ACTA UNIVERSITATIS UPSALIENSIS Studia Linguistica Upsaliensia 8 1

ACTA UNIVERSITATIS UPSALIENSIS Studia Linguistica Upsaliensia 8 1
Studia Linguistica Upsaliensia
Proceedings of the 23rd Scandinavian Conference
of Linguistics
Uppsala University
1 – 3 October 2008
Edited by Anju Saxena & Åke Viberg
© The authors 2009
Grafisk bearbetning: Textgruppen i Uppsala AB
Tryck: Edita Västra Aros, Västerås 2009
ISBN 978-91-554-7594-9
ISSN 1652-1366
Electronical version available at:
Dorothee Beermann and Pavel Mihaylov
TypeCraft – Glossing and databasing for linguists . . . . . . . . . . . . . . . . . . . . . . . . . .
Karl Erland Gadelii
Fusional verb morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Elena Gorishneva
ONE: Between numeral, indefinite marker and intensifier . . . . . . . . . . . . . . . . . . . .
Lutz Gunkel & Susan Schlotthauer
Attribution in Basque, Finnish, Hungarian and Turkish: Morphology vs. Syntax . .
Shinji Ido
An analysis of the formation of the Tajik vowel system . . . . . . . . . . . . . . . . . . . . . .
Leonid Kulikov
Valency-changing categories in Indo-Aryan and Indo-European: A diachronic
typological portrait of Vedic Sanskrit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Anju Saxena, Beáta Megyesi, Éva Csató Johanson & Bengt Dahlqvist
Using parallel corpora in teaching & research: The Swedish-Hindi-English &
Swedish-Turkish-English parallel corpora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Luying Wang
On the Grammaticalization of Mandarin aspect markers . . . . . . . . . . . . . . . . . . . . . 102
Torbjörn Westerlund
The basic case marking of Ngarla, a language of Western Australia . . . . . . . . . . . . 115
Toshiko Yamaguchi
The causative/ inchoative alternation in Icelandic . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Elena Buja
Sociolinguistic aspects of bilingualism among the Moldovan students studying
in Romania . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Angela Falk
Narrative patterns in monolingual and bilingual life-history conversations . . . . . . . 159
Makiko Fukuda
Castilian or Catalan? Linguistic survival strategies of Japanese residents in
Catalonia, Spain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
Christine Johansson and Christer Geisler
The Uppsala Learner English Corpus: A new corpus of Swedish high school
students’ writing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Indira Y. Junghare
Syntactic convergence: Marathi and Dravidian . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Katri Karjalainen
Using communication strategies to gain fluency, accuracy and complexity
in L2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
Sanita Lazdia & Heiko F. Marten
The “Linguistic Landscape” method as a tool in research and education of
multilingualism: Experiences from a Project in the Baltic States . . . . . . . . . . . . . 212
Harry Lönnroth
The multilingual history of an industrial society. The case of Tampere, Finland . . . 226
Magomedkhan Magomedkhanov
Linguistic assimilation and the weakening of ethnic identities in Dagestan . . . . . . . 239
Theodore Markopoulos
Medieval Mediterranean as a multilingual area: the Greek perspective . . . . . . . . . . 245
Elena Nikishina
Language use in Moscow schools with an ethno-cultural component (based on
schools with the Armenian and the Azeri ethno-cultural component) . . . . . . . . . 258
Stefano Rastelli
Lexical Aspect too is learned: data from Italian Learner Corpora . . . . . . . . . . . . . . 272
Paula Rossi
Language changes and language contacts in a 19th century Maritime College and
Commercial College . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
Misuzu Shimotori
Conceptual contrast of dimensional adjectives in Japanese and Swedish:
Exploring the mental lexicon by word-association test . . . . . . . . . . . . . . . . . . . . 296
Bettina Zeisler
Mainstream linguistics for minor(ity) languages? Or: What is it like to speak
Ladakhi? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
Workshop on Readability and Multilingualism
Sofie Johansson Kokkinakis
Workshop on Readability and Multilingualism . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
The 23rd Scandinavian Conference of Linguistics (SCL 23) was held at Uppsala
University 1–3 October 2008 on behalf of the Nordic Association of Linguists
The theme of the conference was Multilingualism, which was intended to subsume cross-linguistic typological studies, linguistic variation and language
change in contact situations as well as studies relating to bilingualism and to second and foreign language learning. In addition, presentations from other areas of
linguistics were welcome.
The conference featured plenary lectures by two invited speakers:
Michael Noonan (University of Wisconsin)
What do we mean by the genetic relatedness of languages?
Anna Siewierska (Lancaster University)
The impersonal-to-passive highway: an instance of bidirectional change
In addition to the general sessions, the conference included four workshops:
• Language Documentation and Language Description (Conveners: Éva Csató
Johanson, Anju Saxena & Åke Viberg)
• Language Change in Bilingual Communities. Focus on the Post-Soviet Countries and their Immigrant Communities Elsewhere (Conveners: Nino
Amaridze, Anne Tamm, Manana Topadze, Inge Zwitserlood)
• New voices – New visions, Swedish Minority Language Policies in Transition
(Convener: Leena Huss)
• Readability and Multilingualism (Convener: Sofie Johansson Kokkinakis)
The workshop entitled Workshop on language documentation and description of
lesser-known languages started as a pre-conference workshop 30 September.
This volume includes studies covering a wide spectrum of approaches to linguistics. For the sake of simplicity, the papers from the general sessions have
been divided into two sections. A general section and a section called multilingualism which includes both studies concerned with multilingualism as a social
and political phenomenon as well as studies concerned with bilingualism and
second language acquisition from a developmental perspective. Several of the
contributions to the general section are concerned with language typology and
areal linguistics which are closely related to the general theme of multilingualism. The volume ends with a special section devoted to one of the workshops:
Readability and Multilingualism.
In February this year, we were reached by the news of the unexpected death
of our colleague and friend Michael Noonan. Mickey will be remembered for his
wide-ranging interests in all things linguistic, for his great sense of humour and
for being such an eminently likeable person. We miss him deeply.
Uppsala in September 2009
Anju Saxena & Åke Viberg
TypeCraft –
Glossing and Databasing for Linguists
Dorothee Beermann and Pavel Mihaylov
Norwegian University of Science and Technology, Trondheim, Norway
Ontotext Lab, Sirma Group, Sofia, Bulgaria
1 Introduction
In a manuscript from 1987, William Labov questions the relation between quantitative and qualitative methods in linguistics:
“…, the number, variety and complexity of linguistic relations are very great, and
it is not likely that a large proportion can be investigated by quantitative means. At
present, we do not know the correct balance between the two modes of analysis:
how far we can go with unsupported qualitative analysis based on introspection,
before the proposals must be confirmed by quantitative studies based on observation and experiment.”
William Labov 1987
In this paper we will address Labov’s question indirectly by looking at the role
that interlinear glossed text plays in linguistic publications. We then will introduce an online ‘glosser’ and knowledge sharing tool called TypeCraft, which allows to generate, preserve and distribute linguistic data in a neat way.
In a short online article on ‘Description in Linguistic Theory’ Magnus (1995)
states that surface grammar is not a thing, but a methodology. Here we will argue
that symbolic rewriting, and, as such, interlinear glossing, leads to a representation that we (at least on one interpretation of the term) can call a surface grammar. Glossing, as it is used today, however, needs to change drastically to generate what one truly can call a linguistic representation.1 One reason is that standardization is still at its beginning, 2 more fundamentally, however, it is simply
unclear what it really would mean for the individual researcher to represent linguistic material in such a way that it serves not only in his own research, but
equally is usable as an independent resource for linguists in general.
See also Lehmann (1999) for a similar view.
But for work towards standardization within a typological context, see for example work in Canonical
Typology by Corbett and The Surrey Morphological Group (
In this paper we will to focus on the role of glossing in the publication of linguistic research. We will discuss ways in which linguists can generate and administer their own linguistic resources as part of a collaborative effort which has
the goal to make multi-lingual in-depth annotated text-based natural language
examples public.
2 Background Information
The role that digital tools play in all fields of modern linguistics can not be underestimated. This is partially due to the success of computational linguistics.
Crucially however this reflects developments in IT and in particular the success of the World Wide Web which has created new standards also for linguistic research. Through the INTERNET our perception of ‘data’ and publication
of linguistic results has changed drastically only in a matter of a few years. Yet,
the situation Labov described in 1987 still captures the situation within descriptive and theoretical linguistics, since also today little agreement has been
reached about the role that empirical data should play in research. Although
several multi-lingual databases3 are open for linguistic research, it is often not
easy to find a way to access data relevant to ones own research. This might be
because it proves to be too costly to find the data that represents the paradigms
that one is interested in, or data access requires a person experienced in the use
of online databases. Not surprisingly so, linguistic examples found in the linguistic literature are often the result of a long citation chain which not in all
cases can be reconstructed. In addition original annotations tend to be partial
or even misleading.4 In short linguistic data is spare, at least for linguistics not
directly involved in field work and/or not connected to networks managing
some sort of data resource. In addition, available data, for example those found
in publications, is hard to validate. In spite of these difficulties it seems evident that the general availability of linguistic material from a larger sample of
languages is not only desirable for individual research, but also could influence
how we think about language, and the way we go about linguistic research in
general. If you are only a few mouse clicks away from showing that a certain
generalization only holds for a limited set of languages, but truly fails to describe a given phenomenon for a wider sample, statements claiming linguistic
generality must be phrased much more carefully. Yet, it is well known that the
preparation of natural language samples enriched by linguistic information
through glosses is a time consuming enterprise, quite independently of the
form that the raw material has and the tools we have chosen to document our
results. Well known are problems connected to the generation and storage of
See the Language Typology Resource Center at
See Beermann (2009).
linguistic data in the form of standard electronic documents or spread sheets
(Bird and Simons 2003). In addition electronic data should be archived and be
made portable so that it can be migrated to different applications and storage
facilities.5 Yet for the individual researcher it is not easy to decide which of the
available tools serves his purpose best. Research is a process and it is therefore
not always clear from the on-set which categories are needed and in which
form the material should be organized and stored. But perhaps even more importantly most tools turn out to be so complex that the goal of mastering them
becomes an issue in its own right. As a result tool management develops into
an independent issue, taking away time and resources from the original task at
hand – the linguistic analysis.
To sum up: on the one hand the linguistic field experiences an uncertainty
about the role of primary data in linguistic research, and on the other hand
there is a fast and often confusing development of digital tools for language
documentation. It therefore appears adequate to make it a linguistic goal in itself to (a) work towards a more general and more straightforward access to linguistic resources, (b) encourage the systematic generation of linguistic data
beyond what emerges from fieldwork and other descriptive studies, and (c) advocate the generation of a common data pool for linguistic research.
In the following we would like to introduce and inform about a linguistic tool
for text annotation called TypeCraft, which we have created through combining
several well-understood tools of knowledge-management. It makes manual text
annotation an easier task by adding efficiency to annotation tasks. Example sentences can be directly imported into research papers. Another important feature
that should be mentioned is that TypeCraft is also a collaboration and knowledge
sharing tool, which sets it aside from other annotation tools. Safe storage and other functionalities connected to digital language documentation will be discussed
in the following.
3 An overview
TypeCraft is a database for natural language texts combined with a tabular text
editor for interlinear glossing, wrapped into a wiki which is used as a collaborative tool and for online publication. Interlinear morphological glossing is a
time consuming and linguistically challenging tasks, and in order for morphological glossing to follow certain standards, it is important that glossing can be
performed efficiently and that the results of prior annotation can be accessed
at any time for further use. Table 1 summarizes TypeCraft’s main functionalities:
For more information on this point see for example
Table 1
Overview over TypeCraft Functionalities
Data Migration
tabular interface for word
level glossing – automatic
sentence break-up
individual work spaces for
users that would like to keep
data private
manual import of text and
individual sentences,
drop down reference list of
linguistic symbols
data sharing for predefined
groups such a research
collaborations research
export of annotated sentence
tokens (individual tokens or
sets) to Microsoft Word, Open
Office and LaTEX
word and morpheme deletion
and insertion
data export from the
TypeCraft database to the
TypeCraft wiki
export of XML (embedded
DTD) for further processing
of data
lazy annotation mode
(sentence parsing)
access to tag sets and help
pages from the TypeCraft wiki
customized sets of sentence
level tags for the annotation of
construction level properties
access to information laid out
by other annotators or
4 Glossing
As pointed out by Lehmann (2004), the use of interlinearized glossing in the representation of primary data became a standard for linguistic publications as late as
in the 1980s, where glossing of sample sentences started to be required for all languages except for those coming from English. However, the use of examples in
written research is not accompanied by a common understanding of its function. It
seems that glosses, when occurring in publications, are seen by most linguists as a
convenience to the reader. Quite commonly, information essential to the understanding of examples is given in surrounding prose, and without any appropriate
reflection of salient linguistic properties in the glosses themselves. In this way language examples of potential interest to research in general become tied to individual research papers, and thus are at risk to become lost to future research.
Let us look at a glossed example taken pretty much at random from the online
database Odin at Fresno State University.6 The Odin database is a repository of
interlinear glossed text which has been extracted mainly from linguistic papers.
The database contains data for a variety of the world’s languages, and all data
contained in the repository has been subject to linguistic analysis. The user interface of the database shows a list of URLs ordered by language. This is exemplified in Table 2. The right hand side fields give the user some idea about the quality of the data and indicate different data formats. ‘Highest’ and ‘Yes’ in this context means that the examples have been manually verified and that one can view
the examples in isolation by clicking on ‘Yes’.
For URL see references.
Table 2 User interface of the Odin database at Fresno State University
We extracted from Odin two examples from a paper on serial verb constructions
in Ewe by Agbedor (1994). The examples below serve to illustrate some of the
problems connected to standard glossing:
(1) Kofi e awua le ka dzi da e x me
K. remove shirt on rope top put LOC. room in
‘Kofi removed the shirt from the line and put it in the room’
(2) Kofi f agbaleawo le xa me da e gota
K. collect book-the-PL. in room put LOC. outside
‘Kofi collected the books from the room and put them outside’
As in (1) and (2) interlinear glosses mainly consist of translational glosses with
some occasional indication of morpho-functional properties or part of speech information. For the sequence le xa me da e in (2) it appears as if le, me and e
express a locative meaning such as ‘on’ or ‘in’ as in le ka dzi ‘on top of the rope’
(see (1)) and x me ‘in room’7(see (1)). Yet neither the category of the lexemes
nor their meaning can be deduced from the examples. Since Ewe is a verb serialisation language one might wonder if these elements are perhaps verbs, yet also
a nominal category or a preposition seems possible.8 Purely translational glosses,
perhaps adequate for text strings which serve as mere illustration, are insufficient
to transform raw material into linguistic data, which to our understanding should
be all paradigms in a publication that are crucial for the evaluation of the theoretical development reported on.
5 Glossing with TypeCraft
Typecraft is an interlinearized ‘glosser’ designed for the annotation of natural
language phrases and small corpora. Figure 1 displays the home page of TypeCraft which serves as an access point to the TypeCraft database and main page
of the TypeCraft wiki (TCwiki). Notice the navigation bar on the left in Figure1.
From here the user can (after login) enter My Texts which is the user’s personal
interface to the TC database and from which he enters into annotation mode.
xa ‘room’ x-a ‘room-DEF’
The role of le in le xa me has not become clear to us.
Figure 1 TypeCraft Main page at
Figure 2 shows the interface for a user that not only has his own data (Own texts),
but also shares data with other users (Shared Texts). At present sharing of text is
a feature set by the database administrator, but in the near future the user will be
able to choose from the TypeCraft user list the people with whom he wants to
share data. Note that data is stored as texts which consist of annotated tokens,
generally sentences. Importantly, Own texts hosts data that is ‘private’, that is, it
can only be seen and changed by its owner. Sharing of data and publish data
online are additional features to which we we will briefly come back later.
Different from Toolbox,9 which is a linguistic data management system,
TypeCraft is a relational database and therefore by nature has many advantages
over file based systems like Toolbox. This concerns both data integrity and data
migration. In addition databases in general offer a greater flexibility for data
search. The other mayor difference between Toolbox and TypeCraft is that
TypeCraft is an online system, which brings many advantages, but also some
disadvantages. An online database is a multi-user system, that is, many people
can access the same data at the same time independent of were they physically
are. Distributive tools represent a distinct advantage for research where collaboration between researchers worldwide is essential. TypeCraft is designed to allow data sharing and collaboration during the process of annotation and for the
presentation and discussion of research results. Yet in spite of the many advantages that a web-based application such as TypeCraft has, to be web-based has a
disadvantages, since not all users work with a stable INTERNET connection; in
particular for work in the field, TypeCraft is at this point not suitable.
Figure 2 below illustrates that different scripts can be easily represented in
TypeCraft, which uses Unicode. Every script that the user can produce on his
PC can be entered into the browser, which for TypeCraft must be Mozilla Fire9
Toolbox can be found at
Figure 2 TypeCraft interface for the administration of personal and shared data
fox. Different from Toolbox, TypeCraft insists on glossing standards as for examples advocated by GOLD10 or given by the Leipzig Convention distributed by
the Max Planck Institute for Evaluationary Anthropology.
TypeCraft supports word-to-word glossing on eight tiers as shown in Figure
3. After having imported a text and run it through the sentence splitter (a process
we will not describe here), the user can select via mouse click one of the phrases
and enter the annotation mode. The system prompts the user for the Lazy Annotation Mode ( in Toolbox called sentence parsing) which will automatically insert (on a first choice basis) the annotation of already known words into the annotation table. TypeCraft distinguishes between translational, functional and
part-of-speech glosses. They are visible to the annotator as distinct tiers called
Meaning, Gloss and POS.
Every TypeCraft phrase, which can be either a linguistic phrase or a sentence,
is accompanied by a free translation. In addition the specification of construction
parameters is possible, a feature that we will not further describe here. Under annotation the user has access to a drop-down menu, showing standard annotation
symbols. These symbols together with short explanations can also be accessed
from the TypeCraft wiki, so that they can be kept open in tabs during annotation.
In Figure 3 we also see the effect of ‘mousing over’ symbols, which displays
their ‘long-names’. Some symbols have been classified. In Figure 3 we see for
example that the feature past is a subtype of the feature TENSE. This classification will in the future also inform search. Further features of the annotation in10
Figure 3 TypeCraft ‘Glosser’
terface that we cannot describe here are the easy representation of non-Latin
scripts, deletion and insertion of words and morphemes during annotation, the
accessibility of several phrases under annotation and the grouping of tokens into
6 Data Migration
Export of data to the main text editors is one of the central functions of TypeCraft. TC tokens can be exported to Word, Open Office and LaTeX. This will
allow the user to integrate datasets relative directly in his research papers. Export can be selected from the text editing window or from the SEARCH interface. Example (3) is a sentence example exported from TypeCraft, illustrates locative inversion in Runyankitara, a Bantu language spoken in Uganda.
Omu nju hakataahamu abagyenyi
CL9 house CL16 PST
“In the house entered visitors”
IV CL2 visitor
Generated in TypeCraft.
The translational and function glosses, which belong to two distinct tiers in the
TypeCraft annotation interface, appear as one line when imported to one of the
word processing programs supported by TypeCraft. Although glossing on several tiers is conceptually more appropriate, linguistic publications require a more
condensed format. As for now we have decided on an export which displays 5
tiers for languages with a Latin script.
Figure 5 below gives an indication of the size and the content of the TypeCraft
database. TypeCraft is truly multi-lingual, and we encourage in particular the annotation and storage of data from lesser know or endangered languages. The
main language at present is Lule Sami with 2497 annotated or partially annotated
Top 15 TypeCraft languages
(by number of phrases)
Lule Sami (2497)
Runyankore-Rukiga (439)
Norwegian (411)
Akan (314)
Nyanja (192)
Ganda (114)
German (94)
Sekpele (87)
Abron (77)
Bini (71)
Koyraboro Senni Songhai (64)
Tumbuka (64)
English (60)
Icelandic (56)
Ga (55)
Figure 5 TypeCraft 15 Top Languages, March 2008
TypeCraft was used in a pilot study which had as a goal to compile a small corpus
of Lule Sami text. The project was supported by the Faculty of Humanities at the
Norwegian University of Science and Technology, and was run as a cooperation
between the Department of Language and Communication Studies and the Lule
Sami Center Arran at Drag in Northern Norway.
The use that the Lule Sami annotation project made of the Tcwiki illustrates
how TC can be used as an information sharing tool. With the TC wiki, projects
as well as individuals can represent their research online. The TCwiki is a mediawiki allowing user driven editing. Special for the TCwiki is that it supports the
loading of annotated examples from the TC database to the wiki. The two most
important functions of the TCwiki are the build-up of background knowledge essential for language annotation and its role as a forum for annotators and other
TypeCraft users. Wikis have very powerful features: articles can be created and
edited at any time, changes can be monitored, and the reversing of changes is
possible. This wiki functionality carries over to TypeCraft.
Figure 6 Annotating Lule Sami – a pilot study directed by the Norwegian University of Science and
Technology in cooperation with the Cultural the Lule Sami Center Arran at Drag in Northern Norway.
7 Conclusion
Symbolic rewriting is a distinct mode of linguistic research. There is no verdict
that forces us to express descriptive generalizations exclusively by evoking a formal apparatus of considerable depth. Instead given simplicity and parsimony of
expression it might well be that symbolic rewriting serves better for some research purposes than the representational modes of standard linguistic theories.
Yet at this point conventional glossing standards prevent us from generating independent linguistic resources.
With TypeCraft we have presented an interlinerized ‘glosser’ accessed online
and connected to a relational database for the storage and retrieval of linguistically processed text data. The application is wrapped in a wiki tool which allows
TypeCraft users to share information related to language annotation and documentation. TypeCraft is designed for linguists with no prior experience in the use
of databases or digital annotation tools. The system allows one to gloss textual
data without having to learn how to install and customize software. It facilitates
the generation of richly annotated linguistic examples directly suited for linguistic publication. With the help of TypeCraft the linguist can keep his linguistic
material in a repository independent of his individual publications.
African Language Technology. Accessed January 1, 2009.
Beermann, Dorothee. 2009. From Interlinear Glossing to Standard Annotation. Creating
Infrastructure for Canonical Typology. Conference hosted by the Surrey Morphological Group. Surrey University. England.
Bird, Steve and Gary Simons. 2003. Seven Dimensions of Portability for Language
Documentation and Description. Manuscript SIL International.
Labov, William. 1987. Some observation on the foundation of linguistics. unpublished
manuscript. University of Pennsylvania, USA.
Lehmann, Christian 1999. Documentation of endangered languages . A priority task for
linguistics. ASSIDUE Arbeitspapiere des Seminars für Sprachwissenschaft der
Universität Erfurt . Nr 1.
Lehmann, Christian. 2004. Interlinear morphological glossing. Morphology Ein Internationales Handbuch zur Flexion und Wordbildung. DeGryter Berlin–New York.
Leipzig Glossing Rules.
Accessed January 1, 2009.
Magnus, Margret. 1995. Description in Linguistic Theory. online manuscript. http:// Accessed January
31, 2009.
Online Resources
Field Linguist’s Toolbox. Accessed January 1,
Language Typology Resource Center
index.htm. Accessed January 30, 2009.
GOLD . Accessed January 31, 2009.
Open Language Archiving Community. Accessed
January 1, 2009.
Odin Database. Online Database for Interlinear Text.
Accessed January 2009.
TypeCraft Accessed January 2009.
Fusional verb morphology
Karl Erland Gadelii
University of Gothenburg/University of Paris–Sorbonne
1. Introduction
The expression “strong”verb has been used in different ways in the linguistic tradition. In Barðdal et al. (1997: 322), eight strong verb paradigms in the Nordic
languages are identified on the basis of the ablauted preterite (cf. also Pettersson
1996: 99–100). The ablauted vowel shows some intra-Nordic variation: as an illustration consider the strong preterite föll ‘fell’ of the Swedish verb falla ‘fall’,
which in other Nordic languages corresponds to fal(d)t/fall or féll. Another difference concerns diphthongized ablaut which is found mainly in Icelandic,
Faroese and Nynorsk, as in breit ‘broke’ (cf. bröt) and beit ‘bit’ (bet). In principle
however, regular patterns can be discerned in seven of the eight groups distinguished by Barðdal et al., whereas verbs of their eighth group exhibit what they
call “other” vowel alternation. Karlsson & Sahlquist (1975, henceforth KS) propose that a strong verb has a non-suffixal preterite and/or a supine on -it (we will
come back to this definition). KS distinguish six groups of strong verbs in Swedish, where the last one is made up of “irregular” verbs (although KS do not use
this term). This group contains 27 verbs belonging to 18 different types. Other
criteria for « strength » which have been used in the Swedish tradition are, citing
KS (1975 :46) : perfect participle suffix -en or -et, null preterites, present suffix
-er, ablaut. Linell, Svensson & Öhman (1971: 92–95, henceforth LSÖ) likewise
distinguish six groups of strong verbs in Swedish which coincide with those of
KS with the exception that they do not single out an “irregular”group. In contrast,
they treat the vowel change <a: o: a:> (I am here using a combination of graphic
and phonetic notation) among the “irregular” verbs as a case of systematic ablaut
(which could be questioned since it is only found in four or five verbs, dra(ga)
‘pull’, fara ‘leave’, gala ‘crow’, ta(ga) ‘take’ and marginally begrava ‘bury’).
Following LSÖ’s classification, we get the following picture as far as Swedish
is concerned:
i: e: i:
y: ö: u:
bita bet bitit
finna fann funnit
frysa frös frusit
‘bite bit bitten’
‘find found found’
‘freeze froze frozen’
u: ö: u:
ä(:) a(:) u(:)
a: o: a:
bjuda bjöd bjudit
bära bar burit
fara for farit
‘bid bid bid’
‘bear bore born’
‘leave left left’
LSÖ formulate quite an abstract rule system in order to account for the ablaut
observed, and in particular collapse (2) and (5) as well as (3) and (4), thereby disregarding vowel length. (5) is derived from (2) by vowel lowering, and (4) from
(3) by palatalization. Barðdal et al. (1997: 322) partially arrive at the same analysis, but since they respect vowel length, they collapse (3) and (4) but not (2)
and (5). LSÖ go on to identify additional regularities (exceptions added by the
present author):
(7) the pattern <y: ö: u:> (group 3) is instantiated after liquids and nasals ([r
l m n]), exceptions byta ‘change’ and dyka ‘dive’
(8) <u: ö: u:> (group 4) mostly after [j], various exceptions: duga ‘suffice’,
hugga ‘cut’, sluka ‘devour’, sluta ‘finish’, stupa ‘fall’, suga ‘suck’,
supa ‘drink’
(9) <ä: a: u:> (group 5) [r] or [l] after the vowel, exceptions kläcka ‘jump’
(particle verb as in det klack till i honom ‘he jumped’), skvätta ‘splash’,
växa ‘grow’
A more complex question concerns the derivation of the distribution of different vowels in the individual ablaut series. In contrast, nominal and adjectival
umlaut as in varm ~ värme ‘hot ~ heat’, mus ~ möss ‘mouse ~ mice’, ung ~
yngre ‘young ~ younger’, etc. is fairly straightforward: the only rule needed is
one of vowel fronting. When it comes to ablaut, LSÖ formulate a constrained
but complex set of rules involving fronting/backing and raising/lowering of
vowels. KS make a less formal but impressive attempt to extract a general picture of ablaut patterns by suggesting that classes (1) through (5) exhibit, roughly speaking, a “high – lowered – raised” pattern. In contrast, their class (6)
(fara for farit ‘leave left left’ + all other “irregular” verbs) rather exhibits a pattern of “low – raised – lowered” (very generally speaking, exceptions giva/ge
gav givit ‘give gave given’, be bad bett ‘pray prayed prayed’, ligga låg legat
‘lie lay lain’). In spite of different counterexamples, KS manage to come to the
conclusion that the patterns of “ordinary strong” verbs and “irregular” ones are
in complementary distribution. They consider the “high – lowered – raised”
patterns of the main classes of strong verbs to conform to the cardinal vowel
pattern /i a u/, and strengthen their argument by citing frozen expressions
which abide by this geometry (Swedish tripp trapp trull ‘one two three’ (never
*trupp trapp trill, *trapp tripp trull, etc.), snipp snapp snut “common way to
end a story”, Finnish hujan hajan ‘helter skelter’, ristiin rastiin ‘to and fro’,
etc, (KS 1975: 71). In English we find cases such as tick-tock, criss-cross,
cling-clang, snip-snap, etc. Potential counterexamples are however the Japanese expressions kasa-koso ‘rustle’ and gata-goto ‘rattle’ (Wikipedia, entry
“Reduplication”). According to KS, the opposed pattern of “irregular” verbs
make them vulnerable to change to a greater degree than verbs from the major
strong classes, which conform to a universal or even productive pattern (see
further below).
Both LSÖ and KS manage to bring home the idea that strong verbs in Swedish
do not just form a list of exceptions, but can be captured by a restricted set of
rules. Although ablaut was originally conditioned by nowadays opaque properties in the phonetic context, it is in principle synchronically productive, and the
relation between the infinitival and the following vowels is always one-to-one or
many-to-one (given that vowel length is considered). Since it is never
one-to-many, this means that there is no ambiguity in the system, and it is possible to describe it in terms of rules. For example, if the infinitival vowel is <u:>
as in njuta ‘enjoy’, the preterite vowel has to be <ö:> as in njöt. Similarly, the
preterite vowel <e:> as in, say, led ‘suffered’ by necessity implies a supine form
comprising <i:>, i.e. lidit ‘suffered’. So in principle, learning ablaut is not more
complicated than acquiring the five principles for plural formation of nouns in
Swedish, or any comparable allomorphic variation. This also means that we
ought to be able to synchronically elicit non-existing but possible ablauted preterites and supines, and this is also what we find in joking cases like “rang” ‘rang’
as opposed to the normal preterite ringde ‘rang’ of the verb ringa ‘ring’, where
forms like “reng” or “röng” are impossible.
In spite of the above-mentioned examples of the regularity and potential productivity of ablaut, it is obvious that this phenomenon tends to disappear when
verb paradigms weaken. In KS’ definition overview of weakening of strong
verbs in Swedish, loss of ablaut is one sign of regularization of the verb paradigm. However, I will argue in the present paper that loss of ablaut is THE most
important ingredient in verb weakening, and I do not think earlier studies have
pointed this out clearly enough. But before I develop this idea I would like to
sketch the behaviour of ablaut from a diachronic perspective.
2. Ablaut and language change
It is commonly believed in diachronic linguistics that ablaut was a Proto-Germanic innovation (cf. Harbert 2007, Venneman 2003, and many others),
whereby a simple Indo-European root alternation involving <ei oi -i>, <eu ou
-i>, or <e o -> (i.e. in principle <e o ->), was extended. As was pointed out in
the previous section, ablaut is rule-governed and therefore in principle productive. Examples such as dykade –> dök ‘dove’, skrivade –> skrev ‘wrote’,
stridde –> stred ‘fought’ were introduced during different periods of history—
Barðdal et al. (1997:326) mention on the one hand the 17th century, on the other the Romantic movement during the 19th century. Innovations such as those
just mentioned have become established in ordinary Swedish usage whereas
other forms have not, such as hang for hängde ‘hung’ and jog for jagade ‘hunted’. Synchronically, ablaut is found in Swedish in dialects as well as in emphatic and joking usages. Some examples apart from the above-mentioned
rang ‘rang’ are böt ‘changed’ (infinitive byta), lös ‘shone’ (lysa), mös
‘cocooned’ (mysa). Further cases mentioned in the radio broadcast Språket,
November 18, 2008, are spek ‘nailed’’ (spika) skröv ‘screwed’ (skruva), net
‘pulled the brakes (nita), fek ‘drank coffee’ (fika). töt ‘sounded one’s horn’
(tuta). Some of these examples seem highly peripheral to me, but respect the
morphophonological rules formulated above.
In a comparative Nordic perspective, ablaut is considered to be most common
in Icelandic, Faroese and Nynorsk, less so in Swedish, Bokmål and Danish. In
the Swedish newspaper corpus used by SAG (1999, vol. 2, p. 567), the percentage of strong verbs amounts to 12% at type level, whereas the number of strong
verb tokens is as high as 28%. Despite various mismatches among the Nordic
languages, such as Swedish weak knäckte ‘broke’ corresponding Danish strong
knak, and Swedish strong pep ‘whined’ corresponding to Icelandic weak pípaði,
Danish seems to be the language which has pushed farthest the regularization of
strong verbs. This impression is strenghtened by the fact that Danish only has
two weak verb conjugations, where verbs end in -ede and -te, respectively,
whereas the other Nordic languages display four (but different) weak paradigms.
When it comes to presento-preterite verbs, i.e. verbs which exhibit ablaut in the
present but not the preterite, these are more numerous in Icelandic and Faroese
than in the other Nordic languages, and include in the former languages verbs
such as mega ~ má ‘may’, muna ~ man ‘remembers’, munu ~ mun ‘will’, þurfa
~ þarf ‘needs’, unna ~ ann ‘likes’, eiga ~ á ‘possesses, has to’ (Barðdal et al.
1997:326). Related to this fact is the possibility in Nynorsk, Icelandic, Faroese
and certain Jutlandic dialects of having ablaut of strong verbs already in the
present. Cf. Nynorsk brista ~ brest ~ brast ~ broste ‘burst bursts burst burst’, as
well as taka tek ‘takes’, gråte graet ‘cries’, fare fer ‘goes’, komme kjem ‘comes’.
In Icelandic we similarly find tekur ‘takes’, grætur ‘cries’, and in Jutlandic dialects [gæ:r] ‘walks’ from gå (ibidem). Icelandic and Faroese also display a
number distinction in the preterite absent in the other Nordic languages: brast ~
brustum ‘(I) burst’ ~ ‘(we) burst’.
The above-mentioned facts indicate that Insular Scandinavian has preserved
ablaut to a greater extent than has Mainland Scandinavian, and among the latter
languages, Danish seems to be the one which has abandoned ablaut to the greatest extent (with the exception of the Jutlandic example cited above). This behaviour of ablaut parallels that of many other morpho-syntactic phenomena in the
Nordic languages, where Insular Scandinavian tends to preserve distinctions
which have been lost in Mainland languages.
It is natural to suppose that preservation of linguistic features is favoured by
geographic isolation, whereas intense language contact with ensuing bilingualism, code-switching, koinéization, etc., tends to lead to paradigmatic regularization. For geographic reasons language contact has been more intense in Denmark
than elsewhere in Scandinavia, which would explain the levelling tendencies
found in Danish verb paradigms. But as we just asked, why should ablaut be
more prone to change than affixal morphology? This obviously has to do with
the fact that ablaut is an instance of fusional morphology whereas affixal morphology is agglutinating. It could be suggested that fusional morphology is
marked per se in comparison with agglutinating morphology, regardless of how
rule-governed or principally productive the former is is. Intuitively this seems to
make sense: agglutination is a simple concatenative operation, whereas fusional
morphology involves a change internal to the verb root.
It is interesting in this respect to compare situations of abrupt language change
such as creolization to those of a more gradual nature, which the Nordic languages have supposedly experienced, in spite of the massive influence of Low
German at the time of the Hanseatic League. During creolization, fusional and
agglutinating verb morphology tends to get lost but reappear in the form of preverbal particles. As a case in point consider the behaviour in Lesser Antillean
French Creole of the verb bwè ‘drink’, reflecting French boire:
Lesser Antillean
Nou Ø bwè.
Nou Ø bwè.
Nou té bwè.
Nou ka bwè.
Nou ké bwè.
Nou té ka bwè.
Nou té ké bwè.
Nou ké ka bwè.
Nou té ké ka bwè.
‘We drink.’
‘Nous buvons.’
’We drank / have drunk.’‘Nous avons bu / bûmes.’
‘We had drunk.’
‘Nous avions bu.’
‘We are drinking.’
‘Nous buvons.’
‘We will drink.’
‘Nous boirons.’
‘We were drinking.’
‘Nous buvions.’
‘We would drink.’
‘Nous boirions.’
‘We will be drinking.’ ‘Nous boirons.’
‘We would have been ‘Nous aurions bu.’
As can be seen, the French verb is highly variable, exhibiting change in the root
vowel as well as different suffixes. In Lesser Antillean, the verb is invariably
bwè, preceded by strictly ordered particles coding tense, mood and aspect. In
case preverbal particles are absent, the verb is interpreted as present or past depending on the semantics of the verb and the context.
Now in ordinary language contact one does not normally witness this kind of
dramatic change, but rather less spectacular cases where fusional morphology
tends to be substituted by agglutinating one. But even in situations of less dramatic language contact, one can observe the fact that preterites are sometimes
avoided to the benefit of supines, as in Jag har bitit av en tand igår “I have broken a tooth yesterday” (supine) ~ Jag bet av en tand igår ‘I broke a tooth yesterday’ (preterite). In a case such as this, a fusional form (bet) is replaced by a construction which is both isolating (Aux + Supine) and agglutinating without ablaut (bit-it). In the following I will however not be concerned with this type of
change, but rather with cases where fusional morphology is replaced by agglutinating ditto within the verb.
We will now apply the idea of fusion –> agglutination to the regularization of
verb paradigms in the Nordic languages, notably Swedish. As a starting point, let
us take Karlsson & Sahlquist’s (1975: 489) definition of “strong verb”:
(19) A strong verb is a verb whose paradigm displays the supine suffix -it
and/or a preterite allomorph other than -de, -te, -dde. [my translation]
I would like to suggest that this definition, useful though it is for KS’ purposes,
suffers from two weaknesses, namely that (i) it does not evoke ablaut explicitly,
and (ii) it assigns too much importance to the supine suffix -it as a characteristic
of paradigm strength.
The second part of the definition identifies the preterite allomorphs -de, -te,
-dde as manifestations of the three weak conjugations in Swedish, as exemplified
by dansa dansade ‘danced’, köp(a) köpte ‘bought’ and bo bodde ‘lived’. As can
be seen, suffixes are here added in a straightforward agglutinating way. Having
another preterite allomorph than these implies displaying ablaut. Problematically, KS’ definition will rule out preterites which have both ablaut and suffix, such
as e.g. dölja dolde ‘hide hid’. Since these preterites display -de, the paradigm will
be considered as weak, although it exhibits ablaut.
The first part of the definition concerns agglutinating morphology, where the
suffix -it, in keeping with Nordic linguistic tradition, is considered to be a characteristic of strong verbs, as in dricka drack druckit ‘drink drank drunk’. However, in non-standard Swedish one frequently encounters supines such as drickit
“drinked” instead of standard druckit. Once again, KS’ definition fails to distinguish between these two supines, which are simply ruled in as (equally) strong.
In my opinion, the non-ablauted form drickit is much “less strong” than the ablauted druckit, but the difference between the two is not signalled in KS’ definition.
In addition, we may among the “irregular” verbs find non-standard cases
where the preterite is weak and the supine does not end in -it and is thus by definition weak, making the whole paradigm weak, even if it exhibits ablaut. A case
in point would be ligga liggde legat ‘lie “lied” lain’ where the preterite is weak
by virtue of the suffix -de, and where the supine is likewise weak since it does
not end on -it. According to KS’ defnition, the presence of ablauted -e- does not
make a difference, wrongly in my opinion.
For KS, the disjunctive definition is useful since it makes it possible to establish two cases of semi-strong verb paradigms, namely those where either the
preterite or the supine is strong. Consider the four possibilities below:
(20) Fully strong:
(21) Semi-strong:
(22) Fully weak:
dricka drack druckit ‘drink drank drunk’
(a) strypa ströp strypt ‘strangle strangled strangled’
(strong preterite, weak supine)
(b) växa växte vuxit ‘grow grew grown’
(weak preterite, strong supine)
döma dömde dömt ‘judge judged judged’
KS’ very exhaustive corpus further shows that the combination of strong preterite and weak supine, i.e. (21a), is far more common than the reverse case,
(21b). In addition, they claim that semi-strong verbs may pick up forms from all
three weak conjugations, leading to various types of semi-strong verbs, but interestingly enough, adoption of forms from the second conjugation (cf. köpa köpte
köpt ‘buy bought bought’) is far more common than that from conjugation 1 and
3. In the above example, the supine strypt ‘strangled’ and the preterite växte
‘grown’ conform to the 2 conjugation pattern. KS propose, following Per Linell,
that this is because strong verbs take the suffix -er in the present, just like verbs
of the second conjugation. KS explain the fact that supines are more prone to
change than preterites by claiming that the supine is marked as compared to the
preterite, and that marked forms adopt changes before unmarked ones do. I do
not find this idea fully clear but have not been able to check their source (Jakobson 1966). The observation put forward above that preterites can be replaced by
supines (bet –> har bitit) but not vice versa is an example of the opposite, i.e. that
the preterite is marked as opposed to the supine. I will not pursue this question
3. An alternative definition of “strong verb”
I would like to suggest that the presence of ablaut makes a paradigm strong,
wherever the ablaut may reside (in the preterite and/or the supine form, together
with a weak suffix in the preterite (cf. dölja dolde ‘hide hid’), together with a
supposedly weak suffix in the supine (as in ligga legat ‘lie lain’), or in the present
of a presento-preterite verb (as in kunna kan kunde kunnat ‘be/is/was/have been
able to’). Different degrees of strength can be discerned: paradigms which have
ablaut also in the present of a strong paradigm (cf. Nynorsk brista ~ brest ~ brast
~ broste ‘burst bursts burst burst’) and those where number distinctions lead to
different ablauts (as in Icelandic brast ~ brustum ‘(I) burst’ ~ ‘(we) burst’) are
“stronger” than those who do not make these distinctions. That is, I find ablaut
to be the hallmark of “strength”, whereas it is not even mentioned in three of the
six major sources on strong verbs in Swedish quoted by KS (1975 :46). One argument for the vulnerability of ablaut is the above-mentioned fact that fusional
morphology is the first phenomenon to disappear in situations of abrupt language
contact, whereas some types of bound morphology may actually survive even in
these contexts (cf. the passionate debate about morphology in creole genesis
which has engaged creolists during the last ten years (see e.g. DeGraff 2002,
McWhorter 2001). DeGraff, desperately trying to prove that there is nothing “exceptional”about creole languages, has managed to come up with several examples of bound morphology in creoles, but non which involve ablaut or umlaut.
This means that I think that the first part of KS’ definition should be omitted
(“A strong verb is a verb whose paradigm displays the supine suffix -it…”) and
I would like to suggest the following alternative definition of “strong” verb:
(23) A strong verb is a verb which displays ablaut SOMEWHERE in its
This means that I don’t consider the supine suffix -it to be able to make a paradigm strong on its own. For one thing, this suffix is affixal rather than fusional.
Also, the fact that it ends on -t makes it formally similar to supine markers in
weak conjugations (cf. dansat ‘danced’, köpt ‘bought’, bott ‘lived’). That is, the
difference between supposedly strong vik-it ‘folded’ vs. weak vik-t of the verb
vika ‘fold’ is minimal as compared to the one between the strong ablauted preterite nös ‘sneezed’ vs. its weak counterpart nys-te (infinitive nysa).
Next, we notice with Bergman (1968:169–70), citing Rydqvist (1850–52),
that colloquial speech in Stockholm at the time displayed the supine ending -i,
by all probability a reduced form of -it, on ordinary weak verbs of the 2 conjugation. Bergman gives the following examples:
(24) Non-standard supine
blödi ‘bleeded’
händi ‘happened’
köpi ‘bought’
skilji ‘divided’
spilli ‘spilled’
Standard supine
Bergman (ibidem) goes on to note that strong verbs in the Stockholm dialect further displayed non-ablauted supines on -i(t) as in
(25) Non-standard supine
bäri ‘born’
bryti ‘broken’
frysi ‘frozen’
sticki ‘stung’
stjäli ‘stolen’
sälji ‘sold’
sätti ‘put’
välji ‘chosen’
Standard supine
(Cf. further Ståhle (1966); Reuter’s ruta April 4, 1986 treats parallel facts in Finland Swedish). We will see later that this type of regularization of the supine is
typical of present-day mixed verb paradigms.
The above examples show that -it is actually productive when it comes to
forming supines of the 2 and 4 (strong) conjugation, which means that it can not
be a strong feature if we by “strong” mean “marked” or “exceptional”. It could
also be ventured that the language of 19th century Stockholm was characterized
by dialect levelling among the large number of newcomers from all parts of the
country (cf. Kotsinas (2000). In such a situation, we may consider supine formation by means of -it to have become the generalized way of inflecting the verb in
the 2 and 4 conjugation. In the 1 and 3 conjugation the operation cannot apply
for phonological reasons (*dansait “danced”, *boitt “lived”)
The same phenomenon can be observed in child language: as anecdotal evidence I would like to cite my nephew who took part in the following conversation:
(26) Adult: Vad har ni gjort för något idag?
‘What have you been doing today?’
(27) Child: Vi har liggit och läsit.
‘We have been reading in our beds.’
(lit. “We have lain and read”, adult Swedish: Vi har legat och
The adult form legat, which comes out as weak by KS’ definition since it does
not end on -it (ablaut in the supine not making the paradigm strong in their opinion), is equalled by liggit which according to KS is strong since it ends on -it, and
since non-ablaut in the verb stem is not relevant for them. I would on the contrary
like to propose that the adult form legat contains TWO strong features: the ablauted stem vowel which the child replaces by the non-ablauted one, and the
marked supine suffix -at, which is substituted by the (on my assumption) unmarked -it. When it comes to läsit, this form is likewise strong according to KS,
whereas they consider läst to be weak, since it ends on -t only. The overall result
is strange: the adult sentence would contain two weak forms but the child utterance two strong ones ! If child language is thought to manifest unmarked morphology, the phenomena illustrated above are very unexpected. However, by the
definition in the present work, both supines in the child language come out as
weak (since they do not display ablaut). In the adult sentence, legat will be analyzed as strong since it exhibits ablaut, whereas läst will be regarded as weak
since it does not. So the variation between -t and -it does not play a role in the
current definition—läst and läsit are equally weak. In fact, it may be argued that
läsit is weaker or at least more basic speaking in terms of universal phonotactic
principles: läsit conforms better to the canonic CV syllable than läst, let alone if
it is pronounced läsi. This type of CVCV word is commonly attested in child language, starting with mama and papa, to which läsi and liggi conform perfectly
well. Another argument for the basicness of CVCV is the paragogue we find not
only in languages of the Bantu and Japanese type (among many others) but also
in creole languages: cf. two examples from Surinamese English creoles: Sranan
taki reflecting English talk and Saramaccan lafu < laugh (Holm 2000:142). KS
(1975:58) represent a diametrically opposed view when they consider forms like
läst to be basic since they are monosyllabic.
All the above examples indicate that the supine suffix -it has no strong property at all but rather constitute an unmarked way of forming supines of verbs of
the 2 and 4 conjugation, whereas, as we have seen, verbs whose stems end in a
vowel (1 and 3 conjugation, cf. dansa ! ‘dance !’, bo ! ‘live !’) receive the -t supine suffix for independent phonological reasons.
I will now look at the six groups of strong verbs discerned by KS (cf.
examples (1) through (6) in the introduction of this paper) in the light of my al30
ternative, ablaut-based definition of “strong verb”. I will thus consider “weakening” to mean “avoidance of ablaut”, and I will primarily be investigating supine
forms, not preterites.
4. Application to material
KS (1975) contains a wealth of data on strong, mixed and weak verb paradigm
in Swedish, and it is difficult to add new specimens to their material. However,
the verb forms treated in KS are judged by the linguistic intuitions of the authors
in their capacity as native speakers, and the present work hopes to be able to complement their intuition-based work with naturalistic data from the Internet, where
the language is often highly informal, thus following KS’ suggestion that an
overview of mixed verbs should not be based on literary language.
4.1 The <i: e: i:> group
The first group of strong verbs identified by KS diplays the <i: e: i:> alternation
as in (1) above, here repeated as (28):
(28) <i: e: i:>
bita bet bitit
‘bite bit bitten’
This group contains 31 verbs, which display notably little weakening: most preterites are ablauted and practically all supines take the -it suffix. It should be
noted that supines in this group do not display ablaut, which means that according to my hypothesis they should be stable non-ablauted forms ending on -it,
where the stem cannot change since it is not ablauted in the first place, and the
suffix should not be prone to change from -it to -t if -it is taken to be the unmarked supine suffix. In the present group, KS identify five cases where they
consider another supine than the one on -it to be equally or more common than
the latter. The very fact that 26 of the 31 supines in this group are not considered
by KS to exhibit variation, let alone preference for the -t suffix, would seem to
constitute an argument for the hypothesis in this paper. The five cases which KS
consider to oscillate are spritt ~ spridit ‘spread’, vritt ~ vridit ‘torn’, vikt ~ vikit
‘folded’, tigit ~ tegat ‘kept silent’ and fisit ~ fist/fisat ‘farted’, where it should be
noted in passing that tegat and fisat unusally enough have picked up the supine
suffixe of the 1 conjugation. However, a search on the Internet shows that KS’
intuitive judgements are not always confirmed. But before looking at statistics
on the five supines in question, it should be pointed out that it is notoriosly difficult to avoid false hits when searching the Internet, so in the following I have
searched for the supine in combination with the auxiliary verb har ‘has’ (as in
e.g. har spridit ‘has spread’) and of course on Swedish pages only, in order to
reduce noise. This means that some potential hits have by definition been excluded, in particular those where the auxiliary and the supine are not adjacent, but on
the other hand no past participles have been included. Since this inconvenience
pertains to all searches, statistical trends should still be of interest. The result of
the current search is illustrated in the table below:
spritt spridit vritt vridit vikt
vikit tegat tigit fist/fisat fisit
‘kept silent’ ‘farted’
30 700 74 800 4 360 14 000 10 300 3 170 277 2 770 412
7 090
As can be noticed, the -it supine is far more common in four of the five cases, the
only exception being vikt ‘folded’ with 10 300 hits vs. vikit ‘folded’ 3 170. This
is somewhat surprising but could have to do with the fact that I got a number of
undesired hits referring to the noun vikt ‘charge’ (where har was interpreted as a
main verb). In spite of this minor counterexample, the current hypothesis is confirmed: contrary to KS’ judgements, the supine suffix -it seems quite resistent to
4.2 The <i a u> group
The second group is illustrated in (30):
(30) <i a u> finna fann funnit
‘find found found’
This category numbers 25 verbs and is particularly relevant for our purposes
since the supine displays ablaut. Our hypothesis thus predicts that funnit above
should be weakened to finnit, with corresponding changes for the other 24 verbs.
However, according to KS only 2 of the 25 verbs in this group display a non-ablauted form on -it, namely slintit ‘slipped’ instead of standard sluntit and stinkit
‘stunk’ (cf. stunkit). However, my Internet search shows that another 19 supines
not mentioned by KS display an alternative non-ablauted -it form (the only cases
not being attested are förnimmit “sensed”, klingit “rang”, stingit “stung”, svinnit
“disappeared”, and tvingit “forced”, but then it should be pointed out that the
verbs förnimma, stinga and svinna are highly infrequent in and of themselves).
The non-ablauted -it supines elicited in my Internet search were: bindit ‘bound’,
brinnit ‘burned’, bristit ‘burst’, dimpit ‘fallen’, drickit ‘drunk’, finnit ‘found’,
hinnit ‘had time’, rinnit ‘flown’, simmit ‘swum’, sittit ‘sat’, slinkit ‘slipped’, slippit ‘hadn’t to’, spinnit ‘spun’, sprickit ‘cracked’, springit ‘run’, sprittit ‘jumped’,
stickit ‘stung’, försvinnit ‘disappeared’, vinnit ‘won’. True, some of these forms
are not particularly frequent (brinnit 8 occurrences, rinnit 6, etc.) but the very
fact that they exist is interesting, and furthermore it should be noted that brinnt
and rinnt are even less frequent (0 and 2 hits, respectively), again indicating that
-it is the unmarked supine suffix.
4.3 The <y: ö: u:> group
The third group is exemplified below:
(31) <y: ö: u:> frysa frös frusit
‘freeze froze frozen’
Our hypothesis predicts that the first change affecting this group should be the
replacement of frusit by frysit rather than by frust or fryst, with parallel changes
affecting the other verbs in this group (25 in all). Here KS acknowledge numerous instances of non-ablauted supines on -it co-existing with ablauted dittoes
(i.e. frysit ~ frusit ‘frozen’), something which yet again underscores the commonality of the non-ablauted -it form. They do however not list bytit ‘changed’
(cf. bytt), knyckit ‘stolen’ (knyckt), and ryckit ‘pulled’ (ryckt), all attested on
the Internet. In the majority of cases, KS consider the ablauted form on -it or
the non-ablauted form on -t to be more common than the non-ablauted one on
-it (i.e. they find frusit and fryst to be more common than frysit). This may be
so, but once again the ubiquitous existence (sometimes amounting to thousands of tokens) of the non-ablauted form on -it is striking, and furthermore 8
verbs display the non-ablauted -it form more often than the ablauted one (contrary to KS’ judgements), namely fnysa ‘frown’, nypa ‘pinch’, nysa ‘sneeze’,
ryka ‘smoke’, rysa ‘shiver’, snyta ‘blow one’s nose’, strypa ‘strangle’, tryta
(32) fnysit ~ fnusit, nypit ~ nupit, nysit ~ nusit, rykit ~ rukit, rysit ~ rusit,
snytit ~ snutit, strypit ~ strupit, trytit ~ trutit
4.4 The <u: ö: u:> group
The <u: ö: u:> group, exemplified by
(33) <u: ö: u:> bjuda bjöd bjudit
‘bid bid bid’
contains 19 verbs. Here again, supines do not display ablaut so this group is less
interesting for us, but we do predict that the -it suffix should be stable, not tending to be replaced by -t. KS’ judgments back up this prediction in all but one of
the relevant cases, namely dugit vs. dugt ‘sufficed’, where they consider the latter form to be more common, which is also confirmed by my Internet search. I
have at present no explanation as to why this particular verb presents a counterexample when supines such as sjungit ‘sang’, sjunkit ‘sunk’ and supit ‘drunk’
on the other hand conform to the rule—they are more common than sjungt,
sjunkt, supt.
4.5 The <ä(:) a (:) u(:)> group
As an illustration to the <ä(:) a (:) u(:)> group, which has 10 members, consider
(34) <ä(:) a(:) u(:)>
bära bar burit
‘bear bore born’
KS do not list a single non-ablauted supine on -it for this group, whereas 9 of
the 10 cases in point are actually attested on the Internet, namely bärit ‘born’,
skvättit ‘splashed’, skärit ‘cut’, smällit ‘banged’, smältit ‘melted’, stjälit
‘stolen’, svältit ‘starved’, växit ‘grown’ (cf. burit, skvätt, skurit, smällt, smält,
stulit, svultit, växt/vuxit). Once again, the non-ablauted forms on -it may not
always be frequent, but their very presence should be noticed. The only nonattested case is skälvit ‘trembled’, where skälvt is the only form present on the
4.6 The group of “irregular” verbs
I am here putting “irregular” between quotation marks because it is a term which
is often avoided, and furthermore it is not fully clear where the demarcation line
should be drawn between “irregular” and strong verbs. In KS’ model, this sixth
group consists of 27 verbs belonging to 18 different types. It is significant that
some of the verbs in this group are elevated to members of specific classes in
Barðdal et al. (1997: 322–23) as well as in Linell, Svensson & Öhman (1971), as
we saw in the introduction. I will in the following stick to KS’ classification, and
treat the 27 verbs in question as one single but heterogeneous group.
Most supines in this class do not display ablaut, as in the model case fara for
farit ‘leave left left’, which means that they are not primarily relevant to our investigation, as they furthermore tend to express supine by the -it suffix. There is
only one case where KS does not list an existing non-ablauted supine on -it in
this group, namely our aforementioned example liggit ‘lain’, cf. legat. Monosyllabic verbs cannot take the suffix -it for phonological reasons, something which
rules out se ‘see’, dö ‘die’, le ‘smile’, stå ‘stand’, slå ‘hit’, få ‘get’, gå ‘go, walk’.
A number of verbs hesitate between mono- and disyllabic infinitives with ensuing effects on the supine:
dra(ga) drog dragit/dratt ‘pull pulled pulled’, ta(ga) tog tagit/tatt ‘take took
taken’, giva/ge gav givit/gett ‘give gave given’, be(dja) bad bedit/bett ‘pray
prayed prayed’. The same behaviour can be found with purely monosyllabic slå
‘hit’: slå slog slagit/slått ‘hit hit hit’. Forms on -tt seem to be phonological reductions of the -it type, but of a different kind than e.g. frysit ~ fryst. I will not
treat these cases further here, cf. the discussion in KS. My prediction is borne out
in the case of hålla höll hållit/hållt ‘hold held held’ where the former supine occurs 275 000 times and the latter 139 000. However, four verbs constitute counterexamples, namely
gala ‘crow’: galit (172) / galt (289) ‘crowed’
häva ‘heave’: hävit (28) / hävt (3 290) ‘heaved’
begrava ‘bury’: begravit (100) / begravt (2 800) ‘buried’
svära ‘swear’: svurit (41 800) / svärit (925) ‘sworn’
At present I have no explanation to offer for these counterexamples, except for
the observation that a non-ablauted supine on -it indeed exists, but is less frequent than the ablauted one.
5. Summary and suggestions for future research
I have in the present paper tried to complement Karlsson & Sahlquist’s (1975)
extremely detailed investigation of verb paradigms in Swedish with authentic
frequency statistics googled from the Internet. It was thereby found that some
of the grammaticality/frequency judgments pronounced by KS were contradicted by Internet data, something which comes as no surprise considering the
fact that the KS paper dates back in time to the 1970’s. The second main point
in this paper has been to argue for the idea that the -it suffix constitutes an unmarked way of expressing the supine, whereas ablaut is marked. I have not
been able to deal with a number of relevant topics which could doubtless shed
more light on the behaviour of strong verbs, in particular the interaction between supines and preterites, presents, imperatives, past participles and
particle verbs. In addition, the “irregular” group should be subject to a more
fine-grained analysis, coupled by a study of presento-preterite verbs as well
as those of the 2 conjugation which display ablaut (as in dölja dolde dolt ‘hide
hid hid’). Future research should also present statistical data in more detail
and ought to involve in-depth studies of strong verbs in all the Nordic languages.
6. References
Barðdal, Johanna, Nils Jörgensen, Gorm Larsen & Bente Martinussen. 1997. Nordiska.
Våra språk förr och nu. Lund: Studentlitteratur.
Bergman, Gösta. 1968. Kortfattad svensk språkhistoria. Stockholm: ePan Prisma.
DeGraff, Michel. 2002. Morphology in creole genesis. Kenstowicz, Michael, ed. Ken
Hale: a life in language. Cambridge, MA: The MIT Press. 53–121.
Harbert, Wayne. 2007. The Germanic languages. Cambridge: Cambridge University
Holm, John. 2000. An introduction to pidgins and creoles. Cambridge: Cambridge
University Press.
Karlsson, Fred & Åsa Sahlquist. 1975. Starka verb i förvandling. Nysvenska studier 54.
Kotsinas, Ulla-Britt. 2000. Kontakt, variation och förändring: Studier i stockholmsspråk. Stockholm: Almqvist & Wiksell.
Jakobson, Roman. 1966. Zur Struktur des russischen Verbums. Martin Joos, ed.
Readings in Linguistics. Chicago : Chicago University Press. 22–30.
Linell, Per, Bengt Svensson & Sven Öhman. 1971. Ljudstruktur. Lund : Gleerups.
McWhorter, John. 2001. The world’s simplest grammars are creole grammars.
Linguistic Typology 5(3/4). 125–156.
Pettersson, Gertrud. 1996. Svenska språket under sjuhundra år. Lund: Studentlitteratur.
Rydqvist, Johan Erik. 1850–52. Svenska språkets lagar 1. K.F. Söderwall, utg.
SAG (Svenska Akademiens Grammatik). 1999. Teleman, Ulf, Staffan Hellberg & Erik
Andersson, red. Stockholm : Norstedts.
Ståhle, Carl Ivar. 1966. Verbböjningstyper och verbfrekvens i stockholmskt berättarspråk. Allén, Sture, utg. Förhandlingar vid sammankomst för att dryfta frågor
rörande svenskans beskrivning 3, Göteborg 1965. 5–27.
Venneman, Theo. 2003. Language in prehistoric Europe north of the Alps. Bammesberg, Alfred & Theo Vennemann, eds. Languages in prehistoric Europe. Heidelberg : Carl Winter. 319–332.
Karl Erland Gadelii
University of Gothenburg
Linguistics Department
Box 200
SE 405 30 Göteborg, Sweden
[email protected]
Département d’Etudes Nordiques
Université de Paris–Sorbonne
108 bd de Malesherbes
FR 750 17 Paris, France
[email protected]
ONE: Between Numeral, Indefinite Marker and
Elena Gorishneva
ZAS, Berlin
The paper will examine the development and the use of the number word ONE1
as indefinite article, focussing on languages with an incomplete article system
like Bulgarian, Hebrew or Icelandic as well as on articleless languages like Russian or Korean. The goal of this paper is to discuss various functions of the indefinite article, giving special attention to the intensificational force of ONE only
sporadically noticed before. This use will be explained as an interplay between
three factors: i) the indefinite article contributing a taxonomic interpretation, ii)
the scalar property provided by a gradable predicate, and iii) the exclamative
prosody indicating the remarkability of the corresponding subkind.
1. Article development
1.1 First stage of article development
According to Givón (1978:300), cf. also Presslich 2000, Leiss 2000, Heine &
Kuteva (2006)2, the development of the numeral ONE into an indefinite article
proceeds in two stages: “the first representing a generalization from [numeral] to
[ref.-indef.], the second from [ref.-indef.] to [indefinite].” At the first stage the
numeral occurs in spatially and temporally defined contexts and marks the corresponding NP as specific-indefinite, e.g. (1) in Russian:
(1) Ja chou
vstretit’sja s
odnim znakomym, kotorogo davno ne videl.
I want-PRS.1SG meet-INF
with ONE
‘I want to meet a friend whom I haven’t seen for a long time.’
not saw
In this paper ONE is used to refer to the numeral and the indefinite article which have the identical
form in languages under discussion.
Heine & Kuteva (2006), based on Heine (1997), posit a more fine-grained classification of article
development in which they distinguish between five stages: the initial stage is the numerical one,
stages 2 and 3 as well as stages 4 and 5 roughly correspond to Givón’s referential-indefinite and indefinite stages respectively.
The same function can be fulfilled by edin ‘ONE’ in Bulgarian, han ‘ONE’ in
Korean (cf. Yeom 2007), xad ‘ONE’ in Hebrew (cf. Borer 2005).
In order to distinguish the specificity marker ONE from the numeral, some
diagnostic tests can be used, cf. Ionin 2007 for Russian. First, the specificity
marker is always destressed while the numerals can have different stress patterns. Second, the specificity marker ONE takes an obligatorily wide scope,
whereas other numerals can have both wide-scope and narrow-scope readings,
regardless of stress. Third, the use of ONE implies that the identity of the individual denoted by the NP is somehow relevant in contrast to other numerals which
do not have such a requirement. Finally, the specificity marker ONE occurs in
contexts where the use of the numeral ONE is awkward (cf. such sentences as
Mary married a linguist).
Some characteristics with regard to the distinction between the indefinite
article and the numeral seem to hold cross-linguistically, even in genetically and
geographically unrelated languages. In many languages which possess an identical form for the numeral and the article the latter doesn’t usually carry stress
(cf. Bulgarian, Korean, Turkish, etc.). Rullmann & You 2003 noticed, for example, the following correspondence between singular indefinites in Mandarin
Chinese und English:
(2) a. yi (stressed) + CL+ noun
b. yi (unstressed) /n + CL + noun a(n)
While the sequence yi CL noun ‘ONE + classifier + noun’ with stressed yi corresponds to the numeral ONE, the meaning of unstressed yi is rather similar to the
English determiner a(n) and in this case yi can also be omitted. However, when
yi occurs in contrast to other numerals it is always stressed and its omission is not
permitted, cf. (3) and (4):3
(3) Zuotian wo mai le
(yi) ben shu.
yesterday I buy ASP ONE CL book
‘Yesterday I bought a book.’
(4) Wo bu zhi shi mai le
(Rullman & You 2003:12)
ben shu, ershi mai le
I not just be buy ASP ONE CL book
‘I didn’t buy one book. I bought five.’
wu ben.
buy ASP five CL
(Rullman & You 2003: 25)
The second tendency which applies cross-linguistically is that the reduced
forms developed from the numeral ONE only function as a determiner. Consider, for instance, a(n) vs. one in English, e / en vs. eine in Swiss German,
’n /’ne (colloquial) vs. ein / eine in German, xad vs. exad in Hebrew, or eden
vs. en in Slovenian. Indonesian also has the numeral satu ‘ONE’ and the form
The tendency that contexts in which the quantity has to be stressed require the explicit appearance
of the numeral displays in many languages which allow the use of bare nouns, cf. Slavic languages,
e.g. Russian:
(i) a. Peter kupil knigu.
b. Peter kupil odnu knigu, a Maria tri.
Peter bought book
Peter bought ONE book but Maria three.
‘Peter bought a book.’
‘Peter bought one book and Maria bought three.’
se developed from satu, which corresponds to the English indefinite article
(5) Dia ada se
ekor kucing.
she have ONE CL
‘She has a cat.’
In contexts in which the quantity is important, the numeral satu ‘ONE’ should be
used, e.g. satu ekor kucing ‘ONE CL cat’, or in colloquial speech with the omission of a classifier, e.g. satu kucing ‘ONE cat’. In sentences with specific indefinite NPs like Mary married a linguist the reduced form se is the appropriate option.
1.2. Second stage of article development
The second stage of article development is characterized by the occurrence of
ONE in episodic sentences as well as in generics and with predicative NPs. At this
stage, the use of ONE is not limited to a specific indefinite function anymore but
is extended to non-specific environments. The languages representing this stage
include the Romance and Germanic languages. We can also observe the presence
of the indefinite article in Bulgarian and Turkish, in spite of some restrictions of
the use of ONE. Though Bulgarian edin ‘ONE’ doesn’t reach the stage where the
indefinite article is completely developed, the range of its possibilities is very extensive, cf. (6)–(9):
(6) Imam
have-PRS.1SG meeting with
‘I’m meeting a friend now. ’
(7) Edin mâž trjabva
edin poznat sega.
da izchranva
man must-PRS.3SG
to feed-PRS.3SG
‘A man must be able to feed his family. ’
(8) Kitarata
semejstvoto si.
e (edin) strunen instrument.
guitar –DEF
is ONE string
‘The guitar is a string instrument.’
As illustrated above, edin ‘ONE’ serves as a specificity marker of the indefinite
NP in (6) or it can appear in generic contexts (7). A different situation arises in
the predicative position where the use of edin ‘ONE’ is more restricted. It can occur in generic classifying sentences like (8); nevertheless, its use is not obligatory
and a bare NP is more common in this case. There is, however, a difference between these two variants. In the case with a bare NP, we are dealing with ascribing a property to a subject (9a). The kind ‘guitar’ has the property ‘string instrument’ because every specimen of the kind ‘guitar’ has the property ‘string instrument’.The use of edin leads to a taxonomic reading which says that the guitar is
a subkind of the kind ‘string instrument’, that is, it is one of the string instruments
(9b). Notice that ‘string instrument’ occurs in two functions here: as a predicate
in (9a) and as a corresponding kind with subkinds in (9b).
(9) a. STRING INSTRUMENT (guitar)4
b. SUBKIND (guitar, string instrument)
In (10) the articleless noun is the only option while edin ‘ONE’ is not allowed:
(10) Peter e (* edin) uitel.
Peter is ONE teacher
‘Peter is a teacher. ’
To sum up, the development and the use of the numeral ‘ONE’ in the languages
analysed can be demonstrated in the picture below:
(11) numeral ‘1’
+episodic, - generic, -predicative
+episodic, +generic, +predicative
Russian, Korean, Hebrew
Bulgarian Turkish
In terms of Givón’s two-stage grammaticalization of the article, Icelandic exemplifies the numerical stage where ONE is used only as a numeral. Such languages
like Russian, Korean, and Hebrew represent the first stage of the article development characterized by the ability of ONE to mark specific indefinite NPs whereas
Bulgarian and Turkish can be posited close to the final stage that displays the
completely grammaticalized indefinite article that is exhibited by German,
French, etc.
1.3. Intensifying ONE
Some unusual findings concerning the use of ONE seem to provide evidence that
ONE has an additional function, i.e. can contribute to an intensifying reading. As
mentioned in the previous section, the Bulgarian edin ‘ONE’ cannot appear with
predicative NPs (cf. 9); nevertheless, its felicitous use has been observed in sentences like (12):
(12) a. Peter e glupak.
Peter is fool
‘Peter is a fool.’
b. Peter e edin glupak!
Peter is ONE fool
‘Peter is such a fool! ’
In comparison with a bare noun in (12a), the NP in (12b) has an intensificational
force and is correctly translated into English as such a (big) fool. This use of edin
has been found only in sentences where a predicative NP introduces a scalar
noun like glupak ‘fool’, chubavets ‘handsome man’, genij ‘genius’, vseznajko
‘know-it-all’, etc. In cases in which edin occurs with a non-scalar noun, it is coerced to a scalar noun:
(13) a. Peter e dete / student.
Peter is child / student
b. Peter e edno dete / edin student.
Peter is ONE child / ONE student
In the formulas, predicates are given in capital letters and names in lowercase letters.
The variants with edin can be accepted, though they are problematical, in contexts dealing with explanation, characterization, or evaluation based on typical
characteristics denoted by an NP, e. g. Peter can’t do something like this. He is
just a child.
Similar data have been found in Turkish, e.g. in contrast to güzel kiz ‘a pretty
girl’ without the indefinite article bir, the NP güzel bir kiz ‘a pretty girl’ with bir
‘ONE’ positioned after güzel ‘pretty’, which gives more prominence to the adjective, has the meaning ‘an especially pretty girl’ (Presslich (2000: 68), cf. also
Göksel & Kerslake (2005). The use of the indefinite article under the intensifying
reading has also been observed in German and Greek with abstract mass nouns
(hunger, thirst, heat, etc.), which usually occur without an article. The appearance of ein or mia leads to the intensification, see (14), (15):
(14) a. Ich habe Durst.
I have thirst
‘I’m thirsty. ’
b. Ich habe einen Durst!
I have ONE thirst
‘I’m so thirsty! ’
(15) a. (Ego) pinao.
‘I’m hungry. ’
b. Exo
mia pina!
have-PRS.1SG ONE hunger
‘I’m so hungry! ’
What the nouns in the examples cited have in common is that they are inherently
gradable, i.e. they possess an inherent scale in their lexical meaning. The questions which now arise are whether the scalarity of a noun indeed plays a role by
intensification and how the intensifier reading of the indefinites can be explained.
2. Scalarity of nouns
2.1. Analysis of scalar nouns
Before moving to the explanation of the intensifier reading of the indefinite
article, we should explore the scalarity of nouns in detail. In order to account for
the data, scalar nouns are understood similarly to gradable adjectives, as relating
entities to a scale, and can be treated in line with the analysis of gradable adjectives proposed by Kennedy (1999), Kennedy & McNally (2005). Their analysis
is based on two principal assumptions:
(i) “Gradable adjectives map their arguments onto abstract representations of
measurement, or degrees.”
(ii) “A set of degrees totally ordered with respect to some dimension (height, cost, etc.)
constitutes a scale.” (Kennedy 2007: 4)
The meaning of a gradable adjective can then be formalized as follows:
(16) ƐAGrAdjƑ = dx.g(x) d, where g is a function from objects to degrees d.
The positive form of an adjective is assumed to contain a phonologically unreal41
ized degree morpheme POS which maps a gradable adjective to a property that
indicates a relation to a context-dependent standard of comparison dStd:
(17) ƐDeg POSƑ = gxmd[g(d)(x) dStd]
The treatment of scalar nouns is closely parallel to the analysis of gradable adjectives, i.e. scalar nouns can be interpreted as denoting a relation between an individual and a set of degrees viewed as a scale. For instance, a predicate such as
fool relates an individual to the scale of “foolishness”. An unmodified scalar
noun is similar to the phonologically unmarked positive degree morpheme POS,
cf. (18):
(18) ޭ fool ޮ = xmd [FOOL(d,x) „d dStd]
The next point which should be considered is how scalar nouns can be distinguished from non-gradable ones. As a simple test suitable for this purpose we
can use the following sentence with bigger used as a degree modifier that can be
combined only with gradable NPs:
(19) I ’ve never seen a bigger egoist, opportunist, tramp, *teacher, *manager before.
A number of other diagnostic tests (cf. Bolinger 1972, Matushansky & Spector
2005) which can be applied are illustrated in table 1 below:
Table 1. Tests for noun scalarity5
paraphrasing with gradable adjectives
appearance with degree modifiers
usage in exclamatives with such5
appearance in the N of an N construction
ability to be used as epithets
appearance in the complement of seem
fool c a foolish person, teacherc n
He is a complete fool.
He is such a genius!
He is an idiot of a doctor.
I hate John. The idiot stole my money.
He seems a genius.
2.2. Degree modifiers of scalar nouns
Degree modifiers used with gradable nouns then scale the degree of the predicate
up or down:
(20) He is a complete fool. = The degree is higher than the standard degree.
(21) He is not such a fool you may think. = The intensity is scaled down.
As a counterargument for the claim that the ability of scalar nouns to occur unmodified in
such-exclamatives confirms their scalarity, one can remark that non-gradable nouns also appear
without an adjectival modification in this type of exclamative sentence, e.g. Such a teacher!
However, this fact doesn’t actually provide the counterevidence. In sentences like above, the scale
is covertly expressed and can be only recovered from the context, consider (i) cited in Grimm
(1987: 109):
(i) Das war ein (z. B. schöner, anstrengender, ereignisreicher,…)Tag!
‘That was a (e.g. good, hard, eventful,…) day!’
Kennedy (1999), Kennedy & McNally (2005) differentiate between three kinds
of degree modifiers of gradable adjectives: (i) true degree morphemes, (ii) scale
adjusters, and (iii) intensifiers. True degree morphemes (measure phrases, e.g.
10 feet, proportional modifiers completely, half, etc.) combine with a gradable
predicate in a head-specifier position and map the adjective (type µd,µe,t¶¶ – a relation between individuals and degrees) to a property of individuals (µe,t¶). As for
noun modifiers, we can identify, for instance, such morphemes as first degree,
schwer ‘seriously’ vs. leicht ‘slightly’ in compounds like first degree burn,
Leicht- / Schwerverwundeter ‘a slightly / seriously wounded person’ as measure
The next class of degree modifiers – scale adjusters of type µµd,µe,t¶¶,µd,µe,t¶¶¶
– establish an ordering relation between two degrees that is expressed by comparative morphology. Consider (22) and (23):
(22) Peter is more Italian than Paulo6.
(23) Bill is twice the mathematician his mother was.7
The largest group of degree modifiers of scalar nouns consists of intensifiers, the
most frequent of them are listed in table 2 below.
Table 2. Intensifiers
intensifying such, so
intensifiers twice, thrice
adjectives such as complete, utter, absolute
indefinites ein, edin‘ONE’, some
He is such an opportunist!
He is thrice a fool!
Affenhitze lit.‘monkey heat’, une faim de loup
an utter fool, a real success
He’s some fool!
J’ai une de ces faims! ‘I’m so hungry!’
Kennedy & McNally (2005) analyze intensifiers as predicate modifiers of type
µµe,t¶,µe,t¶¶ that are characterized by combining only with properties of individuals based on gradable adjectives. This means that the comparison class doesn’t
include all objects but only those that possess a property denoted by a gradable
predicate to the positive degree. When we take an example like Peter is an awful
fool, the comparison class consists not of all individuals but only of foolish people, that is, individuals who have the positive degree of ‘foolishness’.
Though all intensifiers function to reinforce the degree of a property, there is,
however, some difference between them. Thus, augmentatives or such intensifiers as real, awful, etc. boost the degree of the property, see (24):
(24) ޭ an awful fool ޮ = xmd[FOOL(d)(x) „ d š StdFOOL]
As shown in (22), degree modifiers can occasionally be used with non-scalar nouns. In this case we
are dealing with a coercion of a noun like Italian into a gradable predicate, which leads to the reinterpretation of the predicate meaning sentence where Italian doesn’t denote a nationality but rather the
typical features of representatives of the corresponding nation.
For discussion of this type of a sentence see Grosu & Krifka 2008.
The use of intensifiers like complete, total, and absolute with scalar nouns leads
to the interpretation of a scale as closed and indicates a maximum degree, which
can be formalized as follows:
(25) ޭ a complete fool ޮ = xmd[FOOL(d)(x) „ d = max (StdFOOL)]
The meaning of scalar nouns can be demonstrated with the help of the following
a fool
a big fool
an oversize fool a complete fool
sort of a fool
As we can see, some degree modifiers (e.g. big ) mark that the degree of the property exceeds a contextually appropriate standard degree whereas the use of other
groups of intensifiers yield a high or maximum degree reference (cf. awful, oversize, or complete, total, etc. respectively). In contrast to these degree boosters,
such modifiers as sort of, rather, etc. function as diminishers and indicate that
the value of the property is less than the standard that is achieved by pragmatic
restriction of their literal meaning (cf. Krifka 2005).
3. Intensification with ONE
3.1. Characteristics of the use of ONE
Having discussed noun scalarity and the functions of degree modifiers, we can
now turn to the core question of this paper: How is the intensification in sentences like (12)–(15), or (27) achieved?
(27) a. Ich habe Hunger.
I have hunger
‘I’m hungry.’
b. Ich habe einen Hunger!
I have ONE hunger
‘I’m so hungry!’
The role of the indefinite article in intensification has several essential characteristics. As can be seen in the examples below, ONE cannot be used in a question
(28a) and is also infelicitous as an answer to the question (28b). This use of the
indefinite article only occurs in exclamatives (29).
(28) a. Hast
du (*einen) Hunger?
have-PRS.2SG you ONE hunger
‘Are you hungry?’
b. A: Hast duHunger?
B: *Ich habe einen Hunger!
(29) Was für einen Hunger er hat!
‘What a hunger he has!’
Furthermore, the use of ONE for intensification has the following requirements.
First, the NP must be scalar. Second, the intensificational meaning must not be
blocked by the indefinite article in its regular use, otherwise it needs to be ex44
pressed explicitly (e.g. with intensifying particles). For instance, the intensifier
reading of ONE in a sentence like I have a friend can not be obtained due to the
use of ONE as a regular indefinite article. In this case we need to insert such to get
an intensifying interpretation (I have such a friend!). Third, the intensifier reading of ONE must be accompanied by an exclamative-like intonation (cf. (27)).
However, it should be pointed out that the exclamative prosody alone is not a sufficient condition for the intensifying interpretation of a sentence with an indefinite NP. The comparison of two examples like Ich habe Hunger! and Ich habe
einen Hunger!, both marked by an exclamative, makes that clear. Where in the
former case we just emphasize the fact that an individual is hungry, only the latter refers to the high degree of hunger, which supports the claim that the reference to the high degree can not be achieved by means of the prosody alone.
3.2. Analysis of intensification with ONE
In order to account for the data, the intensification in sentences considered in the
previous section is assumed to result from the interplay of three factors whose
contributions are necessary to achieve a high degree reference: (i) the use of the
indefinite article, (ii) the presence of a gradable predicate and (iii) the exclamative force of the sentence.
The indefinite article used with abstract nouns like hunger, thirst, heat, etc.
leads to a subkind interpretation of the corresponding NP, similar to the sort
reading of mass nouns caused by the insertion of the article, cf. Reis ‘rice’ vs. ein
Reis ‘one sort of rice’. Many kinds can be constructed as having subkinds, then
there is a counting criterion that makes them count nouns. The taxonomic relation has the following property (cf. Krifka 1995:74ff.):
(30) SUBKIND (k’, k)„R(x, k’) R(x, k) = if k is a subkind of kind k and the
realization relation R (x, k’) holds, i.e. the object x is an instance of k, then x is also
an instance of k.
The contribution of the indefinite article in sentences like Ich habe einen Hunger
is to make a taxonomic reading available. The indefinite NP refers to subkinds
of hunger (typically weaker and stronger subkinds) like “small hunger”, “regular
hunger”, “big hunger”, “extreme hunger”, i.e. to one natural taxonomy of hunger
according to the strength8.
Another point which plays an important role in achieving the intensifying
reading is that exclamatives are based on evaluation, i.e. they require an evaluative predicate, which is in turn inherently gradable. This requirement explains
then the infelicity of non-scalar nouns in such environments where the reference
to the high degree of the property has to be obtained. The scale for evaluation is
The availability of the subkind reading of abstract nouns is apparent in partitive constructions, cf.
the following examples: (i) J’ai une de ces faims! ‘I’m so hungry!’ or (ii) She was in one of those
provided explicitly by a scalar noun or in cases with a non-scalar noun by a
gradable adjective or implicitly by a phonologically unrealized adjective, cf.
Such a fool!, Such a good teacher! vs. Such a (good, bad) teacher! The requirement that the null adjective has to be gradable can be clarified by a comparison
of exclamatives with a degree operator like such or what that obligatorily takes
a scalar predicate as its argument. See (31):
(31) a. Such / What a (nice, bad) teacher! vs. b.*Such /* What a (Swiss, former)
Finally, the exclamation indicates the remarkability of the corresponding subkind denoted by the indefinite NP, i.e. a subkind must be remarkable in order to
cause an emotional reaction from the speaker. Thus the following generalization
is now possible:
(32) kk [k is subkind of k „ k k1 R (x, k1) >remarkable R (x, k)]
This means that the speaker utters an exclamative ‘x is k!’ where a postcopular
indefinite NP denotes a subkind k1 that is remarkable in comparison with other
subkinds of k, which evokes an emotional attitude from the speaker.
3.3. Evidence for the relationship between taxonomies and
Evidence in favour of the claim regarding the relationship between the taxonomic reference and the exclamatives is provided by the following observations.
First, as exemplified in (33), we can state that the deeper a taxonomy denoted by
an NP, the more infelicitous the appearance of this NP in exclamative expressions, due to the difficulty of deriving other, i.e. deeper, taxonomic hierarchies.
(33) a. DAS ist aber ein Marmortisch! vs. b. ??DAS ist aber ein runder Marmortisch!
‘That is some marble table!’
‘That is some round marble table!’
The reverse observation that kind denoting NP as opposed to taxonomic NPs is
infelicitous in exclamatives, in particular with non-scalar nouns, is demonstrated
in (34):
(34) a. ??Oh! A house!
b. Oh! A straw house!
We can hardly imagine that a person would express an emotional attitude towards a house like in (34a)9. In (34b) the aspect which is noticeable and evokes
the speaker’s emotions is the fact that the house is made of straw in contrast to
other types of houses.
4. Types of exclamatives
4.1. Fact vs. degree exclamatives
I propose to distinguish between two kinds of exclamatives: fact exclamatives
and degree exclamatives, which are exemplified in (35a–38a) and (35b–38b) respectively:
(35) a. Ich habe Hunger!
b. Ich habe einen Hunger!
(36) a. Peter e glupak!
‘Peter is a fool!’
b. Peter e edin glupak!
‘Peter is such a fool!’
(37) a. ??Das ist ein Mann!
‘That is a man!’
b. DAS ist ein Mann!
‘That is such a man!’
(38) a. Das ist ein IDIOT!
‘That is an idiot!’
b. DAS ist ein Idiot!
‘That is such an idiot!’
In (a)-sentences the speaker expresses an emotional attitude towards a fact (for
instance, the fact that someone is hungry, or someone is an idiot). In the case
where the predicate denotes a gradable property (hunger, foolishness, etc.) it
means that an individual has the corresponding property to a standard degree.
Predicates constructed of non-scalar nouns, e.g. (37a), are infelicitous under the
fact reading. In the absence of the blocking regular indefinite article (e.g. (35),
(36)), the insertion of ONE is sufficient to contribute to the intensifying reading
of a sentence, otherwise some additional means are needed in order to achieve
the high degree interpretation. Consider, for example (37b) and (38b), where the
specific exclamative intonation with a stress at the left periphery of the sentence
satisfies this function.
By way of contrast, in (b)-sentences we are dealing with emotional attitudes
towards a high degree of the property possessed by an individual. This reference
to a high degree can be achieved due to the use of the indefinite article, which
induces a taxonomic interpretation, due to the scalar property provided by a
gradable noun or by a phonologically unrealized gradable adjective (e.g. 42b)
and to the exclamative prosody that signals the remarkability of an NP.
Summing up, the intensification with ONE results from the interaction of three
factors: (i) indefinite article making a taxonomic interpretation available, (ii)
gradable predicate providing grade subkinds, and (iii) exclamation indicating a
high value subkind.
(34a) is interpretable when viewed as a realization of a situation which is remarkable in comparison
to other possible situations, cf., for instance, the following scenario: Someone walking in the wilderness suddenly sees a house, then he/she can exclaim: Oh! A house! However, in this case the attitude
is expressed not towards a house itself but towards an unexpected situation: finding a house in the
wilderness. This situation with a house is a remarkable one compared to other possible situations. We
are also dealing here with taxonomies derived from the situation and the gradable element which is
the expectedness which can be viewed as a scale with two poles: expected vs. unexpected.
4.2. Tests for fact and degree exclamatives
At this point I would like to propose some tests which clarify the distinction between two kinds of exclamatives (cf. (39a)–(42a) for fact exclamatives and
(39b)–(42b) for degree exclamatives below):
Test 1. Compatibility with the intensifier such
(39) a. Das ist SO ein IDIOT! vs.
‘That is such an idiot!’
b. DAS ist (*SO) ein Idiot!
As we can see, fact exclamatives can be combined with so ‘such’, which boosts
the degree of the property and yields a high degree reference. This is not so with
degree exclamatives. Because they already express the reference to a high degree
of the property, there is no need in such, whose addition leads then to a conflict
between different methods of the intensification, i.e. specific intonation and the
use of an intensifier. The same holds for Bulgarian where intensifying edin ‘ONE’
cannot appear with an intensifier takâv ‘such’.
Test 2. Compatibility with emphasizers
(40) a. Das ist (*aber / *vielleicht) ein IDIOT! vs. b. DAS ist aber / vielleicht ein Idiot!
‘That is an idiot!’
‘That is such an idiot!’
Degree exclamatives allow the combination with emphatic particles like aber,
auch, and vielleicht, frequently used in colloquial German, which emphasize
emotional attitudes of the speaker and additionally stress the presence of a high
degree reference. However, fact exclamatives display the incompatibility with
such emphasizers. These expressions can appear only in their primary function
as adverbs with a corresponding meaning, i.e. aber ‘but’, auch ‘also’, vielleicht
Test 3. Subject-verb inversion
(41) a. *Ist das ein IDIOT! vs.
b. IST das aber ein Idiot!
While (41a) is infelicitous, (42b) offers a common reversed word order pattern
in degree exclamatives.
Test 4. Negation
(42) a. Er ist kein IDIOT! vs. b. *ER ist aber kein Idiot!
Unlike the fact exclamative (42a), which can be easily negated (and the negated
counterpart can be expressed as an exclamative), the degree exclamative in (42b)
is not compatible with a negation because degree exclamatives which involve the
reference to the high degree of a property presuppose that the property holds for
its argument. This presupposition causes a conflict with the negation where the
presence of the corresponding property is denied. The application of this test to
Bulgarian examples yields the same result.
These tests can also be applied to other degree predicates, i.e. verbs and adjectives (cf. aufstehen ‘stand up’ vs. abnehmen ‘lose weight’, wooden, German vs.
big, nice), which, as a result, show the same differences between two kinds of
exclamatives. This fact supports the idea that the gradability of the property, disregarding by which category (nouns, adjectives, verbs) it is denoted, is a mandatory condition for the degree exclamatives.
5. Conclusion
In discussing the various functions of the indefinite article according to the
stages of its development, our main concern was to capture the data observed in
different languages where the article contributes to the intensification. It has
been argued that this use of an article has the following requirements: the scalarity of the denoted NP and the absence of the blocking regular indefinite determiner. The intensifying reading of a sentence results from the taxonomic reference of the indefinite NP, the gradability of a noun and the force of the exclamation, which indicates the remarkability of the subkind denoted by the NP.
Although the indefinite article cannot be considered on a par with regular intensifiers, its contribution is necessary in order to achieve the reference to the high
degree of the property in exclamative sentences. It should be pointed out that the
use of the indefinite article in the intensificational function has been observed
only in languages which belong to the second stage of article development (cf.
Bulgarian, Greek, German), i.e. the stage where the article can occur in generic
environments and, hence, provide the taxonomic reference. This fact supports
the claim that the intensificational force of the indefinite article is related to its
ability to refer to kinds.
Bolinger, D. 1972. Degree words. The Hague: Mouton.
Borer, H. 2005. Structuring sense: an exo-skeletal trilogy. Book I. In name only. Oxford:
Oxford University Press.
Givón T. 1978. Definiteness and referentiality. In Greenberg J., Ferguson C. & C.
Moravcsik (eds.): Universals of human language, Vol. IV: Syntax. Stanford. 192–230.
Grimm, H.-J. 1987. Lexikon zum Artikelgebrauch. Leipzig: Verlag Enzyklopädie.
Göksel, A. & C. Kerslake. 2005. Turkish: a comprehensive grammar. New York:
Grosu, A. & M. Krifka. 2008. The gifted mathematician that you claim to be. Equational
intensional ‘reconstruction’ relatives. Linguistics and Philosophy 30. 445–485.
Heine, B. 1997. Cognitive foundations of grammar. Oxford, New York: Oxford University Press.
Heine, B. & T. Kuteva. 2006. The changing languages of Europe. New York: Oxford
University Press.
Kennedy, C. 1999. Projecting the adjective. The syntax and semantics of gradability and
comparison. New York: Garland.
Kennedy, C. & L. McNally. 2005. The syntax and semantics of multiple degree modification in English. In S. Müller (ed.): Proceedings of the HPSG05. CSLI Publications.
Kennedy, C. 2007. Vagueness and grammar: the semantics of relative and absolute gradable adjectives. Linguistics and Philosophy 30. 1–45.
Krifka, M. 1995. Genericity: an introduction. In: Carlson C. & F. J. Pelletier (eds.):
The generic book. Chicago, London: The University of Chicago Press. 1–124.
Krifka, M. 2007. Negated antonyms: creating and filling the gap. In Sauerland, U. & P.
Stateva (eds.): Presupposition and implicature in compositional semantics. Palgrave
Studies in Pragmatics, Language and Cognition, Palgrave Macmillan. 163–177.
Matushansky, O. & B. Spector. 2005. Tinker, tailor, soldier, spy. In Maier E., Bary C. &
J. Huitink (eds.): Proceedings of SuB 9. 241–255.
Leiss E. 2000. Artikel und Aspekt. Die grammatischen Muster der Definitheit. Berlin: de
Presslich, M. 2000. Partitivität und Indefinitheit. Frankfurt / Main: Peter Lang.
Yeom, J.-I. 2007. Two specificity markers etten ‘some’ and han ‘one’ in Korean and their
implications. Paper presented at the Workshop “Funny Indefinites”. Berlin.
Elena Gorishneva
Centre for General Linguistics (ZAS)
Schützenstraße 18
D-10117 Berlin
[email protected]
Attribution in Basque, Finnish, Hungarian and
Turkish: Morphology vs. Syntax
Lutz Gunkel & Susan Schlotthauer
Institut für Deutsche Sprache, Mannheim
1. Introduction
In this paper we address the question of what is needed, in terms of morphosyntactic encoding, to relate a so-called verb-specific modifier to a nominal head.
For the purposes of this paper we shall assume that the notion of a verb-specific
modifier includes adverbs and their phrasal or clausal projections, adpositional
phrases, and noun phrases featuring a particular semantic case such as locative
or instrumental. Noun-specific modifiers, in turn, are considered to be first and
foremost adjectives and adjective phrases, next participles and their phrasal projections and, finally, relative clauses.1 The basic motivation underlying this
distinction relates to markedness. As pointed out by Croft (1991: Chap. 2) and
Hengeveld (1992: 37) it can be cross-linguistically observed that when, say, a
verb-specific modifier is used to modify a nominal head or when a noun-specific
modifier is used to modify a verbal head it must in some way be changed –
usually in terms of morphosyntax – before it could be used in the respective
marked case.
In the simplest kind of case, however, no such marker will be necessary. Thus,
for instance, in English or German we may use adverbs as adnominal modifiers
by simply juxtaposing them to their nominal head, cf. ENG the talk yesterday,
GER der Vortrag gestern. A second strategy consists in embedding a verbspecific modifier within a noun-specific modifier, viz. either a participial phrase
or a relative clause, cf. ENG the talk given yesterday, the talk that was given yesterday. As long as the language avails itself of participial phrases or adnominal
relative clauses, this option exists. Therefore it might not appear to qualify as a
true strategy of its own. However, at least with respect to participial phrases it
becomes less trivial when considering languages where the participle to be used
may be an auxiliary (or some other grammaticalized verb) as in, e.g., Hungarian
Whether or not there are additional categories for each type of modifier won’t matter for the purposes of this paper.
and Turkish. Note that the possibility of using auxiliaries as heads of participial
phrases is subject to language-specific constraints, being more or less prohibited
in German, and it can be considered to establish a more grammaticalized, formal
strategy to link verb-specific modifiers to nominal heads. In the third strategy,
then, it is merely a formal marker, henceforth called ‘relational marker (RM)’,
that attaches to the verb-specific modifier within an attributive construction.
Such RMs come in various shapes and with different semantic effects: Formally,
they may be words (der Vortrag von gestern ‘yesterday’s talk’), clitics (yesterday’s talk) or (derivational) affixes (der gestrige Vortrag ‘yesterday’s talk’).2
Semantically, they may preserve or change the semantic type of the modifier, being either instances of what is called ‘function-indicating’ or ‘type-changing’
morphosyntax in Croft (1991: 69). The central question we want to pursue in this
paper is whether or not the degree of bondedness between a RM and the expression it relates to the head correlates with the semantic type of the modifier it
In what follows we investigate four languages, i.e. Basque, Turkish, Hungarian and Finnish, where RMs figure prominently within the syntax of attribution.
Section 2 explores the range of phrasal categories a RM of the respective language may attach to. Section 3 deals with the syntactic scope of each RM, i.e.
whether it is to be analyzed as a clitic or as an affix. Finally, in section 4 we examine the range of semantic relations an attribute introduced by a RM (henceforth RM-attribute) may enter and enquire the semantic type of the RMattributes. The paper closes with a brief conclusion.
2. Relational marking in the languages under
Basque Basque possesses slots for both postnominal and prenominal modifiers.
Adjectives as a rule follow the noun they modify, though a few must precede it
(cf. Trask 2003: 138). Clausal modifiers also occur in postnominal position,
while all other types of modifier are restricted to the prenominal position, among
them possessive attributes and participial phrases, though this type of construction is said to be limited to eastern varieties of Basque (cf. Trask 2003: 142). Categories such as NPs (marked for semantic case, cf. (1)), PPs (cf. (2)), AdvPs (cf.
(3)) and adverbial participles (cf. (4)) have to be linked by the RM -ko in order
to function as adnominal modifiers. To a limited degree -ko also attaches to finite
clauses (cf. (5)).
We are not aware of RMs that are encoded non-segmentally, though those might well exist.
(1) mendieta-ko
mountains.LOC-RM the caves
‘the caves in the mountains’
(Trask 2003: 145)
(2) bakea-ren alde-ko amak
peace-GEN for-RM mothers
‘mothers for peace’
(de Rijk 1993: 148)
(3) atzo-ko
yesterday-RM the newspaper
‘yesterday’s newspaper’
(4) atzo
yesterday I.ERG buy.PRT-RM
‘the book I bought yesterday’
(5) hil dutela-ko
kill AUX.that-RM the report
‘the report that he has been killed’
(Trask 2003: 145)
the book
(Trask 2003: 146)
(Trask 2003: 147)
Turkish Except for finite relative clauses introduced by the subordinator ki – not
to be confused with the RM -ki –, adnominal modifiers precede the head noun
within the Turkish NP. These include adjectives, bare nouns, participial phrases,
genitive- and ablative-marked NPs3 and PPs headed by the postposition gibi
(‘like’; cf. Boeder & Schroeder 2000). In addition, there are RM-attributes, formed
by means of the RM -ki on the basis of NPs (cf. (6)), PPs (cf. (7)) and AdvPs (cf.
(8)).4 Note that those NPs and PPs must be marked for locative case (case suffix
-da/-de), which means that -ki can only attach to ‘possessive-marked postpositions’ as in (7). This type of postposition traces back to nouns, having the form
noun stem + possessive marker + case marker and forming a possessive construction with their complement (cf. Göksel & Kerslake 2005: 240ff.). They include the
central postpositions which express spatial relations. Consequently, PPs with ‘bare
postpositions’ as well as PPs headed by ‘possessive-marked postpositions’ that are
marked for a case other than locative are barred from forming RM-attributes.
(6) bahçede-ki
garden.LOC-RM trees
‘the trees in the garden’
(Göksel & Kerslake 2005: 196)
(7) Harunla
Harun.CONJ space.POSS1PL.LOC-RM tension
‘the tension between Harun and me/us’ (Göksel & Kerslake 2005: 259)
Those prenominal ablative-marked NPs exclusively denote partitive relations (cf. Boeder &
Schroeder 2000: 166).
AdvPs like the one in (8) are sometimes described as PPs (cf. Göksel & Kerslake 2005: 259), since
the adverb occurs with an optional complement.
(8) geri dönülerinden
back return.POSS3PL.ABL after-RM
‘the first year after their return’
(Schroeder 2000: 205)
Hungarian The pictures provided by Hungarian and Finnish are not as clear-cut
as in Turkish. Juxtaposing AdvPs, PPs or NPs marked for semantic case in postnominal position is not completely ruled out, but at least restricted by factors
such as style or the syntactic position of the entire NP within the clause. Relative
clauses occur postnominally as well, while all other types of modifiers, such as
adjectives, possessives, participial phrases and RM-attributes, have to precede
the head noun. RM-attributes are built on the basis of AdvPs (cf. (9)) and PPs (cf.
(10)) by using the RM -i, which is also employed to derive denominal adjectives.
Restrictions on the combination of RMs with postpositions will be dealt with in
sections 3 and 4.
(9) a ma-i
the today-RM
‘today’s meeting’
(10) az ablak
the window under-RM flower
‘the flower under the window’
(Laczkó 2000: 634)
Finnish As in Hungarian, the Finnish NP possesses a postnominal slot for relative clauses and allows NPs marked for semantic case and PPs to be simply
juxtaposed in postnominal position. Other types of modifiers (adjectives, genitive phrases, participial phrases and RM-attributes) occur prenominally. RMattributes are formed by attaching the RM -(i)nen to AdvPs (cf. (11)) and PPs (cf.
(13) & (14)). As in Hungarian, the RM also serves to derive adjectives from
nouns (cf. (12)).
(11) eili-nen
yesterday-RM TV programme
‘yesterday’s TV programme’
(12) joulu-inen
Christmas-RM mood
‘Christmas mood’
As in Hungarian and Turkish most postpositions can be traced back to nouns
(noun stems + case markers). Diachronically the prototypical Finnish PP is analyzed as a NP consisting of a head (P) and a genitive-marked complement (N).
Synchronically, those postpositions are considered to be fully grammaticalized.
Finnish postpositions for the most part belong to the category of adverb as well,
i.e. they can occur also without any complement. Adverbs also often trace back
to case-marked nouns, cf. lähellä ‘nearby’ < -lähi- ‘proximity’ + llA (casemarker for adessive case), where in current Finnish lähi- exists only as a bound
morpheme within compounds (e.g. lähitulevaisuus ‘the near future’). Thus no
clear line can be drawn between adverbs and postpositions. Note that this holds
for the most central class of postpositions, viz. the local ones. Now, when an adverb or a postposition combines with the RM -(i)nen, its case marker is ‘dropped’
and the RM attaches to the ‘bare’ noun stem.5 Appropriate examples and postpositions with optional and obligatory complements are shown in (13a, b) and
(14a, b).
(13a) taka-na
‘at the back, behind’
‘at the back, behind’
(13b) kulissien
taka-inen toiminta
scenes.GEN back-RM activity
‘activity behind the scenes’
(14a) viere-ssä
‘next to’
edge- RM
‘next do’
(14b) postiluukun
viere-inen roskakori
letterbox.GEN edge-RM
litter bin
‘the litter bin next to the letterbox’
(ISK 2005: 605)
3. Syntactic scope
In this section we pursue the question of whether the RM of the respective languages behaves more like a clitic or an affix. We test each RM according to the
following criteria: (i) whether it allows an unrestricted range of host categories
(i.e. categories of word forms it attaches to), (ii) whether it may occur only once
in coordinate structures (‘suspended marking’), (iii) whether its host may be inflected, (iv) whether it undergoes vowel harmony, given that the language has
vowel harmony at all and (v) whether the entire phrase it builds may be further
(i) An unrestricted range of host categories can be considered as indicating that
the marker is a clitic, since clitics should not be sensitive at all to the category of
its host. As we have seen in section 1, the broadest range is found in Basque
where the RM can attach to nouns, postpositions, adverbs, verbs and even adjectives. In Turkish, Hungarian and Finnish the range of categories is more limited,
as the RM may only occur in combination with nouns, adverbs and postpositions.
From a morphological perspective it is even more limited in Finnish, where adverbs and postpositions in combination with the RM correspond (morphologically) to ‘bare’ noun stems.
Note that the case marker may not be synchronically active anymore.
(ii) If suspended marking is allowed the marker behaves more like a clitic than
an affix. Conversely, if it is disallowed the marker should be regarded as an affix.
Suspended marking is generally prohibited in Finnish (cf. (15)) and Hungarian
(cf. (16)), but is obligatory in Basque (cf. (17)). In Turkish it is optional and common for a variety of markers, including case-markers, possessive markers, the
plural marker and the RM (cf. (18)).
(15) *kielen
language.GEN within and outside-RM(PL) reasons
‘intralinguistic and extralinguistic reasons’
(16) *e gondolat mellett és ellen-i
this thought
and against-RM arguments
‘arguments for and against this thought’
(17) gaztelaniatik
‘translations from Spanish into English’
(de Rijk 1993: 148)
(18) Almanya('da-ki)
ve Türkiye'de-ki
Germany.(LOC-RM) and Turkey.LOC-RM education
‘education in Germany and Turkey’
(iii) Since derivational affixes are barred from attaching to inflected stems, the
RM behaves like a clitic in case the host can be inflected. Again, we find a clear
distinction between Basque and Turkish on the one hand and Hungarian and
Finnish on the other. In Basque the host may be inflected for a whole range of
semantic cases (cf. (19) for the ablative case and the examples (1) and (17) given
above), while in Turkish the RM combines only with hosts marked for the locative case (cf. (20)). In Hungarian, the RM is restricted to uninflected hosts, including those that end in what was a local case affix in earlier stages of the language. For instance, in postpositions like mögött (‘behind’) the segment -tt corresponds to a former locative affix (cf. (21)). To put it more generally, the RM is
limited to adverbs and postpositions lacking any inflectional marker (of any
kind). This restriction excludes all not yet fully grammaticalized postpositions,
namely idiomatic participles taking a noun phrase complement (cf. (22)) or
case-marked nouns being the possessee of a possessive construction (cf. (23)).
Likewise, in Finnish the host the RM attaches to can not be inflected. Recall that
in Finnish adverbs and postpositions have do drop their case marker when combining with the RM (cf. section 2).
(19) Santurtziti-ko
the train
‘the train (coming) from Santurtzi’
(Eguzkitza 1993: 167)
(20) cam
glass.GEN side.POSS3SG.LOC-RM bed
‘the bed at the side of the window’
(Boeder & Schroeder 1998: 215)
(21) a polc
the shelf
behind-RM book
‘the book behind the shelf’
(Kenesei et al. 1998: 97)
(22) *a jövre
the future.SUB look.PRT-RM our plans
‘our plans concerning the future’
(23) *a feldolgozott információk alapján-i
the processed
information ground.POSS3SG.SUP-RM
‘the decision on the basis of the processed information’
(iv) As for vowel harmony, we consider absence of vowel harmony in a language that otherwise does show vowel harmony as an argument against treating
the RM in question as an affix. Now, in Basque there is no vowel harmony at all
and in Finnish and Hungarian the relevant vowels of the RMs (i and e, respectively) are not subject to vowel harmony. In Turkish, however, the RM does not
change according to vowel harmony, which is evidence for its status as a clitic
(cf. Almanya'da-ki vs. Türkiye'de-ki). There are some exceptions, where the
vowel (-i) changes to its rounded counterpart (-ü), e.g. dün (‘yesterday’), gün
(‘day’) and words containing gün such as bugün (‘today’; cf. Göksel & Kerslake
2005: 71). Underhill (1976: 212) mentions that in the spoken language we can
find a front-back harmony as well, cf. e.g. Ankaradak instead of Ankaradaki.
(v) We assume that only items marked by a derivational affix may be further derived. Further derivability is given in Finnish as well as in Hungarian. In both
languages abstract nouns can be derived from phrases including the RM, cf. (24)
and (25). In contrast, no such derivations can be found in Turkish and Basque.
(24) rautatieaseman
train station.GEN next to-RM-SUFADJ/NOM
‘the being next-to-the-train station’
The change of -in(en) into -is- is due to a regular morphophonological alternation.
(25) egymás
each other next to-RM-SUFADJ/NOM
Summing up, we found the clearest indication for clitic-hood with the RM in
Basque, followed by the RM in Turkish. As for Hungarian and Finnish, we have
seen that both markers behave like (derivational) affixes and we can safely treat
them as such. The following table summarized the results according to our test
criteria, with a plus sign pointing to clitic-hood, and a minus sign to affixhood.
Range of host categories
No affix suspension
No inflected host
Vowel harmony
Further derivability
not limited
N, Adv, P
N, Adv, P
N, Adv, P
4. Semantic type
Having examined the formal properties of the RMs in the two preceding sections
we move on to their semantic characteristics. To this end, we first explore the
range of semantic relations an RM-attribute may express. Second, we ask to what
extent the type(s) of meanings expressed by RM-attributes correspond to typical
adjectival meanings.
As for the range of semantic relations we concentrate on the basic local relations: locative, ablative, allative, thereby disregarding temporal or more abstract
relations. Furthermore, we restrict ourselves to those RM-attributes that are built
on the bases of NPs or PPs, thereby disregarding AdvP (among others).
In Basque, an RM-attribute may express each type of local relation, cf. (26)–
(26) LOC mendieta-ko
mountains.LOC-RM the caves
‘the caves in the mountains’
Santurtzi.ABL-RM the train
‘the train (coming) from Santurtzi’
the road
‘the road to Bilbao’
(Trask 2003: 145)
(Eguzkitza 1993: 167)
(de Rijk 1993: 148)
Since the Turkish RM -ki only combines with NPs and PPs that include a locative
case marker attributive NPs/PPs must express a locative meaning, cf. (29). To relate an ablative or allative phrase to a nominal head one has to use a participial
phrase (cf. (30), (31)), or a relative clause.
(29) LOC cam
glass.GEN side.POSS3SG.LOC-RM bed
‘the bed at the side of the window’ (Boeder & Schroeder 1998: 215)
cahillikten gelen
silliness.ABL coming a
‘a thing (coming) out of silliness’ (Boeder & Schroeder 2000: 191)
Kuzey Irak'ta-ki
north Iraq.LOC-RM Kurds.DAT going
‘the support for the Kurds in Northern Iraq’
(Boeder & Schroeder 2000: 191)
Interestingly, the Hungarian RM is less restricted than the Turkish one, since it
not only occurs with locative phrases (cf. (32)) but also with ablative phrases (cf.
(33)). Again, allative phrases have to be embedded within participial phrases (or
relative clauses) in order to serve as adnominal modifiers (cf. (34)).
(32) LOC a polc mögött-i
the shelf behind-RM book
‘the book behind the shelf’
(Kenesei et al. 1998: 97)
az íróasztal melll-i
know-all manner
the desk
from next to-RM
‘know-all manner from behind the desk’
a Budapest mellé
való megérkezés
the Budapest (to) near being arrival
‘the arrival near Budapest’
Finnish, in contrast, differs from Hungarian in allowing only locative phrases to
build RM-attributes (cf. (35)), while for both ablative and allative phrases one of
the other two strategies must be used, cf. (36) and (37).
(35) LOC hotellin
behind.RM parking lot
‘the parking lot behind the hotel’
from under coming water
‘water from under the earth’
pääkaupungin lähelle
(to) near to
‘a trip near to the capital’
To summarize, all four languages allow locative phrases to combine with the respective RM. Turkish and Finnish RM-attributes (that are built on NPs or PPs)
are restricted to the locative type. Hungarian also admits ablative phrases. In
Basque all three types of local attributes are possible.
In order to further examine the semantic properties of the RM-attributes we
are going to test them for three features that are characteristic of typical qualitative adjectives: gradability, intensification and predicative use.
In Hungarian grading of RM-attributes is completely ruled out, cf. (38). Examples of graded RM-attributes, which can occasionally be found, have an idiomatic reading, cf. (39).
(38) *a konyha mellettibb
the kitchen next to.RM.COMP room
lit. ‘the room more next to the kitchen’
(39) El sem lehet isten háta
PRV not
can God back.his behind.RM.COMP place.ACC
lit. ‘you cannot imagine a place more behind God’s back than
The same holds true for Finnish, where grading is possible with idiomatic or
even lexicalized RM-attributes as in example (40) involving the RM-attribute
kansainvälinen (‘international’, lit.: ‘between nations’).
(40) tämä kaupunki on kansainvälisempi7
kuin …
this city
is people.PL.between.RM.COMP than
‘this city is more international than }’
Idiomaticity is also crucial for intensification in Hungarian and Finnish, since
only RM-attributes with an idiomaticized meaning may combine with intensifiers, as shown by example (41) from Hungarian. However, neither grading
nor intensification seems to be possible with the corresponding RM-attributes
in Basque and Turkish, regardless of whether their meaning is literal or idiomatic.
Again, -n- alternates with -s- (see fn. 6).
(41) elég történelem eltt-i
quite history
before-RM conditions
‘quite prehistorical conditions’
The predicative use of RM-phrases is generally excluded in Basque, Turkish (cf.
(42)) and Hungarian (cf. (43)), but can occasionally be found in Finnish (cf.
(42) Kitap masan
üstünde. / *masan
n üstünde-ki.
book table.GEN
‘The book is on the table.’
(43) *Ez a
[ Európa mellett-i ] volt.
this the decision Europe for-RM
intended meaning: ‘This decision was in favour of Europe.’
(44) Tämä järjestö
on [ Unescon alainen ].
organization is Unesco.GEN under.RM
‘This organisation is subordinate to the Unesco.’
The following table presents the results of the tests for gradability, intensification
and predicative use. We have found no examples of RM-attributes in Basque and
Turkish that would pass these tests. As for Hungarian and Finnish, only RMattributes having an idiomatic meaning were found to be gradable and/or intensifiable, and the only example of an RM-phrase in predicative function, which
was attested in Finnish, has a metaphorical meaning.
Predicative use
5. Conclusion
As for the formal aspects of RMs we can conclude that the RM in Finnish and
Hungarian qualifies as a derivational affix, whereas the corresponding marker in
Turkish and Basque should be analyzed as a clitic. More importantly, while the
cases of Basque and Finnish seem more or less clear-cut, this is not the case for
Turkish and Hungarian. If we regard bondedness as a scalar concept with Basque
and Finnish marking the endpoints, Turkish and Hungarian will lie somewhere
in between, with Hungarian being located closer to the ‘affixal’ end and Turkish
between Basque and Hungarian.
With respect to the semantic characteristics of RM-attributes we have observed that those RM-attributes denoting directions, i.e. allatives and ablatives,
are restricted as compared to those denoting locations. While the RM-attributes
of all four languages were found to express locations, ablatives were attested
only in Basque and Hungarian and allatives only in Basque, where RMattributes cover a considerably larger semantic range than those of the other
Now recall that prototypical adjectives denote qualities, i.e. entities that are
persistent and gradable (cf. Croft 1991: 65). As we have seen, locations are not
gradable (and neither are directions), but since locations are at least persistent
– in contrast to directions – they are closer to the prototype than directions. Interestingly, in being barred from both predicative use and gradability ‘locational’ RM-attributes exhibit two central characteristics of so-called classifying
(alias relational) adjectives, such as presidential or departmental in expressions like presidential address and departmental issues, respectively. Therefore, it is safe to consider those RM-attributes as classifying adjectives that
qualify as adjectives in terms of form. These are the RM-attributes in Hungarian and Finnish.
Furthermore we have seen that some of those ‘RM-adjectives’ may shift
from denoting locations to denoting qualities when assuming an idiomatic or
at least metaphorical meaning. In this case they can be graded or even used
predicatively. This is also in line with their treatment as classifying adjectives
as these can usually be forcibly transformed into qualifying adjectives by being
graded or put into predicative position, cf., e.g., Her speech was very presidential.
Finally we address the question of the relation between form and function,
which essentially relates to Croft’s notions of function-indicating and
type-changing morphosyntax. As far as Basque and Turkish are concerned
there seems to be no indication that clitics should be able to change the semantic type of an expression they attach to. The clitics in Basque and Turkish act
merely as function indicators. They syntactically transform phrases of different
types into attributes but have no impact on their semantics. Conversely, derivational affixes do not necessarily change the semantic type of their base. The
derivational affixes from Hungarian and Finnish that are at issue here derive
classifying adjectives and with those the semantic type of the base is usually
considered to be preserved. On the other hand we recognized that the expression of locative relations by RM-adjectives was constrained in favour of locations. Since locations are the type of locative relation that comes closest to ‘adjectival meanings’ in terms of semantic type (see above), one may conclude
that the relevant affixal RMs, though not type-shifting, could at least be considered as ‘type-restricting’. That the Turkish RM, which we qualified as a
clitic, apparently shares this kind of ‘type-restricting’ force does not blur the
picture. Recall that we observed that the Turkish RM shares a number of formal properties with affixes and can by no means considered to be a prototypical clitic like the Basque RM.
6. References
Boeder, Winfried & Christoph Schroeder 1998. Attribution und sekundäre Prädikate im
Sprachvergleich: Deutsch, Englisch, Kurdisch, Georgisch, Türkisch. Sprachtypologie
und Universalienforschung (STUF) 51. 207–227.
Boeder, Winfried & Christoph Schroeder 2000. Relational coding in Georgian and
Turkish noun phrases: syntax, derivational morphology, and “linking” by means of
participles. Turkic languages 4. 153–204.
Croft, William 1991. Syntactic categories and grammatical relations: The cognitive organization of information. Chicago: University of Chicago Press.
De Rijk, Rudolf P. G. 1993. Basque hospitality and the suffix -ko. In José Ignacio Hualde
& Jon Ortiz de Urbina (eds.), Generative studies in Basque linguistics (Current Issues
in Linguistic Theory 105), 145–161. Amsterdam etc.: Benjamins.
Eguskitza, Andolin 1993. Adnominals in the grammar of Basque. In José Ignacio Hualde
& Jon Ortiz de Urbina (eds.), Generative studies in Basque linguistics (Current Issues
in Linguistic Theory 105), 163–187. Amsterdam etc.: Benjamins.
Göksel, Asl
& Celia Kerslake 2005. Turkish. A comprehensive grammar. London etc.:
Hengeveld, Kees 1992. Non-verbal predication. Theory, typology, diachrony (Functional
Grammar Series 15). Berlin & New York: Mouton de Gruyter.
ISK = Hakulinen, Auli et al. (eds.). 2005. Iso suomen kielioppi. 3. pain. (Suomalaisen
Kirjallisuuden Seuran toimituksia 950). Helsinki: Suomalaisen Kirjallisuuden Seura.
Kenesei, István, Robert Michael Vago & Anna Fenyvesi 1998. Hungarian. London
Laczkó, Tibor. 2000. Zárójelezési paradoxonok. In Ferenc Kiefer (ed.), Strukturális
magyar nyelvtan. 3. Morfológia, 619–651. Budapest: Akadémiai Kiadó.
Schroeder, Christoph 2000. Attribution in Turkish and the function of -ki. In Asl
& Celia Kerslake (eds.), Studies on Turkish and Turkic languages. Proceedings of the
Ninth International Conference on Turkish Linguistics, Lincoln College, Oxford,
August 12-14, 1998 (Turcologica 46), 205–216. Wiesbaden: Harrassowitz.
Trask, Larry 2003. Morphology. In José Ignacio Hualde & Jon Ortiz de Urbina (eds.), A
grammar of Basque (Mouton Grammar Library 26), 113–170. Berlin etc.: Mouton de
Underhill, Robert 1976. Turkish grammar. Cambridge, MA & London: MIT Press.
6. Abbreviations
7. Acknowledgements
We are grateful to the following people for helpful discussions regarding the
data: Mari Junkkari (Finnish), Andrea Kraus (Hungarian), Ezel Babur (Turkish)
and Latif Durlanik (Turkish). Special thanks are due to Renate Raffelsiefen for
discussions and to Bruce Straub for correcting our English.
8. Contact information
Lutz Gunkel & Susan Schlotthauer
Institut für Deutsche Sprache
R5, 6-13
D-68161 Mannheim
[email protected]
[email protected]
An analysis of the formation of the
Tajik vowel system
Shinji Ido
Tohoku University / University of Sydney
The present paper presents a novel analysis of the formation of the Tajik vowel
system. It proposes the hypothesis that the vowel shifts that took place in the formation of the Tajik vowel system constitute a push chain shift and argues that
Tajik-Uzbek language contact was involved in the chain shift.
0. Introduction
In the present paper, I propose the hypothesis that the vowel shifts that took place
in the formation of the vowel system of Tajik were a push chain shift and were
affected by the language contact between Tajik and Uzbek. For clarity, I decompose this hypothesis into the following three hypotheses:
Hypothesis 1: The vowel shifts that took place in the formation of the vowel system of Tajik / Northern Tajik dialects constitute a chain shift. (The vowel system
of standard Tajik is based on and identical with that of Northern Tajik dialects
(hereafter NTDs), hence the term ‘Tajik/NTDs’ in this statement. )
Hypothesis 2: The Tajik chain shift was a push shift.
Hypothesis 3: The Tajik chain shift was affected by the language contact between Tajik and Uzbek.
In the remainder of this paper, I will argue for the validity of these hypotheses.
My argument will draw on previous scholarship in Tajik and Uzbek dialectology
and the study of vowel shifting.
1. Evidence that supports Hypothesis 1
In this section, I introduce data that support Hypothesis 1, namely the hypothesis that a vowel chain shift took place in the formative period of the Tajik
vowel system.
1.1. Evidence 1
A chain shift in the formation of the Tajik vowel system would be manifest if we
contrast the vowel system of Early New Persian (hereafter ENP), which is the
predecessor of Tajik, with that of Tajik. Windfuhr (1990: 543–544) presents the
following chart1 and states that the Tajik vowel system was formed out of the
ENP vowel system ‘by the merger of the short and long high vowels and the
rounding of long a’.
Thus, for example, ENP rst ‘right, straight’ and rz ‘day’ became rost and r
respectively, in Tajik. (Windfuhr does not say what the value of is.) If we convert the correspondence between ENP vowels and Tajik vowels shown in (1) into
schematic vowel charts, replacing the symbols and o that Windfuhr uses for the
mid-central and mid-back vowels respectively with the symbols /ɵ/ and /ɔ/
(which better represent the actual phonemes of Tajik), the formation of the Tajik
vowel system would look as follows.
(2) Early New Persian
Northern Tajik dialects22 / Standard Tajik
A chain shift is evident in these schematized charts, which is the first reason to
assume that a chain shift occurred in the vowel system of Tajik/NTDs. Another
fact that is apparent in (2) is that the vowel shifts are in accordance with the principles of chain shifting that Labov proposes in Principles of linguistic change
(1994), to which we turn now.
This is the upper half of the chart that Windfuhr provides in page 543. The lower half in which Dari
and Persian vowels are shown is not copied here.
In Tajik linguistics, it is often assumed that there was a ‘classical’ stage in the development of Tajik
1.2. Evidence 2
The second piece of evidence that supports Hypothesis 1 comes from
Labov’s principles of vowel shifting. The movements of ENP // and //
shown in (2) are in line with the principles of chain shifting that Labov
(1994) proposes.
According to Labov, the raising of long vowels in chain shifts is a phenomenon attested in various (chiefly Indo-European) languages. Labov (1994:
116) accordingly posits the raising of long vowels as a principle of chain
shifting, calling it Principle I: ‘In chain shifts, long vowels rise’. Labov proposes a total of seven such principles of vowel shifting, the third of which,
Principle III, is also a principle of chain shifting: ‘in chain shifts, back
vowels move to the front’. These principles of chain shifting are established
on the chain shift data collected from various languages. In other words, the
raising of long vowels and fronting of back vowels are two of the typical
characteristics of chain shifts. The vowel shifts in Tajik/NTDs exhibit these
characteristics – in Tajik/NTDs, the long vowel // rose and the back vowel
// moved to the front. The agreement of the vowel shifting in NTDs with the
principles of chain shifting constitutes another piece of evidence that support
Hypothesis 1.
1.3. Evidence 3
The third piece of supporting evidence for Hypothesis 1 comes from Tajik
dialect groups other than NTDs. Tajik dialects are typically classified into
Northern, Southern, Central, and South-eastern dialects. Southern and Central Tajik dialects are spoken in areas that are adjacent to the areas where
NTDs are spoken. South-eastern Tajik dialects, which comprise dialects spoken in villages near the Panj river in Gorno-Badakhshan, are geographically
isolated from NTDs. Observe the following charts in which the vowel shifts
that took place in Southern, Central, and South-eastern Tajik dialects are
at which the vowel system had the following eight vowels (see, e.g., Eshniyozov (1977: 21) and Rastorgueva (1992: 3)).
The Tajik dialectological literature customarily uses this chart in describing the vowel system of
NTDs because in a limited number of Tajik words, /i/ and /u/ whose ENP counterparts are long
vowels are pronounced perceptively long in (at least some) NTDs. One such example is Bukharan
Tajik [zi«rak] ‘sharp-witted’ in which the /i/ is pronounced longer than the /i/ in [zirak] ‘earring’.
However, apparently these words do not constitute a minimal pair in young Bukharans’ Tajik; an informant tells the author of this paper that [zirak] (for ‘earring’) is an Uzbek word whose Bukharan
Tajik counterpart is halka ‘ring’ and that she does not use [zirak] in her speech in Bukharan Tajik.
This suggests that the phonological vowel length distinction between [zi«rak] and [zirak] exists not
in Bukharan Tajik but in (the Bukhara dialect of) Uzbek, in which there are both ziyrak ‘alert, attentive’ and zirak ‘earring’. Aside from this ‘minimal pair’, none of the twelve words that Kerimova
(1959: 6) lists as words in which there are long vowels has a word with which it forms a minimal pair
in the Bukharan Tajik of today.
schematized.3 (Broken lines indicate conditioned changes. The broken lines
that extend from the arrows in (5) indicate that the vowels moved/changed
further in certain environments.)
(3) Early New Persian
Southern Tajik dialects
(4) Early New Persian
Central Tajik dialects (West to Matcha)
Early New Persian
Central Tajik dialects (East to Matcha)
South-eastern Tajik dialects
(5) Early New Persian4
A glance at the charts in (2), (3), (4), and (5) reveals that the vowel shifts in
Southern, Central, and South-eastern Tajik dialects constitute chain shifts. This
means that chain shifting is not unique to NTDs but is widely observed in Tajik
in general. This supports (albeit in a rather indirect way) Hypothesis 1.
The description of the vowel systems here is based on Rastorgueva (1964) and Eshniyozov (1977).
Their descriptions of vowel systems are essentially phonological and hence the symbols that appear
in (3–5), some of which I replaced provisonally with IPA symbols, may not closely represent the
sound values of the phonemes.
Some of the vowel movements shown here are speculative. The chart below shows alternative
movements of ENP // in South-eastern Tajik dialects.
2. Evidence that supports Hypothesis 2
An examination of the vowel shifts schematized in (2), (3), (4), and (5) reveals
that ENP // was invariably raised in all the four dialect groups of Tajik.
(6) a. Northern Tajik dialects /
South-eastern Tajik dialects
Southern Tajik dialects /
Central Tajik dialects
As can be clearly observed in the charts in (6), ENP // was raised in chain shifts
in every major Tajik dialect group. This is despite the diversity that the vowel
systems of Tajik dialect groups exhibit today. This renders some plausibility to
the assumption that the raising of ENP // had taken place before the diversification of vowel systems in Tajik took place. It therefore seems reasonable (and explanatorily parsimonious) to assume that the raising of ENP // initiated different
chain shifts in different Tajik dialect groups. This assumption naturally supports
the validity of Hypothesis 2.
Some instrumental data also exist that seem to support Hypothesis 2. The
F1-F2 scatter plots presented in Appendix indicate that the vowel system of a
sixty-year-old speaker of Bukharan Tajik, which is a variety that is classified as
an NTD, /ɵ/ differs from that of an eighteen-year-old Bukharan Tajik speaker.
The sixty-year-old man’s /ɵ/ is markedly more back, and hence also much closer
to /ɔ/, than is eighteen-year-old girl’s /ɵ/ (see Appendix). On the other hand, the
position of /ɔ/ relative to those of /i/, /e/, /u/, and /a/ does not vary much between
them, which appears to allow the assumption that the raising of ENP // preceded
the fronting of ENP //. The position of /ɵ/ in the vowel system of the sixtyyear-old Bukharan may then be considered to represent the stage in the Tajik
chain shift where the raised ENP // (Tajik /ɔ/) pushes the mid-back vowel out
of its original position.
In this section, I presented two sets of data that support Hypothesis 2. I will
turn to Hypothesis 3 in the immediately following section.
3. Evidence that supports Hypothesis 3
The preceding sections argued for the validity of the hypothesis that the vowel
shifts that occurred in the Tajik vowel system (i.e. the vowel system of NTDs)
was a push chain shift. In this section, I argue that Tajik-Uzbek language contact
was involved in the chain shift in Tajik/NTDs.
The foremost reason to assume that Uzbek influenced the formation of the
vowel system of Tajik/NTDs is the correspondence between the vowel system
of Tajik/NTDs and that of the Uzbek dialects which are in intensive contact
with NTDs. Observe the vowel system of the Bukhara-Samarqand dialects of
Uzbek, which are the Uzbek dialects that are most intensively in contact with
Bukhara-Samarqand dialects of Uzbek
This vowel system is identical with the vowel system of NTDs (compare (2) with
(7)).5 It is unlikely that NTDs and the Bukhara-Samarqand dialects of Uzbek developed the same vowel system independently from each other.6 The unlikeliness derives from the following facts. Firstly, in Uzbek, the use of the vowel system in (7) is confined to the Bukhara-Samarqand dialects. Secondly, the use of
the vowel system in (2) is confined to NTDs. Thirdly, the Bukhara-Samarqand
dialects of Uzbek are in contact with NTDs. The confinement of the use of a
vowel system that is unique among the dialects of both Tajik and Uzbek specifically to the dialects that are in intensive Tajik-Uzbek contact is unlikely to occur
by chance. It is therefore possible to state with some certainty that Tajik-Uzbek
language contact exerted influence on the formation of the vowel system of
Another reason to assume the validity of Hypothesis 3 is the difference between (6a) and (6b). NTDs diverge from the Tajik dialect groups that are geographically adjacent to them in having moved ENP // not to high-position but
to the front. As we have seen in § 1.2, the movements of ENP vowels schematized in (6) are within the explanatory power of Labov’s Principle I and Principle
III. However, even if Labov’s principles are sound, as I assume in the present paper that they are, the question remains as to what determines the direction of
vowel shifting when multiple principles are in conflict with each other. The two
differing movements of // that are observed in (6) are a case in point – Principle
This is not to say that the vowels in NTDs and their corresponding vowels in the Bukhara-Samarqand dialects of Uzbek are phonetically identical, though each vowel phoneme in NTDs seems to be
within the allophonic variation of its corresponding vowel in the Bukhara-Samarqand dialects of Uzbek. Four of the vowel phonemes of the Bukhara dialect of Uzbek (spoken by Uzbek monolinguials)
have a large degree of allophonic variation, which is not surprising because they came into existence
as a result of mergers (see Mirzaev 1969: 28–33). Bobomurodov’s (1978: 13) assertion that the Tajik
mid-central vowel is more front than its Uzbek counterpart may have to be assessed in this context.
The argument here is based on the assumption that two language varieties in contact can move their
vowels so that their vowels coincide. Data presented in Bullock & Gerfen (2004) and Bond et al.
(2006) appear to support this assumption.
As it probably did on the formation of the vowel system of the Qorluq-Chigil-Uyg'ur dialects of
Uzbek. The Qorluq-Chigil-Uyg'ur dialects, which comprise the Bukhara-Samarqand dialects, constitute a dialect group in Reshetov’s classification of Uzbek dialects (see Reshetov & Shohabdurahmonov 1978: 36–42). See Cokun (2000: xxviii–xxxii), Rajabov (1996: 76–98), and Reshetov &
Shoabdurahmonov (1978: 29–42) for other classifications of Uzbek dialects.
I may raise ENP // (as it apparently did in (6b)) while Principle III can move it
to the front (as it apparently did in (6a)).
Tajik-Uzbek language contact can provide an answer to this question, because
the Bukhara-Samarqand dialects of Uzbek have the mid-central vowel towards
which ENP // moved in NTDs and with which the /ɵ/ of NTDs coincide. Thus,
it is possible to ascribe the movement of ENP // that is responsible for the difference between (6a) and (6b) to utilization of /ɵ/ in the Bukhara-Samarqand dialects of Uzbek, with which NTDs have been in contact for centuries. The ascription of the movement of ENP // in NTDs to the Uzbek influence is not novel.8
Perry (2005: 15) writes that the mid-central position of /ɵ/ in Tajik ‘is due to its
having merged with Uzbek /uį /’ (/uį / is the symbol that Perry uses for the midcentral vowel phoneme in Tajik).9 Thus, the existence of a mid-central vowel in
the Uzbek dialects that are in contact with NTDs can serve as another piece of
evidence that Tajik-Uzbek language contact had an impact on the formation of
the vowel system of Tajik/NTDs.10
4. Summary
In this paper, I presented data that support the hypothesis that the vowel shifts
that took place in the formation of the vowel system of Tajik constitute a push
chain shift and that the chain shift was affected by the language contact between
Tajik and Uzbek. In section 1, I interpreted vowel shifts that took place in Tajik
as a chain shift, drawing on existing descriptions of the vowel system of NTDs.
NTDs are the dialects that are known for (and partly defined by) their intensive contact with Uzbek.
On the other hand, the contact between Central Tajik dialects and Uzbek is limited. Contact with Uzbek is minimal in Southern Tajik dialects. Any linguistic feature that is found in NTDs but not in the
other Tajik dialects geographically adjacent to them is therefore a candidate for a contact-induced
Incidentally, this passage precedes the following bracketed note: ‘(orig. common Turkic vowels
/ö/ and /ü/)’, which he repeats in paraphrase in Perry (2006: 485). This information is incorrect.
Uzbek /ɵ/ (Perry’s /uį/) corresponds to Turkic /ö/ and /o/ (see, e.g., Reshetov and Shoabdurahmonov
1978: 45).
One apparent obstacle to the ascription of the fronting of ENP // in Tajik/NTDs to Tajik-Uzbek
language contact is the fronting of ENP // in South-eastern Tajik dialects, which are geographically
isolated from Uzbek (though ENP // was, in certain environments, raised to high-position via
mid-position in South-eastern Tajik dialects). I do not have a ready explanation for the fronting of
ENP // in South-eastern Tajik dialects. However, considering the fact that they are dialects that are
in contact with the Pamir languages some of which (particularly Šughn, which is the largest among
the Pamir languages in terms of number of speakers) utilize mid-central vowels and also the fact that
at least some parts of the area in which South-eastern Tajik dialects are spoken today were areas
where the Pamir languages were spoken, it is possible that the contact between Pamir language and
Tajik played a role in the fronting of ENP // in South-eastern Tajik dialects. Payne writes that Wan
was ‘formerly spoken in the Van valley, but by the 1920s at the latest replaced by a dialect of Tik’
and that ‘a Pmir languge might once have been spoken … in the Darvaz and Karategin areas of
Tadžikistan’ (Payne 1989: 420). See in Édel'man (1987a: 238–240 and 1987b: 350–351) the vowel
systems of the Pamir languages that are geographically close to South-eastern Tajik dialects (See also
Wurm & Lee-Smith 1996).
In section 2, I analyzed different chain shifts that took place in various Tajik dialects and argued for the hypothesis that the Northern Tajik chain shift was a push
chain shift. In section 3, I used Uzbek dialectological data to show that TajikUzbek language contact is likely to have been responsible in moving ENP // to
the front.
One novelty of the analysis of the Tajik vowel system made in this paper lies
in its asserting that at least some instances of vowel shifts in Tajik are chain
shifts. To my knowledge, no previous work in Iranian linguistics exists that identifies a chain shift in the formation of the Tajik vowel system. While I posit a
chain shift in the formation of the Tajik vowel system, I also argue for the validity of Hypothesis 3. One result of this approach is the identification of the formation of the Tajik vowel system as a ‘joint product’ of internally motivated language change and externally driven language change. Much work on chain shifting has focused on chain shifting as a language-internal phenomenon (see, e.g.
Langstrof 2006, Maclagan & Hay 2007, and Labov 1994). As a result, limited
attention has been paid to ‘external’ factors in chain shifting such as language
contact.11 The analysis of the formation of the Tajik vowel system that I have
proposed in this paper indicates that including language contact as a factor in the
analyses of chain shifting can increase their explanatory, if not predictive, power.
Bobomurodov, Shakar. 1978. Sadonoki “” va mavqei on dar sistemai vokalizmi zaboni
adabii tojik. Dushanbe: Universiteti Davlatii Tojikiston ba nomi V. I. Lenin.
Bond, Z. S. et al. 2006. Sixty years of bilingualism affects the pronunciation of Latvian
vowels. Language variation and change 18, 165–177.
Bullock, Barbara & Chip Gerfen. 2004. Phonological convergence in a contracting
language variety. Bilingualism: Language and cognition 7(2), 95–104.
Cokun, Volkan. 2000. Özbek Türkçesi Grameri. Ankara: Türk Dil Kurumu.
Édel'man, D. I. 1987a. Šugnano-rušanskaja jazykovaja gruppa. IN V. S. Rastorgueva ed.
1987. Osnovy iranskogo jazykoznanija (Novoiranskie jazyki: Vostonaja gruppa).
Moskva: Nauka. 236–347.
Édel'man, D. I. 1987b. Jazguljamskii jazyk. IN Rastorgueva, V. S. ed. 1987. Osnovy iranskogo jazykoznanija (Novoiranskie jazyki: Vostonaja gruppa). Moskva: Nauka. 348–
Éshniyozov, M. 1977. Dialektologiyai tojik (Dasturi Ta"lim) Qismi yakum. Dushanbe:
Universiteti Davlatii Tojikiston ba nomi V. I. Lenin.
Kerimova, A. A. 1959. Govor tadjikov buxary. Moskva: Izdatel'stvo Vostonoy Literatury.
Labov, Willaim. 1994. Principles of linguistic change Volume 1: Internal factors.
Oxford: Blackwell.
Langstrof, Christian. 2006. Acoustic evidence for a push-chain shift in the intermediate
period of New Zealand English. Language variation and change 18, 141–164.
In fact, Labov (1994: 233) also makes use of an ‘external’ factor in accounting for the Valais raising of back vowels.
Maclagan, Margaret & Jennifer Hay. 2007. Getting fed up with our feet: Contrast maintenance and the New Zealand English “short” front vowel shift. Language variation
and change 19, 1–25.
Mirzaev, M. M. 1969. O'zbek tilining Buxoro gruppa shevalari. Toshkent: Fan.
Payne, John. 1989. Pamir languages IN Schmitt, Rüdiger ed. 1989. Compendium
Linguarum Iranicarum. Wiesbaden: Dr. Ludwig Reichert Verlag. 417–444.
Perry, John. 2006. Tajik Persian. IN Keith Brown et al. (eds.) 2006. Encyclopedia of
Language and Linguistics 2nd Edition. Elsevier. 484–487.
Perry, John. 2005. A Tajik Persian reference grammar. Leiden: Brill.
Polivanov, E. D. 1933. Uzbekskie govory i uzbekskiy literaturnyi jazyk. Tashkent: Gosizdat.
Rajabov, Nazar. 1996. O'zbek shevashunosligi. Toshkent: O'qituvchi.
Rastorgueva, V. S. 1964. Opyt sravnitel'nogo izucheniya tadjikskix govorov. Moskva:
Izdatel'stvo Nauka.
Rastorgueva, V. S. 1992. A short sketch of Tajik grammar (translated by Herbert H.
Paper). Bloomington: Indiana University.
Reshetov, V. V. & Shonazar Shoabdurahmonov. 1978. O'zbek dialektologiyasi. Toshkent: O'qituvchi.
Shoabdurahmonov, Shonazar. 1971. O'zbek shevalari va ularni o'rganish haqida umumiy
ma"lumot. IN Shonazar Shohabrahmonov (ed.) 1971. O'zbek xalq shevalari lug'ati.
Toshkent: Fan. 386–407.
Windfuhr, Gernot. 1990. Persian. IN Bernard Comrie (ed.) 1990. The world’s major
languages. London: Croom Helm. 523–546.
Wurm, S. A. & M. W. Lee-Smith. (comp.) 1996. Languages in the Pamir and in Yunnan.
IN Stepphen A. Wurm, Peter Mühlhäusler, and Darrell T. Tryon eds. 1996. Atlas of
Languages of Intercultural Communication in the Pacific, Asia, and the Americas
(Volume 1: Maps). Berlin: Walter de Gruyter. map 99.
I would like to thank Alan Libert for his comments on a draft of this paper. Financial support for this research came in part from MEXT Grant-in-Aid for
Young Scientists (B) (20720101) and the Matsushita International Foundation.
Contact information
Shinji Ido
Research Center for Language, Brain and Cognition
Tohoku University
41 Kawauchi, Aoba-ku, Sendai
[email protected]
Appendix: F1-F2 scatter plots of Bukharan Tajik vowels
pronounced in isolation
iiii i
iiiiiiiiii iiiiiiii iiiiiiiiiiii iiiii i
iiii iii iiii i i
iiiiiii i iiiiiiiii iiii i ii i
iiiiiii i iiiii i i i i
eee e
uuuuuuuu u uuuuu
u uuuuuuu
F1 (Hz)
e eee
ee ee
eeeeee e eeee eeeee
ee e eeee
e eeee
eeeee e ee
eeee eeeeee eeeeeee eeeeeeeeeeeeeeeee
eeeee ee eeee
e eeee
eee e e e
ööööööööööööööööö öö
ö öööööööööööööööööö
ö öööööööööööööööö ö
öö öö ööö öö
öööö ööö öööö ö
ö ööööö
öööööööööööööö öö
öööööööööööööööö ö
öööö öööööö ööö
ö ööö
ööö ö ööö öö
ööö öö ö
öööö ööö
a a
a a
F2 (Hz)
60 yo male Tajik speaker (recorded in Bukhara on 07 August 2006)
F1 (Hz)
iii i
ii iii iiiiiiiii
uu uuu
uuu uu
uuu uu
e eeee
eeee e
eeee e
F2 (Hz)
18 yo female Tajik speaker (recorded in Bukhara on 06 June 2007)
Valency-changing categories in Indo-Aryan
and Indo-European:
A diachronic typological portrait
of Vedic Sanskrit
Leonid Kulikov
Leiden University
1. Preliminaries: An approach to the diachronic
typological study of a linguistic category
This paper concentrates on the diachronic aspects of the typology of transitivity
oppositions and valency-changing categories, focusing on evidence available
from one branch of Indo-European, Indo-Aryan. It also aims to draw attention to
the regrettable imbalance of the synchronic and diachronic typological studies.
On the one hand, we dispose of rich catalogues and a detailed synchronic
analysis of the systems of valency-changing derivations attested in the languages
of the world. On the other hand, a systematic treatment of these categories in a
diachronic perspective is lacking. The rise, development and decline of these
categories mostly remain on the periphery of the typological interests.
It seems advisable to start a diachronic typological research with collecting
evidence from languages (language groups) with a history well-documented in
texts for a sufficiently long period of time (around 1000 years or more). When
approaching the history of a particular valency-changing category, such as passive or causative, it might be useful to outline some kind of diachronic typological portrait of the relevant category in the given language group or family,
tracing it from the earliest attested texts in an ancient language (L0) onwards up
to its reflexes in the daughter languages (L1, L2 etc.). Of particular interest would
also be – if available – evidence from the sister languages of L0, which can serve
as a basis for a tentative reconstruction of the hypothetical history and possible
sources of the category under study in the proto-language.
2. An example of a ‘family portrait’: the case of
One of the best objects for such a diachronic typological study would be, for instance, the Indo-Aryan group of the Indo-European language family. We dispose
of an uninterrupted documented history of Indo-Aryan for a period of more than
3.000 years, starting with the Old Indo-Aryan (OIA), which can be roughly identified with (Vedic) Sanskrit,1 and continued in Middle Indo-Aryan (Pli and
Prakrits) and New Indo-Aryan (Hindi-Urdu, Bengali, Marathi, Sinhalese, etc.).
Thus, in the case of Indo-Aryan, we dispose of rich material for a diachronic
analysis of the valency-changing categories. On the one hand, the rich evidence
collected by the Indo-European comparative linguistics creates a good basis for
hypotheses about the origin and possible sources of the morphological and syntactic categories attested in OIA and thus provides important material for a retrospective diachronic typological study. On the other hand, evidence from late
Vedic and Middle Indo-Aryan texts, as well as from New Indo-Aryan languages,
allows for a prospective diachronic study (how the OIA categories develop into
their reflexes in Middle and New Indo-Aryan). In what follows, I will offer an
overview of several features of the Indo-Aryan, and, particularly, of OIA system
of voices and valency-changing categories, which are relevant in a diachronic typological perspective. The main tendencies which determine the evolution of the
Vedic (OIA) system of transitivity oppositions include: (i) decline of the middle
diathesis, which, as I will argue, amounts to its degrammaticalization; (ii) the
rapid growth of new valency-changing categories, passives and causatives; and
(iii) decline of the labile patterning.
3. Degrammaticalization of the middle
The diathesis, or the active/middle opposition, is a grammatical category of the
Ancient Indo-European verb that surfaces in the type of the verbal personal inflexion. Cf. the present tense Vedic active endings 2sg. -si, 3sg. -ti etc. as opposed to middle endings 2sg. -se, 3sg. -te, etc.
The middle diathesis (also called ‘middle voice’) is usually said to function as
a syncretic marker of several intransitive derivations: passive, anticausative (decausative), reflexive, reciprocal; see examples below. This might indeed be the
case in Proto-Indo-European. However, one of the oldest documented Indo-European languages, Vedic Sanskrit, seems to attest the decay of the original system. Already in the language of the earliest texts, ɰgveda (RV) and Atharvaveda
(AV), we observe the loss of several grammatical functions of the ancient IndoEuropean middle; many of them are taken over by special markers. The only
The most ancient Vedic text, the ɰgveda, dates to the 2nd half of the second millennium B.C. For
the chronology of Vedic texts, see Witzel 1995: 96ff. (with bibl.).
function of the middle which is still quite productive in Vedic is the expression
of the self-beneficent (or autobenefactive) meaning. Instead, Indo-Aryan attests
the rapid growth of new valency-changing categories, foremost in the present
tense system: passives with the suffix -yá- and causatives with the suffix -áya-.
Let us take a closer look at the main alleged functions of the middle.
3.1. Passive
Within the three main tense systems, present, aorist, and perfect, passive is expressed by characterized formations, rather than by non-characterized (bare)
middle forms: (i) in the system of present: by present passives with the accented
suffix -yá-2 (e.g. yuj ‘yoke, join’: 3sg. yujyáte ‘is (being) yoked, joined’, 3pl.
yujyánte ‘are (being) yoked, joined’, participle yujyámna-, etc.); (ii) in the system of aorist: by medio-passive aorists in -i and -ran (-ram) (3sg. in -i, 3pl. in
-ran/-ram; e.g. yuj ‘yoke, join’: 3sg. áyoji, 3pl. ayujran); and (iii) by statives in
-e and -re, which supply passives in the system of perfect (3sg. in -e, 3pl. in -re:
e.g. hi ‘impel’: 3sg. hinvé ‘(it) is / has been impelled’, 3pl. hinviré ‘(they) are /
have been impelled’); for details, see Kümmel 1996; Got 1997. Both mediopassive i-aorists and statives have a defective paradigm.
The system of passive formations attested in early Vedic, first of all in the language of the RV, is schematically represented in Table 1. According to the communis opinio, alongside with characterized passive formations (YA-presents,
I-aorists and statives), there is a plethora of non-characterized middle forms in
all the three tense systems that allegedly function as passives (the shadowed column in the midst of the table).
Table 1. Passive in Old Indo-Aryan: traditional view
aorists in -i/-ran
statives in -e/-re
Below I will argue that non-characterized (bare) middle forms are extremely rare
in passive usages. There are indeed two large groups of non-characterized
middle formations (which I will call ‘bare middles’) employed in passive usages,
middle perfects and middle athematic participles with the suffix -na-. In fact,
Finite verbal forms are normally unaccented except when appearing in a subordinate clause and/or
at the beginning of a sentence or metrical unit (pda), i.e. a verse which forms the minimal constituent
of a stanza.
however, these forms have special paradigmatic status, being morphologically
(grammatically) ambiguous and therefore should be discarded as evidence for
the passive function of the middle voice.
Athematic middle participles with the suffix -na- exhibit unusual syntactic
properties in early Vedic, particularly in the language of the ɰgveda. While the
corresponding finite forms are employed only transitively, the -na-participles
are attested both in transitive and intransitive (passive) constructions (see already
Delbrück 1888: 264).
For instance, the participle hinvná- (root hi ‘impel’), taken by all grammars
as the middle participle of the nasal present with the suffix -nó-/-nu- (class V in
the Indian tradition), occurs 18 times in intransitive (passive) constructions (as
in (1a)), and 10 times in transitive constructions (as in (1b)) in the ɰgveda):
a. (RV 9.12.8)
‘Soma, being impelled, flows.’
b. (RV 9.97.32)
… índrya
pavase …
‘You (sc. Soma) purify yourself for Indra, impelling (your) speech with the (religious)
thoughts of the poets.’
By contrast, the finite middle forms made from the same stem ( hinváte
etc.), with which hinvná- is supposed to belong together can only be employed
transitively, meaning ‘to impel’, as in (2):
(RV 9.65.11)
impel-PRES-1SG.MED price:LOC.PL runner:ACC.SG
‘I spur on this runner [in the race] for prices.’
Likewise, the participle yujná- (root yuj ‘yoke’) occurs 8 times in intransitive
(passive) constructions (as in (3a)) and 14 times in transitive constructions (as in
(3b)) in the ɰgveda:
a. (RV 6.34.2c)
ná mahé
chariot:NOM.SG like great:DAT power:DAT
‘… like a chariot yoked for the great power.’
b. (RV 6.47.19a)
‘... (TvaƪƬar,) yoking two fallow [horses] to the chariot.’
-ó is the same ending as in yujn-á in (3a), resulting from the sandhi before a voiced consonant
(-á h- -ó h-).
Vedic grammars treat yujná- as a middle participle of the root aorist (see, for
instance, Whitney 1885: 132; Macdonell 1910: 370). However, again, as in the
case of hinvná-, the corresponding finite forms (3sg. áyukta etc.) can only be
employed in transitive usages, as in (4):
(RV 7.60.3)
saptá haríta
AUG-yoke:AOR-3SG.MED seven fallow:ACC.PL
‘He yoked (now) his seven fallow (horses).’
Elsewhere I have demonstrated (Kulikov 2006) that the grammatical characteristics of such passive -na-participles should be reconsidered. In my view, these
participles are homonymous, or morphologically (grammatically) ambiguous.
Thus, the participle hinvná- in its transitive usages, meaning ‘impelling’, belongs to the paradigm of the transitive nasal present (hinváte etc.). But it is a
member of the paradigm of the stative = a stative participle (3sg. hinvé, 3pl.
hinviré) when employed intransitively (passively), meaning ‘impelled’. Likewise, yujná- is a member of the paradigm of the (transitive) root aorist (áyukta
etc.) when employed transitively (‘yoking’), but it is a member of the paradigm
of the passive aorist (3sg. áyoji, 3pl. ayujran), that is, a passive aorist participle
when employed in passive constructions (‘yoked’):
(i) hi ‘impel’
(ii) yuj ‘yoke’
3pl. hinv-áte
sg. hinv-é
3sg. á-yuk-ta
3sg. á-yoj-i
Although, traditionally, Vedic grammars do not include participles into the paradigms of statives and medio-passive aorists, the assumption that passive
-na-participles should be added to these paradigms seems quite attractive, since
it easily explains their abnormal syntax.
Another large group of non-characterized middle forms employed in passive
constructions consists of middle perfects. Most remarkably, only 3sg. and 3pl.
middle perfects forms (with the endings -e and -re, respectively) are attested in
passive usages.4 In my view, all such forms should be taken as statives built on
perfect stems, rather than as middle perfects proper.
For instance, the form dadhé (root dh ‘put’) should be taken as a 3sg.form of
the middle perfect when meaning ‘has put’, as in (5a), and as 3sg. of the stative
when meaning ‘is put / has been put’, as in (5b):
For a detailed study of Vedic perfects, see Kümmel 2000.
a. (RV 9.18.4)
yó vívni vry
who all
desirable:ACC goods:ACC hand:LOC.DU put:PF-3SG.MED
‘The one who holds / has put all desirable goods in his hands ...’
b. (RV 1.168.3)
ca sá
hand:LOC.PL brooch:NOM.SG and sward:NOM.SG and together
‘Brooch and sward is put in [your] hands.’
Likewise, the 3pl. form yuyujré is middle perfect of yuj ‘yoke’ when employed
transitively, as in (6a), but stative when employed passively, as in (6b):
a. (RV 5.58.7)
wind:ACC.PL since horse:ACC.PL shaft:LOC.SG PREV-yoke:PF-3PL.MED
‘Since [the Maruts] have yoked the winds as their horses into the shaft …’
b. (RV 1.168.3)
thought:INS.SG yoke:STAT-3PL.MED drop:NOM.PL
‘The [Soma-]drops have been yoked with a religious thought.’
The same holds true, mutatis mutandis, for middle participles made from perfect
stems and employed in passive usages. Such forms should be taken as belonging
with statives rather than with middle perfects, as in the compound yuyujnásapti- ‘with yoked horses’:
(RV 6.62.4)
‘[these two Avins] which have yoked horses’
To conclude this short discussion of the passive paradigm, let it be mentioned
that the sub-paradigm of present is in fact defective, too. We mostly find 3sg. and
3pl. forms of the present tense, as well as participles. Next to present tense forms
proper, there are rare imperatives (some 10 forms in the RV and AV). Only exceptional attestations of other tense-moods are found, which makes the subparadigm of present much more similar to those of the aorist and perfect.
The early Vedic passive paradigm (as attested for yuj ‘yoke’ and su ‘press
(out)’) is summarized in Table 2. Different types of shadowing show the status
of the corresponding forms: dark grey = lacking and morphologically impossible; middle grey = morphologically possible but unattested or only exceptionally attested (underdeveloped part of the paradigm); light grey = morphologically
possible but rare.
-ra is the same ending as in yuyuj-ré in (6a), with a resulting from the sandhi before a vowel
(-e i- -a i-).
Table 2: Passive paradigm in early Vedic
… , yujyáse
1 [-panymahe (?)]
… , yujyante
ásvi, áyoji … , yoji
sunvé, yuyujé
[badhyantm] … , áyujran
syámna-, yujyámna-
s vná-, yujná-
sunvire, yuyujré
sunvná-, yuyujná-
Most importantly, the system almost exclusively consists of characterized forms.
There are only exceptional and isolated non-characterized (bare) middle forms.6
Thus, the middle diathesis cannot be said to serve as the marker of the passive
3.2. Reflexive
The reflexive is another valency-decreasing (intransitivizing) derivation traditionally associated with the middle diathesis. There are indeed some doubtless
instances of the reflexive usage of the middle forms (see Gonda 1979: 50), as in
(RV 2.33.9)
adorn:PERF-3SG.MED golden.decoration:INS.PL
‘[Rudra] has adorned himself with golden decorations.’
(RV 1.36.16)
áty aktúbhir
who:NOM.SG.M mortal:NOM.SG sharpen:PRES-3SG.MED by night
‘The mortal who sharpens himself by night …’ ( who is too nimble …)
Such examples are relatively few, however. In many cases the term ‘reflexive’ is
misleading. In fact, most occurrences of middle forms that are traditionally
called ‘reflexives’, should rather be qualified as anticausatives, cf. pryate ‘becomes full’ (not ‘fills oneself’), pávate ‘becomes clean’ (not ‘purifies oneself’!),
These include, for instance, class IX pres. gté ‘is praised’ or class I pres. stávate ‘is praised’.
stávate and gté are likely to be based on the stems of the statives stáve (see Narten 1969) and ge
‘is praised’, instantiating a sort of back derivation (Rückbildungen). A few sigmatic aorists (mostly
3pl. forms): ayukata ‘(they) were yoked’, adkata ‘(they) were seen, visible, (they) appeared’,
askata ‘(they) were set free’ must be replacements of the medio-passive 3pl.aorists in -ran, which
disappear after the RV.
Furthermore, several non-passive intransitives which may go back to true reflexives, exhibit idiomatic semantic changes, cf. ap ‘curse’: ápate ‘swears’
( *‘curses oneself’); ‘sharpen’: íte ‘is too nimble’ ( *‘sharpens himself’).
The productive markers of reflexive are two pronouns of substantive origin:
tan-, originally meaning ‘body’ (cf. (10)) – in early Vedic; and tmán- (‘breath,
soul’) – in later texts (cf. (11)) (see Kulikov 2007a for details):
(10) (RV 1.147.2)
praiser:NOM.SG your self:ACC.SG praise:PRES:1SG.MED Agni:VOC.SG
‘As your praiser, I praise myself, o Agni.’
(11) (MS 1.6.4:93.3)
tmnam evá
gold:ACC.SG give:PRES:3SG.ACT self:ACC.SG
‘He gives gold; thereby he purifies himself.’
thereby purify:PRES:3SG.MED
3.3. Reciprocal
Again, as in the case of passive or reflexive, the regular markers of reciprocity
include several morphemes which typically (but not always) co-occur with the
middle type of inflexion (see Kulikov 2007b for details): preverbs sám ‘together’
and ví ‘asunder’ as well as the adverb mithás ‘mutually’ and reciprocal pronoun
anyó-(a)nyám (lit. ‘another-another’). Cf. (12), where two of these markers are
(12) (AV 3.30.4)
ná vi-y-ánti
ná‰7 u ca vi-dviŦ-áte
which:INS.SG god:NOM.PL not vi-go:PRES-3PL.ACT not and vi-hate:PRES-3PL.MED
mitháŪ tát k-mo
vo ghé
mutually that make:PRES-1PL.ACT incantation:ACC.SG your house:LOC.SG
‘We perform in your house that incantation by virtue of which the gods do not go apart, do
not hate one another (mutually).’
The reciprocal adverb mithás ‘mutually’ is particularly common as marker of
reciprocity in the language of the RV, cf. p ‘purify’ – punné mithá (RV
4.56.6) ‘purifying each other [earth and heaven]’; hi ‘urge, impel’ – mithó
hinvn (RV 10.65.2) ‘impelling each other’; t
(tr) ‘surpass’ – mithas-túr- (e.g.
RV 6.49.3 mithas-túr ‘(day and night), surpassing each other’). Periphrastic
constructions with anyó (a)nyám (lit. ‘another-another’) become productive in
the middle Vedic period.
3.4. Anticausative
The causative/anticausative distinction is the only valency-changing derivation
which, unlike passive, reflexive and reciprocal, is quite regularly expressed by
the active/middle opposition, at least in early Vedic, as in med. várdhate ‘grows’
The symbol ‰ shows that the sandhi has been undone.
~ act. várdhati ‘makes grow, increases’ or med. réjate ‘trembles’ ~ act. réjati
‘makes tremble’.
However, in most cases, the middle type of inflexion is not the only marker of
anticausative, being supported by the stem opposition – which, eventually,
weakens the functional value of the middle as a marker of anticausative. For instance, transitive-causative presents with nasal affixes with the active inflexion
are mostly opposed to middle thematic root presents (= class I presents in the traditional notation) or class IV presents with the suffix -ya-, cf. pávate ‘becomes
clean’- punti ‘makes clean’; ryate ‘flows, bubbles’- riti ‘makes flow, makes
Moreover, already in early Vedic the binary oppositions of the type med.
várdhate ~ act. várdhati, med. códate ‘rushes, hastens (intr.)’ ~ act. códati
‘urges, impels (tr.)’ are often complicated by a third member, the more characterized causative with the suffix -áya-: vardháyati, códáyati, as shown in the following scheme:
act. várdhati
med. várdhate
act. códati
med. códate
act. vardháyati
act. códáyati
In later texts, the causative meaning is still more regularly rendered by the suffix
-áya-, which decreases the functional weight of the active/middle opposition
even further. In other words, Indo-Aryan becomes a causative-marking language.
3.5. The only functional domain which the middle diathesis does not share with
other markers, is the group of functions which can be called self-beneficent, or
auto-benefactive. The self-benefactive meaning was one of the main functions
of the Vedic (and, in general, ancient Indo-European) middle type of inflexion,
as illustrated in (13):
(13) Vedic Sanskrit
a. brhmao (rjñe)
priest:NOM (king:DAT) sacrifice:ACC worship:PRES-3SG.ACT
‘The priest performs the sacrifice (for the king).’
b. brhmaa prayja
priest:NOM sacrifice:ACC worship:PRES-3SG.MED
‘The priest performs the sacrifice (for his own sake).’
4. Development of the new valency-changing categories
The decay of the middle is compensated by and goes essentially parallel with the
development of the new valency-changing categories, foremost within the system of present.
4.1. Causatives
Causatives with the suffix -áya- dramatically increase their productivity already
within OIA. In early Vedic (and probably in Proto-Indo-European) they can only
be derived from intransitives and intransitive/transitives (I/T) verbs of perception and consumption (d ‘see’, vid ‘know’, p ‘drink’). In middle Vedic (in the
language of Vedic prose, or Brhmaƞas) we find first occurrences of causatives
of transitives, such as k ‘make’ – kráyati (Br.+) ‘cause to make’, vac ‘speak’ –
vcáyati (YVp+) ‘make speak’, h ‘take, carry’ – hráyati (YVp+) ‘make take,
make carry’. Finally, in late Vedic and post-Vedic (Stras, Epic Skt.) earliest attestations of causatives with double characterization in -paya- appear: a ‘eat’
– apayati (MnGS) (~ simple caus. ayati (Br.+)), kal ‘wash’ – opt.
klpayta (S.) (~ simple caus. klayati (Br.+)). These formations correspond
to Middle and New Indo-Aryan double causatives.
4.2. Passives
Passives with the suffix -yá- likewise increase their productivity. In early Vedic,
these formations are attested from some 40 roots, which only include non-derived transitives. In middle Vedic (young mantras, Yajurveda, Brhmaƞas) we
find first examples of -yá-passives derived from secondary stems (desideratives
and causatives of intransitive verbs). Finally, in late Vedic and post-Vedic (from
the rauta-Stras onwards), passives of causatives derived from transitives first
appear (caus. dhpáyati ‘makes put’ – ni-dhpyamna- VaitS, caus. pyáyati
‘makes drink’ – -pyyamna- pS).
To sum up, we observe two parallel tendencies in the history of Indo-Aryan.
The loss of many original functions of the middle and the lexicalization of many
middle forms suggests that the diathesis opposition, albeit physically preserved
in the paradigm, loses a large part of its functional content. Thus, the middle,
supposedly a syncretic marker of several intransitive derivations in Proto-IndoEuropean, loses one by one its intransitivizing functions. In other words, the category of middle can be said to degrammaticalize in Indo-Aryan. This process
runs parallel with, and is supported by, the grammaticalization of several new
categories, such as -yá-passives and -áya-causatives, reflexives with tmán- and
reciprocal constructions with anyo’nya.
5. Decay of labile syntax
The third important tendency which determines the development of the Old
Indo-Aryan verbal syntax is the decline of lability. The term ‘labile’ refers to
verbs or verbal forms which can show a valence alternation with no formal
change in the verb, cf. Eng. The door opened ~ John opened the door; Vedic
rudr tásya sádaneu vvƕdhuŪ ‘Rudras have grown [intransitive] in the residences of the truth’ ~ índram ukthni vvƕdhuŪ ‘The hymns have increased
[transitive] Indra’. The ancient Indo-European languages, such as early Vedic
and (Homeric) Greek, are usually considered as characterized by a high degree
of lability. According to the communis opinio, they had a considerable number
of labile verbs and verbal forms. Being one of the most intriguing aspects of the
(ancient) Indo-European verb, this phenomenon has even caused quite desperate
claims expressed by some Indo-Europeanists, such as:
Que signifiait donc [la forme proto-indo-européenne] *e-liq-ê-s? Était-ce ‘tu laissas’ ou ‘tu restas’? Si l'un des deux, comment est-il devenu l'autre? Si tous les deux,
il faut convenir que nos ancêtres manquaient de clarté (Henry 1893: 121)
Almost a half-century later, H. Hirt in his seminal Indogermanische Grammatik
(VII/II: Syntax) has formulated his views less emotionally, but hardly more optimistically:
Bei den Sätzen mit Verben muß man <…> unterscheiden, ob das Verb allein steht
oder noch eine Ergänzung, ein Objekt, fordert, ob es nach der gewöhnlichen Ausdrucksweise intransitiv oder transitiv ist. <…> Nun ist aber die Unterscheidung
nicht so wesentlich, da intransitive Verben transitiv und transitive intransitiv werden können. Wäre sie von großer Bedeutung, so würden wir wohl eine Verschiedenheit der Form zwischen den beiden Kategorien antreffen (Hirt 1937: 28)
In my view, the productivity of the labile patterning in such ancient Indo-European languages as Vedic is strongly exaggerated. Thus far we have no full treatment of the phenomenon of lability in ancient Indo-European languages in general or in Vedic, in particular. I will of course make no attempt to present the full
inventory of the labile forms attested in Vedic. Rather, I will confine myself to
mentioning several forms of the verbal paradigm where labile patterning was
most common, arguing for the secondary character of lability in most such cases
(for details, see Kulikov 2003).
5.1. Lability of middle present forms
First, in a number of middle forms of the system of present, labile patterning results from the polyfunctionality of the middle diathesis. The middle inflexion can
express either the self-beneficent (auto-benefactive) meaning with no valence
change (cf. the textbook example act. yájati ‘sacrifices’ ~ med. yájate ‘sacrifices
for oneself’, as in (13)), or an intransitivizing derivation, most often, anticausative (decausative). Correspondingly, in the cases where the middle diathesis can
have both functions, its middle forms can be employed either transitively with
the self-beneficent meaning, or intransitively, so that we are confronted with labile patterning, as in the case of verbs svádate ‘makes sweet / is sweet’; códate
‘impels / rushes, hastens’, námate ‘bends’, bhárate ‘brings (for oneself) / brings
oneself’, vahate ‘carries / drives, goes’, ráyate ‘lays, fixes on, fastens / leans
on’. Cf. (14–15):
(14) a. (RV 9.74.9)
be/make.sweet:PRES-2SG.IMPV.MED Indra:DAT.SG Pavamna:VOC.SG drink:INF
‘Be sweet for Indra, O Pavamna (= Soma sap), for drinking.’
b. (RV 3.54.22)
be/make.sweet:PRES-2SG.IMPV.MED oblation:ACC.PL
‘Make the oblations sweet [for yourself].’
(15) a. (RV 1.104.7)
‘Rush [like] a bull for a big contest!’
b. (RV 8.75.6)
bull:DAT.SG impel:PRES-2SG.IMPV.MED good-praise:ACC.SG
‘Send forth your beautiful praise for the bull.’
Labile syntax is also attested for presents with nasal affixes (i.e. with the suffixes
-nó-/-nu-, -n-/-n- and with the infix -ná-/-n- = classes V, IX and VII in the traditional notation), particularly for their thematicized variants (see Kulikov
2000). Cf. the labile thematic middle present páte ‘fills; fills oneself’:
(16) a. (RV 3.33.12)
‘Fill your udders, (o rivers).’
b. (RV 7.37.1)
sómair …
pressing:LOC.PL Soma:INS.PL fill:PRES-2PL.IMPV.MED
‘At the [Soma-]pressings fill yourself with the Soma[-sap].’
5.2. Verbs constructed with content accusatives: type púyati
‘prosper’ / ‘make prosper’
Another type of the Vedic and Indo-European lability is represented by the verbs
of the type púyati, employed both in the intransitive usage ‘prosper, thrive’ and
the transitive-causative usage, meaning ‘make prosper, make thrive’, as in (17a–
(17) a. (RV 7.32.9)
jayati kéti púŦya-ti
fast:NOM.SG only wins dwells prosper:PRES-3SG.ACT
‘Only the one who is fast is victorious, dwells (in peace), prospers.’
b. (RV 8.39.7)
sá mud
he joy:INS.SG poetic.inspiration:ACC.PL many
iva puŦya-ti
everything:ACC earth:NOM.SG like prosper:PRES-3SG.ACT
‘By [his] joy, he (sc. Agni) [makes thrive] many poetic inspirations, as the earth makes
thrive everything.’
Elsewhere (Kulikov 1999) I have argued that only intransitive constructions, as
in (17a), represent the original, authentic usage for this verb. The overwhelming
majority of the occurrences with the accusative are, in fact, either (i) constructions with the ‘etymological’ accusative (puí- ‘prosperity’, póa- ‘prosperous
thing’), or (ii) constructions with the content accusative (Inhaltsakkusativ), referring to some aspect(s), parameter(s) or scope of prosperity; cf. (18–20):
(18) (RV 6.2.1)
tvá ... rávo
ná puŦya-si
you:NOM glory:ACC.SG Vasu:VOC.SG prosperity:ACC.SG as prosper:PRES-2SG.ACT
‘You, o Vasu, prosper in glory [= you are glorious], as [one prospers] in prosperity [= as
one is prosperous].’
(19) (RV 7.56.5)
s ví
marúdbhir as-tu ...
this tribe:NOM.SG good-man:NOM.PL Marut:INS.PL be:PRES-3SG.IMPV.ACT
prosper:PRES-PART.ACT-NOM.SG.F manliness:ACC.SG
‘Let this tribe be full of valiant sons with [the help of] Maruts, ... prospering in manliness.’
(20) (RV 1.81.9)
these your Indra:VOC people:NOM.PL
vívam puŦya-nti
all:ACC prosper:PRES-3PL.ACT desirable.good:ACC.SG
‘These men of you, O Indra, prosper in all desirable goods.’8
The rare transitive-causative usages, as the one illustrated in (17b), are likely to
result from the reanalysis of constructions with content accusative, in accordance
with the following semantic scenario: bhma vívam puyati ‘the earth thrives in
everything [what exists on it]’ ‘the earth makes thrive everything [what exists
on it]’.
5.3. Middle athematic participles and middle perfects
Labile patterning is also very common for middle athematic participles with the
suffix -na-. However, as I argued at the beginning of my paper, the labile syntax
of forms such as hinvná- ‘impelling; impelled’ and yujná- ‘yoking’; yoked’ is
a direct corollary of their morphological (grammatical) ambiguity. The transitive
occurrences of hinvná- belong with the present paradigm, while its intransitive-passive attestations belong to the paradigm of the perfect/stative. Likewise,
yujná- is a middle root aorist participle in transitive usages and a medio-passive
aorist participle in intransitive-passive usages.
The same holds for the allegedly labile 3rd sg. and pl. middle perfects as well
as for the corresponding middle perfect participles. Transitive forms such as
dadhé (dh ‘put’) (‘has put’) or yuyujré (‘have yoked’) should be taken as a 3sg.
or 3pl. forms of the middle perfect, as in (5a), while passive occurrences (‘is put /
has been put’; ‘are yoked / have been yoked’) belong with the stative paradigm.
Such constructions with content accusative are erroneously translated by some scholars as
transitive-causative, for instance, by Geldner – in example (20): ‘Diese Leute hier bringen für dich,
Indra, allen begehrenswerten (Besitz) zur Blüte.’ (Geldner 1951: I, 105).
5.4. Active perfects
Of more authentic character is the labile patterning of the active perfects. Typical
examples are perfects of the verb vdh ‘grow, increase’.9 Both active and middle
forms of this verb can be employed either intransitively or transitively. For instance, the 3rd person plural active form vvdhú occurs in the ɰgveda 6 times
in intransitive usages (as in (21a)) and 14 times in transitive-causative usages (as
in (21b)) (see Kümmel 2000: 469ff. for details):
(21) a. (RV 2.34.13)
Rudra:NOM.PL law:GEN.SG residence:LOC.PL grow:PF-3PL.ACT
‘Rudras have grown in the residences of the truth.’
b. (RV 8.6.35)
Indra:ACC.SG hymn:NOM.PL grow:PF-3PL.ACT
‘The hymns have increased Indra.’
After the ɰgveda, we observe the decay of the labile type. Already in the secondmost ancient Vedic text, the Atharvaveda, we find very few labile forms. Most
of the active perfects which show labile syntax in the ɰgveda are either attested
in intransitive usages only (e.g., () vvárta ‘has turned / has made turn’, both intransitive and transitive in the RV, as opposed to AV -vvarta ‘has turned’ (intr.);
see Kümmel 2000: 462ff.), or in transitive usages only (RV mamda ‘has rejoiced, has been exhilarated / has exhilarated’ (tr.), as opposed to AV 7.14.4
3sg.subj.act. mamádat ‘he should exhilarate’ (transitive); see Kümmel 2000:
356ff.), or do not occur at all (as is the case with RVic vvdhú ‘have grown /
have increased’, rurucú ‘have shone / have made shine’).
6. Concluding remarks: Indo-Aryan within the
Indo-European typological context
To sum up, we observe three main tendencies in the evolution of the Indo-Aryan
syntax, which are partly related to, but not entirely dependent from, each other.
The decay of the labile patterning essentially runs parallel with two processes:
the rise and development of new valency-changing categories, causatives with
the suffix -áya- (see Jamison 1983) and passives with the suffix -yá- (see Kulikov 2001), which brings the language to a more overt morphological marking of
the transitivity oppositions; and (ii) degrammaticalization of the middle diathesis, which amounts to transferring most functions of the (Proto-)Indo-European
middle to specialized markers.
The labile syntax of the early Vedic perfect (especially common in the ɰgveda) may originate in
the predominant intransitivity of the Proto-Indo-European perfect, of which some traces can still be
found in early Vedic and Homeric Greek’; for details, see Kulikov 2003; 2006.
Importantly, these tendencies are not shared with most other branches of IndoEuropean. It will now be in order to consider the situation in Indo-European in a
diachronic typological perspective. On the one hand, several groups of Indo-European, including most Germanic, Romance and Slavic languages, replace the
old syncretic marker of the valency-reducing categories, the middle diathesis,
with a new one, mostly going back to the Proto-Indo-European reflexive pronoun *s(u)e- (for this issue, see, for instance, Cennamo 1993). On the other hand,
a number of Romance and Germanic languages attest the emergence and expansion of the labile patterning (which becomes particularly common and productive in English); the expansion of labile verbs is also well attested in Greek. Furthermore, the Proto-Indo-European causative morpheme *-eie-, still well-attested in Gothic (jan-verbs) and Old Church Slavonic (i-causatives), has left only
few traces in modern Germanic and Slavonic languages. This type of evolution,
well-attested in the Western part of the Indo-European area, might be called
By contrast, several other daughter languages, mostly those which belong to
some Eastern branches of Indo-European, radically abandon the syncretic strategy and develop special markers for several intransitive derivations. These include, in particular, Indo-Aryan and Armenian markers of morphological passive going back to Proto-Indo-European suffix *-ie/o-; Indo-Iranian reflexive
pronouns tan- (originally meaning ‘body’) and Indo-Aryan tmán- (‘breath’);
Indo-Iranian reciprocal pronouns. Furthermore, morphological causatives become quite productive in some Eastern branches, in particular, in Armenian
(causative marker -uc‘anem based on the nasal present derived from a sigmatic
aorist) and Indo-Iranian (productive morphological causative suffixes -áya-,
-aiia- going back to Proto-Indo-European *-eie/o-). An interesting feature (isogloss) shared by several Eastern Indo-European languages of the non-syncretic
type, such as Indo-Aryan, Iranian, and Armenian, is the parallel development of
the new non-syncretic passive and productive morphological causative. The
Proto-Indo-European middle diathesis is degrammaticalized and eventually disappears. The labile syntax, even if attested in some ancient languages of the Eastern branches, tends to disappear in the course of their history. One might call this
type ‘antisyncretic’.
Thus, we observe two basic types of evolution, or two evolutionary types, attested in the history of the system of transitivity oppositions and valency-changing categories in Indo-European: syncretic type found in many Western branches
and anti-syncretic type attested at least in some Eastern branches, in particular,
in Indo-Aryan.
Typologically, the Eastern type, as attested in Indo-Aryan, shares more features with some non-Indo-European families, such as Turkic or Altaic in general,
rather than with the Western Indo-European type, as attested in Germanic or
Greek. Like Indo-Aryan, Turkic has productive morphological valency-changing categories, such as causative or reciprocal, and there is some evidence for the
decline of labile patterning (still present in Old Turkic), as well as the underdeveloped middle voice, as shown in Table 3.
Table 3. Diachronic typological features of some language families
middle voice
East Caucasian
++ (new)
++ (new)
++ (new)
increasing (in productivity, frequency etc.)
The Western type has no such clear non-Indo-European parallels as the Eastern
type, although we probably can observe some affinities with such families, as,
for instance, Kartvelian or Egyptian.
The origins of these features and the anti-syncretic evolutionary type, in general, instantiated by Indo-Aryan is a difficult problem on its own. It may be (partly) due to the influence of the substrate languages of the Altaic or Dravidian type.
These languages could be responsible for some other features of Indo-Aryan as
well, in particular, for the dramatic restructuring of the case system, loss of many
Proto-Indo-European cases and the emergence of the new, agglutinative, case
7. References
Cennamo, Michela. 1993. The reanalysis of reflexives: a diachronic perspective. Napoli:
Delbrück, Berthold. 1888. Altindische Syntax. Halle a. S.: Verlag der Buchhandlung des
Geldner, Karl Friedrich. 1951. Der Rig-veda aus dem Sanskrit ins Deutsche übersetzt...
Bd. 1–3. Cambridge, MA: Harvard University Press.
Gonda, Jan. 1979. The medium in the gveda. Leiden: Brill.
Got, Toshifumi. 1987. Die “I. Präsensklasse” im Vedischen: Untersuchung der
vollstufigen thematischen Wurzelpräsentia. Wien: Verlag der Österreichischen Akademie der Wissenschaften. (2., überarbeitete und ergänzte Aufl. 1996).
–– 1997. Überlegungen zum urindogermanischen «Stativ». In: E. Crespo & J.L. García Ramón (eds.), Berthold Delbrück y la sintaxis indoeuropea hoy. Actas del Coloquio de la Indogermanische Gesellschaft. Madrid: UAM; Wiesbaden: Reichert,
Henry, Victor. 1893. Rev. of: K. Brugmann. Grundriss der vergleichenden Grammatik
der indogermanischen Sprachen II, II, 2. (Strassburg, 1892). Revue critique d’histoire
et de littérature 35 (7). 120–123.
Hirt, Hermann. 1937. Indogermanische Grammatik. Teil VII. Syntax II. Die Lehre vom
einfachen und zusammengesetzten Satz. Heidelberg: Winter.
Jamison, Stephanie W. 1983. Function and form in the -áya-formations of the Rig Veda
and Atharva Veda. Göttingen: Vandenhoeck und Ruprecht.
Kulikov, Leonid. 1999. May he prosper in offspring and wealth: A few jubilee remarks
on the typology of labile verbs and Sanskrit púyati ‘prospers; makes prosper’. In:
Ekaterina V. Rakhilina & Yakov G. Testelets, eds. Tipologija i teorija jazyka: Ot
opisanija k ob"jasneniju. K 60-letiju A.E. Kibrika. [Typology and linguistic theory:
From description to explanation. For the 60th birthday of Aleksandr E. Kibrik].
Moskva: Jazyki russkoj kul'tury, 224–244.
–– 2000. Vedic causative nasal presents and their thematicization: a functional approach. In: J.Ch. Smith & D. Bentley (eds), Historical linguistics 1995. Selected
papers from the 12th International Conference on Historical Linguistics. Vol. 1:
General issues and non-Germanic Languages. Amsterdam: Benjamins, 191–209.
–– 2001. The Vedic -ya-presents. PhD diss. Leiden University.
–– 2003. The labile syntactic type in a diachronic perspective: The case of Vedic. SKY
Journal of Linguistics 16. 93–112.
–– 2006. Passive and middle in Indo-European: Reconstructing the early Vedic passive
paradigm. In: W. Abraham & L. Leisiö, eds. Passivization and typology: form and
function. Amsterdam: Benjamins, 62–81.
–– 2007a. The reflexive pronouns in Vedic: A diachronic and typological perspective.
Lingua 117 (8). 1412–1433.
–– 2007b. Reciprocal constructions in Vedic. In: V. P. Nedjalkov et al., eds. Reciprocal
constructions. Vol. 2. Amsterdam: Benjamins, 709–738.
Kümmel, Martin. 1996. Stativ und Passivaorist im Indoiranischen. Göttingen: Vandenhoeck & Ruprecht. (HS; Ergänzungsheft 39).
–– 2000. Das Perfekt im Indoiranischen. Wiesbaden: Reichert.
Macdonell, Arthur Anthony. 1910. Vedic grammar. Strassburg: Trübner.
Narten, Johanna. 1969. Zum “proterodynamischen” Wurzelpräsens. In: J.C. Heesterman
et al. (eds.), Pratidnam: Indian, Iranian, and Indo-European studies presented to
F.B.J. Kuiper on his sixtieth birthday. The Hague: Mouton, 9–19. [= J. Narten. Kleine
Schriften, I: 97–107. Wiesbaden: Reichert, 1995]
Whitney, William Dwight. 1885. The roots, verb-forms, and primary derivatives of the
Sanskrit language. Leipzig: Breitkopf and Härtel.
Witzel, Michael. 1995. Early Indian history: linguistic and textual parameters. In: G. Erdosy (ed.) Language, material culture and ethnicity. The Indo-Aryans of ancient South
Asia. Berlin – New York: de Gruyter, 85–125.
8. Acknowledgements
I am grateful to A. Lubotsky and I. Serzants for their comments on earlier drafts
of this paper. I also would like to take this opportunity to express my thanks to
the audiences of the 23rd Scandinavian Conference of Linguistics (University of
Uppsala, 1–3 October 2008) and Friday Afternoon Lecture organized by Leiden
University Centre for Linguistics (LUCL) – in particular to Chr. Schaefer, A.
Siewierska, F. Kortlandt, M. Kossmann, A. Lubotsky, M. Mouse, and T. Schadeberg – for suggestions and critical remarks. I acknowledge the Netherlands Organization for Scientific Research (NWO) for financial support, grant 275-70009 (VENI- project).
9. Contact information
Leonid Kulikov
Leiden University Centre for Linguistics (LUCL) / Dept. of Comparative Indo-European
Linguistics (VIET)
Leiden University, Faculty of Humanities,
PO Box 9515
2300 RA Leiden
The Netherlands
[email protected]
Using parallel corpora in teaching & research:
The Swedish-Hindi-English & SwedishTurkish-English parallel corpora
Anju Saxena, Beáta Megyesi
Éva Á. Csató, Bengt Dahlqvist
Department of Linguistics and Philology
Uppsala University
1. Introduction
Proposed methods for the automatic acquisition of linguistic knowledge by computer potentially allow for the rapid creation of language technology resources
with minimal human work, which if realized would be of great help in the case
of many less-commonly taught languages. In this connection, it is important to
note that, arguably, language technology – and consequently also the application
of machine learning methods in language technology – has been shaped by the
typological and other traits of the most explored languages, especially English,
which is in many respects an atypical language from a linguistic point of view,
and quite unlike many other languages. There is, thus, a need to test and refine
these methods on a number of structurally different languages, making languages such as Turkish, Swedish and Hindi a good testing ground, allowing us
to gain a better understanding of the generality or language-specificness of these
“Supporting Research Environment for Minor Languages” is a research program at the Department of Linguistics and Philology, Uppsala University – financed by the Swedish Research Council and Uppsala University – the aim of
which is to provide a research environment for less explored languages by developing parallel treebanks. We are currently working on building a TurkishSwedish-English and a Hindi-Swedish-English parallel treebank. The aim of this
article is to describe briefly our work with these two parallel treebanks and also
to show how these treebanks are a valuable resource in teaching and research.
Section 2 describes briefly some important aspects of the two parallel corpora.
In section 3 we will describe how corpora can be used in linguistic research and
in teaching in general and our experience of using the SwedishTurkish treebank
in teaching in particular.
2. On the Swedish-Hindi-English and
Swedish-Turkish-English parallel treebanks
A parallel corpus is a bi- or multilingual text material containing original texts
in one language and their translations into another language or other languages
(or, alternatively, parallel translations from an original in a language not in the
corpus; this is normally the case with parallel corpora of Bible texts). Often,
parallel corpora are aligned, meaning that corresponding units (sentences,
phrases, even words) from the different language versions are explicitly linked
together. Syntactically annotated corpora are called treebanks.
The focus in language technology has been on English and major Western languages. English has been one of the languages included in many of the existing
parallel treebanks. For example,
• The Prague Czech-English Dependency Treebank (Haji, 2001)
• ISJ-ELAN Slovene-English Parallel Treebank (Erjavec, 2002)
• Swedish-English Parallel Treebank (Ahrenberg, 2007)
Relatively little work has been done on less-commonly taught languages. This
is one major reason for including Hindi, Swedish and Turkish in this project.
Up until now there have been no parallel corpora involving Swedish-Hindi or
The Swedish-Hindi-English parallel treebank
Hindi provides a good testing ground for language technology tools. There are,
at present, two major Hindi corpora available: the EMILLE corpus (<http://>; Hardie et al. 2006) and the Hindi corpus at IIT
Bombay (<>). The IIT Bombay Hindi corpus is a
monolingual corpus, consisting of excerpts of texts each around 2 000 words
(which can be an issue for discourse related work), and it does not seem to have
any linguistic annotation. The EMILLE corpus has both a large monolingual
and a smaller parallel corpus part (Hindi-English), but this corpus, too, is not
linguistically annotated. Keeping in view our goal of creating a trilingual
parallel treebank which is POS-tagged and dependency parsed as well as
aligned at word and sentence levels, and, keeping in view our ambition to use
this treebank in research in linguistics and in teaching makes these existing
Hindi treebanks less useful for our purposes. Therefore decision was made to
start from scratch. The Hindi materials which, at present, are included in our
corpus are.
Bible texts: the 4 Gospels
Texts from the parallel treebank section of the EMILLE project
The UN Declaration of Human Rights
A Hindi novel
Some texts providing information about Sweden (see below)
Hindi texts were typed in manually for lack of usable electronic versions. All
texts have undergone semi-automatic cleaning and converting in order to make
them conform to the requirements of the annotation tools used. The texts are now
available in XML with Unicode (UTF-8) character representation. Initial work
has also been done regarding POS tagging, morphological analysis and chunking. MaltParser (Nivre et al. 2006a) has been trained on a syntactically annotated
Hindi treebank (Saxena et al. 2008).
The following table summarizes the details of the texts included in the
Swedish-Turkish-English and the Swedish-Hindi-English parallel corpora. It
shows that while some texts are unique to one of the two corpora, a large number
of the texts are common in the two corpora:
UN Declaration of
Human Rights
What is unicode
Gospel of Luke
Gospel of Matthew
Gospel of Mark
Gospel av John
1 382
1 911
2 106
3 770
1 831
1 221
4 267
1 604
32 238
19 564
18 872
24 209
133 568
30 621
29 274
18 481
24 907
108 235
288 701
32 238
29 247
18 888
24 625
162 302
Important considerations while building these parallel corpora have been that:
• The texts are high quality texts, where the translations are done by trained
translators. Compiling a proper text corpus entails a much greater amount of
work than merely collecting any kind of text that you can lay your hands on,
especially where other text types than newstext are difficult or impossible to
acquire in electronic form. Texts are morphologically and syntactically analyzed
• They are (semi-)automatically aligned at the sentence and word levels
• Special attention has also been made to explore and to the extent possible, to
use open-source resources and to use the same tools in working with the
Swedish-Turkish-English and the Swedish-Hindi-English parallel corpora.
This has several advantages. For example, this gives us a chance to examine
the usefulness of a given tool across languages. In this way, it provides us a
set of tools which are not language-dependent. This, at times, has meant using
an existing tool as it is (and we have in fact often been able to do just that),
while at other times it has meant, developing an existing tool further.
Since both these parallel corpora have, to a large extent, similar format and they
have used similar sets of tools, we will below describe the building of the two
corpora by describing in more detail the building of the Swedish-Turkish-English parallel corpus.
The Swedish-Turkish-English parallel treebank
The Swedish-Turkish parallel treebank, which is currently under development
and which has been previously described (Megyesi et al. 2006; Megyesi & Dahlqvist 2007; and Megyesi et al. 2008) contains syntactically annotated parallel
texts with various annotation layers from part-of-speech tags and morphological
features to dependency annotation where each layer is automatically annotated,
the sentences and words are aligned, and partly manually corrected.
The treebank material is processed automatically by using various tools making the annotation, alignment and manual correction easy and straightforward for
users with less computer skills. This is necessary, as our ambition is to allow researchers and students of particular languages to enlarge the treebank by automatically processing and correcting the new data by themselves.
In order to build the treebank automatically, we use a basic language resource
kit (BLARK) for the particular languages and appropriate tools for the automatic
alignment and correction of data.
First, the original materials received from the publishers in various formats are
cleaned up. For example, rtf, doc, and pdf documents are converted to plain text
files. After cleaning up the original data, the texts are processed automatically by
using tools for formatting, linguistic annotation and sentence and word alignment.
During formatting, the texts are encoded using UTF-8 (Unicode) and marked
up structurally using XML Treebank Encoding Standard (XCES). The text files
are processed by various tools in the BLARKs developed for each language separately. A tokenizer is used to split the text into tokens such as words and punctuation marks. Sentence segmentation is also performed to break the texts into
Once the sentences and tokens are identified, the data is linguistically analyzed. We use several annotation layers for the linguistic analysis, first on a morphological level, then on a syntactic level. The annotation and the labels for the
linguistic analysis are de facto standard for the involved languages. For the linguistic annotation, external morphological analyzers, part-of-speech taggers and
syntactic dependency parsers are used which are trained on annotated treebanks
developed for the specific languages. For example, for Swedish we use the
Stockholm Umeå Treebank tag set (SUC 1997) for the morpho-syntactic annotation and the functional annotation of Talbanken05 (Nivre et al. 2006b), while
we derive the linguistic annotation from the Metu-Sabanc
Turkish Treebank
(Oflazer et al. 2003) for the syntactic analysis of Turkish. For English, we use
the Penn Treebank tag set.
The Swedish and English texts are morphologically annotated with the open
source HunPoS tagger (Halacsy, et al. 2007). The tokens are annotated with parts
of speech and morphological features and are disambiguated. The results for the
morphological annotation of Swedish show an accuracy of 96.6% (Megyesi
2008). The Turkish material is morphologically analyzed and disambiguated using a Turkish analyzer (Oflazer 1994) and a disambiguator (Yuret & Türe 2006)
with an accuracy of 96%. The English data contains less error, approximately
only 2%–3%.
The other linguistic layer contains information about the syntactic analysis.
We use dependency rather than constituent structures, as the former has been
shown to be well suited for both morphologically rich and free word order languages such as Turkish, and for morphologically simpler languages, like Swedish. The English, Swedish and the Turkish data are annotated syntactically using MaltParser (Nivre et al. 2006a), trained on Penn Treebank and Talbanken05
(Nivre et al. 2006b) and on the Metu-Sabanc
Turkish Treebank (Oflazer et al.
2003). The annotation includes approximately 15%–20% errors, depending of
the language, which need to be manually corrected.
The sentences and words in the languages are aligned automatically by using
standard techniques, such as the length-based approach (Gale and Church 1993)
for sentence alignment, and the clue alignment approach (Tiedemann 2003) with
the toolbox for statistical machine translation GIZA++ (Och and Ney 2003) for
word alignment. The aligned sentences are manually corrected by a student who
speaks both languages. We automatically compare the links before and after the
manual correction and the user gets statistics about the differences. The results
show that between 67% and 94% of the sentences are correctly aligned by the
automatic sentence aligner depending on the text type. We are currently working
the automatic correction of word alignment in the syntactic trees.
In addition, we visualize the treebank in different ways without showing the
structural markup when used, for example, in teaching.
3.1 Use of treebanks in research and in teaching
There has been a growing interest in using natural language corpora in teaching
and in research, partly due to the growing availability of computer-readable linguistic corpora, and partly due to an increase in examining language in its natural
context as opposed to investigating constructed language examples in isolation.
Researchers, teachers and students now have access to different types of language corpora to discover facts about language; for example, which words are
the more frequently used words in a language or a language-type, in which context they predominantly occur and which grammatical patterns are associated
with a particular linguistic item (Ghadessy et al. 2000). There have been two primary approaches for the use of corpora in language teaching/learning: the
“COBUILD approach” and the Data-Driven Learning approach.
Until recently, the COBUILD approach was the predominant approach. Corpora, in this approach, are used by researchers and producers in building dictionaries and other language learning materials. Traditionally it has been very large
corpora which have been used for this purpose. Further, within this approach, the
user (a student, for example) receives results of a project involving corpora as
end-products (for example, in the form of a language learning packet). Learners
do not get to use the corpora themselves in order to come up with their own analyses and learn from that. Another limitation of this approach has been its limited
access. Up until quite recently access to the results of such works has been
limited, primarily because of the high cost of such language-learning tools.
In the Data-Driven Learning approach students use corpora directly in their
own learning. They use the corpora, for example, to discover linguistic patterns
and to organize linguistic patterns which they observe, arriving at generalizations inductively and verifying deductive rules. Such exposure to corpora provides students the chance not only to extract relevant examples of one or the
other linguistic structures, but also provides them material for discussion when
they find gaps, to verify and extend their hypothesis and to arrive at generalizations. In favour of the Data-Driven Learning approach, Johns (1991) states:
What distinguishes the data-driven learning approach is the attempt to cut out the
middleman … and give direct access to the data so that the learner can take part in
building up his or her own profiles of meanings and uses. (Johns 1991:30 in Aston
Johns (1991) mentions three phases in the Data-Driven Learning:
• observation
• classification
• generalization
One advantage of using corpora in teaching is that instead of learning about linguistic theories in vacuum (a more passive learning method, where facts are fed
to students in form of lectures), students have a chance to test these theories
themselves against these corpora and learn about these theories or concepts for
themselves (a more active learning method). When corpora are used by students
as part of their learning, distinction between teaching and research is “blurred”,
as students, by discovery proceedure (thus, research), learn things for themselves
(Knowles 1990). The use of corpora in teaching can, in this way, affect both
teachers’ as well as students’ role. This approach is as equally relevant in a classroom set-up as in self-study situations.
The gap between the COBUILD and DDL approaches is, however, getting
smaller. More access to corpora (especially for non-commercial purposes) provides better (pre-)conditions to use them in producing language learning tools as
well as in using them directly in teaching/learning.
3.2 Using Swedish-Turkish-English treebank in a teaching enviroment
The aim of the Swedish-Turkish treebank (STPC) is to provide Swedish speaking students and researchers with easily accessable annotated linguistic data on
Turkish. The corpus is now being completed with English texts. The webbased
STPC can be used both by regular and distance students in their data-driven acquisition of new vocabulary items and their usage. It functions also as a learning
platform for and for testing hypotheses concerning the morphological and syntactic aspects of Turkish grammar. Further, it helps the students to practice translation between Swedish and Turkish. All this is possible due to the fact that the
Swedish-Turkish parallel texts are available in annotated form. The annotations,
on request, are visualized in pop-up windows. The morphological analyses are
given at present in clumsy, parser-generated formulas but will in the near future
be substituted by labels in more intelligible forms based on the grammatical
terms employed in textbooks and in the Turkish Suffix Dictionary (Csató &
Nathan 2003). The interface for displaying syntactic information is not ready yet.
A search tool assists the students to create concordance lists. They can search
for whole words, beginnings of words, parts of words or ends of words in Turkish or Swedish. The concordance lists display whole sentences in which the target item appears and it is highlighted. The selected sentences are aligned with
their translational equivalents. This form of displaying the linguistic data is much
more suitable for learning than KWIC lists in which only the immediate environment of the target item is shown.
Such lists are used to find frequent patterns of usage, transformational equivalents, different meanings of polysemic words, translational equivalents of Turkish grammatical categories, etc. Different types of exercises are designed and
published in the Internet. Students in Turkic languages also use STPC while writing their theses. Bergdahl (2006) studied the meanings of the Turkish word gölge
‘shadow’ and the corresponding Swedish word skugga. Dadasheva (2005) inves99
tigated how the Turkish indirective category marked by -mI / imi is translated
into Swedish and Russian. Hedman (2009) compared the meanings of the Swedish and Turkish verbs ‘do’ and ‘make’. Haktan
r (2006) reviewed the ambiguous Turkish morphological forms in one of the parallel texts and described different types of morphological ambiguities.
Apart from using STPC in learning environments, it is also being used by researchers. One example of which is the article Rendering evidential meanings in
Turkish and Swedish (Csató 2009), which examined the Turkish evidential category of indirectivity and the less grammaticalized or lexical strategies in Swedish
to render evidential nuances. The description of the strategies used in the two
languages was complemented with an analysis of data in one of the parallel
Turkish-Swedish texts. It was found that although Swedish has means to express
evidential nuances these were much less used in the Swedish translations than
expected. The article describes several reasons for this. One might be that the
Turkish category allows three different types of reading. This ambiguity is significant in certain texts. The Swedish devices may render a particular evidential
nuance but not the whole range of ambiguity of the Turkish forms.
In short, the parallel corpora in general and the Swedish-Turkish-English and
the Swedish-Hindi-English treebanks, in this way, can be used in a variety of
ways in teaching and in research.
Ahrenberg, L. 2007. LinES: An English-Swedish parallel treebank. In Proceedings of
Nordiska Datalingvistdagarna, Nodalida 2007. Tartu, Estonia
Aston, G. 1997. Small and large corpora in language learning.
Bergdahl, E. A. 2006. Shadow. From a relaxing spot to darkness and death. A semantic
study of how the word shadow is used in Swedish and Turkish. C-uppsats. Bat
ve Edebiyat
Bölümü, Bogaziçi University & Department of Linguistics and Philology, Uppsala University
Csató, É. Á. 2009. Rendering evidential meanings in Turkish and Swedish. In: É. Á. Csató
et al (eds.) Turcological letters to Bernt Brendemoen, 77–86. Oslo: Novus
Csató, É. Á. & D. Nathan. 2003. Turkish suffix dictionary.
Dadasheva, S. 2005. Den turkiska indirektiva kategorin. En undersökning av återgivningen av den turkiska indirektiva kategorin i ryska och svenska autentiska översättningar. C-uppsats. (In Swedish.) Department of Linguistics and Philology, Uppsala University
Erjavec, T. 2002. The IJS-ELAN Slovene-English parallel corpus. International Journal
of Corpus Linguistics, 7:1:1–20.
Gale, W. A., & K. W. Church. 1993. A program for aligning sentences in bilingual corpora. Computational Linguistics, 19.1: 75–102
Ghadessy, M., A. Henry & R. L. Roseberry. 2000. Introduction. In: Mohsen Ghadessy,
Alex Henry and Robert L. Roseberry (eds.) Small corpus studies and ELT. Theory and
practice, xvii–xxiii. Amsterdam/Philadelphia: John Benjamins
r, M. 2006. Orhan Pamuk’un Beyaz Kale adl eserinde çok anlamllk. C-uppsats.
(In Turkish.) Department of Linguistics and Philology, Uppsala University
Haji, J., E. Hajiová, P. Pajas, J. Panevová, P. Sgall & B. Vidová-Hladká. 2001. Prague
dependency treebank 1.0 (final production label). CDROM CAT: LDC2001T10.,
ISBN 1-58563-212-0.
Halácsy, P., A. Kornai & Cs. Oravecz. 2007. Hunpos – an open source trigram tagger. In
Proceedings of the 45th Annual Meeting of the Association for Computational Linguistics, Companion volume, proceedings of the demo and poster sessions, 209–212,
Prague, Czech Republic.
Hardie, A., P. Baker, T. McEnery & B. D. Jayaram. 2006. Corpus-building for South
Asian languages. In A. Saxena and L. Borin (eds.) Lesser-known languages of South
Asia. Status and policies, case studies and applications of Information Technology,
211–242. Berlin: Mouton de Gruyter
Hedman, M. 2009. Verbet ’göra’ i svenska och turkiska. C-uppsats. (In Swedish.) Department of Linguistics and Philology, Uppsala University
Johns, T. 1991. Should you be persuaded: Two examples of data-driven learning. In T.
Johns & P. King (eds.), Classroom concordancing. English language research journal
4, 1–16
Knowles, G. 1990. The use of spoken and written corpora in the teaching of language and
linguistics. Literary and linguistic computing. Journal of the Association for literary
and linguistic computing. 5.1: 45–48
Megyesi, B. 2008. The open source tagger HunPoS for Swedish. Department of linguistics and philology, Uppsala University
Megyesi, B. B., A. Sågvall Hein & E. Csato Johanson. 2006. Building a Swedish-Turkish
parallel corpus. In Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC'06). Genoa, Italy
Megyesi, B. B. & B. Dahlqvist. 2007. A Turkish-Swedish parallel corpus and tools for its
creation. In Proceeding of Nordiska Datalingvistdagarna, NoDaLiDa 2007
Megyesi, B., B. Dahlqvist, E. Pettersson & J. Nivre. 2008. Swedish Turkish Parallel
Treebank. In Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC'08), 470–473. Marocco
Nivre, J., J. Hall & J. Nilsson. 2006a. MaltParser: A Data-Driven parser-generator for dependency parsing. In Proceedings of the 5th International Conference on Language
Resources and Evaluation (LREC), 2216–2219
Nivre, J., J. Hall & J. Nilsson. 2006b. Talbanken05: A Swedish Treebank with phrase
structure and dependency annotation. In Proceedings of the 5th International Conference on Language Resources and Evaluation (LREC), 1392–1395
Saxena, A., P. Swaroop Madhyasta, & J. Nivre. 2008. Building the Uppsala Hindi Corpus.
Poster presentation. SLTC 2008. Stockholm. <>
Oflazer, K. 1994. Two-level description of Turkish morphology. Literary and Linguistic
Computing, 9:2
Oflazer, K., B. Say & Hakkani-Tür. 2003. Building a Turkish treebank. In A. Abeillé (ed.)
Treebanks: Building and using parsed corpora, 261–277. Kluwer
Och, F. J. & H. Ney. 2003. A systematic comparison of various statistical alignment
models. Computational Linguistics, 29.1: 19–51
SUC. Department of Linguistics, Umeå University and Stockholm University. 1997. SUC
1.0 Stockholm Umeå Corpus, Version 1.0.
Tiedemann, J. 2003. Recycling translations Extraction of lexical data from parallel
corpora and their applications in Natural Language Processing. PhD Thesis. Uppsala
Yuret, D. & F. Türe. 2006. Learning morphological disambiguation rules for Turkish. In
Proceedings of HLT NAACL'06, 328–334. New York
On the Grammaticalization of Mandarin
Aspect Markers
Luying Wang
Uppsala University
The term grammaticalization was first proposed by French linguist Antoine
Meillet in 1912 with the definition “the attribution of grammatical character to
an erstwhile automous word.” (Hopper & Traugott 1993:19) As Hopper &
Traugott point out, this proposal originates from “earlier speculations …about
the evolutionary development of human speech.” These speculations do not only
exist in the Indo-European linguistic history. Since the 13th century, many Chinese scholars have discussed about the theory that “empty words,” i.e. grammatical particles, come from “full words,” i.e. full lexical items such as verbs.
However, it was not until the 1990s that the research on grammaticalization in
the Indo-European languages attracted the attention of Chinese linguists. Since
1994, a considerable amount of study has been done on Chinese historical linguistics. Several studies were on the origins of some of the four aspect markers
-le, -guo, -zhe, and zai or the aspectual system in Mandarin, for example, Li &
Shi (1997), Sun (1998) and Jiang (2004, 2006). In this paper I will first give a
brief account of the four makers’ usage in modern Mandarin, then discuss about
the motivations, mechanisms and processes of the four markers’ grammaticalization with the help of data from Classical Chinese and theories on grammaticalization in the literature, and provide my insight in this subject.
1. The four aspect markers in modern Mandarin
The four markers are the perfective -le and -guo, and the imperfective -zhe and
zai. The perfective aspect marker -le ‘presents closed, non-stative situation.’
(Smith 1994: 111) So is another perfective marker -guo. However, in contrast
with -le, -guo emphasizes a kind of discontinuity. The suffix -guo “signals that
an event has been experienced at least once at some indefinite time”. (Li &
Thompson 1989:226) Thus it is also referred to as an experiential marker. Here
are some examples given by Chao (1968) to compare these two:
(1) a. wo shuai duan le tui.
break LE leg
“I broke my leg (and it is still broken).”
b. wo shuai duan guo
break GUO leg
“I once broke my leg (and I am fine now).”
We can see that the focus of a sentence with -le is that the event or a series of
events occurred and consequently there is influence to the present time. The focus of a sentence with -guo is that the event has taken place at least once, and it
is over now. Therefore -le hints a result of the event while -guo indicates the
completion and discontinuity of the event.
In Chinese, pre-verbal imperfective marker zai focuses on the continuity of an
action, thus also called a progressive marker. Post-verbal imperfective marker
-zhe focuses on the durativity of an event, thus also called a durative marker. A
major difference between these two markers are that zai cannot be used with stative verbs while -zhe is used mostly with stative verbs. This is because, as Huang
et al. (1989) has observed, that zai in the sense of immediacy asserts an ongoing
activity, while -zhe with its senses of remoteness emphasizes a resultant state or
pragmatically that there is some kind of background information. We can see this
difference of these two markers by the following examples:
(2) a. ta zai dai
he ZAI wear
“He is putting on a hat.”
b. ta dai zhe
he wear ZHE
“He is wearing a hat.”
Table 1 presents the main functions of the four aspect markers in modern
Table 1 The four aspect markers in modern Mandarin
perfective marker
perfective marker
imperfective marker
imperfective marker
“indicates a bounded event that impact the present time”
“indicates the completion and discontinuity of an event”
“focuses on the durativity of an event”
“focuses on the continuity of an action”
2. The process of grammaticalization of the four markers
The pioneer of modern Chinese linguistics, Wang Li, stated in Wang (1958) that
the appearance of verb final particle -le and -zhe was a major even in the history
of Chinese grammar, because it meant that Chinese started to use grammatical
means to express tense-aspect instead of pure lexical means. In this section I will
take a look at the paths that these grammatical particles go through to change in
the linguistic history. First I will give the definitions of two main mechanisms of
change at work here: reanalysis and analogy.
Reanalysis is defined as changing in a language structure without any immediate modification of its surface manifestation (Peyraube 1999: 191). Most
grammaticalization process does involve reanalysis, although it is not obligatory.
Reanalysis happens due to a basic cognitive principle: to use existing item to express new function. (Traugott & Heine 1991)
Analogy is also called extension, which refers to “the attraction of extant
forms to already existing constructions” in grammatical change. (Peyraube
1999:187). As the opposite of reanalysis, analogy changes the surface structure
without changing the underlying structure. Analogy is a main mechanism in
connecting the original lexical item and the future grammatical morpheme.
(ibid. 189)
2.1 The grammaticalization of -le
Table 2 Path of Progression for -le
Change stages
Structure and Meaning
Original structure
1st reanalysis
1st analogy & reanalysis
2nd analogy & reanalysis
2nd reanalysis
NP1+V1+NP2+le (V) “to finish”
NP1+V1+NP2+le (V/VC) “to finish”
V + le (VC) “finished”
V + le (perfective marker) + NP
NP1+V1+NP2+le (CRS*)
V + -le (perfective aspect) + (NP)
Sentence +le (CRS)
3rd century
10th century
10th century
10th century
13th century
modern Chinese
modern Chinese
*CRS=”current relative state”, also called “sentence final le”, terms adopted from Li & Thompson
In the 3rd century le was a verb meaning “to finish”.
gong liu
wo le
stay-behind I
Ming Fu bu neng zhi.
finish COM, Ming Fu no can
“You just let me stay behind (then this matter will be accomplished), Ming Fu cannot stop it.”
(Sanguozhi, 3rd century, quoted from Jiang 2004:137)
This le rarely acts alone. It usually appears in V+O+le structure as a secondary
verb. In this structure, besides the object, there can be various adverbs between
the verb and le. Therefore this le is regarded as a verb instead of a complement.
(Li & Shi 1997)
By the 10th century, there are examples of le in V+le structure as in (4) b. and
from the same text, V+Le+O structure also appears, as in (4) a. and (5). It is possible that the two structures are actually one: V+Le+O with the possibility to omit
the object.
(4) a. jian le
bian ru
see LE fellow apprentices then
enter in
“After seeing your fellow apprentices, then come in.”
b. jian le
chou shen bian que
see LE leave
body then
immediately return
“After seeing (the person), leave immediately and return.”
(Dunhuan bianwen, 10th century, quoted from Jiang 2004:149)
le chun hong, tai congcong.
forest flower wither Le spring red
too soon
“The forest flowers wither after the spring red, too soon.”
(Li Yu poem, 10th century, quoted from Jiang 2004:139)
Li & Shi (1997) find out only 5% of the cases of the V+le structure that appeared
in the texts around the 10th century had adverbs between the verb and le. This
demonstrates a closer and closer tie between the verb and le, which acts more like
a verb complement than a verb. On the other hand, they also find that in the texts,
more and more adverbs that used to appear between the verb and le in the original
V+O+le moved to the front of the main verb, although data from the 7th to 10th
century texts show that 70% of the V++O+le still present an adverb before le, as
in (6). Therefore this le should still be considered a verb.
zhi yi
sigh it
le, ni
already LE adopt into Buddhist text title
“After a sigh about the matter, (they) adopted it into the title of the text.”
(LiuJiang, 10th century, quoted from Li & Shi 1997: 90)
Therefore we know that in the 10th century, le existed in two structures, V+le +
(O) and V+O+le. The le in the first structure acts more like a clitic that is attached
to the verb, and the le in the second structure remains a verb because 70% of the
cases adverbs were inserted in front of it. It is possible that during this time this
le can be explained as either a verb or a verb complement. Grammaticalization
is a continuum with no clear-cut borders during the changing process. Consequently the co-existence of the same morpheme with its original function and
new function is common. Moreover, the coexistence is the evidence of the grammaticalization process. It is natural that there are times that it is difficult to judge
the exact characteristic of the morpheme.
In the 13th century, the usage of le was very common and very close to modern
usage, appearing both with and without an object in the V+ le + (O) structure as
in (7) a. b. and the first le in (7) c., and also as a sentence final le as the second
le in (7) c.
(7) a. bu jie
le xiongdi.
not lend when repel LE brother
“If not lend (it) to the brother, he will hate me.”
b. wo chenyin le banshang bu yu.
LE half day
not speak
“I hesitated for a long time without speaking.”
(Ma Zhiyuan poem, 13th century)
c. mo
que yi
somebody GEN next but
le zhe weici le.
move LE this seat
“However someone who was next to him moved the seat.”
(Zhuzi yulei, 13th century, quoted from Jiang 2004:155)
I agree with Li & Shi (1997) that in the grammatical development of le, there are
actually two different routes that lead to the sentence-final le and the aspect
marker -le in Modern Mandarin from the same verb le. Both routes are presented
in Table 2. One is that the original structure changed through the 1st reanalysis,
when although the surface structure did not change, the le in it can be analyzed
as either a verb or verb complement, depending on if there is adverb in front of
the le. At the same time, i.e. around the 10th century, another route started by 1st
analogy & reanalysis, where le changed to a verb complement and the surface
structure it appeared in also changed to V + le. Through the 2nd analogy & reanalysis, an object can be attached after this le. At this time the le has become a
perfective marker. The final grammaticalization of the sentence final le is not
finished until 2nd reanalysis happened sometime before the 13th century, when the
distinction between the perfective marker -le and sentence final le is clear and
settled, as exhibited in (7) c.
2.2 The grammaticalization of -guo
Table 3 Path of Progression for -guo
Change stages
Structure and Meaning
Original structure
1st reanalysis
2nd reanalysis
3rd reanalysis
NP1+guo (V) “cross”+NP2
NP1+V+guo (V2) “cross”+NP2
NP1+V+guo (V2)”action finished”+NP2
NP1+V+guo(VC) “action finished”
NP1+V+guo (perfective aspect marker)
NP1+guo (V) “pass, cross”+(NP2)
NP1+V1+(Adv)+guo(V2) “cross”
NP1+V+guo (perfective aspect marker)
3rd century
9th century
9th century
9th century
14th century
modern Chinese
modern Chinese
modern Chinese
The timeline for the grammaticalization of -guo is very similar to that of -le.
-Guo appeared in the 3rd century as a transitive verb that meant “cross (space or
time); exceed”. This verb guo is still in use today.
bu gan guo
transportation clothes not dare exceed standard
“(He) does not dare to use excessive transportations and clothing.”
(San Guo Zhi, 3rd century)
In the 9th century the verb guo in (9) a. extended its function through analogy into
the second verb in a serial verb construction, then through 1st reanalysis the
meaning of this guo change to an action had happened and finished, as in (9) b.
This change can be explained by pragmatic inferencing, based on its original
meaning, to have crossed a space or a time could easily mean the action is
finished. The detail of the motivation of this semantic change will be discussed
in section 3. Afterwards, through the 2nd reanalysis, this guo can be analyzed as
a verb complement in V+ guo structure. The reason to this is that adverbs cannot
be inserted in between the verb and guo in this structure.
(9) a. xiamo tiao
que’er yu.
cross bird
“The toad jumped over the bird-bath.”
(9th century poem, quoted from Li & Shi 1997: 91)
b. po
yun: “Shui bufang yin, po
granny say
water can
you yi
drink granny have one question must first
wen guo.”
“The granny said: “You can drink the water, but I have a question that I must ask first.”
(9th century Buddhism text, quoted from Li & Shi 1997: 91)
Overtime the verb complement guo gradually became very attached to the
verb. Li & Shi (1997) also find that only after the 13th century did V+ guo often
precede an object. They think this is the time when it finally grammaticalized
into an experiential marker, when the 3rd reanalysis is completed. In modern
Mandarin, the verb guo and secondary verb guo exists as well as the marker
xian da
private first
beat GUO several
“(They were) beaten up privately several time first.”
(14th century novel, quoted from Li & Shi 1997:92)
2.3 The grammaticalization of -zhe
Table 4 Path of Progression for -zhe.
Change stages
Structure and Meaning
Original structure
1st analogy & reanalysis
2nd analogy & reanalysis
3rd analogy & reanalysis
NP1+zhe (V) “attach to”
NP1+V+zhe (V2) “exist”
NP1+V+NP2+zhe (V2) “reach”+NP3
NP1+V+zhe (aspect marker) +NP2
NP1+zhe (V) “reach”+NP2
NP1+V+zhe (V)“exist” +NP2
NP1+V+zhe “durative marker” +NP2
1st century
5th century
8th century
11th century
modern Chinese
modern Chinese
modern Chinese
-Zhe first was a verb meaning “attach to” in the 1st century:
shumu, bu zhe
no attach grain
“The dew attaches to the trees, not to the grains.”
(Lunheng, 1st century, quoted from Jiang 2006:114)
During the 5th century, the verb started to change a secondary verb that meant
“exist” in existential sentences through 1st analogy & reanalysis. Through prag107
matic inferencing, when something is “attached to” something else it is likely to
stay there, thus “exist”:
Changwen shang xiao, zai
zhe che
carry ZHE carriage inside
“Changwen was still small and was put in a carriage.”
(Shishuoxingyu, 5th century, quoted from Jiang 2006:113)
Through the 2nd analogy & reanalysis, the meaning of secondary verb zhe extended to “reach” followed by a receiver through inferencing principle, to be able
to “attach”, one must “reach”:
gen jie
fei chong
fly bugs
beat at
zhe ren.
“Still more, (the bird) catches flying bugs and shoots at people.”
(Dufu poem, 8th century, quoted from Jiang 2006:113)
Around the 10th century, ooccasionally in a sentence like (14), -zhe indicates
that the event continues. I think this change comes from the verb meaning “existing”. If something exists at one place, it is likely to continue to be there. In
(15), the sentence with -zhe also means two actions are going on at the same
time, one action exists to serve as background information. Furthermore, this
zhe acts more like a grammatical marker than a verb because it is closely attached to the main verb. These usage of zhe is very close that of it in modern
Mandarin. Thus we can say that the 3rd analogy & reanalysis had happened
dui zhe
huangjing wu mai chu.
pile ZHE gold
no buy place
“(my) gold are piling up and yet no where to buy (it).”
(Wangjian poem, 10th century, quoted from Jiang 2006:113)
zhishi xiang
beat ZHE drum just
qian qu.
towards front go
“Beating the drum, (they) are just going forward.”
(Zhuzi Yulei, 11th century, quoted from Jiang 2006:113)
In the 13th century, -zhe commonly appears in sentences like (16) as an aspect
marker indicating an ongoing action:
Feng mama
ta laorenjia, wo yangji ta chuxia shi zhe shou li.
Feng madam, the elderly, I request she kitchen use ZHE hand EXP
“About old madam Feng, she is working in the kitchen under my request.”
(Jing Ping Mei, 13th century, quoted from Li & Shi 1997:93)
2.4 The grammaticalization of -zai
Table 5 Path of Progression for -zai
Change Stages
Structure and Meaning
Original structure
1st reanalysis
NP1+zai (V) “exist”+NP2
NP1+zai (prep.) “at” +NP2+V+NP3
NP1+zai (V) “exist”+NP2
NP1+zai (prep.) “at” +NP2+V+NP3
NP1+zai (progressive marker)+V+NP2
10th-5th century B.C.
10th-5th century B.C.
modern Chinese
modern Chinese
modern Chinese
2nd reanalysis & analogy
Zai is found in Shijing “the Book of Songs”, which records folk songs from about
1000 B.C. to 500 B.C.
guan Guan Jujiu,
zai he
guan Guan fish hawks
are river GEN bank
“Guan, Guan! Cry the fish hawks, which are by the river bank.”
(Shijing, Zhounan, quoted from Zhao 2001:71)
Zai in this sentence is a verb meaning “to exist”. This definition of zai is further
confirmed by its appearance in the following sentence from Lunyu “the Analects
of Confucius”:
yue: “wen
Confucius say
zai, guan
literature exist observe Pron.ambition literature die
observe Pron action.”
“Confucius said: “We can observe a person’s ambition through his writings, and if there is no
writing then we have to observe his action.”
(Lunyu, Xue’er, quoted from Zhao 2001:71)
Zai also appears as a preposition in the same texts.
zai Qishao, san
Confucius Prep. Qishao
bu zhi
rou wei.
three month no know meat taste
“Confucius didn’t eat meat for three month when he was in Qishao:”
(Lunyu, Xue’er, quoted from Zhao 2001:71)
yu zai
qi pu.
fish exist prep. Algae attach prep its leaf
“The fish is in the algae and is attaching to its leaves.
(Shijing, Xiaoya, Yuzao, quoted from Zhao 2001:71)
In (19) zai could be a verb, but it is better to analyze it as a preposition leading a
locative phrase. In (20) the distinction between the two zai is very clear: the first
one is a verb and the second one a preposition leading a locative phrase. Shao
(2005) stated that even though the usage of zai as a preposition appeared very
early, it was not common until about the 10th century. Therefore I predicate that
1st reanalysis happened to change the verb zai to a preposition. This is based on
the fact that as a verb meaning “exist”, it often precedes a locative phrase,
through pragmatic inferencing this zai can be reanalyzed as a preposition leading
a locative phrase. On the other hand, zai as an aspect marker marking an ongoing
action did not appear until early 20th century, in Lu Xun (1881-1936)’s work,
who by the way is regarded as the father of modern Chinese literature. I find in
his work a few examples using zai as an aspect marker.
fangtaitai liaoxiang ta shi zai nao zhe yi
Mrs. Fang guess
he is
wu jiaoyu.
ZAI angry ZHE she GEN no
“Mrs Fang guessed that he was angry about her ignorance.”
(Duan Wu Jie, 1922)
In (21), zai is not ‘acting along’ because it has the help of zhe to signal the ongoing action. But nevertheless it is an aspect marker because it is attached to the
verb after it. During Lu Xun’s time, the written language was under a reform to
represent more the spoken language. Moreover, never before had so much western literature been translated into Chinese. The grammaticalization of zai to an
aspect marker could have happened at that time under the influence of the language reform and contact between Chinese and Indo-European languages.
To summarize, Liu et al. (1995) have identified four factors that trigger the
grammaticalization of lexical items in Chinese. In my opinion these four factors
can be viewed as the stages of change on the path of gammaticalization: change
in syntactic position (analogy), pragmatic influence, meaning shift (reanalysis).
For example, it is very common in Chinese that first a main verb becomes a secondary verb V2 in a serial verb construction, then its meaning shifts, possibly under a specific pragmatic environment, reanalysis will finally happen. As
Peyraube (1999) discussed, although these factors are by no means prerequisite
to the grammaticalization process, their idea is refreshing in the way that it focuses on the contexts where the grammaticalization happens rather than just the
lexical item itself.
3. Motivations and mechanisms of change
Peyrube (1999:194) states that the meanings of the lexical item subject to grammaticalization are usually quite general, for example, verbs like say, move or go,
not verbs with a specific meaning like whisper. Therefore typically more basic
words or words that are easily accessible tend to be grammaticalized. This is true
with all four markers. The reason is that language change is usually motivated by
speakers’ communicative needs, which are led by human cognition process,
which motivates meaning and syntactic change in the most common words.
(Hopper & Trougott, 1993:66)
Cognitive strategies motivate meaning change, which is central in the early
stages of grammaticalization. (Hopper & Traugott, 1993:12, 68; Peyraube, 1999:
184) Why should these four specific verbs grammaticalize into aspect markers?
Because they are verbs that semantically cover wide ranges, flexible in position
and therefore easy to change meaning and induce grammaticalization. Among
the four verbs, guo “cross (space)”, zhe “attach to (location)” and zai “exist (location)” all closely related to space. Cross-linguistically there is a phenomenon
in which temporal grammatical markers often come from spatial terms (Heini et
al. 1991; Bybee et al.1994). This is because in human cognition, space and time
are very close concepts. As for le, its original meaning is “to finish; to accomplish”. Therefore there is a big possibility semantically it turns into a marker
marking the completion of an action.
Semantics and pragmatics are closely related. Traugott (1995:31) finds that
subjectification is most important in semantic change because “meanings become increasingly based in the speaker’s subjective belief/state/attitude toward
the preposition.” So the semantic change is actually a pragmatic one and subjects
to the speaker and hearer’s interpretation. Hopper & Traugott (1993) believe that
pragmatic inferencing is a motivation for grammaticalization. The speaker and
the hearer negotiate meaning in communicative situations. The speaker’s role is
based on the economical principle, which means the speaker always tries to use
least possible words to clearly express most possible information. This is why it
is natural for spatial terms to extent to signify time.
Semantic factors are very important in syntactic change, especially in Chinese
(Liu et al. 1995). In turn the syntactic environments also influence the grammaticalization of the lexical items. The three verbs le, guo and zhe were all, at
one time or another, secondary verbs in serial verb constructions. When a verb
often acts as a secondary verb in a sentence, and its syntactic position is relatively
settled, it can easily become a verb complement, and eventually a grammatical
Take these markers for example. At the beginning le appeared after continuous verbs signifying the ending of an action. At the end -le appeared after instantaneous verbs signifying the completion of an action. This is because in the mind
of language users there are not much difference between continuous verbs and
instantaneous verbs. However, for -zhe, the verbs in front of it changed from spatial verbs to any verbs, the noun phrase after it changed from location phrase to
receiver of the action. This change should be explained by semantic metaphorical extension. When verb zhe had the meaning “attach to”, it was naturally followed by a noun signifying location. Later zhe changes from reflecting the continuality of space to that of time, thus grammaticalized into the zhe that signifies
the continuality of a state. Thus it turned into a verb complement while it is still
followed by a locative noun. Therefore its meaning changed to “existing”. On the
other hand, zhe also changes from signifying the completion of a movement in
the space to the completion of time. The change from spatial relations to time relations makes it possible to break the constrain between the verbs that can appear
before zhe and the noun phrase after zhe. The change is gradual. Thus the verb
zhe “reach” appeared much later than verb zhe “exist”. But once new combinations of phrasal and syntactic structures appeared and used more and more, more
people get use to it. In turn, when a lexical item appears as a grammatical mor111
pheme again and again and is accepted in the relatively settled position, then
pragmatically grammaticalization happens. Moreover, when V2 changes into a
verb complement, its location make it easier for semantic change to becoming
more abstract, then further develop into a grammatical marker.
In conclusion, cognitive process dominated the meaning shift. However, as
Liu & Tang (2004) have argued, the grammaticalization process is not balanced,
therefore some of the markers lose the original meaning, some retains; and some
change meaning, some do not. When new meanings emerge, old meanings are
not necessarily lost, the old and new may coexist and interact with each other.
This has been proved by the four markers’ path of change.
Hopper & Traugott (1993) mentioned analogy, reanalysis, renewal, reinforcement, superposition etc. as mechanisms for grammaticalization. For these Chinese aspect markers, I identified three main mechanisms at work: reanalysis;
analogy and external borrowing. Here I will discuss about how external borrowing helped the grammaticalization of -zai.
Peyraube (1999) points out that borrowing as a mechanism of change was both
the least studied and the most abused. Unstudied, because historical linguists
strongly favored internal mechanisms, and most abused because it was often
evoked without evidence of the source of borrowing. Some universals, principles
and constraints that have been proposed on borrowing are very debatable.
Peyraube believed that it is best to consider borrowing triggered or accelerated a
grammatical phenomenon which was already growing independently. I quite
agree. And I think the grammaticalization of -zai is a good example.
It is the prepositional zai that grammaticalized into an aspect marker. Why?
The prepositional zai phrase most commonly locates before the verb indicating
the action happens at the place. Then it has a tendency to gramamticalize, based
on its syntactic structure and meaning, as in this example:
(22)a. wo zai chufang zuo
ZAI kitchen
make dinner
“I am making dinner in the kitchen.”
b. wo zai zuo fan.
ZAI make dinner
“I am making dinner.”
Based on the economical principle of speech, it is very natural to lose the word
“kitchen” in (22) a. without affecting the meaning of the sentence, thus (22) b.
appears with zai no longer a preposition but a grammatical marker precedes a
verb meaning action in progress. But why was there no example of aspect marker
zai until the 20th century? Why did it not only appeared, but also became very
common in a very short time? My hypothesis is that in spoken language by the
end of the 19th century, there could already be usage of preverbal zai marking an
ongoing action. It most probably appear together with aspect maker -zhe as
shown in example (21). But as written language and spoken language had significant difference, it is very hard to find written records of it. In the beginning
of the 20th century, the language reform to make the written language represent
spoken language made it possible that the “informal” use of aspect marker -zai
became formal in writing. At about the same time, there was large quantity of
translations from the Indo-European literature into Chinese. Rapidity of language change is a characteristic of borrowing. So I also assume that borrowing
of the present tense is another possibility for the mechanism of change in this
case. Most of the pioneers of modern Chinese literature, like Lu Xun himself,
studied or lived abroad, and/or heavily influenced by western language and literature. Language contact was certainly at work here. A quantitative study of zai
of the literature of the end of 19th century and beginning of 20th century could
better prove this hypothesis and find the exact source of borrowing.
In all, these three mechanisms reanalysis, analogy and borrowing were the
main mechanisms at play in the grammaticalization of these four aspect markers
in Chinese.
4. Conclusion
Grammaticalization is a complicated phenomenon and its process is closely related to human cognitive process. By testing the grammaticalization process of
the four aspect markers against some existing theories in this field, I hope this
study will lead to more understanding to language universals in the grammaticalization process cross-linguistically.
Bybee, Joan L., Reverre D. Perkins & William Pagliuca. 1994. The grammaticalization
of grammar: Tense, aspect, and modality in the languages of the world. Chicago: University of Chicago Press.
Chao, Yuanren. 1968. A grammar of spoken Chinese. Berkeley: University of California
Heine, Bernd, Ulrike Claudi & Friederike Hunnemeyer. 1991. Grammaticalization: A
conceptual framework. Chicago: University of Chicago Press.
Hopper, Paul J. & Elizabeth Traugott. 1993. Grammaticalization. Cambridge: Cambridge
University Press.
Huang, Lillian Meei-jin & Philip W. Davis. 1989. An aspectual system in Mandarin Chinese. Journal of Chinese Linguistics 17(1). 128–166.
Jiang, Shaoyu. 2004. Jingdai hanyu yanjiu gaiyao (sketches on modern Chinese studies).
Beijing: Beijing University Press.
Jiang, Shaoyu. 2006. The formation of verb partical /zhe/. Journal of Zhoukou Normal
University 23(1).113–117.
Li, Charles N. & Yuzhi Shi. 1997. On the mechanism behind the birth of the aspectual
markers in Chinese. Zhongguo yuwen (Chinese Langua) 2. 82–96.
Li, Charles N. & Sandra A. Thompson. 1989. Mandarin Chinese: a functional reference
grammar. Berkeley: University of California Press.
Liu, Jian, Guangshun Cao & Fuxiang Wu. 1995. Lun youfa hanyu cihui yufahua de
ruogan yinsu (On some factors that cause Chinese words to grammaticalize). Zhongguo yuwen (Chinese Lingua) 3. 161–169.
Liu Shaozhong & Tang Jianjun. 2004. Cognition, Pragmatics and Grammaticalization.
Journal of Guilin Teachers College 18 (4). 85–87.
Peyraube, Alain. 1999. Hisorical change in Chinese grammar. Cahiers de linguistique
Asie Orientale 28(2). 177–226.
Shao, Hongliang. 2005. On the evolution of /V Zai L/ structure. Journal of Shanghai Normal University 34(4). 119–124.
Smith, Carlota. 1994. Aspectual viewpoint and situation type in Mandarian Chinese.
Journal of East Asian Linguistics. 3(2). 107–146.
Sun, Chaofen. 1998. Aspectual categories that overlap: A historical and dialectal perspective of the Chinese zhe. Journal of East Asian Linguistics 7(2). 153–174.
Traugott, Elizabeth Closs & Bernd Heine (eds.). 1991. Approaches to grammaticalization, vol. 1, Focus on theoretical and methodological issues. vol 2, Focus on types of
grammatical markers. Amsterdam: John Benjamins Publishing Company.
Traugott, Elizabeth Closs. 1995. Subjectification in grammaticalization. In Dieter Stein &
Susan Wright (eds.), Subjectivity and subjectivisation: Linguistic perspectives, 31–
54. Cambridge: Cambridge University Press.
Wang, Li. 1958. Hanyu shi gao (on Chinese language history). Beijing: Kexue Publishing
Zhao, Lingzhen. 2001. Some discussions about the grammaticalization phenomenon in
classical Chinese. Journal of Jiangsu Institute of Education 17(1) 69–71.
Luying Wang
Department of Linguistics and Philology
Uppsala University
Box 635
SE-75126 Uppsala
[email protected]
The basic case marking of Ngarla, a language
of Western Australia1
Torbjörn Westerlund
Uppsala University
1. Introduction
The aim of this paper is to describe the basic case marking on nouns as well as
pronouns in Ngarla (Pama-Nyungan, Ngayarta) – a previously understudied
Australian language. Basic case marking – the marking of core clausal constituents (Dixon 2002:132–137) – is understood to comprise marking on intransitive subject (S), transitive subject (A), and transitive object (P). Some information will also be provided about peripheral case markers; dative, purposive, instrumental, and aversive.2 The analysis will show that the basic case marking in
Ngarla is complex, with no less than three major types of case marking alignment
being represented. As will be shown below the peripheral case marking is typologically interesting too. The marking of dative and possessive have for example
merged in the language, and Ngarla also appears to have a separate aversive suffix, something that is quite unusual in Australian languages.
2. Background
2.1 The Ngarla language: geography, affinity, and publication status
In literature about Australian Aboriginals the Ngarla people and their language
have alternatively been referred to as Ngurla, Ngerla, Gnalla, Wanbarda and
Kudjunguru (Berndt & Berndt 1964:71, von Brandenstein 1967:map 5, Curr
I am indebted to Alan Dench, Brian Geytenbeek, William McGregor & Anju Saxena for reading
and commenting drafts of this article, and to Geytenbeek for proving the Ngarla material on which
the article is based.
Dative marks the complement of intransitive verbs, the indirect object of transitive verbs, and also
expresses the semantic roles of purpose and beneficiary. Purposive expresses the goal of an activity,
instrumental the semantic role of instrument, and aversive marks NPs with “undesirable potential”.
C.f. Dench 1991:145, Dixon 2002:134, 137.
1886:287, Dixon 2002:xxxviii, O’Grady et al. 1966:36, 80). Traditionally, the
language was spoken along the coast of Western Australia east of Port Hedland,
in the northernmost part of the Pilbara region. And while the language had
several hundreds of speakers towards the end of the 19th century, only four
speakers remain today (Curr 1886:288, Brown, p.c.).
In the classification of the Australian languages that was presented by
O’Grady et al. (1966), Ngarla was placed together with the other languages of
the northern Pilbara in a subgroup of the Pama-Nyungan language phylum called
Ngayarta (Dench 2001:110–111, O’Grady et al. 1966:26–57, 82–83). Presently,
the following languages are considered to belong to the subgroup in question:
Jurruru, Martuthunira, Ngarla, Ngarluma-Kariyarra, Nhuwala, Nyamal, Palyku,
Panyjima, Yinhawangka, Yindjibarndi-Kurrama (Koch 2004:37).
As Dench (2001) has demonstrated, the lack of older linguistic material
however makes it impossible conclusively to established if the similarities
noted in Ngayarta languages are caused by common genetic inheritance or by
areal spread.
The Ngarla language is mentioned in a number of articles discussing languages of the Pilbara region, e.g. in O’Grady (1966), von Brandenstein (1967),
Dench (1982, 1994, 1998, 2001) and the phonology of Ngarla was described in
the previously mentioned O’Grady et al. (1966:80–82). However, prior to the
publication of Westerlund (2007) very little had been written about the grammar
and syntax of the language.
Avaliable linguistic sources in the language include a Ngarla-Italian word list,
compiled in the 1860s by one Pietro Ferrara, Charles Harper’s word list which
was included in Curr (1886:292–293), field notes made by O’Grady and by
Dench, and a fairly extensive amount of elicited material recorded and transcribed by Geytenbeek since the early part of the 1980s (Dench p.c., Geytenbeek
p.c., 2006a:3, O’Grady et al. 1966:80–82). In the study on which this article is
based material collected by Geytenbeek has been used.
2.2 The grammar of Australian languages
Australian languages in general are commonly considered to have two main
word classes, the nominal class, and the verbal class (Blake & Dixon 1979:2,
1991:3, Dixon 1980:127–128, 2002:67–68, 553–557, 643–644). The latter category includes verbs, verb phrases, and sometimes also adverbs, and the former
common and proper nouns, pronouns, demonstratives, and often also words with
“adjectival” and “adverbial” meaning. It has been found that the languages of the
Pama-Nyungan phylum all employ case marking suffixes on nominal words, and
both nouns and pronouns are suffixed to show case relationships. Dixon (2002:
131–143) sees fit to divide the generally very extensive nominal morphology of
the Australian languages into four main functional groups: core clausal functions
(covering the marking of the nominal constituents A, S, O), peripheral clausal
functions (see section 1 above), phrasal functions (covering morphology that is
used on an NP that modifies another NP, e.g. possessive), and local functions
(locative “at”, allative “to”, and ablative “from”). In this article, Dixon’s division
is followed.
Although separate markers for many of the functions mentioned above can usually be found in the Pama-Nyungan languages it is important to point out that it is
not uncommon to find instances of what Blake (1977:35–43, 60–61) has called
“case syncretism”, i.e. when individual suffixes get to mark more than one function.
As a general principle in the languages together labelled Pama-Nyungan, the
marking of core clausal functions tends to follow the nominative-accusative pattern on pronouns, and the ergative-absolutive pattern on common and proper
nouns, demonstratives, and words with “adjectival” and “adverbial meaning”, in
the way demonstrated by table 1 (where a dash is used to indicate that no overt
case marking occurs; for this section, see also Blake & Dixon 1979:2, 6, 9, 10,
1991:3, 16, Dixon 2002:66–91, 131, 134–135, O’Grady et al. 1966:80):
Table 1: Split ergativity in Australian languages (Blake & Dixon 1979:7)
Syntactic function
Transitive subject (A)
Intransitive subject (S)
Transitive object (P)
3. Ngarla marking of core clausal functions
3.1 Ergative and locative case marking on nominals
Table 2: Ergative and locative allomorphs in Ngarla3
Ergative marker
Locative marker
on dimoraic stems ending with a vowel
on dimoraic stems ending with a vowel, the
last mora of which contains a homorganic
nasal+stop cluster
on stems of more than two morae, ending with
a vowel3
on words ending with an apico-alveolar
on words ending with a palatal consonant
When it comes to the marking of core clausal functions on common and proper
nominals and demonstratives, Ngarla adheres to the pattern expected in Pama3
Compare to Dench (2001:119–120). Dench states that the -ngura allomorph can only be found in
two of the languages of the Pilbara region, Ngarla & Payungu, and that it in both of them is employed
on dimoraic stems. This is not the case in Geytenbeek’s Ngarla material. See e.g. the following sentence, where the two locative allomorphs -ngka and -ngura are used on yirrpi (“shade”) and jirrkuru
(“Mesquite tree”): Kujarra piyalu marrungu nyiniyanpula yirrpingka jirrkurungura.” There are two
people sitting in the shade of the Mesquite tree.” (Geytenbeek & Westerlund 2008:13)
Nyugan languages in general, since ergative-absolutive case marking can be
shown to be obligatory. While the intransitive subject and transitive object are both
left unmarked, the transitive subject takes one of the five ergative allomorphs
shown in table 2 above (see also Westerlund 2007:30–33). Compare the case
marking on puka when it appears in the O and A position in examples 1a, b below4:
(1) a. Julyajan palakarni mantu puka!
buried in the ground-VBLISER-IRR
mantu puka!
DEM (near) meat
rotten smell
“Bury that stinking meat!” (Geytenbeek 2008:12)
b. Pukangku nganya maturarrijiparnu.
rotten smell-ERG
“The stinking smell made me vomit.” (Idiomatic. Lit.: “stirred me into vomiting”). (Geytenbeek
In the study of the marking of core clausal functions in Ngarla the locative
markers have also been looked at. Ergative and locative allomorphs are usually
related in the Australian languages, and most frequently they only differ on the
final vowel (Blake 1977:51, Dixon 2002:157–166, Hale 1976:414–417). There
is not an abundance of examples of the use of all five ergative allomorphs of the
language available in the linguistic material perused. The distribution of the
locative allomorphs however appears to follow the same rules as the distribution
of ergative allomorphs, and taken together a number of complex distribution patterns emerge.
O’Grady (1966:75) stated that a distinctive feature of what he called “western
languages” (of Australia) is that a “morphophonemic alternation in the form of
the “agent- instrumental suffix” *-lu/-ngku conditioned by the length of the word
stem” can be found. This means that different allomorphs are used depending on
if a vowel-final stem contains two morae or more than two morae (Dench 2001:
118–119). In the Ngarla material, three of the sets of allomorph shown in table
2, -ngku/-ngka, -ku/-ka, and -lu/-ngura, have been found to be distributed according to such rules, the first two sets appearing on dimoraic stems (the second set
of which having a very limited distribution; for an example with -ngku, see sentence 1b above) and the third set on trimoraic and longer stems, as in 2:
The transcription of all the example sentences has been done by Geytenbeek, using the phonemic
alphabet developed for the languages of the eastern Pilbara region by Geytenbeek & Geytenbeek,
Hudson, Richards, and Marsh & Marsh (Geytenbeek, p.c.). The morpheme-by-morpheme analysis
of the sentences is my own. Abbreviations used in examples in this article are: 0: zero-marked morpheme, 1 PL EXCL: 1st person plural exclusive, 1 PL INCL: 1st person plural inclusive, 1SG: 1st person singular, 2SG: 2nd person singular, 3PL: 3rd person plural, 3SG: 3rd person singular, ABS: absolutive, ACC: accusative, ALL: allative, AVERS: aversive, BUFF: buffer morpheme, COM:
comitative, DAT: dative, DEM: demonstrative, ERG: ergative, HORT: hortative, ImmPAST: immediate past, INSTR: instrumental, NEG: negation, NMLISER: nominalizer, IRR: irrealis, NOM:
nominative, PAST: past tense, PERM: permissive, PRES: present, POSS: possessive, PURP: purposive, VBLISER: verbaliser.
Nyampalilu ngajapa juntumarnu, pakurta nyayi jankurna.
Nyampali-lu nga-japa juntu-ma-rnu,
bad, no good
DEM (here) emu
The boss told me, “This emu is no good.” (Geytenbeek 2008:12)
Two other sets of ergative and locative allomorphs used in Ngarla, -tu/-ta and
-ju/-ja, are exclusively employed according to phonological rules. The first set
appears on nominals ending on apico-alveolar consonants, and the second set on
words ending on palatal consonants, in both cases regardless of word-length.5
3.2 Case marking on pronouns
As was shown by Dench (1994), the pronouns of Ngarla do not follow one, but
two different case marking systems. The dominating alignment type is the
nominative-accusative one, which is to be expected for pronouns in languages of
the Pama-Nyungan phylum (see section 2.2 above). All pronouns except first and
second person singular follow this alignment. 1 and 2SG, however, follow the
tripartite system which is by Dench (1994:167–170, 182–185, 187–190, 2001:
123) considered to be the original case marking system for singular pronouns in
the Ngayarta languages.6 The differences in marking should become clear from
the comparison of the marking of the intransitive subject, the transitive subject,
and the transitive object on first and third person singular pronouns in examples
3 a–c and 4 a–c below:
(3) a. Karlinyjarriyan ngaya.
“I am coming back.” (Geytenbeek 2006b:18)
b. Ngaja jaarnu warnta.
“I chopped the tree.” (Geytenbeek 2006b:1)
c. Ngalkarrtu pajirnu nganya.
Ngalkarr-tu paji-rnu
bite, eat-PAST
“The ant bit me.” (Geytenbeek 2006b:51)
Concerning the ergative markers of Ngarla, see also Westerlund 2007:34.
This means that Ngarla does not adhere to the case-marking split described as universal by Silverstein (1976:113, 116–119, 159–164), who states that pronouns higher in animacy take nominative-accusative marking, while pronouns lower receive ergative-absolutive marking (with the possibility of tripartite marking in the “mid section”). The following personal pronouns have been found
in Ngarla: 1st, 2nd, 3rd person singular, 1st person dual inclusive, 1st person dual exclusive, 2nd, 3rd person dual, 1st person plural inclusive, 1st person plural exclusive, 2nd, 3rd person plural. For more information about Ngarla pronouns, see Dench (1994), and Westerlund (2007:42–53).
(4) a. Palura wangkakarriyanu juntu.
3SG:NOM speech, talk-VBLISER-ImmPAST
“He spoke rightly.” (Geytenbeek 2006b:64)
b. Palura ngalila nyanta karrin mantu.
nyanta karri-n
carry-FUT meat, animal, bird
“He'll bring the meat over to us.” (Geytenbeek 2006b:61)
c. Yukurrulu parnunya pajirnu.
parnu-nya paji-rnu.
tame dog-ERG 3SG-ACC
bite, eat-PAST
“The dog bit him.” (Geytenbeek 2006b:32)
4. Ngarla case marking of peripheral clausal functions
4.1 Dative
Two dative markers have been found to be employed on Ngarla common and
proper nouns and demonstratives, -rra, which occurs after vowels, and -ku, used
after consonants. As the examples in 5 show, these suffixes are also used to express
the semantic roles of purpose and beneficiary, which, as stated in section 2.2, is not
uncommon in Australian languages (see also Westerlund 2007:35–36):
(5) a. Nyukapilu nyinpa jankurnarra.
increase site or feature-VBLISER-HORT
“You should do the increase ceremony for the emu.” (Geytenbeek 2006b:64)
b. Ngayinyku ngaya para wajarriyan.
general term for internal organs-DAT 1SG-ABS
“I'm short-winded” (Lit. “I'm searching for breath”). (Geytenbeek 2006b:23)
c. Partanyal-ku jilyarra waan.
Partanyal-ku jilya-rra
child-BEN give-IRR
“Give it to one child.” (Geytenbeek 2008:26)
A clear case of syncretism can however be demonstrated for the Ngarla dative
markers, since it also turns out that they are used to express alienable possession
(the possessive function being one of Dixon’s phrasal functions, see 2.2 above
and example 6 a below).7 The same also goes for the pronominal dative marker
-nga, as shown by examples 6 b & c:
Inalienable possession is expressed by juxtaposition. See Westerlund 2007:52.
(6) a. Palakarnilu pungarnu para karlajangu jarntu Piyitarra.
DEM (near)-ERG hit, kill-PAST 3SG-DAT cattle
tame, friendly Peter-POSS
“That chap killed Peter's pet bull.” (Geytenbeek 2006b:84)
b. Karingarriyan palakarni parnunga murri.
feel sympathetic-VBLISER-PRES
DEM (near) 3SG-DAT
“The man felt very sorry for that chap.” (Geytenbeek 2006b:48, 49)
c. Karrarrulu jungkarri parnunga parru-parru.
Karrarru-lu jungka-rri
parnu-nga parru-parru.
spider-ERG make a net-PRES
“The spider is spinning his net.” (Geytenbeek 2006b:20)
4.2 Purposive
The purposive function, as mentioned above, occurs in phrases expressing the
goal of an activity, e.g. “going out FOR (to catch) KANGAROOS”, “call them
FOR (i.e. to eat) FOOD”. This function is commonly expressed in nominalized
clauses in Australian languages, but only a smaller number of languages have
been found to have a separate purposive suffix, the function instead usually being
expressed as a secondary sense of some other suffix, often the dative marker
(Dixon 2002:134–135).
Despite the relative rarity of separate purposive markers, the studied Ngarla
material indicates that the language does not have only one, but two separate suffixes marking the purposive function, -yartara and -wanti. Both suffixes are used
on nominals and on nominalized verbs, but the latter suffix appears to be more
restricted in use.8 Westerlund (2007:54–58) showed that Ngarla has two main
verbal conjugations, there called the yan and rri conjugations, for their respective
present tense markers. While the purposive suffix -yartara is employed on
nominalized verbs of both conjugations, -wanti in the material only occurs on
nominalized verbs of the rri conjugation. The reason for this difference in use, if
it is a valid difference, has yet to be discovered:
(7) a. Wula ngaja punyjarnu kupalyayartaralu.
Wula nga-ja
punyja-rnu kupalya-yartara-lu.
Water 1SG-ERG drink-PAST
“I drank some water in readiness for a sleep.” (Geytenbeek 2008:116)
In the material analyzed, the suffix -yartara appears 19 times, only 5 of which on verbs carrying
nominalizing suffixes. In the remaining cases it is used on bare verbal stems (11 times) and on
nominal words (3 times). Zero nominalization is however used for Ngarla verbs of the yan conjugation, as shown by example 9, and with that in mind the -yartara suffix needs to be analyzed as a
Ngarla nominal suffix (compare this to the situation in Nyamal; section 5 below).
b. Palakarni puti-putiman yaanmara wulakarni nganyjarranga
Palakarni puti-puti-ma-n
yaa-nmara wula-karni
nganyjarra-nga punyja-l-wanti.
”Make him go and get some water for us to drink.” (Geytenbeek 2008:95; the combination
puti-puti ma-rri means “to make someone do something” (Geytenbeek 2008:95)9.)
4.3 Instrumental
According to Dixon (2002:135–136), all Australian languages mark the instrumental function, a function that is generally employed to describe the use of a
weapon or tool (“he hit it WITH A CLUB”, “she cut it WITH A KNIFE”; see
example 8 a below), and which frequently also extends to body parts (“the
crocodile held me WITH ITS CLAWS”). The number of languages that have distinct instrumental case suffixes are however stated to be few, the function instead
commonly being marked by suffixes having the same shape as the ergative case
The Ngarla ergative and instrumental case markers have indeed been found
to be identical in shape and distribution (see table 2 above, and also Westerlund
2007:30–34). The question of how to label these suffixes in the Ngarla language is however complicated by the fact that they have been spread to be used
also to create different types of adverbials (specifically those expressing manner and cause/reason), and to provide spatial orientation, as illustrated by examples 8 b–c:
(8) a. Kunyjarri pananya payinyjarnu jilamankartalu yartara.
pana-nya payinyGrey Teal (Anas gibberifrons) 3PL-ACC
jilaman-karta-lu yartara.
lying around
“He shot them with the gun, and got the lot, ducks lying around everywhere!” (Geytenbeek
b. Jakarntu kanyin.
care, caution-ERG/INSTR(?) 1) look after, take care of, 2) tread on, kick-IRR
“Tread on it carefully!” (Geytenbeek 2008:1)
c. Putangara nganarna yangarnu yawartakartalu.
Perentie (Varanus giganteus) 1PL EXCL:NOM
follow, chase-PAST
“We used to follow goannas on horseback.” (Geytenbeek 2006c:4)
In Geytenbeek’s database the exact meaning of the nominal element in a nominal + verbaliser-compound is frequently not explicitly stated.
4.4 Aversive
Aversive is another function characteristically marked by nominal morphology
in Australian languages (Dixon 2002:137). An NP carrying the aversive suffix is
considered to have undesirable potential, and the verb of the same clause typically describes what was, can or should be done in order to avoid the referent in
question. The number of Australian languages that have a separate aversive suffix is however stated to be few, the function instead commonly being the secondary sense of some other suffix (usually the locative, ablative, causal, dative/purposive, or allative suffix; Dixon 2002:137, 171).
The study of Geytenbeek’s Ngarla material however indicates that the language does in fact belong among the small group of languages where a separate
aversive marker can be found. In Ngarla the suffix -katangka is used on both
nominals and nominalized verbs carrying undesirable potential10:
Nyapirilu nganya kartuwarramarnu nyirukarrikatangka.
Nyapiri-lu nga-nya
Nyapiri-ERG 1SG-ACC
saviour, rescuer-VBLISER-PAST
”Nyapiri saved me from drowning.” (Geytenbeek 2008:25; the combination nyiru karri-yan
means “to drown” (Geytenbeek 2008:72).)
5. An attempt at typology: a few comparisons to case
marking in other Ngayarta languages
The languages that form “the core” of the Ngayarta group, Martuthunira,
Ngarluma, Panyjima, and Yindjibarndi-Kurrama, have all, quite uncommonly in
Australia, developed a nominative-accusative case marking pattern for all
nominals. Dench (1982:43–47, 2001:120–121, 126–127) has shown that this pattern has developed out of an older “extended intransitive frame” (S-DAT). As
demonstrated in section 3.1 above, however, Ngarla, being the northernmost
Ngayarta language (von Brandenstein 1967:map 5, Dench 1982:44), has, just
like the other languages found towards “the rims” of the Ngayarta area, retained
the original ergative-absolutive case marking pattern for common and proper
nouns and demonstratives.
When it comes to the marking of peripheral clausal functions, it can be noted
that the kind of case syncretism demonstrated for the Ngarla dative markers in
section 4.1 above also is present in neighbouring Nyamal. Nyamal however employs the dative/possessive markers -ku/-yu (Dench 1999:ch. 5, pp. 7–11).
This suffix could be analyzed as a combination of -kata and the locative -ngka. If the aversive suffix should indeed be sub-divided in this way is however presently unclear. For one thing, it is uncertain if the language has a separate -kata suffix. See table 2 above, and compare to the discussion about
aversive marking in Nyamal in section 5 below!
Unlike in Ngarla, however, the dative suffixes of Nyamal also serve to mark
the purposive function (Dench p.c., 1999:ch. 5, pp. 7–8). Suffixes being identical
or similar to the Ngarla purposive suffixes however exist in the language, but
where -wanti in Nyamal is used to fill the function labeled “intentive nominalization” by Dench (1999: ch. 5, pp. 47–50, ch 7, pp. 12–13), i.e. to denote intended future purpose, the suffixes -yartaa/-larta serve as verbal purposive
markers (see section 4.2 above; Dench 1999, ch. 7, pp. 33–34). Separate purposive markers have not been found in the other two Ngayarta languages described
by Dench, Martuthunira and Panyjima (1991, 1995), thus indicating that Ngarla
in this respect stands out among the languages of the group (Dench 1991:137–
153, 1995:63–99).
Separate aversive markers are usually not found in the languages of the Pilbara
region, and consequently no such markers have been reported for Martuthunira
and Panyjima (Dench 1991:136–153, 1995:73–99). For Nyamal the question is
however if the marker -martila should be interpreted as an aversive marker, or
as a combination of the “full laden” suffix -marti plus the locative -la, together
serving the aversive function (Dench 1995:ch. 5, pp 42–43).
6. Conclusions
The nominal morphology of Ngarla is very extensive, just like nominal morphology in many other Australian languages (Dixon 2002:131). The marking
of “core clausal functions” in the language has been shown to be quite complex, with no less that three major types of case marking alignment being represented. On common and proper nouns and demonstratives the ergative-absolutive pattern is followed, and in this regard the language is unlike the central
Ngayarta languages to the south-west of it, but alike Pama-Nyungan languages
in general. Five different ergative allomorphs are in use (-ngku, -lu, -tu, -ju,
-ku), and complicated rules have been demonstrated to govern the use of each
allomorph. When it comes to pronouns, marking on first and second person
singular follows the tripartite system, while remaining pronouns take nominative-accusative suffixes.
Just as in neighbouring Nyamal, the marking of non-core dative and possessive arguments have merged in Ngarla, leaving the same set of suffixes to be
used for both functions (in the case of Ngarla pronouns -nga, for common and
proper nouns and demonstratives -rra (after vowels)/-ku (after consonants)).
However, unlike many Australian languages, the dative suffixes are in Ngarla
not employed to mark purposive (Dixon 2002:134–135). Instead, the language
employs two separate suffixes marking this function (-yartara and -wanti). As in
most other Australian languages, however, the instrumental function is marked
by suffixes having the same shape and form as the ergative suffixes. These suffixes are however in Ngarla also used to fill a couple of other functions, which
leaves some doubt as to how they should properly be labeled. Aversive, finally,
is in most Australian languages the secondary sense of some other suffix, but is
in Ngarla marked by a separate aversive marker (-katangka).
To conclude it can be said that the marking of core and peripheral clausal functions in Ngarla both follow the expected norm for Pama-Nyungan languages, and
at the same time stands out in interesting ways. The nominal morphology of the
language will however have to undergo a more detailed study for all the characteristics of it to be fully understood.
Unpublished material
Dench, Alan. 1999. A Grammar of Nyamal. A language of north western Australia.
Dench, Alan. 2007–08. Personal communication via e-mail.
Geytenbeek, Brian. 2006a. NgarlaDicIntro.
Geytenbeek, Brian. 2006b. Ngarla-EnglishDictionary31Oct06.
Geytenbeek, Brian. 2006c. Ninengla.
Geytenbeek, Brian. 2008. Ngarla-EnglishDictionary29Apr08.
Geytenbeek, Brian. 2006–08. Personal communication via e-mail.
Geytenbeek & Westerlund. 2008. Ngarlatexts.
Published material
Berndt, Ronald M & Catherine H Berndt. 1964. The World of the First Australians, An
introduction to the traditional life of the Australian Aborigines. Chicago: University
of Chicago Press.
Blake, Barry J. 1977. Case marking in Australian languages. Canberra: Australian institute of Aboriginal studies.
Blake, Barry J & R M W Dixon. 1979. Introduction. Handbook of Australian Languages,
Volume I, Barry J Blake & R M W Dixon, editors. Amsterdam: John Benjamins B. V.
Blake, Barry J & R M W Dixon. 1991. Introduction. The Handbook of Australian
Languages, Volume 4 The Aboriginal Language of Melbourne and Other Grammatical Sketches, Barry J Blake & R M W Dixon, editors. South Melbourne: Oxford
University Press Australia. 1–28.
von Brandenstein, C G. 1967. The language situation in the Pilbara – past and present.
Papers in Australian Linguistics, 1967 (2). 1–20a + maps 1–7.
Curr, Edward M. 1886. The Australian Race: Its Origin, Languages, Customs, Place of
Landing in Australia and the Routes by which It Spread Itself over that Continent,
Volume 1. Melbourne: John Ferres & London: Trübner & Co.
Dench, Alan. 1982. The Development of an Accusative Case Marking Pattern in the
Ngayarda Languages of Western Australia. Australian Journal of Linguistics. 2. 43–59.
Dench, Alan. 1991. Panyjima. The Handbook of Australian Languages, Volume 4 The
Aboriginal Language of Melbourne and Other Grammatical Sketches, Barry J Blake
& R M W Dixon, editors. South Melbourne: Oxford University Press. 125–243.
Dench, Alan. 1994. The Historical Development of Pronoun Paradigms in the Pilbara Region of Western Australia. Australian Journal of Linguistics. 14. 155–191.
Dench, Alan Charles. 1995. Martuthunira, A Language of the Pilbara Region of Western
Australia. (Pacific linguistics. Series C, 125.) Canberra: Dept. of Linguistics, Research School of Pacific and Asian Studies, the Australian National University.
Dench, Alan. 1998. What is a Ngayarta Language? A Reply to O’Grady and Laughren.
Australian Journal of Linguistics. 18 (1). 91–107.
Dench, Alan. 2001. Descent and Diffusion: The Complexity of the Pilbara Situation.
Areal Diffusion and Genetic Inheritance, Alexandra Y. Aikhenvald & R. M. W.
Dixon, editors. Oxford: University Press. 103–133.
Dixon, R M W. 1980. The languages of Australia. Cambridge: University Press.
Dixon, R M W. 2002. Australian Languages – Their Nature and Development.
Cambridge: University Press.
Hale, Kenneth. 1976. On Ergative and Locative Suffixial Alternations in Australian
Languages. Grammatical Categories in Australian Languages, R M W Dixon, editor.
Canberra, New York: Australian Institute of Aboriginal Studies, Humanities Press
Koch, Harold. 2004. A Methodological History of Australian Linguistic Classification.
Australian Languages, Classification and Comparative Method, Claire Bowern &
Harold Koch, editors. Amsterdam & Philadelphia: John Benjamins. 17–60.
O’Grady, Geoffrey N. 1966. Proto-Ngayarda Phonology. Oceanic Linguistics. 5(2). 71–
O’Grady, Geoffrey N, C F Voegelin & F M Voegelin. 1966. Languages of the World:
Indo-Pacific Fascicle Six. Anthropological Linguistics. 8(2). 1–161.
Payne, Thomas E. 1997. Describing Morphosyntax, A guide for field linguists.
Cambridge: University Press.
Silverstein, Michael. 1976. Hierarchy of features and ergativity. Grammatical Categories
in Australian Languages, R M W Dixon, editor. Canberra & New York: Australian
Institute of Aboriginal Studies, Humanities Press Inc. 112–171.
Song, Jae Jung. 2001. Linguistic Typology, Morphology and Syntax. Harlow, England:
Pearson Education.
Westerlund, Torbjörn. 2007. A grammatical sketch of Ngarla: A language of Western
Australia. South Hedland: Wangka Maya.
8. Contact information
Torbjörn Westerlund
Department of Linguistics and Philology
Uppsala University
Box 635
751 26 Uppsala
[email protected]
The causative/inchoative alternation
in Icelandic
Toshiko Yamaguchi1
University of Malaya
1. Introduction
This paper examines the behaviour of transitive and intransitive verbs in the diathesis alternation in Icelandic – known as the causative/inchoative alternation. It
examines the extent to which causative verbs alternate with their inchoative
counterparts. We are concerned with such a pair as shown in (1).
(1) a. Jane broke the vase.
b. The vase broke.
The hallmark of the causative/inchoative alternation is that the thematic role of
the direct object in the transitive sentence is identical to that of the subject of the
intransitive sentence (see Levin and Rappaport Hovav 1995: 79). In (1), the vase
is the patient or affected argument in both sentences. More precisely, (1a) refers
to an event in which the causer, Jane, is responsible for the occurrence of the denoted event, that is, the breaking of the vase, while (1b) describes an event constituted by a change from one situation where an entity is present (the vase being
unbroken) to another situation where the entity is absent (the vase being broken).
In this regard, inchoative is synonymous to change-of-state (Frawley 1992: 12).
The crucial difference between the two variants is that the causative sentence linguistically makes it explicit who has caused the event to take place, while the inchoative sentence is silent about it. Hence, unlike passives, inchoatives do not
express the causer/agent (*The vase broke by Jane versus The vase was broken
by Jane) in syntax (see also Arce-Arenales et al. 1994: 14). Because the causer
does not receive linguistic coding, the general consensus shared by specialists in
the field is that inchoatives describe a ‘spontaneous event’, meaning that the
event is conceptualized as occurring without the involvement of an external
I am grateful to Magnús Pétursson who provided me with examples and patiently attended to the
questions that arose during the period in which I prepared and wrote this paper.
causer (Klaiman 1991: 30–1, 74, Kemmer 1993: 142–7, Haspelmath 1993: 90,
Davidse and Heyvaert 2007: 72, Levin and Rappaport Hovav 1995: 92, among
others). (1a) and (1b) can thus be brought about by the same factors but (1b) emphasizes, or brings to the fore, the absence of an agent. Importantly, scholars seek
to identify the semantics of spontaneous events in the meaning of the verb.
Haspelmath (1993: 93) observes the linkage between a spontaneous event and
what he calls ‘the absence of agent-oriented meaning components’. This indicates that verbs expressing change-of-state are those most likely to license the alternation because a change-of-state can be conceptualized without specific reference to the agent. A similar observation is made independently by van Voorst
(1995: 513ff.) who states that the causative/inchoative alternation occurs through
the combination of the event-initiator’s lack of control (or independent process)
and the potential of a change-of-state. Haspelmath (1993: 94) provides the condition, as given in (2), under which the causative/inchoative alternations occur.
This condition explains why English verbs such as cut, build, paint, dig, wash
and read do not permit the alternation. The answer is that these verbs require the
strong agentive involvement in conducting the entire event.
(2) A verb meaning that refers to a change-of-state […] may appear in an
inchoative/causative alternation unless the verb contains agent-oriented
meaning components or other highly specific meaning components that
make the spontaneous occurrence of the event extremely unlikely.
Icelandic exhibits various phenomena that contradict this generalization. First,
an agent, though non-volitional or secondary, can appear in inchoatives (sections
3.1 and 4). Second, instruments and manner adverbials that imply the presence
of the agent (see also Delancey 1984; Schlesinger 1995) serve as obligatory adjuncts to save the grammaticality of the inchoative (section 3.2). Third, verbs
(such as decide, see, explain) that do not express change-of-state, or what Dowty
(1991: 577) calls ‘primary transitive verbs’ (such as build, eat), license the
causative/inchoative alternation (section 3.3). In this paper, I will argue that
spontaneity is not a decisive factor for licensing the alternation in Icelandic although it is an important meaning of the inchoative construction. The decisive
factor is that the inchoative construction treats change-of-state and the resultant-state as two independent semantic categories, and crucially, that the semantics of the resultant-state is based largely on the speaker’s conventional
knowledge of an event (described by the predicate) and/or contextual information available in discourse.
This paper is organized as follows. Section 2 describes the morphosyntax of
alternating verbs. Section 3 looks closely at the semantic factors crucial to the
understanding of the alternation in Icelandic. Section 4 examines, albeit cursorily, some examples from the perspective of the Transitivity Continuum (Hopper
and Thompson 1980). Section 5 concludes the paper.
2. Morphosyntax of causative and inchoative verbs
Causative verbs are formed lexically, while inchoative verbs are formed either
lexically or morphologically. Examples (3) and (4) show cases with lexical
verbs. Where the lexical inchoative verbs are concerned, there are two groups of
verbs: one group takes the same form, while the other the different forms. Note
that inchoative verbs in the second group typically contain the infix -n-, an indicator of change-of-state. Verbs that behave like (4) are listed in (5). The infix -nis in boldface.
(3) Causative and inchoative share the same form:
a. Causative lenda ‘land’
‘The pilot landed the plane’.
b. Inchoative lenda ‘land’
‘The plane landed’.
(4) Causative and inchoative have different forms:
a. Causative brjóta ‘break’
Jón braut
‘Jón broke the window’.
b. Inchoative brotna ‘break’
‘The window broke’.
(5) List of verbs
English equivalent
make wet
make better
make white
make black
English equivalent
get wet
get better
become white
get split
become suffocated
get torn
get interrupted
become untied
get torn
become black
get awaken
Examples (6) and (7) show cases in which inchoative verbs are formed morphologically, that is, by adding the suffix -st to the transitive verb. This suffix is derived from an old Icelandic reflexive sik ‘self’, which corresponds to the reflexive sig ‘self’ in Modern Icelandic. These two forms are used independently,
that is, the reflexive sig does not take the same function as -st but functions only
as a reflexive, as shown in three examples in (8). While lexical inchoatives are
limited in number, morphological inchoatives are abundant in Icelandic, which,
interestingly enough, includes verbs that do not lexically entail change-of-state
in the direct object, as shown in (7), (9), (10) and (11).
(6) Addition of the suffix -st: change-of-state verb
a. Causative
fly.the.NOM.PL damage.3PL.PAST painting.the
‘The flies damaged the painting’.
b. Inchoative with -st
painting.the.NOM damage.3SG.PAST
‘The painting got damaged’.
(7) Addition of the suffix -st: non-change-of-state verb
a. Causative
Ég ákvarðaði
decide.PAST.1SG hotel.the.ACC.SG
‘I chose the hotel’.
b. Inchoative with -st
‘The hotel was chosen’.
(8) Verbs that permit reflexive sín, sér and sig.
a. Anna
Anna.NOM dress.3SG.PAST
‘Anna dressed herself’.
b. Barnið
child.the.NOM wash.3sg.PAST ref.DAT
‘The child washed him/herself’.
c. Kennarinn
be.ashamed.3SG.PAST ref.GEN
‘The teacher was ashamed’.
Examples (9) and (10) exemplify how the case marking operates when transitive
and intransitive verbs alternate. The object of transitive sentences can be marked
by three cases, either accusative, dative or genitive, as shown in the (a) examples.
When intransitive sentences are constructed, the nominative case is employed
for the subject, as shown in the (b) examples. This stands in contrast to passives,
which can retain the genitive and dative (but never accusative) for the subject, as
shown in (10c) and (11c). (12) is a list of verbs taking the genitive case and not
(9) Causative object with accusative; inchoative/passive subject with
a. Gunnar
explain.3SG.PAST the.novel.ACC.SG
‘Gunnar recounted the novel’.
b. Sagan
‘The novel was recounted’.
c. Sagan
novel.the.NOM be.3sg.PAST
‘The novel was recounted’.
(10) Causative object with dative; inchoative subject with nominative; passive
subject with dative
a. Pétur
‘Peter rolled the stone’.
b. Steinninn
‘The stone rolled’.
c. Steininum
‘The stone was rolled’.
(11) Causative object with genitive; inchoative subject with nominative; passive
subject with genitive
a. Gunnar
‘Gunnar wanted help’.
b. Hjálp
help.NOM want+st.3SG.PAST
‘The help was wanted’.
c. Hjálpar
be.3SG.PAST want.3SG.PP
‘Help was wanted’.
(12) sakna ‘miss’, gœta ‘take care of’, vænta ‘expect’, geta ‘mention’, krefja
‘demand’, þurfa ‘need’
3. Explaining the alternation
This section demonstrates that there are essentially three semantic properties that
underlie the alternation. These are ‘change-of-state’ (3.1), ‘agentive involvement’ (3.2) and ‘resultant-state’ (3.3). As noted earlier, spontaneity is not a crucial element in Icelandic. This means that the notion of change-of-state is not a
defining characteristic of the alternation.
3.1. Change-of-state
Verbs such as brjóta ‘break’ or beygja ‘bend’ that lexically encode the meaning
of change-of-state can have an intransitive that expresses an inchoative situation.
What is denoted in (13b) and (14b) is a change from being unbroken/unbent to
being broken/bent. Verbs illustrated under (5) behave in exactly the same way as
do these two verbs.
(13) a. Jón braut
Jón break.3sg.PAST window.the.ACC.SG
‘Jón broke the window’.
b. Glugginn
‘The window broke’.
c. Glugginn
break.3SG.PAST at
‘Jón broke the window undeliberately’.
(14) a. Pétur
bend.3sg.PAST stick.the.ACC.SG
‘Pétur bent the stick’.
b. Stafurinn
‘The stick bent’.
c. Stafurinn
‘Pétur bent the stick undeliberately’.
As shown in (13c) and (14c), however, the entailment of change-of-state does
not necessarily mean that the role of the agent is excluded from a denoted eventuality. Icelandic inchoatives are able to express the agent syntactically when the
agent is non-volitional or secondary to the event. In other words, if the agent carries out the action volitionally, the transitive construction is preferred. Although
(13b) and (14b) can express a spontaneous event in the sense that we do not know
exactly who is responsible for the denoted activity, spontaneity is not cognate
with the ‘non-presence’ of an agent.
3.2. Agentive involvement
Agentive involvement can be examined through the use of an instrumental
phrase. As early as Fillmore (2003 [1970]: 127), linguists were aware that English can allow an inanimate entity such as an instrument to be the subject of a
transitive construction. The reason for this use of an instrument is that the instrument (such as a key) can replace the agent because it is manipulated by the
agent. An interesting fact in Icelandic is that although an instrument does not
appear in a subject position of the transitive, as shown in (15b) and (16b), the
presence of an instrument saves the grammaticality of the inchoative construction, as shown in (15d) and (16d). This behaviour can be accounted for by the
nature of cutting and peeling, since these activities cannot, by definition, be accomplished without the use of a knife. This indicates that the use of an instrument is deemed to be an integral part of an activity, or to put it differently, the
instrument serves as what Schlesinger (1995: 63–4) calls a ‘tool’ in the sense
that it ‘performs an activity in its entirety’. Without using a knife, the activity
of cutting will not be implementable.
(15) a. Bakarinn
‘The baker cut the bread’.
b. *Hnífurinn
knife.the.NOM.SG cut.3SG.PAST bread.the.ACC.SG
‘The knife cut the bread’.
c. *Brauðið
‘The bread cut’.
d. Brauðið
‘The bread was cut with the knife’.
e. Brauðið
cut+st.3SG.PAST well with
‘The bread was cut well with the knife’.
(16) a. Jóa
‘Jóa peeled the apple’.
b. *Hnífurinn
peel.3SG.PAST apple.the.ACC.SG
‘The knife peeled the apple’.
c. * Eplið
‘The apple was peeled’.
d. Eplið
‘The apple was peeled with the knife’.
e. Eplið
‘The apple was peeled with the knife’.
Exactly the same behaviour is observed with the verb ryksuga ‘vacuum’, as
shown in (17d) and (17e).
(17) a. Strákurinn
‘The boy vacuumed the floor’.
b. *Ryksugan
vacuum.cleaner.NOM.SG vacuum.3SG.PAST floor.the.ACC.SG
‘The vacuum cleaner vacuumed the floor’.
c. *Gólfið
‘The floor was vacuumed’.
d. Gólfið
with vacuum.cleaner.the.DAT.SG
‘The floor was vacuumed with the vacuum cleaner’.
e. Gólfið
floor.the.NOM.SG vacuum+st.3SG.PAST well with vacuum.cleaner.the.DAT.SG
‘The floor was vacuumed well with the vacuum cleaner’.
An interesting fact is that, as shown in (15e), (16e) and (17e), the sentence
sounds even more natural when the manner adverbial vel ‘well’ is supplemented.
Since this adverbial verbalizes the state of an entity in conjunction with the accomplishment of an activity, this grammaticality judgment substantiates that Icelandic inchoatives express the resultant-state, the notion we now attend to.
3.3. Resultant-state
An intriguing fact in Icelandic is that agent-oriented verbs can participate in the
alternation. The key explanation for this might be that the inchoative construction serves as a resultative but not merely as a change-of-state. To define what
the resultative is, I adopt Nedjalkov and Jaxontov (1988: 6) who state that the resultative expresses a state that arises as a result of a previous event. This definition stands in contrast to that of the stative which expresses natural or primary
states of things without making any implication as to their origin. In this paper,
I use the expression ‘resultant-state’ in place of ‘resultative’ to make it explicit
that resultative is a state that occurs or takes place obligatorily after the completion of a previous event. Consider examples with borða ‘eat’ and þvo ‘wash’, as
shown in (18) and (19), as well as útskýra ‘explain’, as shown in (9a,b). The
scenarios that both (18b) and (19b) depict are states arising from the events that
previously occurred. The former refers to an empty plate that exists on a table as
a result of someone having eaten the cake. The latter refers to the carpet, which
was dirty before and is now stretched fresh on the floor after having been
(18) a. Anna
‘Anna ate the cake’.
b. Kakan
‘The cake was eaten’.
(19) a. Móðirin
wash.3SG.PAST carpet.the.AC.SG
‘The mother washed the carpet’.
b. Teppið
‘The carpet was washed’.
The behaviour of these verbs runs counter to Haspelmath’s (1993) generalization
(see (2) above) in three aspects. First, all the verbs express an event that includes
a change-of-state.2 Second, each event is controlled clearly by an external agent.
Third, the causative/inchoative alternation is feasible. I propose that this observed incongruence can be resolved if we separate the resultant-state from the
Haspelmath (1993: 93) accepts that ‘wash’ is a change-of-state verb, while Van Voorst (1995: 507)
classifies ‘eat’ as a change-of-state verb.
change-of-state by treating them as two independent semantic categories. A lack
of consideration of result in its own right may lie ultimately in the general trend
in the linguistic literature. As Gorlach (2004: 47–8) rightly claims, the notion of
result has been applied intuitively and peripherally in studies on grammatical
phenomena. She goes on to say that result has often been merged with the perfective aspect or the passive voice, so it has failed to enjoy an independent status.
For example, Levin and Rappaport Hovav (2005: 2) maintain that the passive
voice The window was broken allows both eventive and stative readings. The fact
that the concept of result does not surface in Haspelmath’s generalization might
explain that result is seen to be merged into change-of-state. Against this background, it is worth mentioning Frawley (1992: 183) who distinguishes resultant-state (20b) from change-of-state (20a).
(20) a. My circumstances changed.
b. My circumstances changed into a nightmare.
Intransitive verbs in English such as break may express a situation denoted by
(20a) (see also Fillmore 2003 [1970]: 130), whereas Icelandic intransitive verbs
such as borða ‘eat’ and þvo ‘wash’ may express situations denoted by (20b). It
is important to note here that the semantic scope of the resultant-state is not circumscribed by the lexical semantics of the verb but it includes the speaker’s
world knowledge and/or contextual information. Consider example (9) once
again, as repeated below.
(9) a. Gunnar
explain.3SG.PAST the.novel.ACC.SG
‘Gunnar recounted the novel’.
b. Sagan
‘The novel was recounted’.
The apparent anomaly here is related to the fact that the message conveyed in
(9b) includes a person who is not himself the recounter of the novel but the third
party who is committed indirectly to the act of recounting. There are two resultative interpretations. The most preferred one is that the recounting of the novel enables the listener (i.e., the third party) to understand the novel better than before.
The minor one is that someone, not the person who recounts the novel, explains
one or two subtopics related to the content of the novel and this helps people understand the novel better than before (see also (35)). It is evident that both interpretations go beyond the lexical meaning of telling a story; rather it has a bearing
on the observer’s (or the speaker’s) conventional knowledge of what telling a
story involves.
Note that some verbs such as mála ‘paint’ require an addition of a manner adverbial such as vel ‘well’ or ágætlega ‘very well’ to accept an inchoative variant,
substantiating that (21c) and (22c) express the resultant-state.
(21) a. Einar
paint.3SG.PAST wall.the.ACC.SG
‘Einar painted the wall’.
b. *Veggurinn málaðist.
c. Veggurinn
wall.the.NOM.SG paint+st.3SG.PAST well
‘The wall was painted well’.
(22) a. Anna
‘Anna turned the pages’.
b. *Blaðsíðurnar flettust.
c. Blaðsíðurnar
‘The pages got excellently turned’.
3.4. Cases with opna ‘open’ and loka ‘close’
In this final subsection, I will examine two verbs, opna and loka, whose behaviour provides evidence for the distinction between the semantic components
‘change-of-state’ and ‘resultant-state’. These verbs have two different inchoative
forms, lexical and morphological. As shown in (23) and (27), the causative constructions accept two different types of direct object. (23) has bíóinu ‘cinema’
and dyrunum ‘door’, while (27) has flöskuna ‘bottle’ and verslunina ‘shop’.
However, as shown by the (un)grammaticality of (24) to (26) and (28) to (30),
when these sentences alternate with inchoatives, the choice between grammaticality and ungrammaticality depends on which direct object is taken as the subject.
(23) Jón
‘Jón closed the cinema/the door’.
(24) Bíóið
‘The cinema closed’.?
(25) *Dyrnar
‘The door closed’.
(26) Dyrnar
‘The door closed’.
(27) Jón
‘John opened the bottle/shop’.
(28) Verslunin
‘The shop opened’.
(29) *Flaskan
‘The bottle opened’.
(30) Flaskan
‘The bottle opened’.
I suggest that inchoatives with the suffix -st express change-of-state, whereas
lexical inchoatives express resultant-state. This distinction can be proven by authentic data extracted from newspaper articles (Morgunblaðið). (31) reports that
the road called Circle Road will be reopened later in the evening. The journalist
who reports (32) is concerned with an episode in which actress Kirsten Dunst invites the customers of a pub to her house to hold an after-hours party. It is clear
that the journalist in (31) is merely interested in the change from the closing to
the opening of the Circle Road (i.e., change-of-state), while the journalist in (32)
directs her/his interest to what happens after the closing of the district pub (i.e.,
resultant-state). While in the former the opening is primary information, in the
latter the closing is secondary. (33) is a constructed example, describing that the
pub came to be closed because of a gas explosion. Here, lokaði, the lexical form
of loka, would not be felicitous. When the construction merely includes the
speaker’s interest in the change from one state to the other, loka must take the -st
form. Similarly, alternative forms, opnar for (31) and lokast for (32), would be
(31) Hringvegurinn
opnast / *opnar
fyrr en
‘Circle Road will not be opened until this evening’. (Morgunblaðið 19 February 2008)
(32) Leikkonan
Kirsten Dunst
actress.the.NOM.SG Kirsten Dunst (person‘s name) be.3SG.PRES not
í Islingtonhverfi í London. Þrjár
neighbor.the.NOM.SG. in Islington.district in London three
have.3PL.PRES put
fram kvörtun
committee.ACC I.
þegar lokar / *lokast býður
hún öllum kráargestum
invite.3SG.PRES she
í in
‘The actress Kirsten Dunst is not the most popular neighbour in Islington in London. Three families
have complained about her before the district committee complained. Kirsten spends the evenings in
the district pub, but when it closes, she invites all the customers of the pub to an after-hours party’.
(Morgunblaðið 1 August 2007)
(33) Kráin
lokaðist / *lokaði
pub.the.NOM.SING close+st.3SG.PAST
‘The pub was closed due to a gas explosion in the house’.?
4. Transitivity Continuum
As briefly mentioned in section 3.1, the fact that the inchoative construction in
Icelandic allows the presence of an agent leads us to the notion of the transitivity
continuum introduced by Hopper and Thompson (1980). This notion basically
states, if applied to our discussion here, that causative and inchoative variants
stand in continuum with a prototypical transitive event at its higher end and a
prototypical intransitive event at its lower end. I demonstrate here two cases in
which the prototypical agent (i.e., controlling, volitional, high in potency, primary) turns out to be less prototypical (i.e., lack of control, non-volitional, low in
potency, secondary) by the use of a prepositional phrase headed by hjá ‘at’. (34)
shows that Jón was committed to making the painting wet although he did not do
it deliberately. (35) shows that Jón did a supplementary job to that of the original
agent to clarify the content of the novel. The difference between (34) and (35) is
that Jón in the former is not acting intentionally, while Jón in the latter is acting
intentionally, though his role is secondary, since its scope is dependent on what
the original agent has done. Looked at this way, the transitivity continuum finds
expression in the causative/inchoative alternation in the sense that the alternation
enables the backgrounding of the prototypical agent.
(34) Málverkið
become.wet.3SG.PAST at
‘Jón accidentally made the painting wet’.
(35) Sagan
‘Jón told supplementary information about the novel’.
Another important fact that supports the role of the transitivity continuum is the
contrast we have observed between two alternative verbs opna ‘open’ and loka
‘close’ (section 3.4). Based on this, we might do ourselves justice by stating that
intransitive verbs with -st are more transitive than those lexical alternatives on
the grounds that, in the former, the causer is seen to have direct control over the
change-of-state of the patient, while, in the latter, change-of-state acts as an intermediary to a subsequent event, to which the speaker/writer pays more attention in discourse, and over which the causer has no control.
5. Conclusions
The discussion in sections 2 to 4 has highlighted three points. First, the causative/
inchoative alternation in Icelandic operates based on the distinction between
change-of-state and resultant-state interpretations of the predicate. This distinction was verified by two verbs opna ‘open’ and loka ‘close’, which have two intransitive forms, one being assigned a change-of-state reading, and the other a resultant-state reading. Second, the meaning of the resultant-state is construed not
only by its semantics but also by pragmatic forces that reflect either the language
user’s conventional implicatures or the writer’s narrative organizations. It was
shown that these forces play an especially important role in explaining why
agentive verbs (such as borða and þvo) and the lexical forms of opna and loka
permit the causative/inchoative alternation. Third, the causative/inchoative alternation has a bearing on the continuum of transitivity. This suggestion was made
with respect to an adjunct phrase headed by hjá ‘at’, which presents two methods
of backgrounding the agent in a denoted event. On closer scrutiny, it becomes
evident that not only semantic but also historical explanations may be in order.
For example, the inclusion of the infix -n- in inchoative forms is no longer productive in Modern Icelandic. I have listed under (5) two colour verbs: sverta
‘blacken’, hvítta ‘whiten’, both of which have an inchoative with the infix. Verbs
that belong to the same class such as blána ‘become blue’ or roðna ‘become red’
are only intransitive. According to Berlin and Kay (1999 [1969]), ‘white’ and
‘black’ are the most basic categories in the evolution of colours across languages.
If we apply this account to our data, one might say that the alternation was only
permitted for basic verbs, and it is likely that inchoatives for other colour verbs
arose in their own right. A similar case is observed with verbs expressing ‘internally caused eventuality’ (Levin and Rappaport Hovav 1995; McKoon and Macfarland 2000; Smith 1978) such as blómgast ‘bloom’. This class of verbs typically contains the suffix -st but lacks causative alternatives. This might also mean
that -st forms arose without recourse to morphological derivation. In summary,
the paper has demonstrated that semantic notions such as ‘spontaneous event’,
‘independent process’ or ‘agent-orientedness’ are still simplistic for the treatment of the Icelandic data, implying that they can be further dissected, as exemplified preliminarily by this study. My general conclusion is that the causative/inchoative alternation in Icelandic is heterogeneous, suggesting that indepth studies need to be carried out to fill in the details. I hope this paper has
paved the way for movement in this direction.
Arce-Arenales, Manuel, Melissa Axelrod & Barbara A. Fox. 1994. Active voice and
middle diathesis: A cross-linguistic perspective. In Barbara Fox and Paul J. Hopper
(eds.), Voice: form and function, 1–21. Amsterdam/Philadelphia: John Benjamins.
Berlin, Brent & Paul Kay. Basic color Terms: Their universality and evolution. 1999
[1969]. Stanford, CA: CSLI Publications.
Davidse, Kristin & Liesbet Heyvaert. 2007. On the middle voice: an interpersonal
analysis of the English middle. Linguistics 45(1). 37–83.
Delancey, Scott. 1984. Notes on agentivity and causation. Studies in Language 8(2). 181–
Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67(3). 547–
Fillmore, Charles. 2003 [1970]. The grammar of hitting and breaking. In Form and meaning in language, vol. I: Papers on semantic roles, 123–39. Stanford: CSLI Publications.
Frawley, William. 1992. Linguistic semantics. Hillsdale, NJ.: Lawrence Erlbaum.
Gorlach, Marina. 2004. Phrasal constructions and resultativeness in English: A signoriented analysis. Amsterdam/Philadelphia: John Benjamins.
Haspelmath, Martin. 1993. More on the typology of inchoative/causative verb alternations. In Comrie, Bernard and Maria Polinsky (eds.), Causatives and transitivity, 87–
120. Amsterdam/Philadelphia: John Benjamins.
Hopper, Paul & Sandra Thompson. 1980.Transitivity in grammar and discourse.
Language 56(2). 251–299.
Kemmer, Suzanne. (1993). The middle voice. Amsterdam/Philadelphia: John Benjamins
Klaiman, M. H. 1991. Grammatical voice. Cambridge: Cambridge University Press.
Levin, Beth & Malka Rappaport Hovav. 1995. Unaccusativity: At the syntax-lexical
semantics interface. Cambridge, MA: MIT.
Levin, Beth & Malka Rappaport Hovav. 2005. Argument realization. Cambridge:
Cambridge University Press.
McKoon, Gail and Talke Macfarland. (2000). Externally and internally caused change of
state verbs. Language 76(4). 833–858.
Nedjalkov, Vladimir P. & Sergej Je. Jaxontov. 1988. The typology of resultative constructions. In Vladimir P. Nedjalkov (ed.), Typology of resultative constructions, 3–62.
Amsterdam/Philadelphia: John Benjamins.
Schlesinger, Izchak M. 1995. Cognitive space and linguistic case. Cambridge: Cambridge
University Press.
Smith, Carlotta 1978. Jespersen’s ‘move and change’ class and causative verbs in English.
In Mohammad Ali Jazayery, Edgar C. Polomé & Werner Winter (eds.), Linguistic and
Literary Studies. In Honor of Archibald A. Hill, 101–109. The Hague: Mouton.
Voorst, Jan G. van. 1995. The semantic structure of causative constructions. Studies in
Language 19(2). 489–523.
Contact information
Toshiko Yamaguchi
Faculty of Language and Linguistics
University of Malaya
50603 Kuala Lumpur
[email protected]
New affiliation:
University of Malaya
Sociolinguistic aspects of bilingualism among
the Moldovan students studying in Romania
Elena Buja
Transilvania University of Brasov, Romania
1. Introduction
The former USSR was a pluralistic society. It was a melting pot in which cultural
and ethno-racial groups contributed to the whole, at the same time trying to retain
their original character. In the long period of Soviet domination, most of the ethnic groups living in the former USSR tried to stick to their culture, and to their
mother tongue, too. Once the USSR crumbled, many of the former Soviet bloc
countries insisted that Russian be replaced with their own ethnic languages as official languages as part of the process of reclaiming their own ethnic languages
as official languages. One such ethnic group is constituted by the people of the
Republic of Moldova covering Bassarabia and Northern Bucovina, a large territory which was disputed by Romania and the Soviet Union. In 1944, at the end
of WWII, Moldova became part of the Soviet Union. Under these circumstances,
Russian became the dominant language in this area, the Moldovan speakers of
the non-dominant ‘Moldovan’ having limited opportunities of using their mother
The historical events that took place in Eastern Europe starting with 1989 have
given this part of the world the chance for a new beginning. Thus, one advantage
of the ‘new era’ was that the people belonging to the former Communist bloc
were now able to travel more freely. As a consequence, many young Moldovans
(the majority of them being born little before or at the beginning of the 1990s)
were encouraged to come and study in the mother/sister-country, Romania. This
had both advantages and disadvantages, as the comments of the subjects of the
present study will reveal.
This presentation is actually the ripple effect of my experience with Moldovan
students studying at the Faculty of Letters of Transilvania University of Brasov.
Since all of them speak two languages, the paper will focus on issues pertaining
to bilingualism, defined by L. Bloomfield (1933) as the situation in which a
person has acquired two languages in a native-like manner, the degree to which
s/he masters both these languages being relative. As bilingualism is a very broad
topic, I will most importantly concentrate on some sociolinguistic issues. Of
major importance would be to determine the status of the language spoken by
the Moldovans: is it to be considered a language per se, should it be treated as a
dialect of Romanian or would it be better to classify it as an archaic variety of
Romanian. To this aim, historical and linguistic evidence will be considered, as
well as the opinions expressed in this respect partly by my students and partly by
members of various forums of the Moldovans.
A second issue I intend to focus on is the attitude the Moldovans have toward the two languages they speak, and implicitly toward the monolingual
users of Romanian and Russian, since very frequently attitudes toward a language often reflect attitudes toward the users of that language. Also investigated will be the attitude of the Romanians toward the Romanian-Russian bilingual people.
Since code-switching is an essential speech strategy of bilinguals and since
‘bilingualism and language switching have existed for hundreds of years with
no sign of decline in the use of the languages involved’ (Romaine, 1995:5), I
think that it is also worth investigating. In the Republic of Moldova, the Romanian-Russian bilinguals have developed a new variety of communication
which incorporates both languages, and this variety functions as ‘the basic vernacular of everyday speech of the community’ (Owens, 2008:392). But it
seems that once these people leave home to study abroad, they stick to this variety. One may wonder: why do they keep using this mixture of Russian and
Moldovan/Romanian in a Romanian environment? What factors trigger
2. Research methodology and ethical issues
The tools I have used in order to find answers to the questions above are first
of all interviews which I conducted with 5 of my Moldovan students within
the premises of our university. All interviews were recorded, transcribed and
analysed. Before the interviews I asked my informants whether they would be
willing to provide me with information concerning their bilingual behaviour.
Since all of them proved eager to participate in a research study, I devised a set
of general questions concerning their acquisition and use of Romanian and
Russian. At some points during the interviews I sensed that my interviewees
tended to withhold certain negative opinions. In order to overcome this danger,
I used another tool, namely Moldovan internet forums where people express
their views freely, without any fear, under the safety offered by their nicknames. I was basically interested in their opinion concerning the use of Russian
or regional terms as well as their opinions concerning their mother tongue and
their identity.
Out of ethical reasons, the students I interviewed will be referred to by using
their first names only. As for the forum participants, I have used their nicknames
with their original spelling. To save space, the Romanian fragments excerpted
from interviews and forums will be rendered in English. They will be quoted in
Romanian only in those cases when this is really relevant (i.e. to highlight regional lexical items or syntactic structures).
3. The analysis
3.1. The premises I have started from are the following:
a) With respect to the status of the official language of the Republic of Moldova,
I hope that the subjects consider it to be Romanian, and not Moldovan or Russian. Not only the historical documents or the political and economic issues
should help the Moldovans determine their identity and, implicitly, the language
they call their mother tongue, but more importantly their own perceptions and
b) Due to the long period of Russian domination and the abuses committed by
the Russian communists, I expect Moldovans to adopt an attitude of Russophobia.
c) I also expect some negative attitudes of the monolingual Romanians towards
the Moldovan-Russian bilinguals due to the imperfect use of the Romanian language and, consequently, potential embarrassment experienced by the latter and
lack of self-confidence when using this language.
3.2. The language of the Republic of Moldova
The strongest evidence in favour of the fact that the language spoken in the Republic of Moldova (RM from now on) is Romanian is the two maps presented
below. The first is an ethnographic map1 of Romania in 1938 which shows that
the territory of present-day RM was part of Romania and was inhabited to a very
large extent by Romanians. The second one is a linguistic map2 from the same
year, containing the regional variants of the verb a casca ‘to yawn’. As the map
reveals, there are no considerable differences between the forms employed in
Romania and those in use in Moldova (in the NE on the map). On the basis of
this evidence, one may conclude that the same language, namely Romanian, was
spoken all over this area.
Atlasul Lingvistic Roman, publicat sub inaltul patronaj al M.S. Regelui Carol II de Muzeul Limbii
Romane din Cluj, sub conducerea lui Sextil Puscariu (Profesor la Universitatea din Cluj, Membru al
Academiei Romane), Partea I (ALR I), Vol. I: partile corpului omenesc si boalele lui. Cluj, 1938,
Muzeul Limbii Romane, Str. Elisabeta 23
Fig. 1. Map of the Romanian territory in 1938.
Fig. 2. The linguistic map of Romania in 1938
But once the territory of Bucovina and Basserabia became part of the USSR, the
fate of the Romanian language on both sides of the Prut differed. Thus, the vocabulary of Romanian was enriched by neological terms of various origins
(mainly French, Italian, and German), whereas Moldovan Romanian3 was ‘secluded’ within the political borders of the Soviet Union, having little (or no) contact with other languages except for Russian. This accounts for the ‘archaic’
shade it presents nowadays. The late G. Pruteanu, former senator of the Romanian Government and a linguistic historian described Moldovan Romanian as
follows: ‘It is not another language, but a language that has been maimed,
jeered at. It is as if the Romanian language had come out, after a number of
years, of an Auschwitz of the language. A man who spent some years in Auschwitz is not like he used to be. He is like a shipwreck, he looks awful. Well, this is
the deplorable situation of Romanian over there, a situation that inspires pity.
(interview with G. Pruteanu, in Cotidianul July 30th, 2003).
Various attempts have been made to establish Moldovan as the national language. Thus, Vasile Stati published the Moldovan-Romanian Dictionary (2003).
This dictionary contains 19,000 words, but as it emerges from the comments
made by the Moldovans on various forums, it is not a scientific works, but rather
a political one. Below are some excerpts from a forum which express the opinions of both Moldovans and Romanians concerning this dictionary:
smcristi, Sep 1, 20034: ‘I haven’t seen for a long time such a stupid thing like the
one that appeared in August 2003’
dragos_avram, Nov 15 20055: ‘Funny little book … You change its title into ‘Dictionary of archaisms and regionalisms (from Moldova) and that’s it’.
dorinteodor, Jul 19 20066: ‘the dictionary of this guy is a dictionary of regionalisms. It could have been extended to entire Romania. I don’t think that
the bloke who worked hard to make it is very guilty. It’s been a political manoeuvre. The recipe is simple: someone makes a dictionary of regionalisms.
Good for him, he has made something useful! But then someone else comes
to call it a Moldovan-Romanian dictionary. The author has no choice since
he knows that this is a political manoeuvre, but he is happy because his book
has been published; otherwise the dictionary wouldn’t have stood many
chances of publication. Obviously, if someone would like to see whether, indeed, there is a ‘Moldovan language’, he should have taken 10 newspapers
from “them” and 10 from us and see whether there are indeed differences in
vocabulary or grammar structures. We know the answer … but I was just
suggesting a test’.
From this point on I shall call the language spoken in present-day Romania, Romanian and, by
analogy with American English I shall refer to the language used in RM as Moldovan Romanian.
Not even Stati himself is convinced of what he asserts, namely that the
Moldovans speak the Moldovan language and not Romanian. This becomes obvious in an interview he offered7, where a number of times Stati pointed out that
there are actually no differences between these two languages. Some excerpts
come to illustrate the point:
a. ‘Both Romanian and Moldovan are Romance languages. Undoubtedly, the
literary form, the most elevated form of Moldovan, the learned version, processed by writers and linguists is identical to the literary form of Romanian’.
b. ‘I do not doubt the common background of the Moldovan national literary language and of the Romanian national language; that would be stupid…’
c. ‘Of course, all the grammar issues, such as the declension, conjugation, derivation, prefixes, suffixes, etc. are the same as in Romanian, nobody doubts that’.
Once you come to the end of the interview taken by a Romanian journalist, you
may wonder whether the discussion took place in Romanian or Moldovan, and
if it occurred in the latter, how many times Stati’s interlocutor made use of this
famous dictionary in order to clearly understand the message.
According to my opinion, another political manoeuvre meant to establish
Moldovan as the national/official language of RM is the launching of
Out of the numerous comments I found in connection with this event, only one
was favourable:
Eugen, July 12th, 20068: ‘It’s normal [that the language of Google] should be
Moldovan as long as the national language is Moldovan’.
Most of the other forum participants have hard feelings toward it:
Medar, July 11th, 20069: ‘That’s exactly what Moldova needed right now: Google
in Moldovan…’
M. July 12th, 200610: ‘Now on the internet as well??? Brainwashing goes on in
the respectable republic maldafia’.
Coleaghin Nina, September 1st, 200611: ‘(…) are the people from GOOGLE
THAT STUPID… can’t they find out that there is no such thing as the “Moldovan
language”, that Moldovan is just a dialect…’
An outraged forum participant, Vladislav Namashco12 suggested to the other
members writing a letter of complaint to Google. Eventually, he was the one who
wrote the following:
Vladislav Namashco, July 12th, 2006: ‘We, the undersigned, consider that the introduction of a new option in selecting the Google interface language, namely
“Moldovan” is useless, provoking and political. We want to inform the management of Google that the “Moldovan language” is an invented language which
was imposed by the communist leadership of the Republic of Moldova with the
aim of denationalizing the natives and of creating a false history. (…) By means
of this letter we demand that the “Moldovan language” should be deleted as an
option from the “interface language” and we wish Google management were
more prudent in approving such decisions’.
Many members of the above-mentioned forum subscribed to the letter: Bunduche, Loserville, o.brega, pingback. One of them, Adrian Hancu, even considered using a different search engine.
The status of the language spoken in RP is also questioned on http:// where the topic under debate was: Moldovan!? Below are some of the opinions I had access to:
Saint, 08 Oct, 2006: ‘Is there such a language [i.e. Moldovan], or do we speak
Izolyda, 17 March, 2008:‘some of us, who are less concerned with speaking correctly, coherently and literary, speak the Moldovan dialect, which, of course, is
part of the ROMANIAN language’.
Homka, 12 April, 2008:‘Unfortunately, in the constitution it exists. But what
kind of a language is it … to what extent does it differ from Romanian?’
Nokomment, 18 April, 2008:‘If our leaders have stipulated the Moldovan language in the constitution, this does not mean that we have to speak Moldovan
(…) our mother tongue has been, is and always will be ROMANIAN’.
Two messages concerning the language used by the Moldovans impressed me.
One was written by marcel13 whose tone is similar to that of a former famous Romanian (Moldovan) leader (Stephen the Great) in urging his co-nationals to stick
to Romanian: ‘Dear Bassarabians, Moldovans between the Dnister and the Prut,
bear in mind forever that even if nowadays we are a special state and it is true
that we can be like that many years from now, until fate decides on our destiny,
we have to admit that the language we all speak is the ROMANIAN language’.
The other vote in favour of Romanian as the national language of the Moldovans
comes from a Moldovan woman, Inessa Baban14, who had studied at a Romanian
university. She concludes the article she sent for publication to a monthly Romanian magazine (i.e. Magazin istoric) as follows: ‘If anyone might think that
we will give up on the melody of the Romanian language in favour of an invented
or non-existent Moldovan language, then s/he proves to be Futile (….) or is
dominated by Utopias (…) because we shall not abandon the Celebration of Romanian and we refuse to believe that we are not Romanians’.
To conclude this part, the historical proof and the opinions of the people advocate in favour of Romanian being the official language of the Republic of Moldova. It is only for political reasons that “Moldovan” is mentioned as the national
language in the constitution. As Holmes (2008:99) states, ‘in multilingual countries, the government often declares a particular language to be the national language for political reasons. The declaration may be a step in the process of asserting the nationhood of a newly independent or established nation (…)’.
3.3. Attitudes toward bilingualism
F. Grosjean (1985:117) points out that ‘language is not just an instrument of
communication. It is also a symbol of social group identity, an emblem of group
membership and solidarity’. Both as an instrument of communication and as a
symbol of group identity, language is accompanied by attitudes and values held
by its users and also by persons who do not know the language. What is important to realize is that very frequently the attitudes toward a language often reflect
attitudes towards the users of that language.
• Attitude of the Romanian monolinguals toward the Moldovan-Russian bilinguals
I shall start the study by presenting first the attitudes expressed by the monolingual Romanians toward the Moldovans and their language. The messages posted
on chat forums revealed mixed feelings of the Romanians. Some of them feel
sympathetic. This is the case of a forum visitor called inginerum15 (Sep 2 2003)
who posted the following message: ‘It is the same language … only that the
Moldovans across the Prut speak the Moldovan dialect we have here, only with
another accent and with many archaisms…. Believe me … I live in Iassy and I
have faculty mates from Chisinau, Balti a.s.o.… and they are Romanians indoctrinated by communism, and their language remained for most of them at the
level it was in the 50s and the 60s because of the lack of introduction of neologisms as was our case, and because of many other reasons’.
Some other Romanian monolinguals have a neutral attitude: people simply
express their puzzlement at not understanding what they are told, without making any comments. One such person is Iulia16 who recounts her experience with
a Moldovan fellow student. Most of her story below has been translated into
English, but some sentences have been preserved in Romanian (and translated in
brackets) in order to show the cause of puzzlement: ‘Once she was waiting for
her boyfriend and seeing that he was not coming, she said: “offf, cre’ca s-o
PRAPADIT Gheorghe” [oh, I think that Gheorghe HAS DIED]. She meant to say
that he had got lost. Another time she asked me: "cand ai venit mai era
PITEOARCA aia afara ?" [when you came, was that PITEOARCA still outside?]. Realizing that I did not understand, she confused me even more:
“PITEOARCA, adica JIGULIE” [PITEOARCA, that is a JIGULIE]. What do
you think she was talking about? A car, an old Russian-made Lada’.
There are, nevertheless people who seem to enjoy making fun of the oldfashioned, archaic façade of the Romanian spoken in Moldova. Thus, there is a
site17 which posts a lot of jokes related to the archaic speech of the Moldovans.
The same ridiculing attitude is expressed by Billy18 who makes fun of some TV
news he accidentally came across: ‘I couldn’t help laughing at them. Longunheard archaisms made me laugh my eyes out! (…) The Moldovans call a 24hour shop an EVERLASTING SHOP’. Fortunately, such messages are rare.
• Attitude of the Moldovan Romanian-Russian bilinguals toward Romanian/
In an article entitled ‘On the Bad Habits of the Moldovans’ and posted by Puiu19
on May 14th 2007, the writer presents the feelings experienced by a famous
Moldovan poet, Grigore Vieru, when he first visited Romania. ‘Recently, in a
Basserabian daily, a young poet from Chisinau was sharing his impressions on
how he feels in Romania. The comparison he made was based on a general truth,
namely that we, the foreigners, the immigrants, are regarded in the host-country
as some kind of panda bears. Indeed, no matter how much we strive to look Romanian, in Romania we are seen as something exotic, picturesque, even
unnatural : as some kind of Russian-Moldovan hybrids’.
This first paragraph of the article clearly indicates a state of unease which
many Moldovans coming to our country for various reasons experience. The
same idea emerges from the article written by Inessa Baban20entitled Suntem
români ?i punctum ! [We are Romanian, full stop!]. It begins as follows: ‘Every
day spent in Romania is a new challenge. You challenge the puzzlement you read
in the eyes of those who listen to you and try to decode the message you are trying
hard to send. But despite your efforts to make yourself understood, you fail, so
that the eyes of your interlocutor send back the question: ‘What does s/he mean
to say? Oh God, another cryptic message!’ Could it have been the accent? If it
had been only for the accent…’
Similar embarrassment was reported by three of my interviewees. Veronica, a
first-year student, had to share a room with a Romanian girl. On hearing that Veronica was from the Republic of Moldova the Romanian student became quite
distant until she learned that the only thing that made them different was the accent, not the language or nationality. Vadim, another student of mine, reported
felling uncomfortable when realizing that he spoke with a different accent and
that he used archaic/regional words. He recounted a talk with his university
mates in which he used the expression amu ia which is both archaic and regional,
its present-day form being ‘acum’ [now]. Due to a certain sexual connotation attached to the string of sounds in Romanian, his male fellows started laughing,
which embarrassed Vadim terribly. Finally, Viorica mentioned the puzzlement
caused among her Romanian room-mates by her use of such regionalisms as
prostire (Standard Romanian cearsaf [bed sheet]) and iorgan (Standard Romanian plapuma [quilt])
The unease experienced by many of the Moldovan students/people in Romania due to their pronunciation could be the reason why they do their best to improve their accent and they freely admit that, as the following quotes reveal:
Vadim: ‘I have tried to get rid of my accent, but, still, I am very proud of the language I speak and it feels good when I use it with my compatriots’.
Veronica: ‘I am trying hard to speak without a Moldovan accent, but I am capable anytime to switch back to the Moldovan dialect’.
Carolina: ‘You don’t have to try too hard to get rid of your accent because once
you are in a certain community you get rid of it anyway’.
This phenomenon is known under the name of communication accommodation
(Giles, 1991, Wardaugh, 2002): speakers may be inclined to alter their speech
and speaking style in order to gain social approval from the listener, increase
communicative efficiency, and maintain a positive social identity with their audience. I have encountered similar positive attitudes among the forum participants, some of them being very keen on improving not only their accent but also
their grammar and lexicon. Thus, on the forum suggestively entitled lataifas21
[chatting], a topic suggested in 2003 was Reper grammatical [grammar issue].
Someone launched the following question: Do you think you need to improve
your Romanian? Out of a total of 45 participants, 17 (37%) answered ‘Yes,
definitely’, 23 (51%) answered ‘Yes; although I know it well, repetitio est mater
Two messages posted on this forum confirm the general trend among the
Moldovan youth. Thus, Apollo stated: ‘I am sure that most of us need to brush
up the grammar of the Romanian language. As long as you live in a society
where Romanian is spoken with mistakes and as long as you don’t speak much
Romanian, you have no chance of improving your knowledge of this language’.
The other message was written by Amelie, who is of the opinion that‘(….) we
still have a long way to catch up with them (i.e. the Romanians)… for me the
speech of those in Moldova is not at all an example … I would be glad if somebody corrected me when I say/write something incorrectly, so that the mistake
will not happen again…’
Both messages clearly indicate that the Moldovan youth are aware of the differences between the Romanian language spoken in Romania and the one spoken
in Moldova and that they are not really happy with the latter. Both forum par21
ticipants mention some ‘mistakes’ the Moldovans make, but reading their messages I didn’t encounter any problem in understanding them. This means that
what they call mistakes are simply some minor differences that cannot be ‘language barriers’ defined by Moser (1972) (quoted in Ionescu-Ruxandoiu & Chitoran, 1975:257) as ‘obstacles that hinder communication in the mother tongue’.
• Attitude of the Moldovans towards Russian/Russians
In a community where different language groups coexist, language attitudes play
an important role in the lives of users of these languages. Haugen (1956) (quoted
in Grosjean, 1982:117-118) is of the opinion that ‘whenever languages are in
contact, one is likely to find certain prevalent attitudes of favour or disfavour towards languages involved. These can have profound effects on the psychology
of the individuals and on their use of the languages’.
What is important to realize is that attitudes toward a language are often confounded with attitudes toward the users of that language. One of the premises of
the study was that due to the conflicts between the Russians and the Moldovans,
the latter might not have a positive attitude either toward the language or its
speakers. To a very small extent this was confirmed by some messages posted on
forums. Thus, mihai015, Wed May 2, 2001 6:23 am22 expresses his feelings as
follows: ‘I don’t have anything against the Russians. I would gladly leave them
on Saturn’s orbit and I would mind my own business’. Other russophobic feelings can be encountered on
(see the messages posted by dirtyangel, Apr. 10, 2007 and by Chirila Carol, Apr.
23, 2007).
Rz, a participant in a forum of the Moldovans23 recounts an event that impressed him/her: ‘Subjectively, it seems that Russian is more popular because
when speaking to Russians, the Moldovans use Russian, while the Russians addressing Moldovans don’t speak Romanian. Here is an example that demonstrates that the ‘minority’ dictates … especially because in many institutions the
civil servants are Russian. Once I saw how a woman (Russian) was almost shouting at an elderly woman who did not speak Russian and who could not understand a mistake that appeared in her bill…’
Similar events were recalled by some of my Moldovan students. Viorica, who
comes from Cahul, said that 50% of the population in her home-town are of Russian origin. In shops, customers have to speak Russian, otherwise they are not
served. Carolina, who is a native of Balti, confirmed this.
Despite the reluctance at being imposed the Russian language in certain circumstances, all my interviewees and most of the forum participants seem to like
this language and use it frequently. Vadim, one of my interviewees, told me that
in some places in Moldova the Moldovan people speak Russian because they like
it better than Romanian: both the elderly people and the youngsters prefer Rus22
sian to Romanian. He is of the opinion that culture affected their attitude toward
the language. In their everyday speech, when Moldovans want to emphasize
something, they say: Dup cum spune rusul [as the Russian says]. Veronica, another respondent in the interview, told me that she has no negative feelings either
toward the Russians or their language. Once she heard some tourists speaking
Russian in Brasov and this made her homesick. Carolina, another interviewed
student, said that despite the fact that she hated being imposed the Russian language, she likes it and uses it frequently.
Similar positive attitudes toward both Russian and Romanian were encountered among the messages posted on internet forums.24 Iuliu states ‘I am for writing in Romanian and then in Russian’. Creatza highlights the advantages of being bilingual, irrespective of the languages someone speaks: ‘indeed, we have to
pay more attention to the language spoken in our country (…) do you want to say
that you suffer because you can speak two languages? .. I will tell you that you
have an extraordinary advantage; the Romanians who try to speak Russian do
not succeed so well; I am pleased that I can speak Russian without any accent
and Romanian impeccably’.
To conclude this part, there are a number of reasons why Russian is favoured
in certain circumstances. First, there is tradition: elderly people have studied and
used more Russian than Romanian. Then, there is the need for local integration:
the Moldovans need to speak Russian with those who can’t speak Romanian.
Thirdly, as Creatza put it, it is broadening to have more than one language. Then,
some Moldovans use Russian as an exclusionary tool: they have a secret language that not every Romanian understands. Another factor could be cultural
inertia: people maintain their language simply because it is there. Last but not
least, some people favour Russian because this language is linked to a rich and
varied cultural heritage.
3. 4. Code-switching and code-mixing
The positive attitude of the Moldovans toward the Russian language is also
demonstrated by the frequent code-switching25 and code-mixing26. Vadim recounted some sentences uttered by his mother while he lived at home, in Tighina.
The words originally spoken in Russian, but spelled with Latin letters, are
Ia SHVABRA si spala pe jos (‘Take the MOP and clean the floor’)/Vezi ca
mincarea e in HOLOGHELNIK (‘The food is in the FRIDGE’).
Code-switching is defined loosely by Swann et al. (2000) as the use of two or more languages during the same conversation. According to Owens (2008:414) code-switching is “the shifting from one
language to another. The behaviour is not random, nor does it reflect an underlying language deficit”.
Gumperz (1982:52) defines it as “the juxtaposition within the same speech exchange of passages of
speech belonging to two different grammatical systems or subsystems”.
Code-mixing is defined as the combination of a root/stem of one language with an affix of another
One reason Vadim gave me for using Russian words when speaking Romanian
is that they better express concepts or feelings. ‘Russian is more expressive due
to polysemy’, he said. He considers Russian to be a prestigious language. Thus,
Vadim seems to be in line with Grosjean, who stated that ‘the prestige language
is considered to be more beautiful, more able to express abstract thoughts, and
the other language is felt to be ungrammatical, concrete and coarse’ (1982:122).
Another reason he mentioned was the fact that for most of the tools he used in
the yard he had learned the Russian terms.
Veronica produced a mixed sentence when talking to her room-mate: Da-mi
NOJikul [Give/pass me the.KNIFE!], where noj is the Russian word for knife,
while the ending –ul is the definite article for the masculine/neutral nouns in Romanian. She also confessed that whenever she had to name objects in the household, she inserted the Russian words for them in Romanian sentences.
When asked why Moldovans insert Russian words in Romanian sentences,
Ana answered that it is nicer, whereas Carolina said that ‘it is not nicer, but sometimes the Russian words are handier’. This is in line with Swann et al. (2000)
who are of the opinion that speakers may shift between languages in order to
maintain the flow of conversation; a particular word or phrase from another languages may come to one’s mind faster, and may be more convenient to express
at that moment. Thus, for most of my interviewees code-switching depends on
linguistic factors. It is not that certain vocabulary or grammatical features and
concepts either differ from Russian to Romanian or are non-existent in Romanian. My subjects seem to shift codes in order to convey a message in the most
natural manner possible. They seem to have stronger Russian vocabulary for certain domains and are thus more capable of conversing in the language normally
associated with that particular domain.
Forum participants also shift from one language to the other, but they seem to
do it out of social reasons: they change codes in order to establish a certain social
identity, to express solidarity. Since most of the messages on contain code-switches, one may assume that the
members of this forum don’t want either monolingual Romanians or Russians to
participate in their discussions. Some members post messages entirely in Russian, some others use this language for salutation formulas, as demonstrated below27, whereas a couple of them show a tendency of mixing Russian words in
Romanian sentences28 (again the Russian words will be indicated by capitalised
1. Nishka (01.12.2004): ‘unii chiar parca-s UKURINIIE’ [some (people) indeed
2. KazeolNevozmojnii (01.13.2004): ‘sini aisi ii UKURENII?’[which of you are
For the translation of the Russian words/sentences, I was helped by my student, Vadim, to whom
I am very grateful.
3. Valerika (05.17.2004): ‘salutIK la toti cei noi … bine ati venit shi enjoy’ [hello
to all the new members…. welcome and enjoy]
4. Melania (01.23.2005) ‘PRIVETIK PATZANI! Iaka mai avem un user nou –
Stefan’ [HELLO GUYS! Here we have a new user – Stefan]
5. Melania (01.24.2005) ‘f Meliuga … ce materialist ai ajuns tu … am auzit
c-i placea culoarea roz … la nebunie PRITOM … as putea s-i cumpr niste
sandali rozove … e bun?’ [Meliuga, dear … how materialistic you have become
… I have heard that you like pink …. very much INDEED … I could buy you
some pink sandals … is it OK?]
According to the classification of code-switches given by Poplack (1980)
(quoted in Romaine, 1995:122-123), the examples under 1, 2 (the first two
clauses) and 5 could be considered tag-switches, i.e. ‘he insertion of a tag in one
language into an utterance which is otherwise entirely in the other language’. The
effect of such shifts is that they may add emphasis to a point, and make a certain
statement even stronger and personal. The last part of examples 2 and example
4 are cases of inter-sentential switches, i.e. ‘switches at a clause or sentence
boundary, where each clause or sentence is in one language or another’ (Romaine, 1995:123).
4. Conclusions
What I hoped to find was that the Moldovans do not regard themselves linguistically or culturally inferior to either the Russians or the Romanians and that they
perceive the language they speak, irrespective of whether it is called Romanian
or Moldovan, accepted and respected on both sides of their geographical borders.
But in many cases my hopes were contradicted by my findings. Thus, some of
the Romanian-Russian bilinguals feel stigmatized due to their speech/language
(confirming prediction c, in 2.1). But this stigmatization seems to have, nevertheless, a positive effect in that it reinforces the loyalty and solidarity of the
group toward their language and the people. Grosjean (1982:126) states that
‘even though use of a stigmatized language may be associated with a less prestigious group - at least in the eyes of the majority group – it may reinforce the
group’s positive values and symbolize solidarity for them.’
Despite the negative attitudes some of the Romanian and Russian monolinguals have toward the Romanian-Russian bilinguals, most of the latter seem to
have positive feelings toward Romanian and Russian. On the one hand this is reflected in their efforts to brush up their Romanian, to alter their speech and
speaking style in order to gain social approval from their Romanian listeners, to
increase communicative efficiency, and maintain a positive social identity with
their audience. On the other hand, the positive attitude towards Russian (contrary
to my prediction, see 2.1. b) is shown in the delight the bilinguals take in speaking Russian or in switching from Romanian to Russian in their speech/writing.
One reason why the people I investigated do not seem to have a negative attitude
toward Russian is the fact that they learned this language while still children and
that most of them were born shortly before or after the events that led to the independence of the Republic of Moldova.
Diversity of languages and cultures should be regarded as a good and beautiful
thing in itself. Each language has its own way of seeing the world and is the
product of its own history. All languages have their individual identity and value,
and all are equally appropriate as modes of expression for the people who use
them. To devalue a language or to presume that one language is better, ultimately
devalues individuals and cultures. In order to live peacefully in a multilingual
Europe, there is a need to generate a greater interest in and curiosity about languages. And above all, there is a need to enhance linguistic tolerance both within
and between nations.
Atlasul Lingvistic Roman, publicat sub inaltul patronaj al M.S. Regelui Carol II de Muzeul
Limbii Romane din Cluj, sub conducerea lui Sextil Puscariu (Profesor la Universitatea
din Cluj, Membru al Academiei Romane), Partea I (ALR I), Vol. I: partile corpului
omenesc si boalele lui. Cluj, 1938, Muzeul Limbii Romane, Str. Elisabeta 23
Bloomfield, Leonard. 1933. Language. New York: Holt, Rhinehart, and Winston
Grosjean, Francois. 1982. Life with Two Languages: An Introduction to Bilingualism.
Cambridge, Mass. and London, England: Harvard University Press
Gumperz, John J. 1982. Discourse Strategies. Cambridge: Cambridge University Press
Holmes, Janet. 2008. An Introduction to Sociolinguistics (3rd edition). London, New
York: Pearson, Longman
Ionescu-Ruxandoiu, Liliana. & Dumitru Chitoran. 1975. Sociolongvistica. Bucuresti:
Editura Didactica si Pedagogica
Owens, Robert E. jr. 2008. Language Development. An Introduction. 7th edition. Boston,
Paris, Munich: Pearson Educational
Romaine, Suzanne. 1995. Bilingualism.Oxford, UK: Blackwell
Swann, Joan, Rajend Mesthrie, Ana Deumert, and William Leap. 2000. Introducing
Sociolinguistics. Philadelphia: John Benjamin Publishing Co
Wardhaugh, Ronald. 2002. An Introduction to Sociolinguistics. 4th edition. Oxford:
Internet sources
Elena Buja
Department of Foreign Languages and Literatures
Transilvania University
29 Eroilor Blvd
500036 Brasov
[email protected], [email protected]
Narrative patterns in
monolingual and bilingual life-history
Angela Falk
Uppsala University
1 Introduction
Much is known about narrative patterning in conversations among L1 speakers
(Norrick 1997, 2000). For example, narratives anchored in the present time of the
interaction are catapulted into the past, thanks in part to the particular ordering
of clauses that achieve temporal organization (Labov 1972, 2008).
In this paper, I will look at the patterning that contributes to situating narratives in a moment of time in the past and that creates the possibility for speakers
to mark and accentuate differences between them and their interlocutors (cf. Irvine 2001). A landmark study pertaining to intergenerational conversations is
Coupland, Coupland and Giles (1991), whose work examines how elderly persons formulate their age in relation to other interlocutors. One of their main
viewpoints is that age identity is in part “an intrinsically rhetorical projection”
(66); such a perspective opens the research arena for further exploration. While
the authors briefly mention bilingualism in connection to the communities where
elderly persons live (pp. 6; 65), the data they present in their study appear to have
been extracted from monolingual conversations.
Following Coupland, Coupland and Giles (1991), my focus is also placed on
analyzing age identity work taking place in intergenerational conversations, but
the present study attempts some additional exploratory work examining both L1
and L2 data extracts. I look at some ways elder speakers, most in their 90s and
all living in Swedish-American communities, verbally construct their age identity when speaking with much younger interlocutors. Bilingual, intergenerational
conversations are not uncommon in ethnic communities, even in societies in
which one language appears to dominate, such as English in, for example, the
Midwestern region of the United States. Nonetheless, the research literature has
yet to provide systematic sociolinguistic investigation and analysis of data extracts involving intergenerational experience and bilingual communities. One of
my pilot studies (Falk forthcoming), based on a case study consisting of multiple
recordings with an elder bilingual speaker ‘Greta’ and two younger interlocutors
who were incipient bilinguals at the time of the conversations, reveals that Greta
makes extensive use of comparison and contrast discourse patterning when presenting her experiences to the younger listeners, while she simultaneously frequently glosses Swedish lexical items for benefit of the younger L2 speakers of
Swedish. The present study now draws upon a slightly wider range of data in L1
(English) and L2 (English and Swedish) settings in effort to examine narrative
patterning in intergenerational conversations. My aim in this ongoing investigation is to identify frequent patterns in life-history narratives and thereby gain insights into identity formation that emerges via discourse structure. I will investigate ways that speakers use age categorization processes and temporal framing
processes. These two processes were among those considered by Coupland,
Coupland and Giles (1991) in their exploration of “elderly identity marking” (ch.
3); in addition to these processes, I will draw attention to two other patterns of
discourse topicalization used by the elder speakers to introduce narratives.
2 Data extracts and patterns under investigation
The data were transcribed from life-history interviews recorded in two communities in the United States: Lindsborg, a small town in Kansas, and Minneapolis,
a large city in Minnesota. Most of these particular recordings were made in the
mid-1990s with persons who had strong ties with the Swedish-American communities in the respective areas.
2.1 Data
From a corpus of conversational data, consisting of approximately 70 speakers,
that I compiled (see Karstadt 2003, published under my previous surname), for
the purpose of this particular study I selected eight respondents: four speakers
from the L1 Lindsborg data set (2 male and 2 female), and four speakers, Swedish-born, in the Minneapolis data set (2 male and 2 female). The speakers in this
study were all born in the first decade of the 1900s. When interviewed, they had
or were soon to celebrate their 90th birthdays. The speakers in Lindsborg are
Swedish Americans, whose grandparents and/or parents immigrated to Kansas
and settled in the Smoky Valley region, where the town is situated. While these
respondents understand and speak Swedish to varying degrees, English is clearly
their dominant language. The Minneapolis speakers emigrated from their home
regions in Sweden (Dalarna, Västmanland, and Värmland) in the 1920s and
settled in Minneapolis, where, at the time of the recordings, they were long-time
members of the Swedish-American cultural community, devoting a sizeable portion of their active retirement to participating in Swedish-American clubs and organizations. See Karstadt (2003) for further details relating to the speakers’
Swedish-American networks.
I will examine structures in both monolingual conversations (L1 English) as
well as in bilingual conversations involving speakers of Swedish and English.
The term bilingual will be used to refer to conversations in which one of the interlocutors is using his/her second language. The transcribed data in the study are
collected from life-history interviews with individuals, approximately half in
which the interviewers and respondents share the same L1 and half in which one
of the interlocutors is speaking his or her L2. In this study, the data from Minneapolis are considered bilingual.
The data sets of the two communities are comparable in size. The transcribed
portion of each individual interview ranges from 19 to 25 minutes, though the
full length of each complete, extended recording may run anywhere from one to
two hours. The current number of transcribed words extracted for the purposes
of this particular study is somewhat modest at slightly more than 10, 000 words
from each community, for a total of approximately 20,500 words for the eight
2.2 Some characteristics of life-history conversations
A common dimension in the recorded conversations, whether in the L1 or L2 set,
is the nature of life-history conversations, in which a premium is placed, not surprisingly in view of the genre, on the past. The nature of the life-history conversation can be briefly described as follows: An underlying and guiding principle
in this type of conversational setting is a tacit acknowledgement that the respondent’s life experiences are sufficiently different from the interviewer’s and from
the persons who will listen to the recording in the future. At least part of the motivation behind the interviewer’s questions hinges on this expectation. Stories
told by Swedish Americans have been my focus because they are one major way
elderly Swedish immigrants and elderly Swedish Americans convey their experiences for younger listeners. Stories thus move the listener to different points in
time. In my data, the scope and span of this movement is potentially great as the
interviewers were nearly sixty years younger than the respondents. Thus, compared to the younger interlocutors, the elder speakers have a sizeable reservoir of
memories from which they can draw. In the monolingual conversations, respondents typically call attention to their early connections to the local community. For example, they can refer to their own eye-witness accounts of hard times
during the American Depression, beginning in the year 1929, and they can even
convey memories shared by their parents and grandparents of the pioneer era of
the region, beginning in the last decades of the 1800s. In the L2 conversations
examined here, respondents draw upon memories of their emigration from Sweden, generally in the 1920s, of their early years in the Swedish-American immigrant community, and, as we will see in some data extracts coming up, they draw
upon some linguistic resources from their L1 to introduce some concepts they
link to the past.
2.3 Patterns under investigation
Thanks to the nature of life-history conversations, even fairly limited amounts of
transcribed data for each individual participant (approximately 2,000 words or
more) contain numerous narrative stretches of talk. Analysis of some extended
narrative passages from my material appear in Falk (forthcoming) and Karstadt
(2002, 2003). My purpose in this paper, however, is not to focus on the structure
of entire narratives, but to identify and examine some perceptible patterns in the
beginnings of narratives as they are launched by the elder speakers. After presenting some evidence and brief discussion of the emerging patterns in my data,
I will suggest some directions for further research.
Three patterns appear to be particularly relevant as contributing to the construction of elder identity in the recordings. As mentioned in section 1, these patterns
are age categorization processes, temporal framing processes, and processes
achieving discourse topicalization. Discourse topicalization appears to manifest itself in two ways in the recordings: The use of a mental verb, such as remember,
that draws attention to the information that is subsequently reported as being remembered, or the use of a noun phrase that is used cataphorically as it announces
and launches an upcoming narrative that will offer an elucidation of the particular
noun phrase. This strategy will be exemplified in extract (17), in which the noun
phrase “an odd way” is subsequently explained in the narrative that ensues.
3 Patterns
3.1 Age categorization processes
Among the numerous ways by which speakers can mark themselves as being elderly, Coupland, Coupland and Giles (1991) identify several communication approaches speakers may use to categorize their age. These include telling one’s
chronological age (p. 59); using a label that is marked for age, such as referring
to oneself “an old lady” or “an old man” (60–61); or drawing attention to one’s
failing health and strength, something the authors refer to as “painful self-disclosure” (61–62). Clear instances of age categorization processes in my data are
very rare, a finding quite surprising in light of the fact that the age difference between the interlocutors differed by up to six decades. I located one very brief instance in the data, extract (1) below, uttered with mock-serious intonation by
Anna (a pseudonym, as the case with all names given for my respondents). She
mentions her ailing back before calling herself “an old lady.” Accompanying this
extract and others is the speaker’s pseudonym, the date of the recording, the data
set (L1 or L2), and the community in which the interview was recorded.
I’ve got a bad back. I think I’ve been out working in the yard too much for an old
Anna, 1993.08.25, L1, Lindsborg
In contrast to extract (1), which clearly fulfills the criteria for age categorization
as described, the age categorization process in extract (2) unfolds successively
through the interaction of the speakers. Extract (2) is the sole instance of age
categorization I have located to date in the L2 data set. In extract (2), we see that
Anders obliquely categorizes my age first (“So, so you, you weren’t even born
yet,” line 1) before explicitly establishing his own existence as having been for
“a long time” (line 3). Elsewhere in this extract Anders substantiates his perception that his childhood was dominated by chores, which left little or no time for
play. Excerpt (2) is a conversational extract that shows how age categorization
clearly intersects with temporal framing processes. Anders does not state that he
is “old” per se, but he does provide evidence to support his perception that has
life has been different from mine and my generation. For further discourse
analysis of Anders’ narratives, see Karstadt (2002).
(2) you weren’t even born yet
1 Anders:
so, so you, you weren’t even born yet
2 Angela:
I wasn’t around yet, not yet
3 Anders:
no, so I’ve been around a long time
4 Angela:
yeah, you have
5 Anders:
and I’ve gone through a lot
6 Angela:
uh-huh, do you have a lot of, um, clear memories of Leksand
when you
grew up?
8 Anders:
9 Angela:
do you think about that time?
10 Anders:
oh ja, I uh uh I remember because I had to work!
11 Angela:
12 Anders:
you know when we, when we went to school, well we uh when
we came
home it was to either cut wood or uh in da winter time shovel
da snow and
uh denn we had ta carry in da water and carry out da slop pail
15 Angela:
16 Anders:
and uh and dat was a steady job, you know, you’d never had
dat much
time to play.
Anders, 1996.06.06, L2, Minneapolis
3.2 Temporal framing processes
Apart from the two extracts provided above, age categorization appears to be
very rare in the transcripts examined so far. A clearly productive and prevalent
pattern used by the speakers, however, involves the use of temporal framing
processes. As I examined the transcripts of the life-history conversations, I
looked for temporal framing processes that involved one or more of the follow163
ing four criteria: (1) mentioning a specific year or years in the past. Here is it important to clarify that here I am concerned with years that are mentioned by the
respondents in a narrative—not when a date is mentioned in direct response to a
question posed by the interviewer; (2) making an age or time reference to oneself
concerning earlier periods in the speaker’s life; (3) making an age or time reference concerning earlier periods in the lives of the speaker’s parents or grandparents; and (4) giving age or time reference in relation to the interlocutor. Criteria (1) through (4) are narrative techniques that directly or indirectly are able
to position a speaker as “elderly” (cf. Coupland, Coupland and Giles 1991:62ff.).
According to Coupland, Coupland and Giles (1991), temporal framing processes
place “topics within a clearly time-past frame” (63), they may heighten the impression that the speaker is self-associating with the past (64), and finally, they
may give the impression that the speaker wishes to amplify awareness of “historical, cultural or social change” (65).
Temporal framing processes are readily apparent in the speaker transcripts,
appearing just as frequently in L1 as in L2 conversations. Six brief extracts appear below, three from the L1 data and three from the L2 data.
I’ve lived here, uh, moved here in 1922.
Ray, 1993.08.25, L1, Lindsborg
(4) old as Lindsborg
and they came here early,
they came, Dad’s people came here to Salemsborg in 1868
Angela: uh-huh?
that’s a hundred and twenty-five years ago.
Angela: yeah.
old as Lindsborg
Ray, 1993.08.25, L1, Lindsborg
A temporal framing process which is achieved by referring to another generation
can be found in (5). Agnes employs many such temporal framing processes, primarily by framing a time period around her grandparents’ lives as pioneers and
around her own childhood memories.
and my grandmother talked about walking through that prairie grass.
Agnes, 1994.07.15, L1, Lindsborg
The intensive use of temporal framing processes is also readily observable in the
L2 data. Without having been prompted to mention specific years, Greta carefully relates a timeline pertaining to her family’s immigration to Minnesota. Her
father migrated first, and then he worked at numerous jobs to earn sufficient in164
come to buy cross-Atlantic tickets for his two eldest children, Greta and Margit,
before sending money back to Sweden to pay for the tickets for other family
so in 1924, in March 1924 he immigrated to this country […] and d- by 1926 he
had managed to scrape up enough what he could buy a ticket for Margit and I and
he wanted his girls to come to America.
Greta, 1991.10.19, L2, Minneapolis
In extract (7) we see evidence that Greta anchors an account of a swimming hole
to her young childhood.
we learned da hard way when we were only about three-four years-old.
Greta, 1991.10.19, L2, Minneapolis
well, well, I tell you, when I first, first left, I left Sweden on de ‘leventh of February 1927 Johanna, 1993.05.11, L2, Minneapolis
3.3 Discourse topicalization
Discourse topicalization is a term I use as an umbrella concept to refer to structures that announce that a stretch of narrative talk is immediately forthcoming in
the interaction. One of the frequent topicalization strategies involves the use of
a phrase containing a mental verb, such as remember, never will forget, tell, and
know. Data extracts (9) through (16) are particularly salient because the speaker
in each case teams one of the mental verbs, such as remember or forget (Biber
2006: 59–61), with a lexical item that is then given special focus. The mental
verb aids in topicalizing a forthcoming lexical item that will either be elaborated
upon—as in the case of the L1 data, see extracts (9) through (13) —or be
glossed—as in the case of Swedish lexical items used in the L2 data, see especially extracts (14) and (15)—in the stretch of narrative discourse that follows.
Below, mental verbs are italicized and the items I propose as being topicalized
are underlined.
I remember uh, oh, Good Friday, (be)fore noon…
Agnes, 1994.07.15, L1, Lindsborg
as long back as I can remember my mother always took me
to hear the Messiah every year and that was my desire to sing
Agnes, 1994.07.15, L1, Lindsborg
another thing I remember was, uh, Santa Lucia
Agnes, 1994.07.15, L1, Lindsborg
I never will forget the a cappella choir, how they …
Agnes, 1994.07.15, L1, Lindsborg
wait and do you know? It was hard to get water on that farm, in that area where
we lived.
Ed, 1994.07.14, L1, Lindsborg
and I especially remember one Christmas, we thought dat my grandmother’s
house, dat downstairs was enormously large. Dey had one sal, dey called it, and
dat was da nice rum.
Greta, 1991.10.19, L2, Minneapolis
and I never forget dat, so I tought I would go call on that you know old man, you
know my- målarmästare…
Fredrik, 1988.04.21, L2, Minneapolis
and I’ll never forget dis uh someting dat he call marbleizing, he uh,
Fredrik, 1988.04.21, L2, Minneapolis
In (14), Greta launches her narrative to a great extent with the mental verb remember, which organizes the discourse with its focus on the topical items Christmas, downstairs and the Swedish word sal. In (15), Fredrik achieves topicalization in his forthcoming narrative with the forget verb together with the Swedish
lexical item målarmästare, which refers to a master painter who teaches and
supervises apprentice painters. He creates a similar effect in the sequence identified as (16) when he teams the mental verb forget with the specialist term
The use of mental verbs in narratives comprises one strategy relevant to topicalization. Discourse topicalization is also achieved lexically via a cataphoric
process with a noun phrase. Ray sets in motion the beginning of the narrative in
(17) by observing that he and his wife originally “met in an odd way.” The noun
phrase points ahead to a discourse unit consisting of multiple clauses that will reveal what the “odd way” was. For more on the sequencing of new and old information in noun phrases, see Johnstone (2008:113–116).
(17) we met in an odd way
and uh, we met in an odd way out at Falun.
most of us we didn’t have cars.
we couldn’t go to Lindsborg or to Salina or to anything,
or to the show or anything.
so, Sunday nights you went to church. At church, every Sunday
so that’s when you went
Angela: uh-huh
I was there one Sunday night and my wife was there,
my wife-to-be was there and she winked at a guy,
and I thought it was me
11 Angela: {chuckles} uh-huh?
12 Ray:
and we started dating
13 Angela: uh-huh?
14 Ray
and never quit.
15 Angela: uh-huh
16 Ray:
Ray, 1993.08.25, L1, Lindsborg
4 Summary, discussion, and directions for further study
Analysis of my data sets shows that the elderly speakers readily employ two of
the three patterns for structuring their contributions in the life-history conversations. The elder speakers frequently use temporal framing processes, and a selection of these from the L1 and L2 data sets were provided. In addition, the elder
speakers frequently mobilize discourse topicalization.
With regard to topicalization, the mental verbs, particularly two of them, remember and forget, appear to play central roles. I find evidence that both L1 and
L2 speakers of English launch narratives by employing one of them. Additionally, in extract (13), we saw that Ed calls the interlocutor’s attention to a new, upcoming strand of narrative by asking “Wait and do you know?” My future research will continue to track the discourse patterning associated with these particular verbs. The data set of the Swedish immigrants shows that their strategic
use of certain Swedish concepts, sal in (14) and målarmästare in (15), may be
intended in part to announce that an extended discourse unit is coming up.
Discourse topicalization is also achieved lexically via a cataphoric process
achieved with a noun phrase. Ray, an L1 speaker of English, established the beginning of a narrative in (17) by mentioning “an odd way.” The phrase focuses
the listener’s attention to the impending discourse unit of multiple clauses that
revealed what that “odd way” was. To date, I have not located the similar use of
cataphora with noun phrases in the L2 data. More research is needed to establish
whether the patterning in (17) is more typical for L1 rather than L2 data. Current167
ly, however, my analysis suggests that L1 and L2 speakers use patterns that are
more alike than different, at least when it comes to the structure of their discourse. (See Karstadt 2003 for some syntactic differences between the data sets,
however, involving relative clause patterns and pragmatic particles.)
As mentioned above, a pattern in my data that appears to be quite rare is age
categorization. While Coupland, Coupland and Giles (1991) report that age
categorization processes are highly expected and indeed prevalent in talk between older and younger interlocutors—in fact, it is the first pattern they
present in their overview of “Dimensions of elderly identity marking” (59)—
the appearance of such processes is comparatively rare in my data, a surprising
finding, given the vast age differences between the interlocutors. Two observations are relevant here: First, participants in the conversations arranged and
studied by the Couplands and Giles were new acquaintances participating on a
voluntary basis. Making age categorization explicit in such an arranged meeting may have felt immediately relevant as the older speakers (aged 70 to 87)
were members of two so-called “day centres in Cardiff, Wales” (57). While it
is not clear from the methodological description in Coupland, Coupland and
Giles whether the younger volunteer participants (age 30 to 40) in the study
met the older women in the day centers, readers can logically draw this assumption. By contrast, the participants in my study were already well acquainted, having known each other for years; furthermore, most of the recordings
took place in the homes of the respondents. The second observation that is
relevant to make is that social network integration may have the consequence
that temporal framing processes, frequent in my data, achieve a more useful
purpose than age categorization among friends and acquaintances. Divulging
one’s age to a long-time friend does not typically add anything new to a conversation. There are certainly obvious contrasts in methodology between my
study and the one conducted by the Couplands and Giles. The contrasts between the two studies are nevertheless valuable for the capacity to locate variable patterning in intergenerational communication.
5 Selected references
Biber, Douglas. 2006. University language. A corpus-based study of spoken and written
registers. (Studies in Corpus Linguistics 23). Amsterdam & Philadelphia: Benjamins.
Coupland, Nikolas, Justine Coupland & Howard Giles. 1991. Language, society and the
elderly. (Language in Society 18). Oxford: Blackwell.
Falk, Angela. forthcoming. Narratives at the crossroads of language and generations.
Studia Neophilologica.
Irvine, Judith T. 2001. ‘Style’ as distinctiveness: The culture and ideology of linguistic
differentiation. In Penelope Eckert & John R. Rickford (eds.), Style and sociolinguistic
variation, 21–43. Cambridge: Cambridge University Press.
Johnstone, Barbara. 2008. Discourse analysis. 2nd edn. Malden, MA: Blackwell.
Karstadt, Angela. 2002. Talk over time. Longitudinal analysis of life-history recordings.
In Ulla Melander Marttala, Carin Östman & Merja Kytö (eds.) Samtal i livet och i lit-
teratur. Conversation in life and in literature. Papers from the ASLA symposium, 125–
134. Uppsala: Svenska föreningen för tillämpad språkvetenskap.
Karstadt, Angela. 2003. Tracking Swedish-American English. (Studia multiethnica
Upsaliensia 16). Uppsala: Acta Universitatis Upsaliensis.
Labov, William. 1972. Language in the inner city. Studies in the Black English vernacular. Philadelphia: University of Pennsylvania Press.
Labov, William. 2008. Oral narratives of personal experience. (13 December 2008.)
Norrick, Neal R. 1997. Twice-told tales: Collaborative narration of familiar stories.
Language in Society 26. 199–220.
Norrick, Neal R. 2000. Conversational narrative. Storytelling in everyday talk. Amsterdam & Philadelphia: Benjamins.
Angela Falk
Department of English
Uppsala University
Box 527
SE-751 20 Uppsala
[email protected]
Castilian or Catalan? Linguistic survival strategies of Japanese residents in Catalonia, Spain
Makiko Fukuda
University of Barcelona
INTER-ÀSIA, Autonomous University of Barcelona
0. Introduction
The process of globalization of world economies has brought with it greater mobility of peoples. In Catalonia, the increasing number of foreign workers from
non-European Union countries has raised many questions and been the subject
of several studies.
Japanese migrants, most of whom are the product of Japan’s economic expansion world-wide, constitute one of several foreign communities that have settled
in Catalonia. Little is known, however, about this collective, due mainly to the
fact that it is small in number and highly mobile. Often seen as a closed community of executives of transnational companies who live temporarily in the host
society and speak only Japanese, this image of a “closed” community does not
necessarily reflect reality: there may be a closed community of Japanese expatriates with its own ecosystem, but there are also Japanese nationals who are well
integrated into the host society. One of the main factors contributing to the difference between these two communities is their command of the local language.
This study explores the linguistic situation of Japanese nationals resident in
Catalonia by analyzing their degree of knowledge of two of the official languages of the host society – Catalan and Castilian – and aims to identify the role
of each language in their linguistic survival strategies within the plurilingual
reality of Catalonia.
1. Sociolinguistic situation of Catalonia
1.1 Co-officiality of Catalan and Castilian
Catalonia, whose capital is Barcelona, is historically a distinct polity within
Spain. The Spanish Constitution of 1978 recognizes and guarantees the right to
self-government of Catalonia alongside the Basque Country, Galicia and Andalusia. Since the Statute of Autonomy was passed in 1979, Catalan is one of three
official languages in Catalonia together with Castilian, which is known more
commonly as Spanish and is the official language throughout Spain, and Aranese, a dialect of the Occitan language.
1.2 Immigrants in Catalonia and the languages of the host society
During the sixties and seventies, Catalonia received migration flows from inside
Spain, but these have recently been superseded by a massive wave of “newcomers” (Aguilera, 2001; Jaime, 2002) or “new immigrants” (Boix&Vila, 2006;
Rovira et al.,2004) from several countries with different cultures. The flow of
these non-native migrants, far from being stable, has accelerated, diversified and
extended throughout the territory (Secretaria d’Immigració, 2005). The advent
of such large numbers of newcomers has been the subject of intense debate and
controversy, in particular over the question of language use.
The acquisition of the host society’s language is a significant indicator of social integration. Language, whilst it serves as a tool of communication, constitutes one of the elements which indicate membership of a community. Lack of
knowledge of the host society’s language can therefore exclude an individual
from interaction with that society. By the same token, knowledge of the language
of the host country represents an immigrant’s will to and/or possibility of integration (Llompart, 2007). In Catalonia, however, where the Catalan language coexists with other languages, the situation is somewhat more complicated. Catalan, as occurs in many parts of the world, is a key symbol of ethnic identity and
has served for the past century as a prime symbolic resource of Catalan nationalism (Woolard, 1989:1).
Currently, Castilian serves as the host language or language of communication
amongst newcomers, while Catalan continues to be the language of integration,
learned only by those who wish to settle in Catalonia. There are a number of reasons why this is so. Firstly, these two languages are not found in exactly the same
situation: whilst almost all Catalan speakers speak Castilian fluently, the same
cannot be said about Castilian speakers of Catalan, and Castilian is omnipresent
in the public domain in Catalonia (Rovira et al., 2004). Secondly, the authorities
in Catalonia cannot institute an autonomous immigration policy, for it is the
Spanish government which has competence in matters of immigration (Boix i
Vila., 2006; Ros, 2006; Rovira et al., 2004). The Spanish government favours the
use of Castilian, so that the use of Catalan remains fairly limited in this particular
sphere. Thirdly, traditional language etiquette (Woolard, 1989:69–73) amongst
native Catalan speakers has it that they speak Castilian with foreigners and
strangers, so that there is no need for the newcomers to speak Catalan – something that newcomers actually complain about (Llompart, 2007). Finally, many
newcomers live in surroundings in which Castilian is the predominant language
(Boix&Vila, 2006; Ros, 2006).
1.3 Preference for widely-spoken languages
Castilian maintains an important presence within Catalonia in that it is by defect
the language spoken by newcomers, whilst Catalan is perceived as a secondary
language learned only by those who wish to become fully integrated into Catalan
society. The prevalence of Castilian causes newcomers to perceive it as a primary language. In fact, research into newcomers in Catalonia and the language they
speak coincide in highlighting the fact that Castilian is perceived as indispensable in order to live in Catalonia, whilst learning Catalan is voluntary (Aguilera,
2001; Beltrán&Sáiz, 2001; Boix&Vila, 2006; Llompart, 2007; Ros, 2006 etc.).
Once basic communicative competence has been achieved in Castilian, most
newcomers no longer find any need to learn Catalan, since the limited use of
Catalan, as opposed to the predominant use of Castilian, discourages them from
doing so. In particular, those who arrive in Catalonia expecting to return to their
home country believe that once they leave Catalonia, Catalan will be of little use
to them, if any. Those who decide to learn Catalan do so voluntarily. Most recognize that Catalan is necessary if one wishes to become fully integrated into
Catalan society, even though it is perceived to be a secondary language. The custom of Catalan speakers of addressing foreign interlocutors in Castilian is also
an important factor in perpetuating this perception. In such a complicated situation, Castilian comes between new migrants and Catalan, and makes it harder for
them to access the latter (Boix&Vila, 2006).
It may thus be concluded that newcomers to Catalonia tend to learn Castilian
first, out of necessity. Learning Catalan first as a survival strategy is unusual. In
general, a preference for the use of widely-spoken languages as opposed to minority languages exists, although some authors point out that newcomers’ cultural and social background must be taken into account, for it is not irrelevant to
their learning Catalan (Cf. Torres, 2006; Llompart, 2007, Miralles&Iturrapse,
2005; Rovira et al., 2004). Those who come from a plurilingual society, such as
Africans, tend to learn Catalan, whilst those who have Castilian as their first language are not so motivated to learn another language. Those who choose to learn
Castilian are often from a society where the hierarchical relationship between
dominant language and dominated language is clear, or where a strong monolingual tradition exists. Newcomers who reside temporarily in Catalonia do not
even learn Castilian.
The situation described above raises some important questions: given that
mastering the host society’s language is key to full integration, what happens
when two languages co-exist , each with a different status?; how do expatriates
from Japan, where the myth of monolingualism is well established, react in the
face of the multilingual reality of Catalonia? Who learns or does not learn Castilian and/or Catalan and for what reasons? Is there any correlation between
knowledge of Catalan and that of Castilian?
2. Profile of the Japanese community in Catalonia
2.1 Distribution
During the past few decades, Japan has experienced globalization in several
sectors, in particular the economy. Japanese economic expansion has produced
a great mobility of people, which has led to the dispersion of a large number of
Japanese nationals all over the world. Japanese communities overseas, made
up of workers with temporary residence, are a product of this phenomen. The
number of Japanese resident in Spain continues to increase steadily. According to the Foreign Ministry of Japan, in 2005, this number exceeded 6,000. In
Spain, Catalonia is one of the two autonomous communities with most Japanese residents, together with Madrid. The 2007 census, published by the General Consulate of Japan in Barcelona, showed a total of 2,043 Japanese nationals resident in the Autonomous Community of Catalonia, 80% of whom
were concentrated in Barcelona. The majority were executives of Japanese
companies located in Catalonia, 60% of whom lived in the city of Barcelona.
Compared with communities from other Asian countries such as China, the
Japanese community is not so numerous, but this is due to the fact that their
stay is usually short-term, and turn-over is much more frequent than in the case
of other Asians.
2.2. Some sociocultural features
The daily life of Japanese expatriates is connected to the homeland through the
educational system, through imported consumer goods and services, and through
the media. It is common for expatriate communities to create such “environmental
bubbles”, but it applies more specifically to the Japanese than to other migrant
groups. (Björklund, 2007:12)
The profile of Japanese immigrants in Catalonia is characterized mainly by (1)
temporality of migration, (2) exclusivity within the host society, and (3) a high
concentration of professionals at management level in highly qualified companies. This transitional residence does not motivate members to integrate into the
host society. Instead, they attempt to maintain a Japanese environment, associating exclusively or mainly with their compatriots. Why does this occur?
Lack of knowledge of local languages is often the main obstacle, making integration into the host society difficult. This is one of the main reasons why
Japanese expatriates group together (Shibano, 1983; White, 2003; Lam, 2005).
Given that their interpersonal relations tend to be limited to their compatriots,
there is therefore no need to learn local languages. However, this raises the question of the chicken and the egg: are Japanese not eager to learn local languages
because they associate only with their compatriots, or do they group together because they have no knowledge of local languages?
The value given to local languages is a further factor: if expatriates consider
that the language in question is of little value on their return to Japan, they save
themselves the effort of learning it. Traditional monolingualism in Japanese so173
ciety does not value language variety, which makes it difficult to motivate people
to learn a language which is not official at State level.
Although this profile does not characterize all Japanese residents in Catalonia
– there are some small subgroups of people with a long-stay migratory project
such as mixed families, Japanese families with the right of permanent residence
etc. – the fact is that “transient” members account for most of the members of the
community, which makes subgroups less visible.
3. The degree of knowledge of Catalan and Castilian of
the Japanese community in Catalonia:
3.1. The study
The main objective of this study was to analyze the degree of knowledge of
Catalan and Castilian of Japanese residents in Catalonia. The study was carried
out in cooperation with several individuals and organizations concerned with
Japanese residents or representing Japan in Catalonia (Japanese School of Barcelona, Association of Japanese Companies in Barcelona, etc.)
3.2. Data Collection
Data was obtained through the use of questionnaires and follow-up interviews.
Of a total of 200 questionnaires, 121 were returned completed. Follow-up interviews were conducted with some of subjects (n=34). They were drawn from the
list of subjects who had willingly agreed to be interviewed. Subjects themselves
declared the degree of their knowledge of each language based on self-evaluation; other direct methods such as testing were considered too costly Four basic
skills (listening, speaking, reading, and writing) in Castilian and Catalan were
evaluated, on a scale from zero to five (Level 0: Not at all; Level 1: Some words;
Level 2: Some simple phrases; Level 3: A part of a text or conversation; Level
4: The main part of a text or conversation; Level 5: With total fluency)
3.3 Degree of knowledge of Catalan and Castilian
Analysis of the results obtained from the questionnaires and the follow-up interviews showed subjects’ degree of knowledge of Catalan and Castilian (Fig. 1).
Those who lived in a predominantly Japanese environment made up the largest
subgroup among our subjects, although almost all have some knowledge of Castilian, in different degrees. Only a limited number of subjects had some knowledge of Catalan. It would thus appear that, as far as our subjects are concerned,
Castilian is a language which is learned first as a basic instrument, whilst Catalan
is a secondary language, which is learned optionally.
With regard to the four skills, a marked difference may be observed between
the two languages. Writing is the least developed skill in both languages, and
there is a tendency to show a higher degree of receptive competences. Oral com174
munication skills in Castilian are slightly more developed compared with those
of writing. It must be remembered, however, that the results obtained provide a
general overview, while extreme cases (maximum and minimum) can bias results.
Figure 1 Degree of knowledge of Catalan and Castilian. Mean.
3.4 Who learns or does not learn Castilian? Who learns or does not
learn Catalan?
Data obtained from the questionnaires show that most subjects (83,5%) have
some knowledge of Castilian, in different degrees, whilst a minority (35,5%)
have some knowledge of Catalan. In our interview, the following motivations
were indicated most frequently for having learned Castilian: (i) Castilian is the
language of the country in which they currently live; (ii) Castilian is also spoken
outside Spain; (iii) Castilian would be of use when they go back to Japan;. On
the other hand, the main reasons why subjects had not learned Catalan were: (i)
they can meet basic daily needs with only Castilian; (ii) Catalan is not used outside Catalonia; and (iii) local people address them in Castilian. From their opinions, it may be concluded that subjects’ attitude to Castilian and Catalan is different. As a general rule, those who had a good command of Catalan were from
mixed families consisting of a couple (a Catalan native speaker and a Japanese
native speaker), who considered that Catalan is essential for full integration. It
was deduced that this view is one of the most influential factors in our subjects’
language learning.
3.5 Relation between the degree of knowledge of Catalan and Castilian
From the opinions extracted from the interviews with some subjects, it was hypothesized that some correlation existed between our subjects’ degree of knowledge of Catalan and Castilian.
To verify this hypothesis, data obtained from the questionnaires was again
analyzed, this time using correspondence analysis. This method of statistical
analysis is used to describe relationships of dependence and independence – that
is to say, difference or similarity – between two variables in the multidimensional space based on the cross-tabulation data. This tabulation is formed by crossreferencing at least two variables. For the purposes of our study, the two variables established were: the degree of knowledge of Castilian in four basic skills
(listening, speaking, reading and writing) and the degree of knowledge of Catalan in the same four basic skills (listening, speaking, reading and writing).
Figures 2,3,4,5 show the proximity and the distance between the results obtained for the two variables
First, listening skill is examined. As the Figure 2 shows, the minimum degree
of knowledge of Catalan (Level 1) is found close to the intermediate degree of
Castilian knowledge (Level 3). This suggests that it is difficult for those who understand no more than some simple phrases in Castilian to understand some
words in Catalan. Intermediate degrees of knowledge of Catalan (Levels 2 and
3) are located quite close to Level 4 of Castilian, which suggests that a fairly high
degree of knowledge of Castilian does not necessarily ensure an equivalent degree of knowledge of Catalan. Those who declare higher degrees of knowledge
of Catalan (Levels 4 and 5) are those subjects who declare the same or higher degrees of knowledge of Castilian. That is to say, those who have a good command
of Catalan have a good command of Castilian too. As far as our subjects are concerned, it may be said that it is nearly impossible to understand Catalan without
some knowledge of Castilian.
As for reading skills, a similar tendency is observed, but with a slight difference. The maximum degree of knowledge of Castilian (Level 5) is found close
to Level 4 of reading skills in Catalan. That is to say, many of the informants who
can read a text in Castilian without any difficulty can also read most of a text in
As for productive skills – speaking and writing –, the level of knowledge of
Castilian of a certain number of individuals with a minimum knowledge of
Catalan is higher when compared with the receptive competences, especially in
writing. The vast majority of subjects who declare their oral expression skills in
Castilian to be Level 4 are concentrated at lower levels of Catalan (from 0 to 2).
As Figure 4 shows, the intermediate level of knowledge of Castilian is found beside the lowest level of knowledge of Catalan. It may be interpreted that many of
those who can speak quite fluently in Castilian, can say no more than some
simple phrases in Catalan, or at most, they can maintain conversation partly in
Catalan, and many of the informants who declare that they can speak partly in
Castilian cannot speak in Catalan at all. As in other skills, individuals who declare the maximum degree of Catalan are only those who declare the equivalent
degree in Castilian.
Finally, the degree of skill in writing deviates somewhat from the tendency observed earlier: the degree of a minimum knowledge of Catalan (Level 1) is found
almost adjacent to Level 4 of Castilian. This means that, though subjects can
Figure 2. Relation between the listening skill in Castilian and Catalan.1
Figure.3 Relation between reading skill in Castilian and Catalan.
write the greater part of a text in Castilian, subjects can write no more than some
words in Catalan at most. This tendency is not observed in other receptive competences, for all subjects who declare a higher level of Castilian (Levels 4 and 5)
declare a certain degree of knowledge of Catalan (Levels 2 or 3). Thus, a strong
correlation between the degree of knowledge of Castilian and Catalan (p-value
< 0,0001) is observed in all skills.
Green dots represent the degrees of understanding Catalan, while red dots represent those of
Figure 4. Relation between speaking skill in Castilian and Catalan.
Figure 5. Relation between writing skill in Castilian and Catalan.
4. Conclusions
The results of the study presented in this paper represent only a part of the language reality of the Japanese community in Catalonia. Nevertheless, it provides
us with an overall picture of language skills. Our findings may be summed up as
(1) The Japanese community views Castilian and Catalan differently: Castilian
is a language indispensable for daily life, readily available to the public in general, and widely used, whilst Catalan is a secondary language, learned optionally
because its use is strictly limited, and is not a language open to foreigners.
(2) This perception influences subjects’ language learning, that they continue to
consider Castilian to be the host language and Catalan a secondary one.
(3) A high degree of knowledge of Castilian does not always guarantee an equally high degree of knowledge of Catalan, although all individuals who declare a
high degree of Catalan knowledge also declare a high degree in Castilian. This
shows that knowledge of Catalan is always accompanied by knowledge of Castilian, as if it were a condition sine qua non. Thus, at least in the case of our subjects, Castilian is learned almost necessarily, whilst only some individuals come
to learn Catalan and always at the second stage of integration.
Aguilera, Montserrat. 2001. Reflexions a l’entorn de la llengua, la immigració i la planificació lingüística a Catalunya. Llengua i Ús, 22. 11–14.
Beltrán, Joaquín., Sáiz, Amélia. 2001. Els xinesos a Catalunya. Família, educació, i
integració. Barcelona: Fundació Jaume Bofill.
Boix, Emili., Francesc X., Vila & Núria, Alturo. (eds.) 2006. Integrar, des de la fragilitat?
Societats plurilingües davant els reptes de les immigracions multilingües: Suïssa,
Luxemburg, Brusel·les, Quebec i Catalunya. (Col·lecció XARXA CRUSCAT; 4). Barcelona: Institut d’Estudis Catalans.
Boix, Emili. & Francesc X. Vila. 2006. Les noves immigracions, la integració i la llengua.
In Boix, Emili., Francesc X. Vila & Núria, Alturo.(eds.), 7–12.
Björklund, Krister. 2007. Migration in the interest of nation. Population movements to
and from Japan since the Meiji Era. Web Report no.25, Siirtolaisuusinstituutti.
Fukuda, Makiko. 2007. ¿Globalización sin inglés? Experiencia de la colonia japonesa en
Cataluña. Interculturalidad y lenguaje II: Identidad cultural y pluralidad lingüística.,
Granada: Granada Lingvistica, 139–148.
Iturraspe, Amaia.; Joan Miralles. 2005. L’impacte lingüístic del turisme residencial al Pla
de Mallorca. Llengua i Ús, 34. 75–83.
Llompart, Júlia.2007. L’acolliment i la integració lingüística: Estudi del funcionament. (4 May, 2008)
Jaime i Femenia, Laura. 2002. Entitats catalanes: porta d’entrada a la societat d’acollida
i a l’autonomia personal. Llengua i Ús 24. 28–39.
Lam, Wing S. 2005. Adventures in Hong Kong. Migration Decisions, Adaptations, and
Re-adaptations of Japanese Expatriate Wives. Master’s thesis presented at Chinese
University of Hong Kong.
Marí, Isidro. 2005. De l’acolliment lingüístic a la interculturalitat, Llengua i Ús, 34. 45–
Pereña, Mònica. 2007. Com acollim lingüísticament les persones que arriben a Catalunya:
els plans pilots, Llengua i Ús, 38. 18–26.
Ros, Adela. 2006. Catalunya davant el nou repte migratorti. In Boix, Emili., Francesc X.
Vila & Núria, Alturo. (eds.) 51–56.
Rovira, Marta. et al. 2004. El català i la immigració. Anàlisi de l’oferta de cursos del
català als immigrants adults extracomunitaris. Informe general per a la Fundació
Jaume I.
SECRETARIA PER A LA IMMIGRACIÓ. 2005. El Pla de ciutadania i immigració, un
pla per a tothom. Llengua i Ús, 34. 4–10.
Shibano, Masayama. 1983. Kaigai nihonjin komyuniti to sono kyouiku mondai.
Kobayashi, Tetsuya. (ed.) Ibunkani sodatsu kodomotachi. Tokyo: Yuuhikakusensho.
Torres, Joaquim. 2006. El coneixement de català dels nats a l’estranger residents a
Catalunya l’any 1996. Boix, Emili., Francesc X. Vila & Núria Alturo.(eds.) 57–70.
Vila i Moreno, Francesc X. 2005. Els coneixements linguistics. In Joaquim Torres
(coord.) Estadística sobre els usos lingüístics a Catalunya 2003. Llengua i societat a
Catalunya en els inicis del segle XXI. (Publicacions de l’Institut de Sociolingüística
Catalana Sèrie Estudis, 8) Barcelona: Secretaria de Política Lingüística, Departament
de la Presidència, Generalitat de Catalunya.17–54.
White, Paul. 2003. London ni okeru nihonjin. Komyuniti keisei katei. In Iwasaki, Nobuhiko. et al. (eds.). Kaigaini okeru nihonjin, nihon no nakano gaikokujin. Gurôbaruna
iminryuudou to esunosukêpu. Tokyo: Shôwadô.131–151.
Makiko Fukuda
East Asian Studies
Faculty of Translation and Interpreting
Autonomous University of Barcelona
Campus de la UAB, Edifici K1013
08193 Bellaterra (Cerdanyola del Vallès) Barcelona, Spain
[email protected]
The Uppsala Learner English Corpus:
A new corpus of Swedish high school students’
Christine Johansson and Christer Geisler
Uppsala University
1. Introduction
The present paper introduces a new corpus of learner English, the Uppsala
Learner English Corpus (ULEC). In the paper, we describe both the design of
the corpus and various ways of working with it. ULEC is a collection of essays
by Swedish junior and senior high school students aged between 14 and 19. The
material is collected as part of student teachers’ degree projects. Student teachers
use the material to form ideas on how to teach grammar and contextualise grammar in the classroom. They also use the corpus to investigate the knowledge of
English grammar among junior and senior high school students. Frequent grammar errors are categorized and analyzed. In addition, ULEC is a resource for
scholars interested in different aspects of learner English. In section 4, we report
on some preliminary results from a study on syntactic complexity in the writing
of senior high school students.
2. The design of the ULEC corpus
The ULEC data is collected by student teachers. Junior and senior high school
students write their essays in a simple web interface (Johansson 2006; Geisler
and Johansson 2007). The corpus is constantly growing, thanks to the contributions of new student teachers each semester. Currently, ULEC contains approximately 136,000 words (19,000 words from junior high school and 117,000
words from senior high school, see Tables 1–2).
Each essay in the corpus is coded for various extra-linguistic categories, such
as date of composition, register, year in school, level of English course in senior
high school, type of high school program, gender of the writer, and age of the
writer. A sample of the coding system is given in (1).
(1) <D 20081023><G DESC><Y 2><K B><P S><S M><A 17>
If I would win one million kronor, people around me would probably hear
it. I would save some, for sure, but I would probably spend most of it. Maybe
spend some on a new car, since I'm about to turn 18. … (male student, aged
17, academic program)
The codes in the COCOA format in (1) refer to D = date of composition, G =
genre/register, Y = school year, K = level of English course in senior high school,
P = type of program in senior high school, S = gender, and A = age.
Tables 1 through 4 give sample word counts, number of essays in parenthesis,
and average number of words per essay. Table 1 shows the sample sizes among
junior high school students (school years 7 through 9) across gender.
Table 1. Distribution of essays in the junior high school data.
For example, essays written by girls in year 8 comprise a total of 1523 words, in
8 essays, with an average length of 190.4 words. Table 2 gives the same type of
information for senior high school students. On the whole, boys write shorter essays than girls do.
Table 2. Distribution of essays in the senior high school data.
29417 (116)
12662 (49)
2812 (13)
44891 (178)
48813 (229)
18580 (76)
5505 (12)
72898 (332)
A comparison of the sample sizes across school year in Table 2 shows that the
majority of the data was collected among senior high school students in year 1
with 78230 words and 345 essays (116 girls and 229 boys). The senior high
school students either study in academic programs or vocational programs, as
shown in Table 3.
Table 3. Distribution of essays across academic and vocational programs in the
senior high school data.
34085 (131)
10806 (47)
44891 (178)
51603 (213)
21295 (119)
72898 (332)
Table 4 shows the distribution of the essays across level of English course in
senior high school. The English C course represents an advanced level course in
the third year of senior high school within the academic programs. Students in
vocational programs study English A and B only.
Table 4. Distribution of essays across level of English course in the senior high
school data.
29611 (118)
14054 (55)
1226 (5)
44891 (178)
51649 (244)
17917 (74)
3332 (14)
72898 (332)
3. Students’ use of ULEC
There are two ways of working with the texts in the corpus. First, the student
teachers write about and reflect on “How to teach grammar”. During the time
they spend in the schools as part of their third-term studies in English (school
placement, in Swedish “Verksamhetsförlagd utbildning, VFU”), they often see
grammar teaching carried out in a way that does not appeal to them. Grammar is
not contextualised and not made interesting (see Larsen-Freeman 2001a: 255–
257; Larsen-Freeman 2001b: 11–24; Tornberg 2002: 115–118; Thornbury 1999;
Ellis 1997: 20–32; Ellis 2006). As a consequence of the grammar teaching that
the student teachers have seen in school, they present new ideas in their papers,
designing practical exercises for new methods in grammar teaching. An example
of such a paper is Rohm (2006), which describes an attempt to include grammar
teaching in the writing process. A very detailed timetable with deadlines for
drafts and final versions of the texts is given to the students along with other assignments. If there are still grammar errors or mistakes in the final versions, the
teacher bases his/her grammar teaching on those particular grammar points.
Secondly, student teachers are very interested in investigating junior and
senior high school students’ knowledge of English grammar (the papers may
consist of a comparison between different programs in senior high school or progression from English A to English B). Two examples of such papers include
Grönblom (2007) and Tillenius and Johansson (2007). In the first paper, the
academic and vocational programs were compared. The vocational senior high
school students had, as expected, more problems with grammar, particularly
with verb forms, such as spelling, tense and agreement. The second paper is a
study of progression in the knowledge of grammar between English A and B
among senior high school students. There is progression in that the English B
students get better at grammar points such as verb forms and nouns (the genitive
and plural endings), but seem to become more careless when dealing with spell183
ing in general and punctuation. The study also suggests that the high school students generally write in a way that reflects the students’ spoken language: simple
sentence structure and frequent use of discourse markers (see also section 4).
3.1 Error categories
When the student teachers investigate the proficiency of junior and senior high
school students, frequent grammar errors are categorized, explained and evaluated with error analysis as a model (as described by e.g. Ellis & Barkhuizen
2005: 57; Saville-Troike 2006: 37–38). Often the students give a linguistic taxonomy of the errors (listed below), but there are also papers which discuss errors
in more ‘technical terms’ according to the surface structure taxonomy proposed
by Dulay et al. (1982: 150), see also Ellis & Barkhuizen (2005: 60–61). For example, the use of the wrong preposition would then be classified as misinformation. Errors or mistakes are often found in the following areas:
Subject-verb agreement
Article usage (generic reference)
Capital letters (spelling)
The ’s-genitive
Verb form
Verb form (spelling, tense, and agreement) is the category where most errors occur. The errors in (2)–(3) are not as serious as the ones in (4)–(6) where both
agreement errors and the wrong combination of verb forms are used.
We flyed to the warm Grand Canaria on vacation.
We all get tanned and we have a lot of fun.
… so we can felt the warmth.
The house have a beach, we was there a hole week, it wasen't menny
I doesn't remember much, it's will be fun, my friends dident whant to go
Gröndahl (2007) sets up grammar categories to be tested in a diagnostic test for
English A students in an academic program. This is called ‘controlled production’. Then the student teacher investigates whether the same error categories occur when the high school students write freely, that is, when they write their essays which are intended to be a part of ULEC. Teachers think that certain errors
will occur but sometimes students make other types of errors instead or do not
make them at all. Most of the errors in the diagnostic test and in the ULEC essays
were made in the categories of subject-verb agreement and prepositions. In the
diagnostic test, however, relative pronouns formed a large category of errors.
When the students wrote freely they probably avoided using relative clauses.
The student teachers often include in their papers a discussion of how grammar should be taught and which grammar points they should concentrate on in
their teaching. A comparison of grammar errors between the USE corpus (Uppsala Students’ English Corpus, which is a corpus of university students’ English,
see Westergren-Axelsson 2000) and ULEC has been made in a few papers.
Agreement errors, article usage, the writing of capital letters and the ’s-genitive
cause problems at university level too and could indicate areas in which the
grammar teaching in junior and senior high school should have its focus.
The student teachers also want to know about attitudes to grammar and grammar teaching in their papers. When senior high school students were asked how
they liked grammar, surprisingly many of those in the academic programs, more
than 50%, answered that they liked grammar. They also wanted more of the traditional grammar teaching in which the teacher explains the problem, followed
by exercises and a test as check-up (see Isaksson 2006).
3.2 Different analyses of grammatical errors
As stated earlier, some of the different steps of error analysis are often the model
for the discussion of grammar errors in the student teachers’ papers. Ellis &
Barkhuizen (2005: 53) state that today error analysis (EA) has given way to other
types of analysis, for example, obligatory occasion analysis, but there is still an
interest in EA in applied linguistics. Obligatory occasion analysis includes calculating accuracy scores for how well a certain grammatical morpheme is
learned (Ellis & Barkhuizen 2005: 73–81). If, for example, the third person singular –s is the morpheme to be investigated, all the obligatory occasions for the
use of this morpheme have to be identified and the total number of occasions has
to be calculated. This type of analysis has been carried out in papers dealing with
subject-verb agreement errors. A few papers have dealt only with how high
school students master subject-verb agreement. This is a topic we tend to stress
in our teaching, whether in schools or at university (cf. also Källkvist & Peterson
2006). Obligatory occasion analysis has also been done in studies investigating
whether Swedish students overuse the progressive or not.
Functional analyses, such as form-function analysis (Ellis & Barkhuizen
2005: 111–127), are naturally a part of both error analysis and obligatory occasion analysis in the papers. For example, a discussion of the function of the progressive (when denoting ‘ongoing activity’) must be included. The opposite
analysis, function-form, has been carried out on essay topics that deal with future
events, such as “My life in ten years” (the use of expressions for future time), hypothetical situations, for example “If had 10 million kronor” (conditional constructions) and argumentation “Do you believe in ghosts” (the use of nominal
4. Syntactic complexity in Learner English
The Uppsala Learner English Corpus is also a resource for scholars like ourselves interested in different aspects of learner English. We have started studying
the structure of clauses and sentences in high school students’ writing. A few
term papers by student teachers point out that high school students “write as they
speak”. Preliminary results show that many high school students do use features
of spoken language in their writing. Examples of such spoken features include
run-on sentences, multiple clausal coordination, as in (7), sentence fragments, as
in (8), and the use of discourse markers (well, good-bye), as in (9). Another feature of learner English writing involves the use of subordinate clauses as main
clauses, as in (8) and (10) (see also Jonsson 2008).
(7) If I got 10 million sek. I would save some money and give some away to my
family and I would invest in some big compenys (mady sony) shears and I
would buy stuff to me self like som things to my Playstation 3 to exempel a
headset, optik sound and some new consules…
(male student, aged 16, academic program)
(8) My brother study civil engineering at chalmers, so he has a small student
apartment there, which by the way he aren’t using during the summer because he then lives at home. Therefore me and my loved one stayed there on
our own. A big plus in my opinion.
(male student, aged 18, academic program)
(9) The last hotel was a beach hotel and where about 50m from the beach, pretty
nice to wake up and go down and sleep again ... well... I could tell you more
but I think this is enough, goodbye!
(male student, aged 16, vocational program)
(10) At first i didn’t want to sitt next to it. But I’m very satisfied now when i have
done it. Because the tiger was must bigger then me. So it could eat it me with
only one bite.
(female student, aged 16, vocational program)
4.1 T-units, sentences, and clauses
As a measure of syntactic complexity, we will use the concept of the T-unit,
which is short for “minimal terminable unit” and which was first identified and
described by Hunt (1965 and 1966) and further discussed in Hunt (1968). Each
T-unit consists of one main clause and whatever subordinate clauses happen to
be attached to it or embedded in it (Hunt 1966: 737; see also Biber et al. 1999:
179). The passage written by a fourth grader in example (11) is taken from Hunt
(1966: 737). It shows why sentence length is not a good measure of syntactic
complexity. Example (11) is one sentence; thus this student seems very advanced
since he or she writes long sentences comparable to the ones found in journalistic
or scientific prose, for example. In (12), the sentence is divided into six T-units
of varying length and complexity.
(11) I like the movie we saw about Moby Dick the white whale the captain said
if you can kill the white whale Moby Dick I will give this gold to the one
that can do it and it is worth sixteen dollars they tried and tried but while
they were trying they killed a whale and used the oil for the lamps they almost caught the white whale.
The T-units in (12a–f) begin with a capital letter and end with a period; a slant
line indicates a new clause (Hunt 1965: 36–38). As is obvious from (12), a T-unit
can be quite long and complex, as exemplified in (12b): the nominal clause includes an adverbial if-clause and a relative clause.
(12) (a) I like the movie we saw about Moby Dick, the white whale.
(b) The captain said / if you can kill the white whale, Moby Dick, / I will
give this gold to the one / that can do it.
(c) It is worth sixteen dollars.
(d) They tried and tried.
(e) But while they were trying / they killed a whale and used the oil for the
(f) They almost caught the white whale.
According to Hunt (1965: 45–47), a short T-unit includes a maximum of eight
words and too many short T-units are a characteristic of immature writing, as is
also the co-ordination of short T-units with and or but. A long T-unit is defined
as one containing more than 20 words. Hunt’s (1965) study includes 54,000
words collected from the writings of fourth, eighth and twelfth graders. Perhaps
somewhat surprisingly, the average number of clauses per T-unit does not differ
to a great extent between the three student groups: 1.30, 1.42 and 1.68 respectively. There is not a great difference in average T-unit length either: 8.6 words,
11.5 words and 14.4 words. In a more recent study including 241 third grade students’ essays and 238 sixth grade students’ essays, Biber et al. (1998: 178–180)
reported similar results as Hunt. The sixth graders write longer essays than the
third graders but the average T-unit length barely increases (from 9.6 to 10.8
In the ULEC data, approximately 70% of the dependent clauses in a T-unit are
made up of adverbial clauses. The most common subordinators are when, since,
because, and then. Time, place and reason seem to be important meanings to
convey in the ULEC essays. In example (10) above, the dependent clauses are
treated as main clauses, but there are only two T-units in the passage (marked by
[1 and [2 in example (13)).
(13) [1 At first i didn’t want to sitt next to it.][2 But I’m very satisfied now when
i have done it / because the tiger was must bigger then me / so it could eat it
me with only one bite.]
T-unit length in ULEC increases to some extent (cf. Biber et al. 1998: 179) when
the writing of first-year senior high school students (English A) is compared to
that of third-year students (English C). In a small study of 40 essays from each
group of students, the average length of the T-units was 15.3 words for the
first-year students (English A) and 16.4 for the third-year students (English C).
Example (14) (from a third-year student) looks different, perhaps more mature
than (10) above but both (10) and (14) contain only two T-units.
(14) [1 I will ofcourse have a second personality like all the other superheroes.][2
He is going to work as a florist because no one would suspect that a simple
florist would be a superhero.]
(female student, aged 18, academic program)
The more advanced students argue and put forth their ideas in nominal thatclauses and expand the noun phrases in the sentences with relative clauses. When
the topic of the third-year students’ essays is argumentative, this becomes even
more obvious. Constructions such as I think (that), I don’t believe that, people
say that, and they claim that occur frequently in these essays. Relative clauses
are predominantly with that or zero relativizers. When senior high school students use wh-forms, however, they have mastered the personal/nonpersonal contrast, as in (18)–(19). Examples (15)–(19) are all from third-year students’ essays
and the topic is argumentative (“Do you believe in ghosts?”).
(15) I think we all know somehow thay [that] these supernatural forces exist,
otherwise, we wouldn’t be discussing these matters in the first place, because why would people discuss something that ”doesn’t exist” in many
thousand years, and still be interested in it..
(male student, aged 18, academic program)
(16) At this Halloween holiday I borrowed my grandmothers house since she
was on vacation, she lives in a very old and very big apartment, when I were
going to sleep I had a hard time persuading myself that ghosts does not exist.
(male student, aged 18, academic program)
(17) I have seen and heard about people that can talk to the dead and they say that
they can have a dialog with people that have gone to the other side and i belive that these kind of tv-shows are fake.
(male student, aged 18, academic program)
(18) People who say they’ve seen ghosts and stuff like that is probably over reacting, or just lying. I don’t think you can see a ghost, as a white thing, but
I think that you can feel the energy of it.
(female student, aged 18, academic program)
(19) We want answers, answers that we can’t find and that makes it so much
more interesting. So people start to make their their own opinions and ideas
about things, which leads to thoughts about the unknown.
(female student, aged 18, academic program)
Examples (15)–(19) show that the third-year students’ writing has reached a fairly high level of syntactic complexity. More specifically, they use more subordi188
nate clauses per T-unit, both adverbial clauses and nominal clauses (in other
words, they have a higher “subordinate clause index”, see Hunt 1966: 733).
5. Concluding remarks
The Uppsala Learner English Corpus is steadily growing as student teachers collect material for their papers. In the future we hope to be able to include more
essays from junior high-school, so that a more detailed comparison could be
made between junior and senior high school students’ writing. The analysis of
gender differences in writing is another research area which we will turn to. At
this stage, we can only state that many senior high school students, whether male
or female, “write as they speak”. Syntactic complexity in high school students’
writing does increase with age but not to the extent that could have been expected. Here more detailed research needs to be done with the T-unit as a measure of
the degree of syntactic complexity. We will also investigate whether learner
English is different from first language acquisition as regards syntactic development.
Biber, D., S. Conrad & R. Reppen. 1998. Corpus linguistics. London: Longman.
Dulay, H., M. Burt & S. Krashen. 1982. Language two. Oxford: Oxford University Press.
Ellis, R. 1997. SLA research and language teaching. Oxford: Oxford University Press.
Ellis, R. & G. Barkhuizen. 2005. Analysing learner language. Oxford: Oxford University
Ellis, R. 2006. “Current issues in the teaching of grammar: An SLA perspective”. TESOL
Quarterly 40: 83–107.
Geisler, C. & C. Johansson. 2007. “ULEC – The Uppsala Learner Corpus”. Paper
presented at ASLA, 8–10 November, 2007, at Lund University.
Grönblom, L. 2007. “A comparative study of students’ knowledge of English grammar in
the academic and vocational programmes”. Degree project. Uppsala University:
Department of English.
Gröndahl, S. 2007. “Grammar in action”. Degree project. Uppsala University: Department of English.
Hunt, K. W. 1965. Grammatical structures written at three grade levels. Research report
No. 3. Urbana, IL: National Council of Teachers of English.
Hunt, K. W. 1966. “Recent measures in syntactic development”. Elementary English 43:
Hunt, K. W. 1968. An instrument to measure syntactic maturity. Tallahassee, FL: Florida
State University.
Isaksson, A.-S. 2006. “Grammatikens ställning i läroboken – Vad tycker lärare och
elever?” Degree project. Uppsala University: Department of English.
Johansson, C. 2006. “Supplemental instruction och lärarstudenters intresse för grammatik
i språkundervisningen och i examensarbeten”. In Approaches to teaching and learning
in linguistic research: Papers from the ASLA symposium in Växjö,10–11 November,
2005, edited by J. Einarsson, E. Larsson Ringqvist & M. Lindgren, 101–111. Uppsala:
Svenska föreningen för tillämpad språkvetenskap.
Jonsson, E. 2008. “Coherent, cohesive and clear?” Degree project. Uppsala University:
Department of English.
Källkvist, M & S. Pettersson 2006. “An s or not an s; that is the question”. In Approaches
to teaching and learning in linguistic research: Papers from the ASLA symposium in
Växjö, 10–11 November 2005, edited by J. Einarsson, E. Larsson Ringqvist & M.
Lindgren, 112–140. Uppsala: Svenska föreningen för tillämpad språkvetenskap.
Larsen-Freeman, D. 2001a. “Teaching grammar”. In Teaching English as a second language or foreign language, edited by M. Celce-Murica, 251–266. Boston: Heinle &
Larsen-Freeman, D. 2001b. Teaching language: From grammar to grammaring. Boston:
Heinle & Heinle.
Rohm, S. 2006 “Process writing with grammar teaching: A method for contextualising
EFL grammar”. Degree project. Uppsala University: Department of English.
Saville-Troike, M. 2006. Introducing second language acquisition. Cambridge: Cambridge University Press.
Tornberg, U. 2000. Språkdidaktik. Second edition. Malmö: Gleerups.
Thornbury, S. 1999. How to teach grammar. London: Longman/Pearson Education.
Tillenius, E. & D. Johansson. 2007. “A study of upper secondary students’ writing skills.”
Degree project. Uppsala University: Department of English.
Westergren-Axelsson, M. 2000. “USE – The use of a corpus of students’ production in
university English teaching”. In Corpora in research and teaching: Papers from the
ASLA symposium on corpora in research and teaching, Växjö, 11–12 November 1999,
edited by G. Byrman, H. Lindquist & M. Levin, 293–303. Uppsala: Svenska föreningen för tillämpad språkvetenskap.
We thank Peter Hughes, Department of English, for valuable comments on our text.
Contact information
Christine Johansson & Christer Geisler
Department of English
Uppsala University
Box 527
SE-75120 Uppsala
[email protected]
[email protected]
Syntactic Convergence: Marathi and Dravidian
Indira Y. Junghare
University of Minnesota
1. Introduction
Marathi is the southern-most Indo-Aryan language, but some of its most striking
features resemble those of the neighboring Dravidian languages like Telugu and
Kannada. Scholars, such as Bloch (1914) and Southworth (1971) noted many of
these influences in the area of phonology and morphology.
This paper examines some syntactic structures in Marathi (relative clauses,
passive construction, zero pronouns, etc.) and compares them to Dravidian syntax. This paper claims that these similar structures did not result from simple borrowing, but that they are a case of conversion.
Conversion indicates that Marathi developed as quasi-Creole from pidginized
Prakrit through the socio-cultural interaction between the two linguistic groups,
Maharashtrians (Aryans) and Dravidians.
2. Syntactic Constructions
2.1 Full Relative Clauses:
Like other Indo-Aryan languages Marathi relative clause construction consists of
two clauses containing co-referential NP’s. The relative clause may precede or
follow the main clause.
(1) Marathi:
jo mƞus titha ubh he to mjh bhu he
ĝo dm vah khaƤ hai vah mer bhi hai
who man there standing is he my brother
“The man who is standing there is my brother.”
(2) Marathi:
ĝe pustak tu mal dila te maĝh htt he
ĝo kitb tne muĝhe d vo mere htme hai
which book you to-me gave that my hands-in
“The book which you gave me is in my hands.”
2.2 Reduced Relative Clauses:
From the examples in (1) and (2), it is clear that Dravidian languages do not use a
full relative clause construction. In these languages the only dominant clause construction is the reduced relative clause construction. In addition to the full relative
clause structure, which is similar to other Indo-Aryan languages, such as Hindi,
Marathi has reduced relative clauses. We can categorize them in two types: (1)
relative clauses without the relativized NP (or without the relative pronoun and
NP) and (2) Participial relative clauses. Consider the following example.
(3) Marathi:
titha ubh he to mƞus mjh bhu he
* vah khaƤ hai vo dm mer bhi hai
there standing is that man my brother is
“The man (who is) standing there is my brother.”
Marathi, being more inflectional like the Dravidian languages, allows such a deletion since the confusion of reference does not arise due to agreement patterns.
In addition to this reduced relative clause construction, Marathi makes use of
participial construction. Basically it makes use of three types of participles, past
or perfect, progressive and habitual.
(4) Marathi:
(5) Marathi:
(6) Marathi:
[mal philel] mulg paƐl
[nnnu tssin] pillƀu paripyƀu
*[muĝhe dekh hu] laƤk bhg
me-acc. seen
boy ran away
“The boy who saw me ran away.”
(Rel: Subj.)
m [paƀlely mƞsl] phila
nnu [paƀina vaƞƞi] tssænu
*maine [gire huye dmko] dekh
I - inst. fallen man-acc. saw
”I saw the man who fell.”
(Rel: Dir. Obj.)
[m pustak dilel] mƞus
[nnu ami pustakam iina] ayana
*[maine kitb diy hu] dm
I-inst. book given man
“The man to whom I gave the book.” (Rel: Indr. Obj.)
(7) Marathi:
(8) Marathi:
[kl tina pustak dilel] mƞus
[vaƀu ninna ami pustaklu iina] ayana
*[kal usne kitb diy hu] dmi
yesterday she-inst. book given man
“The man she gave the book to yesterday.”
(Rel: IO)
[rmne bolvlel] mulg t l
[rmuƀu piliƀu] attadini lopliki waƀu
*[rmk pukr hu] laƤk andar y
“The boy called by Ram came in.”
(Rel: DO)
If we compare the structure of Marathi, Hindi, and Telugu, examples, (4), (5),
(6), (7) and (8), Marathi resembles the relative clause structure of Telugu, whereas, Hindi does not allow the relativization or modification of subjects, direct objects, and indirect objects by participle phrases.
2.3 Passive Construction
I have shown (Junghare 1988 & 1985) that both Indo-Aryan languages and Dravidian languages are topic prominent and that Marathi is more topic prominent
than Hindi but less topic prominent than Telugu and Kannada, indicating the direct correspondence between topicality and passivization:
The more topic prominent a language is, the less it uses the passive.
(9) Marathi:
mĝhy kaƀun te km kela gela nhi
the work did went not
“I was not able to do that work.” (Capabilitative)
(10) Marathi:
diwƐi diwši laxmii puj keli jte
diwali of day laxmi’s worship did goes
“Laxmi is worshipped on the day of Diwali.” (Perspective)
An examination of the Dravidian languages shows that they do not have passive
constructions. To quote Caldwell (1956: 463) “The Dravidian verb is entirely
destitute of a passive voice, properly so called, nor is there any reason to suppose
that it ever had a passive. None of the Dravidian dialects possesses any passive
particle or suffix, or any means of expressing passivity by direct inflexional
The function/usage of the passive is to mark the passivity or indirectness of
the action, which is clearly a discourse strategy. But when languages use another strategy for indicating the indirectness of the action, or do not involve the
subject by de-emphasizing it, there is no need for those languages to develop
the passive.
2.4 Deletion of Co-referential Constituent (Use of Zero-NP Anaphora)
This rule of deleting co-referential constituent is governed by pragmatics or discourse considerations. In noting the application of this phenomenon, Gundel
(1980) has made the following generalization:
The more topic-prominent a language, the less restricted its use of Zero-NP
It has been suggested that it is the topic rather than the subject that controls the
deletion of co-referential constituent (Le & Thompson 1976, Gundel 1980).
The more Zero-pronouns a language has the more topic prominent it is. IndoAryan languages are more topic prominent than they are subject prominent
(Junghare 1981). Dravidian languages are more topic prominent than
Indo-Aryan and hence make more use of Zero-pronouns than Indo-Aryan languages. Naturally, Marathi being contiguous to Telugu makes more use of
Zero-NP’s than Hindi.
(11) Marathi:
rm itha he. mi tyl phila
rm yah h. maine usko dekh
ram here is
him saw
rmu ikkaƀa unnƀu. nenu (atanni) chusænu
ram here
“Ram is here. I saw him.”
(12) Marathi:
[tu] kuƞl philas? [mi] tyl phila
tumne kisko dekh? maine usko dekh
you whom saw
him saw
0 evarini chusavu? 0 atanni chusænu
“Q: Who did you see?
(13) Marathi:
A: I saw him.”
[tu] kuƬha cllis?
[mi] deuƐt clli
tu kah j rah?
ma mandir j rah h
you where going
(I) temple going am
0 ekkaƀiki veƐtunnru? 0 guƀiki veƐtunnnu
“Q: Where are you going? A: I am going to the temple.”
2.5 Word Order and Topicalization
The word order both in Indo-Aryan and Dravidian is flexible, which allows any
constituent to occur in the sentence initial position and to become topic. There
does not seem to be any constraint on what can serve as the topic.
(14) “I bought that book for Ram.”
mi rmsƬhi te pustak ghetla
maine rmkeliye vah kitb kharid
nnu rmuƬi ksam pustakam konnnu
ram for
that book bought
(15) “For Ram, I bought that book.”
rmsƬhi te pustak ghetla mi
?rmkeliye vah kitb kharid maine
rmuƀi ksam pustakam konnnu
ram for that book bought I
(16) “That book, I bought for Ram.”
te pustak mi rmsƬhi ghetla
?vah kitb maine rmkeliye kharid
pustakam rmuƀi ksam konnnu
that book
ram for bought I
2.6 Basicness of Topic Comment Structures and Marking of Definiteness
Word order in Indo-Aryan and Dravidian is, to a large extent, determined by
topic-comment relation rather than by grammatical relation. Topic-comment
structure seems to prevail in these languages. Post-positional noun phrases seem
to occupy the sentence initial position when they are topics; whereas subject
noun phrases, when indefinite, occur somewhere else in the sentence. Sentences
in (17) illustrate this point.
(17) “There is a book on the table.”
Ƭeblvar ek pustak he
Ƭebalpar ek kitb hai
Ƭèbulu mda pustakam undi
table on one book is
(18) “The book is on the table.”
pustak Ƭeblvar he
kitb Ƭebalpar hai
pustakam Ƭèbulu mda undi
Several grammatical constructions have been examined: full relative clauses, reduced relative clauses, participial relative clauses, passive, deletion of co-referential NP’s or the use of Zero-NP anaphora, word-order variation, and topicali195
zation in Marathi, Hindi, and Telugu, a representative of Dravidian language
family. The analysis shows that Marathi stands between Hindi and Telugu. There
is no doubt that the influence of Dravidian, particularly of Telugu, on Marathi
grammar is significant. Generally, due to contact, languages borrow at the levels
of phonology, morphology, and lexicon. Marathi seems to have gone further into
the level of syntax.
Clearly, Marathi contains two kinds of relative clauses: (1) Full relative
clauses which resemble Indo-Aryan structure, comparable to Hindi, another
Indo-Aryan language. And (2) Reduced relative clause structure, comparable to
Telugu, a Dravidian language. Synchronically, they seem to represent two levels
of discourse, formal as opposed to informal, written vs. spoken. Full relative
clauses represent Indo-Aryan, and formal speech, whereas, Reduced relative
clauses present Dravidian pattern and are more colloquial. Southworth (1971) remarked that the speech of the uneducated (particularly non-Brahmins) and also
of women is consistently less Sanskritized, or more Dravidianized. Changes in
the direction of the Dravidian are often carried through more consistently in
non-standard speech. Can the syntactic patterning be explained simply on the basis of the borrowings due to cultural contact?
The Marathi syntactic and semantic patterns cannot be satisfactorily explained
by the process of borrowing from Dravidian. These similarities show that the intimate parts of the grammatical structures were relatively secure from outside influence. These structures show non-lexical influence, that is the use of inherited
Indo-Aryan morphemes (in most cases) according to completely Dravidian pattern. The process of borrowing involves primarily the transfer of lexical items
from one language to another, though extensive borrowing may also contribute
to structural changes of various kinds.
Some of the structural similarities, such as the patterning of reduced relative
clauses and other syntactic processes could be explained by the pidginization
process, which is distinct from borrowing in that it involves a sharp break in
transmission and the creation of a new code, which serves for communication between two groups which previously had no common language. Pidgins are
popularly thought to combine the vocabulary of one language with the grammar
of the other. Marathi seems to have the vocabulary of Indo-Aryan and grammar
of Dravidian.
3. Implications of Syntactic Similarities with those of
In order to explain the grammatical structures of Marathi which are similar to
Dravidian, Southworth suggested that Marathi is a quasi-Creole language, meaning it might have developed from a pidgin or pidginized parent language. Southworth states that the present characteristics of Marathi are probably the result of
a prolonged process of mutual adaptation between an Aryan language and a local
Pidgin-Creole (or more likely, a series of pidgin-Creoles).
Marathi, even in its oldest known form (tenth century A.D.) presents a picture
of syntactic and lexical convergence; on lexical grounds, it is Indo-Aryan, and
on grammatical footings, it is Dravidian. Grammatical and semantic resemblances with Dravidian have been massive.
4. Other Morphological, Semantic and Phonological
(1) Morphological: Marathi has developed a whole set of negative auxiliaries on
the Dravidian pattern: karat nhi ‘doesn’t work’ karu nako ‘do not work’ (Southworth 1971). It appears that Marathi constructions consist of inherited IndoAryan material (including the initial morphemes) but have been modeled on the
prevailing Dravidian pattern.
(2) Semantic: The most important resemblances between Marathi and Dravidian
are found in the realm of semantics; for example, the inclusive and exclusive first
person plural pronoun [paƞ] ‘we’ (you and I/we, or just us); and absence of
copula which identifies one Np with another (for example, mjha nv rashm).
Also, Marathi shows the development of verbal sequences, called verbal operators such as khun Ƭk (finish up eating).
(3) Phonological: The development of dental affricates, c, and j, and frequency
of retroflex ƞ and  seem to resemble the phonological features of Telegu and
5. Summary & Conclusion:
The paper has examined some Marathi syntactic structures and compared them
with the similar structures in the neighboring languages: Hindi (Indo-Aryan),
and Telugu (Dravidian). The syntactic constructions included full and reduced
relative clauses, participial clauses, passive constructions, use of Zero pronouns
(or deletion of co-referential constituents), word order variation, topic-comment
structures, and marking of definiteness. The analyses showed remarkable resemblances between Marathi and Telugu syntactic constructions, which lead us
to conclude that such syntactic similarities cannot be attributed to simple borrowings and that they have resulted from the process of conversion.
The complex and elaborate structure of relative clauses in Marathi, particularly the reduced relative clauses which are patterned after Telugu, and which are
not so extensively used in other Indo-Aryan languages, seem to provide additional support to Southworth’s theory of the creolized nature of Marathi and its origin
from a pidginized Prakrit. It is recognized that Marathi was developed around
10th century A.D. from Maharashtri Prakrit which was the language of common
folks; Prakrit meaning “naturally evolved.” Whereas, Sanskrit “well formed”
language was the language of Brahmins and the educated. India has been known
for social stratification. In Sanskrit plays, the language of the low classes and
women characters was Maharashtri Prakrit. The language of upper classes and
men was Sanskrit.
Southworth (1971) claims that pidginized Prakrit resulted as a language of
communication between the Dravidian workers and Indo-Aryan employers.
Later pidginized Prakrit was adopted as mother tongue by both groups and became Creole from which developed present day Marathi. The adoption of
pidginized Prakrit as mother tongue changed its status from pidgin to Creole or
quasi-Creole (not fully Creole.)
The following diagram indicates Southworth’s analysis about the origin of
Full Relative Clauses + Relative clauses without relative pronouns
+ Reduced Relative clauses (participial clauses)
Telugu: (Dravidian) Only reduced or participial clauses
Hindi: Full Relative Clauses + limited reduced relative clauses
Two levels : (1) Formal, Standard (Indo-Aryan)
(2) Informal, Colloquial (Dravidian)
OIA => Prakrit => Maharashtrian (Upper class) Pkt. p
Creolized Pkt. => Marathi
Prakrit => Maharashtrian Pidgin Pkt
local lg.
Whether Marathi qualifies as a true Creole or not, the study of its grammatical
structure and its patterning after Dravidian, which cannot be explained by the
process of simple borrowing is surely a case of convergence. It points to the
socio-cultural interaction between the Dravidians and the Maharashtrians. Initially, the Maharashtrians, as Indo-Aryan, might have been employers and considered themselves to be superior to the Dravidians. But in due course of time,
they must have developed neighborly and brotherly economic and socio-cultural
relations that helped shape the language of basic Dravidian structure with the
lexicon from Indo-Aryan, i.e. Marathi.
Apte, Mahadev L. 1974. Pidginization of a lingua franca: A linguistic analysis of HindiUrdu spoken in Bombay. International Journal of Dravidian Linguistics (IJDL) 3(1).
Bloch, Jules. 1914. La formation de la langue marathe. Dev Raj Chanana (English
Trans.), The formation of the Marathi language.1970. Delhi: Motilal Banarsidass.
Caldwell, Robert. 1956. A comparative grammar of the Dravidian languages. 3rd edn. (J.
L. Wyatt & T. Ramakrishna Pillai). Madras: University of Madras.
Gumperz, John & Robert Wilson. 1971. Convergence and creolization: A case from the
Indo-Aryan/Dravidian border. In Dell Hymes (ed.), Pidginization and creolization of
languages. Cambridge: Cambridge University Press. 151–167.
Gundel, Jeanette K. 1980. Zero NP-anaphora in Russian: A case of topic-prominence. In
J. Kreiman and A. E. Ojeda (eds.), Papers from parasession on pronouns and anaphora.
Chicago Linguistic Society (CLS). 139–146.
Junghare, Indira Y. 1990. Discourse considerations for Marathi and Hindi Syntax. Indian
Linguistics 49(March). 66–80.
Junghare, Indira Y. 1988. Topic prominence and zero NP-anaphora in Marathi and Hindi.
In Mohammad A. Jazayery & Werner Winter (eds.), Languages and cultures. Berlin:
Mouton De Gruyter. 309–328.
Junghare, Indira Y. 1985. Topic prominence in Indo-Aryan and Dravidian,” International
Journal of Dravidian Linguistics (IJDL)14(2).
Lehmann, W. P. 1976. From topic to subject in Indo-European. In Charles N. Li & Sandra
A. Thompson (eds.), Subject and topic: A new typology of languages. New York: Academic Press.
Li, Charles N. & Sandra A. Thompson. (eds.). 1976. Subject and topic: A new typology of
language. New York: Academic Press.
Pandharipande, Rajeshwari. 1979. Passive as an optional rule in Hindi, Marathi, and
Nepali. In Braj B. Kachru (ed.), South Asian languages analysis 1. Urbana, IL:
University of Illinois. 89–106.
Southworth, Franklin. 1971. Detecting prior creolization: An analysis of the historical
origins of Marathi. In Dell Hymes (ed.), Pidginization and creolization of languages.
Cambridge: Cambridge University Press. 255–273.
Thanks are due to Professors Rocky Miranda, and K. V. Subbarao for providing the data
from Dravidian languages.
Contact information
Indira Y. Junghare
Institute of Linguistics, ESL & Slavic Languages & Literatures (ILES)
University of Minnesota
214 Nolte Center
315 Pillsbury Drive S.E.
Minneapolis, MN 55455
[email protected]
Using communication strategies to gain
fluency, accuracy and complexity in L2
Katri Karjalainen
University of Vaasa
1 Background
Language skills and learning have several approaches. This article discusses
three dimensions of language skills, namely fluency, accuracy and complexity.
All these aspects are important for a language learner who wants to be able to
function in a second language in different situations and for different purposes.
Mastery of the language code (i.e. accuracy), different sociolinguistic and discourse contexts (i.e. complexity) and strategic competence that helps in communicating effectively (fluency is part of this), are all considered as parts of communicative competence (see Canale 1993: 6–14) and should, hence, be noted in
effective language learning. In this paper I focus on sequences of native speaker
(NS)–non-native speaker (NNS) discussions where the NNS signals problems
with language production. My aim is to give examples of and discuss (i) how the
NNS (in cooperation with the NS) solves these problems, (ii) if these situations
can promote not only interaction but also learning and (iii) if they seem to promote especially one of the three dimensions mentioned above.
1.1 Tandem learning
The informants in my study are participating in a FinTandem course based on
tandem learning (FinTandem. Language learning through communication with
a native speaker 2008). Tandem learning is language learning via authentic interaction situations with a NS in the target language. Two persons with different
mother tongues learn each other’s languages in reciprocal cooperation (Brammerts 2003: 14). In my study Finnish and Swedish speakers learn their respective
target languages from each other. The time they spend with each other is divided
equally between the two languages and partners switch roles and can thus either
function as a second language learner or as a native language tutor (Rost-Roth
1995: 9). Tandem learning is based on principles of autonomy and reciprocity.
The former implies that each learner decides what and how to learn and monitors
his/her own learning, i.e. acts in accordance with learner autonomy. Autonomy
also implies that each learner decides what kind of support and feedback s/he
wishes from his/her partner. Consequently the NS partner is not a teacher but
rather a tutor who assists when asked. The principle of reciprocity is fulfilled
when both partners in a tandem pair profit from the cooperation and both time
and energy are divided equally between the two languages. Partners are supposed to help each other as best they can, in accordance with the wishes of the
other partner. (Brammerts 2003: 14–19.)
Tandem is a communication and learning situation where these two aspects
are combined in a way that gives the learner the kind of opportunities for learning
which are not found in other language learning situations (Schmelter 2004: 15).
Tandem is a form of NS–NNS conversation that is explicitly meant to be a language learning situation despite the authentic interaction situation, where the
content is highly important. In other discussion situations the NS would seldom
correct the NNS because that could be considered impolite (Lightbown & Spada
2006: 32), but in tandem this is not only allowed but desirable and hence a common trait of the interaction (cf. Rost-Roth 1995: 132). Metalinguistic discussions
initiated by the NNS partner are also desirable in language learning via tandem,
whereas they could be regarded as too time-consuming in other NS–NNS discussions, as the goal is merely to get the message through. In this way tandem combines the advantages of both informal and formal language learning situations
(Rost-Roth 1995: 132).
1.2 Communication and learning strategies
Communication strategies (CS) are different solutions that a language user can
choose between when encountering problems in language production. In L2 production CSs can be used when the NNS does not remember or know the correct
expression for the meaning s/he wants to communicate. The aim with using CSs
is to be able to continue interacting despite the linguistic problems that may occur. (Tarone 1980; Færch & Kasper 1984; Mård 2002.) There are many different
taxonomies of communication strategies. Many researchers separate between
avoidance/reduction strategies when the message does not get through, and
achievement strategies, when the intention is to keep to the original meaning
even if the situation develops differently than first planned (e.g. Tarone 1980;
Færch & Kasper 1984) . In my study I am especially interested in achievement
strategies because they include the possibility of learning. By using achievement
strategies the language user can continue the process of communicating. When
the conversation is kept going, the NNS gets more (modified) input from the NS
which is important for learning (Larsen-Freeman & Long 1994: 126).
Because my study concentrates on both the NS–NNS interaction and on what
the NNS intentions are, I use a taxonomy that is a combination of Tarone’s
(1980) interactional taxonomy and Færch and Kasper’s (1984) psycholinguistic
taxonomy. In addition, my taxonomy is also slightly modified while I do not define whether the strategies are cooperative or non-cooperative (cf. Færch &
Kasper 1984). Neither do I consider that CSs always have to be the two interlocutors’ mutual attempt to solve the problem as Tarone (1980) does.
Reduction strategies
Topic avoidance (reduction of the communicative goal by avoiding topics
where vocabulary or other structure is unknown)
Message abandonment (begins to talk about a topic but stops when unable
to communicate the whole meaning)
Achievement strategies
Intralingual strategies (based on L2)
Approximation (use of a super-, sub- or side ordinate concept that the NNS
knows is not correct but, in the NNS opinion, conveys the same meaning)
Restructuring (breaks off in the middle of an utterance and starts again in
a different way)
Word coinage (constructing a new L2 word that does not exist, at least in
that meaning)
Paraphrase (explaining a word with a phrase)
Direct appeal (explicitly asking for support)
Interlingual strategies (based on L1/L3)
Code switching (switching language to L1/L3)
Foreignizing (phonologically and/or morphologically adapted words
from L1/L3)
Literal translation (verbatim use of word(s) on the basis of L1/L3)
Non-linguistic strategies
Non-linguistic act (mime, gesture etc. non-verbal communication)
Indirect appeal (implicitly asking for support, for example questioning intonation)
CSs can be flagged or unflagged. Flagged CSs are marked by pauses, false starts,
hesitation or repetition, indicating that the speaker has problems in putting the
message s/he intends to communicate into words (Watson 2005: 2325). In my research I concentrate on flagged strategies because those sequences are used
while NNS is encountering problems and consequently show how communication can be successful in spite of lacking language skills. As flagged strategies
also make it obvious for the NS tandem partner that the NNS has problems with
his/her language production, one can also assume that they can trigger the NS to
offer support and feedback to the learner. This leads to opportunities for the
learner to get new information about language by eliciting more input and can
thus enhance learning (cf. Faucette 2001: 4).
Learning strategies (LS) are used by the learner to improve his/her linguistic
and sociolinguistic knowledge in the target language (Tarone 1983: 67). Oxford
(1990: 1) defines LSs as “steps taken by students to enhance their own learning”.
LSs can be viewed and categorized from several points of view (see e.g. Dörnyei
& Scott 1997; Macaro 2006 for a theoretical overview). In my study I follow Oxford’s (1990) categorization which divides LSs into direct and indirect strategies
of which the former is further categorized into memory strategies, cognitive
strategies and compensation strategies. Compensation strategies are divided into
guessing intelligently and overcoming limitations in speaking and writing. For
my study it is the last mentioned subcategory that comes into use, because it
compensates gaps in lexical and grammatical knowledge, i.e. it is used for the
same purposes as CSs.
In my study strategies for overcoming limitations in speaking (and writing)
are emphasized because they function in the same way as CSs. Compensation
strategies are however not included in LSs by all researchers and e.g. Faucette
(2001: 3f) criticizes Oxford’s taxonomy (and other taxonomies which include
compensation strategies) as these strategies can be used for learning purposes
only by a proficient learner who intends to get more input and improve his/her
language skills with the help of the strategy. Oxford points out that she wants to
give a wide variety of strategies that influence learning, and thus also includes
compensation strategies (Oxford 1990: 17, 22). Oxford’s (1990: 19) categorization of strategies for overcoming limitations is as follows:
Strategies for overcoming limitations in speaking and writing
1. Switching to the mother tongue
2. Getting help
3. Using mime or gesture
4. Avoiding communication partially or totally
5. Selecting the topic
6. Adjusting or approximating the message
7. Coining words
8. Using a circumlocution or synonym
When comparing this categorization with the taxonomy of CSs presented in this
article, one notices that these strategies overlap each other. Considering the nature of the language learning situation in tandem (learning through communication), I regard the same strategies to be disposable both as CS and as LS depending on the context and the NNS’ intentions.
2 Three dimensions of second language learning:
fluency, accuracy and complexity
As mentioned in the introduction of this article, I discuss language skills and
learning in relation to three dimensions in language skills: fluency, accuracy and
complexity. I am interested in if and how these can be identified as objectives
behind NNS’ behavior in problematic situations where s/he uses CS, i.e. in situations where at least one of the dimensions is not working as the NNS has
I base my approach on the three dimensions of language skills of the Common
European Framework of Reference for Languages (2001, later referred to as
CEFR) which sets six levels for language proficiency. I observe here the two
highest levels (C2 and C1), also referred to as proficient user level which is regarded as the ultimate objective in language learning and teaching in CEFR. Fluency of qualitative aspects on spoken language use on these levels is defined as
C2: “Can express him/herself spontaneously at length with a natural colloquial
flow, avoiding or backtracking around any difficulty so smoothly that the interlocutor is hardly aware of it.”
C1: “Can express him/herself fluently and spontaneously, almost effortlessly. Only
a conceptually difficult subject can hinder a natural, smooth flow of language.”
(CEFR 2001: 28)
CEFR´s description of fluency is in line with Skehan’s (1996: 22). He defines
fluency as a learner’s “capacity to produce language in real time without undue
pausing or hesitation”. In my study fluency is observed as the learner’s focus of
interest in continuing the communication on the actual topic with as little disturbance as possible. Disturbance can be noticed when the learner shows a need
to use flagged CS. Also metalinguistic negotiations and discussions that interrupt
interaction on the on-going topic are classified as disturbance. If fluency is the
learner’s intention when using a CS, one can anticipate that the learner tries to
minimize disturbance and continue with the same topic as soon as possible.
Accuracy is traditionally a fundamental objective in language learning and
teaching. It is concerned with how well the produced utterances follow the rule
system of the target language (Skehan 1996: 22). Even if the communicative
view on language learning emphasizes interaction, it is also important to develop
accuracy. Especially advanced language learners (proficient user level) are expected to be able to use language accurately, which can be seen e.g. in the CEFR
where the following is said about language accuracy regarding spoken language
use on the highest levels (C2 and C1):
C2: “Maintains consistent grammatical control of complex language, even while
attention is otherwise engaged (e.g. in forward planning, in monitoring others’ reactions).”
C1: “Consistently maintains a high degree of grammatical accuracy; errors are rare,
difficult to spot and generally corrected when they occur.”
(CEFR 2001: 28)
In my study accuracy as an intention includes the efforts of the learner to use accurate language for example in areas like morphology, syntax, word choice and
pronunciation. This phenomenon can be noticed in sequences which include CS
and where the learner is willing to suspend the discussion on the original topic
for even lengthy negotiations to get the correct form of the utterance.
The third dimension, complexity, implies the learner’s willingness to use language in accordance with situational demands, in a more challenging and difficult way, and to expand his/her knowledge and use of the language system (e.g.
Skehan 1996: 22; Skehan & Foster 2001: 190). The ability to vary the language
use and being able to cope even in less familiar situations are part of the complexity. While noting fluency and accuracy, CEFR does not explicitly mention
complexity, but I consider the aspect of range to be equivalent to it. Range in the
highest, proficient user levels of spoken language use is defined as:
C2: “Shows great flexibility reformulating ideas in differing linguistic forms to
convey finer shades of meaning precisely, to give emphasis, to differentiate and to
eliminate ambiguity. Also has a good command of idiomatic expressions and colloquialisms.”
C1: “Has a good command of a broad range of language allowing him/her to select
a formulation to express him/herself clearly in an appropriate style on a wide range
of general, academic, professional or leisure topics without having to restrict what
he/she wants to say.”
(CEFR 2001: 28)
Complexity as a learner intention can be observed in my data for example in metalinguistic conversations. These conversations are caused by the learner’s attempt
to have more control over the language, to have a chance to learn new expressions
which can be used instead of the ones s/he already masters or to gain deeper knowledge about stylistic factors. If complexity is the intention, the learner presumably
cares less about fluency and is willing to try new utterances, even if there is a risk
for inaccuracy (cf. Skehan & Foster 2001: 190f). The learner can also try to get a
model for more complex language use from the NS partner.
Communication strategies are discussed in CEFR and it is emphasized that
they are not only used to overcome deficiencies in language skills but are also
used by native speakers when they are appropriate to the communicative demands in the actual situation (CEFR 2001: 57). It should however be stressed
that this relates to strategy use that is so smooth that “it is scarcely noticeable”
(CEFR 2001: 64) while the CSs that I study are flagged (see chapter 1.2) and thus
clearly noticeable and indicate problems in language production.
Figure 1 presents how Skehan and Foster (2001: 190) relate the performance
dimensions to fluency and form:
Figure 1: Theorising dimensions of performance (Skehan & Foster 2001: 190)
In Figure 1 performance dimensions are divided into fluency and form of which
the latter includes both accuracy and complexity. Communication and learning
strategies (see chapter 1.2) can also be reviewed in accordance with the same
model. CSs include the goal of fluency (communicating without disturbance, i.e.
fluently) while LSs concentrate on form; the goal is to learn to use the language
correctly considering situational demands. If the learner’s intention is better accuracy, the focus is on learning to express oneself as formally correctly as possible while complexity emphasizes learning to use the language in accordance
with the situational demands.
3 Material and method
The material for the study presented in this article consists of audio recordings
with three Finnish-speaking learners of Swedish. Informants have recorded their
tandem meetings themselves at three separate meetings during one academic
year. The recordings give me authentic data that show what goes on during tandem meetings. The dyads act in accordance with the principle of autonomy, so I
have no control over topics or how the participants resolve problematic situations. Even if a more predictable data from test situations would be easier to
analyze, I prefer the authentic data and the chance to study tandem meetings as
real situations. These recordings are also a part of the audio data of my forthcoming doctoral thesis.
After thorough listening to the recordings, I have excerpted sequences where
informants display problems in speech production, i.e. flagged CSs, and transcribed these sequences. I analyze these strategies in their context: what is the
problem, how is it solved, has the solution (CS) been successful and can the
situation promote learning. The opportunity of learning depends on the whole sequence and the NS partner’s participation, not only on the use of CS. I have also
included fluency, accuracy and complexity into the account and present how the
learners’ intention to advance in these three aspects can be observed in the analyzed sequences. My intention is not, however, to quantitatively measure fluency, accuracy or complexity of the NNS utterances but to note if learners aim to
develop these three aspects. For the empirical part of this article I have chosen
some examples that highlight this point of view.
4 Fluency, accuracy and complexity in FinTandem
In the following I will present three examples which each represents one of the
three aspects: fluency, accuracy and complexity. In the examples the Finnish-speaking learner of Swedish is marked with an F and the Swedish native
speaker with an S. For transcription principles, see appendix 1. Example 1 is an
excerpt from a discussion about the Finnish-speaking partner being sick:
Example 1.
S: va e de e de liksom huvudvärk eller e de
what is it is it like a headache or is it
F: de e ö e y ingen ö / ingen / (håller) ingen inne / [allt] kommer ut
it’s er er e no er / no / (keeps) nothing inside / [everything] comes out
[jå] just så jaha på de vise jå
[yeah] right so okay in that way yeah
F: jå de va se- sex timmar tillsammans när ja var (komma)
yeah it was si- six hours altogether when I was (come)
S: jaha
F: slutligen
S: jaså
oh yes
F: å detta var tre på morgonen
and this was at three in the morning
In example 1, F has a problem with the Swedish word magsjuka, (stomach bug).
He manages to get the message through with a paraphrase where he describes
what happened. S does not offer F the word he is looking for but signals that he
understands what F means. F also seems to be satisfied with this signal so they
continue discussing the topic without further disturbance. From the communicative point of view the CS works well and disturbance in fluency is minimized.
From a learning point of view this is not effective; F does not get a chance to advance in his second language, in this case to learn an accurate new word. F could
here have stopped and explicitly asked for the right word, but in this case he prefers fluency in interaction over accuracy.
In contrast to the first example, in example 2, the focus is particularly on accuracy, i.e. on choosing the correct preposition:
Example 2.
F: nå ja ö nå öhh / i skolan naturligtvis
well yes er well ur / in school of course
S: mm
F: å vi- vi har bott i XXX ö / i under ja ha s- svårt med de här pre- preposi[tioner]
and we- we have lived in XXX er / for during I have d- difficulties with these
pre- preposi[tions]
[ja pre]positionerna i svenska e svåra
[yes pre]positions in Swedish are difficult
F: jå jå
yeah yeah
S: i om du säger [hur] många år
for if you say [how] many years
[i] i ö / eller över tjugo år
[for] for er / or over twenty years
S: över tjugo år
over twenty years
In example 2 F has problems with choosing the accurate preposition. Instead of
choosing one preposition and trying to get his message through F stops after he
tried to use a couple of prepositions and he initiates a metalinguistic discussion
about prepositions which he finds difficult in Swedish. When F initiates this discussion it functions as a direct appeal for assistance from S. F’s intention with
the discussion is thus not only to communicate but also to make it clear for S that
he needs feedback and support, which in turn promotes learning. S agrees that
prepositions are difficult and offers the right choice in this context. F repeats the
right preposition, which can promote learning, but then also includes another
way of saying what he wants to say. S confirms that this is also an accurate way
of expressing the concept by repeating it. After this section they continue the discussion on the original topic.
In example three the focus is on expressing oneself as well as possible in this
context and finding the best equivalent to the concept F uses, i.e. complexity:
Example 3.
F: dom har alla varit i fängelse:
they have all been in prison
S: mm
F: å sen dom kommer tillbaka å dom kommer till XXX å dom har / dom använder
mycke / narkoti:ka [frågande]
and then they come back and they come to XXX and they have / they use a lot of
/ narcotics [question]
S: ja narkotika
yes narcotics
F: narkotika päihteet / ö
narcotics intoxicant [in Finnish] / er
S: åp ja va sa du
erm yes what did you say
F: ö va e päihteet
er what’s intoxicant [in Finnish]
F: på svenska / likor å heroin å
in Swedish / liquor and heroin and
S: mm (xxx) olika droger
mm (xxx) different drugs
F: olika droger jå
different drugs yes
In example 3 F starts by using an approximation narkotika (narcotics) because
he does not know or remember the Swedish word berusningsmedel (intoxicant).
He is, however, aware of that it is not exactly the word he wishes to use even if
S accepts the word narcotics. F initiates a metalinguistic discussion. He continues with a code switch to Finnish, päihteet, and after a question from S repeats
the code switch combined with a direct appeal for assistance. When S does not
give any answer and there is a pause in their discussion, F uses yet another CS,
a paraphrase, where he gives examples of different intoxicants. S does not use
the word intoxicant but uses the word drugs, which actually is analogous with
narcotics as both are subordinate concepts to intoxicant. In this example F’s use
of CSs confirms that the word narkotika (narcotics) that he already knows is correct and offers him a chance to learn the word droger (drugs). However, F does
not get exactly the word he was looking for, namely berusningsmedel (intoxicant) and is thus unable to make his vocabulary as complex and varied as he intends to.
5 Conclusions
The three examples presented in this article illustrate that strategies for communication, CS, and learning, LS, can be identified as intended to gain different aspects in language skills. The same strategies can be used both as CS and as LS,
but the intention directs them towards differentiation. When the focus is on fluency, I conclude that the strategies also work as pure CS, i.e. they are used to promote interaction between partners. When the focus is on accuracy or complexity
this causes breaks in fluency, in relation to how well partners can continue discussing the original topic. In these cases, strategies will not work so much as CS,
to promote interaction, but more as LS. The situation and discussion are not used
only to communicate a meaning, but also to focus on language and on developing
language skills (cf. Skehan & Foster 2001: 190). This happens in spite of the fact
that they start from a situation where the NNS has problems in communicating
the meaning s/he intended to. In these cases the strategy use starts as a CS but
changes into LS. This is a natural part of conversations in FinTandem because of
the dual aim: to interact with one’s partner and to learn the target language. Combined use of these strategies can be studied if we regard not only the strategy
used, but also the intention when using it.
Brammerts, Helmut (2003). Språkinlärning i tandem och inlärarautonomi. Om utvecklingen av ett koncept. In: Självstyrd språkinlärning i tandem. En handbok. Rapport nr
13, 13–21. Redaktör: Bengt Jonsson. Härnösand: Department of Humanities, Mid
Sweden University.
Canale, Michael (1993). From communicative competence to communicative language
pedagogy. In: Language and Communication, 2–27. Edited by: Jack C. Richards &
Richard W: Schmidt. First published 1983. London: Longman.
Common European Framework of Reference for Languages: Learning, teaching, assessment (2001). Council for Cultural Co-operation, Education Committee, Moderns
Languages Division, Strasbourg. Cambridge: Cambridge University Press.
Dörnyei, Zoltán & Mary Lee Scott (1997). Communication Strategies in a Second Language: Definitions and Taxonomies. Language Learning 47(1), 173–210.
Faucette. Priscilla (2001). A Pedagogical Perspective on Communication Strategie:
Benefits of Training And an Analysis of English Language Teaching Materials. In:
Second Language studies 19:2, 1–40.
FinTandem. Language learning through communication with a native speaker. 2008. http:
// (29 October, 2008.)
Færch, Claus & Gabriele Kasper (1984). Two ways of defining communication strategies.
Language Learning 34(1), 45–63.
Larsen-Freeman, Diane & Michael H. Long (1994). An Introduction to Second Language
Acquisition Research. London: Longman.
Lightbown, Patsy M. & Nina Spada (2006). How Languages are Learned. Third edition.
Oxford university press.
Macaro, Ernesto (2006). Strategies for Language Learning and for Language Use:
Revising the Theoretical Framework. The Modern Language Journal 90(iii), 320–
Mård, Karita (2002). Språkbadsbarn kommunicerar på andraspråket. Fallstudier på
daghemsnivå. Acta Wasaensia Nr 100, Språkvetenskap 21. Vaasan yliopisto.
Oxford, Rebecca L. (1990). Language Learning Strategies. What Every Teacher Should
Know. Boston, Massachusetts: Heinle & Heinle Publishers.
Rost-Roth, Martina (1995). Sprachenlernen im direkten Kontakt. Autonomes Tandem in
Südtirol. Eine Fallstudie. Unter Mitarbeit von Oliver Lechlmair. Meran: Alpha &
Schmelter, Lars (2004). Selbstgesteuertes oder potenziell expansives Fremdsprachenlernen im Tandem. Giessener Beiträge zur Fremdsprachendidaktik. Herausgegeben von
Lothar Bredella, Herbert Christ, Michael K. Legutke, Franz-Joseph Meiner &
Dietmar Rösler. Tübingen: Narr.
Skehan, Peter (1996). Second language acquisition research and task-based instruction.
In: Challenge and Change in Language Teaching, 17–30. Editors: Jane Willis & Dave
Willis. Oxford: Macmillan Heinemann.
Skehan, Peter & Pauline Foster (2001). Cognition and tasks. In: Cognition and Second
Language Instruction, 183–205. Edited by: Peter Robinson. Cambridge : Cambridge
University Press.
Tarone, Elaine (1980). Communication strategies, foreigner talk, and repair in interlanguage. Language learning 30(2), 417–431.
Tarone, Elaine (1983). Some toughts on the notion of ‘communication strategy’. In:
Strategies in Interlanguage Communication, 61–74. Edited by: Claus Færch &
Gabriele Kasper. London: Longman.
Watson, Greg (2005). A Model for the Functional Interpretation of Code-Switching in
NNS-NS Contact Situations. In: Proceedings of the 4th International Symposium on
Bilingualism, 2324–2337. Ed. James Cohen, Kara T. McAlister, Kellie Rolstad & Jeff
Mac Swan. Sommerville. MA: Cascadilla Press.
Appendix 1.
[ overlapping begins
] overlapping stops
- sudden break or stammering
/ pause in the middle of the utterance
(xxx) illegible utterance
(word) transcription is doubtful
[italic font] transcribers comment inserted in the text
XXX name etc. that is left out
Katri Karjalainen
University of Vaasa
Department of Scandinavian Languages
P. O. Box 700
FIN-65101 Vaasa
[email protected]
The “Linguistic Landscape” Method as a Tool
in Research and Education of Multilingualism:
Experiences from a Project in the Baltic States
Sanita Lazdia
Heiko F. Marten
Rzekne University College
1 Introduction
“Linguistic Landscapes” (LL) is a research method which has become increasingly
popular in recent years. In this paper, we will first explain the method itself and discuss some of its fundamental assumptions. We will then recall the basic traits of
multilingualism in the Baltic States, before presenting results from our project
carried out together with a group of Master students of Philology in several
medium-sized towns in the Baltic States, focussing on our home town of Rzekne
in the highly multilingual region of Latgale in Eastern Latvia. In the discussion of
some of the results, we will introduce the concept of “Legal Hypercorrection” as a
term for the stricter compliance of language laws than necessary. The last part will
report on advantages of LL for educational purposes of multilingualism, and for
developing discussions on multilingualism among the general public.
2 Linguistic Landscapes: Some Aspects of the Method
The LL method investigates societal multilingualism by collecting and analysing
language on signs in public – shop windows, road signs, graffiti etc. Its advantages lie in the relatively easy way of obtaining a large amount of data. This data
can be analysed from a quantitative point of view, to which a qualitative angle
can be added by interviewing the persons responsible for or dealing with these
signs. The data collection takes place through taking photos. Through the wide
spread of digital cameras, LL is therefore an easy and enjoyable way of involving
students into field work and thereby motivating them for research in multilingualism. This applies today even to countries beyond the traditional Western
countries, such as the transformation countries in Eastern Europe.
The LL approach can be used in any territory – in cities and rural areas,
whether they are rather monolingual or traditionally multilingual. It has in recent
years been taken up by scholars from all over the world in areas as diverse as
Amsterdam, the Basque Country, Japan, Israel, or Ethiopia (cf. Gorter (ed.) 2006
and Shohamy & Gorter (eds.) 2009 for an overview of LL pioneer studies and
discussions), thereby dealing with regions with various types of multilingualism,
and a global network of scholars has gathered at two inaugural work-shops in Tel
Aviv in 2008 and in Siena in 2009. Our project used one of the (by now justifiably called) “classic” understandings of the approach by analysing the main
shopping streets of middle-sized towns in regions with a high degree of autochthonous multilingualism. Such areas have been chosen for reflecting upon the relationship between majority and minority languages in areas which have been influenced to a lesser degree by international developments than bigger cities.
The main interests of our project thereby related to the following topics: How
is multilingualism reflected on signs in the public sphere and which patterns of
language prestige do they illuminate? How do people acting in the public sphere
react to the needs and wishes of the population? And do the signs indicate differences between linguistic behaviours at the openly visible and hidden levels?
And, finally, in particular in the context of strong language policies: In what way
reflect language practices and prestige patterns the language laws in an area?
3 Multilingualism in the Baltic States: An Overview
The societies of the three so-called Baltic States of Latvia, Estonia and Lithuania
are characterised by a transformation from post-Soviet to European structures.
As part of this process, language legislation is based on the principle of reversing
societal language shift towards Russian caused by more than 40 years of Soviet
occupation. The number of speakers with another language than the State language as L1 is highest in Latvia (about 40%), second highest in Estonia (over
30%) and lowest in Lithuania (under 20%). The number of Latvian L1 speakers,
for instance, decreased from 77% before World War II to 52% in 1991 (Ozolins
2003: 218).
Languages which have to be considered when analysing multilingualism in
Latvia are therefore:
– Latvian as the official State language;
– Livonian – a small Finno-Ugric language which is gradually becoming extinct, and which is protected by law as a traditional autochthonous language;
– Latgalian – a Baltic language which has traditionally often been considered to
be a dialect of Latvian, and is mentioned by law as a „historical variety of
Latvian“; yet, Latgalian today enjoys increased awareness as a separate language – although this is a view which meets resistance by centralist Latvian
– minority languages with support in education, culture etc. – speakers of these
languages have to be divided into the large group of Soviet times migrants
(mostly Russian speakers) and autochthonous speakers of languages traditionally present on Latvian territory such as Byelorussian, Polish, or Lithuanian,
and also including small groups of Russian speakers; and
– foreign languages – traditionally mostly German, today more English, and increasingly also French and other languages.
The demographic situation of Rzekne as the town that we are focussing on in
this article reflects all instances of this multilingualism except for Livonian. Its
36,000 inhabitants are composed of the following ethnicities, with Russians being just short of an over-all majority (Rzeknes pilsetas dome 2008). The composition in Rzekne is contrasted in Table 1 with the figures for the population
in all of Latvia according to self-assigned ethnicity in 2006 (Council of Europe
2006: 3):
Table 1: Ethnicities in Rzekne and Latvia in Contrast
Latvians (including Latgalians)
Language legislation in all three Baltic States has since the early 1990s been
similar in aiming at a reversal of language shift through a reversal of language
prestige and functions. Speakers of the so-called titular languages of Latvian,
Estonian and Lithuanian shall be given the possibility to use their language
everywhere in their countries (cf. Schmid 2008 for an overview of language
policy since the end of the Soviet Union in Latvia; for Estonia: Siiner 2006; for
Lithuania: Hogan-Brun/Meilut 2003; for an overview of language legislation
development in mostly Latvia and Estonia: Ozolins 2003). The Latvian government today is in a process of integrating the two formerly separated systems
of Russian and Latvian schools, with the aim that all students acquire a reasonable knowledge of Latvian. As a result, many young persons with a Russian
background today also have high competence in Latvian – but statistics and
every-day experience also show that there are still considerable numbers of
Russian L1 speakers without such competence (Ozolins 2003: 230). Latvia, as
Estonia and Lithuania, has not signed the European Charter of Regional or Minority Languages, and the ratification of the Framework Convention for National Minorities in 2005 made explicit reservations regarding languages: Minority languages must not be used in administration, and no language other
than Latvian, with the exception of Livonian, must be used on topographic
signs (Council of Europe 2006/2008). Signs, posters etc. in the public must be
in Latvian if they concern the State’s duties, but in exceptional cases they may
also be in other languages. In practice, this rule is used, for instance, for signs
which inform drivers about traffic regulations in Latvia when entering the
country by road – which notably are in Latvian and English, but not in Russian.
For private signs, on the other hand, there is an “at least in Latvian” rule – they
should generally be in Latvian, but other languages may additionally be used
(cf. Latvian Language Law §§21.4 – 21.6, and Cabinet of Ministers of the Republic of Latvia 2000). For Latgalian, in this context, there is a certain degree
of confusion what the denotation as “a historical variety of Latvian” in the
Latvian State Language Law implies: Whereas central authorities in Riga tend
to ignore any status of Latgalian as a separate language, there is today a strong
movement in the region of Latgale to recognise the language as a full-fledged
variety with all rights.
Therefore, in spite of the strong focus on overcoming the marginalisation of
Latvian, Estonian and Lithuanian during Soviet times, language policy in all
three states does not entirely ban the use of languages other than the respective
State language from the public. In addition, private language use is entirely without any restrictions, and in private enterprises, languages other than the State languages may be used as additional languages – a rule which does not only affect
languages of international business or tourism, but explicitly includes Russian
and other minority languages. Only for public authorities, a monolingual State
language policy is carried out, and even regarding this rule, every-day practices
tend to be rather pragmatic than dogmatic – at least as far as oral communication
is concerned.
4 Our Project
4.1 The Approach
The medium-sized towns investigated in our project all have similar roles as regional centres, albeit with different levels and types of societal multilingualism.
The project was carried out throughout the first half of 2008 together with a small
group of Master students of philology at Rzekne University College. After the
introduction of the LL method to the students and a general discussion on various
aspects of multilingualism, research was conducted individually as home-work
from one lesson to the next in Rzekne, and in the group during week-end trips
to the three other towns within the framework of a small project financed by
Rzekne University College: Alytus in Southern Lithuania, Pärnu in SouthWestern coastal Estonia, and Ventspils at the Latvian coast. These field trips resulted in the collection of large amounts of quantitative data, which was sided by
spontaneous interviews for a better understanding of the background and motivation underlying the LL. After that, a time-consuming data-base creation set in,
which provided the basis for the discussion and interpretation of the results in the
course’s final phase.
The specification of our research interests resulted in a list of 27 parameters
which we used during the data-base creation. The parameters relating to the type
and place of the signs indicated the town and the area within each town, the question if the author or the sign was the government or a private person, or the location of the sign in terms of institution: e.g. at a shop, a restaurant, a bank, on an
advertising billboard etc.. In case that the sign could be assigned to a shop, we
also classified the branch, the question if the shop was part of a chain or independent, and a more detailed account of the location – i.e. if the sign was found
at the door, the window, above or in front of the shop. As an example which illustrates the diversity of the signs, Table 2 shows the list of categories for the parameter “Type of Sign”:
Table 2: The Parameter „Type of Sign“ in our Project
Name of establishment
Other sign of establishment
Sign at establishment not by the establishment
Product on display, e.g. in a shop window
Street sign
Personal name plate
Advertising poster
Security information
Private information
Official information
Sign allowing/prohibiting something
Direction sign on private shop or similar
Credit card sign
Security sign
Political information/slogan
Memorial sign
WiFi sign
Student card sign
Sign at door (interphone, post box or similar)
The parameters of the second type were language-related and collected information regarding the number of languages on a sign, the presence of proper
names, the 1st, 2nd, 3rd etc. languages in order of appearance, the question if
there were differences between the 1st, 2nd, 3rd etc. language in size, the type
of font, or the amount of information given, and aspects of translation and language contact. Finally, there were two parameters relating to the more qualitative part of the research about reactions when taking pictures, and about
whether we had spoken with any persons about the individual sign. The list in
Table 3 is again an example – the parameter relates to the translation of texts
on multilingual signs:
Table 3: Parameter “Translation of Multilingual Signs in our Project”
Word by word translation
Free translation
No translation (i.e. there is no overlapping of the content in the different languages at all)
Partial translation
Not applicable (normally when a sign was monolingual)
As a result, we found 17 languages as the first language in order of appearance
during the research. Many of these were expected, such as Latvian, Latgalian,
Russian, Polish, Lithuanian, Estonian, English, German, French, Swedish and
Finnish, but some were also rather unexpected such as Spanish, Italian, Norwegian, Danish, Japanese, and Latin. On the other hand, some languages which we
assumed to find were missing completely, such as Ukrainian.
4.2 Our Project: Quantitative Tendencies
When addressing some of the results of the project, it should first be mentioned
that there are no considerable differences between the results from the four
towns. It is little surprising that there is an absolute dominance of the State languages in all towns included in the project.
What is more striking is that English is more present than Russian, even in
Rzekne, where Russian speakers outnumber Latvian L1 speakers. Whereas the
lack of Russian on signs from public bodies can easily be explained by existing
language legislation, it is more difficult to account for the behaviour in private
businesses, on private notes etc. Russian is, however, more frequently present in
situations at the border of public and more private domains of language use such
as in the stair-cases of apartment buildings – people apparently feel safer to use
Russian closer to their homes where language use is not as visible as on the
streets. There is also a discrepancy between the inside and the outside of shops
and other institutions such as banks: These often use State language-only signage
outside but have multilingual information inside – mostly in Russian, but sometimes also in English. This applies more frequently to leaflets and brochures than
to information about products on display. The latter observation could also be
made in supermarkets – in spite of the fact that in some of these Russian is much
more present than in others, even if largely the same products are on offer. This,
however, only relates to the language of the notorious background music and to
the language spoken by the shop assistants between each other. In relations with
customers, also in more Russian-dominated supermarkets the State language is
usually chosen as the first option to address someone unknown.
The only domain where “top-down“ signs not only in the titular language were
regularly found is tourism. Occasionally, remains of increasingly rotten Soviet
time road or address signs were spotted at private houses – which are bilingual
or even only in Russian. These were mostly found in hidden corners and were
usually easily identifiable as old signs which have not (yet) been replaced – and
apparently have not received attention by the relevant authorities.
English is very frequently used in advertisements, and sometimes also in the
names of shops. Concrete information in English, on the other hand, is much less
regular. German comes second, with a large gap to English, in the list of most
frequent international languages. In contrast to English, its use is much less tokenistic or aimed at the creation of a prestigious image: Tourism information
plates in Latvia are regularly quadrilingual Latvian-Russian-English-German
and reflect the economic potential of German-speaking tourists as well as geographical and historical connections. Otherwise, German appears in rather unexpected places – for instance at newspaper stands. It can be assumed, however,
that most magazines available in German are bought rather for their illustrations
than for their language – frequent journals on display in German deal with
decorations, gardening or other rather visual topics. At the same time, German
products are omnipresent in shops, thereby creating a clearly German-oriented
atmosphere. In addition, cars, trucks and buses regularly display texts in German
(and occasionally also other Western European languages such as Dutch, French,
Swedish or Danish). This, however, is less caused by traffic connections between
these countries, but because transport companies tend to buy old vehicles in Germany without re-decorating them. This even applies to coaches which drive
around the Baltic States on long-distance connections with the names of German
or French transport companies. They frequently only get marginal refreshments
in Latvian, Lithuanian or Estonian where officially required, e.g. for indicating
emergency exits.
Regarding other international languages, different Scandinavian languages are
occasionally present in all cities – mainly in company names or advertisements.
Other languages visible, mostly in special situations, are French, Spanish or
Italian. These enjoy a high prestige as, in this part of the world, exotic and
glamorous. Given the low spread of competence in these languages it cannot be
assumed that these signs are meant for information. Interestingly, Byelorussian
and other traditional minority languages are almost not to be seen at all – with
some very rare exceptions in cafés or restaurants. As the only minor exception,
Polish has a very limited presence in Rzekne. The three Baltic State languages
can similarly occasionally be found in the respective two other countries. Given
their close geographical proximity, however, these are surprisingly rare – and often they feature on advertisements or products by international companies which
only have one version of their package for the three countries.
It is finally noteworthy that there is hardly any written presence of Latgalian
in Rzekne – in spite of the fact that it can frequently be heard in oral use in the
streets of the town. The instances where Latgalian occurs are very few and highly
marked – such as a local radio station, a traditional café, a stone commemorating
the deportation of parts of the local population to Siberia in the 1940s, and, very
infrequently, in graffiti.
4.3 Our Project: Qualitative Results and Interpretation
The qualitative side of the research and its interpretation also show that all languages can be assigned similar functions and hierarchic positions in all three
States. The following conclusions are based on the spontaneous interviews conducted in the shops, cafés etc., and on reactions and observations experienced
throughout the photographing.
One fundamental result is that there is principally very little awareness of linguistic behaviour in the public. This applies even to situations in which people
are confronted with the presence of a sign in their every-day life. Employees often do not know which languages are on display in the windows of the shop in
which they work. In an extreme case, the employee of a café with the English
name “One more” was not able to translate this seemingly simple phrase – and
had to be helped by a customer who joined the conversation. Other answers revealed a very superficial perception of languages and names: the employee of a
beauty parlour with the name “La Femme”, for instance, reported that the shop
owner thought that French was a “beautiful language” – and considered any further contemplation unnecessary. This incidence confirms the assumption that the
use of “exotic” languages often takes place simply for prestige purposes. Passive
attitudes may also be explained by the fact that window displays often are provided by the headquarters of a company outside the region. Local employees
have little influence on it – but in any case, this tendency remains remarkable
from a point of view of linguistic identification of the employees with their
As indicated above, Russian is – in contrast to English – more often present in
situations where a concrete information shall be transmitted, for instance in bilingual job advertisements at the door of a shop. The presence of English is more
symbolic and for obvious prestige purposes – along the lines of linguistic behaviour in other European countries. In such situations, it is not the meaning of
the English text which is in focus – such as in advertisements in the State language with an added slogan in English. Interestingly, there is a certain tendency
towards trilingualism in Latvia in Latvian, English and Russian, e.g. in telephone
booths: Here, the display of Russian is apparently seen as advantageous by the
phone company – but it tries to avoid to create a picture of a bilingual society
which would be given if only Latvian and Russian were used. Some international
languages repeatedly occur in specific situations only: Norwegian, for instance,
is regularly present in the name and the products connected with a major Norwegian petrol station chain, or Italian in the advertisements for opticians. It was in
one optician’s store also that we found one of the rather rare examples in which
the manager was indeed very aware of the LL in her shop and happily engaged
in a conversation on Italian as a main language of advertisement in her business.
Other seemingly odd situations could be clarified easily through a brief investigation, such as the use of Spanish in the quadrilingual advertisement for an
Italian restaurant featuring text pieces of Latvian, English, Italian and Spanish in
Rzekne. Since it can not be expected that too many residents would recognise
the difference between Italian and Spanish, and the number of Spanish visitors
to the region is marginal, this left us at first with a big question mark. However,
there was no mistake on the poster – the reason was simply that one of the owners
of the restaurant has a Spanish background. This is one other important result of
the research – for many “irregular” or “unexpected” situations of language use
there are concrete explanations which render cases individual rather than prototypical.
Regarding local minority languages, where they are present at all, it should be
noted that they rather appear in their international than their local functions. The
rare stickers in Polish which we found in Rzekne come from Poland and are not
connected with the Polish minority in Latgale, but relate to Polish companies or
cultural events imported from Poland – and, similarly to German, to Polish
products in the shops. Advertisements by Polish or Lithuanian companies were
much more frequent than signs with local information in these languages; one
rare exception was a local Byelorussian café in Ventspils. The only major exception to this is the Polish school in Rzekne. Russian, on the other hand, which in
spite of it decreasing importance can still be called an international lingua franca
in post-Soviet countries, features mostly in its local rather than in its global function. Notes in Russian were usually addressed to locals rather than aimed at international business relations – with some exceptions such as for tourist information purposes. Some situations are difficult to interpret in this respect – such as
the use of Russian in telephone booths or in cash machines mentioned above.
Some of the very few Russian-only signs could be found at Russian Orthodox
As a final result, it is important to remark that the data collection regularly provoked interest by passers-by. Their reactions occasionally turned out to be sceptical or even openly hostile. In several situations, the researchers were literally
chased away during the process of photographing a shop. Here, ethnic tensions
in the Baltic States had their direct influence on the linguists’ work – but also that
is, of course, a part of the whole picture and as such a result of the investigation.
It is not always easy to draw the right conclusions from these situations: This behaviour might be based on negative experiences with controls by the State language inspectors. Alternatively, shop assistants might also have believed that we
were working on behalf of competitors in order to investigate their business strategies. An initially hostile reaction repeatedly relaxed slightly when we explained
carefully our aim and our background as scientists within an academic project –
but hostility usually only turned into scepticism, but not into welcoming. On the
other hand, we have also met occasional true interest in the research and the possibilities for creating discussions around multilingual issues generated by its results – such as by a local radio station in Rzekne which showed active interest
concerning the results regarding Latgalian.
Summarising these qualitative and quantitative results, it is possible to draw
the following hierarchy of languages in the LL of Rzekne – regarding both
their frequency and their functions. Language hierarchies in the other towns
are similar – with the exception of the lack of a language corresponding to Latgalian:
1. Latvian
2. English (Prestige) / Russian (Functionality)
3. Russian (Prestige) / English (Functionality)
4. Other international languages
5. Latgalian
6. Local minority languages
4.4 “Legal Hypercorrection”
As a last aspect of the research results, the findings presented above allow the
conclusion that the LL in the Baltic States is indeed heavily influenced by State
language policies, in particular regarding the relation between Russian and the
titular languages, considering the large number of speakers with Russian as a
mother tongue. In that, language laws regarding written language in public even
seem to be followed to a higher degree than necessary: As explained above, language policy and legislation would allow the use of other languages than Estonian, Lithuanian and Latvian in addition to the respective State language on any
sign which is not subject to the immediate sovereign functions of the state.
Thereby, any information by a private business or any personal note could contain parts in other languages in addition to the State language, including Russian.
The fact that there are only rather few examples where this is done shows a phenomenon which can be labelled as “Legal Hypercorrection”. In analogy to the
classic notion of “hypercorrection” in sociolinguistics, Legal Hypercorrection
shall thus denote a fulfilment of linguistic legal norms by language users to a
higher degree than necessary. Such practices are in the written language of the
Baltic States synonymous to a high degree of linguistic accommodation towards
a variety which is perceived as more prestigious, and related to clear power relations between languages and their speakers. In addition, Legal Hypercorrection
in Rzekne does not only affect the use of Russian – it applies also to Latgalian:
Because of a lack of knowledge about the status of Latgalian in language laws
with a certain (albeit not clearly defined) degree of recognition, people feel insecure about using it in the written public space – and therefore use it only very
Two major reasons for the remarkable lack of awareness about language
legislation can be deduced from the research. First, there is a lack of knowledge
about policies and laws among the population in the Baltic States in general –
based on the lack of a tradition of democratic participation in these post-Soviet
transformation societies. Among large parts of the population there continues to
be a prevailing attitude that decisions are taken from above, that individuals are
not supposed to take initiative and responsibility, and that the general public’s
opinion doesn’t matter. At the same time, what maintains this attitude to languages is the fact that any notion of an active promotion of multilingualism – re221
gardless if that would be relating to Russian, to Latgalian or to other local languages – is not on the agenda. The state does not openly encourage the use of
either of these languages – whereas efforts that people should learn English have
regularly become parts of education programmes in recent years. The latter aspect is connected with the second reason – the fear of authorities and in particular
of the language police, as experienced through the hostile reactions during the
photographing process. Connected with the lack of interest in societal affairs,
many people have little awareness of their rights, including issues relating to language. The Soviet legacy therefore does not only influence the linguistic composition and the LL of Baltic societies – it also influences efforts to overcome linguistic tensions as it renders an open debate more difficult.
It may therefore be concluded that Legal Hypercorrection in the Baltic States
is based both on the wish to participate in prestigious parts of society – and on a
lack of knowledge of laws and a fear of punishment by the authorities.
5 Benefits of the LL Method Beyond the Core Research
5.1 Advantages of the Method: Education
As the last part of our paper, we are now going to reflect on the potential that the
LL method has beyond the core research interest. Our project was explicitly designed to use its opportunities as a tool in education and for raising multilingual
The students involved in the project were benefiting from the project through
enhancing their understanding of linguistic patterns in the society they live in,
and from the discussions of the effects of language legislation and policy in their
practical application in reality on language power relationships and hierarchies.
An additional benefit was that the students were able to start their own field work
projects – instead of studying in a theoretical way only. By getting out of the
classroom, students got the feeling of doing something “concrete” and of exploring something new. They thereby benefited on various levels – through the practical application of a previously studied topic in field work, through the statistical
analysis and other processing of their data, and they finally could practice the
presentation of their research results in essays and at a students‘ conference.
Practical advantages for the project leaders included the fact that the amount of
data which could be collected and processed was increased by the higher number
of active researchers. Of course, it should not be forgotten in this context to give
the students the credit which they deserve for their share of the investigation.
However, the involvement of students requires that they are well chosen, have
enough true intrinsic motivation, and are reliable. Therefore, the use of this
method is suited for a small truly interested group (between 4 and 8 students in
our case) rather than for larger seminars. Our course required more commitment
and time than “ordinary“ seminars – in particular of those students who joined
us on the research trips to the other towns beyond Rzekne. At the same time, as
one of the additional positive outcomes, a team building process set in among
students as well as between students and teachers, from which all participants
could profit in the long run. All that was possible at relatively low costs – even
though students received funding for accommodation, transport and eating. In
our case, it was not too difficult to convince the University to support this project
since it so obviously connected research with the integration of students into
academic life. Finally it should be stressed that, whereas our project was conducted in order to create a synergetic effect for teaching and research, the method
could also be used in secondary, and to a lesser degree, even primary schools.
There, however, the teaching aspect would obviously be more central than the
5.2 Advantages of the Method for the Creation of Debates around Multilingual Issues
In addition to its potential in education, the LL method also carries a high value
for the creation of awareness of language issues among the general public. This
starts during the research itself: It is inevitable that people notice what the researchers are doing – it is simply too obviously unusual to see someone walking
around town taking pictures of any public sign – from every shop window to
every sticker on a rubbish bin. Where it is appropriate, such situations can already be used for creating debates, explaining what linguists do, or speaking
about multilingualism. On the more personal level, this applies also to the opportunity for involving friends and family into one’s work through the presentation
of research examples.
The results of the research may then be used for a visualisation of language
policies, multilingualism, and linguistic work in general on any level. Lecturers
and students much more easily gain a forum for public presentations with such a
topic which is much closer to many persons’ every-day experience than many
others. Beyond linguistic circles, the pictures are very useful for entertaining
conference presentations and at any general university event. Lectures in and
outside the academic world illustrate what linguists are doing – and thereby
people might start to be more conscious of their linguistic environment. Posters
may easily exemplify the work at exhibitions, festivities at university, library
events, or within university marketing campaigns. In addition, they may also be
helpful in creating a connection between academics and public events in the local
city – such as at lectures in local libraries, for exhibitions in museums of local
history, and because of its artistic potential even in art galleries.
One such example where we used our project for linguistic marketing was the
participation of Rzekne University College in the European Researchers‘ Night
in September 2008: We used pictures taken in the four towns of the research for
creating a quiz around them. Questions asked related, for instance, to the languages of the texts on the pictures shown and their meaning, to the specific place
where we found a sign, but also to more interpretive topics such as why a sign
was at a certain place. The audience consisted mainly of High School students
who had come to the University College for getting an insight into its work.
There were different levels of the questions, and some of them proved to be very
difficult – but in general, feed-back was very positive and the young visitors understood that linguistics can be an enjoyable enterprise.
Similarly, also in tourism, LL results may easily be used for creating a valuable tool for visualising cultures and languages in an area. Visitors to a region
interested in cultural tourism often find it rather difficult to experience theoretical readings in reality – but through the use of LL results, concrete examples of
language use can be shown. Both from an aesthetic point of view and for systematic information, examples of multilingualism can be sided by language laws or
information about cultures, and wherever possible set into relation with the explanation of historic events. This can happen at monuments, historically important sites and other places, but also in the context of visualising the situation under which language groups and other minorities live, such as at cultural centres
or in the presentation of organisations of linguistic and cultural minorities. Obviously, this can be followed by a number of side-products such as books, brochures, post cards or posters – which can be important promotional tools for the
development of tourism, and may even carry a certain economic importance, in
particular in rather peripheral areas. The city of Rzekne, for instance, has produced a colourful DVD with short video clips of its minorities: Each of these
films presents one local minority by associating one colour and one song to each
group – the clips show both their traditions and how these find their expression
in contemporary society (Rzeknes pilstas dome & Rzeknes nacionlo
kultras biedrba 2006).
And finally, as the example of the Latgalian radio station has shown, this is
also a method that local media in small towns may show interest in. Also in this
way, discussions resulting from LL research may contribute to the awareness of
a minority language and thereby to the promotion of tolerance for smaller languages.
All these examples show that the Linguistic Landscape method does not only
have a value for understanding hierarchies of language use and prestige. In our
project in the Baltic States, it has also proved to be a useful tool in various settings in education and beyond. We therefore encourage all interested students,
researchers and teachers at any level of education to apply the LL method to their
own environment. As our experience shows, the relation between its easy application and the diversity of results which can be obtained are therein certainly
more favourable than when using many other methods.
Backhaus, Peter. 2009. Rules and Regulations in Linguistic Landscaping: A Comparative
Perspective. In Elana Shohamy & Durk Gorter (eds.): Linguistic Landscape: Expanding the Scenery. London: Routledge. 157–172.
Cabinet of Ministers of the Republic of Latvia. 2000. The Usage of Languages in Information Issued in accordance with Paragraphs 5 and 6 of Article 21 of the State Language L, Regulations No. 292, 22 August 2000. Latvia_LangRegInformation_English.htm (November 30, 2008).
Cenoz, Jasone & Durk Gorter. 2006. Linguistic Landscape and Minority Languages. In
Durk Gorter (ed.), Linguistic Landscape: A New Approach to Multilingualism. Clevedon/Buffalo/Toronto: Multilingual Matters. 67–80.
Council of Europe. 2006. Report submitted by Latvia pursuant of article 25, paragraph 1
of the Framework Convention on the Protection of National Minorities, http://
PDF_1st_SR_Latvia_en.pdf (30 November 2008)
Council of Europe. 2008. List of Declarations Made with Respect to Treaty No. 157, http:
// (30 November 2008)
Edelman, Loulou. 2009. What’s in a Name? Classification of Proper Names by Language.
In Elana Shohamy & Durk Gorter (eds.), Linguistic Landscape: Expanding the
Scenery. London: Routledge. 141–154.
Gorter, Durk (ed.). 2006. Linguistic Landscape: A New Approach to Multilingualism.
Clevedon/Buffalo/Toronto: Multilingual Matters.
Hogan-Brun, Gabrielle & Ramonien Meilut. 2003. Emerging Language and Education
Policies in Lithuania. Language Policy 2/2003. 27–45.
Lazdia, Sanita & Solvita Poseiko & Heiko F. Marten. 2008. Lingvistisks ainavas
metode – netradicionls ceš multilingvisma jautjumu izpt un mcšan. tagad 1/
2008. Zintniski metodisks izdevums. 43–49.
Ozolins, Uldis. 2003. The Impact of European Accession upon Language Policy in the
Baltic States, Language Policy 2. 217–238.
Republic of Latvia. 1999. State Language Law. Latvia/Latvia_Language_English.htm (November 30, 2007)
Rzeknes pilstas dome. 2008. Skaiti un fakti.
(November 30, 2008)
Rzeknes pilstas dome/Rzeknes nacionlo kultras biedrba. 2006. Rzekne un ts
audis. Rzekne: Rzeknes pilstas dome.
Schmid, Carol. 2008. Ethnicity and language tensions in Latvia. Language Policy 7/2008.
Siiner, Maarja. 2006. Planning Language Practice: A Sociolinguistic Analysis of Language Policy in Post-Communist Estonia. Language Policy 5/2006. 161–186.
Shohamy, Elana & Durk Gorter (eds.). 2009. Linguistic Landscape: Expanding the
Scenery. London: Routledge.
Sanita Lazdia & Heiko F. Marten
Department of Philology, Rzekne University College
Atbrvošanas aleja 115
Rzekne LV-4600, Latvia
[email protected]
[email protected]
The multilingual history of an industrial society
The case of Tampere, Finland
Harry Lönnroth
University of Tampere
1 The rationale
In this paper I discuss the multilingual history of an industrial society. The
society in question is the city of Tampere, Finland, founded in 1779 by King
Gustav III of Sweden (on the history of Tampere, see Voionmaa 1929a, 1929b,
1932, 1935; Jutikkala 1979; Rasila 1984, 1992; Alhonen et al. 1988). The point
of departure is that multilingualism in history is of fundamental linguistic,
historical and social interest (cf. also Braunmüller & Ferraresi 2003: 3). In the
sociolinguistic literature on multilingualism, however, the historical dimension
is often forgotten or, at least, not enough emphasized. By historical dimension I
mean here the study of historical disciplines such as urban and social history. The
lack of historical reasoning in modern sociolinguistic research is, of course, most
regrettable. The same is true for linguistic matters in historical research. That is
why the interdisciplinary perspective is of vital importance in historically oriented sociolinguistics or the sociology of language. As Trudgill (2000: 21) points
out, a research on language “totally without reference to its social context inevitably leads to the omission of some of the more complex and interesting aspects of language and to the loss of opportunities for further theoretical
progress”. By social context I refer here to the historical context in which the
members of a language community interact. Hence, a study of the multilingual
history of an industrial society must be interdisciplinary with a focus on the interplay between language and society, that is, history. Tandefelt (2003: 87, see
also 2002), who builds her discussion on the four languages of the former Finnish city of Viipuri (Vyborg) on historical literature, writes: the “historical framework is, however, absolutely necessary in order to be able to understand what
role the different languages and the city’s multilingualism has played in the life
of the inhabitants”. The historical dimension does not, naturally, exclude the fact
that the main perspective of a sociolinguist first and foremost must be on linguistic matters.
1.1 The project
The paper is a part of the postdoctoral researcher’s project “Linguistic Change in
an Industrial Town. Swedishness in Tampere, 1779–2000”, financed by the
Academy of Finland for the period 2008–2010. The project concentrates on societal multilingualism, especially Swedish but also other languages, in the city of
Tampere from the end of the 18th century to the end of the 20th century. The
main question of the project is the socio-historical development of so-called
Swedishness, which includes the study of Swedish language and culture in Finland. I work on the assumption that it is linguistically and socially relevant to
study the social history of Swedish in a Finnish-speaking industrial city in inland
Finland. The frame of reference of the project is the sociolinguistics of society
(e.g. questions concerning language groups and their social structure, national
awakening and nationalist movements, language attitudes, language choice and
language maintenance and shift). The development of Swedishness is analysed
with regard to major historical events in local and Finnish history (Finland as a
part of Sweden until 1809, Finland under Russian rule 1809–1917 and independent Finland from 1917 onwards). The case of Tampere is particularly interesting
because Tampere was the most prominent industrial city in the Nordic countries
until the beginning of the 20th century; nowadays it is the biggest Swedish
speech island in Finland (in Swedish språkö). The Oxford English Dictionary
(1989) defines the word speech island as follows: “a small area inhabited by
speakers of a language or dialect other than that spoken in the surrounding
areas”.1 The starting point of the project is that languages “came” to Tampere
with the advent of heavy industry at the beginning of the 19th century and “disappeared” when the smokestacks started to fail in the 1980’s. (On Tampere as a
language community, see contributions in Lönnroth, forthcoming a; on aspects
on multilingualism in Finland, see Latomaa 2005.)
The project is the first in the Nordic countries to concentrate on the socio-historical analysis of the Swedish-speaking part of the nation’s linguistic, cultural
and social structure in the context of a Finnish-speaking industrial city. I examine
the history of the Swedish language in the light of socially critical periods such
as industrialization, urbanization and globalization. One of the aims of the project is to produce additional information about the linguistic influences of social
changes by explaining the sociolinguistic factors that affected language use and
social status.2 The project focuses on the social history of the written and spoken
Swedish of Tampere. As mentioned above, the linguistic perspective is on the sociolinguistics of society (cf. Fasold 1984), not on the sociolinguistics of language
(cf. Fasold 1990; for criticism of the distinction, cf. Romaine 2000: x). In other
The three other established Swedish speech islands in Finland today are the coastal cities of Kotka
(approx. 580 Swedish-speaking inhabitants), Pori (approx. 380 Swedish-speaking inhabitants) and
Oulu (approx. 290 Swedish-speaking inhabitants).
The development of bilingualism or multilingualism in Finland in general and in Tampere in particular is internationally comparable to the linguistic conditions of Belgium, Canada and Switzerland,
for example (cf. McRae 1999, 2007).
words, I do not concentrate on explaining whether linguistic feature x is “Tampere Swedish” or not. Rather, I am interested in the users of the Swedish language and the status of the language in the development of a Finnish industrial
city during historically critical periods.
1.2 The perspective
In this paper the terms bilingualism and multilingualism are interchangeable (cf.
Romaine 2000: 33). Thus my emphasis is mainly on Swedish language and culture in Tampere. However, I want to highlight the fact that Finnish and Swedish
neither were nor are the only languages spoken in Tampere. In the following I
will therefore discuss other languages and language groups in Tampere as well,
mainly during the 19th century and in the light of historical literature. The underlining of the Swedish language is relevant because of the status of Tampere as a
major Swedish speech island. In other respects, too, Swedish has had a special
status in Finland; nowadays Finland is, as is well known, officially a bilingual
country with two national languages: Finnish and Swedish. In addition to the national languages there are also three official minority languages in Finland today:
Sámi, Romany and sign language. They are not discussed in this paper.
In his book Finlandssvenska (‘Finland-Swedish’), Hugo Bergroth, a Finland-Swedish language planner, divides the written and spoken Swedish in Finland, so-called Finland-Swedish, into four categories.3 His discussion from 1928
is still partially relevant for the perspective of the study:
The Finnish Swedish […] is thus not a unitary conception. There are different kinds
of Finnish Swedish: we can distinguish between Swedish spoken by Finns, the language spoken and written by Finnish-speaking people who have more or less successfully tried to learn Swedish; Viipuri Swedish, strongly influenced by Finnish,
German, and Russian; the Swedish spoken in the manors and towns of the Finnish-speaking regions (“behind Kuopio Swedish”), influenced by Finnish; and, finally, Finland Swedish proper, the idiom of the educated class which, more or less
Swedish in origin, evolved in places where it has not been exposed to overwhelmingly strong Finnish influence. (Bergroth 1992: 17; my translation from Swedish.)4
Over the years, several books and articles have been written on the Swedish of
the Finnish-speaking people (“finsk-svenska”), the Swedish of the city of Viipuri
(“viborgssvenska”) and the so-called Finland-Swedish proper (“egentlig finlandssvenska”). Finland-Swedish proper, in particular, has been studied using
The term Finland-Swedish (in Swedish finlandssvenska) originates in early 20th century language
planning. Therefore, the term is not used to refer to the time prior to the 20th century. The appropriate
term is Finnish Swedish (in Swedish finländsk svenska).
In the original Swedish: “Den finländska svenskan […] är sålunda inte något enhetligt begrepp.
Det finnes olika slag av finländsk svenska: vi ha att skilja mellan finsk-svenskan, det språk som talas
och skrives av finsktalande personer vilka mer eller mindre framgångsrikt sökt lära sig svenska; viborgssvenskan, starkt påverkad av finskan, tyskan och ryskan; herrgårds- och stadssvenskan i landets finska bygder (»bakom-kuopio-svenskan»), påverkad av finskan; och slutligen den egentliga finlandssvenskan, idiomet för den bildade klass som, mer eller mindre svensk till sin härstamning, vuxit
upp på orter där den icke varit utsatt för något överväldigande starkt finskt inflytande.”
various methods and data. Recently these studies have mainly concentrated on
spoken language. The Swedish that was spoken in the manors and towns of the
Finnish-speaking regions of inland Finland (“herrgårds- och stadssvenska i landets finska bygder”) is a little studied phenomenon in linguistics and history.
From a linguistic point of view, the Swedishness of inland Finland is more closely affected by the Finnish language than Finland-Swedish on the whole, which
is seen, for example, in bilingualism, language contact and code switching. The
Swedish that is spoken in the Tampere region, at times called “Tampere Swedish”, belongs to the category of the Swedishness of inland Finland. Therefore,
it forms one secondary Finland-Swedish language community, which is located
outside the primary Swedish-speaking areas of Finland.5
The studies of the Swedish language and Finland-Swedish culture in Finland
have traditionally concentrated on so-called “Svenskfinland”, that is, the language spoken in and the culture of the coastal regions of the provinces of Uusimaa, Turunmaa, the Åland Islands and Ostrobothnia. However, in the histories
of many inland industrial localities, the Swedish-speaking minorities and factory
owners in particular have held a crucial position in society. One can say that
Swedish capital found its way into forests and hydroelectric power; Tampere,
too, was founded on the Tammerkoski rapids. The Swedish language and culture
“appeared” along with factories and smokestack industry, but later on they also
“disappeared” from the originally Finnish-speaking areas, such as the province
of Häme.6 The inland localities developing around industry include for example
the towns of Mänttä, Nokia, Tampere and Valkeakoski. Tampere, which is the
focus locality of my research interest, is the only one of the inland localities
which still remains a speech island, that is, there is a Swedish-speaking bilingual
language minority. At the same time, Tampere is the only speech island in inland
Finland, and also the largest in the whole country.
The selection of Tampere as the focus locality of a sociolinguistic study can
be justified by its distinctive industrial tradition. From the 19th century to the
mid-20th century, Tampere was the leading industrial city in the Nordic countries. During that period many Swedish-speaking people and families with industrial influence settled there and became more or less bilingual. Regardless of
the Swedish-speaking influence, however, Tampere was for a long time characterized as a Finnish-speaking working class city. In 1900, 5.6 per cent of the
people in Tampere were Swedish-speaking; in 1930 they numbered 2.6 per cent.
At the end of 2007 there were 1,066 people with Swedish as their first language,
which was 0.5 per cent of the total population. The number of bilingual Tampere
The word Tampere Swedish first occurs as early as in Voionmaa (1929a: 353, 1929b: 331, 1932:
284). Because there is lack of adequate research information, it is uncertain whether it is even possible to talk about any specific “Tampere Swedish”. I use “Tampere Swedish” as an operationalization
to describe the language of those Swedish-speaking people who originally lived in Tampere or have
moved there from somewhere else.
For example, in the manors Swedish often remained the language that was spoken at home. “Tampere Swedish” represents the language of urban society even though Tampere was for long a small
residents has always been greater: bilingualism is a linguistic phenomenon
which to a considerable extent is characteristic of Swedish-speakers in Tampere
(the case is the same in many other parts of the country as well, e.g. in the cities
of Helsinki and Turku). In addition to the Swedish-speaking people (which also
includes people from Sweden), English- and German-speaking experts came to
the city along with industry, thereby increasing the linguistic diversity of Tampere. Nowadays there is a Swedish day-care centre, school, residential home,
parish, newsletter and about ten associations which maintain the solidarity and
linguistic identity of the Swedish-speaking people not only of Tampere but also
of the whole Tampere region (see Lönnroth, forthcoming b). Thus the notion of
“Tampere Swedish” includes all the Swedish-speaking citizens of Finland who
live or have lived in the Tampere region.7
2 Multilingualism in the history of Tampere
This section focuses on multilingualism and multiculturalism in Tampere past
and present. The following languages and language groups will be addressed:
Finnish and Swedish (2.1) followed by other languages like Russian and German
(2.2). First a few words about the historical setting in which the languages of
Tampere presented in this paper coexisted.
At the end of the period of Finland under Swedish domination, Tampere was
still a modest rural town with confusing conditions of nationality and education.
According to the parish register, in 1800 there were 463 residents in the town. A
decade later, in 1810, there were 682 residents and in 1860 already 5,232 residents in Tampere. During the 20th century the demographic development was
enormous; in 1910 there were 45,442 residents, in 1930 already 55,697, and even
more in 2008, when there were 207,866 residents in Tampere. Finnish had been
the vernacular language of Tampere, and therefore, it was impossible to found
an entirely Swedish-speaking town in this locality. Even the majority of the bourgeois of the town had an indifferent command of Swedish. Divine service was
held on Sundays by turns in Swedish and in Finnish, and on the most important
religious holidays in both languages. (Voionmaa 1929a: passim.)
The image of Tampere has long been, and in part continues to be, that of an
industrial city. For instance, many foremen and machinists at the Finlayson cotton factory had arrived from England and Sweden because of the small amount
An analysis of the Finnish linguistic conditions is given in the political-historical work Conflict and
Compromise in Multilingual Societies by McRae (1999), in which the author discusses linguistic diversity from a historical point of view. He studies language as a part of a wider social and political
history. Other relevant literature includes the study by Harmes (1992) on the population structure of
the Swedish-speaking people in Tampere, their regional mobility and inner identification during the
years 1926–1939 from a socio-historical point of view. Keskinen et al. (2005) study the language
connections between the different districts of Tampere 1918–1940. Kirmo (2005) deals with the Russian population of Tampere 1918–1950. She also discusses its relation to the Swedish minority of the
Finnish expertise available. By the late 1830s Tampere had already acquired the
image of an industrial city. The notion of Tampere as Finland’s Manchester began to be emphasized in literature of a different kind. At the beginning of the 20th
century, too, especially during the 1920s and 1930s, Tampere maintained but
also itself constructed the image of an industrial society, and became the centre
of the Finnish industry. (Hietala & Kaarninen 2005: 187–189, 195.) From the
1940s to the 1960s the self-image of Tampere was the “beautiful city of factories” (Hietala & Kaarninen 2005: 197).
Haapala (2005: 165) states that Tampere was “the first major inland city of its
kind”. According to him it was also “the first and foremost industrial city of Finland in the 19th century and continued to be that in the 20th century” (Haapala
2005: 172). Tampere was also, he continues, “the most typical industrial city but
not the only one”. Factories like Finlayson imported all their technology and
know-how from abroad. Due to this import numerous people came to Finland
and “soon Tampere had a small and a highly respected community of foreigners
who did not associate much with local people”. (Haapala 2005: 170.) The growing industrial activities in Tampere in the 19th century had many features in common with colonial economy (e.g. dominance of outsiders, cheap local labour, important foreign resources). It is obviously of historical and linguistic interest that
the entrepreneurs had an international background or education. (Haapala 2005:
171, 173.)
2.1 Finnish and Swedish
Despite the fact that most of the leading figures of industry, above all, had Swedish as their first language, Tampere was almost entirely a Finnish-speaking city.
Whereas in the 1950s Swedish was a symbol of belonging to a higher social
class, it no longer had the same characteristic in the 1980s. Earlier, knowing
Swedish was a precondition for upward social mobility, even for a Finnishspeaking resident. Nevertheless, it is worth mentioning that in a register of economic management published in the 1970s, there was no-one from Tampere who
gave Swedish as their first language. In practice, however, many people were
still bilingual (cf. language spoken at home and language spoken at work). The
Swedish-speakers had a dominant role in the industry of Tampere, for example,
in Finlayson cotton factory (est. 1820), in Tampereen Verkatehdas broadcloth
factory (est. 1856–1858), in Tako mechanical pulpwood factory (est. 1865), in
the Tampere Linen and Iron Industry Corporation Tampella (est. 1859) and in
Suomen Trikoo tricot factory (est. 1903), to mention some of the most important.
Even though the company names started to change from Finnish into English as
the 21st century approached, English-speaking company managers never actually came to Tampere. (Rasila 1992: 248–249.)
As mentioned above, the Finnish language has always been the language of
the majority in Tampere and that is why historians have considered Tampere a
clearly Finnish-speaking city, for example during the period 1870–1900, which
is under investigation here. During a period of 30 years, the number of Swe231
dish-speaking inhabitants in Tampere became smaller and smaller; in 1880 there
were 7.4 per cent Swedish-speakers, in 1890 approximately 6.9 per cent and in
1900 only 5.6 per cent. At the beginning of the 20th century the decrease continued. (Vähäpesola 2008.)
The speakers of the two major languages in Tampere, Finnish and Swedish,
lived mainly in different parts of the city, which has been seen as a sign of that
the areas were not only linguistically but also socially differentiated. With some
simplification one can say that the Swedish-speakers, who were mainly officials
and other individuals with high social status (such as clerical workers and
managers of the city’s many factories), lived in the central city area. In the areas
occupied by workers the language of the great majority was Finnish. (Vähäpesola 2008; on the Tampere dialect as the oldest form of Finnish, see Kuiri, forthcoming; on the spoken language of Tampere from the 1970s to today, see Mustanoja, forthcoming; different aspects on Swedish in Tampere can be found in
Lönnroth, forthcoming b; see also Lönnroth & Rossi 2008a, 2008b on the Swedish speech islands Tampere and Oulu.)
2.2 Other languages
During the period 1870–1900 there were only approximately 1 per cent of the
inhabitants of Tampere who spoke some other language as their first language
than Finnish or Swedish. It has been calculated that there were only 169 individuals who had another first language. Of these people the majority was either
German- or Russian-speaking. In 1890 there were only 222 people in Tampere
who were not Finns by origin. Of these people the number of Swedes was the
largest: there were almost twice as many Swedes as Russians in Tampere (today the case is the opposite as shown below). The Russians were the second
largest ethnic group in Tampere after the Swedes. (Vähäpesola 2008; on the
Russian language and Russians in Tampere, see Kirmo 2005; Mäkinen 2008c;
Suodenjoki, forthcoming; on German and Germans in Tampere, see von Witzleben, forthcoming.)8
As can be assumed from the discussion above concerning foreigners in Tampere, Finland, too, has for a long time had a relatively small number of foreigners
(cf. also Mäkinen 2008a). Compared to other major cities in Finland, Helsinki,
Turku and Viipuri, the number of foreigners living in Tampere has been small;
during the period 1870–1900 the number of foreigners in Viipuri was almost 21
per cent, in Helsinki 8.2 per cent and in Turku 4.5 per cent of the whole population of the cities. One of the Swedish speech islands in Finland, Pori, also had 1.9
On the notion of “mother tongue”, cf. Latomaa (2005: 91, with reference to McRae 1999: 83–84):
“Language statistics, collected since 1865, are not strictly comparable over time, since the definition
of the language to be recorded has varied. Up to 1930s, it was the language generally spoken; in 1940
it became ‘mother tongue’ which was understood as the language most fully mastered. The census
of 1950 asked for the language spoken best while those of 1960 and 1970 again asked for ‘main language’. Since 1980, every citizen has been classified by language according to the criterion ‘mother
per cent foreigners, which was more than in Tampere. In Tampere, most of the
foreigners were self-employed, like craftsmen. (Vähäpesola 2008.) The foreign
population in Tampere at the end of the 19th century was, as mentioned above,
small. However, when the economic life and international contacts started to
grow during the 19th century, Tampere began to receive inhabitants from abroad.
It is important to note that Finland was a part of Russia, a multicultural empire,
which meant that the Finns were also a minority nationality among others. (Mäkinen 2008a.)
Languages have often been linked with the industrial tradition of Tampere:
there were many foreigners in the industry of Tampere, for example during the
period 1870–1900. According to historical research, there were, actually, “quite
a few” of them. The industry of Tampere consisted both of domestic and international industrial entrepreneurs, and it was especially the bigger industrial concerns which were financed and built up by many people with foreign background. Many of the major foreign business men came to Tampere from Russia.
(Mäkinen 2008b.)
The multilingual and multicultural industrial tradition of Tampere is reflected,
for example, in literature. In his memoirs of 1902 Hermann Kauffmann, the
polyglot engineer, describes the English industrial experts during the latter half
of the 19th century. The examples (1)–(6) below illustrate the life of the English
community of Tampere as seen by a contemporary observer (the translations
from Swedish are mine):
(1) […] an Englishman from T:fors [Tammerfors = Tampere] Mr. Joseph Fielding, a remount master in the firm Finlayson & Co there. (p. 243)9
(2) […] to which he invited all his fellow countrymen with families employed at
the factories in the town, as many people as the boat held […] (p. 245)10
(3) […] because she was like most English wives in Tampere at that time very
precise and fond of housework. (p. 245)11
(4) One or another of the Englishmen at the factories amused himself sometimes
with […] trout fishing […] (p. 278)12
(5) […] one addressed some familiar Englishman at the factories who on occasion ordered the desired item from his native country. (pp. 288–289)13
In the original Swedish: “[…] en engelsman från T:fors Mr. Joseph Fielding, remontmästare hos
firman Finlayson & C:o därstädes”.
In the original Swedish: “[…] hvartill han inbjöd alla vid fabrikerna i staden anställda landsmän
jämte familjer, så mycket folk som rymdes i båten […]”.
In the original Swedish: “[…] ty hon var som de flesta engelska fruar i T:fors på den tiden mycket
noga och hushållsaktig”.
In the original Swedish: “En eller annan af engelsmännen vid fabrikerna roade sig ibland med […]
forellfiske […]”.
In the original Swedish: “[…] vände man sig till någon bekant engelsman vid fabrikerna, som vid
tillfälle skaffade det önskade från hemlandet”.
(6) Some of the Englishmen employed at the factories […] made picnic trips in
summer with families […] and in order to get welcome practice in the language I went willingly with them […] (p. 291)14
Some thirty years later, the manager Napoleon Wesander describes the Tampere
Linen and Iron Industry Corporation Tampella as a multilingual working community. His description concerns the foreign experts of the factory, not the Finnish-speaking workers:
During my walk round the factory I got acquainted with many of the masters in the
different sections: the spinning master Gustafsson, the weaving master Nell, an
Englishman, the bleaching master Miller, a German, the building contractor Asp
and the factory foreman Nyman, all nimble and capable people, each well versed
in their line.
Afterwards I looked at the mechanical pulp mill, the manager of which, the engineer W. H. O. Wallgren, a Sweden-Swede, welcomed me kindly and gladly and also
presented me to the factory foreman Karl Grönmark. (Wesander 1936: 46–47.)15
The factories in Tampere developed into a multicultural industrial milieu with,
for example, their own jargon between the workers (the old vocabulary of the
textile industry of Tampere, for example, is discussed by Vuosara 2003). In addition to the languages mentioned above, other, smaller languages have also been
spoken in Tampere. However, they are not discussed here in detail. (On “other
languages” in general, cf. Edwards 1994: 39, 186; on “other languages” in Tampere, see Latomaa, forthcoming.)
3 Conclusions
In this paper I have given a socio-historical overview of the sociolinguistic landscape, i.e. languages and language groups, of the city of Tampere, Finland, from
the founding of the city in 1779 to the beginning of the 21st century. However,
the main emphasis of the paper has been on the late 19th century.
In sociolinguistics or the sociology of language one usually draws a distinction between individual and societal multilingualism (e.g. Romaine 2000: 33).
This distinction is also relevant for the present study of the multilingual history
of the city of Tampere, formerly a well known industrial city in inland Finland.
In the original Swedish: “Några af de vid fabrikerna anställda engelsmännen […] gjorde under
sommaren med familjerna picknick-färder […] och för att ha en välkommen öfning i språket for jag
gärna med […]”.
In the original Swedish: “Vid min rond genom fabriken blev jag bekant med flera av mästarna i
de olika avdelningarna: spinnmästaren Gustafsson, vävmästaren Nell, engelsman, blekmästaren
Miller, tysk, byggmästaren Asp och verkmästaren Nyman, alla raska och duktiga personer, var och
en väl inne i sitt fack. Jag fick därefter bese träsliperiet, vars föreståndare, ingenjör W. H. O. Wallgren, rikssvensk, hälsade mig kamratligt glatt och även presenterade mig för verkmästaren Karl
In light of this study one can say that the image of multilingualism in Tampere
basically has to do with multilingual individuals, i.e. polyglots, rather than multilingual groups of people on a societal level. Finnish historians of Tampere from
Väinö Voionmaa onwards have emphasized that Tampere has always been a predominantly Finnish-speaking city (cf. for instance Helsinki, Turku and Viipuri).
However, many leading factory owners in Tampere had a multilingual family or
educational background. The most famous examples of these are James Finlayson (1771–1852), a Scot and the founder of the Finlayson factory, and the Nottbeck family, originally of Baltic-German origin. This image of Tampere as a
Finnish society with multilingual individuals is supported, for example, by the
memoirs by Hermann Kauffmann and later by Napoleon Wesander: people
working in industry and factories have been an important factor when we consider the multilingual and multicultural history of Tampere.
The connection between language and society is closely intertwined; so also
in Tampere. According to Haapala (2005: 178) traditional industry collapsed in
the late 1980s and early 1990s. From a linguistic point of view, however, this
meant a changed situation. Although the “old” languages that were connected
with the industrial tradition of the city declined, some “new” smaller languages
arrived in the city. During the period of globalization Tampere has actually many
new languages and language groups. At the end of 2007 the linguistic landscape
of Tampere had many languages, some of which are presented in Table 1
(Vainikainen 2008 whose numbers are from Statistics Finland). Finnish was spoken by 197,993 inhabitants, Swedish by 1,066 and other languages by 8,790.
These new languages are for their part creating the multilingual history of Tampere in the future.
Table 1. Languages spoken in Tampere, 31 December, 2007
Native speakers
Alhonen, Pentti et al. 1988. Tampereen historia 1. Vaiheet ennen 1840-lukua. Tampere:
Tampereen kaupunki.
Bergroth, Hugo. 1992. Finlandssvenska. Handledning till undvikande av provinsialismer
i tal och skrift. A facsimile edition of the 2nd, revised and added edn. from 1928.
Helsinki: Schildt.
Braunmüller, Kurt & Gisella Ferraresi. 2003. Introduction. In Kurt Braunmüller & Gisella
Ferraresi (eds.), Aspects of Multilingualism in European Language History, 1–13.
Amsterdam & Philadelphia: John Benjamins.
Edwards, John. 1994. Multilingualism. London & New York: Routledge.
Fasold, Ralph. 1984. The Sociolinguistics of Society. Oxford: Blackwell.
Fasold, Ralph. 1990. The Sociolinguistics of Language. Oxford: Blackwell.
Haapala, Pertti. 2005. History of Tampere: The Very Long Road to Informational City. In
Antti Kasvio & Ari-Veikko Anttiroiko (eds.), e-City. Analysing Efforts to Generate
Local Dynamism in the City of Tampere, 163–181. Tampere: Tampere University
Harmes, Nina. 1992. Svensk kontakt i Tammerfors 1926–1939. In Kristian von Essen &
Mika Huovinen (eds.), Satakunda från hertig Johan till Duke Ellington, 87–107. Helsinki: Historicus.
Hietala, Marjatta & Mervi Kaarninen. 2005. The Foundation of an Information City –
Education and Culture in the Development of Tampere. In Antti Kasvio & Ari-Veikko
Anttiroiko (eds.), e-City. Analysing Efforts to Generate Local Dynamism in the City of
Tampere, 183–215. Tampere: Tampere University Press.
Jutikkala, Eino. 1979. Tampereen historia 3. Vuodesta 1905 vuoteen 1945. Tampere:
Tampereen kaupunki.
Kauffmann, Hermann. 1902. Muntra minnen från mellersta Tavastland. Små berättelser
om ångbåtar, jakt och fiske. Helsinki.
Keskinen, Jouni et al. 2005. Tamperelaiset. Tehdaskaupungin väestö, alue ja asuminen
1918–1940. Tampere: Tampereen museot.
Kirmo, Outi. 2005. Tampereen venäläinen väestö vuosina 1918–1950. Etninen identiteetti ja sosiaaliset suhteet. In Antti Häkkinen et al. (eds.), Vieraat kulkijat – tutut talot.
Näkökulmia etnisyyden ja köyhyyden historiaan Suomessa, 342–366. Helsinki:
Suomalaisen Kirjallisuuden Seura.
Kuiri, Kaija. Forthcoming. Tampereen suomi vanhinta suomea. In Harry Lönnroth (ed.),
Tampere kieliyhteisönä. [Abstract: The Tampere Dialect is the Oldest Form of Finnish.]
Latomaa, Sirkku. 2005. Multilingualism in Finland: Past and present. In Xoán Paulo
Rodríguez-Yáñez et al. (eds.), Bilingualism and Education: From the Family to the
School, 91–106. München: LINCOM.
Latomaa, Sirkku. Forthcoming. Tataarista teluguun ja tigrinjaan. Tampereen monet kielet. In Harry Lönnroth (ed.), Tampere kieliyhteisönä. [Abstract: From Tatar to Telugu
and Tigrinya. The Many Languages of Tampere.]
Lönnroth, Harry (ed.). Forthcoming a. Tampere kieliyhteisönä. [Abstract: Tampere as a
Language Community.]
Lönnroth, Harry. Forthcoming b. Suomalainen teollisuuskaupunki, ruotsalainen
kielisaareke. Ruotsinkielisyys Tampereella 1700-luvulta 2000-luvulle. In Harry Lönnroth (ed.), Tampere kieliyhteisönä. [Abstract: A Finnish Industrial City, a Swedish
Speech Island. Swedishness in Tampere from the 18th Century to the 21st Century.]
Lönnroth, Harry & Paula Rossi. 2008a. Språkhistorien i stadshistorien. Språk i kontakt i
Tammerfors och Uleåborg. In Jan Lindström et al. (eds.), Svenskan i Finland 10, 197–
211. Helsinki: Institutionen för nordiska språk och nordisk litteratur, Helsingfors universitet.
Lönnroth, Harry & Paula Rossi. 2008b. Språken i föreningsspråk. Finländska fruntimmersföreningar som språkliga miljöer på 1800- och 1900-talen. In Marianne Nordman
et al. (eds.), Svenskans beskrivning 29, 187–194. Vaasa: Svensk-Österbottniska samfundet.
McRae, Kenneth D. 1999. Conflict and Compromise in Multilingual Societies. Finland.
2nd edn. Helsinki & Ontario: The Finnish Academy of Science and Letters & Wilfrid
Laurier University Press.
McRae, Kenneth D. 2007. Toward language equality: four democracies compared. International Journal of the Sociology of Language 187/188. 13–34.
Mustanoja, Liisa. Forthcoming. 1970-luvulta tähän päivään. Tampereen puhekielen
muutoksen tutkimus. In Harry Lönnroth (ed.), Tampere kieliyhteisönä. [Abstract:
From the 1970s to Today. The Study of the Spoken Language of Tampere.]
Mäkinen, Suvi. 2008a. Tampereen ulkomaalainen väestö 1870–1900.
koskivoimaa/kaupunki/1870-00/ulkom.htm (24 November, 2008.)
Mäkinen, Suvi. 2008b. Ulkomaalaiset Tampereen teollisuudessa 1870–1900. http:// (24 November, 2008.)
Mäkinen, Suvi. 2008c. Tampereelle Venäjältä muuttanut väki 1870–1900. http:// (24 November, 2008.)
Oxford English Dictionary, 2nd edn. 1989. Oxford: Oxford University Press.
Rasila, Viljo. 1984. Tampereen historia 2. 1840-luvulta vuoteen 1905. Tampere:
Tampereen kaupunki.
Rasila, Viljo. 1992. Tampereen historia 4. Vuodesta 1944 vuoteen 1990. Tampere:
Tampereen kaupunki.
Romaine, Suzanne. 2000. Language in Society. An Introduction to Sociolinguistics. 2nd
edn. Oxford: Oxford University Press.
Suodenjoki, Sami. Forthcoming. Kauppiasboikotista kansallisuusvihaan? Valtaväestön ja
venäläisten suhteet Tampereella vuosina 1899–1918. In Harry Lönnroth (ed.),
Tampere kieliyhteisönä. [Abstract: Interaction between the Russians and Finns in
Tampere 1899–1918.]
Tandefelt, Marika (ed.). 2002. Viborgs fyra språk under sju sekel. Helsinki: Schildt.
Tandefelt, Marika. 2003. Vyborg. Free trade in four languages. In Kurt Braunmüller &
Gisella Ferraresi (eds.), Aspects of Multilingualism in European Language History,
85–104. Amsterdam & Philadelphia: John Benjamins.
Trudgill, Peter. 2000. Sociolinguistics. An introduction to language and society. 4th edn.
London: Penguin.
Vainikainen, Janne. 2008. Ulkomaan kansalaiset 31.12.2007. (5 May, 2008.)
Wesander, Napoleon. 1936. Levnadsminnen. Helsinki: Söderström.
Witzleben, Brigitte von. Forthcoming. Saksalaisuus Tampereella. In Harry Lönnroth
(ed.), Tampere kieliyhteisönä. [Abstract: German and Germans in Tampere.]
Voionmaa, Väinö. 1929a. Tampereen kaupungin historia 1. Tampereen historia Ruotsin
aikana. 2nd, revised and added edn. Tampere: Tampereen kaupunki.
Voionmaa, Väinö. 1929b. Tampereen kaupungin historia 2. Tampereen historia Venäjän
vallan ensipuoliskon aikana. 2nd, revised and added edn. Tampere: Tampereen kaupunki.
Voionmaa, Väinö. 1932. Tampereen kaupungin historia 3. Tampereen historia itämaisesta sodasta suurlakon aikoihin. 2nd, revised and added edn. Tampere: Tampereen
Voionmaa, Väinö. 1935. Tampereen kaupungin historia 4. Tampereen uusin historia.
Tampere: Tampereen kaupunki.
Vuosara, Pekka. 2003. Kurakolmannesta kyyppikattilaan. Vanhaa tamperelaista kuto-
mateollisuuden sanastoa. Tampere: Suomen kieli ja yleinen kielitiede, Tampereen
Vähäpesola, Johanna. 2008. Äidinkieli, kansallisuus ja uskonto.
koskivoimaa/kaupunki/1870-00/kielijauskonto.htm (24 November, 2008.)
This article is a part of my postdoctoral researcher’s project “Linguistic Change in an
Industrial Town. Swedishness in Tampere, 1779–2000”, financed by the Academy of
Finland, 2008–2010. The site of research of the project is the School of Modern
Languages and Translation Studies, University of Tampere. I would like to thank
Virginia Mattila MA, University of Tampere, for checking my English.
Contact information
Harry Lönnroth
School of Modern Languages and Translation Studies
FI-33014 University of Tampere
[email protected]
Linguistic Assimilation and the Weakening of
Ethnic Identities in Dagestan
Magomedkhan Magomedkhanov
Institute of History, Archeology and Ethnography of
Dagestan Scientific Center of Russian Academy of Sciences
wise man remembering the willingness of the great empires to conquer the
Caucasus once said: “The tragedy of the Caucasus lies in its geographic position”. The other truth is that Caucasus is located at one of the most active and
lively crossroads of Eurasia. Caucasian civilization with its substrata of cultural
components of Turkic, Iranian, Jewish, Slavic, Greek worlds absorbed life experience of settled and nomadic tribes in material and military culture, and the in
economic development of lowland and steppe. It contains spiritual values and
technological achievements of Europe and Asia, of Orient and West, of Islam,
Christianity and Judaism. The historical and cultural unity of the Caucasus very
often became victim of global geopolitics and local conf1icts. Today, unfortunately, it is the reality that “the Caucasus lost conciseness of unity and solidarity”1.
Dagestan for centuries had been an organic and culturally productive part of
Eurasia. Selective innovation of achievements of other civilizations protected
Dagestan both from cultural backwardness and cultural assimilation. Political
and juridical independence of self-governing civil communities (jamaats) was
the main reason for the preservation and survival of the linguistic and ethnic plurality of Dagestan.
Since the second half of the 19th century, Dagestan passed through certain
stage of cultural transformation, determined by the fact of being part of Russian
Empire and Soviet Union.
It becomes obvious that the Soviet concept “Vzaimoobogasheniye yazikov i
kultur” (Mutual enrichment of languages and cultures) does not apply to the period since the 1960s and particularly in the contemporary ethnic situation in Russia. Functional stagnation of ethnic cultures and linguistic assimilation was obvious even without special investigations, but no one could talk about it openly.
Bgazhnokov $. \h. Osnovaniya gumanisticheskoi etnologii. ^., 2003, `. 86.
Except for the Census of 1926, all Soviet Censuses (1939, 1959, 1970, 1979,
1989) contain false information about non- Russian languages when it comes to
identifying the number of natives their ability to speak the native languages.
It used to be said, “Dagestanians are one people with many languages”. I
would add: “And without future’ possibility to preserve them.”
Out of slightly more than the two million population in Dagestan, two thirds
live in ethnically mixed lowlands and towns and about one million Dagestanis
live outside the Dagestan Republic. Such territorial distribution of the population plus the growing percentage of ethnically mixed families (in Dagestan every
tenth family is mixed) the array of functioning of Dagestan languages is steadily
being reduced.
In Dagestan (in towns and settlements with mixed communities) elementary
schooling is in Russian and, in theory, in eleven Dagestanian languages. In a
number of rural schools and departments of philological faculties Dagestanian
languages and literature are offered as a subject· of instruction.
Recent attempts to prepare primers in languages, which did not have writing
and introduce some of them (for example, the Andi, Dido, Tsez) languages) into
school as languages of instruction at the primary level turned into political
action rather than cultural enhancement.
The point is that these initiatives take the form of political actions. It should
also be noted that in the past the speakers of the so-called «non-written» languages used to have enough intellectual potential to transliterate their mother
tongues on the basis of the Arabic graphics or Russian alphabet.
Thus, Professor ^.. Aglarov discovered manuscripts of 19th century written
in the Andiysky language. Recently I have found the religious and instructive
treatises, prose and poetry (344 lines), written in Archi at the beginning of the
19th century. Such manuscripts, written in other «oral» languages may be further
discovered in future. However, these works did not lead to written tradition in
those languages. The reason is in the practicality of such written forms as the
speakers of the so-called «non-written» (oral) languages used to turn to the writing systems of larger nations of Dagestan in their religious, ceremonial, cultural
and creative spheres, to such languages as Avar (for the Archi and Ando-Tsezian
peoples their fluency and efficient use of the Avar language (Bolmats) and Avar
writing system is an objective basis for their cultural identity with the Avars.
Lezginian used to play the same role in southern Dagestan, having at the same
time some particular features as here traditionally Turkic language (Azerbaidjanian) and Turkic orthography had dominant position in the cultural life. The
fact that the number of treatises in Lezghinian written by the Tabasaran, Rutul,
Tsahur and Agul is negligible is indicative of the situation.
As far as initiatives to create orthographies of «oral languages» are concerned,
it may be noted that it is not the issue itself or methods of its solution that are of
urgency but possible practical outcomes of the activities along these lines. In this
respect it is worthy to note the efforts of recent years to create the written form
of the Agul language and introduce it into schools and mass-media. Thus, since
1992 the instruction in Agul schools has been conducted in the mother tongue.
The language is taught in the teacher-training secondary college in Derbent and
Dagestan Pedagogical Institute. Now that the Agul have their own orthography
they have more opportunities for their cultural and spiritual self-expression. But
these opportunities, however, are limited by the area where Agul is functioning
and the number of its native speakers (about 20000 people, of which approximately 14000 live in Dagestan and a third of these live in the towns of the Republic). Besides, as far as their elite with their constitutional status are concerned
they may be more interested in their self-expression in the circles of the local and
national establishment than in their ethnic and linguistic identity at the regional
ethnic level. Therefore the experience with creating Agul orthography should not
be transferred mechanically to Ando-Tsezian languages. Besides, it may also be
useful to study failures in experimenting with Tsahur (1932) and Akhvakh
(1936) writing systems.
From an anthropological point of view every language and every culture is
unique. In constitutions of almost all countries one can find declarations of
equality of people regardless of language, race, religious differences. Majorityminority of ethnic populations is one of the natures of human being. It means that
there is hierarchy of functional ability of certain languages and cultural influence
of certain peoples. In this context a popular statement that any language including one without a written form can be developed to the highest social communicative and social cultural levels should not be taken at its face value. For example
it is hard to image equality of Gihukh and Lezgin languages spoken by 500 and
500,000 people.
Sociolinguistic situations do not always comply with hypothetical virtual
models of management of ethnic-linguistic processes. Efforts to revive languages are short-lived if their social contexts and particular ethnic-cultural needs
of their respective speakers are not taken into consideration.
It seems evident: the smaller a social or ethnic community is, the simpler are
mechanisms and strategies of manipulating with the mentality of people. One of
the indications of this way of thinking is the «noise» around the alleged poor
state of the Tsez (Dido) language as expressed by the activists for Tsezian autonomy in the newspaper «Didoyskiye Vesti». As the experience of bankrupt “national movements”, and “fronts” shows, their pronouncements about the imminent degradation and loss of the ethnic-cultural identity accompanied by slogans
to protect languages (at the early stage of «fronts» and «movements») are very
often motivated not so much by their care for the revival of their ethnic culture
but by their narrow political aims. Gradually with «co-optation» of the leaders of
«ethnic movements» into power structures, the social political spirit of such pronouncements has a tendency to fade out.
The cultural identity of the 400 Ghinuh, 1500 Archi, 3000 Gunzib and the
same number of the Khvarshin, Botlikh, Godoberi peoples can not be described
only in terms of their own ethnicity or ethnic linguistic community. Of equal importance is the issue of reasonable concentration of the intellectual potentialities
of these ethnic groups, their scarce material and demographic resources on solving the problems of physical and spiritual survival without wasting them on endless discussions of the questions of bilingualism which has been there for centuries.
The law of the State Duma of March 5, 1999 «On Guarantees of the Rights of
Indigenous Minorities of the Russian Federation» signed by the President of
Russia on April 30 of the same year is not only illogical but unnatural and can be
regarded as «an ethnographic anecdote». Article 1 of this law says, «taking into
consideration the unique ethnic composition of the population of the Republic of
Dagestan in terms of the number of nations living on its territory, the State Counsel (Gossovet) of the Republic of Dagestan identifies quantitative and other specific features of its indigenous minorities, as well as makes a list of these nations
to add to the Unified Inventory of Indigenous Minorities of the Russian Federation». But is it the business of an authority to identify quantitative and other specific features of indigenous minorities, as well as include them into a certain list?
It is important to point out that the cultural history of Dagestan is characterized
not only by permanent processes of overcoming linguistic barriers and borrowing foreign innovations, but also by contributions of cultural and intellectual elite
of Dagestan to the Eastern and, since the 19th century, Russian and European
culture which are comparable with the process of enrichment of their culture
with the achievements of Eastern and European civilizations.
The social functions and spheres of usage of native languages with firmly established written traditions are steadily growing smaller and tend to be used only
within the family circle and rural areas. This can be illustrated by the example
of the Avar culture serving not only the Avar themselves but also ethnic communities close to them, having no language barriers for centuries and sparing a common, Avar cultural background.
It is doubtful that disregard of centuries-old experience of integration with the
Avar cultural environment providing opportunities for preserving their ethnic
identity and language and a tendency for the isolation and separatism will make
the life of minorities any spiritually and culturally richer.
Many scholars believe that Dagestani languages are endangered and the processes of linguistic assimilation affects almost third of the population. In
Dagestan assimilation does not take such complex forms as observed for example among Mordvinians where there are two opposite directions of assimilation: Russification of the Ersa Mordvinians and Tatarization of the Moksha
Mordvinians accomplished by Russification of Mokshas and Tatars themselves.
But turning ethno-linguistic orientation into direction of assimilation is very
harmful for the Republic.
From the sociological point of view modern towns and a number of settlements around them in Dagestan are in fact largely Russian speaking. The importance of the Russian language for citizens of Russia can not be questioned. In
Dagestan for a century and a half period Russian has been steadily and rapidly
extending its cultural and social functions. For the last decades it has strength242
ened its sociolinguistic positions going beyond the role of a lingua franca in its
importance. For the majority of urban population of Dagestan Russian is the only
means of communication. It is the main language of culture and research for
Dagestanians, like for the rest of the population of Russia.
There is an urgent problem with mother tongue instruction in Dagestanian (!)
schools which so far has not been solved as Russian is required as the language
of instruction, foreign language instruction is a required subject in accordance
with the curriculum while the mother tongue is not required even as a subject of·
instruction «with parents’ agreement». As a result newspapers in minority languages and writers writing in those languages are losing their readers, radio stations and theatres have fewer listeners and fans.
Dagestan is not the only multilingual region in the world and the study of the
world experience of solving ethnic linguistic and cultural problems is quite urgent as it is sure to promote a more responsible decision-making in the sphere of
language politics. It should be added here that well-intentioned but inefficient
programs, such as «Program for Preserving and Promoting Culture and Arts in
the Russian Federation», “Program for Preserving and Promoting the National
Culture of Peoples of Dagestan for the Years of 1996–2000”, «Program for Enriching National Culture with New Forms and Genres and Promoting the Artistic
Elite of the Republic», etc. were made up without experts’ review and due attention to particular cultural needs of the Dagestan peoples.
Measures taken by the state to revive languages and consequently their ethnic
and cultural traditions and moral values may be effectively provided if there is
an understanding of the society that these measures are the urgent concern of the
whole Daghestani community.
The use of one’s mother tongue is fundamental for preservii1g one’s ethnic
and cultural identity simply because many components of the ethnic culture exist
due to language forms (such as folklore, traditional popular songs, literature,
theatre, etc.). In this respect it should be noted that the erosion of features and
characteristic of ethnic, cultural identity, accompanying the processes of linguistic assimilation, lead to shift in the ethnic moral values and become subject
not only of scientific discussion but theme for discussion among family and
friends, factor of the ethnic mentality as well.
The disappearance of features of cultural identity and first of all ethnic language does not automatically lead to loss of ethnicity. But linguistic assimilation
is one of the obvious signs of future of assimilation and ethnic disappearance.
The process of linguistic assimilation of the Dagestani community of nationalities is very active and can be observed everywhere where Dagestanis live. Indifference, pessimism about the future of the ethnic and cultural development and
disregard of mother tongues, observed in Dagestan, are in fact widespread elsewhere.
The analysis of contemporary tendencies of the ethnic and cultural preferences
of the people, especially of young urban Daghestanis, shows that ethnic cultural
heritage and the artistic personalities representing it are more symbolic value
than an objective basis for ethnic and cultural identity. Thus, the character of
“ethnicity” with specifically ethnic qualities and characteristics of an individual
and the ability to speak his/her mother tongue among them, has evolved into
something largely superficial or symbolic.
Magomedkhan Magomedkhanov
Institute of History, Archeology and Ethnography,
Dagestan Scientific Center of Russian Academy of Sciences
ul. Yaragskogo 75-a,
Russian Federation
[email protected]
Medieval Mediterranean as a multilingual area:
the Greek perspective*
Theodore Markopoulos
Uppsala University
0. Introduction
This study focuses on the eastern Mediterranean in the late medieval period (ca.
12th–15th centuries) as a multiethnic, multilingual area. The impressive variety of
people inhabiting this region resulted to the creation of a language contact area,
where many languages could be heard and spoken: Romance languages, prominent among them being various Italian dialects (especially Venetian and Genoese), French and Catalan, but also Greek, Arabic, Turkish and Slavic, among
others. This multilingualism has obviously attracted scholarly interest, but mostly in relation either to the famous lingua franca (cf. Minervini, 1996, among
many others) or to specific phenomena observed for the most part in Romance
languages (cf. e.g. Ramat & Roma, 2007).
On the contrary, this paper attempts to pay the attention due to the Greek aspects of this multilingualism. In the 13th century, especially after 1204, large
parts of the Greek-speaking eastern Mediterranean, formerly belonging to the
Byzantine empire, passed into the hands of Western rulers, as a result of the
fourth crusade. Consequently, situations of language contact were immediately
created, involving mainly Greek and Romance (Italian and French, for the most
part) speakers.
So, the aims of the study are twofold:
a) To elucidate the sociolinguistic aspects of this language contact situation,
through an attempt to give some answers to the basic – but still not dealt with
– question: who was talking to whom and under what circumstances? and,
* This study is part of a Marie-Curie individual research project, named “Language contact in the
eastern Mediterranean in the Late Middle Ages: Greek in a multilingual society”. I would like to
thank all the participants in the 23rd Scandinavian Conference in Linguistics, for their valuable comments.
b) To tie the sociolinguistic investigation with the examination of specific morphosyntactic phenomena of Late Medieval Greek, arguably related to language contact.
Most studies undertaken so far address almost exclusively the lexical borrowing
between Greek and the other languages, primarily the borrowing of Romance
lexical items into Greek (e.g. Triantafyllidis, 1909, Kahane & Kahane, 1982, Stanitsas, 1984, among others), while studies for the opposite direction of borrowing are rare, but still extant (cf. e.g. Cortelazzo, 1970). However, recently new
investigations have come to light, attempting either to describe the sociolinguistic situations in specific areas (cf. Terkourafi, 2005 for Cyprus) or to deal with
specific morphosyntactic phenomena and to relate them with language contact
accounts (cf. Manolessou, 2008). This investigation aims to develop further this
trend of research.
The scarcity of relevant studies is partly due to intrinsic difficulties facing all
researchers dealing with Late Medieval Greek. In particular, the textual sources
available are rather limited and, to make matters worse, the majority of them
have been published in compliance with old philological practices (involving
corrections, ‘mixing’ of the variants found in the manuscripts etc.), rendering
difficult any kind of linguistic research.1 Despite all these problems, linguistic investigations into Late Medieval Greek are very much possible, especially in
cases where more sociolinguistic background is available, regarding not only the
text themselves and the circumstances of their production, but also the social surroundings. Therefore, the present study will investigate two specific sociolinguistic cases related to a great extent to language contact, namely Venetian-ruled
Crete and Frankish-ruled Cyprus, in relation to particular morphosyntactic phenomena.
1. Venetian Crete and the future-referring periphrastic
1.1 The sociolinguistic aspects
Crete became a Venetian territory on 1211, and remained so until 1669, i.e. more
than five centuries. It is known that a Venetian population was established on the
island from the very beginning, although its exact numbers are elusive. It is
rather uncontroversial, though, that their number in comparison to the local,
Greek-speaking population must have been relatively small. Originally, the
Venetians settled only in the major towns of the island, and predominantly in
Chandakas (Candia), the capital of Crete. As a consequence, they lived mostly
For a detailed discussion of these issues, cf. Markopoulos, 2007, 2008, Manolessou, in press,
among others.
apart from the majority of the Greek agricultural population of the island, who
dwelt in small villages in the countryside.
It is therefore obvious that the mere presence of Venetians on the island does
not immediately imply a language contact situation, if the two populations with
distinct native languages stayed apart from each other. Nonetheless, there is
strong evidence that a language contact situation did obtain on Crete, if not from
the very beginning of the Venetian rule, shortly afterwards. To be more precise,
Venetian notarial sources testify that there have been mixed marriages between
the two communities already by the end of the 13th century (Maltezou, 1997).
Apart from marriages, there was the prominent issue of the illegitimate children,
usually involving an Italian father and a Greek mother. Generally, contrary to
earlier ideologically prejudiced accounts that highlighted the separation and the
animosity of the two populations, new research and analyses have shown that the
approach of the two communities on the social level was very much in force in
the 14th century (cf. McKee, 2000). There is an illustrative example on that respect: we know of an Italian notary in Chandakas in the early 14th century (1303–
4), named Stefano Bon, who most probably had some knowledge of Greek, and,
perhaps, more importantly, had a Greek wife and a daughter named Pothiti, undoubtedly a Greek name.
This evidence suggests quite strongly that there had been interaction between
the two communities from a very early stage of the Venetian rule. The phenomenon must have been more common in the urban areas, especially in Chandakas, where the majority of the Italian population resided. It is worth noting that
the bilingualism was not only present in social life, i.e. in trade activities, but in
family life as well, either through mixed marriages or through the presence of
Greek servants in Italian households. On the basis of this evidence, one could expect to find language contact phenomena in Greek from the 14th century, especially in texts produced in the urban centres (which was in any case the most likely scenario for all texts of this period, with the exception of notarial acts produced in the countryside).
The extent of bilingualism among the two communities is very difficult to determine. One can easily assume that, initially, it must have been the Greeks who
learned the Venetian dialect, since the Venetians represented the upper class and
had less motivation to learn the language of the lower classes. This is undoubtedly true, but it is only half of the story: the other half is the attested growing
knowledge of Greek among the Venetians. This is manifested in various ways,
the most important being the fact that almost all literature works produced in the
Venetian-ruled Crete were written by authors of Italian origin, as the names Falieros (15th c.), Dellaporta (15th c.) or Kornaros (16th c.) suggest. This is highly
significant, since it proves that by the 15th century some Venetians not only knew
Greek, but were also willing to and capable of writing poetic works in this language! It is not an exaggeration to assume that they constituted a group – whose
extent is unknown – of true, full bilinguals.
Apparently, Italians not only learned Greek but, by the late 16th century, most
of them must have shifted to Greek as their native language. This is directly reported by travelers passing or visiting the island (such as Schürpf on 1498 and
Sendlitz on 1556), who declared that the language spoken on the island was
Greek (Maltezou, 1997: 41–3). This process of language shift from Italian into
Greek must have been concluded in the 17th century, when the notarial acts were
almost exclusively written in Greek.
In other words, the sociolinguistic situation in Venetian Crete was characterized by an increasing and extensive bilingualism, at least as far as the urban
centres were concerned: On the one hand, Greek learned Venetian as their second language but, on the other, the Venetians not only learned Greek, but eventually shifted into Greek, which thus became their L1.
The results of this language contact situation have been studied, as already
mentioned, solely regarding the lexical borrowings both in Greek and in Venetian. However, almost nothing is known with regard to the other linguistic levels,
especially morphology and syntax. The next section illustrates a possible instance of contact-induced morphosyntactic change in Greek through the influence of Venetian.
1.2 The future-referring ‘the na’: a contact-induced grammaticali
The development of the Greek future-referring form ‘the na + Subjunctive’, originating from a ‘periphrastic form’ based on the verb ‘thelo = want’, constitutes
one of the most famous examples of the phenomenon of grammaticalization in
the relevant literature (cf. e.g. Bybee et al., 1994, Heine, 2003, among many
others). According to the traditional account, the Medieval Greek future-referring ‘periphrastic form’ thelo+Infinitive developed into an impersonal construction thelei na + Subjunctive, after the replacement of the Infinitive by a subordinate clause, and then, due to its grammatical status and its frequency of use, phonological attrition ensued, leading to the ‘the na’ construction. Moreover, this development of medieval Greek is also traditionally linked with the notorious
Balkan Sprachbund, i.e. the phenomenon of linguistic convergence observed in
the Balkan languages with regard to a variety of lexical items and morphosyntactic constructions (cf. Joseph, 1983, Dixon & Aikhenvald, 2007, among others).
However, in a recent investigation (Markopoulos, 2008) I argued that both assumptions are basically wrong, and cannot account for the developments observed in the data, at least in a satisfactory manner. In other words, the future-referring construction did not develop as smoothly as the ‘standard’ grammaticalization literature has led us to believe and, more significantly for our purposes
here, there is evidence that the whole development was partly influenced by language contact with the Italian varieties spoken in Crete. In what follows, a brief
account based on Markopoulos, 2008 will be presented, together with some new
evidence regarding the spread of the construction.
Two main accounts of the development of this construction exist in the literature:
a) According to Horrocks (1997), the construction is the result of the strengthening of the future-referring form ‘na + Subjunctive’, quite common in Medieval
Greek, through the use of the, an abbreviated form of thelei, which was used
in the formation of the thelo future-referring construction.
b) Joseph & Pappas (2002) argue that the ‘the na’ construction is the ultimate
outcome of the thelo future-referring construction, through the following
‘thelo + Infinitive’ ‘thelo na + Subjunctive’ ‘thelei (impersonal) na +
Subjunctive’ ‘the na + Subjunctive’
It is crucial to note that, according to Joseph & Pappas, in all these stages the
construction retains its future reference.
Various problems of these two accounts have been noted (cf. Markopoulos,
2008). Most importantly, neither account can explain examples such as the following:
(1) |
‚‘‚’ˆ“, “‹
Other-GEN destruction wants,
other-ACC want-3rd SING.PRES. that
become-poor-3rd PRES. SUBJ.
“For one he wants destruction, for another he wants him to become poor”
(Sachlikis, Ms. P, f. 146v, ll.4–5)
This example, taken from the poems of Sachlikis, a 14th-century poet from Crete,
illustrates that the ‘the na’ construction has a clear volitional meaning, in accordance with the lexical meaning of the, and not future-referring, as the traditional
accounts would like us to believe. This is not an isolated case, but part of a systematic pattern, as shown in the examples below:
(2) •™š ’‚
‹‹’ Š›œ‚™ “ŸˆŠ•Š,
Š™‹‰ ˆŠ•Š ›Š™ ¡ š
who want-? that listen-3rd SUBJ. songs, riddles
and stories
›Š™ ˆ‹–¢‰Š‹ •†Š –¢‰–‹, ‚™š •‡‹ £Š†Ÿ‰Š‹ Šš ‘“ ‚™!
and hymn
to the Tarsian
prt go-3rd SUBJ.
“Whoever wants to hear songs, riddles and stories and tragedy songs, let him go to Tarsos!”
(AoT, 614–5)
(3) ˆ“ ¡Ÿ
‹“ ˜¤‹‚ŸŠ™
›“•‚˜‚ ›™
but asmuch want-? that hide-2nd SING.SUBJ, know
and were-seen
“but as much as you want to hide, know that you have been seen”
(Falieros, Istoria kai Oneiro, 576)
(4) ¦”‹‚•‚ ¤•™ ’”
¦“ ™
‹“ ¦‚œ ™
Š‘¡ •¡ ¢‰›Š™–‹ ›Š‰ ‹“
that want-3 PRES. that leave-3 SUBJ. from the right
•¡ ‚¢™›¡ ˆ ˆ”
eat-3rd SUBJ. the my-mine
and that
‘‹‡†‰‚Ÿ ˆ‚ “‚š
with deceits
“it seems that he wants to forswear justice and to steal my wealth by deceit”
Sourianos, 126/ d. 1582)
Two main observations can be made on the basis of these examples: First, that
the na could be used with a volitional meaning even as late as 1582, in a text constituting the proceedings from the court, thereby rendering rather accurately the
spoken language of that time (cf. ex. 4). This fact, together with various other details (cf. Markopoulos, 2008: 186–206), renders inadequate the existing accounts
of the development of this future-referring form. Second, the construction could
be used with a volitional meaning in various grammatical persons (e.g. in the 2nd
person singular, ex. 3), not only for the 3rd person singular, the alleged source of
the phonologically ‘impoverished’ the. Given that, cross-linguistically, volitional verbs are quite rarely impersonal (cf. Haspelmath et al, 2005), since their
lexical meaning is directly linked with the intention of an agent, this pattern is
rather hard to explain.
A possible factor to which we can appeal is language contact between Greek
and Venetian. In this respect, it is highly relevant the fact that Old Venetian exhibited a form of the verb ‘want’, namely vol, which could be used for both the
3rd person singular and the 3rd person plural, already in the end of the 12th century,
i.e. before the appearance of the ‘the na’ construction in Greek, while the modern
dialects of Veneto have generalized this pattern to the 2nd person as well (Marcato & Ursini, 1998: 366). It seems quite plausible to assume that this was a pattern
that facilitated the spread of volitional the na to other grammatical persons apart
from 3rd person singular, a fact that ultimately led to its grammaticalization as a
future-referring construction (for more details, cf. Markopoulos, 2008).
This language contact account is based firmly on the language contact situation of Venetian Crete, as illustrated in the previous section. As Venetians
learned Greek and even shifted to Greek, they could well use a pattern found in
their own dialect in their version of Greek, altering thus the path of development
of the ‘the na’ construction. In that line, Thomason mentions (2001:80) that in
cases of ‘shift interference’, it is the phonology and syntax that are usually influenced first, which complies well with the scenario on this phenomenon.
Further corroborative evidence is found both in the areal and the sociolinguistic distribution of the ‘the na’ construction. Regarding the former, one cannot
help but note that this construction was initially attested in Crete and, subsequently, in other Venetian-rules areas (e.g. the Ionian islands), but also in Cyprus. This is clearly in favour of the account proposed here, the only complication being Cyprus: How can Cyprus fit into this Venetian-based account? This
can be easily explained, if one considers the following:
a) Trade between Crete and Cyprus was quite extensive in this period, especially
in the 14th century, when both Cretans and Cypriots traveled back and forth (cf.
Coureas, 2005: 150). In such circumstances, linguistic innovations could easily
spread from the one island to the other and, moreover,
b) Venetians had a strong presence on Cyprus, not only among the westerners
and the royal Frankish court, but also in the rural areas, as there existed numerous
settlements of Venetians already from the 12th century, before the arrival of the
Franks on the island (cf. Baglioni, 2006: 45–6).
Under such circumstances, it is not difficult to imagine common linguistic developments between Cretan and Cypriot Greek, even if they were partly influenced
by the Venetian presence in both islands.
The sociolinguistic distribution corroborates this language contact account as
well. It is relevant to note here that in the instances of the construction in the 16th
century notary books from the whole Greek-speaking world, 15 out of 16 are
found in books written in Chandakas, the capital of Crete, where the denser
Venetian population was established, while the final occurrence comes from the
capital of Kefalonia, another Venetian-ruled area. And in any case, innovations
tend to spread quicker in towns than in villages (cf. e.g. Milroy, 1993: 228–9),
and therefore, language contact in urban areas, such as observed in Venetian
Crete, is much more likely to cause language change.
Therefore, evidence of various kinds (both linguistic and sociolinguistic) suggests that the development of the future-referring ‘the na’ construction is partly
due to language contact. This instance of linguistic change could have occurred
without contact between Greeks and Venetians, but this cannot be taken to argue
against a language contact account. On the contrary, it complies well with Heine
& Kuteva’s (2008) definition of contact-induced change, which states, contra
Thomason (cf. e.g. 2008), that language contact should not be considered a
causative factor only for changes less likely to occur outside a contact situation:
it can easily trigger a frequent instance of language change instead, which should
still be called contact-induced.
2. Cyprus under the Lusignan: multilingualism through
prepositional use
2.1 The sociolinguistic perspective
The second case study of the language contact situation in the medieval eastern
Mediterranean is Cyprus. As is well-known, Cyprus came under the rule of the
Frankish family of Lusignan in 1192, and remained so for another three centuries
up to 1489, when the Venetians managed to gain control of the island for another
century, before it finally passed into the hands of the Ottoman empire. This paper
is mostly focused on the Lusignan period, when the stage for the extensive multilingualism observed in Cyprus was set and developed.
The Franks constituted the main Western population residing on the island,
followed closely by the Italians, especially Venetians and Genoese. As usual, it
is quite challenging to come up with numbers for the medieval populations, but
it is estimated that the Western part comprised no more than 20% of the total
population at any time during this period (cf. Nicolaou-Konnari, 2005: 15). Obviously, the Greek-speaking Cypriots constituted the majority, but there were
also various other ethnic groups on Cyprus, in particular Syrians (Muslims and
Christians) and Armenians. So, it is not an exaggeration to say that Cyprus under
the Frankish rule was a real multiethnic, multilingual society.
The multilingualism involving Greek and Romance speakers must have been
evident predominantly in the urban areas of the island, like in Crete, i.e. in Nicosia and Famagusta, the former being the main residence area of the Frankish
aristocracy, and the latter of the Genoese, who possessed walled quarters in this
port. But walled quarters did not exist in Nicosia, were a true multilingual society
slowly emerged. There is evidence of this ‘mixing’ of populations, involving
both Franks and Greeks: it is telling that Cypriot Greek was used by the Lusignan
rulers in their correspondence with various Turkish rulers of Asia Minor, already
in the early 13th century, in other words only a few decades after the Frankish
settlement on the island. This strongly implies the presence of Cypriot Greeks inside the royal court of Nicosia from the very beginning, a fact corroborated by
evidence that various Greeks worked as bureaucrats in the Frankish administration in the 14th century at the latest (cf. e.g. Edbury, 2005: 76).
It is therefore obvious that a number of Cypriot Greeks learned French as a
second language.2 But there is also sizeable evidence that Franks also came to
use Greek in their every-day language, a trend that resulted in the overall shift of
the Franks to Greek by the late 15th century (cf. Edbury, 2005: 64). It is noteworthy that the last Frankish rulers of Cyprus had Greek as their native language,
while there are various hints suggesting that the Frankish aristocracy was not a
stranger to Greek already in the 14th century (cf. Nicolaou-Konnari, 1995). This
came about through a steady rise of various Cypriot Greeks in the social stratification, since even the ranks of aristocracy were opened to the local powerful elite
in the mid-14th century (cf. de Collenberg, 1982). An aspect of this upward social
movement of the Cypriot Greeks is the translation of a legislative text for the
members of the class of burgesses, the Assises de la Cour des Bourgeois, from
French into Cypriot Greek, probably in the early 14th century.
Finally, it should be added that the presence of Italians, especially of Venetians, was intensely and increasingly felt on the island, as can be seen in the ‘creation’ of a Romance variety that, according to Baglioni (2006), is most probably
a mixture of Venetian and French. This represents perhaps an attempt by the part
of Greek speakers to write French or at least a Romance language, as it is not at
all clear that the Cypriot Greeks could distinguish between French and Italian at
that time. The important role of Venetian in the multilingualism of Cyprus fits
quite neatly with the account for the ‘the na’ construction, as described above.
So, Cyprus constitutes a similar situation to the one on Crete, in terms of the
results, as the ruling class (this time, the Franks) shifted into Greek in the course
of the centuries. On the other hand, the Cypriot Greek variety exhibited a massive influence from French and Italian in its lexicon, which has been the centre
of attention of researchers (cf. e.g. Stanitsas, 1984, Giagkoullis, 1993). Again,
possible contact-induced changes on the morphosyntactic level have not attract2
Note also the much celebrated assertion by Machairas, a 15th century Greek Cypriot chronicler, according to which the Cypriots began to learn French and thus their language became so mixed that
no-one could tell what it was (Machairas, 158).
ed the interest of linguists, with a recent exception (Sitaridou & Terkourafi,
2007), which deals with the famous issue of the loss of the masculine genitive
plural case form of Cypriot Greek. Starting off from this construction, the next
section provides some preliminary observations and comments on the wider issue of language contact in medieval Cyprus, as attested in the morphosyntax of
Cypriot Greek.
2.2 Case and prepositions
In their recent paper, Sitaridou & Terkourafi (2007) have argued that the loss of
the genitive masculine plural and its replacement by the accusative in Cypriot
Greek, already evident in the medieval texts (cf. 5), is partly due to contact between medieval Greek and French, echoing in a sense a previous account put forward by Papadopoullos (1983).
(5) ’‚–†¤‹•Š •‡‹ •“¥™‹ •š
‘“ •™ˆ‡ˆ”‹‡‹
considering the order the-ACC.PL. Templars-ACC.PL. much
“considering the order of the Templars as very honourable”
(Machairas, 13)
I have argued elsewhere that this phenomenon cannot be attributed to the language contact situation, because it occurs even in texts of the 13th century which
do not seem to have been influenced enough by French otherwise, and more significantly, because it is also found in the early Byzantine period, in inscriptions
from Syria, Palestine and even South Russia.3 However, there is evidence that
the language contact situation must have facilitated the diffusion of the construction, especially in the context of prepositional complement:
(6) ‚‹¤‘™‹
›Š™ •š
in front of the-GEN.SING. viscount-GEN. and the-ACC.PL. judges-ACC.
(devant le vesconte et devant les jurés, French Assises, 52)
“in front of the viscount and the judges”
(Assises B, 51)
As is elsewhere explained (cf. Markopoulos, forth.), in this context the replacement of the old genitive by the accusative had lagged behind, and it is probably
examples such as (6) which illustrate how contact with French might have spread
the construction in prepositional contexts as well: French had always an accusative form after a preposition, a fact easily generalized in Cypriot Greek, which
had a number of prepositions still followed by a genitive complement (as
shown in 6), an older pattern that was in any case slowly giving ground
to accusative complements.
On the other hand, evidence for contact-induced change in Cypriot Greek on
the basis of French is found elsewhere in the Assises, a text that facilitates such
conclusions as both the French version and its Greek translation have survived.
For a full account of this phenomenon, cf. Markopoulos, forth.
Consider the following example:
(7) ‘Š†‚•œš ¡•™ • ”‹›‡ˆŠ‹ ‹Š ”‹‡ •š
that the claim
prt made the-ACC.PL. creditors-ACC.PL
‚™š •‡‹ Š¨‹
to the court
(si tost come le clain en sera des acreors en la cort, French Assises 195)
“immediately after the claim is made at the court by the creditors”
(Assises B, 184)
In this case, the Cypriot Greek translator used an accusative form (
) in an attempt to render in Greek the French phrase des acreors.
Such French forms usually constitute a genitive form, thus its translation with the
accusative form in Cypriot, according to what has been said above. Nevertheless,
in (7) the French phrase des acreors is a prepositional phrase with the meaning
of ‘origin, agent’, which is usually translated into Greek with some other prepositional form (e.g. Š‘¡ + Acc.). Therefore, the accusative form in the Greek version in (7) constitutes a direct result of influence from the French original, as the
Greek translator failed to distinguish between a French genitive and a prepositional phrase, and used a Greek accusative form in an innovative way. It is worth
noting that similar examples are found elsewhere in the Assises (B, 184), but it
remains to be seen whether this type of change remained in use in Cypriot Greek
after the medieval period.
Finally, there is another instance of interference from French into medieval
Cypriot Greek, as far as prepositions are concerned. Perhaps it is no coincidence
that the French preposition de is involved, which, due to its multifunctionality,
created problems in its interpretation and translation to the Cypriot Greek speakers. This is illustrated quite nicely in (8):
(8) ª‚†‰ • †“¦•‡ ... ›Š‰ Š‘œ ‘“‹•Šš •š •‚˜‹‰•Šš.
the tailor
… and for
“About the tailor…and about all craftsmen”
(Assises A, 88)
This example is taken from a heading of a chapter, and explains what the ensuing
text is about, in this case the tailors and all craftsmen. Notice that, in the French
text, these prepositional phrases of ‘reference’ are rendered by de, but in the
Greek Cypriot text, the first of them is translated by using the – common and expected – preposition #, but the second phrase is introduced by $ / ,
which is completely unexpected in this context, as it usually introduces prepositional phrases of ‘origin’. The reason for this is fairly evident: the Greek translator used this preposition instead of #, influenced by the equivalent use of the
French de, which is also used for the meaning of ‘origin’. This pattern is often
repeated in the Assises, and therefore it constitutes an instance of systematic influence exerted in Greek by French (cf. for instance Assises A, 22, 38, 56, among
many other occurrences).
Much more research is needed in order to determine the extent of convergence
of the two main languages in contact (French and Cypriot Greek). The results of
this preliminary research lead us to assume that the morphosyntactic level of
Cypriot Greek was also affected by the multilingualism of Cyprus, certainly in
the domain of the prepositional phrases, but possibly elsewhere as well.
3. Conclusions
The multilingualism in the Greek-speaking world in the medieval period must
have been quite extensive. Recent studies on the social level have shown that the
different communities in Crete and Cyprus, which were the case studies of this
paper, had much more in common than previously thought, and were not kept
strictly apart in their everyday life. Moreover, both linguistic and sociolinguistic
evidence converge in the observation that multilingualism, to a varied extent,
was a feature shared by both the Greek- and the Romance-speaking communities.
The multilingualism affected not only the lexicon, as was mainly argued in
earlier studies, but also various morphosyntactic constructions of Greek (and
possibly of the relevant Romance languages as well, although this issue has not
been looked into). Contact-induced change in this level is hard to prove, given
the fact that morphosyntactic convergence is usually not accompanied by borrowed phonological matter, but usually constitutes a –not identical- adaptation
of a semantic-syntactic pattern (cf. e.g. Heine & Kuteva, 2005). This is particularly relevant for the case of the Greek future-referring ‘the na’ construction,
which had a much more complicated development than traditionally acknowledged, involving language contact as a possible factor in the diachronic path it
followed. Nevertheless, such complex accounts are usually preferable to simpler
ones, when the phenomena under investigation occur in such complex and multiethnic societies such as Venetian-ruled Crete and Frankish-ruled Cyprus.
ACC – accusative; GEN – genitive; PL – plural; PRES – present; SING – singular; SUBJ
– Subjunctive.
Primary sources
AoT = Apollonios of Tyre (870 v.). Kehagioglou, Giorgos. 2004. Apollonios tis Tyrou:
Ysteromesaionikes kai neoteres ellinikes morfes. Thessaloniki: Institouto Neoellinikon Spoudon [Idryma Manoli Triantafyllidi].
Assises = Sathas, Konstantinos. 1877. Mesaioniki Vivliothiki VI: Asizai tou Vasileiou ton
Ierosolymon kai tis Kyprou. Paris: Maisonneuve.
Falieros= van Gemert, Arnold. 2006. Marinou Phalierou ‘Erotika Oneira’. Athina:
French Assises= Beugnot, M. le Comte. 1843. Recueil des Historiens des Croisades,
Lois, t. II: Livre des Assises de la Cour des Bourgeois, pp. 1–226. Paris: Académie
Machairas = Dawkins, Richard McGillivray. 1932. Leontios Makhairas: recital concerning the sweet land of Cyprus entitled Chronicle. Oxford: Clarendon Press.
Sachlikis = Manuscript P(arisinus) (unpublished).
Sourianos = Zapanti, Stamatoula. 2001. Giakoumos Sourianos, Notarios Kastrou. Katasticho 1570–1598. Argostoli.
Baglioni, Daniele. 2006. La scripta italoromanza del regno di Cipro. Roma: Aracne.
Bybee, Joan et al. 1994. The Evolution of Grammar: Tense, Aspect and Modality in the
languages of the World. Chicago: University of Chicago Press.
Cortelazzo, Manlio. 1970. L’influsso linguistico greco a Venezia. Bologna: Patron.
Coureas, Nicholas. 2005. Economy. In: A. Nicolaou-Konnari & C. Schabel (eds.),
Cyprus: Society and Culture 1191–1374. Leiden–Boston: Brill.
De Collenberg, Rüdt Wipertus Hugo. 1982. Le déclin de la société Franque de Chypre
entre 1350 et 1450. Kypriakai Spoudai 46. 71–83.
Dixon, Robert M.W. & Alexandra Y. Aikhenvald (eds.). 2007. Grammars in Contact: A
cross-linguistic typology. Oxford: Oxford University Press.
Edbury, Peter. 2005. Franks. In: A. Nicolaou-Konnari & C. Schabel (eds.), Cyprus:
Society and Culture 1191–1374. Leiden–Boston: Brill.
Giagkoullis, Konstantinos. 1993. The Assizes of the Kingdom of Cyprus: Etymological
Glossary, with an introductory note [in Greek]. Nicosia.
Haspelmath, Martin et al. (eds.). 2005. The World Atlas of Language Structures. Oxford:
Oxford University Press.
Heine, Bernd & Tania Kuteva. 2005. Language Contact and Grammatical Change.
Cambridge: Cambridge University Press.
Heine, Bernd & Tania Kuteva. 2008. Constraints on Contact-Induced Linguistic Change.
Journal of Language Contact 2. 57–90.
Heine, Bernd. 2003. Grammaticalization. In: Joseph, B. & R. Janda (eds.), The handbook
of historical linguistics. Oxford: Blackwell.
Horrocks, Geoffrey. 1997. Greek: A history of the language and its speakers. London:
Joseph, Brian & Panayiotis Pappas. 2002. On some recent views concerning the development of the Greek future system. Byzantine and Modern Greek Studies 26. 247–73.
Joseph, Brian. 1983. The Synchrony and Diachrony of the Balkan infinitive: A study in
areal, general, and historical linguistics. Cambridge: Cambridge University Press.
Kahane, Henry & Renée Kahane. 1982. The Western Impact on Byzantium. The linguistic evidence. Dumbarton Oaks Papers 36. 127–53.
Maltezou, Chryssa. 1997. The historical and social environment [in Greek]. In: Holton,
D. (ed.), Literature and Society in the ‘Rennaissance Crete’ [in Greek]. Iraklion:
Panepistimiakes Ekdoseis Kritis.
Manolessou, Io. 2008. Language contact in medieval Greece: the pronoun ‘o opoios’ [in
Greek]. In: Chila – Markopoulou, D. et al (eds.), Glossis Harin. Volume in honour of
Prof. G. Babiniotis. Athina: Ellinika Grammata.
Manolessou, Io. in press. Medieval Grammar and Historical Dialectology: Observations
based on the Cypriot dialect [in Greek]. Proceedings of the 5th International Conference on Modern Greek Dialects.
Marcato, Gianna & Flavia Ursini. 1998. Dialetti Veneti: grammatica e storia. Padova:
Markopoulos, Theodore. 2007. Language contact research on the texts of Medieval
Greek: methodological evaluation of available sources. To be published in the proceedings of the 8th International Conference on Greek Linguistics, Ioannina, 2007.
Markopoulos, Theodore. 2008. The Future in Greek: From Ancient to Medieval. Oxford:
Oxford University Press.
Markopoulos, Theodore. forth. Case overlap in a multilingual society: a new interpretation of the masculine genitive plural puzzle of Medieval Cypriot Greek.
McKee, Sally. 2000. Uncommon dominion: Venetian Crete and the myth of ethnic purity.
Philadelphia: University of Pennsylvania Press.
Milroy, James. 1993. On the social origins of language change. In: Jones, C. (ed.),
Historical Linguistics: Problems and Perspectives. London: Longman.
Minervini, Laura. 1996. La Lingua Franca Mediterranea: Plurilinguismo, mistlilinguismo, pidginizzazione sulle coste del Mediterraneo tra tardo medioevo e prima età moderna. Medioevo Romanzo 20. 231–301.
Nicolaou-Konnari, Angel. 1995. Language on Cyprus under Frankish rule: Means for the
expression of mutual influence phenomena and determination of national identity [in
Greek]. Vyzantiaka 15. 349–387.
Nicolaou-Konnari, Angel. 2005. Greeks. In: A. Nicolaou-Konnari & C. Schabel (eds.),
Cyprus: Society and Culture 1191–1374. Leiden–Boston: Brill.
Papadopoullos, Theodoros. 1983. Appendice II: Les texts grecs du Livre des Remembrances. In: J. Richard (ed.), Le Livre des Remembrances de la Secrète du Royaume
de Chypre (1468–1469). Nicosia: Cyprus Research Centre.
Ramat, Paolo & Elisa Roma. 2007. Europe and the Mediterranean as Linguistic Areas.
Amsterdam: Benjamins.
Sitaridou, Ioanna & Marina Terkourafi. 2007. On the Loss of the Masculine Genitive
Plural in Cypriot Greek: language contact or internal evolution?. To appear in the
Proceedings of the XVIII International Conference on Historical Linguistics,
Montreal, 6–11 August 2007. Amsterdam: John Benjamins.
Stanitsas, S. 1984. Les gallicismes et les occitanismes dans les ‘Assises de Chypre’ et les
Chroniques de L. Machairas et de G. Boustron en comparaison avec les gallicismes de
la ‘Chronique de Morée Grecque. In: Contribution de Chypre à la civilisation
neo-Hellenique: Actes du VIII Congrés International des neo-Hellenistes des Universités Francophones, pp. 80–127. Montpellier: Centre d’études neo-Helleniques.
Terkourafi, Marina. 2005. Understanding the present through the past: Processes of
koineisation in Cyprus. Diachronica 22. 309–372.
Thomason, Sarah Grey. 2001. Language Contact: An introduction. Edinburgh: Edinburgh University Press.
Thomason, Sarah Grey. 2008. Social and Linguistic Factors as Predictors of ContactInduced Change. Journal of Language Contact 2: 42–56.
Triantafyllidis, Michalis. 1909. Lehnwörter der mittelgriechischen Literatur. Marburg.
Theodore Markopoulos
Department of Linguistics & Philology
Uppsala University
Box 635
SE-751 26 Uppsala
[email protected]
[email protected]
Language use in Moscow schools with an
ethno-cultural component (based on schools
with the Armenian and the Azeri ethno-cultural
Elena Nikishina
Department of Modern Russian Language,
Vinogradov Russian Language Institute
of Russian Academy of Sciences, Moscow
This paper investigates the language situation in Moscow schools with an ethnocultural component – a new form of national schools. The analysis is based on
interviews which were recorded in 2007, in two Moscow schools, one of which
had an Armenian ethno-cultural component, and the other, Azeri. All of the interviews were carried out in Russian. The sample included ten students from
each school (five boys and five girls). Each student had to answer questions concerning their language use at school, in the family, with friends, and in public
places. The respondents were 15-18 years old – thus, pupils of the 10th and the
11th grade (the two last years of high school), who came to Russia with their
families 3–15 years ago.
My goal is to analyze the process of linguistic integration of Azeri and Armenian children into modern Russian society. The choice of these two groups is
motivated by the fact that people from the Caucasus, including the Azeris and
Armenians, suffer from xenophobic attitudes in Russia, more so than any other
national minority of the former USSR. The comparison between these two
groups is particularly appealing, because the effects of Soviet Russification, and
the language situations in general, were different in Armenia and in Azerbaijan.
I will show that this difference influences the use of language by Azeri and Armenian children.
The paper is organized as follows. In Section 1, I briefly outline the history of
Soviet Russification in Armenia and Azerbaijan, and its consequences. Section
2 sketches the history of national schools in the Soviet Union and Russia, and de258
scribes the characteristic features of the teaching of Russian and the national languages in the two schools under analysis. In Section 3, I discuss the degree of
knowledge of Russian and of the national language, at the moment of immigration, and at the present time. Sections 4 and 5, which are the main focus of this
paper, analyze the use of Russian and the national languages, in specific situations involving oral (Section 4) and written communication (Section 5). This
analysis allows me to discover important dimensions in the use of Russian, and
of the national language, in the everyday life of Azeri and Armenian school students.
1. Russification in Soviet society
It is well-known that the Soviet government paid special attention to language
politics. However, these politics changed significantly from one period to another.
In the 1920s, the Bolsheviks carried out the politics of korenizatsija (‘indigenization’), which encouraged each person to use his/her national language in
all communicative situations. They tried to reach this aim by promoting the national languages of the USSR. For instance, all legal documents. and cultural
productions were translated into each national language. Special committees of
linguists created alphabets for unwritten languages, on the basis of the Latin alphabet, and some written languages were Latinized. This period of language
politics was called ‘language building’ (jazykovoje stroitel’stvo).
The situation changed radically under Stalin. The 1930s witnessed the politics
of Russification, the opposite to language building. Studying Russian became
compulsory, and the government changed the alphabets of the languages which
had just been latinized, to alphabets based on Cyrillic. The Soviet leaders after
Stalin continued to pursue this policy of Russification.
This policy was abandoned only during perestroika, a period of liberalization in
the 1980s. The government passed a number of laws which gave more freedom to
all of the nationalities of the Soviet Union, including cultural and linguistic autonomy. After the collapse of the USSR in 1991, the subsequent years were marked
by mass immigration from former Soviet republics to Russia, especially to Moscow. Given the long period of Russification in the history of the USSR, it is natural
to expect that the new migrants did not have difficulty speaking Russian.
Turning to Azeri and Armenian: these two languages, just like all of the other
national languages of the Soviet republics, were considered secondary to Russian in the hierarchy of language use (Alpatov 1991). This hierarchy reflects the
use of different languages in various spheres of life. National languages were
used in many spheres: for instance, in newspapers, magazines, literature, or in
the local administration. Therefore, Azeri and Armenian had a relatively high
status, and stable traditions. However, their fate in the Soviet period was very
The politics of Russification affected Azeri more than Armenian. Traditionally, this difference is explained by the developed national consciousness of Armenians, the high level of development of their language, and by the fact that Armenia was the most ethnically homogeneous republic of the Soviet Union. All of
these features made Armenia more resistant to Russification. However, this was
not the case in Azerbaijan. In contrast to Armenia, where the script has not
changed since its creation by Mesrop Mashtots in the 5th century, the Azeri
script has been changed three times in the 20th century.First, the Latin alphabet
replaced the Arabic writing; then, Cyrillic replaced the Latin alphabet; and finally, in 1991, when Azerbaijan became an independent country, a change to the
Latin script was effectuated again. All of these facts prove that language traditions were stronger and more stable in Armenia, than in Azerbaijan.
2. History of national schools, and modern schools with
an ethno-cultural component
Schools with an ethno-cultural component, which emerged in the Post-Soviet period, are a new form of the national schools that used to exist in the beginning of
the 20th century. The history of national schools in Russia (especially in Moscow) can be divided into three periods: 1) before Stalin; 2) from Stalinism until
perestroika; and 3) from the end of perestroika until now (see Alpatov 2000 and
Daucé 2007 for details).
National schools existed in Russia long before the Soviet period. After the
Bolsheviks came to power, they began to implement new language policies.
They founded, throughout the whole territory of the USSR, many national
schools where the education was held in the national languages. Under Stalin,
ethnic groups were deprived of their official status, and all ethnic schools and
ethno-cultural associations were banned. They completely ceased to exist during
the Second World War.
Only under Gorbachev, when liberalization began, did national movements in
the USSR become stronger. In 1989, the first national school opened, for the victims of the earthquake in Armenia. Then, in 1990–1991, during the “Parade of
Sovereignties,” many laws were passed which gave all sorts of liberties to national minorities, including language sovereignty, and national schools began to
emerge again. In 1997, the Moscow government adopted the Regulation on
Educational Establishments with an Ethno-cultural (National) Component (Položenije ob obseobrazovatel’noj škole s etnokul’turnym (nacional’nym) komponentom obrazovanija v g. Moskve1), so the ethnic schools acquired the status of
State institutions.
The text of the regulation is accessible at
The education in these schools is based on an application of the federal educational programme, common to all schools in Russia, and of the ethno-cultural
component, which includes courses of the national language, history and culture.
Although each school has only one ethno-cultural component, these schools can
accept all children, irrespective of their nationality, including Russian children.
Schools with an ethno-cultural component are mostly financed by the Moscow
Department of Education, but they may also get financial support from the countries whose language is studied at these schools.
All subjects, except those which are included in the ethno-cultural component
of education, must be taught in Russian. The government (for instance, the government in Moscow) demands that the schools give particular attention to teaching Russian as a vehicular language. Besides this, these schools give the children
of immigrants the opportunity to attend a course of Russian as a foreign language, at the beginning of their stay in Moscow.
It should be noted that schools with ethno-cultural components correspond
better to the needs of national minorities than Russian schools. Alpatov (2000)
observes that each language speaker has two “needs” which must be satisfied:
the need for identity, and the need for mutual understanding. The need for identity is the desire of each individual to speak his/her mother language, whereas the
need for mutual understanding is the need to be able to speak to everyone without
any problems. Schools with ethno-cultural components satisfy both needs: on the
one hand, children’s identities are preserved because they can speak to each other
in their mother tongue, and study their own language and culture; on the other
hand, children study Russian in order to be able to communicate with everyone
in Russia. In standard Russian schools, Azeri and Armenian children satisfy their
need for mutual understanding by studying Russian, but they lose their identity,
because they do not study and/or speak their mother tongue. However, the disadvantage of schools with an ethno-cultural component is that it is more difficult
for children to study there than at standard schools, because the pupils have to
follow a double educational programme: the federal programme, and the ethnocultural component.
Both schools at which the present research was conducted have the official
status of schools with an ethno-cultural component. However, there are some differences between them. The school with the Armenian ethno-cultural component
is almost mono-national, whereas the school with the Azeri component is
multi-national: the number of Azeri children at this school is equal to the number
of Russian children. Furthermore, at the Azeri school, there are more Russian
teachers than Azeri ones; by contrast, at the Armenian school there are almost no
Russian teachers. In all other aspects, these schools are organized in the same
way: at both schools there are centers of national culture, which offer different
activities to the children (circles of national dance and song, lessons of national
musical instruments, and so on). Both schools have a museum, where national
dress and other cultural items are exhibited. The number of children in both
schools is relatively small, in comparison with other Moscow schools.
The teaching of national languages is organized in different ways in the Azeri
and the Armenian schools, which is due to the fact that one of the schools is
multi-national, and the other one mono-national. At first glance, it seems that the
aim of preserving the national language and culture is more important to the administration and teachers of the Armenian school, than to those of the Azeri one.
At the Armenian school, the national language is taught more thoroughly. First
of all, at the Armenian school, each class has four lessons of the national language (Armenian) per week, whereas in the Azeri school there are only two lessons per week, and in the last two years, children do not study Azerbaijani at all.
Moreover, at the Azeri school, the Azeri lessons often take place at the end of the
day, after all of the other courses, when the children are already tired. At the Armenian school, Armenian lessons are scheduled at a more convenient time. Beside this, at the Azeri school, many children in my sample did not show any interest in Azerbaijani. My sample even included Azeri children who confessed to
shirking the lessons of the national language, because they did not believe them
to be really useful. This is why they do not progress in their language, and completely lose their knowledge of written Azeri. In contrast, Armenian children are
really interested in studying their national language. The level of their knowledge of oral and written Armenian is really high; for instance, in the last years of
school, they can easily write compositions in courses of Armenian literature.
To analyze these differences, I use the conception proposed by C. Baker
(1993). Baker analyzes possible forms of bilingual education, and divides them
into two subtypes: “strong forms” and “weak forms”.
Strong forms of bilingual education serve the purpose of making children
bilingual. At the end of their education, they are supposed to speak two languages perfectly: their mother tongue, and the official language of the country
where they live (in our case, it is Russian). This form of education presupposes
that children not only study the official language in order to achieve professional
success in the “accepting country,” but also to maintain their mother tongue. In
contrast, bilingualism is not the ultimate aim of weak forms of education. It is,
rather, regarded as an intermediate stage between monolingualism in the mother
tongue, and monolingualism in the official language of the “accepting country”.
I propose that in the school with the Azeri ethno-cultural component, the
weak form is used. Since knowledge of Russian becomes sufficient for Azeri
children, and comes to the foreground in their practical life, they do not see any
need to continue learning their national language. Even the school cannot
stimulate their interest in the Azeri lessons. In contrast, the school with the Armenian component uses the strong form of bilingualism, because it offers the
children equal opportunity for studying Russian and the national language.
Therefore, Armenian children should eventually become bilingual, whereas
many of Azeri children gradually lose their knowledge of Azerbaijani.
The Regulation on Educational Establishments with an Ethno-cultural (National) Component requires that all children at schools with an ethno-cultural
component pass the same exams as pupils in standard Moscow schools; that is,
they have to take all of the exams in Russian. As the administrations of the Armenian and the Azeri schools are interested in the success of their pupils, this
causes them to give particular attention to teaching Russian. However, despite
all of these efforts, children at the two schools sometimes do not reach the level
of Russian that is considered to be sufficient to pass the Common State Examination (graduate exam) in the Russian language.
All of these facts show that education in schools of this type is not unproblematic. Below, I will show that these schools, which contribute much to the integration of the children into Russian society, nevertheless cause some tension between Russian and the national languages.
3. Knowledge of Russian and of the national language
In this section, we want to compare the children’s current knowledge of Russian
and of the national language, with that at the moment of immigration, in order to
estimate the dynamics and the rhythm of their integration into Russian society.
Most of the families of our Azeri and Armenian informants left their countries
because of economic problems. In almost all cases, the emigration was forced
and unexpected; therefore, the emigrants did not have time to learn Russian. This
is why children who arrived in Russia did not speak Russian at all, or had insufficient knowledge of Russian to use it for everyday communication. However,
most of the Azeri children spoke Russian better than the Armenian children. This
fact results from some particularities of the current educational system in
Azerbaijan. In this country, some schools belong to the so-called “Azeri sector,”
where most subjects are taught in Azerbaijani, and others to the “Russian sector,”
where pupils learn Russian, and most subjects are taught in Russian. Some of the
Azeri children in our sample studied in the Russian sector prior to their immigration. In Armenia, a strong tradition of teaching Russian exists, but there is no division into two sectors parallel to the Azeri one. There are only Armenian
schools, where Russian is taught as one of the subjects.
A better knowledge of Russian by the Azeri children is also related to the fact
that Russification was stronger in Azerbaijan than in Armenia. Even children
who did not study in the Russian sector acquired an elementary knowledge of
Russian from their parents. Therefore, there is nobody among our Azeri informants who did not know Russian at all at the moment of immigration, whereas
among the Armenian informants, there are such children. However, even the
children who knew Russian had only a passive knowledge of this language – they
could understand something in Russian, but could not speak it. This caused many
of them to take private lessons of Russian, at the beginning of their stay in Moscow.
With time, this difference in the knowledge of Russian between Azeri and Armenian children disappeared. This was due to the fact that children who arrived
in Russia at an early age managed to learn basic Russian in the shortest amount
of time (3 months for children of 6–10 years). For those children who were older
at the moment of their arrival to Russia, learning Russian could take a bit more
time. In general, the knowledge of Russian by our informants is comparable to
that of Russian children of the same age.
As to the national language: at the moment of emigration, the children did not
have any problems with their national language. They spoke it well, and were
monolingual. The problems began later, when Russian started to drive out their
national languages from the spheres of everyday communication. Currently, the
children are bilingual, but their bilingualism can take different forms. As far as
the time of acquisition of Russian and of the national language is concerned, in
the case of most informants we can speak of the consecutive, but not simultaneous, form of the bilingualism – i.e., the national language is learnt, and is spoken
in the family, before the children start to learn Russian at school (see Belikov,
Krysin 2001, on different types of bilingualism).
Concerning the correlation between the two languages, we can say that most
children have the coordinative form of bilingualism (they have equal knowledge
of both languages). 15 informants in our sample can pretend to the status of ideal
bilinguals: they have perfect knowledge of both languages. However, the problem is that in fact, the status of Russian and of the national language is not entirely equal: 13 out of 15 pupils admit that it is already easier for them to use Russian in most situations. For instance, in the emotional situations of quarrels and
disputes, when they do not control their choice, they often switch to Russian.
The remaining 5 children in our sample, whose school results are below average, have mastered spoken Russian, but do not understand that this is insufficient, and that they should also learn the written Russian language. Even after
many years in Russia, their Russian vocabulary is limited, and lets them touch
only upon a restricted range of subjects. They continue to make serious mistakes
in writing. Thus, their bilingualism is subordinative: the notions of Russian are
perceived through the notions of the native language.
Therefore, all informants know Russian and their national language very well,
but all of them feel the tension which occurs between the two languages. Below,
we will analyze the distribution of the languages across the spheres of communication.
4. Use of Russian and of the national languages in oral
Our analysis of the use of Russian and of the national languages by the Azeri and
Armenian children is based on the scheme proposed by Joshua Fishman. According to Fishman 1965, the use of a particular language is determined by several
parameters or dimensions: membership in a social group; the situation of communication; the subject of conversation; the “channel” of communication (oral
vs. written communication), and so on. Using the responses of the informants to
my questions, concerning their use of the languages in each of the spheres, I
analyzed the use of Russian and of Armenian/Azerbaijani within the family, with
teachers, with classmates and friends, and in public places. I also investigated the
use of language in written communication.
4.1. Family communication
18 out of the 19 children I interviewed said that they speak their mother tongue
in the family, and only one Azeri girl speaks Russian (in fact, the whole family
speaks Russian – they chose to speak Russian in order to integrate more easily
into Russian society). The case of this girl is unique: all of the teachers at the two
schools assert that normally, all pupils speak their national languages within the
family. These statistics are illustrative, because they show that the sphere of
family communication is the most resistant to the difficulties that accompany the
process of integration. It is also resistant to time, since even those families which
immigrated to Russia more than ten years ago have not turned to Russian in
family communication. Nevertheless, even this sphere could not remain unchanged.
The general tendency is that most children sometimes use Russian within their
families when they speak to their younger brothers and sisters, who do not know
Armenian and Azerbaijani very well (they were born in Russia). The elder members of some families (mostly women who are housewives, and do not have to
use Russian at their work place) do not speak Russian very well, and therefore,
do not understand their children when they are speaking Russian to each other.
The knowledge of Russian increases and the knowledge of the national language
decreases from the oldest generation to the youngest one. Thus, in this case, Russian becomes a factor in the segregation of the older and the younger generations.
Most children use Russian when they just cannot find the right word: for instance, they do not know scientific terms, the Russian variants of which they are
exposed to at their schools, in their mother tongue (it is possible to say that they
use a sort of pidgin). This proves that the mother tongue is used only for everyday
communication, whereas Russian is used for more complicated subjects. The
parents have different attitudes toward their children’s use of Russian in the
family. Most of them do not mind it, because they understand that Russian is
more important for the future of their children than Azerbaijani or Armenian.
However, there are parents who rebuke their children, and ask them to speak
their national language at home.
The length of time that the children have lived in Russia also plays a significant role: only two persons in our sample – Armenian children who moved to
Russia less than five years ago – never switch to Russian when speaking to their
family members.
Communication within the family is a situation where the parents determine
the “tonality” (decide which language to speak), and where the children should
accommodate to their choice. The use of the ‘pidgin’ helps them in this situation.
4.2. Communication with teachers
Communication with teachers at school is mostly carried out in Russian. According to the Regulation on Educational Establishments with an Ethno-cultural (National) Component, all subjects must be taught in Russian, except those which
are included in the ethno-cultural component (national language and culture).
The ideology of these schools is based on the primacy of Russian in the process
of education; therefore, all teachers must watch the children in order to realize
this idea.
In general, the children assert that their teachers observe these rules. For instance, one of the Armenian girls told us that when she greets the teacher of Russian, who is Armenian, with the Armenian phrase Barev dzez! ‘Hello!’, the
teacher tells her that she should speak Russian at school. However, sometimes
teachers in both schools switch to the national language, in order to be better understood by the children. As I have said in Section 2, in the Armenian school almost all of the children are Armenian, whereas in the Azeri school, only 50% of
pupils are Azeri, and the rest are mostly Russian. Because of this, Russian is used
as a vehicular language in the Azeri school. Thus, we could expect that in the
Azeri school, the Regulation is always observed, because the teachers would also
violate the pedagogical ethics if they speak Azerbaijani in the presence of Russian pupils. Nevertheless, the teachers sometimes speak the national language in
both schools. Moreover, in the Azeri school, Russian children usually learn some
Azerbaijani words, in order to understand what their teachers tell the Azeri children. As to the Armenian school, my informants say that the course of mathematics is simultaneously taught in both languages, contrary to the Regulation, to
help the children understand the subject better.2
4.3. Communication with friends and classmates
I combine communication with friends and classmates into one type, because in
most cases, the friends of our informants are at the same time their classmates,
and communication with them usually takes place at school. However, it is primarily informal, which is why the formal school rules imposed by teachers are
no longer in force. The children are relatively free in their choice of language,
but this choice depends on multiple factors.
First, the use of language depends on the interlocutor: whether (s)he knows
Azerbaijani/Armenian, or not. Of course, the children in our sample use Russian
as a vehicular language in communication with those of their friends who do not
speak Azerbaijani or Armenian. In contrast, the use of language in communica2
This manner of teaching mathematics ideally corresponds to the representation of the mother
tongue in linguistics. Some linguists suppose that to find out which language is more important for a
bilingual person, it is sufficient to ask him her which language (s)he counts in mentally, because each
bilingual always chooses the language which is best fixed in his/her mind for counting (Zemskaja
2001). During my research, I asked all informants which language they count in. As it turned out, all
of them use their national language for this purpose, even those children who assert that it is easier
for them to speak Russian than Azeri or Armenian.
tion with Azeri/Armenian friends is conditioned by the topic of the conversation,
and by the presence/absence of pupils who do not speak Azerbaijani or Armenian.
Note that in this sphere, the difference between the two schools under analysis
is also significant. As I have shown, in the Armenian school, all children are Armenian, so they can use Armenian during the breaks. But in fact, what the pupils
use is a mixture of their mother tongue and Russian, including many Russian
words for which they do not know the equivalents in their mother tongue (the
same sort of pidgin which they use in family communication). Nevertheless,
rather often, our Armenian informants completely switch to Russian, mostly
when they discuss the material learnt at school, or their homework. The only reason is that the principal language of education at this school is Russian, so they
know many terms, such as “Newton’s laws,” or “Periodic table of chemical elements,” only in Russian, and can hardly translate them into their mother tongue.
As far as the Azeri children are concerned, the distribution of the two languages is the same: Russian is used for discussing school subjects, and Azeri (or
rather, a mixture of Azeri and Russian) for all other topics.
Now I pass to the second factor – the presence/absence of pupils who do not
speak Azerbaijani or Armenian. Situations in which communication takes place
in the presence of children who do not understand Armenian are very rare at the
Armenian school. But at the Azeri school, everything is more complicated, because Azeris study side by side with Russians. Therefore, unlike Armenian children, the necessity of respecting the Cooperative Principle (Grice 1975) – that is,
speaking a language which can be understood by all participants (Russian) –
weighs upon all Azeri children. The problem is that this principle is not always
respected. Sometimes the children do not control themselves, and continue to
speak Azerbaijani even in the presence of Russians. Moreover, it sometimes happens that they speak Azerbaijani on purpose, when they want to tell their Azeri
friends something that ought to be kept secret from Russians. However, they
claim that this does not lead to any conflicts.
4.4. Communication in public places
Communication in public places is a particular sphere. I have found that the
choice of language in this situation, by Azeri and Armenian children, obeys other
rules than those relevant for family communication, or communication with
friends and classmates. If an Azeri or an Armenian child is in a public place (in
public transport, or in the street), it is not important to him/her who (s)he speaks
to, and what (s)he speaks about. The crucial thing which determines the characteristic features of the communication is that a child is surrounded by strangers
who do not speak his / her national language.
In this situation, the children use different arguments to explain their choice of
language. Most of them say that they must speak Russian to show respect to the
Russian-speaking people around them (that is, the Cooperative Principle is also
important here) – accordingly, they always speak Russian in public places. This
argument is used mostly by the Armenian children, especially by girls, and more
rarely by the Azeri children.
Another argument used is that the children speak Russian because they are
afraid to be identified, and of differing from the majority of people: this is explainable, taking into account the large number of nationalists in Moscow, and
especially the xenophobic attitude towards Caucasian immigrants within Russian society. Very often, it is the parents fearing for their children, who insist that
they speak Russian in public places. The children do not always follow their parents’ advice.
In contrast, there is an argument which makes many children speak their national languages in public places: they say that they are proud to be Azeris or Armenians, and consider it to be humiliating to hide their ethnicity. This argument
is mainly used by Azeri children, and much less by Armenian children. Hence,
one could conclude that in this situation, the national identity is more important
for Azeri children than for Armenian ones, and that the latter are more polite than
the former. However, these conclusions would be hasty and superficial. It is
more precise to say that for Azeri children, who choose their national language
for communication in public places, affirmation, and sometimes even proclamation, of their national identity is the crucial thing in this situation – whereas for
Armenian children, it is more important to avoid a conflict.
5. Use of Russian and of the national languages in
written communication
In this paper, we use the term ‘written communication’ in a broad sense: it covers
all situations of everyday life where our informants have to use the written form
of the language, either for emission (writing), or for reception (reading).
When the families of the children in our sample moved to Russia, their use of
the written form of the national languages began to decline, whereas their use of
Russian in written communication increased: the children prepare their homework in Russian; it is difficult to find any books or magazines in their national
languages in Moscow, and so on. As it turned out, the most frequent situation in
which the national language is used, is during lessons of this language. Thus,
knowledge of the written language depends directly on the manner of teaching
the language at school. The forms of bilingual education at the two schools under
analysis have been examined above (see Section 2). It is obvious that the pupils
of the Armenian school, where the strong form is represented, know the written
form of the national language better than those of the Azeri school (weak form).
Among our Armenian informants, only one boy cannot read and write in Armenian, and this is due to the fact that he used to study at a Russian school, and entered the school with the Armenian ethno-cultural component only one year ago.
As to the Azeri children, cases like this are more frequent, even for children who
have always studied at the Azeri school. Their answers show that they have additional difficulties, related to the recent change from the Cyrillic alphabet to the
Latin one. There has not been a change of the alphabet in the history of the Armenian language.
In situations involving written communication, the importance of knowing not
only the grammatical and lexical mechanisms, but also the whole system of the
communication norms of a language, becomes more evident. In other words, if
a person is able to apply the grammatical rules of a language, and write without
mistakes, it does not mean that (s)he can understand the classical literature in this
language, or express his/her thoughts in the form of a school essay. Our informants can distinguish these two types of knowledge of the written language.
I asked the children to estimate the level of their written expression in Russian,
and in the national language, and I found that the Armenian children have an
equal level in the two languages: many of them can easily write texts in Russian
and in Armenian without mistakes. As to the Azeri children, they assert that their
level in written Azerbaijani is rather low; almost all of them make some mistakes
in writing. By contrast, most of them (except those whose school results are below average) say that they make much fewer mistakes in written Russian. Because of this, the Azeri children feel embarrassed when writing Azerbaijani.
However, in situations involving informal written communication (SMS, short
letters), when nobody can accuse them of spelling errors, the children may use
their national language. Zemskaja (2001) claims that emigrants who have lived
abroad for a long time often feel embarrassment at writing in their national language, because they may make a mistake.
In order to find out whether my informants know the whole system of communicative norms in the two languages, and whether they can express their thoughts
in the written form, I asked them to think in which language it is easier for them
to write a school composition. As it turned out, for most Azeri and Armenian
children, it is already easier to do it in Russian, despite the fact that in the Armenian school, children often write compositions in their national language. Many
of the informants referred to the fact that they have a limited vocabulary in their
national language, which does not allow them to write anything on serious
topics. This is not the case with children who have lived in Russia for less than
five years – their communicative skills in the national language still outrank their
skills in Russian.
If we consider, in more detail, our informants’ ability to compose written texts
in Russian, it turns out that everything is not so good. Many of the children confess to using templates of school compositions, which one can buy in any bookshop in Moscow. Thus, we can conclude that their communicative competence
is insufficient for writing connected texts by themselves.
As to the reception of written texts in the national language, and in Russian,
the situation is grosso mode the same as with the emission of written texts: almost all children find it easier to read in Russian than in their national language,
irrespective of genre, be it classical literature, school-books, magazines, Internet
websites, and so on. However, many informants have problems with comprehension of Russian classical literature: some of them read Russian classics in the
brief adapted version. This proves that it is difficult for immigrants’ children to
master the communicative norms of written Russian.
I conclude that Azeri and Armenian children gradually lose their knowledge
of the written form of their language, whereas the written form of Russian comes
to the forefront. However, only a very small number of pupils attain a high level
of written Russian. Moreover, one should not forget that knowledge of the written language depends directly on the length of time since emigration.
6. Conclusions
In the present paper, I have shown that schools with an ethno-cultural component, while assisting the children in integrating into a new environment, at the
same time cause significant tension related to language use. The children have to
follow a double programme, and since time does not always allow them to concentrate equally on studying Russian and the national language, they are forced
to choose between them. In this situation, all of the informants pay more attention to Russian, because they associate it with their future success at school and
at work. However, they do not refuse their national language, either, because
they need it to “feel their roots”. For Armenians, the two aims – to integrate into
the Russian society, and to preserve their language and culture – are equally important. This is why the Armenian school is almost mono-ethnic, and the Armenian children are motivated in learning Armenian. Although Azeris also realize
that preserving their language and culture is crucially important, in practice, the
aim of integrating into Russian society appears to be much more important.
The comparison between knowledge of Russian and the national language, at
the moment of emigration, and the present time, showed us that the Azeri and the
Armenian children integrate into the Russian society rather quickly. The rhythm
of the integration depends first of all on the ambition of the children and their
parents, which manifests itself in the effort they make to learn both languages
(persevering work at home, private lessons vs. passive attitude).
In oral communication, Russian and the national language are distributed
among the spheres of communication. The analysis of the spheres (communication with family members, with teachers, with friends and classmates) allows us
to conclude that the use of the national language is mainly restricted to everyday
life, whereas Russian is employed in a wider range of situations (for Armenians,
this distribution is more apparent than for Azeris). The case of communication
in public places is an exception to this rule, because in this particular situation,
children often switch to Russian due to fear of xenophobia. As to the written
forms of the languages, it is possible to say that this form of the national language
cannot develop normally in emigration. It retreats to the background, and surrenders its position to written Russian.
The distribution of the languages among the spheres is not complementary. In
the family, at school, and in other spheres, the children in our sample may switch
from Russian to Azerbaijani/Armenian (in other words, we see a situation of interference). Therefore, language use by the Azeri and Armenian children in our
sample can be regarded as a case of bilingualism, with some elements of diglossia (functional distribution of languages).
Alpatov, Vladimir M. 2000. 150 jazykov i politika. 1917–2000. Sociolingvistieskije
problemy SSSR i postsovetskogo prostranstva. Moscow: Kraft+, Institut vostokovedenija RAN.
Belikov Vladimir I. & Leonid P. Krysin. 2001. Sociolingvistika. Moscow: Rossijskij
gosudarstvennyj gumanitarnyj universitet.
Zemkaja, Elena A. (ed.). 2001. Jazyk russkogo zarubež’ja: Obsije processy i reevyje
portrety. Moscow: Jazyki slavjanskoj kul’tury.
Baker, Colin. 1993. Foundations of Bilingual Education and Bilingualism. Clevedon:
Multilingual Matters.
Daucé, Françoise. 2007. The Reinvention of National Schools in Moscow: Between
Liberal Claims and State Practices. In Proceedings of the conference, organized by the
Association for the Study of Nationalities. In press.
Grice, H. Paul. 1975. Logic and Conversation. In Peter Cole & Jerry L. Morgan (eds),
Syntax and Semantics, vol. 3. Speech Acts, 41–58. New York: Academic Press.
Fishman, Joshua. 1965. Who speaks what language to whom and when? La Linguistique
2. 67–88.
I thank Alexander Letuchiy for his help in preparation of the English text of this paper.
Elena Nikishina
Russia, 127273, Moscow, Dekabristov street, 38–134,
[email protected]
Lexical Aspect too is learned: data from Italian
Learner Corpora
Stefano Rastelli
University of Pavia
The hypothesis presented in this work is based on the observation of two facts:
(a) the actional content of initial L2 learners’ predicates is often underspecified;
(b) L2 learners’ predicates often fail the actional tests designed for mature languages. As a consequence, it is claimed that L2 Actionality is a separate and late
component of the verb meaning and that the properties of L2 Actionality are not
innate, rather they are learned. Since probably L2 Actionality has to be focalized
gradually by learners, it cannot be held as the triggering factor for paving the way
to the acquisition of the whole Tense-Aspect system, as it is supposed by the proponents of the Aspect Hypothesis. In fact, sampling data from five different
Italian learner corpora possibly shows that when learners use – say – a telic verb,
it is far from being unquestionable that they know the properties of telicity. On
this grounds, it’s very hard to assess whether a verb is telic or not (or stative or
durative etc.) in interlanguages. Finally, since L2 Actionality cannot be said to
pre-exist to other semantic traits, its acquisition and the acquisition of Tense and
of grammatical Aspect (the language-specific encoding of aspectual morphemes) probably progress at the same time.
1. Key-Issue
According to the Aspect Hypothesis (AH) (Andersen & Shirai, 1994), verb Actionality (also referred to as Aktionsart or lexical Aspect) is a determining factor
in the acquisition of Tense and Aspect. Shirai & Andersen (1995) claim that the
distribution of aspectual morphemes is biased by an inner predisposition which
leads learners to first acquire and use traits which are semantically close (i.e.,
pastness, perfectivity and telicity vs. non-pastness, imperfectivity and atelicity).
Also data from Italian learner corpora seems to confirm this bias (Bernini, 2005;
Giacalone Ramat, 1995). Furthermore, a large number of studies confirms that
the traditional vendlerian quadripartition of Aktionsart upon which the AH is
based is accepted as a psychological realis, along the line of Dowty (1991) and
van Valin (1990). The AH, though, attracted some criticism, mainly in three respects: Firstly, the AH is at risk of committing closeness fallacy because the actional value often attributed to L2 verbs is that of their translation in the targetlanguage (or in English, see Ebert, 2005; for a discussion see Lardiere, 2003;
Shirai, 2007). Secondly, at least in its early strong version, the AH downplays
the decisive role of input: if the distributional bias attested in learner data simply
reflected a tendency in the input, there would be no need to postulate a cognitive
predisposition in order to explain why some associations are more frequent than
others. Thirdly, the AH is at risk of downplaying the contribute of the phrasal
context. Just to make an example, in native Italian the Actionality of about 75%
of verbs is compositional, that is to say, it depends on the context (see Lenci &
Zarcone, in press). It has been recently acknowledged by one of the AH proponents (Shirai, 2007:10) that trying to interpret the way learners mentally represent the actional content of L2 verbs could turn out to be highly misleading. The
hypothesis that Actionality too is learned stems also from the implications of this
statement. The claim is that verb L2 Actionality undergoes the same learning
process which involves all the other traits belonging to the Tense-Aspect system
(aspectual grams, time expressions, PP etc.). The actional content of L2 verbs too
would be in reconstruction and would be subject to fine-tuning and also to great
variation over time. From a methodological point of view, Actionality should be
looked at and accounted for inductively and compositionally (at a VP level or at
P levels) especially at early stages of the learning process. Finally, as far as the
explanatory and descriptive adequacies are concerned, the current hypothesis
does not contradict the claim that Actionality is a parameter whose value has to
be fixed by learners. This might be true, on condition that this process is much
longer and much more complex than it is supposed by Slabakova (2001).
2. Aspect Hypothesis re-opened
Four issues at least can be raised along the line of the recent criticism to the Aspect hypothesis sketched in §1. The first one concerns comparability of different
analyses. Bardovi-Harlig (2002:129), admonishes that: “the differences in these
analyses could lead us to support or reject the Aspect Hypothesis on the basis of
the very same data”. In fact, the criteria adopted in different works may strongly
affect the results and compromise their mutual comparability. For instance, the
researchers usually classify the actional content of learners’ data by basing on the
well-known grid of four vendlerian classes (state, activity, accomplishment,
achievement), which – in their turn – are established mainly on the basis of the
classical actional tests. To give an example, Housen (2002:166) classifies the
English verb ‘grow up’ as atelic, despite Vendler (1967:108) placing it among
telic verbs. We too could classify the correspondent Italian verb crescere as a
gradual completion verb (see Bertinetto & Squartini, 1995) and – as such – fully
telic. As long as misalignment between parameters of different analysis is not
clearly recognized and handled adequately, non-comparability among results
won’t represent an isolated case. The second issue concerns the reliability of Actionality tests. There exists a set of Actionality tests (for an exhaustive and
up-to-date view, see van Valin, 2005: 34–42) which is assumed of detecting the
vendlerian classes in many (if not all) languages. The assumption of the actual
cross linguistic effectiveness of such tests make some scholars conclude that the
vendlerian classes are psychological realia or something very close to language
universals (Andersen & Shirai, 1994:532) or to semantic primitives. For instance, Weist (2002:36) maintains that “the Vendler-like categories have broad
cross-linguistic semantic and syntactic implications”. Quite differently, Tatevosov (2002:324) claims that the vendlerian classes are not semantic primitives or
language universals but that “Actionality is a parameter that allows for different
settings in different languages”. While underlying the language-specific orientation of terminology used in Actionality studies, he quotes Ebert (1995:186):
“most often it is assumed that a verb or verb phrase has the same actional
character as its closest English counterpart”. Verkuyl too (2005:9) reinforces the
claim of non-universality of actional traits aiming at the non-reliability of the
tests and observing for instance that the well-known telicity-check test based on
the imperfectivity paradox can’t work with languages such as German or Dutch
which lack a morphological encoding for the progressive form (Dutch at least
has a periphrastic form). As long as researchers won’t declare under which (often
peculiar) conditions and tests a certain verb is assumed as belonging to a certain
actional class, there is the risk of exchanging a bundle of language-specific traits
for universally shared properties of language. The third issue concerns the intrinsically elusive character of Actionality. Actional categories are highly elusive
and the judgement on the belonging of a verb to a class or another is far from being clear-cut also for native speakers. The actional value of a predicate seems to
be a matter of compositional induction rather than a lexical deduction. That’s
why Lenci & Zarcone (in press) offer a stochastic model of Actionality in order
to underline its high variability and its strong dependence on the context. As far
as L1 Italian is concerned at least, the actional shift (or ‘actional hybridism’ as it
is called in Bertinetto, 1986) seems to be the norm rather than the exception. The
hypothesis put forth by Lenci & Zarcone (in press) is that the interpretation of
the actional value of a verb in context is the result of a complex process of integration of hybrid linguistic constraints which act as “probabilistic soft constraints”. If we take a look at the data presented in their article, one is easily
driven to believe that, under the appropriate conditions, almost every Italian verb
in the corpus may be regarded as belonging to two, sometimes three and even
four different actional classes. For instance, among 3429 verbal tokens of 33 V
types extracted from the TreSSI corpus (a corpus of written modern Italian), the
cases of univocal or almost univocal actional assignment amount to about 25%,
that is, the absolute minority. The last issue concerns data elicitation. Bertinetto
et al. (in press) find important exceptions and unexpected non-prototypical
matchings between stative and activity verbs with perfective morphemes in a
corpus of L1 Italian children. For instance, the activity verbs vedere ‘see’ and
disegnare ‘draw’ are among the first verbs to be marked with the past participle
suffix (besides other telic verbs such as rompere ‘break’, cascare ‘fall down’, arrivare ‘arrive’). If the authors are right, this behaviour might be directly triggered
by adult speech and – I would add – also by the kind of task in which the children
are involved (e.g., if they are asked to recount something they are doing or something they have just stopped doing). The additional factors brought into play by
the context of interaction and by the way in which spoken and written data are
elicited leave us once more with the difficult task of deciding whether it makes
sense or not to assign an actional class to a verb on a priori basis or to look for a
kernel actional meaning for L2 verbs. As far as respectively first and second language learning is concerned, Bertinetto et al. (2006) and Giacalone Ramat &
Rastelli (2008) advance what might be called a ‘syncretic account’, according to
which the re-construction of Tense-Aspect system in a second language takes
place starting from syncretic and vague categories, none of them seeming to
work as a guiding principle. These authors – even though from different perspectives – share the assumption that at the initial stage of the learning process, learners move for a syncretic dimension in which the pertinent Tense-Aspect traits appear under-determined and tightly intertwined. Learners’ task will be to disentangle such categories. As far as second languages are concerned, Rastelli (2008)
uses the expression “zero-degree stage” to convey the idea that a theory of
Tense-Aspect acquisition which aims at being explanatorily adequate should not
require any temporal or aspectual notion to be pre-existent to the initial moment
in which learners begin to map temporal, aspectual and actional values together
into a single predicate.
3. The data and the method
The hypothesis advanced in this paper is based on a sampling from performance
data. The analysis of interlanguage data adopted privileges contexts where the
same V is combined with different aspectual morphemes or/and with time expressions. In this kind of analysis, I consider L2 Actionality ‘in action’, that is,
as it results when tested out against complex sentences and not in isolation. All
the sentences presented in this paper are sampled from five different well-known
(some of them not yet published though) Italian learner corpora:
(a) Corpus ISA (Italiano Scritto di Americani – Italian writings of American Students). Undergraduate students spending a semester abroad in Milan were asked
to re-write some scenes from the film “Pane e tulipani” (see Rastelli, 2006).
(b) Corpus Rosi: oral and written retelling of short film scenes by Erasmus students collected at the University of Pisa by Fabiana Rosi.
(c) Corpus LIPS (Lessico dell’Italiano Parlato di Stranieri – Lexicon of Spoken
Italian by foreigners), collected at the University of Siena (see Barni & Gallina,
(d) Corpus Prin Verona: oral retelling plus personal narratives (supervised by
Camilla Bettoni and Gabriele Pallotti).
(e) Corpus Marco Polo: oral retelling of the “frog story” and animated puppet
scenes by Chinese students (collected by Michela Biazzi and Cecilia Andorno at
the University of Pavia).
These corpora are very different from each other and thus the sampling is of no
significance from a statistical point of view. Nevertheless, I do not agree that the
sampling procedure is not attractive to SLA research. What may be appealing is
the fact that the phenomena I’m going to describe occur in all these learner corpora, regardless of parameters such as learners’ first languages, instructional and
social backgrounds, age, kind of elicitation task, medium and textual genre. It is
true that designing and evaluating data from a sampling procedure is something
different than evaluating data from longitudinal or quantitative analyses. Paradoxically though, the heterogeneity and the relative scarcity of samples are an argument in favour of the comparability of the inductive convergences they might
show. In fact, the less the data is constrained by extra-linguistic factors (such as
those mentioned above) the more the analogies are likely to be attributed to linguistic factors alone. In general, sampling data is falsifiable because it is incomplete by definition. Admittedly, sampling data is more helpful to build a hypothesis than to prove it. In this respect, the data I’ll be presenting needs to be confirmed or discharged by quantitative analysis before it can be compared to experimental data.
4. The hypothesis that L2 Actionality is learned
As tentatively assumed in §§ 1 and 2, the hypothesis that L2 Actionality is
learned does not exclude the existence of an actional blueprint. Actionality may
be conceived as being a parameter to be set again by learners over time. Instead,
the hypothesis claims that an actional bootstrapping is highly unlikely. The divorce between the nature of L2 Actionality and its potential as the triggering factor for the acquisition of the Tense-Aspect system implies that other impairment
factors (both processing and acquisitional) should be invoked in order to explain
how the three dimensions of Tense, Aspect and Actionality are acquired in combination. Thus the claim that L2 Actionality is learned must fully take into account the interaction of those factors and be split into different assumptions:
– L2 Actionality is focalized gradually and it is subject to refinements and
changes over time.
– L2 Actionality is likely to change in a learner‘s mind as long as the whole
tense-aspect system is in reconstruction.
– L2 Actionality could be regarded not as an inherent component of the verb
meaning, but as a separate and – in some cases – delayed learnable component.
– What has to be learned are not only the new values of universal actional parameters (see Slabakova, 2001), of the kind: “sapere ‘know’ in Italian is stative”.
What have to be learned and discovered by learners are firstly and foremost the
properties of L2 Actionality.
– These properties can be detected in terms of gradually changing form-function
mappings through the whole course of interlanguage.
Let’s make an example of two actional properties which are not innate and that
have to be acquired by learners. For example, L2 Italian learners must learn what
happens when a V undergoes the perfective/imperfective shift that can make a V
change its actional content to stative to telic (or the other way round) (Bertinetto,
1986:109). In sentence (1) the English mother-tongue learner does not realize
that using the past perfective ha indossato ‘wore’ instead of the past imperfective
indossava ‘was wearing’ causes the actional content of the verb to shift from stative to telic:
(1) C’era una donna che faceva una passagiata a Venezia sul treno. Ha indossato
occhi di sole verdi
‘there was a woman taking a walk in Venice on the train. She wore [was
wearing] green eyes’ [sunglasses] [L1 English]
A L2 Italian learner must learn whether a V is compatible or not with certain expressions of time. In sentence (2) the expression per un po' di più giorni ‘for a
few more days’ is not compatible with the telic verb ‘arrive’:
(2) Quando arriva per un po’ di più giorni ti telefono
‘when she arrives for some more days I’ll call you’ [when she stays for a few
more days...] [L1 German]
A methodological issue raises here and two positions face. The first position assumes that learners have clear semantic representations, but poverty of vocabulary, L1 pressure, and mistakes due to the performance or other factors lead them
to a wrong lexical choice. If this is the case, only mental representation measures
of L2 Actionality could tap directly learners’ actional competence. Interlanguage
data should not be expected to confirm the theory because it is too noisy
(processing impairments and performance factors might override and blur learners’ actional competence). The alternative position assumes that – in order to reconstruct what’s on a learner’s mind (the aspectual traits of a predicate) – we
can’t disregard or completely bypass the words actually produced by learners.
The data somehow must reflect some kind of competence. According to the data,
the actional competence often performs poorly when ‘put in action’. In fact, performance measures of L2 Actionality are aimed at tracking down indirectly L2
Actionality by scaling down the theoretical assumptions only to what the interlanguage can tell us. In this paper I follow the latter position and I choose to investigate the actual L2 Actionality, not the virtual L2 Actionality. So, instead of
keep on detecting the actional content of L2 verbs, I address the problem of what
kind of Actionality is L2 Actionality. It is worth underlining that all the phenomena I’m going to describe now are scattered through all these learner corpora.
5. The peculiar characters of L2 Actionality
At least five peculiar characters of L2 Actionality can be found through the
analysis of Italian learner corpora. Firstly, L2 Actionality is something hard to
detect in isolation without relying on the presupposed, ontological properties of
the situation described by the verbs. Sentence (3) was written by a Chinese student of Italian and refers to a famous “frog story” scene in which a frog at night
escapes from the glass jar in which it was kept by a boy:
(3) dopo il bambino ha dormito la rana ha provato di scapare
‘After the child has slept, the frog tried to escape’[L1 Chinese].
Even if we know perfectly the scene that sentence number (3) refers to in the frog
story, anyway, we still find it very difficult to interpret correctly the sentence
since it might mean ‘after the child has fallen asleep’ (telic predicate) or ‘while
the child is sleeping’ (activity predicate). Secondly, L2 Actionality is hard to
single out without committing comparative fallacy, that is, without assuming that
the Actionality is the same of that of the target language or of that of English
translation (Ebert, 1995; Tatevosov, 2000; Verkuyl, 2005). Thirdly, L2 Actionality is hard to detect without relying on actional tests designed for mature languages that very often L2 learners fail. This is exactly what happens in sentence
(4) da tre alle quattro abbiamo inisiato a ballare. Fino alle nove siamo cominciati
a ballare
‘From three to four we started to dance. Until nine we started dancing’ [L1
Here in (4) the punctual telic event of ‘starting to dance’ is first improperly
stretched by a durative expression of time and then prolonged by a culminative
adverbial (fino alle nove, ‘until nine’). As a consequence, it’s impossible to actionally label these predicates basing on actional tests. Furthermore, L2 Actionality might also turn out to be something that – especially initial learners – seem
to do without. In fact, it happens that initial learners over-extend the so called basic verbs (Viberg, 2002) as opposed to other verbs which roughly belong to the
same category but which have a different Actionality. In cases like these andare
‘go’ is used instead of: venire, ‘come’, tornare ‘return’, trasferirsi ‘move’, arrivare ‘arrive’, uscire ‘go out’, entrare ‘enter’, volare ‘fly’. For instance, in sentence (5) the verb andare ‘go’ stands for ‘travel’:
(5) quando lui è andato da una città all’altra città
‘When he went from a city to another city’ [‘when he was travelling’, ‘during
the journey’] [L1 Tigrinya]
In cases like the one above, when a verb is actionally underspecified, its actional
content has to be derived inductively and never taken for granted on a lexical basis. Finally, L2 Actionality is something that learners build up at a VP or P level
by-passing V. In sentence (6) the periphrasis ‘to go from Italy to other countries’
stands for the lexical entry ‘leave’ or possibly ‘move’
(6) noi non voliamo andare di Italia per l’altri paesi
‘we don’t want to go of [from] Italy for [to] other countries’ [we don’t want
to leave Italy] [L1 Tigrinya]
If ‘leave’ is the missing word in a learner’s lexical competence, everything that
learners already know comes in useful and nothing is wasted in order to express
the equivalent meaning of ‘leave’ not lexically, but at a syntactic level. in cases
like this, L2 Actionality seems not to be deductable from a supposed kernelmeaning of V. Rather; it is often built by learners starting from a closed set of
‘building blocks’: basic verbs, aspectual morphemes, PP, time expressions.
6. Explanatory adequacy and falsifiability criteria
It is common knowledge that a hypothesis is explanatorily adequate if it links together apparently unrelated grammatical phenomena. In this case, the phenomenon at issue is the auxiliary selection in compound tenses, which in Italian is held
to be the main diagnostic for a verb of being either unaccusative or inergative.
Antonella Sorace (2004) maintains that auxiliaries are learned first for ‘core’ unaccusative and ‘core’ inergative verbs since the choice of either avere ‘have’ or
essere ‘be’ as auxiliaries in these cases is strictly and simply determined by verb
semantics (respectively, telicity and agentivity). The analysis of Italian learner
corpora has shown that the progression in learners’ mistakes follows the predictions of the auxiliary selection hierarchy (see Jezek & Rastelli, 2008). What is
unexpected though is a high percentage of mistakes in core verbs both in beginner and intermediate learners (especially in telic unaccusative core verbs like arrivare ‘arrive’ and fermarsi ‘stop’). Furthermore, in all the corpora, auxiliaries
are omitted very frequently (see Rastelli, in press). The presence of these mistakes and omissions in performance data could be correlated with the fact that L2
Actionality is learned. If this process takes some time, there could exist a period
of latency during which learners have an impaired access to the rules that govern
the selection of auxiliaries. The existence of these rules is not in question here
and neither the existence of an actional blueprint could be disputed on these
grounds. On the other hand though one on no account should expect that these
rules are fully operative and clearly reflected and so recognized in performance
data. Finally, the hypothesis that L2 Actionality is learned undergoes two falsifiability criteria at least. The first one is that performance data are at risk of restricting the analysis to obligatory contexts and to replicate the comparative fallacy mentioned above: as a matter of fact, no comparison between performance
data and competence data is run. The second factor is that the role of the complex
interaction between the Tense-Aspect system in L1 and in L2 is not taken into
account. The risk that L1 semantics may affect heavily learners’ acquisitional
pace of L2 Actionality is partially counterbalanced by the sampling design: in
fact, the phenomena illustrated by examples (1)–(6) are not exclusive of a language or of a set of tipologically close languages. On the contrary, these phenomena hold across very different languages, such as English, Chinese, Tygrinia,
Spanish, and German.
7. Conclusions and further research
To sum up: L2 actional blueprint and L2 actional bootstrapping are two different
facts. The existence of the former does not necessarily imply the existence of the
latter. The analysis of interlanguage can tell us something about the actional
bootstrapping, but very little about the existence of an actional blueprint (more
direct measures of Actionality would be needed). Indirect measures (performance data analysis) allow us to conclude that the role of Actionality should be
scaled down. In fact, initial l2 learners do not seem to know some basic actional
properties, rather they learn them little by little. These properties are learned at
the same time when the whole Tense-Aspect system is learned too. Thus, even if
an actional blueprint exists, it is too weak to trigger the acquisition of the
Tense-Aspect system. Further research will be devoted to collect experimental
data such as eye-tracking, grammaticality judgements, magnitude estimations.
Since these are meant to be more direct measures of actional competence, probably they will contribute to draw the bigger picture.
Alexiadou, Artemis – Anagnostopoulou, Elena & Everaert, Martin, (eds.). 2004. The
Unaccusativity Puzzle. Exploration of the Syntax-Lexicon Interface, Oxford – New
York: Oxford University Press.
Andersen, Roger & Shirai, Yasuhiro. 1994. Discourse motivations for some cognitive
operating principles. Studies in Second Language Acquisition 16. 3–156.
Barni, Monica & Gallina, Francesca (forthcoming). Le parole degli stranieri: il LIPS, il
primo lessico di frequenza dell’italiano parlato da stranieri. Siena: Università per
Bardovi Harlig, Katherine. 2002. Analyzing Aspect. In R.Salaberry & Y.Shirai (eds.).
Bernini, Giuliano. 2005. La seconda volta. La (ri)costruzione di categorie linguistiche
nell’acquisizione di L2. In: L. Costamagna & S. Giannini (eds.). Acquisizione e mutamento di categorie linguistiche. Proceedings of the conference of the Società Italiana
di Glottologia. Roma: Il Calamo. 121–149.
Bertinetto, Pier Marco. 1986. Tempo, aspetto e azione verbale nel verbo italiano. Il sistema dell’indicativo. Firenze: Accademia della Crusca.
Bertinetto, Pier Marco – Lenci, Alessandro – Noccetti, Sabrina & Agonigi, Maria. 2006.
Metodi quantitativi nello studio dell’acquisizione delle strutture tempo-aspettuali.
Quaderni del laboratorio di linguistica 6. Pisa: Scuola Normale Superiore.
Bertinetto, Pier Marco & Squartini, Mario. 1995. An attempt at defining the class of
Gradual Completion Verbs. In: P.M. Bertinetto – V.Bianchi – Ö.Dahl & M.Squartini
(eds). Temporal reference, Aspect and Actionality. Vol. I: Semantic and Syntactic
perspectives. Torino: Rosenberg & Sellier.
Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67–3. 547–
Ebert, Karen. 1995. Ambiguous perfect-progressive forms across languages. In P.M.
Bertinetto – V.Bianchi – Ö. Dahl – M.Squartini (eds.). Vol. 2. 185–204.
Giacalone Ramat, Anna. 1995. Tense and aspect in learner Italian. In: P.M. Bertinetto –
V.Bianchi – Ö. Dahl – M.Squartini (eds.). Vol. 2. 289–309.
Giacalone Ramat, Anna & Rastelli, Stefano. 2008. Learning Actionality: an investigation
on data of L2 Italian. In B. Ahrenholz – U. Bredel – M. Rost-Roth & R. Skiba (eds.).
Empirische Forschung und Theoriebildung. Beiträge aus der Soziolinguistik,Gesprochene-Sprache-Forschung und Zweitspracherwerbsforschung. Eine Festschrift für
Norbert Dittmar zum 65 Geburtstag. Frankfurt: Peter Lang. 239–250.
Housen, Alex. 2002. The development of Tense-Aspect in English as a second language
and the variable influence of inherent Aspect. In R.Salaberry & Y.Shirai (eds.). 155–
Ježek, Elisabetta & Rastelli, Stefano. 2008. Gradiente di Inaccusatività e verbi pronominali in italiano L2. In: G.Bernini – L.Spreafico & A.Valentini (eds.) Competenze lessicali e discorsive nell’acquisizione di lingue seconde. Perugia: Guerra. 79–99.
Lardiere, Donna. 2003. Revisiting the comparative fallacy: a reply to Lakshmanan and
Selinker 2001. Second Language Research 19. 129–143.
Lenci, Alessandro & Zarcone, Alessandra. In press. Un modello stocastico della classificazione azionale. In: Proceedings of the XL international congress of the Società di
Linguistica Italiana. Roma: Bulzoni.
Rastelli, Stefano. 2006. ISA 0.9 – Written Italian of Americans: syntactic and semantic
tagging of verbs in a learner corpus. Studi Italiani di Linguistica Teorica e Applicata
(SILTA) 1. 73–100.
Rastelli, Stefano. 2008. La competenza lessicale-azionale nei dati di un corpus elettronico
di italiano scritto di americani. In E. Cresti (ed.) Prospettive nello studio del lessico
italiano. Firenze: University of Florence Press. Vol. 2. 661–668.
Rastelli, Stefano. 2008. A Compositional Account of L2 Verb Actionality and the Aspect
Hypothesis. Lingue e Linguaggio 2. 261–289.
Salaberry, Rafael & Shirai, Yasuhiro. 2002. The L2 Acquisition of Tense-Aspect
Morphology. Amsterdam–Philadelphia: John Benjamins.
Shirai, Yasuhiro. 2007. The Aspect Hypothesis, the comparative fallacy and the validity
of obligatory context analysis: a reply to Lardiere, 2003. Second Language Research
23–1. 51–64.
Shirai, Yasuhiro & Andersen, Roger. 1995. The acquisition of Tense-Aspect morphology:
A prototype account. Language 71. 743–762.
Slabakova, Roumyana. 2001. Telicity in the second language, Amsterdam–Philadelphia:
John Benjamins.
Sorace Antonella. 2004. Gradience at the Lexicon-Syntax Interface: Evidence from
Auxiliary Selection and Implications for Unaccusativity. In: A.Alexiadou et al., (eds.).
Tatevosov, Sergej. 2002. The parameter of Actionality. Linguistic Typology 6. 317–401.
van Valin Robert. 1990. Semantic parameters of split intransitivity. Language 66. 221–
van Valin, Robert 2005. Exploring the Syntax-Semantics Interface. Cambridge: Cambridge University Press.
Vendler, Zeno. 1967. Linguistics in Philosophy. Ithaca: Cornell University Press
Verkuyl, Henk – De Swart, Henriette – van Hout, Angeliek. 2005. Perspectives on Aspect,
Dordrecht: Springer.
Verkuyl, Henk. 2005. Aspectual composition: surveying the ingredients. In: H.Verkuyl et
al. (eds.). 19–39.
Viberg, Åke. 2002. Basic verbs in lexical progression and regression. In: P.Burmeister –
T. Piske and A.Rohde (eds.). An Integrated View of Language Development: papers
in honor of Henning Wode.Trier: Wissenschaftlicher Verlag. 109–134.
Weist, Richard. 2002. The first language acquisition of Tense and Aspect: A review. In:
R.Salaberry & Y.Shirai (eds). 21–78.
Stefano Rastelli
Department of Theoretical and Applied Linguistics
University of Pavia Carlo Alberto 5
27100 – Pavia
[email protected]
Language changes and language contacts in a
19th century Maritime College and Commercial
Paula Rossi
University of Oulu
1. Introduction
In 1880, a teacher of the Oulu Maritime College wrote to the College Board of
Governors as follows: “In my opinion the students will not benefit at all from being taught in two different languages during one lesson. If it is not economically
feasible to set up parallel classes – one in Finnish and one in Swedish – I recommend that only one language, i.e., the Finnish language, be declared the language
of instruction.”
Even though the official language of instruction at the Maritime College was
Swedish, many of the students were native Finnish speakers. The above citation indicates that during the lessons two languages were used side by side, and
this, in the opinion of this teacher, was not a functional solution. Swedish was
used with Swedish-speaking students and Finnish with those who only spoke
This is one example of the language contacts and language changes that the
speakers of different languages experienced in the 19th century school life of the
city of Oulu, which was one of the most important cities for foreign trade and
shipping in Finland in the middle of the 19th century. It was the administrative,
economic and cultural centre of Northern Finland. Its affluence was based on tar,
timber, salmon and butter trading. The thriving trade and maritime commerce of
the city of Oulu boosted the establishment of industrially important colleges.
Due to the efforts of Oulu burghers, first a Commercial College and then a
Maritime College was founded in the city in 1862 and 1863, respectively.
(Hautala 1976: 523, 526.)
2. The aim, data and method of the study
In this study I will discuss the choices made for the languages of instruction and
administration at the Commercial and Maritime Colleges from their foundation
until the beginning of the 20th century (see Rossi 2007). The colleges were bi- or
even multilingual communities where the representatives of various language
groups met and became involved in interaction.
In the same way as Tampere, Pori and Kotka, the city of Oulu formed a
Swedish-speaking linguistic island. In the 19th century, Oulu had a Finnishspeaking majority alongside an economically and administratively significant
Swedish minority. Besides Finnish and Swedish, other languages had also had a
role in the history of Oulu. When the bi- and multilingualism of the city is investigated, it is important to find out what kind of linguistic communities were
formed by the two prominent educational institutes of the city, the Commercial
College and the Maritime College.
When the colleges started operation, the administrative language of the city
was Swedish, but during the following few decades Finnish gained momentum.
This situation resulted from the legislation of the country and the language debates that waged in the city. Such debates were not only reflected in the city’s
political life but also in the languages adopted by the two colleges. My purpose
is to investigate how the municipal and state-level language-related decisions affected the choice of the administrative language and the language of instruction
in the two colleges. The focal questions of the study include: What was the process like through which the administrative language and the language of instruction of the colleges changed from Swedish to Finnish? What were the consequences of the change of language? What kind of language skills were required
from the teachers and students?
The data used for the study consisted mostly of the minutes and annual reports of the target schools as well as of other documents available in the archives. The documents of the Oulu Commercial College are stored in the City
Archives of Oulu while those of the Maritime College are located in the Oulu
Provincial Archives. According to Hyltenstam and Stroud (1991: 72–74) language change can be studied through different methods, e.g., through conducting interviews and participatory observation. Because neither of these options
could be used in my study, I investigated language contacts through literary
sources (cf. Hyltenstam & Stroud 1991). I have systematically studied archived materials relating to the above two schools and paid particular attention
to those parts of the documents which deal with language issues. This study
presents a content analysis, not a linguistic analysis of the relevant documents.
Language contacts were investigated from the point of view of individual language users (e.g., the language choices of the students are recorded in the annual reports of the colleges), the target schools (e.g., the minutes of College
Board meetings report on the decisions made on the languages of instruction),
and the surrounding society (e.g., the city histories of Oulu give accounts of
the general language situation in the city). This paper constitutes a part of my
wider study on the historical language contacts of the city of Oulu as well as a
part of the city language collaboration project of the cities of Oulu (Paula Rossi) and Tampere (Harry Lönnroth). (See Lönnroth & Rossi 2008a, 2008b, Rossi 2005, 2007.)
3. Theoretical background
This study pertains to the field of historical sociolinguistics with a focus on historical multilingualism. The linguistic description of multilingual communities
such as that of the above two 19th century colleges is based on the following
framing concepts: language contact, language competition, and language conflict. The linguistic climate of a community may be defined by investigating the
nature and volume of the language contacts (when and where speakers of different languages meet each other), language competition (which language is chosen
as the language of communication), and language conflicts (do the languages
collide rather than coexist) that take place in the community. (Tandefelt 2002:
It is typical of multilingual communities that the languages occupy unequal
positions. The dominant language of the community may not, however, be the
language of the majority. The language of a minority is likely to assume a
dominant position if it has, for example, been established as the language to be
used in political and economic contexts. (Tandefelt 1988: 15.)
Multilingual contexts and communities produce various language contacts
which may result in either language shift or language maintenance. In language shift the speakers of one language prioritize another language in some
communicative contexts. In this process, the native language may even become permanently superseded. (Romaine 2000: 51, Johansson & Pyykkö
2005: 14–15.)
The theoretical frame of my study largely draws on contact linguistics. According to Börestam and Huss (2001: 7–8), contact linguistics deals with language contacts, which can be investigated from the perspectives of the individual, the language and society. When language contacts are studied from the
perspective of an individual bi- and multilingualism constitute essential conceptual tools. Bi- and multilingualism may be seen as phenomena with both individual and societal aspects (Romaine 2000: 33). There are several definitions
for bilingualism, for example, the researchers may place varying emphasis on
the knowledge of a language and the pragmatic communicative mastery of a
language. (See, e.g., Hyltenstam & Stroud 1991: 49–52.) A community may
also be bi- or multilingual either through legislation or because its members
speak different languages (Hyltenstam & Stroud 1991: 46–47). When language
contacts are studied from the perspective of the languages involved, the essential concepts include notions such as ‘loan words’, ‘code switching’, and
‘pidgin’ and ‘creole’ languages. The societal study of language contacts focuses on the speakers of different languages who interact with each other within society. ‘Society’ in this context may consist of a political unit, such as a
state, or of a sociological unit where individuals get together to form a community of their own. In the latter case, the choice of language plays an important
role in the creation of the community’s identity. (Börestam & Huss 2001: 66,
67, 110.)
4. Oulu – a Swedish language island
Ever since it was founded in 1605 on the eastern cost of the Gulf of Bothnia by
the Swedish king Charles IX, the city of Oulu has been the centre of this Finnish-speaking region. However, the linguistic pattern of the city became more diverse when the developing city attracted increasing numbers of merchants, artisans, and civil servants with a native language other than Finnish. The majority
of the inhabitants still consisted of Finnish speakers, but the Swedish speakers
constituted a strong and influential minority. (Lönnroth & Rossi 2008b.)
The old merchant class of Oulu was traditionally Swedish-speaking and consequently Swedish was spoken on public occasions (Hautala 1976: 88). In the
middle of the 19th century Oulu had about 6000 inhabitants, 10 % of whom were
Swedish speakers (Salo 2005: 62, Ulfvens 1997: 15). The members of distinguished merchant and bourgeois families had had opportunities to study abroad,
for example in Sweden and Germany. However, the burghers were frequently bilingual and spoke both Swedish and Finnish. It was necessary to have a command of both languages because Swedish was the administrative language and it
was needed in trading. On the other hand, the bulk of the Oulu population was
Finnish-speaking because the city was surrounded by a countryside populated by
Finnish speakers. Moreover, both Finnish and Swedish were part of the Oulu
way of life because many artisans were Swedish-born and many sailors had
learnt Swedish during their travels. (Hautala 1975, 1976, Lönnroth & Rossi
For a long period, Swedish also was the language of administration and education. The use of Swedish was considered a drawback by the Finns and the disagreement resulted in extensive language conflicts. The language conflicts
manifested themselves in the political life of the city as well as in the school system, various associations and the banking world. The newspapers issued in the
city in the late 19th and early 20th centuries were primarily in Finnish but they also
incorporated texts in the Swedish language. This brings evidence for the assumption that the speakers of Swedish occupied a strong position in the city. (Hautala
1976.) According to Kaukiainen (1998: 16), the fact that the Finnish newspaper
of the city of Oulu reported news on weather, annual crop production, and farming but seldom on trade and seafare, which were the origins of affluence, highlights the nature of the existing language border. The native Swedish speakers
directed their attention to the outside world while the Finnish educated class
found greater interest in domestic and local issues.
Although Finland had become a part of Russia in 1809, Swedish was maintained as the administrative and only official language of Finland due to the
country’s earlier history as a part of Sweden. This situation continued until the
year 1863 when Finnish was given an equal status with Swedish and the authorities were obliged to receive official documents and other written materials
in Finnish. Twenty years later all documents also had to be written in Finnish, if
this was requested. (Suomen historian käsikirja 1949: 189–190, Vahtola 2003:
230–231.) The above decisions naturally had an effect on the choice of the administrative language in the city of Oulu.
5. The language of instruction and the linguistic
background of the students
Both of the colleges accepted only male students. The regulations of the Commercial College provided that the students should be proficient in Swedish and
Finnish, but the language of instruction was officially Swedish. This may be one
reason for the fact that the number of students remained low for the first few decades of the college’s existence. During the first academic year only four students
received instruction. The school admitted its first female students in 1871. The
decision of the college board to accept female students was probably not easy because the decision was prohibited from being announced in the local newspapers.
In fall 1871, 7 of the 29 students of the college were girls. All those girls had a
Swedish educational background. This was a year before the first Finnishonly-speaking students were admitted.
In spring 1880 the College Board was concerned about the number of students in the Commercial College: There were only 13 students in the college
during the academic year 1879–1880. Since the foundation of the college up to
the year 1880 only 173 students had attended the college and about a half of
them had graduated. The board decided to enchance the teaching of both Swedish and Finnish. The board’s decision suggests that language problems were
considered to be one significant reason for the small number of students and
the low graduation percentage. Swedish remained still the official language of
The educational practices of the Commercial College were hampered by the
entry of the students who only spoke Finnish. The teachers are reported to have
taught in two languages, in Swedish and in Finnish. Gradually the situation became intolerable; in 1886 a local newspaper, Kaiku (May 15th,1886), criticized
the school and wondered how the school could provide instruction for over 20
Swedish-speaking students when the Oulu region was otherwise primarily Finnish. According to the newspaper the students did not really command Swedish
but only wrote and copied words whose meanings they did not understand. The
principal and the teachers responded by asserting that learning Swedish was the
most urgent concern for the Finnish-speaking students because without knowing
Swedish they would not be able to get work.
The situation changed in 1889 when the College Board decided that instruction should be provided in the language that the majority of the students spoke.
Should any student have problems with the language of instruction, he or she was
allowed to use his or her native language. At that time, most of the students were
Finnish-speaking: Finnish was spoken as the native language by 23 students and
Swedish by 5 students.
In 1891, new bylaws were drafted for the Commercial College. The new bylaws provided that the language of instruction be Finnish because almost all of
the students were native Finnish speakers. The decision incorporated, however,
a provision stating that the students who did not know Finnish were entitled to
be taught in Swedish. Besides the Finnish-speaking students, there were at that
time only one Swedish-speaking and one Norwegian-speaking student in the college. The change in the language of instruction was due to a change in the linguistic background of the students.
Since the foundation of the Oulu Maritime College, two languages of instruction were used in the college. Swedish was the official language of instruction, but those students who did not understand Swedish were taught in
Finnish. For example in 1879, 13 students were taught in Swedish while 5 students received instruction in Finnish. Because both Swedish and Finnish were
used during the same lessons, the academic years often lasted longer than was
provided by the bylaws. In addition, there were separate lessons for the speakers of Swedish and Finnish. The principal of the college advocated in 1880
that the school be assigned one language of instruction. At that time there
were in the college Finnish-speaking students who did not understand Swedish and also students who had not mastered Finnish. The latter students were
not from the college’s proper region of influence. The Oulu Maritime College
also had students from other parts of the country, e.g., from Southern Finland
and Åland. This situation lasted for several years. When the College Board
again discussed the language of instruction in spring 1899, 13 of the college’s
28 students were taught in Swedish while 15 of them received instruction in
Finnish. From the Swedish-speaking students only one was from the Province
of Oulu. The board decided that it was justified in choosing Finnish as the language of instruction in the Maritime College of a Finnish-speaking neighbourhood.
The aim of a single language of instruction was reached as a results of the College Board’s determined efforts to that end. In 1899, 10 years later than in the
Commercial College, Finnish was made the language of instruction in the Maritime College. This means that for the whole of the 19th century, both Finnish and
Swedish were used for instruction in the Maritime College. The year 1891 made
an exception because then instruction was provided only in Finnish due to the
students’ native-language background. The language issue was considered important: for example in the annual reports of the Maritime College, the students
who received instruction in either Finnish or Swedish are separately mentioned
by name.
According to the annual reports of the Maritime College, Finnish-speaking
and Swedish-speaking students alternated in constituting a majority. The table
below shows the numbers of students receiving instruction in only Swedish or in
only Finnish in the 1880’s and 1890’s.
Table. The numbers of students receiving instruction in only Swedish or in only
Finnish in the Maritime College of Oulu in the 1880’s and 1890’s
Spring term
6. The language skills of the teachers
Many of the Maritime and Commercial College teachers were able to teach in
both Swedish and Finnish. They also had occasionally had living experience of
more international languages than Swedish and Finnish. One of these teachers
was Aug. Ernst Biese from Germany. He was elected the first principal of the
Commercial College in 1864. He had studied economics in Rostock and then
moved to Sweden. His curriculum vitae indicates that he had taught languages
and economics in the cities of Kalmar and Norrköping. He knew at least German,
Swedish and English. His election to the principal’s post was not, however,
unanimous because he did not speak Finnish, the language of the majority of the
population living in the Oulu region. Mr. Biese, Principal of the Commercial
College, also taught English and accounting in the Maritime College. Obviously,
he never learnt Finnish even though he lived in Oulu for several decades. For instruction he only used Swedish, and accordingly his Finnish students had to attend classes taught in Swedish. According to the documents of the Maritime College, the English skills of his students were tested through English-Swedish and
Swedish-English translations. In the Commercial College, the language of instruction was officially changed from Swedish into Finnish only after the death
of Mr. Biese in 1889. On the basis of documentary evidence it can be inferred
that the use and status of the Finnish language in the Commercial College was
closely connected to the stand of Mr. Biese, the principal.
The election of the first teacher of the Maritime College, Alexander Cannelin,
on the contrary, was based on his ability to speak Finnish. His good command of
Finnish was highlighted as an important qualification that supported his election.
Even though the significance of Finnish skills increased in the 20th century, a
knowledge of Swedish still remained an important qualification for college
faculty. When a new principal was elected for the Commercial College in 1913,
it was stressed that oral and written skills in both Finnish and Swedish be required from the new holder of the office.
The female teachers of the colleges also commanded several languages. One
of the English teachers of the Commercial College, Elin Liljeblom, had graduated from a Swedish high school in Oulu, taught at several schools in Helsinki
and made excursions to England. When she applied in 1891 for an English
teacher’s post she had written her application in Swedish. According to the application, she new both Swedish and Finnish but other documents suggest that
she had a stronger background in Swedish. Like the Commercial College
faculty, the Maritime College teachers also had experiences of foreign countries because many of them had worked in various capacities on board international ships.
7. The language of college administration
Because the administrative language of the surrounding society was Swedish,
the official documents of the colleges, the minutes and annual reports, were written in Swedish. The boards of the colleges consisted of respectful citizens such
as the provincial governor and merchant burghers whose native language was
Finnish was first introduced for administrative purposes in the Commercial
College. At the Commercial College, the administrative language was Swedish
for the whole term of the first principal, Mr. E.A. Biese. From the year 1891 onwards, the annual report of the college was published in both Finnish and Swedish. This change took place on the initiative of the new principal, Santeri Dahlström. Similarly, the college bylaws from the same year were written in both languages.
At the same time, the role of Finnish also increased in town administration.
In Oulu, Swedish-speaking authorities were increasingly considered a drawback by the Finnish-speaking citizens who constituted a majority of the Oulu
inhabitants. The turn of the year 1893–94 became a milestone for the establishment of the Finnish language in the city administration of Oulu. After abundant
and even fierce discussions, Finnish was chosen to be the language of minute
taking for town council meetings. The decision, which came into effect as of
the beginning of the year 1894, was not unanimous: while 15 members of the
town council voted in favour of Finnish, 6 voted against it. Some members of
the board of the Commercial College also sat in the town council and some of
them voted for Finnish while others voted against it. The new principal, on
whose initiative college annual reports were published in Finnish, was among
the proponents. (Westerlund 1975: 49–50, Rossi 2007: 312.) The decision also
had a wider effect. Because Oulu was the capital of the province, Finnish became the official language of the whole province. (Westerlund 1975: 50.)
However, in spite of many obstacles, the Swedish-speaking citizens had a
strong hold over the town’s political life until the first decade of the 20th century (Hautala 1976: 350).
The above-mentioned town council decision had a clear effect on the choice
of the administrative language at the two colleges. The first annual report of the
Maritime College that was written in Finnish was published in the same year,
1894. Finnish also took over as the language of the minutes in the Commercial
College where the annual reports had already been drafted in both languages for
the previous three years. The first Finnish Maritime College minutes, on the
other hand, are from the year 1896. However, in the archives of the Maritime
College, annual reports and minutes can be found which even after 1896 were
written both in Swedish and Finnish. The same is also true of the official letters
sent from the Maritime College.
The annual reports of the Commercial College were printed in both languages
for several years. In 1913, the principal of the college, Santeri Dahlström, proposed that the annual reports be printed only in Finnish. The board did not accept
the proposal: 4 members voted for bilingual reporting and 2 supported the use of
only Finnish in the annual reports. However, in 1915 Finnish took over in annual
reporting because, in the opinion of the principal, Otto Valmari, there was no
longer any demand for Swedish reports and accordingly bilingual printing of the
annual report would have been a waste of resources. This time the proposal was
not objected. In this manner, the economic situation of the colleges and world
history also affected the language policy that each of them conducted.
8. The languages taught at the Maritime and Commercial
The merchant burghers of Oulu had good trade relations with both the commercial houses of Stockholm in Sweden, and those of other countries. For this reason
it was important that the future business people at both the Commercial and
Maritime College learnt foreign languages.
In the 1890’s, when the language of instruction in the Commercial College
was Finnish, the weekly number of lessons in the Swedish language was considerably greater than that in the Finnish language. If the college graduates
wanted to pursue a successful career in business, they had to be proficient in
Swedish. English and German were originally included in the curriculum of the
Commercial College. The teaching of Russian was initiated in 1892 and half
of the students then chose the subject. The choice of Russian was probably due
to the fact that since the year 1891 civil servants were required to speak Russian. Those who had a good command of the Russian language were given a
priority when governmental posts were filled. At the end of the 19th century,
however, interest in Russian studies declined as a result of the russianising
policy of the Russian government. In spite of this objection Russian became a
compulsory school subject in 1904. The decision was opposed because it entailed that the students who were poor at languages and not able to study several languages side by side could study only Russian. The decision was overruled but Russian became a compulsory school subject again in 1915. (See also
Herva 1965.)
For the first decade of the 20th century, Swedish enjoyed a strong position in
the Commercial College. All students wanted to study Swedish. German became
more popular in the Commercial College than English in spite of the fact that
England was Oulu’s most important export trading partner (see Hautala 1976:
In the Commercial College the aim of language teaching was that the students
should learn to read and write easy business language needed in trading. The importance of practical exercise and a knowledge of grammar was stressed. (Herva
1965: 78, Rossi 2007: 313.)
At the Maritime College, English had been taught since the foundation of the
college. The documents pertaining to the first years of the Maritime College’s
operation demonstrate that the students’ knowledge of English was poor and
consequently the number of classes had to be increased from what it had previously been. English teaching involved familiarization with the rules of pronunciation and grammar. The students translated Swedish and Finnish texts
into English and vice versa. Many students had already travelled on board
ships to foreign countries and learnt the basics of English. An effort was made
to systematize the students’ fragmentary knowledge of English by means of instruction.
The need for Russian studies was introduced at the end of the 1870’s. At that
time Finnish sea captains and mates had good work opportunities on board Russian ships. A precondition for being employed was, however, at least a satisfactory knowledge of Russian. Thus the students of the Maritime College submitted
an application to the College Board demanding instruction in the Russian language. They wanted to have instruction in the basics of Russian to be able to continue developing their skills later through independent study and to have, consequently, better chances of being employed. The college board took a positive
stance towards the application: It could well be supported because trade with
Russia had increased and further increase was expected.
In the 1880’s, Russian became an optional subject. The popularity of Russian
increased towards the end of the century, and even former graduates returned to
study Russian at the Maritime College. In the academic year 1897–1898, 18 of
the school’s 20 students studied Russian. During the first few years of the 20th
century all students of the Maritime College studied Russian as an optional sub292
ject. According to the annual reports, after having completed the course the students were able to translate with the help of a dictionary relatively easy Russian
texts into Finnish.
9. In conclusion
The written materials that were available for the present researcher indicate that
the official language of instruction at both of the colleges was originally Swedish. In the Maritime College, however, it was possible for students who did not
speak any Swedish to get instruction in Finnish. The number of these students
was, however, very low during the first years of the Maritime College’s operation.
During the latter half of the 19th century the role of Swedish decreased in the
city administration of Oulu. In 1894 Finnish became the administrative language
of the city. This was not only due to the general language policy of the country
but also to the great number of the Finnish-speaking inhabitants of the city. Many
old Swedish commercial houses got into financial difficulties and, consequently,
the influence of the Swedish bourgeois class weakened in the city. The city became a scene for language conflicts which also had an effect on the schools including the two colleges that were the target of the present study. The Swedish
language gave way for Finnish as a language of both instruction and school administration. (Hautala 1976, Lönnroth & Rossi 2008b.)
It is difficult to reconstruct a picture of the languages used for real communication in the two colleges. The materials available for the present researcher do not reveal the nature of the real communication that was conducted by the students and staff of these colleges. One can only guess that the students may have switched language during their school day. This may have
been the case with those students whose native language was Finnish and
who had to attend classes taught in Swedish. Similarly, the Swedish-speaking students probably could not avoid hearing and even using Finnish within
the college community.
At least many teachers used both Finnish and Swedish for instruction. The
documents included in the data of this study show evidence that the teachers used
both languages for teaching even during the same lesson. For instance, in a newspaper article (Kaiku 15.5.1886) one of the teachers of the Commercial College,
Mr. Snellman, was praised for having used, in consideration for the Finnish students, Finnish alongside Swedish for instruction.
Even though Finnish became the official language of the colleges, Swedish
was still highly appreciated, especially in the Commercial College. Neither of the
colleges could overlook Russian because Finland was a part of the Russia.
Finally, it is worth mentioning that teaching through two languages inspired a
vivid discussion about the teachers’ salaries. In 1888, the Maritime College was
allocated an additional sum of 1200 marks to be paid to teachers who regularly
taught in two languages during the school year. Captain Ekholm claimed the
whole sum because he was the only teacher who had provided support lessons in
Finnish. Captain Lundelin also claimed compensation because he had taught
classes bilingually using both Finnish and Swedish. The College Board could not
reach a decision immediately. At the next meeting it decided, however, that Captain Ekholm should get 900 marks of the additional revenue while Captain Lundelin was to receive 300 marks. Thus, providing support lessons in the student’s
mother tongue was considered to be a more valuable activity than teaching a lesson bilingually.
Börestam, Ulla & Leena Huss. 2001. Språkliga möten. Tvåspråkighet och kontaktlingvistik. Lund: Studentlitteratur.
Hautala, Kustaa. 1975. Oulun kaupungin historia 3. 1809–1856. Oulu: Oulun kaupunki.
Hautala, Kustaa. 1976. Oulun kaupungin historia 4. 1856–1918. Oulu: Oulun kaupunki.
Herva, Yrjö. 1965. Oulun kauppaoppilaitos 1864–1964. Oulu: Oulun kauppaoppilaitos.
Hyltenstam, Kenneth & Christopher Stroud. 1991. Språkbyte och språkbevarande. Om
samiskan och andra minoritetsspråk. Lund: Studentlitteratur.
Johansson, Marjut & Riitta Pyykkö. 2005. Johdanto: monikielisyys ja kielipolitiikka. In
Marjut Johansson & Riitta Pyykkö (eds.), Monikielinen Eurooppa. Kielipolitiikkaa ja
käytäntöä, 9–26. Helsinki: Gaudeamus.
Kaiku 15.5.1886.
Kaukiainen, Yrjö. 1998. Laiva Toivo, Oulu. Helsinki: Suomalaisen Kirjallisuuden Seura.
Lönnroth, Harry & Paula Rossi. 2008a. Språken i föreningsspråk. Finländska fruntimmersföreningar som språkliga miljöer på 1800- och 1900-talen. In Marianne Nordman, Siv Björklund, Christer Laurén, Karita Mård-Miettinen & Nina Pilke (eds.),
Svenskans beskrivning 29. Förhandlingar vid Tjugonionde sammankomsten för
svenskans beskrivning, 187–194. Vasa: Svensk-Österbottniska samfundet r.f.
Lönnroth, Harry & Paula Rossi. 2008b. Språkhistorien i stadshistorien. Språk i kontakt i
Tammerfors och Uleåborg. In Jan Lindström, Pirjo Kukkonen, Camilla Lindholm &
Åsa Mickwitz (eds.), Svenskan i Finland 10. Föredrag vid tionde sammankomsten för
beskrivningen av svenskan i Finland, 197–211. Helsingfors: Institutionen för nordiska
språk och nordisk litteratur.
Romaine, Suzanne. 2000. Language in Society. An Introduction to Sociolinguistics.
Oxford: Oxford University Press.
Rossi, Paula. 2005. Språkfrågor i handelssocieteter. In Kristina Nikula, Harry Lönnroth,
Kaisa Alanen & Carl-Eric Johansson (eds.), Svenskan i Finland 8, 234–246. Tammerfors: Nordiska språk, Institutionen för språk- och översättningsvetenskap, Tammerfors
Rossi, Paula. 2007. Handelsskolan i Uleåborg – en flerspråkig skolmiljö. In Nina Niemelä
& Esa Lehtinen (eds.), Översättningsteori, fackspråk och flerspråkighet. Vakki-symposium XXVII, 305–316. Vasa: Vasa universitet. Forskargruppen för översättningsteori, fackspråk och flerspråkighet.
Salo, Matti. 2005. Oulu laivojen ja tervan vuosisadalla (1809–1879). In Reija Satokangas
(ed.), Oulun vuosisadat 1605-2005, 60–80. Oulu: Pohjois-Suomen Historiallinen
Suomen historian käsikirja II. 1949. Helsinki: Suomalainen Tiedeakatemia.
Tandefelt, Marika. 1988. Mellan två språk. En fallstudie om språkbevarande och språkbyte i Finland. Uppsala: Acta Universitatis Upsaliensis.
Tandefelt, Marika. 2002. Fyra språk på en marknad. In Marika Tandefelt (ed.), Viborgs
fyra språk under sju sekel, 11–67. Helsingfors: Schildts.
Ulfvens, Levi. 1997. Elin och Karl Pentzin. Donatorer och kulturfrämjare. En minnesteckning. Helsingfors: Svenska Kulturfonden.
Vahtola, Jouko. 2003. Suomen historia. Jääkaudesta Euroopan unioniin. Helsinki: Suuri
Suomalainen Kirjakerho oy.
Westerlund, Jorma. 1975. Oulun kaupunginvaltuusto 1875–1975. Piirteitä kaupunginvaltuuston satavuotisen toiminnan ajalta. Oulu: Oulun kaupunki.
Paula Rossi
Nordic Philology
Faculty of Humanities
University of Oulu
BOX 1000
FIN-90014 University of Oulu
[email protected]
Conceptual contrast of dimensional adjectives
in Japanese and Swedish:
Exploring the mental lexicon by
word-association test
Misuzu Shimotori
Umeå University
1. Introduction
This paper discusses the semantics of dimensional adjectives that basically refer
to three-dimensional extensions of objects and between objects, such as high,
long, and far. The basic question in the present essay is how dimensional adjectives are represented in the mental lexicons of native speakers of Japanese and
Swedish. My aim is to explore, by using a word-association test, the question of
whether there are any conceptual contrasts between the two languages in the understanding of dimensional adjectives.
The semantics of dimensional expressions is said to be universal (Dixon
1982). Indeed dimensional expressions are widely and frequently used in our
daily communication. Usages of dimensional expressions seem to be common
and necessary across languages. When used literally, for instance, they intend to
express the external appearances of concrete objects that we perceive in the
world. In metaphorical usages our conceptual image of size is transferred to other
semantic domains such as time extensions, valuations of one’s work and so on.
Concerning linguistic form, dimensional expressions are lexicalized in different word classes. In English, for instance, we see at least dimensional verbs, e.g.
expand and enlarge, dimensional nouns, e.g. size, height, and dimensional adjectives, e.g. high, long and big. In the same way Japanese and Swedish dimensional
expressions are lexicalized in varying word classes. In the present essay, however, I focus on adjectival expressions which refer to objects that are placed in
three-dimensional space.
2. Word-association test
A word-association test was conducted in order to tap subject’s understanding of
dimensional adjectives. Word-association tests were originally used in the field
of psychology, as a methodology for exploring individual idiosyncrasies. The
most common way of procedure is to ask for the first word that comes up in the
subject’s mind when s/he hears the stimulus word. This methodology has been
used frequently in the field of linguistics as well, especially in the study of bilingualism and the process of language acquisition. In the field of bilingualism
word-association tests are used, for instance, to measure the development of
mental lexicons in a second language and to explore the differences of representations of words in mental lexicons between native speakers and non-native
speakers, or between adults and children.
In my view the word-association test is a suitable method for exploring how a
dimensional adjective is represented in our mind. The notion of ‘representation
of words in the mind’ indicates our ways of organising and remembering words
in a lexical network. It is assumed that a word is not left on its own in the lexicon.
A word has multiple relations to other lexical entities. When a language speaker
hears a word, then s/he can almost simultaneously relate that word to another
concept. This connection is made by various types of association links. I believe
that the word-association test is a good tool for revealing interrelationships of
words in the mind. To sum up, I quote a statement by Clark (1975): “Unlike conversation or the other language games we play daily, the word association game
is an artificial, derivative phenomenon, important not because it is interesting in
itself, but because it reveals properties of linguistic mechanisms underlying it.
Our ability to produce associations is presumably derived from our ability to understand and produce language. For this reason, language must play a central role
in the explanation of these associations”. Exploring the associations in depth is
likely to reveal aspects of the lexical network in the mind, both in general and in
specific ways.
Aitchison talked about the lexicon as ‘a gigantic multi-dimensional cobweb,
in which every item is attached to scores of others’ (1987:72). She analyzed word
association types based on the word-association study conducted by Jenkins
(1970). The most important associative links in the word-web are co-ordination
(e.g. salt – pepper), collocation (e.g. salt – water), superordination (e.g. red –
colour) and synonymy (e.g. hungry – starved). Conventionally the types of association responses are categorized into three main classes. These are syntagmatic
links, paradigmatic links and phonological links. Fitzpatrick (2006) addressed
some problem areas in the categorisation of association responses into those
three classes. She proposed a new classification of association responses, which
are based on four main categories: meaning-based responses (i.e. those determined by semantic characteristics), position-based responses (determined by
syntactic and collocational characteristics), form-based responses (determined
by phonological, orthographical or morphological characteristics), and erratic re297
sponses (where no link between cue and responses was apparent, or where no response at all was given) (Fitzpatrick 2007). Lexical linkages which are elicited
by association tests have different degrees of associated strength1. There are very
likely and common associations which are rapidly elicited and made frequently
by language users. On the other hand there are uncommon and very individual
associations which often are affected by the speaker’s own experience. Such relations have relatively weak associated strength.
Many findings have been reported in previous research on word-associations,
and the responses have been analysed from various perspectives. For a long time
it was believed that there is a marked linguistic phenomenon which is called the
syntagmatic-paradigmatic shift. This shift implies that ‘young children often respond with a word normally following the stimulus word in a sentence
(GO-HOME), whereas older children and adults frequently respond with a replacement word (GO-WENT)’ (Entwisle et al. 1964). This hypothesis was supported by researchers such as Brown & Berko (1960), Ervin (1961) and Entwisle
et al. (1964). However this hypothesis has met with counterexamples in later
studies (Söderman 1993, Nissen and Henriksen 2006). In their study on word
class influence on word associations, Nissen and Henriksen (2006) concluded
that nouns seem to be predominantly organized in paradigmatic relations, whereas verbs and adjectives are characterised by syntagmatic relations. They claim
that word class type clearly affects the test results.
Different views have been expressed on the methodology ofword assotiation
tests. It is clear that different ways of conducting the study may draw different
results. Concerning the response time there are some discussions of the time effects on the responsesobtained. Clark (1975) stresses the importance of the fastest responses in order to get the common associations, i.e. the responses other
people are most likely to give. If an informant is allowed to take his time, he normally reacts with his own imaginations, experiences and memories. It results in
giving idiosyncratic, individual and personal responses.
3. Overview of the study
The word-association tests were administered in the form of a questionnaire
which was sent via e-mail to the informants. The response time was not limited.
The stimuli words were given in the written form, so the informants got only
visual stimuli. Thirty native speakers of Japanese who live presently in Japan and
30 native speakers of Swedish who live presently in Sweden participated in the
test. They are both men and women in the age range from 17 to 60. All were
educated at least including high school. Most of them are employed. All informants were asked to answer by giving three associations for each stimulus word.
The term associative strength is quoted from Deese (1965: 14).
3.1. Stimulus items
The well-known stimulus for word-association test is Kent and Rosanoff’s
(1910) list of 100 very common English words. The stimuli included both nouns
and adjectives. The word’s commonality is measured by its frequency. This
stimuli list has been used by many researchers in previous studies (Söderman
1993, Namei 2002). In the present study the stimuli are basic dimensional expressions in Japanese and in Swedish. The Japanese dimensional expressions
consist of 14 adjectives and the Swedish dimensional expressions consist of 14
adjectives as well. Table 1 illustrates dimensional expressions in English and
their correspondences in Japanese and Swedish.
Table 1: Dimensional adjectives in English, Japanese and Swedish
Wienold and Rohmer (1997) insisted that lexicalization of degree of extension
includes elative lexicalizations ‘exhibiting an extension to a very high degree’,
e.g. E. huge, Germ. riesig ‘huge’ or winzig ‘very small’. Unsurprisingly there are
such elative lexicalizations both in Japanese, e.g. dekai ‘big’ and grov ‘big’ in
Swedish. However I eliminate such expressions in my study. Instead I use basic
dimensional adjectives which are highly frequent words in both languages. They
should be semantically neutral, i.e. without any connotation of very, too and so.
3.2. Expressions of DISTANCE
Many studies on dimensional expressions include the category of DISTANCE as
a dimensional expressions (Wienold and Rohmer 1997, Kushima 2001). It is
clear that DISTANCE words express a dimensional extension of both concrete
and abstract distance between objects. However I excluded DISTANCE words
from my stimuli because this study concerns only adjectives. DISTANCE words
are lexicalized differently in the surveyed languages. Two Japanese expressions
of DISTANCE are given that are adjectives, i.e. tôi ‘far’ and chikai ‘near’. Yet
in Swedish the word class of dimensional expressions is varied. There is one adverbial phrase långt borta ‘far’, one adjective avlägsen ‘far’ and one adverb fjärran ‘far’.
3.3. The first assumption
Based on previous studies there are two opposite association pattern I could anticipate. If we posit the hypothesis of word class influence which is examined by
Nissen and Henriksen (2006) (see 2 above), it can be assumed that dimensional
adjectives are associated mostly with nouns. The greater part of the responses
would be nouns which are modified by dimensional adjectives, e.g. high associated with building. However, contrary to the hypothesis of word class influence,
most associations must be adjectives if we posit the syntagmatic-paradigmatic
shift. According to Entwisle (1966) the shift will take place between 6 to 8 years
of age. My informants consist of 60 adults, so the expected associations would
consist of paradigmatic relations to dimensional adjectives, so that high might
associate with low for instance.
4. Result
All in all, I got 1260 associations from Japanese informants and 1242 associations from Swedish informants. Some informants did not give three associations
to each stimulus, and missing responses were not counted. In Table 2 the results
are analyzed depending on their word class.
Table 2: Word class of associations in L1
Word class
% of Total
% of Total
The first finding we see in Table 1 is that the proportion of each word class to the
total are different in the two languages. More than 99% of the Japanese associations are nouns. On the other hand Swedish associations are given both as nouns
(62,30%) and adjectives (36,20%). Verbs and other word classes like adverbs are
very few in Swedish, e.g. smal ‘thin’ associates with banta ‘go on diet’, but nonexistent in Japanese. Table 3 below exemplifies some types of associations to
nouns elicited from Japanese informants.
Many Japanese associations refer to concrete objects which have noticeable
extension in three-dimensional space. Semantically, as well as syntactically, they
are quite commonly-used combinations. Abstract nouns are associated with dimensional adjectives interpreted metaphorically. They refer to temporal aspect
of extensions, e.g. nagai ‘long’ associates with jikan ‘time’ and our cognitive activities, which normally are given a positive and/or negative estimation. For instance hiroi ‘large, wide’ associates with chishiki ‘knowledge’ and the phrase hiroi chishiki is understood as a positive judgment by native speakers of Japanese.
To conclude I found the modifier-modified relation in most cases of Japanese as300
Table 3: Association types of nouns elicited from Japanese native speakers
Associations of noun
concrete object
abstract object
cue word
takai ‘high’
yama ‘mountain’
nagai ‘long’
fukai ‘deep’
kawa ‘river’
umi ‘sea’
body part
nagai ‘long’
futoi ‘thick’
kami ‘hair’
ashi ‘legs’
hiroi ‘large’
nagai ‘long’
ie ‘house’
hashi ‘bridge’
hiroi ‘large’
asai ‘shallow’
chishiki ‘knowledge’
kangae ‘thought’
semai ‘narrow’
fukai ‘deep’
kokoro ‘heart’
aijô ‘love’
sociations. In contrast, Swedish associations are of several different types compared to Japanese associations. They are nouns, adjectives, verbs and other word
classes such as adverb. The associations to nouns show quite a similar pattern to
the Japanese examples in Table 3. They refer to both concrete and abstract objects. What is very interesting here is the distribution of adjectival associations.
Table 4 below illustrates different types of adjectival associations.
Table 4: Association types of adjectives elicited from native speakers of Swedish.
Associations of adjective
Dimensional adjectives
cue word
stor ‘big, large’
liten ‘small, little’
hög ‘high’
lång ‘long’
låg ‘low’
kort ‘short’
stor ‘big, large’
enorm ‘enormous’
bred ‘broad, wide’
tunn ‘thin’
vid ‘wide’
mager ‘lean’
hög ‘high’
lång ‘long’
stor ‘big, large’
lång ‘long’
tjock ‘thick’
stor ‘big, large’
Adjectives with related
stor ‘big, large’
bökig ‘untidy, tiresome’
djup ‘deep’
smal ‘narrow, thin’
mystisk ‘mysterious’
snygg ‘tidy, attractive’
Colour terms
stor ‘big, large’
röd ‘red’
djup ‘deep’
liten ‘small, little’
svart ‘black’
grön ‘green’
Adjectival associations are categorized into antonyms, synonyms, dimensional
adjectives, other adjectives with related meaning, and most interestingly, colour
terms. Antonyms and synonyms account for a large share of the total of adjective
associations. Dimensional adjectives are normally lexicalized as antonymous
pairs. A pair consists of a positive term, e.g. lång ‘long’ and a negative term, e.g.
kort ‘short’. It is a very common combination in word association test. Jenkins
(1970) showed that his informants nearly always associated to the partner if the
stimulus item is one of a pair, e.g. king – queen, or has an obvious opposite, e.g.
high – low.
Swedish dimensional adjectives are sometimes associated with another member of a dimensional pair as well, e.g. hög ‘high’ associates with lång ‘long’ and
grund ‘shallow’ associates with låg ‘low’. It is worth noticing that when a dimensional adjective is associated with another dimensional adjective except its
antonym, it is always with an adjective with the same markedness, so that an adjective with positive evaluation is associated with another positive term, and the
negative term is associated with another negative term.
Adjectives with a meaning related to dimensional adjectives are also found,
e.g. stor ‘big, large’ associates with bökig ‘untidy’. This kind of sense relation is
not normally included in our existing dictionaries. It is apparently established by
a language user’s own experience. Hence the associations sometimes have a very
subjective aspect and are not acceptable in a general way. Among all the responses by Swedish informants the most fascinating associations were colour
terms. It is probably one form of synaesthesia, i.e. a neurologically based phenomenon in which stimulation of one sensory or cognitive pathway leads to automatic, involuntary experiences in a second sensory or cognitive pathway (from
Wikipedia, The common connections in this field
are known as grapheme colour synaesthesia or number colour synaesthesia. The connection between a dimensional adjective and a colour term is, however, not exactly made involuntarily. We could call it a synesthetic relation, in
the way that some metaphorical expressions use cross-sensory transforming, e.g.
an icy stare, dark secret and bittersweet memory.
5. Discussion
In this essay I have described a word-association test and tried to figure out
how Japanese and Swedish dimensional adjectives are represented in mental
lexicons. The result shows that there is a clear conceptual contrast between the
two languages. Japanese informants associate dimensional adjectives mostly
with nouns (99,40%) which normally are combined with dimensional adjectives that describe their spatial size. The nouns refer both to concrete and abstract objects as shown in Table 3. The results confirm the statement that adjectives are characterised by syntagmatic relations, in other words, that adjectival stimuli are predominantly associated with nouns, as discussed by Nissen
and Henriksen (2007). In addition the present results support the notion by
Clark (1975) who talked about the selectional feature realization rule, which
means that ‘the list of features for a word often contains selectional features
that partially characterize the meaning of the potential context of that word’.
This means that the adjective young has selectional restrictions on what it can
modify and thus its associations can be for instance boy, girl and child. Japanese dimensional adjectives did specify the nouns they modify, in more than
99% of the syntagmatic responses.
Swedish informants, on the other hand, associate dimensional adjectives with
both nouns (62,30%) and adjectives (36,20%). Like the Japanese associations,
the associated nouns in Swedish are normally combined with dimensional adjectives when their three-dimensional sizes are described. The associated adjectives
can be categorized into antonyms, synonyms and other adjectives that have related meanings in the native speaker’s mental lexicon. This result shows that
Swedish associations violate the hypothesis by Nissen and Henriksen (2007) of
word class influence.
As a result, the findings from the present word-association test is an apparent
counterexample to the hypothesis of a syntagmatic-paradigmatic shift. According to this hypothesis, the adult native speakers who have good competence in a
language should associate words paradigmatically. In my results, however, both
Japanese and Swedish informants associate dimensional adjectives with a syntagmatic relation in a fair percentage of the total responses.
Besides, my results do not apply to any hypothesis of association pattern
which has been discussed before. This is simply due to the fact that association
patterns of dimensional adjectives are clearly organized differently in the mental
lexicons of the native speakers of Japanese and Swedish.
Why is it so? Is this contrast language internal or culturally grounded?
Aitchison uses the term ‘linguistic habits’, which form the links between words
(1987:73). With the use of this term I should add another question to future
studies. What is lying at the root of those linguistic habits? What distinguishes
the association patterns of Japanese and Swedish linguistic habits? Fitzpatrick
(2007) studied the association pattern of L1 and presented varying types of associations elicited from native speakers of English. On this basis she expressed
doubts of the studies which refer to ‘native-speaker-like’ responses. It is possibly true that word associations can be changed easily by context, and that
some responses are nonpredictable. Therefore it is possibly wrong to assume
that we can ever lay down fixed and detailed pathways that link specific words
in the mental lexicon. However from this study on the associations of dimensional adjectives there is considerable contrast between Japanese and Swedish
associations. A further question is whether this is a unique phenomenon or not.
This is at present an open question. In future studies more data sets and a detailed analysis of the conceptual contrasts in mental lexicons of different languages are required.
Aitchison, Jean. 1987. Words in the Mind. An Introduction to the Mental Lexicon. Oxford.
Brown, Roger., and Berko, Jean. 1960. Word association and the acquisition of grammar.
Child Develpm, 31, 1–14.
Clark, Herbert H. 1975. Word associations and linguistic theory. In New Horizons in
Linguistics. John Lyons (Ed.), Penguin Books
Deese, James. 1965. The structure of associations in language and thought. Baltimore :
The Johns Hopkins Press.
Dixon, Robert M. W. 1982. Where Have All the Adjectives Gone?: and other essays in
semantics and syntax. Berlin, Mouton.
Entwisle, Doris R., Forsyth, Daniel F., and Muuss, Rolf. 1964. The syntactic-paradigmatic shift in children’s word associations. Journal of verbal learning and verbal
behaviour 3, 19–29.
Ervin, Susan M. 1961. Changes with age in the verbal determinants of word association.
American Journal of Psychology 74, 361–372.
Fitzpatrick, Tess. 2006. Habbits and rabbits. Word associations and the L2 lexicon.
EUROSLA Yearbook, 6, 121–145.
Fitzpatrick, Tess. 2007. Word association patterns: unpacking the assumptions. International Journal of Applied Linguistics, Vol. 17, No. 3.
Jenkins, J.J. 1970. The 1952 Minnesota word association norms. In L. Postman and G.
Keppel (eds.), Norms of word association. London: Academic Press.
Kent, G. H. and A. J. Rosanoff. 1910. A study of association in insanity. American Journal of Insanity 67: 37–96, 317–90.
Kushima, Shigeru. 2001. Mono to basho no tairitsu [Conflict between things and places].
Kuroshio shuppan, Tokyo.
Namei, Shidrokh. 2002. The Bilingual Lexicon from a Developmental Perspective. A
word association study of Persian-Swedish bilinguals. Doctoral Dissertation, Stockholm University, Sweden.
Nissen, Henriette Bagger. and Henriksen, Birgit. 2006. Word class influence on word association test results. International Journal of Applied Linguistics. Vol. 16, No.3
Söderman, Tove. 1993. Word associations of foreign language learners and native speakers – different response types and their relevance to lexical development. In
Hammarberg, B. (ed.) Problems, process and product in language learning. Åbo:
Wienold, Götz. and Rohmer, Ulrich. 1997. On implications in lexicalizations for
dimensional expressions. In The locus of meaning. Papers in honor of Yoshihiko
Ikegami. (Eds.) Yamanaka, Kei. & Ohori, Toshio. Kuroshio publishers, Tokyo.
Misuzu Shimotori
Department of Language Studies
Umeå University
SE - 901 87 Umeå
[email protected].se
Mainstream linguistics for minor(ity) languages?
What is it like to speak Ladakhi?
Bettina Zeisler
formerly SFB 441, Universität Tübingen
1 General background
A member of the large family of Tibetan languages, Ladakhi is spoken by approximately 180.000 speakers throughout Ladakh, one of the three main regions
of the state Jammu & Kashmir in India. Together with Balti (spoken mainly in
Baltistan, Pakistan) it forms the western-most branch of the Tibeto-Burman languages. The Ladakhi dialects fall into two main groups, which differ quite substantially on the grammatical level: Shamskat, spoken in Lower Ladakh (including Balti and the dialects of Purik and Nubra), Kenhat, spoken in Leh and Upper
Ladakh (including Zanskar; for more details see Zeisler forthcoming). Despite a
considerable number of speakers, the language is threatened by Urdu, the state
language, English, the language of education, and by Classical Tibetan, the language of religious books, held up by Buddhist scholars as the only standard of
writing. First sketches and descriptive grammars have been available since the
beginning of the last century (for a brief discussion of the literature cf. Bielmeier
1985: 16-22 and Zeisler 2004: 600-604). Elicited data and data from free speech
were collected and transcribed by the present author in various field stays from
six weeks in 1994 to three or four months each in 1996, and 2002-2008. The elicited data discussed here, was mainly collected in 2007 in collaboration with a
partner project.1
Field work in 2002-2008 was part of a research project within the Sonderforschungsbereich 441 at
the Universität Tübingen, supported by the Deutsche Forschungsgemeinschaft (see http:// For our project B11 on Tibetan and the partner project B17 on comparatives, please visit the respective sites.
2 Mainstream linguistics and minority languages2
The somewhat remote relation suggested in the header can be described as a
combination of lyric and prose. Lyric is widely used when it comes to formulate innovative projects in order to procure research funds. E.g., the newly
launched EuroBABEL project (Better Analyses Based on Endangered Languages) suggests that linguistic theory can profit from the description of hitherto non-described or little-known endangered languages (but why only of
endangered languages and not from lesser-known languages in general?).
Less explicitly but for similar reasons, I was invited into a joint research
project seven years ago (see note 1 above), but the prosaic facts are that what
might be new exciting data in relation to our ignorance or very basic knowledge of the grammar of a badly described language, does not necessarily meet
with interest among the ordinary linguistic community. In my particular case,
when I first turned up with many quite exotic sentence patterns that are not
necessarily expected by linguistic theory, such as case-marked sole arguments of one-place verbs or three non-case- marked arguments of three-place
verbs, etc. (Zeisler 2007, see also
b11fieldwork05.html#Clauses), the benevolent reaction was something like
“Sentence construction plans?! Shouldn’t you go on to proto-roles?”, a statement that did not even take into account that proto-roles, which are not much
more than subject and object in disguise, cannot work even for ‘normal’ ergative languages.
My colleagues cannot really be blamed: there is obviously a wide gap between
what mainstream linguists are doing and what we, the linguists concerned with
lesser known or poorly documented languages, are. The reasons for this gap are
likewise quite obvious: while they come in hundreds, drawing upon a long history
of linguistic discussions, and may have many young and academically, even linguistically, trained informants, we are usually working alone on a given language,
almost always starting from the very beginning, often lacking appropriate terminology, and may be left with the last handful of old, often toothless speakers, or with
farmers, nomads, and hunters not acquainted with our standard of abstract reasoning.
Improvement is possible – if, and only if, we become many, working on the
same language or dialect (which might not yet be very endangered), and if we all
go beyond mere documentation of the most obvious facts, that is, deeper into de2
By ‘mainstream’ I refer to approaches that are, intentionally or not, biased towards the greater languages of the world, i.e., some of the Indo-European languages. Present approaches in Typology are
included, since all generalisations tend to level out minor and not so minor distinctions. The term ‘minority language’ is used here rather loosely to refer to all languages that have some kind of minority
status because of: (a) a limited number of speakers, (b) their status as languages of ethnic minorities,
(c) their low status in a community (spoken versus written varieties, dialects versus standard languages), (d) their poor documentation or low level of recognition, (e) having features not recognised
or discussed in the majority discourse of linguistics (or lacking features that are prominent in this discourse), or (f) their own linguistic traditions being unknown or neglected. I shall focus especially on
those notions of ‘minority’ that have some relevance for the linguistic discourse itself (d–f).
tails, working thus upon one and the same variety for quite some time. Only if
we build up a research tradition of our own, can we attempt to reach a level that
would make our findings more meaningful for other linguists. I certainly do not
want to deny that language documentation, specifically of highly endangered
languages, is a very important research goal, but one should also accept that a
language is not just ‘done’ after a preliminary documentation on the base of two
or three years of post-degree research. Many of the more interesting phenomena
(whether the indigenous name of the half-forgotten plant that cures cancer, the
very construction that challenges all that we know about human conceptualisation or possible languages, to put it lyrically, or the details presented below)
might come to light only after a long period of research. I wonder how one could
even seriously expect postgraduate students without any previous knowledge of
the language or at least a related specimen to come up with such crucial data in
the usually allotted time frame.3
3 What is it like to speak Ladakhi?
If we try to understand the meaning of a particular expression or construction
in a language that is structurally quite different from what we already know,
such as English, Latin, or perhaps also Sanskrit, this will have at least three
dimensions: one is the question of the meaning or function for the speaker
and his or her audience, another is the question of how this specific meaning
can be transferred into another language with the least losses, and finally we
want to know what are the shared features to be compared cross-linguistically.
According to my own experience in the field, not even all regular constructions are open to elicitation or a systematic, that is, logical approach, and they may show up only gradually. E.g., I have
been recording the speech of one and the same person over a period of 12 years while transcribing it
during the last 7 field stays, besides conducting more focused elicitation work. In each year we came
across at least one if not several new constructions or idiomatic expressions that did not appear in the
previous texts (and often also not in the following ones). In some of these cases, I simply could not
imagine the existence of such patterns and thus had no means of including them in the elicitation
work, in others I had, in fact, tried to elicit them, but since I did not provide an adequate context or
for other minor faults, the existence of the pattern was denied. Sometimes, as in the case of examples
(3) and (6) below, the non-elicited data even contradicts and thus enriches the elicited data.
Even when it comes to such rather simple features as consonant migration in compounds, the systematic approach (taking a list of possible candidates, e.g. all combinations with the noun hu ‘water’ as
first element) yields not as much result as when you are able to confront your consultants with ‘real’
data from another speaker. Even in case of a negative answer, the consultant will then, more often
than not, come up with one or two new compounds, which might again be rejected by the first consultant in exchange for one or two new items, and so on, almost endlessly. Only after some time you
will have a list you can go through more systematically.
3.1 Whose meaning? – Some problems of linguistic discourse
To start with the last question, Haspelmath (2004: 572) states quite laconically that “the definition of categories for cross-linguistic comparison […] must
be based on meaning”. This should be a matter of course, but several questions
arise immediately: How do we define the meaning of an utterance in a language we are not native speakers of? How much are we actually allowed to
abstract from the language or utterance-specific context? Does the focus on
meaning, and only meaning, imply that we can abstract from structural
I am asking the last question, not because I disagree with Haspelmath, but because I observe with respect to the Tibetan languages that the meaning or function of an expression is typically defined by its translation into English. To give
only one example: Tibetan embedded nominalised clauses are often treated as
‘relative clauses’.4 In fact, since nominalisation is quite restricted in English,
relative clauses are almost always the only possible choice when translating such
constructions into English. But should not the very fact that English as well as
Tibetan have both constructions and that these constructions not only have different formal properties but also different distributions, tell us that they might be
functionally different?
Even if English had no nominalisation and Tibetan no relative clauses at all,
would it be enough to say that both constructions serve a roughly similar purpose, such as adding background information or specifying some item, to conclude that they have the same function or even the same meaning for the respective speakers? From a philosophical (as well as from a scientific) perspective it would certainly not be enough to say that the bat sonar system serves
to “make precise discriminations of distance, size, shape, motion, and texture
comparable to those we make by vision” in order to call it a visual system
(Nagel 1974: 438). And, except for the most general statements, it would be
not very satisfactory if the different kinds of perception, by virtue of being
perception, were not further discriminated. The word ‘perception’, however,
is in fact nothing more than a cover term for what goes by various different
names or descriptions. It is neither expected that these subcategories are func4
Cf., e.g., Beyer (1992: 309 ff.) for Classical Tibetan. Typically, the notion of ‘relative clause’ is
not defined. Strictly speaking, we should not even apply the term ‘nominalisation’: all varieties have
several ‘nominalisers’, the choice of which depends on the semantic role of the derived noun. In most
cases, the original function of the ‘nominalisers’ as nouns is still transparent (e.g. -sa ‘ground, place’,
yielding ‘location of verb-ing’). All of them, even the opaque ‘nominaliser’ {pa} can be found in
combination with nouns, cf. Classical Tibetan as well as Ladakhi rta-pa ‘horse-PA’, i.e. ‘rider’ (note
that the nominal derivation by {pa} is no longer productive and the morpheme might have lost all
functions as in the case of khapa ‘house’ for which no underlying lexeme kha seems to exist synchronically). Finally, Tibetan verb stems have some nominal properties (e.g. they can appear in the
place of verbal nouns or can be combined with case or even definiteness markers). It is not very intuitive to talk of nominalisation of nouns or other items with nominal properties. However, an alternative description in terms of compounds or derivations is even less adequate, because the initial
elements can still govern arguments.
tionally equivalent, nor that any one of the subcategories may serve as cover
term for all others. Whereas by using the term ‘relative clause’ for embedded
nominalisation it is automatically implied that there is no functional difference and that nominalisation is the same as a relative clause or at least a subcategory thereof. This implication might not have been intended originally,
but it is symptomatic for the general terminological laisser faire and all-inclusive attitude in linguistics.
Haspelmath (2008) might have felt a similar uneasiness, as he states that
“we should approach any language without prejudice and describe it in its own
terms, […] overcoming possible biases from our native language, from the
model of a prestige language (such as Latin or English), or from an influential
research tradition (such as that of Donatus’s Latin grammar, or Chomsky’s
generative grammar).” This should be a common place. Yet again, two important questions arise immediately: First, given the many minority languages
without a linguistic tradition of their own and even without a single linguist
in their population, how do we, the outsiders, often very bad performers, define a language’s own terms? Secondly, although it is more than deplorable
that native traditions are typically neglected by typological approaches (see
also further below), would it not be the end of cross-linguistic comparison, if
each grammar followed but its own idiosyncratic terminology? There is, in
fact, an indigenous grammatical tradition and terminology for Classical
Tibetan.5 But besides a small elite among the scholars of Indian and Tibetan
studies, who has ever heard of it? We have not yet managed to speak to a general linguistic audience about these concepts as concepts in their own right
and without being forced to reinterpret them and translate them into some sort
of linguist-ese.
Furthermore, I cannot help but observe in Haspelmath’s statement a blatant
bias against non-European prestige languages such as Sanskrit or Classical
Chinese and even more against the corresponding grammatical traditions,
which is all the more surprising as Pƞini’s most influential Adhyy, recognised as the first ever generative grammar, seems to have inspired some of our
modern theories. This is quite telling for a linguistic discourse that is basically
a majority discourse: minorities do not have a voice, at least not a voice of their
From the viewpoint of a minority language, one could ask many odd questions, such as: Why is it the case that embedded nominalisation is described
in terms of ‘relative clauses’ and not the other way round: why are relative
Not of ‘its own’ in the strict sense, as it is strongly inspired by Indian Buddhist grammars. Nevertheless, Tibetan grammarians made all efforts to reconceptualise the inherited terminology time and
again in order to make it account for facts of their own language (cf. Verhagen 2001 for an overview)
– to the extent that it creates the greatest confusion if one tries to interpret these terms in their original
sense (cf. Zeisler 2006 for examples of how much lateral thinking we need only to understand why
the Indian terminology is used the way it is and why this creates difficulties within the Tibetan grammatical tradition itself).
clauses never called ‘(a kind of) nominalisation’ or at least ‘(a kind of) embedding’ or ‘dependent modification’? Why did Indian linguists never try to
apply, test, or reconceptualise one of the Indian grammatical systems as a
whole with respect to some European language? Why is it not a matter of
course that a grammar of an Indian language, co-authored by an Indian linguist, is written “in a more traditional indological format” and uses a
“non-standard transcription”, as Liljegren (2008) states with a tone of complaint? After all, the second largest population of the world might have more
Indologically trained scholars than there might be ever typologists – who, in
absolute numbers, might perhaps not really become “an increasingly important target group” (ibid.).
As it is so often the case with minorities in the real world, we, the researchers
of minority languages, lack strong allies. For want of a linguistic tradition or
in-depth studies, we cannot compare our seemingly ‘strange’ and inexplicable
phenomena with those of other minority languages of the same area, hence we
lack the opportunity to draw upon the lucky brilliant insight of one of our colleagues. We are left with our own biased intuitions and with the bias of bilingual
informants towards linguistic concepts of the dominant language they are taught
in school (if they receive any adequate teaching at all).
Trying to understand what a particular expression really means for the speakers themselves, thus amounts at trying to understand what it is like to be a bat for
a bat.6 Not only because of the different cultural background, not only because
informants and researcher, due to differences in education, dispose over quite
different levels of abstract thinking (a constant source of errors for the latter), but
mainly because the researcher speaks a structurally different language. Mainstream linguistics does not provide us with any practical or theoretical means to
overcome this bias or to close the hermeneutic gap between one’s own and the
other’s language. The Sapir/Worff-hypothesis is best corroborated by the linguistic discourse itself.
“It will not help to try to imagine that one has webbing on one’s arms, which enables one to fly
around at dusk and dawn catching insects in one’s mouth; that one has very poor vision, and perceives
the surrounding world by a system of reflected high-frequency sound signals; and that one spends the
day hanging upside down by one’s feet in an attic. In so far as I can imagine this (which is not very
far), it tells me only what it would be like for me to behave as a bat behaves. But that is not the question. I want to know what it is like for a bat to be a bat. Yet if I try to imagine this, I am restricted to
the resources of my own mind, and those resources are inadequate to the task. I cannot perform it
either by imagining additions to my present experience, or by imagining segments gradually subtracted from it, or by imagining some combination of additions, subtractions, and modifications” (Nagel
1974 p. 439). “The problem is not confined to exotic cases, however, for it exists between one person
and another. The subjective character of the experience of a person deaf and blind from birth is not
accessible to me, for example, nor presumably is mine to him” (ibid. p. 440). As the citation shows,
the problem of intersubjectivity is one of degrees. But the problem described in the following is certainly somewhat more intricate than the standard situation of communication, which typically functions well, even though, or perhaps rather because, one does not fully understand (and does not want
to understand) what the communication partner really means.
3.2 Hermeneutics fails when you least expect it
As an illustration for the above discussion I have chosen non-equative expressions of comparison, as one should think that the situation in the outside world
is quite manageable and hence cross-linguistic comparison should not pose much
problems: We have two entities, A (the item to be compared) and S (the standard
to which something is compared), to which we ascribe, for the sake of simplicity,
a perceptible and measurable, i.e. scalable property X. Furthermore, we restrict
ourselves to a situation where A and S do not have the same amount or degree of
X. Whether the expression is grammaticalised as in English or not as in Ladakhi,
we should not find much difficulty in establishing what situations the speakers
refer to.
The structure of Tibetan non-equative comparative expressions can be described as follows:
with -REL being a non-specific relational morpheme, -(b)asa in Shamskat,
-(e)sa in Kenhat. This morpheme is not only found in comparisons, but can
also be used to express contrastive, hence non-comparative relations, such as
not only - but, beyond, hardly, instead, or rather than. X stands for an adjectival, i.e. a verbal expression of the property in question (e.g. ri ‘be long’, the
verbal noun of which is ri(b)a ‘being long’). With the exception of Balti and
Ladakhi, which both suffer heavy influences from the dominant language
Urdu, the derived nominal adjectives (such as rimo ‘long’) cannot be used in
non-equative comparisons. In most varieties a nominalised form of the verbal
adjectival plus auxiliary is used, a construction that may or may not be formally identical with a normal present tense form. In Old and Classical Tibetan we
still find the adjectival stems used without nominalisation and auxiliaries and
the same is true for the speakers of the more peripheral dialects of Ladakhi. In
this case, the use of the tensed verb seems to always indicate some sort of comparison or some sort of dynamicity (implying a comparison between earlier
and later states). But one should not forget, that even the simple property
ascription implies some sort of comparison. We would not ascribe a property
except when a certain threshold level has been reached (Stefan Hofstetter,
B17). Example (1) shows the three available forms (for abbreviations, cf. p.
318 below).
zgo(-e ri%bo)-basa
Tsheri% ri-ok.
door(-G length)-REL
/ ?ri-(b)a duk.
/ rimo duk.
/ long(ADJ) be.EXP
/ be.long(VA)-NOM be.EXP
‘With respect to (the length of) the door, Tshering long-s (generic) / is long-ing7 / is long (visual
evidence).’ ~ Tshering is taller than (the height of) the door.8
As it appears, structural differences in basic linguistic modules, such as
nominal versus verbal adjectivals or lack of comparative morphology, can be
ignored in the case of simple comparisons, so that the Ladakhi expressions
and their English counterparts can be matched easily. Before continuing with
one of the most complex cases, I should like to draw the attention to possible
differences in conceptualisation between those languages that (also) use comparative morphology and those that do not (at all): The comparative morpheme -er in German seems to automatically invoke a concept of difference
in a way periphrastic constructions with the help of quantitative expressions,
such as more, do not. Such differences in conceptualisation may have conse7 8
quences for the analysis and understanding of a foreign utterance.
Comparisons involving an integral factor, thus A having the property X of
S y times, are conceptually somewhat between equative and non-equative
comparisons. Different languages may have different solutions to this conceptual problem, but it seems that languages with overt comparative morphology
like German rather use an equative construction: A is y times as X as S, or a
possessor construction with an abstract noun: A has y-times the X of S, lest an
explicit interpretation in terms of difference is intended: A is y times X-er than
S, which (for some German speakers at least) actually means that A is y+1
times as X as S.9
The equative option does not seem to be possible (or necessary) for languages that can only use some sort of paraphrase, thus the French expression
A est y-fois plus X que S has only one reading, namely A is y times as X as S.
This comes quite as a surprise to me. Apparently deeply biased by the hard7
This rendering should not be taken as an equivalent for the English progressive. A better representation would be German *ist größend which differs from a present tense *größt as well as from the
vernacular progressive expression *ist am größen. The question mark on the corresponding form applies only for the proper dialectal form riba, which is perceived as being out-dated, while the form
ria, borrowed from the Leh dialect, is perfectly acceptable, although rejected in the elicitation context by virtue of not belonging to the speaker’s dialect.
In order to keep the underlying structure as transparent as possible and not to lose possible semantic
differences, no attempt will be made to smoothing out the translation, even on the risk of making the
language appear more exotic than it is. Even such ‘literal’ translations may not be very objective,
since they contain my personal interpretations. Only for better understanding I will add interpretations in terms of a standard situation or a possible intended meaning in English. This will be marked
by italics. It should be clear that such an interpretation is not necessarily a representation of the intended meaning in Ladakhi. Furthermore, it should never be mistaken for the only possible interpretation. It is only expected to be the most common one in a most common situation.
In English, the expression is ambiguous between y times and y+1 times (Sam Featherston, p.c.).
This may perhaps be facilitated by the frequent use of the paraphrase with more or less.
wired implication of the German comparative morpheme, I would have expected at least some sort of ambiguity. Quite similarly, Ladakhi, lacking overt
comparative morphology, makes use of the non-equative or contrastive relator
-(b)asa% / -(e)sa%, and not of any available equative relator or a possessor construction, and the resulting expression is likewise not ambiguous at all. In contrast to the European languages, however, the property in question is typically
not specified with an adjectival.
(2) tso%-e
LEHa onion-G
‘In relation to the width of the onion one should prick having made two times a depth (generic).’ ~
One should prick the onion, having made a hole two times as deep as its width.
When it comes to more complex situations, structural differences and hence
subtle non-correspondences in meaning for each linguistic module involved
add up or even multiply, so that it might even be impossible to translate one
sentence into another. This is exactly what happens when we think of a situation where A is compared with a non-existing standard: A is X-er than nobody.
The resulting mismatch in such complex cases may indicate that the apparent
matching in the simple cases was only superficial or accidental.
In the case of Ladakhi, the structural differences in the module of comparison combine with even more fundamental differences in the module of negation: Ladakhi like all other Tibetan languages has no n-words or negative
indefinites. The negation markers are obligatorily bound to a verb or its auxiliary and thus always operate on the whole clause. In the case of constituent
negation, a negative polarity item, typically consisting of an indefinite or
limiting quantifier combined with a focus marker, is used together with sentence negation, e.g. Anybody ever / A single person ever not-verbs. Most probably this paraphrase does not have the same logical entailments as the presumed English counterpart (but this is one of the many features that have
never been researched).
In order to obtain a rough equivalent of the sentence A is X-er than nobody,
we have thus not only to transform the adjective into a verb, but also to transform the negated noun phrase into a negated clause, which in turn has to be
nominalised and embedded, so that we may eventually arrive at a formulation
such as:
In relation to anybody ever non-existing, A X-es.
Understandably, such constructions are not very natural, and the researcher
finds some difficulties to get elicited examples without some more modifications, which all bring us further away from the intended meaning that the original model sentence could have had. Elicitation is not made easier by the fact,
that the model sentence does not have any meaning for the researcher or for
most speakers of English.10 In some languages, however, a sentence of this
type implies that there is no S compared to which A is X-er. This paraphrase
may again have two readings, which I would call tentatively: (a) ‘anti-superlative’: A is the most non-X of all and (b) ‘super-superlative’: A is so extremely
X that there simply does not exist a standard to which A could be compared.
The first interpretation might be somewhat more frequent and is found among
other languages in Turkish (Stefan Hofstetter, B17), the second interpretation
has so far only be observed in Guarani (Stefan Hofstetter, B17) and, as it
seems, in Ladakhi.
In the elicitation context we get two slightly different construction types: the
first one (3), by explicitly negating the property X of S, is interpreted by most
informants as ascribing property X to A only in a limited quantity or quality: in
relation to S who does not X at all, A X-es, that is, A is not very X either, hence I
will call it tentatively the modest excess variant. However, as the different readings given by the Leh informants as well as example (6) below show, the interpretation depends, as in so many other cases, not so much on formal clues than
on contextual features. Some informants, however, preferred a super-superlative
interpretation as in the second construction (4), the extreme excess variant. This
construction is somewhat elliptical and thus hard to analyse. If accepted at all, it
gets the interpretation: A is extremely or unnaturally X.
(3) SKI su-a
LEHa/b su-ig-a
*(rimo) met-kan-eesa
who-(LQ)-FM long(ADJ)
Tsheri (rimo)
Tsiri rimo/ri-a
long/be.long… be.EXP
‘In relation to whosoever not being tall (assimilated knowledge), Tshering is (tall) (visual evidence).’
~ modest excess variant: Tshering is taller than anybody who is not tall. (SKI: %, LEHa: %%,
LEHb: *)
~ extreme excess variant: Tshering is taller than the limit of tallness that no one ever reaches
(i.e. extraordinary tall). (LEHb: , LEHa: %)11
According to my colleague Stefan Hofstetter from the partner project B17, there might be some
variation in acceptance. Likewise speakers from the northern parts of Germany tend to reject the sentence as having no sense at all, while speakers from the southern regions might get a superlative reading, at least if the particle sonst ‘else’ is added: A ist X-er als niemand/keiner sonst. To my ears (socialised in Bavaria) this seems to be much less shocking than the corresponding expression without
the particle, but I do not want to preclude that this kind of ‘improvement’ is merely due to an interference with the equative construction A ist so X wie niemand / keiner (sonst). The latter certainly
improves a lot through the use of the particle. If I am forced to give the former sentence some sense
and if I think (too) much about it, I will get a somewhat different reading implying two comparisons,
one non-equative, the other equative: A is in a way X-er [than S] that nobody else is X-er [than S] or
somewhat simpler: A exceeds S with respect to X as much as nobody else. Stefan Hofstetter, who
grew up in Swabia, says he has no objections at all against the construction.
The symbol indicates acceptance without qualification. % indicates that the acceptance or interpretation depends on a suitable context, %% signals here that the other possible interpretation must
be ruled out by explicitly stating that all people are very small and that Ts. is not really tall, but at
least taller than the rest. Note that the sentence is not acceptable for the Leh informants without adding a limiting quantifier to the pronoun.
(4) a su-basa
who-REL ø long(ADJ) NG2.exist/have/be.N.EXP-NOM
Tsheri% (ri%mo)
long(ADJ) be.EXP
‘In relation to anyone, without having / [there] existing ø = [anyone] tall / without ø = [anyone]
being tall (assimilated knowledge), Tshering is (tall) (visual evidence).’ ~ Tshering is taller
than anybody (else) in a way no one is tall / without having a tall match. (%/)
b su-b
Tsheri rimo
long(ADJ) be.EXP
‘In relation to anyone, without having / [there] existing ø = [anyone] ø = [tall] / without ø =
[anyone] being ø = [tall] (assimilated knowledge), Tshering is tall (visual evidence).’ ~ Tshering
is taller than anybody (else) in a way no one is / without having a match. ()
The intended meaning was paraphrased by the informants as: “Nobody is as tall
as an elephant, but Tshering is” (SKI), that is, s/he is supernaturally (SKI) or
extraordinarily (GYA) tall, “a giant” (SKI); s/he is surpassing the upper height
limit expected for human beings. According to the informants, there is no or not
much difference to an as big as nobody (else) construction (SKI, GYA). Sentences (4a/b) are, in fact, based on expressions, such as:
ikrhten-ekane met-kan
LEHa/b world-upon
Tsiri rimo
/ *ri-a
long(ADJ) / be.long(VA)… be.EXP
‘[In a way/as one that] does not exist in the (whole) world, Tshering is (tall) (visual evidence).’
~ Tshering is [as] tall [as] nobody else in the whole world.
The following example from a personal narrative would fall at least formally into
the modest excess category (the inverted order is not relevant for the interpretation):
den do_
_rdemo dug_ _ja, i-a
KHAL then that.DF beautiful be.EXP CONJ what-FM NG2.exist/have/be. house-LQ-REL
But what does it actually mean?
The solution is not getting easier in view of the fact that evidential marking
overrides the distinction between attributive and existential linking verbs (in ‘be’
vs. yot ‘exist’ or ‘have’), both becoming duk in an experiential context and yot
in a non-experiential one, so that we get at least three possible translations for the
embedded negation ia metkhan: ‘not being any-thing’ or perhaps also ‘not being beautiful at all’, ‘not having anything’ or perhaps also ‘not having any beauty
(or grain) at all’, and ‘not existing at all’. The least I can say, is that in this very
utterance it is not implied that A is only relatively beautiful (modest excess
variant). Nor is it implied that A is extremely beautiful (extreme excess variant).
In the immediately preceding context of (6), the narrator first describes a representative room that, although possessing attributes of wealth and modernity, is
not very beautiful in his eyes, because it is ‘empty’. He then contrasts it with a
room of a more traditional house where barley is heaped up in the corners (as if
this could make a room more homelike) and continues with the above sentence.
Quite apparently, the relational morpheme is used here with a contrastive function. My rather tentative ‘literal’ and ‘intensional’ translations thus run as:
‘Then THAT one is/was beautiful, again, in contrast to / instead of the house that
does not have anything.’ ~ Now THIS one is (really) beautiful, NOT the other house
that hasn’t anything.
As it seems to turn out, negation may – depending on the context – constitute a
subset of universal quantification,12 which in turn is used to formulate absolute
comparisons. But the boundaries between simple and absolute comparison and
between comparison and positive property ascription are not well defined.
‘Comparison’ as a linguistic category is not a grammaticalised category in
Ladakhi, perhaps not even a valid psychological concept. With the changing context, an expression may vary between quite contradicting values: absolute comparison, relative (qualified) property ascription, and contrastive (unqualified)
property ascription. The original model sentence, whatever meaning it may or
may not have, cannot be translated, because it is based on a grammatical category
that does not exist in Ladakhi. Hence the conceptual problem associated with the
grammatically marked form does not arise. Since any positive property ascription implies a comparison with a culturally defined, and thus always relative
threshold level, the ‘real’ model for Ladakhi might be three-fold:
A is X – modified by an underspecified negative relation with S:
a) with no comparable S available, b) compared to a non-X S,
c) in contrast to a non-X S
“Mit einem Ablativ des Vergleichs übersetzt man also die Ablativpartikel mit ‚als’
und bildet den Komparativ des betreffenden Adjektivs oder Adverbs. Diese mechanische Übersetzungsregel darf natürlich nicht dazu führen, dem Tibetischen
eine Steigerungsform zuzuschreiben, denn vom tibetischen Standpunkt aus liegt in
diesen Fällen lediglich ein spezielles modales Verhältnis vor, d.h. der Geltungsbereich von Adjektiven und Adverbien wird in bestimmter Weise eingeschränkt”
(Hahn 1985: 97, emphasis added). What could have been rejected by Haspelmath
as a merely formal approach, further biased towards traditional Latin grammar,
appears to be fully justified by a closer look at the linguistic facts. Form matters. –
But exactly when and how much, that remains the question.13
As the question was raised in the discussion, I should like to make it clear that embedded negation
is not an equivalent or the closest translation of the English expression any, nor is any an equivalent
or the closest translation of a Ladakhi embedded negation. The English expression any is usually best
translated by an indefinite pronoun, such as su ‘someone’ or ‘who’ plus focus marker {ja} ‘even,
also’, or by an allquantor.
It is certainly justified to treat different alignment systems as expressing the same relations, at least
in the majority of their instantiations. I am quite convinced that a sentence like the man sees (perceives) a dog, does not change in meaning or does not change its semantic roles, whether the experiencer is encoded in the nominative, as in English or German, in the ergative, as in most Tibetan
languages, or in the aesthetive (dative-allative), as in Ladakhi. But there are obviously many cases
where different representations of the seemingly same situation lead to subtle or less subtle differences in meaning.
It goes without saying that I am not absolutely convinced to have found the
right analysis and I would have been very happy to find a construction similar to
those in examples (3), (4), and (6) be discussed for any of the Tibetan languages,
but all that we can find in the reference grammars (if comparison is mentioned,
at all) is the basic facts as described in the beginning of this section. I would also
be happy to find such discussion with respect to any other language. But given
the extreme difficulties in understanding a construction in a language I researched for many years and speak, although imperfectly, I am somewhat sceptical that without further knowledge of any such language I would really understand what is going on.
Before concluding I should like to refer to some discussions we had within our
joint research project. Depending on our individual research socialisation we
seem to have quite different expectations about what would be the most natural
interpretation of A is X-er than nobody. While for the members of the partner
project B17 there seems to be little doubt that the sentence should by preference
get an ‘anti-superlative’ interpretation, I myself have the feeling that the underlying logical operation is not very natural for a speaker of a natural language outside an academic context (but I do not want to preclude a conventionalised
usage). On the other hand, the ‘super-superlative’ interpretation, although not
logically transparent, might perhaps be better motivated psychologically: Many
languages have expressions such as matchless, countless, or even incomparable
that express the superior quality or quantity of A by explicitly negating the
feasibility of comparison. Such expressions imply that the gap between A and
any ordinary standard with respect to the property X is so extreme that it is not
worth to even mention S. One could say that S apparently lacks a comparable
amount of X.
4 Possible lessons
While the documentation of endangered or not-so-endangered lesser known languages is certainly an important task, I hope to have shown that the writing of
standard grammars alone cannot be enough, and that one has to go much deeper
into every detail, e.g., by testing border-line cases.
Further more, linguistic data, however exciting it might be, cannot be selectively collected or discussed without having or giving an idea of how this data
relates to the overall structure of the language.14 It is, therefore, not possible to
work with a standard questionnaire not specifically designed for a particular language, as it is inherently biased and thus forces its conceptualisations on languages when they do not apply (such questionnaires may still give valuable hints,
what to look for or how to interpret one’s findings). Consequently it is also not
Of course, this holds also for the present paper, but more details will be provided in Zeisler (in
possible to let just any linguist conduct such survey without a substantial knowledge of the language to be researched. Even more: elicitation can fruitfully be
used to confirm data, but it cannot be the first choice for testing or establishing
categories in highly context dependent languages.
Cross-linguistic comparison, neglecting the language specific details and even
more how they relate to the overall structure of a language, may lead to theoretical artefacts, just as when we analyse the bat sonar system as a visual system.
Dialects and interlocutors or language consultants
Gya (Kenhat), elicited data (Menggyur Tshomo)
Khalatse (Shamskat), free speech (meme ‘grandfather’ Tondup Tshering)
Leh (Kenhat), elicited data (a: Thrinlas Wangmo, b: Tsewang Dorjey;
Skindiang (Shamskat), elicited data (Choron Angmo)
Descriptive terms
derived adjectival of nominal character (in some cases, an additional
derivational morpheme is inserted before the nominaliser)
definiteness marker
clause chaining
experiential construction (direct visual evidence, new knowledge)
focus / emphatic marker
genitive marker
generic statement (also: inferential future)
limiting quantifier ‘a, some’
non-experiential construction (assimilated knowledge)
negation marker or negated form (NG1 = mi NG2 = ma)
relational marker (relator)
verbal adjectival
x_ _y
phonetically conditioned features across word boundaries
Beyer, Stephan V. 1992. The Classical Tibetan language. New York: State University of
New York. Reprint 1993. (Bibliotheca Indo-Buddhica series 116.) Delhi: Sri Satguru.
Bielmeier, Roland. 1985. Das Märchen vom Prinzen <obza. Eine tibetische Erzählung
aus Baltistan. Text, Übersetzung, Grammatik und westtibetisch vergleichendes Glossar. (Beiträge zur tibetischen Erzählforschung 6.) St. Augustin: VGH Wissenschaftsverlag.
Hahn, Michael. 1985. Lehrbuch der klassischen tibetischen Schriftsprache. (Indica et
Tibetica: Monographien zu den Sprachen und Literaturen des indo-tibetischen Kulturraumes 10.) Bonn: Indica et Tibetica.
Haspelmath, Martin. 2004. Does linguistic explanation presuppose linguistic description?
Studies in Language 28(3). 554–579.
Haspelmath, Martin. 2008. Framework-free grammatical theory. (15 Oct. 2008)
Liljegren, Henrik. 2008. A Grammar of the Shina Language of Indus Kohistan. By Ruth
Laila Schmidt and Razwal Kohistani … (Review) Himalayan Linguistics Review 6. 1–
Nagel, Thomas. 1974. What is it like to be a bat. The Philosophical Review 83(4). 435–
Verhagen, Pieter Cornelius. 2001. A history of Sanskrit grammatical literature in Tibet.
Vol. 2 Assimilation into indigenous scholarship (Handbuch der Orientalistik: Abt. 2,
Indien 8.) Leiden etc.: Brill.
Zeisler, Bettina. 2004. Relative Tense and Aspectual Values in Tibetan languages. A
Comparative Study. Berlin etc.: Mouton de Gruyter.
Zeisler, Bettina. 2006. The Tibetan understanding of karman: Some problems of Tibetan
case marking. In Christopher I. Beckwith (ed.), Medieval Tibeto-Burman languages
II, Tibetan studies: Proceedings of the Tenth Seminar of the International Association
for Tibetan Studies, Oxford 2003, 57–101. Leiden: Brill.
Zeisler, Bettina. 2007. Case patterns and pattern variation in Ladakhi: a field report. In
Roland Bielmeier & Felix Haller (eds.), Linguistics of the Himalayas and beyond,
399–425. Berlin etc.: Mouton de Gruyter.
Zeisler, Bettina. Forthcoming. Kenhat, the dialects of Upper Ladakh and Zanskar. In
Mark Turin (ed.) Selected papers of the 11th Himalayan Languages Symposium,
Chulalongkorn University Bangkok, Thailand, 6–9 December 2005. Leiden etc.: Brill.
Zeisler, Bettina. In preparation. mindra mindra (incomparably different): Property ascription and expressions of difference (and comparison) in Ladakhi Tibetan.
I should like to express my deepest gratitude to all informants for their patiently fostering
of my slow understanding, besides those mentioned above also to Tshewang Tharchin,
Thrinlas Chospel, Lobzang, Rinchen Lhamo (Domkhar), Tsering Dolkar (Tia), Rigzin
Samdup (Nyoma), Sonam Tundup (Sachukhul), and Tsering Samdup (Pipcha). I should
also like to thank Sam Featherston for improving the grammatical acceptability of this
Since submission of the paper, the author arrived at the conclusion that the Ladakhi
expressions are best described in terms of contrasting, implying not a gradual, but a
categorical difference between A and S: property X is positively ascribed only to A, but
not to S, for which the property might even be denied. This might hold also for other
Tibeto-Burman languages, e.g. for Newer (Kazuyuki Kiryu, p.c.). Accordingly,
categorical contrasting and gradual non-equative comparison should be seen as part of a
broader concept of differentiating property ascription.
Bettina Zeisler
formerly SFB 441, Universität Tübingen
[email protected]
Readability and Multilingualism
Workshop on Readability and Multilingualism
Sofie Johansson Kokkinakis ([email protected])
Institute for Swedish as a Second Language, Språkbanken, Department
of Swedish Language, University of Gothenburg, Sweden
The workshop”Readability and Multilingualism” took place October 1-3 2008 in
Uppsala as part of the 23rd Scandinavian Conference of Linguistics. The workshop aimed at covering different aspects of the subject “Readability and multilingualism”.
Heavy focus is today put on texts for different purposes. They must be readable and even in some cases easy-to-read and understand. This is partly dependant upon the fact that we live in a society where information is intensive and
where it is getting more and more important to find the information and to retrieve the information available especially on the Internet. That some texts in certain situations have to be easy-to-read and understand is due to the fact that the
information has to be accessible for everyone irrespective of educational background or mother tongue. Texts in this situation are often directed to adults and
concern information from the government, the society, news, social and medical
care etc.
Similar demands on readability are raised for texts directed to children and
youth, where texts might be literature and school textbooks and texts in different
kinds of test situations.
In this workshop we were discussing issues concerning readability on a more
general level where multilingualism was the common denominator. The issue of
readability was treated from a broad perspective covering aswell adults as children and youth. The invited participants of the workshop presented their research
from different perspectives of readability.
Eight contributors to the workshop were initially invited to give a 30 minutes
presentation. A general open discussion followed the presentations. The discussion brought up questions like “What common issues are there in our research?”,
“How can we co-ordinate research based on similar ideas and interests on readability?” and “What should we aim for in the future?” Summarizing the discus323
sion shows an obvious common interest in different aspects of readability and
also multilingualism. Research would typically benefit from sharing and distributing language resources such as lexica and various computer-based tools. Aims
for the future and what to follow up from this workshop we agreed that we should
meet again in the near future to continue sharing ideas and resources, starting by
forming a web-page with related information on research and resources and also
working towards forming a network to ensure and aid information exchange. We
could also confirm that there is a great need in schools for instance for resources
to use for educational purposes and assessment.
There was a surprisingly great interest in this workshop, both from contributors, visitors and others. The concluding remarks will be to keep an open mind
in these issues and to work together towards a common goal, namely to continue
doing research on reading and writing issues.
Sofie Johansson Kokkinakis
Workshop webpage:
Readable, Legible or Plain Words
– Presentation of an
easy-to-read Swedish corpus
Katarina Mühlenbock
University of Gothenburg
1 Introduction
International studies (OECD 1995, PISA 2003) of literacy and text comprehension among the adult population reveal that approximately 7.5% of Swedish
adults are able to search information only in familiar, very simple and well-structured texts. These persons belong to a highly heterogenic group with varying underlying difficulties such as dyslexia, aphasia or cognitive disabilities, or they
might be second language learners; all with individual specific needs. Target
group for the present study is people with mild intellectual disabilities. For thirty
years the term easy-to-read has been used to denote simplified written material,
targeted at these persons. Production of easy-to-read text has, however, so forth
not aimed at any individual adaptation. Considering the variation in individual
reading skills and prior knowledge, this presentation focuses on readability factors that may be crucial when producing text for a certain reader.
Scientific studies approach the process of reading from various points of view,
and research activities within the field often focus on typographical issues, layout or rhetorical matters, i.e. the legibility of a text. This aspect is not considered
in this study. Instead, I adopt a general definition of the concept readability, and
regard a text from which maximum information is obtained with the minimum
cognitive effort needed, as easy-to-read.
The corpus LäSBarT, comprising ~1.2 million tokens, was compiled mainly
as a basis for readability studies, but also for extraction of a simple Swedish base
vocabulary. The corpus was divided into four subcorpora, according to text
types, distributed as to Table 1. Preceding further studies the corpus was annotated, POS-tagged and lemmatized.
A large amount of American studies focusing on a multitude of readability
factors and indices have been carried out. Chall (1958) and Klare (1963) have
made extensive overviews of these studies, involving for instance word length,
amount of syllables or characters, frequency of subordinate clauses, etc., all of
which, separately or in combination, were expected to give a rough estimate of
the syntactic, lexical and semantic complexity. For Swedish, Björnsson (1968)
developed the LIX formula and Platzack (1974) put forward his thesis discussing
the issue of readability in relation to the reader and his efforts, e.g. psycholinguistic factors.
In this presentation, the factors selected for focus concern LIX, extra long
words, nominal quote (NQ) and the lexical variation, revealing different linguistic aspects of a text. All calculations are based on figures of the lemma/lexem
LIX is calculated by adding the average number of words per sentence to the
percentage of long words, i.e. t 7 characters. A general assumption is that if the
number of long words increases, so does text complexity. The same assumption
is made for high value of long sentences. Simple texts typically have a LIX value
between 25 and 30.
The extra long words ( 14 characters), indicate the proportion of long compounds, a property which is also partially provided for in the LIX calculation.
The nominal quote (Hultman & Westman 1977), presents the degree of information density, and is calculated as: NQ =
nouns + prepositions + participles
pronouns + adverbs + verbs
Standard information texts normally reveal a high lexical density depending on
a large number of nominalizations. LäSBarT presents low values in average, particularly the Fiction subcorpora. An NQ of 25 is typically found in a speech corpus whereas school textbooks might have an NQ of 120.
Finally, the lexical variation (Laufer & Nation 1995), calculated as the variation between different word types and tokens in the semantically disambiguated
text, shows a low degree in average, corresponding to speech. The results of all
the calculations above are shown in Table 1.
Table 1. Composition and readability measures for the LäSBarT corpus
Easy-to-read texts
No. of tokens
No. of types
LIX value
Extra long words %
NQ value
Lexical variation
Ordinary texts
News text
Children’s fiction
After comparing these values across the four subcorpora I found that a measure
based on more factors on lexical, syntactic and semantic levels contributes
strongly to a more appropriate weighing of text difficulty. Texts adapted to the
specific needs of an individual reader are valuable assets for various types of applications connected to research and education, and constitute a necessary prerequisite for the integration into society of language impaired persons.
LäSBarT will be made available for research at Språkbanken, Department of
Swedish Language, University of Gothenburg.
Björnsson, C.H. 1968. Läsbarhet. Stockholm: Liber.
Chall, J. 1958. Readability. An appraisal of research and application. Ohio: Bureau of
Educational Research.
Hultman, T.G. & Westman, M. 1977. Gymnasistsvenska. Lund: Liber läromedel.
Klare, G. 1963. The measurement of readability. Iowa State University Press.
Laufer, B. & Nation, P. 1995. Vocabulary size and use: Lexical richness in L2 written
production. Applied Linguistics 16(3).
Organization for Economic Co-operation and Development. 1995. Literacy, economy,
and society. Paris: Organization for Economic Co-operation and Development, Centre
for Educational Research and Innovation.
PISA. 2003. <>
Platzack, C. 1974. Språket och läsbarheten: en studie i samspelet mellan läsare och text.
Lund: Gleerup.
Contact information
Katarina Mühlenbock
Department of Swedish Language
University of Gothenburg
Box 200
405 30 Göteborg
[email protected]
Text and language in assessment of
mathematics and science
Caroline Liberg, Eva Forsbom
Uppsala University
1 Introduction
The current research is part of a larger project, where we aim to study what kind
of knowledge, linguistic or subject matter, is really measured in international assessments of students’ knowledge of mathematics and natural science. The main
purpose of the study is to measure readability of questions in the Swedish part of
TIMSS 2003 (Mullis et al. 2003) and PISA 2003/2006 (OECD 2003 & 2006), in
connection with the actual test results, to see what effect language has for the results, given a multivariate analysis.
TIMSS1 and PISA2 are two international assessment frameworks, both including tests of mathematics and science in regular cycles. In addition to the actual
tests, students, their teachers, and their school principals complete questionnaires
about the contexts for learning mathematics and science, e.g. individual background such as sex and L1/L2. These contextual variables are recorded in the assessment databases, together with test results, and a characterisation of the cognitive level and domain of each test item.
The item is therefore the statistical unit. An item could be a main question, or
a sub question, which, in turn, could share information with other sub questions.
In order to compute readability measures for an item, we therefore first need to
single out what information is relevant for that item. As a first step, we have defined a macro format, i.e. a model of the logical structure of assessments. The
macro format and a genre analysis are used to characterise the items' readability
on a macro level. Linguistic annotations on the micro level, such as vocabulary
usage and information packaging, are under way.
Trends in International Mathematics and Science Study. URL:
Programme for International Student Assessment. URL:
2 Macro format
We have adapted the generic TEI3 P5 annotation model (TEI 2007) to suit our
needs (Forsbom et al. 2007a & 2007b). The adaptation consists of a set of divisions representing the logical structure of the assessments and questions, and a
number of relationships between various question components. The logical components can be visualised as boxes through a web application.
We wanted the logical structure to be straightforward to annotate, so logical
components should therefore optimally be supported by graphical signals. If not,
or in case of a clash, the logical structure was mostly chosen as the primary annotation, while graphical signals were kept as secondary annotations (links and
At a top level, both the TIMSS and PISA material can be classified into three
types of question structures: simple questions (one item), multipart questions
(two or more items), and grouped questions (several simple or multipart questions grouped around a common topic).
In our model, all types have the same basic overall structure, the question set.
The question set contains one or more items. The item is the statistical unit in all
assessments. Each item has a unique ID, which is connected to an item-specific
assessment guide, and to all statistical variables collected for that item.
A question set can also have meta information, i.e. information on the testing
context or the test itself, such as End of Metal Crown section. In addition, it can
contain figures, tables, and leading text, i.e. sections that put the actual question
into context or contain information needed to solve the question, but not directly
states what the student should answer or do. A question set can furthermore be
recursive. Multipart and grouped questions, for example, are represented in this
The item, in turn, contains exactly one core question, and can have zero or
more prompt types. The core question is the central question, stating what the
student should answer or do. The prompt extends the core question, and prompts
the student to react in a certain way, such as formulating an answer or filling in
a table. We have specified the following prompt types: answer line (for open
constructed responses), multiple choice (for option lists), order (for statements
to be ordered), figure (for a figure to draw something in), table (for a table to be
completed), specification (for a more detailed instruction on how to respond),
and true or false (for a set of true/false or yes/no statements). The item can also
include any leading text, figures, and tables that are relevant for that item only.
In some cases, particularly in TIMSS, the answer to one item is required to understand or answer a following item. In those cases, we use a pointer within the
second item to link that information to the previous item.
Although our focus is on textual information, graphical elements, such as figures and tables, are often essential for the understanding of a question, and some3
Text Encoding Initiative. URL:
times contain a fair amount of textual information. As a first step towards integrating textual and graphical information, we have also classified the type of relations between graphical elements and items.
3 Genre analysis
We have used the macro format together with a preliminary genre analysis of the
science items in TIMSS 2003. The items covered the sub areas Life Science (57
items), Physics (46), Earth Science (32), Chemistry (31), and Environmental
Science (28). Most of the items (60%) were multiple-choice questions, while the
rest were open constructed response questions, or question complexes (only in
Life Science, Chemistry and Physics). The distribution, however, varied for the
sub areas, from 74% in Chemistry down to 35% in Environmental Science.
Multiple-choice and open constructed response questions were further subdivided into simple questions (no leading text or diagrams, figures or tables), questions with leading text, and questions with leading text and diagrams, figures or
tables. For Life Science, over half of the multiple-choice questions were simple,
while for Physics, the proportion was only about 10%. Physics, on the other
hand, had the highest proportion of diagrams, figures or tables, particularly in
multiple-choice questions.
According to Knapp & Waters (1994), genres are social processes that describe, explain, instruct, argue, or narrate. The result is used in products such as
reports, explanations, processes, expositions, or stories. Veel (1997) argues that
the domains of language use in school science are doing science by either explaining events scientifically or organising scientific information, and thereby
challenging science. These domains correspond to the main genres explanation,
report and procedure, with a number of sub genres, e.g. procedural recount.
There is also a cognitive progression: procedures and recounts -> sequential explanations -> causal explanations -> factorial explanations -> theoretical explanations.
In TIMSS 2003 Science, procedure was the least frequent genre for all sub
ares (Physics 13%, Earth Science 0%) and report the most frequent genre (except
for Environmental Science 50%). Procedures were almost exclusively recounts,
and reports mainly descriptive, for all sub areas. In Life Science, Chemistry and
Environmental Science, however, there was also a small proportion of taxonomic reports (about 5-10%). The various types of explanations are more evenly
distributed within each sub area. We are planning to extend this preliminary
analysis to all sub corpora to get more substantial results.
4 Readability measures
We are currently annotating the material with micro-level linguistic information,
such as sentences and words. These features will be used in readability measures,
e.g. LIX (Björnsson 1968). For words, we are also adding information on baseform, part-of-speech, together with morphosyntactic and vocabulary features.
These features will be used to measure information packaging (e.g. nominal quotient and number of finite verbs), coherence (types of connectors), and vocabulary usage (word frequency/dispersion and subject-specific terms). We use two
corpora for vocabulary reference: the text book corpus Ord i läromedel (Lindberg & Johansson Kokkinakis 2007) and the general language corpus Stockholm-Umeå Corpus (Ejerhed et al. 2006).
5 Concluding remarks
Our aim is to study readability of the questions in TIMSS 2003 and PISA 2003/
2006, to see what effect language has for the test results. For that purpose, we
have defined a macro format describing the logical structure. The macro format
and a genre analysis have then been used to preliminary characterise the items'
readability on a macro level. Linguistic annotations on the micro level, such as
vocabulary usage and information packaging, are under way.
6 References
Björnsson, C.H. 1968. Läsbarhet. Stockholm: Liber.
Ejerhed, Eva, Gunnel Källgren & Benny Brodda. 2006. V. 2.0. Stockholm-Umeå Corpus.
Dep. of Linguistics, Stockholm University & Dep. of Linguistics, Umeå University.
Forsbom, Eva, Ebba Gustavii & Markus Saers. 2007a. Uppmärkning av TIMSS-frågor
[Mark-up for TIMSS questions]. Uppsala University: project report.
Forsbom, Eva, Ebba Gustavii & Markus Saers. 2007b. Uppmärkning av PISA-frågor
[Mark-up for PISA questions]. Uppsala University: project report.
Knapp, Peter & Megan Watkins. 1994. Context – text – grammar: teaching the genres
and grammar of school writing in infants and primary classrooms. Broadway:
Lindberg, Inger & Sofie Johansson Kokkinakis (eds.). 2007. OrdiL: en korpusbaserad
kartläggning av ordförrådet i läromedel för grundskolans senare år [OrdiL: a
corpus-based study of the vocabulary in secondary school text books]. The institute
for Swedish as a second language, University of Gothenburg. Rapporter om svenska
som andraspråk (ROSA) no 8.
Mullis, I.V.S., M.O. Martin, T.A. Smith, R.A. Garden, K.D. Gregory, E.J. Gonzalez, S.J.
Chrostowski & K.M. O’Connor. 2nd edn. 2003. TIMSS assessment frameworks and
specifications 2003. Boston College, Chestnut Hill, MA: TIMSS International Study
OECD. 2003. The PISA 2003 assessment framework: mathematics, reading, science and
problem solving knowledge and skills.
OECD. 2006. Assessing scientific, reading and mathematical literacy: a framework for
PISA 2006.
TEI Consortium (eds.). 2007. TEI P5 guidelines for electronic text encoding and
interchange. Version 1.0. TEI Consortium.
Veel, Robert. 1997. Learning how to mean – scientifically speaking: apprenticeship into
scientific discourse in the secondary school. In Frances Christie & J.R. Martin (eds.),
Genre and institutions. social processes in the workplace and school, 161-195.
London: Cassell.
7 Acknowledgements
The project is funded by The Swedish Research Council, and is a collaboration between
the Department of Educational Measurements, Umeå University, the Department of Curriculum Studies and the Department of Linguistics and Philology, both Uppsala University.
8 Contact information
Caroline Liberg
Department of Didactics
Uppsala University
Box 2136, 750 02 Uppsala
[email protected]
Eva Forsbom
Department of Linguistics and Philology
Uppsala University
Box 635, 751 26 Uppsala
[email protected]
Testing a reading test –
Construction of reading comprehension in
international reading surveys.
Jenny Folkeryd, Åsa af Geijerstam, Caroline Liberg
Uppsala University
1 Introduction
Although Swedish students from an international perspective read well, a slight
decrease in students’ reading ability has recently been shown in national as well
as international reading tests. Among other results, it has been shown that students with Swedish as a second language show significantly lower results than
students with Swedish as a first language (Skolverket 2004a). It has also been
concluded that girls generally perform better than boys (Rosén, Myrberg & Gustafsson 2005; Skolverket 2004b).
When trying to find explanations for this negative trend, focus has been placed
on socioeconomic changes in society as well as changes within the educational
system. The type of reading ability tested within these studies is more seldom
discussed. The aim of the project discussed in this paper is therefore to investigate the type(s) of reading which is construed through the tests and what thereby
characterizes a student who fails or succeeds. The type(s) of reading construed
through the tests is also investigated in relation to national educational curricula
and reading repertoires developed within and out of school activities. The reading test in focus is the PIRLS-study 2006 (Progress in International Reading Literacy Study). The research focus outlined above is formulated as a set of general
research questions as follows:
• What characterizes the texts used in the reading test?
• What characterizes the answers given by the students? What characterizes
the answers that are accepted/non-accepted according to the scoring guide?
• What characterizes the reading that is construed in the test in relation to national educational curricula and reading repertoires developed within and
out of school activities
Texts, questions, students’ answers, text practises, questionnaires and scoring
guide in PIRLS 2006 are analyzed from a number of perspectives. To investigate how reading is construed, text analyses are performed of text structure and
language resources used to engage readers. Structural analyses are inspired by
methods such as Rhetorical Structure Theory (RST) (Mann & Thompson
1988), Relief analysis (Evensen 2005) and Logic-semantic structure analysis
(Halliday & Matthiesen, 2004). Language resources used to potentially engage
readers are investigated through Appraisal analysis (Martin & White, 2005,
Folkeryd, 2006). Within this semantic framework, language resources for constructing emotion, judging behaviour in ethical terms and valuing objects aesthetically are described. Another question addressed is that of how attitudinal
meaning is intensified, thus creating greater or lesser degrees of positivity or
negativity associated with the feelings (graduation of evaluation).
In the following example such evaluative language resources are exemplified.
Penguins in Antarctica
There are more penguins in the Antarctic than any other bird.
They cannot fly but use their short wings as swimming flippers. They are superb
swimmers. On land, they waddle upright, or move in short hops.
Penguins have many feathers that overlap each other. These, together with
woolly down feathers and a thick layer of fat, keep out the cold air, wind and water. For extra warmth, penguins huddle together in groups.
(PIRLS 2006)
In the text above there is an example of grading the swimming ability of the penguins (superb swimmers) and a grading of the hops as short, thus construing an
association to something small, cute and lovable.
In the next paragraph there is also grading that implies evaluation in various
• Quantification implying importance and focus (many feathers, thick layer of
• Intensification of a situation through the choice of verbs (keep out the cold
air, penguins huddle together in groups) or repetition (cold air, wind and
water). In both of these groups of examples the negative evaluation of the
weather conditions are implied.
Results from the different text analyses are furthermore related to questions in
the tests and how answers are evaluated according to the scoring guide.
Preliminary results show that evaluative language is often found in the paragraphs where students (according to the scoring guide) are expected to find the
correct answers. The answer to one of the questions asked about the penguin text
above (Give three ways penguins are able to keep warm in Antarctica) is thus
expected to focus on the evaluatively intensified passages (many feathers which
overlap, thick layer of fat, huddle together in groups). We have also found that
the questions asked, are often in themselves pointing to passages in the text with
intensified evaluative language resources (Antarctica is the coldest place on
Earth. What other records does it hold?).
When relating the questions of the test to results from the structural text
analysis, it is noted that questions seldom focus on main themes in the foreground of the text but rather on information expressed in the background.
Results from the different text analyses are finally also related to the student
answers and their language background.
Through qualifying and giving a more nuanced picture of what it means to succeed or fail on these tests, the project will contribute theoretically as well as practically to the discussion of what is meant by terminology such as ‘reading’ and
‘reading ability’. This dynamic discussion of reading will also contribute to the
interpretation of results from studies of reading ability.
2 References
Evensen, Lars Sigfred (2005). Perspektiv på innhold? Relieff i ungdomsskoleelevers
eksamensskriving. I: Kjell Lars Berge, Lars Sigfred Evensen, Frøydis Hertzberg och
Wenche Vagle (red.) Ungdommers skrivekompetanse. Bind II. Norskeksamen som
text. Oslo: Universitetsforlaget.
Folkeryd, W. Jenny (2006). Writing with an Attitude. Appraisal and student texts in the
school subject of Swedish. Doktorsavhandling. Uppsala: Institutionen för lingvistik
och filologi.
Halliday, M.A.K. & Matthiesen, Christian, (2004) An Introduction to Functional Grammar. London: Arnold Publishers.
Mann, W.C., & Thompson, S.A. (1988). Rhetorical Structure Theory: Toward a functional theory of text organization. Text, 8 (3) s. 243–281.
Martin, J.R. & White, P.R.R. (2005). The language of evaluation-Appraisal in English.
Palgrave Macmillan.
Pirls 2006 Sample passages, Questions and Scoring Guide
Rosén, Monica, Myrberg, Eva & Gustafsson, Jan-Eric (2005). Läskompetens i skolår 3
och 4. Nationell rapport från PIRLS 2001 i Sverige. The IEA Progress in International
Reading Literacy Study. Göteborgs universitet: Acta Universitattis Gothoburgensis.
Göteborg Studies in Educational Sciences 236.
Skolverket (2004a). Nationella utvärderingen av grundskolan 2003. Huvudrapport –
svenska/ svenska som andraspråk, engelska, matematik och undersökningen i årskurs
Skolverket (2004b). Internationella studier under 40 år. Svenska resultat och analyser.
3 Contact information
Jenny Folkeryd
Department of Didactics
Uppsala University
Box 2136, 750 02 Uppsala
[email protected]
Åsa af Geijerstam
Department of Didactics
Uppsala University
Box 2136, 750 02 Uppsala
[email protected]
Caroline Liberg
Department of Didactics
Uppsala University
Box 2136, 750 02 Uppsala
[email protected]
Word Type Grouping in Swedish Secondary
School Textbooks – An Inventory of Words
from a Second Language Perspective
Inger Lindberg, Sofie Johansson Kokkinakis
Stockholm University, University of Gothenburg
1 Introduction
When investigating questions like “What characterizes the vocabulary of school
book texts in general and in particular subjects at different levels?” and “Which
words present particular problems for students studying in their second language?”, school text book corpora offer important empirical data. We have therefore compiled and analyzed a one millon word corpus of Swedish secondary
school textbooks (OrdiL). The texts represent eight different school subjects
contributed by three different publishers (Lindberg & Johansson Kokkinakis
2007). In order to carry out comparative studies and analyses on the vocabulary
of these texts we have performed computer-based disambiguation and semantic
analyses at a lemma- and lexeme-level. Moreover, we have carried out word frequency analyses, including comparative analyses of relative frequency and dispersion between different subjects. To be able to distinguish characteristic features of school book vocabulary we use an equally sized reference corpus of
easy-to-read texts (Mühlenbock, 2008) as a point of comparison. The vocabulary
of this corpus is representative to the kind of written language that children at this
age can be exposed to out of school.
2 Word type grouping
To identify and categorize various types of words in textbooks, we propose a
model based on earlier research by Coxhead & Nation (2001) and Hyland & Tse
(2007) modified to account for the word types of potential difficulty for second
language secondary school students. As for the classification of word types Coxhead & Nation group text words into word families and furthermore classify
them into three groups identified in relation to frequency, register and range:
(1) high frequency words (belonging to the 2000 most frequent and widely used
word families, covering about 80 % of most texts),
(2) academic vocabulary (words which are reasonably frequent in academic
writing and corresponding to some 8-10 % of the running words) and
(3) technical vocabulary (which differs by subject area and covers up to 5 % of
texts) Coxhead & Nation (2001).
In our study we have chosen to disambiguate words at a semantic level to capture both
homographic and polysemic words. Based on analyses of frequency, range and distribution we propose two main groups of words divided into at least two subgroups. The
main division is between cross-disciplinary and disciplinary specific words.
Cross-disciplinary words:
(1) the 1000 most frequent words (lemmas) hereby counting only nouns, verbs,
adjectives and adverbs. These are words that would be found among the most
frequent ones in most texts. Examples of these words are: göra ‘to do’, vara ‘to
be’, ha ‘to have’, hur ‘how’, stor ‘big’, inte ‘not’, år ‘year’.
(2) School-related words often associated with formal written language corresponding to the so called ‘academic words’ referred to in earlier research. Examples of these words are: motsvara ‘correspond to’, utbredning ‘extension’,
föremål ‘object (noun)’, avta ‘decline’.
Disciplinary specific words:
(3) domain-related words also appearing in every day language and linked to a
specific school subject. Examples are: blandning ‘mixture’, klimat ‘climate’,
muskel ‘muscle’, helig ‘holy’.
(4) technical terms often unique for a particular subject. Examples of such words
are: produktionsfaktor ‘factor of production’, kopplingsschema ‘wiring (connection) diagram’, kromosom ‘chromosome’.
3 Vocabulary tests
To explore the question “Which words present particular problems for students
studying in their second language?” we also need empirical data on students’ actual
vocabulary difficulties. For this purpose we are constructing a number of computerized tests focusing various types of lexical knowledge in relation to the different word types. These tests will be administered to students with different background characteristics to make comparisons of quantitative as well as qualitative
aspects of vocabulary knowledge between different groups of students possible.
In order to cover different aspects of word knowledge we have designed a
model of lexical knowledge. which forms a matrix representing depth – width as
well as receptive – productive qualities of lexical knowledge. The idea is to have
different vocabulary tests measuring the different aspects covered in the matrix
of lexical knowledge. Test data consist of vocabulary occurring in the seconday
school text book corpora (OrdiL).
The need for evidence-based assessment tools focusing the academic language proficiency of second language students is becoming increasingly obvious. In many schools in multilingual settings decisions whether second language
students are ready to meet the language demands of content area instruction and
assessment are based on the results of language tests constructed for the assessment of communicative and more social language skills which are not good predictors of language performance for more academic purposes (Bailey 2007).
4 References
Bailey, A.(ed.), 2007. The language demands of school. Putting academic English to the
test. New Haven and London: Yale University Press.
Coxhead, A. & Nation I. S. P. 2001. “The specialized vocabulary of English for academic
purposes”. In J. Flowerdew & M. Peacock (Eds.), Research perspectives on English
for academic purposes:252-267. Cambridge: Cambridge University Press.
Hyland, K., Tse P. 2007. “Is There an “Academic Vocabulary?”. In TESOL Quarterly,
Volume 41, Number 2, June 2007: 235-253(19).
Lindberg I., Johansson Kokkinakis S. 2007. (Eds.), OrdiL - en korpusbaserad kartläggning av ordförrådet i läromedel för grundskolans senare år. ROSA-rapport 8. Institute
for Swedish as a Second Language, Gothenburg University.
Mühlenbock K. 2008. Legible, readable or plain words - presentation of an easy-to-read
Swedish corpus. Readability and Multilingualism, workshop at 23rd Scandinavian
Conference of Linguistics.
5 Contact information
Inger Lindberg
Department of USOS
Stockholm University
SE-106 91 Stockholm
[email protected]
Sofie Johansson Kokkinakis
Department of Swedish Language
University of Gothenburg
Box 200
SE-405 30 Göteborg
[email protected]
Studia Linguistica Upsaliensia
Department of Linguistics
Editors: Joakim Nivre and Åke Viberg
Jörg Tiedemann, Recycling translations. Extraction of lexical data from parallel
corpora and their application in natural language processing. 2003
Agnes Edling, Abstraction and authority in textbooks. The textual paths towards
specialized language. 2006
Åsa af Geijerstam, Att skriva i naturorienterande ämnen i skolan. 2006
Gustav Öquist, Evaluating Readability on Mobile Devices. 2006
Jenny Wiksten Folkeryd, Writing with an Attitude. Appraisal and student texts
in the school subject of Swedish. 2006
Ingrid Björk, Relativizing linguistic relativity. Investigating underlying assumptions about language in the neo-Whorfian literature. 2008
Joakim Nivre, Mats Dahllöf and Beáta Megyesi, Resourceful Language Technology. Festschrift in Honor of Anna Sågvall Hein. 2008
Anju Saxena & Åke Viberg, Multilingualism. Proceedings of the 23rd Scandinavian Conference of Linguistics. 2009
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF