Exploring microtoponyms through linguistic and geographic

perspectives
Julia Villette Ross S. Purves
Department of Department of
Geography Geography
University of Zurich University of Zurich
Zürich, Switzerland Zürich, Switzerland
julia.villette@geo.uzh.ch ross.purves@geo.uzh.ch

Abstract

T oponyms are an important part of our cultural heritage and are thus more than names, they also encode history. Flurnamen, or
microtoponyms, are names given to natural features and they have been argued to be relatively closely linked to properties of landscape. In
this paper we analyse a large databases of microtoponyms using a mix of linguistic information and simple text analysis to explore patterns
of naming of natural features in the canton of St. Gallen in Switzerland. We observe that meaningful elements grounded in descriptive
cultural and natural readings of landscapes are common, and that usage of microtoponyms closely follows Zipf’s law.
Keywords: Flurnamen, ethnophysiography, microtoponyms, landscape

1 Introduction aims of the emerging field of ethnophysiography, as set out by

M ark & Turk (2003). Indeed, in this context, Burenhult &
Toponyms, or place names, have been the subject of study in a Levinson (2008, p. 136) posed the question ‘What is the
relation between landscape terms (common nouns) and place
range of fields including linguistics, geography, history and
political science (e.g. Burenhult & Levinson 2008; Zelinsky names (proper nouns)?’
1955; Derungs et al. 2013; Fagúndez & Izco 2016; Feng & Understanding patterns of toponym usage, and relationships
M ark 2017; Fagúndez & Izco 2016; Radding & Western, between, for example, generic landscape terms and place
2010). From a geographic perspective they offer, at first names requires some form of classification related to ways in
which names are given. Tent & Blair (2011), in a detailed
glance, a beguiling window to understanding the physical
environment and its development through time (Zelinsky review, explore many of the criteria which can be used in
1955). Linguistically, naming is clearly not arbitrary, since the naming, while Tyroller (1996) does so specifically in the
action of naming a place is a conscious one (Radding & context of microtoponyms. Tyroller makes a useful distinction
Western 2010). Naming also has a purpose, allowing us to between natural and culturally determined influences on
naming (for example Rotberg (Red M ountain) versus Rüti (A
distinguish one salient location from another, facilitating
communication and reducing ambiguity (Coates 2006). In this place cleared of trees)). We assume that microtoponyms are,
paper we concern ourselves with Flurnamen sensu Tent & Blair (2011, p. 85), more likely to be either
(microtoponyms), defined by Tyroller (1996) as toponyms descriptive (‘indicating an inherent characteristic of the
given to non-populated places such as fields, mountains, feature’) or associative (‘indicating something which is
always or often associated with the feature or its physical
forests and so on.
M icrotoponyms are particularly interesting for geographers, context’) and thus offer a relatively direct link to landscape.
since they refer to natural features, and thus may allow us to In this paper, in contrast to previous work, we start from a
understand ways in which landscapes are partitioned into spatially contiguous set of microtoponyms for the canton of St
meaningful elements, while also providing clues as to the Gallen in Switzerland, and analyse these with respect to a
detailed lexicon prepared for the same region by linguists. In
history of, for example, land use. However, microtoponyms,
like all toponyms can act as referents without conveying any doing so, we argue that we come closer to an interdisciplinary
meaning (Coates 2006). For example, over time language can study bridging the gap between linguistics and geography.
evolve, spelling may be normalised, events can be forgotten,
and landscapes and their usage can change making the link
between a name and its origins opaque. In linguistics, the field 2 Data and Methods
of onomastics explicitly seeks to etymologically disentangle
the original meanings of toponyms, typically by exploring 2.1 Study Area and Data
historical sources to find and explain the first documented
usage of a toponym as referent to a place. Geographical S tudy area. Our study area is the canton of St Gallen located
investigations of toponyms have often focussed on a few in north-eastern Switzerland which has an area of 1951km2.
concepts, captured through generic terms in compound Some 48% of the canton is used for agriculture, 32% is
toponyms, and explored, for instance, their spatial distribution forested and only around 10% urban. The topography ranges
(e.g. Zelinsky 1955; Derungs et al. 2013; Fagúndez & Izco from plains to high mountains (from 398m to 3247m) (Kanton
2016; Feng & M ark 2017). To date, few studies have sought of St Gallen 2017). The official language of the canton is
to combine deeper linguistic etymological studies with German, but the inhabitants speak a range of Swiss German
automated analysis of large numbers of toponyms in space. dialects, and historically the language of Romansh was also
M aking this link more explicit, and carrying out multi- and spoken and has influenced toponyms.
interdisciplinary research on toponyms is one of the explicit

1
 AGILE 2018 – Lund, June 12-15, 2018

Microtoponyms of St Gallen. M icrotoponym data were underlying number of occurrences of each term. The 26 most
provided by the canton itself and consist of names assigned to commonly occurring meaningful elements, used as a basis for
individual polygons for the whole canton. These names are the exploration of collocation, included different sorts of
collected in a bottom-up process by individual communes, and terms such as: natural features (e.g. Berg, Wald), cultural
provided to the cantonal authorities for a variety of purposes. features (e.g. Hof (farmyard), Dorf (village), adjectives (e.g.
Importantly, the data are spatially contiguous, meaning that lang, neu (new)), spatially modifying terms (e.g. ober, hinder
microtoponyms are assigned even to land parcels in urban (behind)) and animals (e.g. Geiss (goat)).
areas. A total of 17598 individual names are contained in the
dataset, of which some 54% (9489) are unique. On average,
parcels associated with microtoponyms have an area of 3 Results and Discussion
11.2±2.3 hectares.
Lexicon of meaningful elements. To analyse the After analysing the microtoponyms using the lexicon, we
microtoponyms data we used a lexicon of meaningful found that 15153 (86%) of microtoponyms contained at least
elements. This lexicon was created by linguists tracing the one lexicon term. These were matched by 1409 different
etymology of individual toponyms and their components. lexicon terms. Thus, we can surmise that meaningful
Terms included range from generic landscape terms (e.g. Berg elements, which are interpretable in terms of landscape
(mountain) or Wald (forest), adjectives (e.g. lang (long) or rot properties (e.g. used in descriptive and associative ways Tent
(red)) and spatially modifying terms which Leino (2005) & Blair (2011)) are indeed common in St Gallen’s
postulated were used in inductive toponyms (e.g. ober (upper) microtoponyms. Furthermore, microtoponym uniqueness is
whose usage implies another similar toponym (e.g. Upper the result of combinations of meaningful elements, rather than
Town implies a nearby place named simply Town or Lower the use of a single term in isolation (though indeed the most
Town)). The lexicon consists of 3378 meaningful elements, common microtoponyms are meaningful elements used in
with links allowing meaningful elements with the same isolation (compare Figure 1 and Figure 2)).
meaning to be analysed together (e.g. Berg and Bärg are
alternative spellings or forms with the same roots). It is Figure 1: Frequency against rank of microtoponyms.
important to note that an individual microtoponym may
contain no, one or more than one of these meaningful
elements. Furthermore, the lexicon is not exhaustive, meaning
that not all toponyms or parts thereof are contained.

2.2 Analysing microtoponyms using a lexicon of

meaningful elements
Our analysis of microtoponyms data focussed on using the
lexicon of meaningful elements to, firstly, explore which
meaningful elements were most commonly used and,
secondly, explore with what other terms the most common
meaningful elements were associated. Thirdly, by performing
a frequency analysis using the lexicon, we were also able to
identify microtoponyms with no meaningful elements
currently contained in the lexicon. Our basic approach to Figure 2: Frequency against rank of meaningful elements.
frequency analysis used simple string matching. However, we
first removed all two letter strings from the lexicon, since
these led to a high proportion of false positives.
Our matching process was iterative, and since the
same microtoponyms could match onto multiple meaningful
elements, we first sorted meaningful elements by length such
that the longest possible matching meaningful element from
the lexicon was matched onto a microtoponym. Having found
a match, the matching part of a microtoponym was deleted,
and the process repeated allowing further matches to be made.
This is especially important since German contains many
compound nouns (e.g. Rotberg would match both rot and
Berg).
To explore the relationships between meaningful elements, To illustrate this result we show in Figures 1 and 2 plots of
we used collocation frequencies to identify combinations of meaningful element frequency and microtoponym frequency
meaningful elements occurring together more or less than one as a function of rank on a logarithmic scale. Both plots are
would expect by chance. We therefore compute χ-squared relatively linear, thus following (more or less) Zipf’s law.
values for collocated frequent meaningful elements. χ-squared However, the relationship is clearly stronger for
essentially measures whether or not co-occurrence is more or microtoponyms (r2=0.99) than meaningful elements (r2=0.94)
less than we would expect by random chance, given the

2
 AGILE 2018 – Lund, June 12-15, 2018

and we postulate that this has two potential reasons. Firstly, meaningful choices in naming. For example Buech (beech)
microtoponyms follow general patterns of language, which relates a particular type of tree to Wald (forest) or Holz
have been shown to follow a power law in many examples (timber) which is particularly common in the area. Egg (edge)
(c.f. Zipf 1935; Piantadosi 2014). Secondly, as we have is associated with length (lang) and Dorf (village) with neu
shown, many microtoponyms use meaningful elements, which (new) and a range of spatial qualifiers (under, hinter, vorder,
in turn reflect landscape elements and their properties. These ober). It is also worth noting that Dorf would typically be
meaningful elements however, are not likely to be distributed associated with settlement names, which are not considered by
according to a power law (i.e. we have no reason to assume linguists to be microtoponyms, however, the example allows
there are twice as many mountains as forests in St Gallen), but us to show the utility of our approach well.
rather according to the inherent properties of the landscape. Underrepresented co-occurrences are also instructive. Thus,
Thus, though microtoponyms follow broadly properties of Feld (field) is specifically not chosen with Berg, Wald or Holz
language, the lexicon of meaningful elements is constrained presumably because these features exclude one another.
by the properties of landscape, in other words as put by Furthermore, Feld is also note associated with Wis (meadow)
Piantadosi (2014, p. 9): ‘The lexicon did not have much in this case more likely because these are related categories.
freedom in how it labelled the terms in these categories since
the referents of these terms are salient, fixed natural binds.’
In Figures 1 and 2 we mostly concentrate on the frequencies 4 Conclusions
of microtoponyms and meaningful elements, without paying
much attention to their semantics. In Figure 3, we show a In this paper we set out to show how a multidisciplinary
wordcloud of the 200 more frequent meaningful elements approach could allow us to explore toponyms from both a
according to frequency. geographic and linguistic perspective. However, it is also
important to set out some key limitations. Firstly, we analyse
Figure 3: 200 most frequent meaningful elements. microtoponyms in the aggregate, that is to say our approach
does not include any examination of the etymologies of
individual usages. Thus, although we base our analysis on a
detailed, linguistic lexicon, our methods are relatively crude
and rely on their application to a relatively large dataset. This
in turn means we are limited to exploring, at least
semantically, relatively common toponym occurences.
Secondly, we analysed a contiguous dataset for the canton of
St. Gallen. These data include settlement areas, something
which would not typically be the case in data collected by
linguists. Thirdly, our results are specific to a particular region
and its landscape, and thus the occurrence of particular
meaningful elements could be argued to be trivial.
The ten most frequent terms are a mix of natural (Berg However, we believe that there are also some important and
(mountain), Wald (forest), Egg (edge), Tobel (ravine), Bode useful results in this work, which will form the basis for our
(level place on or between slopes), cultural (Wis (meadow), ongoing research. Firstly, we were able to show that
Weid (pasture), Hof (farmyard), Holz (timber) and one term microtoponyms in St. Gallen are dominated by meaningful
used in inductive toponyms (ober (upper)). Indeed, in general elements, and that these meaningful elements are combined to
this list is composed of many terms which appear to be related produced microtoponyms which broadly follow Zipf’s law,
to the physical properties of the canton of St Gallen, again even though the landscape producing them (and thus the
suggesting the strong link between the microtoponyms and the supply of concepts used in their production) probably does
physical environment. Note that a small number of terms in not. In future work we will test this hypothesis by more
fact refer to, for example, person names (e.g. nau or gar). directly linking geographic properties to microtoponyms, and
Their appearance in the word cloud is most likely the result of exploring their distributions.
both their misclassification in the lexicon and erroneous
matching on very short terms in our matching process, and Acknowledgments
demonstrates one of the limitations of our approach.
Furthermore, some terms are associated with a high degree of The first author would like to thank Olga Chesnokova and
semantic ambiguity – thus val, a Romansh term can refer to Pawel Biernacki for their help getting started with Python. We
valley, forest or a ditch. would also like to thank Linda Steiner, Stefan Wuerth and
Figure 4 illustrates our final result, co-occurrence of a Elvira Glaser for access and explanations of the lexicon, and
selection of meaningful elements in our microtoponyms list. valuable discussions on this topic. This research is part of a
Importantly, we illustrate not only raw co-occurrence counts project ‘Local Toponyms of the Canton of St. Gallen’ funded
which are influenced in turn by the overall frequency of by the Swiss National Science Foundation and the Canton of
occurrence of a term, but also the statistical significance of the St. Gallen (SNF project no. 100015_162591). We also
co-occurrence according to a χ-squared test. Thus, all bold acknowledge the support of the university research priority
values in Figure 4 occur significantly more than would be programme language and space.
expected by chance. An examination of these co-occurrences
quickly demonstrates that they are also semantically

3
 AGILE 2018 – Lund, June 12-15, 2018

Figure 4: Co-occurrence counts for selected meaningful elements. Counts marked ** occur significantly more (p < 0.01)
than expected by chance, *- and **- significantly less (p < 0.05 and p < 0.01 respectively).

References
Piantadosi, S.T. (2014) Zipf’s word frequency law in natural
Burenhult, N. & Levinson, S.C. (2008) Language and language: A critical review and future directions.
landscape: a cross-linguistic perspective. Language Sciences. Psychonomic bulletin & review. 21 (5), 1112–1130.
[Online] 30 (2–3), 135–150. Available from:
doi:10.1016/j.langsci.2006.12.028. Radding, L. & Western, J. (2010) What’s in a name?
Linguistics, geography, and toponyms. Geographical Review.
Coates, R. (2006) Properhood. Language. 356–382. 100 (3), 394–412.

Derungs, C., Wartmann, F., Purves, R.S. & M ark, D.M . Tent, J. & Blair, D. (2011) M otivations for naming: The
(2013) The meanings of the generic parts of toponyms: use development of a toponymic typology for Australian
and limitations of gazetteers in studies of landscape terms. In: placenames. Names. 59 (2), 67–89.
International Conference on Spatial Information Theory.
2013 Springer, 261–278. Tyroller, H. (1996) Typologie der Flurnamen
(M ikrotoponomastik): Germanisch. Eichler et alii. 1434–
Fagúndez, J. & Izco, J. (2016) Spatial analysis of heath 1441.
toponymy in relation to present-day heathland distribution.
International Journal of Geographical Information Science. Zelinsky, W. (1955) Some Problems in the distribution of
30 (1), 51–60. generic terms in the Place- names of the northeastern United
States. Annals of the Association of American Geographers.
Feng, C.-C. & M ark, D.M . (2017) Cross-Linguistic Research 45 (4), 319–349.
on Landscape Categories Using GEOnet Names Server Data:
A Case Study for Indonesia and M alaysia. The Professional Zipf, G.K. (1935) The psycho-biology of language. Oxford,
Geographer. 1–12. Houghton, M ifflin.

Leino, A. (2005) In search of naming patterns: A survey of Kanton of St Gallen. (2017) Kopf und Zahl [Online] Available
Finnish lake names. Denominando il mondo. Dal nome from:
comune al nome proprio/Naming the world. From Common https://www.statistik.sg.ch/home/publikationen/ksgmz/2017.h
Nouns to Proper Names. 355–367. tml [Accessed 20th January 2018].

M ark, D.M . & Turk, A.G. (2003) Ethnophysiography. In:

Workshop on Spatial and Geographic Ontologies, 23
September 2003. Kartause Ittingen, Switzerland.