You are on page 1of 476

Connectivity in Grammar and Discourse

Hamburg Studies on Multilingualism (HSM)


Hamburg Studies on Multilingualism (HSM) publishes research from
colloquia on linguistic aspects of multilingualism organized by the Research
Center on Multilingualism at the University of Hamburg.

Editors
Jürgen M. Meisel
Monika Rothweiler
Juliane House
University of Hamburg
Research Center on Mulitlingualism

Volume 5
Connectivity in Grammar and Discourse
Edited by Jochen Rehbein, Christiane Hohenstein and Lukas Pietsch
Connectivity in
Grammar and Discourse

Edited by

Jochen Rehbein
Christiane Hohenstein
Lukas Pietsch
Middle East Technical University, Ankara
and University of Hamburg

John Benjamins Publishing Company


Amsterdam / Philadelphia
TM The paper used in this publication meets the minimum requirements of
8

American National Standard for Information Sciences – Permanence of


Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data

Connectivity in grammar and discourse / edited by Jochen Rehbein, Christiane


Hohenstein and Lukas Pietsch.
p. cm. (Hamburg Studies on Multilingualism, issn -3363 ; v. 5)
Includes bibliographical references and indexes.
1. Grammar, Comparative and general--Connectives. 2. Discourse analysis. I.
Rehbein, Jochen. II. Hohenstein, Christiane. III. Pietsch, Lukas.
P302.27.C66   2007
415--dc22 2007060663
isbn 978 90 272 1925 1 (Hb; alk. paper)

© 2007 – John Benjamins B.V.


No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands
John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
The production of this series has been made possible
through financial support to the Research Center on Multilingualism
(Sonderforschungsbereich 538 "Mehrsprachigkeit")
by the Deutsche Forschungsgemeinschaft (DFG).
Table of contents

Connectivity as an object of linguistic research in multilingualism 1


Jochen Rehbein, Christiane Hohenstein, Lukas Pietsch

Section 1. Aspects of language change and language acquisition


Grammaticalization of converb constructions: The case of Japanese –te
conjunctive constructions 21
Masayoshi Shibatani

Contact, connectivity and language evolution 51


Yaron Matras

Allora: On the recurrence of function-word borrowing in contact situations


with Italian as donor language 75
Thomas Stolz

Some notes on the syntax-pragmatics interface in bilingual children: German in


contact with French / Italian 101
Natascha Müller

Section 2. Pronouns, topics and subjects


Distribution and function of clitic object pronouns in popular 16th-18th
century Greek narratives: A synchronic and diachronic perspective 139
Chrystalla A. Thoma

Nominative subjects of non-finite clauses in Hiberno-English 165


Lukas Pietsch

Section 3. Finiteness in text and discourse


Aspectotemporal connectivity in Turkic: Text construction, text subdivision,
discourse types and taxis 187
Lars Johanson

Connectivity by means of finite elements in monolingual and bilingual


Turkish discourse 199
Birsel Karakoç
 Connectivity in Grammar and Discourse

Section 4. Subordination – coordination


Alternative subordination strategies in Turkish 231
Celia Kerslake

Studying connectivity with the help of computer-readable corpora:


Some exemplary analyses from modern and historical, written and
spoken corpora 259
Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner
and Ludger Zeevaert

Discourse coordination in Turkish monolingual and Turkish-German bilingual


children’s talk: işte 291
Annette Herkenrath

Section 5. Adverbials, particles and constructions


Modal adverbs as discourse markers: A bilingual approach to the study
of indeed 329
Karin Aijmer

„So, given this common theme ...“: Linking constructions in discourse across
languages 345
Kristin Bührig and Juliane House

An utterance-transcending connector: Particle to in utterance-final position


in Japanese business reporting 367
Yuko Sugita

Between connectivity and modality: Reported speech in interpreter-


mediated doctor-patient communication 395
Thomas Johnen and Bernd Meyer

Matrix constructions 419


Jochen Rehbein

Language index 449

Name index 451

Subject index 457


Connectivity as an object of linguistic
research in multilingualism

Jochen Rehbein, Christiane Hohenstein, Lukas Pietsch


Middle East Technical University, Ankara, and University of Hamburg

The contributions in this volume investigate the role played by various linguistic ele-
ments in interconnecting units of text and discourse. We discuss this role, the linguistic
forms involved and their functions under the term of ‘connectivity’. Connectivity con-
cerns issues of linguistic interaction in its different aspects of grammar, prosody, text and
discourse. Since it has repercussions on issues of multilingualism and language contact,
as well as language change and language acquisition, the topic of connectivity is of rele-
vance to various research projects working within the Research Center 538 ‘Multilin-
gualism’ in Hamburg, out of which the present collection has developed (cf. House and
Rehbein 2004). Of particular interest are the various linguistic means individual lan-
guages use to express connectivity. To date, even though connectivity is an area of re-
search that has produced a large number of studies, the field is still rather opaque. Thus,
a few structuring and systematic introductory remarks are in order.
To begin with, we discuss (1) a general outline of an analytic approach to linguistic
connectivity, and (2) different aspects through which it manifests itself in language. In
an outline of the book (3), we relate specific fields of linguistic research pertaining to
connectivity to the article sections and their main topics. As eminent fields of linguis-
tic research, both language change and language acquisition offer a more general per-
spective on the role of connectivity and connectives (3.1). The subsequent chapters –
‘Pronouns, topics and subjects’ (3.2), ‘Finiteness in text and discourse’ (3.3),
‘Subordination – coordination ’ (3.4), and ‘Adverbials, particles and constructions’
(3.5) – deal with particular linguistic means which are highly language specific. Ac-
cordingly, connectives relating to various grammatical concepts are discussed with
regard to their discourse functions and effects.

1. Linguistic connectivity

During the 1980s, connectivity was addressed frequently as a connection between


“clauses”, and discussed in terms of “subordination” (Mackenzie 1984, König and van
der Auwera 1988). Phenomena such as “(hypotactic) clause combining” (Matthiessen
 Jochen Rehbein, Christiane Hohenstein, Lukas Pietsch

and Thompson 1988), and procedures of “clause chaining” (Slobin 1988) formed a
focus of interest. That is, connectivity between linguistic units internal to the sentence
or an utterance, mostly between their constituents, ranked high with a discernible re-
search interest in ‘surface’ structures of language. However, under a contrastive per-
spective, aiming at the discovery of universals in propositional encoding, entities be-
yond the clause came into view. Forms used to encode a sequence of propositions were
discussed e. g. with regard to the notions of “nexus” (of clauses) and of “linkage” (Foley
and Van Valin 1984, Van Valin and LaPolla 1997). Characteristically, utterance-inter-
nal connections serve to integrate their conjunct elements into a single, complex hier-
archical whole and submit them to a common illocutionary force (cf. Foley and Van
Valin 1984: 239f.).
Since connectivity is organised in ways highly specific to individual languages,
perspectives from non-Indo-European languages and contrastive studies have led to
insights in effects on text and discourse structure according to the type of connectivity.
For instance, gerund constructions in Japanese, especially the converbial “-te” form,
have received a steady and growing interest (cf. Myhill & Hibiya 1988, Hasegawa 1996,
Grein 1998, Shibatani, this vol.). Equally, Turkish and Turkic converbs (Johanson
1995) continue to offer new vistas on intricate patternings in discourse up to the point
of setting up specific discourse types (cf. Karacoç, this vol.; Johanson, this vol., Kers-
lake, this vol.).
Studies on the phenomenon of textual ‘connexity’ (cf. Conte et al. 1989, Heydrich
et al. 1989) and, more prominently, the notions of ‘coherence’ and ‘cohesion’ (Halliday
and Hasan 1976) helped to develop a body of research on interrelated topics, dealing
with linguistic devices overarching the boundaries of single sentences and single utter-
ances (cf. also, in terms of text linguistics, van Dijk 1979, de Beaugrande 1980, contri-
butions in Cambourian 2001, and others). At the same time, research on “connexion”
in terms of information-processing within the confines of sentence-based linguistic
units, linking “sentence content”, and leading to “textual coherence”, is still growing (cf.
Fabricius-Hansen 2000). – In their extensive study of the language development in
English, German, Spanish, Hebrew, and Turkish, based on narratives of frog stories,
Berman and Slobin (1994) propose connectivity of the informational structure of dis-
course as an essential concept of analysis: “... we are studying the development of the
capacity to describe situations. We are also concerned with the ways in which individ-
ual events are related to each other. That is, we are studying the development of linguis-
tic means to connect events and syntactically “package” them into coherent structures
– at the level of scene, episode, and overall plot.” (Berman and Slobin 1994: 1–2).
We can conclude that ‘connectivity’ as a notion comprises both utterance-internal
and utterance-external linguistic devices of various categories (cf. Rehbein 1999).
Cross-linguistically, it can be observed that connectivity manifests itself on the ‘lin-
guistic surface’ along a continuum from subordinating connectives through “serialisa-
tion” by means of converbs, to coordinative devices which may span not only clauses
and phrases, but may link complete utterances, sentences, even parts of text and dis-
Connectivity as an object of linguistics 

course to each other (cf. Rudolph 1996, Bisang 1995, 1998b, Johanson 1995, Haspelmath
1995, Kortmann 1996).
The phenomenon of connectivity has been receiving increasing attention from
linguists, especially from a typological and contrastive perspective (van der Auwera
and Bultinck 2001; Behrens and Fabricius-Hansen 2002; Berman 1998; Rehbein 2002;
Unger 1996; Verhagen 2001; Grein 1998; Hasegawa 1996; Horie 2000 etc.). Aspects of
meaning and formation of discourse particles more and more come into focus (e.g.
Mori 1999; Onodera 2004; Park 1998; Fischer 2006). A new domain of connectivity
was opened up by the debate about the pragmatics of “small words”, in the English
speaking community about ‘discourse markers’ (Schourup 1985, Schiffrin 1985, Fraser
1990 a.o.) and ‘gambits’ (Edmondson 1978, House and Kasper 1981), in the German
speaking community about ‘Partikeln’ (cf. e.g. Weydt (ed.) 1979, 1983, 1987).
Moreover, connectivity pertains not only to ‘surface’ elements of language such as
phrases, clauses, or sentences. It also applies to illocutions, propositions and mental/
cognitive entities, such as particular elements of knowledge, to patterns of (inter-)ac-
tion and to types of discourse (cf. Hohenstein 2005, 2006 a,b,c, ch. 6.2). This type of
mental/cognitive connectivity is established incessantly during the interaction process
between speakers and hearers, writers and readers. Thus, “connectivity” is not to be
identified with “context” (cf. Rehbein and Kameyama 2003; House 2006). Rather, the
development and expansion of linguistic means of connectivity – in diachronic lan-
guage change and in ontogenetic language development – actually render ‘text’, at least
partially, independent of ‘context’.
In the following discussion, emphasis will be laid on discursive and textual con-
nections established by the grammatical and interactional potential of more or less
complex linguistic devices. While they all serve to connect linguistic actions in their
various dimensions, they belong to different formal and functional categories. They
are of different internal make-up, and thus constitute different linguistic activities and
processes. The guiding hypothesis in this is that connectivity frequently constitutes
hinges within the flow of discourse and text (s. Bührig and House, this volume; cf. the
German term ‘Scharnier’ as well as Jakobson’s 1957 concept of ‘shifter’).

2. Aspects of connectivity

There is no clear-cut division of labour between grammar and discourse in the sense that
grammar would exclusively concern the utterance internal connectivity whereas dis-
course/text concerned that between utterances. Rather, one may say that there is an in-
tertwining of grammar and discourse/text in linguistic connectivity (or: connectives).
Therefore, if one looks at the characteristics of connectivity, one should account for both,
grammar as well as discourse and text.
As for discourse and text, linguistic connectives can generally be described as join-
ing a linguistic action X with a linguistic action Y. Since what is joined are frequently
 Jochen Rehbein, Christiane Hohenstein, Lukas Pietsch

not full-fledged linguistic actions, but specific dimensions of these, we speak of “di-
mension X” and “dimension Y” of different linguistic actions being joined.
Dimensions of linguistic actions are:
a. Constellation: Speech situation, categories of action space (motivations, needs, in-
teractional space, control field etc.)
b. Utterance act: Syntax, morphology, phonetics/phonology, prosody etc.
c. Illocutionary act: Illocutions, utterance mode, etc.
d. Propositional act: Semantics, argument, predicate, predication, symbol fields, etc.
e. Knowledge: Presupposition, elements of knowledge, knowledge space, knowledge
structure, expectations, knowledge differential, linguistic knowledge, etc.
f. Other mental processes: Evaluation, imagination, understanding, planning, etc.
g. Cooperation: Back-channel activities, interaction management, speaker-hearer-
steering devices, reception, etc.
All the dimensions mentioned above can become either starting point or target point of
a connective: A connective starts out at a dimension X of one linguistic action as a
starting point, and joins it to a dimension Y as the target point of another linguistic
action. For instance, in the case of a phoric pronoun the starting point is a nominal or
another propositional element, while the target point is the phoric expression, which
serves to copy the nominal or propositional element into a different syntagm. Con-
nectivity then manifests itself in the process of coreferring, the phoric procedure. In the
case of coordination, one propositional element is the starting point, while the target
point is a second propositional element of the same grammatical category; in that case,
the connective itself implements the process of joining them under a common category
of propositionally based knowledge. Thus, the connective acts as a morphosyntactic
trigger element of discursive/textual knowledge building.
Prosody may be described as a formal domain of its own, since it interacts with the
grammatical components, especially in spoken discourse. Quite often we are dealing
with a connection between prosodic units joined into a larger schema (cf. Chafe 1988,
1994). Some of the contributions in this volume take into account the connective func-
tions of prosodic and phonetic / phonological resources (e.g. Kerslake, Matras, Thoma,
Sugita, cf. this volume).
Thus, connectivity between what we may term as the dimensions of a linguistic
action X and a linguistic action Y is established by means of different linguistic devices.
With regard to their linguistic form, i. e. to grammar, they can be described according
to their phonetics/phonology, morphosyntax, constructional make-up and other crite-
ria, depending on the theoretical background of the analyst. As for the functions of
linguistic devices realizing connectivity, these may be found – and diagnosed – if one
looks at what is done with the forms on the level of discourse and text. Then, functions
of connectives can be characterised as the purpose inherent in their forms. Frequently,
these inherent purposes are located in the mental-cognitive processes on the part of the
hearer/reader which establish connections between different dimensions of linguistic
Connectivity as an object of linguistics 

actions as aspects of their own (Givon 1995 pointed out that the connectivity of a text
is based on its “structured mental representation... The more connections a node has...
the more mentally-accessible it is.” (Givon 1995: 64)). The majority of connectives in-
volves procedures that are operative, in so far as they work on linguistic dimensions,
instructing a hearer/reader to process parts of the propositional act in relation to each
other (for the ‘field’ characteristics of linguistic procedures cf. Rehbein this vol.). Com-
posite deictic procedures (Rehbein 1995a; for contrasts in English-German transla-
tions, cf. Bührig and House 2004, for a comparative analysis in German and English
academic articles, cf. Fandrych and Graefen 2002), and connectives of formal coopera-
tion (back channel cues and the like, cf. Ehlich 1986, 1987), can be regarded as typical
cases, since, from the point of view of procedural pragmatics, connective devices are
rather complex structures, encompassing multi-procedural components as combina-
tion, ensemble, integration etc. of linguistic procedures.

3. Outline of the book

3.1 Language change and language acquisition

Several contributions in this volume investigate the relation between connectivity and
language change. For instance, Masayoshi Shibatani demonstrates how Japanese con-
verb constructions with motion verbs (-te iku/kuru, conv go/come) undergo several
stages of change before they grammaticalize into a new grammatical class of aspectual
suffixes. There are, for instance, stages of a solidification of constructions on the pho-
netic and syntactic level – depending on sucessive changes in the main verb, e.g. a loss
of valency – but not on the morphological level. In other cases there is a decreasing
degree of congruity of verb and converb. Actually, both grammaticalization types,
gradual and instantaneous grammaticalization, are observed with these Japanese con-
verb constructions. That is, the steps leading up to the point where a construction or a
part of it finally changes its grammatical category can be revealed only by fine-grained
syntactic analysis and if differential linguistic context is taken into account.
If, as Bisang 1998a states with regard to serial and relative clause constructions in
various African and Asian languages, “constructions and the human equipment are
somehow involved in the propagation of linguistic changes” as well as “the factor of
sociolinguistics” (Bisang 1998a: 51), in multilingual communities, language change
may be induced by contact due to plurilingual people communicating with each other in
the contact languages (cf. Stolz and Stolz 1996/97; Stolz, this volume, Matras 2002 and
this volume). Many of the connective procedures are sensitive to contact, insofar as
they import linguistic devices from a contact language, copy them, integrate them or
modify their own forms under conditions of contact (cf. Johanson 1992b, 1999a,
Muysken 2000, Clyne 2003, Heine and Kuteva 2005). New connectives are frequently
created in mixed languages (cf. Bakker and Mous 1994, Matras 1996, 2000, Matras and
 Jochen Rehbein, Christiane Hohenstein, Lukas Pietsch

Bakker 2003, Bakker 2003, Stolz 2003, etc.). Apparently it is due to the complexity of
their linguistic structure (both synchronically and diachronically speaking) and func-
tion that connectives are not only sensitive to contact-induced language change but
vulnerable to it (Müller 2003). According to Yaron Matras (this volume), certain lan-
guages such as Turoyo, Domari, Romani and others have used various languages from
their respective environments as sources of new connectives, which belong to the types
of coordination, illocutionary modification, and interaction management. The shift of
one expression, structure or construction into a new category or a new paradigm often
happens in a complex process involving functional shift, splits in syntax, morphology,
word boundaries and semantics, to the point of their fusion. Matras also spells out
some sociolinguistic conditions for replication and points towards the mental base of
contact-induced change: “Connectivity structures, then, appear particularly vulnera-
ble not just to the actual borrowing for formal material such as conjunctions and other
particles, but also to the replication of patterns of constituent ordering, agreement and
overall form-function mapping that form the mental blueprint from the respective
construction.” (Matras, this volume).
Thomas Stolz (this volume) points to the interesting fact that the item allora, bor-
rowed from Italian into (Italo-)Greek, (Italo-)Albanian, Cimbrian, Molise Slavic and
Maltese, failed to acquire the full range of functions in the receiving languages that it
had in the donor language. His findings point to a dominant role played in contact by
the category of temporal deixis, which later develops further into functions of coordi-
nation and others. In a similar vein, Herkenrath (this volume) observes that the item
işte in German Turkish develops a mixed deictic-coordinative type of use, pointing
towards an instance of partial contact-induced grammaticalisation. Stolz explains dif-
ferences observed between contact results in different languages with language-typo-
logical as well as sociolinguistic factors. In this respect, his approach differs from that
of Matras, who regards interaction management as the dominant category governing
contact-induced effects.
Ultimately, linguistic elements may shift into a new formal class and thereby change
their functional behaviour with respect to connectivity too. Under a functional-prag-
matic perspective, grammaticalization can be characterized as a process whereby com-
plex procedures are transposed within a linguistic field or to a new linguistic field (such
as procedural ensembles in the case of matrix constructions or composite deictics; cf.
Rehbein, this vol., Johnen and Meyer, this vol.). Restructuring within the utterance may
then be the consequence of an increase in context dependence, as in the case of shift to-
wards aspect semantics of the converb constructions in Japanese (Shibatani, this vol.).
In bilingual language acquisition, connective elements seem to play an important
role, all the more so since connective elements make up a large part of all functional
elements (Muysken 2000). Natascha Müller (this volume) notes that in certain gram-
matical domains – e.g. null-subject properties, object clitics, complementation – there
is not an absolute separation of languages in bilingual acquisition (cf. Meisel 1989,
Genesee 1989, Genesee, Nicoladis and Paradis 1995 etc.), but some amount of positive
Connectivity as an object of linguistics 

cross-linguistic influence, which may result in a partial acceleration of acquisition


processes on both sides. She relates these findings to the concept of a common syntax-
pragmatics interface, where a shared pragmatic module may help in preparing the
ground for syntactic acquisition in both languages. Müller also argues, based on Lopez
(2003), that “phases” in the sense of Chomsky’s (2001) theory of syntactic derivation
can be interpreted pragmatically. Thus, both in German and in French matrix con-
structions, pragmatically interpreted subordinated structures can be classified as con-
nectives according to factors of presuppositions and language contrast, independently
of their formal-syntactic means of expressing subordination (with or without the com-
plementizer dass/que).
A comparable finding is the reinterpretation of Turkish aspecto-temporal finite
forms as deictic terms in Turkish–German bilingualism (Rehbein and Karakoç 2004,
Karakoç, this vol. cf. below).

3.3 Pronouns, topics and subjects

Topic as a central formative element of thematic organization is a subdomain of lin-


guistic information processing. The manifestation of topic connectivity (see e.g. Aars-
sen 1998) is strongly sensitive to typological parameters. Languages can be ordered
typologically along a continuum between “subject-prominent” and “topic-prominent”
(cf. Li and Thompson 1976). Of these, the subject-prominent languages make more
heavily use of phoric expressions and can therefore also be characterized as “phoric
languages”. On this scale, languages like Japanese and Turkish are more on the topic-
prominent side, while German, English and French are more on the subject-promi-
nent (phoric) side. This distinction partly correlates with the distinction made in UG-
based approaches between “Pro-drop” (or “Null Subject”) and “Non Pro-drop” (or
“Non-Null Subject”) languages (Jaeggli and Safir 1989, Haegeman 1997, Rizzi 1997).
Phoric expressions (~ PRO-elements) are the “pronouns” of the 3rd person (German
er, sie, es; English he, she, it; French il, elle; including their case-inflected forms). In
many languages these have no free-word counterparts. The difference between phoric
and deictic “pronouns” (“pronoms”) was pointed out first by Benveniste (1956). The
purpose of phoric expressions is to encode the continuity of a topic in such a way as to
keep it mentally activated for the hearer/reader (Hoffmann 1997: 844f.)
PRO-elements are thus linguistic devices of connectivity typical of phoric languag-
es. Their syntactic distribution is often identical with that of free noun phrases; in par-
ticular, they can take up subject or object positions, typically depending on the valency
of the verb. Grammatically, these phoric elements are pronouns in the narrow sense.
On the other hand, phoric elements may be clitics (as in the Romance languages,
some Semitic languages, or in spoken German): In Greek (Chrystalla Thoma, this vol-
ume), cliticized, weak phoric pronouns have developed. They are employed for topi-
calization procedures under “clitic object doubling” (cf. Haberland and van der Auwera
1990). Thoma states an interrelationship between the preverbal position of unstressed
 Jochen Rehbein, Christiane Hohenstein, Lukas Pietsch

third person (weak) pronouns and the deployment of discourse pragmatic functions
within the historical language development. “In Early Modern Greek, fronted elements,
not necessarily contrastive, attract clitic pronouns to pre-verbal position, until that po-
sition becomes the norm and is grammaticalised in Standard Modern Greek. A similar
process has taken place in the Romance languages...” (Thoma, this volume; cf. also Janse
2000). It seems to be of high interest that phoric elements of this kind have been devel-
oped in an otherwise topic-based language, and despite the fact that other topic based
languages were spoken in the environment of the Greek language.
An instance where the coding of connectivity is affected by the case-marking on
pronouns comes from Irish English (Lukas Pietsch, this volume). In Irish English (Hi-
berno-English), gerund constructions are found to have changed their syntactic be-
haviour under contact with Irish. While Standard English in such constructions dis-
plays case marking properties that overtly reflect the subordinate character of the
construction (either accusative or genitive), Hiberno-English developed a usage of
nominative subject pronouns instead. Pietsch argues that this morphological change
reflects a change in the underlying syntactic properties of the constructional pattern,
obliterating the effect of “subject-to-object raising” (or “exceptional case marking”)
that motivates the predominant accusative marking in the Standard English system.
Hiberno-English has replicated a pattern from Irish where subject pronouns fail to
mark their subordinate status by means of case. Interestingly, Irish itself, while lacking
the device of case marking to make this distinction, employs another, syntactic one,
marking subordinate clauses by means of different word order. As this syntactic dis-
tinction was not replicated in English, the net effect is that the contact variety shows an
overall reduction in the overt coding of subordination, having fewer overt formal dif-
ferences between non-finite subclauses and finite clauses than either of the two source
languages have.

3.4 Finiteness in text and discourse

The connectivity created by finite elements (cf. Bisang 2001, Johanson 1994, 1999b)
results from the fact that a finite element is characterised by its role as the carrier of
predication, the procedure whereby a predicate is assigned to an argument. Lars Johan-
son (this volume) discusses a type of “chain sentence” characteristic of Turkish literary
language, where there is a series of non-finite utterances anchored in the narrative con-
stellation only through the basic finite element at the very end, thus establishing aspec-
tual/temporal, modal and illocutionary features of the earlier utterances only in retro-
spect. The role of finiteness may also be sensitive to parameters of text and discourse
type. For instance, as Johanson remarks, there was an influence of French prose on the
structuring of connectivity in late Ottoman literature. (Cf. also Johanson 1971, 1992a).
Birsel Karakoç (this volume) finds that bilingual German-Turkish children acquire
finite elements in Turkish later than monolingual Turkish children. In monolingual Turk-
ish these finite elements have a function of establishing connectivity on a discourse-type
Connectivity as an object of linguistics 

level. In contrast, the bilinguals develop a stronger use of a range of temporal deictic ex-
pressions to extend their linguistic capabilities in Turkish. The difference is most marked
in the use of the Turkish evidential suffixes. (Cf. also Rehbein and Karakoç 2004).

3.5 Subordination – coordination

Subordination includes phenomena such as attribution, relativization, and complemen-


tation. Depending on the language, these phenomena may manifest themselves in gram-
matical elements such as complementizers, relativizers, wh-elements, case, non-finite-
ness, or syntactic position (Mackenzie 1984, Haiman and Thompson 1988, Redder 1990,
Fabricius-Hansen 1992, Müller 1993, Verhagen 2001 a.o.). An important typological
parameter in this respect is that of “deranking” of verb forms, which can be described
along a cline from finite through subjunctive to nonfinite (cf. Croft 2001, ch. 9).
Celia Kerslake (this volume) presents a classification of subordinate clauses in
Turkish. Of special interest is a type of finite subordinate clause with a subordinator in
the initial position, because it does not fit the canonical typological classificatory as-
sumptions of Turkish as a left-branching word order language. This phenomenon is
contact-related, as the subordinator in question, ki, which introduces subordinate
clauses with a postnominal or postpredicative placement, is borrowed from Persian. In
Persian its meanings include ‘that’, ‘who’, ‘which’, ‘when’, ‘so that’, etc. Kerslake con-
cludes: that “wherever Turkish (or another Turkic language) is exposed to prolonged
contact with a politically and/or demographically dominant Indo-European language
(Iranian, Slavic, Germanic, etc), there is a strong tendency for attrition to occur in the
indigenous (left-branching, non-finite) type of subordinate clause in favour of finite
right-branching clauses introduced by a subordinating conjunction [...]” (Kerslake,
this vol.; for the grammatical background, s. Göksel and Kerslake 2005).
In Turkish, ki-constructions serve the language-internal management of pragmat-
ic and rhetorical purposes, in a way that meets psycholinguistic (universal) require-
ments of language processing in spoken language. This renders ki a linker of finite
clauses to a postpredicative subordination and, in this way, enables speakers to process
unplannend propositions in a syntactic-semantic relationship.
Coordinating connectivity is often described as symmetric, but it may actually link
any kind of element (conjunct) to any kind of preceding element (conjunct) (cf.
Haspelmath ed. 2004). Coordination is mostly concerned with the linking of proposi-
tional elements (cf. Redder 2006). Through coordination, that conjunct which is set off
from the coordinator through a structural cesura is categorially upgraded and turned
into the lead category for the other conjunct (cf. Matras 1997, 1998). While coordinat-
ing connectivity may join individual words or phrases within an utterance, it can also
join entire utterances, illocutionary acts or even whole linguistic actions to each other.
Annette Herkenrath (this volume), in a corpus-based study comparing bilingual
acquisition of Turkish and German in Germany with monolingual acquisition of Turk-
ish in Turkey, finds differences in the functional diversification of the discourse con-
 Jochen Rehbein, Christiane Hohenstein, Lukas Pietsch

nective işte. A specific coordinative use abundant in Turkish narrative discourse seems
to be lost or play a minor role in German Turkish (cf. Herkenrath, Karakoç and Reh-
bein 2003). Acquired late even by monolinguals as a linguistic device of sequential and
concatenative discourse connectivity (cf. Özbek 2000, Yılmaz 2004), this kind of dis-
course coordination thus may be undergoing functional change in German Turkish.
Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and
Ludger Zeevaert (this vol.) conduct a corpus-based study of functional characteristics
of coordinating elements in diverse languages – Modern English and German, Old
Swedish and Turkish. They demonstrate how such an approach can lead to conclusions
with regard to more general insights into coordination. Thus, they hypothesize that
coordinating connectives may undergo functional innovation or diversification trig-
gered by coordination on macrosyntactic textual and discourse level. They also con-
sider general issues regarding computer­aided quantitative and qualitative methods of
corpus analysis.

3.6 Adverbials, particles, and constructions

Adverbial constructions (cf. e.g. van der Auwera 1998) serve to relate a base construc-
tion to preestablished knowledge, and to add a new knowledge relation or evaluation to
it. Karin Aijmer (this volume) focusses on the range of functions displayed by the item
indeed, drawing particular attention to a number of rhetorical functions associated
with it (cf. White 2003). Through a process of gradual grammaticalization, this item has
developed functions as a discourse marker (cf. Schiffrin 1987, Fraser 1990, Aijmer and
Simon-Vandenbergen 2006). Observing textual choices made by translators, Aijmer
finds that indeed tends to attract special attention when used as an adverbial, as it tends
to be explicitly rendered more regularly than other types of connectives, which are of-
ten omitted in translation.
Complex composite adverbials, like in fact, any time, in addition, often stand extra­
posed at the sentence periphery, joining the base construction to which they are at-
tached, with overarching large-scale textual segments. Thus they act as hinges, joining
larger propositional structures (such as argumentations) with the more specific speech
actions of which they are part. Kristin Bührig and Juliane House (this volume) call such
connectives “linking constructions” (cf. also Bührig 2003, Leuschner 1998 on options of
these constructions for grammaticalization). In complex texts, they can also be used as
advance organizers which vary cross-linguistically (cf. Gülich 1970, Clyne 1987 a.o.).
Particles form another subgroup of connectives to be mentioned here (for litera-
ture in a cross-linguistic/contrastive perspective, cf. Fischer 2006, Weydt 1983, 1987,
König 1991, Liedke 1994, Nekula 1996, Cardenes Melián 1997, van der Wouden, Foo-
len and van de Craen 2003, and others). Generally speaking, particles stand in a three-
fold relation: (1) they establish a relation to a preceding linguistic action, mostly in
propositional and illocutionary terms; (2) they modify the propositional content of the
utterance of which they are part, by taking a certain part of the propositional content
Connectivity as an object of linguistics 

of their utterance into their scope; (3) in addition, they link the content of the utter-
ance with the knowledge of the hearer, especially with respect to the evaluation of the
propositional content being received; in doing so they act upon the illocution of the
speech action in which they are used.
The Japanese language is notorious for its manifold particles, whose functions
range from purely grammatical operations, such as ‘case’, to illocutionary tasks, such as
‘interrogative’, and many more. Yuko Sugita (this volume) investigates a functional
expansion observed in the Japanese mono-moraic and pitchless ‘quotative’ particle to
in oral business reports. She finds that this complementizer is used recurrently in ut-
terance final position, serving as a device for non-linear information processing over
larger parts of interrelated concatenative discourse.
Matrix constructions, a class of complex formulae which often comprise construc-
tions of reported speech, have connective functions not necessarily limited to the
scope of a single utterance. The study of Thomas Johnen and Bernd Meyer (this vol-
ume) on interpreted doctor-patient communication shows that different grammatical
parts of the reported speech construction may realize different functions in discourse.
Especially the lexical (symbol field) component of the matrix reflects the attitude of the
interpreter towards what s/he reports in order to qualify the reported information for
the hearer-patient: “Whereas Turkish diyor and Portuguese dizer que (to say) are main-
ly used to trace propositional parts to their source, thus indicating reliability, the Ger-
man meinen with third person (he/she means) is used mainly to refer to some kind of
deficiency in the source language discourse.” (Johnen and Meyer, this vol.).
Matrix constructions establish utterance-internal connections between proposition-
al acts and expressions of thinking, believing, speaking, feeling and the like (i.e. the sym-
bol field of knowledge and its verbalization), employing a wide possible range of lan-
guage-specific means (for a comparative analysis of Japanese and German matrix
constructions cf. Hohenstein 2004). Their utterance-external connective function in dis-
course and text is concerned with joining hearer knowledge unto the speaker knowledge
related to the current utterance. Following a proposal by Bührig (2002), this function can
be characterized as “interaction coherence” (Jochen Rehbein, this volume).
This book assembles a broad range of topics in the domain of connectivity in mul-
tilingual constellations and/or in a cross-linguistic perspective. Future research, then,
might be directed at establishing a theoretically and empirically based classification of
connectives and at elaborating the systematic (and maybe partly universal) architec-
ture of their grammatical forms and communicative functions across languages. Such
a classification might also contribute to a better understanding of multilingual lan-
guage acquisition, of language contact and of the finely granulated stages of gram-
maticalization, whether instantaneous or gradual.
We wish to conclude by deeply thanking our colleagues who agreed to anony-
mously review the articles in this collection and who by their critical comments and
their suggestions helped to achieve a higher degree of clarity and precision. It goes
 Jochen Rehbein, Christiane Hohenstein, Lukas Pietsch

without saying that the responsibility for any mistakes that may have been overlooked
lies with the authors and ourselves.

References

Aarssen, J. 1998. Acquisition of topic continuity in Turkish children’s narratives. In The Mainz


Meeting. Proceedings of the Seventh International Conference on Turkish Linguistics, L. Jo-
hanson (ed.), 501–516. Wiesbaden: Harrassowitz.
Aijmer, K. and Simon-Vandenbergen, A. (eds). 2006. Pragmatic Markers in Contrast. Amster-
dam: Elsevier.
Bakker, P. and Mous, M. (eds.). 1994. Mixed Languages. 15 case studies in language intertwining.
Amsterdam: IFOTT.
Bakker, P. 2003. Mixed languages as autonomous systems. In The Mixed Language Debate. Theo-
retical and empirical advances, Matras, Y. and Bakker, P. (eds.), 107–151. Berlin: Mouton de
Gruyter.
Behrens, B. and Fabricius-Hansen, C. 2002. Connectives in Contrast: A discourse semantic
study of elaboration based computer research. In Information Structure in a Cross-Linguistic
Perspective, H. Hasselgard, S. Johansson, B. Behrens and C. Fabricius-Hansen (eds), 45–61.
Amsterdam: Rodopi.
La nature des pronoms. In Problèmes de linguistique générale, 251–257.
Benveniste, É. 1956. ��������������������������
Paris: Gallimard.
Berman R. A. 1998. Typological perspectives on connectivity. In Issues in the theory of language
acquisition, N. Dittmar and Z. Penner (eds), 203–224. Bern: Peter Lang.
Berman R. A. and Slobin, D. (ed.). 1994. Relating Events in Narrative. A crosslinguistic develop-
mental study. Hillsdale, NJ: Lawrence Erlbaum.
Bisang, W. 1995. Verb serialization and converbs – differences and similarities. In Converbs in
Cross-Linguistic Perspective. Structure and meaning of adverbial verb forms – Adverbial parti-
ciples, gerunds –, M. Haspelmath, and E. König (eds), 137–188. Berlin:
��������������������������
Mouton de Gruyter.
Bisang, W. 1998a. Grammaticalization
�������������������������������������������������������������������������
and language contact, constructions and positions. In
The Limits of Grammaticalization, A. G. Ramat and P. J. Hopper (eds), 13–58. Amsterdam:
John Benjamins.
Bisang, W. 1998b. Structural similarities of clause combining in Turkic, Mongolian, Manchu-
Tungusic and Japanese – a typological alternative to the hypothesis of a genetic relation-
ship. In The Mainz Meeting. Proceedings of the Seventh International Conference on Turkish
Linguistics, L. Johanson (ed.), 199–223. Wiesbaden: Harrassowitz.
In Language Typology and Language Universals,
Bisang, W. 2001. ������������������������������������
Finite vs. non finite languages. ���
M. Haspelmath, E. König, W. Oesterreicher and W. Raible, W. (eds), 1400–1413. Berlin:
Walter de Gruyter.
Bührig, K. 2002. Interactional coherence in discussions and everyday story telling: On consider-
��� Rethinking Sequentiality, A. Fetzer and
ing the role of ‘auf jeden Fall’ and ‘jedenfalls’. In
C. Meierkord (eds)�������������������������������������
, 273–290. Amsterdam:
��������������������������
John Benjamins.
Bührig, K. 2003. �����������������������������������������������������������������������������
Zur Funktionalität von ‘auf jeden Fall’ und ‘jedenfalls’. Untersuchungen
�������������������
zur
Zusammenhangbildung in Text und Diskurs. Universität Hamburg: Habilitationsschrift.
Connectivity as an object of linguistics 

Bührig, K. and House, J. 2004. Connectivity in translation. Transitions from orality to literacy.
In Multilingual Communication, J. House and J. Rehbein (eds), 87–114. Amsterdam: John
Benjamins.
������ Textkonnektoren und andere textstrukturierende Einheiten. Tübin-
Cambourian, A. (ed.). 2001.
gen: Stauffenburg.
Cardenes Melián, J. 1997. Aber, denn, doch, eben und ihre spanischen Entsprechungen. Eine
funktional-pragmatische Studie zur Übersetzung deutscher Partikeln. Münster:
����������������
Waxman.
Chafe,W. 1988. Linking intonation units in spoken English. In Clause Combining in Grammar
and Discourse, J. Haiman and S. A. Thompson (eds), 1–27. Amsterdam: John Benjamins.
Chafe, W. 1994. Discourse, Consciousness, and Time. The Flow and Displacement of Conscious
Experience in Speaking and Writing. Chicago: University of Chicago Press.
Chomsky, N. 2001. Derivation by phase. In Ken Hale: A life in language, Kenstowicz, M. (ed.),
1–52. Cambridge MA: The MIT Press.
Clyne, M. 1987. Cultural differences in the organization of academic texts. Journal of Pragmatics
11: 211–247.
Clyne, M. 2003. Dynamics of Language Contact. English and immigrant languages. Cambridge:
CUP.
Conte M., Petöfi, J. and E. Sözer, E. 1989. Text and Discourse Connectedness. Amsterdam: John
Benjamins.
Croft, W. 2001. Radical Construction Grammar. Oxford: OUP.
de Beaugrande, R. 1980. Text, Discourse, and Process. Toward a multidisciplinary science of texts.
London: Longman/Ablex.
Edmondson, W.J. 1978. A note on pragmatic connectives. Interlanguage Studies Bulletin, Utrecht
1978: 100–106.
Ehlich, K. 1986. Interjektionen. Tübingen: Niemeyer.
Ehlich, K. 1987. Kooperation und sprachliches Handeln. In Kommunikation und Kooperation, F.
Liedtke and R. Keller (eds), 17–32. Tübingen: Niemeyer.
Ehlich, K. 1991. Funktional-pragmatische Kommunikationsanalyse. Ziele und Verfahren. In
Verbale Interaktion, D. Flader (ed.), 17–32. Stuttgart: Metzler.
Fabricius-Hansen, C. 1992 Subordination. In Deutsche Syntax. Ansichten und Aussichten, L.
Hoffmann (ed.), 458–483. Berlin: Walter de Gruyter,
Fabricius-Hansen, C. 2000. Formen der Konnexion. In Text- und Gesprächslinguistik, Vol. I., K.
Brinker, G. Antos, W. Heinemann and S. F. Sager (eds), 331–343. Berlin: Walter de Gruyter.
Fandrych, C. and Graefen, G. 2002. Text commenting devices in German and English academic
articles. Multilingua 21: 17–43.
Fischer, K. (ed.). 2006. Approaches to Discourse Particles [Studies in Pragmatics 1]. Amsterdam:
Elsevier.
Foley, W. A. and Van Valin, R. D. 1984. Functional Syntax and Universal Grammar. Cambridge: CUP.
Fraser, B. 1990. An approach to discourse markers. Journal of Pragmatics 14: 383–395.
Genesee, F. 1989. Early bilingual development: One language or two? Journal of Child Language
16: 161–179.
Genesee, F., Nicoladis, E. and Paradis, J. 1995. Language differentiation in early bilingual devel-
opment. Journal of Child Language 22: 611–631.
Göksel, A. and Kerslake, C. 2005. Turkish: A comprehensive grammar. London: Routledge.
Givón, T. 1991. Serial verbs and the mental reality of ‘event’: Grammatical vs. Cognitive Packag-
ing. In Approaches to Grammaticalization, Vol. I, E. Traugott and B. Heine (eds). 81–127.
Amsterdam: John Benjamins.
 Jochen Rehbein, Christiane Hohenstein, Lukas Pietsch

Givón, T. 1995. Coherence in text vs. coherence in mind. In Coherence in Spontaneous Text, M.
A. Gernsberg and T. Givón (eds), 59–115. Amsterdam: John Benjamins.
Grein, M. 1998. Mittel der Satzverknüpfung im Deutschen und im Japanischen. Eine typologisch-
kontrastive Analyse. Wiesbaden: Deutscher Universitätsverlag.
Gülich, E. 1970. Makrosyntax der Gliederungssignale im gesprochenen Französisch. München: Fink.
Haberland, H. and van der Auwera, J. 1990. Topics and clitics in Greek relatives. International
Journal of Linguistics, Acta Linguistica Hafniensia. Copenhagen: The Linguistic Circle of
Copenhagen: 127–157.
Haegeman, L. (ed.). 1997. The New Comparative Syntax. London: Longman.
Haiman, J. and Thompson, S. A. (eds). 1988. Clause Combining in Grammar and Discourse.
Amsterdam: John Benjamins.
Halliday, M.A.K. and Hasan, R. 1976. Cohesion in English. London: Longman.
Hasegawa, Y. 1996. A study of Japanese clause linkage: The connective TE in Japanese. Stanford:
CSLI.
Haspelmath, M. 1995. The converb as a cross-linguistically valid category. In Converbs in Cross-
Linguistic Perspective. Structure and Meaning of Adverbial Verb Forms – Adverbial Partici-
ples, Gerunds –, M. Haspelmath, and E. König (eds), 1–55. Berlin: Mouton de Gruyter.
Haspelmath, M. 2004. Coordinating Constructions: An overview. In Coordinating Construc-
tions, id. (ed.), 3–39. Amsterdam: John Benjamins.
Heine, B. and Kuteva, T. 2005. Language Contact and grammatical Change. Cambridge: CUP.
Herkenrath, A., Karakoç, B. and Rehbein, J. 2003. Interrogative elements as subordinators in
Turkish – Aspects of Turkish-German bilingual children’s language use. In (In)vulnerable
Domains in Multilingualism [Hamburg Studies on Multilingualism I], N. Müller (ed.), 219–
267. Amsterdam: John Benjamins.
Heydrich, W., Neubauer, F., Petöfi, J. S. and E. Sözer (eds). 1989. Connexity and Coherence: Anal-
ysis of text and discourse. Berlin: Walter de Gruyter.
Hohenstein, C. 2004. A comparative analysis of Japanese and German complement construc-
tions with matrix verbs of thinking and believing: To omou and ich glaube. In Multilingual
Communication [Hamburg Studies on Multilingualism 3], J. House and J. Rehbein (eds),
303–341. Amsterdam: John Benjamins.
Hohenstein, C. 2005. Interactional expectations and Linguistic Knowledge in academic expert
discourse (Japanese/German), International Journal of the Sociology of language, 175/176:
285–306.
Hohenstein, C. 2006a. Kausale konnektivität in der deutschen un der japanischen Wissen-
schaftssprache. In Mehrsprachige individuenrechtsprachige Gesellschaften, D. Wolff (ed.),
155–178. Frankfurt: Peter Lang.
Hohenstein, C. 2006b. Sind Handlungsmuster mehrsprachig? Erklären im Wissenschaftlichen
Vortrag deutsch/japanisch. In Praxen der Mehrsprachigkeit. K. Ehrlich and A. Hornung
(eds), 155–194, Münster-Waxmann.
Hohenstein, C. 2006 c. Erklärendes Handeln im ‘Wissenschaftlichen Vortrag’: Ein Vergleich des
Deutschen mit dem Japanischen. München: Iudicium.
Hoffmann, L. 1997. Thematische Organisation von Text und Diskurs. In Grammatik der deut-
schen Sprache, G. Zifonun, L. Hoffmann and B. Strecker (eds), 507–591. Berlin: Mouton de
Gruyter.
Horie, K. (ed.). 2000. Complementation: Cognitive and functional perspectives. Amsterdam: John
Benjamins
House, J. 2006. Text and context in translation. Journal of Pragmatics 38: 338–358.
Connectivity as an object of linguistics 

House, J. and Kasper, G. 1981. Politeness markers in English and German. In Conversational
Routine. Explorations in standardized communication situations and prepatterned speech, F.
Coulmas (ed.), 157–185. The Hague: Mouton.
House, J. and Rehbein, J. 2004. What is ‘multilingual communication’? In Multilingual Commu-
nication [Hamburg Studies on Multilingualism 3], J. House and J. Rehbein (eds), 1–17.
Amsterdam: John Benjamins.
Jaeggli, O. and Safir, K. J. 1989. The null subject parameter and parametric theory. In The Null
Subject Parameter, O. Jaeggli and K. J. Safir (eds), 1–44. Dordrecht: Kluwer.
Jakobson, R. 1957 [1971]. Shifters, verbal categories, and the Russian verb. In Selected Writings
II. 130–147. The Hague: Mouton.
Janse, M. 2000. Convergence and divergence in the development of the Greek and Latin pro-
nouns. In Stability, Variation and Change of Word-Order Patterns over Time, R. Sornicola,
E. Poppe and A. Shisha-Halevy (eds) 230–258. Amsterdam: John Benjamins.
Johanson, L. 1971. Aspekt im Türkischen. Vorstudien zu einer Beschreibung des türkeitürkischen
Aspektsystems. Uppsala: Alquist and Wiksell.
Johanson, L. 1992a. Periodische Kettensätze im Türkischen. Wiener Zeitschrift für die Kunde des
Morgenlandes 82: 201–211.
Johanson, L. 1992b. Strukturelle Faktoren in türkischen Sprachkontakten. Stuttgart: Steiner
Johanson, L. 1994. Türkeitürkische Aspektotempora. In Tense Systems in European Languages,
R. Thieroff and J. Ballweg (eds), 247–266. Tübingen: Niemeyer.
Johanson, L. 1995. On Turcic converb clauses. In Converbs in Cross-Linguistic Perspective. Struc-
ture and meaning of adverbial verb forms – Adverbial participles, gerunds –, M. Haspelmath,
and E. König (eds), 313–347. Berlin: Mouton de Gruyter.
Johanson, L. 1999a. The dynamics of code-copying in language encounters. In Language En-
counters across Time and Space. Studies in language contact, B. Brendemoen, B. and E. Lan-
za (eds), 37–62. Oslo: Novus.
Johanson, L. 1999b. Viewpoint operators in European languages. In Tense and Aspect in the
Languages of Europe, Ö. Dahl (ed.), 27–187. Berlin: Mouton de Gruyter.
Kerslake, C. 1996. The role of connectives in discourse construction in Turkish. In Modern Stud-
ies in Turkish Linguistics, A. Konrot (ed.), 77–104. Eskişehir: Anadolu University.
König, E. 1991. The Meaning of Focus Particles: A comparative perspective. London: Routledge.
König, E. and van der Auwera, J. 1988. Clause integration in German and Dutch conditionals,
concessive conditionals, and concessives. In Clause Combining in Grammar and Discourse,
J. Haiman and S. A. Thompson (eds), 101–133. Amsterdam: John Benjamins.
Kortmann, B. 1996. Adverbial Subordination. A typology and history of adverbial subordinators
based on European languages. Berlin: Mouton de Gruyter.
Leuschner, T. 1998. At the boundaries of grammaticalization. What interrogatives are doing in
concessive conditionals. In The Limits of Grammaticalization, A. G. Ramat and P. J. Hopper
(eds), 159–187. Amsterdam: John Benjamins.
Li, Ch. N. and Thompson, S. A. 1976. Subject and topic: A new typology of language. In Subject
and Topic, C. N. Li (ed.), 457–489. New York: Academic Press.
Liedke, M. 1994. Die Mikro-Organisation der Verständigung. Diskursuntersuchungen zu grie-
chischen und deutschen Partikeln. Frankfurt: Peter Lang.
López, L. 2003. Steps for a well-adjusted dislocation. Studia Linguistica 57(3): 193–231.
Mackenzie, J. L. 1984. Communicative functions of subordination. In English Language Re-
search: The Dutch contribution 1, J. L. Mackenzie and H. Wekker (eds), 67–84. Amsterdam:
Vrije Uni Press.
 Jochen Rehbein, Christiane Hohenstein, Lukas Pietsch

Matras Y. 1996. Prozedurale Fusion: Grammatische Interferenzschichten im Romanes. Sprachty-


pologie und Universalienforschung 49(1): 60–78.
Matras Y. 1997. The function and typology of coordinating conjunctions: Evidence from dis-
course and language-contact situations. In Discourse and Pragmatics in Functional Gram-
mar, J. H. Connolly, R. M. Vismans, C. S. Butler and R. A. Gatward (eds), 177–191. Berlin:
Mouton de Gruyter.
Matras, Y. 1998. Utterance modifiers and universals as grammatical borrowing. Linguistics 36:
281–331.
Matras, Y. 2000. Mixed languages: A functional-communicative approach. Bilingualism: Lan-
guage and Cognition 3: 79–99.
Matras, Y. 2002. Romani. A linguistic introduction. Cambridge: CUP.
Matras, Y. and Bakker, P. (eds) 2003. The Mixed Language Debate. Theoretical and empirical ad-
vances. Berlin: Mouton de Gruyter.
Matthiessen, C. and Thompson, S. A. 1988. The structure of discourse and ‘subordination’. In
Clause Combining in Grammar and Discourse, J. Haiman and S. A. Thompson, S. A. (eds),
275–329. Amsterdam: John Benjamins.
Meisel, J. M. 1989. Early differentiation of languages in bilingual children. In Bilingualism across
Lifespan: Aspects of acquisition, maturity and loss, K. Hyltenstam and L. Obler (eds), 13–40.
Cambridge: CUP.
Mori, J. 1999. Negotiating Agreement and Disagreement in Japanese: Connective expressions and
turn construction. Amsterdam: John Benjamins.
Müller, N. 1993. Komplexe Sätze. Der Erwerb von COMP und von Wortstellungsmustern bei bi-
lingualen Kindern (Französisch/Deutsch). Tübingen: Narr.
Müller, N. (ed.) 2003. (In)vulnerable Domains in Multilingualism [Hamburg Studies in Multilin-
gualism 1]. Amsterdam: John Benjamins.
Muysken, P. 2000. Bilingual Speech. A typology of code-mixing. Cambridge: CUP.
Myhill J. and J. Hibiya 1988. The discourse function of clause-chaining. In Clause Combining in
Grammar and Discourse, J. Haiman and S. A. Thompson (eds), 361–398. Amsterdam: John
Benjamins.
Nekula, M. 1996. System der Partikeln im Deutschen und Tschechischen. Unter besonderer Berück-
sichtigung der Abtönungspartikeln. Tübingen: Niemeyer.
Onodera, N. 2004. Japanese Discourse Markers: Synchronic and diachronic discourse analysis.
Amsterdam: John Benjamins.
Özbek, N. 2000. Yani, işte, şey, ya: Interactional markers of Turkish. In Studies on Turkish and
Turkic Languages, A. Göksel and C. Kerslake (eds), 393–401. Wiesbaden: Harrassowitz.
Park, Y.Y. 1998. A discourse analysis of contrastive connectives in English, Korean, and Japanese
conversation: With special reference to the context of dispreferred responses. In Discourse
Markers: Descriptions and theory, A.-H. Jucker and Y. Ziv (eds), 277–300. Amsterdam: John
Benjamins.
Redder, A. 1990. Grammatiktheorie und sprachliches Handeln: ’denn’ und ‘da’. Tübingen: Niemeyer.
Redder, A. 2006. Konjunktor. To appear in Handbuch Wortarten, L. Hoffmann (ed.). Berlin: de
Gruyter
Rehbein, J. 1995a. Über zusammengesetzte Verweiswörter und ihre Rolle in argumentierender
Rede. In Wege der Argumentationsforschung, H. Wohlrapp (ed.), 166–197. Stuttgart-Bad-
Cannstatt: Frommann-holzboog.
Rehbein, J. 1995b. Grammatik kontrastiv – am Beispiel von Problemen mit der Stellung finiter
Elemente. Jahrbuch Deutsch als Fremdsprache 21: 265–292.
Connectivity as an object of linguistics 

Rehbein, J. 1999. Zum Modus von Äußerungen. In Grammatik und mentale Prozesse, A. Redder
and J. Rehbein (eds), 91–139. Tübingen: Stauffenburg.
Rehbein, J. 1999a. Konnektivität im Kontrast. Zu Struktur und Funktion türkischer Konverbien
und deutscher Konjunktionen, mit Blick auf ihre Verwendung durch monolinguale und
bilinguale Kinder. In Türkisch und Deutsch im Vergleich, L. Johanson and J. Rehbein (eds),
189–243. Wiesbaden: Harrassowitz.
Rehbein, J. 2002. Pragmatische Aspekte des Kontrastierens von Sprachen – Türkisch und Deut-
sch im Vergleich. Arbeiten zur Mehrsprachigkeit, Folge B (Nr. 40). In Eröffnungsreden und
Tagungsbeiträge des VII. Türkischen Germanistikkongresses, S. Yıldız, N. Ülner, K. Halm
Karadeniz u.a. (eds), 15–58. Ankara: Hacettepe University.
Rehbein, J. and Kameyama, S. 2003. Pragmatik. In Sociolinguistics / Soziolinguistik. An interna-
tional handbook of the science of language and society, (2nd ed.), U. Ammon, N. Dittmar, K.
Mattheier and P. Trudgill (eds), 556–588. Berlin: Walter de Gruyter.
Rehbein, J. and Karakoç, B. 2004. On contact-induced language change of Turkish aspects: Lan-
guaging in bilingual discourse. In Languaging and Language practices [Copenhagen Studies
in Bilingualism 36], C. B. Dabelsteen and N. Jørgensen (eds), 129–155. Copenhagen: Uni-
versity of Copenhagen.
Rizzi, L. 1997a. A parametric approach to comparative syntax: Properties of the pronominal
system. In The New Comparative Syntax, L. Haegeman (ed.), 268–285. London: Longman.
Rudolph, E. 1996. Contrast. Adversative and concessive relations and their expressions in English,
German, Spanish, Portuguese on sentence and text level. Berlin: Walter de Gruyter
Schiffrin, D. 1987. Discourse markers. Cambridge: CUP.
Schourup, L. 1985. Common Discourse Particles in English Conversation: like, well, y’know. New
York NY: Garland.
Slobin D. I. 1988. The development of clause chaining in Turkish child language. In Studies on
Turkish linguistics, S. Koç (ed.), 27–54. Ankara: Middle East Technical University
(METU).
Stolz, T. 2003. Not quite the right mixture: Chamorro and Malti as candidates for the status of
mixed language. In The Mixed Language Debate. Theoretical and empirical advances, Y.
Matras and P. Bakker (eds), 271–315. Berlin: Mouton de Gruyter.
Stolz, C. and Stolz, T. 1996/1997. Universelle Hispanismen? Von Manila über Lima bis Mexiko
und zurück. Muster bei der Entlehnung spanischer Funktionswörter in die indigenen
Sprachen Amerikas und Austronesiens. Orbis XXXIX, 1996–1997, 1–77.
Unger, C. 1996. The scope of discourse connectives. Implications for discourse organization.
Journal of Linguistics 32(2): 403–438.
van der Auwera, J. 1998. Adverbial Constructions in the Languages of Europe. Berlin: Walter de
Gruyter.
van der Auwera, J. and Bultinck, B. 2001. On the lexical typology of modals, quantifiers, and
connectives. In Perspectives on Semantics, Pragmatics, and Discourse, I. Kenesei, R. Harnish
and J. Gervain (eds), 173–186. Amsterdam: John Benjamins.
van der Wouden, T., Foolen, A. and van de Craen, P. (eds.). 2003. Particles. Belgian Journal of
Linguistics 16.
van Dijk, T. 1979. Pragmatic connectives. Journal of Pragmatics 3: 447–456.
van Valin R. D. and R. J. LaPolla 1997 Syntax. Structure, meaning and function, Cambridge:
CUP.
Verhagen, A. 2001. Subordination and discourse segmentation revisited: Or, why matrix claus-
es may be more dependent than complements. In Text Representation: Linguistic and psy-
 Jochen Rehbein, Christiane Hohenstein, Lukas Pietsch

cholinguistic aspects, T. Sanders, J. Schilperood and W. Spooren (eds), 337–357. Amster-


dam: John Benjamins.
Weydt, H. (ed.). 1979. Die Partikeln der deutschen Sprache. Berlin: Walter de Gruyter.
Weydt, H. (ed.). 1983. Partikeln und Interaktion. Tübingen: Niemeyer.
Weydt, H. 1987. Partikel-Bibliographie: Internationale Sprachenforschung zu Partikeln und Inter-
jektionen. Frankfurt: PeterLang.
White, P. 2003. Beyond modality and hedging: A dialogic view of the language of intersubjective
stance. Text 23(2): 259–284.
Yılmaz, E. 2004. A Pragmatic Analysis of Turkish Discourse Particles: yani, işte, and şey. PhD dis-
sertation. Ankara: Graduate School of Social Sciences of Middle East Technical University.
section 1

Aspects of language change


and language acquisition
Grammaticalization of converb constructions
The case of Japanese -te conjunctive constructions

Masayoshi Shibatani
Rice University, Austin

This paper examines the patterns of grammaticalization of motion verbs (verbs


of coming and going) in Japanese converb complex predicate constructions. The
-te converb form is the most widely used connective device in Japanese, and
the forms combining with motion verbs are among the most frequently used
converb constructions in the language. Among the -te converb constructions
involving motion verbs, those that appear to have grammaticalized are most
numerous in the corpus data. Detailed examinations of the decategorialization
pattern of the motion verbs involved indicate a clear cline of grammaticalization,
which contradicts some earlier studies such as Teramura (1984) and
Hasegawa (1996). The paper also addresses a number of important issues in
grammaticalization studies and questions some of the current understandings
of them, which include the following: (1) What drives grammaticalization
– frequency of use and metaphor (Traugott and Heine 1991; Heine, Claudi, and
Hünnemeyer 1991)? (2) What are the possible paths of grammaticalization in
the development of temporal meanings from motion verbs (Bybee, Perkins and
Pagliuca 1994)? (3) Are grammaticalization processes always gradual (Bybee,
Pagliuca and Perkins 1991; Brinton and Traugott 2005)?

1. Introduction

Japanese -te conjunctive constructions, illustrated in (1) below, share functional simi-
larities with both converb constructions of the Altaic family (e.g., Turkic converb con-
structions) and verb serialization constructions widely observed across languages.
Indeed, a legitimate typological question is whether there is any substantial difference
between these two types of constructions.1
(1) a. Wareware-wa eki-kara densya-ni not-te Ochanomizu-made it-ta.
we-TOP station-from train-on ride-CON Ochanomizu-up to go-PAST
‘We rode on a train and went to Ochanomizu.’
 Masayoshi Shibatani

b. Midori-wa sanzi-sugi-ni kaet-te ki-ta.


Midori-TOP three-past-LOC return-CON come-PAST
‘Midori returned (return come) at past three o’clock.’
(Haruki Murakami Norwegian Wood)
A major reason for our using the label “converb construction” (as opposed to “serial
verb construction”) for the Japanese construction involving -te – a conjunctive particle in
the parts-of-speech classification of traditional Japanese grammar – is to highlight the
presence of this particle, which marks the non-finiteness of the clause, thereby signaling
the existence of a following finite verb or (truncated) clause. The presence of the marker
with varying names such as conjunction, linker, converb ending, non-finite marker, etc.
is relevant to the general discussion of the grammaticalization process of converb and
serial verb constructions. In his discussion of the grammaticalization of motion and
other types of verbs in Lhasa Tibetan, DeLancey (1991) recognizes a stage where the
non-finite marker is lost and a contiguous verb sequence is created. This is an important
step in the chain of reanalysis from clause chaining constructions to verb serialization
and to auxiliarization of relevant types of verb. A similar verb contiguity requirement is
stipulated by Foley and Olsen (1985:45–46) in their reanalysis of core layer junctures to
nuclear layer junctures, whereby a verb sequence is created from VP-like conjunctions
(see also Lord 1993, Hopper and Traugott 1993). This paper shows that, contrary to these
assumptions, both complex predicates and their grammaticalization obtain across a con-
junctive or non-finite marker (see Shibatani and Chung 2005).
The grammaticalization of certain converb constructions is implied by the term
hozyo-doosi ‘helping verbs’ in traditional Japanese grammar in reference to the finite
verb involved in these constructions. That is, by calling the verbal form ki-ta ‘come-
PAST’ in (1b) above a helping verb as opposed to a (main) verb, traditional grammar
recognizes some degree of decategorialization of this form in contradistinction to the
full verb status accorded to it-ta ‘go-PAST’ in (1a), where no grammaticalization is
involved. This paper is concerned with the grammaticalization of converb construc-
tions involving specifically motion verbs of coming (kuru) and going (iku) as in (1)
above. The rationale for this focus is manifold, but we first note the fact that converbs
are among the most widely used clause linkage devices in Japanese (see Table 1 below).
Secondly, according to Grein’s (1998:284) count, 136 out of 203 occurrences of con-
verb forms in her data were instances of the -te form.
Grammaticalization of converb constructions 

Table 1.  Distribution of clause-linkage devices in Japanese


(adapted from Grein 1998:286)

Device Number Percentage

Converbs 203 40.1%


Functional nouns 154 30.4%
Adverbial
subordinators 68 14.2%
Complementizers 50 9.6%
Coordinators 31 5.7%
TOTAL 506 100.0%

Thirdly, it is also noteworthy that grammaticalized -te constructions most frequently oc-
cur in this construction type, as can be seen from the following table from Schmidt (2004),
which describes the distribution of different functions in -te converb constructions.

Table 2. ����������������
Function of the -te converb (based on Schmidt 2004)

Function Text A Text B Text C

Grammaticalized 181 54.4% 335 71.4% 80 65.1%


Narrative 102 32.9% 100 21.3% 31 23.5%
Subordinate 27 8.7% 34 7.3% 15 11.4%
Symmetric 0 0.0% 0 0.0% 0 0.0%
TOTAL 309 100.0% 468 100.0% 132 100.0%

Finally, constructions involving verbs of coming and going are among the most fre-
quently seen in the grammaticalized usage of the -te converb form (see Table 3).
The goal of this paper is to explicate the nature and pattern of grammaticalization
in converb constructions involving the two motion verbs of iku ‘go’ and kuru ‘come’.
This, in turn, will help us better understand the process of “auxiliarization” implicit in
the term hozyo doosi ‘helping verb’ in traditional Japanese grammatical terminology.
Theoretical questions of wider relevance raised in this paper include the form-func-
tion correlation of connected events, instantaneous vs. gradual grammaticalization,
contexts facilitating grammaticalization, and the paths of grammaticalization.
 Masayoshi Shibatani

Table 3.  Token frequency of grammaticalized converb constructions in three texts (ac-
cording to Christopher Schmidt’s unpublished survey)

V-te iru ‘exist’ 301 (progressive/resultative)


V-te oru ‘exist’ 1 (progressive/resultative)
V-te iku ‘go’ 48 (deictic)
V-te kuru ‘come’ 47 (deictic)

V-te simau ‘finish’ 27 (completive)

V-te kureru ‘be given’ 23 (benefactive)


V-te yaru ‘give’ 4 (benefactive)
V-te sasiageru ‘give’ 1 (benefactive)
V-te miru ‘look’ 18 ‘try doing V’
V-te goran ‘look’ 1 ‘try doing V’
V-te miseru ‘show’ 1 ‘show off V-ing’
V-te morau ‘receive’ 12 (self-benefactive)
V-te ii ‘good’ 12 ‘may V’
V-te oku ‘put’ 10 ‘V in preparation of X’

V-te aru ‘exist’ 9 (resultative)


A-te naranai I ‘won’t 5 ‘cannot help but being A’
V-te hosii ‘want�’ 1 (desiderative)

Topics bearing on the central theme of this paper include the relationship between the
grammaticalized constructions and clause chaining constructions of the following type:
(2) Sorekara watasitati itumono-yooni syokudoo-de
and then we as usual dinning room-in
gohan-o tabe-te, ohuro hait-te, sorekara totteokino
meal-ACC eat-CON bath enter-CON and then choicest
zyootoono wain ake-te hutari-de non-de,
superior wine open-CON two-by drink-CON
watasi-ga gitaa hii-ta no.
I-NOM guitar play-PAST FP
‘And then we ate our meal in the dining room, took a bath,
and then opened the choicest super-quality wine (and we)
two drank, (and) I played the guitar. ’
(Haruki Murakami Norwegian Wood)
Grammaticalization of converb constructions 

Prevailing views (DeLancey 1991 and others) in the field hold that grammaticalized
converb/serial verb constructions arise from clause-chaining constructions via rean-
alysis. While it is necessary for us to contrast grammaticalized constructions with
clausal conjunctions throughout this paper, we will not directly address this important
question here and refer the readers to Shibatani and Chung (2005), which casts doubt
on the reanalysis hypothesis.

2. Grammaticalized converb constructions as complex predicates

Before discussing the grammaticalization patterns of converb constructions, it is nec-


essary to point out that there are two kinds of possible verb sequence which are ren-
dered similarly in orthographic or transliterated form.
(3) a. Taroo-wa ringo-o tabe-te it-ta.
Taroo-TOP apple-ACC eat-CON go-PAST
‘Taro ate an apple (and went away).’
b. Taroo-wa gakkoo-e ringo-o tabe-te, it-ta.
Taroo-TOP school-to apple-ACC eat-CON go-PAST
‘Taro, having eaten an apple, went to school.’
c. [ Taroo-wa [ ringo-o tabe-te] gakoo-e it-ta]
Taroo-TOP apple-ACC eat-CON school-to go-PAST
‘Taro, having eaten an apple, went to school.’
While (3a) and (3b) have the same verbal series of tabe-te it-ta ‘eat-CON go-PAST’,
only the former instantiates a grammaticalized construction whose formal property is
in constituting a single complex predicate. (3b) is a paraphrase of (3c), obtainable by
moving the goal nominal gakoo-e ‘to school’ to position immediately before the subor-
dinate clause. Indeed, the verb sequences in (3a) and (3b) differ both phonologically
and grammatically. In the former, the sequence manifests the pitch pattern of a single
word, whereas in the latter the two verbs retain their own pitch patterns:
(4) a. tabe-te it-ta ‘ate (and went away)’
LHHHH
b. tabe- te it-ta ‘ate and went’
H L L LH
The Tokyo dialect is typical of many Japanese dialects in allowing a phonological word
to have at most one stretch of high-pitched morae; accordingly, once a pitch fall occurs
within a word, it can never rise in the subsequent morae of the same word. In (4b) for
(3b) we see two high-pitched morae divided by a stretch of low-pitched morae indicat-
ing two (phonological) words. In (4a) for (3a), on the other hand, there is only one
 Masayoshi Shibatani

stretch of high-pitched morae indicating that the verb sequence here forms a single
phonological word.
There is further indication that the verb sequence in (3a) forms a single predicate,
while the similar sequence in (3b) and (3c) does not. Indirect evidence for this comes
from the fact that (3a) is a single clause expression, while (3b) and (3c) involve a bi-
clausal structure. The negative polarity item sika ‘only, save’ requires the presence of
the negative morpheme –nai ‘not’ in its own clause at some stage of derivation.2
(5) a. *Taroo-wa [ Hanako-sika kit-a] =koto-o sir-ana-katta.
Taro-TOP [ Hanako-only come-PAST] -COMP-ACC know-NEG-PAST
‘Taro didn’t know that only Hanako came.’
b. Taroo-wa [ Hanako-sika ko-na-katta] =koto-o sit-ta.
Taro-TOP [ Hanako-only come-NEG-PAST] -COMP-ACC know-PAST
‘Taro found out that only Hanako came/(lit.) Taro knew that everyone save
Hanako didn’t come.’
c. Taroo-sika [ Hanako-ga kita] =koto-o sir-ana-katta.
Taro-only [ Hanako-NOM come-PAST] -COMP-ACC know-NEG-PAST
‘Only Taro knew that Hanako came/ (lit.) Everyone save Taro didn’t know that
Hanako came.’
Now (6a) based on (3a) below is perfectly fine, indicating that sika and the negative -
ana occur in a single clause. But (6b) and (6c) – based on (3b) and (3c), respectively
– are ungrammatical, indicating that sika and the negative morpheme occur in two
separate clauses.
(6) a. Taroo-wa ringo-sika tabe-te ik-ana-katta.
Taro-TOP apple-only eat-CON go-NEG-PAST
‘Taro ate only an apple (and went away)/(lit.) Taro didn’t eat-go anything but
an apple.’
b. *Taroo-wa gakkoo-e ringo-sika tabe-te ik-ana-katta.
Taro-TOP school-to apple-only eat-CON go-NEG-PAST
‘Taro ate only an apple and didn’t go to school.’
c. *Taroo-wa ringo-sika tabe-te gakkoo-e ik-ana-katta.
Taro-TOP apple-only eat-CON school-to go-NEG-PAST
‘Taro ate only an apple and didn’t go to school.’
What is particularly interesting about grammaticalized constructions in Japanese
is that while they are words in both phonological and syntactic senses, they do not
form morphological words. That is, they do not show the property of lexical integrity
characteristic of morphological words. For example, (complex) morphological words
involving suffixation and compounding do not permit insertion of a particle or word
even at the morphological boundaries, but the converbs under consideration do, as
shown below:
Grammaticalization of converb constructions 

(7) a. tabe-rare-ru a’. *tabe-wa-rare-ru


eat-PASS-PRES eat-TOP-PASS-PRES
‘to be eaten’
b. tabe-aruk-u b’. *tabe-wa-aruk-u
eat-walk-PRES eat-TOP-walk-PRES
‘walk around eating here and there’
c. tabe-te ik-u c’. tabe-te-wa ik-u (keredo)
eat-CON go-PRES eat-CON-TOP go-PRES (although)
‘eat (and go away)’ ‘(although) (I) eat (and go away)’
(7a’) and (7b’) are impossible forms, whereas (7c’) is a perfectly formed expression
with the topic marker wa inserted between the two verbal forms. Similarly, while mor-
phological words do not permit modification of their parts, converbs permit such
modification. For example, the compound form in (8a) below does not allow its sec-
ond member to undergo the honorification process, although the honorification of the
whole word is possible, as in (8a”). In contrast, an honorified finite verb within a con-
verb construction is perfectly well-formed, as in (8b’). Notice furthermore that an en-
tire converb construction cannot be honorified as a whole – see (8b”).
(8) a. moti-age-ru a’. *moti-o-age-ni nar-u a”. o-moti-age-ni nar-u
hold-raise-PRES
‘hold up (something)’
b. mot-te ik-u b’. mot-te o-iki-ni nar-u b”. *o-mot-te iki-ni nar-u
hold-CON go-PRES
‘take (away)’
The morphological modification patterns of the above type provide crucial tests for the
degree of grammaticalization of the relevant converb constructions below. Grammati-
calized converb constructions are thus interesting linguistic objects; they are words
under both phonological and syntactic criteria, but not under morphological criteria.

3. The syntax of grammaticalized converb constructions

3.1 Teramura (1984) and Hasegawa (1996)

Teramura (1984) represents an early effort toward the understanding of the gram-
matical properties of grammaticalized converb constructions. First, with regard to the
question of combinatorial possibilities of iku ‘go’ and kuru ‘come’ with various con-
verbs, he presents the following table, intended to show the combinations of the two
motion verbs with different kinds of verb as converbs:
 Masayoshi Shibatani

Table 4.  Combinatorial possibilities of grammaticalized converb constructions according


to Teramura (1984:163)

umareru sinu mieru/kikoeru toozakaru


‘to be born’ ‘die’ ‘visible/audible’ ‘go away’

-te kuru 〇 × 〇 ×
‘come’

-te iku × 〇 × 〇
‘go’

While the restrictions noted by Teramura appear to be correct, a quick Google


search reveals that in fact all combinations noted in the table are possible. Teramura’s
intuitions on the combinatorial possibilities appear to reflect the typicality of perspec-
tive chosen by native speakers rather than grammatical restrictions. For example, in
the combinations of umareru ‘to be born’ and the two motion verbs, we are likely to
view the birth of an entity from the perspective at which an entity emerges; hence the
choice of kuru ‘come’ indicates that a movement toward the speaker is preferred. For
verbs of disappearance such as sinu ‘die’, the entity disappearing is viewed from the
perspective of existence; hence iku ‘go’ is preferred with such expressions. The follow-
ing examples from Google indicate that speakers of Japanese can easily reverse these
conventional perspectives, however, and allow various combinations that are identi-
fied as impossible by Teramura.
(9) a. inoti-o sukuu-tabi kiboo-ga umare-te ik-u.
life –ACC save-every.time hope-NOM be.born-CON go-PRES
‘Each time a life is saved, a hope is born.’
b. gen-ga sin-de ku-ru =to rensyuusuru
string-NOM die-CON come-PRES =when practice
ki-ga use-te ku-ru...
motivation-NOM lost-CON come-PRES
‘When strings [of a musical instrument] begin dying, motivation to practice
becomes lost…’
c. tamenteki, bunsekiteki-ni miru =koto-niyotte, zinkenkyooiku-no
many.sided analytical-ADV look =COMP-by human.right.education-GEN
genzyoo-to kadai-ga yori senmeini mie-te ik-u
current.state-COM issues-NOM more clearly visible-CON go-PRES
‘By looking (at the problems) from a many sided and analytical angle, the cur-
rent state and the issues of human rights education become visible even more
clearly.’
Grammaticalization of converb constructions 

d. kyuukyuusya-ga hanare-te ik-u =to sono sairen-no oto-wa


ambulance-NOM go.away-CON go-PRES =when its siren-GEN sound-TOP
tiisaku kikoe-te ik-u
faintly audible-CON go-PRES
‘When the ambulance goes far way, the sound of its siren goes on being audi-
ble ever more faintly.’
e. haru-ga tikazui-te samusa-ga toozakat-te ku-ru=to…
spring-NOM come.near-CON coldness-NOM go.far-CON come-PRES=when
‘When spring comes near and the coldness goes far away…’
These examples indicate one functional domain in which converb constructions have
grammaticalized with a functional shift from a physical spatial domain to an abstract
aspectual domain. Our initial focus in this paper, however, is on the degrees of gram-
maticalization seen in the spatial domain, where their patterns are harder to discern,
due to the fact that the relevant motion verbs retain to a greater or lesser extent the
original meaning of physical motion. In particular, we are concerned with the follow-
ing types of expression, for which Teramura (1984) offers an analysis (indicated in the
parentheses):
(10) a. Taroo-wa koohii-o non-de ki-ta. (V-V)
Taro-TOP coffee-ACC drink-CON come-PAST
‘Taro drank coffee (and came).’
b. Taroo-wa gakkoo-e arui-te it-ta. (v-V)
Taro-TOP school-to walk-CON go-PAST
‘Taro walked (walk went) to school.’
c. Taroo-ga heya-ni hait-te ki-ta. (V-v)
Taro-NOM room-to enter-CON come-PAST
‘Taro came (enter came) into the room.’
According to Teramura (1984: 157ff), expressions like (10a) depict two sequen-
tially ordered actions; accordingly, both the converb and the finite motion verb func-
tion like main verbs (indicated by the capital V) connected by -te. In (10b), on the
other hand, the converb modifies the motion verb, specifying in this example the man-
ner of the going motion; accordingly, the relevant verbs are connected in a subordi-
nate-main relationship, where the converb is subordinated (as indicated by the small
v) to the main motion verb (indicated by V). In the case of (10c), the subordinate-
main relationship is reversed, such that the converb (V) expresses the main action
whose deictic orientation is indicated by the motion verb (v). According to Teramura’s
analysis, the finite motion verbs in the expression type (10c) are most advanced in
grammaticalization, turning themselves into something like deictic markers, whereas
those in the expression types (10a) and (10b) retain the original verb status.
 Masayoshi Shibatani

More recent work of Hasegawa (1996) is couched in the framework of Role and
Reference Grammar, which pays special attention to issues of clause-linkage type.
Hasegawa’s classification differs somewhat from Teramura’s reviewed above. The main
syntactic division she draws is between nuclear coordination and nuclear subordina-
tion types. In the former, the converb and the finite motion verb are in a coordinate
relation, together forming a complex nuclear (predicate). This analysis resembles Tera-
mura’s V-V analysis in that the full verbal status of both the converb and the finite
motion verb is recognized. The nuclear subordination type is similar to Teramura’s V-v
analysis; hence the converb alone functions as a nuclear (predicate), with the finite
verb functioning as a deictic operator. Hasegawa would classify the three types of ex-
pression in the following manner:
(11) a. non-de kur-u (Nuclear coordination; Teramura’s V-V)
drink-CON come-PRES
‘drink (and come)’
b. arui-te ik-u (Nuclear coordination; Teramura’s v-V)
walk-CON go-PRES
‘go (by) walking’
c. hait-te ku-ru (Nuclear subordination; Teramura’s V-v)
enter-CON come-PRES
‘enter (come in)’
There are both similarities and differences between the two analyses. Both Tera-
mura and Hassegawa recognize a full verbal status for motion verbs in (11a) and (11b).
While Teramura recognizes a diminished autonomy of the converb in (11b), Haseg-
awa considers the converb and the finite verb to be of equal status. As for (11c), Tera-
mura recognizes some degree of verbiness in the finite motion verb, but Hasegawa
considers it to have lost the predicate function completely, perhaps along with con-
comitant decategorialization from verb to deictic operator.3 Whether or not a degree
of verbiness is intended in the representations of “v” and “V” by Teramura, or com-
plete decategorialization is recognized for certain motion verbs in the relevant converb
constructions by Hasegawa, the following discussion shows that their analyses do not
accord well with the facts. In particular, it will be shown below that the iku/kuru ‘go/
come’ verbs in the non-de iku/kuru ‘drink-CON go/come’ type (10a, 11a) are least
verb-like (contrary to Teramura and Hasegawa’s analyses), that motion verbs in the
arui-te iku/kuru ‘walk-CON go/come’ type (10b, 11b) are most verb-like (agreeing
with Teramura and Hasegawa’s analyses), and that those in the hait-te iku/kuru ‘enter-
CON go/come’ type (10c, 11c) are verb-like to a considerable extent, contra Teramura’s
and Hasegawa’s analyses.
Grammaticalization of converb constructions 

3.2 Cline of grammaticalization

While there appear to be both instantaneous and gradual aspects to the grammaticaliza-
tion process (see Section 4 below), the framework of grammaticalization is especially
attuned to the task of capturing gradual patterns of change in the status of lexical items.
In particular, the changes to the motion verbs of coming and going in converb construc-
tions within the grammaticalization framework reveal that these verbs do not change
their category membership instantaneously, and that a synchronic description of them
must recognize a cline of categoriality along a path of change from verb to auxiliary. The
remainder of this section is devoted to a demonstration that the three relevant construc-
tion types are distributed along the following cline of grammaticalization:

(12) Less grammaticalized (More V-like)

arui-te iku/kuru (Teramura’s v-V; Hasegawa’s


walk-CON go/come Nuclear coordination)

hait-te iku/kuru (Teramura’s V-v; Hasegawa’s


enter-CON go/come Nuclear subordination)

non-de iku/kuru (Teramura’s V-V; Hasegawa’s


drink-CON go/come Nuclear coordination)

More grammaticalized (Less V-like)

Of the various possible combinations of converbs and motion verbs, we single out
the above three patterns, since they exhaust Teramura’s representations of types in
terms of the combination of V and v symbols – V-V, v-V, and V-v. In reality, the pos-
sibility exists that there are numerous intermediate types of combination interspersed
among the three representative ones dealt with here. Of the three combination types,
the arui-te kuru/iku ‘walk-CON come/go’ type represents combinations involving a
motion verb and a manner of motion verb ( aruku ‘walk’, hasiru ‘run’, hau ‘crawl’, oyogu
‘swim’, tobu ‘fly’, etc.). In the following discussion, this type will be exemplified by ex-
pressions and glosses such as arui-te iku ‘walk come’ and hasit-te kuru ‘run come’. The
hait-te iku/kuru ‘enter-CON go/come’ type represents combinations of a motion verb
and a verb of change-of-location (deru ‘exit’, agaru ‘ascend’, oriru ‘descend’, otiru ‘fall’,
etc.), and examples such as hait-te kuru ‘enter come’ and de-te iku ‘exit go’ illustrate this
type of construction. Finally, the non-de iku/kuru ‘drink-CON go/come’ type repre-
sents combinations of a motion verb and a non-translational action (taberu ‘eat’, asobu
‘play’, sake-o nomu ‘drink sake’, eiga-o miru ‘see a movie’ etc.). In the following discus-
sion, examples such as non-de kuru ‘drink come’ and tabe-te iku ‘eat go’ illustrate this
type of construction. Thus even if the examples in the following discussion have the
 Masayoshi Shibatani

form of arui-te iku ‘walk go’, non-de kuru ‘drink come’, it should be understood that the
relevant points of discussion apply equally to other instances of the same type such as
hasit-te kuru ‘run come’ and tabe-te iku ‘eat go’.
The methods used in arriving at the pattern of distribution of different construc-
tion types seen in (12) typically involve comparison of the grammatical properties of
iku/kuru ‘go/come’ in their full-fledged main verb function vs. those in converb com-
plex predicate constructions. The greater the similarity is between them, the less gram-
maticalized the finite motion verbs in converb constructions are.

3.2.1 Mieru suppletion


As mentioned in section 2, grammaticalized converb complex predicates do not show
the morphological property of lexical integrity, such that finite motion verbs may un-
dergo morphological processes on their own right. The first of such processes exam-
ined here is the suppletion of kuru ‘come’ by the honorific expression mieru ‘(lit.) vis-
ible’ shown below, where lexical kuru is suppleted by mieru:
(13) Yoru-ni naru-to, takusan-no kyaku-ga ku-ru/mie-ru
night-to become-when many-GEN customer-NOM come-PRES/come.HON-PRES
rasii.
EVI
‘I hear that many customers come when night falls (speaking of a restau-
rant).’
Kuru ‘come’ of the arui-te kuru ‘walk come’ and hait-te kuru ‘enter come’ type eas-
ily supplete with mieru, but that of the non-de kuru ‘drink come’ type does not. This
indicates that while kuru in the former two types is similar to the main verb kuru, the
one in the latter type is not.
(14) a. Yamada-sensei-wa gakkoo-ni arui-te ki-ta/mie-ta.
Yamada-Prof.-TOP school-to walk-CON come-PAST/come.HON-PAST
‘Professor Yamada walked (walk come) to school.’
b. Yamada-sensei-ga kyoositu-ni hait-te kit-ta/mie-ta
Yamada-Prof.-NOM classroom-to enter-CON come-PAST/come.HON.-PAST
toki...
when
‘When Professor Yamada came into (enter come) the classroom…’
c. Yamada-sensei-wa ippai non-de ki-ta/*mie-ta.4
Yamada-Prof.-TOP a.drink drink-CON come-PAST/come.HON-PAST
‘Professor Yamada had a drink (and came).’
Grammaticalization of converb constructions 

3.2.3 Rassyaru truncation


That there is no pragmatic conflict in the honorific expression of non-de kuru ‘drink
come’ is shown by the fact that irassyaru, another honorific suppletive form of iku/
kuru ‘go/come’, is perfectly acceptable in this type of converb construction. Indeed, the
three types of converb construction under study are all compatible with this honorific
form. What is particularly interesting about the irassyaru form is that it truncates to
rassyaru by deleting the initial vowel.
(15) a. Yamada-sensei-wa gakkoo-ni arui-te ki-ta/irassyat-ta/rassyat-ta.
Yamada-Prof.-TOP school-to walk-CON come-PAST/come.HON-PAST
‘Professor Yamada walked (walk come) to school.’
b. Yamada-sensei-ga kyoositu-ni hait-te kit-ta/irassyat-ta/rassyat-ta
Yamada-Prof.-NOM classroom-to enter-CON come-PAST/come.HON.-PAST
toki...
when
‘When Professor Yamada came into (enter come) the classroom…’
c. Yamada-sensei-wa ippai non-de ki-ta/irassyat-ta/rassyat-ta.
Yamada-Prof.-TOP a.drink drink-CON come-PAST/come.HON-PAST
‘Professor Yamada had a drink (and came).’
While the truncated rassyaru versions in (14) are all acceptable, there is a difference in
the degree of felicity in these expressions. Rassyaru truncation is least favored with
aruite-kuru ‘walk come’ in (14a), though it is not impossible.5 On the other hand, non-
de kuru ‘drink come’ is the favored type for a contracted rassyaru expression. The hait-
te kuru ‘come enter’ type falls in the middle of these two polar types. These impres-
sionistic evaluations of the three rassyaru forms are corroborated by the ratio of
contracted rassyaru forms against their non-contracted irassyaru counterparts. The
following table is based on a Google search which excluded irrelevant rassyaru/
irassyaru forms such as honorific versions of iru ‘be’.6
Table 5 contains polite imperative expressions (e.g. Go wash your face). While the
truncated forms appear regularly with the arui-te kuru ‘walk come’ type, the ratio of
truncated expressions is not as high as with the non-de kuru ‘drink come’ type, for
which the favored versions are clearly truncated rassyaru forms.
The significance of the truncation phenomenon is that the lexical verbs kuru/iku
‘come/go’, which easily supplete for irassyaru, resist truncation:
(16) a. Yamada-sensei-ga kyoositu-ni ki-ta/irassyat-ta/*rassyat-ta.
Yamada-Prof.-NOM classroom-to come-PAST/come.HON-PAST
‘Professor Yamada came to the classroom.’
b. Yamada-sensei-ga Amerika-ni it-ta/irassyat-ta/*rassyat-ta.
Yamada-Prof.-NOM America-to go-PAST/go.HON-PAST
‘Professor Yamada went to America.’
 Masayoshi Shibatani

The fact above that the honorific irassyaru of the arui-te kuru ‘walk come’ type
least favors the truncation indicates that the verb kuru here is most similar to lexical
kuru, whose honorific irassyaru version never contracts to rassyaru. On the other
hand, kuru of the non-de kuru ‘drink come’ type is least verb-like in that it favors the
rassyaru truncation. That kuru of the arui-te kuru ‘walk come’ type permits the rassyaru
truncation, though less frequently, is significant in that this verb has a function differ-
ent from the main verb kuru. This may mean that the motion verb has been gram-
maticalized to some extent even in arui-te kuru ‘walk come’ type constructions, as the
reduction in form permitted in them is certainly an indication of grammaticalization.
At any rate, it is clear from the pattern of the truncation of irassyaru that kuru of the
non-de kuru ‘drink come’ type is least verb-like in favoring the truncation, contrary to
Teramura’s and Hasegawa’s analyses that this form is similar to lexical kuru.

Table 5.  Truncation ratio of rassyaru vs. irassyaru

(kotti-ni) aruit-te irassyai 27


(kotti-ni) arui-te rassyai 23
‘ (here-to) walk-CON come.HON’

(kotti-ni) hasit-te irassyai 5


(kotti-ni) hasit-te rassyai 5
‘(here-to) run-CON come.HON’

de-te irassyai 494


de-te rassyai 1,060
‘exit-CON come.HON’

non-de irassyai 34
non-de rassyai 84
‘drink-CON come.HON’

tabe-te irassyai 57
tabe-te rassyai 155
‘eat-CON come.HON’

kao-o arat-te irassyai 23


kao-o arat-te rassyai 115
‘face-ACC wash-CON come’

3.2.3 Valency property


The lexical motion verbs iku ‘go’ and kuru ‘come’ sanction both goal and source argu-
ments, though the goal argument is rarely seen with the verb of coming due to the lexi-
cal specification of the deictic information in it – that the motion is directed toward a
deictic center, typically the place of speech. This valency property of these motion verbs
Grammaticalization of converb constructions 

plays an important role in Japanese converb constructions. Unlike their English coun-
terparts, manner-of-motion verbs in Japanese such as aruku ‘walk’, oyogu ‘swim’, and
hau ‘crawl’ do not allow the specification of a goal argument. In order to express a goal
with these verbs, they must be combined with iku ‘go’ or kuru ‘come’, as shown below:
(17) a. *Taroo-wa gakkoo-e arui-ta.
Taro-TOP school-to walk-PAST
‘Taro walked to school.’
b. Taroo-wa gakkoo-e arui-te it-ta/ki-ta.
Taro-TOP school-to walk-CON go-PAST/come-PAST
‘Taro walked (walk went/came) to school.’
The fact that iku of the arui-te iku ‘walk go’ construction licenses a goal argument
means that it retains the valency property of lexical iku. Compare this with iku in the
other types of converb construction.
(18) a. Taroo-wa zibun-no heya-o de-te, Hanako-no heya-ni it-ta.
Taro-TOP self-GEN room-ACC exit-CON Hanako-GEN room-to go-PAST
‘Taro exited his room and went to Hanako’s room.’
a’. *Taroo-wa Hanako-no heya-ni zibun-no heya-o de-te it-ta.
Taro-TOP Hanako-GEN room-to self-GEN room-ACC exit-CON go-PAST
‘(lit.) Taro went out (exit went) of his room to Hanako’s room.’
b. Taroo-wa ringo-o tabe-te, gakkoo-e it-ta.
Taro-TOP apple-ACC eat-CON school-to go-PAST
‘Taro ate and apple and went to school.’
b’. *Taroo-wa gakkoo-e ringo-o tabe-te it-ta.7
Taro-TOP school-to apple-ACC eat-CON go-PAST
‘(lit.)Taro eat-went an apple to school.’
While clausal (or VP) conjunction converb constructions in the (a) and (b) forms
are perfectly grammatical, the parallel complex predicate forms (a’) and (b’) do not
permit a goal expression, indicating that the motion verbs in the latter have lost the
verbal property of licensing a goal argument. These should also be compared with the
forms in (17b), which show that motion verbs in it retain this argument licensing
property, and thus behave more like lexical iku/kuru. In order to better appreciate the
property of valency loss seen in (18), compare them to the following pattern seen in
English, where the verb go in combination with eat in (19b) has lost the argument li-
censing property associated with its lexical counterpart in (19a).
(19) a. Let’s go to McDonald’s to eat.
b. *Let’s go eat to McDonald’s.
c. Let’s go eat at McDonald’s.
 Masayoshi Shibatani

3.2.4 Fragments
A converb in an ungrammaticalized construction can form a sentence fragment in
response to a yes-no question. Compare the following examples:
(20) a. Zitensya-ni not-te gakkoo-e kita=no?
bicycle-to ride-CON school-to come-PAST=Q
‘(You’ve) come to school riding a bicycle?’
b. Un, zitensya-ni not-te.
yeah bicycle-to ride-CON
‘Yeah, riding a bicycle.’
Again, grammaticalized converb constructions do not behave uniformly. The arui-te
kuru ‘walk come’ type barely allows a fragment expression, while the other two types
resist it altogether.
(21) a. Arui-te ki-ta=no?
walk-CON come-past=Q
‘(You) came walking?”
b. *?Un, arui-te.
yeah walk-CON
‘Yeah, by walking.’
(22) a. De-te ki-ta=no?
exit-CON come-PAST=Q
‘(You) came out?’
b. *Un, de-te.
yeah exit-CON
‘Yeah, (having) exited.’
(23) a. Ippai non-de ki-ta=no?
a.drink drink-CON come-PAST=Q
‘(You) had a drink (and came)?’
b. *Un, non-de.
yeah drink-CON
‘Yeah, (having) drunk.’
There may be a number of ways of interpreting this phenomenon, but it can be
understood as showing that the converb complex arui-te kuru ‘walk come’ is structur-
ally ‘less tight’, hence the converb and the finite verb are morphologically more au-
tonomous than those involved in the other two construction types. This interpretation
is corroborated by a pattern of contracting of iku ‘go’ to -ku. According to the Google-
based survey summarized in Table 6 below, iku of the arui-te iku ‘walk go’ type con-
tracts least while that of the non-de iku ‘drink come’ type contracts most frequently. 8
Grammaticalization of converb constructions 

Table 6.  Contraction of iku to -ku

arui-te iku=to 328,000


arui-te-ku=to 956
walk-go=when 0.003 %
de-te iku=to 58,200
de-te-ku=to 637
exit-go=when 0.01 %
tabe-te iku=to 17,400
tabe-te-ku=to 751
eat-go�����
=����
when 0.04 %

Notice that lexical iku never contracts to ku. The fact that iku of arui-te iku ‘walk go’
contracts less frequently compared to that of tabe-te iku ‘eat go’ indicates that the
former is structurally more autonomous than the latter. This contraction fact also
shows that the tabe-te iku ‘eat go’ type is at a more advanced stage of grammaticaliza-
tion than the other two types.

3.2.5 Scope of negation


The status of the motion verbs in the non-de kuru ‘drink come’ type noted above is
consonant with the fact that these verbs cannot support negative scope, indicating that
they are turning into a suffix status. Observe:
(24) a. Daremo basu-ni not-te ko-na-katta.
no one bus-to ride-CON come-NEG-PAST
‘No one came riding a bus.’
(No one came on the bus = wide scope reading)
(They came but no one came by bus = narrow scope reading)
b. Daremo heya-kara de-te ko-na-katta.
no one room-from exit-CON come-NEG-PAST
‘No one came out of the room.’
(No one came out = wide scope reading)
(No narrow scope reading possible; They came but not exiting a room)
c. Daremo gohan-o tabe-te ko-na-katta.
no one meal-ACC eat-CON come-NEG-PAST
‘No one had a meal (and came).’
No wide scope reading is possible; No one came after having eaten a meal)
(They came but no one had a meal = narrow scope reading)
It is clearly seen that while the finite motion verb of the arui-te kuru ‘walk come’
can support negative scope also giving rise to a wide scope reading for an expression
like (24a), those of the non-de kuru ‘drink come’ type cannot – (24c) above can never
 Masayoshi Shibatani

mean that no one came. The motion verb here is thus behaving more like an affix,
which cannot support negative scope independently of the root to which it is attached.
The reason that the de-te kuru ‘exit come’ does not allow a narrow scope reading is
unclear, but it may have to do with the a higher degree of lexicalization of this form.

3.3 Summary of the syntactic properties of the three types of converb construction

The following table summarizes the findings discussed in this section.

Table 7.  Summary of syntactic patterns

mieru rassyaru -ku Valency Fragment Neg. scope

Lexical kuru
‘come’ 〇 ✕ ✕ 〇 〇 N/A
arui-te kuru
‘walk come’ 〇 △ △ 〇 △ wide/narrow
de-te kuru
‘exit come’ 〇 〇 〇 ✕ ✕ wide
non-de kuru
‘drink come’ ✕ ◎ ◎ ✕ ✕ narrow
(◎ = super, 〇= O.K., △=grudgingly, ✕ = no)

Table 7 clearly substantiates our earlier claim that converb construction types are
distributed along a cline of grammaticalization as depicted in (12) above. In particular,
it shows that motion verbs of the non-de kuru ‘drink come’ type are most advanced in
grammaticalization, and are least verb-like – contrary to the analyses of Teramura
(1984) and Hasegawa (1996). Moreover, those of the de-te kuru ‘exit come’ type still
retain some verbal properties, indicating that they have not decategorialized com-
pletely as implied in Hasegawa’s work. Finally even motion verbs of the arui-te kuru
type, being susceptible to phonological reduction, behave differently from their lexical
counterparts, indicating that they are on the way to being (further) grammaticalized.

4. Methodological and theoretical implications

On the methodological front, the discussion above shows the importance of syntactic
analysis in grammaticalization studies. On the one hand, intuitive semantic analyses
seldom carry over to proper syntactic analyses. For non-de kuru ‘drink come’ type ex-
pressions, Teramura (1984: 157) – while recognizing a subtle difference between the
relevant converb constructions and their paraphrases – suggests that they can be un-
derstood as expressing sequential events, e.g. an action of drinking followed by a mo-
Grammaticalization of converb constructions 

tion of coming. In grammaticalized form, however, these expressions do not express


sequentially connected events. Specifically, finite motion verbs no longer express the
whole meaning of a motion event. The meaning of motion verbs of going and coming
comes in two dimensions, one expressing movement of an entity along a spatial path,
and the other a deictic orientation of the movement. In the non-de kuru ‘drink come’
construction type, the dimension of physical movement has been completely lost, as
reflected in the loss of a goal-sanctioning valency property. The motion verbs in these
constructions now have the sole function of indicating deictic information, i.e. indi-
cating where the actor ends up after the execution of the action expressed by the con-
verb. Thus, sake-o non-de ki-ta (sake-ACC drink-CON come-PAST) does not assert
that someone performed an act of translational motion toward the deictic center after
having drunk sake. Rather, it means that someone ended up in the speaker’s sphere
after having drunk sake somewhere away from it. It is for this reason that the transla-
tions of relevant examples in this paper parenthesize the motion component as ‘drank
sake (and came)’.
Syntactic analysis, furthermore, is crucial in showing that the phenomena under
consideration fall in the category of grammatical change known as grammaticaliza-
tion. Meaning shifts and extensions by themselves do not constitute grammaticaliza-
tion phenomena per se. In order for a semantic change to qualify as a case of gram-
maticalization, the form in question must be grammaticalized, i.e., change its status
from lexical to grammatical. In other words, decategorialization is an important com-
ponent of the grammaticalization process. Thus, when verbs are involved as they are
here, it is important to show that they lose certain verbal properties.
There are many theoretical implications of the present study, but here we restrict
them to those that appear to be intimately connected. The first concerns the question
of the gradualness of grammaticalization.

4.1 Gradual vs. instantaneous grammaticalization

From the time of the Neogrammarians, the question of gradualness of historical


change (phonetic or otherwise) has been controversial, the issue itself having multiple
facets ranging from the actualization of change to its spread. The same issue arises with
changes characterizable as grammaticalization. The pattern in Table 7 indicates that
gradual loss of verbal characteristics in finite motion verbs affects different types of
converb constructions. However, the question remains as to whether those at an ad-
vanced stage in grammaticalization realize each step of the grammaticalization paths.
That is, did iku ‘go’ and kuru ‘come’ in the non-de kuru ‘drink come’ construction type
once had a stage where they behaved like those in the arui-te kuru ‘walk come’ type?
Was there a stage where kuru in the former type could be suppleted with mieru or
where iku sanctioned a goal argument as those in the latter type now do? Alternatively
could the change be instantaneous such that as soon as a construction of the non-de
kuru ‘drink come’ is formed, the finite motion verbs contained therein are used as
 Masayoshi Shibatani

deictic markers losing their verbal properties instantaneously?9 Both positions have
been maintained in the field. Brinton and Traugott (2005) and others (e.g., Bybee,
Pagliuca and Perkins 1991) generally believe that grammaticalization is gradual, while
the possibility of instantaneous grammaticalization is entertained by Givón, as shown
by the following quotations:
Grammaticalization is gradual in the sense that it is non- instantaneous and pro-
ceeds by very small and typically overlapping, intermediate, and sometimes
indeterminate, steps. (Brinton and Traugott 2005:100)
[Instantaneous grammaticalization] involves the mental act of the mind rec-
ognizing a similarity relation and thereby exploiting it, putting an erstwhile
lexical item into grammatical use in a novel context. The minute a lexical item
is used in a frame that intends it as grammatical marker, it is thereby gram-
maticalized. (Givón 1991:122)
While changes in the formal properties of a grammaticalized item may be gradual,
as shown above in the contraction phenomena of motion verbs (see Tables 5 and 6),
there seems to be at least clear cases where instantaneous grammaticalization occurs
accompanying the functional shift from the spatial to the temporal domain. As alluded
to earlier, Japanese converb constructions of coming and going have also developed a
function of aspectual marking. Examples include the following:
(25) a. soredake syabet-te simau-to hidoku hara-ga het-te
that.much talk-CON finish-when very stomach-NOM decrease-CON
ki-ta.
come-PAST
‘(lit.) When I finished talking all that much, my stomach began to decrease/I
began to become hungry.’
b. sikamo sono aida kekkoo ta-no onna-to ne-te
moreover that interval well.enough other-of woman-with sleep-CON
ki-ta.
come-PAST
‘Moreover during that period (I) have been sleeping well enough with other
women.’
c. sooyatte iki-te ik-u.
that.way live-CON go-PRES
‘(I) will go on living that way.’
d. Atama-ga dandan okasiku nat-te it-te…
brain-NOM gradually funny become-CON go-CON
‘My brains gradually keep on turning funny…’
(Haruki Murakami Norwegian Wood)
Grammaticalization of converb constructions 

These forms can be considered to have an aspectual function in that iku/kuru ‘go/
come’ add a span to the event expressed by the converbs. The finite expression hara-ga
hetta ‘(lit.) the stomach decreased/got hungry’, for example, locates change of state at
one point in time. When it combines with kuru ‘come’ as in (25a) hara-ga het-te ki-ta,
the whole expression entails change of state over a span of time. Moreover, this expan-
sion of change over time span is conceived to be directed toward the speaker by virtue
of the use of kuru ‘come’. When iku is used as in (24c,d), the expansion of the event is
viewed to take place in a direction away from the speaker, giving a future-oriented
interpretation if the finite form has a present tense ending as in (25c).
The widely received scenario for the development of temporal meanings from mo-
tion verbs developed by Bybee, Perkins and Pagliuca (1994: 269) holds that meaning
elements, e.g. inferences, giving rise to this development already exist in the lexical use
of motion verbs. For example, “[w]hen one moves along a path toward a goal in space,
one also moves in time.” What happens is the gradual loss of the spatial meaning with
increased dominance of the temporal meaning. Under this scenario the only change
necessary is the generalization to contexts in which the subject is not moving, as in
(25) above, where an aspectual meaning uniquely obtains. This scenario of gradual
grammaticalization, however, faces a serious challenge when confronted with a wide
range of cases of the development of temporal meanings from motion and other types
of verbs. Crucial cases in Japanese relate to examples such as the following, where no
aspectual reading is possible:
(26) Boku-wa kore-kara-mo eki-e arui-te ik-u.
I-TOP this-from-also station-to walk-CON go-PRES
‘I will walk to the station from now on too.’
As detailed in the next section, it is precisely in those contexts where the meaning of
spatial movement does not obtain that give rise to the aspectual meaning in question.
That is there is no stage where the aspectual meaning has become dominant with con-
comitant bleaching of the spatial meaning that bridges the gap between literal spatial
expressions and aspectual ones, which are purported to derive via generalization of the
intermediate expressions. The shift in meaning here, instead, appears best considered
as a case of instantaneous metaphorical transfer. That is, once an image schema of mo-
tion is metaphorically transferred from the spatial domain to the temporal one, then
the grammaticalization takes place instantaneously, as in Givón’s characterization
above. Whether a meaning shift such as this would count as a case of grammaticaliza-
tion is debatable, but if the change were accompanied by a decategorialization process,
it would be considered as a case of grammaticalization. Indeed, the Japanese aspectual
iku ‘go’ and kuru ‘come’ do not behave like lexical iku/kuru. For example, aspectual
kuru ‘come’ does not supplete for honorific mieru, unlike lexical kuru, and honorific
irassyaru of aspectual iku contracts to rassyaru, unlike the irassyaru form suppleting
lexical iku (see earlier discussions on these phenomena). Observe:
 Masayoshi Shibatani

(27) a. Satoo-sensei-wa syoozikini iki-te ki-ta/*mie-ta.


Satoo-Prof.-TOP honestly live-CON come-PAST
‘Prof. Sato has been living honestly.’
b. syoozikini iki-te ik-eba/irassyar-eba/rassyar-eba...
honestly live-CON go-if/go.HON-if/go.HON-if
‘If (you) keep on living honestly…’

4.2 Contexts facilitating grammaticalization

Our study raises an important question that past studies do not concerning the context
or environment which propels grammaticalization. The earlier discussion shows that
grammaticalization of motion verbs does not take place uniformly. The verb iku ‘go’ in
non-de iku ‘drink go’ constructions is far more advanced in grammaticalization than in
the arui-te iku ‘walk go’ constructions. Generalizing the case to include instantaneous
grammaticalization, the question can be formulated as: “Which contexts facilitate
grammaticalization?” By “facilitate” we refer to two situations, one where a specific
context propels grammaticalization at a faster rate than others, the other concerning
the context in which instantaneous grammaticalization – including metaphorical ex-
tension – is easier to take place.
Past grammaticalization studies have raised a similar question but at a very gen-
eral level. For example, Heine, Claudi and Hünnemeyer (1991) show that metaphori-
cal extension or transfer of an imagine schema from one (concrete) domain to another
(abstract) is a major driving force for semantic change and grammaticalization. We
have also recognized the effect of such a mechanism in the development of the aspec-
tual use of iku/kuru ‘go/come’ in Japanese. But metaphorical extension does not ex-
plain the cline of grammaticalization exhibited by iku/kuru in the spatial domain,
where metaphorical extension is not involved. A question also remains as to which
contexts favor/disfavor metaphorical extension or instantaneous grammaticalization
in the sense of Givón (1991).
Another factor invoked in the discussion of both the actualization and spread of a
grammaticalization process is textual frequency. Here is a relevant quote from Traugott
and Heine (1991: 9):
Given that a form A is a candidate for grammaticalization both because of its se-
mantic context and its salience, a further condition has to apply for grammati-
calization to take place: The form has to be used frequently. The more grammatical-
ized a form, the more frequent it is...The seeds of grammaticalization are therefore
in a correlated set of phenomena: Semantic suitability, salience and frequency.
Only the third actually leads to grammaticalization and hence to fixing, freezing,
idiomatization, etc. (Emphasis added)
Grammaticalization of converb constructions 

Does this apply to our case? Not really. First, situations concerning frequency are not
as straightforward as Traugott and Heine (1991) put it. That is, determining the token
frequency of constructions is not simple, since the frequency of use of a construction
varies greatly depending on the specific lexical items involved. For example, in one
recent Google count hat-te iku (crawl-CON go) occurred 963 times, whereas the same
Manner of motion + Motion combination of arui-te iku (walk-CON go) yielded
623,000 instances. In ascertaining the token frequency, then, it is necessary to compare
the frequency of different types of verb combination. The following table, from Shiba-
tani and Chung (to appear), shows the token frequencies of the three combination
types in a modern fiction in Japanese and Korean:

Table 8.  Token frequency of three combination types (Japanese; Ito Sei Hanran: Korean;
Kim Joo Yong Kokicapinun kaltaylul kkekkci anhnunta)

Manner Location change Action

Japanese
V-te iku/kuru 23 109 6
V-CON go/come
Korean
V-e/ko ota/kata 68 358 11
V-CON come/go

Our case may also be at variance with a generally observed correlation between


frequency of use and the size of a grammatical item. For example, Bybee, Perkins and
Pagliuca (1994: 20) draw the following conclusion:
There is a link between frequency of use and phonetic bulk such that more fre-
quently used material, whether grammatical or lexical, tends to be shorter (pho-
netically reduced) relative to less often used material.

Determining the frequency of use across construction types is rather difficult, as noted
above, but if Tables 5 and 6 above suggest anything, it is that in those expressions that
are more advanced in grammaticalization but still occur comparatively less frequently
we observe higher rates of occurrence of phonologically reduced forms.10
Neither metaphorical extension nor frequency is thus able to account for the pat-
tern of grammaticalization examined in this paper at both the functional and the for-
mal level. In the balance of this paper, we will explore semantic factors differentiating
the contexts that facilitate grammaticalization from those that hinder it.
Verbs conjoined by the -te conjunctive in Japanese express different kinds of event
combinations. Combinations of the motion verb iku ‘go’ or kuru ‘come’ and manner of
motion verbs (aruku ‘walk’, hasiru ‘run’, oyogu ‘swim’, etc.) result in a semantically con-
gruous whole in that there is complete spatio-temporal overlap, such that when one
 Masayoshi Shibatani

walks, s/he necessarily moves (goes or comes), and when one goes or comes to some
place, s/he does so in some manner (e.g., walk, run, swim). Indeed, English lexicalizes
macro-events comprised of these sub-events in manner-of-motion verbs such as walk
and swim, which – unlike their Japanese counterparts – sanction a goal argument (e.g.
walk to the station, swim to the shore). Combinations of location change and motion are
similarly congruous in that there is partial spatio-temporal overlap. But the overlap is
not to the full extent as in the Manner + Motion combinations, since it obtains only at
the point of threshold in the case of Location change + Motion combinations.
Compared to these combinations, those of sequentially ordered events, such as
drinking and going to some place or eating and coming back, i.e., those not tied by
such unifying relations as cause, purpose, and result lack semantic congruity and do
not usually form a macro-event that is lexicalized in a single verb. Indeed, they do not
even enter into verb serialization in many languages.11
The three types of converb construction examined in this paper thus have the fol-
lowing patterns of event combination, which can be characterized in terms of degree
of semantic congruity:

More congruous

Manner + Motion

Location change + Motion

Action + Motion

Less congruous

Notice that in Japanese, the combination of action and motion represents a pre-gram-
maticalization stage. That is, current forms such as non-de kuru ‘drink come’ and tabe-
te iku ‘eat go’ do not combine action and motion in the literal sense, since they have
already undergone grammaticalization, as described above. Our characterization of
the three types of event combination then allows us to draw the following hypothesis
on the context that propels or facilitates grammaticalization:
(28) Grammaticalization is facilitated in semantically less congruous environ-
ments
This hypothesis makes sense for the process of semantic bleaching, whereby a meaning
component is more easily lost when there is no semantic support or reinforcement
from the surrounding environment. Interestingly it is precisely this kind of semanti-
cally incongruous environment where metaphorical extension is also facilitated. As
mentioned above, Japanese converb constructions involving verbs of going and com-
Grammaticalization of converb constructions 

ing have also developed aspectual meanings (see examples in (25)). In combination
with a converb, however, the aspectual meaning of iku/kuru ‘go/come’ obtains most
readily in semantically incongruous combinations of events. Indeed, a large majority
of the tabe-te iku ‘eat go’ and non-de iku ‘drink go’ forms found in Google are of the
aspectual type illustrated in the following examples:
(29) a. eigo-de gohan-o tabe-te ik-u
English-with meal-ACC eat-CON go-PRES
‘(lit.) to go on eating meals with English/to go on living on the job using English’
b. dondon karee-o tabe-te ik-u
steadily curry-ACC eat-CON go-PRES
‘to keep on eating curry steadily’
(30) a non-de ik-u utini azi-ni-mo nare….
drink-CON go-PRES while taste-to-also get.used.to
‘(you) also get used to the taste while (you) keep on drinking’
b. kare-mo dondon biiru-o non-de ik-u
he-also steadily beer-ACC drink-CON go-PRES
‘He too keeps on drinking beer.’
There are more instances of spatial expressions than aspectual ones involving non-
de kuru ‘drink come’ and tabe-te kuru ‘eat come’ forms, but one still encounters the
following kinds fairly frequently:
(31) a. Nan-no tame-ni ima-made gyuunyuu-bakari non-de
what-GEN reason-to now-up.to milk-only drink-CON
ki-ta=n. desu=ka
come-PAST=NMLZ COP=Q
‘For what reason have you been drinking only milk?’
b. Ningen-wa nani-o tabe-te ki-ta=ka.
human-TOP what-ACC eat-CON come-PAST=Q
‘What have humans been eating?’
In contrast, there are no instances of the aspectual type for combinations like arui-te
iku ‘walk go’ and de-te iku ‘exit go’. Indeed, native speakers of Japanese find a clear dif-
ference in grammaticality between the (a) and (b) sentences below:
(32) a. korekara-mo dondon sake-o non-de ik-u
from.now-also steadily sake-ACC drink-CON go-PRES
‘From now on too (I will) keep on drinking sake steadily.’
b. *korekara-mo dondon eki-e arui-te iku
from.now-also steadily station-to walk-CON go-PRES
(Intended for) ‘From now on too (I will) keep on walking to the station.’
 Masayoshi Shibatani

(33) a. Ano kooen-o sanzyuu-nen-rai zutto arui-te ki-ta.


that park-ACC thirty-year-over steadily walk-CON come-PAST
‘(I) have walked in that park steadily for over thirty years.’
b. *Sanzyuu-nen-rai zutto uti-kara gakkoo-ni arui-te ki-ta.
thirty-year-over steadily house-from school-to walk-CON come-PAST
‘(I) have walked from the house to the school for over thirty years.’
Notice that in (33a) we have an action expression ‘walk (in) the park’ rather than
a translational motion event ‘walk to some place’. The combination of the former with
a motion verb expresses an incongruous event combination. In (33b), we have a con-
gruous combination of manner and motion events, which fails to yield an aspectual
interpretation. In other words, where a literal interpretation of the V-te iku/kuru form
yields an incongruous event combination, aspectual interpretation obtains. This means
that instantaneous grammaticalization (or a case involving metaphorical extension) is
facilitated in the same environment where non-metaphorical and gradual grammati-
calization is propelled, namely a semantically incongruous context. This situation has
parallels in the temporal domain in other languages such as Formosan language Atay-
al and Thai, calling into question the widely accepted scenarios of the development of
tense/aspect morphemes offered by Heine et al. (1991), Hopper and Traugott (1993),
and Bybee et al. (1994), who suggest that tense/aspect meaning develops first in a se-
mantically congruous environment, where inferences of intention and prediction ob-
tain (see Shibatani and Huang 2006).

5. Summary and conclusion

This paper has examined two patterns of grammaticalization in Japanese. One repre-
sents a case of gradual grammaticalization in which a meaning component of transla-
tional motion verbs of coming and going is lost as they occur as a finite verb in converb
constructions. The loss of the motion meaning in these verbs has entailed a loss of
valency property such that they can no longer sanction a source or goal argument in
constructions at an advanced stage of grammaticalization. Advancement in grammat-
icalization is also accompanied by a formal reduction in motion verbs, as predicted by
the general principle of the grammaticalization process. It was then shown that neither
metaphorical extension nor frequency of use is a factor propelling grammaticalization
in these converb constructions with motion verbs. Instead, an environment facilitating
grammaticalization was identified as one in which an incongruous event combination
obtains. It was further demonstrated that plausible cases of instantaneous grammati-
calization involving metaphorical extension are also facilitated in an environment
where a literal interpretation of the expressions leads to semantically less congruous
event combinations. Congruous event combinations, on the other hand, inhibit meta-
phorical extension and instantaneous grammaticalization.
Grammaticalization of converb constructions 

This paper has also shown the importance of syntactic analysis in grammaticaliza-
tion studies. In addition, it underscores the importance of a meticulous investigation
of currently ongoing changes, which are much easier to document than less accessible
ones. Finally, dealing with grammatical constructions as a whole as units of analysis
affords a more fine-grained perspective than dealing with individual forms in isolation
since the constructions provide a linguistic context that may play an important role in
grammaticalization.

Notes

1. See Bisang (1995) on the similarities and differences between converb and serial verb
constructions and Shibatani and Huang (2006) for the demonstration that the converb complex
predicates and serial verbs do not form distinct types of complex predicates.
2. This caveat is necessary because a structure that raises the negative -nai may contain sika
and -nai in two separate clauses; e.g. Boku-wa [Taroo-ni-sika aitai]-to omow-anai ‘(lit) I don’t
think (I) want to meet only Taro/I think I don’t want to meet anybody but Taro/I don’t think I
want to meet anybody but Taro’. See Matsumoto (1996) for similar arguments that (3a) involves
a simplex structure.
3. Actually, the categorial status of what she calls operator is not entirely clear in Hasegawa’s
work. But if the central definitional property for verbs is their ability to function as a predicate,
then Hasegawa’s deictic operators do not count as verbs, since they do not function as predica-
tes.
4. The form non-de mieta is fine if understood as an honorific version of non-de iru (drink be)
‘is drinking’.
5. Be aware that arui-te kuru ‘walk come’, for example, can be ambiguous between a transla-
tional motion reading, as in (14a) with a specified goal or source argument, and a nontranslatio-
nal action reading, as in kooen-o arui-te kuru ‘walk (in) the park (and come)’, which belongs to
the non-de kuru ‘drink come’ type of (14c).
6. The table lists de-te irassyaru/rassyaru ‘exit come’ rather than hait-te irassyaru/rassyaru ‘en-
ter come’ because the latter are involved in a large number of ambiguous cases between the in-
tended hait-te iku type and the non-de iku type.
7. Expressions like the following are possible with the interpretation that the effect of drinking
persists, hence overlaps with the motion event:
Gakkoo-ni sake-o non-de ku-ru=to-wa nanigoto=ka?
school-to sake-ACC drink-CON come-PRES=COMP-TOP whatever =Q
‘Whatever (is the matter with you) to come to school (after) drinking sake?’
This is similar to the following type of expression, where again the result of an action overlaps
with the motion event.
Hanako-wa gakkoo-ni mizikai sukaato-o hai-te ki-ta.
Hanako-TOP school-to short skirt-ACC wear-CON come-PAST
‘Hanako came to school wearing a short skirt.’
 Masayoshi Shibatani

To a large extent, these expressions behave similarly to the arui-te kuru ‘walk come’ type, in
which there is a complete spatio-temporal overlap between the manner component and the
motion component of the motion event. See the relevant discussion in Section 4.
8. The expressions cited here are framed in a complement construction headed by the formal
noun to ‘when, as’ in order to exclude extraneous instances of the relevant phenomena.
9. In answering the questions posed here we need historical evidence. Another type of evi-
dence comes from languages having similar construction types. See Shibatani and Chung (2005)
for Korean evidence suggesting a gradual loss of valency property in the motion verbs of the
non-de kuru ‘drink come’ type construction.
10. As mentioned in the text, determining the correlation between frequency and form size is
tricky in that certain expressions occur more frequently across the board. For example, the im-
perative form De-te ik-e! (exit-CON go-IMP) ‘Get out!’ occurs very frequently despite being of
the de-te iku type. The ratio of reduced forms of this expression indeed increases more dramati-
cally than with the other expressions of the same construction type. Google count: de-te ik-e=to
(exit-CON go-IMP=COM) ‘(e.g., I was told) to get out’ 16,700; de-te-k-e=to 16,500.
11. In Mandarin, Thai and Atayal, for example, simple sequentially ordered events such as ea-
ting and then going to some place do not serialize.

References

Bisang, W. 1995. Verb serialization and converbs – differences and similarities. In Converbs in
Cross-Linguistic Perspective, M. Haspelmath and E. König (eds), 139–188. Berlin:
���������������
Mouton
de Gruyter.
Brinton, L. J. and Traugott, E. C. 2005. Lexicalization and Language Change. Cambridge: CUP.
Bybee, J., Pagliuca, W. and Perkins, R. 1991. Back to the future. In Approaches to Grammatical-
ization, Vol. 2, E. C. Traugott and B. Heine (eds), 17–58. Amsterdam: John Benjamins.
Bybee, J., Perkins, R. and Pagliuca, W. 1994. The Evolution of Grammar: Tense, aspect, and mo-
dality in the languages of the world. Chicago IL: University of Chicago Press.
DeLancey, S. 1991. Origins of verb serialization in Modern Tibetan. Studies in Language 15: 1–23.
Foley, W. and Olsen, M. 1985. Clausehood and verb serialization. In Grammar Inside and Out-
side the Clause: Some approaches to theory from the field, A. Woodbury and J. Nichols (eds),
17–60. Cambridge: CUP.
Givón, T. 1991. Serial verbs and the mental reality of ‘event’: Grammatical vs. cognitive packag-
ing. In Approaches to Grammaticalization Vol. 1, E. C. Traugott and B. Heine (eds), 81–127.
Amsterdam: John Benjamins.
Grein, M. 1998. Mittel der Satzverknüpfung im Deutschen und im Japanischen: Eine Typologisch-
Kontrastive Analyse. Wiesbaden: Deutscher Universitäts-Verlag.
Hasegawa, Y. 1996. A Study of Japanese Clause Linkage: The connective TE in Japanese. Stanford
CA: CSLI Publications/Tokyo: Kurosio Publishers.
Heine, B., Claudi, E. and Hünnemeyer, F. 1991. Grammaticalization: A conceptual framework.
Chicago IL: University of Chicago Press.
Hopper, P. and Traugott, E. C. 1993. Grammaticalization. Cambridge: CUP.
Lord, C. 1993. Historical Change in Serial Verb Constructions. Amsterdam: John Benjamins.
Grammaticalization of converb constructions 

Matsumoto, Y. 1996. Complex Predicates in Japanese: A syntactic and semantic study of the notion
‘Word’. Stanford CA: CSLI Publications/Tokyo: Kurosio Publishers.
Schmidt, C. 2004. Cosubordinate converbs in Japanese. Paper presented at the First Rice Univer-
sity and UT Austin Workshop on Language in Use. October 23–24, 2004. Rice University.
Shibatani, M. and Chung, S. Y. 2005. On the grammaticalization of motion verbs: A Japanese-
Korean comparative perspective. The 15th Japanese/Korean Linguistics Conference. Octo-
ber 7–9, 2005. University of Wisconsin, Madison. To appear in the proceedings.
Shibatani, M. and Huang, L. 2006. Serial verb constructions in Formosan languages: their form
and grammaticalization patterns. The Linguistics of Endangered Languages (Third Oxford-
Kobe Seminar in Linguistics), April 2–5, 2006. Kobe, Japan. To appear in the proceedings.
Teramura, H. 1984. Nihongo-no Imi-to Sintakusu (Syntax and Semantics of Japanese) Vol. II.
Tokyo. Kurosio Publishers.
Traugott, E. C. and Heine, B. (eds). 1991. Approaches to Grammaticalization. Amsterdam: John
Benjamins.
Contact, connectivity and language evolution

Yaron Matras
University of Manchester

This paper is concerned firstly with introducing a preliminary sketch of a


function-oriented theory of language contact. Second, it will address the position
of connectivity devices as mental or cognitive operations, which constitute part
of the bilingual’s communicative activity. From the behaviour of connectivity
devices in bilingual communication, I will be drawing some conclusions as to
their position in the grammatical apparatus. The final issue to be addressed
is a rather speculative addendum, a commentary on the possible position of
connectivity operations in the ancient architecture of the language faculty itself.

1. Preliminaries: A function-oriented theory of language contact

Linguists normally view contact-related change, and often also synchronic bilingual
behaviour, through the prism of self-contained linguistic ‘systems’: Contact is envis-
aged as a partial convergence of two systems. I propose an alternative approach to
language contact, one that consistently takes the perspective of the bilingual speaker,
and which places at the top of its agenda the task to investigate how speakers act in
multilingual settings. To the bilingual speaker, languages are not analytical ‘systems’.
They are, rather, components of an overall repertoire of forms, constructions, experi-
ence, and skills on which the speaker draws in order to communicate. The bilingual
speaker, very much like the monolingual speaker who is familiar with various registers
of just one language, acquires rules on the selection of components within this reper-
toire through experience and socialisation. These rules tend to point toward the selec-
tive usage of distinct repertoire components in separate environments, or for separate
sets of communicative activities. The key to communicating in a bi­-lingual repertoire
is therefore the acquisition and the application of skills that enable the bilingual speak-
er to draw demarcation boundaries within the repertoire, and to maintain those de-
marcation boundaries in line with the communicative conventions and social expecta-
tions of the surrounding speech community.1 Rather than talk about the bilingual’s
languages, I therefore choose to discuss the distinct components of a bilingual’s overall
repertoire of linguistic-communicative structures.
 Yaron Matras

Investigating language contact is therefore not about describing how systems con-
verge, but about investigating how successful bilingual speakers are in maintaining
demarcation lines within their linguistic repertoires: Where will speakers fail to main-
tain the boundary between components of the repertoire? Around which grammatical
categories or constructions? For which language-processing operations? When will
they license themselves consciously to lift the boundary? And under which circum-
stances will either the lifting of a demarcation boundary, or the failure to maintain it,
become habitual, and lead ultimately to a permanent shift in the position of the bound-
ary – i.e. to language change?
My basic assumption is that both the maintenance of demarcation boundaries
within the repertoire, and their partial lifting or even longer-term removal around
specific structures, are functional to communication in multilingual settings. This is a
view of language contact that does not discriminate between diachronic and syn-
chronic aspects of contact, or between idiosyncratic occurrences and widespread vari-
ation, but takes an integrated approach to the theme of acquiring, maintaining, and
re-setting demarcation boundaries within the repertoire. This is the view that is taken,
because any changes in the conventions of linguistic behaviour that become accepted
by the speech community will have had their roots in individual innovations that
speakers introduce at the level of the single utterance, embedded into a specific com-
municative event. As in monolingual speech (cf. Croft 2000), changes in bilingual set-
tings too are individual innovations which are imitated, replicated, and ultimately
propagated successfully throughout the speech community or in certain sectors of it.
Bilingual behaviour is a way to navigate through the challenge of communicating, us-
ing a relatively wide and diverse repertoire of linguistic structures, and under a com-
plex set of contextual constraints on variant selection. Contact-induced language
change is a set of solutions to this challenge, which have become favoured by a signifi-
cant number of speakers in a significant set of communicative constellations.
The task of contact linguistics is to describe and explain speakers’ communicative
navigation strategies in multilingual settings, and the conditions under which they
lead to longer-term solutions, and so to change. This begins with the gradual process
of the acquisition of the skill that enables the speaker to actively set demarcation
boundaries within the repertoire. In respect of the infant acquiring two languages
from birth, the agenda for investigation has long been defined as a need to analyse the
acquisition of boundaries as a social skill (Ochs 1988, Lanza 1997). In respect of more
mature speakers, who are generally self-confident in the skill of maintaining bounda-
ries among repertoire components, the agenda of the contact linguist is to define the
conditions under which those speakers license themselves to lift boundaries and make
discourse-strategic use of the repertoire in its entirety, or indeed of the contrast be-
tween repertoire components (cf. e.g. Maschler 1994, Auer 1999). Discourse-strategic
bilingual activities such as situational, conversational, and metaphorical codeswitch-
ing, insertions, loan translations or loan blends may be considered in this light.
Contact, connectivity and language evolution 

But even in respect of the mature and self-confident speaker, the question arises
whether difficulties in maintaining separation between repertoire components might
still occur, either in certain situations, or indeed around particular processing func-
tions of language, manifesting themselves around certain grammatical categories or
constructions. This would point toward a cognitive trigger, rather than a discourse-
strategic one, for certain kinds of language mixing. Are there instances of particular
language-processing operations, where under certain conditions even mature and
skilled bilingual speakers fail to keep apart their repertoire components? The hypo­
thesis put forward here is that this is indeed the case with certain types of connectivity
markers: When engaging in the linguistic-mental operation known as ‘connectivity’,
under certain conditions, bilingual speakers may ‘lose control’ of the selection mecha-
nism that maintains demarcation between the different components of the linguistic
repertoire, and select a structure that is cognitively functional (in that it triggers the
relevant processing function) but socially inappropriate (in that it is not an acceptable
variant in the particular setting). Some types of connectivity devices stand out in their
vulnerability in such situations.
From this it is firstly possible to derive an hypothesis of contact-induced language
change: Should the difficulties in maintaining a boundary between the repertoire com-
ponents persist around a particular function of language beyond just spontaneous and
idiosyncratic use, and should the resulting pattern of ‘language mixing’ become ac-
cepted and even propagated within the speech community, then a cognitive trigger may
be said to have sown the seeds for language change. Second, it can be argued that lan-
guage contact can shed light on the internal structure of grammar; such ‘pathological’
errors in navigating through the bilingual repertoire indicate that the mechanism in-
volved in selecting connectivity structures is more error-prone, and so that it is struc-
tured differently from that which governs the selection of other grammatical opera-
tions or lexical items. On this basis we might hypothesise that connectivity occupies a
different kind of position in the speech production mechanism.

2. Connectivity devices in multilingual language acquisition

I base this section on informal observations of the language acquisition process of a


trilingual child, whom we shall call ‘Ben’. Born and raised in England, Ben is exposed
to two languages at the home: German, which he hears from his mother, and Hebrew,
which he hears from his father. Both parents speak their respective languages consist-
ently to Ben, consciously trying to avoid mixing. Between the ages of 0:4 and 4:4, input
was balanced: During the first two years of his life, Ben spent four days a week with an
English-speaking child minder, and was cared for at home during roughly half of the
working week primarily by his father, and during the other half primarily by his moth-
er, while weekends were spent with both parents. At the age of 1:11, Ben’s parents split
up and moved into separate households, in separate towns. Ben stayed with his moth-
 Yaron Matras

er, spending three to four working days at an English-speaking nursery, while six days
out of a fourteen-day cycle were spent with his father. Holiday time was split equally
among the two parents. Most of the holiday time with the mother was spent in Ger-
many, and around half the holiday time with the father was spent in Israel – in both
countries with family and relations. On the whole, then, between the age of 0:4 and 4:4,
Ben spent roughly equal amounts of time with each of the two parents (each speaking
his/her language consistently) and at the English-speaking nursery, with exposure
during holidays to monolingual contexts of German and Hebrew.
Ben’s active language acquisition history begins between the age of 1:3–1:6, with
the acquisition of lexical items from all three input languages. In the early phase, the
lexicon appears mixed, characterised, not surprisingly, by the absence of bilingual syn-
onyms and the indiscriminate use of individual lexical items, irrespective of addressee
or context (cf. already Volterra & Taeschner 1978). A consistent desire to separate the
languages becomes clear however by the age of 1:9, by which time much of the vocabu-
lary is used selectively according to setting (setting being defined primarily by the par-
ent-addressee, and gradually also by a number of other adults who are associated with
one language rather than the other). Bilingual synonyms begin to appear during this
period, while there is also a set of preferred lexical items, containing elements from all
three languages, which the child continues to use irrespective of context.
By the age of 2:4 – 2:6, Ben has acquired a fairly fluent command of both his do-
mestic/parental languages, German and Hebrew, with English lagging somewhat be-
hind in active use. In both German and Hebrew, Ben is able to conjugate verbs in
various tenses, to make full use of person inflection and inflected pronouns or pro-
nominal clitics, and to produce complex sentences with the full range of conjunctions.
He is fully aware of the context-bound separation of languages, and pursues it consist-
ently. However, lapses do occur: In situations immediately following the transition
from one parental household to another, i.e. within a few hours or on the first day, or
in other situations of partial ambiguity, such as when speaking on the phone to one
parent while in the care of the other, Ben occasionally slips into the ‘wrong’ language,
that is, into the language that is not the language of the addressee (i.e. the slips are ei-
ther into the language of the parent with whom he had been spending the past few
days prior to the transition, or, in the case of phone conversations, into the language of
the parent at whose house he is currently staying). The slips are unintentional; some-
times they are noticed and self-repaired, but quite often they remain unnoticed by the
child, and usually un-commented on by the hearer.
The most interesting aspect of these bilingual slips is the fact that they involve al-
most exclusively a particular class of discourse and connectivity markers. Most fre-
quently affected are the particles ‘yes’ (Hebrew ken/ German ja) and ‘no’ (lo/nein), the
conjunctions ‘because’ (ki/weil), ‘and’ (ve/und), ‘or’ (o/oder), and ‘but’ (avál/aber), and
occasionally focus particles such as ‘too’ (gam/auch), ‘even’ (afílu/sogar) or ‘at all’ (bix-
lál/überhaupt):
Contact, connectivity and language evolution 

(1) Hebrew; age 2:3, first few days in the father’s care after returning from a 3-
week holiday in Germany; inspecting the shell of a snail in the garden:
báyit šel xilazón aber éyn xilazón bifním
house of snail but [German] is-no snail inside
‘A snail-shell, but there is no snail inside’
(2) German; age 7, on the phone to the mother while at the father’s house, de-
scribing a collection of insects:
dann gibt’s Butterfly, ve/ und zwei Craneflies
then is and [Hebrew] and two
‘Then there’s a butterfly, and/ and two craneflies’
(3) Hebrew; age 5:2, reaching to inspect his trousers while planning a game in
which a toy is to be hidden in a pocket:
yeš li überhaupt kis?
there.is to.1sg at all [German] pocket
‘Do I even have a pocket?’
(4) Hebrew; age 7:2, on the phone to his father while on holiday in Germany,
about a sports event that was being shown on television there:
Jan Ulrich hayá be’érex mispár šéva oder šmóne/ o šmóne
was approximately number seven or [German] eight or eight
‘Jan Ulrich was about number seven or eight/ or eight’
Language choice errors, or rather, speech production errors of the type illustrated in
(1) – (4) occur in both directions – German connectors in Hebrew utterances and vice
versa – especially during the age period 2:3 – 4:6, but occasionally also later. Of par-
ticular interest is the history of the adversative conjunction, Hebrew avál, German
aber, at an earlier phase during this period. Both language forms of the conjunction (as
well as the English form) had been acquired and used regularly in the individual lan-
guages before the age of 2:6. At 2:6, Ben spent a three-week holiday with his mother in
Germany. Upon his return, and for the next three months, German aber consistently
replaced the Hebrew adversative conjunction in Hebrew discourse. It appears as
though the two languages underwent a fusion of the structure expressing contrast be-
tween propositional units in discourse.2 The demarcation line between the repertoire
components collapsed around the particular processing operation of contrast, allow-
ing aber to function independently of context or setting and so independently of ‘lan-
guage’ selection. This situation prevailed until the age of 2:10, when Ben left for a
three-week holiday with his father to Israel. Within a week of interacting in the mono-
lingual Hebrew environment, Hebrew avál was reinstated in Hebrew discourse. Then,
upon Ben’s return home, and for the next 2 – 3 weeks, avál replaced aber in German
discourse. Thus, a repetition of the process of fusion took place, with the languages in
reverse roles.
 Yaron Matras

How can we make sense of the turbulent fate of the contrastive conjunction during
this period? We might first establish that, while an entire set of connectivity markers is
prone to instability around the transitions between contexts (= between sets of extral-
inguistic factors that condition the selection of particular elements within the linguis-
tic repertoire), contrast is affected in a special way, and there is greater difficulty in
keeping apart the repertoire components, and less flexibility to select the appropriate
form and so to accommodate to the new extralinguistic setting. The behaviour of con-
trast is similar to, but is more extreme than that of the other connectivity markers.
Next, we can establish that the child’s difficulty in maintaining separation between
repertoire components is not due to the overall dominance of any one set of linguistic
structures (= one ‘language’). Rather, ‘dominance’ or loyalty to a particular set of struc-
tures is variable and fluid. It is, in a way, this changing loyalty which triggers the confu-
sion in the first place. The child undertakes an effort to accommodate to the constraints
of the new setting; consequently, a language becomes dominant – not within the rep-
ertoire as a whole, but pragmatically, as the preferred set within the repertoire for a
series of ongoing activity domains. It is, in other words, pragmatically dominant (see
Matras 1998).
In this section we saw that even a bilingual child with maximum domain separa-
tion between the languages and a high level of linguistic awareness and avoidance of
mixing, encounters certain difficulties in keeping apart two of his languages around
connectivity functions, especially contrast. There is a certain difficulty, in situations of
relative ambiguity which surround the transition between settings, to disassociate con-
nectivity devices from the pragmatically dominant language – the language in which
communicative performance took place until the transition. The speaker/child is thus
unable to maintain the demarcation boundary consistently. Instead, the boundary col-
lapses occasionally around a specific function or set of related functions, leading to an
instantaneous or temporary fusion – i.e. non-separation – of the structures of connec-
tivity (or certain functions thereof) in the two languages.

3. Fusion in adult speech

In this section I will demonstrate that the difficulties encountered by the bilingual
child and discussed in the previous section are not particular to children, but may just
as well affect adult or mature bilingual speakers. Consider first the case of a young, but
linguistically mature Hebrew-English bilingual, brought up partly in Israel and partly
in the United States. At the age of around eight, he is addressed, in Israel, by a mono-
lingual Hebrew speaker of a similar age group, who asks him a yes/no question:
(5) Hebrew; asked a yes-no question by Hebrew monolingual of same age group:
S: uh-húh. [ʔʌ̃ˈhʌ̃]
‘Yes’.
Contact, connectivity and language evolution 

H: ((looks puzzled)) ma?


‘What’?
The speaker’s reply is an affirmative signal which he attributes to a non-language-spe-
cific sector of his linguistic repertoire – a kind of language-independent linguistic ges-
ture. In fact, this sequence of segments is not understood in the Hebrew context/set-
ting, i.e. it is not familiar to the hearer with a monolingual Hebrew repertoire.
Virtually the same kind of communicative breakdown occurs at a somewhat later age,
here in the United States, when the speaker is confronted with a question by an English
monolingual adult:
(6) English; having been asked a yes-no question by English monolingual adult:
S: ts-. [ʇə]
‘ No’.
H: ((awaits response))
The click-related implosive sound indicating a negative reply, which is common in the
Mediterranean region, is meaningless to the English monolingual hearer, who contin-
ues to await a response. In both situations, the speaker mistook operators that help
regulate the interaction, for gestures that are language-independent. This kind of col-
lapse of the demarcation boundaries differs from the type discussed in the previous
section; here, there is no realisation that demarcation exists around a particular func-
tion. The result in both cases, however, is a fusion of both components for a particular
language processing function (or group of functions).
Consider now the following excerpts (7) – (8) from a television interview, broad-
cast in the United Kingdom in the documentary series ‘Dispatches’ on Channel 4 tel-
evision, on 8 June 1997. The speaker is German, and is being interviewed in Berlin
about his part in an arms smuggling scheme:
(7) S: Well jus/ just the way ǝ ǝ the m/ the weapons ǝ brought ǝ/ I/ I have brought
to London, nǝ, und I/ I have told them the truth, nǝ, that they were brought
by car, nǝ and/ and ǝ...
H: Were they very interested?
S: Yes, they were very interested, nǝ, to know how, nǝ.
(8) At the border in England were by the custom/ they have investigated this car,
very very ǝ ǝ thoroughly, and they have removed the panels from the doors,
the panels from the luggage room and they in/ investigated in the engine com-
partments aber they didn’t find anything but they/ they have forgotten to get
unten ǝ/ ǝ •/ ((clears throat)) they/ they forgot to look under the • car.
The first excerpt, (7), is characterised by the frequent insertion of the (northern) Ger-
man tag nǝ. Its consistent appearance suggests that it is not attributed by the speaker to
any single component of his repertoire, but has again the status of a sector-independ-
ent structure within the repertoire, a kind of universal communicative gesture, appli-
 Yaron Matras

cable with no setting-related constraints. In reality, of course, nǝ is completely foreign


to English speakers and fails to achieve its communicative function when directed at a
monolingual English addressee. In addition to this absence of realisation of a demarca-
tion boundary, in both (7) and (8) the speaker fails to maintain the boundary around
altogether three forms/structures: Two of them are connectors – und ‘and’ in (7), and
aber ‘but’ in (8).3 It is not clear that in either of the cases the speaker actually noticed
his error, as no obvious self-repair is inserted. Still, one can safely assume that the
speaker is aware of the German origin of the two words, and aware of their English
counterparts. The German insertions convey the impression as though chaining de-
vices in the narrative are being carried out in German, while the actual propositional
part of the discourse – though possibly planned in German4 is on the whole commu-
nicated successfully in English.
The conclusion is that there is a set of connectivity devices which at times escapes
the speaker’s control with respect to conscious language choice, and that another sub-
set of connectivity items – those that are less ‘lexical’, and appear more ‘gesture-like’ –
are in fact processed by the speaker as ‘untagged’ for any particular language, i.e. as
part of the overall linguistic repertoire, rather than of a particular component of that
repertoire, constrained to a specific set of communicative domains.
More evidence for lapses in adult bilingual’s control of the language selection
mechanism around connectors follows. In (9), a Czech academic whose foreign work-
ing language is generally English gives a brief survey of her research activities at a Ger-
man-speaking academic forum:
(9) ...in Norwegen/ über diese zwei Sprachen, Bokmål and Nynorsk, und so weiter.
‘...in Norway/ about these two languages, Bokmål and Nynorsk, and so on.’
In (10), a native speaker of Polish who has been living in Germany for the past decade
or so meets friends from Germany in London, where she has been attending an ad-
vanced English language course and living with an English family during the past two
weeks. She crosses the street from the Café where the party is sitting to inspect a res-
taurant, which is said to be decorated in Polish style. Having returned and confirmed
that the style is on the whole indeed Polish, she says:
(10) ...bis auf/ bis auf die Tischdecken, because/ eh weil sie...
‘...except/ except for the tablecloth, because/ uh because it...’
The error was noticed by the speaker, and a self-repair is inserted. Note that in both (9)
and (10), the target of the error – the prevailing form – does not come from a language
that is generally dominant in the respective speaker’s life, but rather from one that has
pragmatic dominance in the particular communicative domain: Thus for the Czech
academic in (9), English is the usual language of international academic events. For
the Polish-German student in (10), English has been the focus of her communicative
efforts in London over the past couple of weeks.
Contact, connectivity and language evolution 

Consider finally two more examples. In (11), a group of four Israelis – the speaker,
her husband, and two friends – are having lunch at a Chinese restaurant in Manches-
ter, England. As they are engaged in conversation in Hebrew, a waiter appears to take
their order. The speaker’s husband takes the active role in ordering most of the items,
and in (11) the speaker intervenes and adds:
(11) …and one Won Ton soup avál/ eh/ the vegetarian one.
The Hebrew contrastive conjunction avál appears here, rather than English but; it is
possible, though not obvious, that the error was noticed by the speaker, for there is a
pause and hesitation, but no explicit repair. Finally, in (12), the Saudi Ambassador to
the United Kingdom gives an interview on BBC television (Newsnight, 7 June 2004),
commenting on plans for municipal elections in Saudi Arabia. In response to a ques-
tion challenging him about the authorities’ ability to ensure fair elections, he says:
(12) I would beg to say that yaʕni/ the Kingdom is a very big territory.
The Arabic discourse marker yaʕni in fact derives from an Arabic inflected verb mean-
ing ‘it means’. Nonetheless, as a discourse marker it appears to be assigned to those
gesture-like items of speech that are not consciously assigned to a particular compo-
nent of the linguistic repertoire; yaʕni appears frequently in foreign-language dis-
course of Arabs. For a trained diplomat speaking on national television about matters
of state, it is certain that the insertion of yaʕni constitutes a non-volitional, potentially
embarrassing clash with his own and his audience’s expectations in respect of the well-
formedness of his discourse. We can safely assume that the insertion is not intentional
nor strategic in any way, but that it represents a lapse in the speaker’s control over the
mechanism that governs selection over repertoire components.
This brings us to the principal point of this section, and that is, that in all instanc-
es documented here, the failure to maintain demarcation boundaries between reper-
toire components surrounding connectivity devices is not in any way discourse-strate-
gic: It is not pre-planned, it is not conscious, in fact it is often self-repaired, and it does
not yield any overt communicative advantages for the speaker, either in terms of pres-
tige or recognition on the part of the hearer, or in terms of efficiency of communica-
tion and the ability to express a function or relation that would otherwise have to re-
main unexpressed (this is already excluded through the potential breakdown in
communication or at least the lack of any guarantee that the hearer will be able to de-
code the foreign structure successfully). We can therefore rule out two factors that are
often regarded as principal triggers for language mixing, namely prestige-related or
emblematic mixing, and functional gap-filling. There remains therefore what must be
regarded as a language-processing related or cognitive trigger, a non-volitional and
sub-conscious mechanism which, in its quasi-pathological character remotely resem-
bles some symptoms of bilingual aphasia: the inability, around specific language-
processing functions (mental operations), to maintain a demarcation boundary be-
tween components (sets) within the linguistic repertoire, and to select structures in
 Yaron Matras

accordance with the context- or domain-particular constraints that govern their nor-
mal use in the speech community.

4. Variation in connectivity marking in bilingual communities

It has been argued that in bilingual communities speakers will use discourse markers
to flag their linguistic abilities (Poplack 1980), or that they will make use of bilingual-
ism to mark out boundaries between utterances (Maschler 1994), or that frequent use
of discourse markers will flag cultural integration (Clyne 2003).5 In the previous sec-
tion we saw examples of bilingual speech production in which cognitive difficulties of
language processing and language production override social-setting related and con-
versational constraints. I suggest on this basis that bilingual speakers, while generally
able to maintain demarcation boundaries within their complex linguistic repertoires,
encounter inconvenience when it comes to connectivity devices. The removal of de-
marcation boundaries is a solution of convenience, and hence at least covertly advan-
tageous to the speaker. This ‘covert’ advantage of course competes with the overt com-
municative disadvantages of losing face when failing to select correctly (i.e. according
to the social norms of the community) among the repertoire components, or of risking
a breakdown in the efficiency of communication.
But let us assume a situation where a) there is no loss of face, since most speakers
are bilingual, hence slips of the kind documented in section 3. are commonplace; and
b) there is no breakdown of communication, since all members of the speech com-
munity understand the pragmatically dominant language into which the slips are like-
ly to occur. In such cases, the inconvenience of maintaining demarcation boundaries
around utterance-organising operations might not be counteracted by factors such as
prestige or efficiency of communication. Where can we expect to encounter such situ-
ations? They are, in fact, common in minority language settings where several condi-
tions are met: 1) The minority language is an ethnic language, whose primary function
is to flag loyalty to a clan or ethnic group; it is mainly used for group-internal com-
munication; 2) It is primarily an oral language, and enjoys minimal institutional sup-
port (i.e. little or no support in schools, media, written tradition or other institutions,
and no norm dictating language purism); 3) Bilingualism is widespread among all
adult speakers of the minority language, and interaction with persons outside the
community, and so use of the majority, dominant language, is frequent.
Although members of the community might well value the maintenance of their
own language, the regular appearance of connectors that are identical to those of the
second, external or dominant language might not be regarded as endangering the co-
herence of the ethnic (minority) language. Indeed, we saw above that in the case of
some connectivity structures, especially those that are interaction-related and hearer-
directed and have a remote relation to the content of adjacent propositions, the struc-
tures involved are sometimes regarded as non-component-specific, i.e. as part of the
Contact, connectivity and language evolution 

overall repertoire of linguistic structures, but untagged for a specific component that is
limited to particular communicative domains or settings. The motivation to remove
the inconvenience of having to maintain strict demarcation boundaries around utter-
ance-organising operations, might then override the wish to preserve the internal, aes-
thetic coherence of a separate ethnic language, or might indeed not be seen at all as
contradicting the latter.
Consider the example of Romani, a typical oral minority language lacking the
protection of any institutions of literacy or other kind of language regulation, in a per-
manent situation of bilingualism, its key functions being to serve as family- or clan-
language, flagging ethnic separateness. Examples (13) – (14) stem from a corpus of
recordings of the Lovari dialect, spoken here by a second-generation immigrant from
Poland to Germany:
(13) Laki familija sas also kesave sar te phenav, artisturi, nǝ?
her family were part [German] such how comp I-say artists part
[German]
Her family were like such how shall I say, showpeople, right?
(14) Taj žasas ande veša taj rodasas, taj dikhasas, khelasas
and we-went in woods and we-searched and we-saw played
ame halt, nǝ.
we part [German] part [German]
‘And we used to go into the woods and search, and look around, we like used
to play, right.’
The speaker belongs to the first generation in her family to grow up with both lan-
guages, Romani and German (previous generations spoke Romani and Polish, Slovak,
and Romanian). Note that the German discourse particles also, halt, nǝ are treated as
an integral part of the Romani discourse. They are not regarded as foreign, and they do
not disturb the subjective (from the speaker’s point of view) well-formedness of the
Romani discourse. Needless to say, they do not cause either loss of face or barriers to
comprehensibility within the extended family, which is where Romani is mainly used.
On the other hand, there is nothing to suggest that acceptance of German culture, or
flagging of German-language competence, are motivating triggers behind the inser-
tion of German particles. Bilingualism is an accepted reality in Romani communities,
and at least among members of the second generation of immigrants it is a self-evident
norm which does not entail any privileged position. Cultural adaptation is often func-
tional, but a culture of contrast is cherished as a key to the preservation of a diasporic
identity. There is therefore, in the Romani-speaking context, no obvious advantage of
flagging any form of cultural or emotional accommodation to the outside, non-Rom-
ani environment; if anything, community members who do appear strongly integrated
into the dominant, surrounding culture are looked down upon as semi-assimilated by
other members of the community. What we witness in (13) – (14), therefore, is the
 Yaron Matras

emerging acceptance of regular insertions of utterance-organising devices from the


contact-language into the home-language discourse.
In (13 – (14) we witness the acceptability of those connectors that operate exclu-
sively at the interactional level, like fillers and tags. They convey processing instructions
in respect of the roles of the participants and the sequencing arrangement of speech
activities, and have a special function as an invitation to the hearer to accept the speak-
er’s role and point of view. These connectors tend to be less overtly lexical or even gram-
matical, and we might conclude that it is for this reason that they are more volatile to
the mixing of boundaries between repertoire components during conversation, and so
also more prone to propagation and ultimately acceptance within the speech commu-
nity. There is indeed evidence in support of such an hypothesis. Thus, the speaker cited
in (13) – (14) uses German sentence particles, but the Polish contrastive conjunction
ale ‘but’ (recall that the speaker’s family immigrated to Germany from Poland).
However, those connectors that make reference to logical relations among propo-
sitions that are part of a propositional chain, are also susceptible when it comes to
crossing demarcation boundaries and gaining acceptability. Consider examples
(15) – (16) from conversations with two native speakers of Low German originating
from the province of Schleswig-Holstein in northern Germany, who immigrated to
the United States as teenagers in the mid-1950s:6
(15) IF: (a) und dor arbeidet ik den för hunnertfofti Dåler
and there worked I then for hundredfifty dollar
de Monat
the month
DH: (b) jåå, dat weer al wat anners.
yes that was already something else
IF: (c) dat weer wat anners, und he weer uk en Düütsche
that was something else and he was also a German
een, ni, and äh ik weer de drüdde, drüdde or
one no and uh I was the third third or
feerte Düütsche, wat för em arbeiden dee.
fourth German who for him work did
IF: (a) ‘And there I then worked for one hundred and fifty dollars a month
DH: (b) Yes that was already different.
IF: (c) That was different, and he was also a German, right, and uh I was the
third, third or fourth German who worked for him.
(16) dat weer’n Ünnericht för süstein Stunnen, but ik hef bloos
that was a lesson for sixteen hours but I have only
acht Stunnen måkt, åber dor hef ik uk nix leert.
eight hours made but there have I also nothing learned
‘That was a sixteen-hour class, but I only did eight, but I also didn’t learn any-
thing there.’
Contact, connectivity and language evolution 

In (15) we find both Low German und and English and (as well as or), and in (16) both
Low German åber and English but. Although some effort is made to select connectors
in line with the overall choice of linguistic set (lexical, grammatical), i.e. in line with
the selected ‘language’ of the discourse, this effort is not consistent, and the infiltration
of English connectors does not seem to be regarded by the speaker as a clash with the
norms and expectations on well-formedness of communication structures. Thus, in
both the Romani examples and the Low German ones, speakers are licensing them-
selves to lift the demarcation boundaries within their repertoire around a specific class
of grammatical operations.
That the adoption of such a licence is gradual, makes common sense, and is also
supported by evidence from speech communities where original (native, or inherited)
devices continue to be used alongside those markers that are borrowed from the sur-
rounding dominant language. Mosetén is spoken by approximately 800 people in the
foothill-region of the Bolivian Andes. It has been in contact with Spanish since the
17th century, but Spanish influence has been particularly strong since the 19th centu-
ry. The following examples (from Sakel, in press) illustrate the acceptability of Spanish
connectors in Mosetén discourse:
(17) Wënjö’ khö’ï mömö’ jishyiti’ y
move f only.f comb and [Spanish]
me’me’ shiphki’ raej dyaba.
so leave all peanut
‘She must have come, [and] she just combed herself and so the peanuts came
out [of her hair].’
(18) Chhibin o tsiis ji’jaem’te penne chapatikdyetyi’.
three or [Spanish] four good raft big.raft
‘Three or four rafts to make a big raft.’
(19) Me’ jïmë mö’ pero mö’ majjo’ me’.
so close 3f.sg but [Spanish] 3f.sg much so
‘This (water-source) is closer, but that one has more (water).’
(20) Tyiñetyi’ pero-ki pen’-ki jai’bai.
semi.red but [Spanish]-co side-co white
‘It (the peanut) is semi-red, but one side is white.’
While examples (17) – (19) show plain integration of the Spanish connectors y ‘and’, o
‘or’ and pero ‘but’ into Mosetén discourse, in (20) the Spanish contrastive connector
pero is combined with a Mosetén marker of contrast. The two devices are, typologi-
cally and apparently also functionally, not incompatible with one another, so that the
lifting of boundaries here may be conceived of as making full use of the entire reper-
toire, with no constraint on domain-specific selection.
Anders (1993) documents the speech of ethnic Germans in Russia. Despite the
preservation of some literary tradition in German, and a closed and rather isolated
 Yaron Matras

community structure, Russian discourse particles are infiltrating German-language


discourse and gaining acceptability; this can be seen by examining their frequency, as
well as by paying attention to the fact that such elements are rarely accompanied by
any self-repairs:
(21) T: Der Mann war krank und wir konnten doch nicht.
the man was ill and we could part not
P: Ja.
yes
T: Nu und • • wie er gestorben ist, und dann • • • hat es nicht
part [Russian] and as he died is and then has it not
so lang gedauert.
so long lasted
T: ‘The man was ill and so we couldn’t.
P: Yes.
T: Well and • • when he died, and then • • • it took so long.’
Other Russian connectors, such as no ‘but’, appear in the corpus alongside their Ger-
man equivalents (in this case aber):
(22) Ma: Sie wollen jetzt noch Brot holen?
you want now still bread fetch
No ich hab etwas zu Hause!
but [Russian] I have some at home
Ba: A, ich han Brot. Aber auf morge.
oh [Russian] I have bread but for tomorrow
Ma: Do you still want to get bread now? But I’ve got some at home!
Ba: Oh, I’ve got bread. But for tomorrow.
We have established that a) connectivity devices are prone to lapses in control over the
choice of repertoire component in early bilingualism, and to speech production errors
involving control over the choice of repertoire component in adult bilingualism, and
that b) connectivity devices are subject to variation in communities with stable bilin-
gualism, with markers from the surrounding majority or dominant language gradu-
ally gaining acceptance as part of the inventory of linguistic structures of the in-group,
ethnic, immigrant or minority language.

5. Connectivity structures and grammatical borrowing

My claim in this paper (see also Matras 1998) is that variation of the kind seen in the
previous set of examples is an outcome of the gradual acceptance and conventionalisa-
tion of spontaneous and unintentional lapses in bilingual speakers’ overall effort
Contact, connectivity and language evolution 

to maintain demarcation among their repertoire components. We shall return to a dis-


cussion of the reasons behind the pressure triggering such lapses specifically around
connectivity devices in the discussion below. Once variation sets in, it motivates change:
Where too different forms compete for the same function, speakers are likely to regard
one of them as more fashionable than the other. Change is thus likely to affect the in-
ventory of connectivity devices; one might say, therefore, that the connectivity appara-
tus is particularly prone to grammatical borrowing in language contact situations where
the recipient language is the weaker or rather the group-internal language.
Turoyo (Christian Neo-Aramaic of Tur Abdin, southeastern Turkey; Jastrow 1992)
for example, a minority language in a multilingual region, shares its connective and
discourse particles amma ‘but’, ya ‘or’, faqat ‘however’, and ya … ya ‘whether/either …
or’, ḥetta ‘even’ and balki ‘maybe’ with the surrounding languages Kurmanji (Kurdish),
Turkish, and Arabic, the particles û ‘and’ with Kurmanji and Arabic, the connectors
ǝnkān ‘if ’ and lašān ‘in order to’ with Arabic, and the particle disa ‘again’, hēš ‘still’, edi
‘then, so’, šxwa ‘anyway’, tǝ ‘at all’, veğa ‘now’, and žnu ‘only then’ with Kurmanji. Such
wholesale adoption of the inventory of connective devices (including coordinating and
some subordinating conjunctions, interjections, discourse particles, focus particles,
and phasal particles) is common in other languages under similar sociolinguistic con-
ditions (stable and widespread multilingualism, primarily oral tradition, restriction of
the language to in-group or family communication). Domari, for instance, the Indic
language of the Dom peripatetics of the Middle East, employs the full inventory of Ara-
bic conjunctions and discourse particles, including all coordinating and subordinating
conjunctions and co-particle (co-relatives), relative particles, phasal and focus parti-
cles, interjections, fillers, and tags (items in italics are grammatical loans from Arabic):
(23) law ēr-om xužoti kān laher-d-om-s-a
if [Arabic] come.past-1sg yesterday cond [Arabic] see-past-1sg-3sg-ant
‘If I had come yesterday I would have seen him’.
(24) warik-ar-a mlāy-ēk minšān mā džan-ad-is
wear-3sg-ant veil-pred.m so.that [Arabic] neg [Arabic] know-3pl.subj-3sg
‘She used to wear a veil so that one would not recognise her’
(25) qabel-mā dža-m xałłaṣk-ed-om kam-as
before-comp [Arabic] go-1sg.subj finish-past-1sg work-obl
‘Before I left I finished my work’
(26) iza wars-ari, n-aw-am-eʔ
if [Arabic] rain-3sg neg-come-1sg-neg
‘If it rains, I shall not come’
(27) na kil-d-om bara liʔann-hā wars-ari
neg go.out-past-1sg out because-3sg.f [Arabic] rain-3sg
‘I did not go out because it is raining’
 Yaron Matras

(28) ū daʔiman/ yaʕnī/ kunt ama


and [Arabic] always [Arabic] that.is [Arabic] was.1sg [Arabic] I
kury-a-m-ēk wala kil-šami wala aw-ami.
house-obl-loc-pred.f and.not [Arabic] exit-1sg and.not [Arabic] come-1sg
‘And I was always/ I mean/ at home, not going out nor coming’
(29) mana illi to-r-im iyyā-h
bread rel [Arabic] gave-2sg-1sg res-3sg [Arabic]
‘the bread that you gave me [it]’
As a result of this wholesale incorporation of Arabic connectors, along with the syntac-
tic convergence of Domari with Arabic and its replication of Arabic word order and
clause structure features, sentence organisation and clause combining is identical in the
two languages. For Domari speakers, the in-group component of their linguistic reper-
toire, or what makes up the ‘Domari language’, therefore does not include a specific in-
ventory of connectors or connectivity structures, but is limited to the internal organisa-
tion of clauses or even just constituents. As far as speakers are concerned, speaking
‘internally’ (speaking Domari) and speaking ‘externally’ (Arabic) are identical when it
comes to connectors and connectivity devices. From a diachronic viewpoint, Domari
has undergone complete fusion with Arabic in the domain of connectivity.
Romani, too, tends to borrow connectivity markers from the various surrounding
languages with which Romani dialects have come into contact. Most prone to early bor-
rowing are, as we saw above, discourse particles, fillers, and tags. Coordinating conjunc-
tions form an interesting hierarchy of borrowing in Romani. All dialects of Romani
borrow the conjunction ‘but’ from a current or recent contact language (e.g. Slavic no, po
and ali/ale, Hungarian de, Turkish ama, German aber). A pre-European expression for
‘or’, vaj, is retained in some dialects. Elsewhere, ‘or’ is often found to be a more stable
borrowing than ‘but’: For instance, the Ajia Varvara Romani dialect of Athens has ja ‘or’
from its Recent L2 Turkish, but ala ‘but’ from its Current L2 Greek; Helsinki Romani has
elle ‘or’ from its Recent L2 Swedish, but mut ‘but’ from its Current L2 Finnish. Pre-Euro-
pean ta(j) ‘and’ is retained in many dialects, though often alongside a borrowed conjunc-
tion for ‘and’. The additive conjunction too may be a more conservative, earlier loan than
the contrastive or alternative conjunction; thus Bugurdži Romani hem/em from the Re-
cent L2 Turkish, alongside pre-European thaj, but ili ‘or’ and po/ali ‘but’ from the Cur-
rent L2, Serbian. There is a clear implicational hierarchy for the borrowing of coordinat-
ing conjunctions, based on contrast (cf. Matras 1998): If ‘and’ is borrowed, then ‘or’ is
also borrowed, and if ‘or’ is borrowed, then ‘but’ is also borrowed.7
Romani is not unique in this regard. A sample of languages under the cultural
influence sphere (directly, or via another language) of Arabic (Matras 1998) show a
similar hierarchy, with Arabic-derived contrastive expressions such as lakin or amma
most widespread (cf. Urdu, Punjabi, Persian, Turkish, Lezgian, Albanian, Somali, Hau-
sa, Ful), followed by Arabic-derived disjunctive expressions (ya), followed by additive
conjunctions from Arabic (such as ve/w). In a sample of 29 Mesoamerican languages
Contact, connectivity and language evolution 

discussed by Stolz & Stolz (1996), 21 languages borrow Spanish pero ‘but’, 16 borrow o
‘or’, and 12 borrow y ‘and’. The results are by and large implicational: If a language in
the sample borrows y, it also borrows pero. There are no exceptions to this rule. If a
language borrows y, with only two exceptions it also borrows o. If a language borrows
o, it is also very likely to borrow pero, there being only three exceptions. We return to
the significance of this hierarchy in the Discussion below.
The Macedonian dialect of Turkish has borrowed not only individual connectors
from its contact language, Macedonian, but also the design of clause combining, which
relies on free-standing, clause initial conjunctions introducing finite clauses (whereas
in Turkish clause combining is often achieved through converbal structures, with con-
verbs marked as suffixes to verbs):
(30) i şimdi onların atları kaldi “Yürük”.
and [Macedonian] now 3pl.gen name.pl.poss stay.past Yürük
And so they kept the name “Yürük”
(31) a gemide para yok
whereas [Macedonian] boat.loc money none
whereas on the boats there’s no money
In (30) – (31), we find the Slavic (Macedonian) clause-initial conjunctions i ‘and’ and
a ‘and, however; whereas’, and so also a contrast between two distinct connectivity
types – addition, and contrastive-addition – which do not exist in (Standard) Turkish.
The organisation of clause combining structures in Macedonian Turkish replicates the
Macedonian blueprint even in those cases where no actual formal material is bor-
rowed. Thus in (30), the first embedded clause (‘what I want to say’) is finite, and is
introduced by a conjunction ne, deriving from the interrogative ‘what’8; this conjunc-
tion also serves as a relativiser. The second, modal clause (‘I want to say’) is also finite,
showing the typical Balkan infinitive-loss and its replacement through a finite (in this
case, subjunctive) verb describing the target action (cf. Matras 2003/2004):
(32) bilirsin ne istiyom deyim
know.aor.2sg what want.prog.1sg say.subj.1sg
‘You know what I want to say.’
Note that Macedonian Turkish remains, in contrast to the other languages of the Bal-
kans, an agglutinating, postpositional, and predominantly SOV language, and so typo-
logically quite distinct. In terms of syntactic typology, the structure of clause combin-
ing might be regarded as the first (and so far principal) domain in which convergence
with the neighbouring languages takes place. Connectivity structures, then, appear
particularly vulnerable not just to the actual borrowing for formal material such as
conjunctions and other particles, but also to the replication of patterns of constituent
ordering, agreement and overall form-function mapping that form the mental blue-
print for the respective construction. We find evidence for this tendency toward con-
 Yaron Matras

vergence of clause combining structures in other linguistic areas, too. Consider the
structure of complementation in a number of languages in the northern Levant / east-
ern Anatolian region – Kurmanji (Iranian), Neo-Aramaic (Semitic), Arabic (Semitic),
and Domari (Indo-Aryan).
(33) Kurmanji:
ez di-xwaz-im her-im mal-ê
I prog-want-1sg go.subj-1sg home-obl
(34) Neo-Aramaic (Zakho, Iraq):
ana g-ib-ǝn āz-in l-bēsa
I prog-want-1sg go.subj-1sg to-home
(35) Arabic:
ana biddī a-rūḥ ʕa-l-bēt
I want.1sg 1sg-go.subj to-def-home
(36) Domari:
ama biddī dža-m kury-ata
I want.1sg go.subj-1sg home-dat
‘I want to go home’
Note that the languages differ in basic word order rules, Kurmanji showing predomi-
nantly SOV, the others showing mainly SVO. As is the case in Macedonian Turkish,
the structure of connectivity appears more prone to convergence than other aspects of
syntactic typology.
Based on a selection of case studies, we can arrive at the following tentative con-
clusion about the position of connectivity devices in language contact situations:
a. Connective particles are prone to direct borrowing (i.e. replication of form), par-
ticularly from a dominant language (whereby the prestige of the dominant lan-
guage is the key to long-term acceptability of the forms, but not necessarily the
trigger for their initial insertion at the utterance level). ‘Smaller’ or ‘weaker’ lan-
guages will tend toward fusion of the inventory of connective particles with that of
the dominant language. The susceptibility of connective particles to borrowing
tends to be hierarchical, with those expressing contrast as well as those with min-
imal lexical content (i.e. gesture-like) among the most likely to be affected.
b. Connectivity structures that operate at the level of the entire organisation of the
complex clause are prone to replication of that organisation pattern. Such conver-
gence or pattern replication occurs under the influence of a dominant language,
but also in linguistic areas (where relations of prestige and dominance may not be
obvious). What is replicated here is primarily the rules of form-function mapping,
the ordering of elements and their grammatical meaning. In other words, we are
dealing here with the replication or fusion of constructions (in the sense of Con-
struction Grammar; cf. Goldberg 1995, Croft 2001).
Contact, connectivity and language evolution 

6. Discussion: The triggers of fusion around connectivity devices

Why is connectivity prone to fusion (of the inventory of operators and their functions)
and convergence (of the inventory of constructions) in language contact situations? If
our hypothesis is correct, and fusion begins as speech production errors where speak-
ers select the correct function, but fail to observe the constraints on the selection of the
appropriate form (‘language’), then we must assume that there is a property of the
mental processing operation that is associated with connectivity, and which interferes
with the mechanism that controls the selection of items within the overall repertoire of
forms and constructions in accordance with context- and situation-bound constraints
(i.e. the demarcation mechanism). We must therefore search for the answer to the
question of the susceptibility of connectivity devices to borrowing in the function of
connectivity in processing language in discourse.
Approaches to discourse agree that connectivity does not just concern the relation
between the content of one clause or phrase and that of another that is adjacent to it.
Rather, it concerns the complex communicative interrelations between the speaker
and the hearer during the sequencing of units of discourse (cf. already Schiffrin 1987;
see also Rehbein 1979, Redder 1989, Ehlich 1986). The speaker has to guide the hearer
into processing relations between utterances, or between the propositional compo-
nents of utterances. With some combinations of propositions, the processing track is
more problematic than with others, i.e. some connections are harder to accept against
the set of existing presuppositions. This is where the speaker has to ‘work harder’ in
order to sustain the supportive participation of the hearer, and the hearer’s acceptance
of the speaker’s authority as speaker and point of view. Similarly, the speaker has to
‘work harder’ when his/her own authority is at stake for other reasons, for instance for
failing to maintain the flow of the turn, or for failing to provide convincing clarifica-
tion, on in instances where the speaker may simply wish to ascertain the support of the
hearer (circumstances that are often highlighted by means of fillers, hesitation mark-
ers, or tags).
The inventory of structures and forms involved in such communicative proce-
dures constitute an apparatus through which the speaker can monitor hearer-sided
participation in the discourse, guide the hearer and intervene with hearer-sided
processing of the discourse – a kind of ‘monitoring-and-directing’ apparatus. It con-
tains the more obvious functions of connectors –inviting the hearer to accept a propo-
sition as a supplementary continuation of a previous proposition (addition), urging
the hearer to revise a possible course of processing and accept a broken causal chain
(cf. Rudolph 1995) proposed by the speaker (contrast), or inviting the hearer to accept
the speaker’s explanation of a state-of-affairs and its relevance (causation/ result).
They may also include the more strictly interaction-oriented functions of connectors
– inviting the hearer to continue to anticipate the conclusion of the speaker’s turn de-
spite an interruption in the speaker’s flow of speech or clarity of expression (fillers), or
inviting the hearer to signal reassurance of having accepted the speaker’s subjective
 Yaron Matras

chaining of propositions or portrayal of an overall point of view (tags). The monitor-


ing-and-directing apparatus involves complex processing in that it depends on the
speaker’s anticipation of the hearer’s mental processing of and reactions toward the
discourse and communicative behaviour; thus, it calls on the speaker to make predic-
tions about the interlocutor’s activities, and to organise actions of speech accordingly.
In this light, the borrowing hierarchy of connectors discussed in the previous sec-
tion has more than just coincidental value. At the top of the cline contrast>disjun-
ction>addition we find, in discourse-pragmatic terms, maximal intervention on the
part of the speaker in the hearer-sided processing of propositions, in effect blocking a
pathway for processing the proposition that is available (based on presupposition),
and limiting interpretation to a particular direction. With disjunction, the option of
diverting a course of interpretation is introduced by the speaker, who offers the hearer
an alternative pathway. Addition may be considered the most supportive of all three
positions on the cline, in terms of the firm and clear direction of processing which the
speaker offers to the hearer, and its harmonious relationship to contextual presupposi-
tions. Borrowing thus appears to mirror the intensity of tension between anticipated,
presupposition-based hearer-sided processing of propositional content, and the direc-
tion that the speaker’s following speech activity takes.
If we trace borrowing of connectors back to a lapse in exercising control over the
selection mechanism through which bilingual speakers maintain context-bound de-
marcation among repertoire components, then we come across a link between interac-
tional tension of this kind, and failure at a given instant to maintain control over the
demarcation boundary. Compromising demarcation boundaries is thus, we hypothe-
sise, triggered by the mental processing effect of intense monitoring-and-directing by
the speaker of hearer-sided participation in the discourse. This explains why alongside
connectors and connectivity structures, other operators whose function it is to proc-
ess, explicitly, hearer-sided expectations, and so to highlight explicitly the relevance of
propositional contents (in the sense proposed by Blakemore 2002; cf. also Sperber &
Wilson 1986), such as focus particles (even, too) or phasal particles (still, no longer), are
similarly prone to bilingual speech production errors (i.e. ‘wrong’ choices) as well as to
long-term borrowing in contact situations.
It appears, then, that the monitoring-and-directing apparatus is tightly associated
with the forms that trigger the individual operations (but for contrast, you know for
tags/fillers, and so on), which reduces the ‘reflection time’ that the speaker has at his/
her disposal to contemplate the selection of an appropriate element, and which makes
this selection more automaticised or gesture-like, compared with other linguistic se-
lections. At the same time, the procedures involved in monitoring-and-directing ap-
pear more abstract and so more loosely associated with any particular subset within
the repertoire of linguistic forms and constructions, and so more prone to evading
control over the context-bound demarcation among repertoire components.
In practice, what this means is that the bilingual speaker, when initiating a monitor-
ing-and-directing procedure, is prone to select the structure that encodes this procedure
Contact, connectivity and language evolution 

from the pragmatically dominant language (i.e the language to which strongest mental
attention has recently been directed), rather than from the selected language of the ongo-
ing interaction. If the structure is a word-form, that word-form is likely to be inserted. If
the structure is a construction, that is a pattern of form-function mapping, then this pat-
tern will be replicated drawing on elements of the selected language of the interaction.

7. Epilogue: Connectivity and language evolution

If our interpretation of the data discussed in the sections above is correct, then we can
conclude that the position of connectivity devices at the top of the cline of structural bor-
rowing reflects the particular position of connectivity in the grammatical apparatus.
What, then, is that particular position, and what can we learn from it about the ancient
architecture of the language faculty? The evolution of language is probably the most
speculative domain in current linguistic discussions (cf. Wray 2002, Tallerman 2005, for
an overview). This final section does not pretend to offer any methodological break-
through in the study of language evolution. Rather, it relies on an interpretation of the
empirical evidence discussed in the previous sections, and joins the pool of speculations
on the significance of empirical evidence to theories of the emergence of language.
The monitoring-and-directing apparatus, we argued, is more difficult to control as
part of the demarcation of repertoire components (= ‘separation of languages’ in the
bilingual repertoire). This positions the monitoring-and-directing apparatus in a differ-
ent category from other types of linguistic structures and constructions, not just func-
tionally, but from the point of view of the nature of its retrievability or mental activa-
tion. Indeed, its potential detachability from a closed subset of the repertoire places it
in the vicinity of other means of communication, such as gestures, that are tightly as-
sociated with a particular interactional function, but only more loosely associated with
just a particular set of contexts or settings (a particular ‘system’ or ‘language’). Which
other elements of language might be regarded as belonging to this class of gesture-like
functions, and is there any evidence linking them with connectivity devices?
There is, in fact, some experimental evidence based on neuroimaging of the brain,
indicating that the processing of clause conjoining is linked to greater right-hemi-
spheric activity (Friederici 2001). There appears to be rather clear evidence linking the
processing of prosody to greater right-hemispheric activity, compared to the process-
ing of melodic features of language (Schirmer et al. 2001; Friederici 2001; cf. also Mc-
Mahon 2005). There is also cross-linguistic evidence that discourse particles, includ-
ing clause-initial coordinating conjunctions, tend to be produced correctly by patients
suffering from agrammatism (Menn & Obler 1990: 1370), which again supports the
impression of some kind of neurofunctional separation of discourse-organisational
operators from other function words. Brain lateralisation could therefore be a factor
promoting susceptibility to bilingual production errors in some linguistic procedures
– those procedures that are more difficult to control in an analytical fashion. Not sur-
 Yaron Matras

prisingly, this involves functions that show greater dependency on non-analytical,


right-hemispheric processing.
We might assume that the source of the neurofunctional distinctness of connec-
tors and discourse markers is related to their function as a kind of verbal gesture ex-
pressed by the speaker to accompany the organisation of speech at the discourse level,
and especially to direct the attention and participation of the hearer. These are ele-
ments with which the speaker may be said to reach out to the hearer. McMahon (2005)
argues for a primary function of prosody in an evolutionary sense, one that emerged
prior to the emergence of language as an analytical faculty. It is noteworthy that pros-
ody, like connectivity, is particularly susceptible to contact-induced change, and to
convergence in linguistic areas. Arguably, contact-susceptibility in these domains is
linked to instinctive (rather than analytical), gesture-like communicative activities. If
there is indeed a right-hemispheric orientation for connectors and discourse markers,
as there is for prosody, one would be tempted to assume that monitoring-and-direct-
ing functions evolved prior to the emergence of language as a more analytic (and left-
hemisphere-oriented) faculty.
The assumption that monitoring-and-directing operations emerged early in the
sequence of evolution of communicative functions is not incompatible with Tomasel-
lo’s (1999) suggestion that language evolved is response to the emergence in humans of
the capacity to ‘identify’ with fellow humans as intentional agents. It is possible that,
much like prosody, the gesture-like nature of grammatical elements such as connec-
tors and discourse markers is a residue of the earliest mental ‘tools’ used by human to
capture an interlocutor’s attention and to try and influence the course of an interlocu-
tor’s intentions and actions. The interactional aspects of connectivity might therefore
belong to the more primitive, more ‘instinct’-driven and less analytical components of
human communication; as such, they are more prone to escaping the analytical con-
trol of the bilingual who struggles to maintain what is, effectively, a socially construct-
ed demarcation boundary within a repertoire of communicative tools.

Notes

1. Grosjean’s (2001) speaks of ‘language modes’, which involve a cline of different degrees of
activation of each language within the bilingual’s repertoire.
2. See Matras (1998) for a discussion of fusion in this sense. Cf. Meisel (1990), who characte-
rises the lack of grammatical separation between languages in Bilingual First Language Acquisi-
tion as fusion.
3. The third – unten ‘below’ – reveals that at least part of the utterance was planned in Ger-
man, and the local expression which serves as a crucial reference point escapes the simultaneous
transposition of the original mental plan of the proposition into English. This slip does not es-
cape the speaker’s attention, however, and he pauses to save face, then repairs.
Contact, connectivity and language evolution 

4. See previous footnote, and note German word order patterns in … were by the custom, the
use of the perfect tense, and the lexical composition luggage room < German Kofferraum.
5. But see also Salmons (1990) for a different approach to bilingual discourse markers, one
that emphasises the convergence of communicative strategies.
6. Examples provided by Dörte Hansen-Jaax, 1995.
7. Aikhenvald (2002: 182) mistakenly cites this hierarchy in reverse order: “In the hierarchy
of borrowing suggested by Matras (1998: 303), the coordinating conjunction ‘and’ is considered
more likely to be borrowed than ‘or’”, and goes on to say that Tariana is a counterexample, since
the coordinating conjunction is not borrowed, neither is the Portuguese contrastive conjunction
mas borrowed, although Portuguese o ‘or’ does appear.
8. In Standard Turkish a paratactic construction resembling this structure is possible, but the
default embedded clause would be nominal: bil-ir-sin de-mek iste-diğ-im-i know-aor-2sg say-
inf want-past.part-1sg-acc.

References

Aikhenvald, A.Y. 2002. Language Contact in Amazonia. Oxford: OUP.


Anders, K. 1993. Einflüsse der russischen Sprache bei deutschsprachigen Aussiedlern [Arbeiten
zur Mehrsprachigkeit 44]. Hamburg: Germanisches Seminar.
Auer, P. (ed.). 1999 Codeswitching in Conversation. Language, interaction and identity. London:
Routledge.
Blakemore, D. 2002. Relevance and Linguistic Meaning. The semantics and pragmatics of dis-
course markers. Cambridge: CUP.
Clyne, M. 2003. Dynamics of Language Contact. Cambridge: CUP.
Croft, W. 2000. Explaining Language Change: An evolutionary approach. Harlow: Longman.
Croft, W. 2001. Radical Construction Grammar. Syntactic theory in typological perspective. Ox-
ford: OUP.
Ehlich, Konrad. 1986. Interjektionen. Tübingen: Niemeyer.
Friederici, A. 2001. ������������������������������������������������������������������������
Syntactic, prosodic, and semantic processes in the brain: Evidence from
event-related neuroimaging. Journal of Psycholinguistic Research 30: 237–250.
Goldberg, A. E. 1995. Constructions: A construction grammar approach to argument structure.
Chicago IL: Chicago University Press.
Grosjean, F. 2001. The bilingual’s language modes. In One mind, Two languages. Bilingual langu-
age processing, J.L. Nicol (ed.), 1–22. Oxford: Blackwell.
Hansen-Jaax, D. 1995. Transfer bei Diglossie. Synchrone Sprachkontaktphänomene im Nieder-
deutschen. Hamburg: Dr. Kovac.
Jastrow, O. 1992. Lehrbuch der Turoyo -Sprache. Wiesbaden: Harrassowitz.
Lanza, E. 1997. Language Mixing in Infant Bilingualism. A sociolinguistic perspective. Oxford:
OUP.
Maschler, Y. 1994. Metalanguaging and discourse markers in bilingual conversation. Language
in Society 23: 325–366.
Matras, Y. 1998. Utterance modifiers and universals of grammatical borrowing. Linguistics 36:
281–331.
 Yaron Matras

Matras, Y. 2003/2004. Layers of convergent syntax in Macedonian Turkish. Mediterranean Lan-


��������������������������������������������������������������
guage Review 15: 63–86.
McMahon, A. 2005. Heads I win, tails you lose. In Headhood Elements, Specification and Con-
trastivity, P. Carr, J. Durand and C. Eden (eds), 255–275. Amsterdam: John Benjamins.
Meisel, Jürgen M. 1990. ���������������������������������������������������������������������
Code switching and related phenomena in young bilingual children. In
Papers for the Workshop on Concepts, Methodology and Data [European Science Founda-
tion Network on Code Switching and Language Contact], 143–168.
Menn, L. and Obler, L. K. 1990. Cross-linguistic data and theories of agrammatism. In Agram-
matic Aphasia: A cross-language narrative sourcebook, L. Menn and L. K. Obler (eds),
1369–1389. Amsterdam: John Benjamins.
Ochs, E. 1988. Culture and Language Development: Language acquisition and language socialisa-
tion in a Samoan village. Cambridge: CUP.
Poplack, S. 1980. Sometimes I’ll start a sentence in Spanish y termino en español: Toward a ty-
pology of code-switching. Linguistics 18: 581–618.
Redder, A. 1989. Konjunktionen, Partikeln und Modalverben als Sequenzierungsmittel im Un-
In Dialoganalyse II, E. Weigand and F. Hundsnurscher (eds), 393–407.
terrichtsdiskurs. ���
Tübingen: Niemeyer.
Rehbein, J. 1979. Sprechhandlungsaugmente. Zur Organisation der Hörersteuerung. In Die Par-
tikeln der deutschen Sprache, H. Weydt (ed), 58–74. Berlin:
�������������������
De Gruyter.
Rudolph. E. 1995. Contrast. Adversative and concessive expressions on sentence and text level.
Berlin: Walter de Gruyter.
Sakel, J. In press. The borrowing of Spanish discourse markers in Mosetén. International Journal
of Bilingualism.
Salmons, J. 1990. Bilingual discourse marking: Code switching, borrowing, and convergence in
some German-American dialects. Linguistics 28: 453–480.
Schiffrin, D. 1987. Discourse Markers. Cambridge: CUP.
Schirmer, A., Alter, K., Kotz, S. and Friederici, A. 2001. Lateralization of prosody during lan-
guage production: A lesion study. Brain and Language 76: 1–17.
Sperber, D. and Wilson, D. 1986. Relevance. Communication and cognition. Oxford: Blackwell.
Stolz, C. and Stolz, T. 1996. Funktionswortentlehnung in Mesoamerika. Spanisch-amerindischer
Sprachkontakt [Hispanoindiana II]. Sprachtypologie und Universalienforschung 49: 86–123.
Tallerman, M. (ed.) 2005. Language Origins. Perspectives on evolution. Oxford: OUP.
Tomasello, M. 1999. The Cultural Origins of Human Cognition. Cambridge MA: Harvard Uni-
versity Press.
Volterra, V. and Taeschner, T. 1978. The acquisition and development of language by a bilingual
child. Journal of Child Language 5: 311–326.
Wray, A. (ed.) 2002. The Transition to Language. Oxford: OUP.
ALLORA
On the recurrence of function-word borrowing in
contact situations with Italian as donor language

Thomas Stolz
University of Bremen

The article addresses the issue of function-word borrowing from Italian into a
number of languages spoken in Italy or neighbouring regions. Among the most
frequently borrowed items, the discourse particle allora stands out as the most
widespread Italian loan. The article explores the properties which make allora
a prime candidate for being borrowed. The borrowability of allora is compared
to two competing relative chronologies of function-word borrowing whose
empirical basis does not contain evidence from language-contact situations with
Italian. It is argued that the success story of allora calls for a partial revision of
those models which ascribe particular importance to adversative conjunctions
and discourse particles (translating English but and however) because the
ubiquitous Italian discourse particle is a serious competitor for the role of prime
mover in the chronology of function-word borrowing.

1. From Spanish to Italian

Some (though by far not all) function words are known for being particularly prone to
being borrowed in language contact situations.1 Especially discourse markers / parti-
cles lend themselves easily to transfer from one language to the other. On Thomason’s
(2001: 70) borrowing scale, they occur as early as stage 2 which is described as a phase
of “slightly more intense contact (borrowers must be reasonably fluent bilinguals, but
they are probably a minority among borrowing-language speakers)”. Unsurprisingly,
discourse markers / particles also count among the most prominent elements in codes-
witching because they most often occupy positions at the periphery of clauses, sen-
tences and paragraphs – positions which are frequently involved in codeswitching
(Myers-Scotton 1993). Recent research has revealed that discourse markers / particles
of one and the same prestigious superstrate or adstrate language are indiscriminately
borrowed by recipient languages with very diverse genetic, areal and typological back-
 Thomas Stolz

grounds. Matras (1998: 301–305) for instance has studied the almost ubiquitous bor-
rowing of adversative and coordinating conjunctions corresponding to English but /
however and and / or in numerous contact constellations. He provides an intriguing
cognitive explanation for the prominence of not only adversative conjunctions in the
inventories of borrowed function words across languages:
[T]he trigger behind language mixing around discourse-regulating grammatical
elements is cognitive, not social, in the sense that it derives from the mental-
processing functions associated with the linguistic expression, or in plain terms
from its communicative-interactional function. The ‘donor’ language is the one
that is pragmatically dominant in the particular instance in which transfer occurs,
that is, it is a linguistic system to which speakers show a special situative commit-
ment and to which their efforts at norm-conforming linguistic behavior are cur-
rently directed. Typically, such a donor system belongs to the language in which
group-external communication takes place, and where the speakers under con-
sideration, being outsiders, do not participate in defining the norm. Conversely,
their own group may be tolerant toward changing norms and may accept such
cognitively triggered transfers. If such a constellation persists, recurring trans-
fers may be conventionalized. Prestige is thus not itself a motivation for the trans-
fer of discourse markers […], but at most a background precondition for the
wholesale, long-term borrowing of this class of items [inverted commas original]
(Matras 1998: 396).

These rather appealing ideas, however, have not been left unchallenged. In her chapter
on Portuguese influence on the Amazonian language Tariana, Aikhenvald (2002: 179–
86) observes that her evidence runs counter to the patterns familiar from other contact
constellations where discourse markers are prime candidates for being borrowed. Tar-
iana does not seem to have borrowed any discourse markers from Portuguese (al-
though it has borrowed the preposition até ‘until’ [a somewhat doubtful case though],
the conjunction ou ‘or’ and the negator nem ‘nor’). Aikhenvald (2002: 182) emphasises
that the notorious adversative conjunction has not been borrowed into Tariana. More-
over, she claims that the borrowing of the disjunctive ou ‘or’ without prior borrowing
of the coordinating e ‘and’ disproves the validity of Matras’s borrowing hierarchy.
However, this latter claim is caused by a misinterpretation of the implication and ⊃ or
⊃ but.2 Matras (1998: 303 and elsewhere) clearly states that but is the unmarked case
whereas the borrowed or implies borrowed but and borrowed and implies borrowed
or and but. Thus, only the absence of Portuguese mas ‘but’ in Tariana is a challenge to
Matras’ model. In the light of such potential counter-evidence it is advisable to extend
our empirical basis by reviewing more contact constellations with different donors and
recipients and, other sets of function words.
With a view to testing Matras’s hypothesis, it is thus necessary to look at function
words other than adversative conjunctions and functionally closely related items from
a comparative perspective. These additional function words have to be checked in de-
Allora: On the recurrence of function-word borrowing in contact situations with Italian 

tail for their borrowability in order to determine whether or not there is something
special about say, adversativity, which facilitates the borrowing of its expressions pro-
vided by the prestigious (or in Matras’s terms “pragmatically prominent”) language. If
a certain class of markers behaves differently from other function words fulfilling dis-
course-organizing tasks when it comes to being borrowed, then this evidence has to be
checked against Matras’s model and the hierarchies proposed therein. If however there
is no evidence for a special status of one class of markers in comparison to other dis-
course particles then a revision of the model is called for. This will also be the case if it
turns out that certain languages – be they recipient or donor languages – yield diver-
gent results in language contact situations. Thus, in this contribution, I will give special
attention to a different set of discourse particles whose functional domain does not
prototypically revolve around adversativity (see below). I will also look at languages
which hitherto have not been prominent in the literature on our topic. In order to
guarantee comparability with Matras (1998), I will largely refrain from discussing so-
cial factors, whose potential effects in function-word borrowing will be focused upon
in a follow-up study.
The paradigm case of parallel borrowing behaviour is the Hispanicisation of (many
of) the indigeneous languages presently spoken in those territories which once be-
longed to the erstwhile Spanish colonial empire or which now form part of the offi-
cially Spanish-speaking successor states of the former colonies in Latin America (Zim-
mermann 1987; Th. Stolz 1996, 1997, 1998, 2002; Ch. Stolz 1998; Stolz and Stolz 1995,
1996a, 1996b, 1997, 2001). Owing to the common exposure to Spanish as their prestig-
ious donor language, languages of different genetic affiliation and typological classes on
three continents have become similar in spite of their original structural diversity. The
features they now share are the ones which they have borrowed independently from
Spanish: more often than not, languages in Austronesia (Philippines, Marianas, Easter
Island), the Americas (from Mexico to Tierra de Fuego) and – though probably to a
lesser extent – Africa (Northern Maroc, West Sahara and Equatorial Guinea) display
almost identical and relatively extended sets of Spanish-derived function words with
discourse particles ranking highest. The widespread use of Spanish entonces ‘then,
therefore, thus’ in Autronesian and Amerindian languages exemplifies this tendency. In
(1)-(4) I present sentences from Philippinian, Polynesian, Tupi-Guaranian and To-
tonac-Tepehuan languages which have integrated entonces into their systems without
ever having been in direct contact with each other.3 In these sentences, the usage to
which the borrowed versions of entonces are put in the recipient languages largely con-
forms to the patterns provided by the donor language Spanish, although functional and
distributional equivalence is not necessarily strict, as recipient languages may employ
borrowed items in a way which is at odds with the rules of the donor language. Note too
that direct Spanish influence on Hiligaynon ceased more than a century ago whereas it
is very strong (and ever increasing) in the three remaining cases.4
 Thomas Stolz

(1) Hiligaynon [Austronesian, Philippinian – Philippines] (Wolfenden 1971: 79–80)5


Madamo ang bulak dira’ intonsis manguha kita
many Det flower at_there therefore Fut:get 1Pl
‘There are lots of flowers there, so let’s get some.’
(2) Rapanui [Austronesian, Polynesian – Easter Island] (Makihara 2001: 198)
entonces ka u’i mo hakapiri o tātou
entonces Imp look Subord unite Poss 1Pl.Incl
‘Therefore, let’s look to unite ourselves.’
(3) Guaraní [Amerindian, Guaraní-Tupi – Paraguay] (Gregores and Suárez 1967: 188)
e<n>tónse ko’éro ‘ó ko’è=muaéro še po-visitá ta
entonces tomorrow or day_after_tomorrow I you-visit Fut
‘Then, tomorrow or the day after, I will visit you.’
(4) Totonac [Amerindian, Totonac-Tepehuan – Mexico] (Levy 1990: 42)
entonces tuku li:-cha:’lhka:tnan-a chi?
entonces what Ins-work-2Sg.Incomp now
‘What then are you working on now?’
Chances are that we would encounter similar phenomena not only in Rifeño or other
varieties of Berber exposed to Spanish influence in Northern Maroc, but also in heav-
ily Hispanicized substandard / regional varieties of Basque where the normative influ-
ence of the unified standard Batua is still rather restricted. Given that the recipient
languages, or their immediate predecessors, made use of autochthonous means fulfill-
ing functions similar to those of the borrowed Spanish items during pre-contact times
(and, in many cases, continue to do so even today), it makes no sense to attribute the
astonishing homology among the recipient languages to incidentally shared structural
“deficiencies” (since, of course, there are no deficiencies in natural languages). Like-
wise, it is impossible to identify any “positive” structural feature common to so many
dissimilar languages which might have triggered the borrowing process. Is there then
anything about the Spanish discourse markers / particles which makes them irresisti-
ble when languages come into contact with Spanish?
In order to answer this deliberately naive question, it is necessary to turn our at-
tention to a different donor language. If the discourse markers / particles of other do-
nor languages prove less successful in language contact situations, the idea becomes
more tenable that there is something about the Spanish elements (or the socio-com-
municative conditions in the Spanish sphere of influence) which makes them espe-
cially attractive for being borrowed. However, if it turns out that (certain classes of)
discourse markers / particles rank high on the borrowability hierarchy independent of
the donor language, then the Spanish elements are nothing extraordinary – they only
corroborate a general tendency. In this contribution, therefore, I look at the impact
Italian and its regional varieties have been exerting on co-territorial minority languag-
es and the somewhat outlier heavily Italianized Maltese language spoken on the Mal-
Allora: On the recurrence of function-word borrowing in contact situations with Italian 

tese archipelago just south of Sicily. I haven chosen Italian as the test case because of
the following reasons:
(i) Like the seemingly problematic Portuguese (according to Aikhenvald 2002, see
above), Spanish and Italian are both Romance languages and are thus easy to compare.

(ii) The periods of Hispanicization and Italianization are almost co-extensive as


both have been in existence for at least half a millennium in some parts of the ter-
ritories under scrutiny. In addition, there is also an almost identical division into
sub-periods of different contact intensity: for the first centuries, both Spanish and
Italian were only sporadically present in the non-Romance speech-communities
– a situation which changed drastically in the 19th century when official language
policy started to aim at an enforced cultural and linguistic homogenization.

(iii) The Italian case is different from the Spanish one as it does not reflect a typi-
cal colonial situation but a socially assymmetric adstrate constellation within the
zone of cultural and / or demographic dominance of the prestigious group.

(iv) Except Maltese, the languages influenced by Italian are all of Indo-European
stock and thus distantly related to and structurally not too dissimilar from Italian
and its dialects. Thus, the Italian situation differs considerably from the ones de-
scribed in Matras (1998), Stolz and Stolz (1995, 1996a, 1996b, 1997, 2001) and
Aikhenvald (2002) whose focus is on genetically diverse contact constellations.
On the basis of a genetically less heterogeneous sample, one might expect differ-
ent behaviour on the part of the recipient languages. If this hypothesis is falsified
however, claims with a language-independent scope can be formulated.

(v) As with Spanish, the contact situations both actually involve bilingual ones
(with Italian or a local variety thereof being one of the languages mastered by the
speakers) and historical ones (for which Italian has almost completely disappeared
from the inventory of actively used languages of the population).

The languages to which I mostly refer in the remainder of my contribution are in al-
phabetic order: (Italo)-Albanian (also called Arbëresh), Cimbrian (Germanic) a heav-
ily Italianised variety of Upper German (Bavarian), (Italo-)Greek (also called Griko),
Maltese and Molise Slavic, a South Slavic variety.6 (Italo)-Albanian, Cimbrian,
(Italo-)Greek and Molise Slavic are minority languages spoken in relatively small and
isolated village varieties scattered about various Italian regions, namely the Veneto and
Trentino for Cimbrian, Molise, Calabria, Sicily and Puglia for (Italo)-Albanian, Calab-
ria and Puglia for (Italo-)Greek, and Molise for Molise Slavic. Owing to the better ac-
cessability of reliable data for Calabria, my discussion of (Italo-)Albanian and
(Italo-)Greek focusses on their Calabrian varieties. (Italo)-Albanian has the largest
speech-community with about 88,000 speakers whereas Cimbrian, (Italo-)Greek and
Molise Slavic together count no more than 20,000 speakers. Maltese is of course the
official language (co-official with English) of the Republic of Malta with slightly less
than 400,000 resident native speakers. Maltese boasts of a normative grammar and a
 Thomas Stolz

relatively rich literary tradition dating back to the late 18th century. It has been expe-
riencing a constant increase in functions and domains ever since the 1930s when its
status was made official (to the detriment of the erstwhile dominant Italian). With
Malta’s admission to the EU, Maltese has also become one of the official languages of
this supra-national organization whereas the above minority languages of Italy are
more or less endangered notwithstanding recent legal provisions aiming at their pres-
ervation. Literacy in these languages is limited. Practically all speakers of the minority
languages also know Italian and / or a regional Romance variety. On Malta, active
knowledge of Italian is restricted although a sizeable number of Maltese are still able to
understand Italian fairly well (which continues to be present on the islands because of
mass tourism, TV and radio). In sociolinguistic terms, the speech-communities of the
minority languages in Italy can be compared to the ones of many indigeneous lan-
guages in modern Latin America where Spanish is omnipresent in daily life outside the
local residence and the autochthonous Amerindian languages are confined to home,
family and perhaps the village. Maltese on the other hand compares with those lan-
guages which, as a result of the vicissitudes of political history, were able to emancipate
themselves from the foreign dominance, for instance Tagalog in the Philippines (where
English ousted Spanish as it did with Italian on Malta).

2. Italian

The Italian equivalent of Spanish entonces is allora. Allora is multifunctional in the sense
that it may be employed as temporal adverb meaning ‘then, at that time’ (Renzi, Salvi and
Cardinaletti 1995: 284–286), as a paragraph-connecting discourse particle with a range
of consecutive-causal-temporal readings corresponding to English then / thus / therefore
and as a marker of turn-taking (Renzi, Salvi and Cardinaletti 1995: 245). Furthermore, it
is also used with a certain adversative flair especially in emphatic questions (comparable
to English but / still / yet), as a marker of the apodosis in emphatic (often counterfactual)
conditional clauses (Renzi, Salvi and Cardinaletti 1991: 781–784) and as a marker of
emphatic imperatives. Some of these functions are secondary in the sense that allora
fulfills them only if it occurs in combination with other more dedicated elements. In (5),
I give examples for most of the above functions of allora. The majority of the examples is
taken from the Italian translation of Antoine de Saint-Exupéry’s Le petit prince – a source
that will be important in subsequent sections too.
(5) Italian [Indo-European, Romance – Italy]
(5.1) Temporal adverb (LPP Italian XIV.34)
E dopo di allora è cambiata la consegna?
and after of allora be.3Sg change:Part:F Det:F task:F
‘And has the task been changed since then?’
Allora: On the recurrence of function-word borrowing in contact situations with Italian 

(5.2) Temporal discourse particle (LPP Italian XIV.11)


Allora veniva il turno dei lampionai
allora come:Past:3Sg Det turn of:Det.M.Pl lamp_lighter:Pl
della Russia e delle Indie
of:Det.F Russia and of:Det.F.Pl Indies
‘Then it was the turn of the lamp-lighters of Russia and the Indies.’
(5.3) Temporal-consecutive discourse particle (LPP Italian I.35)
E allora non parlavo di boa
and allora Neg speak:Past:1Sg of boa
‘And thus I did not talk about boas.’
(5.4) Causal discourse particle (LPP Italian II.35)
Non avevo mai disegnato una pecora
Neg have:Past:1Sg ever paint:Part:M Indef:F goat
e allora feci per lui uno di quei due disegni
and allora make:Perf:1Sg for him one:M of this:M.Pl two painting:Pl
che avevo fatto tante volte
which have:Past:1Sg make:Part:M so_many:F.Pl time:F.Pl
‘I had never painted a goat and therefore I produced for him one of the two
paintings I had made so often.’
(5.5) Emphatic imperative (LPP Italian IX.51)
Hai deciso di partire e allora vattene!
have:2Sg decide:Part.M of leave:Inf and allora go:you:Loc
‘You have decided to leave and thus go away!’
(5.6) Adversative support in questions (LPP Italian VII.8)
Ma allora le spine a che cosa servono?
but allora Det.F.Pl thorn:F.Pl to what thing serve:3Pl
‘But then what purpose do the thorns serve?’
(5.7) Turn-Taking (LPP Italian X.68–69)
King: …i miei ordini sono ragionevoli.
Det.M.Pl my:M:Pl order:Pl be.3Pl reasonable:M.Pl
Prince: E allora il mio tramonto?
and allora Det.M my:M sundown
‘King: My orders are based on reason. Prince: And what about my sun-
down?’
(5.8) Conditional (Renzi / Salvi / Cardinaletti 1991: 781)
Se fossi un marziano
If be:Conj.Past:1Sg Indef Martian
allora avrei le orecchie verdi.
allora have:Cond:1Sg Det.F.Pl ear:F.Pl green:Pl
‘If I were a Martian, then I would have green ears.’
 Thomas Stolz

Owing to this wide variety of functions, it comes as no surprise that allora counts
among the most frequent words of modern Italian. In the Italian frequency dictionary
(Bortolini, Tagliavini and Zampolli 1971: 232), allora occupies rank 78. Unfortunately,
the various functions of allora are not distinguished in the frequency count. Of the 42
occurrences of allora in the Italian version of Le petit prince, 21 (= 50%) are of a causal-
consecutive nature, 13 (= 30%) can be classified as temporal whereas the other “sec-
ondary” functions are somewhat underrepresented.7 In four cases (= 9.5%), allora
adds adversative emphasis to questions. It is also used twice in conditional clauses,
once in turn-taking and once for the purpose of rendering an imperative more em-
phatic. I reckon that this distribution of allora over functions is not too far remote
from the one a more extended corpus analysis would yield. The fact that it is often
rather difficult to determine whether an instance of allora counts as temporal, con-
secutive or causal will be important in the subsequent sections. At this point, it suffices
to state that allora tends to invite several readings at a time according to the principle
of post hoc ergo propter hoc.
Allora has a number of partial synonyms which cover different sections of the
functional domain of allora. All of these partial synonyms are multifunctional them-
selves. For reasons of space, I restrict my further comments to three of the many par-
tial synonyms which are particularly interesting for the present purpose. This does not
mean that other partial synonyms such as quindi ‘thus’, perciò ‘therefore’, ora ‘now, thus,
then’ etc. have nothing on offer for a contact-linguistically inspired investigation.
Dunque ‘thus, therefore’ is responsible for the causal-consecutive part but may also be
used for the purpose of reinforcement in emphatic questions (Cusatelli 1979: 580), poi
‘then’ is a temporal adverb although it is not free of consecutive and emphatic func-
tions similar to the ones mentioned for dunque (Cusatelli 1979: 1285) and dopo
‘after(wards), behind’ is a temporal (and spatial) adverb, preposition and discourse
particle (Cusatelli 1979: 571). In (6), each of these partial synonyms is exemplified in
their prototypical function.
(6) Italian [Indo-European, Romance – Italy]
(6.1) poi (LPP Italian II.46)
Lo guardó attentamente e poi disse
it.Obj watch:Perf.3Sg attentively and poi say.Perf.3Sg
‘He watched it carefully and then said:…’
(6.2) dopo (LPP Italian XXIV.32)
e dopo un silenzio disse ancora
and dopo Indef silence say.Perf.3Sg again
‘And after a moment of silence he said again:…’
(6.3) dunque (LPP Italian III.23)
Tu vieni dunque da un altro pianeta?
you come:2Sg dunque from Indef other:M planet
‘Thus you come from another planet?’
Allora: On the recurrence of function-word borrowing in contact situations with Italian 

As to their frequency, two of the partial synonyms are more frequent than allora itself,
namely poi, which occupies rank 54 in the frequency dictionary, and dopo which ranks
74 (Bortolini, Tagliavini and Zampolli 1971: 232). Both are thus high frequency items.
Dunque however is much more seldomly used and thus winds up on rank 325 (Borto-
lini, Tagliavini and Zampolli 1971: 237). Surprisingly, all three partial synonyms fail to
come close to the values reported for allora in the above sample text: poi accounts for
25 cases followed by dopo with 14 and dunque with just seven attestations. Only when
added up they can compete with allora in terms of statistics. In what follows, I investi-
gate the fate of allora in language contact situations with Italian as the donor language.
For this purpose, I compare allora to its supposed competitors in the above mentioned
five recipient languages.

3. Contacts

3.1 Quantities

Given the general validity of my above corpus-based layman’s statistics, one expects a
similar statistical disproportion for the discourse-regulating items when it comes to
borrowing.8 My working hypothesis assumes that if any of the elements discussed in
section 2 is borrowed into the languages which have been in contact with Italian, al-
lora is the most likely candidate. Moreover, if more than one of the items are borrowed
then allora will not only form part of the set of borrowed items but it will also be the
most frequently used. And thirdly, the use to which allora is put in the recipient lan-
guages will largely reflect the uneven distribution of this item over its various func-
tions. I check these predictions on the basis of a relatively heterogeneous corpus. The
corpus is made up of:
anthologies of folklore including recorded native speaker interviews
Gradilone (1970) for (Italo-)Albanian spoken in Calabria; 268 pages of origi-
nal texts;
Bellotto et al. (1978: 38–333) for Cimbrian; all prayers, proverbs, sayings and
songs have been excluded; where there are two versions of the same story, I
have opted for the one recorded in Lusern; 150 pages of original texts;
Rossi Taibbi and Caracausi (1994) for (Italo-)Greek spoken in Calabria; 218
pages original texts (discounting prayers, proverbs, sayings and songs);
Breu and Piccoli (2000: 421–44) for Molise Slavic, just 12 pages of original texts;
All text editions are bilingual and provide faithful Italian translations which allow di-
rect comparison. All sources comprise texts produced by a variety of native speakers of
the recipient languages. For (Italo-)Albanian, Cimbrian and (Italo-)Greek, this prac-
tice also implies that different village varieties are represented in the collections. The
texts were recorded in the decades after the 2nd World War (the [Italo-]Greek ones al-
 Thomas Stolz

ready in the 1950s, some of the ones documenting Molise Slavic as late as the 1990s)
and thus reflect contemporary usage.
Short written texts for religious purposes: these stem exclusively from the (Italo-)
Greek dialect spoken in Bova (Rossi Taibbi and Caracausi 1994: 424–486);
the translation of Le petit prince into Maltese by Tony Aquilina, 85 pages.
Thus, the corpus contains an uneven assortment of very few written texts and an over-
whelming majority of recorded spoken language. Owing to this imbalanced mixture of
registers and also owing to the difference in size, the direct statistical correlation of the
individual results is ruled out as the margin of potential errors is too wide. Neverthe-
less the absolute figures and the statistical weight of the borrowed items per language
are telling in themselves.
In table 1, I present the token frequencies of the four items mentioned in the pre-
vious section complemented randomly by the functionally unrelated intanto standard
Italian ‘(mean)while’; colloquial Italian also ‘however’ (Cusatelli 1979: 878) as an ad-
ditional control category. Intanto is number 410 in the Italian frequency dictionary
(Bertolini, Tagliavini and Zampolli 1971: 239) and is thus an item of medium fre-
quency. For easy reference, I repeat the Italian values in the first line of table 1. The
recipient languages follow from top to bottom according to how many of the dis-
course-regulating elements are attested in the corpus. Boldface marks the highest-
ranking token frequency values per language. Zero frequency is additionally high-
lighted by grey shading.
A comment is called for as to the results I have obtained for Cimbrian. In his descrip-
tive grammar of Cimbrian, Tyroller (2003: 88) states that
[o]kkasionell aus dem Italienischen übernommene Wörter, wovon vor allem
Konjunktionen wie alora [a’lo: a] dann‘ und dunque [ ] ‚also‘ betroffen
sind, werden hier nicht berücksichtigt, da sie nicht zum festen System des Zimbr-
ischen von Lusern gehören, sondern nur gelegentlich in bestimmten Situationen
gebraucht werden.9

Table 1.  Token frequencies

language allora dopo intanto dunque poi

Italian 42 14 1 7 25
(Italo-)Greek 1 130 3 26 153
(Italo-) 170 65 34 1 0
Albanian
Cimbrian 118 18 8 0 0
Molise Slavic 12 3 0 0 0
Maltese 38 0 0 0 0
Allora: On the recurrence of function-word borrowing in contact situations with Italian 

At least for allora, this statement is clearly contradicted by the statistics. Allora is much
too frequent to be just an “occasional” transfer from Italian. The absence of dunque
from my Cimbrian corpus however is not only in line with Tyroller’s observation but
also fits in with a general trend (see below).
The first part of my working hypothesis is borne out by the numerical values. Allora
is indeed attested in each and every recipient language. As a matter of fact, allora is the
only element which has been reported for all of the languages.10 Were it not for the unique
case of (Italo-)Greek, the second part of the working hypothesis would also hold because
allora is the most frequently used of the items under scrutiny in four out of five recipient
languages. Wherever dopo is attested, it is always second best – outnumbered either by
allora (with ratios ranging from 3-to-1 [= identical to the one calculated for standard
Italian] to 6.5-to-1) or by poi (with a ratio of 1.7-to-1) in (Italo-)Greek.
The position of allora in (Italo-)Albanian, Cimbrian, Molise Slavic and Maltese is
very strong whereas it is particularly weak in (Italo-)Greek where it is attested only
once. This discrepancy is of course statistically significant as the frequency count for
the (Italo-)Greek varieties is based on almost 220 pages of text. Rohlfs (1977b: 142)
surveys the (Italo-)Greek equivalents of Italian allora for all (Italo-)Greek varieties
including the ones spoken in the Salento. According to his description, all (Italo-
)Greek varieties make use of autochthonous elements instead of borrowing allora. This
marginal status of allora in (Italo-)Greek goes along with exceptionally high figures for
poi, dopo and dunque in the same varieties. Poi is the topmost item in terms of token
frequency. Moreover, (Italo-)Greek is the only recipient language for which poi is at-
tested as a borrowing in the corpus. Apart from (Italo-)Greek, dunque has made it only
into (Italo-)Albanian where I have found one isolated example. Thus, (Italo-)Greek
diverges from the rest of the recipient languages (and to some extent also standard
Italian) in the sense that it attributes more importance to items which elsewhere are
less prominent as to frequency.
An explanation that comes immediately to mind is that the differences observed
can be accounted for if, in lieu of the standard language, Italian dialects are considered
primary contact partners of the recipient languages. The high frequency of dunque in
(Italo-)Greek could perhaps reflect the fact that in the southernmost Romance varie-
ties, the cognates of Italian dunque are more frequently used to the detriment of allora.
This is a fair guess as for instance Sardinian – a distinct Romance language often as-
sociated with the southern Italian varieties in the pertinent literature – avoids allora
while it makes ample use of a cognate of dunque. In the Sardinian translation of Le
petit prince, there is no trace of allora at all whereas duncas – the Sardinian equivalent
of Italian dunque – turns up 24 times. Its functional range in Sardinian seems to cover
much of the fuzzy zone of temporal-consecutive-causal relations for which allora is
responsible in Italian (as duncas is used in many of the sentences where the Italian ver-
sion has allora). However, there seem to be some problems with this kind of explana-
tion. Dunque is especially frequent in the (Italo-)Greek dialect of Bova whereas, at
Roccaforte, dunque is a low frequency item. Moreover, the Calabrian varieties of (Ita-
 Thomas Stolz

lo-)Albanian, which are spoken to the north of the (Italo-)Greek ones, do not favour
dunque at all. This supposed riddle is easily solved however. While it is true that both
(Italo-)Albanian and (Italo-)Greek varieties in Calabria are spoken in what Italian dia-
lectology calls la zona meridionale estrema (= southernmost dialectal area), they are
not immediate neighbours. The (Italo-)Albanian varieties are located in the province
of Cosenza i.e. in the northern part of the dialectal area whereas the (Italo-)Greek va-
rieties are spoken in the distant southern tip of Calabria, in the province of Reggio. As
it happens, the extant Calabrian dialect dictionaries clearly show that the Romance
dialects spoken around Cosenza differ from the ones spoken in the vicinity of Reggio
as to the presence of cognates of allora and dunque. Rohlfs (1977a: 72 and 246) reports
that dunca (= dunque) is used throughout Calabria whereas allura (= allora) is re-
stricted to the varieties spoken near Cosenza. Accordingly, there is no entry for allura
in the dictionary of the Romance variety of Reggio compiled by Malara (1970) where-
as dunca forms part of the recorded items. In his comparative grammar of the
(Italo-)Greek varieties, Rohlfs (1977b: 146) explicitly states that for Calabria the equa-
tion “allora = […] addunca” holds. Thus, the (almost complete) absence of allora from
(Italo-)Greek has a trivial explanation: the Romance varieties with which (Italo-)Greek
has been in contact most intensively simply lack this item. In its stead, the cognates of
dunque and poi are more frequently employed both in the local Romance varieties and
their (Italo-)Greek partners in contact. The fact that dunca is common to all Calabrian
varieties of Romance may also explain why it is at least marginally attested in Calab-
rian (Italo-)Albanian. Outside the extreme southern dialectal area, dunque is likely to
occur less frequently as opposed to allora. The frequency values given for the standard
language support this idea (see above).
Independent of the diachronic developments which have to be scrutinized in a
separate study, it is possible to interpret the outcome of my frequency count in table 1
as suggestive of an implication, viz. wherever allora is favoured, dunque is much less
frequent and vice versa. We may speculate therefore that allora has remained marginal
in (Italo-)Greek even after standard Italian became more important in language con-
tact because dunque already occupies good part of the territory. The single occurrence
of allora in (Italo-)Greek most probably reflects a recent development. For those varie-
ties where allora is strong however, the reverse is true: dunque is an exception because
allora has already been there before.
Before I discuss what these facts teach us in more general terms, it is necessary to
look at a selection of examples in order to determine which functions of the borrowed
items dominate in the recipient languages.

3.2 Qualities

In the recipient languages, allora may come in different phonological shapes – most


commonly allura, but also as reduced alure, alor and lor. What strikes the eye immedi-
ately is the fact that despite the relatively high token frequency of allora in the various
Allora: On the recurrence of function-word borrowing in contact situations with Italian 

recipient languages, it normally does not have the full range of functions as in the do-
nor language. The overwhelming majority of cases belongs to the area of temporal-
consecutive-causal relations which also dominate in the donor language. Thus, the
third part of the working hypothesis successfully stands the test. (Italo-)Albanian and
Maltese display the widest ranges of functions of allora; the only function shared by all
recipient languages including (Italo-)Greek is the consecutive-causal one. Examples
for the various functions of borrowed allora are given in (7)-(12). For contrastive pur-
poses, I add fitting examples for the use of (Italo-)Greek dunca in (13).
(7) Temporal
(7.1) (Italo-)Albanian [Indo-European, Albanian – Italy] (Gradilone 1970: 52)
Allura u martuen e mbetëtin kutjènd
allora Pass marry:Aor.3Pl and remain:Imperf.3Pl happy
‘Then they married and remained happy.’
(7.2) Cimbrian [Indo-European, Germanic – Italy] (Bellotto et al. 1978: 217)
Alóra ‘s püable is gãnt huam is o’.
allora Det.Nt boy be.3Sg go:Part home it too
‘Then the boy went home too.’
(7.3) Maltese [Afro-Asiatic, Semitic – Malta] (LPP Maltese XXVI.23)
Allura jien stess baxxejt rasi…
allora I self lower:Perf:1Sg head:Por.1Sg
‘Then I lowered my head…’
(7.4) Molise Slavic [Indo-European, Slavic – Italy] (Breu and Piccoli 2000: 429)
Alor su prisegl
allora be.3Pl marry:Part
‘Then they married.’
(8)
Temporal-causal
(8.1)
(Italo-)Albanian [Indo-European, Albanian – Italy] (Gradilone 1970: 88)
Allura ajo e zeza vate t’ i rrëmbit zogun
allora that Det black:Def.F go.Aor.3Sg Subord him catch bird:Def.M:
Acc
‘Thus the black one went to catch the bird for him.’
(8.2) Cimbrian [Indo-European, Germanic – Italy] (Bellotto et al. 1978: 258)
Alóra issar aldar darschrakht
allora be.3Sg:he right_there frighten:Part
‘Thus he immediately was shocked with fear.’
(8.3) Maltese [Afro-Asiatic, Semitic – Malta] (LPP Maltese I.35)
Allura la kont noqgħod nitkellem fuq sriep boa
allora Neg be.Perf:1Sg 1Sg:stand.Imperf 1Sg:speak.Imperf on snake:Pl boa
‘Thus I spoke neither about boas…’
 Thomas Stolz

(8.4) Molise Slavic [Indo-European, Slavic – Italy] (Breu and Piccoli 2000: 432)
Alor je mu da ovi praščič
allora be.3Sg him give.Part that piglet
‘Thus he gave him this piglet.’
(9) Consecutive-causal
(9.1) (Italo-)Albanian [Indo-European, Albanian – Italy] (Gradilone 1970: 19)
Allura u poxhár ditë e natë nd’ atë finester
allora Pass put day and night near that.Acc window
e nëng tundej
and Neg move
‘Thus she stood at the window day and night and did not move.’
(9.2) Cimbrian [Indo-European, Germanic – Italy] (Bellotto et al. 1978: 310)
alóra sain-sa ghebeest naünzekh.
allora be.3Pl-they be:Part ninety
‘Well then they were ninety [wolves].’
(9.3) (Italo-)Greek [Indo-European, Greek – Italy] (Rossi Taibbi and Caracausi
1994: 163)
Allúra su kófto tin ģefali
Allora you behead:1Sg Det.Acc head
‘Thus I decapitate you!’
(9.4) Maltese [Afro-Asiatic, Semitic – Malta] (LPP Maltese III.18)
Allura inti wkoll ġej mis-sema!
allora you also come.Part from:Det-heaven
‘Then you too come from the heavens!’
(9.5) Molise Slavic [Indo-European, Slavic – Italy] (Breu and Piccoli 2000: 422)
tvoja divojka lor je ndelidžend
your:F daughter:F allora be.3Sg intelligent
‘Thus, your daughter is intelligent!’
(10) Emphatic question
(10.1) (Italo-)Albanian [Indo-European, Albanian – Italy] (Gradilone 1970: 231)
Allura çë kem’ bëmi nanì?
allora what have:1Pl do:1Pl mommy
‘What then do we have to do, mum?’
(10.2) Maltese [Afro-Asiatic, Semitic – Malta] (LPP Maltese V.10)
Mela allura jieklu l-baobabi wkoll?
then allora 3:eat.Imperf:Pl Det-baobab:Pl also
‘Well then, do they eat the baobabs too?’
Allora: On the recurrence of function-word borrowing in contact situations with Italian 

(11) Turn-Taking
(11.1) (Italo-)Albanian [Indo-European, Albanian – Italy] (Gradilone 1970: 298)
A: Ma u ngë të njoh ngë të pé maj!
but I Neg you know.1Sg Neg you see.1Sg.Aor ever
B: Allura gjegjnje gjithë njo çë thotë ki…
allora listen all know.Imp what say.3Sg this
‘A: But I do not know you, I have never seen you! B: Well, listen everybody,
that’s what this one says…’
(11.2) Maltese [cf. 5.7 above] [Afro-Asiatic, Semitic – Malta] (LPP Maltese X.68–69)
King: …l-ordnijiet tiegħi huma raġonevoli
Det-order:Pl of:1Sg they reasonable:Pl
‘My orders are based on reason.’
Prince: Allura ser tniżżilhieli x-xemx
allora Fut 2Sg:go_down:Caus:DO.3Sg.F:IO:1Sg Det-sun
li tlabtek?
that beg.Perf:1Sg:DO.2Sg
‘Well, are you going to let the sun go down like I asked you to do?’
(12) Emphatic imperative
(Italo-)Albanian [Indo-European, Albanian – Italy] (Gradilone 1970: 126)
Allura nga me mua bír’ im nga
allora come.Imp with me son my come.Imp
‘Then come with me, my son, come!’
(13) (Italo-)Greek [Indo-European, Greek – Italy]
(13.1) temporal (Rossi Taibbi and Caracausi 1994: 434)
O Ġoséppi dúnka eyái apissu to lleiδíondu
Det.M Joseph dunque go.3Sg behind Det.M.Gen.Pl brother:Gen.Pl: Por.3Sg
‘Then Joseph followed his brothers.’
(13.2) Temporal-causal (Rossi Taibbi and Caracausi 1994: 430)
Ećino dúnka eyérti
this dunque raise:Aor.Pass.3Sg
‘He thus raised himself.’
(13.3) Consecutive-causal (Rossi Taibbi and Caracausi 1994: 394)
Dúnka ipiye pánd’ ambró
Dunque go:Imperf.3Sg always forward
‘Thus he progressed continually.’
(13.4) Emphatic question (Rossi Taibbi and Caracausi 1994: 303)
Addúnka esù ise o δyávolo?
dunque you be.2Sg Det.M devil
‘Then you are the devil?’
 Thomas Stolz

(13.5) Emphatic imperative (Rossi Taibbi and Caracausi 1994: 93)


Addúnka δotému ta rúxa!
dunque give:Imp.2Pl:me Det clothes
‘Then give me my clothes!’
In (Italo-)Greek, dunca covers most of the functions which elsewhere among the re-
cipient languages form part of the domain of allora.
With the proviso that some of the languages are clearly underrepresented in my
corpus because of the limited size of the available texts, I summarize the results of my
investigation of allora in the five recipient languages in table 2. Attested functions are
labeled yes in the appropriate cells whereas empty cells – additionally shaded grey –
indicate that there is no evidence for such a function of allora in the corpus.

Table 2.  Attested functions of allora per recipient language

language consecu- temporal temporal- emphatic turn- emphatic


tive-causal causal question taking imperative

(Italo-) yes yes yes yes yes yes


Albanian
Maltese yes yes yes yes yes
Cimbrian yes yes yes
Molise Slavic yes yes yes
(Italo-)Greek yes

Unsurprisingly, what seems to be most attractive in the borrowing process is the quirky
area of temporal, consecutive and causal relations between which the boundaries are
often blurred. I admit that there might be better ways of classifying some of the above
examples. However, the reclassification would most probably take place within the
said quirky area and not go beyond it.
Two things have to be made clear in connection with the above findings. First of
all, the examples of allora in the recipient languages are by no means restricted to nar-
rative passages. A considerable number of examples stem from direct speech. The dia-
logues are very often lively or dramatic and thus invite markers of emphasis. Neverthe-
less, emphatic allora is relatively rare in my corpus. Thus, there is no additional bias to
the detriment of spoken language. Secondly, in none of the recipient languages is al-
lora without autochthonous competitors. For instance, in Cimbrian, alóra and asó
(standard New High German also ‘thus’) compete even in one and the same idiolect
and in one and the same text. In Maltese, there is Semitic mela ‘thus’ which may even
co-occur with allura in a kind of doublet construction (see [10.2] above) familiar from
the context of Hispanicization (Ch. Stolz 1998). In (Italo-)Greek too, the combination
of borrowed dunque and autochthonous arte ‘then’ is commonplace at least in the few
extant written texts. This co-existence of functionally related autochthonous elements
Allora: On the recurrence of function-word borrowing in contact situations with Italian 

supports the view that there is no structural reason for the borrowing of allora and
other function words because the recipient languages have a repertoire of adequate
means of their own. Irrespective of the difficulty to determine the exact make-up of the
class of discourse-regulating elements for the early stages of the recipient languages, it
is legitimate to assume that there were no structural gaps even at the time when con-
tact with Romance varieties first began. The modern standard varieties to which the
minority languages of Italy are directly related of course display the full array of dis-
course markers of all kinds among which functionally partial equivalents of Italian
allora are well represented.

3.3 Hierarchy

The above is suggestive of a rather strong attraction not only of allora itself but of ele-
ments which have a bundle of vaguely contoured temporal-consecutive-causal func-
tions. The ubiquity of such items in the inventories of borrowed function words of the
recipient languages of my sample makes it necessary to check whether or not these
elements are in an implicational relation to the discourse-regulating elments discussed
by Matras (1998), namely coordinating and, disjunctive or and adversative but. Table
3 surveys the presence / absence of Italian e ‘and’, o ‘or’ and ma ‘but’ / però ‘however’ in
the recipient languages. Brackets indicate that the evidence is exclusively corpus-
external. Grey shading marks the absence of a borrowing in a given language.

Table 3.  Distribution of coordinating, disjunctive and adversative elements over recipient
languages

language coordinating disjunctive adversative

e o ma però

Molise Slavic yes yes yes yes


(Italo-)Albanian yes yes yes
(Italo-)Greek (yes) yes
Cimbrian yes yes
Maltese (yes)

Before I comment on these findings, examples for the borrowings are presented in
(14)-(17). For (Italo-)Greek o ‘or’ and Maltese però ‘however’, I was unable to find any
evidence in my corpus and thus had to resort to additional sources. The (Italo-)Greek
example is taken from the dialect spoken in the province of Otranto in the southern-
most part of Puglia, the so-called Salento. For (Italo-)Albanian, I have refrained from
interpreting the frequent conjunction e ‘and’ as a loan from Italian because there is an
identical conjunction e ‘and’ in standard Albanian – one of two allomorphs of edhe
‘and’ (the other being dhe ‘and’). In Cimbrian, e occurs twice as part of the expression
 Thomas Stolz

e bèn ‘well then’ (= Italian ebbene ‘well then’) which does not allow us to count e as a
separate morpheme.
(14) Coordinating e
Molise Slavic (Breu and Piccoli 2000: 431)
E je donija doma
e be.3Sg bring:Part home
‘And he carried (it) home.’
(15) Disjunctive o
(15.1) (Italo-) Albanian (Gradilone 1970: 203)
Ngapu dí o tri dit e shoqja i thot
after two o three day Det.F woman:Def him say.3Sg
‘After two or three days, his wife said to him…’
(15.2) (Italo-) Greek (Rohlfs 1977b: 210)
krási o gála
wine o milk
‘wine or milk’
(15.3) Molise Slavic (Breu and Piccoli 2000: 424)
ti maš arsolvit si pula je
you have:2Sg decide:Inf if filly be.3Sg
do trajina o je do bešče
of coach:Gen o be.3Sg of she_donkey:Gen
‘You have to decide whether the filly belongs to the (coach-pulling) horse or
to the donkey.’
(16) Adversative ma
(16.1) (Italo-) Albanian (Gradilone 1970: 7)
Ma si vate e erdhe?
ma how go.Aor.3Sg and come.Aor.3Sg
‘But how did he go and come?’
(16.2) Cimbrian (Bellotto et al. 1978: 42)
Ma alls in an stròach ghit’s-en an schüttlar
ma all in one strike go.3Sg:it-him Indef push
‘But all of a sudden, he was hit by something that made him tremble.’
(16.3) (Italo-) Greek (Rossi Taibbi and Caracausi 1994: 303)
Ma kánnome to xartí!
but make:1Pl Det.Acc contract:Acc
‘But let us make the contract!’
Allora: On the recurrence of function-word borrowing in contact situations with Italian 

(16.4) Molise Slavic (Breu and Piccolo 2000: 436)


Ma sa jidu cukar
ma now eat:3Pl sugar
‘But now they eat sugar.’
(17) Adversative però
(17.1) (Italo-) Albanian (Gradilone 1970: 102)
però mos u harrò!
però Neg.Imp it forget.Imp
‘But don’t forget it!’
(17.2) Cimbrian (Bellotto et al. 1978: 202)
Di laüt però hãm-en alle pensaart ke…
Det people però have-they all think:Part that
‘The people however have all thought that…’
(17.3) Maltese (Aquilina 1990: 1050)
jien niġi però trid tweghedni li…
I 1Sg:come.Imperf però 2Sg:must 2Sg:promise.Imperf:me that
‘I will come but you have to promise that…’
(17.4) Molise Slavic (Breu and Piccoli 2000: 423)
però ti inim na pat
però you make:1Sg Indef pact
‘However, I make a pact with you.’
Ma is a high frequent item in all of the recipient languages whereas però occurs only
occasionally. It comes as no surprise that però is absent from (Italo-)Greek as the ex-
tant descriptions of the Romance varieties spoken in Calabria do not mention any
cognates of però at all. The mere difficulty in finding appropriate examples for some of
the borrowings in relatively sizeable collections of texts is a telling fact. This is also true
of the astonishing results obtained for Molise Slavic. The limited character of the avail-
able texts notwithstanding, all candidates for borrowing are indeed attested. This dif-
ference between Molise Slavic and the other recipient languages is indicative of an
advanced stage of Italianization of Molise Slavic.
In what way do allora-like elements correlate with the ones surveyed in table 3? The
answer to this question is straightforward: wherever allora is borrowed we also find an
element with adversative functions and vice versa. Both categories are in an equipollent
implication. Synchronically, it is next to impossible to decide whether or not there is a
chronology which still gives precedence to one of thet two competing categories. Our
present state of knowledge does not allow for any substantial diachronic speculations.
Suffice it here to mention some interesting evidence from the early period of written
Maltese. Kontzi (1999: 431) looks at the Italianisms in religious texts by the cleric Pan-
zavecchia dating back to the early 19th century. In these old texts, he noticed occur-
rences of Italian e ‘and’ and o ‘or’ – two elements which today have disappeared from
 Thomas Stolz

the Maltese lexicon. As the author of the texts examined by Kontzi had ideas of his own
about what should be part of the Maltese language (Kontzi 1999: 428–429), it is difficult
to interpret the presence of the two conjunctions in the said texts. As the two conjunc-
tions are also attested in the 18th century religious texts by Ignazio Saverio Mifsud
(Zammit Ciantar 2005: 694 and 702) it is possible that they were part of the common
Maltese lexicon of the times, and thus lend some support to Matras (1998) as and and
or occur together and thus fulfill the required implication. However, the evidence also
supports Aikhenvald (2002) because but is never expressed by an Italianism in Panza-
vecchia’s writings where the Semitic imma and iżda (both ‘but, however’) abound in-
stead, i.e., but is missing from the list of early borrowings whereas and and or are at-
tested. However, in the 18th century, some rare instances of ma però occur in the above
mentioned religious texts (Zammit Ciantar 2005: 701): a combination of two adversa-
tive conjunctions of Italian, viz. ma ‘but’ and però ‘but, however’. The situation is incon-
clusive: either but was borrowed from Italian in Maltese already in the 18th century but
remained infrequent, and thus did not figure in the texts of Panzavecchia, or the early
attestations of but are idiosyncrasies of the Maltese cleric who wrote the texts. Inde-
pendent of which of the two possibilities actually applies, one thing is clear: e / o – like
ma – only had an ephemeral existence in Maltese, and thus chances are that they never
were fully integrated elements. This in turn suggests that they were latecomers, whereas
other elements – such as the adversative però – became well-established members of the
Maltese system of which they still form part today.

4. Conclusions

In (18), I summarize the results of the above discussion in the shape of three implica-
tional scales.

(18) Implicational hierarchies

(18.1) Material borrowings

poi ⊃ dunque ⊃ intanto ⊃ dopo ⊃ allora

(18.2) Functions of allora

turn-taking temporal-

causal
emphatic ⊃ ⊃ ⊃ consecutive-

imperative causal
poi ⊃ dunque ⊃ intanto ⊃ dopo ⊃ allora

Allora: On the recurrence of function-word borrowing in contact situations with Italian 

(18.2) Functions of allora

turn-taking temporal-

causal
emphatic ⊃ ⊃ ⊃ consecutive-

imperative causal

emphatic temporal

question

(18.3) Classes of discourse-regulating elements

adversative

coordination ⊃ disjunctive ⊃
consecutive-causal

With the notable exception of (Italo-)Greek where the Italian loans o ‘or and ma ‘but’
have ousted completely the autochthonous elements (Rohlfs 1977b: 210), most of the
recipient languages add the Italianisms to their inherited inventories and thus enlarge
them. They thus have several options among more or less synonymous alternatives.
Pragmatical prominence drives the selection mechanism which determines the actual
choice the individual speaker makes in a given situation. However, it is still largely
unclear to what extent pragmatical prominence correlates with certain functions and
thus with clearly defined classes of discourse-regulating elements. While it seems plau-
sible that not anything goes, the evidence at hand suggests that there is more than one
possibility for the status of the pragmatically most prominent category. The Italian
scenario resembles the Spanish one where temporal-consecutive-causal elements (=
entonces and pues) are most widely spread together with the adversative pero ‘but’. The
disjunctive o ‘or’ is also quite common as a Hispanism in recipient languages whereas
coordinating y ‘and’ is a rare bird among the borrowings (Stolz and Stolz 1996b). In a
way, Italian and Spanish as donor languages seem to constitute a common “type” when
it comes to borrowing. Matras’s and Aikhenvald’s cases potentially represent different
types. Whether the idea that there is a typology can be substantiated is a question
which can be answered only by extending our empirical basis.
 Thomas Stolz

Future research also has to reveal whether there is a moderate kind of situation-
dependency of function-word borrowing (as the differences in borrowing behaviour
of Matras’s sample, Tariana and the Italian case seem to suggest). Clearly, the issue of
universality vs. idiosyncrasy can be tackled properly only if we have access, in addition
to more crosslinguistic evidence, to sufficient diachronic data. These data will then
help to test the validity of the implicational scales put forward in this section.

Notes

1. This article builds on ideas I have developed in Stolz (2005) and fits in with a large-scale
research project on processes of Romanicization worldwide. I am grateful to Ermenegildo Bi-
dese, Walter Breu, James R. Dow, Elvira Glaser, Miki Makihara, Tamar Khizanishvili, Steven
Roger Fischer, and Aina Urdze for their kind help with technical and bibliographical matters. A
word of thanks is also due to the two anonymous readers of the first draft of this paper and to
the editors of this volume for their kind invitation to participate in this undertaking.
2. I am especially grateful to the two anonymous readers for making me aware of the fact that
I followed Aikhenvald’s mistaken interpretation too closely in the earlier version of this paper.
3. A word of caution is in order: the term language is used in a very unspecific sense. This is
of course a simplification, as the reported cases may be typical only of one variety of a diatopical
system, whereas other related varieties do not make use of the borrowed items. It cannot be ru-
led out completely that, in some cases, idiosyncrasies of an individual speaker come to the fore-
ground and are not representative of any interpersonal variety at all. In the absence of the neces-
sary socio-linguistic clues, I have taken all examples at face value, accepting them as instances of
common communicative practice (see also Stolz 2003).
4. For a general outline of the extra-European presence and fate of Spanish, I refer the reader
to Quilis (1992).
5. In the examples, boldface marks those items which are focused upon in the discussion. The
borrowed function words are glossed with their equivalent form in the donor language. The
examples are given in the graphic representation they have in the sources on which I draw. The
abbreviations I use in the transmorphemization are spelled out in the appendix. For each exam-
ple, the object language is identified by its glossonym accompanied by genetic and geographic
information in square brackets.
6. This sample of languages is identical to the one I used for my earlier paper on related issues of
Italianization (Stolz 2005). For further socio-linguistic information on these varieties, the reader is
advised to consult Bellinello (1998). The collection of articles edited by Breu (2005) assembles
numerous studies of processes of Italianization in the co-territorial languages of Italy. The Maltese
situation and its historical background are described in some detail in Brincat (2000).
7. In some substandard varieties of Italian, the frequency of allora is probably even higher.
The translation of Le petit prince into the dialect spoken in the city of Milan contains 47 attesta-
tions of alora, the local equivalent of standard Italian allora. Since I have not yet checked spoken
language corpora and more sizeable dialect materials, these assumptions of mine must remain
conjecture.
Allora: On the recurrence of function-word borrowing in contact situations with Italian 

8. Note that I do not claim that frequency generally dictates borrowing behaviour as one of
the anynonymous readers suggests. However, if certain words such as discourse markers are
borrowed at all, it makes sense to assume that their frequency in the donor language has a say in
their borrowability. Meaning: it is more likely that the most frequent discourse marker is bor-
rowed before a (considerably) less frequent one. On the other hand, it is by no means necessary
that the most frequent items of a potential donor language are borrowed at all (especially be-
cause the top-ranking elements such as definite and indefinite articles, etc. often lack salience).
9. “Occasional transfers from Italian, especially conjunctions like alora [] ‘then’ and dunque []
‘thus’, will not be considered here as they do not belong to the established system of the Cimbrian
variety of Lusern but are used only occasionally in determined situations” [my translation].
10. As has been pointed out by one of the anonymous readers, my sources are partly based on
oral usage of speakers and thus the possibility is high that at least some of the instances of allora
are cases of code-switching and cannot be counted as proper borrowings. Apart from the fact
that it might be difficult to determine exactly where to draw the dividing line between the two
categories, the evidence is overwhelmingly indicative of borrowing as very often allora comes in
a phonological shape that presupposes adaptation and integration. Admittedly, the isolated ins-
tance of allora in (Italo-)Greek is problematic and might turn out to instantiate code-switching.
These difficulties notwithstanding, the hapax allora in (Italo-)Greek lends additional support to
the idea of different adstrate constellations on Italian soil (cf. below).

Abbreviations

Acc = accusative, Aor = aorist, Caus = causative, Cond = conditional, Conj = conjunctive, Def =
definite, Det = determiner, DO = direct object, F = feminine, Fut = future, Gen = genitive, Imp =
imperative, Imperf = imperfect(ive), Incl = inclusive, Incomp = incompletive, Indef = indefinite
article, Inf = infinitive, Ins = instrumental, IO = indirect object, Loc = locative, M = masculine, Neg
= negation, Nt = neuter, Obj = object, Part = participle, Pass = passive, Past = past tense, Perf =
perfect(ive), Pl = plural, Por = possessor, Poss = possessive, Sg = singular, Subord = subordinator,

References

(a) Translations of Le petit prince


LPP Italian: Nini Bompiani Bregoli. 1994. Il piccolo principe. Milano: Bompiani.
LPP Maltese Toni Aquilina. 2000. Iċ-ċkejken prinċep. Msida: Mireva
LPP Milanese Lorenz Banfi. 2002. El princip piscinin. Gressan: Wesak.
LPP Sardinian Andria Deplano. 1997. Su prinzipeddu. Cagliari: Artigianarte.

(b) Quoted literature


Aikhenvald, A. Y. 2002. Language Contact in Amazonia. Oxford: OUP.
Aquilina, J. 1990. Maltese-English Dictionary. Vol. II: M-Z and addenda. Malta: Midsea Books.
Bellinello, P. F. 1998. Minoranze etniche e linguistiche. Cosenza: Editoriale BIOS.
Bellotto, A. et al. 1978. I racconti di Luserna. Vicenza: Circolo Culturale M. Ghandi di Luserna,
Istituto di Cultura A. dal Pozzo di Roana.
 Thomas Stolz

Bortolini, U., Tagliavini, C. and Zampolli, A. 1971. Lessico di frequenza della lingua italiana
contemporanea. Milano: IBM Italia.
Breu, W. (ed.). 2005. L’influsso dell’italiano sulla grammatica delle lingue minoritarie (problemi
morfologici e sintattici). Rende: Università della Calabria.
Breu, W. and Piccoli, G. 2000. Dizionario croato molisano di Acquaviva Colleroce. Campobasso: s.l.
Brincat, J. 2000. Malta – elf sena ta‘ storja. Malta: PIN.
Cusatelli, G. 1979. Dizionario Garzanti della lingua italiana. Milano: Garzanti.
Field, F. W. 2002. Linguistic Borrowing in Bilingual Contexts. Amsterdam: John Benjamins.
Gradilone, G. 1970. Novellistica albanese: Racconti popolari die S. Sofia d’Epiro, S. Demetrio Cor-
one, Macchia Albanese, S. Cosmo Albanese, Vaccarizzo Albanese, S. Giorgio Albanese.
Firenze: Olschki Editore.
Gregores, E. and Suárez, J. A. 1967. A Description of Colloquial Guaraní. The Hague: Mouton.
Kontzi, R. 1999. Wort und Schrift. Des Kanonikus Fortunato Panzavecchia Bibelübersetzung ins
Maltesische, nach den Handschriften des Kathedralarchivs in Mdina. Tübingen: Narr.
Levy, P. 1990. Totonaco de Papantla, Veracruz. México: El Colegio de México.
Makihara, M. 2001. Modern Rapanui adaptation of Spanish elements. Oceanic Linguistics 40(2):
191–223.
Malara, G. 1970. Vocabolario dialettale calabro-reggino-italiano. Bologna: Forni Editore.
Matras, Y. 1998. Utterance modifiers and the universals of grammatical borrowing. Linguistics
36(2): 281–331.
Myers-Scotton, C. 1993. Duelling Languages: Grammatical structure in codeswitching. Oxford:
Clarendon Press.
Quilis, A. 1992. La lengua española en cuatro mundos. Madrid: MAPFRE.
Renzi, L., Salvi G. and Cardinaletti, A. 1991. Grande grammatica italiana di consultazione. II. I
sintagmi verbale, aggetivale, avverbiale. La subordinazione. Bologna: Il Mulino.
Renzi, L., Salvi G. and Cardinaletti, A. 1995. Grande grammatica italiana di consultazione. III.
Tipi di frase, deissi, formazione delle parole. Bologna:Il Mulino.
Rohlfs, G. 1977a. Nuovo dizionario dialettale della Calabria. Ravenna: Longo Editore.
Rohlfs, G. 1977b. Grammatica storica dei dialetti italogreci (Calabria, Salento). München: Beck.
Rossi Taibbi, G. and Caracausi G. 1994. Testi neogreci di Calabria. Palermo: Istituto Siciliano di
Studi Bizantini e Neogreci.
Stolz, C. 1998. Hispanicisation in modern Yucatec Maya: Grammatical borrowing. In Conver-
gencia e individualidad. Las lenguas mayas entre hispanización e indigenismo, A. Koechert
and T. Stolz (eds), 165–194. Hannover: Verlag für Ethnologie.
Stolz, C. and Stolz, T. 1995. Spanisch-amerindischer Sprachkontakt: Die ‘Hispanisierung’ mes-
oamerikanischer Komparationsstrukturen. Iberoamericana 58/59(2/3): 5–42.
Stolz, C. and Stolz, T. 1996a. Spanisch-amerindischer Sprachkontakt: Funktionswortentlehnung
in Mesoamerika. Sprachtypologie und Universalienforschung 49(1): 86–123.
Stolz, C. and Stolz, T. 1996b. ‘Transpazifische Entlehnungsisoglossen’. Hispanismen in Funk-
tionswortinventaren beiderseits der Datumsgrenze. In Areale, Kontakte, Dialekte. Dynamik
der Sprache in mehrsprachigen Situationen, W. Enninger et al. (eds), 262–291. Bochum:
Brockmeyer.
Stolz, C. and Stolz, T. 1997. Universelle Hispanismen? Von Manila über Lima bis Mexiko und
zurück: Muster bei der Entlehnung spanischer Funktionswörter in die indigenen Sprachen
Amerikas und Austronesiens. Orbis 39(1): 1–77.
Allora: On the recurrence of function-word borrowing in contact situations with Italian 

Stolz, C. and Stolz, T. 2001. Hispanicised comparative constructions in indigenous languages of


Austronesia and the Americas. In Lo propio y lo ajeno en las lenguas austronésicas y amer-
indias, K. Zimmermann and T. Stolz (eds.), 35–56. Frankfurt: Vervuert.
Stolz, T. 1996. Sprachkontakt, Sprachwandel und ‘Grammatikalisierung’ als Faktoren bei der
Hispanisierung von indigenen Sprachen in hispanophonen Einflußzonen. In Grammatika-
lisierung in der Romania, S. Michaelis and P. Thiele (eds), 81–106. Bochum: Brockmeyer.
Stolz, T. 1997. Grammatical Hispanisms in Amerindian and Austronesian languages. Amerindia
21: 137–160.
Stolz, T. 1998. Die Hispanität des Chamorro als sprachwissenschaftliches Problem. Iberoameri-
cana 70: 5–38.
Stolz, T. 2002. General linguistic aspects of Spanish-indigenous language contacts with special
focus on Austronesia. Bulletin of Hispanic Studies 79: 133–58.
Stolz, T. 2003. Not quite the right mixture: Chamorro and Malti as candidates for the status of
mixed language. In The Mixed Language Debate. Theoretical and empirical advances. Y.
Matras and P. Bakker (eds), 271–315. Berlin: Mouton de Gruyter.
Stolz, T. 2005. Italianisierung in den alloglotten Sprachen Italiens. In Das Zimbrische zwischen
Germanisch und Romanisch. E. Bidese, J. R. Dow and T. Stolz (eds), 43–68. Bochum: Brock-
meyer.
Thomason, S. G. 2001. Language Contact: An introduction. Washington DC: Georgetown Uni-
versity Press.
Tyroller, H. 2003. Grammatische Beschreibung des Zimbrischen von Lusern. Wiesbaden: Franz
Steiner.
Wolfenden, E. P. 1971. Hiligaynon Reference Grammar. Honolulu HI: University of Hawai’i
Press.
Zammit Ciantar, J. 2005. Il-priedki bil-Malti ta’ Ignazio Saverio Mifsud. Eddizzjoni kkumentata bi
studju kritiku. PhD dissertation, Fakultà ta’ l-Arti, Università ta’ Malta.
Zimmermann, K. 1987. Grammatisch bedeutsame Entlehnungen aus dem Spanischen in Otomí.
Ein Beitrag zur Theorie des Sprachkontakts. Neue Romania 5: 20–58.
Some notes on the syntax–pragmatics
interface in bilingual children
German in contact with French / Italian

Natascha Müller
Bergische Universität Wuppertal

The present article investigates finite verb placement in German subordinate


clauses. It will argue that there is cross-linguistic influence in the case of
bilingual German-French and German-Italian children. The children use a
Romance syntactic derivation for German finite subordinate clauses which is
based on an analysis of Romance infinitival clauses in which the prepositional
complementizers enter the derivation above the VP, and not as sister to the
IP they are associated with. The relation between complementizer and IP is
expressed by movement of the IP from within the VP to the specifier position
of the complementizer. The generalization of this type of derivation to all kinds
of complementizers is enforced by the existence of constructions in the German
input of the children which are compatible with this kind of analysis. The result
of the child analysis is root word order in German subordinate clauses. We
will argue that the children chose the Romance analysis since it minimizes the
interaction between syntax and pragmatics.

1. Introducion: Crosslinguistic influence and language separation

Research in bilingual first language acquisition has been guided by two main approach-
es: Either it has been argued that bilingual children are not able to separate their two
languages from early on, since the two languages influence each other (e.g. Taeschner
1983), or it has been shown that bilingual children are able to separate their two lan-
guages from early on and that there is no evidence for crosslinguistic influence (Meisel
1989, 1994, Genesee, 1989, Genesee, Nicoladis and Paradis 1995). Put differently, lan-
guage separation and crosslinguistic influence have been considered as being mutually
exclusive in describing early child bilingualism. The main reason for the assumption of
mutual exclusiveness is that most research has conceptualized separation and influence
as involving whole language systems (or languages). Recently, some researchers have
 Natascha Müller

become interested in the possibility that only some domains in simultaneous bilingual-
ism are vulnerable for cross-linguistic influence, while others are developed by the chil-
dren without any sign of influence. The present paper looks at the development of finite
verb placement in German subordinate clauses. In particular, it will be argued that the
absence of root-/non-root asymmetries in German child grammar is motivated by as-
sumptions which pertain to the syntax–pragmatics interface. More specifically, it will be
argued in line with Herkenrath, Karakoç and Rehbein (2003), that cross-linguistic influ-
ence may have the effect that “the grammar of the bilingual children becomes ‘larger’
than that of monolinguals” (ibid.: 257), which represents a problem if the child has to
unlearn this grammatical option. The kind of cross-linguistic influence discussed in this
paper is driven by pragmatics, i.e. by the function of the subordinate clause, and is “cre-
ated” by the bilingual children as a grammatical option which generalizes to all cases of
subordination. Thus, a pragmatically motivated option, which is well-defined for the
recipient language (the language which is being influenced), is overgeneralized to the
effect that the option is treated as if it was an obligatory syntactic rule.

1.1 Two conditions for cross-linguistic influence

Hulk and Müller (2000) and Müller and Hulk (2000, 2001) have defined two condi-
tions under which cross-linguistic influence is likely to occur. (a) The vulnerable gram-
matical phenomenon is an interface property, e.g., a grammatical property located at
the interface between syntax and pragmatics; (b) The surface strings of the two lan-
guages are similar for the expression of the vulnerable grammatical phenomenon.
What is meant by interface property? As a first example, let us have a closer look at the
null-subject phenomenon. Languages like Spanish and Italian do not express subject
pronouns unless subjects carry contrastive stress. In non null-subject languages like
English and German, sentences like duerme, dorme are translated as ‘he/she/it sleeps’
or ‘er / sie / es schläft’. The correct interpretation of the Spanish and Italian sentence
implies the interpretation of the subject. Thus, null-subjects are part of the syntactic
description of these sentences, namely pro. The existence of null-subjects is one prop-
erty of the so-called null-subject parameter. Null-subjects in null-subject languages are
not regulated by syntactic constraints. They are possible whenever the subject can be
considered as known to the hearer. Chomsky (1981: 65) has formulated a conversa-
tional maxim, the Avoid Pronoun Principle (APP), in order to guarantee that known-
subjects are indeed omitted in null-subject-languages. In other words, syntax opens
the possibility to have null-subjects, whereas pragmatics regulates their use. Violation
of the APP leads to unacceptable (not ungrammatical) strings, e.g. Giannii ha detto che
luii va venire alla festa ‘John has said that he will come to the party’ without contrastive
stress of lui, is unacceptable but a grammatical Italian construction. The APP also
guarantees that expletive subjects will not be phonetically realized in languages like
Spanish and Italian. The interaction between syntax and pragmatics can be summa-
rized as follows: In null-subject-languages like Spanish and Italian, syntax allows for
Some notes on the syntax–pragmatics interface in bilingual children 

an additional option apart from phonetically realizing subjects, namely to drop them.
The choice between the two options, realization or omission, is regulated by pragmat-
ics, not by syntax. The universal principle APP, a conversational maxim, determines
that if no further conditions are specified, pronouns will be avoided (for a thorough
discussion cf. Pillunat, Schmitz and Müller 2006). In non-null-subject-languages, the
subject can also be omitted: Ø Hab das schon gemacht, ‘Have done it already’. One
prerequisite is that the subject is the topic. So far, non-null-subject-languages seem to
pattern similarly with null-subject-languages. However, contrasting with null-subject-
languages, the subject in non-null-subject languages can only be omitted in root claus-
es, i.e. in first position. This syntactic restriction is unknown in null-subject-languages.
The question is whether the APP also regulates subject omissions in non-null-subject
languages. In German, sentences with a dropped subject and with an overt pronomi-
nal subject are discourse-pragmatically equivalent; they reflect a different speech style,
the subject omission belonging to an informal register: Ø Hab das schon gemacht / Ich
hab das schon gemacht. In other words, it is not pragmatics which regulates subject
omissions in non-null-subject-languages, but syntax proper. The interaction of both
components, syntax and pragmatics, is minimal in languages like German. We prefer
the term ‘minimal’ instead of ‘non-existent’ since it goes without saying that pragmat-
ics regulates whether subjects can be expressed as pronouns or must be expressed as
lexical DPs. Notice that the assumption about minimal interaction between syntax and
pragmatics in the domain of subjects can also account for the observation that exple-
tives do exist in non-null-subject-languages and that their presence / absence is regu-
lated by syntax, not by the APP.
The second example for the interaction between syntax and pragmatics in the
above non-trivial sense are object pronouns in the Romance languages. Interestingly,
object clitic pronouns (‘weak’ object pronouns which cannot be stressed, co-ordinated,
modified, etc., ‘le’/ ‘l’ ‘ in example (1) below) can be construed with noun phrases con-
taining an indefinite article and assume an interpretation as a ‘type’: i.e., ‘un homme’ (‘a
man’) is intended as ‘a man in general’. The important point is that the antecedent, the
noun phrase the pronoun refers to (and with which it agrees in gender and number),
must be an entity that can be presupposed. This explains why the utterance of speaker
B in (2) is incorrect. Here, the noun phrase ‘un verre’ (‘a glass’) cannot be presupposed
from what has been said before:
(1) Un homme, on le reconnaît par sa façon de parler.
A man, one him recognises by his way to speak
‘A man can be recognized by his way of speaking.’
(2) Speaker A: Je voudrais boire du vin.
I’ d like to- drink some wine
‘I’d like to drink some wine.’
Speaker B: D’accord. *Un verre, je l’ai dans l’ étagère.
Alright. A glass, I it have in the shelf
 Natascha Müller

Thus, object clitics can only be construed with a noun phrase which can be presup-
posed. We can push our analysis of object clitics even further and assume that the func-
tion of the clitic is that of marking presupposition. López (2003) noticed that Catalan
clitics occur in constructions which contain language material which is known (‘pre-
suppositional’). The two constructions discussed are CLLD (Clitic Left Dislocation, in
which a noun phrase has been positioned at the left of the clause as in (1) and in (3) ‘les
tables’/‘le tavole’) and CLRD (CLitic Right Dislocation, in which a noun phrase is being
placed at the right edge of the clause as in (4) ‘ton crétin de stylo’/‘la tua stupida penna’).
Clitics appear in order to mark the dislocated noun phrase as presupposed either in the
discourse or by the hearer. This is nicely shown by the example (5), indicating that the
construction without the object clitic is only grammatical once the object is not presup-
posed. If the object is presupposed, the clitic must be present; otherwise the construc-
tion becomes ungrammatical, as in (6). The examples (3) and (4) are taken from López
(2003: 199) and have been translated from Catalan into French and Italian.
(3) Speaker A French: Qu’est-ce que tu as fait avec les meubles?
Speaker A Italian: Cosa hai fatto con i mobili?
‘What did you do with the furniture?’
Speaker B French: Les tables, je les ai réparées le matin, mais les chaises je les ai
réparées le soir.
Speaker B Italian: Le tavole, le ho riparate la mattina, ma le sedie
Lit.: The tables(I) them repaired in the morning, but the chairs(I)
le ho riparate la sera.
them repaired at night.
(4) Speaker A French: Qu‘est-ce que tu as fait avec le stylo?
Speaker A Italian: Cosa hai fatto con la penna?
‘What did you do with the pen?’
Speaker B French: Je l‘ai oublié sur la table, ton crétin de stylo.
Speaker B Italian: L’ho dimenticata sulla tavola, la tua stupida penna.
Lit.: (I) it forgot on the table, your stupid pen.
(5) Speaker A French: Qu‘est-ce qu’il y a?
Speaker A Italian: Cos’ è?
‘What happened?’
Speaker B French: J’ ai oublié ton crétin de stylo sur la table.
Speaker B Italian: Ho dimenticata la tua stupida penna sulla tavola.
Lit.: (I) forgot your stupid pen on the table.
Speaker B French: *Je l‘ai oublié sur la table, ton crétin de stylo.
Speaker B Italian: *L’ ho dimenticata sulla tavola, la tua stupida penna.
(6) Speaker A French: Qu‘est-ce que tu as fait avec le stylo?
Speaker A Italian: Cosa hai fatto con la penna?
Some notes on the syntax–pragmatics interface in bilingual children 

Speaker B French: *J’ ai oublié sur la table, ton crétin de stylo.


Speaker B Italian: *Ho dimenticata sulla tavola, la tua stupida penna.
Lit.: (I) forgot on the table, your stupid pen.
These examples illustrate that object clitics are different from normal pronouns. Nor-
mal pronouns, although being used under certain pragmatic conditions, are not ob-
ligatory in syntax. Instead of a normal pronoun, a noun phrase can be used and the
sentence remains grammatical, although it becomes pragmatically odd, marked as ‘?’,
as shown in (7).
(7) Speaker A German: Hast du das Buch auf dem Tisch schon gelesen?
‘Have you already read the book on the table?’
Speaker B German: Das habe ich schon gelesen, ja.
It have I already read, yes
Speaker B German: ?Das Buch auf dem Tisch habe ich schon gelesen, ja.
The book on the table have I already read, yes
In other words, French and Italian clitics are not just syntactic categories which are
used under certain pragmatic conditions which favor pronominalisation, but they also
mark certain syntactic constituents in the languages as presuppositional. In other
words, pragmatics and syntax interact in such a way that pragmatic factors restrict the
possibilities offered by the syntactic system.
Both examples, the one of subject omissions and the one of object clitic realiza-
tions, illustrate a non-trivial interaction between syntax and pragmatics. These gram-
matical domains are characterised by the invasive nature of pragmatics onto syntax. It
is this non-trivial interaction between syntax and pragmatics which bilingual children
− as well as monolingual ones − find problematic in acquisition. In the case that the
second language does not present this kind of interface property, the bilingual
child may use the grammatical analysis of the less complex language not showing this
kind of interaction for both of her/his languages. Grammatical phenomena like the use
of object clitics in the Romance languages are vulnerable to cross-linguistic influence
due to their interface properties.
The second condition which favours cross-linguistic influence is a property of the
surface strings of the two languages. Hulk and Müller (2000) assume that the surface
strings are similar for the expression of the vulnerable grammatical phenomenon.
What is meant by 'similarity of surface strings’ in (b)? In order to illustrate similarity of
surface strings, we will take the well-known example of SVO. German is a verb-second
language, whereas Romance languages like Italian are generally treated as an SVO lan-
guage. However, both languages share the word order SVO in some contexts: the Ger-
man main clause in (8) can be translated directly into the Italian sentence in (9) below:
(8) Maria liest das Buch.
Mary reads the book
 Natascha Müller

(9) Maria legge il libro.


Mary reads the book
In other words, two otherwise syntactically different languages may share word order
on the surface.
Generally speaking, the hypothesis of Hulk and Müller (2000) is that word se-
quences in constructions which represent interface phenomena of grammar and dis-
play similarity in both languages are more difficult to process. Confronted with this
major computational complexity, the bilingual child will have recourse to the less com-
plex analysis of the grammatical property in question. If language A offers a less com-
plex analysis than language B for the grammatical phenomenon Z, the bilingual child
will make use of the less complex analysis for both languages. For example, a German-
Italian bilingual child may opt for one syntactic analysis for SVO constructions in both
languages, which requires the less complex analysis.
Taking computational complexity as the reason for why bilingual children use one
grammatical analysis for both languages implies that monolingual children will also
have problems with a grammatical phenomenon which is ‘complex’ in the above men-
tioned terms. Indeed, Müller and Hulk (2001) make this observation for monolingual
children with the more complex language background.

1.2 Cross-linguistic influence and language dominance

Recent studies observing crosslinguistic influence have explained the influence in


terms of language dominance, not in terms of properties of the grammatical phenom-
enon mentioned above (Bernardini Röst 2001, Döpke 1992, Gawlitzek-Maiwald and
Tracy 1996, Hulk 1997, Schlyter 1993, Tracy 1995): The stronger or “more developed”
language influences the weaker or “less developed” language. Most authors have de-
fined language dominance on the basis of a comparison of MLU values in the two
languages of the bilingual child and / or language use (the amount of utterances in the
two languages used during a recording session). The first attempts to define ‘language
dominance’ were made at the end of the 1980s and during the early 1990s. The authors
of these studies chose both quantitative and qualitative ways of measuring language
dominance, although there was no clear line with respect to which criterion should be
considered as quantitative and which as qualitative. Quantitative criteria were associ-
ated with language performance and linguistic production, while the investigation of
grammatical phenomena in both languages (considered to be areas of competence)
was considered to be qualitative. ‘Language dominance’ was, in this sense, considered
to be a type of ‘grammatical predominance’ (see, e.g., Lindholm & Padilla 1978, Pe-
tersen 1988, Schlyter 1993). The use of grammatical phenomena to determine lan-
guage dominance, however, may be criticised because their emergence may differ in
the two languages of the bilingual child, even in the respective monolingual children,
and it is subject to cross-linguistic variation. An illustrative example is a bilingual child
Some notes on the syntax–pragmatics interface in bilingual children 

who develops the accurate gender system later in German than in French. It goes with-
out saying that we cannot conclude from this situation that the bilingual child has a
dominant language, namely French, because it might be the case that monolingual
German children take more time for the acquisition of gender than monolingual
French children. Furthermore, the later development of gender in German may be due
to cross-linguistic influence in the bilingual child. Another example comes from the
article system. As Kupisch (2004) shows, the development of the article system is sub-
ject to cross-linguistic influence, with German developing faster in bilinguals than in
monolinguals. The result that, due to cross-linguistic influence, convergence to the
target is achieved around the same age in both languages does not allow us to conclude
that the bilingual does not have a dominant language. Among the quantitative meas-
ures for language dominance range are MLU, Upper Bound (the longest utterance in a
recording session), absolute number of utterances, MMU (Multi Morphemic Utter-
ances, i.e. the number of utterances with more than one word/morpheme), quantita-
tive aspects of the lexicon (number of verb/noun types) and the amount of mixed ut-
terances (which contain lexical material from language A and language B).
Due to space limitations, the present article cannot discuss the issue of language
dominance at depth; cf. Genesee, Nicoladis and Paradis (1995), Cantone, Kupisch,
Müller and Schmitz (2006). Furthermore, it is impossible to present a more thorough
analysis of the bilingual children’s language dominance. Therefore, it will compare bi-
lingual children with different degrees of balance measured on the basis of MLU.

1.3 Effects of cross-linguistic influence

After having explained the conditions for cross-linguistic influence to occur, we may
ourselves ask which effects it will have on the language acquisition process. This influ-
ence may be quantitative in nature, accelerating or delaying the bilingual’s develop-
mental process in comparison with monolingual peers. It may also have qualitative
effects, often called ‘transfer’ in the literature. The term ‘transfer’ has been used in con-
nection with second language acquisition, i.e., the successive acquisition of more than
one language, and was applied to the phenomenon in which previously acquired
knowledge is extended to a new domain (a new language). Depending on whether the
first and the second language are similar or different for the grammatical phenomenon
in question, transfer may be positive or negative. To summarise, if quantitative in na-
ture, the acquisition process should be delayed or speeded up in the bilingual, but the
data should not contain evidence for the extension of linguistic knowledge from the
first L1 onto the second native language. If qualitative in nature, the data should con-
tain evidence for the extension of a grammatical phenomenon X of language A onto
language B, e.g., one should find different types of extension, XA, XE, XG, in language B,
if we vary the other language, A, E, G, and by this the properties of the grammatical
phenomenon, or the developmental path in bilinguals should be different from that in
 Natascha Müller

monolinguals in showing a particular type of construction which is absent in mono-


lingual development.
The most surprising aspect of cross-linguistic influence is indeed the effect of ac-
celeration of the language acquisition process. A good example for accelerating effects
in language development is finite verb placement in German. From monolingual lan-
guage acquisition we know that children pass through a stage during which they place
the finite (and non-finite) verb clause-finally, even in main clauses − which does not
correspond to adult language (Clahsen 1982). A typical monolingual child is Chantal,
who has been studied by Schmitz (2004): until the age of 2;7, the verb-final pattern
prevails, i.e. she produces utterances like 'ich auch mache’, ‘I also make’ (instead of ‘Ich
mache [es] auch’) or ‘die Puppe schlafen will’, ‘the doll sleep wants’ (instead of ‘Die
Puppe will schlafen’) in 60% of her utterances. This amount will be reduced to 10% at
the age of 2;9 and to 5% at the age of 3;0.
Schmitz compared the development of verb placement in Chantal with that in
German-Italian bilingual children. The child Lukas is representative for the path cho-
sen by German-Italian bilingual children, namely that they seem to skip the stage in
the development of German which is characterised by the predominance of verb-final
patterns: Lukas reached a developmental stage at 2;3 which Chantal reaches at the age
of 2;10. The interesting observation is that the stage with predominant verb-final place-
ment is absent in the German of the bilingual child. Also evident is the fact that verb-
second placement (VS) is used earlier in bilingual child development, to an extent
which the monolingual child only reaches after the age of 3.
Delay effects are visible in other grammatical domains, like gender marking. In
the monolingual French and Italian corpora studied by Kupisch, Müller & Cantone
(2002) a very low number of gender errors has been observed (1.3% (385:5) for L1-
French and 1.9% (362:7) for L1-Italian). The observed children were all younger than
2;7. These monolingual Italian and French children mastered their gender systems
very early (before 2;3) without making a significant number of target-deviant gender
assignments. This was not the case for bilingual children speaking French-German
and Italian-German, respectively, which have been studied by Kupisch, Müller & Can-
tone (2002) and Müller & Kupisch (2003). Especially the French gender system ap-
peared to be acquired with a delay. Although the bilingual children were also quite
accurate with their gender marking if we consider the whole period from 1;6 – 2;6 in
Kupisch, Müller & Cantone (2002) and the period from 2;2 – 4;0 in Müller & Kupisch
(2003), in particular the German-French children reached an accuracy rate of only
60% at some points in development. One of the French-German children studied by
Müller & Kupisch (2003) reached the accuracy rate of 95% only at the age of 3;9.
Finally, even qualitative effects of cross-linguistic influence are visible in the cor-
pora of the bilingual children. The grammatical domain presented in this article will
be an example of this kind of cross-linguistic influence in bilingual acquisition.
Some notes on the syntax–pragmatics interface in bilingual children 

2. The research project

The data to be presented are part of the research project “Die Architektur der frühkind-
lichen bilingualen Sprachfähigkeit: Italienisch-Deutsch und Französisch-Deutsch in
Italien, Deutschland und Frankreich im Vergleich” (The architecture of the early child-
hood bilingual language faculty: Comparing Italian-German and French-German in
Italy, Germany and France).1 The bilingual children are presented in table (1). All chil-
dren have been raised bilingually (German-Italian/French) from birth in Germany. The
parents decided to raise them according to the “une personne – une langue”- strategy
(Ronjat 1913). In most cases (Céline is the exception) the mother speaks the Romance
language to the child and the father German. The corpus consists of video recordings
(made every fortnight). The recordings started, for the most part, at age 1;6. The lan-
guages were separated during the recordings. The German interviewer spoke German
with the child, while the Romance interviewer interacted with the child in the respective
Romance language.

Table 1.  Children under investigation

Alexander French/German recordings started at 2;2,6


Céline French/German recordings started at 2;0,9
Carlotta Italian/German recordings started at 1;8,28
Lukas Italian/German recordings started at 1;7,12
Jan Italian/German recordings started at 2;0,11

2.1 Language dominance

In the following we will discuss the German-French children Alexander and Céline
(for Céline cf. Cordes 2001). We will confine ourselves to the MLU-values of the chil-
dren in order to determine their language balance. More details are provided in Can-
tone, Kupisch, Müller and Schmitz (2006). Figure (1) shows that Alexander is a faily
balanced bilingual child, his French MLU being slightly higher at all points in his de-
velopment as compared to the values reached in German. In contrast to Alexander,
Céline has a dominant language, namely German. Her French MLU values differ con-
siderably from those in German, as can be seen in figure (2). In some recordings, the
difference between the two languages amounts to more than 1.
 Natascha Müller

7,0
6,0
5,0
4,0 German
MLU

3,0 French
2,0
1,0
0,0
25

19
3

22
6

7
24

28

11
0
,2
2,

7,

6,
3,
5,

1,

4,
3,

8,

8,
10
2;
2;

3;

3;
2;

2;

2;

3;

3;

3;
2;

age

Figure 1.  MLU: Alexander

6,0

5,0

4,0
German
MLU

3,0
French
2,0

1,0

0,0
13

12
9

2
19

,3

0
14

20

23

12
0,

6,

8,

8,
11
4,

0,

3,
1,

9,

4,

6,
2;

2;

3;
2;

2;
2;

2;

3;

3;

3;

3;

age

Figure 2.  MLU: Céline

Let us turn to the German-Italian children (Loconte 2001). Figure (3) shows Carlotta’s
MLU in both languages. Even more so than Alexander, she can be regarded as a bal-
anced bilingual child. Figure (4) presents graphically the MLU-values for Lukas. Until
the age of approx. 3;3, he can be considered as the most balanced child in the project.
After that age, he uses Italian to a declining degree, most of his utterances in the Italian
recordings being German or mixed utterances. Finally, figure (5) contains Jan’s MLU-
values. He can be compared with Céline with respect to the considerable difference
between the Italian and the German MLU until approx. age 3. His German is stronger
than his Italian. However, in contrast to Céline who does not use French but German
Some notes on the syntax–pragmatics interface in bilingual children 

with the French interviewer, Jan likes to speak Italian and mostly uses Italian during
the Italian recording session.

4
German
MLU

3
Italian
2

0
28 12 4 7 13 25 27 13 8 17 27 ,6
8, 1, 2, 4, 7, 9, 1, 2, 4, 6, 8, 11
1; 1 2; 2; 2; 2; 1 3; 3; 3; 3; 3;
1; 2;
age

Figure 3.  MLU: Carlotta

6,0

5,0

4,0
German
MLU

3,0
Italian
2,0

1,0

0,0

12 17 1,
3 23 15 18 27 3,
2
5,
8 15 20 22
7, 0, 2; 4, 7, 9, 1, 3; 3; 7, 9, 1,
1; 1;
1 2; 2; 2; 2;
1 3; 3; 3;
1

age

Figure 4.  MLU: Lukas


 Natascha Müller

4,5
4
3,5
3
2,5 German
MLU

2 Italian
1,5
1
0,5
0
1 5 7 2 7 9 3 1 5 7 4
,1 ,1 7, ,1 ,2 ,1 ,2 ,1 8, 0, ,1
2;
0
2;
4 2; 2;
9
;1
1 3,
2
3;
4
3;
6 3; 3;
1
4;
0
2
age

Figure 5.  MLU: Jan

3. Crosslinguistic Influence

3.1 Finite verb placement in German subordinate clauses

Müller (1993, 1994, 1998) has concluded, on the basis of a review of the relevant lit-
erature, that bilingual children with French, Italian or English as one of their two na-
tive languages have considerable problems with finite verb placement in German sub-
ordinate clauses.
Some notes on the syntax–pragmatics interface in bilingual children 

Absolute number

10

12

14
0

8
2;2,6

2;2,27

2;4,6

2;5,25

target-deviant
Vfinal
2;6,25

2;7,27

2;8,28

2;10,2

2;11,6

3;1,22

3;2,16
Age

3;3,22

3;4,19

3;5,24

3;6,21

3;8,11

3;9,7

3;10,6

3;11,10

4;1,26

4;3

4:4.25

Figure 6.  Subordinate clauses Alexander


 Natascha Müller

Absolute number

10

12
14

16
18

20
0
2

4
6

8
1;8,28
1;10,8

1;11,12
2;0,11
2;2,4

target-deviant
Vfinal
2;3,2
2;4,7

2;6,9
2;7,13
2;8,21
2;9,25

2;10,30
2;11,27
3;1,16
Age

3;2,13
3;3,11

3;4,8
3;5,6
3;6,17
3;7,13

3;8,27
3;10,22
3;11,26
4;1,14
4;2,11

4;3,9
4;4,6

Figure 7.  Subordinate clauses Carlotta


Some notes on the syntax–pragmatics interface in bilingual children 

Adult German is a V2-language and exhibits clause-final finite verbs in subordinate


clauses: Heute habe ich Geburtstag ‘Today have I birthday’ – Ich habe dir gesagt, dass ich
heute Geburtstag habe ‘I have you told that I today birthday have’. German exhibits a
root/non-root-asymmetry with respect to finite verb placement. For the SVO languag-
es like French and Italian, a root/non-root symmetry with respect to word order of fi-
nite verbs can be observed, i.e. the verb in subordinate clauses follows the subject, as
in main clauses. Müller (1998) concludes her review with the observation that about
half of the population of bilingual children have problems with finite verb placement
in German subordinate clauses if they acquire a second native language which is an
SVO language, i.e. French, Italian or English. Interestingly, two of the five children
studied in the present article have considerable problems with finite verb placement in
subordinate clauses: Alexander in the French-German study (cf. figure 6) and Carlotta
in the Italian-German study (cf. figure 7).
As can be seen from the figures (6) and (7), target-deviant verb placement ceases
to appear in both children only at the age of 4, i.e. German verb placement in subordi-
nate clauses can be regarded as a grammatical domain with late acquisition.
For the other three children, there are either no problems at all, i.e. the placement
of finite verbs in subordinates is ‘error-free’ or the development marginally shows root
word order in subordinate clauses. This is shown in figures (8), (9) and (10) for the
children Jan (German-Italian), Lukas (German-Italian) and Céline (German-French).
Since we are dealing with longitudinal data, absence of data should be avoided as
an argument for or against hypotheses concerning language development. Let us
therefore concentrate on the children Alexander and Carlotta who exhibit problems in
subordinate clauses.
 Natascha Müller

Absolute number

10

15

20

25
0

5
2;1,3

2;5,26

2;7,7

2;8,18

target-deviant
Vfinal
2;10,8

2;11,27

3;1,1

3;3,8

3;4,23

3;6,11
Age

3;7,22

3;9,15

3;10,27

4;0,14

4;2,25

4;4,6

4;5,17

4;7,5

4;9,16

4;10,27

Figure 8.  Subordinate clauses: Jan


Some notes on the syntax–pragmatics interface in bilingual children 

Absolute number

10

12

14
0

8
1;7,12

1;9,13

1;10,17

1;11,22

2;1,3

target-deviant
Vfinal
2;3,6

2;4,23
2;5,20

2;7,15

2;8,12

2;9,18

2;10,22

2;11,27
Age

3;1,30

3;3,2
3;4,7

3;5,8

3;6,13

3;7,15

3;8,17

3;9,20

3;10,17

3;11,22

4;1,20
4;3,14

4;4,12

Figure 9.  Subordinate clauses: Lukas


 Natascha Müller

Absolute number

10

12

14

16

18
0

8
2;0,9

2;1,6

2;3,15

2;4,19

target-deviant
Vfinal
2;5,25

2;6,21

2;8,2

2;8,29

2;10,5

2;11,3

2;11,29
Age

3;0,27

3;3,12

3;4,9

3;5,15

3;6,12

3;7,17

3;8,14

3;9,18

3;10,18

3;11,15

Figure 10.  Subordinate clauses Céline


Some notes on the syntax–pragmatics interface in bilingual children 

3.2 The “errors”: The influence of the other language

In her review of the literature, Müller (1998) finds that the kind of “errors” in German
subordinate clauses differs as a function of the second language. Whereas children
with French or English as first languages use COMP (complementizer) – X (any con-
stituent) – Vfin –Y (any constituent), dass heute habe ich Geburtstag ‘that today have I
birthday’, children with Italian use COMP – Vfin – S – X – Y, dass habe ich heute Geburts­
tag ‘that have I today birthday’. This is also true for the two bilingual children studied
here: Whereas Alexander makes use of 1.9% of VSX(Y) patterns in his German subor-
dinate clauses (only 2 out of 104), Carlotta uses it to a high degree, 32% of her subor-
dinate clauses showing exactly this pattern until the age of 3;7. All bilingual children
use SVfinX in their subordinate clauses, dass ich habe heute Geburtstag ‘that I have today
birthday’, i.e. this pattern occurs independently of the specific second language, French
or Italian.2 A further important result of Müller’s review is that finite verb placement in
the Romance language (or in English) is not problematic, i.e. we are dealing with an
error-free grammatical domain in these languages.
Before we turn to the discussion of the acquisition facts, we will introduce a new
view of word order in subordinate clauses and the syntax–pragmatics interface.

3.3 Complementizers in French, Italian and German- adult and child grammar

In generative syntax and other frameworks, complementizers are generally assumed to


form a constituent with the infinitival IP they are associated with, e.g. in the sentence
Il est important de chanter ‘It is important de sing-inf.’, de is often analyzed as forming
a constituent with chanter, since de requires an infinitival, and therefore is not compat-
ible with a finite phrase: *Il est important de vous chantiez ‘It is important de you sing-
subjunc.’ (Kayne 1999: 41). The same can be proposed for Italian di which, like its
French counterpart, is only compatible with an infinitival. Recently, Kayne (1999) has
argued that the French and Italian prepositional complementizers de / di enter the
derivation above the VP, that is, not as a sister to the IP they are associated with. The
relation between complementizer and IP is expressed by movement of the (infinitival)
IP to the specifier position of the complementizer. Subsequent movement of the com-
plementizer to a head W, is followed by phrasal movement to Spec, Spec,W accounts
for the observed word order in these languages. If prepositional complementizers do
not form a constituent with the infinitival IP they are associated with, the following
derivational steps are necessary for a sentence like Gianni ha tentato di cantare, ‘Gi-
anni has tried di sing-inf ’: The infinitival IP cantare is merged with the main verb
tentato, not with di. Di is subsequently merged with tentato cantare, the result being di
tentato cantare. Di then attracts the infinitival IP cantare to its Spec, resulting in canta-
rei di tentato ti. Di further raises to an immediately higher head W, dij+W cantarei tj
tentato ti. Di+W then attracts VP to its Spec, [tentato ti]k dij+W cantarei tj tk.
 Natascha Müller

Gianni ha WP

[tentato t i] k W'

di j DiP

cantare i Di'

dij VP

[ tentato cantare ]
i k

Figure 11.  Derivation of [...] tentato di cantare

Kayne (1999) advances several arguments for his analysis. The first type of argument relates
to the nominal character of French and Italian infinitives. French de and Italian di require
an infinitival, and they are not compatible with a finite phrase (Kayne 1999: 40f.).
(10) Il est important de chanter
It is important de sing-inf.
(11) *Il est important de vous chantez
It is important de you sing-subjunc.
(12) Gianni dice di aver capito
Gianni says di have-inf. understood
(13) *Gianni dice di (lui) ha capito
Gianni says di (he) has understood
Both Romance languages have developed a complementizer for finite clauses, namely
que and che. The same generalization extends to other prepositional complementizers,
like pour and per in the Romance languages. They are restricted to infinitival clauses.
If a finite clause is expressed, French uses pour que, Italian perché.
The second type of argument in favor of the derivation in figure (11) relates to the
observation that in French and Italian, infinitives do not occupy DP positions. Thus,
despite the nominal character of French and Italian infinitives, they do not occupy
ordinary DP positions. In conclusion, French and Italian infinitives have to be licensed
in a special way, one way being to have them preceded by de / di, or, extending Kayne’s
proposal, by other prepositional complementizers like pour / per. The new idea of
Kayne’s approach is that the prepositional complementizers do not form a constituent
with the infinitival IP they are associated with. As a consequence, the complementizer
Some notes on the syntax–pragmatics interface in bilingual children 

is not part of the argument of the (main) verb, but enters the derivation subsequent to
the merger of the infinitival IP and the main verb.
Let us assume that Romance language children have the correct analysis of infini-
tival clauses from the beginning. If Kayne’s analysis is extended to pour / per, which is
the first prepositional complementizer in child grammar (Müller 1993), and which
emerges about 6 months before the first adult-like finite complementizers, we would
have the derivation in figure (12) for a French child sentence: c’est pour dormir, ‘this is
for to-sleep’.

c' WP

[est t i ]k W'

pour j PourP

dormir i Pour'

pour j

[est dormir i ] k

Figure 12.  Derivation of infinitival introduced by pour

This structure would conform to the adult system. Bilingual Romance-German chil-
dren have been reported to use German für, the equivalent of pour / per as one of the
first elements which introduces clauses (cf. Müller 1993). This is also the case for the
two children of the present study who exhibit considerable problems with finite verb
placement in German subordinate clauses:3 für setzen, ‘for to-sit’ (Carlotta, 2;7,13), für
grattieren, ‘for to scratch’ (=kratzen, Alexander, 2;7,6). Since these German construc-
tions are created by the children on the basis of their Romance language – the con-
struction as such does not exist in German –, it seems plausible to extend the pour / per
analysis to für in German. Thus, a child sentence like das is für einkaufen ‘this is for
to-shop’ has the derivation in figure 13:
 Natascha Müller

das WP

[ist t i]k W'

für j FürP

einkaufen i Für'

für j

[ist einkaufen i] k

Figure 13.  Derivation of infinitival introduced by für

Kayne (1999) mentions the possibility that this kind of analysis extends to the real
complementizers que / che in the Romance languages. Let us assume that this is indeed
the case. We would like to exemplify this possibility with a French construction: Je sais
que tu lis ce livre ‘I know that you read this book’.

je WP

[sais t i] k W'

que j CP

IP[tu lis ce livre] i C'

que j

[sais IP [tu lis ce livre]i ] k

Figure 14.  que/che clauses: the Romance derivation

If children extend the analysis of pour / per clauses to für, it is plausible that they also
use their analysis of subordinate clauses introduced by que / che for German subordi-
nates. As a consequence, they would merge a finite clause with matrix word order with
the matrix verb and subsequently insert the complementizer dass (cf. figure 15), thus
the absence of sentence final placement of the finite verb in these children is predicted.
In other words, there is a link between the first prepositional complementizer für
Some notes on the syntax–pragmatics interface in bilingual children 

which is syntactically analyzed like French pour and Italian per, and the persistent
problems with finite verb placement in German subordinate clauses: Children who
exhibit problems in German subordinate clauses adopt a course to the complemen-
tizer system via using für with the same syntactic derivation as pour / per. Indeed, a
syntactic derivation of the prepositional complementizer introducing infinitival claus-
es is extended to real complementizers, as is indicated by the observation that prepo-
sitional complementizers can also introduce finite clauses in the children’s grammar,
e.g. für das macht musik, ‘for it makes music’ (Carlotta, 2;10,30), für die bahn kann, ‘for
the train can (go there)’ (Alexander, 2;11,20).

ich WP

[weiß t i] k W'

dass j DassP

IP[du kannst nicht mitkommen]i Dass'

dass j

[weiß IP[du kannst nicht mitkommen]i]k

Figure 15.  Derivation of finite subordinate clause in German with a Romance pour/per-
analysis:
*Ich weiß dass du kannst nicht mitkommen, ‘I know that you can not with-come’

Interestingly, the bilingual children have positive evidence for the Romance analysis of
complementizers in German. German subordinate clauses are verb-final if they are
introduced by a complementizer like dass.

ich weiß CP

dass IP (head-final)

du nicht mitkommen kannst

Figure 16.  dass clauses: the German derivation

German subordinate clauses which are not introduced by any lexical element do not
show clause-final placement of the finite verb, but main clause word order, as shown in
(14) in contrast to (15).
 Natascha Müller

(14) ich weiß, du kannst nicht mitkommen


I know you can not with-come
‘I know you cannot join us.’
(15) ich weiß, dass du nicht mitkommen kannst
I know that you not with-come can
Thus, for a sentence without a lexical complementizer, the für-derivation would yield
the correct analysis. In other words, the Romance syntactic derivation does not trans-
fer “blindly” into child German, but is supported by positive evidence.

ich WP

[weiß ti]k W'

Øj DassP

IP[du kannst nicht mitkommen] i Dass'

Øj

[weiß IP[du kannst nicht mitkommen]i]k

Figure 17.  Derivation of finite subordinate clause in German without lexical complemen-
tizer in terms of a Romance pour/per-analysis

One may object that we do not need dassP and WP for the derivation of such sentences.
But since these non verb-final clauses have to be distinguished from reported speech
(which becomes visible in spoken language from the type of pronoun used: mich fragst
du willst du brot oder würstchen, ‘me ask you want you bread or sausages’ (reported
speech) versus mich fragst du ob ich brot oder würstchen will, ‘me ask you if I bread or
sausages want’), the functional projections may be necessary, but remain empty in the
course of derivation. Spoken adult German also has a small class of elements which can
introduce subordinate clauses with main clause word order, weil, ‘because’ and obwohl,
‘although’ among them: Ich bin wütend weil ich kann nicht mitkommen, ‘I am angry
because I can not with-come’.
The next construction which might be regarded as evidence for the Romance für-
derivation in German child grammar is the conditional. Rehbein (1992) shows that
the German conditional has two possible realizations. Either as a wenn... dann, if...
then’ construction, in which wenn would trigger the verb-final pattern typical of Ger-
man subordinate clauses, or an alternative with a finite verb in clause-initial position:
hätte ich im Lotto gespielt, dann hätte ich diesmal bestimmt gewonnen, ‘had I in the lot-
Some notes on the syntax–pragmatics interface in bilingual children 

tery played, then would I this time for sure won’. Thus, the verb-final pattern competes
with non-verb-final alternatives. These alternatives might constitute positive evidence
for the Romance für-derivation applied to German complementizers in general.
In section 3.2 we have mentioned that the Romance languages, French and Italian,
have a slightly different effect on finite verb placement in German subordinate clauses.
Whereas Alexander (French-German) did not use VSX(Y) patterns in his German
subordinate clauses, Carlotta uses this pattern in about one third of her subordinate
clauses. It is not the purpose of the present article to investigate the syntactic deriva-
tions for the different word orders. Notice, however, that those languages, which in
bilingual combinations regulate that German subordinate clauses are COMP-Vfin-S-X,
e.g. Italian, are characterized by the absence of that-t-effects. In turn, language combi-
nations which exclude this pattern, like French, manifest that-t-effects.4 Müller (2005)
argues that this difference explains word orders which are specific to the language
combinations. We have mentioned the kind of errors bilingual children make in Ger-
man subordinate clauses, because we wanted to give further evidence for the assump-
tion that we are dealing with cross-linguistic influence, the proof being that a certain
language combination matters for a certain kind of error to occur.
Let us now turn to the syntax–pragmatics interface, and approach the question
why the Romance analysis is extented to German subordinate clauses.

3.4 The syntax–pragmatics interface

In the Principles & Parameters framework and in early work in Minimalism (Chomsky
1995), the syntactic derivation could interface with the interpretive systems only at one
point, which was called Spell Out. This operation strips the phonological features off the
lexical items and maps them onto a structure called Phonetic Form, which in turn inter-
faces with the auditory-perceptual system. Spell Out is also the earliest point at which,
given that all uninterpretable features are checked off, a structure called Logical Form
can be assumed, which in turn interfaces with the conceptual-intentional system.
More recently, some researchers like Epstein, Groat, Kawashima and Kitahara
(1998) and López (2003) have challenged the view that interfacing with the interpre-
tive systems must wait until the derivation is finished. Epstein et al. (1998) provide the
most radical approach in the sense of assuming that the interpretive systems interface
invasively with the syntactic derivation after every application of Merge and Move. As
a result, there are no PF or LF structures. López (2003), building on recent proposals
by Chomsky (2000, 2001) and Uriagereka (1999), defends a more moderate view on
the invasiveness of the interpretive systems. He assumes that the derivation spells-out
at particular points called phases. More specifically, a phase is headed by the light verb
v or by C. These represent the two points at which the derivation can be handed over
to the interpretive systems, once vP / CP is being derived. Once a phase is spelled-out,
it is opaque. The only exception to this is the edge of the phase, which is defined as the
head of the phase and its Specs.
 Natascha Müller

It is the more moderate framework which we will adopt in this article. In particular,
we are interested in the point CP at which López allows the syntactic derivation to inter-
face with pragmatics, “the interpretive module that deals with focus/ presupposition
structures, contrast, and possibly other notions […]” (López 2003: 195). In his approach,
pragmatic values – like presupposition and contrast – are regarded as features which are
assigned to constituents in the syntactic component and remain as parts of the feature
matrix of these constituents. In other words, pragmatics can assign features to constitu-
ents in the CHL (computational system for human language) which were absent in the
lexical array. The pragmatic component can thus inspect a syntactic structure and attach
a (pragmatic) feature to it. Pragmatics can invade the computational system only at two
points in the derivation, when the vP / CP phase is being completed.
What pragmatic value(s) could be involved in subordinate clauses? Let us exem-
plify the relevance of pragmatic values by using the example of conditionals. If en-
coded in a wenn... dann construction, or in the Romance equivalent construction with
si / se, the feature “conditional” is introduced via the lexical array of wenn, si, and se.
However, German as well as the Romance languages exhibit an alternative construc-
tion for which it would be difficult to imagine a feature “conditional” in the lexical ar-
ray of one element in the clause. This is the following construction hätte ich im Lotto
gespielt, dann hätte ich diesmal bestimmt gewonnen which exists in French as a con-
struction with a clause-initial finite verb as well as aurais-je le temps.... ‘had-I the
time...’.5 In this case it is necessary to assign features to constituents which are absent in
the lexical array.6
López (2003: 204) discusses the case of clitics which spell out the pragmatic fea-
ture “presupposition” or “[p]”. The feature [+p] is assigned to the EPP of v. Pragmatics
can “invade CHL and assigns [+p] only after the vP phase is completed.” Spec,v also
becomes [+p] since this specifier is licensed by an EPP with a [+p] feature. If it is plau-
sible to assume that in case of conditionals the whole sentence containing the verb in
the conditional is marked as such, it would be the functional category C or FürP/DassP
in Kayne’s analysis, which receives this feature. Pragmatics would have to assign this
feature to C/Dass/Für (or the EPP of C/Dass/Für) and the specifier of C/Dass/Für
would have the same feature since it is licensed by an EPP with a feature “conditional”.
In other words, we would like to suggest that there are two possibilities for features to
enter the derivation: either via the lexical array or via pragmatics.
Some notes on the syntax–pragmatics interface in bilingual children 

ich WP

schrie W'

wenn CP

ich war baby C'

C+EPP=wenn [+temporal]

schrie ich war baby


'I cried when I was a baby'

Figure 18.  Derivation of a German subordinate clause with [+temporal] inserted


via pragmatics

Let us turn to child grammar. Research on early child language has shown that pragmat-
ics, apart from its identifying function, can also license empty constituents (Müller and
Hulk 2001). If this is plausible, this would have the following consequences for the gram-
matical domain of subordinate clauses. We would predict that children will first choose
the possibility for features to enter the derivation via pragmatics, and only later study the
lexical array in more detail. In fact, researchers find that young children, monolingual
and bilingual, start to produce subordinate clauses long before they use complementiz-
ers; the position of the complementizer in the target language either remains empty in
child language (Müller 1993: sitz da un pa auf papa komm nich (=sitzt da und passt auf
dass der Papa nicht kommt), ‘sit there and pay-attention daddy comes not’) or is filled
with a dummy complementizer, like là, ‘there’ (Müller 1993: demander maman là il est
(=demander maman où il est), ‘ask mummy there it is’). This view would contradict
learning models which suppose that lexical learning is the driving force in syntactic
development. If the features needed for subordinates are introduced into syntax via
pragmatics, or put differently, if the conjunction does not encode the respective features,
it is the whole clause which marks the features. The (adult) Romance analysis of subor-
dinates nicely reflects this relation between syntax and pragmatics. In Romance pour /
per-clauses, the subordinate clause is merged with the matrix verb and later moved into
the functional projection which hosts pour / per. As a consequence of the application of
the Romance analysis to German subordinates, bilingual children use root word order
with finite verbs in subordinate clauses. Unlike adult German, child language comple-
mentizers are not analyzed as elements inserted into a syntactic position where, in ma-
 Natascha Müller

trix clauses, the finite verb checks off finiteness. A German-French or German-Italian
bilingual child will thus ignore the importance of the complementizer and assume a
symmetric analysis of root and non-root clauses due to the insertion of features needed
for subordination via pragmatics. Children might opt for invasive pragmatics even in
cases for which the feature of subordinate clauses, conditional, temporal, causal, etc. is
encoded in the lexical array of a conjunction in adult grammar. A child sentence with an
invasively introduced feature [+temporal] would have the structure in figure (18), inde-
pendently of whether a conjunction is present or not.
Müller (1993) shows that the bilingual children who have problems with finite
verb placement in German subordinate clauses acquire the target-like verb-final order
with each conjunction separately. In other words, the children have to analyze and
acquire the features for subordination with each lexical array. The last conjunctions to
be integrated into the target-like verb-final pattern are the semantically empty comple-
mentizers dass, ‘that’ and ob, ‘if/whether’. The way children revise their analysis of Ger-
man subordinate clauses may be interpreted as a further indication that the kind of
cross-linguistic influence we are facing with finite verb placement in German subordi-
nates is what is defined as transfer in section 1.3.
What prediction would follow from the approach outlined here? The prediction
would be that monolingual children indeed have problems with those subordinate
clauses for which the adult language offers an alternative with features introduced in-
vasively via pragmatics. These are conditionals. Stern & Stern (1928) reported these
problems in conditionals in the German monolingual children they studied: wenn ihr
würdet immerfort in Berlin geblieben sein, so würdet ihr immerfort Berliner gewesen
sein, ‘if you would always in Berlin stayed to-be, then would you always people-from-
Berlin been to-be’.
Finally, let us come back to the two conditions for cross-linguistic influence to oc-
cur. To reiterate, the two conditions presented in section 1.1 were: (a) The vulnerable
grammatical phenomenon is an interface properties, e.g., a grammatical property lo-
cated at the interface between syntax and pragmatics; (b) The surface strings of the two
languages are similar for the expression of the vulnerable grammatical phenomenon.
We have already mentioned that adult German exhibits subordinate clauses with root
word order. These clauses might lead the children into generalizing the Romance anal-
ysis onto German subordinates. The first condition is also fulfilled since the gram-
matical domain studied represents an interface property. As discussed in section 1.1, it
is also true for subordinates that pragmatics and syntax interact in such a way that
pragmatic factors restrict the possibilities offered by the syntactic system. In section
3.3 we have mentioned the possibility that (adult) German subordinate clauses with
root word order can be analyzed à la Kayne, as Romance subordinates in general. If
this assumption turns out to be correct, German would be characterized by a non-iso-
morphic interplay between syntax and pragmatics: If the features needed for subordi-
nation are inserted via pragmatics, the finite verb of the subordinate must check off
these features (residually, this option exists for Romance conditionals). If the features
Some notes on the syntax–pragmatics interface in bilingual children 

needed for subordination are inserted via the lexical array of the complementizer, it is
the complementizer which guarantees that the construction is well-formed. It is the
role of pragmatics which restricts the possibilities offered by the syntactic system. Un-
der the assumption that children find this kind of interaction complex, they will opt
for the Romance analysis of German subordinates which minimizes interaction be-
tween syntax and pragmatics. Under the Romance analysis, nothing follows for syntax
if features for subordination are inserted via the lexical array of complementizers, or
via pragmatics.
What other consequences follow from the analysis presented here? Cantone (2004)
studies code-switching in simultaneous bilingual Italian-German children from age
2;6 to 5. Her results show that children do mix between the functional head C and its
complement:
(16) perché ihr seid böse (Lu, 3;11,2, Italian recording)
because you are bad
(17) pecché ich war kleiner =perché (Lu, 4;0,5, Italian recording)
because I was younger
(18) wir sind aus- perché wir sind aus- aus- aus- auf deutsch-auf
deutschland
We are from – because we are from- from- from in German- in
Germany (Ja, 3;1,1, German recording)
(19) hai visto che geht leicht
have(you) seen that (it) goes easy (Lu, 3;4,25, Italian context)
(20) guarda che war hier
Look that (it) was here (Lu, 3;10,3, Italian context)
Cantone hypothesizes that the language of the complementizer determines the word
order of its complement. Subordinate clauses are a good testing ground for the hypoth-
esis since German subordinate clauses are verb-final, whereas Italian subordinates are
SVO. Since German weil (‘because’)-clauses can be non-verb-final in spoken language,
(16) – (18) do not constitute clear examples. (19) and (20), however, might show that an
Italian complementizer determines Italian word order in the subordinate clause, al-
though the clause is made up of German lexical items. With German word order, the
child should have said hai visto che leicht geht and guarda che hier war. Furthermore, in
these examples the German finite verb in the subordinate clause is not accompanied by
an overt subject, a possibility characteristic of Italian, not of German, given that Italian
is a null-subject language. Cantone specifies that if the complementizer is Italian, then
the structure below the functional head C is Italian, too, independent of whether its
syntactic positions are filled with Italian or with German elements. In the examples,
Italian C can be combined with an underlying structure which contains a null-subject.
Under the approach of subordinates presented here, the conclusion that the lan-
guage of the complementizer determiners word order in the subordinate clause fol-
 Natascha Müller

lows rather naturally. If the complementizer is Romance, the Romance analysis will
apply: It merges the subordinate with the matrix verb first and later moves the subor-
dinate into the functional projection which hosts the complementizer. Romance word
order will result, which is corroborated by the data. The code-switching data thus lend
support to the analysis presented here.

3.5 A note on language dominance

In what follows, we would like to exclude the possibility that language dominance is
the driving factor for why the children Alexander and Carlotta extend the Romance
analysis of subordinates to German. We have illustrated the MLU values of the two
children. For Alexander, we could argue that since his German is weaker than his
French, he will try the Romance option for German. However, notice that this argu-
ment does not generalize to Carlotta’s case. She can be considered as a balanced bilin-
gual, nevertheless, she opts for the Romance analysis.
What about the children Céline, Jan and Lukas? Lukas can be compared with Car-
lotta, at least until the age of 3;3. In both children, the MLUs of both languages nearly
match. Jan and, to a greater extent, Céline have a dominant language, namely German.
We may speculate that those children who do not show problems in subordinate claus-
es are children who develop German as a dominant language. Schmitz (2004), analyz-
ing dative case errors in the children Alexander, Céline, Lukas and Carlotta comes to
this result, namely that only those children who develop German as a dominant lan-
guage reach the target stage as quickly as children who are balanced for grammatical
domains which are vulnerable for cross-linguistic influence. Notice, however, that Lu-
kas can be considered a balanced bilingual child until the age of 3;3, the age span dur-
ing which he very rarely produces verb placement errors in subordinate clauses. Under
the assumed relation between cross-linguistic influence and language dominance, we
would wrongly predict that he should pattern like Carlotta. In conclusion, the balance
between the two languages, although an interesting research area as such, does not
explain why bilingual children show signs of cross-linguistic influence in their devel-
opment of verb placement patterns in German subordinate clauses.

4. Conclusion

The present investigation has attempted to show that some bilingual children may use a
derivation of language A for language B as well. This was the case since language B con-
tained two constructions of which one was the common construction in language A. A
further important factor which comes into play is the syntax–pragmatics interface. Chil-
dren tend to minimize the interaction between pragmatics and syntax if they interact in
a way that pragmatic factors restrict the possibilities offered by the syntactic system.
Some notes on the syntax–pragmatics interface in bilingual children 

The remaining problem is why only half of the bilingual population exhibit the
kind of developmental path outlined here. We have no solution to offer for the children
who do not show signs of cross-linguistic influence. What seems plausible, when look-
ing at the differing degree of balance of the children is that language dominance can be
excluded as a reason for absence / presence of language influence.

Notes

1. The project is financed by the Deutsche Forschungsgemeinschaft and directed by the pre-
sent author at Bergische Universität Wuppertal (nmueller@uni-wuppertal.de). For further de-
tails cf. Müller, Cantone, Kupisch and Schmitz (2002), Müller, Kupisch, Schmitz and Cantone
(2006) and Cantone, Kupisch, Müller and Schmitz (2006).
2. Müller (1998) also presents results from monolingual German children on the develop-
ment of finite verb placement in subordinates. One reviewer formulates the prediction that mo-
nolingual German children would have similar problems with verb placement in subordinate
clauses as bilinguals in German. One could also predict that there would be a bilingual-mono-
lingual difference in the magnitude of the problem, with Romance-German bilinguals produ-
cing significantly more deviant verb placements than monolinguals because of the strengthe-
ning effect of the Romance language. It is not possible to discuss this issue in more depth, since
the situation is complex. Monolingual children from Northern parts of Germany (the same re-
gion as the bilingual children studied here) acquire finite verb placement without error. They use
finite verbs clause-finally even at a time when they have not acquired complementizers yet.
Monolingual children from Southern parts of German take some time to acquire the correct
verb placement in subordinates; this may be the case, because their input contains evidence for
non-V-final placement of finite verbs with a considerable amount of lexical elements introdu-
cing subordinate clauses. Müller (1998) argues that these monolingual German children have an
input in German which the Italian-German bilingual children have in the Romance language,
favoring COMP – Vfin – S – X – Y.
3. Für does not figure among the prepositions which introduce infinitival clauses in standard
German. Instead, the particle zu is used in German infinitival constructions. für-infinitives are
in use in the Ruhr-area: 360 Eier für mitten Auto fahrn ‘360 eggs for with-the car drive’ (Gawlit-
zek-Maiwald 1997:35). The preposition für replaces um in für-infinitives, the common preposi-
tion to introduce infinitival clauses; the infinitival particle zu is often not realized. für-infinitives
allow for the realization of the subject, which is otherwise impossible in infinitivals (Voyles
1983, mentioned in Gawlitzek-Maiwald 1997:35): *Menschen zu irren ist möglich ‘People to err
is possible’ versus Für Menschen zu irren ist möglich ‘For people to err is possible’. The bilingual
children investigated in the present study have been raised in the Northern parts of Germany;
they did not have für-infinitives in their input.
4. French and Italian differ with respect to the presence / absence of the so-called that-t-effect:
In French, it is impossible to move the subject of an embedded clauses into the front position of
the main clause, thus the ungrammaticality of Qui crois-tu que viendras ‘who think you that
will-come’. The sentence is grammatical in Italian Chi credi che verrà. Most researchers assume
that fronted subjects cannot be moved in one step if two clauses are involved, but that they have
 Natascha Müller

to land in the specifier position of the functional projection which hosts the complementizer,
CP. We can interpret the examples in the sense that the French complementizer que does not
license a position preceding it which can host the subject (Bošković & Lasnik 2003 assume that
such a head does not have an EPP (Extended Projection Principle) feature (guaranteeing that a
specifier will be projected) and therefore will not project a specifier position) whereas Italian che
does. The example is ungrammatical because the subject, in order to be fronted, would have to
move in one step, which is impossible though *Quii crois tu que ti viendras. An alternative ac-
count would be to analyze the trace following the complementizer que as the offending trace:
*Quii crois tu ti que ti viendras. Rizzi (1990) has argued on the basis of minimality that the com-
plementizer que does not carry agreement features and therefore cannot properly govern the
subject trace, the result being an ungrammatical sentence. French qui and Italian che carry
agreement feature, explaining the grammaticality of Qui crois tu qui viendras and Chi credi che
verrà. “If Agr is selected for the head of Comp, it must be coindexed with its specifier; […] here
the subject trace is properly head-governed by Agr in the head of Comp and antecedent-gover-
ned by the specifier of Comp.” (Rizzi 1990:52f.) He summarizes that “a C0 endowed with such
features is an intrinsic governor.” (Rizzi 1990:122) Interestingly, Rizzi (1997) has noted that the
that-t-effect can also be voided if a TopP (Topic Phrase) intervenes between the position which
hosts the complementizer (Force in his approach) and the functional projection which hosts the
trace of the fronted subject (Fin in his approach): Who did Leslie say that *(for all intents and
purposes) was the mayor of the city?
5. In French, these constructions belong to the register of written language. It would be inte-
resting to study the occurrence of conditionals in the German input of the children.
6. One reviewer remarks that the finite verb is morphologically marked for the conjunctive
(II). This observation generalizes to the French example. Notice however that the morphological
marking on the verb could not explain the word order. The feature “conditional” would be nee-
ded independently of the morphological marking of the finite verb.
7. One reviewer suggests that the variation among children is hard to accommodate with the
presented account, given that the reinforcing effect from the Romance language is present in all
children. Recent studies in the literature have suggested that the bilingual child’s problem may
be related to the processing resources necessary to integrate syntactic and pragmatic knowledge
(see Avrutin 1999). The processing domain is open to developmental effects, and therefore to
inter-subject variation. The nature of the data, spontaneous production data, makes it impossi-
ble to disentangle a representational and a processing approach. Notice however, that a proces-
sing approach could not predict the occurrence of different target-deviant word orders, depen-
ding on the language combination. In fact, the presence of COMP-Vfin-S-X in German
subordinates of German-Italian children and its absence in French-German children reinforces
the competence-driven approach presented here.

References

Avrutin, S. 1999. Development of the Syntax-Discourse Interface. Dordrecht: Kluwer.


Bernardini Röst, P. 2001. Lo squilibrio nell’acquisizione di due lingue nell’infanzia: Indagine longitu-
dinale sullo sviluppo della sintassi nominale. Lund : Licentiatavhandling, University of Lund.
Some notes on the syntax–pragmatics interface in bilingual children 

Bošković, Z. and Lasnik, H. 2003. On the distribution of null complementizers. Linguistic In-
quiry 34(4): 527–546.
Cantone, K. 2004. Code-Switching in Bilingual Children. PhD dissertation, University of Ham-
burg.
Cantone, K., Kupisch, T., Müller, N; and Schmitz, K. 2006. Redefining language dominance in
bilingual children. Submitted
Chomsky, N. 1981. Lectures on Government and Binding. The Pisa lectures. Foris: Dordrecht.
Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press.
Chomsky, N. 2000. Minimalist inquiries: The framework. In Step by Step: Essays on minimalist
syntax in honor of Howard Lasnik, R. Martin, D. Michaels and J. Uriagereka (eds), 89–156.
Cambridge MA: The MIT Press.
Chomsky, N. 2001. Derivation by phase. In Ken Hale: A life in language, M. Kenstowicz (ed.),
1–52. Cambridge MA: The MIT Press.
Clahsen, H. 1982. Spracherwerb in der Kindheit. Eine Untersuchung zur Entwicklung der Syntax
bei Kleinkindern. Tübingen: Narr
Cordes, J. 2001. Zum unausgewogenen doppelten Erstspracherwerb eines deutsch-französisch
aufwachsenden Kindes: Eine empirische Untersuchung. MA thesis, University of Hamburg.
Döpke, S. 1992. One Parent, One Language: An interactional approach. Amsterdam: John Ben-
jamins.
Epstein, S., Groat E., Kawashima R., and Kitahara H. 1998. A Derivational Approach to Syntactic
Relations. Oxford: OUP.
Gawlitzek-Maiwald, I. 1997. Der monolinguale und bilinguale Erwerb von Infinitivkonstruktio-
nen. Ein Vergleich von Deutsch und Englisch. Tübingen: Niemeyer.
Gawlitzek-Maiwald, I. and Tracy, R. 1996. Bilingual bootstrapping. Linguistics 34: 901–926.
Genesee, F. 1989. Early bilingual development: One language or two? Journal of Child Language
16: 161–179.
Genesee, F., Nicoladis, E. and Paradis, J. 1995. Language differentiation in early bilingual devel-
opment. Journal of Child Language 22: 611–631.
Herkenrath, A., Karakoç, B. and Rehbein, J. 2003. Interrogative elements as subordinators in
Turkish. In (In)vulnerable Domains in Multilingualism, N. Müller (ed.), 221–269. Amster-
dam: John Benjamins.
Hulk, A. 1997. The acquisition of French object pronouns by a Dutch/French bilingual child. In
Language Acquisition: Knowledge, representation and processing. Proceedings of the GALA
‘97 Conference on Language Acquisition, A. Sorace A., C. Heycock and R. Shillcock (eds),
521–526. Edinburgh: University of Edinburgh Press.
Hulk, A. and Müller, N. 2000. Crosslinguistic influence at the interface between syntax and
pragmatics. Bilingualism: Language and Cognition 3(3): 227–244.
Kayne, R. 1999. Prepositional complementizers as attractors. Probus 11: 39–73.
Kupisch, T. 2004. The Acquisition of Determiners in Bilingual German-Italian and German-
French Children. PhD dissertation, University of Hamburg.
Kupisch, T., Müller, N. and Cantone K. 2002. Gender in monolingual and bilingual first lan-
guage acquisition: Comparing Italian and French. Lingue e Linguaggio 1: 107–149.
Lindholm, K. J. and Padilla A. M. 1978. Language mixing in bilingual children. Journal of Child
Language 5: 327–335
Loconte, A. 2001. Zur Sprachdominanz bei bilingual deutsch-italienischen Kindern. MA thesis,
University of Hamburg.
López, L. 2003. Steps for a well-adjusted dislocation. Studia Linguistica 57(3): 193–231
 Natascha Müller

Meisel, J. M. 1989. Early differentiation of languages in bilingual children. In Bilingualism across


the Lifespan: Aspects of acquisition, maturity, and loss, K. Hyltenstam and L. Obler (eds),
13–40. Cambridge: CUP.
Meisel, J. M. (ed.) 1994. Bilingual First Language Acquisition: German and French. Amsterdam:
John Benjamins.
Müller, N. 1993. Komplexe Sätze. Der Erwerb von COMP und von Wortstellungsmustern bei bi-
lingualen Kindern (Französisch/Deutsch). Tübingen: Narr.
Müller, N. 1994. Parameters cannot be reset: Evidence from the development of COMP. In Bilin-
gual First Language Acquisition. French and German grammatical development, J.M. Meisel
(ed.), 235:269, Amsterdam: John Benjamins.
Müller, N. 1998. Transfer in bilingual first language acquisition. Bilingualism: Language and
Cognition 1(3): 151–171.
Müller, N. 2005. SELECT und funktionale Kategorien im bilingualen Erstspracherwerb: Franzö-
sisch, Italienisch und Deutsch im Vergleich. Paper presented at Hamburg University.
Müller, N., Cantone, K., Kupisch, T. and Schmitz, K. 2002. Zum Spracheneinfluss im bilingualen
Erstspracherwerb: Italienisch – Deutsch. Linguistische Berichte 190: 157–206.
Müller, N. and Hulk, A. 2000. Crosslinguistic influence in bilingual children: Object omissions
and root infinitives. In Proceedings of the 24th Annual Boston University Conference on Lan-
guage Development, C. Howell, S.A. Fish and T. Keith-Lucas (eds), 546–557. Somerville
MA: Cascadilla.
Müller, N. and Hulk, A. 2001. Crosslinguistic influence in bilingual language acquisition: Italian
and French as recipient languages. Bilingualism: Language and Cognition 4(1): 1–21.
Müller, N. and Kupisch, T. 2003. Zum simultanen Erwerb des Deutschen und des Französischen
bei (un)ausgeglichen bilingualen Kindern. Vox Romanica 62: 145–169.
Müller, N., Kupisch, T., Schmitz, K. and Cantone, K. 2006. Einführung in die Mehrsprachigkeits-
forschung. Deutsch, Französisch, Italienisch. Tübingen: Narr.
Petersen J. 1988. Word-internal code-switching constraints in a bilingual child‘s grammar. Lin-
guistics 26: 479–493.
Pillunat, A., Schmitz, K. and Müller, N. 2006. Die Schnittstelle Syntax-Pragmatik: Subjektauslas-
sungen bei bilingual deutsch-französisch aufwachsenden Kindern. Zeitschrift für Literatur-
wissenschaft und Linguistik 143, 7–24.
Rehbein, J. 1992. Zur Wortstellung im komplexen deutschen Satz. In Deutsche Syntax. Ansichten
und Aussichten, Hoffman L. (ed.), 523–574. Berlin: Walter de Gruyter.
Rizzi L. 1990. Relativized Minimality. Cambridge MA: The MIT Press.
Rizzi L. 1997. The fine structure of the left periphery. In Elements of Grammar. Handbook in
generative syntax, L. Haegeman (ed.), 281–337.Dordrecht: Kluwer.
Ronjat, J. 1913. Le développement du langage observé chez un enfant bilingue. Paris: Champion.
Schlyter, S. 1993. The weaker language in bilingual Swedish-French children. In Progression and
Regression in Language, Hyltenstam K. and Viberg A. (eds), 289–308. Cambridge: CUP.
Schmitz, K. 2004. Erwerb der Verben mit zwei Objekten durch bilingual deutsch/französisch und
deutsch/italienisch aufwachsende Kinder. Unpublished PhD Thesis, University of Hamburg.
Stern, C. and Stern W. 1928. Die Kindersprache, neu abgedruckt 1975. Darmstadt: Wissen-
schaftliche Buchgesellschaft.
Taeschner, T. 1983. The Sun is Feminine: A Study of Language Acquisition in Bilingual Children.
Berlin: Springer.
Tracy, R. 1995. Child Languages in Contact: Bilingual Language Acquisition in Early Childhood.
Habilitationsschrift: University of Tübingen.
Some notes on the syntax–pragmatics interface in bilingual children 

Uriagereka, J. 1995. Aspects of the syntax of clitic placement in Western Romance. Linguistic
Inquiry 26 (1): 79–124.
Voyles, J. 1983. Ansätze zu einer deutschen Grammatik. Göppingen: Kümmerle.
section 2

Pronouns, topics and subjects


Distribution and function of clitic
object pronouns in popular
16th-18th century Greek narratives
A synchronic and diachronic perspective

Chrystalla A. Thoma
University of Hamburg and Berlitz International, Costa Rica

The phenomenon of variation in weak object pronoun placement in Early


Modern Greek has been the object of a number of studies. This paper presents
some results from a quantitative and qualitative analysis of the phenomenon
based on a varied corpus. The perspective taken is functional, focusing mainly
on Talmy Givón’s model of referential cohesion in narrative discourse. The
findings reveal a consistent pattern of variation that is statistically significant,
and they provide a plausible explanation both for a functional distribution as
well as the historical change leading to today’s relatively stable patterns.1

1. Introduction

In the 16th-18th centuries (with few and sometimes controversial earlier cases) we find
increased production of ‘low’ narrative texts, i.e. texts written in a popular, oral-like reg-
ister in a language that is very similar to Modern Greek (cf. Eideneier 1999). Here ‘low’
is seen as opposed to ‘high’, a written register which attempted to emulate ancient Greek
of the Classical and sometimes even of the Homeric period (cf. Toufexis in prep.). The
Greek of the 16th-18th centuries is referred to as “Early Modern Greek”2 and this is the
term we shall be using throughout this paper. In the ‘low’ narrative texts of our corpus
we find great variation in the placement of unstressed, object pronouns before and after
the verbal phrase. This ample variation is found in certain texts of the previous period as
well (Late Medieval Greek) and has been the object of a number of studies. To date, no
conclusive solution has been offered to the problem of possible rules, patterns or func-
tions of the variation. Furthermore, most studies have concentrated on data from the
earlier period, mostly poetic in nature, since available ‘low’ prose texts of that time are
not very common.
 Chrystalla A. Thoma

Our aim with this study is to concentrate on the phenomenon in Early Modern
Greek, offering new data and a new perspective to the ongoing discussion. We claim
that in order to understand the diachronic mechanisms of language change we need to
give more attention to this (immediately intermediate) stage of the language between
Late Medieval and Modern Greek. We have little data as to the linguistic situation of
this time,3 a situation partly amended through this paper. Hereby we present a quanti-
tative and qualitative analysis of the phenomenon in declarative finite clauses in a cor-
pus comprising new texts4 which we tested with statistical methods. The results were
compared to the Escorial Digenis Akritis of the Late Medieval period, the text on which
Mackridge (see below) based his pioneer study of the phenomenon.
A second important contribution of this study is that it offers a first functional
explanation to the phenomenon according to Givón’s Syntax ([1990] 2001). With this
study we are able to offer evidence for the distribution and function of the placement
of the clitic5 object pronoun in Early Modern Greek in a synchronic analysis, but also
a diachronic perspective on the development of its placement and function. We believe
that our study complements other studies on the topic discussed below.

2. State of the art

The variable placement of the clitic object pronoun in Early Modern Greek narratives
has received the attention of a number of scholars who investigated it by means of differ-
ent methods. Mackridge (1993, 2000) was the first one to formulate rules for the phe-
nomenon concentrating on texts from the 11th to the 15th centuries. Pappas (2001,
2004a, b) was the first to undertake a quantitative analysis of the phenomenon on a large
scale and to challenge certain claims concerning the phenomenon in Late Medieval
Greek. He focussed on texts from the 12th-16th centuries, unfortunately poetic texts,6
and in one occasion also compared them to data (from both poetic and prose texts) from
the 17th century. Condoravdi and Kiparsky (2001, 2004) approach the problem from the
point of view of dialect groups of Greek and maintain that Late Medieval Greek appears
to be a union of type A dialect syntax (such as Greek Cypriot, generally postverbal clitic
placement except with focalized elements) and Type B/C dialect syntax (such as Pontic
and Western dialects). Philippaki-Warburton (1995) and Horrocks (1990, 1997) focus
on the generative rules governing clitic placement in Late Medieval Greek, therefore we
shall not discuss their work here (for a critical view see Pappas 2004a: 93–99).
Although Rollo (1989) was the first scholar to write about the phenomenon under
study here, noting a similarity between Italian and Greek in this respect, we shall only
mention his work very briefly for reasons made explicit in the following lines: Rollo
was apparently content with merely offering a list of isolated examples from Early
Modern Greek in which the clitic object pronoun appears in preverbal position (OV)
and the Greek Cypriot dialect in which just the opposite happens (VO). Neither is any
attempt made to offer any explanation as to the possible syntactic or functional rules
Distribution and function of clitic object pronouns in 16th–18th century Greek narratives 

governing its placement, nor are preverbal object pronouns (OV) examined at all (cf.
also Pappas 2004a: 31).7 His work is considered to be pioneer; nevertheless, the topic
has since seen a lot of development and analysis from the scholars mentioned above to
whom we turn our attention below.

2.1 The beginning: Mackridge (1993, 2000)

Mackridge mentions that he drew his inspiration from a book by Newton (1972) on
the Greek Cypriot dialect which prompted him to formulate rules for the Early Mod-
ern Greek clitic object pronoun placement.8 His belief was that the rules followed in
medieval prose texts we have from Cyprus (mainly Makhairas’s Chronicle), as well as
in the modern dialect of Cyprus, are the same as in Early Modern Greek. He proposed
that, until the 15th century, texts follow strict rules of clitic placement, something
which they somehow stop doing later.9
Our intention is to investigate our data keeping in mind Mackridge’s rules, which
were formulated for texts older than the ones under analysis here, in order to see
whether they still apply to Early Modern Greek clitic placement variation. As most
studies take these rules as their starting point and as there is no other existing set of
rules for the phenomenon, we shall largely base the categorisation of our data on them.
Our aim is to offer a quantitative and qualitative analysis in search of consistent use
patterns of the variable position of clitic object pronouns in Early Modern Greek.
However, we also examine the text Mackridge analyses (the Digenis Akritis verse epic,
the Escorial version) in order to compare our results with a text of the Late Medieval
period in search of similarities and differences.
Concerning the Greek Cypriot dialect (in the Late Medieval period as well as today),
we, along with other scholars (discussed below), disagree that the rules concerning the
phenomenon in question are the same as in Late Medieval or Early Modern Greek. Greek
Cypriot represents in all probability an earlier stage in language, also in the domain of
unstressed object pronoun placement (cf. also Condoravdi/Kiparsky 2001,10 see also
Pappas 2004a: 125) from which Standard Medieval Greek began diverging already at the
time the Escorial Digenis Akritis epic was produced (see Discussion below). In Cypriot
Greek, unstressed object pronouns are preverbal only in cases of clearly contrastive top-
ics such as finite declarative clauses with strong subject pronouns or focalised NPs, ob-
jects and adverbs, or in cases of negation, interrogative, subordination and the subjunc-
tive. If we accept that the Escorial Digenis Akritis indeed originated in the 12th century,
then the divergence must have taken place even before that time.
We do not wish to reproduce here the list of rules proposed by Mackridge as gov-
erning the placement of unstressed object pronouns in Late Medieval Greek (Mack-
ridge 1993: 325). Let us just mention that according to him, postverbal is the default
word order, with the verb usually at clause-initial position and following certain con-
junctions and complementizers, as well as when preceded by an object to which the
unstressed pronoun refers (object is fronted and the pronoun serves a doubling func-
 Chrystalla A. Thoma

tion). The order is postverbal with certain elements such as the subjunctive, concessive
and future particles να (na), ας (as), θα (tha), the negative particles μη (mi), μηδέν
(miden), ουδέν (ouden), δεν (den) ‘not’, interrogative pronouns or adverbs, and se-
mantically and syntactically focalised words or phrases (non-temporal adverbs, nouns
or adjectives in an object noun phrase, subject or object complements), while the
placement remains variable after subjects or temporal adverbs.
Interestingly enough, Mackridge notes that the Escorial Digenis Akritis is the only text
that consistently follows the rules.11 This of course poses intriguing questions as to the ab-
solute force of these rules. Since we take a variationist approach to language, we propose
that variation is to be expected in language and that, as we discuss below, variation can have
a functional role which, we believe, it already played in Late Medieval Greek.

2.2 The quantitative dimension: Pappas

Pappas (2001, 2004a,b) gives a quantitative analysis with the purpose of re-examining
and describing the phenomenon and adding a criticism to some previous proposals.
Unfortunately Pappas not only uses mainly poetic works due to difficulties in finding
prose texts of the period he examines (12th to 16th centuries), admittedly a problem if
one wants to say something about the language of the period (cf. also Chila-Marko-
poulou 2004), but also editions that, to a great extent, are outdated or contain major
editorial changes.12
Pappas offers the only statistical analysis of the phenomenon so far. He finds no
statistical difference between fronted and focussed elements in determining clitic pro-
noun placement and remarks that it is in fact impossible in the texts under study to
effectively tell them apart.13 Furthermore, in his statistical analysis, he finds no signifi-
cant difference between the factors “preverbal subject” and “fronted temporal adverb”,
possible factors influencing word order proposed by Mackridge (1993). He however
finds a tendency for OV with preverbal subjects, which he sees as an epiphenomenon
of metrical constraints, a problem particular to poetic discourse and a plausible expla-
nation in view of the corpus he uses. In general, though, he tends to reaffirm Mack-
ridge’s rules, although he states that they point to “poetic convention”, being therefore
“stylistic” and not rules of the language as such; a claim however that he has not been
able to prove (cf. again Chila-Markopoulou 2004).

2.3 Condoravdi and Kiparsky (2001, 2004)

Condoravdi and Kiparsky examine the phenomenon under study in the generative
grammar framework. They focus on a theory of Greek dialects and their differences
which, as they propose, explain the particularities of Late Medieval Greek. As men-
tioned above, they propose that Late Medieval Greek is the result of a mixture of type
A and type B/C dialects. We see absolutely no contradiction between their study and
Distribution and function of clitic object pronouns in 16th–18th century Greek narratives 

ours; indeed, we believe that they complement each other very well, as do the findings
of the other studies mentioned above. Condoravdi and Kiparsky endorse the view that
preverbal placement of the unstressed object pronoun depends on fronted focalized
elements,14 a view not shared by Pappas as we have mentioned above. Indeed, as Con-
doravdi and Kiparsky (2004: 167) go on to mention, “a constituent may be focussed if
the speaker or writer thinks of it as contributing a particularly noteworthy or surpris-
ing piece of information, or wishes to represent it as such, but in the absence of enough
syntactic and intonational cues one would have to be a mind-reader to predict when
that is the case.” The contexts in which the reading is focussed beyond any doubt are
not very common in written texts.

3. Type of data and methodology

For this study we have opted for the analysis of 3rd person clitic object pronouns. We
only analysed narrative parts, to the exclusion of dialogue, examining only finite de-
clarative clauses in order to restrict our analysis. No particular attention was paid to
the already relatively fixed word order in the cases of:
OV: (δ)εν (den) ‘not’, temporal clauses, ότι (oti) ‘that’ and διότι (dioti) ‘because’,
the relative pronoun όστις (ostis) ‘who’, and all subjunctive forms such as ας
(as), να (na), θα (tha)‚ and μην (min).
VO: participles, (the rare cases of) infinitives, imperatives
In Mackridge’s list of rules, in finite clauses, focalised elements apparently correlate
with preverbal clitic object pronoun placement, while absence thereof (with the excep-
tion of fronted objects) correlates with postverbal clitic object pronoun placement.
With preverbal subjects and fronted adverbs there apparently exists variation in the
placement. From this list it is easy to see why scholars who have based their studies on
Mackridge’s work are convinced that there is no pattern in the correlations. With this
set of possible correlations in mind, we have chosen to examine the correlations of
placement, in our texts, with:
a. Clause-initial verbs,
b. Preverbal subjects,
c. Preverbal adverbs,
d Preverbal objects.
A point that needs to be discussed, albeit briefly, is Mackridge’s claim that there is
variation in the placement of the pronoun after temporal adverbs. This is a tricky point,
as Condoravdi and Kiparsky (2004) also mention, but in this case for different reasons
to the ones they name. In narrative and especially oral narrative research, one-word
temporal adverbials, such as τότε (tote) ‘then’, ευθύς/παρευθύς (efthis/parefthis) ‘im-
mediately/then’ etc., are termed “discourse markers”.
 Chrystalla A. Thoma

Discourse markers are “usually lexical expressions, (which) do not contribute to


the propositional content of a sentence but signal different kinds of messages” (Fraser
1999: 936). They can be temporal (e.g. ‘then’), coordinating (e.g. ‘and’) or conjoining
(e.g. ‘well’); characteristically, their propositional meaning has been attenuated over
time due to a process of grammaticalisation (Georgakopoulou 1997: 93). The role of
temporal discourse markers is not so much to anchor any statement in time, since
paratactically joined clauses in narrative are expected to follow each other chronologi-
cally (Berman 1988: 473) as to express subjectivity. Their function is very different to
that of long temporal fronted adverbials which give a clear temporal line to the text.
These one-word temporal discourse markers show continuity in the same sense that
the additive marker και (ke) ‘and’ does: they add similar, non-exceptional information.
Thus their distribution between the two word orders is mixed. Temporal discourse
markers were therefore not accounted for in our counting of fronted adverbs.
The first part of the analysis is quantitative; it consists in a search of trends and
patterns in the placing of the pronouns in the chosen texts (given below). For the sta-
tistical analysis, Student’s t-Test was used, as the results tested belonged to two distinct
categories (i.e. preverbal and postverbal clitic object pronoun placement).

4. Texts under analysis

As mentioned in section 2 (“State of the art”), other studies have made use of texts
which present some serious problems for the analysis: They are poetic, answering
therefore to metrical constraints, and they have undergone major changes by editors.
Our aim is to avoid these problems by using a corpus of narrative ‘low’ (i.e. popular)
texts that fulfil the requirements of (1) being prose and (2) having undergone virtually
no editorial changes. Furthermore, we tried to examine enough texts in order to avoid
findings that reflect an author’s personal style or one scribe’s errors. Although they are
not as numerous as the texts analysed in Pappas (2004a), they sufficed for significance
tests to be run, therefore representing a sufficient minimum. Most of these texts are
brought together and examined in this perspective for the first time, thus offering new
data to the ongoing discussion.
The chosen texts consist of three ‘low’ versions of the Life of Aesop (D:1644, I: 1617
and K: 1600, edition in progress by Hans Eideneier), Maximos Kallioupolitis’s version
of Luke’s Gospel (printed 1638) (Kasdaglis 1999), and part of an anthology by G. Ke-
hayoglou (Kehayoglou 2001) which we used as one long text in order to see general
tendencies (Ioannis Morezinos’s religious narrative about the Virgin Mary, Nikos
Gabrielopoulos’ Chronicle about the plague of 1688, Ierotheos Avvatios’ chronicle
about the earthquake of 1648, Dimitrios Pirris narrative about the plague of 1728,
Papa-Lavrentios’ personal narrative from 1580, Parthenios Peloponisios’ religious nar-
rative printed 1765, the anonymous narrative about the destruction of the island San-
torini from the year 1669, the anonymous vitae of the blessed Josef Pangalos, Ioannis
Distribution and function of clitic object pronouns in 16th–18th century Greek narratives 

Eleimon and Ieromartyr Vlassios probably from the 17th century, the family chronicle
of Konstantinos Theodosis from the end of the 1400s, Ioannikios Kartanos’ religious
narrative printed 1536, Agapios Landos’ religious narrative printed 1664), as well as
Papasynadinos’s memoirs from the mid 1600s (Odorico 1996). We have also included
the verse epic Digenis Akritis, the Escorial version (Alexiou 1985), probably composed
in the 12th but transmitted in a single manuscript of the 15th c., in order to compare
the phenomenon with that of the previous era.

5. Results

In this section we give the findings from our quantitative investigation of the phenom-
enon, examining each of the four factors of influence on the placement of the clitic
object pronoun mentioned in section 3 (“Type of Data and Methodology”).
In Table 1 and Figure 1 we give the variation in placement of the object clitic pro-
noun when the verb is in initial (thematic) position in the clause, either with elided or
postverbal subject. This correlation is only strong in the case of Digenis Akritis Escorial
version, the one text given by Mackridge as consistent in following the list of rules he
sets up, and the three ‘low’ versions of the Life of Aesop.15 However, when our corpus
was tested as a whole with Student’s t-Test, no significant variation was found in the
distribution of the two word orders with clause-initial verb.

Table 1.  Clause-initial verb (tokens), in the first four texts very
significant variation (** p<0.01)

Clause-initial verb VO OV

Aesop D 180 ** 7
Aesop I 164 ** 25
Aesop K 161 ** 32
Digenis Akritis 54 ** 9

Kallioupolitis: Luke 99 104


Anthology Kehayioglou 154 165
Papasynadinos Memoirs 17 37
 Chrystalla A. Thoma

Figure 1.  Percentages of clause-initial verb in relation to the placement of the clitic object
pronoun before (OV) or after (VO) the finite verb

The second step, as given in Tables 2, 3 and 4 (below), was to examine the correlation
in placement of clitic object pronouns with fronted, preverbal elements.
In Table 2 and Figure 2 we give the correlation of clitic object pronoun placement
with preverbal subject, as illustrated in example 4. Student’s t-Test showed that prever-
bal pronouns correlated very significantly with preverbal subjects.16

Table 2.  Overt subject in preverbal position (tokens),


*significant (p<0.05)

Preverbal subject VO OV *

Aesop D 7 3
Aesop I 4 5
Aesop K 6 54
Digenis Akritis 2 16
Kallioupolitis: Luke 2 85
Anthology Kehayioglou 5 69
Papasynadinos Memoirs 1 19
Distribution and function of clitic object pronouns in 16th–18th century Greek narratives 

Figure 2.  Percentages of overt subject in preverbal position in relation to the placement of
the clitic object pronoun before (OV) or after (VO) the finite verb

As shown in Table 3 and Figure 3, preverbal pronouns correlated very significantly with
preverbal objects with doubling pronouns.17 The results of preverbal direct object pro-
nouns were pooled with preverbal indirect object pronouns in genitive and the rare
cases of non-doubling pronouns following a preverbal object (referring to another ele-
ment). The latter type occurred almost exclusively in Digenis Akritis, (5 cases with OV
and 1 with VO); the reading of these cases, which are probably due to the poetic nature
of the text, is not necessarily one of focalisation.18 Not only does this method not affect
in any way the significance readings but it is also in accordance with the position that in
our texts the placement of the clitic pronoun depends on the position of the fronted ele-
ment and not on its being focalised (cf. also Pappas 2004a: 38 for a similar statement).
 Chrystalla A. Thoma

Table 3.  Preverbal object with doubling pronoun (tokens),


*significant (p<0.05)

Fronted object VO OV *

Aesop D 0 1
Aesop I 0 5
Aesop K 2 1
Digenis Akritis 1 21
Kallioupolitis: Luke 0 3
Anthology Kehayioglou 0 24
Papasynadinos Memoirs 0 4

Figure 3.  Percentages of preverbal object with doubling pronoun in relation to the place-
ment of the clitic object pronoun before (OV) or after (VO) the finite verb

As shown in Table 4 and Figure 4, when tested for significance with Student’s t-Test,
the predominant preverbal placement of the clitic object pronoun with fronted ad-
verbs was found to be significantly more frequent.19 As mentioned in the Introduction,
one-word adverbs, in reality discourse markers marking continuity such as “then”,
were not counted among the adverbs in this table and were taken to have the same
function as coordinating conjunctions such as “and” (cf. also Condoravdi and Kipar-
sky 2001). Indeed, their distribution is similar to that of coordinating conjunctions
with clause-initial verbs.
Distribution and function of clitic object pronouns in 16th–18th century Greek narratives 

Table 4.  Totals of correlation with fronted adverb,


*significant (p<0.05)

Fronted adverb VO OV*

Aesop D 3 8
Aesop I 7 4
Aesop K 1 7
Digenis Akritis 1 47
Kallioupolitis: Luke 0 6
Anthology Kehayioglou 1 44
Papasynadinos Memoirs 3 17

Figure 4.  Percentages of fronted adverb in relation to the placement of the clitic object
pronoun before (OV) and after (VO) the finite verb

Finally, we pooled together all the above discontinuity markers (overt subject in pre-
verbal position, preverbal adverb and object) and counted the occurrences, as given in
Table 5 and Figure 5. As expected, the tendency to have preverbal object pronouns was
very strong, as a Student’s t-Test showed.20
 Chrystalla A. Thoma

Table 5.  Discontinuity markers (tokens),


**very significant (p<0.01)

Discontinuity marker VO OV **

Aesop D 10 12
Aesop I 11 14
Aesop K 9 62
Digenis Akritis 4 84
Kallioupolitis: Luke 2 94
Anthology Kehayioglou 6 147
Papasynadinos Memoirs 4 40

Figure 5.  Percentages of discontinuity markers in relation to the placement of the clitic
object pronoun before (OV) and after (VO) the finite verb

6. Discussion

6.1 Placement patterns

As seen in section 5, Results, there is variation in the placement of the clitic object
pronoun when the verb is in initial (thematic) position in the clause, either with elided
or postverbal subject (given in Table 1 of Results). This correlation, as given in the list
of Mackridge’s rules, was expected to be very strong. Indeed, the tendency is strong in
the one text Mackridge presents as consistently following the rules, the Escorial Di-
Distribution and function of clitic object pronouns in 16th–18th century Greek narratives 

genis Akritis, and in the three ‘low’ versions (D, I and K) of one text, i.e. the Life of
Aesop. However, when the whole corpus used for this study was tested, no significant
tendency was verified. This finding indicates that older texts such as the Escorial Di-
genis Akritis probably followed this rule quite consistently. This is in fact corroborated
by the same strong tendency found in the three versions of the Life of Aesop, which are
all Early Modern Greek adaptations of older, linguistically more archaic narratives
(Notis Toufexis, pers. comm.). If the tendency we see is not the author’s personal style,
then it can be taken to be a remnant of an older feature. However, in our corpus of
Early Modern Greek texts no significant tendency was found supporting this rule, sug-
gesting that an important change was already underway: the position of the object
clitic was at the time not yet consistently preverbal; still, neither position was preferred
over the other. This finding is in direct contrast with Pappas’s findings (Pappas 2004a:
118) for this factor. In his counting, in the beginning of the 16th century we already
find a relatively fixed preverbal placement of the clitic when the verb is clause-initial, a
fact not corroborated by our texts.
An instance from the corpus is given in example (1), taken from Digenis Akritis, in
which the clitic object pronoun ‘το’ refers to the word θαμπούριν ‘tamboura’ in the previ-
ous clause and is postverbal. As described by Mackridge and also as the findings of this
study reconfirm, this is typical in this text when the verb occurs at thematic position or
when preceded by a coordinating conjunction such as (in this case) και (ke) ‘and’.
(1) Digenis Akritis Escorial (15. century)
(Και επήρεν το θαμπούριν του) και αποκατάστησέν το
(And he took his tamboura) and string-3SG-PAST it-PRO-ACC
And he took his tamboura and strung it
In contrast, in Kallioupolitis translation of the Gospel of Luke from Hellenistic into
early Modern Greek, we find great variation in the placement of the clitic object pro-
noun. In example (2) we see a sentence in which all placements occur postverbally
with the verb at clause initial position.
(2) Kallioupolitis: Gospel of Luke (1638)
Και έφερέ τον εις την Ιερουσαλήμ και έστησέν τον απάνου εις την άκρην του
ιερού και είπε τον:
And bring-3SG-PAST him-PRO-ACC to Jerusalem and place-3SG-PAST
him-PRO-ACC on the edge of the altar and say-3SG-PAST- him-PRO-ACC:
And he brought him to Jerusalem and placed him on the edge of the altar and
said to him:
In example (3) from the same text and by the same author/translator, in an apparently
comparable co-text, the clitic object pronoun τον occurs preverbally.
(3) Kallioupolitis: Gospel of Luke (1638)
Και τον έδωκαν βιβλίον του Ησαίου του προφήτου·
 Chrystalla A. Thoma

And him-PRO-ACC give-3PL-PAST book of Isaiah the prophet;


And they gave him the book of Isaiah the prophet;
We then showed that the factor “preverbal subject” correlates significantly with proc-
litic object pronouns (as given in Table 2 of Results). Mackridge (2000: 133) mentions
that further investigation is needed in this respect.21 Although Pappas (2001) mentions
numbers that speak of the same tendency in his data, he ascribes it to poetic con-
straints. Indeed, in his corpus analysis he only found limited evidence of such a ten-
dency even though his core corpus reaches into the 16th century, thus already into
Early Modern Greek (Pappas 2004a: 42). In our corpus the tendency is significant; the
finding is also valid for Digenis Akritis, indicating that already in Late Medieval Greek
this factor played an important role in the case of preverbal placement of the object
clitic. A functional explanation is given below (section 6.2).22
Examples (4), (5) and (6) illustrate this significant tendency in our corpus. The
preverbal subjects are not focussed (i.e. not contrastive) and this is the norm of prever-
bal subjects in our corpus (cf. also Pappas 2004b)23 despite the fact that the position is
still not fixed.
(4) Life of Aesop K (1600)
Και ο δεσπότης τον λέγει
And the master-NOM him-PRO-ACC say-3SG-PRESENT
And the master tells him
(5) Vita of the blessed Josef Pangalos (ca. 17th century)
Και ο άγιος τα εδέχθη και έφαγε
And the saint them-PRO-ACC accept-3SG-PAST- and eat-3SG-PAST
And the saint accepted and ate them
(6) Papasynadinos`s memoirs (1600s)
Και οι τούρκοι το εγύρισαν αλλέως, το πως είπεν τον τούρκον, οπού τα
πουλεί, χριστιανόν
And the Turks-NOM it-PRO-ACC turn-3PL -PAST differently, it-PRO-ACC
that say-3SG-PAST Turk-ACC, who sell-3SG-PAST them-PRO-ACC, Chris-
tian-ACC
And the Turks distorted it, saying that he had called the Turk who sold them
a Christian
Preverbal object pronouns correlated significantly with preverbal object with (dou-
bling) pronoun (as shown in Table 3 of Results). This is important because preverbal
objects are rarely focussed24 thus dispelling the idea that focusing is the main factor in
preverbal placement of the clitic pronoun already in Late Medieval Greek. Pappas
(2004a: 42–43) also found the same tendency for preverbal placement, although in a
previous study of his (Pappas 2001: 81) he found that in “most of the instances in
which the ‘doubling pronoun’ appears preverbally, the doubled element is some form
of the adjective όλος.” Still, his latest view, corroborated by our findings, suggests that
Distribution and function of clitic object pronouns in 16th–18th century Greek narratives 

already in Late Medieval Greek the pattern existed and was retained through Early
Modern to Modern Greek, illustrated here by examples (7), (8), (9), and (10).
(7) Kartanos (1536)
Την τετάρτην την είχαν δια τα ζώα
The-ART-DEF-FEM-ACC Wednesday-ACC it-PRO-FEM-ACC have-3PL-
IMPERFECT for the animals-PL-ACC
As for Wednesday, they had it for the animals
(8) Gabrielopoulos (1688)
Τούτον τον λογισμόν τους τον έβαλαν και εις πράξην
This-PRO-MASC-ACC the-ART-MASC-ACC thought-ACC it-PRO-MASC-
ACC put-3PL-PAST and in action-ACC
This thought of theirs, they translated it into action
(9) Ioannis Morezinos’s religious narrative about the Virgin Mary (ca. 17th cen-
tury)
Τούτα τα είπεν εις όλους την νύκταν απού την επήρασιν
These-PRO–ACC-PL them-PRO-ACC say-3SG-PAST to all-ACC the night-
ACC that her-ACC take-3PL-PAST
These, she talked about them to all the night that they took her
(10) Kallioupolitis: Gospel of Luke (1638 )
Και πολλούς τυφλούς τους εχάρισε το φως τους
And many-ACC-PL blind-ACC-PL them-PRO-ACC give-3SG-PAST their
light-ACC
And to many blind, he gave to them their sight
The correlation of preverbal object pronouns with preverbal adverbs was also found to
be significant (as given in Table 4 of Results). This difference in our findings from the
findings of other scholars (for example Pappas 2004b) is probably due to the fact that
we pooled together all adverbs since we propose that any preverbal adverb can and
does affect the placement of the clitic. Furthermore, since we filtered temporal dis-
course markers such as “then”, the picture changed even more, finally resulting in a
strong tendency. Examples (11), (12), (13) and (14) illustrate the patterning of a pre-
verbal clitic object pronoun following a preverbal adverb.
(11) Life of Aesop I (1617)
Πολλά του εβαρυφάνη
Much-ADV him-PRO-MASC-GEN (heavy appear)-PAST
He took it to heart
(12) Kallioupolitis: Gospel of Luke (1638)
Και με τα μαλλία της κεφαλής της τα εσφούγγιζε
And with the hair-ACC of her head them-PRO-ACC-PL wipe-3PL-IMPERFECT
And she wiped them with her hair
 Chrystalla A. Thoma

(13) Ioannis Morezinos’s religious narrative about the Virgin Mary (ca. 17th cen-
tury)
Και τυφλώνουνται από τον εχθρόν, και μίαν ημέραν τηνε πιάνουσι και
λέγουσίν της
And blind-3PL-PR-PASSIVE by the enemy-ACC, and one day her-PRO-ACC
take-3PL-PRESENT and say-3PL-PRESENT her-PRO-GEN
And they are blinded by the enemy and one day they take her and tell her
(14) Kallioupolitis: Gospel of Luke (1638)
Και από τα κεραμίδια τον εκατάβασαν μαζί με το κρεβάτι εις την μέσην και
έβαλάν τον μπροστά εις τον Ιησούν
And from the roof tiles-ACC him-PRO-ACC lower-3PL-PAST with the bed-
ACC in the middle-ACC and place-3PL-PAST him-PRO-ACC before Jesus-
ACC
And from the roof tiles they lowered him with his bed in the middle and
placed him before Jesus
Supported by our findings given in Table 5 (cf. “Results”), a relatively clear picture of
the placement of the clitic object pronoun in Early Modern Greek emerges in our
study. Preverbal placement appears with highly significant frequency with preverbal
(fronted) elements, which in their majority appear not to be focussed (see Pappas
2004a: 118 for similar findings in a brief overview of 17th century texts). These prever-
bal elements mentioned above correspond to discontinuity markers in Givón’s (1983,
1992) model of referential cohesion in narrative discourse and are supposed to have a
functional role: they indicate junctures in the narrative line which correspond to “ma-
jor topics”. Givón’s theory as it concerns our findings and the question of the func-
tional role of clitic object placement pronoun is examined below (section 6.2).
Our findings present another perspective on the ongoing discussion concerning
the factors influencing the placement of unstressed object pronouns in earlier periods
of Greek. Let us now see if the patterns presented above fit into a functional, both syn-
chronic and diachronic, framework.

6.2 A functional synchronic explanation

Givón is one of the main representatives of the school of Typological discourse analy-


sis (TDA). TDA pays particular attention to word order phenomena in a variety of
languages and connects them to information distribution in discourse and to iconicity
principles inside the boundaries of the clause, and in discourse (interclausally and in-
tersententially). TDA pays close attention to referent tracking mechanisms and epi-
sodic memory retrieval and activation. According to Givón (1992: 6) “the grammar of
referential coherence is not primarily about reference. Rather it is about identifying
and activating the locations (“mental files”, “storage nodes”) where verbally coded text
is stored in episodic memory.”
Distribution and function of clitic object pronouns in 16th–18th century Greek narratives 

Narrative discourse is about “referents”, entities stable in time which can be topical
(nominal) or adverbial (Givón 1979: 320–324). Referential cohesion tracks these refer-
ents and brings them back into the foreground when needed after having been relayed
to a backgrounded position for a while or when the mention of other referents inter-
feres and makes their mental retrieval difficult. The junctures in discourse where ref-
erents are brought back into the foreground are marked by long fronted discontinuity
elements, i.e. nominal and adverbial groups mostly, certainly not pronouns and espe-
cially not elements for which mental retrieval is the easiest such as clitic pronouns.
According to Givón (2001), in discourse we find chains of ‘primary’ (subjects) and
‘secondary’ (objects) topics. They are tracked according to the principles of iconicity
and referential cohesion. The maximum degree of continuity (facility of retrieval) for
subjects corresponds to zero anaphora, whereas for objects it corresponds to clitic pro-
nouns (in Greek at least), as illustrated here for Early Modern Greek in an example
from The Life of Aesop (also valid for today’s Modern Greek).
Continuity (accessible topics):
(Ø) ήφερεν τα
(no subject) (verb) (clitic pronoun)
(He) brought them
Discontinuity (non-accessible/salient topics):
Κάποιος γεωργός ήφερεν σύκα
(explicit subject) (verb) (explicit object)
A farmer brought figs
We hereby propose a cohesive function of the variation in pronoun placement, an ex-
tra dimension added to the already cohesive referential nature of clitic pronouns as
continuity markers of secondary topic cohesion, as illustrated above. As we have seen,
preverbal clitic object pronouns co-occur with fronted elements in our Early Modern
Greek corpus. The larger the fronted element, the more important the break in dis-
course, and the greater the processing difficulty and mental effort, creating therefore
discontinuity junctures. Given by Givón as the “code quantity principle”, the definition
is as follows: “The less predictable the information is, or the more important, the more
prominent, distinct or large will be the code element(s) that convey it” (Givón 2001:
249). The association of this principle with word order becomes clear in the following
statement: “the less predictable the information is or the more important, the more
likely it is to be placed earlier in the clause (or in whatever relevant unit of structured
information)” (Givón 2001: 250).
We suggest that the function of preverbal placement of the clitic object pronoun is
to intensify the effect of a discontinuity, or break, in discourse. The placement follows not
only prosodic reasons (although this might also be the case), but certainly pragmatic
reasons: the preverbal placing of the clitic, which causes it to appear between the fronted,
discontinuous element and the verbal group, effectively demarcates the former and plac-
 Chrystalla A. Thoma

es it apart, giving it more emphasis. The relation to memory is defined as follows: “more
prominent and more distinct coding attracts more attention”, and “information that at-
tracts more attention is memorized, stored and retrieved more efficiently” (Givón 2001:
250). The demarcation is not only effectuated by means of prosody, since the clitic is
unaccented, but also by means of a “low” point in information flow: the clitic pronoun is
the entity that is easiest to retrieve from the previous clause or a clause that occurred not
far back in discourse. The author takes for granted the fact that the reader/listener will
not find it difficult to infer which referent is being talked about. With all the attention of
the reader/listener already turned to the initial, longer and information-loaded element,
a “break” in the reader/listener’s attention is helpful so that another element might follow
with new information: a verb perhaps, or an object.
We propose, therefore, that by means of syntactic and prosodic management, dis-
continuous elements (preverbal subjects, objects and adverbs) are made to stand out,
creating clear junctures in discourse, a function in which the placement of clitic object
pronouns appears to play an important role. This functional role was possible due to
the variation in the placement of the clitic for a long period of time until late in the
history of the Greek language; indeed, as our study clearly shows, important variation
exists at least until the 18th century, as opposed to Pappas’s (2004a: 117) claim (cf. also
Chila-Markopoulou 2004 for the same criticism), a claim perhaps based on his choice
of poetic texts.

6.3 Diachronic perspective on clitic placement

The lack of stress and placement of clitic pronouns in the clause are very probably in-
tertwined. For some time it has been known that certain positions in the clause cor-
relate with the placement of clitics: mainly clause-second and postverbal (cf. Janse
2000: 237). The first case (clause-second placement) has received most of the attention.
The famous Wackernagel’s law, which proposed that the clause-second position is re-
served for clitic elements, is quoted in the majority of studies on clitics and is con-
cerned with exactly this position in the clause. This has since been interpreted as “the
first position is reserved for stressed elements”. Now, as Janse (2000: 234) accurately
observes, the stress on the first position is closely linked to discourse-pragmatic rea-
sons. It seems that prosody, like word order, follows mainly pragmatic reasons.25 The
first constituent of a clause is sometimes focalised (contrastive), and for this reason
receives the strongest stress in the clause. Most often, though, the literature does not
mention anything about the stress of preverbal, non-focalised elements. Yet it is clear
that they also carry stress, albeit not as strong as the one of focalised elements. The
term “unstressed pronoun” emphasises just that: other elements in the clause carry the
stress. The placement of the clitic right after preverbal elements follows, apart from
discourse-pragmatic, also prosodic reasons, an opinion shared by Condoravdi and Ki-
parsky (2004: 175), albeit, in their view, due to syntactic reasons.
Distribution and function of clitic object pronouns in 16th–18th century Greek narratives 

Now we shall turn our attention to the second option, the postverbal placement. As
Givón (2001: 423) points out, weak elements often cliticise on the verb because it is the
element that is usually present in a clause; participants such as subjects and objects are
often elided in cases of referential continuity. Verbs are hardly ever focalised in Greek,
except perhaps in cases of imperative or subjunctive expressing command, and they do
not carry strong stress, even if they are at thematic position in the clause. But verbs do
carry stress, especially if they stand at clause-initial position with no preverbal, fronted
element present. That must be the reason behind the placement of the clitic pronouns in
postverbal position; see also Condoravdi and Kiparsky (2004: 179).26
Essentially the first concern of word order is to serve discourse cohesion and
therefore the episodic memory (Givón 1992: 5 et passim, for Ancient Greek see Dik
1994, 1995, for Modern Greek see Thoma 2006). Clause prosody, just like word order,
serves the same function and places certain groups of grammatical words in certain
positions in the clause; in those positions they do not disturb the function of lexical
words carrying most of the informational content and cohesive power of the clause in
discourse. These positions become later grammaticalised and fixed (as it happened in
German and English for instance) and over time lose their discourse-pragmatic power
in order to become clitics and affixes, resulting in a renewal of the circle. This “fixing of
discourse strategies” (Traugott 1994) allows for new elements to take over the dis-
course-pragmatic role.
This development of clitic pronouns seems to mirror the development of Greek
pronouns from ancient Greek to this day. In ancient Greek we find clitics that tend to
appear in second position in the clause, enclitic on the first constituent (Horrocks
1990: 36, Janse 2000: 233). The situation apparently changes in Hellenistic Greek, when
clitic pronouns (1st and 2nd person, still no 3rd person clitic pronouns) get attached
to the verbal group as a result of reanalysis (Horrocks 1990: 41), and appear postver-
bally even in cases of focalised fronted constituents, probably as a result of language
contact with Aramaic and Hebrew, languages which have pronominal suffixes instead
of clitic pronouns (Janse 2000: 237). In Medieval Greek we see the emergence of 3rd
person clitic object pronouns (Mackridge 2000: 135). As Mackridge (1993: 225) aptly
remarks, the placement of clitic pronouns is not really ‘free’. There is a very real possi-
bility that the placement initially followed a pattern similar to today’s Greek Cypriot
dialect, in which finite declarative clauses present fixed postverbal placement except in
the case of focalised (contrastive) fronted constituents which, for prosodic reasons,
attract the clitic to preverbal position. The proposal for this phase in the development
of the placement of clitic object pronouns will remain for the time being tentative,
since neither is there any proof that postverbal placement became fixed at any one
point, offering ground for comparison with Greek Cypriot, nor that preverbal place-
ment correlated at any given time in the history of Standard Greek only with focalised
constituents. In Early Modern Greek, fronted elements, not necessarily contrastive,
attract clitic pronouns to preverbal position, until that position becomes the norm and
is grammaticalised in Standard Modern Greek. A similar process has taken place in
 Chrystalla A. Thoma

the Romance languages, most probably due to a similar process (cf. Martins 1994 for
Portuguese, Rivero 1991 and Fontana 1993 for Spanish, cf. also Pappas 2004a).

6.4 Possible causes for patterning OV with clause-initial verb

In today’s Standard Modern Greek (in contrast to dialects such as Greek Cypriot), clitic
object pronouns have become strictly proclitic on the verb, except in the case of the im-
perative or participle. In our Early Modern Greek corpus, the placement of the clitic
object pronoun after clause-initial verb is variable. However, in certain, especially older
texts, there is a tendency for the clitic pronoun to follow the clause-initial verb (see sub-
section 6.1), while in other, more recent texts we find a tendency to have the clitic pro-
noun come before the verb. If we accept that the first placement (postverbal) is the older
one and that preverbal placement originated in cases of focalised and then fronted ele-
ments, the question remains why we find such variation with clause-initial verbs, a vari-
ation which obviously led to today’s relatively fixed proclitic placement. We shall propose
here very briefly two possible reasons:
1 Analogy (“echo structures”, Janse 1994: 439, 1998: 262): There is the possibility of
patterning the clitic object pronoun after the first clause of clause complex, as in
example (15):
(15) Kallioupolitis: Gospel of Luke 1638
Και ο Ιησούς τον επίασε και τον ιάτρευσε και τον απόλυσε
And Jesus-NOM him-PRO-ACC take-3SG-PAST, him-PRO-ACC heal-3SG-
PAST and him-PRO-ACC let go-3SG-PAST
And Jesus took him, and healed him and let him go
2 Hypercorrection: There is the possibility of patterning the clitic object pronoun
after the subjunctive structure which is always used with preverbal pronouns
when the declarative clause can be interpreted as such:
(16) Pirris (1728)
Έτσι έδωκαν θέλημα και την έθαψαν (= έτσι έδωκαν θέλημα να τη θάψουν)
Thus give-3PL-PAST order-ACC and her-PRO-ACC bury-3PL-PAST
Thus they ordered and they buried her (= that they should bury her)
The above are indications of the process that finally led to the grammaticalisation of
the preverbal position of the clitic object pronoun in Standard Modern Greek.

7. Conclusions

The findings of this study can be summarised as follows: In the time period under in-
vestigation, clitic object pronouns appear quite consistently in postverbal position
Distribution and function of clitic object pronouns in 16th–18th century Greek narratives 

with verb in clause-initial position in some texts which probably originated in the
earlier period; otherwise, the placement appears to be random, indicative of a period
of change. A significant correlation was shown to exist between preverbal clitic object
pronouns and discontinuity elements (preverbal subjects, objects and adverbs) in
clause-initial position. We propose that this correlation can be explained as due to the
function of the position of clitic object pronouns which consists in allowing the de-
marcation of stressed, emphasized chunks of information (mostly topics, i.e. nominal
or adverbial groups but also very often verbs in initial position) which indicate junc-
tures of new episodes by re-introducing central referents of a narrative line. Due to
their placement and lack of stress and information load, clitic pronouns enable the
reader/listener to manage information in the clause better, by creating a “low” in infor-
mation, reflected also in their lack of stress, signalling the end of an important unit, as
well as by creating expectations for more important information to follow. As our find-
ings also show, in Early Modern Greek texts the placement of clitic object pronouns
still varies, especially in cases of verbs at clause-initial position; however in some texts
they do show a predisposition for preverbal placement, foreshadowing their fixed pre-
verbal position in Standard Modern Greek. The findings of this study are reliable be-
cause the editions chosen are without major changes and the texts are prose, avoiding
thus metrical constraints on the placement of the clitics. Moreover, the findings are
also valid for the Late Medieval epic Digenis Akritis.
From the point of view of diachronic change, we propose that postverbal place-
ment of clitic pronouns is probably older (following Wackernagel’s law) than the
preverbal. As in Greek Cypriot and other eastern dialects (as Mackridge already pro-
posed), preverbal placement was probably once exclusive to focalised contrastive con-
structions, such as focalised subjects, objects and adverbs. Later, in the time under
study here, it obviously became the norm with general discontinuity patterns of pre-
verbal subjects, objects and adverbs. This pattern was generalised in the end, resulting
in today’s fixed preverbal placement in Standard Greek, whereas dialects such as Cyp-
riot or Cappadocian retained the older word order, probably due to linguistic isolation
(see Pappas 2004a). We also suggest that the grammaticalisation process of clitisation
in Greek was halted due to the emergence of a written (narrative) register which imi-
tated in many ways spoken language (cf. Eideneier 1999, 2002) and by choosing to
write clitic object pronouns as separate words stopped the cycle which probably would
have led to them being seen as affixes.

Notes

1. This investigation has taken place in the framework of “Research Centre 538: Multilingua-
lism” under the auspices of the German Research Foundation (DFG). Our thanks to Prof. Dr.
Hans Eideneier (Hamburg), Dr. Carlos Jiménez (Bremen), Dr. Notis Toufexis (Cambridge) and
Dr. Ludger Zeevaert (Hamburg) and the anonymous reviewers for their advice and comments.
 Chrystalla A. Thoma

2. According to Horrocks (1997: 291ff.) these texts belong to Early Modern Greek and not to
Late Medieval, a period ending with the fall of Constantinople. However, these are conventional
distinctions; language does not always change immediately or entirely when a new historical
period begins. We thank the anonymous reviewer for their comment.
3. This period has been under intensive scrutiny in the historical project H4: “Forms of Writ-
ten discourse in Byzantine and Modern Greek Diglossia” under the direction of Prof. Dr. Hans
Eideneier at “Research Centre 538: Multilingualism” (University of Hamburg).
4. We thank both anonymous reviewers for their comments on this point. Pappas (2004a)
indeed looks at texts from the 17th century, although his sources are not mentioned and our
findings are not exactly the same. Most texts here are studied for the first time, and their editions
are reliable.
5. Clitic here is understood as bound to the finite verb (cf. Mackridge 1993: 317, Janse 2000:
244) but not as part of it.
6. Pappas (2004a) states that in his main corpus he examines the phenomenon in both Late
Medieval and Early Modern Greek. However, his texts of Early Modern Greek are only three
poetic texts from the beginnings of the 16th century (Pappas 2004a: 21).
7. “Rollo overestimated the importance of postverbal pronoun placement in a number of en-
vironments, most notably those of preceding subject and adverb” (Pappas 2004a: 33).
8. “The realization of the medieval rules came to me after reading Newton (Newton 1972:
64).” (Mackridge 2000: 135).
9. “[...] the rules seem to display a certain instability.” (Mackridge 1993: 326).
10. “[...] the system of the A-type dialects must be the most archaic of the three” […] “the syn-
tax of A-dialects is closest to the medieval Greek system, as sketched out in Mackridge (1993)”.
It is mentioned later on that A-type dialects include Greek Cypriot.
11. “[...] the redactor and/or scribe of the Escorial Digenis Akritas displays a remarkable
consistency in his application of the rules governing the position of the clitic pronoun, in
contrast with the situation prevailing in certain other texts […]” (Mackridge 1993: 338).
12. See also Chila-Markopoulou (2004: 2). It is common knowledge among editors of Late
Medieval and Early Modern Greek texts that 19th c. editions (especially those by W. Wagner
used by Pappas) are unreliable and should only be used with great caution. For a general discus-
sion of modern editorial practices see Eideneier, Moennig and Toufexis (2001).
13. “In fact, without the necessary prosodic information (i.e. information about sentence
stress), this distinction between topic and focus is hard to confirm based on the surrounding
context alone.” (Pappas 2001: 86).
14. “If it is a topic, it occurs with postverbal clitics; if it is a focus, it occurs with preverbal cli-
tics.” (Condoravdi and Kiparsky 2004: 166, here referring to preverbal subjects and adverbials).
15. Significance (one tail): df= 4, p= 0.000050872, t= 2.77644511
16. Significance (one tail): df= 8, p= 0.0135777277, t= -2.69806094
17. Significance (one tail): df= 8, p= 0.01768578, t= -2.52788826
18. “A further complication is that in certain types of presentational sentences, even modern
Greek literary style allows topicalization of objects without clitic doubling” (Condoravdi and
Kiparsky 2004: 167).
Distribution and function of clitic object pronouns in 16th–18th century Greek narratives 

19. Significance (one tail): df= 8, p= 0.0219835339, t= -2.3883769


20. Significance (one tail): df= 14, p= 0.003740025, t= -3.12316895
21. “Another area that needs investigation is the situation after subjects.” (Mackridge 2000: 136).
22. Janse also brings up the possibility of discourse pragmatic reasons governing the preverbal
placement of clitic object pronouns with subject in initial position in two papers (Janse 1994,
1998); in a line of argument similar to that of Mackridge he claims that the reason is that prever-
bal subjects constitute information focus. As Pappas (2004a: 58) points out, this is not shown to
be a sufficient condition for preverbal placement of the clitic; furthermore, he claims, this is
apparently not the case in Late Medieval Greek.
23. “[...] fronted constituents are more likely to be focused elements than subjects“ (Pappas 2004b).
24. This is in reality a case of so-called ‘left dislocation’ of the object (Givón 2001), as in the
English example “the house, he sold it”, in which the object is not New information, only topi-
calised. This also explains why it can not be seen as a focussed construction (see also Tzartzanos
1931 for the same idea, i.e. that doubling pronouns are connected to Given information, as well
as Philippaki-Warburton 1995 and Haberland and van der Auwera 1990 for the connection of
doubling pronouns to Topicality. See discussion of Objects as Secondary Topics in Givón’s mo-
del. We thank the anonymous reviewer for this comment).
25. For a different view on this matter see Revithiadou (2006). We are grateful to the anony-
mous reviewer for bringing this study to our attention.
26. In this sense we take weak object pronouns to be enclitic (so do Condoravdi and Kiparsky
(2004) as opposed to Pappas (2004a).

References

Alexiou, S. (ed.). 1985. Βασίλειος Διγενής Ακρίτης (κατά το χειρόγραφο του Εσκοριάλ) και το
Άσμα του Αρμούρη. Κριτική έκδοση. Athens.
Berman, R. 1988. On the ability to relate events in narrative. Discourse Processes 11: 469–497.
Chila-Markopoulou, D. 2004. Review of Panayotis A. Pappas. Variation and Morphosyntactic
Change in Greek: From clitics to affixes. Journal of Greek Linguistics 5: 199–212.
Condoravdi, C. and Kiparsky, P. 2001. Clitics and clause structure. Journal of Greek Linguistics
2:1–39.
Condoravdi, C. and Kiparsky, P. 2004. Clitics and clause structure: The late medieval Greek
system. Journal of Greek Linguistics 5: 159–180.
Dik, H. 1994. Ancient Greek warfare: A case study in constituent ordering. In Function and
Expression in Functional Grammar, E. Engberg-Pedersen, L. Falster Jakobsen and L. Schack
Rasmussen (eds), 197–213. Berlin: Mouton de Gruyter.
Dik, H. 1995. Word Order in Ancient Greek. A pragmatic account of word order variation in Hero-
dotus. Amsterdam: Gieben.
Eideneier H. 1999. Von Rhapsodie zu Rap: Aspekte der griechischen Sprachgeschichte von Homer
bis heute, Tübingen: Narr.
Eideneier, H., Moennig, U. and Toufexis, Ν. (eds.). 2001. Θεωρία και πράξη των εκδόσεων της
υστεροβυζαντινής, αναγεννησιακής και μεταβυζαντινής δημώδους γραμματείας. Πρακτικά
του διεθνούς συνεδρίου Neograeca Medii Aevi IVa. Αμβούργο 28.-31.1.1999, Ηράκλειο:
 Chrystalla A. Thoma

Πανεπιστημιακές Εκδόσεις Κρήτης (Theory and practice in the editions of Late Medieval
Renaissance and Post-Byzantine popular literature). In Proceedings of the International
Conference Neograeca Medii Aevi Iva, Hamburg 28.-31.1.1999. Herakleion: University Press
Crete.
Eideneier, H. 2002. Rhetorik und Stil – der griechische Beitrag [AZM, Working Papers in Multi-
lingualism. Series B 43]. Hamburg: University of Hamburg, Research Centre 538 Multilin-
gualism.
Fontana, J. 1993. Phrase Structure and the Syntax of Clitics in the History of Spanish. PhD dis-
sertation, University of Pennsylvania.
Fraser, B. 1999. What are discourse markers? Journal of Pragmatics 31: 931–952.
Georgakopoulou, A. 1997. Narrative Performances: A study of Modern Greek storytelling, Am-
sterdam: John Benjamins.
Givón, T. 1979. On Understanding Grammar. New York NY: Academic Press.
Givón, T. [11990] 2001. Syntax: A functional-typological introduction. Amsterdam: John Ben-
jamins.
Givón, T. 1983. Topic continuity in discourse: The functional domain of switch reference. In
Switch Reference and Universal Grammar, J. Haiman and P. Munro (eds.), 51–82. Amster-
dam: John Benjamins.
Givón, T. 1992. The grammar of referential coherence as mental processing instructions. Lin-
guistics 30: 5–55.
Haberland, H., and van der Auwera, J. 1990. Topics and clitics in Greek relatives. Acta Linguis-
tica Hafniensia 22: 127–157.
Horrocks, G.. 1990. Clitics in Greek: A diachronic review. In Greek Outside Greece II, M. Rous-
sou and S. Panteli (eds), 35–52. Athens: Diaspora.
Horrocks, G. C. 1997. Greek: A history of the language and its speakers. London: Longman.
Janse, M. 1994. Son of Wackernagel: The distribution of object clitic pronouns in Cappadocian.
In Themes in Greek Linguistics I, I. Philippaki-Warburton, K. Nicolaidis and M. Sifianou
(eds), 435–442. Amsterdam: John Benjamins.
Janse, M. 1998. Cappadocian clitics and the syntax-morphology interface. In Themes in Greek
Linguistic II, B. Joseph, G. C. Horrocks, and I. Philippaki-Warburton (eds), 257–281. Am-
sterdam: John Benjamins.
Janse, M. 2000. Convergence and divergence in the development of the Greek and Latin clitic
pronouns. In Stability, Variation and Change of Word-Order Patterns over Time, R. Sorni-
cola, E. Poppe and A. Shisha-Halevy (eds), 231–258. Amsterdam: John Benjamins.
Mackridge, P. 1993. An editorial problem in the Medieval Greek texts: The position of the object
clitic pronoun in the Escorial Digenes Akrites. In Neograeca Medii Aevi, N. M. Panayiotakis
(ed.), 325–342. Venice.
Mackridge, P. 2000. The position of the weak object pronoun in medieval and modern Greek.
Yazik I Rechevaya Deyatel’Nost’ [St. Petersburg] 3.
Martins, A. M. 1994. Clíticos na história do português. PhD dissertation, University of Lisbon.
Newton, B. 1972. Cypriot Greek: Its phonology and inflections. Hague: Mouton.
Pappas, P. 2001. Weak object pronoun placement in later Medieval Greek: Intralinguistic pa-
rameters affecting variation. OSUWPL 56: 79–106.
Pappas, P. 2004a. Variation and Morphosyntactic Change in Greek: From clitics to affixes. Hamp-
shire: Palgrave.
Pappas, P. 2004b. Medieval Greek weak object pronouns and analogical change: A response to
Condoravdi and Kiparsky (2001). Journal of Greek Linguistics 5: 127–158.
Distribution and function of clitic object pronouns in 16th–18th century Greek narratives 

Philippaki-Warburton, I. 1995. Διαχρονική θεώρηση της θέσης των εγκλιτικών μέσα στην
πρόταση. Studies in Greek Linguistics 15: 123–134.
Revithiadou, A. 2006. Prosodic filters on syntax: An interface account of second position clitics.
Lingua 116: 79–111.
Rivero, M. 1991. Clitic and NP climbing in Old Spanish. In Current studies in Spanish linguistics,
H. Campos and F. Martínez-Gil (eds.), 241–282. Washington DC: Georgetown University
Press.
Rollo, A. 1989. L’ uso dell’ enclisi nel Greco volgare dal XII al XVII secolo e la legge Tobler-Mus-
safia. Italoellinika 2: 135–146.
Thoma, C. 2006. Combining Functional Linguistics and Translation Theories: A case study of
Greek Cypriot and British folktales. Frankfurt: Peter Lang.
Traugott, E. C. 1994. Grammaticalization and lexicalization. In The Encyclopaedia of Language
and Linguistics, R. E. Asher and J.M.Y. Simpson (eds), Vol. 3, 1481–86. Oxford: Pergamon
Press.
Toufexis, N. In prep. Defining HIGH and LOW from a typological perspective. A case-study of
the Early Modern Greek intralingual translations of the Life of Aesop. University of Ham-
burg.
Toufexis, N. In prep. Register variation and the grammar of Medieval Greek. University of Cam-
bridge.
Tzartzanos, A. A. [11931] 1999. Νεοελληνική Σύνταξις (της Κοινής Δημοτικής) (Modern Greek
Syntax of demotic Greek) Vol. I – II, Athens: Kyriakides Bros Ltd.

Sources

Digenis Akritis: Jeffreys, E. (ed.) 1998: Digenis Akritis: The Grottaferrata and Escorial versions.
Cambridge: CUP.
Kallioupolitis: Kasdaglis, E. C. (ed.) 1998: Η Καινή διαθήκη του Κυρίου Ημών Ιησού Χριστού.
Athens: MIET, 1995–1998.
Kehayoglou Anthology: Kehayoglou, G. (ed.) 2003: Πεζογραφική Aνθολογία: Αφηγηματικός
γραπτός νεοελληνικός λόγος. Τόμος Α΄: Από τα τέλη του Βυζαντίου ως τη Γαλλική
Επανάσταση, σελ. 1–760, Τόμος Β΄: Από τη Γαλλική Επανάσταση ως τη δημιουργία του
ελληνικού κράτους, σελ. 761–1504 Aristoteles University of Thessaloniki, Institute for Mod-
ern Greek Studies, Manolis Triantafyllides Institute.
Life of Aesop: Papathomopoulos, M. (ed.) 1999: Πέντε δημώδεις μεταφράσεις του Βίου του
Αισώπου. Athes: Papadimas.
Der Äsoproman: Eideneier, H. (ed.) In prep.: (electronic text).
Papasynadinos Memoirs: Odorico, P. (ed.) 1996: Αναμνήσεις και συμβουλές του Συναδινού,
ιερέα Σερρών στη Μακεδονία, 17ος αιώνας. Association Pierre Belon: Athens, Paris.
Nominative subjects of non-finite clauses
in Hiberno-English

Lukas Pietsch
University of Hamburg

Hiberno-English (Irish English) dialects have developed an innovative


grammatical pattern where the pronominal subject of an embedded gerund
clause is marked with nominative instead of accusative or genitive case. This
paper traces the use of this structure in older forms of Hiberno-English and
contrasts it with corresponding patterns in Standard English and in Irish. It is
argued that the development of the pattern is indirectly due to structural transfer
from Irish, although superficially the resulting distribution of nominative and
accusative pronoun forms in Hiberno-English differs crucially from that in both
the substrate and the superstrate language.

1. Introduction1

The language contact situation between Irish and English in Ireland has long been
recognised as a veritable hotbed of contact-induced language change phenomena.
Starting in the 17th and culminating in the mid-19th century, a situation of migration,
societal bilingualism and subsequent mass language shift have led to the emergence of
a set of highly divergent dialects of English which show very marked traces of influ-
ence from the Irish substrate (Filppula 1999, Hickey ed. 2005). Given the intensity of
the contact, it is not surprising that these effects have permeated virtually all domains
of linguistic organisation, from discourse through phonology and grammar.
According to Matras (this vol.), clause-combining strategies, such as complementa-
tion constructions, appear to be a domain particularly susceptible to contact-induced
change. As Matras notes, such convergence effects need not involve actual borrowing of
formal material, but may also happen through a replication of more abstract properties
of the constructions involved, or in his words, a “fusion” of rules of form-function map-
ping. It will be one effect of this kind that will form the focus of this paper.
One of the means of coding inter-clausal connectivity employed prominently by
English is the use of non-finite verb forms (infinitives, gerunds and participles) in
subordinate clauses. As in many other languages, the use of non-finite clauses in Eng-
 Lukas Pietsch

lish involves not only a special morphological coding of the verb, but in many in-
stances a re-organisation of the basic clausal architecture as a whole, especially with
respect to the licensing and case marking of nominal arguments of the verb. This is a
feature that English shares – although with characteristic differences in detail – with its
neighboring and contact language, Irish.
Certain forms of Hiberno-English dialects seem to have diverged from Standard
English in a way that may strike the observer as surprising. While in Standard English,
subject arguments of gerund clauses are coded exclusively as either genitives or accu-
satives, Hiberno-English instead allows nominative-marked subjects in this position,
as shown in the 19th-century attestation in (1):
(1) My sister Bridget stopped with her old missus after I leaving [Normil04,
1855]
Although there is a considerable amount of literature on the grammar of Hiberno-
English, and on the question of possible contact-related influences on it, this phenom-
enon has so far not featured prominently in the discussion. If at all, it has usually been
mentioned only in passing, in connection with the somewhat better-known issue of
the subordinating use of and (see Section 2.2.2 below). However, the issue of non-fi-
nite nominative subject (henceforth NNS) constructions is interesting in its own right,
and it has ramifications well beyond the subordinating and cases.
In this article, I will provide a contrastive analysis of the relevant aspects of the
grammars of Irish, Standard English, and Hiberno-English, focussing first on the mor-
phological system of case-marking, and then on the syntactic structures encountered
in non-finite clauses. In doing so, I will also provide a first descriptive outline docu-
menting NNS constructions in older forms of Hiberno-English, based on data from a
corpus of subliterary 18th and 19th century texts.2 In conclusion, I shall then sketch
out a scenario of how the NNS constructions may have developed in Hiberno-English.
I will argue that the use of nominative pronouns is not explainable in a straightforward
way as a simple transfer feature from Irish, because (at least at first sight) it would seem
that the corresponding structures in Irish should, if anything, have reinforced the use
of accusative pronouns in these positions, as they are used in many other present-day
varieties of English. I will then argue, on the other hand, that transfer of a rather more
subtle kind did take place after all: it proceeded not from a simple equivalence relation
between surface grammatical forms of the two languages, but rather from more ab-
stract properties of the constructional frames in which these forms appear. This ab-
stract, indirect transfer seems to have led to a rather paradoxical result, insofar as it has
led to a surface distribution of forms that differs crucially both from Standard English
and from Irish.
Nominative subjects of non-finite clauses in Hiberno-English 

2. English and Irish in contrast

English and Irish, while differing sharply from each other in their general typological
characteristics in many important ways, nevertheless share a good deal of common
ground in the use of non-finite verb forms in embedded clauses. A central role in Irish
grammar is played by the verbal noun (VN). Besides being used in a periphrastic con-
struction to mark a progressive, it corresponds in many of its other uses to either an
English gerund or a to-infinitive. While there is a great deal of functional overlap be-
tween these categories in both languages, both of them also display a fair amount of
internal variation in their use. It is with respect to one aspect of this variation that I will
compare the grammars of the two languages in this paper: the question of case-marking
and syntactic licensing of overt subject arguments of the gerund/verbal-noun clauses.

2.1 The case systems of English and Irish

The apparent paradox which one needs to face in attempting a contact-related expla-
nation of the Hiberno-English NNS constructions lies in the fact that Irish, in the cor-
responding positions of non-finite constructions, uses forms that are often identified
as accusative (object), not nominative (subject) forms. Hence, one might be led to ex-
pect that any Irish influence on Hiberno-English should re-inforce the use of accusa-
tives rather than support the innovation of nominatives in these positions. The identi-
fication of the relevant Irish forms as accusatives is, however, based on the assumption
of a fundamental equivalence between the relevant domains of English and Irish mor-
phology. In order to show that this identification is indeed mistaken (cf. Genee 1998:
15), it is necessary to give a brief analysis of the case systems of both languages.
As is well known, English lacks a nominative-accusative case distinction in lexical
nouns and in most pronouns, retaining only the genitive as an overtly marked case,3
contrasting with a common or default case. Beyond this distinction, it has preserved a
nominative-accusative contrast only in the closed set of personal pronoun forms I/me,
he/him, she/her, we/us, they/them (and marginally who/whom). However, this contrast
does not follow the functional pattern of a typical nominative-accusative distinction eve-
rywhere either, namely that of marking subjects and objects/complements respectively.
There has been a rather strong tendency, in many modern varieties of English, to intro-
duce seemingly accusative forms into traditionally nominative environments, especially
in those that differ markedly from the prototypical subject position, which is immedi-
ately preverbal, topical, and unstressed. Sentences such as me and John are going or it’s me
illustrate this tendency, which may be interpreted as a trend towards reanalysing the ac-
cusative forms into strong and the nominative forms into weak pronouns. It should be
noted that in most environments where accusative forms encroach on traditionally
nominative territory, they bear focus and are often phonologically stressed.
Irish, too, rather like English, lacks a nominative-accusative case distinction in
lexical nouns and in most pronouns. There exists, however, a formal contrast, restrict-
 Lukas Pietsch

ed to third person pronouns. This contrast, which has traditionally often been de-
scribed as one of nominative vs. accusative case, is today treated in the more neutral
terms of ‘conjunct’ vs. ‘disjunct’ forms by most descriptive grammars of Irish (follow-
ing Bráithre Críostaí 1960). However, modern discussions in the generative frame-
work, based on the assumption of a universal repertoire of ‘structural cases’, have con-
tinued to discuss it in terms of nominative and accusative. When seen under this
perspective, there is a perceived mismatch in ‘case’ usage between Hiberno-English
and Irish in the environments in question here.
The disjunct (or ‘accusative’) form of Irish is morphologically unmarked and (for
the most part) historically older. The conjunct form (or ‘nominative’) is created fairly
transparently from the disjunct by the addition of a prefixed element s-. Both disjunct
and conjunct forms can be further expanded by another suffix to yield an emphatic,
contrastive form:

Table 1.  Pronoun forms in Irish

disjunct conjunct disj. contr. conj. contr.

Sg. Masc. é sé eisean seisean


Sg. Fem. í sí ise sise
Pl. (M/F) iad siad iadsan siadsan

The conjunct forms are used exclusively in the immediately post-verbal subject posi-
tion of finite lexical verbs (Bráithre Críostaí 1960: 138). The disjunct forms are used
everywhere else: as objects of finite verbs, as subjects and predicates in copula clauses
(which in Irish have a syntax very different from that of normal verbs); in non-finite
clauses; and also as parts of syntactically complex subject NPs after finite verbs, for
instance as the second of two co-ordinated subjects. It is the contrast between the use
of conjunct forms as finite subjects and disjunct forms as finite objects that provides
the obvious grounds for their traditional identification as, respectively, nominative
and accusative ‘case’ forms. However, the additional use of the disjunct forms in so
many other environments, especially in what are clearly subject positions in copula
clauses, fairly strongly suggest that their identification as accusative/object-case forms
is misleading. It should be noted that the apparent mismatch between subject syntactic
position and seemingly ‘object-case’ form cannot be motivated along the same lines as
in the English examples of the it’s me type: the disjunct/object forms display no signs
of being ‘strong’ pronouns, since they tend to be just as unstressed as their conjunct
counterparts; and both stand in the same formal opposition to the extra set of em-
phatic pronouns, which are obligatorily used whenever emphasis is required.
Thus, the obvious analysis is to say that the disjunct forms are an unmarked de-
fault form, neutral with respect to structural case assignment. The conjunct forms, in
contrast, are only a surface morphological variant, conditioned, like clitics, under con-
Nominative subjects of non-finite clauses in Hiberno-English 

ditions of linear adjacency to the finite verb (for a similar argument, see Carnie 1995:
160f.).4 This analysis also happens to fit the diachronic facts better, since the conjunct/
disjunct contrast is not historically a reflex of inherited case morphology, as it is in
English, but rather a fairly recent innovation (Pedersen 1913: § 491): the disjunct pro-
nouns are a continuation of a single, unified paradigm of common-case pronouns that
existed in older forms of the language, while the conjunct forms represent an innova-
tion that has happened in Irish but not even in the very closely related sister language
Scottish Gaelic.
Summing up, we may characterise the relations between the pairs of forms in-
volved in Irish and in English in terms of their markedness relations, which in some
sense are just the reverse of each other. Whereas in English, the structural markedness
relation between the two forms can be analysed, fairly straightforwardly, as the accusa-
tives being the marked forms (notwithstanding the fact that they today appear in a
greater range of positions, because that is arguably motivated by independent factors
such as their emergent function as strong pronouns), it seems to be exactly the reverse
in Irish: here, the form that happens to be used in object (and other) positions is clear-
ly the unmarked, default form, and the one that happens to be used in (most) subject
positions is a special, marked form. I will argue below that it is mainly this markedness
relation, and not the superficial equivalence between seemingly subject and object
forms, that became decisive in the mapping of correspondence relations between
forms in both languages, during the syntactic transfer processes that led to the emer-
gence of the nominative usage in non-finite clauses in Hiberno-English.

2.2 Subject marking in English ‘–ing’ clauses

2.2.1 Standard English


In order to explain the contact processes in Hiberno-English, we first need to provide
a broad sketch of the development of subject marking in mainstream English -ing
clauses. In doing so, it will be useful to distinguish the following types of constructions
(2): gerund clauses that stand in object or complement position within the matrix
clause, i.e. those that are outwardly governed by either a verb or a preposition (2a);
gerund clauses in subject position, where such a governing element is not present (2b);
so-called absolute constructions, which have the function of (usual causal) adverbials
(2c); and finally the special type of so-called subordinating-and clause, which is found
in certain non-standard dialects of English (2d).
(2) a. I hate [[John’s/his/John/him] being late]
b. [[John’s/his/John/him] being late] was bad news for all of us.
c. [[John/him/he] being late], I decided not to wait longer.
d. Why should we keep waiting, [and [John/him/he] being late again]!
As is well known, in clauses of the types (2a–b) the historically older option of genitive
marking alternates with the more modern option of accusative/common-case mark-
 Lukas Pietsch

ing in Present-Day English. While the genitive is structurally motivated by the original
nominal properties of the gerund, the innovative use of the accusative can be regarded
as a sign of the gradual strengthening of the verbal properties of this category in Eng-
lish (Fanego 1998, 2004a). In contrast to this, the absolute constructions of the type in
(2c) and apparently also those of (2d) have a different origin. Here, the original case
was the nominative (“nominative absolute”), and the verb forms in these clauses were
clearly participles, not gerunds, displaying no signs of noun-like syntactic behaviour.
For the intrusion of non-genitive forms in the types (2a–b), Fanego (2004a) de-
scribes two unrelated, distinct sources. Non-genitive forms in environments like (2a)
developed via reanalysis from instances that were morphologically ambiguous be-
tween accusative/common case and genitive case. As relevant cases Fanego (2004a: 43)
mentions: the lexical class of Middle English uninflected genitive singulars, nouns
ending in -s, plural nouns in general, and the pronoun her. This reanalysis was clearly
structurally motivated by the presence of potentially accusative-assigning governing
elements adjacent to these subject positions in the matrix clause, i.e. the governing
verbs or prepositions. In terms of formal grammar, the mechanism licensing the in-
novative accusative case forms can therefore plausibly be characterised as one of ‘ex-
ceptional case-marking’ (ECM).
In contrast to this, the non-genitive forms in sentence-initial (subject-clause) po-
sitions such as (2b) arose mainly through an extension of the superficially similar
nominative-absolute type (2c). Accordingly, it is first found with common-case lexical
nouns but not with unambiguously accusative pronouns.
Examples of non-genitives in type (2a) can be found, though at first very rarely,
from Middle English onwards (Fanego 2004a: 10; cf. also Fanego 2004b). Quoting
Tajima (1996), Fanego provides some early examples from c.1400 both with common-
case lexical nouns and with unambiguously accusative pronouns in such environ-
ments, but she also notes that the latter remained much more infrequent than the
former throughout Early Modern English (2004a: 8f.). Overt accusatives became rea-
sonably common only during Late Modern English, roughly from the 18th century
onwards (2004a: 43). As for structures of type (2b), Fanego’s earliest quoted examples
are from the 17th century (2004a: 44). Unambiguous accusative pronouns in these
positions seem at first to be excluded and are attested only much later, in the early 20th
century. In their stead, some early examples attest to the occasional use of nominatives,
as in (3). Note the great semantic similarity between this use of a gerund subject clause
as a causal subject with make, and the causal use of a nominative absolute:
(3) I having a great esteem for your honour […], makes me acquaint you of an
affair that I hope will oblige you to know. [From the Spectator, 1711–1712,
quoted in Fanego (2004a: 45) after Jespersen (1961–1970: V, §§ 9.8.3)]
Unlike these structures, gerunds in non-initial (object/complement) positions like
(2a) seem never to have been used with nominative subjects during the development
of mainstream English, a fact that accords well with their status as ECM structures.
Nominative subjects of non-finite clauses in Hiberno-English 

This makes the later appearance of nominatives in just these environments in Hiber-
no-English appear all the more striking.
As for the late spread of overtly accusative forms into the sentence-initial clauses
in mainstream English, it can apparently be seen in conjunction with the general
spread of accusative pronouns into other nominative environments, including those of
the John and me type and also those of the absolute construction (2c) itself.
Summing up, we can recognise the combined effects of three interrelated proc-
esses in the development of non-finite subject case marking in Standard English -ing
clauses: the development of ‘exceptional case-marking’ as a mechanism motivating
overt accusatives in environments governed by a case-assigning matrix element; the
emancipation of sentence-initial gerund clauses (with a subject position not overtly
governed by a matrix case assigner, and therefore allowing only for common-case lex-
ical nouns as subjects); and finally the generalisation of formally accusative pronouns
as strong pronouns, intruding into various syntactic positions, at last including those
in the sentence-initial gerund clauses.

2.2.2 Subordinating ‘and’


Before we turn to the use of nominatives in the Hiberno-English equivalents of clauses
of the type in (2a), we should first give some consideration to the special type of ad-
junct clauses introduced by and, as in (2d) above. The term “subordinating and” is here
understood as referring to clause-level constructions that are marked by the lack of a
finite verb as being dependent clauses, but nevertheless linked by and with a finite
matrix clause. This structure plays an important role in Hiberno-English and has at-
tracted somewhat more attention in the literature than the other types under discus-
sion here (Ó Siadhail 1984: 132–134; Häcker 1999; Filppula 1999: 200; Corrigan 2000).
Subordinating and occurs both with -ing or -en participles, and with verbless predicate
clauses. There has been some discussion about the subordinating function of and
found in this construction, about its semantics, and whether it was a feature of older
forms of English outside Ireland.
Corrigan (2000: 77) attempts to trace the history of the construction as far back as
Old English. However, the evidence she cites from the early periods only concerns
construction types with finite subordinated clauses, and Corrigan provides no argu-
ment to support her implicit suggestion that these types are in any way historically
related to the non-finite type at issue here. Filppula (1991) finds some seemingly more
pertinent examples in Early Modern English texts between 1500 and 1710, but even
this list does not convincingly prove a special role of and as a subordinator in a special-
ised construction type, since most of the examples can be analysed simply as absolute
constructions of the type (2c) that are coordinated among each other, as in (4):
(4) all physitians having given him over and he lying drawing his last breath there
came an old woman unto him. [1635, quoted in Filppula 1991: 624]
 Lukas Pietsch

It seems, in short, that the role of subordinating and in Standard English and in its
older forms has sometimes been over-stated in the literature. If anything, subordinat-
ing and with non-finite clauses was quite marginal in Early Modern English and emer-
gent Standard Modern English. However, evidence of comparable structures is more
solid in certain forms of dialectal English. Häcker gives some northern Middle English
and Older Scots examples, such as (5):
(5) Lorde,... I thank the... that thou to daie hase giuene me grace Almous to take...
Off thaim that was wont to be myne awne menne and seruid me, And i vnkna-
wen vnto thaim
‘... at a time when I was unknown to them’ [Häcker 1999: 42]
Ó Siadhail (1984: 133–134) quotes plausible examples from literary representations of
the 19th-century spoken vernacular of Warwickshire in the English Midlands, taken
from novels by George Eliot. Similar examples from late 19th-century Lincolnshire are
quoted by Häcker:
(6) I thought then and I think now, it fell strange and hard on her, and her nobbut
seventeen [Häcker 1999: 42]
The most solid evidence, according to Häcker, comes from modern Scottish varieties,
both from the varieties of the Highlands and western islands (where English has been or
used to be in contact with Gaelic until very recently), and from Lowlands Scots. Finally,
Quirk et al. (1985: § 11.44) recognise the construction, in its verbless form, as a feature
even of Standard Present-Day English, but describe it as a marginal type of “irregular
sentence”, a formal idiom that is in some sense extraneous to regular clausal syntax:
(7) a. How could you be so spiteful and her your best friend?
b. They left without a word, and he so sensitive.
While Quirk et al. recognise both the use type with the accusative and, “less com-
monly”, that with the nominative, the picture that emerges with respect to case usage
in the older varieties is as follows: The oldest attestations, from Middle English, Early
Modern English and literary early 19th-century Scots, invariably have nominative
marking, and are in this respect formally identical to nominative absolutes. Later ex-
amples from British varieties, especially from more colloquial, dialectal forms, mostly
have accusative marking. This is also generally true for modern Scots. Häcker (1999:
43) plausibly interprets this as a diachronic development; it seems to be yet another
manifestation of the overall trend for accusative pronouns to encroach on nominative
domains and to take on strong-pronoun functions.
As for the Irish varieties, the picture is mixed. Ó Siadhail (1984: 131–132) provides
evidence of a preference for the nominative in data from the southern Irish counties of
Clare (Munster), and Carlow and Wicklow (Leinster), the nominative being used in all
of six examples cited by him. Likewise, Filppula (1999: 200) finds only nominative
marking in his data from Clare, Kerry, Wicklow and Dublin, all in the south. In con-
Nominative subjects of non-finite clauses in Hiberno-English 

trast to this, Henry (1957: 206) reports that in the more northern county of Roscom-
mon (Connacht) both cases were used. Finally, Corrigan (2000: 87) finds an exclusive
use of the accusative in her data from Armagh (Ulster), likewise in the north. This sug-
gests that there is a geographical north-south divide across Ireland, with the northern
varieties siding with the British dialects whereas the southern ones have developed a
new, specialised use of the nominative.
This north-south divide apparently corresponds with a semantic innovation also
found mainly in the south of Ireland (Corrigan 2000: 79). The structures attested in
Standard English, those from the modern northern British dialects, and also most of
those found in the northern parts of Ireland, all have a narrow, pragmatically special-
ised function, which – following Filppula – we may call the “exclamative” use type.
These sentences express concessive or causal meanings, and are used emphatically to
signal a stance of personal emotional involvement, such as reproach or regret, as in the
examples (6) and (7) above or in the following example (8):
(8) he is never in work and when he was he give Me nothing & him nicely in
health
[Doorle07, 1905]
In contrast to this, southern Hiberno-English uses subordinating and in a more neu-
tral fashion, expressing purely temporal relations (9):
(9) a. I only thought of him there and I cooking my dinner
(‘...while I was cooking …’) [Häcker 1999:38, quoting Filppula 1991: 618]
b. One of them James came over and borrowed 3 pounds of me and he going to
the diggings
(‘… while he was going …’) [Hogan_04, 1857]
c. and once our waggon overset and we in it but received no material injury ex-
cept the children’s faces a little scratchd
(‘… while we were in it …’) [WrighH01, 1802]
This use type is generally thought to be a syntactic transfer feature from Irish, which, in
very similar fashion, uses the connector agus (‘and’) both as a co-ordinating conjunction
and as a non-finite clausal subordinator, with a very similar range of meanings.
Summing up, the picture that emerges from these findings is the following: while
certain uses of subordinating and, together with the marginal use of nominatives in
absolute participle constructions, were present in various forms of Late Modern Eng-
lish and plausibly also in those that served as input to the language contact situation in
Ireland between the 17th and the 19th centuries, southern forms of Hiberno-English,
apparently under the influence of a formally comparable syntactic pattern in Irish,
radically extended the use of subordinating and by making it take over certain new
functions which it had in Irish but not previously in English. In doing so, Hiberno-
English consolidated the use of the nominative, developing it from the minor and
rather marginal use type that it was in English, into an apparently much more frequent
 Lukas Pietsch

and more deeply entrenched feature of its syntactic system. This increased use of the
nominative found a parallel or further extension in the innovative use of the same case
in another group of constructions too, namely in those non-initial gerund clauses that
were governed by potentially accusative-assigning verbs or prepositions. It is to these
constructions that we will turn next.

2.2.3 Non-initial gerund clauses in Hiberno-English


Unlike the use of subordinating and, the other non-finite nominative subject (NNS)
constructions in Hiberno-English have up to now found little or no attention in the
literature. Hence, to the best of my knowledge, appropriate data regarding their exist-
ence and distribution in modern dialects of Irish English is lacking at present,5 and the
presentation below will be confined to 19th-century attestations from our corpus.
Apart from that, there are only occasional remarks about some dialectal constructions
that are somewhat similar, but not quite identical, to the type to be documented here.
For instance, Henry (1957: 190) documents cases of nominative subjects with to-in-
finitives in a 20th-century Connacht dialect (10), but does not mention parallel cases
with gerunds.
(10) It’s a point o’ law for she to put him out
(‘Her right to put him out (of the house) is legally debateable’) [Henry 1957: 190]
Another rare hint is found in Filppula (1999: 197), who quotes examples with -ing
forms following when; however, since when in Standard English can introduce both
non-finite (participial) and finite subclauses, both these examples could be understood
simply as finite clauses with copula drop, a feature that is independently well attested
in some Hiberno-English dialects. As will become clear shortly, an analysis as finite
copula drop is excluded in many of the cases to be presented below. The when cases
also differ from most of the others discussed in this section insofar as when does not
govern case in Standard English and would therefore not be a candidate for ECM-style
licensing of accusative case either.
(11) a. indeed I walked it myself when I young [Filppula 1999: 197]
b. I remember when I going to school, I remember three of my uncles went away
[...] she and most of her family went to America, when I going to school. [Filp-
pula 1999: 197]
In our corpus, nominatives in gerund clauses in potential ECM environments are at-
tested with a variety of different governing elements. Most often they occur as comple-
ments of prepositions, both prepositional adverbial adjuncts and as parts of preposi-
tional phrasal-verb constructions. There are also a few attestations of object clauses
governed directly by a verb. The most frequently attested governing subordinator is after.
In Standard English, after would require subjectless constructions under conditions of
subject control, i.e. obligatory co-reference between the matrix subject and the non-overt
subject of the embedded clause. Strikingly, in our examples (12), overt nominative sub-
Nominative subjects of non-finite clauses in Hiberno-English 

jects are included both with and without co-reference with the matrix subject. They also
occur with the after-clause both preceding and following the main clause.
(12) a. If a an Irishman goes to drive horses or Bullocks here after he comming out
from home, he might as well go whistle a gig to a milestone [Normil03, 1863]
b. After he returning the Creditors Came and took charge of all the goods [Nor-
mil04, 1863]
c. My Sister Bridget stoped with her old Misses after I leaving, [Normil04, 1863]
d. Their Father & Mother felt very uneasy after they going but they are delighted
now [Dunne_12, 1874]
e. Mrs Farrell had high mass for the repose of his Soul in a few day after She get-
ting word from his wife [DunneJ01, 1868]
Similar constructions are found with various other prepositional subordinators (13):
(13) a. I recd. your letter about the first of Jany. last and would have written an answer
to you ere now were it not for I being paying Michl. Moores passage as re-
quired by you [Hogan_04, 1857]
b. Her uncles & aunt was very much dissapointed in She not coming. [Maho-
ny02, 1887]
c. Michl. Gready Patt McGrath and Bridget Neylon were as glad as if we Gave
them a thousand pound for we being along with them. [Normil01, 1854]
d. Moreland Sold his property twelve Months ago to his tennants & agreed to
Sell this Evicted farom to the Evicte{****} by he Getting one thousant pounds
from the Land Commissioners [MarshM05, 1907]
e. what is the Cause of we not Getting the possession of this farom [MarshM04,
1907]
f. The place was finally handed over to me by my niece on 1st Nov ‘03 on I pay-
ing her the sum of £42 and £3 Cost [ReillP01, 1907]
The following examples show NNS constructions in object clauses of verba sentiendi
(14). (14b) is particularly interesting, as the nominative subject is a resumptive pro-
noun syntactically duplicating an extracted wh- element:
(14) a. When I heard she being in this place I went to see her directly [Normil04, 1863]
b. I have a friend which I did not know she being in this Town until of late,
Michael Healys daughter from Ballanagun. [Normil04, 1863]
All in all, the NNS constructions in question are not particularly frequent in the cor-
pus data: in a corpus of c.227,000 words, there are a total of 19 unambiguous tokens
(counting only those tokens where the subject is unmistakably a nominative pronoun,
excluding cases of you and it as well as all tokens involving full lexical nominals.) To
these we may add 9 tokens of subordinating and with nominative subjects, and 10 to-
kens of nominative absolutes of the type also found in mainstream English (these lat-
ter forms being apparently a feature of formal, conservative written registers in gen-
 Lukas Pietsch

eral, and probably not particularly characteristic of the spoken dialects reflected in
these writings). Given the small numbers, it is clear that not much can be inferred with
statistical certainty about their distribution across the corpus. Nevertheless, the attes-
tations provide ground for the hypothesis that the pattern is a regular and productive
one, not restricted to idiomatic lexical idiosyncrasies of certain subordinating ele-
ments but rather a general syntactic pattern characteristic of gerund clauses at large, at
least in certain forms of Hiberno-English.
Moreover, the distribution of the attestations across the different individuals in the
corpus indicates that within the spectrum of varieties and registers represented, the
NNS structures are characteristic of specifically Irish-influenced dialectal varieties.6
Eight of the nine informants who use the NNS construction come from the southern
provinces of Ireland, and except for one (or possibly two) they are all Catholics. More-
over, most of the individuals involved here display heavy signs of other dialectal phe-
nomena in their writings which can be also characterised as Hibernicisms and/or
which can also be plausibly linked to Irish contact effects. NNS constructions are con-
spicuously absent from other writers in the corpus whose linguistic profiles display
signs of different dialectal backgrounds, especially those influenced by northern, Ul-
ster-Scots speech forms. The southern Irish dialects are usually regarded as the his-
torical outcome of the development of a mixture of mainly southwest English settler
dialects with a strong contact-induced admixture of Irish substrate features. It is there-
fore reasonable to assume that contact effects are likely to have played some role in the
development of the NNS constructions too.

2.3 Subject marking in non-finite clauses in Irish

Having discussed the use of nominatives in English non-finite clauses, we can now
turn to Irish in order to investigate in what ways it would have provided any relevant
structural parallels. As is well known, Irish non-finite clauses are formed with the help
of a verbal noun (VN). There is a variety of structural configurations: the non-finite
clause may be headed by a bare verbal noun, or by a verbal noun preceded by the par-
ticle a (15); or, in yet other environments, the verbal noun may appear as part of a
progressive periphrastic construction headed by the progressive marker ag (16).
(15) roimh í a dhul
before she going
‘before she went …’ [Bráithre Críostaí 1960: 248]
(16) agus é ag tíocht ó bhainis
and he at coming from wedding
‘… when he was coming from a wedding’ [Ó Siadhail 1989: 284 ]
There is also some variation in the coding of subject (and object) arguments of these
clauses. Owing to its nominal, gerund-like quality, the VN may take genitive argu-
Nominative subjects of non-finite clauses in Hiberno-English 

ments in certain environments, realised as either a genitive-marked full NP following


the VN, or a possessive pronoun preceding it. However, unlike in English, it is regu-
larly the object argument, not the subject, that is coded in this way.7 Subjects of verbal-
noun clauses can be expressed as oblique agent phrases, using the preposition do ‘to’
(17, 18). This preposition is used elsewhere in Irish to express possessors too (similar
to French à). As a means of coding subjects/agents, it has no obvious parallels in Eng-
lish. The do-agents can appear either after (17) or before (18) the verbal noun.
(17) Le linn an chaint sin a rá dó
while this talk saying to.him
‘while he was saying this …’ [Bráithre Críostaí 1960: 256]
(18) Le linn dúinn a bheith ag fanacht leo …
while to.us being at waiting with.them …
‘while we were waiting for them …’ [Ó Dónaill 1977, v.s. “do”]
An alternative strategy for the coding of either objects or subjects is the use of a struc-
ture that has variously been called “raising” (Noonan 1995, Disterheft 1982), “promo-
tion” (Armstrong 1977) or “displacement” (Genee 1998). Here, a nominal or pronom-
inal argument, in the unmarked common case, is placed in a preposed (“promoted”)
position, to the left of the verbal noun, i.e. in a position superficially resembling that of
the subject of English gerund clauses (19,20). This position can hold either object or
subject arguments in Irish.8 The following VN in these constructions is usually marked
by the particle a and lenition, historically a reflex of a former preposition do (‘to’).9 Ac-
cording to Ó Siadhail (1989: 277) and Genee (1998: 442), there has been an historical
trend away from the genitive/do to the promotion structures.
(19) tar éis iad féin a shábháil na gcéata
after they themselves saving hundreds.gen
‘… after they themselves saved hundreds’ [Ó Siadhail 1989: 256]
(20) an bun a bhí le mé féin a thógaint geite
the reason that was with I myself taking fright.gen
‘… the reason of my becoming frightened’ [Ó Siadhail 1989: 256]
Of the three coding strategies for nominal arguments with verbal nouns in Irish, two
diverge quite radically from English: the genitive coding because it picks out the wrong
argument, and the prepositional coding because it has no structural counterpart in
English at all.10 It is thus only the third, the promotion construction with the morpho-
logically unmarked NP to the left of the verbal noun clause (15, 19, 20), that offers it-
self as a basis for cross-linguistic identification and may be implicated in a possible
transfer process regarding the nominative-subject structures in Hiberno-English. It is
therefore useful to compare briefly the structural properties of the promotion con-
struction in Irish with those of the subject position of gerund clauses in English.
 Lukas Pietsch

The position of the “promoted” nominal argument to the left of the verbal noun is
unique within the overall system of Irish grammar, as it diverges radically from the
normal verb-initial (and, more generally speaking, head-initial) pattern that Irish dis-
plays everywhere else. In finite clauses, neither objects nor subjects can usually pre-
cede their verbs in Irish, unless fronted through special syntactic processes. There are
basically three approaches that can be found in the literature in treating the anomaly
of this pre-VN position formally. Some, like Disterheft (1982) and Noonan (1995),
have characterised the construction as an instance of “raising”, assuming that the pro-
moted nominal is moved out of the embedded clause to occupy a structural object
position in the matrix clause. For Noonan (1995: 72f.), the crucial argument for this
analysis is the occurrence of the disjunct pronoun forms in these positions, which he
identifies as “object forms”. As I argued above I consider that morphological identifica-
tion misleading.
Furthermore, there is strong structural evidence that indeed the promoted NP be-
haves syntactically as part of a single constituent together with the following a-VN
phrase, a fact that rallies against a raising analysis in the strict sense. This is suggested,
among other things, by the fact that they can occur together as sentence fragments in
elliptical sentences (Bráithre Críostaí 1960: 251). Moreover, when the promoted element
is a pronoun and the subclause is governed by a preposition in the matrix domain, the
preposition fails to fuse with the pronoun into a so-called inflected preposition form, as
would be expected if the pronoun was formally an element of the matrix construction.
This clearly shows that the governing relation is not between the preposition and the
nominal argument as a standalone constituent of the matrix clause, but between the
preposition and the NP-a-VN construction as a whole. See also Genee (1998: 451), refer-
ring to Armstrong (1977) for further arguments against the raising analysis.
It should be noted that the analysis of the promoted nominal as part of a common
NP-a-VN constituent is not contradicted by cases in which it is visibly moved away
from the verbal-noun construction, as this seems to happen through independently
motivated mechanisms, for instance by wh-extraction (22), focussing extraction to the
right (23), or in an ‘easy-to-please’ construction (25) (examples adapted from Bráithre
Críostaí 1960: 251).
(21) Dúirt sí leis na leabhair a dhíol
said she with.him the books selling
‘She told him to sell the books’
(22) Cá bhfuil na leabhair a dúirt sí leis a dhíol
where are the books rel said she with.him selling
‘Where are the books that she told him to sell?’
(23) Dúirt sí leis gan a dhíol ach na leabhair
said she with.him neg selling but the books
‘She told him to sell only the books.’
Nominative subjects of non-finite clauses in Hiberno-English 

(24) Is furasta é a mhealladh


is easy he deceiving
‘It is easy to deceive him.’
(25) Tá sé furasta a mhealladh
is he easy deceiving
‘He is easy to deceive.’
If one accepts the analysis that the promoted element (such as iad féin in example (19))
is part of a single NP-a-VN constituent, this leaves as a further issue of formal analysis
the question of headedness within this construction. Studies that have dealt primarily
with the diachronic development of the construction since Old Irish (Gagnepain 1963:
18, Genee 1998: 451) have assumed that the promoted NP is the head, implying that the
a-VN phrase is in some way adjoined to it on the right. This view is primarily motivated
by the historical roots of the construction: the particle preceding the VN was originally
the preposition do (‘to’), and thus the syntax of the construction [NP [do [VN]] overtly
resembles that of an NP postmodified by an adjoined do-possessor phrase [NP [do [NP]].
Moreover, as Genee (1998: 104) points out, the promoted nominal originally carried
overt case marking (genitive or accusative), reflecting the case assigned to the embedded
construction as a whole by the governing matrix construction (cf. Gagnepain 1963: 129–
134, 235–240). However, genitives in this position are reported to have survived only as
an optional structure in the modern dialects, having been largely replaced by a structure
where the promoted nominal is in the invariant common case regardless of the matrix
construction (Gagnepain 1963: 240, Ó Siadhail 1989: 276).
In contrast to the view that regards the promoted NP as the head, recent analyses
of present-day Irish conducted within a generative approach have chosen to interpret
it as occupying a specifier position at the left edge of a clausal structure projected by
the VN. This analysis obviously fits more easily with the common assumption of claus-
al constituent order being determined by elements moving upwards across a universal
grid of positions defined by a set of functional heads c-commanding the verb in an
X-bar scheme. According to different versions of this view (see the survey in Carnie
1995: 90–98), the particle a is assumed to represent a functional head position, vari-
ously identified as Agr (Duffield 1991) or AgrO (Noonan 1992, Bobalijk/Carnie 1992),
and the promoted object and/or subject is either in its specifier position or else, a step
higher up in the tree, in Spec-T.
The formal accounts differ technically in how they handle the question of case. As
far as I am aware, none of those analyses that see the promoted NP in a specifier posi-
tion deals with those alternative (conservative) structures where it bears overt genitive
case-marking assigned by the governing matrix element. As for the other, more mod-
ern option, where the promoted NP is in the common unmarked case, all the genera-
tive analyses seem to agree that the structural case (in the abstract sense of Chomskyan
case theory) assigned to it is indeed an accusative, not a nominative (Tallerman 2005:
850). This assumption is upheld even by Carnie (1995: 88), who elsewhere acknowl-
 Lukas Pietsch

edges that the morphological facts – i.e. the use of disjunct pronoun forms – are not
necessarily evidence for accusative as opposed to nominative status (1995: 160f.).
While the assignment of accusative case to promoted object NPs presents no theoreti-
cal problem within this framework, the assignment of assumedly accusative case to
promoted subjects requires some exceptional mechanism. Different analyses assume
either accusative assignment via non-finite properties of T, or an ‘exceptional case-
marking’ (ECM) mechanism – but the latter not involving any overt matrix element as
in English, but a hypothesised non-overt complementiser at the left edge of the em-
bedded clause.
If one abstracts away from the purely theory-internal sides of these analyses, one
finds them to agree on one point: they are all based on the intuition that the structural
licensing conditions for the promoted nominals are strictly clause-internal and can be
descriptively stated in terms of linear order of elements within the bounds of the em-
bedded clause alone, particularly in terms of adjacency with the following a-VN con-
stituent. There is no structural dependency between the promoted nominals and any
particular syntactic configuration involving overt elements further up in the matrix
domain. In particular, if one ignores the optional genitive structures mentioned earlier,
there is no evidence of anything openly resembling the English ECM mechanism. If,
moreover, one discards the theory-internally motivated assumption of a universal
nominative-accusative case distinction, taking into account that Modern Irish lacks
any overt morphological expression of such a contrast, then it seems fair to sum up that
the constructional frame responsible for the licensing of these VN arguments is defined
in purely positional terms and does not involve case-marking in the sense of an overtly
expressed morphological mechanism at all. The syntactic relation between the nominal
and the VN is not morphologically coded on the nominal – neither by case morphol-
ogy proper nor by the other principal means of morphological coding available in Irish,
initial consonant mutations. The case properties specified for this position can thus best
be described as being simply the maximally unmarked, common case.

3. Conclusions: Transfer and markedness in the


development of Hiberno-English NNS structures

We can now summarise the essential structural difference between the Irish and the
English systems with respect to the subject position in non-finite clauses: Standard
English requires subjects in the initial position of gerund and infinitive clauses to be
licensed through a case-assignment relation with an overt, adjacent governing ele-
ment, using either genitive assigned by the gerund itself, or accusative assigned via
ECM from outside the clause. Present-day Irish has a structural subject slot found in
superficially the same position, before a verbal noun, but its structural conditioning is
different: subjects in this position are not sensitive to licensing through a government
Nominative subjects of non-finite clauses in Hiberno-English 

relation from outside the clause, and they bear no morphological marking to reflect
such a licensing relation.
It seems to have been this relationship of deceptive structural similarity between
English and Irish that provided the ground for a transfer-induced structural reanalysis
of the English gerund clauses by Irish learners. In their English input, Irish learners in
the 18th and 19th centuries were obviously faced with a situation of great variability in
pronoun marking in a number of closely related constructions, a situation which
would easily lead to systemic simplification or restructuring. In gerund clauses, there
was variation between accusative and genitive marking. Moreover, although initially
rare, there was also the superficially similar absolute participle construction, which
already had the nominative. We may hypothesise that this nominative usage first in-
creased in frequency as Irish speakers co-opted the absolute-participle pattern in or-
der to create the Hiberno-English subordinating-and clauses of the type “I saw John
and he going home”, on the model of Irish “Chonaic mé Seán agus é ag dul abhaile”. This
nominative usage then could also serve as an analogical trigger for an extension of
nominatives into the – superficially similar – subject positions of gerund clauses prop-
er. What made this extension structurally motivated was, in turn, the reanalysis of the
gerund clauses on the model of the corresponding Irish verbal-noun clauses: Irish
learners apparently failed to acquire the rules of the morphosyntactic licensing rela-
tion between the governing matrix element and the accusative-marked subject posi-
tion in these clauses, because such a licensing relation was not (or at least not obliga-
torily) a feature of the corresponding Irish structures. Thus, Irish speakers could opt
for the nominative, which they perceived as the maximally unmarked case-form of
English, and which in this respect corresponded to the Irish disjunct (é, í, iad) pro-
noun forms. This correspondence held even though both sets of pronouns had other-
wise completely different, almost reverse, distributions in finite clauses. It was thus
abstract structural relations between constructions and between morphological para-
digms, rather than superficial equivalence relations between individual items, that
were decisive in contact-induced structural transfer.

Notes

1. This work, including the ongoing compilation of the corpus on which its results are based,
was funded in the framework of the Collaborative Research Centre No. 538 ‘Multilingualism’ by
the Deutsche Forschungsgemeinschaft (German Research Foundation).
2. The Hamburg Corpus of Irish Emigrant Letters (Pietsch, i. prog.) is being compiled at the
University of Hamburg. Some of the material is also available in previously published collec-
tions: Miller et al. (2003), Fitzpatrick (1994), and O’Farrell (1984); the rest was collected from
various archival sources in Ireland. In the preliminary form used for this study, the corpus
consisted of some 230,000 words produced by 162 writers, ranging from the early 18th to the
beginnings of the 20th century. It consists of letters, diary entries and other similar sub-literary
 Lukas Pietsch

text types, most of them written in the context of emigration from Ireland to America and Aus-
tralia or related in some other ways to the political and social upheavals of the 19th century.
3. For the present discussion, we can set aside the discussion about the genitive -s being a
phrasal/syntactic rather than morphological category. I shall assume, for the purposes of this
paper, that both the possessive -s on nouns or noun phrases, and the possessive forms of pro-
nouns, are representative of a common grammatical category ‘genitive’ in English.
4. An anonymous referee rightly points to a similar line of argumentation in Wigger (1970:
36). Wigger also mentions the survival of lenited thú as a specifically accusative alternant of tú,
as being the only isolated remnant of a genuine morphological nominative-accusative contrast
in the language. The same referee also mentions an older approach proposed by Hartmann
(1960: 11), who attempts to justify the nominative-accusative distinction on the basis of a uni-
fied semantically-based characterisation of the two forms. According to him, the “nominative”
denotes “an entity presented as differentiated and identical to itself, from which the event origi-
nates” (“einen als differenziert und mit sich selbst identisch gesetzten Gegenstand, von dem der
Vorgang seinen Ausgang nimmt”), while the “accusative” denotes “a participant which is neutral
with respect to the feature of ‘differentiated, positive origin’” (“Beteiligung eines Gegenstandes
am Vorgang bei Indifferenz gegenüber dem Merkmal ‘differenzierter, gesetzter Ausgangspu-
nkt’”). Based as it is in a theory of linguistic relativism (1960: 8), this proposal seems of little
relevance in the present debate.
5. In an internet discussion forum in 1991, Marion Gunn attested to the use of nominatives in
after-clauses in present-day Dublin English: “after he going out”.
[http://www.smo.uhi.ac.uk/~smacsuib/gaelic-m/logfiles/91/log9109.txt]
6. This hypothesis will be the subject of a separate, more extensive study elsewhere.
7. Only occasionally, subjects of intransitive verbs can also be coded as genitives: bhíomar ag
súil lena dteacht (lit. ‘we.were at hoping with.their coming’; ‘we were hoping for them to come’).
(Bráithre Críostaí 1960: 260).
8. According to Ó Siadhail (1989: 256), there is some dialectal variation: while in northern
dialects of Modern Irish, both a subject and an object (in this order) can be promoted in front
of a VN, southern Irish has only a single structural slot that can hold either a subject or an object
but not both.
9. Here, too, there is dialectal variation, as northern Irish dialects allow the verbal noun wit­h­
out the particle if the promoted nominal is an intransitive subject (Ó Siadhail 1989: 257f.).
10. Henry (1957: 186) quotes some examples in mid-20th-century Roscommon English where
the Irish do-agents have apparently led to a calque using of-agents in English: I saw him, goin’ to
Mass o’ me (‘… when I was going to Mass’); They were there comin’ away o’ me (‘… when I was
coming away’). Nothing resembling these structures is attested in our corpus.

References

Armstrong, J. 1977. The Syntax of the Verbal Noun in Modern Irish Prose 1600–1650. PhD dis-
sertation, Harvard University.
Bobalijk, J. D. and Carnie, A. 1992. A minimalist approach to some problems of Irish word or-
der. Proceedings of the Harvard Celtic Colloquium 12: 110–134.
Nominative subjects of non-finite clauses in Hiberno-English 

Bráithre Críostaí [The Christian Brothers]. 1960. Graiméar Gaeilge na mBráithre Críostaí. Dub-
lin: Mac an Ghoill.
Carnie, A. H. 1995. Non-Verbal Predication and Head Movement. PhD dissertation, MIT.
[http://dingo.sbs.arizona.edu/~carnie/papers/Carnie95Dissertation.pdf]
Corrigan, K. 2000. What are small clauses doing in South Armagh English, Irish and Planter
English? In The Celtic Englishes II, H. Tristram (ed.), 318–338. Heidelberg: Winter.
Disterheft, D. 1982. Subject raising in Old Irish. In Papers from the 5th International Conference
on Historical Linguistics, A. Ahlqvist (ed.), 44–53. Amsterdam: John Benjamins.
Duffield, N. 1991. Particles and Projections. PhD dissertation, University of Southern California.
Fanego, T. 1998. Developments in argument linking in Early Modern English gerund phrases.
English Language and Linguistics 2: 87–119.
Fanego, T. 2004a. On reanalysis and actualization in syntactic change: The rise and development
of English verbal gerunds. Diachronica 21: 5–55.
Fanego, T. 2004b. Some strategies for coding sentential subjects in English: From exaptation to
grammaticalization. Studies in Language 2: 87–119.
Filppula, M. 1991. Subordinating and in Hiberno-English syntax: Irish or English origin? In
Language Contact in the British Isles: Proceedings of the Eighth International Symposium on
Language Contact in Europe, P. S. Ureland and G. Broderick (eds.), 617–631. Tübingen:
Niemeyer.
Filppula, M. 1999. The Grammar of Irish English: Language in Hibernian style. London:
Routledge.
Fitzpatrick, D. 1994. Oceans of Consolation: Personal accounts of Irish migration to Australia.
Cork: Cork University Press.
Gagnepain, J. 1963. La syntaxe du Nom Verbal dans les langues celtiques. Vol. 1: Irlandais. Paris:
Klincksieck.
Genee, I. 1998. Sentential Complementation in a Functional Grammar of Irish. The Hague:
HAG.
Häcker, M. 1999. ‘And him no more than a minister’s man’: The English subordinating and-con-
struction in crosslinguistic perspective. Leeds Working Papers in Linguistics and Phonetics 7:
36–50.
Hartmann, H. 1960. Der Typus ocus é im Irischen. In Indogermanica: Festschrift für Wolfgang
Krause, H. Hartmann and H. Neumann (eds), 8–23. Heidelberg: Winter.
Henry, P. L. 1957. An Anglo-Irish Dialect of Roscommon: Phonology, accidence, syntax. Dublin:
University College.
Hickey, R. (ed.). 2005. Legacies of Colonial English: Studies in transported dialects. Cambridge: CUP.
Jespersen, O. [1909–1949] 1961–1970. A Modern English Grammar on Historical Principles. 7
Vols. Reprint. London: George Allen & Unwin.
Miller, K., Schrier, A., Boling, B., and Doyle, D. 2003. Irish Immigrants in the Land of Canaan:
Letters and memoirs from colonial and revolutionary America. Oxford: OUP.
Noonan, Maire. 1992. Case and Syntactic Geometry. PhD dissertation, McGill University, Montreal.
Noonan, Michael. 1995. Complementation. In Language Typology and Syntactic Description,
T. Shopen (ed), vol. 2. 42–140. Cambridge: Cambridge University Press.
O’Farrell, P. 1984. Letters from Irish Australia, 1825–1929. Belfast: Ulster Historical Foundation.
Ó Siadhail, Mícheál. 1984. Agus (is) / and: A shared syntactic feature. Celtica 16: 125–137.
Ó Siadhail, Mícheál. 1989. Modern Irish: Grammatical structure and dialectal variation. Cam-
bridge: CUP.
 Lukas Pietsch

Pedersen, H. 1913. Vergleichende Grammatik der keltischen Sprachen. Göttingen: Vandenhoeck


und Ruprecht.
Pietsch, L. In prog. Hamburg Corpus of Irish Emigrant Letters. Electronic corpus files.
Quirk, R., Greenbaum, S., Leech, G., and Svartvik, J. 1985. A Comprehensive Grammar of the
English Language. London: Longman.
Tajima, M. 1996. The common-/objective-case subject of the gerund in Middle English.
NOWELE 28/29: 569–578.
Tallerman, M. 2005. The Celtic languages. In The Oxford Handbook of Comparative Syntax, G.
Cinque and R. S. Kayne (eds), 841–879. Oxford: OUP.
Wigger, A. 1970. Nominalformen im Conamara-Irischen. Hamburg: Lütke.
section 3

Finiteness in text and discourse


Aspectotemporal connectivity in Turkic
Text construction, text subdivision,
discourse types and taxis

Lars Johanson

The paper deals with connectivity phenomena relevant for construing and
subdividing Turkic texts, particularly the ways in which aspect, actionality
and tense interact to connect utterances. The issues addressed include
aspectotemporal discourse types as textual cooccurrence patterns, the
contribution of aspectotemporal items to the expression of taxis, and
serialization by means of non-modifying converbial junctors in periodic chain
sentences, where long series of predications represent events of equal thematic
ranks. While periodic chain sentences were typical of the narrative styles of
older non‑Europeanized Turkic varieties, their modern use is strongly limited.
The dominant modern written registers construct texts according to patterns
in which subordination strongly covaries with modification. This linguistic
Europeanization has led to considerable atrophy of Turkic converbial syntax.

1. Introduction

The present paper is devoted to connectivity phenomena of relevance for text con-
struction and subdivision in Turkic languages, in particular the interaction of aspect,
actionality and tense. After a short account of the nature of this interaction, specific
aspectotemporal discourse types in the form of textual cooccurrence patterns will be
discussed. The paper will also address the question of how the interaction of aspect,
actionality and tense may contribute to the expression of taxis relations. Finally, com-
ments on the combinatory properties necessary for construing and subdividing text
portions, in particular the role of serialization by means of converbial junctors in peri-
odic chain sentences will be presented. While the issues addressed are relevant for
Turkic languages in general, the examples will be chosen from Turkish.
 Lars Johanson

2. Aspect-actionality-tense

The following discussion of the possibilities to connect utterances by means of aspec-


totemporal interaction in Turkic languages is based on a model that distinguishes the
functional layers tense, aspect and actionality (Johanson 1971, 1994a, 2000, 2001a,
2001b). Aspect is seen as operating on actionality, whereas tense operates on aspect +
actionality.
Aspect is expressed by viewpoint markers which are grammatical devices convey-
ing different perspectives of linguistically represented events relative to the event’s
limit. Events are conceived of as having an initial limit, the terminus initialis, a final
limit, the terminus finalis, and a more or less salient interval between them, the cursus.
Turkic languages express intraterminality and postterminality, two perspectives gram-
maticalized as viewpoint operators in the verbal morphology.
The intraterminal perspective, +intra, envisages, at a given vantage point, an event
within its limits, intra terminos, i.e. after its beginning and before its end. Intraterminals
are marked imperfectives offering an introspective manner of presentation that allows to
view an event from inside, and not in its totality. Combination with past tense yields
+past (+intra), ‘intraterminal-in-past’, e.g. Turkish geliyordu ‘was coming’.
The postterminal perspective, +post, envisages, at a given vantage point, an event
after the transgression of its decisive limit, post terminum, i.e. after its beginning or its
end. Postterminality is typical of resultatives and perfects. Though the event is totally
or partly absent from the view, it is still relevant at the vantage point, possibly through
observable results or traces. Combination with past tense yields +past (+post), ‘post-
terminal-in-past’, e.g. Turkish gelmişti ‘had come’.
Intraterminals and postterminals possess unmarked opposition partners that ne-
gate the respective concept, or are indifferent towards it: non-intraterminals (‑intra),
non-postterminals (‑post). The adterminal perspective, as represented by Russian and
Polish perfectives, envisages an event at the very attainment of its crucial limit, ad
terminum. Turkic languages, which lack adterminals, +ad, and non-adterminals, ‑ad,
use non-intraterminal and non-postterminal items to cover adterminal situations, e.g.
Turkish geldi ‘came’.

3. Aspect-sensitive actional categories

A classificatory framework for the analysis of the interplay of aspectual and aspect-
sensitive actional categories has been suggested in Johanson 1971: 194–233 and 2000:
145–169. The main distinctions are as follows:
Aspectotemporal connectivity in Turkic 

Table 1.

Actional structure The actional content is conceptualized

Transformative as implying transformation


Finitransformative as implying final transformation
Initiotransformative as implying initial transformation
Nontransformative without transformation

Nontransformatives do not imply any limit that must be attained in order for the ac-
tion to count as accomplished, e.g. yürü‑ ‘to move’. Transformatives do imply such a
crucial limit. With finitransformatives, this limit is the final one, e.g. gel‑ ‘to come’.
With initiotransformatives, it is the initial one, e.g. otur‑, which denote two evolu-
tional phases: a transformational ‘to sit down’ and a resulting posttransformational ‘to
sit’. The lexical meaning of initiotransformatives comprises both an initial transforma-
tion and a resulting state. The first phase stands for a telic and dynamic action, the
second one for an atelic and static action. The verb thus corresponds to both a fini-
transformative and a nontransformative verb in English, e.g. to sit down and to sit. It
can thus occur in constructions such as ‘(has) V-ed and is still V-ing’. This type can be
exemplified with an English verb such as to hide.

4. Temporally localized perspectives

A given aspectual perspective is presented as valid at some vantage point. This may be


the primary point of view, the moment of speaking, e.g. English I am writing, I have
written. An anteriority relation, +past, may establish a secondary point of view, situ-
ated prior to the primary point of view, at which a corresponding perspective is opened.
What +past markers localize on the time-axis is not a narrated event as such, but the
aspectual perspective on it. This principle is essential for the contribution of tense-as-
pect to connectivity. In a narrative ‘text world’, the deictic centre represents the ‘topic
time’ to which the plot has advanced. We are led from one point to another, each one
constituting a potential vantage point for an aspectual perspective.

5. Aspectotemporal discourse types

The contribution of tense-aspect markers to connectivity is heavily dependent on the


cooccurrence situation in various discourse types. In Johanson (1971: 76–87) it was
claimed that the semantic analysis of aspectual oppositions require an account of as-
pectotemporal discourse types, the cooccurrence patterns in which the relevant
 Lars Johanson

items may occur and compete. Discourse types with limited inventories do not allow
for the realization of all relevant values of a given system.
Modern Turkish texts offer a broad choice of aspectotemporal discourse types.
The historical narrative uses +past items at the basic discourse level. The vantage point
can move variably along the time-axis. The ‑di-based historical narrative is the most
differentiated discourse type with numerous items providing optimal contrasts. The
‑miş-based historical narrative is a discourse type used in traditional story-telling.
For the synchronous report, present tense markers such as ‑iyor describe events
simultaneous to the speech event. Each event is viewed in its course and is subse-
quently out of view. No event is viewed in its totality. Relative anteriority is signalled
by +past markers, e.g. items such as ‑di. There is also a minimal discourse type con-
sisting of a present tense in ‑mektedir and a postterminal item ‑miştir.
There are also discourse types whose basic level is not simultaneous with the
speech event, although the predications are not marked for +past. In non-deictic
‑iyor-based and ‑ir-based narratives with variable vantage points, a kind of praesens
historicum may be expressed. These discourse types have a modest inventory of items.
They do not employ the simple past in ‑di, but a retrospective view may be represented
by the postterminal item ‑miş(tir).
Evidential (indirective) discourse types are characterized by a reduced inventory
of aspectotemporal items. The indirective copula particle imiş is temporally indiffer-
ent, i.e. ambiguous between past and present time reference, e.g. Ali geliyor ‘Ali comes/
is coming’, Ali geliyordu ‘Ali was coming’ vs. Ali geliyormuş ‘It appears that Ali comes /
is coming / was coming’.

6. Aspect in dependent clauses

Aspectual values are also marked in dependent clauses, i.e. constituent clauses, relative
clauses, converb clauses and secondary predications. Different items compete with
each other and give rise to different oppositions. There are thus intraterminal converb
markers such as Turkish ‑(y)erek. Most converbs carry other semantic values than as-
pectual values. Turkic converbs have been studied from an aspectual point of view in
Johanson (1990, 1995, etc.). The connective role of aspect in converb clauses will be
dealt with below.

7. Contributions to taxis

Another kind of connectivity concerns taxis, commonly defined as information about


the temporal localization of narrated events relative to other events with respect to si-
multaneity and non-simultaneity (Jakobson 1975). Taxis relations are represented by
Aspectotemporal connectivity in Turkic 

pairs of clauses: constructions consisting of two independent clauses or containing one


dependent clause (Nedjalkov & Otaina 1987, Mal’čukov 2000). Descriptions of taxis
relations may be based on models of temporal information as dealt with in formal se-
mantics or logic-based temporal models from artificial intelligence. For a set of primi-
tive time elements related by order relations see, for example, Ma & Knight 1994.
How may actionality-aspect-tense items contribute to the expression of taxis?
They can often be interpreted as localizing one event relative to another event, to ex-
press simultaneity and non-simultaneity in ways that are deceptively similar to the
expression of taxis relations. Intraterminality may be interpreted as simultaneity, and
postterminality as anteriority. However, signalling temporal localization of events is
not the primary task of viewpoint markers. They present the narrated events as envis-
aged aspectually, specifying how they come into view at a given vantage point. One
reported event may localize the aspectual viewpoint of another reported event, i.e.
serve as the vantage point for it, e.g. Gülerek girdi ‘He entered laughing’, where girdi
‘entered’ sets the vantage point for gülerek ‘laughing’.

8. Simultaneity

Two events may be simultaneous in the sense that both occupy the same temporal
interval. Such cases can be expressed by aspectually equal independent clauses, e.g. Ali
piyano çaldı, Ahmet dinledi ‘Ali played the piano, [and] Ahmet listened’ or construc-
tions with dependent converb clauses, e.g. Ali piyano çalarken Ahmet uyudu ‘While Ali
played the piyano, Ahmet slept’. The two events may be envisaged intraterminally or
postterminally, e.g. Ali piyano çalıyor, Ahmet dinliyordu ‘Ali was playing the piyano,
[and] Ahmet was listening’; Ali piyano çalmış, Ahmet dinlemişti ‘Ali had played the
piano, [and] Ahmet had listened’.
Two events may be partially simultaneous in the sense of temporal inclusion: One
event is localized within the temporal interval occupied by another event. Both
events may be expressed by independent clauses, e.g. Ahmet girdi. Ali piyano çalıyordu
‘Ahmed entered. Ali was playing the piyano’. One of the events may be expressed by a
dependent clause, Ali piyano çalarken Ahmet girdi ‘While Ali was playing the piano,
Ahmet entered’ or Ahmet girince Ali piyano çalıyordu ‘When Ahmet entered, Ali was
playing the piano’. The including event is typically expressed by intraterminals, where-
as non-intraterminals would suggest successive events. But not even constructions
with explicit intraterminals express taxis relations accurately. The temporal relation
suggested does not create a connection between the two events, but between the inter-
val occupied by the including event and an included point of time serving as a vantage
point. What is localized is just the aspectual viewpoint. The temporal location of the
event represented by girdi or girince serves as the vantage point for an introspective
view of the including event. The construction states that the including event was in
 Lars Johanson

progress at this point, but it does not tell us when it had started or whether it was con-
tinued after that point.

9. Anteriority

One reported event may precede another event or follow it, immediately or some time
after its completion. Both events may be expressed by independent clauses, e.g. Ahmet
girdi. Ali gitmişti ‘Ahmet entered. Ali had left’. One of the events may be expressed by a
dependent clause, e.g. Ali gittikten sonra Ahmet girdi ‘After Ali had left, Ahmet entered’.
Chains of aspectotemporally equal clauses may, according to the iconic principle
of linear successivity (Johanson 1971: 246–248), be interpreted in terms of successive
events. They are, however, far from stable indicators of taxis relations, since they can
also be used for non-successive events, e.g. Ali çaldı, Ahmet dinledi ‘Ali played, [and]
Ahmet listened’.
Postterminals may often seem to suggest anteriority. ‘Postterminals-in-past’ local-
ize a postterminal perspective at a past vantage point, e.g. Ahmet girince, Ali gitmişti
‘When Ahmet entered, Ali had left’, where girince ‘when entering’ sets the vantage
point for gitmişti ‘had left’. They may often be interpreted in terms of two anteriority
relations (‘past-in-past’), suitable for ‘flashbacks’ in narratives. However, this interpre-
tation is only valid for finitransformatives. With other actional types, the event may
well continue after the vantage point. For example, with initiotransformatives, whose
crucial limit is the initial one, postterminals just signal that the initial limit is trans-
gressed. The event itself may still be in progress, e.g. Ahmet girdi. Ali uyumuştu ‘Ahmet
entered. Ali had fallen asleep/was sleeping’. Compare English had hidden, which may
also be interpreted as ‘was still hiding’.
Postterminals are thus not stable indicators of taxis relations. The temporal rela-
tion suggested does not create a connection between the two events, but rather be-
tween the interval occupied by the postterminal view of the event and a viewpoint.
Postterminals refer in an indirect way to an event that has already, entirely or partly,
disappeared from the range of vision, but is still relevant. They do not tell us when the
postterminally viewed event started or whether it continues after the vantage point.
Comrie’s definition of the aspectual ‘perfect’ type as a relation between a state and a
previous situation (1976: 52) has been criticized. The description of something as being
prior to some point on the time-axis seems to concern a temporal relation according to
Comrie’s own criterion, “tense is grammaticalised expression of location in time” (1985:
9). This apparent contradiction disappears if the posterminal perspective of the event,
not the event itself, is situated anterior to the vantage point (Johanson 2000: 104).
Aspectotemporal connectivity in Turkic 

10. Taxis in Modern Uyghur

The contribution of aspectotemporal items to connectivity in the terms of taxis is sim-


ilar in other Turkic languages. Rentzsch (2005), analyzing a passage of a literary Uy-
ghur text with respect to taxis relations, shows that the interpretation of the order of
events is totally dependent on cotextual and contextual information. The aspectual
values of the occurring items contribute to the decoding only as far as they do not
contradict the interpretations determined by cotext and context. Certain aspectoac-
tional orderings tend to suggest particular taxis interpretations, which are, however,
not compulsory. The author concludes that aspectual operators signal perspectives,
and that the taxis interpretations are just facultative secondary readings of successive
aspectoactional complexes occurring in the text.

11. Text construction and text division

It is important to try to determine the combinatory properties by means of which aspec-


totemporal items may contribute to connectivity in the broad sense of text construction.
The con­structional casting moulds are relatively similar across the Turkic languages.
The independent sentence with its ‘canonical’ structure and optimal marking for
aspect, tense, mood and person is the central item of text construction. It can be com-
pared to a tree that is firmly rooted in the discourse base. Independent sentences are
mostly of equal narrative rank and combine with each other—juxtaposed or connect-
ed by junctors—with relatively few restrictions.
Independent sentences can expand internally through subordination, other pred-
ications functioning as their dependent clauses marked by subjunctors. Dependent
clauses may express certain aspectual notions, but they are typically unmarked for
tense, mood and person. Unable to strike root in the discourse base, they remain
branches or twigs on the trunk of the independent sentence tree (cf. Johanson 1994b).
They are essential for the construction of texts, but information from the root of the
sentence is necessary for their interpretation. They normally do not express contents
of the same narrative rank as their superordinate clauses.
Converbial clauses mostly modify their matrix clauses semantically, e.g. ‑(y)ince in
Ali gelince Ahmet gitti ‘When Ali came, Ahmet left’, or Ali düşerek yaralandı ‘Ali hurt
himself by falling’ where the dependent clause functions as an adverbial modifier. Cer-
tain converbs, however, are suitable for non-modifying uses. Converb markers such as
Turkish ‑(y)ip are normally non‑modi­fying, suggesting equal narrative values in the
sense of ‘to do and [then] to do’, e.g. Ali düşüp yaralandı ‘Ali fell and hurt himself ’
rather than ‘Ali, having fallen, hurt himself ’. They can thus represent chains of events
in narrative texts, suggesting equal thematic values and sequential relations. Non-
 Lars Johanson

modifying converbs are subject to interpretations of linear successivity and may thus


have a serializing and propulsive (‘plot-advancing’) force.
Non-modification is not signalled in an unequivocal way. Certain converb clauses
vacillate beteen modifying and non‑modifying readings depending on contextual deter-
mination. Even converbs that clearly tend towards modification, e.g. ‑(y)ince, may allow
non‑modifying reinterpretations: Ali düşünce yaralandı ‘When Ali fell, he hurt himself ’
or ‘Ali fell and hurt himself ’, Ali düşerek yaralandı ‘By falling Ali hurt himself ’ or ‘Ali fell
and hurt himself ’. In narrative texts, forms that are primarily mo­difying are sometimes
used to represent events of equal rank and must be interpreted accordingly.

12. Periodic chain sentences

Non-modifying converbs may play important connective roles for construing and


subdividing larger text portions. They form the basis of the techniques of periodic
chain sentences, long series of predications representing events of equal thematic rank.
The last member of a periodic chain sentence, the independent basis, optimally rooted
with respect to tense, mood and personal reference, offers information essential for the
interpretation of the preceding members. The need for temporal-modal and illocu-
tional independence of each member is relatively limited in narrative discourse types.
In many types of older Turkic chain sentences the members can even have different
subjects (first actants) without any overt marking of subject switches.
These narrative patterns are widely spread in the world of Turkic texts, though the pre-
cise functions of the converb markers may vary to a certain extent. The patterns in question
are found in popular narrative prose as well as in highly elaborated older literary texts.
Converb markers of the Turkish type ‑(y)ip suggest sequences of events according
to the principle of linear successivity. Their function can be compared to the finite
non-intraterminal and non-postterminal -di, which can cover adterminal situations
and have a propulsive effect.

13. An example of a periodic chain sentence

Sequences of events of equal rank may be subdivided by means of non-modifying


items that signal major and minor incisions in the flow of events.
Ottoman prose often displays highly complex ramifications in the sense of divi-
sions and subdivisions, arrangements of branches and twigs. In the following long
periodic chain sentence, quoted from Evliya Çelebi’s ‘Book of Travels’, the non-intra-
terminal and non-postterminal converbs in ‑n a and ‑dïkda, which normally get mod-
ifying readings (‘when doing’), serve as non-modifying items marking major incisions,
i.e. dividing the chain into major sections (cf. Johanson 1994b).
Aspectotemporal connectivity in Turkic 

1. Emma: ne idügi maclu:mïmïz olmayup [minor incision]


2. Paša-yï müdebbirden ol maha:lle adam varïn a [major incision]
3. bizim yaγma ïlardan ikiyüz adamï Xa:n askeri ema:n vermeyüp
[minor incision]
4. kïlïjdan gečirüp [minor incision]
5. ümlesi a:m-ï šeha:deti nu:š ve γam-ï dünya:yï fera:mu:š ėdüp
[minor incision]
6. kellelerin Xa:na götürürken bu xaber-i mu:hiš pašaya gelin e [major incision]
7. ‘Bire bin-e!’ deyüp [minor incision]
8. tarfetü’l-cayn ičre bunlara kari:b vardïkda [major incision]
9. ümle bašlarï bïrakdïrup [minor incision]
10. atlarïmïza tekli:f-i ma:-la:-yuta:klar ėdüp [minor incision]
11. Mah:mu:di: begi Ibra:hi:m Beg ileride bulunup [minor incision]
12. xa:nlïlar da Bitlis derelerine girme sadedinde iken Ibra:hi:m Beg anlarï
čevirüp [minor incision]
13. čoγïnuŋ atlarï kalup [minor incision]
14. ‘Bitlis deresine a:n atam’ derken anlarï da kayd bend ėdüp [minor incision]
15. četa:čet bir xaylisi tüfengleširken Paša daxi yetišüp [minor incision]
16. altïyüz Xa:n askerinden an ak iki yüz ellisi piya:de sengista:na düšüp
[minor incision]
17. xala:s oldïlar [chain basis]
Example 1

In this text, representing a specific historical discourse type, converbs that are nor-
mally modifying may be interpreted as non-modifying items subdividing the periodic
chain sentence. The predications are ordered successively as narratively equal informa-
tion units. Though the converbial subjunctors are postpositive, they indicate relations
to the subsequent predications right down to the final basic predication. We may thus
analyze the sentence as if it would convey the following successive events:
1. But we did not know what was happening
2. By order of the prudent Pasha a man went to that place
[Then this happened:]
3. The Khan’s soldiers mercilessly killed two hundred of our pillagers
4. They put them to the sword
5. All of them emptied the cup of martyrdom and forgot the pleasures and sor-
rows of this world
6. While they took their heads to the Khan, this horrible news reached the Pasha
[Then this happened:]
7. He said: ‘Mount!’
 Lars Johanson

8. In the twinkling of an eye we got close to them


[Then this happened:]
9. We made them drop all the heads
10. We pressed our horses to the utmost
11. Ibrahim Beg, the emir of Mahmudi, was right in the front
12. Just as the men of the Khan were about to enter the Bitlis river-beds, Ibrahim
Beg made them turn around
13. The horses of most of them were left behind
14. Just as they wanted to enter the Bitlis river-bed, he caught and bound them
15. While quite a few of them were firing their muskets, the Pasha arrived
16. Of the Khan’s six hundred soldiers only 200 escaped on foot into the rocky hills
17. They saved themselves
Example 1.  Translation

In the rough analysis given here, I have taken the liberty of rendering each single pred-
ication as an independent sentence, ignoring the normal patterns of modification.
Compare the stylistically elegant translation in Dankoff (1990: 204–205):
“We had no idea of what was happening. By the time the prudent Pasha had sent
a man there to investigate, the Khan’s soldiers had mercilessly put to the sword two
hundred of the pillagers, making them quaff the wine of martyrdom and forget the
tribulations of this world, and were taking their heads to the Khan. When this horrible
news reached the Pasha, he cried ‘Mount!’ and in the twinkling of an eye we nearly
caught up with them and made them drop the heads. As we pressed our horses from
behind, Ibrahim Beg, emir of Mahmudi, appeared in front and made them turn around
just as they were about to enter the Bitlis river-beds. Many of them abandoned their
horses in order to plunge into the river and save their lives, but they were caught and
bound. Quite a few stood fast and began firing their muskets. Just then the Pasha came
up. Of the Khan’s six hundred soldiers only two hundred managed to escape into the
rocky hills, minus their horses.”

14. Connective techniques of chain sentences

Chains of the connective kind dealt with here are typical of the narrative styles of
older non‑Europeanized Turkic varieties. In these chains, converbs offered ideal de-
vices for serialization and subdivision. On the other hand, texts written in European
languages exhibit other connective devices for the construction of narrative text­s. Se-
ries of independent sentences play an important role. Rightbranching subordination
makes it possible to combine propulsive predications by means of conjunctio relativa
techniques (cf. Johanson 1975), i.e. to form chains with a progressive direction of ref-
erence, e.g. ‘X occurred, whereupon Y occurred, whereupon Z occurred’, etc. This
Aspectotemporal connectivity in Turkic 

technique is impossible within the leftbranching Turkic subordination system, but the
patterns just demonstrated have a similar effect.
Since the text structures demonstrated here defy a discursive analysis of the kind
applied to Europeanized varieties, they are very often thought not to be rule-governed
at all. The great Ottoman writer Evliya Çelebi has been criticized for his allegedly cha-
otic style, sometimes characterized as a ‘verbal jungle’. However, his prose texts ex-
hibit clear regularities which should in fact also be describable. A discourse-oriented
study of narrative techniques must take this clause combining strategy more seriously
and try to determine its intricate rules. The aspect-based connective conventions of
non‑Europeanized Turkic prose styles are still largely undescribed.
The use of these sentence types is strongly limited in modern Turkic texts. In writ-
ten Ottoman, the typical Turkic periodic chain sentences disappeared in the 19th cen-
tury under the influence of French prose styles. In most other Turkic languages, this
sentence type vanished under Russian influence. Its decline is thus an effect of linguis-
tic Europeanization. The dominant registers of modern written varieties use the sub-
ordination systems to construct texts according to European patterns in which subor-
dination strongly covaries with modification. This development has led to considerable
atrophy of Turkic converbial syntax.

References

Comrie, B. 1976. Aspect. An Introduction to the Study of Verbal Aspect and Related Problems.
Cambridge: CUP.
Comrie, B. 1985. Tense. Cambridge: CUP.
Dankoff, R. 1990. Evliya Çelebi in Bitlis. Leiden: Brill.
Jakobson, R. 1957. Shifters, Verbal Categories and the Russian Verb. Cambridge MA: Department
of Slavic Languages and Literatures, Harvard University.
Johanson, L. 1971. Aspekt im Türkischen. Vorstudien zu einer Beschreibung des türkeitürkischen
Aspektsystems [Studia Turcica Upsaliensia 1]. Uppsala: Almqvist & Wiksell.
Johanson, L. 1975. Some remarks on Turkic ‘hypotaxis’. Ural-Altaische Jahrbücher 47: 104–118.
(Reprinted in Johanson, Lars 1991. Linguistische Beiträge zur Gesamtturkologie, 26‑70. Bu-
dapest: Akadémiai Kiadó).
Johanson, L. 1990. Zur Postterminalität türkischer syndetischer Gerundien. Ural-Altaische Jahr-
bücher N. F. 9: 137–151.
Johanson, L. 1994a. Türkeitürkische Aspektotempora. In Tense Systems in European Languages,
R. Thieroff and J. Ballweg (eds), 247–266. Tübingen: Niemeyer.
Johanson, L. 1994b. Ein türkisches Erzählsatzmuster. In Gedenkschrift Wolfgang Reuschel, D.
Bellmann (ed.), 165–174. Stuttgart: Franz Steiner.
Johanson, L. 1995. On Turkic converb clauses. In Converbs in Cross-linguistic Perspective. Struc-
ture and meaning of adverbial verb forms – Adverbial participles, gerunds [Empirical Ap-
proaches to Language Typology 13], M. Haspelmath and E. König (eds), 313–347. Berlin:
Mouton de Gruyter.
 Lars Johanson

Johanson, L. 2000. Viewpoint operators in European languages. In Tense and Aspect in the Lan-
guages of Europe, Ö. Dahl (ed.), 27–187. Berlin: Mouton de Gruyter.
Johanson, L. 2001a. On three dimensions of aspectual terminality. In Aspects of Typology and
Universals [Studia Typologica 1], W. Bisang (ed.), 53–62. Berlin: Akademie Verlag.
Johanson, L. 2001b. The aspectually neutral situation type. In Aktionsart and Aspectotemporality
in Non-European Languages [Arbeiten des Seminars für Allgemeine Sprachwissenschaft
16], K. H. Ebert and F. Zúñiga (eds), 7–13. Zürich: ASAS-Verlag.
Ma, J. & Knight, B. 1994. A general temporal theory. Computer Journal 37: 114–123.
Mal’čukov, A. L. 2000. Opyt isčislenija taksisnyx značenij (na materiale tungusskix jazykov).
Issledovanija po jazykoznaniju. K 70-letiju člen-korrespondenta Rossijskoj Akademii Nauk A.
V. Bondarko, S. A. Čubik (ed.), 186–196. Sankt-Peterburg.
Nedjalkov, V. and Otaina, G.A. 1987. Tipologičeskie i sopostavitel’nye aspekty analiza zavisi-
mogo taksisa (na materiale nivxskogo jazyka v sopostavlenii s russkim). In Teorija
funkcional’noj grammatiki. Vvedenie. Aspektual’nost’. Vremennaja lokalizovannost’. Taksis,
A. V. Bondarko (ed.), 296–319. Leningrad: Nauka.
Rentzsch, J. 2005. Aspekt im Neuuigurischen [Turcologica 65]. Wiesbaden: Harrassowitz.
Connectivity by means of finite elements in
monolingual and bilingual Turkish discourse*

Birsel Karakoç
Uppsala University Sweden

The topic of this article concerns the acquisition of the connective role of finite
elements in Turkish by monolingual Turkish children and Turkish children
growing up as bilinguals in Germany. First, the structural and typological
properties of the Turkish finite verbal forms and their connective role are
briefly presented. After having discussed the research up to now on the topic of
acquisition of the Turkish finite forms, I will formulate the assumptions of the
paper, and present the data of our mono- and bilingual groups of children, which
will be analysed and compared with regard to the connective role of finite forms.

1. Introduction

In this study I deal with narratives of mono- and bilingual children between the ages
of 5 and 8. Narratives embrace different kinds of discourses, such as reports, stories,
retellings, descriptions, fairy tales and others. Since the use of the respective structures
strongly depends on the acquisition of certain discourse categories (cf. Rehbein and
Grießhaber 1992), I want to explore to what extent functions of syntactical structures
specific to discourse develop between the ages of 5 and 8 in a bilingual context. More
specifically, it is to be investigated how children acquire the ability to form concatena-
tions of speech actions in discourse.1 Discourse concatenation in Turkish, particularly,
can be established by certain uses of finite aspectotemporal elements. It is expected
that this ability of the child can be optimally analysed by means of retold narrations
(cf. Rehbein 1999b).

2. Connectivity by means of finite forms in Turkish

In Turkish, a finite predication consists of a lexeme, a predicate core, a bounded finite


element and personal suffixes. On one side, the lexical components, functionally to be
described – according to Bühler (1934) – as symbol field components, bear the predi-
 Birsel Karakoç

cates and carry the knowledge about the schemes of events, action, process of the sen-
tence. The finite elements, on the other side, enable the predicate to be anchored to the
speaker (S) and the hearer (H) for the processing of knowledge at the level of the
predication. By means of finite elements the finite verb (‘finitum’) fulfils its purpose of
anchoring the utterance as a whole in the perceived or imagined time and place of
speech (Rehbein 1995: 275–278; Rehbein and Karakoç 2004).
According to Johanson (e.g. 1971, 1994, 2000a, 2000b and this volume), the Turk-
ish verbal system as a whole is characterized by two aspectual ideas. Without any mod-
ification of the lexical content (‘symbol field’ according to Bühler and the Functional
Pragmatics framework) of a lexeme, an action can be looked at from a different point
of view, relating to the initial stage, course and end stage. It is a question of different
subjective forms of construal of on one and the same action. Intraterminality (+IN-
TRA) and postterminality (+POST) are the two possibilities of subjective points of
view. Intraterminality is the aspectual perspective which “envisages, at a given vantage
point, an event within its limits, intra terminos, i.e. after its beginning and before its
end”. Postterminality is the perspective which “envisages, at a given vantage point, an
event after the transgression of its decisive limit, post terminum, i.e. after its beginning
or its end” (Johanson, this vol.).

Figure 1.  Morphological structure of finite verbal forms and their values

From a formal-structural point of view, there are two kinds of finite suffixes: morpho-
logically simple finite suffixes which signal, in terms of content, an aspectual idea; and
morphologically complex finite forms, which are composed of a simple form plus a
copula element. The copula elements, which can occur both as free morphemes and as
suffixes, are used to express functions such as temporality,2 evidentiality or epistemic
modality. While morphologically simple finite forms only express an aspectual mean-
ing, the compound forms are marked both for aspectual value and for anteriority
(+ANT), evidentiality (+EVIDENT) or presumptivity (+PRESUM) (s. figure 1).3
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

Connectivity can be established in Turkish by means of finite aspectotemporal


structures of utterances (s. Johanson, this vol.). This kind of connectivity links two or
more different utterances, thereby forming a discourse or text. In accordance with
Rehbein (1999a), I call this kind of connectivity ‘utterance-external connectivity’. Thus,
in Turkish, connectivity of the propositions results from a consistent perspectivity of
the utterances (Rehbein 1999a: 189).
The finite verbal forms in Turkish show certain properties specific to discourse
type. Linguistic structures of similar content can differ from each other with regard to
the discourse types in which they are used.4 Within such discourses, connectivity is
utterance-externally realized by means of finite verbal forms and is subject to certain
rules. The exact functional quality of the respective finite form cannot be understood
within an individual utterance, but, rather, within larger linguistic units.5
In German, according to Redder (1992) (s. also Rehbein 1995; 1999b), the finite
verb is a combination of procedures in which the verbal lexical element is the expres-
sion of symbol field and the finite flexives occur as expressions of different field affilia-
tions (for the theory of linguistic fields with linguistic procedures in the framework of
Functional Pragmatics cf. Ehlich 1986; 1999; Rehbein 2001b and Rehbein, this vol.).
Typologically, narrations in German are constituted by deictic procedures. The con-
stellation is established by means of deictic procedures and is often continued both by
these and by refocusing procedures (s. Rehbein 1977, 1995; Redder 1992; Rehbein and
Karakoç 2004). In comparison, in Turkish the setting of the origo moving and the in-
stallation of narration space or, rather, the orientation of the H takes place via the finite
aspectual predication, and continuity in action space is ensured by sequences of these
forms. Thus, the Turkish finite aspectual forms are constitutive for discourse and text
type (s. e.g. Johanson, this vol.).
Below, I will use the term ‘aspectual discourse type’ in order to explain the quality
of certain discourse types constituted by the consistent use of aspecto-temporal forms
(s. also Rehbein and Karakoç 2004). Aspectual discourse type is strongly dependent on
the constellation of talk (s. Rehbein 1977: § 11). In this sense, I call the basis of an as-
pectual discourse type ‘discourse basis’. A finite form is called ‘discourse basis-consti-
tutive’ if it constitutes a discourse basis. I call a form ‘discourse basis-relative’ if its ex-
istence depends on the respective basis level. Thus we have two different potentials of
certain forms. Each aspectual discourse type has a specific copula form which is used
in nominal predications and marks the discourse basis. I call this the ‘basis copula’ (cf.
Johanson 1971; Karakoç 2005).
We will now look into three aspectual discourse types which have been found
empirically in the project data I have examined for this study:6 the –mIş-based dis-
course type, the -DI-based discourse type, and the –(Ø)Iyor-based discourse type.7
The -mIş-based discourse type: I speak of a mIş-based discourse type, if the finite
form –mIş provides continuity in the narrative space and is responsible for the succes-
sive concatenation. In this case, the mIş-form constitutes the discourse basis of narra-
tion together with the intraterminal form –(Ø)Iyormuş. While the mIş-form carries
 Birsel Karakoç

the action, the intraterminal form –(Ø)Iyormuş presents the action in its course. In this
way, an aspectual contrast takes place. The use of other finite forms serves for various
purposes of discourse organisation, such as anteriority (e.g. –mIştI). This means that a
temporal contrast within the given discourse can be interpreted as being background
information, previously unverbalized “forgotten” information or as an insertion. The
mIş-based narrative gives a certain framework, and a certain contour within which
only specific forms for aspectual and temporal oppositions are appropriate.
The mIş-based narrative does not give any information on the internal process of the
actions. The narrator does not involve him- or herself in the story, but gives the whole
account postterminally. The mIş-based narrative shows the features of collective
knowledge and is used in traditional story-telling.
In example (1), the five-year old monolingual child Melike (Mek) recounts the
story of Snow White on the basis of the –mIş form.8
It is interesting that Melike initially, after a short while of deliberating, begins her
utterance, but repairs it after vardı “there was/it was once” (sc 3). That means that she
thinks of which narrative type she would like to establish for her retelling. After this
attempt she decides to tell the story in a mIş-based narrative and, therefore, begins
with the predication varmış “there was/it was once, (as told, as heard)”. By means of the
nominal copulative predication varmış (var + imiş), a narration space is installed. In
this example the mIş-form is used as a discourse basis-constitutive form.
After deciding in favor of the mIş-based discourse type for the story and moving
the origo into an imagination space the child has the possibility to open a different
perspective on individual actions within this narrative basis, i.e. to use intraterminal
forms like –(Ø)Iyormuş or -(V)rmIş in order to represent the individual actions as in
the process. The predication in score areas 15–16 functions just like this Ormanda
gezinip gezinip dururmuş “He was walking in the forest”.
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 
 Birsel Karakoç
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

The -DI-based discourse type: The –DI and the intraterminal form –(Ø)Iyordu constitute
another discourse type of narration. The form –DI provides for the continuity in the
imagination space, i.e. the sequence of actions is carried out by the use of –DI. The in-
traterminal form –(Ø)Iyordu presents actions in process. A change of –DI into –
(Ø)Iyordu serves the two-dimensional aspectual representations. Other finite forms
which express relative anteriority can also be used within the –DI-based narrative, e.g.
–mIş –mIştI or -DIydI. The use of such forms can mean, as mentioned, a temporal con-
trast within the narrative. The -DI-based narrative gives a direct view of the actions and
reports on narrative actions experienced by S. The speaker is directly involved in the
story and verbalizes the events in the succession in which they originally occurred.
 Birsel Karakoç
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

The –(Ø)Iyor-based discourse type: A third narrative discourse type which occurs in my
data is constituted by the intraterminal form –(Ø)Iyor, which forms the discourse basis
in these discourses.9 In an –(Ø)Iyor-based narrative –mIş, -mIştI or –DI can express rela-
tive anteriority. The communicative function of an -(Ø)Iyor-based narration is to make
the narrative more lively. S and H seem to be directly involved in the story told. An ex-
ample for the –(Ø)Iyor-based narrative occurs in the fragment of discourse given above.
In example (1) Melike retells the story read out by the interviewer in a -mIş-discourse
type which she continuously keeps from (sc 2) to (sc 49) forming a concatenation of
speech actions. After having finished the –mIş-based discourse Melike says that she also
watched a film version of this story which she starts to retell. She retells the film version
in a –(Ø)Iyor-based narrative (s. the following example (2)).10 The postterminal form mIş
would not be suited to the retelling of a film watched directly.11
In principle it is possible to perform a retelling of a story read out by the interlocutor
in each of the discourse types mentioned. However, it is obviously first and foremost ex-
pected that the child retells a story that was told or read out to him on the basis of –mIş.
The mIş-form, as already seen in example (1), is used as a discourse basis-constitu-
tive form. It can also express the relative anteriority within the other aspecto-temporal
narrative structures. In this respect it is important to differentiate between the differ-
ent aspecto-temporal discourse types. It is especially important to consider this point
in regards to the function of -mIş. For the analysis, we have to consider exactly which
potential of that form is used and whether the form activates either its discourse basis-
constitutive function or its discourse basis-relative function. It should be considered
whether a shift from –mIş to another finite form means a shift of discourse types (a
shift from a –mIş-based discourse into another discourse type) or marks only a tempo-
ral contrast moving within the given discourse without any discourse type change.
 Birsel Karakoç

In narrative recountings, characteristically, the succession of actions is presented lin-


early (plot-line). For instance the -(Ø)Iyor-predications within a -(Ø)Iyor-based dis-
course type continue the succession of narration. If another form like –mIş is used
within this succession of actions, the action represented with this form interrupts the
succession and thus inserts a background information, a memory, a later information
etc. That means that in this case there is a temporal–like relationship between the form
of the discourse basis (-(Ø)Iyor) and the form for the relative anteriority (–mIş). The
action expressed with –mIş can only denote anterior information which is to be con-
ceived as being relative to the discourse basis. It cannot cause any absolute shift of
constellation (s. figure 2).

Figure 2.  Propositional realization of succession of actions in narrative discourse und the
role of aspectual forms in it

3. Research on the topic

There has been some research carried out on the topic of the acquisition of finite ele-
ments by monolingual Turkish children. Aksu-Koç (1988) gathered data from three
monolingual children between the age of 1;9 (21 months) and 2;6 (30 months) analys-
ing how they acquire the forms –DI and –mIş which she categorizes from a temporal-
deictical perspective of view as “past tenses”.12 According to this data, Aksu-Koç (1988)
identified -DI as being acquired at an earlier stage. The form already appears within the
first recordings, so she concludes that it is acquired before an age of 21 months. The
form –mIş appears at a later stage in her data. Looking through her data she suggests
that the hearsay is the last function of –mIş to be acquired. Taking the verbalization of
certain picture sequences as a basis, she identified that only one child aged 2;1 (25
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

months) reveals some narrative utterances with –mIş (cf. Slobin and Aksu-Koç 1982;
Aksu-Koç and Slobin 1985; Aksu-Koç 2000).
Aksu-Koç (1994) investigates the data from older children and adults. This data is
elicited with the help of pictures from the frog story (Slobin 1985). The children were
compiled within the context of a cross sectional analysis in the age groups of 3, 5 and
9 and the adults within an age group of 20–24. Aksu-Koç (1994) does not apply the
conception of aspectotemporal discourse types known since Johanson (1971), but
rather interprets the aspects as temporal deixis, noting that one finite form functions
as “anchor tense” or “favored tense” in the Turkish narratives. Taking this as a basis she
points out that three forms –(Ø)Iyor, -DI and –mIş have been investigated from the age
of three onwards, but without a clear function of discourse organization. The children
are not able to maintain the dominant tense over several utterances. More than half of
the children from the group of 5-year olds remain on the line of dominant form and
can realize an adequate form shift. In the group of children aged 9 all speakers are ca-
pable of mastering the discursive and connective use of forms.
Aarssen (1996) investigates how children in the age range from 4 to 10 develop the
ability to conceptualise and express “temporal structures” in narratives in Turkish and
Dutch and analyses the narratives of 20 bilingual informants per age group. He com-
pares the findings with those of monolingual informants (10 per age group) (1996:
123). In the data of bilinguals as well as of monolinguals he finds that the children have
difficulty in establishing a “stable temporal frame” in which they can tell their stories
and that a lot of “unmotivated tense shifts” are found. He remarks that for the bilingual
children a development towards the use of –DI and –mIş as “anchor tenses” was found,
while the monolingual informants retell their stories mainly in –(Ø)Iyor-form (Aars-
sen 1996: 137–156).
Rehbein and Karakoç (2004) deal with how the finite elements in the retellings of
Turkish-German bilingual and Turkish monolingual children are used for the purposes
of discursive connectivity. Analysing a large data basis they find that the number and the
range of different finite forms are significantly higher with monolinguals than with bilin-
guals13 and that the –(Ø)Iyor form in monolingual Turkish is highly preferred. Further-
more they observe that shifts of aspecto-temporal forms occur in the sample of bilingual
informants and that two thirds of these switches are unacceptable (2004: 142).

4. Assumptions and research questions

The present study is based on the general assumption that the acquisition of linguistic
forms takes place within the framework of larger units (discourse forms) due to the
acquisition of their function (Rehbein and Grießhaber 1992). The emphasis lies on the
following research questions.
 Birsel Karakoç

1 Which special constellation do the children mentally refer to, i.e. hearsay, report of the
film version of the story, or the reconstruction of the story with the help of pictures?
2 Do the children establish an aspectual discourse structure? Which discourse types
occur in children’s retellings? How do they handle the specific complex require-
ments of each (individual) discourse type? What do the realizations of the constel-
lations look like?
3 How long are the concatenations of utterances of the children? In this context, we
first have to precisely investigate whether the child’s utterances are elicitated by the
interlocutor, or whether the children are able to initiate concatenations of speech
actions themselves. How is the continuity in imagination space maintained? Are
these discourses characterized as sequences or as concatenations of utterances?
With respect to this, within the specific constellation of the narration, I will look into
the proportional relationship between the children’s and the interlocutor’s utterances.
4 How do the children handle the principle of serialisation within a choosen dis-
course type? Do they, in an appropriate way, verbalize different perspectives on
specific actions, or express temporal oppositions within a discourse structure?

5. The data

I will investigate the questions mentioned above on the basis of linguistic data which
include the accounts derived from a picture book (Snow White).14 The story is read out
to the child from the picture book by the interviewer, and then the child is asked to
retell the story (cf. footnote 10).

Table 1.  Age groups

monolingual bilingual

Age group Number of children Age group Number of children


5 2 5 2
6 2 6 2
7 2 7 2
8 2 8 2

As shown in table 1, I have analysed sixteen discourses from sixteen different children.
Eight discourses are from monolingual children, eight from bilingual children. For
both the mono- and the bilingual group, four groups of 5, 6, 7 and 8-year-olds have
been formed. This allows us both to analyse the development of the children between
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

the ages of 5 and 8 longitudinally and to compare corresponding age groups between
the mono- and bilingual children.

6. On the acquisition of aspecto-temporal discourse structures by


monolingual children

Let us take a look at the monolingual group.


1. Concerning the question of special constellation, I observed that 55 percent of the
stories retold by monolinguals refer to the story that was read by the interviewers.
27 percent refer to the film version of the story. The remaining 18 percent retell the
story with reference to the pictures in the book. However, the stories referring to the
pictures were told in addition to the original stories (table 2).

Table 2.  Occurrences of each constellation (monolingual children)

Reliance on read- Reliance on film Reliance on pictures in


aloud story version of the story the story book

5-year-olds 2 1 1
6-year-olds 1 1
7-year-olds 2 1
8-year-olds 1 1

Total 6 = 55 % 3 = 27 % 2 = 18 %

2. As far as the diversity in the use of aspecto-temporal discourse types is concerned, I


found that, within the monolingual groups, there are a number of different types (table
3). Already in the category of 5-year-old children, all three of the possible discourse
structures occur. This seems to match the findings of Aksu-Koç (1994) that more than
half the children from the group of 5-year old monolinguals investigated in her study
remain on the line of “dominant form” constituting a discourse structure (but cf. Aars-
sen 1996 which finds that both bilinguals and monolinguals have difficulties in estab-
lishing a “stable temporal frame”). To sum up, 11 different discourse types can be found
among the eight children in the monolingual group. I noticed that children in the
group of 5-year-olds usually retell the story twice by changing the constellation. As
shown in example (1) above, the 5-year-old child first refers mentally to the story read
to her, retelling her own story in a mIş-based discourse. Then she retells the story once
more (example 2), basing it on a film version she had seen (s. the remark in footnote 10).
In this case, she retells the story according to a different discourse structure, namely an
–(Ø)Iyor-based narrative. This means that she is also able to change the discourse types
according to the different patterns in her knowledge.
 Birsel Karakoç

Table 3.  Occurrences of aspectotemporal discourse types (monolingual children)

-mIş-based -DI-based –(Ø)Iyor-based Discontinuous


discourse discourse discourse discourse

5-year-olds 2 1 1 -
6-year-olds 1 1 -
7-year-olds 2 1 -
8-year-olds 1 1 -

Total 3 4 4 -

3. Table 4 below gives us an impression of the quantitative distribution of utterances.


We find that the discourses of monolinguals from the age of 5 and onward were estab-
lished by means of aspectual forms following one another and therefore building utter-
ance concatenations. They were able to form reasonably long narratives, as can be seen
in the examples above. The 5-year-old monolingual child Melike retells the story from
score area 2 up to 49 all by herself (example (1)). She carries out a long concatenation
of aspecto-temporal forms, a narrative discourse structure in –mIş. We observed that
the other 5-year-old child displayed a similar result, although she does not retell the
story in a detailed and lengthy version, as she only remembers certain events. How-
ever, she keeps the aspectual narrative within a long concatenation without the help of
the interviewer.
In the 5-year-old monolingual children, the utterances of the children form 91 per
cent of the total utterances. In the 6-year-old children, we have calculated 90 per cent,
in the 7-year-old children 86 percent, and in the 8-year-olds 98 percent. This means
that the interlocutor does not co-constitute the narrative retelling.

Table 4.  Percentage of child utterances in total of utterances (monolingual)

Total of utterances in absolute Child utterances %


numbers in absolute (Utterances of child)
(Interlocutor + Child) numbers

5-year-olds 147 134 91%


6-year-olds 153 139 90%
7-year-olds 155 134 86%
8-year-olds 175 172 98%

Taking the results from table 4 as a basis, I would like to sum up certain tendencies
within the monolingual group. On the whole, the concatenations of speech actions are
typical for this group. However, this should not be strictly interpreted in a way that the
interlocutor does not participate in the narrative story of the child at all. Rather, it
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

should be considered as a tendency in development that the children make up the se-
quences of actions on their own in 86 to 98 per cent of the cases (s. table 5).

Table 5.  Overall characterization of discourses with regard to the initiation of concatena-
tions (monolingual children)

Interlocutor-initiated Child-initiated
Concatenation Concatenation

5-year-olds - +
6-year-olds - +
7-year-olds - +
8-year-olds - +

Table 6.  Shifts of finite forms within the given aspectual discourse type (monolingual children)

-mIş-based discourse -DI-based discourse –(Ø)Iyor-based discourse

5-year- Discourse type occurs yes Discourse type occurs yes Discourse type occurs yes
olds Shift of forms 8 Shift of forms 1 Shift of forms 4
Inadequate shifts - Inadequate shifts - Inadequate shifts -

6-year- Discourse type occurs no Discourse type occurs yes Discourse type occurs yes
olds Shift of forms Shift of forms 13 Shift of forms 1
Inadequate shifts Inadequate shifts - Inadequate shifts -

7-year- Discourse type occurs no Discourse type occurs yes Discourse type occurs yes
olds Shift of forms Shift of forms 9 Shift of forms 10
Inadequate shifts Inadequate shifts - Inadequate shifts -

8-year- Discourse type occurs yes Discourse type occurs no Discourse type occurs yes
olds Shift of forms 2 Shift of forms Shift of forms 2
Inadequate shifts - Inadequate shifts Inadequate shifts -

4. On the level of given discourse structures, I am investigating the possible diversity


of finite forms which are used to realize aspecto-temporal oppositions. The finite forms
which occur within each particular discourse type are used adequately by the whole
group of monolingual children (s. table 6). This finding corresponds with the result in
Aksu-Koç (1994) that the monolingual children in the same age range are capable of
establishing aspecto-temporal narratives and performing adequate form shifts within
these structures (By way of contrast, Aarssen 1996 found that in the data of bilinguals
as well as of monolinguals a lot of “unmotivated tense shifts” occurred).
 Birsel Karakoç

7. On the acquisition of aspecto-temporal discourse structures by bilingual


children

We will now look at the results from the bilingual group:


1. In the bilingual data, I found that the children in 75 per cent of the constellations
retell the story with reference to the pictures of the book. In contrast to the monolin-
gual group, the bilingual group makes more use of the procedure of co-construction
and uses the pictures in the book as a basis more often. They describe what they see in
the pictures (from the original speech situation).15

Table 7.  Occurrences of each constellation (bilingual children)

Reliance on read- Reliance on film Reliance on pictures in the


aloud story version of the story story book

5-year-olds 2
6-year-olds 2
7-year-olds 1 1
8-year-olds 1 1

Total 1 1 6 = 75 %

2. In the data of 5-year-old children, we find a kind of question-answer pattern where the
answers consist mostly of one utterance. Connected sequences and utterance-spanning
with finite elements can hardly be found. Thus, connectivity is established mainly by
means of co-construction between H and S and with question-answer sequences.
In the group of 6-year-olds, an attempt to build concatenated sequences of speech
actions that go beyond the scope of one utterance can be observed, although in gen-
eral most discourses seem still to be interlocutor-initiated and the question-answer-
character is predominant. Characteristic of these discourses is that the 6-year-old chil-
dren establish discontinuous aspecto-temporal discourses. This means that they often
show shifts from one discourse basis to another. The basis established initially is not
continued and the origo is not maintained. In comparison with the group of 6-year-
olds, the children aged 7 and 8 years reveal longer child-initiated concatenations and
realizations of aspectual narratives (s. table 8). At this age range of bilingual children I
found –DI and –(Ø)Iyor-based narratives. –mIş-based narratives do not occur in the
bilingual data I have investigated for this study. By way of contrast, Aarssen (1996)
remarks that for the bilingual children a development towards the use of –DI and –mIş
as “anchor tenses” was found (Aarssen 1996: 137–156).
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

Table 8.  Occurrences of aspectotemporal discourse types (bilingual children)

-mIş-based -DI-based –(Ø)Iyor-based Discontinuous


discourse discourse discourse discourse

5-year-olds - - - -
6-year-olds - - - +
7-year-olds - 1 - -
8-year-olds 1 1 -

Total - 2 1 +

As already mentioned, we observed discontinuous narrative structures in the data of


the 6-year-old bilingual children. Through such unmotivated shifts of finite forms H
must repetitively orientate herself/himself again and again, and at least a monolingual
Turkish H will get confused, being forced to focus and re-focus a new point in time
from utterance to utterance. The perpetuation of the established discourse basis struc-
ture is important in that the origo is anchored and an imagination space is established
through it. Therefore the unmotivated shift of forms means that the origo is newly es-
tablished and consequently the H cannot easily grasp the succession of actions (Reh-
bein and Karakoç 2004).
In the following I will give an example of this phenomenon: example (3) is taken
from the discourse of the 6-year-old bilingual boy Kubat. It shows different finite ele-
ments following each other. In score area 52, a DI-predication occurs, whereas in score
area 54 a –mIş, and in score area 55 an –(Ø)Iyor-predicate are used.
In this case the question is: why can the use of these different finite forms not be
understood as shifts of forms for an aspecto-temporal purpose within an established
discourse structure (e.g. in the sense of relative anteriority), but rather as a shift of the
discourse basis itself? In narrative discourses, it is characteristic that the sequences of
action are presented in a linear form. The forms used, one after another, continue the
sequence of the story. If another finite element is used within this plot, the actions
presented with this element may interrupt the sequence. These interruptions may ex-
press background information, an insertion, something remembered by S and so on
(cf. figure 2 and the explanation for it above).
 Birsel Karakoç
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

I would now like to return to example 3. We must answer the question whether we are
dealing with an inadequate shift of discourse basis, or with a normal shift of forms
within the discourse structure which has already been established. Every action is
combined with different finite forms, first dedi, [say-DI] “he said” (score area 52), then
demiş [say-mIş] “he said” (score area 54) and then söylüyor [say-(Ø)Iyor] “he says”
(score area 55) (s. figure 3). But in this case the plot (sequences of action) is linear.
None of these actions interrupt the succession in terms of the content. One event hap-
pens after another. This sequence in terms of content is not linguistically reproduced
adequately by using the finite predications. In contrast to this we saw in examples (1)
and (2) that continuity is linguistically maintained by means of the consistent use of
finite elements of the given discourse basis.16

Action1 [-DI] + Action2 [-mIş] + Action3 [-(Ø)Iyor]

de-di + de-miş + söyl-üyor

say-DI + say-mIş + say-(Ø)Iyor

Figure 3.  Realisation of succession of actions in 6-year-old bilingual child Kubat

In the groups of 7 and 8 year old children (especially one 7-year-old bilingual child in
my data), I have observed a stabilization of discourse basis. Finite elements play a role
 Birsel Karakoç

by establishing and continuing the origo. Long consistent concatenations of finite ele-
ments can be found both in –DI- and in –(Ø)Iyor-narratives.
In the entire bilingual group, no discourse structure with -mIş can be found, al-
though -mIş as a finite verbal form in single utterances sometimes occurs (cf. Karakoç
2006 on the realization of evidential forms in bilingual Turkish children; Herkenrath
and Karakoç 2006; cf. also Pfaff 1993 and Boeschoten 1990).
3. In the group of 5-year-olds, we found a high level of speech activity on the part of
the interlocutor. However, in this group a question-answer-pattern with a high H-ini-
tiated activity occurs, yet no narrative concatenations of utterances. In the group of
6-year-old children, we have a result of 56 percent and in the group of 7-year-old chil-
dren a result of 84 percent of child utterances. In this context the high number of ut-
terances can be explained as follows: one child in this group shows an advanced ability
to retell and can keep the basis line of discourse over longer utterances. It is of course
important to consider the individual abilities of the children. In the 8-year-old chil-
dren, we find a lower percentage rate again (54 percent) (s. table 9).

Table 9.  Percentage of child utterances of the total utterances (bilingual)

Total of utterances in absolute Child utterances in %


numbers absolute numbers (Utterances of
(Interlocutor + Child) child)

5-year-olds 125 50 40 %
6-year-olds 229 129 56 %
7-year-olds 141 118 84 %
8-year-olds 149 79 54 %

If we sum up the tendencies within the whole bilingual group in the data under inves-
tigation, we observe much more co-construction and H-initiated activities in contrast
to the monolingual group of children (table 10).

Table 10.  Overall characterization of discourses with regard to initiation of concatenations


(bilingual children)

Interlocutor-initiated Child-initiated
Concatenation Concatenation

5-year-olds + -
6-year-olds (+) (+)
7-year-olds - +
8-year-olds - +
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

4. If we closely consider the various types of aspecto-temporal structures that occur on


a whole, we can see that form shifts occur for temporal purposes. All in all, we have
two types of aspecto-temporal discourses within the whole bilingual group. Within
these discourses, there are numerous adequate occurrences and a couple of inadequate
shifts of forms (s. table 11).

Table 11.  Shifts of finite forms within the given aspectual discourse type (bilingual children)

-mIş-based discourse -DI-based discourse –(Ø)Iyor- based discourse

5-year- Discourse type occurs no Discourse type occurs no Discourse type occurs no
olds
Shift of forms Shift of forms Shift of forms
Inadequate shifts Inadequate shifts Inadequate shifts

6-year- Discourse type occurs no Discourse type occurs no Discourse type occurs no
olds
Shift of forms Shift of forms Shift of forms
Inadequate shifts Inadequate shifts Inadequate shifts

7-year- Discourse type occurs no Discourse type occurs yes Discourse type occurs no
olds
Shift of forms Shift of forms 17 Shift of forms
Inadequate shifts Inadequate shifts 3 Inadequate shifts

8-year- Discourse type occurs no Discourse type occurs yes Discourse type occurs yes
olds
Shift of forms Shift of forms 1 Shift of forms 20
Inadequate shifts Inadequate shifts - Inadequate shifts 3

8. Three levels of linguistic development

Looking at the results presented up until now, the discursive properties of connectivity,
which are constituted by finite elements, can be compared by characterizing them on
three different levels of linguistic development (s. tables 12 and 13).
 Birsel Karakoç

Table 12.  Discursive features of finite connectivity in monolingual children

discursive features of finite connectivity

Child-initiated (Concatenation)

Interlocutor- Discontinuity in the Continuity in the discourse basis


initiated discourse basis Problems with the Adequate
(Sequence) perspectivization shift of forms
and temporal
contrast
5-year-olds +
6-year-olds +
7-year-olds +
8-year-olds +

Table 13.  Discursive features of finite connectivity in bilingual children

discursive features of finite connectivity

Child-initiated (Concatenation)

Interlocutor- Discontinuity in the Continuity in the discourse basis


initiated discourse basis Problems with the Adequate
(Sequence) perspectivization shift of forms
and temporal
contrast
5-year-olds +
6-year-olds +
7-year-olds +?
8-year-olds +?

On the first level of linguistic development, we ask the question whether narrative
concatenations occur that are adequate to the constellation of the retelling. All in all,
the child-initiated concatenations predominantly occur in the monolingual group,
compared to an age dependent sequence of development in the bilingual group. With-
in the bilingual group the percentage rate of co-operation by the interlocutor is rather
high in the 5-year-old group. The connectivity is established up to 60 per cent by the
H. We are confronted with a question-answer-pattern rather than with narrative ac-
counts. The actions of the story to be retold are reconstructed step-by-step with the
help of questions and individual answer utterances by the child.
On the second level of linguistic development, the question is which features the
existing concatenations show. Another question is whether the forms of discourse ba-
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

sis are consistently continued in order to present a coherent plot, or whether there are
unmotivated shifts. And finally, we should ask whether there are any unmotivated
shifts from one discourse basis to another. In the group of 6-year-old bilingual chil-
dren we find, as mentioned earlier, attempts at establishing an aspectual discourse ba-
sis. Yet the children fail maintain the chosen basis consistently. We can see these chil-
dren have already acquired the morphological structures but not the complete
discursive functions of the finite forms.
The third level of linguistic development concerns the use of different possible fi-
nite forms within an established discourse type. In the bilingual group, it seems that
there are some unmotivated, inadequate shifts and thus some problems regarding the
perspectivization and the linguistic realization of temporal contrast.

9. Conclusions

The study found that in the data of bilingual informants the Turkish finite aspecto-
temporal elements function much less as discourse type-constitutive elements than in
the data of monolingual informants. Especially –mIş-based narratives do not occur in
the bilingual data investigated for this study. One could say the bilinguals seem to
prefer a dual system with –(Ø)Iyor and –DI (Rehbein and Karakoç 2004; Karakoç 2006;
Herkenrath and Karakoç 2006; cf. the finding by Aarssen 1996 which seems not to
correspond with this result). By contrast, we observe a highly frequent use of tempo-
ral-deictic expressions such o zaman, sonra, ondan sonra etc. in the bilingual group (cf.
example (3) above). Consequently, the utterance-external connectivity seems to
be maintained by means of these elements.17 In this context we may talk about a func-
tional expansion of the deictic elements in contact Turkish (cf. Rehbein 2001a; Her­
kenrath, Karakoç and Rehbein 2003; Rehbein and Karakoç 2004).
The results are based on the analysis of data of the evocative field experiment (EFE
1) “retelling a story / fairy tale which is read out”. Although it is expected that the children
verbalize the retellings of a story read out by the adult person in all discourse types men-
tioned above, it would be interesting to analyse more data collected by using other evoc-
ative field experiments (homileïc discourses, biographic narratives, movie retellings, re-
telling of a self-read book, talking about a picture, talking about games played during the
recording etc.) in order to investigate which discourse type is preferred by children for
which constellation type and which acquisition patterns can be observed. By just taking
e.g. the description of a picture sequence presented to S and H (e.g. Aksu-Koç 1994;
Slobin 1985) or a movie retelling (e.g. Erguvanlı Taylan 1987) as a basis the whole poten-
tial range of connective use of these forms cannot be completely presented.
The study only dealt with the use and function of Turkish aspecto-temporal ele-
ments in spoken narratives. It would be also very interesting to see how the acquisition
and conceptualization of written aspecto-temporal text types take place in bilingual
school constellations.
 Birsel Karakoç

Notes

* This work is a result of the research project SKOBI (Linguistic connectivity in bilingual
Turkish-German children – Principal investigator: Prof. Dr. Jochen Rehbein), which is settled in
the SFB 538 Mehrsprachigkeit (Collaborative Research Center No. 538 Multilingualism), fun-
ded by the Deutsche Forschungsgemeinschaft (German Research Foundation). The project
SKOBI deals with the linguistic devices of connectivity acquired by Turkish children in a bilin-
gual environment through their interaction with German. The results are compared with those
of monolingual Turkish children. I would like to express my special thanks to Prof. Dr. Jochen
Rehbein and Annette Herkenrath, and two anonymous reviewers for their valuable comments
and suggestions.
1. I would like to briefly explain the theoretically important terms of “concatenation” vs. “se-
quentialisation”, and “Hearer-initiation” (“H-initiation”) vs. “Speaker-initiation” (“S-initiation”),
which are used for categorization in the following data analysis. These terms were established by
Rehbein (1984). In the linguistic activities of narrating, describing and reporting, we deal not
with sequences of interactional activities between speaker (S) and hearer (H). Instead, the main
emphasis is placed on a long lasting activity of an individual speaker. For sequences of linguistic
activities carried out systematically one after the other by the same speaker, the term “concate-
nation” (concatenation of speech actions) has been proposed within the theoretical framework
of Functional Pragmatics (FP, cf. Ehlich and Rehbein 1986). If child utterances are elicited by an
interlocutor, discourse intention as well as the retold narrative story are H-initiated (interlocu-
tor-initiated). In this case, we have to deal with a condition of sequentialisation (sequentialised
concatenation). If, however, the utterances of the child follow each other, the narration is S-ini-
tiated (child-initiated), and we are dealing with the situation of concatenation of speech actions
(S-initiated concatenation) (Rehbein 1999b: 13).
2. «Anteriority» as a temporal notion is conceived as «Abstandnahme von der nächstliegen-
den Realität» (self-distancing from immediate reality) and “eine Relation der Reihenfolge, die
zwei Terme voraussetzt” (a relation of sequence based on two terms as a prerequisite, cf. Johanson
1971: 52–55).
3. The idea of “focality” is not discussed here because it does not play a role in the data ana-
lysed in this study (cf. on this point e.g. Johanson 2000b).
4. For the discursive distribution of some finite verbal forms in Noghay, another Turkic lan-
guage, cf. Karakoç (2000) and (2005).
5. Johanson (1971) introduced the notion that in Turkish all important text and discourse
types are based on the cooccurence of specific finite forms.
6. For further discourse and text types constituted by use of other aspectotemporal forms cf.
Johanson (1971: 76–87).
7. In Turkish, another narrative type exists which is based on the finite form –(V)r. One func-
tion of this discourse type is to make the narrative more lively than the –(Ø)Iyor-based dis-
course type, but it usually occurs in traditional story telling. I have not found any narrative dis-
course type which is based on -(V)r in the data I investigated for this study.
8. Examples in the general section on Turkish finite elements are taken from the project data
analysed in this study. However, they are not discussed from the acqusitional point of view here,
but rather serve as empirical illustrations of the linguistic phenomena mentioned. It should also
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

be mentioned that the child language data are presented just as the children speak, i. e. the mis-
takes of children are not corrected by us.
9. The –(Ø)Iyor-based discourse type is primarily used to form descriptions, accounts of facts,
reports etc. –(Ø)Iyor also may constitute narrative discourse (cf. Johanson 1971: 80). In this
study –(Ø)Iyor-based descriptions, accounts etc. do not play a role. They are thus not dealt with
further.
10. In the present study, I observed that some of the children wanted to retell the film version
of the story about Snow White, even though the interlocutor had read out the story from the
book. The film version of the story was not shown to the children, but is well known to most of
the children from their own experience.
11. It is thus not surprising that Erguvanlı Taylan (1987) did not observe any –mIş-based nar-
ratives in her data of retellings of films.
12. These forms differ from each other, as mentioned above, with regard to their aspectuality
but not with regard to their temporality. -mIş is a form expressing evidentiality. In comparison,
-DI does not show any evidential meanings (s. e.g. Csató 2000 and Johanson 2000a).
13. Cf. the results in Herkenrath and Karakoç (2002) in which a reduced number and range of
Turkish subordination elements can be observed in bilingual informants.
14. The experiment book used for the data collection is not a picture book where the individual
sequences of the events are presented by means of the pictures (as used e.g. by Slobin 1985), and
includes few pictures only.
15. According to Aksu-Koç (1988) the verbalization of knowledge gained by hearsay is the
function of –mIş that children acquire last. The ability to verbalize complex discourse forms is
necessary in order to depict facts that are absent in terms of time and space and are thus not
directly visible to a communication partner in the field of perception. In this regard narration
and narrative abilities seem to be very important in language acquisition (Rehbein 1999b).
16. This partial discontinuity of the ongoing plot line is, according to my investigations, also
caused (or motivated) by the integration of reported speech into the action line of the basis
discourse level. The integration of direct speech into the real story basis seems to be a very com-
plex matter for the children. I will not deal with this point in any further detail.
17. For the use of deictic elements in bilingual children s. Rehbein (2001a); for the use of o
zaman in bilingual children s. Baumgarten et al. (this vol.).
 Birsel Karakoç

Abbreviations

3SG 3rd person singular


ABL Ablative
ACC Accusative
ANT Anteriority
CAU Causative
CDCOP Conditional copula
COP Copula
CV Converb
CVCOP Converbial copula
DAT Dative
DEI Deixis
DIM Diminutive
FUT Future
GEN Genitive
IJ Interjection
ICOP Indirective copula
INS Instrumental
LOC Locative
MOD Modal
NEG Negation
PAR Participle
PAS Passive
OPT Optative
PCOP Past tense copula
PL Plural
POP Postposition
PRS Present tense
PST Past tense
PSS Possessive
PTC Particle
PTE Postterminality
PVERB Postverb
Q Interrogativity marker
REF Reflexive
REC Reciprocal
⇔TEMP Temporal opposition
VN Verbalnoun
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

References

J. 1996. Relating Events in Two Languages. Acquisition of cohesive devices by Turkish-


Aarssen,����������
Dutch bilingual children at a school age. Tilburg: Tilburg University Press.
Aksu-Koç, A. 1988. The Acquisition of Aspect and Modality. The case of past reference in Turkish.
Cambridge: CUP.
Aksu-Koç, A. 1994. Development of linguistic forms: Turkish. In Relating Events in Narrative. A
crosslinguistic developmental study, R. A. Berman and D. Slobin (eds), 329–392. Hillsdale
NJ: Lawrence Erlbaum.
Aksu-Koç, A. 2000. Some aspects of the acquisition of evidentials in Turkish. Evidentials. Turkic,
Iranian and neighbouring languages, L. Johanson and B. Utas (eds), 15–28. Berlin: Mouton
de Gruyter.
Aksu-Koç, A. and Slobin, D. I. 1985. The acquisition of Turkish. In The Crosslinguistic study of
Language Acquisition. I: The data, D. I. Slobin (ed.), 839–878. Hillsdale NJ: Lawrence Erl-
baum.
Baumgarten, N., Herkenrath, A., Schmidt, T., Wörner, K. and Zeevaert, L. This volume. Study-
ing connectivity with the help of computer-readable corpora: Some exemplary analyses
from modern and historical, written and spoken corpora.
Boeschoten, H. E. 1990. Acquisition of Turkish by Immigrant children. A multiple case study of
Turkish children in the Netherlands aged 4 to 6. Wiesbaden: Harrassowitz.
Bühler, K. 1934. Sprachtheorie. Die Darstellungsfunktion der Sprache. Jena: Fischer (repr. 1978
Frankfurt/M.: Ullstein).
Csató, É. Á. 2000. Turkish miş- and imiş-items. Dimensions of a functional analysis. Evidentials.
Turkic, Iranian and Neighbouring Languages, L. Johanson and B. Utas (eds), 29–43. Berlin:
Mouton de Gruyter.
Ehlich, K. 1986. Funktional-pragmatische Diskursanalyse – Ziele und Verfahren. Verbale Inter-
aktion. Studien zur Emprie und Methodologie der Pragmatik, D. Flader, D. (ed.), 127–143.
Stuttgart: Metzler.
Ehlich, K. 1999. Funktionale Pragmatik – Terme, Themen und Methoden. Deutschunterricht in
Japan 4: 4–24.
Ehlich, K. and Rehbein, J. 1986. Muster und Institution. Untersuchungen zur schulischen Kom-
munikation. Tübingen: Narr.
Erguvanlı Taylan, E. 1987. Tense variation in Turkish narratives. In Studies on Modern Turkish.
Proceedings of the Third Conference on Turkish Linguistics, H. E. Boeschoten and L. Th.
Verhoeven (eds), 177–188. Tilburg: Tilburg University Press.
Erguvanlı Taylan, E. (ed.) 2001. The Verb in Turkish. Amsterdam: John Benjamins.
Herkenrath, A. and Karakoç, B. 2002. Zum Erwerb von Verfahren der Subordination bei türkisch
- deutsch bilingualen Kindern - Transkripte und quantitative Aspekte [Arbeitspapiere zur
Mehrsprachigkeit, Folge B, 37]. Hamburg: Universität Hamburg, SFB 538 Mehrsprachig-
keit.
Herkenrath, A. and Karakoç, B. 2006. Linguistic connectivity in German-Turkish language con-
tact. Talk given in Jahrestagung der Societas Linguistica Europaea 2006: Relativism and
Universalism in Linguistics, Workshop on Multilingualism and Universal Principles of Lin-
guistic Change Universität Bremen, 30. August -02. September 2006.
Herkenrath, A. and Karakoç, B. and Rehbein, J. 2003. Interrogative elements as subordinators in
Turkish – aspects of Turkish-German bilingual children’s language use. (Non)vulnerable
 Birsel Karakoç

Domains in Bilingualism [Hamburg Studies on Multilingualism 1], N. Müller (ed.), 221–


269. Amsterdam: John Benjamins.
Johanson, L. 1971. Aspekt im Türkischen. Vorstudien zu einer Beschreibung des türkeitürkischen
Aspektsystems [Acta Universitatis Upsaliensis. Studia Turcica Upsaliensia I]. Uppsala: Uni-
versity of Uppsala.
Johanson, L. 1994. Türkeitürkische Aspektotempora. In Tense systems in European languages, R.
Thieroff and J. Ballweg (eds), 247–266. Tübingen: Niemeyer.
Johanson, L. 2000a. Turkic indirectives. Evidentials. Turkic, Iranian and neighbouring languages,
L. Johanson and B. Utas (eds), 61–87. Berlin: Mouton de Gruyter.
Johanson, L. 2000b. Viewpoint operators in European languages. In Tense and Aspect in the
Languages of Europe, Ö. Dahl (ed.), 27–187. Berlin: Mouton de Gruyter.
Johanson, L. and Utas, B. (eds). 2000. Evidentials. Turkic, Iranian and neighbouring languages.
Berlin: Mouton de Gruyter.
Karakoç, B. 2000. The finite copula bol- in Noghay and its functional equivalents in Turkish.
Studies on Turkish and Turkic Languages, A. Göksel and C. Kerslake (eds.), 143–149. Wies-
baden: Harrassowitz.
Karakoç, B. 2005. Das finite Verbalsystem im Nogaischen [Turcologica 58]. Wiesbaden: Har-
rassowitz.
Karakoç, B. 2006. Evidentiality in Turkish-German bilingual children. Talk given at the 13th
International Conference on Turkish Linguistics, Uppsala University, 16–20. August 2006.
Pfaff, C. 1993. Turkish language development in Germany. In Immigrant Languages in Europe,
G. Extra and L. Verhoeven (eds), 119–157. Clevedon: Multilingual Matters.
Redder, A. 1992. Funktional-grammatischer Aufbau des Verb-Systems im Deutschen. In Deut-
sche Syntax. Ansichten und Aussichten, L. Hoffmann (ed.), 128–154. Berlin: Walter de Gru-
yter.
Rehbein, J. 1977. Komplexes Handeln. Elemente zur Handlungstheorie der Sprache. Stuttgart:
Metzler.
Rehbein, J. 1984. Beschreiben, Berichten und Erzählen. Erzählen in der Schule, K. Ehlich (ed.),
67–124. Tübingen: Narr.
Rehbein, J. 1995. Grammatik kontrastiv – am Beispiel von Problemen mit der Stellung finiter
Elemente. Jahrbuch Deutsch als Fremdsprache 21: 265–292.
Rehbein, J. 1999a. Konnektivität im Kontrast. Zu Struktur und Funktion türkischer Konverben
und deutscher Konjunktionen, mit Blick auf ihre Verwendung durch monolinguale und
. In Türkisch und Deutsch im Vergleich, L. Johanson and J. Rehbein (eds),
bilinguale Kinder�����
189–243.�������������������������
Wiesbaden: Harrassowitz.
Rehbein, J. 1999b. Erzählen in zwei Sprachen – auf Anforderung. Kinderkommunikation – ein-
sprachig und mehrsprachig. Mit einer erstmals auf Deutsch publizierten Arbeit von Lev S.
Vygotskij, Zur Frage nach der Mehrsprachigkeit im kindlichen Alter, K. Meng, K. and J. Reh-
bein (eds), In press. Münster: Waxmann.
Rehbein, J. 2001a. Turkish in European Societies. Lingua e Stile XXXVI-2/2001: 317–334.
Rehbein, J. 2001b. Konzepte der Diskursanalyse. Text- und Gesprächslinguistik. 2 Vols., K. Brink-
er et al (eds), 927–945. Berlin: Walter de Gruyter.
Rehbein, J. and Grießhaber, W. 1992. L2-Erwerb versus L1-Erwerb: Methodologische Aspekte
ihrer Erforschung. Kindliche Sprachentwicklung. Konzepte und Empirie, K. Ehlich (ed.),
67–119. Opladen: Westdeutscher Verlag.
Connectivity by means of finite elements in mono- and bilingual Turkish discourse 

Rehbein, J. and Karakoç, B. 2004. On contact-induced language change of Turkish aspects: Lan-
guaging in bilingual discourse. Languaging and Language Practices, C. B. Dabelsteen and J.
N. Jørgensen (eds), 125–149. Copenhagen: Copenhagen Studies in Bilingualism.
Slobin, D. I. (ed.). 1985. The Crosslinguistic Study of Language Acquisition I: The data. Hillsdale
NJ: Lawrence Erlbaum.
Slobin, D. I. and Aksu, A. A. 1982. Tense, aspect, and modality in the use of the Turkish
evidential. Tense – Aspect: Between semantics and pragmatics, P. J. Hopper, (ed.), 185–200.
Amsterdam: John Benjamins.
section 4

Subordination – coordination
Alternative subordination
strategies in Turkish*

Celia Kerslake
University of Oxford

This article examines the distribution of non-finite and finite strategies within
each of the major categories of subordinate clause (nominal, adjectival/relative
and adverbial) in Turkish. Noting the association of the various non-finite
strategies with less formal registers, it seeks psycholinguistic, pragmatic and
rhetorical explanations for the popularity of these alternative structures
alongside a fully developed system of non-finite subordination.

1. Introduction

In structural terms, Turkish subordinate clauses are of two broad types:


a. Finite (F): the predicate may be verbal or nominal, and is marked in the same way
as if it were the predicate of a main clause.
b. Non-finite (NF): the predicate is verbal, and is marked by distinctive subordinat-
ing morphology.
Although NF is generally regarded as the predominant subordination strategy in
Turkish, F is also widespread in particular types of subordinate clause, and often cor-
relates with particular registers of language use, and/or with the operation of prag-
matic or psycholinguistic factors.1 In this paper I propose to examine the distribution
of contrasting subordination strategies within each of the three major functional class-
es of subordinate clause: nominal, relative and adverbial. I shall consider to what ex-
tent the distribution overlaps or is complementary in semantic terms, and suggest how
particular areas of overlap can be accounted for by pragmatic, sociolinguistic/stylistic
and psycholinguistic factors. This is not claimed to be an exhaustive survey, but I at-
tempt to cover the most salient phenomena within each functional class. In the final
part of the paper I shall attempt to draw some conclusions about Turkish subordina-
tion as a whole.
 Celia Kerslake

1.1 Formal characteristics of F and NF clauses in Turkish

Before proceeding to look at the specifics of particular functional categories of subor-


dinate clause, it will be useful to identify more specifically the distinctive features of
Turkish F and NF clauses as defined above.

1.1.1 F subordinate clauses


F subordinate clauses nearly always contain, either at the beginning or at the end, a
subordinator constituent that is structurally a separate word. The only exception to
this is that certain verbs (expressing perception or desire) can take an F noun clause as
their complement without the intervention of a subordinator:
(1) Mehmet artık başka bir yerde otur-uyor.
Mehmet now other a place-loc live-impf
‘Mehmet now lives somewhere else.’
(2) [Mehmet/Mehmet-i2 artık başka bir yerde oturuyor] san-ıyor-um.
-acc think-impf-1sg
‘I think [Mehmet now lives somewhere else].’
(3) and (4) are examples of F noun clauses containing subordinators, one of which,
diye, occurs obligatorily at the end of its clause and the other, ki, obligatorily at the
beginning of it:
(3) [Bu oda ders için çok uygun diye] düşün-üyor-um.
this room lesson for very suitable sub think-impf-1sg
‘I think [this room is very suitable for teaching].’
(4) Anla-dı-k ki Necla artık biz-den ayrıl-mış-tı.
understand-pf-1pl sub Necla now we-abl separate-pf-p.cop
‘We realized that Necla had now parted company with us.’
The Persian-derived general subordinator ki3 (whose meanings include ‘that’, ‘who’,
‘which’, ‘when’, ‘so that’, etc), and other items that contain or may contain it, such as nasıl
ki ‘just as’ and madem(ki) ‘since’, are the only subordinators in Turkish that stand at the
beginning of their clause.4 The subordinators that appear at the end of their respective
clauses, principally diye ‘that’, ‘so that’, etc and gibi ‘as if ’, are of native Turkish origin, and
their position conforms to the left-branching word order typology of Turkish. Of these,
diye is in origin a converbial form of the verb de- ‘say’,5 and gibi is basically a postposition
meaning ‘like’. I shall divide the category of finite subordinate clauses into the subtypes
F-Bare (i.e. with no subordinator), F-Sub-Final and F-Sub-Initial.
A few remarks should be made at the outset about the peculiar features of F-Sub-Initial
clauses introduced by ki. Unlike other F-Sub-Initial clauses (such as those introduced
by madem(ki)), ki-clauses obligatorily follow the main clause to which they are at-
tached. In an article first published in 1975, Johanson drew attention to the problem-
Alternative subordination strategies in Turkish 

atic status of these structures, which he suggested should probably be viewed as para-
tactic rather than hypotactic. He observed (1991: 212; also 1992: 109) that ki-introduced
clauses cannot themselves form part of an NF subordinate structure embedded in the
typical Turkish manner. This is illustrated in my examples (5) – (6):
(5) a. İzmir-de [anne-m-in pek sev-diğ-i]
İzmir-loc mother-1sg.poss-gen very.much love-part-3sg.poss
bir komşu-muz var-dı.
a neighbour-1pl.poss existent-p.cop
b. İzmir’de bir komşumuz vardı ki annem on-u pek sev-er-di.
s/he-acc love-aor-p.cop
‘In İzmir we had a neighbour that my mother was very fond of.’
(6) a. [İzmir’de [annem-in pek sev-diğ-i] bir komşumuz ol-duğ-un]-u
exist-vn-3sg.poss-acc
anlat-tı-m.
tell-pf-1sg
b. * İzmir’de bir komşumuz olduğunu [ki annem onu pek severdi] anlattım.
‘I told [them] that in İzmir we (had) had a neighbour that my mother was very
fond of.’
(5a) and (5b) contain, in addition to their main clause, respectively the NF and F-Sub-
Initial (ki) versions of a relative clause modifying the noun phrase bir komşumuz ‘a
neighbour of ours’. (6) shows that while the version with the NF relative clause can be
readily converted into an NF complement clause, this is impossible with the version
involving a ki-clause. In view of their somewhat indeterminate status, clauses intro-
duced by ki will not be enclosed in square brackets in the examples from now on.

1.2 NF clause morphology

The most fundamental aspect of the finite/non-finite distinction in Turkish subordi-


nate clauses is morphological, and is articulated at the extreme right-hand end of the
chain of suffixes attaching to the root.6 To facilitate the recognition of non-finite struc-
tures in the examples used in this article the subordinating suffixes that mark a clause
as NF will be shown in bold italics. The examples below contrast the finite verb of a
simple sentence (7) with three types of NF subordinate clause: adjectival/relative (8),
nominal (9) and adverbial (10).
(7) Onlar-a bu kart-ı göster-iyor-sun.
them-dat this card-acc show-impf-2sg
‘You show them this card.’
(8) [onlara göster-diğ-in] kart
show- part-2sg.poss
‘the card (that) you show them’
 Celia Kerslake

(9) [Onlara bu kartı göster-me-n]-i söyle-miş-ti-m


show-vn-2sg.poss-acc tell-pf-p.cop-1sg
‘I told you to show them this card.’
(10) [Onlara bu kartı göster-erek] gir-iyor-sun.
show-cv enter-impf-2sg
‘You get in by showing them this card.’
In Turkish, finite verbs are obligatorily marked for person. Some NF verb forms (like
those illustrated in (8) and (9)) equally include, in addition to the appropriate subor-
dinating suffix, a person marker, while others, like that exemplified in (10), are un-
marked for person. Even where a person marker does occur in an NF form, however,
this is selected not from one of the sets that occur on finite predicates, but from the
quite separate set of nominal suffixes that indicate possession (poss).
At the cost of some over-simplification, I present below a summary of the
morphological marking of Turkish NF subordinate clauses.
(11) Noun clause morphology:7 verbal noun (vn) suffixes
-mA(K)8 ‘infinitive’ (no person marking)
-mA + poss used in evaluative, resultative and volitional
utterances
-(y)Iş + poss in relatively restricted use
-DIK/-(y)AcAK + poss used mainly in indirect statements/questions
(12) Relative clause morphology:9 participle (part) suffixes
-(y)An, –mIş, -(y)AcAk ‘subject participle’ (no person marking)
-DIK/-(y)AcAK + poss ‘object participle’ (possessive marking
indicates the subject)
(13) Adverbial clause morphology: converbial marking
Either a distinctly converbial suffix such as -(y)ArAk, -(y)IncA, etc
(glossed cv) or a composite form based on one of the vn/part forms:
-mAk + (…) e.g. -mAk için ‘in order to’
-mA + poss + (…) e.g. -mA-sI için ‘in order for X to’
-DIK/-(y)AcAK + poss + (…) e.g. -DIğ-I zaman ‘when X’

1.3 Conditional clauses

Conditional clauses occupy a position intermediate between the F and NF types, but
closer to the former. They are marked either with the stressed suffix -sA or the un-
stressed copular clitic -(y)sA. All conditional clauses are marked for person in the same
way as finite predicates, and -sA forms (but not those with -(y)sA) can also occur
as main clauses, expressing wishes.
Alternative subordination strategies in Turkish 

(14) Conditional clause morphology


(a) Verb stem + tense/aspect/modality OR nominal predicate
+ (y)sA + person markers (‘real’ conditions)
(b) Verb stem + -sA + person markers (hypotheticals)
(c) Verb stem + -sA + past copula + person markers (counterfactuals)
The presence of a wh-phrase within any type of conditional clause produces a univer-
sal conditional:
(15) a. Sen bahçe-de ol-acak-sa-n ben de ora-ya gel-e-yim.
you garden-loc be-fut-cond.cop-2sg I too there-dat come-opt-1sg
‘If you’re going to be in the garden, I’ll come there too.’
b. Sen nere-de olacaksan ben de oraya geleyim.
where-loc
‘I’ll come wherever you’re going to be.’

2. Noun clauses

The overall picture within the noun clause category is as follows:


a. NF structures are available across the entire range of semantic sub-types. Where
the subject of the NF clause is different from that of the superordinate clause and
is represented in the NF clause by a noun phrase, this noun phrase receives geni-
tive marking.
b. F-Bare structures are in complementary distribution with F-Sub-Final structures
containing diye, and are available only as complements (or subjects) of a small
number of verbs.
c. F-Bare/F-Sub-Final structures, while widely available, are confined to informal
registers.10
d. The use of F-Sub-Initial structures (introduced by ki) seems to be motivated main-
ly by pragmatic factors.

2.1 Sentence types involving non-person-marked NF noun clauses

The only verbal noun suffix that regularly occurs without person marking is -mAK.

2.1.1 Types of sentence in which non-person-marked NF noun clauses are obligatory


Perhaps predictably in view of the fact that finite verbs have to be marked for person,
the type of sentence from which F noun clauses are most clearly excluded are those
that predicate something of an activity, experience or state viewed in the abstract,
without reference to any subject:
 Celia Kerslake

(16) [Hediye al-mak] her zaman hoş ol-mu-yor.


present get-vn every time nice be-neg-impf
‘It’s not always nice to get presents.’
In contexts where the superordinate clause expresses the attitude or behaviour of the
subject of that higher clause towards his/her own potential or actual involvement in
some activity, experience or state, the noun clause expressing that activity, experience
or state again is usually obligatorily NF:
(17) Meryem [yalnız kal-ma]-yı sev-mi-yor.
Meryem alone be.left-vn-acc like-neg-impf
Meryem doesn’t like being left alone.’
(18) Ali [bu araba-yı kullan-ma]-ya başla-dı.
Ali this car-acc use-vn-dat begin-pf
‘Ali has begun to use this car.’

2.1.2 One type in which non-person-marked NF alternates with F-Bare


The only context of the kind treated in 2.1 in which the use of an F noun clause is pos-
sible is one expressing desire, with the verb in the superordinate clause being almost
invariably iste- ‘want’. If an F clause is used it is of the F-Bare type, its verb is marked
for optative (or 3rd person imperative) modality, and its person marking reflects the
standpoint of the speaker, not the subject (unless the two are the same). As in all simi-
lar pairs of examples to follow, (a) and (b) below express the same proposition, using
respectively an NF clause and an F clause:
( 19) a. Çocuk [anne-sin-e yardım et-mek] iste-miş-ti.
child mother-3sg.poss.dat help aux-vn want-pf-p.cop
b. Çocuk [anne-sine yardım et-sin] istemişti.
aux-imper.3sg
‘The child had wanted to help his/her mother.’

2.2 Sentence types involving person-marked NF noun clauses

The verbal noun suffixes that can or must be followed by person markers are -mA,
-(y)Iş, -DIK and -(y)AcAK.

2.2.1 Types of sentence in which person-marked NF noun clauses are obligatory


On the basis of a combination of syntactic and semantic criteria several categories of
sentence can be identified for which, although the NF strategy involves person mark-
ing for subject, there is no alternative F strategy available for the noun clause. Among
the most important are those where the noun clause is:
Alternative subordination strategies in Turkish 

a. the subject of an evaluative or descriptive sentence:


(20) [Bura-ya kadar gel-me-miz] zor ol-uyor.
here-dat as.far.as come-vn-1pl.poss difficult be-impf
‘It’s difficult for us to come all this way.’
b. the subject of a sentence expressing the impact of a situation on some object:
(21) [Kemal-in bir şey söyle-me-yiş-i] ben-i şaşırt-tı.
Kemal-gen a thing say-neg-vn-3sg.poss I-acc surprise-pf
‘It surprised me that Kemal didn’t say anything.’
c. the complement of a copulative sentence:
(22) Bu-nun neden-i [sen-i kıskan-ma-sı-dır].
this-gen reason-3sg.poss you-acc be.jealous-vn-3sg.poss-mod
‘The reason for this must be that s/he is jealous of you.’
d. the subject or object of an act of causation, facilitation or obstruction:
(23) Bu yöntem-le [hasta-nın daha az acı çek-me-si]
this method-ins patient-gen more little pain suffer-vn-3sg.poss
sağla-n-abil-ir.
ensure-pass-possib-aor
‘By this method it can be ensured that the patient suffers less.’
e. the oblique object of an act of helping or permitting:
(24) [Çocuk-lar-ın bahçe-de oyna-ma-ların]-a izin ver-ir
child-pl-gen garden-loc play-vn-3pl.poss-dat permission give- aor
mi-siniz?
q-2pl
‘Would you [please] let the children play in the garden?’
f. an event or state to which a cause, purpose, benefit, etc is attributed:
(25) [Öyle yap-ma-ların]-ın neden-i belli.
so do-vn-3pl.poss-gen reason-3sg.poss obvious
‘The reason for their doing this is obvious.’
The fact that all of the above patterns involve -mA (or -(y)Iş11 rather than
-DIK/-(y)AcAK is not fortuitous. The pair of suffixes -DIK/-(y)AcAK (differentiated by
tense) mark a clause as expressing a proposition rather than an event/state. They are
overwhelmingly associated with the concepts of factivity, cognition and communica-
tion, and their finite counterparts, as we shall see below, therefore attract the F-Sub-
Final subordinator diye, derived from the verb de- ‘say’.
 Celia Kerslake

2.2.2 Alternation of F-Bare and F-Sub-Final clauses with person-marked NF clauses


2.2.2.1 Wishes. The same F-Bare structure seen in (19b) above can be used to
express a wish for something to be done or undergone by a subject other than the
subject of the act of wishing. In this case the NF form receives person marking in
the form of a possessive suffix:
(26) a. [Çocuğ-un anne-sin-e yardım et-me-sin]-i
child-gen mother-3sg.poss.dat help aux-vn-3sg.poss-acc
iste-miş-ti-m.
want-pf-p.cop-1sg
b. [Çocuk annesine yardım et-sin] istemiştim.
aux-imper.3sg
‘I had wanted the child to help his/her mother.’

2.2.2.2 Directives. Where the superordinate verb expresses not the mere desire for
something to happen but an articulated instruction, request, recommendation, etc
for it to be brought about, there is again a choice available between an NF and an F
strategy for the noun clause. In this case, however, the finite construction requires
the inclusion of the subordinator diye (literally ‘saying’), and the subordinate clause
is treated as if it were direct speech, its person deixis being centred on the person
uttering the directive, not on the speaker of the sentence as a whole. Thus although
the verbal noun in (27a) is marked for 1st person. its finite counterpart in (27b) is
2nd person imperative.
(27) a. Tülay [doktora git-me-m]-i tavsiye et-ti.
Tülay doctor-dat go-vn-1sg.poss-acc recommendation aux-pf
b. Tülay bana [doktor-a git diye] tavsiye etti.
I.dat doctor-dat go[2sg.imper] sub
‘Tülay advised me to go to the doctor.’

2.2.2.3 Cognition, communication and factual status. Noun clauses in sentences


that deal with knowledge or communication of, make a claim about or question
the factual status of a proposition present a slightly complex picture in terms of
the distribution of F and NF strategies. In the majority of cases the most generally
acceptable strategy is an NF noun clause marked with the subordinator -DIK
(relative present or past tense) or -(y)AcAK (relative future tense):
(28) [Siz-in-le konuş-acağ-ımız]-ı onlar-a söyle-me-di-k.
you-gen-ins talk-vn-1pl.poss-acc they-dat tell-neg-pf-1pl
‘We didn’t tell them (that) we were going to talk to you.’
Alternative subordination strategies in Turkish 

(29) [Adam-ın endişeli ol-duğ-u] belli-ydi.


man-gen anxious be-vn-3sg.poss obvious-p.cop
‘It was obvious that the man was anxious.’
The distribution of alternative (informal) F-Bare/F-Sub-Final strategies in this catego-
ry is as follows:
a. F-Bare clauses
The direct juxtaposition of a bare finite clause occurs regularly with the verb san-
‘think’ (in the sense of ‘be under the impression that’), as seen in (30b):
(30) a. [Herkes-in git-tiğ-in]-i san-ıyor-du-m.
everyone-gen go-vn-3sg.poss-acc think-impf-p.cop-1sg
b. [Herkes git-ti] sanıyordum.
go-pf
‘I thought everyone had gone.’
Where the content of the noun clause is interrogative, the F-Bare strategy may also
occur if the superordinate clause expresses cognition:
(31) a. [Ayşe-nin ne kadar iste-dig-in]-i bil-mi-yor-uz.
Ayşe-GEN how.much want-VN-3SG.POSS-ACC know-NEG-IMPF-1PL
b. [Ayşe ne kadar isti-yor] bilmiyoruz.
want-IMPF
‘We don’t know how much Ayşe wants.’
However, an F clause is not acceptable where the superordinate verb is sor- ‘ask’:
(32) a. [Ayşe-nin ne kadar iste-dig-in]-i sor-du.
b. *[Ayşe ne kadar isti-yor] sordu.
‘S/he asked how much Ayşe wants.’
b. F-Sub-Final clauses with diye
Outside the contexts illustrated in (30) – (31), noun clauses in sentences expressing
cognition or communication can usually be articulated in F form by the interposition
of diye between the subordinate and superordinate predicates:
(33) a. [Sınav-ın ertele-n-diğ-in]-i duy-muş-tu-k.
exam-gen postpone-pass-vn-poss.3sg-acc hear-pf-p.cop-1pl
b. [Sınav ertele-n-di diye] duymuştuk.
-pf sub
‘We had heard that the exam was being postponed.’
c. The role of factivity
With factive predicates both the F-Bare and the F-Sub-Final strategy become unacceptable:
 Celia Kerslake

(34) a. [Adam-ın endişeli ol-duğ-u] belli-ydi. (See (29))


b. *[Adam endişeli] (diye) belli-ydi.
‘It was obvious that the man was anxious.’
(35) a. [Mehmet-in sen-i sev-diğ-in]-i anla-mı-yor mu-sun?
Mehmet-gen you-acc love-vn-poss.3sg-acc understand-neg-impf q-2sg
b. *[Mehmet seni seviyor (diye)] anlamıyor musun?
‘Don’t you realize that Mehmet loves you?’
However, in the case of the matrix verbs anla- ‘understand’, ‘realize’ and bil- ‘know’, these
factive readings give way to the non-factive ‘think’, ‘be under the impression that’ (i.e.
become synonymous with san-) when stress is placed on the complement rather than on
the matrix verb. In such cases the matrix verb is almost always in a past tense form:
Factive
(36) [Mustafa-nın evli ol-duğ-un]-u biL-İyor-du-m.12
Mustafa-gen married be-vn-3sg.poss-acc know-impf-p.cop-1sg
‘I knew Mustafa was married.’
Non-factive
(37) a. [Mustafa-nın evLİ ol-duğ-un]-u biliyordum.
b. [Mustafa evLİ diye] bil-iyor-du-m.
sub
‘I thought Mustafa was married.’
(38) a. [Sen-in Mehmet-i sev-diğ-in]-İ anla-mış-tı-m.
you-gen Mehmet-acc love-vn-2sg.poss-acc think-pf-p.cop.1sg
b. [Sen Mehmet’i sev-İyor-sun diye] anlamıştım.
-impf-2sg sub
‘I thought you loved Mehmet.’
(35) – (38) show that the factive readings of these two verbs are obtained only by a
combination of NF subordination with stress on the verb itself. Where, as in (37) and
(38), stress is placed on the complement clause, this produces non-factive meaning
with both NF and F-Sub-Final clauses.

2.2.2.4 Emotion. For the expression of complements of verbs of emotion, F-


Sub-Final clauses with diye are freely available as informal alternatives to NF
constructions:13
(39) a. [Sen-in gel-e-me-diğ-in]-e üzül-dü-k.
you-gen come-possib-neg-vn-2sg.poss-dat be.sad-pf-1pl
b. [Sen gel-e-me-di-n diye] üzül-dü-k.
come-possib-neg-pf-2sg sub
‘We were sorry you couldn’t come.’
Alternative subordination strategies in Turkish 

2.3 Finite (F-Sub-Initial) noun clauses formed with ki

Unlike other noun clauses in Turkish (F and NF), which normally precede the main
predicate (but may be placed after it if backgrounded, i.e. context-retrievable), F-Sub-
Initial clauses introduced by the enclitic ki obligatorily follow it. The fact that ki in
noun clauses is enclitic, generating stress on the preceding constituent and often fol-
lowed by a brief pause,14 reinforces the highlighting of the main predicate that is
achieved by this linear re-ordering. By placing the main predicate at the beginning of
the sentence the speaker draws attention to the status of what s/he is about to present:
a desire, a conjecture, a deduction, etc. In (40) – (41) the unmarked sentence employ-
ing the NF strategy is given first, as (a).
(40) a. [Bütün çocuk-lar-ımız-ın bilgisayar kullan-abil-me-lerin]-i
all child-pl-1pl.poss-gen computer use-possib-vn-3pl.poss-acc
isti-yor-uz.
want-impf-1pl
b. İsti-yor-UZ ki, bütün çocuk-lar-ımız bilgisayar kullan-abil-sin-ler.
use-possib-opt-3pl
‘We want all our children to be able to use computers.’
(41) a. [Çevre kirliliğ-i-nin önemli ölçü-de
environment pollution-nc-gen important measure-loc
azal-dığ-ı] iddia ed-il-iyor.
decrease-vn-3sg.poss claim aux-pass-impf
b. İddia edil-iYOR ki, çevre kirliliği önemli ölçüde azal-mış.
decrease-ev/pf
‘It is claimed that environmental pollution has substantially decreased.’
In terms of their semantic range, F-Sub-Initial noun clauses introduced by ki can be
used in the categories of sentence discussed in 2.2.1 and 2.2.3, namely those expressing
wishes, cognition, communication and factual status. In the case of wishes, the verb in
the ki clause, like its counterparts in F-Bare and F-Sub-Final clauses (see 2.1.2, 2.2.1) is
given optative or 3rd person imperative marking, as seen in (40b). The main clause to
which a ki-introduced noun clause stands as subject or complement is almost invari-
ably declarative and affirmative.
In contrast to the situation with the F-Bare and F-Sub-Final structures examined
above, the use of F-Sub-Initial noun clauses is not register-bound. The occasional ap-
pearance of a nominal ki-clause in formal writing is not regarded as stylistically inap-
propriate. But it is in spoken discourse that its pragmatic possibilities can be exploited
to greatest effect. Apart from the rhetorical effect of using an inversion of the canonical
sequence of complement-main clause to highlight the speaker’s attitude or stance, this
strategy can also win him/her time for organizing the substantive content of his/her
 Celia Kerslake

communication. Furthermore, it can be a highly effective device for getting oneself


listened to in a competitive speech arena:
(42) Ben di-yor-um [ki, bu öneri-ler-in hiçbiri-si-nin fayda-sı
I say-impf-1sg sub these proposal-pl-gen none-3sg.poss-gen use-3sg.poss
ol-maz].
be-neg.aor
‘What I say is, none of these suggestions will do any good.’
Because of their decidedly marked status, nominal ki-clauses do not usually occur re-
peatedly within a single utterance or paragraph.

3. Relative clauses

The standard relativization strategy in Turkish is a left-branching clause whose verb is


marked with one of two types of subordinating suffix (shown in (12) above). The func-
tions of these two sets of suffix are illustrated in (43) – (45):
(43) [ben-i yemeğ-e çağır-an] arkadaş
I-acc meal-dat invite-part friend
‘the friend who has invited me to lunch/dinner’
(44) [yemeğ-e çağır-dığ-ım] arkadaş
meal-dat invite-part-1sg.poss friend
‘the friend (that) I’ve invited to lunch/dinner’
(45) [arkadaş-ım-ı çağır-dığ-ım] yemek
friend-1sg.poss-acc invite-part-1sg.poss meal
‘the meal to which I’ve invited my friend’
The -(y)An type of suffix, seen in (43), is used predominantly where the relativized
constituent is the subject, or the possessor of the subject, of the relative clause. The
-DIK type seen in (44) and (45), on the other hand, is used where the relativized con-
stituent has a non-subject function (object or adverbial) in the relative clause. The
-DIK type of suffix is always followed by a person marker (possessive suffix) referring
to the subject of the relative clause (which may additionally be represented by a geni-
tive-marked noun phrase at the beginning of the clause).
NF relative clauses can be non-restrictive as well as restrictive.15 The non-restric-
tive type, while rare in speech, is a regular feature of the written language. The follow-
ing example is taken from a bio-bibliographical dictionary of writers:
Alternative subordination strategies in Turkish 

(46) [Küçük yaşta şiire ilgi du-yan] İbrahim Alâettin [aruzla yaz-dığ-ı] ilk şiirlerini
çeşitli dergilerde yayımladı.
‘İbrahim Alâettin, who became interested in poetry as a young child, published
his first poems, which he wrote in ‘aruz’ metre,16 in various journals.
As far as relative clauses are concerned,17 the only F structure that appears to have
some sort of functional resemblance to the standard NF strategy is an F-Sub-Initial
clause with ki. Apart from the fact that its verb is finite, this type of ‘relative’ clause is
post-nominal, in contrast to the canonical pre-nominal position of the participial
clause. A recent study by Schroeder (2002) has shed important new light on the con-
trastive distribution of these two strategies. There appear to be two major factors de-
termining the distribution of ki-introduced ‘relative’ clauses:
a. They are essentially a feature of unplanned spoken discourse (or of informal writ-
ing styles that imitate this).
b. Semantically they are overwhelmingly non-restrictive.18 For the vast majority of
restrictive relative clauses there simply is no F strategy available.
The use of the quasi-relative ki-structures is convincingly explained by Schroeder
(2002: 81) in terms of “the tendency of spoken Turkish towards a stronger paratactic
organization of the text”. This he sees as part of a cross-linguistic tendency of un-
planned/spoken discourse “to organize information in smaller, syntactically autono-
mous units”, which has various advantages for interactive communication and also
“allows the speaker to plan his/her talk ‘as s/he goes along’”. These imperatives of the
speech situation, Schroeder notes, cannot always be accommodated by left-branching
syntactic structure. We shall return to this crucial issue in section 5.
There are two kinds of antecedent for ki-introduced ‘relative’ clauses: (i) a noun
phrase, or (ii) the entire preceding clause.19 Only the first type is even potentially capa-
ble of reformulation with a pre-nominal NF relative clause, and the pragmatic impact
of the utterance may be reduced by such a substitution. In (47) there seems very little
to choose between the two strategies:
(47) a. Ev-de bir sorun çık-tı, ki (on-u) hemen hallet-me-m
home-loc a problem crop.up-pf ki (it-acc) at.once solve-vn-1sg.poss
gerek-iyor.
be.necessary-impf
b. Evde [hemen halletmem gerek-en] bir sorun çıktı.
-part
‘A problem has cropped up at home that I’ve got to attend to at once.’
On the other hand, applying the same transformation to one of Schroeder’s examples,
reproduced as (48a), produces a far less felicitous result:
(48) a. Güzel, doğru cevap-lar ver-di-ler… ki cevap-lar-ı BEN söyle-miş-ti-m
good correct answer-pl give-pf-3pl ki answer-pl-acc I tell-pf-p.cop-1sg
 Celia Kerslake

onlar-a.
they-dat
‘They gave good, correct answers… It was I who had told them the answers.’
(Schroeder 2002: 78)
b. ?[Benim onlar-a söyle-miş ol-duğ-um] bazı güzel, doğru
I-gen they-dat tell-pf aux-part-1sg.poss some good correct
cevap-lar ver-di-ler.
answer-pl give-pf-3pl
‘They gave (some) good, correct answers, which I had given them.’
The syntactically autonomous structure of the ki-clause in (48a) makes it possible for
ben ‘I’ to receive maximal focal stress by being placed in the immediately pre-verbal
position. Not only is this possibility lost in the substitution of a pre-nominal NF rela-
tive clause in (48b), but so is also the dramatic effect of saving the information about
the source of the answers until after the bland statement about the giving of correct
answers has been made.

3.1 Paratactic features of ki-introduced relative clauses

One notable characteristic that distinguishes the ‘relative clause’ function of ki from its
function of introducing complement or adverbial clauses is the supra-segmental pho-
nology involved. Whereas in the nominal and adverbial functions ki is enclitic and
pre-stressing, being usually followed by a slight pause in speech and a comma in writ-
ing, exactly the opposite is true in the type of clause we are considering here, where ki
has no pre-stressing effect and is usually preceded by a pause.20 Developing the argu-
ment of Johanson mentioned in 1.1.1, Schroeder (2002: 75) states clearly that ‘relative’
ki-clauses are not actually subordinate clauses at all, since they are not syntactically
embedded in the sentence, but rather paratactically conjoined to it. Here we will con-
sider briefly three features of this type of structure that support Schroeder’s claim.

3.1.1 Anaphoric reference to the antecedent


As seen in (5b) and (47a) above, a relative clause introduced by ki may include a re-
sumptive pronoun referring to the antecedent. ‘Pronoun retention’ is one of the rela-
tivization strategies recognized in the typological literature, and is found, for example,
in Arabic and Persian (Song 2001: 218). As far as ki-introduced Turkish relative claus-
es are concerned, the principles governing the use of resumptive pronouns would ap-
pear to be roughly as follows:21
a. Where the antecedent is the subject of the relative clause, use of a resumptive pro-
noun occurs almost exclusively with [+ human] referents, and is optional. (Note
its non-use in (51) below.)
Alternative subordination strategies in Turkish 

b. Where the antecedent is the direct object of the relative clause, use of a resumptive
pronoun is strongly favoured with [+ human] referents (5b), and is optional with
[– human] referents (47a).
c. Where the function of the antecedent within the relative clause is that of an ob-
lique object or adverbial, the use of a resumptive pronoun22 is obligatory:
(49) Bazı öğretmen-ler – ki onlar-dan özellikle nefret ed-er-di-m-
some teacher-pl ki they-abl particularly hate aux-aor-p.cop.1sg
başarısız öğrenci-ler-i hep azarla-r-dı
unsuccessful student-pl-acc always scold-aor-p.cop
‘Some teachers – whom I particularly hated – always scolded the unsuccessful
students.’
The presence in some Turkish post-nominal ‘relative’ clauses of an item that refers
anaphorically to the antecedent noun phrase underlines the syntactic autonomy of
these structures. However, pronoun retention is not the only type of anaphoric refer-
ence encountered in ‘relative’ ki-clauses. Several of Schroeder’s recorded-speech exam-
ples, such as (48a) above, involve actual repetition of the head noun of the antecedent
noun phrase. This noun may be preceded, as in (50), by a demonstrative determiner:
(50) Bir çığlık duydum… ki bu çığlığ-ı çok çok iyi bil-iyor-um…
a scream hear-pf-1sg ki this scream-acc very very well know-impf-1sg
‘I heard a scream… and I know this scream very very well…’
(Schroeder 2002: 75)

3.1.2 Post-predicate position


In most of the examples of ki-introduced ‘relative’ clauses given so far, the ki-clause has
been positioned not immediately after its antecedent, but after the main predicate. It
appears that the examples occurring in Schroeder’s corpus are exclusively of this kind,
since his definition of the noun phrase type of antecedent locates this in “the preceding
clause” (Schroeder 2002: 78). Similarly, Erguvanlı (1980–81: 128) explicitly states that
a relative clause introduced by ki never precedes the predicate. The possibility of de-
tachment from its antecedent is another factor contributing to the autonomy of the
ki-clause, reflected in the fact that some of the translations provided above contain not
an English relative clause but rather a second juxtaposed or conjoined main clause.
The overwhelming predominance of post-predicate positioning for relative ki-
clauses with a noun phrase antecedent arises, I suggest, from the fact the vast majority
of these noun phrase antecedents are indefinite. In verbal sentences in Turkish indefi-
nite items, unless referring to components of a previously mentioned or implied set,
almost always have to be placed in the immediately pre-verbal position. Similarly, in
existential sentences the subject always stands immediately before the predicative con-
stituent var ‘existent’ or yok ‘non-existent’.
 Celia Kerslake

In cases where the noun phrase antecedent is definite, on the other hand, the ki-
clause may be positioned either directly after the antecedent, or after the predicate:
(51) a. Anne-m bile, ki acı sev-mez, bun-u zevk-le yi-yor.
mother-1sg.poss even ki ‘hot’ like-neg.aor this-acc pleasure-com eat-impf
‘Even my mother, who doesn’t like ‘hot’ [food], enjoys eating this.’
b. Annem bile bunu zevkle yiyor, ki genel-de acı sevmez.
general-loc
‘Even my mother enjoys eating this – and she doesn’t generally like ‘hot’
[food].’
As mentioned above, there is a second type of relative ki-clause – claimed by Schroed-
er (2002: 78) to be the more common – which instead of a noun phrase has the entire
preceding clause as its antecedent:
(52) Tanıdık çevre-den bir tip ol-ma-sı lazım… ki o da çok
known circle-abl a person be-vn-3sg.poss necessary ki that top very
zor yani.
difficult I.mean
‘It should be someone from [your] circle of friends… which is really difficult,
of course.’23
(Schroeder 2002: 78)
This type of F-sub-initial clause, which is functionally equivalent to the English ‘sen-
tential relative clause’,24 has no NF counterpart in Turkish, because there is no noun
phrase constituent to which an NF relative clause could appropriately be preposed. The
assertion on which the appended clause provides a comment has to be fully articulated
before the comment is presented to the hearer. As seen in (52), the demonstrative with
anaphoric reference occurs in this type of ki-clause also, although it does not, as sug-
gested by Kissling (1960: 145) have to be bu ‘this’. In addition to o ‘that’, formations
with the determiner öyle ‘such’ are also possible:
(53) Ali bilgisayar-ın-ı çal-dır-mış, ki öyle aksilik-ler
Ali computer-3sg.poss-acc steal-caus-ev/pf sub such misfortune-pl
sık sık gel-ir on-un baş-ın-a.
often come-aor s/he-gen head-3sg.poss-dat
‘It seems Ali has had his computer stolen; such misfortunes happen to him all
the time.’

3.1.3 Inter-clausal parenthetical comments


Where the ki-clause has the entire preceding clause as its antecedent, this antecedent
need not be a main clause. A common function of relative ki-clauses is to provide a
parenthetical25 comment on the content of an adverbial clause, before the speaker goes
Alternative subordination strategies in Turkish 

on to present the main clause that the adverbial clause modifies. The ki-clause is struc-
turally quite extraneous to the syntactic structure of the sentence:
(54) [Bu ev sat-ıl-ır-sa,] ki ev sahib-imiz-in niyet-i
this house sell-pass-aor-cond ki landlord-1pl.poss-gen intention-3sg.poss
öyle, taşın-ma-mız gerek-ecek.
so move-vn-1pl.poss be.necessary-fut
‘If this house is sold – and that is our landlord’s intention – we shall have to
move.’
(55) [Defter-im-de yazılı ol-ma-dığın-a göre,] ki ben
notebook-1sg.poss-loc written be-neg-vn-3sg.poss-dat seeing.that ki I
bu konu-da oldukça dikkatli-yim, bugün toplantı yok-tur.
this matter-loc quite careful-1sg today meeting non-existent-mod
‘Seeing that it’s not written in my diary – and I’m quite careful about these
things – there’s presumably no meeting today.’

4. Adverbial clauses

In principle all the semantic types of adverbial clause can be expressed in NF form in
Turkish, using either specifically converbial morphology or a combination of other NF
morphology with lexical items such as postpositions (e.g. için ‘for’, göre ‘according to’) or
nouns (e.g. zaman ‘time’, hal ‘state’), with or without case marking (see (13) above). How-
ever, within most semantic categories there are F alternatives available, involving subor-
dinators such as clause-final diye and gibi, and clause-initial ki, madem(ki) and nasıl ki.
As noted in 1.3, conditional clauses, which are a type of adverbial clause, have their own
quasi-finite morphology. The conditional copula -(y)sA is also used (with the addition of
other items such as the enclitic particle DA or the interrogative nasıl ‘how’) as an alterna-
tive means for expressing adverbial clauses of concession or similarity.
For reasons of space, the coverage of adverbial clauses here will necessarily be
quite selective.

4.1 Adverbial clauses that can only be expressed by NF means

The following are examples of the very few adverbial clause types for which no F alter-
native is available:
a. Manner (in sense of accompanying action or state)
(56) Bun-u [üzül-erek] söylü-yor-um.
this-acc be.sad-cv say-impf-1sg
‘I say this with regret.’
 Celia Kerslake

b. Substitution
(57) [Şikâyet ed-eceğ-iniz-e] konu-yu ben-im-le
complaint aux-vn-2pl.poss-dat matter-acc I-gen-com
konuş-say-dı-nız.
talk-cond-p.cop-2pl
‘If only you had talked to me about the matter instead of making a complaint.’
Some other types of NF adverbial clauses are paraphrasable by universal conditionals
(see 1.3), but not by fully finite clauses:
c. Manner (in sense of conformity of one action to another)
(58) a. [İste-diğ-in gibi] yap.
want-part-2sg.poss like do
‘Do as you want.’
b. [Nasıl iste-r-se-n] (öyle) yap.
how want-aor-cond.cop-2sg (so) do
‘Do as you want.’ (lit. ‘However you want, do (so).’)
It should be added that within the major category of time adverbials the role of F clauses
is very marginal. The repertoire of relevant NF resources allows for the articulation of a
wide range of temporal relationships obtaining between the main clause situation and
that expressed by the embedded clause (including ‘when’ (successive), ‘when’ (simulta-
neous), ‘while’, ‘before’, ‘after’, ‘until’, ‘as soon as’, ‘as long as’). The only finite structures
available, involving Sub-Initial ki and ne zaman ki26 respectively, are marked construc-
tions that speakers select for the achievement of certain pragmatic goals.

4.2 F-Sub-Final clauses with diye

As in the case of noun clauses, the use of an F strategy based on the subordinator diye
is a feature of informal registers. The adverbial usage that reflects most directly the
original sense of diye (‘saying’ or, by extension, ‘thinking’) is in clauses that express
“how the subject of the main clause understands a situation that is relevant to the per-
formance of the action in the main clause” (Göksel and Kerslake 2005: 463). This is
often an understanding that the speaker knows to be mistaken; in other words diye is
non-factive:
(59) Biz [yemeğ-e çağr-ıl-dı-k diye] git-miş-ti-k.
we meal-dat invite-pass-pf-1pl sub go-pf-p.cop 1pl
‘We had gone thinking we had been invited to a meal.’
Alternative subordination strategies in Turkish 

This type of adverbial diye-clause does not have a direct NF equivalent. To express the
sense of ‘under the impression that’ by non-finite means would involve the embedding
of a noun clause within an adverbial clause:
(60) Biz [[yemeğ-e çağr-ıl-dığ-ımız]-ı düşün-erek] gitmiştik.
vn-1pl.poss-acc think-cv
‘We had gone thinking we had been invited to a meal.’
In the other main adverbial usages of diye as a subordinator the notion of ‘thinking’
has been transformed into one of motivation: cause, as in (61), or purpose, as in (62).
These two senses are distinguished morphologically by the use of optative marking on
the verb when purpose is intended. As in section 2 above, in the remaining examples
of this section wherever alternative NF and F strategies are available the NF strategy is
presented first:
(61) a. Osman, [Zeliha beğen-me-diğ-i için] sakal-ın-ı
Osman Zeliha like-neg-vn-3sg.poss for beard-3sg.poss-acc
tıraş et-tir-miş.
shave aux-caus-ev/pf
b. Osman, [Zeliha beğen-me-di diye] sakalını tıraş ettirmiş.
-pf sub
‘Osman has had his beard shaved off because Zeliha didn’t like [it].’
(62) a. Osman, [Zeliha-nın artık kendisiyle alay et-me-me-si
-gen any.more s/he-ins mockery aux-neg-vn-3sg.poss
için] sakalını tıraş ettirmiş.
b. Osman, [artık Zeliha kendisiyle alay et-me-sin diye] sakalını tıraş ettirmiş.
-3sg.opt sub
‘Osman has had his beard shaved off so that Zeliha won’t make fun of him any
more.’
In the case of the causal example (61) there is a subtle difference in meaning between the
NF and F strategies, and this again involves factivity (cf. 2.2.2.3). Whereas the NF strat-
egy used in (61a) presupposes the truth of the asserted cause (Zeliha’s not liking Osman’s
beard), the use of diye in (61b) places a distance between the speaker and the subject of
the main clause, presenting the proposition that Zeliha didn’t like Osman’s beard as a
perception on the part of Osman, to the truth of which the speaker is not committed.

4.3 F-Sub-Final clauses with gibi27

The postposition gibi ‘like’ may occur at the end of a clause marked with any of the
following predicate markers: aorist -(A/I)r, evidential/perfective -mIş, and evidential
copula -(y)mIş. The resulting structure, corresponding to a clause introduced by ‘as if ’
in English, expresses manner by evoking similarity with another (imagined) event, or
 Celia Kerslake

suggesting an underlying motivation or emotion. These clauses have a status some-


what indeterminate between F and NF, since person marking of their predicates is
absent where -(A/I)r is used and optional in the case of -mIş/-(y)mIş.28 Semantically
these structures are replaceable by the corresponding NF forms -(A/I)rcAsInA,
-mIşçAsınA and -(y)mIşçAsınA. However, in today’s Turkish these converbial struc-
tures are in much less frequent use than gibi-clauses,29 which for their part are not
subject to any register restrictions.
(63) a. [Sonuç-lar-ı şimdiden bil-iyor-muşçaşına] konuş-uyor-sun.
result-pl-acc already know-impf-cv talk-impf-2sg
b. [Sonuçları şimdiden bil-iyor-muş(-sun) gibi] konuşuyorsun.
know-impf-ev.cop(-2sg) sub
‘You talk as if you know the results already.’
Both of the above clause types may optionally be introduced by the subordinating con-
junction sanki ‘as if ’, whose function is to “provide early warning to the hearer of the
non-factual status of the content of the clause” (Göksel and Kerslake 2005: 466):
(64) a. [( Sanki) her şey bit-mişçesine] gevşe-di-ler
(as.if) every thing finish-cv slacken-pf-3pl
b. [(Sanki) her şey bit-miş gibi] gevşe-di-ler
finish-ev/pf sub
‘They slackened off as if everything was over.’

4.4 F-Sub-Initial clauses with ki

The phonological properties of ki when it introduces an adverbial clause are the same
as when it introduces a noun clause (2.3): it generates stress on the preceding constitu-
ent, and is usually followed by a slight pause. As in the case of noun clauses, this qual-
ity gives sentences with ki a marked status. In an unmarked Turkish sentence primary
stress falls on the constituent before the main verb, and anything placed after this
predicate is context-recoverable and unstressed. In sentences with an adverbial ki-
clause, on the other hand, the ki not only throws a strong stress on to the main verb
itself, but also announces that a second statement, of equal weight to the first, is to fol-
low. The sense of expectation is heightened not only by the mid-sentence pause, but
also by the fact that ki itself, while unstressed, has rising intonation. This will be illus-
trated with reference to two of the main types of adverbial ki-clause, both of which are
popular in unplanned discourse, but avoided in formal registers.
a. Dramatic presentation of an event
In this type of sentence the main clause presents a situation that provides the temporal
context for the sudden occurrence of an event to which the speaker gives dramatic
prominence by means of the ki-clause. The verb in the main clause typically has imper-
fective (-(I)yordu) or past perfect (-mIştI) marking, which contrasts with the perfective
Alternative subordination strategies in Turkish 

marking (-DI) of the event verb in the ki-clause. The anticipation of a surprising or
perverse turn of events is often heightened by adding an adverbial such as tam or yeni
(both meaning ‘just’) to the main clause:
(65) Tam sen-i konuş-uyor-DU-K ki, birden oda-YA gir-di-n.
just you-acc talk-impf-p.cop-1pl ki suddenly room-dat enter-pf-2sg
‘We were just talking about you, when you suddenly came into the room.’
(66) Okul-a yeni başla-mış-TI-M ki, bir grip salgın-ı ol-du.
school-dat just start-pf-p.cop-1sg ki a flu epidemic-nc happen-pf
‘I had just started school when a flu epidemic broke out.’
As will be noted from the translations, this type of ki-clause bears a remarkable struc-
tural and functional resemblance to a particular kind of English when-clause that oc-
curs in sentence-final position. Almost all the observations made about the English
sentence pattern in question by Quirk et al. (1985: 1084) apply equally to the Turkish
structure with ki that we are considering here. These authors note that (i) the more
important information is given in the subordinate clause; (ii) the when-clause is non-
restrictive, i.e. makes a separate assertion; (iii) it is usually separated by intonation and
punctuation from the main clause; and (iv) the overall effect is to produce a “dramatic
and emphatic climax” in narrative.
In Turkish this kind of sentence can be paraphrased by turning the original main
clause into an NF subordinate clause marked with the temporal converb marker
-(y)ken. Although the two situations are still presented in the same order (temporal
context followed by highlighted event), the syntactic subordination and prosodic un-
marking of the first clause significantly reduces the ‘climactic’ impact produced.
(67) [Tam (biz) sen-i konuş-uyor-ken] birden oda-YA gir-di-n.
just (we) you-acc talk-impf-cv suddenly room-dat enter-pf-2sg
‘Just as we were talking about you, you suddenly came into the room.’
(68) [Ben okul-a yeni başla-mış-ken] bir grip salgın-ı ol-du.
I school-dat just start-pf-cv a flu epidemic-nc happen-pf
‘When I had just started school, a flu epidemic broke out.’
b. Result
This type of sentence also has a close English parallel in the pattern ‘so/such… that …’.
The main clause in Turkish includes one of the following deictic items, which modifies
some constituent of that clause: o kadar ‘so (much)’, ‘such’, or öyle/öylesine ‘so’, ‘such’, ‘in
such a way’. The function of the ki-clause is to supply cataphoric reference for this
deictic modifier, by presenting a situation resulting from the (extreme) degree that it
expresses.
(69) Oda o kadar karanlık-TI ki hiçbir şey gör-E-mi-yor-du-k.
room so dark-p.cop ki no thing see-possib-neg-impf-p.cop-1pl
‘The room was so dark that we couldn’t see a thing.’
 Celia Kerslake

It is possible to paraphrase such sentences using an NF adverbial clause of degree:


(70) Oda, [[herhangi bir şey gör-me-miz]-i engelle-yecek kadar]
room any a thing see-vn-1.pl.poss-acc prevent-cv
karanlık-tı.
dark-p.cop
‘The room was dark enough to prevent our seeing anything.’
Even more than in the type of sentence discussed in 4.4(a), the result of such a conver-
sion is to neutralize the rhetorical impact produced. The alteration in the shape of the
sentence is more radical, with cause and effect no longer presented in their real-world
sequence but as a pre-packaged whole. In view of the undoubtedly much greater pro-
duction difficulty involved, (70) would come across as almost ludicrously pedantic in
a conversational context.

4.5 Other F-Sub-Initial clauses

Clauses introduced by madem(ki) and nasıl ki are much more clearly ‘subordinate’
than the ki-structures just considered. Madem(ki) ‘since’ presents the information base
for a question or modalized utterance articulated in the main clause. It is interchange-
able with the NF form -DIğInA göre:
(71) a. [Ali-yi sev-me-diğ-in-e göre] on-u çağır-ma.
Ali-acc like-neg-vn-2sg.poss-dat according.to s/he-acc invite-neg
b. [Madem Ali-yi sev-mi-yor-sun] onu çağır-ma.
since like-neg-impf-2sg
‘Since you don’t like Ali, don’t invite him.’
Nasıl ki ‘just as’ is used in sentences expressing the similarity of one situation to another.
Its usage is paralleled by one of the functions of the universal conditional construction
nasıl … -(y)sA. The equivalent NF structure is the converbial combination -DIğI gibi:
(72) a. [Ayşe-nin her gün baba-sın-ı yokla-dığ-ı gibi] sen de
Ayşe-gen every day father-3sg.poss-acc check-vn-3sg.poss like you top
anne-n-le daha yakından ilgilen-meli-sin.
mother-2sg.poss-ins more closely take.interest-oblig-2sg
b. [Nasıl ki Ayşe her gün babasını yoklu-yor,] sen de annenle daha yakından
just.as check-impf
ilgilenmelisin.
c. [Nasıl Ayşe her gün babasını yoklu-yor-sa] sen de annenle daha
how check-impf-cond.cop
yakından ilgilenmelisin.
‘Just as Ayşe checks up on her father every day, you should take a closer inter-
est in your mother.’
Alternative subordination strategies in Turkish 

In the case of both madem(ki) and nasıl ki/nasıl …-(y)sA, one clear advantage that
these F-Sub-Initial structures offer to the speaker is pragmatic, similar to that of intro-
ducing an optional sanki at the beginning of an F-Sub-Final ‘as if ’ clause (see 4.3). The
presence of the subordinator at the very beginning of the sentence conveys to the hear-
er the earliest possible indication that s/he is to interpret the first clause that comes not
as an assertion but as the presupposition on which the main utterance will semanti-
cally depend.

5. Concluding remarks

It is clear that there is a robust coexistence in Turkish of finite and non-finite subordi-
nation strategies. The fact that the finite strategies are encountered mainly in the spo-
ken language might seem to support the claims made by various scholars (see Johan-
son 1992: 261) of psycholinguistic difficulty involved in both the production and the
processing of Turkish NF clauses. The growing body of evidence from the study of
language contact phenomena also appears to point in the same direction. By now it is
indisputable that wherever Turkish (or another Turkic language) is exposed to pro-
longed contact with a politically and/or demographically dominant Indo-European
language (Iranian, Slavic, Germanic, etc), there is a strong tendency for attrition to
occur in the indigenous (left-branching, non-finite) type of subordinate clause in fa-
vour of finite right-branching clauses introduced by a subordinating conjunction. (For
an expert overview and synthesis see Johanson 1992: 259–73.)
The fact remains, however, that Turkish NF clauses are very much holding their
own in Turkey itself, and that for some important semantic subtypes of all three of the
major syntactic categories there is no alternative strategy available. Moreover, while
many of the F structures are eschewed in formal writing, the converse is not true. The
majority of the native NF structures are regularly encountered in natural conversation.
The present survey suggests that one needs to distinguish between the different struc-
tural types of finite subordinate clause available in Turkish in order to identify the
reasons for their attractiveness in real-life speech situations.
In a study published in 1986, Slobin found (a) that acquisition of relative clauses
by children took place in Turkish at a later age than in English, and (b) that in natu-
rally occurring conversation Turkish-speaking adults used relative clauses less than
half as frequently as English-speakers. He identified (1986: 278–284) the “non-canon-
ical” structure of Turkish relative clauses as a major explanatory factor for these find-
ings, pointing in particular to the following features of Turkish relative clauses as di-
verging from the morphosyntax of simple sentences: (i) different verbal morphology;
(ii) reversal of SV order (where the subject is relativized); and (iii) genitive marking of
the subject (where a non-subject constituent is relativized).
Let us consider to what extent this analysis helps us to understand the phenomena
presented in our survey. We may take (i) and (iii) together. All the F subordinate claus-
 Celia Kerslake

es of Turkish replicate, by definition, the finite predicate morphology of simple sen-


tences. Because of this, there is no need for genitive marking of the subject, which is a
feature directly linked to the possessive marking of the nominalized predicates of
many types of NF clause.30 The combination of these two major points of resemblance
to simple sentences undoubtedly accounts for the fact that all types of F clause have a
homely, accessible feel compared with NF clauses in general and those with genitive-
marked subjects in particular. As for (ii), it is only the relative clauses of Turkish that
violate the SV (or OV) order of simple sentences. NF nominal and adverbial clauses
are completely canonical in their internal word order, as can be seen in all the NF/F
pairs of examples presented here. So this cannot be a factor in a speaker’s choice of an
F strategy in either of these categories.
If we turn from the internal ordering of the clause to its position within the sen-
tence as a whole, we find that NF clauses of all types behave exactly like the corre-
sponding non-clausal structures (noun phrases, adjectivals, adverbials). Relative
clauses precede the noun phrase that they modify, just like all other adjectival con-
stituents in Turkish. Nominal and adverbial clauses precede the predicate to which
they stand as arguments or adjuncts, except that any item that is context-recovera-
ble may be placed after the predicate (‘backgrounded’). There is then, nothing “non-
canonical” in the positioning of NF clauses within the sentence. On the contrary, it is
in a certain category of F clauses, namely those introduced by ki, that we encounter a
clear violation of the left-branching word order of Turkish, in respect both of post-
nominal and post-predicate placement.
It is a commonplace of linguistic typology that left-branching structures require
more information to be held in memory than right-branching ones. Hawkins (1994:
321) observes that certain “left-right asymmetries” in word order in languages “are
ultimately consequences of the fact that language is produced and comprehended in
an item-by-item manner from left to right, i.e. in a temporal sequence”. The ‘relative’
clause usage of ki is the clearest example of Turkish-speakers making compensatory
use of a right-branching strategy for ‘on-line’ additions or comments to information
that they have already presented. Most other usages of ki, as demonstrated above, can
be seen as a deliberate exploitation, for pragmatic or rhetorical purposes, of the pro-
sodic features and non-canonical clause sequence that this imported item offers.
Apart from ki-clauses, the regular position for all other finite subordinate clauses
(F-Bare, F-Sub-Final and F-Initial) is before the main predicate, exactly as is the case
with NF clauses. Like NF clauses, and under the same conditions, they can also be
placed after the predicate. (73) is a transformation of (39):
(73) a. Üzül-dü-k [sen-in gel-e-me-diğ-in]-e
be.sad-pf-1pl you-gen come-possib-neg-vn-2sg.poss-dat
b. Üzül-dü-k [sen gel-e-me-di-n diye]
come-possib-neg-pf-2sg sub
‘We were sorry you couldn’t come.’
Alternative subordination strategies in Turkish 

Finally, if we consider the ability of a subordinate clause to be made the focus of inter-
rogation by the postposed clitic mI, we find that while this is universal among NF
clauses, among the F clauses it is found only in the F-Bare and F-Sub-Final types. (74)
is based on (61) above:
(74) a. Osman, [Zeliha beğen-me-diğ-i için] mi sakal-ın-ı
Osman Zeliha like-neg-vn-3sg.poss for q beard-3sg.poss-acc
tıraş et-tir-miş?
shave aux-caus-ev/pf
b. Osman, [Zeliha beğen-me-di diye] mi sakalını tıraş ettirmiş?
-pf sub q
‘Is it because Zeliha didn’t like [it] that Osman has had his beard shaved off?’
By contrast (75), based on (41) above, shows that, while an NF noun clause can be the
focus of interrogation like any other NP, this is completely impossible for the corre-
sponding ki-clause:
(75) a. [Çevre kirliliğ-i-nin önemli ölçü-de
environment pollution-nc-gen important measure-loc
azal-dığ-ı] mı iddia ed-il-iyor?
decrease-vn-3sg.poss q claim aux-pass-impf
b. *İddia edil-iYOR ki, çevre kirliliği önemli ölçüde azal-mış mı?
decrease-ev/pf
‘Is it claimed that environmental pollution has substantially decreased?’
This interrogation test demonstrates that, within the finite alternatives available, it is
those that conform to the branching order of Turkish that are most fully integrated
into the structure of the sentence.

Notes

* This is a revised version of a paper presented at the International Colloquium on Connec-


tivity in Multilingual Settings held in Hamburg in November 2004. Thanks are due to two ano-
nymous reviewers for their constructive suggestions. I take full responsibility for any deficien-
cies that remain.
1. Finite subordinate clauses in Turkish have not generally been treated collectively. They re-
ceive only brief treatment in Kornfilt 1997: 46–47. An early exploration of the topic is that of
Johanson 1991 (first published 1975). Subgroups that have received particular attention are (i)
clauses introduced by ki (see Haig 1998: 116–128, Schroeder 2002) and (ii) the type of comple-
ment clauses displaying Exceptional Case Marking (ECM) (the so-called “small clauses”), for
which see note 2 below. An overview of Turkish subordinate clause structures (nominal, adjec-
tival and adverbial) that brings together the finite and non-finite resources available within each
 Celia Kerslake

category is given in Göksel and Kerslake (2005), Chapters 24–26; see also the remarks about
subordination in general on pp. 135–137.
2. The optional use of accusative marking on the subject of finite complement clauses (mainly of
the matrix verb san- ‘think’) has been the subject of much analytical discussion within the genera-
tive framework since the 1970s. See, for example, Brendemoen and Csató 1986, Özsoy 2001.
3. The Persian ki is related to French qui/que, Italian che, etc.
4. For a historical survey of the use of ki in Turkish and the Turkic languages in general see
Erguvanlı 1980–1981.
5. Cognates of diye occur as F-Sub-Final subordinators in virtually all the Turkic languages
(Johanson 1992: 275–276).
6. For full details of Turkish verbal morphology see Göksel and Kerslake 2005, Chapter 8.
7. For a generative analysis of Turkish non-finite complement clause morphology see Kural
1998. For the criteria determining the choice of one morpheme rather than another see Taylan
1998.
8. In the citation of suffixes, capitals indicate consonants or vowels that are subject to phono-
logical alternation. The K of -mAK is usually deleted in the accusative and dative forms, which
surface as -mA-yI and -mA-yA (see examples (17), (18)).
9. The morpho-syntax of Turkish non-finite relative clauses has been a subject of great interest
to generative linguists since the 1970s. For a recent contribution by one of the leading partici-
pants see Kornfilt 2000. On Turkish relative clauses in general see the comprehensive mono-
graph by Haig (1998), written from a functional-typological standpoint.
10. With the exception of Schroeder’s work on quasi-relative ki-clauses, referred to in section
3, I am not aware of any empirical study on which to support the claims that I make about the
stylistic status of particular constructions. Although some quantitative documentation would be
desirable, in its absence I remain confident that these claims are non-controversial.
11. -(y)Iş has a much more restricted field of application than -mA. It is used almost exclusively
for single or countable instances of actualized events/states, never for those that are merely en-
visaged. See Erdal 1998.
12. In this and some other examples capitalization is used to indicate the position of focal
stress.
13. NF noun clauses functioning as subjects or complements of verbs of emotion may be
marked by -mA or -(y)Iş instead of -DIK/-(y)AcAK. See Göksel and Kerslake 2005: 429–30.
14. See Schroeder 2002: 74–75.
15. On non-restrictive relative clauses in Turkish see Johanson 1991: 221, Erkman-Akerson
and Ozil 1998: 134–5, Haig 1998: 127 and Schroeder 2002: 80.
16. The Arabo-Persian quantitative metrical system, used in classical Ottoman poetry.
17. The discussion of relative clauses here is confined to those that have an adjectival function;
it has not been possible to consider the ‘headless’ or ‘pronominal’ type.
18. There is a marginal and rather literary usage of ki in restrictive relative clauses, which limi-
tations of space did not permit to be discussed in this article; see Göksel and Kerslake 2005:
458–9.
Alternative subordination strategies in Turkish 

19. Cf. Schroeder 2002: 78, where the “antecedent” is conceptualized not as the antecedent of
the ki-clause itself but of its “topic”.
20. This phonological distinction between different usages of ki appears to have been first recog­
nized by Bainbridge (1987: 48–54). It is mentioned in Haig 1998: 123–125, and systematized in
Schroeder (2002: 74–75).
21. These tentative formulations of an as yet under-researched issue are a further development
of Göksel and Kerslake 2005: 458; cf. also Kissling 1960: 145.
22. Or an anaphoric possessive suffix.
23. In this and the other examples borrowed from Schroeder (which are transcripts of natu-
rally occurring speech) the presentation of the Turkish text has been adapted to the conventions
of written Turkish, and the translations have undergone some modification.
24. The ‘sentential’ type of relative ki-clause was recognized by Deny (1921: 852).
25. The parenthetical function of ki-clauses is noted by Lewis (2000: 212).
26. The uses of ne zaman ki ‘when’ are in need of research, and it has not been possible to consi-
der this item further here. It seems likely that it plays a ‘forewarning’ role analagous to that of
nasıl ki (4.5).
27. On finite clauses with the subordinator gibi see van Schaaik 1998: 443–457.
28. The implication in Göksel and Kerslake 2005:466 that person marking is obligatory in
-(y)mIş gibi is erroneous.
29. Internet searches on www.google.com on18.09.2006 of forms such as yapmışçasına versus
yapmış gibi, alıyormuşçasına versus alıyormuş gibi yielded results for the constructions with gibi
that were markedly greater (often by a factor of 100–3,000) than those for the forms with the
converbial suffix.
30. Genitive marking of the subject is required in most NF clauses whose predicate is marked
with a possessive suffix attached to -mA, -DIK, -(y)AcAK or -(y)Iş. The main exceptions are noun
clauses whose subject denotes unspecified members of a category, and most adverbial clauses
marked with -DIK or -(y)AcAK.

References

Bainbridge, M. 1987. Loan conjunctions versus native syntax. In Studies on Modern Turkish, H.
E. Boeschoten and L. T. Verhoeven (eds), 42–56. Tilburg: Tilburg University Press.
Brendemoen, B. and Csató, É. Á. 1986. The head of S in Turkish: A comparative approach to Turk-
ish syntax. In Proceedings of the Turkish Linguistics Conference, August 9–10 1984, A. Aksu
Koç and E. Erguvanlı Taylan (eds), 85–100. Istanbul: Boğaziçi University Publications.
Deny, J. 1921. Grammaire de la langue turque (dialecte osmanli). Paris: Leroux.
Erdal, M. 1998. On the verbal noun in -(y)Iş. In Doğan Aksan Armağanı, K. İmer and L. Subaşı
Uzun (eds), 53–68. Ankara: Ankara Üniversitesi Dil ve Tarih-Coğrafya Fakültesi.
Erguvanlı, E. 1980–1981. A case of syntactic change: ki constructions in Turkish. Boğaziçi Uni-
versity Journal, Humanities 8–9: 111–139.
Erkman-Akerson, F. and Ozil, Ş. 1998. Türkçede Niteleme: Sıfat İşlevli Yan Tümceler [Türk Dilleri
Araştırmaları Dizisi 22]. Istanbul: Simurg.
 Celia Kerslake

Göksel, A. and Kerslake, C. 2005. Turkish: A Comprehensive Grammar [Routledge Comprehen-


sive Grammars]. London: Routledge.
Haig, G. 1998. Relative Constructions in Turkish [Turcologica 33]. Wiesbaden: Harrassowitz.
Hawkins, J. A. 1994. A Performance Theory of Order and Constituency. Cambridge: CUP.
Johanson, L. 1991[1975]. Some remarks on Turkic ‘hypotaxis’. Reprinted in Linguistische Bei-
träge zur Gesamtturkologie [Bibliotheca Orientalis Hungarica XXXVII], 210–224. Buda-
pest: Akadémiai Kiadó.
Johanson, L. 1992. Strukturelle Faktoren in türkischen Sprachkontakten. [Sitzungsberichte der
Wissenschaftlichen Gesellschaft an der J. W. Goethe-Universität Frankfurt am Main XXIX
5]. Stuttgart: Franz Steiner.
Kissling, H. J. 1960. Osmanisch-türkische Grammatik [Porta Linguarum, Neue Serie III]. Wies-
baden: Harrassowitz.
Kornfilt, J. 1997. Turkish [Descriptive Grammars]. London: Routledge.
Kornfilt, J. 2000. Some syntactic and morphological properties of relative causes in Turkish. In
The Syntax of Relative Clauses [Linguistik Aktuell/Linguistics Today 32], A. Alexiadou, P.
Law, A. Meinunger and C. Wilder (eds), 121–159. Amsterdam: John Benjamins.
Kural, M. 1998. Subordinate Infls and Comp in Turkish. In The Mainz Meeting: Proceedings of
the Seventh International Conference on Turkish Linguistics [Turcologica 32], L. Johanson et
al. (eds) 404–421. Wiesbaden: Harrassowitz.
Lewis, G. 2000. Turkish Grammar (2nd ed.). Oxford: OUP.
Özsoy, A. S. 2001. On ‘small’ clauses, other ‘bare’ verbal complements and feature checking in
Turkish. In The Verb in Turkish [Linguistik Aktuell/Linguistics Today 44], E.Erguvanlı Tay-
lan (ed.), 213–237. Amsterdam: John Benjamins.
Quirk, R., Greenbaum, S., Leech, G. and Svartvik, J. 1985. A Comprehensive Grammar of the
English Language. London: Longman.
Schaaik, G. van 1998. On the usage of gibi. In The Mainz Meeting: Proceedings of the Seventh
International Conference on Turkish Linguistics [Turcologica 32], L. Johanson et al. (eds)
422–457. Wiesbaden: Harrassowitz.
Schroeder, C. 2002. On the structure of spoken Turkish. ELISe (Essener Linguistische Skripte –
elektronisch) 2: 73–90. [http://www.elise.uni-essen.de]
Slobin, D. I. 1986. The acquisition and use of relative clauses in Turkic and Indo-European lan-
guages. In Studies in Turkish Linguistics, D. I. Slobin and K. Zimmer (eds) [Typological
Studies in Language 8], 273–294. Amsterdam: John Benjamins.
Song, J. J. 2001. Linguistic Typology: Morphology and Syntax [Longman Linguistics Library].
Harlow: Pearson Education.
Taylan, E. E. 1998. What determines the choice of nominalizer in Turkish nominalized comple-
ment clauses? In Proceedings of the XVIth International Congress of Linguists, 20–25 July,
1997, B. Caron (ed.). New York: Elsevier (CD Rom).
Studying connectivity with the help of
computer-readable corpora
Some exemplary analyses from modern and
historical, written and spoken corpora

Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner


and Ludger Zeevaert
University of Hamburg

This paper discusses methodological aspects of the use of electronic language


corpora for the study of connectivity. We demonstrate how a corpus-based
approach was used to investigate functional characteristics of coordinating
elements in sentence- or utterance-initial position across different languages
(English, German, Old Swedish and Turkish), across different modalities
(written and spoken) and across the diachronic dimension (historic and modern
languages). Our focus is on the difficulties we encountered in this study when
attempting to transfer corpus-based methods developed for the analysis of
corpora of modern, written language to the analysis of corpora of historic or
spoken language. We suggest an abstract corpus-linguistic workflow and discuss
where and how this workflow differs according to the corpus type, and how well
its individual steps are supported by current corpus technology.

1. Introduction

The prerequisites for the comparative analysis of language use across different lan-
guage modes and historical periods are comparable data and a common method in
approaching – though not necessarily in analyzing – the data. By the example of lan-
guage-contrastive and diachronic-contrastive investigations into the use of coordinat-
ing elements in sentence- and utterance-initial position, this paper aims at showing
how disparate types of data can be aligned within a corpus-driven approach to linguis-
tic research. The paper is structured as follows: First, a brief general account of our
interest in macrosyntactic coordination (discourse coordination) will be given and the
common method which was eventually applied to the different corpora will be sum-
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

marized. We will then proceed to give detailed descriptions of how the comparatively
well-established methods of analysing modern written corpora in text and register
linguistics were adapted to the particular conditions determining spoken language
corpora and corpora of historical texts. The goal was to make the data comparable
without compromising the constitutive formal differences between modern and his-
torical and written and spoken language use, which are necessarily reflected in the
corpora compiled for linguistic analysis. The paper concludes with a summary of the
first results which we were able to obtain through the application of a common method
to diverse corpora.1

1.1 Research question: Coordinating elements and linguistic variation

The idea for the project reported on in this article developed from the shared interest in
the emergence of linguistic innovation in situations of language contact, and the wish
to compare the diachronic development of the use and function of coordinating con-
junctions in sentence- and utterance-initial position. Qualitative analyses of English
and German original texts and translations from English into German and bilingual
and monolingual spoken Turkish, carried out in different project groups, as well as
observations gathered from the investigation of Old and Early Modern Swedish texts
from a further project group, seem to suggest that sentence-initial And and Und, the
Turkish expression o zaman and the Old Swedish sentence-initial och are all involved in
processes of linguistic variation. More detailed accounts of the different kinds of varia-
tion observed in the individual corpora will be provided in Sections 2, 4 and 5. The
focus of the present article, however, is on the methodological aspects of linguistic ana­
lysis across different types of data as they appear in modern, historical, written and
spoken corpora. In the remaining part of this section we will briefly describe the major
methodological challenge which hitherto made both the step from qualitative case
studies to the analysis of larger corpora and the comparison of results from work on
related research questions with different (i.e. non-comparable) corpora difficult.
In order to be able to describe the observed variation in the context of the coordi-
nating elements And/Und, o zaman, and och more closely and to assess its validity with
respect to a larger amount of data, it was essential to find a way of transferring the re-
sults of the analysis of single text and discourse exemplars to the analysis of the corpora
as a whole. Because the number of texts and discourses which can be analyzed manu-
ally in a reasonable amount of time is limited, it is necessary to formalize the process of
analysis such that parameters for the (semi-)automatic search for the relevant linguistic
phenomena in the corpora can be deduced. In a simplified manner, such a combined
qualitative and quantitative approach involves using the lexical and grammatical fea-
tures which the qualitative analyses point out as salient in the investigation of a particu-
lar linguistic phenomenon to query the corpus as a whole with the help of software
tools. The total of the occurrences (usually in the form of concordances) is then sub-
jected to further qualitative and recontextualizing analysis. These analyses establish
Studying connectivity with the help of computer-readable corpora 

whether all occurrences of the linguistic phenomenon in the corpus coincide with the
analysis deduced from the case studies or whether there is functional or formal varia-
tion which was not uncovered by the analysis of single text and discourse exemplars.
Although clearly not all corpus-driven linguistic investigations follow such a combined
qualitative and quantitative approach, the use of computerized corpora and software
search tools has developed into a comparatively established practice in text and register
linguistics dealing with modern written texts over the last ten to fifteen years. As will be
presented in Section 3 below, the situation is radically different for virtually all other
types of linguistic data – e.g. spoken discourse and historical texts. It is this incompat-
ibility of the data which often hinders what otherwise might become a fruitful coopera-
tion and constructive coordination of research efforts. In the following section, the in-
vestigation into the function of sentence-initial And and Und in a corpus of English and
German popular scientific writing will be presented as an example of a ‘conventional’
corpus-linguistic approach to variation. We will then go on to describe whether and
how the method applied to modern texts can be transferred to similar research ques-
tions addressed to different – spoken and historical – data.

2. Studying connectivity in a corpus of modern, written language

2.1 Research question: Language contact induced changes in the conventions


of use of And and Und in English and German popular scientific texts

This investigation into the use and function of sentence-initial And and Und in English
texts, German texts and German translations from English was carried out within the
project ‘Covert Translation – Verdecktes Übersetzen’ (cf. Baumgarten to appear). The
project starts from the hypothesis that English, due to its status as a global lingua franca
and prestige language, exerts an influence on those languages with which it is in pro-
longed contact. In the case of German, in particular, the contact with English seems to
result, in certain genres, in an adaptation of German communicative preferences and
textual norms to the ones operative in English texts. It is assumed that this influence is
initially most marked in German translations from English because translations can be
considered as one of the primary interfaces between the English and German language
systems; they are the ‘gateways’ through which non-native, i.e. English, source text influ-
enced, uses of certain linguistic elements can enter German language use.
The differences between the English and German communicative conventions and
stylistic norms can be summarized as follows: Converging evidence from a variety of
English-German contrastive pragmatic studies investigating different spoken and writ-
ten genres points towards a general hypothetical pattern of differences in the communi-
cative behavior of native speakers of both English and German (cf. for example Byrnes
1986; Clyne 1987, 1994; Doherty 1996, 2002; House 1989, 1996, 2004; Kotthoff 1989).
This pattern can be displayed along five dimensions of communicative preferences:
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

ENGLISH GERMAN
Indirectness Directness
Orientation towards other Orientation towards self
Orientation towards persons Orientation towards content
Implicitness Explicitness
Use of verbal routines Ad-hoc-formulation

Figure 1.  Dimensions of communicative preferences between German and English


(House 1996)

These dimensions are clines rather than clear-cut dichotomies with absolute values,
reflecting tendencies in language use rather than categorical distinctions. One can say,
however, that native speakers of German tend to realize linguistic features which are
associated with the values on the right of figure 1. Native speakers of English, in the
same situations, prefer to use linguistic structures associated with the values on the left
of this figure.2 At the same time, due to language typological reasons, English and Ger-
man display different patterns of sentence-internal and macrosyntactic cohesion (Hal-
liday & Hasan 1976). Fabricius-Hansen (1999), for example, describes this difference
in terms of differences between conventionalized levels of informational density in the
individual languages. She suggests that, in contrast to German, English favors incre-
mentatility of discourse information. That is, in English, an additive organization of
discourse information, portioned into smaller chunks in order to reduce information-
al density, is preferred, whereas in German, informationally dense, hierarchical infor-
mation packaging is more common. This finds expression in different preferences in
the use of connecting devices.
Considering the functional descriptions provided for structural coordination with
and (e.g. Quirk et al. 1985; Blakemore & Carston 1999) and the use of sentence- and
utterance-initial And (cf. Halliday & Hasan 1976; Schiffrin 1986, 1987; Redeker 1990;
Biber et al. 1999), And can be described both as a marker of a semantically vague con-
nection between propositions and of the speaker’s interpersonal involvement in the
communicative event. Hence, one can conclude that the use of sentence-initial And in
English texts coincides with the preference for an all in all more ‘oral’, interpersonally-
oriented written style, which arranges discourse information in a linear, additive man-
ner, signaling a connection between the conjuncts but leaving the exact nature of this
relation implicit. In contrast, the use of the coordinating conjunction Und, which has
a functional profile roughly comparable to that of the English And (Ehlich 2001; Wein-
rich 2003, Zifonun et al. 1997),3 in sentence-initial position does not coincide with the
typical pattern of communicative norms in written genres in German, which demand
a lesser extent of interpersonal orientation and greater referential explicitness and pre-
cision. The use of Und to express macrosyntactic coordination in written texts would
therefore have to be considered as a stylistically marked choice.
Studying connectivity with the help of computer-readable corpora 

On this basis, one would expect the use of sentence-initial And to be on the whole
more frequent in English texts than the use of Und in German texts. With respect to the
project’s hypothesis, the use of Und, should, first, be more frequent in German transla-
tions from English than in German original texts. Secondly, as evidence for a shift in
German communicative preferences in the direction of the English ones, in a diachron-
ic perspective, the use of Und in German translations should catch up with the use of
And. These assumptions were investigated in a diachronic English-German translation
and parallel text corpus, which will be introduced in the following section.

2.2 Corpus: Diachronic English-German translation and parallel text corpus

The corpus is a diachronic translation and parallel text corpus for the language pair
English and German (approx. 800000 words). It consists of texts from popular scien-
tific journals and external business communication (mainly letters to shareholders).
The texts cover the years from 1978 to 2002.

Figure 2.  English-German translation and parallel text corpus

The translation part contains English original texts and their translations into German
and German original texts and their translations into English.4 It serves to investigate
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

the particular features of the translation relation English-German and German-Eng-


lish. The parallel part contains
1. English original texts and German original texts. They are analyzed with respect
to the linguistic realization of language- and genre-specific conventions of text
production;
2. German translations from English and German original texts, and English trans-
lations from German and English original texts. These are analyzed with respect
to the (potentially different) characteristics in the use and combination of lexico-
grammatical features in original texts and translations.
The frequency and distribution of sentence-initial And and Und was initially investi-
gated in a subset of the popular scientific part of the corpus. The database consists of
the following sets of texts covering the time frames 1978–1982 and 1999–2002:
1. English original texts from the years 1999–2002 (122866 words).
2. The German translations of these English texts (113420 words).
3. German original texts from the years 1999–2002 (100648 words).
4. English original texts from the years 1978–1982 (42497 words).
5. The German translations of these English texts (37830 words).
6. German original texts from the years 1978–1982 (82480 words).

2.3 Results

The use of sentence-initial And and Und in English and German popular scientific
texts in the two time frames is displayed in tables 1 and 2 below:

Table 1.  Occurrences of And and Und (normalized frequencies on the basis of 10000 words)

1978-1982 1999-2002 relative increase

English (And) 3,1 4,5 + 45,2%


German translations (Und) 2,3 6,3 + 173,9%
German (Und) 0,9 3,1 + 244,4%

Table 2.  Use of And and Und in relation to the total of sentences

1978-1982 1999-2002 relative increase

English (And) 7,4 ‰ 9,7 ‰ + 31,1%


German translations (Und) 5,1 ‰ 11,3 ‰ + 121,6%
German (Und) 2,0 ‰ 5,8 ‰ + 190,0%

The figures show a diachronic development in the use of And and Und which is equiv-
alent to the starting hypotheses. In both English and German texts the use of mac-
Studying connectivity with the help of computer-readable corpora 

rosyntactic coordination by means of And and Und increases. The increase, however,
is more pronounced in the German texts with the German translations even surpass-
ing their source texts in the use of the coordinating conjunction in sentence-initial
position. It follows that the use of Und in the newer translations cannot be triggered by
the presence of a sentence initial And in the source texts alone. This has tentatively
been interpreted as a dissociation of the text conventions determining the use of the
stylistically marked option of Und as a means of textual cohesion in German transla-
tions from those operative in German original texts.5 A difference between the textual
function sentence-initial And and Und express in the texts can be observed in the con-
text phenomenon of what can be termed as ‘subject switches’. According to Halliyday
& Hasan (1976) sentence-initial And is very often used to facilitate a shift in the par-
ticipants from one sentence to the next as for instance the shift from “E. coli” to “our
defenses” in example 1 below:6
(1) <seg> Conflicts with Other Organisms </seg>
<seg> Natural selection is unable to provide us with perfect protection against
all pathogens, because they tend to evolve much faster than humans do. </seg>
<seg> E. coli, for example, with its rapid rates of reproduction, has as much
opportunity for mutation and selection in one day as humanity gets in a mil-
lennium. </seg>
<seg> And our defenses, whether natural or artificial, make for potent selec-
tion forces. </seg>
Such shifts in the participants in subject position in the context of And often co-occur
with changes in the tense or the modality of the sentence as shown in example 2.
(2) <seg> In the absence of medications capable of dependably eliminating the
virus, the NIH recently embarked on a study to determine whether long-term
administration of alpha interferon can slow liver damage in patients who fail
to clear the virus. </seg>
<seg> And we and other researchers are studying the simple expedient of
taking a pint of blood from patients on a regular basis. </seg>
The distribution of these phenomena in the context of And and Und is displayed in table 3.7
Like the figures for the diachronic development of the use of And/Und, the results
for the synchronic distribution of the context phenomena suggest a difference between
the conventions of using Und in the German translations and the German original
texts. While the frequency counts presented in tables 1 and 2 above describe the dia-
chronic development of the use of And and Und in terms of their occurrence on the
linguistic surface of the texts, the proportions presented in table 3, which result from a
recontextualizing analysis of the total of occurrences, provide a view on the functional
diversification of And and Und in the texts. The German translations in both time
frames consistently show more occurrences of Und than the German original texts and
an increase in the use of Und which results in a larger number of occurrences of Und
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

than are licensed by the corresponding linguistic structures in their source texts. How-
ever, the position of the German translations between the English and the German
original texts with respect to the realization of the context phenomena strongly indi-
cates that identical surface forms need not necessarily express the same textual func-
tions. In terms of the introduction of subject, tense and modality switches, the transla-
tions show similarities with the English texts, whereas considering the particular
phenomenon of combined subject and tense/modality switches, the translations are
more similar to the German original texts. In other words, the use of And and Und in
English and German texts and German translations may be partly differently moti-
vated. Hence, at least in the case of And/Und an alignment of German text conventions
with English ones has to be considered on two levels: first, on the level of the linguistic
forms and secondly, on the level of their textual function which derives from the con-
text of their occurrence. In Section 5 below, we will see another particularly striking
example for functional diversification behind an apparently identical surface form. In
the following part of this section the method of corpus creation and the technology
used for its analysis will be presented.

Table 3.  Context phenomena in relation to the total of occurrences of And/Und in the time
frame 1999-2002

English German trans- German


(And) lations (Und) (Und)

subject switch 67,8 % 62,5 % 53,1 %


subject switch + tense/ 37,5 % 19,4 % 15,6 %
modality switch
tense-/modality switch 8,9 % 8,3 % 28,1 %
without subject switch
no switch (referential 23,3 % 29,2 % 18,8 %
continuity)

2.4 Method and technology

2.4.1 Corpus creation and processing


The corpus consists of published articles from popular scientific journals and texts
taken from the external business communication of internationally operating compa-
nies. The texts are digitized through scanning, using OCR-software. After manual ed-
iting they are saved as text files. The texts are then automatically tokenized and anno-
tated with part-of-speech-tags using the TnT-tagger and the SUSANNE and STTS
tagsets for English and German, respectively.8 The translation part is aligned on the
basis on translation-equivalent sentences (manual alignment).
Studying connectivity with the help of computer-readable corpora 

2.4.2 Corpus analysis


The corpus analysis is carried out with commercially available ready-to-use corpus
software.9 The software allows to search the corpus for strings of characters (i.e.
‘words’), regular expressions and combinations of these. The search results are dis-
played in the form of KWIC-concordances.

Figure 3.  Screenshot of a query for “And*” in ParaConc. The upper part of the window
displays the matches on the query expression, the lower part the translation equivalents

The concordances are manually checked for accuracy and the ensuing frequency
counts are normalized on the basis of 10000 words. The normalized frequencies facili-
tate the comparison of the lexical and grammatical features across different data sets.
Each of the occurrences is then subjected to qualitative recontextualizing analyses.
This is carried out in order to assess the function of each instance of use of one par-
ticular linguistic item in its context of occurrence, i.e. with respect to the information
organization on both sentence and textual levels and its contribution to the function of
the textual whole. For this purpose, each occurrence and the immediately preceding
and following context (5 to 10 orthographic sentences) are extracted into separate text
files. These files form a subcorpus of the total of the linguistic items under investiga-
tion embedded in their contexts of occurrence. This subcorpus can be annotated with
further information which characterizes the use of the linguistic item in question, for
example, in terms of patterns of colligation, i.e., the co-occurrence of grammatical
choices, and of collocation, i.e., the co-occurrence of lexical choices at clause, sentence
and text level. This can be carried out with the help of a coding tool (e.g. Systemic
Coder, O’Donnell 2004, figure 4) which furthermore allows to perform some statistical
operations on the annotated subcorpora. The goal of this additional analytical step is
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

to provide close linguistic descriptions of each single occurrence in order to detect


variation behind identical surface forms.

Figure 4.  Screenshot of the Systemic Coder annotation interface

3. Some methodological and technological issues in the study of computer


readable corpora

3.1 Corpus linguistics as a method

The kind of linguistic research outlined in the previous section contains many ele-
ments that are typical of “corpus linguistics”. As Leech (1992) points out, “the term
‘corpus linguistics’ does not refer to a domain of study, but rather to a methodological
basis for pursuing linguistic research.” – in other words: corpus linguistics is a research
method, that is a manner of proceeding which generalizes over the individual case of a
scientific investigation in order to provide a guideline for obtaining comparable results
for related, but varying research questions or objects of investigation. The study pre-
sented here is an instance in which such a common method could be advantageous,
seen as the research questions are similar in all three cases: they aim at finding quanti-
tative evidence for language change in the area of coordinating conjunctions. A com-
mon research method could therefore help to make the results comparable. The ob-
jects of investigation, however, are highly different, not only in terms of the languages
Studying connectivity with the help of computer-readable corpora 

(English/German, Old Swedish, Turkish) involved, but also because we are dealing
with modern, written language in the first case, with historical, written language in the
second case and with spoken language in the third case. This diversity, in turn, makes
a method transfer a non-trivial matter. Therefore, the purpose of this section is to
identify some salient characteristics of the corpus-linguistic method and to analyze if
and how the manner of proceeding outlined above can be successfully transferred to
other objects of study, and where and why such a transfer is found to be problematic.

3.2 The workflow of a corpus linguistic study

On a sufficiently abstract level, we can agree that a corpus linguistic study invariantly
consists of four distinct phases as displayed in figure 5:

I IIa IIb IIc III

Corpus Corpus Corpus Corpus Corpus


Design Creation Processing Annotation Analysis
Model

Application level

Selection / Query
Query result 1

Sample
Model

Conceptual level
Query result 2
Population
Model

Physical level
Query result 3

Corpus
Language Material

Quantification
Interpretation

IV

Figure 5.  Workflow of a corpus linguistic study

In phase I – corpus design – the researcher defines a population of language material


that he wants to make a statement about and picks out a sample that he considers to be
representative of that population. In phase II – corpus processing – he then starts by
creating a corpus in which the salient information of the sample is represented in a
number of digital files. In further processing steps additional structure or informa-
tion may be added to that digital representation (e.g. tokenization or POS-annota-
tion). When this process is completed, phase III – the actual corpus analysis – can
start. Typically, this step consists in querying the corpus for the digital representation
of a specific phenomenon, i.e. selecting those parts of the corpus that are relevant to
the research question. Finally, in phase IV, this set of query results can then be quanti-
fied and interpreted with respect to the original population defined in phase I. The
following subsections will discuss two aspects of this workflow which reveal differ-
ences with respect to the three corpus types.
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

3.3 The relationship between sample and corpus

One of the most fundamental differences with respect to the three corpus types – mod-
ern, written language, historical, written language and spoken language – lies in rela-
tionship between the sample and its digital representation.
In a corpus of modern, written language, the transition from the former (e.g. a
journal article) to the latter (e.g. a text file) is usually a simple mapping of printed sym-
bols to the entries of a character chart (e.g. ASCII or Unicode). Although some ab-
straction is involved in this process, it is restricted to rather superficial properties (for-
matting and layout) of the original, and hence hardly requires an elaborate theoretical
justification.10
This is different for corpora of historical, written language. In this case, the transi-
tion from the original text (e.g. a 16th century handwriting of a Bible text) to its digital
representation (e.g. an XML file) will usually mean more than a simple abstraction
over physical properties; it will also involve a number of theoretically motivated deci-
sions and interpretations. These decisions – such as the writing out of abbreviations or
a segmentation of the text into entities not provided by its author – may have been
anticipated by the edition that the corpus is based on, or they may have to be made by
the corpus linguist himself. In any case, they make the relationship between sample
and corpus more complex than in the case of modern, written language.
For corpora of spoken language, finally, it is an uncontested part of linguistic
methodology that this relationship can only be characterized with respect to an under-
lying theory. The sample, in this case, consists of a number of audio recordings, where-
as the corpus consists of a set of corresponding transcription files. The transcription
process in which these recordings are transformed into a set of computer-readable files
is widely acknowledged to be of a highly selective and interpretative nature (see, for
instance, Rehbein et al. 2004 or Ochs 1979).
Thus, when applying the corpus linguistic method to these different objects of
study, one has to take into account that the transition from sample to corpus is a mod-
elling process, determined by different degrees of simplification, abstraction and inter-
pretation (see Schmidt 2005 for a more detailed account of this aspect).

3.4 Data models, file formats and software tools

Besides assessing which parts of the original to leave out, to simplify, to interpret etc,
the decision on an adequate digital representation of the sample also involves the
choice of an appropriate data model.
Since a corpus of modern, written language can be viewed (at least in the corpus
creation phase, i.e. before other processing steps are carried out) as a simple sequence
of characters, it can be represented with the most straightforward of data models – a
flat text. In addition to the explicit linear structure inherent in such a flat text, there is
also an implicit hierarchical structure “encoded” in the punctuation marks.
Studying connectivity with the help of computer-readable corpora 

For a corpus of historical, written language, this data model will usually not be suf-
ficient: on the one hand, there is a need to represent those pieces of information that
have been added to the original text in the editing process – for instance, it may be nec-
essary to have both abbreviated and full forms or orthographic variants and their nor-
malized lemmas side-by-side. On the other hand, a historical text may not have its hier-
archical structure implicitly represented in its punctuation, so that boundaries of words,
sentences etc. will have to be added explicitly and in a manner that allows the computer
to distinguish between them and the alphanumerical characters that represent symbols
of the original text. A data model that meets these more advanced requirements is the
so-called OHCO (Ordered Hierarchy of Content Objects) model (DeRose et. al 1990).
It sets the rules for ordering and markup of the beginning and the end of content objects
and is nowadays closely associated with XML-based text processing.
Finally, in addition to these requirements, an adequate data model for spoken lan-
guage also has to provide a means of digitally representing temporally parallel relation-
ships, as can be found, for instance, in overlapping speech or between different mo-
dalities. From the data models suggested for this purpose, graph-based frameworks
like the AG formalism (Bird & Liberman 2001) or EXMARaLDA seem to be among
the most promising at this moment in time.11
For reasons inherent in the various objects of investigation, different corpus types
will thus require data models of different complexity. This has significant consequenc-
es for the practical conditions under which a corpus linguistic study is carried out: the
more complex a data model is, the less support is available for it in terms of standard-
ized data formats and, probably even more important, ready-to-use software for cor-
pus processing and analysis.
A linguist dealing with modern, written language can more or less act on the as-
sumption that all the technical tools he might require for creating, processing and ana-
lyzing his corpus – such as taggers, tokenizers and concordancers – are readily availa-
ble in some form or other. Therefore, the main challenge in his work is to put these
tools to a meaningful use and to provide an adequate interpretation of the results that
the chosen tools lead him to.
The situation for a historical linguist is far more complex. Although XML-based
data processing is gaining ground also in corpus linguistics, there is – to our knowl-
edge – no integrated set of tools that would adequately support a researcher in all the
phases of a corpus linguistic study. The fact that XML is now a widely used standard,
supplemented by the TEI guidelines which specifically address linguistic concerns of
XML based text encoding, certainly constitutes a solid basis for this technological side
of things. Working with an OHCO-based data model, however, still implies that a
substantial part of a corpus linguistic study will include finding and implementing ap-
propriate methods of corpus processing and analysis.
Finally, in the case of corpora of spoken language, it is debatable whether or not
the currently available software allows a successful passage through all the phases of a
corpus linguistic study at all. There are, of course, a number of software tools for tran-
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

scribing and annotating spoken language and for querying the resulting corpora.
However, since there is as yet no standard that can ensure the same basis to graph
based data models as XML and TEI do for the OHCO based one, these partial solu-
tions often lack the inter-operability that would be necessary in order to profitably
combine their abilities. The work of a linguist interested in spoken language will there-
fore be determined by the deficiencies of current corpus technology even more than
that of a historical linguist.
In the next two sections these considerations will not only be illustrated in more
detail, we will also describe two partial solutions to such deficiencies.

4. Studying connectivity in a corpus of historical, written language

4.1 Research question: Functional characterization and diachronic develop-


ment of word order patterns after sentence-initial conjunctions in Old Swedish

The primary research question for the analysis of the Old Swedish text corpus was to
investigate the role of contact with foreign languages (Latin, Middle Low German,
Early Modern High German) on the diachronic development of syntax in Swedish.
This investigation aims both at describing and quantifying the different word order
patterns that are found in conjunctional main clauses and at identifying a possible
functional distribution and a diachronic development in the use of those patterns.
The starting point for this research question is provided by some general state-
ments in the literature on word order in Old Norse: Christoffersen (2002, 2003) de-
scribes a difference in word order between main clauses beginning with ok (and) and
non-conjunctional main clauses. After the conjunction ok so-called narrative inver-
sion in declarative sentences is very common with the finite verb placed in initial posi-
tion in front of the subject; the usual order with the subject in the fundamentfield and
the finite verb in the first position of the nexusfield is rather rare.12 Some first unsys-
tematic observations on the Old Swedish corpus indeed showed word order variation
after ok, and this investigation is aimed at quantifying this variation.
Based on the hypothesis of syntactic change through language contact we aim at
describing a diachronic development of word order. The first step for such an attempt,
which is described in the following, concentrates on developing a suitable method of
analysis leading to a quantitative confirmation of our basic assumption. Moreover, by
comparing our results with those of the other participating researchers working on
different languages and situations of language contact we want to check the plausibil-
ity of our findings.
Studying connectivity with the help of computer-readable corpora 

4.2 Corpus: Bible texts and chronicles

The investigation is based on the following corpus of six Old and Early Modern Swed-
ish prose texts altogether containing about 70.000 words:
The Revelation of St John and The Gospel according to Mark from the Swedish
translation of the New Testament from 1526, The Revelation of St John and The Gospel
according to Luke from the Swedish translation of the Bible from 1541 (so called Gus-
tav-Vasa-Bible), the municipal records of Kalmar (Kalmar stads tänkebok, 1381–1560)
and the Old Swedish version of the story of Charlemagne (Karl Magnus, ca. 1430).

4.3 Result

As already mentioned in Section 4.1 Old Norse main clauses usually exhibit the word
order VS (finite verb before substantial) after the conjunction ok (‘and’),13 whereas in
non-conjunctional main clauses the unmarked word order SV (substantial before fi-
nite verb) is used. In several publications (Platzack 1985, Halldór 1990, Wessén 1992,
Christoffersen 1993), however, it was possible to show differences regarding the fre-
quency of this word order. Therefore, Christoffersen (2002: 185) discusses geographi-
cally motivated differences between Old West Norse (Norwegian and Icelandic) and
Old East Norse (Old Danish and Old Swedish).
The following results shown in table 4 were derived from the investigation of the
Old Swedish corpus:

Table 4.  Position of finite verb (first or second) in main clauses after och (‘and’)

sentences sentences with V1 after initial V2 after initial


initial och och och

Revelation 1526 840 403 (48%) 9 (2%) 394 (98%)


Revelation 1541 848 458 (54%) 13 (3%) 445 (97%)
Kalmar 502 95 (19%) 23 (24%) 72 (76%)
Charlemagne 864 77 (9%) 26 (34%) 51 (66%)
Luke 1541 2302 550 (24%) 24 (4%) 526 (96%)
Mark 1526 1354 366 (27%) 42 (11%) 328 (89%)

In addition to this the word order in main clauses after the conjunction men/æn (‘but’)
was investigated.14 Eventhough it has to be stated that the frequency of occurrences of
this conjunction varied greatly across the different texts, not a single case of a finite
verb in first position could be found. A comparison with non-conjunctional main
clauses could not be performed at this stage as it would require manual tagging of all
finite verbs in the corpus. Hence, this has to be reserved for a later investigation.
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

A very interesting finding which can be seen in table 4 is that none of our texts
show the initial position of the finite verb after the conjunction och (‘and’) to be the
most frequent word order. This supports Christoffersen’s claim that East Norse (Swed-
ish in this case) texts exhibit differences compared to West Norse (Icelandic and Nor-
wegian) texts. Moreover this highlights the necessity of a precise quantification of the
different word order patterns. Counting the absolute number of main clauses after och
(‘and’) with verb initial word order is only of very limited use seen as the sentences in
the bible texts in our corpus begin with och in up to 50% percent of the cases. In the
story of Charlemagne, however, this only holds true for about 10% of the sentences. At
the same time this text reveals the highest frequency of verb initial main clauses after
och (34%).
In order to confirm the hypothesis indicated in this pilot study, namely that in och-
initiated main clauses a development can be shown from verb-first to verb-second
structures, however, this investigation would have to be extended to a more repre-
sentative corpus.

4.4 Method and technology

4.4.1 Creation
To test our hypothesis we investigated the word order in conjunctional main clauses
taken from a corpus of texts from the Old Swedish period.15 A digital corpus providing
all relevant syntactic information (case, finiteness etc.) would have been the best pre-
requisite for such a test. However, such a complete part-of-speech-tagging (POS) can
not be done automatically for Old Swedish texts. Computer software was developed
for tagging languages such as English or German, but not for Old Swedish because for
the latter an application e.g. in modules of text processing programs (hyphenation,
grammar check) is commercially unattractive.
Manually tagging the whole corpus would go far beyond the available temporal and
personnel resources. In cooperation with the other participating researchers we devel-
oped methods for a preparation of our corpus, enabling us to voice representative state-
ments about special syntactic structures and make diachronic comparisons between the
different stages of Swedish, but also between Swedish and possible contact languages.

4.4.2 Processing
The texts in the Old Swedish corpus do have punctuation. However, in contrast to the
contemporary written language this punctuation was not aimed at a syntactical struc-
turing of the texts, but rather at representing shorter or longer breaks of its spoken
equivalent. In contrast to contemporary text corpora in which sentence boundaries can
be identified by full stops, question marks or exclamation marks, the preparation of the
Old Swedish texts required a rather high amount of time consuming manual work.
One of the texts (the story of Charlemagne) was taken from the fornsvenska text-
banken (Old Swedish text bank),16 two other texts (the Gospel according to Luke and
Studying connectivity with the help of computer-readable corpora 

the Kalmar municipal records) exist in editions suitable for optical character recogni-
tion, but the remaining three texts had to be typed in from facsimile reproductions. In
order to obtain a structural basis for a grammatical analysis those texts existing as
Microsoft Word texts had to be divided into sentences.
This segmentation was mainly based on the identification of finite verbs. However,
the final classification of syntactical units as separate sentences or clauses had to be
based on the habits of contemporary written language, causing some variation seen as
different annotators participated in the work. Due to the less unambiguous formal
marking of clause types in Old Swedish a sentence like
(3) Uerdher kyrkia brutin oc mæſſu fat ſtolen. þet er niþingſværk.
is church broken and mass utensils stolen that is act.of.sacrilege
can be analyzed either as two main clauses or as a conditional clause and a main clause:
‘A church is broken into and the mass utensils are stolen. This is an act of sacrilege’ or
‘It is an act of sacrilege if a church is broken into and the mass utensils are stolen’.17
In a next step the texts were converted to an XML format based on the TEI-com-
pliant menota-standard.18 Using the Z2-tagger (cf. figure 6 below), the XML-texts were
checked for occurrences of main clause-initial conjunctions. This search was based on
word lists containing all the words searched for in all graphic variants existing in the
corpus. The graphic variants had been taken from word lists created for each text with
the software WordSmithTools.19
The need to quickly and efficiently identify and mark occurrences of main clause-
initial conjunctions and other phenomena led to the development of the Z2-Tagger.20
It was first used to generate TEI-conformant XML-files by tokenizing text files with the
help of finite state machines. Since the segmentation into sentences had already been
done manually (cf. above), only words and punctuation marks were identified in this
step. The tagger was then used to manually assign attributes to the identified words.
The tagger localizes words from a given wordlist, matches them against an XPath-
expression, displays them with their surrounding and offers to assign values from a given
set to attributes by clicking on the appropriate buttons or pressing keyboard shortcuts.
The example in the screenshot in figure 6 shows an occurrence of Och, tagged with
pos=’xCC’ (the tag for coordinating conjunctions) and the surrounding text. In this
case the wordlist consists only of one form, [Oo]ch, which is the regular expression for
either Och or och. All options and the wordlist can be changed while tagging and saved
for later use.
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

Figure 6.  Screenshot of the tagging of the conjunction och with the Z2tagger

Thus, for every occurrence of the entries from the applied wordlist it was possible to
verify the item as a conjunction on the basis of the given context. This was necessary
to exclude occurrences of och (‘and’) connecting constituents and not sentences, but
also to separate the conjunction och from graphic variants of the adverb ock (‘also’).
Every form identified as a main clause-initial conjunction was tagged in order to
enable a later search and quantification. Then all finite verbs that followed the already
tagged conjunctions were manually tagged with the help of an XML-editor (<oxygen/>
XML Editor 4.2). The conjunctions could very easily be identified using the search-
routine included in this software.21
With the help of different style sheets it was possible to display all sentences from
the thus tagged corpus which exhibit a special word order pattern in a browser, to save
them in a text file and finally to carry out automatic counts of the different patterns.
For example, one of the style sheets first identifies all sentences with initial con-
junctions. It then displays them in different colours, distinguishing between the ones
followed directly by a verb and those with the verb located in a later position.
Another style sheet does the same thing, counting the relevant results and pre-
senting the numbers for further analysis. On accurately tagged texts, style sheets like
these are easily written, establishing the possibility of all kinds of empirical analysis.
Since the actual texts and the style sheets are based on standardized XML-technology,
they can be used on almost every platform without specialized tools.
Studying connectivity with the help of computer-readable corpora 

Figure 7.  Position of the finite verb after sentence initial conjunctions identified with a
style sheet

5. Studying connectivity in a corpus of spoken language

5.1 Research question: differences of functional profile and procedural com-


bination in the use of o zaman by Turkish monolinguals and Turkish-German
bilinguals

The following is an exploration of ways to contrast the usage of the Turkish expression
o zaman ‘at that time’ as a means to achieve discourse coordination, in bilingual as op-
posed to monolingual child Turkish. The term ‘discourse coordination’ is understood
as an integration of utterances at an inter-utterance level (Rehbein 1995). The study is
based on the following three pre-assumptions: first, coordinating conjunctions cross-
linguistically develop from linguistic expressions that primarily fulfil other functions
(Mithun 1988, 2003). Second, Turkish, typologically seen, lacks the prototypical, Indo-
European, construction-initial conjunctions as a homogeneous word class and uses
other, more diverse, means of coordination instead.22 Third, while coordination takes
place at several syntactic levels, it is at the level of discourse coordination (cf. Rehbein
1995) where, possibly due to a lesser degree of grammaticization, innovations and
functional expansions occur. Therefore, it is in this area that one will find linguistic
elements of various kinds taking over coordinating functions.
Preliminary research on the basis of the project data has shown that the bilingual
children tend to functionalize knowledge-marking and planning-signalizing linguistic
elements for the purposes of discourse coordination to a lesser degree than the mono-
linguals do. They do, however, use more deictic, especially temporal-deictic, elements
at this level. When comparing the two groups it is above all the usage of the expression
o zaman ‘at that time, in that case, then’ which increases, namely from 1.7% of all dis-
course-connective means used in the monolingual data to 12.8% in the bilingual data
(cf. Herkenrath, Özdil & Rehbein, in preparation).
It can be assumed that the expression o zaman contains a deictic procedure caus-
ing a refocusing, anadeictic effect. This effect can unfold in different ways, thus ac-
counting for at least two different usages of o zaman correlatable with different groups
of speakers. As both a first qualitative interpretation of individual occurrences and the
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

observation of functional expansions in deictic elements in other areas suggest, the


bilingual usage of o zaman should reveal a different functional profile and as a conse-
quence a different combination of procedures than that of the monolinguals.23 In order
to test this assumption, contrastive functional profiles of the different usages are to be
established, taking into account both their different functions with respect to the
structuring of the narrative or discourse and their specifically deictic effects.24
The research question calls for a suitable analytical method and tool in order to
allow for a quantitative testing of the assumptions introduced above. With respect to
the data analyzed in Sections 2 and 4, there is a common interest in the emergence of
innovation in contact situations, specifically in the area of discourse coordination.
While in the present contribution innovation is studied from the point of view of data
of spoken language, we are interested in comparing its findings with those emerging in
corpora of a different kind, namely modern and historical written data, such as those
analysed in Sections 2 and 4, respectively. In his context, one distant aim is to see to
which extent one can draw conclusions (or formulate hypotheses) about the interac-
tion of spoken versus written modes of language with respect to structural innovation
and diachronic change, even with only a selection of samples (spoken Turkish versus
written English, German, and Old Swedish) being studied.

5.2 Corpus: oral story retellings by monolingual and bilingual children

For the purposes of a primary application of the search method, a subcorpus compris-
ing 18 Turkish and German discourses was chosen. This subcorpus is based on the
evocative field experiment (Evokatives Feldexperiment) EFE 04 Küçük Köstebek/Der
kleine Maulwurf ‘The little mole’, consisting of a short silent cartoon that the children
watch and reproduce in words for someone else. This subcorpus contains nine Ger-
man and nine Turkish discourses with a total of 15849 words (6780 words in the Turk-
ish and 9069 words in the German-based discourses). One German discourse is with
monolingual and eight with bilingual children (1172 against 7897 words); among the
Turkish discourses, three involve monolingual and six involve bilingual children (1732
and 5048 words respectively). The evocative design chosen requires that the children
verbalize units of visual information, thereby creating coherence and connectivity ir-
respective of any previous verbalization, which they might otherwise rely on as a mod-
el. This method thus offers a good opportunity to study their spontaneous choice of
discourse-coordinating elements.

5.3 Results

The occurrences of o zaman can be attributed to individual speakers, who were di-
vided into three groups for the present purposes: bilingual children, monolingual chil-
dren, and adults. The obtained occurrences of o zaman are split up among the three
Studying connectivity with the help of computer-readable corpora 

groups as shown in table 5: with the one exception of a monolingual child using the
expression o zaman, all the findings are can be attributed either to bilingual children
(47 occurrences) or to adult interactants (16 occurrences). This seems to be quite a
clear picture, even when taking into consideration that the Turkish-based part of the
subcorpus comprises more bilingual than monolingual data (see above paragraph).

Table 5.  Occurrences of o zaman divided into speaker groups in absolute figures

Occurrences

bilingual children 47
monolingual children 1
adults 16

In a next step, the occurrences were categorized into different types, taking into ac-
count their communicative functions as well as the demonstration space to which the
inherent deictic procedure points. With respect to this categorization, we receive the
following results as displayed in table 6:

Table 6.  Types of usage of o zaman, in absolute figures

Type Communicative Demonstration space Procedure Amount


function

0 – – – – – – – (aborted utterance) – – – – – – – 3


1 refocuses on a state of action and discourse deictic, 15
facts at level of action space symbolic
and discourse space
2 refocuses on a previous verbalized units, ? 44
stretch of narration, narration space
announcing progression
3 refocuses on one single knowledge based on deictic, 1
point of time in self- idiosyncratic experi- symbolic
experienced past ence, action space
4 refocuses on an knowledge based on deictic, 1
extended stretch of idiosyncratic experi- symbolic, operative
time in self-experi- ence, action space
enced past

Disregarding the three aborted utterances (type 0) and the one occurrence of type 4,
the findings can be divided into three communicative categories. Type 1 refocuses on
some state of facts at the level of the action and discourse space, as a condition for
further actions. These usages often concern the negotiation of the communicative situ-
ation (e.g. who is to narrate what and when). Type 2 refocuses on some part of the
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

preceding narration, at the same time announcing a progression. Type 3 refocuses on


a point of time in a self-experienced past. The demonstration space of the deictic pro-
cedure can be identified as the action and discourse space in the case of type 1, the
narration space or verbalized units therein in type 2, and finally action space-related
knowledge based on idiosyncratic experience, in type 3. When correlated with the
three different groups of speakers, the types of usage are attributed to the speakers as
shown in table 7: the bilingual children as a group seem to show a tendency towards
type 2, whereas the adult speakers seem to prefer type 1. The monolingual children
obviously do not use type 1 or 2 at all, as the sole monolingual occurrence of o zaman
in the entire subcorpus is of type 3. This finding seems to corroborate native speaker
intuition, but it still has to be investigated whether this pattern remains consistent
when a larger amount of data is analyzed.

Table 7.  Usage types related to the groups of speakers (absolute figures)

Bilingual Monolingual Adults Total


children children

Type 1 2 - 13 15
Type 2 43 - 1 44
Type 3 - 1 - 1

These figures seem to suggest that one can indeed speak of an innovative usage of the
originally temporal-deictic expression o zaman in the bilingual children’s data, a usage
that refocuses on a verbalized unit in the narration, taking it as a starting point from
which further stretches of verbalization are announced. Such a usage can neither be
found in the monolingual children’s nor in the adults’ data. It seems to suggest a func-
tional shift at the level of the procedural quality as well as at the level of the demonstra-
tion space.25 At the same time, the findings, even though acquired in a different way,
seem to correspond to those obtained in Section 2 in that they suggest a functional
shift concealed behind an apparently unchanged surface.

5.4 Method and technology

5.4.1 Corpus creation


The creation of a corpus of spoken language is a complex affair. The data originally ap-
pears in the form of audio-taped conversation consisting of stretches of monologic,
dialogic, and simultaneous speech. In order to obtain a computer-readable corpus, this
audio-taped data has to be manually transcribed, a process which takes time and re-
quires theoretical reflection at various levels (see Rehbein et al. 1993, Rehbein et
al. 2004). To the extent that spoken discourse involves several interactants, the linear-
ity characteristic of written language corpora, such as the ones analyzed in Sections 2
Studying connectivity with the help of computer-readable corpora 

and 4, has to be abandoned so that even at a minimal level of complexity, is it required


that the digital structure represent the simultaneity of linguistic actions in multi-inter-
actant events. The partiture editor EXMARaLDA offers a tier-based structure in which
the complexity of spoken language can be retained and any given linguistic phenom-
enon can be located, viewed, and analyzed within the constellation in which it origi-
nally occurred (see Schmidt & Wörner 2005).

5.4.2 Corpus analysis


The SKOBI-specific aim of the cooperation was to work out a method suitable for its own
complex analytical needs. These needs were threefold: first, it was necessary to be able to
quickly obtain an overview of all the occurrences of o zaman in a selected and prepared
subcorpus. Second, the format should allow for a recontextualization of any occurrence
in its original transcript at any given time. Finally, it should allow for an ad hoc indexa-
tion, ordering, quantification and correlation at the level of the individual occurrence
according to flexible criteria, at the same time allowing for a fast quantitative testing of a
diversity of analytical categories. The method to be developed needed to combine a fast
search for all occurrences of any expression with an accessibility of the original discourse
constellation in its full complexity, as well as a tool for indexation.
An automatized search along these lines had so far only been applicable to digital
corpora with a linear structure. For non-linear data, the only possibility had been to
transform them into linear formats, such as utterance lists, and then then apply an au-
tomatic search tool. This had the disadvantage that, once obtained, the findings would
have to be viewed out of context. One of the most important desiderata in this area is a
concordance tool applicable to spoken-language corpora with a complex nonlinear
structure and allowing for an immediate switching between an ordered list of findings
on the one hand and these same findings as they occur in their original non-linear
constellation in the discourse, on the other. Such a tool had so far not been available.
The search tool Zecke (Schmidt & Wörner 2005) allows for an immediate search
for a given linguistic expression in an entire EXMARaLDA corpus consisting of any
number of files. The search result is presented in the upper part of the window in the
form of a table listing the individual occurrences plus a surrounding context of flexible
size (see figure 8). In the lower part of the window, the programme also allows for a
direct access to the findings in their original transcripts so that speaker constellations,
simultaneity relations, and information concerning the discourse and the individual
speakers can be viewed at any time.
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

Figure 8.  Zecke window with KWIC-Concordance (upper part) and Score representation
(lower part)

The obtained list can be saved in various formats, such as an Excel file (see figure 9), in
which new columns for further information and categorization can be added. This
format also makes it possible to group the occurrences according to whatever catego-
rizations one may have chosen (such as type of usage, in the presented case) and to test
for possible co-occurrences with other categories (such as group of speakers).

Figure 9.  Excel window with analytical categories added and list reordered according to
types of usage

Thus reordered, the table then visualizes the overwhelming co-occurrence of type 1
usages with members of the group of adult speakers on the one hand and type 2 us-
ages with members the group of bilingual children on the other.
Studying connectivity with the help of computer-readable corpora 

6. Conclusion

The individual investigations of coordinating elements described in this paper set out
from the results of qualitative text, discourse and syntax analyses which all seemed to
point to variation in the context of macrosyntactic conjunctions in written and spoken
language use in situations of language contact. By applying a simple, unified method of
querying larger computerized data bases for the occurrences of English And, German
Und, Old Swedish och and Turkish o zaman we were able to confirm the hypotheses
derived from the qualitative case studies, to find commonalities between the individual
findings and, finally, to form an explanatory hypothesis as to what lies behind the func-
tional diversification of the coordinating expressions. However, due to the pilot charac-
ter of the investigations of o zaman and och in the Turkish and Old Swedish corpora
respectively, for the time being this interpretation will remain a tentative one.26
The findings can be summarized as follows: First, we found evidence for an in-
crease in the use of Und in sentence-initial position in German translations. This was
interpreted as a dissociation of the text conventions determining the use of the stylisti-
cally marked option of Und as a means of textual cohesion in German translations
from those operative in German original texts. Judging from the frequency of occur-
rence of Und in German translations it seems that the option of using Und as a mac-
rosyntactic conjunction in written texts has become more readily available. Secondly,
the investigation of the corpus of Old Swedish bible texts and chronicles revealed word
order variation after och in sentence-initial position. This finding corroborates earlier
assumptions suggesting geographically based differences between Old West Norse and
Old East Norse. Thirdly, it was also possible to substantiate the claim of an innovative
discourse-coordinative use of the Turkish temporal-deictic expression o zaman re-
stricted to German-Turkish bilingual children.
The analyses also revealed that these coordinating elements – even though they
occur in the same syntactic position – are not always associated with the same context
phenomena. In other words, English texts, German texts and German translations
only partially overlap with respect to the triggering of subject, tense and modality
switches across sentences conjoined by And and Und. Likewise, sentence-initial och
does not always trigger narrative inversion. Apparently, variation in word order after
och in Old Swedish varies at least across text types (e.g. biblical, biographic, adminis-
trative). But since we also find word order variation after och in different parts of the
same text (cf. the books of the 1526 and 1541 versions of the Bible in table 4 above) it
is maybe just as likely that this differentiation is related to the particular role which the
sentences connected by och play with respect to the overall information organization.
Finally, German-Turkish bilingual children, Turkish monolingual children and Turk-
ish adults seem to employ the expression o zaman for different linguistic purposes.
Among these, the innovative usage of the expression involves a function of refocusing
on a unit at the level of the verbalization, as opposed to the level of the present or a past
action and/or discourse space. This innovative usage, which mainly indicates a pro-
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

gression in the narration, is almost exclusively restricted to the speaker group of Turk-
ish-German bilingual children.
Our findings lead us to hypothesize that macrosyntactic (or discourse) coordina-
tion could indeed be a level at which functional innovation or diversification of coor-
dinating elements takes place. This functional diversification appears most clearly in
texts and discourses from situations of direct language contact – i.e. in the present case
German-Turkish bilinguals’ speech and translational German. The analyses of And/
Und and o zaman suggest that this functional diversification is associated with con-
verging preferences for patterns of information organization/distribution in situations
of language contact between English and German, and Turkish and German, respec-
tively. Even though word order variation after och has not been correlated with lan-
guage contact, the Old Swedish data nevertheless allow the hypothesis that it might be
conventions of information structuring in different text types or particular stretches
within texts which license or motivate narrative inversion after och.
From a methodological point of view, these findings confirm that corpus linguis-
tics, when seen as a method and not as a linguistic subdiscipline in its own right, holds
the potential for a common perspective on highly different objects of investigation.
The step from a qualitative analysis of a few examples to a quantification of a whole
corpus not only increases the validity of the single study, it can also serve as a means to
make the results of different studies more readily comparable. The circumstances un-
der which these comparable results are achieved, however, still differ greatly for mod-
ern, written, for spoken and for historical corpora. As we have argued, this is initially
due to a theoretical issue, namely the model-like relationship between a language sam-
ple and its digital representation in a corpus, which is much more pronounced in cor-
pora of historical or spoken language than in corpora of modern, written language.
More relevant for the practical work and hence for the actual outcome of a corpus-
linguistic study, though, is the fact that this theoretical issue entails far-reaching prac-
tical consequences with respect to the technological formats and tools that a linguist
requires for corpus processing and analysis. The availability (or lack of availability, re-
spectively) of ready-to-use software determines how large a corpus can become within
a given amount of time, how well such a corpus can be queried in search of a certain
phenomenon and how easily the result of such a query can be analyzed in its original
context. Ultimately, the tools will thus have a decisive impact on the scientific insight
itself.27 In order to tap the full potential of corpus linguistics as a common method, it
is therefore necessary to develop technological frameworks for the study of historical
and spoken corpora that are comparable in functionality and ease-of-use to those
available for modern, written corpora. With the tagging and style sheet methods dem-
onstrated in section 4 and the query tool described in section 5, this paper has tried to
give an idea of what such a technological framework should contain.
Studying connectivity with the help of computer-readable corpora 

Notes

1. This investigation was carried out at the Research Center on Multilingualism at the Univer-
sity of Hamburg. The participating projects are: ‘Mehrsprachige Datenbank’ (Multiligual Data-
base), ‘Verdecktes Übersetzen – Covert Translation’, ‘Skandinavische Syntax im mehrsprachigen
Kontext’ (Scandinavian syntax in multilingual contexts) and ‘Sprachliche Konnektivität bei bi-
lingual türkisch-deutsch aufwachsenden Kindern’ (Linguistic connectivity in bilingual Turkish-
German children (SKOBI)). The Center is funded by the Deutsche Forschungsgemeinschaft
(DFG).
2. Another way of categorizing the differences is that in German, a ‘transactional’ style focu-
sing on the content of the message is frequently preferred, whereas in English, speakers tend to
prefer an ‘interactional’, addressee-focused style.
3. The descriptions of sentence- and utterance-intitial Und are on the whole much more scarce.
4. The corpus contains translations from English into German for the genre of popular scien-
tific texts only, and translations from German into English exclusively for the genre of external
business communication. This is due to the fact that the two areas of text production are appa-
rently determined by different language policies. In the area of popular scientific journalism
there are virtually no translations from German into English. Conversely, in the area of interna-
tional business communication, translations from English into other languages, and in particu-
lar into German, are increasingly rarely produced. German companies, however, generally pro-
vide English translations of their annual reports and mission statements.
5. A more detailed interpretation of the results can be found in Baumgarten (to appear).
6. A similar observation is made in Givón (1990).
7. The table only presents the synchronic view of the time frame 1999–2002. The total of oc-
currences in the time frame 1978–1982 is too small to render reliable results for proportions.
8. The TnT-tagger is available at http://www.coli.uni-saarland.de/~thorsten/tnt/. For informa-
tion on the tagsets see: http://www.grsampson.net/SueDoc.html and http://www.sfs.nphil.uni-
tuebingen.de/Elwis/stts/stts.html
9. For example ParaConc and MonoConc (http://www.athel.com/mono.html; Barlow 2001;
Reppen 2001).
10. In the work with corpora of modern, written language, the difference between sample and
corpus is therefore often ignored totally, for instance in Sinclair’s (1991) definition of a corpus as
„ a collection of naturally occurring language text, chosen to characterize a state or variety of a
language”.
11. For details on EXMARaLDA see Schmidt, Wörner (2005)
12. Theoretically the syntactic analysis was based on Paul Diderichsen’s (1971) topological mo-
del developed in the context of the Copenhagen school after 1935. Diderichsen’s description of
the linear order of constituents in a topological scheme is based on the relation between finite
verb and subject (nexus). The investigation is aimed at quantifying different word order patterns
in main clauses that can be described on the basis of this model, concentrating on the area
between sentence initial conjunction and finite verb (fundamentfield). For a detailed description
of Diderichsen’s topological model cf. Diderichsen (1976).
13. Cf. Christoffersen (2003: 17).
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

14. Besides och (‘and’) this is the most frequent conjunction in our corpus.
15. Of course the representativeness of this subcorpus consisting of approximately 70.000
words which is equivalent to about 10% of the overall corpus of the project is limited; yet we
tried to avoid general conclusions based on structures that are typical only for a special register
and to prevent interferences from a different language by choosing texts from different text types
and by including translations from different source languages and also originally Swedish texts.
16. http://www.nordlund.lu.se/Fornsvenska/Fsv%20Folder/ [20.4.2005], edited by Lars-Olof
Delsing, Lund.
17. Example from Äldre Västgötalagen (The Older Law of Västergötland, 1296), cf. Collin/
Schlyter (1827): 5.
18. Cf. http://gandalf.aksis.uib.no/menota/.
19. Michael Scott: Word Smith 4.0.0.165. Oxford University Press.
20. The Z2-Tagger is available as a free download from http://www.exmaralda.org/z2tagger.
21. Lüdeling/Poschenrieder/Faulstich (2004) describe an approach of annotating a historical
German text corpus that is also semi-automatic but uses different kinds of annotations in diffe-
rent text levels that are aligned with each other. The menota-standard that is utilized here, howe-
ver, is aimed at encoding all information in one document and providing the user with style
sheets in order to enable him/her to display the amount of information she/he wishes.
22. For a thorough syntactic and semantic classification of different kinds of connectors (inclu-
ding conjunctions, sentence adverbials, as well as particles) in German see Pasch et al. (2003).
23. See Rehbein (to appear) as well as (2001) for a more thorough analysis of refocusing proce-
dures in Turkish-German bilingual children’s narratives.
24. The authors are grateful to Kristin Bührig for a valuable reminder of the category of de-
monstration space (Verweisraum), as well as to Jochen Rehbein, Shinichi Kameyama and Rainer
von Kügelgen for useful suggestions concerning a possible de-deictification (p.c.).
25. Thanks to Kristin Bührig (p.c.) for the idea of a linkage between these two at this level of
innovation.
26. At the moment the ENDFAS/SKOBI corpus is in the process of being converted into the
EXMARaLDA format. After that, the tools and methods tested in this study will be applicable to
the entire data. It will then be possible to study occurrences of o zaman in a corpus of more than
200 transcribed discourses.
27. Orlandi (2002) puts this as follows: “[Some] colleagues refer to the computer as “just a tool”
or “simply a bunch of techniques”, as if ways of knowing did not have much to do with what is
known. Because the computer is a meta-instrument – a means of constructing virtual instru-
ments or models of knowing – we need to understand the effects of modelling on the work we
do as humanists.”

References

Barlow, M. 2001. ParaConc: Concordance software for multilingual parallel corpora. http://www.
athel.com/paraweb.pdf
Studying connectivity with the help of computer-readable corpora 

Baumgarten, N. To appear. Converging conventions? Macrosyntactic conjunction with English


and and German und. In Text and Talk.
Biber, D., Johansson, S., Leech, G., Conrad, S. and Finegan, E. 1999. Grammar of Spoken and
Written English. London: Longman.
Bird, S. and Liberman, M. 2001. A formal framework for linguistic annotation. Speech Commu-
nication 33: 23–60.
Blakemore, D. and Carston, R. (1999). The Pragmatics of and-Conjunctions: The non-narrative
cases [UCL Working Papers in Linguistics 11]. London: University College London.
Byrnes, H. 1986. Interactional style in German and English conversations. Text 6: 89–106.
Christoffersen, M. 1993. Setning og sammenheng. Syntaktiske studier i Magnus Lagabøters land-
slov [Adh-serien 65]. Kristiansand: Agder distrikthøgskole.
Christoffersen, M. 2002. Nordic language history and research on word order. In The Nordic Lan-
guages. An international handbook of the history of the North Germanic languages [Hand-
bücher zur Sprach- und Kommunikationswissenschaft 22:], O. Bandle et al. (eds), 182–191.
Berlin: Walter de Gruyter.
Christoffersen, M. 2003. Noe om bindeord i eldre norsk, særlig om ok, en og nema. In Språk i en-
dring. Indre norsk språkhistorie, J. T. Faarlund (ed.), 13–28. Oslo: Novus.
Clyne, M. 1987. Cultural differences in the organization of academic texts. Journal of Pragmatics
11: 101–124.
Clyne, M. 1994. Intercultural Communication at Work: Cultural values in discourse. Cambridge: CUP.
Collin, D. H. S. and Schlyter, D. C. J. (eds). 1827. Westgöta=Lagen. Stockholm: Haeggström.
DeRose, S., Durand, D., Mylonas, E. and Renear, A. 1990. What is Text, Really? Journal of Com-
puting in Higher Education 1(2): 3–26.
Diderichsen, P. 1976. Logische und topische Gliederung des germanischen Satzes. In Ganzheit
und Struktur. Ausgewählte sprachwissenschaftliche Abhandlungen [Internationale Bibliothek
für allgemeine Linguistik 30], P. Diderichsen (ed.), 44–58. Munich: Wilhelm Fink Verlag
Doherty, M. 1996. Passive perspectives: Different preferences in English and German – A result of
parameterized processing. Linguistics 34(3): 591–643.
Doherty, M. 2002. Language Processing in Discourse: A key to felicitous translation. London:
Routledge.
Ehlich, K. 2001. Konsekutives ‘und’ – mündliches Erzählmerkmal. Lecture given at the University
of Hamburg, Research Center on Multilingualism.
Fabricius-Hansen, C. 1999. Information packaging and translation: Aspects of translational sen-
tence splitting (German-English/Norwegian). In Sprachspezifische Aspekte der Informations-
verteilung, M. Doherty (ed.), 175–214. Berlin: Akademie Verlag.
Givón, T. 1990. Syntax: A functional-typological introduction. Amsterdam: John Benjamins.
Halldór Á. Sigurđsson. 1990. V1 declaratives and verb raising in Icelandic. In Modern Icelandic
Syntax [Syntax and Semantics 24], J. Maling and A. Zaenen (eds), 41–69. San Diego CA:
Academic Press
Halliday, M. A. K. and Hasan, R. 1976. Cohesion in English. London: Longman.
Herkenrath, A., Özdil, E. and Rehbein, J. In preparation. Code-switching as a discourse phenom-
enon: Evidence from Turkish-German bilingual children’s speech. University of Hamburg,
SFB 538 Mehrsprachigkeit: SKOBI Working Paper, Ms.
House, J. 1989. Politeness in English and German: The functions of please and bitte. In Cross-Cul-
tural Pragmatics: Requests and apologies, S. Blum-Kulka, J. House and G. Kasper (eds),
96–122. Norwood NJ: Ablex.
 Nicole Baumgarten, Annette Herkenrath, Thomas Schmidt, Kai Wörner and Ludger Zeevaert

House, J. 1996. Contrastive discourse analysis and misunderstanding: The case of German and
English. In Contrastive Sociolinguistics, M. Hellinger and U. Ammon (eds), 345–361. Berlin:
Mouton de Gruyter.
House, J. 2004. Linguistic aspects of the translation of children’s books. In Translation – Übersetzung
– Traduction / Ein internationales Handbuch zur Übersetzungsforschung / An international
handbook of translation studies /Encyclopédie internationale de la recherche sur la traduction
[HSK 26:1], H. v. Kittel, A. P. Frank, N. Greiner, T. Hermans, W. Koller, J. Lambert, F. Paul, in
cooperation with J. House and B. Schultze (eds), 683–697. Berlin: Walter de Gruyter.
Kotthoff, H. 1989. Pro und Kontra in der Fremdsprache. Frankfurt: Peter Lang.
Leech, G. 1992. Corpora and theories of linguistic performance. In Directions in Corpus Linguis-
tics, J. Svartvik (ed.), 105–122. Berlin: Mouton de Gruyter.
Lüdeling, A, Poschenrieder, T. and Faulstich, L. (2004). DeutschDiachronDigital – Ein dia-
chrones Corpus des Deutschen. In Jahrbuch für Computerphilologie 6, G. Braungart, K. Eibl
and F. Jannidis (eds), 119–136. Paderborn: Mentis.
Mithun, M. 1988. The grammaticisation of coordination. In Clause Combining in Grammar and
Discourse, J. Haiman and S.A. Thompson (eds), 331–359. Amsterdam: John Benjamins.
Mithun, M. 2003. Functional perspectives on syntactic change. In The Handbook of Historical
Linguistics, B. D. Joseph and R. D. Janda (eds), 552–572. Oxford: Blackwell.
Nyström, I. 1985. Studier i äldre nysvensk syntax II. Ledstruktur och ledföljd i bisatser. Helsinki:
Yliopistopaino.
Ochs, E. 1979. Transcription as theory. In Developmental Pragmatics, E. Ochs and B. Schieffelin
(eds), 43–72. New York NY: Academic Press.
O’Donnell, M. 2004. Systemic Coder Software. http://www.wagsoft.com
Orlandi, T. 2002. Is humanities computing a discipline? In Jahrbuch für Computerphilologie 4, G.
Braungart, K. Eibl, and F. Jannidis (eds), 51–58.. Paderborn: Mentis.
Pasch, R., Brauße, U., Breindl, E. and Waßner, E. 2003. Handbuch der deutschen Konnektoren.
Linguistische Grundlagen der Beschreibung und syntaktische Merkmale der deutschen Satz-
verknüpfer (Konjunktionen, Satzadverbien und Partikeln) [Schriften des Instituts für Deut-
sche Sprache 9]. Berlin: Walter de Gruyter
Platzack, C. 1985. Narrative inversion in Old Icelandic. Íslenskt mál og almenn málfræði 7:
127–144.
Quirk, R., Greenbaum, S., Leech, G. and Svartvik, J. 1985. A Comprehensive Grammar of the
English Language. London: Longman.
Redeker, G. 1990. Ideational and pragmatic markers of discourse structure. Journal of Pragmatics
14: 367–381.
Rehbein, J. 1995. Aber, also, und, auch – Prozeduren der Diskurskoordinierung. Ms. Deutsche
Fassung von: Aber, also, und, auch – elements of chaining in German discourse. Paper read at
the Workshop on Discourse Markers and the Representation of Text, January 9–11, 1995 in
Egmond aan Zee.
Rehbein, J. 2001. Turkish in European Societies. Lingua e Stile 3: 317–334.
Rehbein, J. To appear. Erzählen in zwei Sprachen – auf Anforderung. In Kinderkommunikation –
einsprachig und mehrsprachig. Mit einer erstmals auf Deutsch publizierten Arbeit von Lev. S
Vygotskij, Zur Frage nach der Mehrsprachigkeit im kindlichen Alter, K. Meng and J. Rehbein
(eds), Münster: Waxmann.
Rehbein, J., Grießhaber, W., Löning, P., Hartung, M. and Bührig, K. 1993. Manual für das com-
putergestützte Transkribieren mit dem Programm SyncWRITER nach dem Verfahren der
Studying connectivity with the help of computer-readable corpora 

Halbinterpretativen Arbeitstranskriptionen (HIAT). Universität Hamburg: Institut für Ger-


manistik I.
Rehbein, J., Schmidt, T., Meyer, B., Watzke, F. and Herkenrath, A. 2004. Handbuch für das com-
putergestützte Transkribieren nach HIAT [Arbeiten zur Mehrsprachigkeit / Working Papers
in Multilingualism, Series B, No 56].
Reppen, R. 2001. Review of MonoConc Pro and Wordsmith tools. Language Learning and Tech-
nology 5(3): 32–36.
Schiffrin, D. 1986. Functions of AND in discourse. Journal of Pragmatics 10: 41–66.
Schiffrin, D. 1987. Discourse Markers. Cambridge: CUP.
Schmidt, T. 2005. Modellbildung und Modellierungsparadigmen in der computergestützten Kor-
pusanalyse. In Sprachtechnologie, mobile Kommunikation und linguistische Ressourcen. Bei-
träge zur GLDV-Tagung 2005 in Bonn, B. Fisseni, H. Schmitz, B. Schröder and P. Wagner
(eds), Frankfurt: Peter Lang.
Schmidt, T. and Woerner, K. 2005. Erstellen und Analysieren von Gesprächskorpora mit EX-
MARaLDA. Gesprächsforschung (Online-Zeitschrift zur verbalen Interaktion) 6. http://www.
gespraechsforschung-ozs.de/heft2005/heft2005.htm.
Sinclair, J. 1995. Corpus, Concordance, Collocation. Oxford: OUP.
Weinrich, H. 2003. Textgrammatik der deutschen Sprache. Hildesheim: Olms.
Wessén, E. 1992. Svensk språkhistoria. III. Grundlinjer till en historisk syntax [Nytryck i nordiska
språk och svenska 6] (1st. ed. 1956). Edsbruk: Akademitryck.
Zifonun, G., Hoffmann, L. and Strecker, B. 1997. Grammatik der deutschen Sprache. Berlin: Wal-
ter de Gruyter.
Discourse coordination in Turkish
monolingual and Turkish-German bilingual
children’s talk: işte*

Annette Herkenrath (Universität Hamburg)

The paper examines systematic uses of the linguistic expression işte in elicited
narrative conversations of Turkish monolingual and Turkish-German bilingual
children as well as of a few exemplary adult speakers. The theoretical framework
is functional-pragmatic. The main hypothesis is that işte, originally functioning
as a discourse particle on a biprocedural basis, with both deictic and incitive
procedures, is on its way to develop into a coordinating expression, working
both above and below utterance level. In this last use, its functional potential
integrates an additional, operative, procedure. The method combines qualitative
analyses of empirical occurrences with a quantitative comparison of usage
types in the different groups of speakers. As a preliminary result, it can be said
that the connective, arguably coordinative, use of işte is a function of narrative
and homileïc discourse competence, prevalent in older children and adult
speakers and generally in monolingual children, whereas it occurs to a lesser
extent in the data of the younger and the bilingual children. While the primary
functions of işte are made use of by all informants and seem to be part of early
acquisition, the realization of its fully expanded functional potential, including
the coordinative usages, seems to be part of later acquisition and can be subject
to loss or delay under the influence of language contact.

1. Goals and questions

The present paper is interested in the workings of one linguistic expression, Turkish
işte1, at the level of utterance -internal and -external connectivity2 in monolingual and
bilingual spoken Turkish. The phenomena will be referred to as ‘discourse coordina-
tion’ in the sense of Rehbein (1995)3, this concept being chosen because it is neutral to
any precategorizations along the lines of traditional terms such as conjunctionhood or
discourse markership. The central idea is that the functional area of discourse coordi-
nation can be realized by linguistic elements derived from a variety of word classes and
serving a multitude of functions.
 Annette Herkenrath

The linguistic expression işte has been chosen for study out of a variety of candidates for
discourse coordination. What makes it an interesting object of study is its wide range of
discourse functions (from knowledge-marking to discourse coordination) and its
markedly differentiated functional profile in the monolingual versus bilingual data. It
will be hypothesized that its connective, i.e. coordinating, function emerges out of the
knowledge-marking and deictic functions that it originally fulfilled and continues to
fulfill alongside a newly developed coordinating one. It seems that both language con-
tact and individual preferences can influence diachronic development in this precise
area (cf. Mithun 1988). Innovations may also take the form of individual changes lead-
ing to a more systematized language change, first taking place in smaller groups or
networks of speakers, as shown in Backus (1996) and Boeschoten (1997: 22). Finally,
some studies (Matras 2000a, 2000b, Maschler 1997, 2000 and others) argue in favour of
a fusion of discourse organization systems in bilingual communication. These proc-
esses are taken to be closely linked to the oral and interactive use of the language in
spoken discourse and to be a function of advanced discourse competence.
The research questions are the following: What are the general discourse-structur-
ing functions of işte? How do they emerge in empirical spoken data and how can the
given occurrences be categorized? How does the knowledge-marking potential of işte
expand into the realm of discourse-coordination? Can the differentiated usages4 be
shown to be a function of the activation and deactivation of specific procedures? And
finally: how do preferences for the different functions of işte distribute over different
speakers or groups of speakers? The goal is diachronic in perspective, in that the func-
tional profile of işte in a situation influenced by language contact will be contrasted
with that in a situation without this contact. The study attempts to relate its findings to
the ongoing discussion about grammaticization (Heine and Kuteva 2005 and others),
taking however a critical view from a functional-pragmatic theoretical framework.
The paper is organized as follows: Section 2 gives an overview of the state of the art
with respect to the relevant terms and concepts, models of grammaticization of coor-
dination, the functional-pragmatic concept of discourse coordination, and Turkologi-
cal studies on işte. Section 3 presents the corpus from a quantitative view. Section 4
presents data displaying işte in generally discourse-structuring functions; section 5
focuses on more specifically coordinating usages. Section 6 theorizes and categorizes
the contrast between the different types of findings in functional-pragmatic terms, and
section 7 and 8 are a quantitative comparison between the monolingual, the bilingual
and some adult data.

2. State of the art

The following is an overview of some analyses of coordinating means that have shaped
the research goals of the present paper. Among these figure the concepts of discourse
particle and discourse markership, the concept of discourse coordination, functional-
Discourse coordination in Turkish monolingual and Turkish-German bilingual children’s talk 

pragmatic studies of coordinating expressions, grammaticization studies as well as


Turkological work in the area concerned.

2.1 Conjunctions, particles, discourse markers, and discourse coordinators in


grammaticization perspectives

Coordination applies at various levels. Utterance-internally it takes place between


small units such as NPs, PPs etc., between propositional contents sharing one illocu-
tion, and between illocutionary phrases of equal status within one utterance. Above
the utterance level it serves to connect utterances (and their illocutions) to other,
mostly preceding, utterances (by the same or a different speaker), and to the discourse
at large, i.e. to patterns, discourse topics, or units of discourse knowledge. Connective
devices may in some languages differ according to the level at which the coordination
applies (Mithun 1988: 336ff, see also Haspelmath 2004, for a recent overview). When
applying at the discourse level, coordinating elements have a referential scope extend-
ing well beyond the boundaries of the utterances to which they originally belong (cf.
Haiman and Thompson 1988: ixf). Kerslake (1996:78) makes an according distinction,
terming ‘conjunctions’ those elements that are involved in utterance-internal coordi-
nation and talking about ‘conjuncts’ when referring to inter-utterance coordination.5
With respect to Turkish, she mentions connectives such as ama ‘but’, ya da ‘or’ and ve
‘and’, which fulfill a double function in that they work as connectives both within and
across utterance boundaries, connecting NPs, adjectives, adverbials on the one hand
and utterances on the other. The question of whether or not Turkish uses conjunctions
has been a matter of controversy. This controversy extends well beyond the area of
coordination to include the realm of subordination as well. Kerslake (p.c., November
19th, 2004) takes the view that Turkish does use conjunctions, arguing for a conjunc-
tion status for elements such as ve ‘and’ or ama ‘but’. Johanson, within his copying
framework (1975, 1996), presents a different view, suggesting a range of formal criteria
for conjunctionhood some of which these elements fail to meet. Rehbein (2006a) ex-
cludes conjunctionhood for the element ki in monolingual Turkish, this being in con-
trast both with its original use in Persian and with the innovations they describe for
their bilingual (Turkish-German) data. The question of conjunctionhood thus needs
to be left suspended for the time being, later to be approached from a different ana-
lytical perspective.
The terms ‘discourse marker’ and ‘discourse particle’, although somewhat differ-
ent, are used synonymously in the present paper. Current definitions of the term ‘dis-
course marker’ (e.g. Schiffrin 1987, 2001: 57) hold that discourse markers are related
to the ongoing discourse or text, operating at different levels or planes of discourse,
such as knowledge and interaction, planning, and textuality/connectivity.6 Another
characteristic is their multifunctionality, with coordination, in this case, being just one
of several discourse-organizing functions they may fulfill (Schiffrin 2001), others be-
ing the marking of a proposition as part of common discourse knowledge, the location
 Annette Herkenrath

of an event in space or time, the marking of evaluation and speaker attitudes, or the
marking of a planning interval. Yılmaz (2004: 51ff) uses the term ‘discourse particle’
(söylem belirleyicisi) for his analysis of Turkish expressions, referring to the general
discussion and highlighting deictic (‘indexical’, cf. Schiffrin 1987) as well as floor-hold-
ing functions. By hypothesis, the way the coordination is achieved in each instance can
be related to this general profile. E.g., if a given element can be observed to be used as
a marker of common discourse knowledge and if it can also be established that it is
used in a coordinating textual function, then it will be assumed that the way the coor-
dination is realized is related to its otherwise functioning as a marker of discourse
knowledge. In other words, the link between the new utterance and the previous dis-
course is at the level of discourse knowledge and this characteristic distinguishes the
coordinating element in question from other expressions which establish links at oth-
er levels, such as the level of discourse planning or the level of the temporal structure
of the narration. It is this special instance of one coordinative function emerging out
of another, more generally discourse-marking one that this paper is interested in.
Mithun (1988, 2003) in her diachronically typological analyses of Northern Iro-
quoian languages (such as Cayuga and Mohawk) describes instances of grammaticiza-
tion of discourse markers into coordinating conjunctions under the influence of lan-
guage contact, involving the colonizing languages English and French, within the last
hundred or two hundred years. Mithun (1988) distinguishes several levels at which
coordination may occur: between noun phrases, between predicates, between clauses,
and at the level of independent utterances, linking these to previous utterances or to
the previous discourse at large (Mithun 1988: 332ff, 347).7 The languages mentioned,
in which coordination originally used to be realized by means of intonation alone,
started developing coordinative conjunctions by way of a functional expansion of ex-
pressions originally used for other purposes, using them to connect ever smaller, low-
er-hierarchy syntactic entities as the grammaticization proceeded, finally turning into
general coordinators operating at all syntactic levels. Schiffrin (1987: 247) in her em-
pirical study of several so-called discourse markers in American English draws a
boundary between a discourse-marking and a conjunction-like usage along similar
lines, defining the utterance boundary as a domain into which a marker cannot oper-
ate and vice versa, the boundary within which coordination can work on syntactically
defined units. This distinction, traditional though it may seem, will have to be taken
up again in the later sections, where the question will have to be asked of whether işte
can at all be shown to operate in relation to any concretely definable units of a mor-
pho-syntactic category. Taking an opposite perspective, several studies argue for an
emergence of all-purpose coordination out of more narrowly syntactic functions,
more specifically from comitative relational nouns or adpositions, which then also
take on clause-combining functions (Stolz 1998, Haspelmath 2004, Heine and Kuteva
2002, 2005: 16, 110 etc.). Haspelmath also offers a series of semantic and syntactic
criteria for coordinatorship, roughly based on a refined notion of symmetry concepts.
Under the terminological heading of ‘replication’, Heine and Kuteva view instances of
Discourse coordination in Turkish monolingual and Turkish-German bilingual children’s talk 

contact-induced change from a wide range of languages, which involve the “transfer of
grammatical information (…) without involving any linguistic forms” (2005:79) and
in which so-called minor use patterns turn into major ones both in terms of frequency
and in terms of conceptual changes, under the influence of a contact language. Heine
and Kuteva view these changes as processes of grammaticization, in the course of
which discourse-related phenomena become structures with a grammatical function,
thus taking a somewhat dichotomizing view on the relationship between discourse
and grammar.
Approaching from a more discourse-based analytical perspective, Matras offers
the idea that discourse management systems might undergo an economically moti-
vated process of fusion in situations of language contact, i.e. multilingualism. This fu-
sion may take the form of one language assuming pragmatic dominance over another,
depending on the discourse constellation (Matras 2000a: 521), such that the language
material for discourse-marking items may be taken from just one language. Matras
contrasts fusion against convergence in terms of the degree of separation between the
systems involved. Fusion then is “the wholesale, class-specific non-separation of two
systems”. Applied to the area of discourse markers, this means that for the speakers of
a given bilingual variety, there remains only one set of discourse markers to opt for.
Matras characterizes fusion as a cognitive process which, like convergence, allows for
a “[reduction] of the processing load on bilinguals” (Matras 2000a: 510ff). Maschler
(1997, 2000 etc.) documenting functions of discourse markers in Hebrew-English
casual talk, argues in favour of a similar process, with discourse markers from both
languages feeding into a unified bilingual system and acquiring additional, more dif-
ferentiated functions, coming to express new categories of communicative contrast.

2.2 Previous Turkological work on işte

Traditional Turkish grammars such as Gencan (1975) have classified expressions such
as işte and others as particles (ilgeç, bağlaç). Others list işte under the heading of gös-
terme belirteci (deictic or demonstrative adverb), thereby ascribing it a deictic function.
In this connection, işte is also analyzed as an intensifier of meaning and of stance (con-
veying such attitudes as sureness, pride, or challenge), the intensification being a qual-
ity of its deictic function (see Şimşek 2003 for a critical overview). In her study on dis-
course connectives, Kerslake (1996) groups işte (together with neyse and sonra) under
the heading of ‘organizational’, and on a more detailed level, as ‘resumptive’. Özbek
(2000), contrasting işte against yani and şey, highlights several distinct functions,
among them a reference to shared or assumed knowledge, a closing of discourse units
or topics, a resumption of previously abandoned topics, as well as a marking of empathy
when stating a point. Özbek places her categorization in the traditional context of a
multifunctionality approach while at the same time assigning işte the core function of a
reference to shared knowledge, which, she argues, is realized in every occurrence.
 Annette Herkenrath

Şimşek (2003), working within a functional-pragmatic framework (see next sub-


section), analyses işte as a discourse particle operating on different kinds of knowledge
and functioning in the communicative areas of summarizing, evaluating, assessing,
and verbalizing results. According to Şimşek’s analysis, işte is a deictic expression serv-
ing to focus on units of knowledge verbalized in a preceding utterance or about to be
verbalized in what is yet to come. İşte consequently has both anadeictic and catadeictic
characteristics. Both the deictic and the knowledge-related characteristics of işte are
directly responsible for its functioning in structuring the discourse. Its knowledge-re-
latedness shows effects of highlighting a unit of verbalized knowledge and of situating
it in relation to the general discourse knowledge built up in the course of an interac-
tion, thereby effecting a restructuring of this knowledge. The general function of işte is
to bring about a change of focus in discourse, something that tends to happen at phase
or pattern boundaries. This change of focus is realized by way of a summary of knowl-
edge verbalized in previous utterances. İşte also has an illocutionary function, which
consists in contextualizing verbalized states of events with respect to the built-up dis-
course knowledge. This contextualization globally relates the new knowledge elements
to larger units of knowledge, verbalized in longer stretches of discourse. According to
the data analysis by Şimşek (2003), işte can fulfill the following functions: (1) pointing
at something instead of verbalizing it, (2) pointing at something for assessment, (3)
concluding some part of discourse, (4) introducing or initiating the verbalization of
something new, and (5) planning and reorganizing.
Yılmaz (2004), also starting from the assumption that işte is a deictic8 expression,
or even a demonstrative pronoun, marks it off against şey and yani, working out both
framing, highlighting, reporting, and connective functions. The connective function
here interestingly applies at an intra-utterance level, between “distant pieces of utter-
ances” (birbirinden uzak sözceleri birleştirme, Yılmaz 2004: iiif). Yılmaz’ analysis docu-
ments the following functions of işte, relating them to the larger functional domains of
conversational structure and content: (1) claiming and initiating a turn, thereby mark-
ing the previous turn as completed and “disengaging from its informational projection”,
there by serving as a disjunction marker “alerting recipients that what follows might not
be related to what preceded” (Yılmaz 2004: 139f, Schegloff 1996), (2) projecting a new
turn as extended and suspending, in cooperation with the other interactants (floorhold-
er function), (3) marking a given information as an exemplification or detail of a previ-
ously established topic, (4) closing a topic, (5) highlighting an information unit, (6)
introducing reported speech, (7) marking a tie-back of information, and (8) as an an-
swer preface, signalizing that a piece of information delivered in answer to a question is
not complete or in some other way dispreferred (in the sense of Levinson 1983, Yılmaz
2004: 55, 166f). These functions are all more or less derived from a basically deictic
character of işte, with işte pointing at relevant interactional circumstances, at units co-
natining propositional information, or at aspects of a knowledge-related constellation.
About the etymology of işte nothing certain seems to be known. As the preceding
paragraphs have shown, Şimşek (2003) and Yılmaz (2004) holistically refer to its deic-
Discourse coordination in Turkish monolingual and Turkish-German bilingual children’s talk 

tic quality without analyzing it into any smaller parts. Fuat Bozkurt (p.c. to Jochen
Rehbein, 22nd December 2005) suggests the following derivation: işte < uş + ta (ta =
locative), uş > şu (by way of metathesis), cf. Eyuboğlu (1991: 355), who however as-
sumes an identity of the first part iş/uş not with şu, but with iş as in işbu/uşbu. Where-
as these latter suggestions seem to lack a thorough documentation, Clauson’s (1972:
254) carefully documented etymology lists a linguistic expression oş as “an exclama-
tion used to call attention to something or someone, ‘look, see here’ (...), not noted as
an independent word after about XVI, but fr[equently] used about XIII fused with the
Demonstrative Pron[noun] ol ‘that’ and bu ‘this’ to add emphasis to them”. By way of
distortion these expressions seem to have turned into şu and işbu in Ottoman Turkish,
and to also have contributed to the Modern Turkish expression şimdi < oş emdi ‘now
at once’, in contrast to imdi ‘now’ in Old Anatolian Turkish. With the help of Clauson’s
documentation, both the Turkic origin of işte and a deictic or incitive9 10, quality of the
morpheme iş- can be retraced. Hacıeminoğlu (1992: 307f) mentions usages of uş in
several historical Turkic varieties, which he parallels to uşda (= uş + da)/uşta/işte in
Western Turkic and finally işte in modern Turkish. The origin of the second part of işte,
-te, however, still seems to be less well documented; assumptions hinting at a locative
case suffix (e.g. Ergin 1993: 351) lack documentation and therefore must be regarded
as hypothetical for the time being. The segmentation into two components does not
seem to be part of the linguistic knowledge of present day speakers, so that its morpho-
logical structure can no longer be regarded as agglutinative, but should rather be per-
ceived as amalgamated, no longer analyzable, opaque.11

2.3 Theoretical framework

From a functional-pragmatic theoretical perspective, what is of interest is the exact


procedural composition of individual coordinating elements, irrespective of traditional
word class membership. The functional-pragmatic notion of ‘procedure’ refers to the
communicative characterization of the smallest parts of a given linguistic element as
the “smallest units of action” mediating between “a formal, an interactional, and a men-
tal side” and “unfold[ing] a specific action dynamic between speaker and hearer” (Reh-
bein 2006b, this volume; see also Rehbein and Kameyama 2004, Ehlich 1991). Proce-
dures belong to one of several linguistic fields12 (symbolic, deictic, operative, incitive
etc.).13 With respect to the present study, the differentiation between ‘deictic’, ‘incitive’
and ‘operative’ procedures will prove to be relevant. Expressions containing deictic pro-
cedures (generally demonstratives of different kinds, i.e. spatial, temporal, and person-
related) are employed by the speaker to directly focus the hearer’s mental attention on
some object or concept, without reference to any symbolic/lexical verbalization thereof.
Incitive expressions (such as imperatives, vocatives, or interjections) serve to interfere
with the actional apparatus of the hearer, without recurrence to propositional verbali-
zations. Operative procedures (for instance phoric expressions, determiners, many in-
stances of verbal morphology, discourse particles, as well as syntactic complementiz-
 Annette Herkenrath

ers) realize a processing of language at a formal and knowledge-categorizing level (see


Rehbein, this volume, for the most recent overview). The crucial difference between
deictic and incitive procedures on the one hand and operative procedures on the other
is that while deictic procedures point at and incitive procedures interfere with aspects
of an extra-linguistic reality, operative procedures are units of language processing oth-
er units of language. Expressions having undergone a change of field are referred to by
means of the prefix ‘para-’; e.g. a formerly deictic or incitive expression having taken on
an operative procedure is referred to as ‘para-operative’ to indicate that its operative
quality is based on a secondary procedure superseding an historically earlier, but not
necessarily inactive, deictic (or incitive) procedure. Processes of linguistic change, in-
stead of being referred to by the term of ‘grammaticization’ or ‘grammaticalization’, is
conceived of in terms of such field transpositions, in consideration of the fact that a
grammatical function, albeit of a different kind, may have existed before as well as after
the change (e.g. Rehbein 2001, Rehbein and Karakoç 2004).
Redder’s (1990) analysis of the German expressions denn ‘then’ and da ‘there’ and
Rehbein’s (1995) analysis of the four German discourse-coordinating elements aber
‘but’, also ‘well’, und ‘and’, and auch ‘also’14 are illustrations of a procedure-analytical
approach applied to connective expressions in German data. Redder (1990) presents a
model for analyzing the inner (procedural) and outer (morpho-syntactic) structure of
the two German connective expressions denn and da. The method followed there is to
systematically present a given expression in the variation of its empirical occurrences
and to work out its contribution to a given linguistic action in a given morphosyntactic
environment, also taking into account the functional etymology (or history of usage,
Verwendungsgeschichte) of the expression. Rehbein (1995) develops the concept of dis-
course coordination as a way of establishing a relation between discourse knowledge
built up until a certain moment and the knowledge to be subsequently added. The
coordinating effect is conceived of as unfolding against the background of a commu-
nicative constellation involving mental processes on both the speaker and hearer sides
and a domain of common discourse knowledge shared by both participants, as well as
their verbalization. It connects a new utterance to the immediately previous one, to a
larger segment of discourse, such as a communicative pattern in the functional-prag-
matic sense, or to a segment of such a pattern (pattern position). Discourse coordina-
tion in such a sense is about “operating on the knowledge” and the “coordination of
knowledge structures between a speaker and a hearer”. The discourse-coordinating
element may have functional scope over one or several utterances, where an utterance
can be defined as an actional unit containing both a proposition and an illocution of
its own.15 The workings of the coordinative effects generally consist of a qualification
of an already verbalized utterance in retrospect and the opening of an expectation re-
garding the next utterance to be verbalized. The new utterance is then integrated into
the larger discourse by way of being contextualized with regard to built-up or shared
knowledge and expectations. Coordination understood in this way is directly con-
nected to planning, hearer-direction, and to the marking of discourse elements as es-
Discourse coordination in Turkish monolingual and Turkish-German bilingual children’s talk 

tablished units of common knowledge. Coordinating elements often fulfill more than
one function in this respect: their functional compositionality can be quite complex
and integrate several procedures.
In functional-pragmatic terms, coordinating elements realize para-operative pro-
cedures. The coordinating elements analyzed in Rehbein (1995) historically each
emerged from the fusion of a deictic and a phoric16 procedure, respectively. Their func-
tioning in discourse crucially relates them both to the previous part of discourse,
which they requalify as something to be completed, elaborated on, contrasted or re-
stated in other terms, and to the new part of discourse yet to be verbalized. The differ-
ences between them are related to the specific ways in which they relate the previous
and the following part of discourse.17 Crucially and interestingly, this approach is ag-
nostic with regard to a categorization into word classes. The action-theoretical notion
of procedures applies to units originating below the word boundary18 and at the same
time operating within and beyond the utterance boundary. The approach also tran-
scends a neat distinction between discourse markers and conjunctions, thus overcom-
ing in a way the categories that formed the basis for a grammaticization view in the
first place. The following sections will present işte under a quantitative perspective
(section 3), then proceed to an illustration of several functions of işte by looking at
some examples from the ENDFAS/SKOBI data.19

3. Discourse coordination in the data: more o zaman, less işte

With respect to the ENDFAS/SKOBI data, an overall screening of a sample of approx-


imately 60,000 utterances for markers of discourse coordination has shown that the
knowledge-marking particle işte is one of the most frequently occurring forms used
for this purpose (8.9% of all the discourse-coordinating items), in the monolingual
data. The bilingual data have shown a reduced employment of işte (1.7 % of all the
discourse-coordinating items) in favor of more temporal deictics such as o zaman
‘then, at that time, in that case’ and conjunction-like elements such as ama ‘but’, a ten-
dency of preferences corresponding to the findings in monolingual and bilingual Ger-
man, cf. table 1.
 Annette Herkenrath

Table 1.  Preferred coordinating elements in monolingual and bilingual data (Herkenrath,
Özdil & Rehbein 2004)

Turkish German

monolingual bilingual monolingual bilingual

dA 27.6% dA 20.7% und 20.9% und 33.8%


ama 9.2% sonra (+) 14.7% aber 17.6% und dann 17.0%
işte 8.9% o zaman (+) 12.8% und dann 17.2% dann 11.7%
ondan sonra 9.1% ama 12.7% dann 9.8% aber 11.2%
sonra 5.5% ondan sonra 10.4% auch 9.0% auch 9.7%
ee 4.6% şey 4.5% da 6.9% also 3.5%
böyle 3.9% bir de 4.5% und da 5.3% weil (+V2) 2.9%
bir de 3.8% böyle 3.1% also 4.5% oder 2.5%
yani 3.4% hani 1.9% weil (+V2) 4.1% da 2.0%
şey 3.3% işte (-) 1.7% oder 1.6% und da 1.1%
şimdi 2.3% çünkü 1.7% danach 1.1%
ya 1.9% şimdi 1.4% nur 1.1%
çünkü 1.8% ve 1.4%
o zaman 1.7% ee (-) 1.3%
tabii 1.5% ya 1.1%
ve 1.3%
tamam mı 1.0%

The presently investigated subcorpus comprises 39 Turkish tape-recorded discourses


(47,500 transcribed utterances20), 22 discourses (28,593 utterances) involving bilingual
and 17 discourses (18,907 utterances) involving monolingual children. The discourses are
of an average length of 60 minutes (some 1,200 transcribed utterances). Speakers include
the child informants, their friends and siblings, interviewers, as well as other members of
the families or neighborhoods. The counted utterances are those made by the main in-
formants, i.e. 8,215 utterances from the monolingual and 11,744 utterances from the bi-
lingual subcorpus. Table 2 shows the distribution among individual children.
Discourse coordination in Turkish monolingual and Turkish-German bilingual children’s talk 

Table 2.  İşte realized by monolingual and bilingual speakers

monolingual bilingual

speaker age utterances işte % speaker age utterances işte %

Azim 8;09 1,730 13 0.75 Binnaz 5;05 593 0 0


Emel 5;08 251 4 1.59 7;03 625 1 0.16
Gözde 8;03 678 14 2.06 Burçin 12;01 500 0 0
Halime 6;09 475 4 0.84 12;02 226 0 0
Melike 5;11 586 60 10.23 Emrah 5;10 321 4 1.24
6;11 464 22 4.74 Enis 8;01 306 15 4.90
7;09 169 0 0 9;00 612 7 1.14
Memduh 5;06 327 1 0.30 Erkin 7;11 202 2 0.99
Musa 7;02 169 2 1.18 Esen 4;01 78 0 0
Seda 6;09 1,288 67 5.20 Faruk 5;02 325 2 0.61
Simge 7;11 279 2 0.71 Handan 5;08 377 1 0.26
Şeyda 9;02 118 4 3.38 Kubat 5;00 598 1 0.16
Tahir 4;03 359 0 0 Özer 5;02 17 0 0
4;10 224 3 1.33 Pervin 10;05 983 0 0
Taner 8;02 441 13 2.94 11;00 238 6 2.52
9;02 291 3 1.03 Sebat 6;04 186 0 0
9;11 366 9 2.45 Selim 12;02 2,158 61 2.82
12;08 737 6 0.81
Sema 11;00 890 0 0
11;01 6 0 0
11;07 223 0 0
Sezen 5;07 814 10 1.22
Şehmuz 9;01 729 2 0.27

∑ monolingual 18,907 ∑ bilingual 28,593


corpus corpus

∑ monolingual 8,215 221 2.69 ∑ bilingual 11,744 118 1.00


children children

Table 2 lists the monolingual (on the left side) and the bilingual children (on the right
side) as well as their respective ages at recording time. For some children, the subcor-
pus comprises transcribed discourses recorded at several ages. The table also lists the
number of utterances attributable to each child, the occurrences of işte (regardless of
their functions, to be looked at below) in absolute numbers as well as the number of
occurrences per 100 utterances. As an overall picture, it can be seen that there are
fewer recordings without işte on the monolingual than on the bilingual side and that
the overall frequency of işte per 100 utterances is 2.69 times as high in the monolingual
 Annette Herkenrath

data (2.69) as in the bilingual data (1.00). The quantification also reveals some indi-
vidual variation, some children using işte a lot more than others. There are, however,
tendencies which do suggest that the employment of işte is an, albeit indirect, function
of age as well as of language contact: of the twelve most frequent users of işte, only
three are bilingual children21 (Enis at 8;1, Selim at 12;0, and Pervin at 11;0); besides,
there seems to be an age level at around six years (for the monolinguals, eight years for
the bilinguals), after which işte starts to become more frequent. In the inverse, among
the twelve children whose data display no occurrences of işte, there are only two mono-
lingual ones, and they are rather at the lower end of the age scale (Melike at 7;9 and
Tahir at 4;3).
The corpus additionally contains 99 occurrences of işte by adult speakers. These
usages seem to be highly contingent on the discourse constellations, the highest den-
sity arising in homileïc discourses among the adults only, such as casual side talks in-
cidentally recorded independently of the original research interests. This group, which
is made up of speakers belonging to a continuum of L1-L2 constellations (ranging
from monolingualism in Turkish through L1 Turkish/adult or adolescent L2 German
to successive bilingualism in Turkish and German), cannot be looked at systematically
in the framework of this study. The set of işte occurrences will however be considered
in its internal distributive structure in section 8.
The following sections will present individual occurrences of işte under a qualita-
tive perspective, aiming at a functional categorization. The preliminary assumption is
that discourse samples with frequent occurrences of işte could be viewed as potential
contexts for its systematic use in a connective, concatenating function. Samples with a
less frequent use might be more likely to display isolated occurrences in passages of a
more sequential kind.

4. Usages of işte in the data

The usages of işte documented in Şimşek (2003) and Yılmaz (2004) (subsection 2.2)
can also be found in the ENDFAS/SKOBI data. For reasons of space, however, they
cannot be documented within the limits of this paper. The following qualitative analy-
ses will therefore be based on just two discourse passages illustrating one basic, non-
connective, and several expanded, connective, usages. (E1) illustrates a typical non-
connective functioning of işte, in which it focuses the hearer’s attention on a
non-linguistic concept, in a holistic, summarizing fashion, calling upon her to recog-
nize the element in question as part of a shared or common knowledge. The knowl-
edge pointed at need not have been verbalized in the preceding discourse, nor does it
need to be about to be verbalized. Its verbal form may be irretrievable at the actual
moment of speaking, or the knowledge itself may be so well established in the shared
system of assumptions, beliefs, and known facts that it seems self-evident to all par-
ticipants. In a slightly different sense, işte can also help to suspend a verbalization or to
Discourse coordination in Turkish monolingual and Turkish-German bilingual children’s talk 

altogether avoid the elaboration of some unpleasant fact, simply by hinting at the dis-
course knowledge and thereby summarizing the non-verbalized events as somehow
being negligible details. In this case, işte functions as a placeholder or highlighter of
knowledge that will not be verbalized. (E1) is a conversation between five-year-old
Emrah22 and an adult visitor to the family, Ferda. The topic concerns the children at
Emrah’s kindergarten.
(E 1) EFE07tk_Emr_b_0677_f_SKO_010301, doc 2, score areas 87-100
010301/SKOBI/EFE7tk/Familie/Taşdemir/Sony WM-GX670/639
110402/Babur/1:60/Yaman/1:30 / / // Ende der Transkription/Sony TC-KB 920 S
Fer Ferda, female interviewer
Emr Emrah, bilingual boy, 5;10 years old, attends kindergarten
[87]
Emr I’ıh˙ • Doch!
Emr [TL] IJ PTC

Emr [eng] Uh uh • I do!


Fer Sevmiyosun. • Peki Kubat’ı
Fer [TL] like-NEG-PRS-2SG okay Kubat-ACC

Fer [eng] You don’t like him. • Okay then, do you like
negating
[88]
Emr I’ıh˙  I’
Emr [TL] IJ IJ

Emr [eng] Uh uh
Fer seviyo musun? Onu da sevmiyosun.
Fer [TL] like-���
PRS Q-2SG DEI-ACC also like-NEG-PRS-2SG

Fer [eng] Kubat? You don’t like him either.


negating
[89]
Emr ıhh˙  İşte.
Emr [TL] PTC

Emr [eng] Uh uh İşte.


Fer Niye? Anlaşmıyo musunuz
Fer [TL] what-DAT understand-REC-NEG-PRS Q-2PL

Fer [eng] Why? Don’t you get on so well with each other?
negating
 Annette Herkenrath

[90]
Emr I’ı˙  İştee.
Emr [TL] IJ PTC

Emr [eng] Uh uh İşte.


Fer onunla? Niye? Ama sen hep “İşte,
Fer [TL] DEI-�������
GEN-COM what-DAT but DEI2SG always işte

Fer [eng] Why? But you keep saying “İşte, işte”,


[91]
Emr
Emr [TL]
Emr [eng]
Fer işte” diyosun, bi şeyin nedeni olması lazım.
Fer [TL] işte say-PRS-2SG one thing-GEN reason-PSS3SG be-PAR-PSS3SG necessary
Fer [eng] a thing must have a reason.
[92]
Emr İşte.
Emr [TL] PTC
Emr [eng] İşte.
Fer Seni kızdırıyo muu, yoksa seni
Fer [TL] DEI2SG-ACC be angry-CAU-PRS Q or DEI2SG-ACC

Fer [eng] Does he make you angry, or does hit you,


[93]
Emr Dövüyo. Dövüyo.
Emr [TL] beat-PRS beat-PRS

Emr [eng] He hits. He hits.


Fer dövüyo muu, yoksa sana bi şeyler mi
Fer [TL] hit-PRS say-PRS Q or DEI2SG-DAT one thing-PL Q

Fer [eng] or does he say things to you?


[94]
Emr Dövüyo. Hıhı˙ 
Emr [TL] beat-PRS IJ

Emr [eng] He hits. Uhuh


Fer diyo? Dövüyo. Pekii • • Carola
Fer [TL] say-PRS beat-PRS okay Carola

Fer [eng] He hits. Okay, and • • Carola and


[95]
Fer falan onlar kızmıyolar mı seni dövdüğü
Fer [TL] and so on DEI-PL be angry-NEG-PRS-PL Q DEI2SG-ACC beat-PAR-PSS3SG

Fer [eng] the others, don’t they get angry when he hits you?
Discourse coordination in Turkish monolingual and Turkish-German bilingual children’s talk 

[96]
Emr Doch.
Emr [TL] PTC

Emr [eng] They do.


Fer zaman? Peki sen de onu • dövüyo
Fer [TL] time okay DEI2SG also DEI-ACC hit-PRS

Fer [eng] Okay, and you, • do you hit him


[97]
Emr Hıı̀˙ • • I’ıh˙ 
Emr [TL] IJ IJ

Emr [eng] Uhuh • • Uh uh


Fer musun geri? Sen ona
Fer [TL] Q-2SG back DEI2SG DEI-DAT

Fer [eng] back? You don’t interfere with


affirmative negating
[98]
Emr Hıı˙ 
Emr [TL] IJ

Emr [eng] Huh


Fer karışmıyosun, hep o sana karışıyo.
Fer [TL] interfere-�����������
NEG-PRS-2SG always DEI DEI2SG-DAT interfere-PRS IJ

Fer [eng] him, it’s always him who interferes with you.
[99]
Emr Şimdi yeter! ((2s)) Yeter şimdii!
Emr [TL] now enough enough now

Emr [eng] Now that’s enough! ((2s)) Enough for now!


Fer Hmm˙  Peki
Fer [TL] okay

Fer [eng] Hmm Okay then,


[100]
Fer bugün n’aapcaksın? Bugün hep benim
Fer [TL] today what-do-FUT-2SG today always DEI1SG-GEN side-

Fer [eng] and what are you gonna do today? Are you gonna stay with me

In what precedes the quoted passage, Ferda asks Emrah about the children at his kin-
dergarten. Emrah briefly mentions a boy named Kubat, who happens to be a relative
of Ferda’s, and then tells her about some other boy he does not like and a third one he
does like even though he spoils his drawings. Ferda then asks Emrah about Kubat: Peki
Kubat’ı seviyor musun?‘ Okay, do you like Kubat?’ Emrah laconically responds in the
negative I’h I’ıh (score area23 87). Being questioned about his relationship with Kubat
puts Emrah in a double-bind situation involving conflicting politeness requirements:
 Annette Herkenrath

on the one hand, principles of communicative cooperation require that he attend to


Ferda’s questions and subject proposals. On the other hand, a different politeness re-
quirement towards the adult visitor and her family prevents him from making negative
statements. The only adequate solution would seem for Emrah to keep his verbaliza-
tions to a minimum. When Ferda is not content with the obtained answer and tries to
suggest an expansion of the topic (by restating Emrah’s answer in more elaborate word-
ing: Onu da sevmiyorsun ‘You don’t like him either’, score area 88), Emrah still does not
venture any more explicit details, but rather remains with his original version: I’h I’ıh
(score area 89).
What can be observed here is an opposition between Emrah’s reticence on the one
hand and Ferda’s determination to engage him in a conversation on the other. Ferda’s
determination to make Emrah talk is doubly motivated; first by an elicitation interest
and second by her personal interest in learning about her own relative’s standing at the
kindergarten. Ferda then continues with two identical, overtly hearer-directed wh-ut-
terances in score areas 89f: Niye? ‘Why?’ Emrah, trying to cut the conversation short
while at the same time trying to avoid a violation of the cooperation requirement, uses
işte twice (in score areas 89 and 80). When the interviewer is not content with this
nonverbalization strategy (cf. her meta-communicative comment on Emrah’s evasive-
ness in score areas 90f: Ama sen hep “İşte, işte” diyorsun, bi şeyin nedeni olması lazım
‘But you keep saying “işte, işte”, a thing must have a reason’), Emrah realizes işte a third
time (score area 92).
In this sense, Emrah’s detopicalizing strategies remain unsuccessful, and even after
the third işte (in score area 92), Ferda insists on a full verbalization: Seni kızdırıyor mu,
yoksa seni dövüyor mu, yoksa sana bi şeyler mi diyo? ‘Does he anger you, does he beat
you, or does he say things to you?’ (score areas 92ff). Thus coerced, Emrah surrenders
a laconic, but at the same time emphatically repeated Dövüyor. Dövüyor. Dövüyor ‘He
hits. He hits. He hits’ (score areas 93f). Ferda nevertheless continues asking questions
and suggesting possible answers for the remaining score areas 94 to 98, without ever
again obtaining anything more elaborate than monosyllabic affirmative or negating
utterances. It is this obvious lack of success of Emrah’s employment of işte that in the
end calls for a more directive plea for a closure of the topic – Şimdi yeter! ((2s)) Yeter
şimdi:! ‘Now that’s enough! ((2s)) Enough for now!’, score area 99 – a plea that Ferda
finally has to accept, cf. her change of topic in score area 99f: Peki bugün n’ a:pcaksın?
‘Okay then what are you up to today?’.
The way işte is used in this example is deictic in that it demonstrates to the listener
(who is an inquisitive listener incessantly verbalizing – ostensible – knowledge defi-
cits) knowledge elements that would fit more or less precisely to fill the verbalized gaps
and to thereby supply the knowledge in the requested way. It is also incitive, in that the
hearer is brought to agree on an assumption to be taken as shared or common knowl-
edge. The required knowledge elements are available in the form of incidental experi-
ental knowledge (‘partikuläres Erlebniswissen’) shared in principle by the two inter-
locutors, yet for some reason, Ferda seems reluctant to focus her attention on her own
Discourse coordination in Turkish monolingual and Turkish-German bilingual children’s talk 

domain of specific knowledge and this is why Emrah uses işte in order to hint at the
relevant shared knowledge and to tell her to accept it as a given fact.

5. İşte as an individually used discourse coordinator

The following section presents an example in which işte is used in a different way,
namely to concatenate chains of events. This employment of işte principally relies on
the deictic, incitive, and summarizing procedures that bring about the effects described
in the above section. It will be argued that işte can also unfold this multifunctional
potential to achieve a kind of coordinating effect, linking a piece of verbalized know­
ledge with the surrounding discourse. In other words, it is the very knowledge-catego-
rizing quality of işte that makes it a good candidate for coordinatorship. The example
presented in this section is a discourse passage from a talk with two monolingual chil-
dren, in which işte is used in a concatenating function. There are three levels at which
işte is employed in a coordinating function. One is the level of the turn, at which a new
turn is integrated within the surrounding discourse. This integration is achieved by
relating the new contribution to the previously verbalized knowledge, thereby making
it interactionally relevant. The knowledge in turn serves as a background against which
the information offered in the new contribution can be highlighted. The second is the
level of the integration of an utterance within a turn, and the third, the integration of
parts of utterances within one utterance. For reasons of space, only one example (E2)
can be given. It contains illustrations of both the second and the third cases.
(E2) illustrates the frequent, swift and routinized use of concatenating işte. In the
passage in question, monolingual Melike speaks about some toys her father bought for
her and her brother on the occasion of their visiting his workplace and about what
happened to the toys (they accidentally broke one after the other, during play). The
passage contains ten occurrences of işte, seven of which (in score areas 62, 66, 69, 70,
71, 73, and 76) achieve an effect of coordination at a turn-internal, inter-utterance
level, the remaining three (in score areas 58, 66, and 83) at an utterance-internal level.
The present subsection briefly comments on both in order to show the interplay of
both kinds of usage in an extended concatenative discourse passage. The constellation
involves Melike, her brother, and both her parents, all of whom have first-hand know­
ledge of the events she is going to narrate, plus the as yet uninformed interviewer
Anna. Score area 57f begins a longer narrative realized by Melike, who has been asked
whether she had ever seen her father’s workplace. The passage starts with Melike claim-
ing the right to an uninterrupted turn from her mother, who has been offering to
coach her along by asking questions (Sussana yaa sana ne! ‘Oh, come on, shut up, it’s
none of your business!’). From then on she realizes an extended turn, interrupted only
twice, by an unsuccessful attempt at changing the topic (score areas 74f), and an at-
tended-to question (score area 84f), both realized by her brother Taner (Anna contrib-
uting a few hearer signals in the form of interjections).
 Annette Herkenrath

(E 2) EFE07tk_Mek_m_0658_f_SKO_290100, doc 5, score areas 57-85


290100/SKOBI/EFE7tk/Familie/Herkenrath/Sony WM-F2041/0826
170402/Babur/1:60/Esen/1:40/Babur/1:50/Taşdemir/1:40/270902/SonyTC-KB920S
Mek Melike, monolingual girl, 5;11 years old, attends preschool
Tan Taner, Melike’s brother, 8;2 years old, third form of primary school
Mut the children’s mother
Ann Anna, interviewer, Turkish L2, L1 German
[57]
Mek Sussana yaá sana ne!
Mek [TL] shut up-IMP PTC DEI2SG-DAT what

Mek [eng] Oh, come on, shut up, it’s none of your business!
Mut çöp şiş yediniz.
Mut [TL]
Mut [eng] eat, you had çöp şiş.
[58]
Mek Babam bize • çöp şiş yedirdi, • • işte bize
Mek [TL] father-PSS1SG DEI1SG-DAT straw spit eat-CAU-PST PTC DEI1PL-DAT
Mek [eng]My Dad gave us • çöp şiş to eat, işte he bought
[59]
Mek oyuncak aaldı bi de. • Eve getirdik. Bebek
Mek [TL] toy buy-PST one also house-DAT bring-PST1PL doll

Mek [eng]us toys, too. • We took them home. He bought me


[60]
Mek aldı banaa, sonra bana ütü aldı • dimi anne,
Mek [TL] buy-PST DEI1SG-DAT then DEI1SG-DAT iron buy-PST NEG-Q mother
Mek [eng]a baby doll, then he bought me an iron, • didn’t he, Mum,
[61]
Mek ütü aldı? Bi de dikiş makinası aldı.
Mek [TL] iron buy-PST one also sew-DER machine-PSS3SG buy-PST

Mek [eng]he bought an iron? And he bought a sewing machine.


Ann Ihım˙ 
Ann [TL] IJ

Ann [eng] Hm
[62]
Mek İşte o kadar şey al dık, • eve gel dik…
Mek [TL] PPPTC DEI degree thing buy-PST1PL house-DAT come-PST1PL

Mek [eng]İşte we bought so many things • we got home…


Discourse coordination in Turkish monolingual and Turkish-German bilingual children’s talk 

[63]
Mek Böyle… Oyuncak ev de
Mek [TL] DEI toy house also

Mek [eng] Like… He had also bought like a


((microphone problem))
[64]
Mek böyle almıştı şöyle, • • ağbime tabanca
Mek [TL] DEI buy-PTE-PCOP DEI elder brother-PSS1SG-DAT gun

Mek [eng]toy house, like that, he bought my brother a gun,


[65]
Mek aldı, kamera aldı, böyle oyuncak teyip aldı
Mek [TL] buy-PST camera buy-PST DEI toy tape recorder buy-

Mek [eng] he bought a camera, and he also bought like a toy tape
[66]
Mek işte ağbime bi de. İşte • o/
Mek [TL] come-PST PTC elder brother-PSS1SG-DAT one also PTC DEI

Mek [eng]recorder, işte for my brother. İşte • that/ we


Ann Hm̀hm˙
Ann [TL] IJ

[67]
Mek geldik, • evde o oyuncak evi bozduu,
Mek [TL] come-PST1PL house-LOC toy house-ACC break-PST

Mek [eng]came, • at home he broke that toy house, I tried to,


[68]
Mek yapmaya çalıştım, yaptım ((yutkunur)) tabii.
Mek [TL] do-VN-����
DAT try-PST1SG do-PST1SG of course

Mek [eng] I did ((swallows)) of course.


[69]
Mek İşte yap tım, • • sonra oyuncaklarımlan
Mek [TL] PTC do-PST1SG then toy-PL-PSS1SG-INS

Mek [eng]İşte I did, • • then I played with my toys.


[70]
Mek oynadım. İşte • böyle ağbim bigün’ •
Mek [TL] play-PST1SG PTC DEI elder brother-PSS-1SG one-day

Mek [eng] İşte • like my brother • played with them one


[71]
Mek oynuyodu onlarla. İşte be/ ağbim
Mek [TL] play- PRS-PCOP DEI-PL-INS PTC PSS1SG elder brother-PSS1SG
Mek [eng]day. İşte my/ my brother was/ my brother
 Annette Herkenrath

[72]
Mek benim/ ağbim benim oğlumdu mahsuz.
Mek [TL] DEIPSS1SG elder brother-PSS1SG DEIPSS1SG son-PSS1SG-PCOP on purpose
Mek [eng]was my son, for the sake of
for: mahsus
[73]
Mek İşte oyuncak evinen oynuyodu
Mek [TL] PTC toy house-INS play-PRS-PCOP

Mek [eng]the game. İşte he played with the toy house, with the mahsus
for: ev ile 
[74]
Mek oyuncak evlen. Böyle yere attı, oy
Mek [TL] toy house-INS DEI ground-DAT throw-PST toy

Mek [eng]toy house. He like threw it on the ground, the toy/


Tan Bi tuvalete gelsene, bi şey
Tan [TL] one bathroom-DAT come-IMPone one thing
Tan [eng] Come to the bathroom, will ya, I wa/ I want
for: ev ile 
[75]
Mek unca k/ oyuncak ev kırıldı’
Mek [TL] toy house break-PAS- PST

Mek [eng] the toy house broke.


Tan de/ bi şey göstericem!
Tan [TL] also one thing show-FUT-1SG

Tan [eng] to show you something!


[76]
Mek İşte sonra • bebek vardı, onu ben
Mek [TL] PTC then doll exist-PST DEI-ACC DEI1SG

Mek [eng] İşte then • remember that baby doll, I was rocking it,
[77]
Mek sallıyodum, dedim “Ağbicik, bunu •
Mek [TL] rock-PRS- PCOP1SG say-PST1SG elder brother-DIM DEI-ACC cradle-
Mek [eng] I said: “My dear brother, put this one • to
[78]
Mek beşiğine yatır!” dedim, böyle beşik vardı
Mek [TL] PSS3SG-DAT lie down-CAU say-PST1SG DEI cradle exist-PST

Mek [eng]sleep in her cradle! “ I said, there was like a cradle


Discourse coordination in Turkish monolingual and Turkish-German bilingual children’s talk 

[79]
Mek ((anl.)). Beşik ((anl.)) (böyle yapmışım böyle)
Mek [TL] cradle DEI make-PTE-1SG DEI

Mek [eng] ((inc.)). The cradle (like I had like made it)
[80]
Mek masaya beşik yaptım. Bi geldim “Onu
Mek [TL] table-DAT cradle make-PST1SG one come-1SG DEI-ACC
Mek [eng]I made
����������������������������
a cradle on the table. So I went like “Put that one to
[81]
Mek yatır!” dedim. • • Bebek böyle bi düştü, •
Mek [TL] lie down- CAU say-PST1AG doll DEI one fall-PST

Mek [eng]bed! “ I said. The baby dollshe like fell down, the
[82]
Mek beşik/ (geldi şöyle tam aşağı) bebeği
Mek [TL] cradle come-PST DEI exactly down doll-ACC

Mek [eng]cradle/ (came like this right downwards) he rolled the baby doll
[83]
Mek yuvarla • dı bebek kırıldı işte ayağı. Ayağı
Mek [TL] roll-PST doll break-PAS-PST PTC leg-PSS3SG leg-PSS3SG

Mek [eng]• about, the baby doll broke işte her leg. Her leg
[84]
Mek kırıldı, ((anl.)) iki ayağı da kırıldı.
Mek [TL] break-PAS-PST two leg-PSS3SG also break-PAS-PST

Mek [eng]broke, ((inc.)) both her legs broke.


Tan Gerçek
Tan [TL] real

Tan [eng] Your real baby?


[85]
Mek A’a’˙, • mahsuzdan. Sonra
Mek [TL] IJ purpose-ABL then neck-

Mek [eng] Uh uh, in the game. Then her neck,


Tan bebeğin?
Tan [TL] doll- �������
PSS2SG

Tan [eng]
negation 

Melike first mentions the food her father treated them to, thereby answering her moth-
er’s question and completing the frame of common memories of that special day. This
frame having been established, other memories come up, with the food and the toys as
two new subtopics. Melike utterance-internally coordinates these two subtopics by
 Annette Herkenrath

means of the procedures of işte described in the above sections, i.e. a storing-away of
the food-related memories and a steering of the listeners’ attention onto something new
to come, namely the toy-related memories and their verbalization. Melike then adds
memories as they come up, generally following a temporal linearization, but also add-
ing elements of knowledge as they come to her mind. Her general principle of discourse
coordination consists in packaging the verbalization up into processable units, to incite
her listeners to store them away in memory and then to add new elements.
In score areas 59ff, Melike lists the things her father bought: Bebek aldı bana, son-
ra bana ütü aldı • dimi anne, ütü aldı? Bir de dikiş makinası aldı ‘He bought me a baby
doll, then he bought me an iron, didn’t he, Mum, he bought me an iron? And he bought
a sewing machine’. Then, summing up, she finishes this package of knowledge ele-
ments and passes on to information of a different category, an advancement in time
and a change of place: İşte o kadar şey aldık, • eve geldik ‘İşte we bought so many things
• we got home’ (score areas 62f). The function of işte here is to state the achieved de-
velopment of the knowledge constellation and to thereby create the conditions to fur-
ther advance the plot. Similarly, the two occurrences in score areas 65ff (… böyle oyun-
cak teyip aldı işte ağbime bi de. İşte • o/ geldik • evde o oyuncak evi bozdu... ‘he also
bought like a toy tape recorder, işte for my brother. İşte • that/ we came • at home he
broke that toy house...’), although working on different syntactic levels, both mainly
help to advance the plot of the narration by concatenating pieces of knowledge. The
concatenation relies on the relevance of the elements added to the background of what
has up to then been established as verbalized and thereby shared knowledge.
The passage from score area 68 to 72 again contains four occurrences of işte: ...
yapmaya çalıştım, yaptım tabi. İşte yaptım, • • sonra oyuncaklarımlan oynadım. İşte •
böyle ağbim bigün • oynuyordu onlarla. İşte be/ ağbim benim/ ağbim benim oğlumdu
mahsuz. İşte oyuncak evinen oynuyordu oyuncak evlen ‘... I tried to, I did, of course. İşte
I did, • • then I played with my toys. İşte • like, my brother • played with them one day.
İşte my/ my brother was/ my brother was my son, for the sake of the game. İşte he
played with the toy house, with the toy house’. Each realization here occurs in an utter-
ance-initial position, such that each in this series of utterances is initiated by means of
işte. The progression of the narrative is realized in a stepwise fashion and what has
been established becomes fixed in memory as a base on which Melike then proceeds:
the unspecified activity at the beginning of the passage is first introduced as something
in the process of being attempted (yapmaya çalıştım ‘I tried to do’), then as something
achieved (yaptım tabi ‘I did, of course’), then, at the beginning of the new utterance,
the result of the transformation is resumed and stored away as shared knowledge (işte
yaptım ‘işte I did’), before the next step of the verbalization is undertaken, the playing
with the toys (İşte yaptım, • • sonra oyuncaklarımlan oynadım ‘İşte I did, • • then I
played with my toys’). This again is stored away at the beginning of the next utterance
(... oynadım. İşte....‘...I played. İşte...’), which then introduces the subtopic of the broth-
er: İşte be/ ağbim benim/ ağbim benim oğlumdu mahsuz ‘İşte my/ my brother was/ my
brother was my son, for the sake of the game’. So, one by one, Melike verbalizes and
Discourse coordination in Turkish monolingual and Turkish-German bilingual children’s talk 

thus establishes as shared knowledge first, her having achieved some attempted activ-
ity, second, the situation of her playing with the toys, as well as third, her brother being
there and being her son for the purposes of the game. Resuming these three estab-
lished elements of knowledge with her listeners, by means of işte (score area 73), Me-
like finally goes one step further and advances the crucial aspect of her brother actu-
ally playing with the toy house: İşte oyuncak evinen oynuyordu oyuncak evlen ‘İşte he
played with the toy house, with the toy house’ (score area 73f), which is the very pre-
requisite for the delivery of the scandalon of the narrative (Fienemann 2005, Rehbein
1980), the brother’s accidentally breaking the toy house during his play. The last occur-
rence of işte at this level of utterance coordination is in score area 76. The scandalon
has been delivered (Böyle yere attı, oyuncak/ oyuncak ev kırıldı ‘He like threw it on the
ground, the toy/ the toy house broke’, score areas 74f) and Melike proceeds to the next
narrative unit (in fact, her narrative displays a succession of like scandalons, her broth-
er apparently breaking several of her newly acquired toys): ... oyuncak ev kırıldı. İşte
sonra • bebek vardı... ‘... the toy house broke. İşte then • remember that baby doll...’.
(E2) also contains three usages of işte as a coordinator working below utterance
level. In the first occurrence – Babam bize • çöp şiş yedirdi, • • işte bize oyuncak aldı bi
de, • eve getirdik ‘My Dad gave us çöp şiş to eat, • • işte he bought us toys, too, • we took
them home’ (score areas 58f) – I would like to argue, işte serves to coordinate intra-ut-
terance constituents depending on a finite verb each (IPs so to speak): the IP Babam
bize • çöp şiş yedirdi ‘My Dad gave us çöp şiş to eat’ is coordinated with the IP bize oy-
uncak aldı bi de ‘he bought us toys, too’, işte concluding the first and announcing the
second IP. The other two occurrences are Oyuncak ev de almıştı şöyle, • • ağbime taban-
ca aldı, kamera aldı, böyle oyuncak teyip aldı işte ağbime bi de ‘He had also bought like
a toy house, • • he bought my brother a gun, he bought a camera, and he bought like a
toy cassette recorder işte for my brother, too’ (score areas 63ff) and Bebek böyle bi
düştü, • beşik/ (gerçi ) bebeği yuvarla • dı bebek kırıldı işte ayağı ‘The baby doll like fell
down, • the cradle ( though) he • rolled the baby doll about, the baby doll broke işte her
leg [broke]’ (score areas 81ff). In both occurrences, the coordination takes place at the
level of NPs; however, there are no two parallel NPs, but rather, the respective NPs are
added as in devrik cümle constructions, additional afterthought information or preci-
sion delivered after the finite verb.24 What distinguishes the usages of işte in (E2) is
their being used not only very routinely, swiftly, and frequently, but crucially the func-
tional extension of işte allowing it to work below the utterance-internal level. By being
used in a series, their deictic and incitive effects seems to not entirely disappear, but to
fade away so as to mainly show a concatenative, coordinating effect. In functional-
pragmatic terms, these instances of işte contain an operative procedure, whose func-
tional scope includes units of verbalization, be it at an utterance-internal syntactic
level or at a level transcending the utterance boundary, and whose communicative
function is to categorize the knowledge verbalized in these units as having become
shared knowledge.25
 Annette Herkenrath

6. Functional analysis and typification

In the following, I am assuming three types of occurrences of işte, each of them show-
ing a differentiated unfolding of its basic fuctional potential. This basic functional po-
tential, as realized in both the monolingual and the bilingual data (cf. E1) and in ac-
cordance with the etymological information available (cf. subsection 2.2) is
biprocedural since işte is a morphologically fused expression: it is incitive, in that it di-
rectly interferes with the actional apparatus of the hearer, calling upon her to recognize
a specific element as part of a knowledge to be shared. It is also partly deictic in focusing
the hearer’s attention on this element in the first place. Depending on which of the two
procedures was integrated later to supersede the original one and to become responsi-
ble for the main communicative effect, işte would have to be analyzed as either para-
deictic or para-incitive. Since the etymological knowledge about işte is sparse, however,
it is at present difficult to decide whether işte is to be analzyed as para-deictic or para-
incitive. I am henceforth referring to occurences of this type as type 1 occurrences.
The unit referred to can be an element in the perception space or in one of several
involved domains of knowledge. In many cases, however, işte takes into its scope stretch-
es of previous verbalization, which thereby become marked as having become shared
knowledge. As the previous sections have shown, the coordinating effects of işte come
about as side effects to their primary functions, but they may also involve the integration
of an additional procedure, so that the communicative function of each occurrence be-
comes even more complex, comprising a combination of functional procedures.26 With
respect to the realizations in E2, one can argue in favour of a connective function of işte,
both at the inter-utterance and the utterance-internal level. Since the connectivity of işte
is based on a processing of the verbalization activity and on a categorization of verbal-
ized units as representing what has become shared knowledge, it seems justified to speak
of an operative procedure in this connection. Those occurrences, in which connectivity
is utterance-internal, i.e. between parts of utterances of different syntactic status can also
be argued to have reached an advanced stage in a process of monologization, as de-
scribed in Redder (1990). I will henceforth refer to examples of inter-utterance connec-
tivity as type 2 and to those of utterance-internal connectivity as type 3 realizations.27 A
schematization of this categorization is given in Table 3.
Type (1) contains examples of the kind illustrated in subsection 4.1, where işte
primarily works in a biprocedural way, focusing the hearer’s attention on non-verbal-
ized elements of knowledge and inciting her to recognize them as being part of a
shared or common knowledge. İşte in this function helps to avoid verbalization, and its
functional scope contains non-linguistic rather than linguistic elements (part of the
perception or imagination space).
Discourse coordination in Turkish monolingual and Turkish-German bilingual children’s talk 

Table 3.  Categorization of usage types of işte

connectivity functional scope procedures changes

1 none