" (43a) or by a relative
Wh-phrase in SpecC" (43b):
We must thus assume that in embedded clauses the domain for the place-
ment of weak pronouns may be extended from F" to C". This is confirmed
by examples such as the following, where the pronoun eum precedes a
topic/frame constituent, supposedly in SpecTop":
452 Giampaolo Salvi
This conclusion perhaps finds its rationale in the fact that the embedded
clause may be considered as a whole (as C") or in its core part (as F"): both
domains may serve as the placement domain for weak pronouns (but not
the intermediate domain Top", either in embedded or in main clauses).
In the light of the analysis sketched above, the treatment of the following
examples is not straightforward: 12
specifier the focussed constituent is placed. In (45b) the weak pronoun mihi
is placed between the topic/frame constituent mortem and the interrogative
Wh-phrase cur. the placement domain for weak pronouns in embedded
clauses could perhaps be extended to Top", besides C " (contrary to our
hypothesis in 3.4). But ex. (45c) is resistant to every kind of adjustment of
our general approach because the complementizer cum apparently occupies
a position too low in the structure: the constituent Curio ad focum sedenti
may well be a topic (in SpecTop"), magnum auri pondus is probably fo-
cussed (it is anyway rhematic and stands therefore in SpecF" or in I"),
Samnites, too, is rhematic (in I"), so in this clause the complementizer cum
seems to be placed in the interior of I". With a more c o m m o n word order
cum would have been in C (as in (33a)/(42c): cum Curio ad focum sedenti
magnum auri pondus Samnites attulissent) or in Top (as in (32b)/(42b):
Curio ad focum sedenti cum magnum auri pondus Samnites attulissent).
Looking at this example from a different point of view, we may notice
that in (45c) the subordinator is strikingly placed before the verb at the end
of the clause, and this analysis might also be extended to (45ab), although
these clauses are too short for us to draw firm conclusions from them. This
construction is possible not only with complementizers (45ac), but also
with interrogative (45b) and relative (46) Wh-words: 1 3
Examples (46), from Early Latin, are all the more striking because, other-
wise, relative Wh-phrases are always the first constituent in embedded
clauses. In all probability we have here an alternative form of the embed-
ded clause, a relic of the archaic age where the movement of constituents to
the front of an embedded clause was freer and, unlike Classical Latin, was
also possible in relative clauses (Watkins 1993: 9.3.1).
454 Giampaolo Salvi
Notes
1. Unless otherwise indicated, the examples in this chapter were collected by us.
The non-literal translations of Latin examples are those of Loeb's Classical
Library. Abbreviations used in the glosses: Abl(ative), Acc(usative), D(ative),
G(enitive), N(ominative). 1 am grateful to Katalin E. Kiss and to Alessandro
Parenti for helpful discussion.
2. Ex. ( 1 H 2 ) from Skärup (1975: 192, 431, 440).
3. Ex. (3) from Skärup (1975: 372, 191).
4. Ex. (4) from Fraenkel (1933 [1964]: 352 [127]) and Jones (1991: 86).
5. Doubling may make use of a stressed pronoun too, a possibility we have not
stated explicitly above.
6. Ex. (5c) from Skärup (1975: 195).
7. Ex. (6) from Skärup (1975: 438).
8. This would be, I suppose, Benincä's (in press) analysis, which develops
Rizzi's (1997) approach, although she does not cite examples of the type of
(6). Alternative analyses are possible, but we do not develop on this problem
here, as it does not concern the main topic of this study.
9. A note of caution is in order here: the pairing of linear word order arrange-
ments and pragmatic functions is traditional in the literature on Latin syntax
(and in studies written in the frame of the functional approach of the Prague
School), but these studies do not in general try to give a structural explanation
for the complex interaction of pragmatic and syntactic facts as happens in the
generative approach and in the present study. For a critique of the traditional
approach to diachronic explanation in syntax and an evaluation of the function
of syntactic theory in the diachronic domain cf. Salvi (2003).
10. Ex. (15) from Ostafin (1986: 6.2).
11. Ex. (19b) from Adams (1994b: 39).
12. Ex. (45c) from Jones (1991: 84).
13. Ex. (46) from Somers (1991).
References
Adams, James N.
1994a Wackernagel's Law and the Position of Unstressed Personal Pronouns
in Classical Latin. Transactions of the Philological Society 92: 103—
178.
1994 b Wackernagel's Law and the Placement of the Copula Esse in Classi-
cal Latin. Cambridge: The Cambridge Philological Society.
Benincä, Paola
In press On the functional structure of the left periphery: evidence from
medieval Romance. Probus.
Some firm points on Latin word order: The left periphery 455
Brody, Michael
1990 Remarks on the Order of Elements in the Hungarian Focus Field. In
Approaches to Hungarian, Vol. 3: Structures and Arguments, Istvän
Kenesei (ed.), 95-121. Szeged: JATE.
Calabrese, Andrea
1980 Sui pronomi atoni e tonici delPitaliano. Rivista di Grammatica
Generativa 5: 65-116.
Cardinaletti, A n n a
1992 On Cliticization in Germanic Languages. Rivista di Grammatica
Generativa 17: 65-99.
Chomsky, N o a m
1986 Barriers. Cambridge, Mass.: The MIT Press.
Diderichsen, Paul
1966 Logische und topische Gliederung des germanischen Satzes. In
Helhed og Struktur, 52-63. Kebenhavn: Akademisk Forlag.
Fraenkel, Eduard
1933 Kolon und Satz: Beobachtungen zur Gliederung des antiken Satzes
II. Nachrichten der Göttinger Gesellschaft der Wissenschaften. Phi-
losophisch-historische Klasse 319-354.
[1964] [reprinted in: Kleine Beiträge zur klassischen Philologie. Vol. I.
Roma: Edizioni di Storia e Letteratura, 93-130.]
Giacalone Ramat, Anna and Paolo Ramat (eds.)
1993 Le lingue indoeuropee. Bologna: il Mulino.
Graffi, Giorgio
1994 Sintassi. Bologna: il Mulino.
Haegeman, Liliane (ed.)
1997 Elements of Grammar. Dordrecht: Kluwer.
Hofmann, Johann Baptist and Anton Szantyr
1965 Lateinische Syntax und Stilistik. München: Beck.
Jones, F.
1991 Subject, Topic, Given and Salient: Sentence-Beginnings in Latin.
Proceedings of the Cambridge Philological Society 37: 81-105.
E. Kiss, Katalin
1987 Configurationality in Hungarian. Budapest: Akademiai kiado.
Kenesei, Istvän (ed.)
1990 Approaches to Hungarian. Vol. 3: Structures and Arguments.
Szeged: JATE.
Kühner, Raphael and Carl Stegmann
1955 Ausführliche Grammatik der Lateinischen Sprache. II: Satzlehre.
Leverkusen: Gottschalk.
Lenerz, Jürgen
1992 Zur Syntax der Pronomina im Deutschen. Sprache und Pragmatik
1-54.
456 Giampaolo Salvi
Meyer-Lübke, Wilhelm
1897 Zur Stellung der tonlosen Objektspronomina. Zeitschrift für ro-
manischen Philologie 21:313-334.
Ostafm, David Mark
1986 Studies in Latin Word Order: A Transformational Approach. Unpub-
lished Ph.D. dissertation, University of Connecticut, Storrs.
Pinkster, Harm
1988 Lateinische Syntax und Semantik. Tübingen: Francke.
Polo, Chiara
2002 Word Order in Latin, Italian and Slovene between Morphology and
Syntax. Unpublished Ph.D. dissertation, University of Padua.
Rizzi, Luigi
1997 The Fine Structure of the Left Periphery. In Elements of Grammar,
Liliane Haegeman (ed.), 281-337. Dordrecht: Kluwer.
Salvi, Giampaolo
1993 Ordine delle parole e struttura della frase nelle lingue romanze an-
tiche. In Grammatikographie der romanischen Sprachen, Christian
Schmitt (ed.), 455^177. Bonn: Romanistischer Verlag.
1996 From Latin Weak Pronouns to Romance Clitics. Budapest: MTA
Nyelvtudomänyi Intezete.
2003 Teoria sintattica e spiegazione diacronica. In Syntaxtheorien. Modelle,
Methoden, Motive, Elisabeth Stark and Ulrich Wandruszka (eds.),
291-309. Tübingen: Narr.
2004 La formazione della struttura di frase romanza. Ordine delle parole
e clitici dal latino alle lingue romanze antiche. Tübingen: Niemeyer.
Schmitt, Christian (ed.)
1993 Grammatikographie der romanischen Sprachen. Bonn: Romanis-
tischer Verlag.
Schwan, Edmond and Dietrich Behrens
1932 Grammaire de l'ancien frangais. Leipzig.
Skärup, Povl
1975 Les premieres zones de la proposition en ancien franqais. Essai de
syntaxe de position. Kobenhavn: Akademisk Forlag.
Somers, Maartje H.
1991 Theme and Topic: The Discourse Function of Initial Contituents in
Latin. [Paper presented at the VI th International Colloquium on Latin
Linguistics, Budapest.]
Stark, Elisabeth and Ulrich Wandruszka (eds.)
2003 Syntaxtheorien. Modelle, Methoden, Motive. Tübingen: Narr.
Watkins, Calvert
1993 II proto-indoeuropeo. In Le lingue indoeuropee, Anna Giacalone
Ramat and Paolo Ramat (eds.), 45-93. Bologna: il Mulino.
Classical Sanskrit, "wild trees", and
the properties of free word order languages
In this study, we investigate the word order of Classical Sanskrit, making use of
two searchable corpora and taking as our theoretical point of departure Staal's
(1967) "wild tree" proposal for the language. We show that this proposal, accord-
ing to which the daughters of a phrase are unordered in the syntax, can capture
much, but not all, of the attested word order variation in the language. We also
show, however, that these additional word orders can still be captured on a "wild
tree" analysis that (i) incorporates a "flatter" treatment of sentential structure, in
which the verb does not project a VP; and (ii) recognizes that long-distance de-
pendencies make a contribution to word order freedom distinct from that related to
"wild trees" themselves.
the phonological component" rather than the syntax itself (Chomsky 1995:
413) strongly echo Staal's concerns, despite otherwise great differences
between these approaches.
More to the point, a "wild tree" analysis permits a very natural descrip-
tion of many of the word orders attested in Classical Sanskrit. Yet, as Staal
(1967: 34) himself acknowledges, such an analysis by itself cannot capture
the full range of word orders attested even for simple transitive sentences in
Classical Sanskrit - in particular, capturing neither those due to long-distance
dependencies nor those due to verb fronting. What we shall show, however,
is that the basic "wild tree" claim can be readily extended to account for the
latter orders and further supplemented to account for the former. These
revisions preserve Staal's insight that the ordering of sister nodes within a
Classical Sanskrit phrase is essentially free, but (following his own sugges-
tions) admit into the language operations that underwrite dependencies
between phrases "longer-distance" than those attributable to "wild trees"
themselves.
Of course, a "wild tree" approach to Classical Sanskrit can also be seen
to have more general consequences for the analysis of word order permuta-
tions across languages. While it is true that at least some of the word order
patterns that this approach seeks to capture are particular to languages with
substantial word order freedom, others, such as that involving verb front-
ing, are found in a much greater range of languages. Thus, the proposal to
capture this and other patterns without invoking a "basic" structure repre-
sents a distinct mechanism for deriving such patterns. The question that
arises, then, is whether this mechanism should be understood as coexisting
in Universal Grammar with others that derive the same ordering effects or
as a theoretical alternative to approaches that recognize little or no variation
in base word orders and make prolific use of movement to derive attested
word orders. Sharpening our own interest in this question are two different
observations about word order. One, which pertains specifically to Sanskrit,
is that certain preferred or typical word orders have commonly emerged
from the examination of Sanskrit texts. The other, not unrelated, pertains to
the substantial body of research on "information packaging" (see e.g. Vall-
duvi 1993 for a survey). This research has revealed that truth-conditionally
equivalent sentences with different word orders may have substantially
different information-packaging properties and thus not be "interpretatively
equivalent in absolute terms" (Vallduvi 1993: 2). These considerations raise
the additional question of how one captures the fact of more "marked"
word orders in Classical Sanskrit and other languages, which correlate
460 Brendan Gillon and Benjamin Shaer
omitted altogether. Nouns and adjectives are inflected for gender (mascu-
line, feminine, or neuter), number (singular, dual, or plural), and case
(nominative, accusative, instrumental, dative, ablative, genitive, locative, or
vocative), with adjectives agreeing with the nouns that they modify. (Al-
though we take the grammatical function of these cases to be largely famil-
iar, it is worth noting that instrumental case, beyond indicating the means
by which or instrument with which an action is performed, also indicates,
for example, the person with whom an action is performed, i.e. a comitative
use, and the agent of the action in passive constructions.) The noun repre-
senting a clause's subject, though it need not appear (making this a "pro-
drop" language), must agree in number with the clause's main verb when it
does. Personal pronouns are inflected for case and number. Some of these
features of the language are illustrated in the following examples: 4
At the heart of the question of word order in Classical Sanskrit is the frequent
observation that a sentence like that in (2) can be equivalently expressed by
any of the five other possible linear orderings of its three words:
about the extent of such freedom permitted by the language more generally.
One way to highlight this is with a craftily chosen sentence from English,
which would - contrary to the intuitions of native speakers and substantial
linguistic research - suggest that it, too, tolerated considerable word order
freedom with little or no effect on meaning:
3. Staal(1967)
As just noted, Staal (1967) adopted in his own analysis of Classical Sanskrit
the traditional Indian view that the language's word order had essentially
no grammatical significance. Central to this analysis was his rejection of
what was long a standard assumption in generative grammar: namely, "that
order is not only a feature of the terminal strings and finally derived sen-
tences... but also figures in the deep structures underlying t h e m " (Staal
1967: 7-8). 7 Appealing to Pänini's concept of the karaka relations —
equivalent to current notions of thematic relations or valences - Staal took
such relations rather than word order to be grammatically basic, asking why
it was necessary to "preserve order in the deep structure" instead of intro-
ducing rules of the base that imposed no such order and left this to be de-
termined by other rules.
Staal's attempt to specify such rules forms the basis of his "wild tree"
analysis of Sanskrit word order. What he proposed was that the set of linear
orderings of words in a clause is precisely the set obtained from the free
transposition of all of the sister constituents in the clause, where such trans-
position induces no tangling. Staal referred to phrase markers that under-
specified linear order in this way "wild trees" (Staal 1967: 15). The phrase
structure rules that he proposed for generating such trees looked like those
in (3), where heads and phrases form unordered sets, as indicated by brace
brackets: 8
When applied to a Classical Sanskrit sentence like that in (2) above, these
rules result in the following four ordering possibilities (where " N P S " indi-
cates the nominative-marked subject and " N P 0 " the accusative-marked ob-
ject): 1 0
466 Brendan Gillon and Benjamin Shaer
(4) a. N P - V P permutations
i. S NPS-V-NP0
[HPS R ä m a h ] [Vp apasyat [Np0 Govindam]]
ii. S - > N P s - N P 0 - V
[HPS R ä m a h ] [Vp [NP0 Govindam] a p a s y a t ]
b. V P - N P permutations
i. S V-NPo-NPs
[vp a p a s y a t [NP0 Govindam ]] [ NPs R ä m a h ]
ii. S —* N P 0 - V - N P S
[VP [NPO Govindam apasyat ] [ NPs Rämah]
Significantly, all of the patterns in (4) are attested in the Apte corpus de-
scribed a b o v e , " if we include among VPs non-verbal predicate expressions
with no overt copula. As Gillon (2004: § § 4 . 2 - 4 . 3 ) shows, this corpus re-
veals (i) non-copular VPs preceding and following subject N P s ; (ii) VPs
with overt copula f o r m s that express existence preceding and following
subject N P s ; and (iii) predicates with no overt copula preceding and follow-
ing the subject noun phrase. These are illustrated in (5), (6), and (7), respec-
tively: 12
Note that this restriction may simply be an artefact of the relatively small
size of the corpus, and can be verified only through the investigation of
additional data. (Another possibility, however, is that this gap represents
something more systematic, which is related to the status of VP in the
grammar of Sanskrit. We shall be sketching this possibility in §4.3 below.)
One aspect of Staal's "wild tree" analysis that seems squarely at odds with
much recent research is his claim that word orders in Sanskrit (and many
other languages) "have no grammatical significance" (Staal 1967: 59).
Such a claim is a contentious one, since it appears to deny the relation
commonly observed between word order and the information-packaging
properties of sentences. As it happens, Staal (1967: 6 3 - 6 4 ) does not see his
claim as incompatible with the possibility of "two Sanskrit sentences...
[differing only] in the arrangement of the words" but not being inter-
changeable, the sentences being understood to give rise to "a difference in
meaning" or not to "fit into larger contexts equally well." This is because
such differences in word order, though clearly relevant to interpretation
broadly construed, are for Staal best treated in a theory of performance
rather than competence. N o w , such a conclusion might be consistent with
views of the competence/performance distinction current at the time that
Staal was writing (see e.g. Chomsky 1965: 11 for some remarks on this).
However, the detailed investigation of information packaging conducted
since then (see e.g. Birner and Ward 1998; Lambrecht 1994; Prince 1981;
Vallduvi 1993) strongly suggests that at least some reflex of speakers'
knowledge of such properties is very much part of their linguistic compe-
tence and not plausibly relegated to performance after all.
W e might still be able to reconcile these diverging views of word order
by observing that on a "modular" view of linguistic competence, the inclu-
sion of information-packaging properties of sentences within the grammar
should have little bearing on the "wild tree" claim in any case. This is be-
cause the determination of possible word orders by the syntax can, on this
view, be seen as orthogonal to the information-packaging functions to
which different word orders are put (granting the possibility of certain word
orders correlating directly with certain information-packaging functions).
Moreover, the strategy of deriving less common word orders with special-
ized information-packaging functions from more basic syntactic structures,
though widely employed in describing the relation between syntactic and
information structure, seems to us considerably less plausible than ones
consistent with the "wild tree" claim. 15 One strategy of the latter kind would
Classical Sanskrit, "wild trees ", and the properties of FWO languages 469
A second and, for some, again undesirable consequence of the "wild tree"
analysis of Classical Sanskrit is that it offers no characterization of phrases
in the language in terms of the position of their heads relative to their other
daughters, even though such a characterization has been widely seen as
basic to the description of phrase structure across languages. In fact, though
we have not yet subjected this matter to detailed scrutiny, what we have
found so far gives us little reason to include a specification of head position
in the syntactic description of Classical Sanskrit. For example, such a speci-
470 Brendan Gillon and Benjamin Shaer
As it happens, such corpus findings are broadly consistent with some key
results of Minimalist research, including its dispensing with primitive dif-
ferences in the direction of case and theta role assignment in deriving cross-
linguistic word order differences and its analysis of certain word orders,
such as those associated with PP extraposition, in terms of processes "in the
phonological component" (Chomsky 1995: 426; see also Chomsky 2005)
Classical Sanskrit, "wild trees ", and the properties of FWO languages 471
rather than the syntax. Taken together, such conceptual and empirical con-
siderations point to a treatment of head position in Classical Sanskrit
phrases not as reflecting any basic syntactic properties of the language, but
rather as determined largely by non-syntactic factors, including those of the
kind enumerated by Gonda (1952), as described in §2.1. Although these
remarks are obviously preliminary, they do at least suggest that the head-
final tendency observed of Sanskrit phrases may be better captured in terms
consistent with a "wild tree" analysis than in terms of extraction operations
that apply to "basically" head-final phrases or of some bifurcation between
" u n m a r k e d " and " m a r k e d " head positions, respectively related to the "core"
and the "periphery" of the grammar.
Now, Staal was not unaware of this problem, and observed that while "the
points dominated by a single node" in a wild tree can "be arranged in any
desired order", "rearrangements are required" to derive additional orders
"which go beyond the confines of single constituents or elements domi-
nated by a single node or category symbol" (Staal 1967: 34). In other words,
Staal clearly recognized the existence of dependencies between pairs of
phrases that were "longer-distance" than could be captured on his "wild
tree" analysis, even though he did not actually pursue this point in his
study. 20
Since there is, in fact, ample evidence for a range of "longer-distance"
dependencies in Classical Sanskrit, which (as we shall be suggesting in §5)
can be handled in terms of extraction, it would seem straightforward to treat
sentences like those in (12)—(13) in similar terms. Accordingly, the word
order exemplified in (12a)-(13a) would arise from leftward extraction of
the object N P out of a sentence-final VP; and that exemplified in (12b)—
(13b) from rightward extraction of the object NP out of a sentence-initial
VP.
Of course, while an appeal to extraction here does serve to account for
the observed word orders, it does so at a price: namely, that it robs the
"wild tree" analysis of some of its cogency. This is because such as analy-
sis seeks to offer an alternative to movement-based approaches of the word
order facts outlined in (4), so that the case for recognizing "wild trees" in
the first place is somewhat undermined by a recourse to movement in ac-
counting for similar word order facts. It is also because the recourse to
movement here introduces an asymmetry between these word orders and
those given in (4), which is at odds with the original observation that all of
the permutations in question are syntactically and semantically equivalent.
As it happens, an alternative to movement exists for these cases which is
very much in the spirit of Staal's (1967) basic "wild tree" analysis. This is
to take the Sanskrit sentence to be even flatter than Staal claimed, and in
particular to take the verb not to project to the phrasal level - making it, its
sisters, and the subject N P all immediate constituents of the sentence. Such
Classical Sanskrit, "wild trees ", and the properties of FWO languages 473
Interestingly, this structure is not only consistent with the word orders in
(12)—(13) - which are accordingly treated not in terms of discontinuous
V P s but simply as t w o available linear orderings of the immediate constitu-
ents of a " f l a t " sentence - but also offers a solution to certain other word
order puzzles. O n e is related to our failure so far to uncover sentences in
the Apte and PV corpora in which the copula and a predicate adjective pre-
cede the subject, as illustrated in (8b) and repeated in (15a) below; and
more generally to uncover sentences in which the verb and its sisters be-
have as a constituent. If we reanalyse (15a) as in (15b), then the copula and
the predicate adjective simply do not form a constituent and thus would not
be expected to behave as one:
O f course, this "no V P " analysis would not actually rule out such a struc-
ture, given the "wild tree" claim. At this stage, this is what we want, since a
more extensive corpus search might still reveal such a structure. However,
the claim that there is no V P constituent would arguably m a k e the infre-
quency of a sentence like that in (15) less surprising, particularly given the
possibility of other ordering factors conspiring to m a k e such a sentence
infrequent or even unavailable.
Another word order puzzle that a "no V P " analysis would help to explain
is the anomaly reported by Gillon (1996) in the restrictions on extraction
474 Brendan Gillon and Benjamin Shaer
If there is, in fact, no VP node here, then this AP can be analysed as having
simply been extracted from the instrumental NP that is an immediate con-
stituent of the sentence.
Admittedly, the claim that Classical Sanskrit had no VP is a strong one,
with far-ranging implications for its grammar - predicting, in particular, the
absence of a range of subject/object asymmetries that have provided sup-
port for VP structure. These asymmetries include ones related to referential
and quantificational dependencies between subjects and objects and the
possibility of pro-forms and of ellipsis processes that include the verb and
object but exclude the subject (see e.g. Baker 2001: 407-418). So far, we
have uncovered no compelling evidence for such asymmetries in our cor-
pora. We have, however, discovered the following sentence, with an object
reflexive occurring to the left of the verb:
dependency relation are the putative daughters. Although we take the ulti-
mate analysis of these dependencies to be an open question, our findings
nevertheless suggest that a useful distinction can be drawn between the free
transposition of a phrase's daughters at the heart of the "wild tree" analysis
and word order permutations that are the result of other processes, thus
providing strong indirect support for "wild trees" themselves.
The long-distance dependencies that we have observed in the Apte and
PV corpora, and the constraint that appears to be operating on them, can
best be illustrated with constructed examples like those in (18). In this set,
(18a) represents a sentence with no such dependencies, while (18b) repre-
sents one with a licit dependency and (18c), which is unattested and thus by
hypothesis ungrammatical, represents one with an illicit dependency:
phrases in the language gives rise to a kind of island effect - albeit an un-
familiar-looking one, a point to which we shall be returning below.
Now, despite the robustness of the dependency patterns just described,
which we shall be describing in more detail below, various considerations
make it unclear that they should, in fact, be assimilated to standard extrac-
tion processes. The most obvious one is the very restricted nature of these
dependencies in Classical Sanskrit: unlike extraction in English and other
languages in which such processes have been studied extensively, the de-
pendencies in Classical Sanskrit are always clause-bound whether the dis-
placed element occurs on the left or the right edge of the clause. 22 This
makes the extraction in question a peculiarly local process, and thus more
like the clause-bound phenomenon of extraposition - which, as noted
above, has been seen by Chomsky (e.g. 1995, 2005) and others as not even
part of the syntax proper - than like unbounded extraction processes such
as wh- and N P movement. Adding to the doubts about a treatment of these
dependencies in terms of extraction are the island effects that we have of-
fered in support of such a treatment. Notwithstanding the unfamiliar form
of these island effects (which arguably makes the connection to well-
studied extraction processes even more tenuous), there has been a growing
recognition that even familiar island effects are more likely the result of
processing factors rather than the violation of purely syntactic constraints
(see e.g. Kluender 1998). Thus, while it may well be that the hypothesized
island effects in Classical Sanskrit, as described above, have the same
source as more familiar ones, the source in question is not likely to be
purely syntactic.
Of course, even setting aside the possibility that extraposition does not
represent true extraction, it must still be noted that extraction processes are
not the only grammatical means to establish long-distance dependencies,
nor therefore the only possibility for describing patterns like that in (18).
One intriguing alternative, for example, is that these dependencies are es-
sentially anaphoric ones not established through extraction. 23 On such an
analysis, the two phrases in a dependency could, syntactically speaking,
have simply an appositive or correlative relation, with case-marking estab-
lishing the anaphoric connection between them. As applied to (18b), this
would suggest a literal paraphrase along the lines of Ά man is going to the
forest, he taller than Rama'. While this possibility is indeed an intriguing
one, we have not yet subjected it to sufficient scrutiny to determine how
well it generalizes to all of the long-distance dependencies that we have
uncovered in the Apte and PV corpora - many of which still look, to our
478 Brendan Gillon and Benjamin Shaer
Let us first turn to cases of leftward extraction. What we can report here is
that these occur in all types of clauses, including main, relative, infinitival,
participial, gerundial, sentential subject, and absolutive clauses (see Gillon
1996 for descriptions and illustrations of these various clause types).
For example, we find phrases extracted leftward from main clause sub-
ject and verb complement N P s in sentences like those in (19a) and (19b),
respectively:
The cases of leftward and rightward extraction given here are among the
many that we have found in the Apte and PV corpora, including certain
more complex ones that await further analysis. We take these examples to
provide solid evidence that such processes did exist in the grammar of
Classical Sanskrit, and understand these processes to have operated on the
phrases in "wild trees". One question that naturally arises from our recogni-
tion of extraction in the language is how this was related to the language's
"wild tree" architecture. At this stage we can offer only a tentative answer
to this question, although one that we consider promising, given the paral-
lelism displayed by leftward and rightward extraction. This is that extracted
phrases were similarly unordered with respect to the rest of the sentence.
What Classical Sanskrit might have had, in other words, was wild trees all
the way up.
Of course, there are many other questions that our discussion of extraction
in Classical Sanskrit has so far left unanswered. Among these is the status
of various elements found at the periphery of the clause, which, as Gillon
(1996) observes, include such elements as interjections, vocatives, clausal
conjunctions, and sentential adverbials. Some of these peripheral elements
are illustrated in (22):
c. Adverbials:
[NP sarvatra ][ N P audarikasya],
in every case-LOC glutton-GEN
[ [ N P a b h y a v a h ä r y a m eva] [nps ι visayah]]
food only object
'In every case, a glutton's object is only food.' (SG 1.1.2)
d. Vocatives:
[isipsakhe] [Ap sä eva ]! [ AP dhänyä ]
friend-VOC the-NOM just lucky-NOM
[NPS Ι gänikä- därikä ]
prostitute-daughter-NOM
[s' [NP yarn] evam [ NPs bhavan-manas ] abhinivisate]s
to w h o m - A C C in this way y o u - m i n d - N O M is devoted
'Friend, the very same prostitute's daughter to whom your mind is
devoted in this way is the lucky one.' (SG 30.1.3)
Given our distinction between word order permutations due to extraction and
those due to "wild trees", the question of what positions such elements occu-
py turns out to be a relevant one, as we shall be suggesting in what follows.
Inspection of the examples in (22) reveals that most of these elements
appear at the left periphery of the clause but are not confined to this periph-
ery, given that we also find cases of right-peripheral conjunctions like that
in (22b). Moreover, since in their left-peripheral occurrence these elements
precede extracted elements, we might take them to occupy positions be-
yond the landing sites of extraction, which are themselves standardly as-
sumed to be beyond S.
Yet, the question arises whether these peripheral elements actually do
mark sentence edges, 2 4 given cases like the following ones, in which
phrases hypothesized to be extracted occur to the left of these elements:
The examples in (24) clearly indicate that vocatives can interrupt the sen-
tence at various points, and thus that any material to their left (other than
interjections or other extrasentential expressions) might still be contained in
the sentence. This indicates, in turn, that vocatives can reliably detect sen-
tence boundaries only when no material occurs to their left.
But here another complication arises, this one related to clausal conjunc-
tions, another in the set of peripheral expressions described above. The
complication here is that vocatives may occur to the left of clausal conjunc-
tions l i k e y a d i ' i f , as shown in (25):
What is still left to determine, though, is the status of conjunctions that may
occur non-initially but are not known to be clitics. These include not only
484 Brendan Gillon and Benjamin Shaer
yadi, as just noted, but also tar hi 'then', as given in (23c). Given the latter's
apparent function of relating one sentence to another in a discourse, it
seems reasonable to assign it an "extrasentential" status and take its non-
initial occurrence to be explainable in much the same way that such an
occurrence of vocatives is. (Note, though, that the possibility of its being
more tightly integrated into the sentence cannot be ruled out, particularly
given analogous elements in German, like dann 'then', that are clearly inte-
grated into sentence structure.) 27 As regards yadi, however, its meaning in-
dicates a function as a clause-introducer, just like its counterparts in English,
German, and other languages, and thus a syntactic status as an element
outside S but still within the sentence periphery. 28 Such a status is further
supported by its occurrence to the right of another peripheral element, the
adverb tat 'then', as in (27):
Of course, a great deal of work remains to be done to sort out the analysis
of conjunctions and other peripheral elements. Yet, the importance of doing
so is clear, given the role of such elements in determining sentence bounda-
ries and in helping to distinguish between extracted and freely arranged
phrases. Among the many questions left unanswered here is whether we
can, in fact, isolate a class of sentence-internal conjunctions in Classical
Sanskrit, which cannot occur higher than extracted elements; and, relevant
to the first question, whether the language permits multiple extraction of
phrases. Interestingly, the latter question, in turn, is closely related to the
question of whether Classical Sanskrit had a VP, as discussed in §4.3. That
these three questions are indeed interrelated becomes clear if we once again
consider the sentence given in (23c), which is repeated below:
6. Conclusion
Notes
* This study grows out of the first author's research on Classical Sanskrit, as
reported, in particular, in Gillon (1996, 2004), which the second author has fol-
lowed closely over many years. The first author wishes to express his gratitude
to his teacher and friend Sri Venugopalan for assistance in the research on
which the study is based; and to the late Pandit Sri T. S. Snniväsa Sästri for
helpful advice on the analysis of various sentences. The second author wishes
to thank Philippa Cook, Werner Frey, Ewald Lang, and Luis Lopez for helpful
discussion. Both of us wish to thank Steven Schäufele for very graciously
sending us a copy of Schäufele (1990), Ruth Kempson for detailed comments
on an earlier version of this chapter, and Katalin E. Kiss for her considerable
efforts to put this volume together.
1. Interestingly, Ross's original proposal took this operation to be part of a "sty-
listic component" rather than of the transformational component of the gram-
mar (on this, see Thräinsson 2001: 154-255). As such, recent work on such
operations that conceives of them as part of the PF component brings them
more in line with Ross's original conception and with the ideas presented here.
2. This is perhaps easier said than done; and inevitably our sketch lends itself
more easily to constraint-based than derivation-based formalization.
3. Although adjectives and nouns in Sanskrit are sometimes thought to be non-
distinct categories, Dash and Gillon (1995: 291-294) present some reasons for
distinguishing them.
4. We use the following abbreviations in the glosses of this and other examples:
(word classes) ADV: adverb, CNJ: conjunction, INJ: interjection, PRT: particle;
(cases) ABL: ablative, ACC: accusative, DAT: dative, ERG: ergative; GEN:
genitive, INST: instrumental, LOC: locative, NOM: nominative, VOC: vocative;
(numbers) SG: singular, DU: dual, PL: plural; (others) INF: infinitive; PRIOR-
COMP: prior completion; PRT: particle.
5. In these and other examples given in the text, we suppress sandhi.
6. A difference is generally recognized between the word order variation permis-
sible in Sanskrit poetry and that permissible in Sanskrit prose. Since we take
the former, on the one hand, to permit greater freedom than prose allows and,
on the other, to be subject to artificial constraints imposed by metrical con-
siderations, we shall restrict our discussion of word order to that of sentences
in prose texts.
7. Of course, as already noted in the text, such an assumption is rejected in Mini-
malist research, as indicated by Chomsky's (1995: 413) remark that "[t]here is
no clear evidence that order plays a role at LF or the computation from Ν to
LF."
8. Note that Staal's (1967) rules make use of only zero and maximal levels of
phrase structure, consistent with assumptions current at the time he was
writing - and, as it happens, with Chomsky's (e.g. 1995: 396) hypothesis that
488 Brendan Gillon and Benjamin Shaer
only minimal and maximal but not intermediate projections are visible "at the
interface and for computation". Because, like Schaufele (1990: 116), we have
found no language-internal evidence that leads us to abandon this assumption,
we shall make use of it throughout our discussion.
9. Staal's original rules also make reference to case and number morphology,
which we omit here.
10. In fact, Staal offers no explicit statement for deriving word orders in Classical
Sanskrit such that they are free up to the free transposition of sisters within a
constituent. Such a statement is, however, offered by Gillon (1996, 2004), who
conjectures that the words in a given sentence may be linearized in any way
whatsoever, modulo the constraint that he formulates as in (i), which he dubs
"Staal's Constraint":
(i) Let Τ be a labelled, order-free phrase marker associated with a sentence S;
let rij and nj be distinct sister nodes in T. Then, if any word in S dominated
by nj precedes any word in S dominated by nj, then every word in S domi-
nated by nj precedes every word in S dominated by η(.
Given this constraint, constituents will be continuous and word order will be
free among the sisters of a given constituent. Gillon's conjecture has subse-
quently been proven by Nielsen (2004).
As it happens, Staal (1967: 34, 79) recognizes the necessity of additional
extraction rules that create discontinuous constituents from these continuous
ones, in order to capture additional word order patterns attested in Classical
Sanskrit. On this, see §4.3 below. However, this constraint on linearization ap-
pears to apply equally well to the output of the syntax in which such extraction
rules also figure.
11. Short references to this corpus are indicated with "SG" (abbreviating the title of
Apte's work, Student guide to Sanskrit composition), followed by the relevant
chapter, exercise set, and example sentence.
12. Note that VP and NP are separated by the conjunction tu in (5b), which occurs
in second position in the sentence. The question of what structural position
such elements occupy is taken up in §5.3.
13. We are assuming for present purposes that the negative element na 'not' is
dominated by VP. That this analysis is not the only one available will emerge
in §4.3.
14. Short references to this corpus are indicated with "PV", followed by the rele-
vant line and page of this work as edited by Gnoli (1960).
15. Staal (1967: 56) was already expressing scepticism of rules that apparently
served "to derive less frequent expressions from more frequent ones".
16. Note that compounds in Classical Sanskrit are always head-final (see e.g. Gillon
1995; Killingley and Killingley 1995: 42-46). However, because word-level
and phrase-level operations do not obviously make reference to the same set of
principles (pace Marantz 1997 and others; see Smirniotopoulos and Joseph
Classical Sanskrit, "wild trees", and the properties of FWO languages 489
27. Thanks to Ewald Lang for his timely reminder of "integrated" and "non-
integrated" possibilities for peripheral elements in German.
28. Note that we have resisted labelling such elements in Sanskrit as complemen-
tizers, since it is unclear whether they have the same formal characteristics as
counterparts in other languages that are identified as such. On this matter as it
pertains to Vedic Sanskrit, see esp. Schäufele (1990: 146).
References
Kroeger, Paul R.
1993 Phrase structure and grammatical relations in Tagalog. Stanford:
CLSI.
Lambrecht, Knud
1994 Information structure. Cambridge: Cambridge University Press.
Laughren, Mary
1989 The configurationality parameter and Warlpiri. In Configurationality:
The typology of asymmetries, Laszlo Maracz and Pieter Muysken
(eds.), 319-353. Dordrecht: Foris.
Lazzeroni, Romano
1998 Sanskrit. In The Indo-European languages, Anna Giacalone Ramat
and Paolo Ramat (eds.), 98-124. London/New York: Routledge.
Marantz, Alec
1997 No escape from syntax: Don't try morphological analysis in the
privacy of your own lexicon. In Proceedings of the 21s' annual Penn
Linguistics Colloquium, Alexis Dimitriadis et al. (eds.), Working
Papers in Linguistics 4 (2): 201-225. Philadelphia: Department of
Linguistics, University of Pennsylvania.
McCawley, James D.
1998 The syntactic phenomena of English. 2 nd edition. Chicago/London:
University of Chicago Press.
Nielsen, Janet
2004 A proof and discussion of Staal's theorem. Honours thesis, Depart-
ments of Linguistics and Mathematics, McGill University, Montreal.
Pollard, Carl and Ivan A. Sag
1994 Head-driven phrase structure grammar. Chicago: University of
Chicago Press.
Prince, Ellen
1981 Toward a taxonomy of given-new information. In Radical pragmatics,
Peter Cole (ed.), 223-56. New York: Academic Press.
Ross, John R.
1967 Constraints on variables in syntax. Unpulished doctoral dissertation,
MIT, Cambridge, Mass.
Schachter, Paul and Fe Τ. Otanes
1972 Tagalog reference grammar. Berkeley, CA: University of California.
Schäufele, Steven W.
1990 Free word-order syntax: The challenge from Vedic Sanskrit to con-
temporary formal syntactic theory. Doctoral dissertation, University
of Illinois, Urbana-Champaign.
Smirniotopoulos, Jane C. and Brian D. Joseph
1998 Syntax versus the lexicon: Incorporation and compounding in Modern
Greek. Journal of Linguistics 34: 447-^488.
Classical Sanskrit, "wild trees ", and the properties of FWO languages 493
Speyer, J. S.
1896 Vedische und Sanskrit-Syntax. (Grundriss der Indo-Arischen Philo-
logie und Altertumskunde 1 (6).) Strassburg: Trübner.
Staal, J. F.
1967 Word order in Sanskrit and Universal Grammar. (Foundations of
Language, supplementary series, 5.) Dordrecht: Reidel.
Thräinsson, Höskuldur
2001 Object shift and scrambling. In Handbook of contemporary syntactic
theory, Mark Baltin and Chris Collins (eds.), 148-202. Maiden,
Mass./Oxford: Blackwell.
Vallduvi, Enric
1993 Information packaging: A survey. Research Paper HCRC/RP-44.
Edinburgh: University of Edinburgh.
A particular coordination structure
of Indo-European flavour
1. Preamble
As Ross (1967: 90-91) has already observed, (1) shows that the conjunction
forms a constituent with the second conjunct and definitely not with the
first. Moreover, a symmetrical structure cannot explain why an R-expression
in the first conjunct may be co-referent with a pronoun in the second and
not vice versa, as is revealed by (2), which we have again taken from Munn
(1992: 20):
calls the maximal projection of this head a Boolean phrase (BP), the term
Boolean standing for any type of logical operator. In (3) the entire BP,
whose head B° has the second conjunct, the YP, as its complement, is as-
sumed to be adjoined to the right of the first conjunct, the XP:
(3) [XP[XP][BPB°[YP]]]
Kayne (1994: 143, note 3) suggests structure (6) as a source from which
cases like (5) are obtained:
In his view the head of the YP rises leftward to the higher head position,
where it is left-adjoined to que·}
With regard to (8), however, two structural alternatives, (9) and (10), are
possible:
498 Emanuele Lanzetta and Lucio Melazzo
In (9) the &P 2 is complement of the &Pi, whereas in (10) the &Pi occupies
the Spec,&P2 position. W e shall soon see the reasons why the structure
illustrated in (10) is preferable to that presented in (9).
As expected, K a y n e ' s analysis of the enclitic conjunction is perfectly
appropriate for coordination by both a single and a double ca in the Indo-
Iranian languages, (j)a in Hittite 3 , te in Ancient Greek:
Let us take another example from Ancient Greek to note that the analysis
we have been proposing has the advantage of explaining without any further
stipulations such coordination phenomena as (13), in which the enclitic
conjunction te and the stressed kai are adjacent:
Here the first conjunct remains in its original position of complement of te,
and in PF this conjunction leans on the participle opsomenai4.
Among Indo-Iranian languages Vedic Sanskrit and Avestan exhibit a
peculiar type of coordination by the enclitic form ca. In this usage such an
enclitic conjunction is called inverse ca , for in the pronunciation it rests on
the first conjunct rather than on the second 6 :
The idiosyncratic character of this kind of coordination has not slipped the
attention of several scholars, who have dealt with the question either by
referring to positional problems connected with Wackernagel's Law (see
Klein (1985: 193), who, however, argues that there are three other ways of
explaining this phenomenon), or alternatively by discovering a likely bidi-
rectional property of the enclitic conjunction (Melazzo 1997) 7 . As regards
the origin of inverse ca, any diachronic investigation which tends either to
establish the precedence of the non-inverse over the inverse use of ca or
describe the latter as secondary compared with the former may be left out
of consideration. However, Klein's (1985: 191-192) hypothesis arouses
some interest. According to this scholar, the use of inverse ca might have
arisen out of coordination structures with a double ca, probably as a result
of the deletion of the enclitic conjunction leaning on the second conjunct
when uttered. Such a hypothesis may in truth constitute an important start-
ing-point for an attempt to determine the precise structural configuration of
coordination by inverse ca from the antisymmetric perspective we have
adopted. In point of fact, such a phenomenon as inverse ca could create
some problems for an approach to coordination from an antisymmetric per-
spective on syntax. Let us assume that in a coordination structure the con-
junction is the head of a phrase of which the positions of specifier and
complement are occupied by the first and the second conjunct respectively.
The representation of this phrase, whose head is a non-inverse ca, will
therefore be the following:
(17) &P
XP &P
YP
ca
(18) &P
XP &P
YP
ty°
In (19) dyaiis e prthivi are coordinated by an inverse ca. Besides, dyaiis and
ca happen to precede the relative pronoun yäm. The occurrence of an ele-
ment on the left of either a relative pronoun or a complementizer is usually
not an isolated phenomenon in Vedic Sanskrit or in other Indo-European
languages. As for Vedic Sanskrit, Hale (1987: 41) puts the fact at issue
down to a topicalization movement to the initial position of the sentence.
Rizzi (1997) has recently described topicalization as a syntactic operation
which moves a phrase endowed with a [topic] feature into the specifier po-
sition of a projection of the left periphery of the sentence, whose head is
provided with the same feature so as to realize a specifier-head agreement
configuration (see also Bianchi 1999: 183). Should this approach be ap-
plied to (19), the sequence [dyaiis ca\ will structurally represent a non-
terminal node. It is therefore clear that the movement of a head cannot de-
termine the correct configuration of the constituents in (19). 8 As a further
502 Emanuele Lanzetta and Lucio Melazzo
consequence this observation makes it unlikely that the order of the con-
juncts dyaüs and prthivi originates from such a structure as (18). Hence
coordination by inverse ca cannot derive through a syntactic operation from
a structure where coordination is expressed by non-inverse ca. If inverse ca
is base-generated, we must ask ourselves what structure could be assigned
to the string devebhyas ca pitribhya in (16). Most importantly, it must be
excluded that the conjunction ca dominates both the constituents; otherwise
there would be room for a symmetric structure, the possibility of which is
not allowed on theoretical grounds. We prefer to remain within the anti-
symmetric framework and hypothesize a structure with two heads project-
ing two separate &Ps. Naturally, one of these heads is ca and the other will
be null. This means that coordination by inverse ca will involve both the
structural alternatives represented in (9)—(10), except that the second con-
junction is not phonetically realized:
Now that we have seen that the first conjunct of a structure containing an
inverse ca can be topicalized and that a topicalized element always has to
be a phrase, it is only the representation in (20b) that reproduces a coordi-
nating phrase with inverse ca correctly. Indeed since it stands in Spec,&P 2 ,
only in (20b) can the &Pi still move after devebhyas has been left-adjoined
to ca. Conversely in (20a), the &P| with head ca has the &P 2 as its com-
plement: in this state of things the &P| cannot move as a maximal projec-
tion without the pied piping of its complement. If we are on the right track,
(21), which exhibits the adjunction of the head of the XP to ca, describes an
immediate transition from the structure in (20b).
A particular coordination structure of Indo-European flavour 503
The correctness of (21) remains to be settled in that the grounds for justify-
ing the empty head of the &P2 still need to be given. Klein (1985) suggests
that in the instances of coordination that we have regarded as being ob-
tained from such a structure as (21) there should be the deletion (of course
at PF) of ca which is the head of &P2, but this cannot be an adequate expla-
nation because no reasons for the need for such a deletion are adduced by
that scholar. On the other hand, the existence of coordination by non-
inverse ca points to the absolute arbitrariness of any deletion phenomena in
the inverse ca construction. Let us briefly consider the well-known asym-
metry in (22) taken from Kayne (1994: 57) which exemplifies the coordina-
tion of more than two conjuncts:
As stated by Kayne (1994: 57) and shown by (23), in (22a) a subpart [Bill
[and Sam]], with and as the head, has to be taken into account. To this John
can be adjoined at the left, licensed by another head, which in English may
not be overt:
There could also be the same covert head in (22b), and (24) ought to repro-
duce its structure:
Kayne (1994: 58) ascribes the grammatical ity of (22a) and (23) to the LF
raising of and, which licenses the phonetically unrealized X°, the ungram-
maticality of (22b) and (24) being due to the non-existence of a parallel LF
lowering of the overt conjunction capable of licensing the covert X°. If we
were to apply the same line of reasoning to the inverse ca coordination that
we have described in (21), we should paradoxically conclude that such a
construction is ungrammatical. This conclusion clashes, however, with the
evidence of the data which testify that such a coordination pattern of the
type in (21) belonged to the grammatical competence of the Vedic Sanskrit
and Avestan speakers.
504 Emanuele Lanzetta and Lucio Melazzo
It is now the time to put forward our solution to the question of how the
phonetically unrealized head is licensed in such cases as (21). As (25) shows
clearly, our solution consists in assuming that the head of the second con-
junct moves leftward in order to be left-adjoined to the covert head of
9
the coordinate phrase:
(25)
devebhyas x ca tx pitribhya y
In point of fact, as (27) makes clear, the Indo-Iranian and other Indo-
European languages repeatedly present constituents coordinated without
recourse to any conjunction:
The sentences in (27), where the conjuncts are in bold type, are taken from
Old Avestan, Old Persian 1 0 , Hittite and Latin texts. The type of coordination
illustrated by (27) can also involve more than two conjuncts. (28) is a good
Old Avestan example of this possibility: 11
UZU UZU
(30) ser=a=ssan MG.GIG SA
ser-ADV-and-CONJ-ssarc-PARTC meat-liver meat-heart
kuirzi dai
CUt-PRES-3SG put-PRES-3SG
'and (he) cuts the liver (and) the heart (and) puts (them) here'.
(Wisurijanza Rs. 7 - 8 )
Here it is not immediately clear which phrasal categories are involved in the
coordination. Two VPs or at most two IPs might come into play. Naturally,
the assumption must be made that the complement in the second conjunct is
deleted because of its identity with the complement in the first conjunct.
Nothing in (30) allows one to hypothesize an instance of the coordination
of full sentences. However, cases like the following in which the verb is
left-adjoined to a sentence connective are to be found in Hittite:
In (31) the verb tarueni is clearly attached to the left of the adversative en-
clitic conjunction ma. It must be considered that the unmarked word order
of the Hittite sentence requires the verb to be in final position and this con-
dition on the sequence of constituents is fairly strict in that language. A
similar instance is also to be found in archaic Latin:
A particular coordination structure of Indo-European flavour 507
In (32) the verb duis is also adjoined to the left of que functioning as a
clause conjunction. Similarly, in archaic Latin the verb is subject to the
same rules of position that apply to the Hittite verb. A s far as the asyndetic
coordination of sentences is concerned, (33) proves even more revealing:
Here the negative item natt(a) occurs at the beginning of the second coor-
dinated sentence. Interestingly, the same sequence is previously found in
the same text with an important variation:
What marks the difference between (33) and (34) is the occurrence in the
latter of the enclitic particle ku, which Watkins (1994) has shown to be the
Hittite continuation of the Indo-European enclitic *kwe. A fair inference can
be drawn f r o m this evidence: in (33) natt(a) performs the identical process
by which its twin in (34) is left-attached to the conjunction ku, but the head
at which it has landed, is phonetically unrealized. 1 3
508 Emanuele Lanzetta and Lucio Melazzo
Something similar to what has been noted in Hittite is found in Old Persian
when the negative form naiy is involved. Thus, in (35) the second negative
seems to play the same role as e.g. Latin neque:
There is not enough evidence here to prove this. However, the comparison
between (36a) and (36b) will arouse greater interest:
Notes
* The two authors worked jointly on the whole content of this article; however,
as far as legal requirements are concerned, E. Lanzetta takes official responsi-
bility for sections 1 and 3, and L. Melazzo for the remaining parts.
As regards the evidence that the authors took into account, the following
critical editions were used: Klingner (1970) and Mazzarino (1982), Kellens
and Pirart (1989), Carruba (1966) and Hoffmann (1984), Monro and Allen
(1902-1920 3 ), Aufrecht (1968) and Kent (1953).
In the examples taken from Carruba (1966) and Hoffmann (1984) we have
used continuous transcriptions instead of syllabic transcriptions, as used by the
authors.
The anonymous reviewer raises the objection that the examples we cited
were taken from poetic texts. Apart from the fact that this is not true, we would
like to state that although poetry may exhibit some unusual constructions,
these never actually break the rules of a given language. There may be differ-
ences of register between prose and poetry, but not of syntax. Though poetry
may contain constructions no longer in current usage, these nevertheless be-
long to the syntax of that language. In any case, our examples exhibit rules that
are valid in both prose and poetic texts.
1. The graphic form ca transliterates the (Vedic) Sanskrit conjunction. The Iranian
equivalent is now transcribed cä with a long vowel due to an Iranian ortho-
graphic rule. For the sake of brevity we shall write ca for both the Vedic San-
skrit and Iranian occurrences of this enclitic conjunction.
A particular coordination structure of Indo-European flavour 511
2. It must be noted that the same explanation also holds good for disjunctive co-
ordination in those Indo-European languages which possess an enclitic dis-
junctive conjunction going back to the reconstructed form *ve.
The phrase category of nominal conjuncts is not specified any further here,
for we want to concentrate our attention on coordination structure. Only in (iv)
of note 8 will it be necessary to make use of precise structural hypotheses in-
volving the category of determiner to account for the peculiar order of the con-
stituents of the sentence. Thus in Indo-European languages the Ν Ρ can be
found in a complex structure in which there are various functional projections
dominated by a DP. We do not intend to assume this as a postulate on the basis
of the analyses conducted in modern languages, so we shall discuss the ques-
tion in a forthcoming paper.
3. The Hittite conjunction ij)a displays the same syntactic behaviour as the enclitic
conjunctions derived from *k" e in several languages having Indo-European as
their common ancestor, but it cannot be assimilated to them as far as etymology
is concerned. It is perhaps worth noticing that in ( l i b ) the conjunct to which
the enclitic form is attached is graphically represented by a Sumerogram.
4. In (15) the occurrence of te in second position might seem to be due to Wacker-
nagel's (1892) Law. Since it must be supposed that te has not undergone any
movement operations, the main question to solve is that of establishing the po-
sition of the participle opsomenai. If opsomenai has remained in situ since the
original formation of the sentence, then te attaches itself to this participial form
undoubtedly on account of phonological and/or prosodic reasons. If, however,
opsomenai is believed to have previously occupied the final position of the
sentence as expected in an SOV language like Ancient Greek, then the initial
position of this participle in the sentence has to be explained by assuming a
movement operation. Yet a left-adjunction of opsomenai to te has to be ex-
cluded, for such a head movement would be hindered by the presence of kai,
the tonic conjunction. Such being the case, the cliticization of te to opsomenai
will take place once all the syntactic operations have been performed, and con-
sequently it will again depend on phonological and/or prosodic reasons.
5. No sure case of inverse ca is testified in Old Persian texts, but see Schmitt
(1963).
6. The following citation from a Hittite text must be regarded as an interesting
case of multiple coordination which could contain a construction parallel to
Indo-Iranian inverse ca:
LÜMES
(i) DUMUme5-SU SESme5-SU gaenas=ses=sa
CHILDREN-HIS BROTHERS-HIS MEN-relatives-NOM-his-NOM-and-CONJ
LlJMEShassannas=sas Ü ERINmeS-SU
MEN-kin-GEN-his-GEN AND TROOP-HIS
"his children, his brothers and his relatives and the men of his kin (and)
his troop". (Telipinus I 3)
512 Emanuele Lanzetta and Lucio Melazzo
In ( i ) t h e l o g o g r a m s D U M U M E S - S U and S E S M E Ä - S U are c o o r d i n a t e d a s y n d e t i -
cally. They are followed by the conjunct LUMES gaenas=ses, to which the enclitic
conjunction (j)a in the assimilated form sa is attached. Finally, the conjuncts
LU hassannas=sas and ERIN MES -SU are joined by a conjunction denoted by
the logogram Ü. This sign is employed in Akkadian script to write the coordi-
nating conjunction, which is not enclitic in that language. A by no means trivial
question now arises. If the scribe is thought to have utilized the Akkadian word
with the intention of making use of a non-enclitic conjunction, then the type of
coordination in (i) merits consideration only on account of this unusual employ-
ment of a foreign form. Such being the case, it should be established whether
(i) would have a structure compatible with the antisymmetric analysis of coor-
dination. This does not seem to be the actual state of the matter. In point of fact,
providing the first two conjuncts are left out of consideration, as their structural
representation can be viewed as irrelevant at this moment, the remaining three
conjuncts should be imagined as having a structure of the following type:
0 0 Up, -JA [&p2 [χρ LU MES gaenas=ses] [ &P2 [ &Pl [ YP HJ MES hassannas=sas]
UP, Ü [zp ERIN m e S -SU]]]]]]
In (ii), where the adjunction of LU MES gaenas=ses to (j)a has been omitted, &P,
has the &P 2 as its complement. The head of this &P 2 has no phonetic content,
while Ü is the head of the &P 3 complement of the &P 2 . Now the question is
whether the covert head of the &P 2 can be licensed or not. For the time being,
there is room for doubt. As a matter of fact, (iii) shows that an analogous struc-
ture would be ungrammatical in languages like e.g. English (as to cases similar
to (ii) in Old Avestan, see Lanzetta 2 0 0 3 : section 2 . 0 ) :
(iii) *I saw both Mary, John and Luke.
If, however, on the basis of a perfectly plausible assumption the Akkado-
gramm Ü is held to hide the Hittite enclitic conjunction, then (i) will turn out
to contain an instance of inverse ca and therefore prove the existence of this
coordinate structure in that language. It is quite right to emphasize that on no
account can the Akkadogram Ü be the equivalent of the non-enclitic connective
nu, for this form occurs only at the beginning of the Hittite sentence.
7. In a more or less direct manner the phenomenon of inverse ca has been taken
into consideration by Zwolanek (1970: 80) and Dunkel (1982: notes 2, 38, 48).
Dressier (1965: 7 6 - 7 7 ) argued that the Greek and Latin enclitic conjunction
was not used in a way similar to that of the cognate Indo-Iranian conjunction.
8. The impossibility of a head movement operation shifting a non verbal element
to the left periphery of the sentence might seem to be belied by the following
quotation:
(iv) püro yäd asya sampinäk
fortresses-ACC when-CONJ he-GEN destroy-AOR-2SG
'when you destroyed his fortresses.' (RV, IV 30.13d)
A particular coordination structure of Indo-European flavour 513
The syntactic relationship between püro and asya could cause one to think that
the latter has remained in the position of either the specifier or the complement
of the phrase with the former as a head. No matter which of these two alterna-
tives is the right one, the operation by which püro has been shifted should be
depicted as a head movement, for a maximal projection cannot move and leave
its specifier or its complement (or both of them) behind. The sheer difficulty of
puzzling out the answer to the question that (iv) raises depends in truth on our
poor knowledge of the constituency relationship within ancient languages, es-
pecially in consideration of the diffusion of such a phenomenon as hyperbaton
in them (see Devine-Stephens (2000) for some careful examination of hyper-
baton in Ancient Greek). A way of regarding (iv) as an instance of topicaliza-
tion of a maximal projection consists in assuming an extended nominal structure
and a DP node in which the N P with püro as a head occupies the complement
position. Under these conditions, following Giusti (2001: 164-166), the demon-
strative pronoun asya can be generated in Spec,NP or in a specifier position
within that extended structure and then move into Spec,DP. If we are on the
right track, the topicalization of püro in (iv) implies either an N P or some func-
tional projection dominating the NP and dominated by the DP, and not a head.
Such a hypothesis is corroborated by Cinque's (1995: 298) serialization of
APs. As stated by this Italian scholar, the modifiers of an NP are arranged in a
set order, each of them filling the specifier position of different functional pro-
jections placed between the DP and the Ν P. As for nouns denoting objects, the
first modifier immediately to their left turns out to be a possessive. In (iv) the
genitive form asya corresponds to a possessive, and thus can be held to remain
in the specifier position of that FP between the DP and the N P that is adjacent
to the latter.
9. If so, we could think of a kind of mechanism of specifier-head agreement of
the type proposed by Kayne (1994: 143, note 2) to account for those cases in
which the &P has a distributive value.
10. (27b) can be compared to (v), where nearly the same terms occur and are
joined by ca:
(v) baga vazraka Auramazdä [...] haya uvaspa
god-NOM great-NOM Auramazda-NOM who-NOM good-horses-ACC
uraSäcä kunautiy
good-chariots-ACC-and-CONJ make-PRES-3SG
"Ahuramazdä (is) the great god [...] who makes good horses and good
chariots". (DSs 1 - 6 )
It must be noted that, even though the passage quoted in (v) is seriously dam-
aged, the string uradäcä can be easily read (see Kent 1953: 146). Nevertheless,
Lecoq (1997: 241) rounds off the passage in a different manner, spotting a
three membered coordination very similar to that of (vii) in note 11, except for
the order of the conjuncts.
514 Emanuele Lanzetta and Lucio Melazzo
11. A comparison can be made between (28) and the following sentence (vi),
where a three-membered coordination has ca as its end:
(vi) höi scantü manarjhä, uxöäis
they-NOM devote-themselves-AOR-iMP-3PL thought-iNSTR statements-iNSTR
siiaoSanäiscä xsnüm mazda
actions-lNSTR-and-CONJ choice-ACC Mazda-GEN
"let them devote themselves with thought, with statements and with (ritual)
actions to the choice of Mazda". (Y 53.2aa'b)
As can be easily seen, the last two conjuncts in (vi) would be identical to those
in (28), but for the negligible difference in number displayed by the noun
siiao9ana-. In both instances this noun happens to occur in the instrumental
Case but the number is singular in (28) and plural in (vi).
An Old Persian instance of three-membered asyndetic coordination similar
in meaning to (27b) and (v) of note 10 might be the following:
(vii) hausaiy xsagam fräbara [... ] taya
he-NOM-he-GEN kingdom-ACC bestow-iMPF.iND.3sG which-ACC
uraSaram uvaspam
possessed-of-good-charioteers-ACC possessed-of-good-horses-ACC
umartiyam
possessed-of-good-men-ACC
"he bestowed upon him this kingdom [...] possessed of good charioteers,
possessed of good horses (and) possessed of good men". (DSp 2 - 3 )
However, the text of (vii) is the result of some conjecture and supplementation.
The form ura&aram seems certainly readable to Kent (1953: 146), whereas
Lecoq (1997: 240) reads and interprets (vii) on the basis of his interpretation of
(v) of note 10.
12. Humbach (1991: 94, part I) considers the inverse ca of (29) to coordinate only
amaratas and hauruuatäs, for he regards utaiiuitT as a sociative instrumental
(see Humbach 1991:97; 112, parti).
13. The following sentence deserves some discussion:
(viii) brähmä ca giro dadhiri
p r a y e r s - N O M - a n d - C O N J songs-NOM b e - c o n c e n t r a t e d - P E R F - 3 P L
säm as min
together-ADV him-LOC
"the prayers and the songs are concentrated in him". (RV, VI 38.3c)
Klein (1985: 176) regards the use of ca in (viii) as potentially ambiguous. He
thinks that its interpretation remains uncertain, an inverse ca being recognizable
as well as a ca which coordinates (viii) to the previous sentence in the text. Let
us suppose that ca joins two sentences together. If so, the constituents brähmä
and giro would be coordinated asyndetically with the same grammatical pattern
that we already know. Moreover, the position of ca in (viii) could contradict
A particular coordination structure of Indo-European flavour 515
H a l e ' s (1987) view that when joining sentences ca and va (the latter is the dis-
junctive conjunction) are subject to Wackernagel's Law and reach their position
in the string only after the process of topicalization of a constituent has taken
place. A s a matter of fact, on no account do there seem to be the discourse
conditions that would allow one to suppose brähmä to have been topicalized. It
m u s t also be noted that (15) does not support H a l e ' s theory in any way because
the enclitic conjunction te definitely does not join sentences there.
14. In Old Avestan negative constituents are coordinated by nöit ... naedä. The
spelling naedä results f r o m a process of adjunction of the negative nöit to the
enclitic *u, w h o s e existence in the language of the sacred writings of Zoroas-
trianism can be inferred f r o m Vedic Sanskrit. The following passage appears
to prove that the adjunction of the negative to is operative even if this head
is phonetically unrealised:
Here the alternation of naedä and nöit, i.e. of coordination by the conjunction
*u and the asyndetic coordination, m a y hint at the m o v e m e n t of nöit to an
empty
15. A first attempt to tackle this problem has been very recently m a d e by Guardiano-
Longobardi (2003).
References
Aufrecht, W. T h e o d o r
1968 Die Hymnen des Rigveda. 2 Teile. Wiesbaden: Otto Harrassowitz.
Bianchi, Valentina
1999 Consequences of Antisymmetry. Headed Relative Clauses. Berlin/
N e w York: Mouton de Gruyter.
Boley, Jacqueline
1985 Hittite and Indo-European Place Word Syntax. Sprache 3 1 : 2 2 9 - 2 4 1 .
516 Emanuele Lanzetta and Lucio Melazzo
Carruba, Onofrio
1966 Das Beschwörungsritual für die Göttin Wisurijanza. Wiesbaden:
Otto Harrassowitz.
Chomsky, Noam
1964 Current Issues in Linguistic Theory. In The Structure of Language, J.
Fodor and J. Katz (eds.), 50-118. Englewood Cliffs, New Jersey:
Prentice Hall.
Cinque, Guglielmo
1995 Italian Syntax and Universal Grammar. Cambridge: Cambridge Uni-
versity Press.
Devine, A. M. and Laurence D. Stephens
2000 Discontinuous Syntax. Hyperbaton in Greek. New York/Oxford:
Oxford University Press.
Dressier, Wolfgang
1965 Inverses te, inverses -que. Glotta 43: 76-78.
Dunkel, George
1982 The Original Syntax of Conjunctive *-k w e' Sprache 28: 129-143.
Fodor Jerry A. and Jerrold J. Katz (eds.)
1964 The Structure of Language. Englewood Cliffs, New Jersey: Prentice
Hall.
Giusti, Giuliana
2001 The Birth of a Functional Category: From Latin ILLE to the Ro-
mance Article and Personal Pronoun. In Current Studies in Italian
Syntax. Essays Offered to Lorenzo Renzi, Guglielmo Cinque and
Giampaolo Salvi (eds.), 157-171. Amsterdam/London/New York/
Oxford / Paris / Shannon / Tokyo: Else vier.
Guardiano, Cristina and Giuseppe Longobardi
2003 Parametric Syntax as a Source of Historical-Comparative Generali-
zation. Unpublished Ms., Universities of Pisa and Trieste and CNRS
Paris.
Hale, Mark
1987 Notes on Wackernagel's Law in the Language of Rigveda. In Studies
in Memory of Warren Cowgill (1929-1985), Calvert Watkins (ed.),
38-50. Berlin/New York: Mouton de Gruyter.
1988 Old Persian Word Order. Indo-lranian Journal 31: 27-40.
Hoffmann, Inge
1984 Der Erlaß Telipinus. Heidelberg: Carl Winter Universitätsverlag.
Humbach, Helmut
1991 The Gäthäs of Zarathushtra parts 1-2. Heidelberg: Carl Winter Uni-
versitätsverlag.
Kayne, Richard S.
1994 The Antisymmetry of Syntax. Cambridge, Mass.: The MIT Press.
A particular coordination structure of Indo-European flavour 517
Munn, Alan
1992 A Null Operator Analysis of ATB Gaps. The Linguistic Review 9:
1-26.
Rizzi, Luigi
1997 The Fine Structure of the Left Periphery. In Elements of Grammar.
A Handbook of Generative Syntax, Liliane Haegeman (ed.), 281-337.
Dordrecht: Kluwer.
Roberts, Ian
1993 Verbs and Diachronic Syntax: a Comparative History of English and
French. Dordrecht: Kluwer.
Ross, John R.
1967 Constraints on Variables in Syntax. Cambridge, Mass.: M I T disser-
tation.
Schmitt, Rüdiger
1963 Ein altpersisches 'ghostword' und das sog. 'inverse ca'. Orientalia
32: 4 3 7 - 4 4 8 .
van Kemenade, Ans and Nigel Vincent
1997 Parameters of Morphosyntactic Change. Cambridge: Cambridge
University Press.
Wackernagel, Jacob
1892 Über ein Gesetz der indogermanischen Wortstellung. Indoger-
manische Forschungen 1: 333^136.
[1969 2 ] [Reprinted in Wackernagel, Jacob, Kleine Schriften. 2 Bände. Göt-
tingen: Vandenhoeck & Ruprecht, 1969 2 , 1. Band: 1-104],
Watkins, Calvert
1994 Indo-European *-k w e 'and' in Hittite. In C. Watkins: Selected Writ-
ings, Lisi Oliver (ed.), 300-306. Innsbruck: Wolfgang Meid.
Zwolanek, Renee
1970 'väyav indrasca'. Studien zu Aurufungsformen im Vedischen, Awes-
tischen und Griechischen. Münchener Studien zur Sprachwissen-
schaft 5: 51-64.
Index
accusativus cum infinitivo, accusative, bridging contexts, 265, 270, 271, 272
with infinitive, Acl, 4, 7, 8, 301, causative, causativisation, 193, 194,
342-367 195, 196, 198, 199, 201,214, 222,
active-stative language, 22, 24,279-284 223, 224, 225, 226, 227, 228, 231,
adjunction, 61, 81, 82, 92, 130, 153-154, 232,287
398,420,499-503, 515 causee, 194
agreement, 16, 17, 18,31-33,44-65,69, cislocative, 167, 183
81,87, 88,90,91,92, 118-122, 147, Classical Greek (see also Ancient
156, 239-241, 244, 245, 248-261, Greek), 5, 6, 7, 11, 265-293, 295-
278, 292, 296, 322, 323, 327, 329, 337
332,436 cleft, 15, 16, 105-133, 185
Akkadian, 2, 3, 5, 6, 20, 21, 22, 161, clitic, enclitic, 4, 5, 14, 16, 17, 20, 30,
189-236, 512 46, 54, 58, 68, 70, 72, 73, 74, 75, 87,
Aktionsart, 33, 90, 270, 277, 278, 279, 91,93, 120, 129, 132, 163, 169, 171,
281,282, 287, 288 172, 173, 177, 180, 183, 184, 185,
Ancient Greek (see also Classical 229, 256, 257, 260, 353, 386, 389,
Greek), 4, 8, 19, 24, 28, 29, 3 0 , 2 6 5 - 390, 400, 401, 404, 406, 409, 413,
293, 295-337, 342, 347, 353, 354, 430, 431, 432-434, 437, 483, 491,
365,457, 498,499, 511, 513 496, 497, 498, 499, 500, 506, 507,
anticipatory genitive, 9, 161, 162, 168, 509, 510, 511, 512, 515
179, 180, 182, 183 complementizer, 7, 8, 15, 16, 17, 47, 55,
antisymmetry of syntax, 349,424, 496 56, 57, 58, 59, 75, 76, 77, 78, 80, 88,
apophony, 225, 226, 227, 270, 277 89, 92, 126, 127, 243, 259, 304, 309,
Arabic, 22, 44, 45, 92, 105, 111, 114, 317,319,320, 321,331,347,352,
128, 129, 189, 191,205,214,221, 357, 359, 360, 364, 367, 382, 409,
228, 230, 231, 233, 234, 239, 241, 413, 443, 446, 448, 449, 450, 451,
246, 248, 250, 252, 259, 260, 261 452, 501
article, 4, 25, 140, 141, 143, 155, 353, concord, 139, 147, 155, 272, 282, 343
366, 367, 421 converter, 154
definite, 10, 115, 116, 141, 142, 143, circumstantial, 141, 155
145, 147, 148, 149, 150, 151,300, relative, 141
307,353,367 Coptic, Coptic Egyptian, 6, 15,29, 105-
indefinite, 10, 141, 143, 147, 150, 134, 139-159
151, 155 copula, 5, 16, 26, 116, 117, 118, 119,
possessive, 141, 142, 143, 150, 151 120, 121, 122, 127, 129, 130, 132,
asyndetic coordination, 496, 506, 507, 177, 179, 184, 185,422, 461,464,
508, 510, 514, 515 466, 467, 473
520 Index
corpus, Classical Sanskrit, 486 extraction, 55, 57, 63, 64, 91, 463, 471,
defmiteness, 22, 92, 110, 129, 141, 143, 472, 473,476, 477, 478, 479, 480,
147, 148, 151, 154, 241,421 481,484, 485,486,488, 489
demonstrative, 10, 89,115,116,129,141, focus movement, focussing,
142, 145, 150, 154,353,367,513 focalization, 12, 15, 105, 113, 123,
determiner, 9, 10, 11, 115, 140, 141, 125, 127, 269, 300, 392-397, 420
143, 146, 147, 148, 150, 156,356, focus, focussed, 12, 14, 15, 16, 17, 27,
357,367,461,511 32, 54, 68, 69, 72, 74, 84, 85, 87,
determiner phrase, 147 105-134, 185, 210, 351, 377, 382,
diachrony, 284, 406, 408 383, 389, 391, 392-397, 403, 406,
discontinuity, 405, 412, 413 410, 413,416, 417, 420, 424, 427,
distributional context, 268, 271, 272 433, 435,440, 441, 442, 444, 447,
double accusative, 342, 344, 365 448, 451,452
D-stem, 20,21, 189-236 contrastive, 13, 50,67, 107, 108, 109,
dual subcategorization, 22, 265, 267, 111,393,397,409,416,417
280, 281,284 formal features,
embedded clause, 15, 76,77,78, 80, 126, interpretable, 65, 81, 139, 151, 212,
243, 248, 259, 298, 317, 323, 359, 219, 220,265,280, 282, 283, 323
429,446, 448, 449, 451, 452,453 uninterpretable, 16, 51, 65, 81, 84,
enclitic conjunction, 496,497,498,499, 139, 146, 147, 148, 150,265,280,
500, 506, 5 0 9 , 5 1 0 , 5 1 1 , 5 1 2 , 5 1 5 284,287
event, (sub-)event, 6, 13, 16, 21,28, 31, frame, 14, 35, 360, 376, 429, 430, 447,
32, 34, 36, 37, 52, 54, 59, 65, 87, 148, 448,449,451,453,454
177, 191, 198, 200,208,209,212, French, Old, 420, 430, 432, 434
215, 218, 219, 220, 226, 227, 231, gemination, geminate, 190, 215, 223,
232, 236, 2 9 8 , 3 1 1 , 3 1 8 , 3 3 1 , 3 8 2 , 224, 225,226,235
403 gender, 15, 16, 17, 44, 45, 47, 54, 55,
eventive, 17, 18, 32, 33, 34, 43, 53, 71, 57, 87, 89, 91, 92, 118, 119, 120,
75, 77, 88 121, 122, 139, 140, 141, 146, 147,
eventual plural, 191,214,219,220,223, 148, 150, 161, 162, 167, 168, 169,
224, 225, 226 173, 177, 183,229, 278,350,353,
Exceptional Case Marking, ECM, 4, 7, 361,461,462
17, 46, 47, 70, 71, 82, 83, 303, 305, genitive case, 9, 10, 148, 149, 153,
306, 307, 309, 317, 332, 333, 346, 163, 169, 181, 182, 183, 266, 267,
347, 359 268, 272, 273, 275, 276, 282, 286
exclamatory infinitive, 358 G-stem, 21, 22, 189-234
Extended Projection Principle, EPP, Hebrew, 5, 6, 16, 18, 22, 25, 26, 44,
16, 18, 31, 64, 65, 69, 71, 72, 74, 76, 49, 54, 62, 65, 85, 92, 94, 95, 96,
81, 84, 86, 87, 88,239,240,251,252, 102, 129, 132, 143, 148, 153, 158,
253, 258, 259, 260, 335 189, 191, 205, 214, 221, 228, 230,
external argument, 16, 21, 60, 196, 197, 231,232,234, 236, 239-264
198,200, 201, 208, 209, 214, 225, historical infinitive, 358
229, 232, 233, 279, 281, 306, 356
Index 521
Hittite, 5, 19, 342, 496, 498, 505, 506, NumP, 11,21, 138, 146, 147, 149, 150,
507, 5 0 8 , 5 1 0 , 5 1 1 , 5 1 2 , 5 1 5 , 5 1 8 151, 153
Homeric Greek, 265, 277,279, 282, 367 object shift, 61, 66, 67, 68, 70, 73, 82
incorporation, 44, 46, 47, 55, 56, 58, Old Avestan, 505, 512, 515
90,91, 179,282,317,318,365 Old Persian, 505, 508, 511, 514, 516,
information 517
packaging, 459, 460, 468 Older Egyptian, 6, 16, 1 7 , 2 8 , 2 9 , 3 1 -
structure, 17, 29, 69, 72, 76, 84, 107, 101
161, 162,468 Osco-Umbrian, 342
intensive stems, 191 overlap, 269, 270
inverse ca, 496,499, 500, 501, 502, 503, passive, passivization, 4, 54, 57, 77, 78,
505,511,512,514,518 79, 80, 89, 90, 93, 189, 206, 208, 231,
Iranian, 342, 510 233, 297, 305, 306, 313, 331, 364,
iterative stems, 190, 213, 222, 232, 234 462
Latin, 4, 6 , 7 , 8, 11, 12, 13, 14, 15, 19, phrase structure, 28, 134, 140, 158, 382,
22, 24, 25, 26, 27, 28, 29, 30, 278, 457, 458, 460, 461, 463, 465, 469,
287,290,316,317,319,331,333, 473,486,487,490, 491,492
334, 335, 339-372, 373^126, 4 2 9 - Classical Sanskrit, 460
456, 457, 458, 496, 505, 506, 507, pied piping, 502
508,512,516 polysemy, polysemous, 49, 265, 267,
left dislocation, 387 269, 270, 286
left periphery, 9, 13, 15, 29, 88, 107, possession
108, 124, 128, 171,336,346,347, external, 9, 10, 161, 162, 168, 172,
350, 426, 429, 436, 445, 446, 454, 179, 180, 181, 183
481, 501, 512 inalienable, 9, 10, 179, 181
lexical root, 65, 285 internal, 181
lexicalization, 269, 270, 286 predication of, 152
LF lowering, 503 possession, possessed noun, 9, 10, 11,
LF raising, 503 32, 37, 53, 138, 141, 144, 146, 148,
light verb, 60, 278 149, 150, 151, 152, 153, 154, 156,
Linear Correspondence Axiom, 411,496 161, 172, 175, 177, 178, 180, 185,
nominative-accusative language, 24, 219
282,284,288, 365 possessive
nominativus cum infinitivo, construction, 9, 10, 145, 159, 179
nominative with infinitive, Ncl, expression, 142
301,302,362 morpheme, 149
null expletive, 50, 54, 92, 240, 241 possessor, 9, 10, 11, 37, 138, 141, 142,
number, 17, 44, 45, 49, 54, 55, 57, 87, 148, 149, 150, 151, 153, 154, 155,
91, 112, 118, 119, 120, 121, 122, 139, 156, 161, 162, 163, 168, 169, 171,
140, 141, 146, 147, 148, 150, 162, 172, 173, 175, 179, 180, 181, 184,
167, 168, 169, 173, 177, 178,210, 185
212, 214, 230, 253, 278, 461, 462, possessor, pronominal, 142, 144, 145,
488,514 151, 156
522 Index
PossP, 10, 11, 149, 150, 151 root, 8, 21, 22, 23, 32, 37, 43, 51, 59, 60,
predication, 89, 90, 109, 122, 123, 152, 65,69,73,78, 81, 82, 87,90, 91, 189,
221,414, 467 190, 202, 206, 224, 225, 229, 230,
pro-drop, 7, 46, 47, 48, 49, 50, 83, 88, 232, 235, 262, 265, 266, 267, 270,
260,462 271, 272, 277, 278, 284, 287, 358,
pronominal object shift, 66, 68, 82 359,360
pronoun, 8,9, 14, 15, 25,44,46,47, 50, Indo-European, 267
54, 55, 56, 58, 61, 67, 68, 70, 71, 75, Sanskrit, 5, 6, 11, 12, 26, 277, 292,457,
8 2 , 9 1 , 9 2 , 9 3 , 106, 107, 109, 110, 458, 459, 460, 461, 462, 463, 464,
111, 112, 114, 115, 116, 118, 120, 465, 467, 468, 469, 471, 472, 475,
121, 122, 129, 131, 143, 154, 155, 476, 477, 480, 483, 484, 485, 486,
156, 167, 171, 183,229,285,346, 487, 488, 489, 490, 491, 492, 493,
353, 357, 360, 361, 362, 364, 365, 501,510
367, 393, 430, 431, 432, 437, 451, Classical, 11,12, 457,458, 459,460,
478,495, 501, 513 461, 462, 463,465,467,469,471,
atonic, 432 472, 474, 475, 476, 477, 480, 484,
clitic, 4, 14, 93, 430 485,486,487,488,491
resumptive, 9, 15, 29, 56, 57, 58, 72, Vedic, 19, 29, 342, 461, 470, 489,
113, 122, 126, 127, 131, 155, 171, 490, 492, 499, 501, 503, 510, 515
430, 4 3 1 , 4 3 2 , 4 3 5 scrambling (see also word order, free),
stressed, 106, 454 398, 458,493
strong, 15, 433 Semitic, 6, 20, 26, 44, 94, 96, 102, 161,
weak, 14, 15, 29, 93, 419, 432, 433, 189, 1 9 2 , 2 0 5 , 2 1 8 , 2 2 7 , 2 2 8 , 2 2 9 ,
434, 436, 437, 438, 439, 440, 441, 231,233,235,236,239,256
451,452,453 situation aspect, 33
Proto-Indo-European, 265, 268, 277, small clause, 15, 16, 90, 107, 108, 122,
279, 286, 344, 357, 370 123, 124, 128, 260, 336, 356, 398
psych-verbs, 203, 204 S-stem, 189, 190, 191, 192, 193, 194,
quantifier, 70, 85, 256, 264, 438 195, 196, 198, 199, 200, 201, 223,
Quechua, 364 225, 226, 227, 228, 229, 231, 232,
relative clause, 9, 15, 16, 26, 27, 52, 55, 233,235
57,58, 105, 106, 107, 109, 111, 112, state, stative, 1, 16, 17, 22, 23, 32-52,
113, 114, 115, 116, 117, 118, 121, 75, 80-84, 88, 144, 153, 189, 203,
122, 123, 126, 127, 128, 129, 130, 204, 208, 209, 210, 215, 216, 217,
131, 134, 141, 152, 155, 169, 171, 219, 220, 221, 225, 231, 233, 268,
357, 364, 398, 420, 453,478 269, 270, 272, 273, 279, 281, 285,
Romance, 4, 13, 14, 25, 29, 102, 157, 286, 287
239,252, 259, 342, 353, 367, 369, structural case, 6, 23, 47, 144, 152, 273,
371, 372, 374, 409,423,426,431, 274, 283, 288, 304
432,433,436, 454, 456,516 subject-to-object raising, 345
Early, 371, 444 subordinate clause, 58,76,345,347,364,
Medieval, 13, 14, 429,430,433, 434 365,400, 401, 406,447,464, 478
Index 523