Dynamics of Contact-Induced Language Change 9783110271430, 9783110271331

Citation preview

Dynamics of Contact-Induced Language Change

Language Contact and Bilingualism 2 Editor

Yaron Matras

De Gruyter Mouton

Dynamics of Contact-Induced Language Change

edited by

Claudine Chamoreau Isabelle Le´glise

De Gruyter Mouton

ISBN 978-3-11-027133-1 e-ISBN 978-3-11-027143-0 ISSN 2190-698X Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.dnb.de. ” 2012 Walter de Gruyter GmbH & Co. KG, Berlin/Boston Cover image: Anette Linnea Rasmus/Fotolia Typesetting: RoyalStandard, Hong Kong Printing: Hubert & Co. GmbH & Co. KG, Göttingen 앝 Printed on acid-free paper 앪 Printed in Germany www.degruyter.com

List of contributors Alexandra Y. Aikhenvald Cairns Institute James Cook University Australia [email protected] Carla Bruno Dipartimento di Scienze Umane Universita` per stranieri di Siena Italy [email protected] Claudine Chamoreau CNRS (SeDyL/CELIA – CEMCA) France and Mexico [email protected] Patience Epps University of Texas at Austin USA [email protected] Zarina Estrada Ferna´ndez University of Sonora Mexico [email protected] Ana Ferna´ndez Garay CONICET – UNLPam Argentina [email protected] Anthony P. Grant Edge Hill University UK [email protected]

Bernd Heine University of Cologne Germany [email protected] Sibylle Kriegel CNRS (Parole et Langage) France [email protected] Isabelle Le´glise CNRS (SeDyL/CELIA) France [email protected] Julen Manterola University of the Basque Country Spain [email protected] Yaron Matras University of Manchester UK [email protected] Thomas Stolz University of Bremen Germany [email protected]

Table of contents List of contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

v

A multi-model approach to contact-induced language change . . . . . Claudine Chamoreau and Isabelle Le´glise

1

An activity-oriented approach to contact-induced language change . . Yaron Matras

17

Contact-induced change as an innovation . . . . . . . . . . . . . . . . . . . . Claudine Chamoreau

53

Language contact in language obsolescence. . . . . . . . . . . . . . . . . . . Alexandra Y. Aikhenvald

77

The emergence of a marked-nominative system in Tehuelche or Aonek’o a jen: a contact-induced change? . . . . . . . . . . . . . . . . . . . Ana Ferna´ndez Garay On polysemy copying and grammaticalization in language contact . . Bernd Heine

111 125

The attraction of indefinite articles: on the borrowing of Spanish un in Chamorro. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Thomas Stolz

167

On form and function in language contact: a case study from the Amazonian Vaupe´s region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Patience Epps

195

The Basque articles -a and bat and recent contact theories. . . . . . . . Julen Manterola

231

Contact phenomena/code copying in Indian Ocean Creoles: the post-abolition period . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sibylle Kriegel

265

Grammaticalization of modal auxiliary verbs in Pima Bajo: an internal or a contact-induced change?. . . . . . . . . . . . . . . . . . . . . Zarina Estrada-Ferna´ndez

285

viii

Table of contents

Contact, convergence, and conjunctions: a cross-linguistic study of borrowing correlations among certain kinds of discourse, phasal adverbial, and dependent clause markers . . . . . . . . . . . . . . . . . . . . Anthony P. Grant

311

On a Latin-Greek diachronic convergence: the perfects with Latin habeo/Greek e´cho# and a participle . . . . . . . . . . . . . . . . . . . . . . . . . Carla Bruno

359

Author index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Language index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Subject index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

377 384 390

A multi-model approach to contact-induced language change Claudine Chamoreau and Isabelle Le´glise

This volume deals with some never before described morphosyntactic variations and changes appearing in settings involving language contact. The primary purpose of the articles it presents is to identify di¤erent factors in language change. These changes are not treated as phenomena amenable to explanation from a single source: they constitute a dynamic domain of complex, complementary, and correlated processes that have to be treated with a fine-grained approach. The development of morphosyntactic structures in a situation of language contact should not be analyzed through a single lens. Contact-induced changes are generally defined as dynamic and multiple, involving internal change as well as historical and sociolinguistic factors. The identification and consideration of a variety of explanations constitutes a first step; analyzing their relationships forms a second. Only a multifaceted methodology enables this fine-grained approach to contact-induced change. A range of methodologies are proposed in the following chapters, but they generally have their roots in a typological perspective. The contributors recognize the precautionary principle: for example, they emphasize the di‰culty of studying languages that have not been described adequately and for which diachronic data are not extensive or reliable, and they warn of the dangers of hypothesizing beyond the evidence and identifying possible tendencies that can never be confirmed definitively. Three main perspectives on contact-induced language change are presented here, corresponding to three possible approaches to discussing the subject as part of a complex whole. The first explores the role of multilingual speakers in contact-induced language change, especially their spontaneous innovations in discourse. The second explores the di¤erences between ordinary contact-induced change and change in endangered languages. The third discusses various aspects of the relationship between contact-induced change and internal change.

2

Claudine Chamoreau and Isabelle Le´glise

The role of speakers and settings Historical linguists claim that change is unpredictable; even the most common or frequent change does not inevitably occur in a particular language or in a particular situation (Faarlund 1990; Lass 1980). This is also true for contact-induced changes: ‘‘any search for deterministic predictions of language change is bound to fail, whether the focus is on internallymotivated change or on contact-induced change’’ (Thomason 2000: 173). Language changes are thus unpredictable partly because speakers’ attitudes are unpredictable, but above all because ‘‘there are no linguistic constraints on interference’’ (Thomason 2001: 85). Contact-induced change and communicative goals Social factors are fundamental to the definition of contact phenomena. Thomason (2001; see also Thomason and Kaufman 1998) has proposed a typology of interference mechanisms, establishing distinctions between language shift and language maintenance, language learning and language creation. It is crucial to take these factors into account, but the correlation between a specific type of social setting and a structural modification due to language contact is not always clear. The same e¤ect may be observed with respect to language shift and language maintenance (see for example the rise of definite and indefinite articles in various languages, as discussed by Matras, Stolz, and Manterola, this volume). Yaron Matras discusses the role of the social prestige of a language, often defined in terms of political, economic, or public dominance. He gives evidence to show that asymmetry in the social roles of the languages may determine the direction of change, but does not necessarily explain the motivation for structural change. The relationship between social settings and structural factors in contact-induced change is a crucial question, which Matras tackles through an integrated approach that links social context, conversational pressure and communicative intent, and the specific functional role of the structure or category in question. He examines the linguistic attitudes of multilingual speakers who make use of a complex repertoire in order to attain their communicative goals. One of Matras’ objectives is to identify the relationship between spontaneous innovations in discourse and the processes of language change through the propagation and stabilization of these innovations in communication. His hypothesis is that innovations are not arbitrary but driven by a communicative purpose, and that contact-induced change is the product

A multi-model approach to contact-induced language change

3

of the creativity of speakers who seek new ways to achieve goal-oriented tasks in communicative interaction. He claims that contact-induced language change is the result of speakers’ creativity in exploiting the full range of options available in their complex linguistic repertoire, and explores the ways in which lexical insertions may become lexical borrowings when they become a regular feature of the language in which they are inserted or when they are used in monolingual contexts. The innovator’s social potential to influence others is another factor in play here. Matras thus shows that social and structural factors are involved in facilitating or constraining the successful propagation of innovations throughout a speech community. Contact-induced change as an innovation Heine (2006) argues that ‘‘speakers recruit material available in R (the replica language) to create new structures on the model of M (the model language) and . . . rather than being entirely new, the structures created in R are built on existing use patterns and constructions that are already available in R.’’ This creation is understood as a process by which the speakers of the receiving language look for methods of establishing equivalence relations between their language and the source language, generally appropriating a feature or structure of a source language and adapting it in their own language. Creative activity is an important part of contactinduced change, as is well-known and described in many studies in which informants are portrayed as ‘‘unpredictable speakers’’ (Thomason 2001) or ‘‘language builders’’ (Hage`ge 1993). However, some studies make a distinction between the creation and the simple addition of a new structure. The former is a well-known activity, which adopts the model of the source language and may modify it to adapt its structure to the receiving language. The latter, less attested, is characterized by the emergence of a structure that is clearly a consequence of contact, but is not produced on the model of the receiving language nor on that of the source language. Claudine Chamoreau describes the structural and typological consequences of the contact between Purepecha (isolate, Mexico) and Spanish in the domain of comparative constructions. It is clear that Purepecha has been modified in this domain under the influence of Spanish in three di¤erent ways. Firstly, the Spanish particle type ma´s . . . que has been borrowed and replicated. Another particle type may be associated with an original construction attested in Lengua de Michoacan (a pre-contact replica lan-

4

Claudine Chamoreau and Isabelle Le´glise

guage), the coordinated type with negation (Lit. ‘It is warmer inside the house and not outside.’). This type is a creation resulting from contactinduced and internal changes. A third particle type is also accompanied by a locative phrase, as in Spanish ma´s . . . de . . . que. However, another specific construction was created on the model neither of the receiving language nor of Spanish, the contact language: a construction in which the Spanish preposition entre is used in order to form a comparison. This construction is clearly influenced by Spanish, but it displays a use in Purepecha that deviates from the patterns of comparative construction in Lengua de Michoacan and in Spanish, and from the use of the morpheme entre in Spanish. The transfer of Spanish entre allows Purepecha to innovate in the expression of the comparison of superiority and in the context of use of this Spanish preposition. In the contact linguistics literature, it is rare to find a feature described as a new structure that diverges from both the languages in contact. In Purepecha, Chamoreau links this innovation with two factors: an identity issue, that is, the desire of the speakers of the villages in which this construction is found to distinguish themselves from others on linguistic and cultural levels, and also a cross-linguistic tendency to connect comparison with location and to express comparison through a locative type. Chamoreau claims that innovative activity as a choice seems to be caused both by sociolinguistic factors and cross-linguistic tendencies.

Contact-induced change and endangered languages Another topic explored is the di¤erence between ordinary contact-induced change and that occurring in endangered languages. Many specific linguistic changes have been cited as markers of obsolescence, in particular reduction of paradigms, reduction in the use of grammatical categories, and loss of grammatical categories or of optional mechanisms in morphology or syntax (for example Dorian 1981; Sasse 1990). However, these same processes are also attested as contact-induced changes (Thomason 2001). Both language contact and language obsolescence may promote structural changes, but specific criteria have not yet been established to distinguish between changes that can be seen as signs of obsolescence in process and changes that might occur under language contact or multilingual settings. The view that contact-induced changes and the consequences of language decay have to be distinguished is relatively unusual among specialists in the field; it has often been said that the types of change observable in an obsolescent

A multi-model approach to contact-induced language change

5

language do not di¤er from those occurring in other kinds of contact settings (Dorian 1981: 151; Romaine 1989: 71). Campbell and Muntzel (1989: 195) try to draw a distinction between obsolescent processes and changes that can be attributed to language contact, while acknowledging that it is not always an easy distinction to make. They use examples from Pipil, but note that ‘‘one might suspect that these Spanish-influenced structural mutations away from relational nouns reflect the kind of change that would only take place in Pipil’s moribund state. However, completely parallel changes have taken place in other completely viable Nahua dialects, Pipil’s sister languages.’’ Other authors, such as Hill (1989: 149) and Tsitsipis (1989: 117), see rapidity as a feature that distinguishes change during obsolescence from ordinary processes of change. For example, Hill (1989) provides a careful study of the frequency of use of relative clauses in Mexicano and Cupen˜o (both Uto-Aztecan languages) and the correlation of these frequencies with the degree of obsolescence of the languages. Dorian (1981: 151) observes that although the types of linguistic change are the same in obsolescence and contact settings, the rate of change may be atypical in the case of language death. Clairis (1991: 9) claims that it is not the presence of a specific feature that is to be considered as a symptom of obsolescence but rather its frequency, compared with the frequency of the feature in healthy languages. Aikhenvald claims that the di¤erence between language change in ‘‘healthy’’ and in endangered or obsolescent languages very often resides in the quantity of change (a massive influx of borrowed forms and patterns as a result of the encroachment of one language on the other), and also in the speed with which this type of language changes. In other words, ‘‘an obsolescent language may tend to rapidly become structurally similar to the dominant one’’ (Aikhenvald, this volume). In this volume, Alexandra Aikhenvald and Ana Ferna´ndez Garay illustrate cases of ‘‘gradual death’’ (Campbell and Muntzel 1989), that is, of languages no longer actively used nor transmitted to the next generation. They observe that speakers of an obsolescent language vary in their proficiency, from ‘‘fluent language speakers’’ to ‘‘semi-speakers’’ and ‘‘rememberers’’ with very limited competence (see also 1998: 441–469). Sasse (1990: 51) gives some evidence at a linguistic level in favor of a distinction between language contact and language obsolescence, relative to structural changes involving loss of linguistic material. He claims that ‘‘Theoretically, contact-induced loss can easily be distinguished from loss due to decay, because the former is motivated by the absence of the respective categories in the contact language, while decay involves loss of

6

Claudine Chamoreau and Isabelle Le´glise

categories not motivated in this way.’’ This distinction is not always easy to show, since some types of loss and reduction in obsolescence are associated with types of loss or reduction that can be attributed to contact. Aikhenvald (this volume) and Ferna´ndez Garay (this volume and 1998: 441–469) also state that simplification of syntactic structure, reduction and loss of linguistic material, phonetic fluctuations, and the existence of optional syntax are all consequences of language obsolescence. Aikhenvald notes that ‘‘categories absent from the dominant language are particularly endangered.’’ Both illustrate the consequences of contact-induced change in contact settings with di¤erent domains. Drawing on synchronic data, Aikhenvald shows that in Tariana, an Arawak language spoken in the multilingual Vaupe´s area in Brazil, obsolescence is accompanied by a rapidly increasing number of calqued forms and constructions from Tucano, the dominant language of the area. She claims that ‘‘before passing into extinction, an obsolescent language may become a ‘carbon copy’ of the dominant idiom.’’ She explores in particular the domain of personal pronouns, showing that languages that do not have the inclusive versus exclusive opposition in the first person plural may adopt it, as has happened in the case of two Arawak languages, Mawayana and Resı´garo, which, like other languages of this family, do not distinguish an inclusive from an exclusive form. The speakers of Mawayana introduced the Waiwai (Carib family) first person plural exclusive pronoun amna and reinterpreted the original first person plural prefix wa- as inclusive in order to express this opposition. The speakers of Resı´garo have also adopted this opposition from Bora (Bora-Witotoan group), borrowing the Boran first person plural exclusive. In these cases, pronouns seem to have been borrowed to fill a perceived gap in the pronominal paradigm. Borrowing a personal pronoun or a category that deals with a pronominal domain, such as the inclusive/exclusive category, is not very common, although it has been described in certain languages as a result of di¤usion in a specific situation of contact (Jacobsen 1980; Thomason and Everett 2005). Thomason and Everett (2005: 307–308) stress the relevance of speakers’ decisions: ‘‘the crucial point in all these cases is that social factors, not linguistic ones, determine the likelihood of pronominal borrowing. If speakers want to borrow one pronoun or a whole set of pronouns, they can do so; and sometimes speakers do want to do this. The borrowed pronouns may change the structure of the pronominal system significantly, as when a new category of inclusive vs. exclusive ‘we’ is introduced or lost through borrowing. . . . extensive lexical and structural borrowing is neither inevitable nor impossible in the most intense contact

A multi-model approach to contact-induced language change

7

situations.’’ The important point is that pronominal borrowing seems not to be especially unusual under certain social circumstances, such as intense contact situations. In her contribution, Aikhenvald suggests that these types of borrowing could be the result of a considerable influx of non-native elements (loanwords and replication) and drastic restructuring, which characterize obsolescent languages. Ferna´ndez Garay argues that the existence of a marked-nominative system in Tehuelche, which was probably an ergative language (like the proto-language Proto-Chon), is due to contact with other languages, but that the variations attested and the speed of the process were probably due to the situation of obsolescence. Ferna´ndez Garay bases her analysis on language reconstruction and synchronic data. The process, which involves a realignment resulting from the reanalysis and/or extension of an adposition, may be an internal one. Nevertheless, it seems probable that in the case of Tehuelche, the influence of another language in the area helped to transform an ergative language into a marked-nominative one. The coexistence of Tehuelche with Mapudungun, a nominative-accusative language, led the ergative marker or agent marker of the transitive clause to be extended to the intransitive agent, leading to the transformation of this ergative system into a marked-nominative one. Ferna´ndez Garay points out that the long and intensive contact with Mapudungun (over at least four centuries) in Tehuelche, an obsolescent language (almost extinct when it was described), may have led to important changes and restructuring in its morphosyntactic structure, showing a loss of a syntactic characteristic. The rise of a marked-nominative system formed part of this restructuring.

Contact-induced change and internally motivated change Contact-induced change and principles of grammaticalization Contact-induced language change has often been related to the presence or absence of constraints that may explain the borrowing of di¤erent kinds of structures (Thomason 2001; Winford 2003). Bernd Heine gives an example of the constraints of principles of grammaticalization on replication in Slavic languages, and Thomas Stolz gives an example of borrowing in Chamorro. They both claim that contact-induced grammaticalization proceeds along a largely predictable sequence of stages and that the stage of grammaticalization in the receiving language never seems to reach the stage of grammaticalization of the source language. They demonstrate

8

Claudine Chamoreau and Isabelle Le´glise

that speakers appear to choose a complex strategy going through the whole process from numeral to article. The case of the indefinite article illustrates this position. Heine examines language contact situations in which grammatical meanings or structures are involved. Using three examples (articles, possessive perfects, and the auxiliation of ‘‘threaten’’ verbs) from a range of European languages, he argues that contact-induced grammatical change is constrained by universal principles of grammaticalization. He explains that the constraint on contact-induced grammatical replication suggests that, at least in cases like those discussed in his article, there really is no polysemy copying and the borrowings are not really complete replicas of their models. He suggests that what language contact triggers is a gradual process from a lesser to a greater degree of grammatical structure. In order to illustrate this process, Heine presents an example from Upper Sorbian, a Slavic language which, like other Slavic languages (with the possible exception of Macedonian), is known for the absence of indefinite articles. Language contact seems to have played some role in the rise of the indefinite article in Upper Sorbian. This receiving language seems to have reached the same degree of development as its German model, but Upper Sorbian displays a number of contexts where the replica category is less grammaticalized than the source. Heine develops another example of the rise of the indefinite article in Molisean, probably due to contact with Italian. It is interesting to note that the two Slavic languages (Upper Sorbian and Molisean) exhibiting the most intensive contact with languages that do have indefinite articles are also the ones that have created corresponding articles. Stolz looks at the use of the indefinite article in Chamorro in order to demonstrate the extent to which the Austronesian morpho-syntax of this language has been a¤ected by the introduction of the indefinite article. He compares his findings with the evidence drawn from other languages whose indefinite articles might turn out to be at least partially the product of language contact with Spanish. The rise of the indefinite article in Chamorro is a consequence of the contact with Spanish – the indefinite article morpheme un is directly borrowed from that language – and its development in the Austronesian language is constrained by universal principles of grammaticalization (see the five-stage scale of Heine 1997, and Heine, this volume). As in other cases discussed by Heine (this volume), the grammaticalization of un has not reached the stage of grammaticalization of Spanish un. However, the indefinite article in modern Chamorro also deviates from the patterns of the Spanish etymological source: the

A multi-model approach to contact-induced language change

9

borrowing, integration, and internal development of the article un has generated a category that is neither completely Spanish nor purely Austronesian. This is an example of partial copying (Heine and Kuteva 2005). Stolz demonstrates that there is a preference in language contact situations for an item to replicate first on a low level of grammaticalization in the receiving language, no matter how far the item has advanced on the grammaticalization scale in the source language, and then to continue the process according to known principles of grammaticalization. ‘‘Conspiracy’’ between contact-induced phenomena and internal phenomena Generally, studies on language change only take into account some of the types of mechanism and process reflecting grammatical changes – either internal phenomena or contact-induced phenomena, but not both. Nevertheless, a century ago, Meillet (1982 [1906]: 4, 1982 [1912]: 130–131) argued that the evolution of grammatical structures would imply the presence of processes due to internal change (analogy and grammaticalization) as well as processes related to language contact (borrowing). Recently, researchers using a variety of approaches have rethought the distinction between these types of mechanism and have proposed a multicausal or multi-factorial perspective (Harris and Campbell 1995: 50; Heine and Kuteva 2005; Peyraube 2002; Kriegel 2003; Thomason 2007; Matras 2007; Chamoreau, Estrada, and Lastra 2010; Chamoreau and Goury in press). These studies re-examine multi-causality and the distinction of the two types of mechanism. Heine and Kuteva (2003, 2005) have explored what they called contactinduced grammaticalization, in which language-contact phenomena work in conspiracy with grammaticalization (2008: 218). If the causes, processes, and consequences of language change are multiple, their explanation must be too. This multiplicity reveals both di¤erences and complementarities between the internal mechanisms and the contact-induced ones. The examination of relevant data is a first step, the analysis of their di¤erences and complementarities a second one. The two types of explanation are not contradictory or mutually exclusive; they interact in a complementary manner to produce language change. It is also necessary to show that these two types of change can act and interact in the language processes and at the outcome level. Four articles here focus on the relationship between contact-induced and internal changes in the causes, processes, and outcomes of change.

10

Claudine Chamoreau and Isabelle Le´glise

Patience Epps argues for a multiple causation approach to language evolution, involving a typological understanding of language contact and language change and the exploration of the possible interaction between these two processes. She o¤ers examples from the Vaupe´s region of northwest Amazonia, a linguistic area characterized by grammatical di¤usion among languages from three families (East Tukanoan, Nadahup [Maku´], and Arawak). The Vaupe´s region can also be considered a ‘‘grammaticalization area,’’ that is, a region where several languages have undergone (and are currently undergoing) similar processes of grammaticalization. The region is known for its unusual language contact situation, in which resistance to the borrowing of lexical and morphological forms is coupled with a widespread di¤usion of grammatical structures and categories that has driven grammaticalization within the recipient languages to generate new forms from existing (native) material to fulfill new functions. In such a context, it is unclear what role, if any, is played by cross-linguistic similarities of form either to limit or promote the transfer of grammatical structures. Epps explores this question through a case study of the etymon ni in Hup (Nadahup/Maku´ family) and other Vaupe´s languages. She points out that the similarities among the forms and lexical functions of the ni etymon across the Vaupe´s languages suggest that contact has played an important role in shaping the current picture, although precisely what should be attributed to contact and what to internal change remains unclear. Nevertheless, she shows that, unusually for this region, ni is represented by a similar constellation of forms and functions across these languages. The case of ni suggests that, in keeping with wider trends of language contact, even in the exceptional context of the Vaupe´s, elements of shared form may precede and even promote structural borrowing. Julen Manterola explores Basque definite and indefinite articles and the role of contact in their diachronic evolution. He points out some problems with the ways Basque data have been used in recent contact theories, in particular the Heine and Kuteva contact-induced grammaticalization thesis. He discusses three specific problems. Firstly, the use of only one source does not take distinctive dialectal data into account; empirical knowledge about Basque needs to be brought up to date. Secondly, historical data have been neglected. Thirdly, the function of the ancient plural indefinite article batzu has never been explored. Manterola emphasizes the importance of knowledge of the history of the language. He argues for the precautionary principle in language contact studies, especially when diachronic information is not available and no clear data have been found to determine whether a change is contact-induced or internal. He shows that contact e¤ects can

A multi-model approach to contact-induced language change

11

conceal the typical diachronic paths of other e¤ects (for example the role of the singular/plural marking overt distinction), and points out an interesting direction for further studies, focusing on the time dimension of language development. Sibylle Kriegel also argues for the precautionary principle when diachronic data are not available, calling attention to the problem of indeterminacy in dating the copying of elements into Creoles. She analyzes an interesting but neglected case of code copying from languages other than the base language into Creole. For elements that come from the base languages it is very often possible to tell whether they date from the period of creolization or more recently, but the absence of data from the period of constitution of Creoles and the sparse data on their later evolution do not allow for a definite answer. Kriegel sheds new light on the notion of creolization. She demonstrates that two function words of Mauritian and Seychelles Creole, two closely related French-based Creoles, are instances of code copying (Johanson 2002), resulting from the di¤erent language contact situations to which these languages were exposed after the abolition of slavery in 1835. The use of depi as an ablative marker in IndoMauritian Creole varieties is interpreted as a covert copy from Bhojpuri, an Indic language which has been in contact with Mauritian Creole since the migration of indentured laborers from Asia. The use of pourdir as a complementizer in some varieties of Seychelles Creole is interpreted as a covert copy from Eastern Bantu languages in contact with Seychelles Creole in the late nineteenth century. Zarina Estrada Ferna´ndez demonstrates that, in the absence of diachronic information, internal reconstruction is an important step to be undertaken in cases where grammatical patterns are involved in language contact situations. In her analysis she takes into consideration not only universal principles of grammaticalization but also the historical changes and typological properties of the language family studied, here the UtoAztecan family. She emphasizes the necessity of distinguishing internal and contact-induced change when the processes occur within a family, while recognizing that this is often di‰cult. She traces the emergence of modal auxiliary verbs in Pima Bajo, one of the Uto-Aztecan languages of northwestern Mexico, as the result of processes involved in verbal complementation, performing a fine-grained exploration of the di¤erent possibilities for encoding verbal complements in various languages of this family. She adopts a cautious approach, concluding with two hypothetical explanations for the processes in question but not opting for either one: it is impossible to determine if the development of modal verbs in Pima Bajo

12

Claudine Chamoreau and Isabelle Le´glise

should be explained as the result of a structural replication from Spanish or as the result of an internal process with di¤erent diachronic pathways. Anthony P. Grant’s article discusses borrowed mechanisms and implicational hierarchies of grammatical borrowing. He too adopts the precautionary principle, in situations where no diachronic data are available or when alternative explanations are possible. Implicational hierarchies show how likely it is that a structural category will be a¤ected by contactinduced change (Matras 2007b). Matras (2007b: 32) explains that two types of generalization may be proposed for the borrowing of grammatical categories. One is ‘‘the frequency with which a category may be a¤ected by contact-induced change’’; the other type suggests ‘‘an implicational relationship between the borrowing of individual categories: the borrowing of one category is understood to be a pre-condition for the borrowing of another.’’ Implicational hierarchies show the borrowing tendencies that take place in language contact. Grant examines major borrowed mechanisms in processes including clause-linking, coordination, complementation, conditionality, and causality in various languages, and discusses the extent to which hierarchies of dependent clause marker borrowing can be established and empirically validated. He notes that several of the languages are documented in considerable chronological depth, while others are less well-described varieties of welldocumented languages, a di‰culty for his approach. He explores the processes in question in a global cross-linguistic sample of 22 languages from a wide range of families. In a majority of the languages, the domains of discourse markers, phrasal adverbs, and coordinating, especially subordinating, conjunctions seem to be amenable to language contact. Grant also discusses the implicational hierarchy of conjunction borrowing, since conjunctions are known to be widely borrowed in many of the world’s languages. In agreement with studies of much linguists, he demonstrates that general hierarchies of grammatical borrowing have to be seen simply as tendencies. For example, the implicational hierarchy but > or > and is a general tendency confirmed in a large number of languages, but Grant o¤ers counter-examples to the expected pattern: in Livonian and Garifuna, the form meaning ‘and’ is borrowed while the one meaning ‘or’ is inherited. Lastly, Carla Bruno’s article focuses on two languages for which diachronic data are available; however, she shows that even in this situation the precautionary principle should be invoked. Against the background of the socio-cultural relations between the Roman and Greek worlds, she proposes a linguistic convergence in Latin and Greek diachrony, that is, the rise of periphrastic constructions consisting of a so-called ‘‘possessive’’

A multi-model approach to contact-induced language change

13

verb form (Lat. habeo and Gr. e´cho# ) and a past participle. Pre-existing structural similarities, due to the genetic relationship of the two languages, may have favored mutual shifts of linguistic features as well as their subsequent integration; Bruno compares the extent to which this periphrasis is integrated into each system. Languages change only in accordance with the possibilities given by their system, and Latin and Greek are instances of this rule. We have put this volume together with a number of goals in mind. First, we have aimed at presenting a number of linguistic phenomena that have not hitherto been described (variations and changes, at a morphosyntactic level, drawn from many diverse languages) and that appear in language contact settings. This diversity of languages and phenomena allows us to test, drawing on contact outcomes already described in the literature, the possibilities and preferences of various languages. Second, we have sought to include cross-linguistic and cross-dialectal perspectives, whatever the specificities of the languages and settings involved. Third, we have tried to show how contemporary approaches and methodologies take into account di¤erent (social and linguistic) factors in order to explain contact-induced language change. Multiple causation – a generally accepted phenomenon in the field – identifies both internally motivated changes and contact-induced processes, but the role played by each process and their precise relationship to each other is not always clear. This has led us to favor a multifaceted methodology and a multi-model approach to explaining contact-induced language change. Finally, the studies presented here argue for caution in proposing explanations of contact-induced language changes, both in historical situations, since limited linguistic or sociohistorical knowledge is available, and in contemporary situations, where to date very few social factors have been taken into account (but see Le´glise and Chamoreau, to appear).

References Campbell, L. and Muntzel, M. 1989. The structural consequences of language death. In Investigating Obsolescence: Studies in Language Contraction and Death, N. Dorian (ed.), 181–196. Cambridge: Cambridge University Press. Chamoreau, C., Estrada Ferna´ndez, Z., and Lastra, Y. (eds). 2010. A New Look at Language Contact in Amerindian Languages. Munich: Lincom Europa.

14

Claudine Chamoreau and Isabelle Le´glise

Chamoreau, C. and Goury, L. (eds). In press. Contact de langues et changement linguistique. Paris: CNRS Editions. Clairis, C. 1991. Le processus de disparition des langues. La Linguistique 27.2: 3– 14. Dorian, N. 1981. Language Death: The Life Cycle of a Scottish Gaelic Dialect. Philadelphia: University of Pennsylvania Press. Faarlund, J. T. 1990. Syntactic Change. Toward a Theory of Historical Syntax. Berlin: Mouton de Gruyter. Ferna´ndez Garay, A. 1998. El Tehuelche. Una lengua en vı´as de extincio´n. Valdivia: Universidad Austral de Chile. Hage`ge, C. 1993. The Language Builder. Amsterdam: John Benjamins. Harris, A. and Campbell, L. 1995. Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Heine, B. 1997. Cognitive Foundations of Grammar. Oxford: Oxford University Press. Heine, B. 2006. Contact-induced word order change without word order change. In Working papers in Multilingualism. Arbeiten zur Mehrspra¨chigkeit. 76. Hamburg: University of Hamburg. Heine, B. and Kuteva, T. 2003. On contact-induced grammaticalization. Studies in Language, 27.3: 529–572. Heine, B. and Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Heine, B. and Kuteva, T. 2008. The explanatory value of grammaticalization. In Linguistic Universals and Language Change, J. Good (ed.), 215–230. Oxford: Oxford University Press. Hill, J. 1989. The social functions of relativization in obsolescent and nonobsolescent languages. In Investigating Obsolescence: Studies in Language Contraction and Death, N. Dorian (ed.), 149–164. Cambridge: Cambridge University Press. Jacobsen, W. H. 1980. Inclusive/exclusive: A Di¤used Pronominal Category in Native Western North America. In Papers from the Parasession on Pronouns and Anaphora, J. Kreiman and A. E. Ojeda (eds), 204–230. Chicago: University of Chicago Press. Johanson, L. 2002. Contact-induced change in a code-copying framework. In Language Change. The Interplay of Internal, External and Extra-Linguistic Factors, M. C. Jones and E. Esch (eds), 285–313. Berlin: Mouton de Gruyter. Kriegel, S. (ed.). 2003. Grammaticalisation et re´analyse. Approches de la variation cre´ole et francaise. Paris: CNRS Editions. Lass, R. 1980. On Explaining Language Change. Cambridge: Cambridge University Press. Le´glise, I. and Chamoreau, C. In press. The Interplay of Variation and Change in Contact Settings – Morphosyntactic Studies. Amsterdam: John Benjamins.

A multi-model approach to contact-induced language change

15

Matras, Y. 2007. Socio-cultural and typological factors in contact-induced change. Plenary talk presented at the ‘Language Contact and Morphosyntactic variation and change’ workshop, Paris, 20–24 September. Matras, Y. 2007b. The borrowability of structural categories. In Grammatical Borrowing in Cross-Linguistic Perspective, Y. Matras and J. Sakel (eds), 31– 73. Berlin: Mouton de Gruyter. Meillet, A. 1982 [1906]. Comment les mots changent de sens. In Linguistique historique et linguistique ge´ne´rale, 230–271. Paris/Geneva: Champion-Slatkine. Meillet, A. 1982 [1912]. L’e´volution des formes grammaticales. In Linguistique historique et linguistique ge´ne´rale, 131–148. Paris/Geneva: Champion-Slatkine. Peyraube, A. 2002. L’e´volution des structures grammaticales. Langages 146: 46– 58. Romaine, S. 1989. Bilingualism. Oxford: Basil Blackwell. Sasse, H.-J. 1990. Theory of Language Death and Language Decay and Contactinduced Change: Similarities and Di¤erences. Arbeitspapier No. 12. Cologne: Institut fu¨r Sprachwissenschaft, University of Cologne. Thomason, S. G. 2000. On the unpredictability of contact e¤ects. Estudios de Sociolingu¨ı´stica 1.1: 173–182. Thomason, S. G. 2001. Language Contact: An Introduction. Edinburgh: Edinburgh University Press. Thomason, S. G. 2007. On internally- and externally- motivated morphosyntactic change in contact situations (and how to tell which is which). Plenary talk presented at the ‘Language Contact and Morphosyntactic variation and change’ workshop, Paris, 20–24 September. Thomason, S. G. and Everett, D. 2005. Pronoun borrowing. Berkeley Linguistics Society 27: 301–315. Thomason, S. G. and Kaufman, T. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley: University of California Press. Tsitsipis, L. 1989. Skewed performance and full performance in language obsolescence: the case of an Albanian variety. In Investigating Obsolescence: Studies in Language Contraction and Death, N. Dorian (ed.), 117–137. Cambridge: Cambridge University Press. Winford, D. 2003. An Introduction to Contact Linguistics. Oxford: Blackwell.

An activity-oriented approach to contact-induced language change1 Yaron Matras 1. Introduction Mechanisms of language change have usually been categorized in terms of either their structural or their societal properties. At the structural level, a well-established distinction is that between loans or transfers of concrete phonological shapes (or linguistic matter), and restructuring, replication, or calques of form-meaning alignments, constructions, or patterns (see Weinreich 1953; Haugen 1956; Heath 1984; Matras and Sakel 2007). Drawing on earlier work (Haase 1991; Nau 1995; Matras 1998b), Heine and Kuteva (2005) point out that pattern-replication or grammatical calques can often be analyzed as cases of language-internal grammaticalization triggered or inspired by a model construction in the contact language. This suggests a two-dimensional process: the first dimension involves the creative formation of new structures and categories, while the second involves the motivation to set such creative processes in motion. In much of the literature on language change, these two dimensions are understood as ‘‘language-internal’’ and ‘‘language-external’’ respectively. A further point of interest to structural approaches is the relative likelihood of borrowing of individual structural forms and categories (see for example Moravcsik 1978; Thomason and Kaufman 1988; van Hout and Muysken 1994; Field 2002). Particular attention has been given to the status of bound and unbound morphemes, to inflectional and agglutinative morphology, and to paradigmaticity as factors that may facilitate or 1. Earlier versions of this paper were presented at the workshop on ‘‘Language contact and morphosyntactic variation and change’’ in Paris, and as seminar presentations at the Research Centre for Linguistic Typology at La Trobe University, Melbourne; Australian National University, Canberra; University of Sydney; the Hebrew University of Jerusalem; and Charles University, Prague. I am grateful to the participants and audiences for inspiring questions and comments. For a full discussion of the argument and some of the examples see also Matras (2009).

18

Yaron Matras

constrain borrowing. It has been suggested that ease of formal integration will contribute to the likelihood of borrowing. More recently, sample-based surveys have attributed a semantic-pragmatic motivation to borrowability and have postulated meaningful hierarchies among paradigm values that are susceptible to borrowing (cf. Matras 1998a, 2007). Social mechanisms of contact-induced change are often described along the lines of Thomason’s (2001) typology of interference mechanisms. A distinction is made between language shift and language maintenance, language learning, and deliberate language creation (cf. Thomason and Kaufman 1988; see also Winford 2003). Social settings seem crucial to the definition of at least some types of contact phenomena. One example is pidginization, which is widely understood as being rooted in restricted communication, drawing selectively on the lexical structures of a superstrate language or lingua franca in a multiethnic setting.2 The result is described variably as the use of a superstrate lexicon either with substrate grammar or with makeshift grammar. Bakker’s (1996, 1997) notion of ‘‘language intertwining’’ also relies on a particular kind of social context, one in which ethnic identity is being negotiated in a situation of linguisticethnic hybridity. For such situations, Bakker predicts an outcome in the form of a mixed language that draws its lexicon from one source and its grammar from another. More generally, it has been suggested that prolonged, reciprocal bilingualism, or language ‘‘equilibrium,’’ is likely to lead to the emergence of pattern-similarities among languages, while diglossia and dominance are more likely to result in the borrowing of word-forms (cf. Aikhenvald 2002: 265¤.). However, it is not always obvious that a direct correlation can be drawn between the type of social setting and the structural outcomes of contact. It may be the case that in the lexical domain, items such as place-names, field-names, and agricultural terms are more likely to be carried over from a substrate to a superstrate language during language shift, while lexical transfers from a superstrate language will tend to cover technical innovations and trade vocabulary. Low German in East Friesland, for example, contains (substrate) Frisian terms connected with agriculture, the sea, dikes, and drainage, and (superstrate) Dutch vocabulary for trade, commerce, and engineering (cf. Remmers 1997). But the loss of a definite article is common both to Russian learners’ varieties of English and to language maintenance in Romani in contact with (superstrate) Russian, while the emergence of a new definite article is common both to language maintenance in Sorbian in 2. Some Pidgins, such as Russo-Norsk, apparently involve more or less equal input from two participating languages.

An activity-oriented approach to contact-induced language change

19

contact with (superstrate) German and to convergence – possibly through language shift, perhaps through a prolonged ‘‘equilibrium’’ – in the context of the Balkan languages Romanian, Bulgarian, and Macedonian. Two principal factors are often said to motivate contact-induced change: ‘‘gaps’’ in the recipient system, on the structural side, and the overall ‘‘social prestige’’ of the donor system, on the societal side. ‘‘Gaps’’ are understood as asymmetries in the structural representation of semantic-pragmatic functions in the two languages in contact. At the structural level, the notion of ‘‘gaps’’ implies that multilingual speakers aim at availing themselves of a uniform system of form-function mapping across their various languages. This does not, however, provide a direct explanation for the borrowing of word-forms to replace forms that had already existed in the language prior to contact, as in the wholesale replacement of the system of discourse markers (see Matras 1998a; Salmons 1990). The ‘‘prestige’’ assumption in turn fails to explain why some structural categories, such as connectors, are more likely to be borrowed than others, such as personal pronouns. If a language is generally ‘‘prestigious,’’ why should its prestige be more easily flagged through connectors than through personal pronouns? ‘‘Prestige’’ is therefore better understood as a license to employ forms and structures from a language which, as a result of unidirectional bilingualism, is more widely used in the community and which, thanks to institutional support and its role in the public domain, is subject to tighter normative control. The language of a monolingual majority that is used in the public domain and is protected by institutions (such as literacy) will often be regarded as the ‘‘dominant’’ language. The non-dominant, smaller, or minority language is less likely to enjoy institutional protection and literacy. Its speakers are therefore less likely to be aware of norms and rules and to enforce them; in fact, they are more likely to be multilingual and tolerant of structural variation. Asymmetry in the social roles of the languages (that is, in their ‘‘prestige’’) may therefore determine the direction of change, but it does not necessarily explain the motivation for structural change. With these considerations in mind, we are left with the question of how best to explore the link between social reality and the role of structural factors in contact-induced change.

2. Multilingual conversation as repertoire negotiation The answer lies, I propose, in understanding the communicative acts that multilingual speakers engage in, and in examining the value that particular

20

Yaron Matras

linguistic constructions have in allowing speakers to achieve their communicative goals. Like any other process of language change, contact-induced change begins with innovations introduced by an individual speaker as part of communicative interaction. My assumption is that such innovations are not arbitrary, but follow goal-oriented tasks. This is based first of all on a view of language as the practice of communicative interaction and of grammatical categories as triggers and operators of language processing tasks involved in communication. According to this approach, the selection of structures by the speaker is not random, but defined by the linguistic task-schema that the speaker wishes to carry out. This, in turn, is subordinate to the goal-oriented activity that the speaker pursues by means of verbal communication, organized at the level of discourse.3 Becoming ‘‘bilingual’’ is an extension of an individual’s contexts of interaction, as a result of that individual’s repertoire of communicative structures. Multilingualism from infancy means exposure to a complex repertoire. This requires gradually sorting out the sets of contexts and contextual conditions under which various sets of structures from within this repertoire are considered appropriate. Thus, even bilinguals from birth do not acquire two language ‘‘systems’’ natively; rather, they acquire a repertoire of linguistic structures and forms, and are left to gradually master the rules on appropriate, context-bound selection of one form over another as part of a process of linguistic socialization (see Lanza 1997). Some contexts allow greater flexibility of choices – what Grosjean (2001, 2008) termed the ‘‘bilingual mode.’’ These are the contexts in which bilinguals can make the most e¤ective use of their full repertoire, exploiting nuances as well as contrasts between variants of equivalent or near-equivalent meaning. Other sets of contexts are more exclusive of the selection of items and groups of items within the repertoire. The existence of selection rules as part of the bilingual’s communicative competence triggers a series of associations between a particular subset of structures and interaction context set A, between another and interaction context B, and so on. This association is what we identify as our socially constructed notion of a ‘‘language’’ or a ‘‘language system.’’ It is thanks to this socially broadcast notion that bilingual children learn, around the age 3. This view of language is inspired by a range of theoretical approaches to communication and discourse (e.g. Gumperz 1980; Sacks, Scheglo¤, and Je¤erson 1974; Rehbein 1977; Ehlich 2007), as well as to speech production (Green 1998).

An activity-oriented approach to contact-induced language change

21

of three, that they speak two ‘‘languages’’; until then, their use of wordforms and constructions is governed by a prolonged process of trial and error, usually unaccompanied by any explicit analytical labeling or other overt classification of the elements of their repertoire. Such an association between structure and set of interaction contexts does not necessarily exist for each and every element of the linguistic repertoire. German-English bilinguals, for example, accept that their repertoire contains only one single word-form for concepts such as internet, download, computer (subjected of course to embedding in di¤erent phonological and morphosyntactic environments). Such category-specific inseparability of the two linguistic subsets in a bilingual’s repertoire is part of the definition of ‘‘borrowing’’ which I pursue in this paper. The definition can be extended to those constellations where a structure continues to di¤use, reaching a monolingual population that has never experienced the need to interact in a new set of contexts. While this aspect of borrowing – di¤usion to monolinguals – is a property of some borrowing situations, such as the German ‘Internet,’ ‘Computer,’ it is not necessarily typical of all. How does borrowing come about? And how is it linked to other contact phenomena? Communication in a language contact setting is the product of the interplay of two primary factors (Figure 1): loyalty to a set of norms that regulate the context-bound selection of elements from the repertoire, and a wish to be able to exploit the repertoire in its entirety irrespective of situational constraints. The balance between these two factors is determined by a need to remove hurdles that stand in the way of e‰cient communication. When loyalty prevails in a strict manner, then ‘‘interference’’ or compromises are likely to be minimal. But when the wish to exploit the full repertoire is given some leeway, then strict context-bound separation of repertoire components might be compromised. Individual words that are usually reserved for interaction in context set A might, for example, be employed (‘‘inserted’’) also in interaction in context set B. Second-language learners might draw on the phonology of their native language while communicating in a second language, bilingual children might employ constructions from one language that are not usually used in the chosen language of conversation, and adult bilinguals might insert discourse markers from one language when communicating in another. All this suggests that multilingual speakers do not ‘‘block’’ or ‘‘switch o¤ ’’ one of their languages when communicating in another, but that they have the full, complex linguistic repertoire at their disposal at all times.

22

Yaron Matras

Figure 1. The interplay of factors in communication in language contact settings

Language contact phenomena are seen in the model outlined here as the outcome of function-driven choices through which speakers license themselves, while interacting in a context of type B, to select a structure (word-form, construction, meaning, phonological features, and so on), despite its association primarily with interaction context set A. When claiming that choices are function-driven, I am not suggesting that selection of A-structures in B-contexts is necessarily always conscious, deliberate, or strategic. Instead, I propose that contact phenomena are arranged on a continuum, from those that are in fact not at all voluntary, indeed even counter-strategic in their origin, to those that are conscious and deliberate. All, however, are functional in the sense that they are the product of language-processing in goal-oriented communicative interaction. The susceptibility of certain structural categories to contact-related change is therefore not accidental, but inherently bound up with the function that those categories have and the way they support language processing in discourse. Contact phenomena are in this respect seen as enabling rather than interfering with communicative activity.

An activity-oriented approach to contact-induced language change

23

My principal claim in this article, then, is that innovative strategies occur in pursuit of specific communicative goals. The challenge that I take on is to identify the connection between spontaneous innovations in discourse, through to the emergence of stable variants in communication in multilingual settings, and on to processes of language change. I will focus on four principal types of innovation: the insertion of lexical wordforms and lexical borrowing, replication of patterns or constructions, fusion of grammatical operators, and playful or ‘‘theatrical’’ mixing. The compilation of data from a trilingual child, from adult bilingual speech, from stable multilingual settings, and from cases of contact-induced language change will illustrate the close a‰nity between spontaneous innovations and long-term change, and show that all types of innovation strategies are already available to the very young bilingual in the very early steps of managing a complex linguistic repertoire in a multilingual setting. 3. From lexical insertions to lexical borrowing Consider an example from the speech of a trilingual from infancy, whose home languages are German (with the mother) and Hebrew (with the father), while English is the language of the environment, including school (from Matras 2009): (1) German; age 7:6, when reminded of a past event Da war ich noch in year one deic was.1sg I still in ‘I was still in year one then.’ In example (1), the child is using events from school life as points of reference. The school is an English-speaking environment, key elements of which are treated as unique referents, or what Backus (1996) calls ‘‘specific’’ entities. Although the child is in principle able to translate or paraphrase the concept year one, use of the English form amounts to an activation of the world of associations represented by the original term. The insertion of the English term thus acts as a discourse device that supports the transposition of imagery of the original setting into the ongoing conversation. The uniqueness of the English term as part of the regulated vocabulary associated with the English-speaking school environment gives year one the status of institutional terminology, for which translations are not appreciated as equivalents because they are dissociated from the original setting.

24

Yaron Matras

Institutional terminology is commonly involved in bilingual insertions in the speech of adults too. There is, in other words, nothing specific to the child acquisition context that promotes the insertion of institutional terminology. Consider example (2), from the speech of a (Lovari) RomaniGerman bilingual (Matras, fieldwork): (2) Romani (Lovari)-German bilingual; biographical narration Aj akana, obwohl kadka meres ke muljas tuke and now although here die.2sg because died.3sg. 2sg.dat varekon, hacˇares, du bist total fertig, tu si te somebody understand.2sg you are totally devastated 2sg is comp zˇas inke te des tu gindo kaj te praxov go.2sg still comp give.2sg 2sg thought where comp bury.1sg les, kudka si te zˇav, Bestattungsinstitut, ehm/ pa/ pa/ pa 3sg.obl here is comp go.1sg funeral home on on on Meldeamt, eh Geb/ Sterbeurkunde, registration o‰ce bir death certificate ‘And now, although you’re dying here because one of your people died, you understand, you’re totally devastated, you still have to go and think about where should I bury him, I have to go there, funeral home, ehm/ to/ to/ to the registration o‰ce, eh birth/ death certificate.’ The speaker inserts German terminology to describe institutions and institutional activities associated with the burial of a relative in Germany, where she lives. From a strictly formal perspective one might regard these insertions as gap-fillers, since they have no Romani equivalent. However, it is precisely the fact that no Romani equivalent is created by speakers that demands our attention. Speakers could calque or paraphrase or otherwise create compounds or terms that would allow them to describe the relevant concepts without having to resort to word-forms that are derived from a non-Romani interaction context. But in this case, the e¤ect of the association evoked precisely with the non-Romani interaction context is purposeful and fills a function. It is a powerful discourse-level tool in emphasizing the contrast between the intimate feeling of mourning and distress, which engulfs the individual and her family following the death of a loved one, and the anonymity of bureaucratic errands carried out in an indi¤erent and potentially hostile environment. The replication of the original German terminology is thus not only a matter of convenience, it is also instrumental to the overall message conveyed by the speaker.

An activity-oriented approach to contact-induced language change

25

Lexical insertions of the types illustrated in these two examples appear to operate precisely on the ambiguity of the context-separation of subcomponents of the speaker’s overall linguistic repertoire. On the one hand, the insertion of words from a di¤erent ‘‘language’’ appears to defy the demarcation of sub-components of the speaker’s overall linguistic repertoire and so to suggest that the speaker is at liberty to make full use of the entire repertoire irrespective of any situational or contextual constraints. On the other hand, it is precisely the association of these particular insertions with another set of interaction settings – that belonging to the public and institutional domain, outside the home, and so on – that creates a special e¤ect in the ongoing discourse, that of authentication and contrast with the more intimate sphere of the chosen language of the ongoing interaction, an e¤ect the speaker exploits for stylistic purposes. This special e¤ect of lexical insertions may become eroded when a word becomes a regular part of the language into which it is inserted, or when it is adopted in monolingual contexts and the contrast of associations with di¤erent interaction settings is thus lost. Nonetheless, special e¤ects may arguably still be detected even following the stabilization of loan vocabulary. The diglossic origin of the contrast among the famous English lexical pairs pig-pork, sheep-mutton, cow-beef, chicken-poultry in peasant (Saxon) English and aristocratic French is still apparent in their domain specialization as livestock versus culinary dishes. Likewise, most German speakers who use English loans in domains such as computer technology, as in example (3), media, and management are aware of their English origins and associations with international communication settings: (3) German: Lexical borrowing Ich muss es vom Internet downloaden I must it from.DEF internet download ‘I have to download it from the internet.’ The discourse-strategic insertion of lexical items pertaining to institutional and other cultural or social domain-specific terminology fits nicely, of course, with the overall picture of lexical borrowing. At the top of the list of typical lexical borrowings we find terms for institutions, specialized instruments, culture-specific practices, and innovations. Most of those are represented by nouns, which appear universally to be the most frequently borrowed word class. Statistics for Japanese (4) and Romani (5) provide an example.4 4. See the two sources cited here for details on the corpora and the mode of calculation.

26

Yaron Matras

(4) Percentage of English loans in Japanese by selected, specialized semantic domains (from Loveday 1996) computer (99%) > broadcasting (82%) > journalism, marketing (75%) > engineering (67%) > flowers (52%) > vegetables (35%) > animals (24%) > colors (9%) (5) Percentage of loanwords by semantic domain in Selice Romani (Elsˇ´ık 2009): household, modern world, agriculture (over 90%) > clothing, warfare (over 80%) > animals, social and political relations, the physical world (over 70%) > religion and belief, speech and language, law, technology, food and drink (over 60%) > time, the body, motion, perception, emotion, cognition, values (over 50%) > spatial relations (over 40%) > quantity, kinship (over 30%) Examining both the synchronic, discourse-based behavior of bilinguals and diachronic data on contact-induced language change in an integrated approach, we are in a position to explain both some structural facts of lexical borrowing – the predominance of nouns among borrowed word classes – and the semantic distribution of lexical borrowings: The roots of lexical borrowings are in bilinguals’ attempts to integrate into an ongoing interaction concepts associated primarily with an environment in which a di¤erent language is spoken. The need to do so arises in particular with reference to unique structures of that environment, such as specific practices or names of specific institutions that are not replicated in the activity domain or community in which the language of the ongoing interaction is spoken. The liberty to draw on such insertions is given in turn only in those interaction settings where the speaker may resort to the bilingual mode, that is, where the other participants are also bilingual and where the language of the interaction is not tightly regulated but at least some flexibility is allowed for. In situations of unidirectional bilingualism, this limits the opportunities for using insertions to interactions among members of the bilingual group only. The insertion serves to activate knowledge of the original set of interaction contexts in which the word is normally used. As such, it has a strategic e¤ect on the structuring of the discourse, apart from facilitating the speaker’s access to concepts by licensing the activation of words and terms from the entire repertoire, irrespective of the interaction context in which they originally appear. In due course, some bilingual insertions may find their way into a monolingual population,

An activity-oriented approach to contact-induced language change

27

carried by a group of innovators whose terms, concepts, or simply stylistic choices are being adopted by others in the speech community. 4. Replication of patterns / constructions Convergence of form-function mapping, semantic meaning representation, constructions, or ‘‘patterns’’ is sometimes regarded as a prolonged process involving not just gradual dissemination within the speech community but also gradual evolution or grammaticalization of the construction itself. In fact, while there is no disputing that a time factor is crucial to the propagation of an innovation throughout the speech community, the emergence of an innovation may certainly be a spontaneous act. Consider the following example, from the German-Hebrew bilingual child. Around the age of four, the child acquires a new construction in German – the politeness term of address Sie. The German second-person polite form Sie is identical to the 3pl pronoun sie, and carries the same 3pl agreement marker on the verb. The context in which the child acquires this construction is a game which he plays with his mother, in which the child is a storekeeper and the mother is a customer coming to the shop, who addresses the shopkeeper in the polite form when inquiring about certain products (haben Sie X ? ‘do you.polite have X?’). The child’s acquaintance with the German politeness form is, at this stage, limited to this particular context. Strictly speaking, he does not acquire a politeness marker as such, but a construction that is employed in a particular slot within the predefined pattern of speech activities that characterizes the game ‘‘shop.’’ By acquiring this new construction, the child has extended his overall communicative repertoire. In this case, this is a more accurate description than suggesting that he has learned a new ‘‘structure,’’ since he is already familiar with the form of the 3pl pronoun and agreement marker, and it is only this use of the structure to refer to the addressee under strictly-defined communicative circumstances that is novel to him. When the child is spending time with his father, a similar game is played in Hebrew. That is the context of example (6). (6) a.

Hebrew; age 4:1, during role-play as a customer addressing a grocer: yesˇ lahem tapuxı´m? there.is to.3pl apples [intended] ‘Do you have apples?’ [expressed] ‘Do they have apples?’

28

Yaron Matras

b.

German model construction for polite form of address: haben Sie A¨pfel? have.3pl you.polite/3pl apples ‘Do you have apples?’

Note that the ‘‘generic’’ shop-game, from the child’s perspective, is played with the mother, and that it is in her household (the parents live in two separate households) that the child has a range of accessories, including a toy counter and till, to facilitate the game. The shop-game in the father’s household is thus a ‘‘replica.’’ Having enriched his linguistic-communicative repertoire as part of mastering the shop-game, the child is eager to repeat the acquired pattern of activity associated with it. This repetition of the activity pattern may be regarded as the child’s communicative goal in the interaction (see Figure 2). It includes the organization of the question which the child, now playing the role of the customer, puts to the storekeeper, this time the Hebrew-speaking father. For this particular task within the overall interaction pattern, the child has recently acquired a specific task-e¤ective construction. However, Hebrew lacks a politeness pronoun of the kind found in German; completion of the task or parts of it in German would be against the rules of compliance with the selection of context-appropriate (Hebrew) word-forms (and might therefore be rejected by the interlocutor, or interpreted as an attempt to create a special conversational e¤ect). Aware of the constraints on the interaction context in the father’s household, namely the need to choose overt word-forms that conform with the context – or ‘‘Hebrew’’ word-forms – the child is keen to comply by selecting a construction that is contextually appropriate. At the same time, the child is keen to communicate most e¤ectively and to exploit the functionality of new constructions in his overall linguistic repertoire. The two seemingly conflicting motivations are reconciled through a creative procedure. The child picks up a single – albeit ‘‘pivotal’’ – feature of the German construction, namely the use of the 3pl. He matches this feature to a counterpart structure in Hebrew, replicating the German construction by employing a Hebrew possessive construction in the 3pl (Figure 2). The combination renders the construction both contextually appropriate and seemingly e¤ective for the communicative task that has been selected. Its actual e¤ectiveness will of course depend on the ability and willingness of the interlocutor to understand and accept the meaning of the new construction. Its chances of becoming propagated within the speech community and so to lead to language change will in turn depend on the inno-

An activity-oriented approach to contact-induced language change

29

Figure 2. Construction replication through ‘‘pivot-matching’

vator’s potential to influence others, on the degree of normative control on language that is exercised in the speech community, and of course on the existence of a community of interlocutors. In the case of the present example, the construction may be understood by the interlocutor, but there are no Hebrew-speaking peers among whom the innovation can be propagated, and parental intervention in the child’s speech is regular and is likely to prevent even the innovator himself from adopting the construction on a regular basis. Nevertheless, while the propagation chances will vary considerably among speech communities, the creative process by which pattern-replication first appears and the discourse-functional motivation behind it can be regarded as similar in principle. Constructions are selected as advantageous and worthy of replication through pivot-matching when they are perceived as particularly taske¤ective. We saw this in example (6), where the child’s selection of the ‘‘politeness construction’’ was motivated by a recently acquired rule to use the ‘‘politeness marker’’ in a particular position of the interaction scheme of the role-play ‘‘shop.’’ But task-e¤ectiveness can also be associated with simple task routines where the selection of a particular construction is motivated primarily by the fact that it is the most readily available, or in the

30

Yaron Matras

absence of secure knowledge about an alternative appropriate construction. Example (7) shows how an adult native speaker of German, whose English is considered fluent, resorts to pattern-replication to activate German construction patterns while giving a formal interview to British television. (7) German/English bilingual, in a British television interview: At the border in England, were by the custom/ They have investigated this car very very eh/ eh/ thoroughly and they have removed the panels from the doors, the panels from the luggage room. So-called subject-verb inversion following an occupied first constituent position in the sentence is replicated in English by following the same word order. It is a specialized word order pattern, applied in German when a constituent other than the subject is employed to create the perspective of the sentence. The speaker is over-di¤erentiating a semanticsyntactic context here, attempting to select a very specialized construction for a very specific task. In what follows, verbs appear in the perfect tense with have auxiliaries, as they would in German in the description of simple events whose outcome does not necessarily extend into the present. Finally, the construction of the lexical term luggage room for boot or trunk is a replication of German Ko¤erraum. Like the child’s pattern replication in (6), these examples too are spontaneous, triggered by an appreciation of task-e¤ective constructions for the given communicative tasks, coupled with (or constrained by) an appreciation of the need to comply with the selection of word-forms from a particular inventory or sub-component of the linguistic repertoire, word-forms that would be understood and accepted by the interlocutor in the ongoing interaction context. Here too, the potential for these makeshift replica constructions to become propagated and stabilized within a larger speech community is small, indeed minimal; but one might just imagine the potential in a community consisting to a large degree of second-language learners for whom the target language becomes the language of choice, in situations of ongoing language shift. There are also numerous observable cases of pattern-replication in smaller or minority languages, brought about by bilinguals imitating constructions from the neighboring dominant language. Here we have, in other words, not a situation of language learning nor one of language shift, but one of maintenance of a community language; nor do we have a linguistic equilibrium, but a case of clear diglossia and dominance. And yet the outcome is the replication of patterns or constructions, sometimes

An activity-oriented approach to contact-induced language change

31

on a massive scale. Domari, for instance, the archaic Indo-Aryan language spoken by a tiny ethnic minority in the Old City of Jerusalem, is in the process of generalizing a new possessive construction, modeled on the one found in colloquial Palestinian Arabic, its principal contact language. (8) a.

b.

‘‘Canonical’’ Domari by-im kuri father-1sg house Palestinian Arabic: be#t-o la-abu#-y house-3sg.m.poss to-father-1sg

c. New Domari construction: kury-os by-im-ki house-3sg.poss father-1sg-abl ‘my father’s house’ In this situation, a variant construction has emerged which has now become stabilized as a regular option, probably even the preferred option, by most speakers. We can assume that its roots were in a spontaneous innovation of the type seen in examples (6) and (7). The motivation for such an innovation will have been the frequency of use of the Arabic construction in interaction outside the Domari-speaking household. The need for tight control in producing correct Arabic constructions among Arab interlocutors will have contrasted with the relatively lenient attitude that speakers of the now moribund Domari have toward their own language, in which variation, flexibility, and mixture with Arabic are commonplace, adding yet another factor in the perception of the Arabic-based construction as more task-e¤ective. A process of language change that has come to its conclusion can be found in the dialect of Gulf Arabic spoken in the Iranian province of Khuzistan, especially in urban communities, where Persian has become the dominant language of institutions and the public domain (cf. Matras and Shabibi 2007). Here, the attributive construction involving nominals (or nominal-possessive construction) has fused with the adjectival attributive construction, following the Persian model. (9) a.

‘‘Canonical’’ Arabic t abaqa-t il-mustasˇfa i-a#niyya ˙floor-constr.f def-hospital def-second.f

32

Yaron Matras

b.

Persian taba'a-ye dovom-e bımaresta# # n floor-attr second-attr hospital

c. Khuzistani Arabic t abaqa-t i-a#niyya-t il-bımaresta# # n ˙floor-constr.f def-second.f-constr.f def-hospital ‘The second floor of the hospital building’ Note that the canonical Arabic construction expresses nominal attribution by attaching a so-called construct ending (with feminine nouns only) to the head, and by attaching a definite article to the following possessor noun. Adjectival attribution is expressed by the postpositioning of the adjective and its agreement with the head in gender, number, and definiteness (a head that is determined through nominal attribution in the possessive construction being considered as definite). In Persian, both types of attribution are represented by the positioning of an attributive particle -(y)e between the head and the attribute (whether nominal or adjectival). Khuzistani adopts the Persian model, and generalizes one single attributive construction to both nominal and adjectival attributes. Moreover, it draws on the linear combination of a construct state ending (with feminine nouns) on the head and the definite article of the following dependent attribute and equates those with the Persian attributive marker, which in Persian is arguably the pivotal feature of the attributive construction. This combination is then transferred to the Khuzistani adjectival construction as well. The result is a one-to-one or isomorphic correspondence between the Persian and the Khuzistani Arabic constructions. Finally, Macedonian Turkish has undergone a series of radical changes to its overall typology of clause linking, which have given rise to devices linking finite clauses of the type that is common in the languages of Europe in general and surrounding languages of the Balkans, Macedonian, Albanian, and Greek in particular (cf. Matras 2004; Matras and Tufan 2007). (10) Macedonian Turkish relative clause adam ne gel-di man rel come-3sg.past ‘The man who came’ The language has developed postposed, finite relative clauses as well as a relativizer, modeled on the interrogative ne ‘what,’ much like its Mace-

An activity-oriented approach to contact-induced language change

33

donian counterpart sˇto; and it has done away with Turkish gerundial constructions of the type gel-en adam ‘the man who came.’

5. Selection malfunctions Above it was proposed that bilingual speakers do not ‘‘switch o¤ ’’ one of their language ‘‘systems’’ during monolingual conversation, but that the entire repertoire of linguistic structures remains available to them in each and every communicative interaction setting. Following Green (1998) and others we might assume that the production of both lexemes and constructions undergoes a monitoring procedure as part of which those structures that comply with the constraints on context-appropriateness (structures that are expected and so are likely to be accepted and understood by the interlocutor) are selected, while those that are not deemed contextually appropriate are blocked. Example (11) below shows how this control and selection mechanism, which is already operational in a multilingual child as young as two, may occasionally malfunction. As a result, a structure is produced that is functionally correct in terms of its semantic-pragmatic value, but contextually inappropriate as it belongs to the ‘‘wrong language.’’ In this example, the child is known to have mastered the production of a series of di¤erent clause combining structures in both Hebrew and German, including the contrastive combination with elements equivalent to English but. But in (11), following a three-week holiday in which German was the only everyday language of interaction, he produces the German conjunction aber during an interaction in Hebrew: (11) Hebrew; age 2:3, first few days in the father’s care after returning from a three-week holiday in Germany; inspecting the shell of a snail in the garden: ba´yit sˇel xilazo´n aber e´yn xilazo´n bifnı´m house of snail but is-no snail inside ‘A snail-shell, but there is no snail inside.’ Clearly, whatever di‰culty the child is having in retrieving the correct Hebrew conjunction is not a¤ecting the semantics of the clause linking device as such, which is correctly produced drawing on the German equivalent, nor is it a¤ecting the retrieval of other Hebrew structures, be they content words or grammatical items. The absence of any hesitation of correction indicates that the child is not aware of the ‘‘error’’ in the choice

34

Yaron Matras

of word, and indeed the interlocutor in this case does not intervene but accepts the construction in its entirety; the construction is thus viable from a communicative viewpoint, at least in this instance. What is the motivation behind the child’s selective failure, around the contrastive conjunction, to control and select the appropriate word-form? Confusion on the basis of the structural similarity between the two conjunctions – German a´ber and Hebrew ava´l – cannot be ruled out entirely, although the di¤erence in prosody appears to make the distinction between the two quite salient. More likely, the source of the malfunction can be attributed to the specific semantic-pragmatic value of the connector. The function of the contrastive conjunction is to signal a break in the expected propositional causal chain (Rudolph 1996). It is inserted by the speaker in anticipation of a disharmony between the expectations of the interlocutor about the subsequent course of the proposition and the speaker’s own intentions concerning the exposition of the proposition. Moreover, it constitutes a direct intervention by the speaker with the interlocutor’s ongoing processing of the proposition. The clash of expectations and the speaker’s e¤ort to intervene and redirect the listener’s processing course constitute a tense moment in the interaction, one during which the speaker’s authority is at stake and a concentrated e¤ort on the part of the speaker is called for in order to maintain the listener’s confidence and possibly even the floor. Elsewhere (Matras 1998a, 2000) I have argued that the mental e¤ort that is required in order to solve this tension comes at the expense of the e¤ort that is directed toward the smooth and continuous operation of the selection and inhibition mechanism, which controls the selection of contextappropriate forms from the multilingual repertoire. There is therefore a direct correlation between ‘‘high-tension’’ mental processing operations such as contrast and other argumentative connectors, and the likelihood of malfunction of the selection and inhibition mechanism, and therefore a direct correlation between such operations and bilingual speech production errors where the functionally correct form is selected, but from the ‘‘wrong language’’ (that is, from the contextually non-appropriate component of the linguistic repertoire). When such malfunctions occur, they tend to be directed towards a language that has recently been activated on a routine basis and therefore constitutes the default fall-back option for routine task-management of the relevant processing operation. For the young bilingual child who has just returned from a three-week stay in Germany, this ‘‘pragmatically dominant language’’ is German. Not just children are prone to selection malfunctions of this kind. Consider the following examples, all recorded from bilingual adults in a multi-

An activity-oriented approach to contact-induced language change

35

lingual setting, all involving a similar class of operators. In (12), a group of Hebrew-English bilinguals is speaking Hebrew at a restaurant. They are approached by the waiter, who takes their order in English. One person from the group then adds an item to the order, choosing the Hebrew contrastive connector instead of English but. (12) Hebrew/English bilingual at a (Chinese) restaurant in England: . . . and one Won Ton soup ava´l/ eh/ the vegetarian one. but The hesitation and seeming self-repair that follows indicates that the speaker has become aware of her production of an incorrect form and, moreover, that the form that had been produced was indeed not intended. Selection malfunctions are counter-strategic; they do not serve a goal in shaping or influencing the message key for any stylistic special e¤ect. They are nevertheless functional in the sense that they are non-random in their distribution and direction; in other words, they can be systematically accounted for and explained through a model of multilingual language processing and speech production, as attempted above. The important thing to note here is that malfunctions are not motivated by ‘‘gaps,’’ either in the system itself or necessarily in the speaker’s command of the system; nor are they motivated by ‘‘prestige,’’ as there is no prestige gain to the speaker who confronts a Chinese waiter in England with a Hebrew conjunction, nor to the young child who fails to conform consistently with the selection constraints that operate in the context of his interactions with his father. (In the latter case a gap can be excluded when there is evidence that the child has used the appropriate Hebrew word or construction on previous occasions.) The fact that malfunctions tend to defy prestige constraints is perhaps best exemplified by (13), where a trained diplomat slips into his native language, Arabic, during a formal television interview. (13) Saudi Ambassador to the UK during a television interview: I would beg to say that yani/ the Kingdom is a very big territory. The slip is the failure to control the production of the Arabic discourse marker yani, which might be translated as ‘I mean’ or even ‘you see,’ and whose function is to grab the interlocutor’s attention and to make sure that it continues to be focused on the speaker’s turn and propositional content. Thus, yani has a somewhat similar tension potential to the contrastive marker, allowing the speaker to regulate roles in the interaction and intervene directly in the hearer-side processing of the ongoing discourse.

36

Yaron Matras

The following examples indicate the volatility in principle of the directionality of selection malfunctions. Above I referred to the ‘‘pragmatically dominant language’’ (see Matras 1998a) as the fall-back option for routine task constructions. In the previous two examples, the lapse in selection control happens to favor the speakers’ respective native languages. However, in (14) the speaker is a Polish native speaker residing in Germany. She is speaking German to two friends she is meeting up with in London, during her stay there on a three-week language course. (14) Polish/German bilingual, on ‘‘language holiday’’ in England: . . . bis auf/ bis auf die Tischdecken, because/ eh weil sie . . . ‘. . . except/ except for the tablecloth, because/ uh because it . . .’ The selection of English because during a portion of German conversation targets the language toward which the speaker has been directing her uppermost intellectual attention during the past weeks. Once again, we are dealing with an argumentative connector, one that is inserted in order to intervene with and influence the hearer’s course of processing propositions and deriving conclusions from them, and at the same time a connector that operates at the interactional level, announcing the speaker’s justification of a preceding statement; thus, because captures the speaker yet again in a position of potential vulnerability on the interaction plane. Example (15), from a German/Hebrew bilingual residing in England, underlines yet again the relevance of the pragmatically dominant language – the language in which routine tasks have most recently been handled – as the fall-back option. (15) German/Hebrew bilingual living in England: ani xosˇevet sˇe ze lo knesiya any more I think.sg.f that this neg church ‘I think that this is no longer a church.’ Here the selection malfunction targets an indefinite expression, which operates at the level of established presuppositions. At the pragmatic level of the interaction, indefinites serve to delegate to the listener the task of supplementing relevant information based on shared presuppositions. In the case of (15), the discontinuity signaled by the speaker through any more (in a negated phrase) presupposes the availability of information on an earlier state of a¤airs that is being discontinued. This information is not made explicit, however, and the hearer is expected to retrieve it from the context. By explicitly delegating to the listener this procedure of mentally

An activity-oriented approach to contact-induced language change

37

supplementing information, the speaker is once again intervening with hearer-side processing (beyond the mere default routine of supplying information to the listener). In so doing, the speaker puts him/herself in a position of vulnerability with respect to the listener’s potential discontent. We therefore find, once again, a link between high-tension constructions, distraction of the mental processing e¤ort, and weakening of the selection and inhibition mechanism, and as a consequence the selection of a functionally adequate but contextually non-appropriate structure. Finally, example (16) shows how a speaker of Ladino (Judeo-Spanish) living in Israel falls back on the pragmatically dominant language – here, Hebrew, the principal language of interaction outside the home – during a connectivity construction linking events into a consequential chain. (16) Judezmo (Ladino)/ Hebrew bilingual: a. S: Los eh/ mekomiyı´m, los lokales, eran relasiones midzˇores the uh locals the locals were relations better de los 'rexos ke vinieron de la turkı´a. from the Greeks who came from Turkey b.

Por ke los ke vinieron de Turkı´a eran ublixados because those who came from Turkey were obliged de tomar lavoros de los eh/ sitadinos/ siudadinos, si. to take jobs from the uh citizens citizens yes

c. H: Mhm, mhm. d. S: Az/ eh es/ entonses empeso´ la/ la kel/ la enemistad so-then uh so-then began the the that the rivalry la ma´s grande. the most great a. S: ‘[With] The uh/ locals, the locals, relations were better than [with] the Greeks who came from Turkey. b.

Because those who came from Turkey were obliged to take jobs from the uh/ citizens/ citizens, yes.

c. H: Mhm, mhm. d. S: So then/ uh/ th/ then the/ uh/ the greatest rivalry emerged.’ Note the speaker’s self-repair in segment d., which follows the slip into Hebrew (both Hebrew az and Ladino entonses have both a temporalsequential and a consequential meaning), confirming that the speaker is

38

Yaron Matras

not trying to avail himself of the contrast of languages for any stylistic purposes or other special e¤ect, but that he has genuinely lost control over the speech production mechanism around the relevant expression. What is the meaning of self-repairs of this kind? At first glance we might at the very least dismiss any chances of further repetition, let alone propagation, of this one-o¤ error, and so any chance of its stabilization as an integral part of the speaker’s inventory of expressions, forms, and constructions potentially selected during conversation in Ladino. However, Berk-Seligson (1986) in fact documents the exact same Hebrew-derived feature – az ‘and then,’ ‘and so’ – in the speech of other Israeli Ladinospeakers. It is quite clear that some selection malfunctions, such as those represented in examples (12)–(15), are unlikely to become propagated throughout a speech community and lead to language change even if they do happen to be repeated by the speaker, or even by another speaker. In all these settings the potential for a sector within the speech community to find the innovation advantageous for communication is virtually nonexistent. Nonetheless, this is not to say that the act, or rather event, of selection malfunction itself cannot lead to language change. Given a sector of individuals with similar bilingual skills and a similar repertoire, frequent occurrence of selection malfunctions targeting similar expressions or even sets of expressions, and lax normative control over performance in the relevant (recipient) language and tolerance of change, the targets of selection malfunctions may indeed become stabilized within a speech community. A prerequisite seems to be the established status of the pragmatically dominant language as a powerful contact language that is both widely understood and widely accepted. Consider the following example, from a speaker of Low German, originating in Schleswig-Holstein in northern Germany, who was recorded in the United States some 30-odd years after his emigration to that country:5 (17) Low German speaker, 35 years in USA ¨ nnericht fo¨r su¨stein Stunnen, but ik hef bloos Dat weer’n U that was a lesson for sixteen hours but I have only acht Stunnen ma˚kt, a˚ber 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 anything there.’

5. I am grateful to Do¨rte Hansen-Jaax for sharing this material with me.

An activity-oriented approach to contact-induced language change

39

Two variants for the connector ‘but’ – Low German a˚ber and English but – appear in this person’s speech alongside one another; the short excerpt reproduced here is typical of longer stretches of discourse documented for the speaker, which show that both variants have become an established part of his Low German speech. We can attribute this to repeated selection malfunctions that have gone unrepaired and uncorrected, and have finally become an accepted and integral feature of the speaker’s idiolect. The following example illustrates the adoption into Lovari Romani of German discourse particles by a speaker who belongs to the first generation in her family to be raised in Germany. (18) Lovari Romani, born in Poland, first generation in Germany: Laki familija sas also kesave sar te phenav, artisturi, n? her family were part such how comp say.1sg artists part ‘Her family were like such how shall I say, showpeople, right ?’ The speaker in (18) licenses herself to freely integrate German discourse particles, thereby accepting on a wholesale basis situations in which German operators of this class slip into her speech in an involuntary and unplanned manner; in other words, she compromises the selection and inhibition mechanism entirely for a complete class of functional operators, ridding herself of the burden to have to engage in suppressing ‘‘wrong language’’ choices in positions of high interactional tension and intense mental e¤ort to monitor and direct the hearer-side processing of the discourse. In earlier work (Matras 1998, 2000) I have referred to this process as ‘‘fusion,’’ as a way of capturing the resulting wholesale, category-specific merger of forms in one language (here Romani) with those of the contact language (here German). Naturally, long-term fusion of this kind presupposes the acceptance by a relevant sector of the Romani speech community of regular insertions of German word-forms into Romani discourse. Significantly, although Romani speakers may be said to operate by default in the bilingual mode (since bilingualism is the rule, and lexical insertions such as those discussed in Section 3 are frequent), this acceptance does not amount to a wholesale license to randomly insert just any German word. Rather, it applies specifically to the extended class of discourse operators, indefinites, particles, and connectors – or ‘‘utterance modifiers’’ (see Matras 1998). It is this kind of scenario that one can postulate as the background for language change and the borrowing of an entire class of operators, as is the case in Domari discussed above. In this language all connectors, most

40

Yaron Matras

focus and modal particles, and most indefinites have been borrowed from the contact language, Palestinian Arabic. The systems of monitoring and directing the interaction are thus identical or almost identical in the two languages. For Domari speakers, speaking their native language is therefore characterized by employing a particular set of vocabulary items and inflections, but not by employing a particular system of clause linking or interaction-level directing. Much like those German-English bilinguals who have only a single word-form for internet, computer, design, and so on, Domari bilinguals have but one system of clause linking and utterance modifying. (19) Domari: Fusion of clause linking devices (Arabic-derived forms are italicized): u# da iman/ yanı #/ kunt ama kury-a-m-e#k wala and always that.is was.1sg I house-obl-loc-pred.f and.not kil-sˇami wala aw-ami. wala waddik-ar-m-i mah all-ak ˙ exit-1sg and.not come-1sg and.not bring-3sg-1sg-pres place-indef ya par-ar-m-i wa#sˇ-ıs# kamk-am, u# par-ar-i or take-3sg-1sg-pres with-3sg work-1sg.subj and take-3sg-pres ple#-m. u# gisˇtane#-san ka#nu yanı # amilk-ad-m-a money-1sg and all-3pl was.3pl that.is treat-3pl-1sg-rem misˇsˇ gha#y kury-am-a bass ka#nat da#y-os neg good house-obl-loc but was.3sg.f mother-nom.3sg h ayyat-e#-ki gha#y wa#sˇ-ım. # pandzˇi rabbik-ed-os-im. ˙obl-abl good with-1sg she bring.up-perf-nom.3sg-obl.1sg yanı # lamma ka#nat h ayya#t far-m-a wila ‘isˇ i that.is when was.3sg.f ˙ hit-1sg-rem or something ka#nat h azzirk-ar-s-a. was.3sg.f ˙warn-3sg-3sg-rem ‘And I was always/ I mean/ at home, not going out nor coming nor did she take me anywhere. Or else she used to take me with her to work, and she used to take my money. And they all used to treat me badly at home. But Hayyat’s mother was nice to me. She brought me up. I mean, whenever Hayyat used to beat me or anything she used to tell her o¤.’ With relatively high frequency, discourse markers are subjected to language selection errors or malfunctions of the type illustrated above. They are also

An activity-oriented approach to contact-induced language change

41

frequently adopted from a contact language into the regular, stable idiolect of bilingual speakers (see for example Maschler 1994, 1997; Poplack 1980). They are frequently borrowed by ‘‘smaller’’ or ‘‘weaker’’ languages, that is, languages whose population tends to be bilingual or was bilingual at some stage in its history. The source of the forms in such cases is a ‘‘dominant’’ language, a language that was used in the public domain, often supported by institutions and literacy, and often spoken by a large population of monolinguals. All this allows us to postulate a direct link between selection malfunctions, their acceptability and stabilization in certain kinds of bilingual settings, and long-term language change (again, under certain sociolinguistic conditions). As argued already in the opening remarks of this article, I disagree with the direction of research that simply attributes borrowing of this kind to ‘‘social pressure,’’ ‘‘prestige,’’ or ‘‘social circumstances,’’ without spelling out the precise link between the social setting, conversational pressure and communicative intent, and the specific functional role of the structure or category in question. Let me therefore summarize the case, again, for an integrated, activitybased approach to the borrowability of the class of discourse markers and related structures. Bilingual speakers are under pressure to conform to monolingual rules on discourse formation, at least in some interaction settings (though a bilingual mode may well be the default conversational mode in some communities). This requires them to select those structures, constructions, word-forms, and so on from within their multilingual linguistic repertoire that are contextually appropriate, and to suppress or inhibit those that are not (cf. Green 1998; Paradis 2004). While this selection and inhibition mechanism is a normal and integral part of bilingual proficiency, it is not immune to occasional malfunctions. Being, essentially, disruptions in the mental processing procedure of language, these malfunctions are more likely to occur under circumstances of distress, fatigue, or confusion (such as a recent move from one environment to another), as well as around interaction management tasks that are particularly demanding and require increased mental e¤ort on the part of the speaker. Direct intervention in the listener’s processing of language and direction of the listener’s participation in the discourse belong to these demanding tasks. Selection malfunctions around the structures, and wordforms that trigger the relevant operations, are more likely to occur than around other structures. This explains the borrowing hierarchies presented in (20)–(25), which are based on examinations of frequency of borrowing patterns and the correlation between the borrowing of individual paradigm values, based on a sample of over 80 dialects of Romani in contact

42

Yaron Matras

with various languages (Elsˇ´ık and Matras 2006), as well as on a crosslinguistic sample of languages in contact (Matras 2007): (20) contrast > disjunction > addition (‘but’ > ‘or’ > ‘and’; ‘only’ > ‘too’; concessive > most other subordination markers; ‘except,’ ‘without,’ ‘instead of ’ > most other adpositions) (21) superlative > comparative (22) discourse markers (including fillers, tags, interjections) > focus particles, phasal adverbs > other function words (23) indefinites > interrogatives > deixis, anaphora (24) modality > aktionsart > future tense > other tense/aspect (25) obligation > necessity > possibility > ability > desire Contrast, and related semantic-pragmatic dimensions such as restriction (‘only’), exemption (‘except’), concession (‘even if,’ ‘although’), and substitution (‘instead’), are prone to tension and so to selection malfunctions due to the clash between the speaker’s communicative intentions and the listener’s expectations (based on shared contextual presuppositions). A similar contrast between an individual case and a set, and hence between the speaker’s chosen thematic focus point and an expected, presuppositional context, is conveyed by the superlative. Discourse markers and related operators participate in the management of interaction roles and the relations between speaker and listener, in particular by monitoring and directing the listener’s participation (for example tags, fillers, and hesitation markers), and are thus instrumental in processing (mental) clashes between speaker and hearer expectations. Indefinites delegate, as argued above, extensive processing work to the listener, risking the latter’s inability or refusal to cooperate and hence a breakdown in the e¤ectiveness and e‰ciency of the communicative interaction. Modality conveys the speaker’s relative weak authority to guarantee the truth-value of a proposition and therefore opens a potential window, yet again, for the listener’s refusal to cooperate and a communication breakdown. Naturally, it is not suggested that speakers are in any way aware of these e¤ects of functions such as indefiniteness and modality on the hearer, or that they pre-empt potential breakdown in the communication. Rather, the pragmatically outstanding function of these categories will lead to pressure in the processing procedure, which in turn may frequently trigger malfunctions. These malfunctions eventually become tolerated, in some speakers’ communities at least, and are no longer subjected to self-repair. At that

An activity-oriented approach to contact-induced language change

43

stage, they compete as variants with inherited forms, or they simply enrich the inventory of forms in these categories. Note that of the tenses, the most prone to contact-induced change is the future tense, which due to its project of unverified events is itself close to modality (cf. Comrie 1989). Finally, the hierarchy of borrowing for modality categories themselves (25) reveals that the association of events and actions with external pressures, which are beyond the speaker’s control, is more likely to trigger borrowing (and the underlying process of disruptions of the selection mechanism) than those associated with internal attitudes or aptitude. The susceptibility of linguistic operations triggering increased tension and mental e¤ort to be subject to lapses in control over the language selection mechanism is thus directly reflected in the likelihood that categories representing these very same linguistic operations will undergo structural borrowing. How, then, can we connect what is an event a¤ecting the individual speaker’s performance in discourse (selection malfunction and bilingual speech production error) with what is by necessity and definition a social process, namely an alteration to the permanent shape of a community’s language? Many, perhaps even most, individual lapses will not lead to language change; they will either be self-repaired by the speaker, corrected by the listener, or ignored by the participants. But an e¤ort will be made by the speaker to avoid them in order to avoid a breakdown in communication or simply to avoid the embarrassment of apparent ineptness. However, frequently occurring selection malfunctions may become stabilized in an individual’s idiolect as they are left uncommented upon and at the same time understood by a regular audience of interlocutors. Such a situation may arise in groups of bilinguals whose default conversation mode is bilingual, and where normative intervention in language use is not intense and flexibility in linguistic choices is tolerated. Normally, such a situation would tend to point toward a diglossic imbalance in the roles of the two languages, with one, the source of borrowings or donor language, being the more dominant and institutionally protected language of a monolingual community and of public life, and the other, the borrowing or recipient language, being limited to a bilingual minority, typically used in informal domestic situations and possibly confined to oral use only, and hence coupled with a low awareness of language structure and low motivation to intervene and consciously shape language use. If the same expressions become stabilized among a group of individuals in such a community, then the process of language change will have been set in motion.

44

Yaron Matras

6. Language manipulation (deliberate mixing) In Section 3 I discussed lexical insertions. I mentioned their ambiguous status: they are quasi-violations of the rules on the selection of contextappropriate linguistic material, and thus e¤ective triggers of special conversational emphasis or specific associations. At the same time they are understood by the interlocutors, and so are communicatively e¤ective. They are also sanctioned by the availability of the bilingual mode (Grosjean 2001) as an option for structuring conversation among the group of interlocutors. My final section is devoted to an even more daring violation of the rules on contextual well-formedness (that is, monolingual selection), one that does not even enjoy the exemption that the bilingual mode is normally able to provide. Consider example (26), from the same trilingual (Hebrew-GermanEnglish) child discussed above. The child’s default languages of interaction are German with the mother, Hebrew with the father, and English outside the home. As we saw earlier, lexical insertions are acceptable in a bilingual mode within the household, as long as the insertions represent concepts that trigger specific associations with the English-speaking (or other, as the case may be) environment. In (26), the child is addressing the father, but violating the normal pattern of language choice by using English. On top of that, he further violates even the norms of a hypothetical (reconstructed or imitated) English-speaking interaction setting by inserting into his utterance in (c.) everyday words from German that carry no individual special e¤ect of their own; that is, unlike ‘‘normal’’ insertions, they do not represent concepts that are associated specifically with a German-speaking environment, and so are not intended to evoke such associations through some kind of original context-bound authenticity: (26) (Hebrew); age 8:6, calling to his father from the bathroom when washing his face before going to bed in the evening (insertions in segment c. from German): a. Child: Aba! b. Father: Hmm. c. Child: Where do I get a Lappen so I can wisch my Gesicht? wash-cloth wipe face The German insertions are thus not selected individually because of their content. Rather, their purpose is to have a wholesale ‘‘humorous’’ e¤ect on the utterance. This e¤ect is brought about precisely by highlighting

An activity-oriented approach to contact-induced language change

45

the deviation from the full range of any expected, norm-compliant utterance. To this end, the assembly of the utterance is being manipulated to deviate from the usual types of either monolingual or bilingual (mixed) utterances that might occur in interaction between the child and his father. Such a step requires a high degree of linguistic skill and linguistic awareness. It requires first of all an awareness of the full range of possible, permissible mixture types, in order to identify a type that is deviant from those. Furthermore, it requires the self-confidence of a skilful and competent speaker to experiment with a type of mixed utterance that is seldom experienced in everyday interaction, and for which there is hardly an existing model in either parental or peer speech. This in turn requires a subtle feel for the immediate context and setting, anticipation of the possible and likely responses on the part of the listener, and a positive assessment of the chances of the utterance to achieve its intended key e¤ect rather than result in a communication breakdown. It is noteworthy that a multilingual child as young as eight is already equipped potentially with the necessary skills and linguistic confidence that allows him to manipulate language mixing in this way. Similar patterns of deliberate mixing for special e¤ect are documented from the speech of adult bilinguals by Golovko (2003). The following example, from the Hebrew of Israeli students in Germany, illustrates a structurally rather subtle approach to language manipulation, once again for the purpose of humor. It involves the imitation of the German morphophonological rule on the formation of conditionals (the so-called second subjunctive) through umlaut – German ich konnte ‘I could,’ wenn ich ko¨nnte ‘if I could’ – and its importation into Hebrew, which lacks not only the phenomenon of umlaut altogether, but even the resulting rounded phoneme itself. The basis is the Hebrew past-tense form yaxo´lti ‘I could’: (27) Israeli students in Germany; imitation of German subjunctive: ´ılu yaxo¨lti if I-could Neither of the two preceding examples is likely to lead to language change, but this is not because of the nature of the communicative task or the structural strategies being pursued by the bilingual speakers in order to achieve it. The improbability of language change stems from the restrictions on potential propagation of the innovative structure to general and regular use among a sector of the speech community. This in turn has to do strictly with the number of bilingual interlocutors in the community and

46

Yaron Matras

their frequency of interaction, the frequency with which they resort to overtly marked humorous keys in conversation, and the availability of other modes, apart from language manipulation, to mark out that humor, and so the overall degree of utility that the construction may have to its users, as well as the degree of flexibility for carrying on shaping patterns of language use free from any restrictive, normative intervention. All this, which we might attribute to the ‘‘sociolinguistic circumstances’’ of the speech community or perhaps just a small sector within it, are not, however, triggers for contact-induced change, nor are they mechanisms for change. Rather, they are factors that may or may not facilitate or contain the spread of innovations throughout a population of speakers, and the consequent emergence of change. The communities of Jewish cattle-traders in pre-war southwestern Germany had developed an in-group speech mode, called Lekoudesch, that was characterized by the insertion of Hebrew-derived word-forms, largely lexical content words, recruited via the community’s exposure to the study of Hebrew and Hebrew scriptures, into their regional dialect of German.6 (28) Lekoudesch: Der scha¨¤t de ganze Jomm im Uschpiss, un duat immer harme he sits the whole day in pub and does always much schasskenna und meloucht lou. drinking and works not ‘He sits all day in the pub, and drinks a lot, and doesn’t work.’ The principal goal for which Lekoudesch was employed was as a mode of humor and entertainment for members of the group of traders, and as a means of setting them apart from outsiders, including using the language to camouflage content. The term for the variety itself, Lekoudesch, is a humorous word-play based on the traditional Ashkenazic term for the Hebrew language, loshn koudesch meaning ‘‘the sacred language.’’ The social setting prerequisites for stabilization are thus present in the existence of a group of interlocutors – not even a speech community in the usual sense of the term – who find continuous use for a consciously created, humorous speech mode based on their multilingual competence, through 6. Data from recordings among Jewish survivors as well as non-Jewish farmers who had learned the variety in their youth while being employed by Jewish cattle-traders.

An activity-oriented approach to contact-induced language change

47

which they can set themselves apart from others and thus rea‰rm their group identity and group a‰liation. My final example is a case in which this stabilization of language mixing patterns has led to the creation of a stable repertoire that is the property of an entire ethnic minority, and thus of a speech community, and is sometimes even referred to as a ‘‘mixed language,’’ the English-Romani mixture or Angloromani of English and Welsh Gypsies. (29) Angloromani: Ol the obben coz when the raklis jels I’m gonna mor yas. ‘Eat the food coz when the girls go I’m gonna kill you!’ Speakers or rather users of Angloromani have at their disposal a special inventory of Romani-derived words covering typically anywhere between 150 and 700 distinct word meanings. Some meanings are covered directly by Romani words that have been preserved, others through compounding and extensions of inherited Romani base-words. The Romani lexical legacy dates back to the use of inflected Romani as the everyday community language of the Gypsy minority, until the shift to English during the second half of the nineteenth century left no fluent speakers of Romani in the community. Holding on to a selection of Romani lexical items thus amounts to a preservation of an important token of older community heritage. Functionally speaking, the availability of a form of speech that sets group members apart from others continues to be an important motivation for the preservation and cultivation of Romani-derived vocabulary. As such, Romani words are now inserted in order to obtain a special conversational key. While there is no obligation on the speaker to insert any particular word, or even clusters of words, from Romani, the insertion of a Romani element will give the entire speech act the flavor of a special emotive mode that will activate the cultural bond between speaker and listener, and call on the listener to interpret what has been said in light of that bond. Romani thus becomes an instrument for intimate comments as well as for warnings, for threats that are based in empathy, as in (30), as much as for secretive content that is to be concealed from bystanders (cf. Matras et al. 2008). 7. Conclusion I hope to have shown in this article that neither categorizations of contact phenomena based on formal-structural factors nor those based strictly on social factors can explain the motivation behind contact-induced language

48

Yaron Matras

change. Social factors are involved in facilitating or constraining the successful propagation of innovations throughout a speech community. Structural factors may also play a role in constraining or facilitating change, but structures are there in the first instance as triggers of language-processing tasks. Contact-induced change is the product of the creativity of speakers seeking new ways to achieve goal-oriented tasks in communicative interaction. Their creativity results in innovations, which in turn may or may not be replicated by other speakers, or even by the same speaker on subsequent occasions, and hence they may or may not lead to language change. But while not every innovation will lead to change, there is no change that is not the product of a task-bound, goal-oriented innovation, introduced into discourse by a creative speaker.

non-conscious

conscious

selection malfunctions > pattern-replication > lexical insertion > speech manipulation no special e¤ect special e¤ect

Figure 3. The continuum of contact-induced creativity and innovation

Speakers’ creativity may be arranged on a continuum ranging from those processes that are non-conscious but nevertheless the product of a function-driven strategy for coping with task-specific language processing, through to those that involve a degree of deliberate and conscious defiance of general rules that normally govern the structuring of communicative interaction. All four types of creativity discussed here – selection malfunctions, pivot-matching, lexical insertions, and speech manipulation – involve some form of negotiation of two opposing pressures: the constraints on selecting only those structures that are considered appropriate and so acceptable in the ongoing interaction context, and full exploitation of the speaker’s overall repertoire of linguistic forms and structures, within which it is impossible to completely deactivate one of the ‘‘language systems.’’ The conscious negotiation and manipulation of the two poles usually targets some kind of special conversational e¤ect, whereas less conscious and less deliberate negotiation is more concerned with ease of the processing load and relaxation of the constraints on context-appropriate selection. The more extreme case of non-conscious negotiation, that of selection malfunctions, represents the speaker’s way of giving in to competing pressures by allowing one – the need to select a functionally e¤ective structure –

An activity-oriented approach to contact-induced language change

49

to triumph over the other – the need to select context-appropriate structures. The less extreme case, pattern replication, constitutes a compromise, with a functionally e¤ective construction selected despite the fact that it fails to satisfy the constraint on context-appropriateness, but adapted using word-forms that do satisfy context-appropriateness. While we may continue to regard most formations of this latter type as non-conscious, some, such as luggage room in example (7), may indeed be semi-conscious formations, while others still, such as the nativization of loanwords (for example Turkish okul ‘school,’ from French e´cole modified to utilize the Turkish verb-root oku- ‘to read’), are carefully planned. At the ‘‘conscious’’ far end of the continuum we find deliberate mixing that targets the entire speech act, rather than just individual references within it, and is usually designed to alter its key rather than its content, thus achieving an e¤ect at the level of the interaction as a whole, rather than at the propositional level of the message content. The speaker’s goal in such instances is obviously to exploit the contrast between components of the multilingual repertoire, so as to direct the key of the interaction and so the special relations with the interlocutor (or audience of interlocutors). The milder form of conscious or deliberate mixing involves mere insertions of word-forms that are deemed to capture the contextual nuances associated with the type of interaction settings for which they are normally reserved. Here, the goal is to maximize precision of expression, rather than to influence the key of the utterance itself, although taking the liberty to insert word-forms from another language will inevitably also signal the speaker’s reliance on the listener’s solidarity and empathy in accepting and supporting the choice of mixing as a legitimate speech mode. Indeed, some insertions, certainly those involving the mere naming of institutions or procedures, as in example (2), will occur spontaneously and with little or no planning at all, placing them perhaps at a similar level of semiconsciousness as some types of pivot-matching. To sum up, I have argued that contact-induced language change is a product of the propagation of creative innovations introduced by speakers as task-e¤ective means to achieve communicative goals. The key to understanding the position of individual structures and structural categories in the process of contact-induced change is to interpret the role of those structures in triggering linguistic-mental processing operations that support specific communicative tasks. The role of social and societal aspects in the process of contact-induced change is, in turn, to act as facilitators in the propagation of an innovation, allowing it to gain acceptability and replication and ultimately a useful role among the inventory of structuralcommunicative devices that a speech community has at its disposal.

50

Yaron Matras

Abbreviations 1 2 3 abl acc attr comp constr dat def deic f loc m neg nom obl part past pl poss pred pres rel rem sg subj

First person Second person Third person Ablative Accusative Attributive Complementizer Construct state Dative Definite (article) Deixis Feminine Locative Masculine Negation Nominative Oblique Particle Past tense Plural Possessive Predication Present tense Relativizer Remote Singular Subjunctive

References Aikhenvald, A. Y. 2002. Language Contact in Amazonia. Oxford: Oxford University Press. Backus, A. 1996. Two in one. Bilingual Speech of Turkish Immigrants in the Netherlands. Tilburg: Tilburg University Press. Bakker, P. 1996. Language intertwining and convergence: typological aspects of genesis of mixed languages. Sprachtypologie und Universalienforschung 49: 9–20. Bakker, P. 1997. A Language of our own. The Genesis of Michif – the Mixed CreeFrench Language of the Canadian Me´tis. New York: Oxford University Press. Berk-Seligson, S. 1986. Linguistic constraints on intra-sentential code-switching: a study of Spanish/Hebrew bilingualism. Language in Society 15: 313–348.

An activity-oriented approach to contact-induced language change

51

Comrie, B. 1989. On identifying future tenses. In Tempus-Aspekt-Modus, W. Abraham and T. Jansen (eds), 51–63. Tu¨bingen: Niemeyer. Ehlich, K. 2007. Sprache und sprachliches Handeln. Berlin: Mouton de Gruyter. Elsˇ´ık, V. and Matras, Y. 2006. Markedness and Language Change: the Romani Sample. Berlin: Mouton de Gruyter. Elsˇ´ık, V. 2009. Loanwords in Selice Romani, an Indo-Aryan language of Slovakia. In Loanwords in the World’s Languages: A Comparative Handbook, M. Haspelmath and U. Tadmor (eds), 260–303. Berlin: Mouton de Gruyter. Field, F. W. 2002. Linguistic Borrowing in Bilingual Contexts. Amsterdam: John Benjamins. Golovko, E. V. 2003. Language contact and group identity: the role of ‘folk’ linguistic engineering. In The Mixed Language Debate, Y. Matras and P. Bakker (eds), 177–207. Berlin: Mouton de Gruyter. Green, D. 1998. Mental control of the bilingual lexico-semantic system. Bilingualism: Language and Cognition 1: 67–81. Grosjean, F. 2001. The bilingual’s language modes. In One Mind, two Languages. Bilingual Language Processing, J. L. Nicol (ed.), 1–22. Oxford: Blackwell. Grosjean, F. 2008. Studying Bilinguals. Oxford: Oxford University Press. Gumperz, J. 1982. Discourse Strategies. Cambridge: Cambridge University Press. Haase, M. 1991. Sprachkontakt und Sprachwandel im Baskenland. Die Einflu¨sse des Gaskognischen und Franzo¨sischen auf das Baskische. Hamburg: Buske. Haugen, E. 1953 [1969]. The Norwegian Language in the Americas: A Study in Bilingual Behavior. Bloomington: Indiana University Press. Heath, J. 1984. Language contact and language change. Annual Review of Anthropology 13: 367–384. Heine, B. and Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Lanza, E. 1997. Language Mixing in Infant Bilingualism. A sociolinguistic Perspective. Oxford: Clarendon. Loveday, L. 1996. Language Contact in Japan. A Socio-linguistic History. Oxford: Clarendon Press. Maschler, Y. 1994. Metalanguaging and discourse markers in bilingual conversation. Language in Society 23: 325–366. Maschler, Y. 1997. Emergent bilingual grammar: the case of contrast. Journal of Pragmatics 28: 279–313. Matras, Y. 1998a. Utterance modifiers and universals of grammatical borrowing. Linguistics 36: 281–331. Matras, Y. 1998b. Convergent development, grammaticalization, and the problem of ‘mutual isomorphism’. In Sprache in Raum und Zeit, W. Boeder, C. Schroeder and K.-H. Wagner (eds), 89–103. Tu¨bingen: Narr. Matras, Y. 2000. Fusion and the cognitive basis for bilingual discourse markers. International Journal of Bilingualism. 4.4: 505–528. Matras, Y. 2004. Layers of convergent syntax in Macedonian Turkish. Mediterranean Language Review 15: 63–86.

52

Yaron Matras

Matras, Y. 2007. The borrowability of grammatical categories. In Grammatical Borrowing in Cross-linguistic Perspective, Y. Matras and J. Sakel (eds), 31–74. Berlin: Mouton de Gruyter. Matras, Y. 2009. Language Contact. Cambridge: Cambridge University Press. Matras, Y. and Sakel, J. 2007. Investigating the mechanisms of pattern-replication in language convergence. Studies in Language 31: 829–865. Matras, Y. and Sakel, J. (eds). 2007a. Grammatical Borrowing in Cross-linguistic Perspective. Berlin: Mouton de Gruyter. Matras, Y. and Shabibi, M. 2007. Grammatical borrowing in Khuzistani Arabic. In Grammatical Borrowing in Cross-linguistic Perspective, Y. Matras and J. Sakel (eds), 137–149. Berlin: Mouton de Gruyter. Matras, Y. and Tufan, S¸. 2007. Grammatical borrowing in Macedonian Turkish. In Grammatical Borrowing in Cross-linguistic Perspective, Y. Matras and J. Sakel (eds), 215–227. Berlin: Mouton de Gruyter. Matras, Y., Gardner, H., Jones, C., and Schulman, V. 2008. Angloromani: a di¤erent kind of language? Anthropological Linguistics 49.2: 1–44. Moravcsik, E. 1978. Universals of language contact. In Universals of Human Language, J. H. Greenberg (ed.), 94–122. Stanford: Stanford University Press. Nau, N. 1995. Mo¨glichkeiten und Mechanismen kontaktbewegten Sprachwandels – unter besonderer Beru¨cksichtigung des Finnischen. Munich: Lincom. Paradis, M. 2004. A neurolinguistic Theory of Bilingualism. Amsterdam: John Benjamins. Poplack, S. 1980. Sometimes I’ll start a sentence in Spanish y termino en espan˜ol. Linguistics 18: 581–618. Rehbein, J. 1977. Komplexes Handeln. Elemente zur Handlungstheorie der Sprache. Stuttgart: Metzler. Remmers, A. 1997. Plattdeutsch in Ostfriesland. Die Mundart von MoormerlandWarsingsfehn. Leer: Sollermann. Rudolph, E. 1996. Contrast. Berlin: Mouton de Gruyter. Sacks, H., Scheglo¤, E. A., and Je¤erson, G. 1974. A simplest systematics for the organization of turn-taking in conversation. Language 50: 696–735. Salmons, J. 1990. Bilingual discourse marking: code switching, borrowing, and convergence in some German-American dialects. Linguistics 28: 453–480. Thomason, S. G. and Kaufman, T. 1988. Language Contact, Creolization and Genetic Linguistics. Berkeley: University of California Press. Thomason, S. G. 2001. Language Contact. An Introduction. Edinburgh: Edinburgh University Press. van Hout, R. and Muysken, P. 1994. Modelling lexical borrowability. Language Variation and Change 6: 39–62. Weinreich, U. 1953. Languages in Contact. The Hague: Mouton. Winford, D. 2003. An Introduction to Contact Linguistics. Oxford: Blackwell.

Contact-induced change as an innovation1 Claudine Chamoreau 1. Introduction Generally, in a situation of language contact, the syntactic e¤ects on replica language (or receiving language) structure seem to be related to features that have come from one of the languages in contact, frequently the model language (or source language). For example, Thomason’s typology of morphosyntactic changes in contact situations shows three types of e¤ects on a receiving language structure: loss of features as a result of language contact; addition of linguistic features through contact-induced changes; and partial or total replacement of old native linguistic features by interference features (2001: 60, 85–91). Heine (2006) indicates that generally in the situation of language contact, ‘‘speakers recruit material available in R (the replica language) to create new structures on the model of M (the model language) and . . . rather than being entirely new, the structures created in R are built on existing use patterns and constructions that are already available in R.’’ This paper explores a specific contact-induced change, that is, innovation, defined as structure that emerges as a consequence of contact between two languages and that diverges from the patterns of both the model language and the replica language. In other terms, these new innovated linguistic features are not created on the model of the model language. In this paper I investigate the development of new features as consequences of the contact between Purepecha,2 the replica language, and 1. This is a revised version of a paper that was originally presented in September 2007 at the Workshop on Language Contact and Morphosyntactic Variation and Change, Paris. I am very grateful to members of this audience who provided relevant comments, in particular Sally Thomason. I also would like to acknowledge with gratitude the comments of Marianne Mithun, Salome´ Gutierrez, and Evangelia Adamou on an earlier draft. 2. Purepecha (formerly known as Tarascan) is classified as a language isolate spoken in the state of Michoacan, with approximately 110,000 speakers (Chamoreau 2009). There are di¤erent ways of spelling the name of this language. In the literature, it is possible to find it as Purepecha, Pure´pecha, Purhe´pecha, P’ure´pecha, P’urhe´pecha, Phurhe´pecha, P’orhe´pecha, Phorhe´picha, etc.

54

Claudine Chamoreau

Spanish, the model language, with which it has been in contact for nearly five centuries. According to the types of contact-induced changes described by Thomason and Kaufman (1988), Purepecha presents a situation of intense contact and the characteristics of a shift situation, since the changes are mainly in phonology and morphosyntax (Chamoreau 2007, 2010). I specifically examine the domain of comparative constructions of superiority in Purepecha. In this language, almost all superiority comparative constructions clearly show the consequences of contact with Spanish. Certain constructions, such as example (1a), constitute borrowing or replication of the less marked construction in the model language, the particle construction with the degree marker ma´s ‘more’ and the relator que ‘than’ shown in example (1b). Another construction, example (1c), formed by the degree marker sa´ni¼teru, the relator ke, and the preposition de, is created by adapting the model of the Spanish construction with ma´s . . . de . . . que, example (1d). (1) a.

b.

enrike mas §epe-s-ti ke Pedru Henry more be lazy-aor-ass3 than Peter ‘Henry is lazier than Peter.’ (Cuanajo-Evaristo9: 208) mi padre baila ma´s ra´pido que mi madre pos1 father dance.pres3 more fast than pos1 mother ‘My father dances faster than my mother.’

c. Gervasio sa´niFteru prontu ni-ra-s-ti ke de ima Gervasio few¼more quickly go-ft-aor-ass3 than of dem ‘Gervasio went more quickly than him.’ (Lit. ‘Gervasio went more quickly than of him.’) (Cuanajo-Evaristo9: 102) ´ d. El es ma´s feliz de lo que pensaba 3ind be.pres3 more happy of dem than think-past.impf1 ‘He is happier than I thought.’ But an original structure has been conceived on the model neither of the replica language nor of Spanish. This structure employs the preposition entre for comparison, for example: (2) Puki mas kokani xano-nka-ti ke entre ima Puki more quickly arrive-centrip-ass3 than between dem ‘Puki arrives more quickly than him.’ (Lit. ‘Puki arrives more quickly than between him.’) (San Andre´s Tziro´ndaro-nana1: 101)

Contact-induced change as an innovation

55

In example (2), we recognized the Spanish particle construction with mas . . . ke, but the presence of entre is original, and impossible in Spanish for a comparative construction. The specific innovation studied in this article is not a partial copy (Heine and Kuteva 2005) but an innovation: speakers attribute to a Spanish morpheme a new function not attested in either the model language or the replica language, inventing a new structure. The interesting fact is that on the one hand contact makes a syntactic innovation possible, while on the other hand this innovation seems to correspond to cross-linguistically cognitive tendencies (Matras 2007). This paper is organized into the following sections: Section 2 introduces some basic typological properties of Purepecha and essential information on data collection procedures. Section 3 presents comparative constructions in Spanish, the model language, and Lengua de Michoacan, the pre-contact replica language.3 Section 4 illustrates the diversity and complexity of comparative constructions in Purepecha. Section 5 gives a detailed analysis of the innovative construction in Purepecha. Section 6 shows the absence of similar constructions in other Mesoamerican languages. The discussion in section 7 assigns the phenomenon under scrutiny a place in the catalogue of contact-induced structural changes.

2. Essential information about Purepecha 2.1. Basic typological properties Purepecha has nominative-accusative alignment, where the subject of a transitive verb, like Selia ‘Celia,’ in (4), is encoded like the subject of an intransitive verb, anima, ‘soul,’ in (3). This is a case-marking language in which the nominal subject has no overt marker. In an intransitive construction, as in (3), the single argument anima-it§a ‘the souls’ has no specific marker. The object is generally marked by the objective case marker -ni. This morpheme encodes the object of a transitive verb, misitu-ni ‘the cat,’

3. In order to distinguish the pre-contact replica language from the contact replica language, I adopt the traditional name, Lengua de Michoacan, for the former, the language spoken in the sixteenth century, and the current name, Purepecha, for the latter (Ma´rquez Joaquı´n 2007).

56

Claudine Chamoreau

in (4), and both objects of a ditransitive verb, such as inte-ni wantantskwa-ni and Puki-ni, in (5).4 (3) ya§fi¼k§fi tsfima anima-it§a tsı´pi-pa-ntha-§a-ti now¼3pl dem.pl soul-pl be glad-centrif-it-prog-ass3 ‘Now these souls are leaving happily. . . .’ (Jara´cuaro-animas5: 10)5 (4) xo Selia ata-§-ti imeri misitu-ni yes Celia beat-aor-ass3 pos3 cat-obj ‘Yes, Celia beat her cat.’

(Jara´cuaro-Alfredo25: 94)

(5) xo Selia a›i-§-ti inte-ni wantantskwa-ni Puki-ni yes Celia tell-aor-ass3 dem-obj story-obj Puki-obj ‘Yes, Celia told Puki a story.’ (Jara´cuaro-Alfredo25: 36) Purepecha is an agglutinative and synthetic language, and is almost exclusively su‰xing. It has an elaborate derivational verbal system. Although bare stems exist, there is a very productive derivational system in which a basic stem can take voice, causative, locative, positional, directional, and adverbial derivative su‰xes. Inflectional su‰xes follow the stem to mark aspect, tense, mood, and person (Chamoreau 2009; Monzo´n 2004; Nava 2004). Subject and object pronouns are expressed by pronominal enclitics that are generally attached to the last element of the first immediate constituent of either the main or the subordinate clause, such as ¼k§fi, in example (3) or ¼ni and ¼kini in example (6). They can also be attached to the verb. Oblique complements are marked by postpositions, such as it§orita ximpo in (6). (6) no¼t§ ka¼ni xi¼thu¼kini xa›oa-ta-s-ki pasari-ni neg¼well¼1 1ind ¼too¼2obj help-caus-aor-int go though-inf it§orita ximpo canoe inst ‘Well, have I not also helped you to cross by canoe?’ (Zipiajo-Emelia4: 71) 4. The presence or absence of the object case marker depends on di¤erent hierarchies: (i) the inherent semantic properties of the referent (human, animate); (ii) properties related to grammatical features (definite, count noun vs. mass noun, generic vs. specific, etc.); and (iii) pragmatic strategies (topic, focus). 5. The examples of Purepecha come from my own fieldwork data. The first name corresponds to the pueblo, here Jara´cuaro; after the hyphen there appears the name of the speaker (real or invented, in accordance with the wish of the speaker) or the name of the narrative, here animas, and then the reference of the recording, here 5: 10.

Contact-induced change as an innovation

57

Purepecha is basically a SV and SVO constituent order language, as illustrated by examples (4) and (5). This order, that is, the order that is pragmatically unmarked, is the basic order in the region of Lake Patzcuaro (Capistra´n 2002 and Chamoreau 2009: 55–58). Other orders indicate specific pragmatic properties. Studies on constituent order in the other regions do not as yet exist. However, Purepecha shows traits of a SOV language: (i) tense, aspect, and modal markers following the verb; (ii) postpositions; (iii) the almost exclusive use of su‰xes; (iv) enclitics; (v) case markers; (vi) main verbs preceding inflected auxiliaries. SVO and SOV constituent orders are attested in the sixteenth century, and the former has progressively increased since then. The change is probably due to areal contact (Smith, personal communication). Spanish has been the principal contact language for many centuries; however, prior to the Conquest there were speakers of other languages in this territory, mostly from Nahuatl (Uto-Aztecan family) and Otomı´ (Otopamean family), two languages with verb-initial structure. The change probably began under the influence of these languages; Spanish, an SVO language, continued the process, for example by introducing prepositions (Chamoreau 2007). 2.2. Data collection procedures This investigation is part of a project6 which aims to document the di¤erent ways of speaking Purepecha. So far, I have studied 60 villages located in 21 municipalities, accounting for 70 percent of the villages in which the language is spoken. In each village, I recorded three men and three women, belonging to three age groups (15–29, 30–49, 50 and older). The method I adopted was to record five types of data (during approximately 15 hours in each village): i)

Traditional narratives, descriptions of specific situations, spontaneous speech ii) Conversations between two or three people from the same village or from di¤erent villages 6. This research was made possible through financial support from the French Center for American Indigenous Languages Studies, CELIA (CNRS-INALCOIRD-Paris VII), the French Center for Mexican and Central American Studies (CEMCA), and the National Institute for Indigenous Languages of Mexico (INALI). Aid from these institutions is greatly appreciated. This research would not have been possible without the support of Teresa Ascencio Domı´nguez, Puki Lucas Herna´ndez, Celia Tapia, and all our Purepecha hosts.

58

Claudine Chamoreau

iii) 200 sentences (translated from Spanish), designed to cover all relevant areas of morphosyntax iv) Sociolinguistic questionnaires (about each village and each speaker) asked in Purepecha v) Attitude questionnaires (perceptual dialectology) also asked in Purepecha. 3. Comparative constructions in model and pre-contact replica languages This paper deals with the e¤ects of language contact in the di¤erent villages where Purepecha is spoken. We observe these consequences from a synchronic perspective. Nevertheless, in order to understand the di¤erent constructions, and to analyze the di¤erence between the impact of contact and that of internal change, it is relevant to show the diversity of constructions attested in Spanish, the model language, and in Lengua de Michoacan, the pre-contact replica language. 3.1. Comparative constructions in Spanish, the model language Spanish has had and has various types of comparative constructions. I will present here the most frequent constructions that were used in the sixteenth century, the time of contact between Lengua de Michoacan and Spanish. The most frequent and less marked is the particle construction which has a degree marker ma´s ‘more’ and a relator, que ‘than’ (Galant 1998; Price 1990; Rojas Nieto 1990a, 1990b). The comparee NP is the subject and the standard NP is expressed after the quality and appears after the relator que. In (7), the degree marker comes before the quality with the be-verb and the adjective. In (8) the position is the same, with quality expressed by the adverb ra´pido ‘fast.’ In (9) where quality is expressed by the verb corre ‘run,’ the degree marker comes after the verb and beside the comparative marker. (7) Spanish Marı´a es ma´s alta que Juan Mary be.pres3 more tall.fem than John ‘Mary is taller than John.’ (8) Spanish mi perro corre ma´s ra´pido que tu gato pos1 dog run.pres3 more fast than pos2 cat ‘My dog runs faster than your cat.’

Contact-induced change as an innovation

59

(9) Spanish mi perro corre ma´s que tu gato pos1 dog run.pres3 more than pos2 cat ‘My dog runs more than your cat.’ This particle type is most widespread in Europe: 93 percent of European languages possess it (Stassen 1985; Heine 1994, 1997). In Spanish this type coexists with another type, described as a marked type, in which the de preposition appears, as can be observed in (10). (10) Spanish (Rojas Nieto 1990b: 226) es ma´s grande de lo normal be.pres3 more tall of dem normal ‘He is taller than the normal one.’ The di¤erence is that the ma´s . . . que construction appears before all clause types, whereas the ma´s . . . de construction shows restriction in use. Rojas Nieto (1990b) notes that this construction is found before temporal NPs, relative clauses (11), indefinite clauses, and others, but never before demonstratives (12), possessive NPs, or relative clauses introduced by quien ‘whose’ (13). (11) Spanish (Rojas Nieto 1990b: 229) mandaron ma´s libro-s de lo-s que pedimos send.past.3pl more book-pl of dem.masc-pl than ask.for.past.1pl ‘They sent more books than those we asked for.’ (Lit. ‘They sent more books of those than we asked for.’) (12) Spanish (Rojas Nieto 1990b: 230) *Vino ma´s gente de estos estudiantes. *More people came of these students. (13) *Vino ma´s gente de quien nos dijeron. *More people came of who they told us. This ma´s . . . de construction shows the cognitive relation between comparison and location meaning (Rojas Nieto 1990b; Stassen 1985). The standard NP is conceptualized in terms of spatial relationships. This type is very frequent in languages worldwide. A third comparative construction exists in Spanish, a lexical structure which is seldom used. It can be classified as belonging to the verbal type since this construction involves lexical concepts that use the idea of surpassing as a degree marker, as in (14). The comparee NP is the subject el duque and the standard NP is the object lo.

60

Claudine Chamoreau

(14) Spanish (Rojas Nieto 1990a: 449) el duque so´lo lo supera en linaje the duke only him surpass-pres3sg in lineage ‘Only the duke surpasses him in lineage.’ This type is widespread in languages that are more verb-like, that is, in which the adjectival category is less developed than in the Indo-European languages, for example (Bath 1994: 184–209). But in Spanish, this construction is marked and generally used when speakers want to insist on the meaning of the verb, for example an action verb which carries the notion of ‘surpass’ as in example (14). 3.2. Comparative constructions in Lengua de Michoacan, the pre-contact replica language There are two types of constructions; both have a xats- ‘surpass’ verb which expresses degree. These constructions correspond to the synthetic and the derivational morphological characteristics of the language: the verb is modified by the causative -ta and by a su‰x expressing transfer -ma. This first construction is a clear verbal type. In (15), the comparee NP, Pedro is the subject and the standard NP Xwano-ni is the direct object. The quality is expressed by a non-finite verb ampake-ni which functions as an argument of the main verb, forming a complement clause (Noonan 1985). The quality appears after the standard NP which is generally a sign of OV languages (Andersen 1983: 99–138; Dryer 2007). This is the opposite word order to that found in Spanish (examples 7, 8, 9). (15) Lengua de Michoacan (Isolate, Gilberti 1987 [1558]: 109) Pedro hatztamahati Juanoni ambaqueni 7 Pedro xats-ta-ma-xa-ti Xwano-ni ampake-ni Peter put-caus-transf-pres-ass3 John-obj be good-inf ‘Peter is better than John.’ (Lit. ‘Peter surpasses John (in) be(ing) good.’) The second construction is a mixed type which combines a verbal type and a coordination type. In (16), the first clause contains the comparee NP, the

7. When an example is quoted, I reproduce the author’s transcription in the first line.

Contact-induced change as an innovation

61

subject Pedro ‘Peter,’ the verb xats ‘surpass,’ and the object, the non-finite verb ampake-ni ‘be good’ which functions as an argument of the main verb, a complement clause. The second clause is introduced by the coordinator ka. The negation no indicates that the standard NP lacks the property. The adverb is ‘like that’ and the negation no operate the semantic reference with the verb xats ‘surpass.’ In the second clause, there is no verb. This construction is similar to what Galant describes as stripping (1998: 242). It refers to a process in which all material is eliminated in the second clause except a nominal constituent, here the standard NP, Xwanu, a special adverb is, and the negative element no. The (lexical) verb is identical in each clause and the overall structure is parallel. (16) Lengua de Michoacan (Isolate, Gilberti 1987 [1558]: 109) Pedro hatztamahati ambaqueni ca noys Juan Pedro xats-ta-ma-xa-ti ampake-ni ka no is Xwanu Peter put-caus-transf-pres-ass3 be good-inf and neg so John ‘Peter is better than John.’ (Lit. ‘Peter surpasses in being good, and John (is) not like that.’) 4. Comparative constructions in Purepecha In Purepecha comparison of superiority is mapped out by means of ten constructions, which can be grouped into four types: Type A. Particle type; Type B. Particle type with a locative phrase; Type C. Mixed coordination and particle type; Type D. Applicative type. The presentation of these types will follow their frequency as primary and secondary options: only the first two types, the particle type (type A) and the particle type with a locative phrase (type B), may be a primary choice. Type A is the primary choice in almost all the villages except a few north of Lake Patzcuaro where type B is the primary choice and type A the secondary choice. The other two types, the mixed coordination and particle type (type C) and the applicative type (type D), always appear as a secondary choice. In this study, I sum up the characteristics of these four types, in order to understand the organization of the expression of comparison in Purepecha. In another article (Chamoreau, under consideration), I propose a detailed typological analysis of the four types. Type A. Particle type Andersen (1983: 118), Stassen (1985: 45, 491), and Heine (1994: 63) stress that the so-called particle construction is heterogeneous. A typical charac-

62

Claudine Chamoreau

teristic of this construction is the presence of a specific comparative marker that accompanies the standard NP (see also Rivara 1990, 1995). In Purepecha, it is identical to the Spanish marker ke8 or to the particle that introduces a complement clause i§ki or to one of its variants (Chamoreau 2009: 259–262). In examples (17) through (20), the particle construction consists of one clause with complex structure, in which the comparee NP is encoded as the subject of the predicate, whereas the standard NP, which has no case marker, appears after the comparative marker. The quality is generally encoded by a verb, as in (18) and (19), but also by an adverb, as in (17), or an adjective, as in (20). The order follows the Spanish order when quality is expressed by an adjective or an adverb (see examples (7) and (8)). The degree marker may be the Spanish marker mas or the Purepecha morpheme sa´niteru, which means ‘more.’ This type presents four subtypes. Subtype A1. Particle constructions with the degree marker mas and the comparative marker ke The first subtype is a clear grammatical borrowing in which both the structure and the phonetic substance appear in the replica or recipient language. The particle type and the two Spanish morphological elements mas and ke are borrowed. (17) ima xu-›a-§-ti mas yo´ntakwa ke t§ i wa´tsfi-ti dem come-ft-aor-ass3 more late than pos2 son-kpos2 ‘He came later than your son.’ (Jara´cuaro-Celia28: 170) Subtype A2. Particle constructions with the degree marker sa´niteru and the comparative marker eska or e§ki This construction is a grammatical replication (also known as a calque), that is, it is produced when speakers create a new grammatical structure 8. One possible hypothesis is that ke is borrowed from Spanish because the form and the function are similar to the Spanish particle que. Nevertheless, another possibility is convergence or syncretism between the Spanish ke and a native Purepecha element. Purepecha also had a relator with the form ki, and a subordinator encoded as ka, attested in the sixteenth century. They now function in various particles such as i§ ka, i§ ki, enka, enki, and their variants. Convergence or syncretism between the two elements might have been favored because they presented a similar form and functioned in similar contexts. This topic has not yet been studied. Nevertheless, in the comparative constructions, we can consider that ke is borrowed for this function, as the entire comparative construction is borrowed or replicated.

Contact-induced change as an innovation

63

based on a model of another language, using the linguistic resources available in their own replica language (Heine and Kuteva 2003, 2005). This type of transfer does not involve phonetic substance of any kind. This is a grammatical replication in which we recognize the Spanish construction but the specific morphological elements are taken from the native language, Purepecha. In (18) the degree marker sa´niteru is analyzed as sa´ni ‘few’ and the clitic ¼teru ‘more,’ while the particle e§ ki ‘than,’ or its variant eska, is a complementizer which may introduce a complement clause (Chamoreau 2009). The degree marker is placed before the quality. (18) nanaka-et§a sa´ni¼teru tere-kuri-§ in-ti eska¼ni xi girl-pl few¼more laugh-mid-hab-ass3 than¼1 1ind ‘The girls are laughing more than me.’ (Arantepacua-Esperanza7: 99) Subtype A3. Particle constructions with the degree marker sa´niteru and the comparative marker ke In (19), we find a particular situation in which only one grammatical item is borrowed, namely the marker ke, while the degree marker is the Purepecha morpheme sa´niteru. It is thus a mixture of borrowing and grammatical replication. Logically, two possibilities exist: borrowing the degree marker mas and using the marker i§ ki, or using the degree marker sa´niteru and borrowing the marker ke. In the data, only the second option is found. In (19), we observe the same order as presented in the examples above; the quality is between the degree marker and the marker. (19) i kamisa sa´ni¼teru xuka-para-s-ti ke i§u dem shirt few¼more put-shoulder-aor-ass3 than here anapu-e-s-ti origin-pred-aor-ass3 ‘This shirt is more expensive than the one made here.’ (Ihuatzio-Agustina1: 39) Subtype A4. Particle constructions with the degree marker sa´niteru and the comparative markers ke and e§ka In this fourth subtype, the two comparative markers ke and e§ ka coexist in the same construction. This is perhaps additional evidence that ke is borrowed from Spanish in this context (see footnote 8). This redundancy

64

Claudine Chamoreau

may be explained as a ‘Purepechization’ of the subtype A3, that is, the construction with sa´niteru . . . ke, the unmarked construction. It seems that the goal of this construction is to give it a more Purepecha-like feel (Chamoreau, under consideration). (20) i§u sa´ni¼teru khe´ri-i-§-ti ke e§ ka xiniani here few¼more big/tall/old-pred-aor-ass3 than than there ‘It’s bigger here than there.’ (Ocumicho-Rutila7: 82) These four subtypes are clear examples of contact-induced restructuring. The constituent order is the same as in Spanish. The encoding of both the degree marker and the comparative marker is borrowed or replicated from the model language, Spanish. Purepecha has adopted the unmarked and more frequent Spanish comparative construction of superiority with ma´s . . . que. The particle type has superseded the verbal type (see 3.2, example (15)). This process shows clear convergence with Spanish and also indicates that the language has come to use a new strategy, exploiting morphological categories to express the degree marker and the comparative particle. The other consequence is that the quality is no longer expressed by a non-finite verb but by a verb, an adjective, or an adverb. Type B. Particle type with locative phrase The basic construction here is that of the particle type (see Type A above). The original feature of type B is the presence of a preposition accompanying the standard NP. Two possibilities exist: (i) A source-subtype (B1), in (21), with the Spanish preposition de ‘from’; the standard NP is marked as the source of a movement. (2) A static-subtype (B2), in (22), with the Spanish preposition entre ‘between’; this preposition is a particular illustration of the static locative type. Subtype B1. Source subtype. Particle type with the degree marker mas as in example (21a) (or sa´ni¼teru, as in example (21b)) and a locative phrase (21) a.

inte at§ a mas khe´ri-e-s-ti ke de §o anapu yamintu dem man more old-pred-aor-ass3 than of here origin all ‘This man is older than anyone else here.’ (Lit. ‘This man is older than of all the others from here.’) (Teremendo-Cleotilde1: 301)

Contact-induced change as an innovation

b.

65

Gervasio sa´ni¼teru prontu ni-ra-s-ti ke de ima Gervasio few¼more quickly go-ft-aor-ass3 than of dem ‘Gervasio went more quickly than him.’ (Lit. ‘Gervasio went more quickly than of him.’) (Cuanajo-Evaristo9: 102)

Subtype B2. Static subtype. Particle type with the degree marker mas and a locative phrase (22) i§u mas khe´-§-ti ke entre xini here more be big/tall/old-aor-ass3 than between there ‘It’s bigger here than there.’ (Lit. ‘It’s bigger here than between there.’) (San Andre´s Tziro´ndaro-Valentı´n4: 71) The particle type with a locative phrase shows the creation of a new type, using a process that is not attested in Lengua de Michoacan: the use of a locative phrase with the particle construction. This construction with the preposition de is attested in Spanish but the order and the conditions of use are di¤erent from Purepecha. This construction does not have the semantic and syntactic restrictions it shows in Spanish (see 3.1), and it is the dominant type in various villages, while in Spanish it is a marked construction (see Chamoreau, under consideration). Furthermore, the construction with entre is not found in Spanish to express comparative meaning (see section 5 for the analysis of this innovative construction). Type C. Mixed coordination and particle type This mixed type presents the combination of two constructions. The basic construction is the coordinated positive-negative polarity in which the comparee NP has the property while the standard NP lacks the property. The basic construction is defined as the complete one, that is, the coordination construction, a structure similar to the one attested in the sixteenth century (see example (16)); the particle construction combines with the coordination one, but presents only some features of this type. In this case, the particle construction is represented only by the presence of the degree marker. In example (23), this mixed type is formed by two clauses; the first one contains the comparee NP kumant§ ikwa›u int§arini, the degree marker mas (it is also possible to find sa´niteru), and the quality xo›epekwa xa›asti. The second clause is introduced by the coordinator ka. The negation no indicates that the standard NP lacks the property. This clause has a stripping structure; the verb is deleted, signifying that it is identical to the verb in the first clause.

66

Claudine Chamoreau

(23) kumant§ikwa-›u int§arini mas xo›epekwa xa-›a-s-ti house-loc inside more warm be there-ft-aor-ass3 ka no we´rakwa and neg outside ‘It is warmer inside the house than it is outside.’ (Lit. ‘It is warmer inside the house and not outside.’) (Janitzio-Simon1: 29) This construction is a clear consequence of the restructuring of the comparative construction domain in Purepecha. This mixed type shows interaction between internal evolution and contact-induced change. The former is shown by the fact that the coordination construction is maintained (see example (16) in Lengua de Michoacan); the latter is illustrated by the process in which verbal type is lost in favor of particle constructions. Type D. Applicative type This type has only one construction, expressing quality through a synthetic derivative structure. In (24a), the basic construction, the quality is expressed by an adjective khe´ri ‘big/tall/old,’ accompanied by a predicativizer e. In (24b), khe´ri is modified by the applicative morpheme ku, which increases the valence and introduces another argument imeri pirimpani ‘his sister,’ which is the syntactic object, and which has the role of the possessor of the quality. The subject Petu ‘Peter’ is the comparee NP, while the object imeri pirimpani ‘his sister’ is the standard NP. The superiority degree is a consequence of the modification by the applicative morpheme. Petu khe´ri-e-§-ti Peter big/tall/old-pred-aor-ass3 ‘Peter is tall/big/old.’ b. Petu khe´ri-e-ku-§-ti imeri piri-mpa-ni Peter old-pred-3appl-aor-ass3 pos3 sister-kpos3-obj ‘Peter is older than his sister.’ (Lit. Peter applies his old age to his sister.’) (Cucuchucho-Francisco3: 401)

(24) a.

This construction was not described in the grammars of the sixteenth century and is now seldom found. It shows the generally agglutinative and derivative character of the language. It is possible to hypothesize that this construction existed in Lengua de Michoacan, but then fell into disuse, until it survived only in a few villages and only with the adjective khe´ri.

Contact-induced change as an innovation

67

The four types and the di¤erent constructions are summed up in Table 1. Table 1. Comparison. Types and sub-types Type A

Type B

Type C

Type D

Particle type

Particle type with a locative

Mixed coordination and particle type

Applicative type

A1 Borrowing mas . . . ke

B1 Source localization – Borrowing mas . . . ke . . . de

Borrowing mas . . . ka no

Applicative -ku

A2 Replication sa´niteru . . . e§ ka

Source localization – Replication sa´niteru . . . ke . . . de

Replication sa´niteru . . . ka no

A3 Replicationþ borrowing sa´niteru . . . ke

B2 Static localization – Borrowing mas . . . ke . . . entre

A4 Replicationþ borrowing and replication sa´niteru . . . ke . . . e§ ka

5. An innovative construction in Purepecha In this section I analyze the constructions in type B, demonstrating that subtype B1, in examples (25) and (26), is a creation on the model of Spanish, whereas subtype B2, in example (27), constitutes an innovation. Subtype B1. Particle type with a degree marker mas/sa´ni¼teru and a locative phrase with de (25) ka Enrike mas §epe-h-ti ke de Carlos and Henry more be lazy-aor-ass3 than of Charles ‘And Henry is lazier than Peter.’ (Lit. ‘And Henry is lazier than of Peter.’) (San Jero´nimo-Adelaida1: 170)

68

Claudine Chamoreau

(26) pedru sa´ni¼teru prontu xano-nku-ti ke de t h u Peter few¼more quickly arrive-centrip-ass3 than of 2ind ‘Peter arrives more quickly than you.’ (Lit. ‘Peter arrives more quickly than of you.’) (Cuanajo-Evaristo9: 102) Subtype B2. Particle type with the degree marker mas and a locative phrase with entre (27) Pedro mas sesi-e-s-ti ke entre Xwanu Peter more good-pred-aor-ass3 than between John ‘Peter is better than John.’ (Lit. ‘Peter is better than between John.’) (San Andre´s Tziro´ndaro-Valentı´n2: 11) The constructions in type B are contact-induced changes. There can be no doubt that the four morphemes mas, ke, de, entre are taken from Spanish. But the two prepositions de and entre are not direct borrowings: although they are Spanish prepositions, since Purepecha has only postpositions, they never occur alone with the semantic features which they have in Spanish. They only appear in code-switching Spanish phrases like de veras ‘really, truly,’ la seis de la man˜ana ‘six in the morning.’ The preposition entre only occurs in Purepecha in the comparative construction, as in (27). Purepecha has postpositions and case markers which generally satisfy the use contexts of the Spanish prepositions de and entre. 5.1. Subtype B1: a creation on the model of Spanish In Spanish, as in example (28), the de-construction encoded with ma´s . . . de . . . que has specific characteristics. First, the order is the degree marker ma´s, then the quality temprano, the preposition de, the object pronoun lo, and the relative clause introduced by the relator que and the verb esperabas. Second, the use of the demonstrative lo is obligatory; this is an anaphoric strategy. Third, in this construction in Spanish a verb is obligatory after the relator (this is a relative clause). Fourth, this construction is marked and not frequent; this is a pragmatic strategy used to stress specific information (Rojas 1990b). (28) Spanish El presidente regreso´ ma´s temprano de lo que tu´ the president return.past3 more early of dem than 2ind esperabas expect.past.impf2 ‘The president returned earlier than you expected.’

Contact-induced change as an innovation

69

These four characteristics are absent in Purepecha: in subtype B1, the order is, first, the two markers mas . . . ke, and then the preposition de, which appears after the comparative marker. There is no demonstrative anaphoric pronoun, no verb after the comparative marker (this is not a relative clause as in Spanish), and in many villages north of Lake Patzcuaro, in the municipality of Quı´roga, this construction is the unmarked and dominant one, used in all contexts. We can hypothesize that the speakers have adopted the Spanish construction, adapting it with a particular strategy: they have conserved the unity and the order of the mas . . . ke particle type construction (type A), but have created a new construction, adding the standard NP in a locative phrase introduced by de. 5.2. Subtype B2 5.2.1. An innovative construction The subtype B2 strategy is di¤erent from the construction in subtype B1. Purepecha displays a use of entre which deviates from the patterns of its Spanish use. No similar construction has been found among the local Spanish speakers, nor among bilingual Spanish speakers. Purepecha speakers have apparently innovated the construction with entre, since Spanish has no comparative construction of superiority with the preposition entre. One may find a superlative construction as in (29), but the NP with the preposition entre is not obligatory in a superlative construction; it is merely additional information. (29) Spanish [Entre esto-s nin˜o-s], Juventino es el ma´s inteligente. between dem-pl boy-pl Juventino be.pres3 the more intelligent ‘Between these boys, Juventino is the more intelligent.’ In (29), with the NP introduced by entre, the nominal must be plural (or at least involve two entities), since it indicates a possibility of choice between various elements. This is not the case in Purepecha (see examples (2), (22), and (27)). The morpheme entre appears before singular items: a demonstrative ima, in example (2), an adverb xini in example (22), and a proper name Xwanu in example (27). The morpheme entre changes in meaning content (it does not indicate a possibility of choice). The use of entre in Purepecha has been extended to a new context (absent in Spanish and original in Purepecha).

70

Claudine Chamoreau

There is no correlation with other structures in Purepecha, as entre is only used in this construction, and there is no comparative construction in pre-contact replica Lengua de Michoacan with a locative pattern that might be used as a model. 5.2.2. Sociolinguistic particularities It is relevant to point out that the construction with mas . . . ke . . . de (subtype B1) essentially appears to the north of Lake Patzcuaro, in the eastern area. More specifically, this construction is attested in the four villages studied in the municipality of Quı´roga and in some villages of the Zacapu region which are in contact with the villages to the north of the lake. In the four villages (Santa Fe de la Laguna, Chupı´cuaro, San Jero´nimo P’urhenchecuaro, and San Andre´s Tziro´ndaro) this is the dominant unmarked choice, used by all speakers. San Andre´s Tziro´ndaro is the only village that also uses the innovative construction with mas . . . ke . . . entre (subtype B2). These four villages, along with Azajo, constitute a sub-area of the eastern area which exhibits great vitality (unlike the rest of the area). All of these villages include more than 87 percent Purepecha speakers (except San Jero´nimo P’urhenchecuaro, with 50 percent), and the people, even the young people, speak Purepecha in everyday conversation. This original sociolinguistic situation, in a region where language diversity is generally losing ground, is revealed through a strategy by which speakers try to distinguish themselves from others. This is also mirrored on historical, social, and cultural levels, especially in the village of Santa Fe de la Laguna, showing that they explore and use the vitality and creative possibilities of Purepecha. The B constructions constitute a distinctive characteristic of this sub-area to the north of Lake Patzcuaro. 5.2.3. A cross-linguistic tendency I consider the construction of subtype B2 to be an innovation, because the Purepecha speakers have ‘‘tinkered’’ with the Spanish constructions but have not created a construction on the model of a specific comparative Spanish construction. The motivation behind the use of the preposition entre is perhaps its meaning: it involves location (like the preposition de in subtype B1), and indicates the cognitive relation between comparison and location meaning. This leads us to a second complementary explanation: there is a general tendency to connect comparison with location and to express comparison through the locative type. This is the largest class in the typology of comparatives, comprising nearly 50 percent of Stassen’s

Contact-induced change as an innovation

71

(1985) and Heine’s (1994) samples. It could thus very easily have developed in the domain of comparison in Purepecha, since in this language spatial expressions are highly relevant in various domains (Chamoreau 2009; Friedrich 1971; Monzo´n 2004). Furthermore, this construction is in accordance with the relations between location and particle constructions developed in several languages (Andersen 1983: 168–185; Stassen 1985: 49).

6. Similar constructions in other Mesoamerican languages Stolz and Stolz claim that ‘‘Hispanicization of comparative constructions was almost commonplace among the indigenous languages of Mexico, Guatemala, Belize, Honduras, Nicaragua, and El Salvador’’ (2001: 38). Particle type is attested in all Mesoamerican families (Chamoreau 2008), generally showing the transfer of the Spanish particle construction as in Purepecha in type A – for example, borrowing in Totonac in (30) and replication in Nahuatl de Xaltipan in (31). (30) Totonac (Totonac-Tepehua, Mexico, Levy 1990: 131) pa:caps xa-tabique mas ta’:la´ que ta-pa:lhta:m wall det-brick more endure than ingr-clay ‘The wall made of bricks is stronger than the one made of clay.’ (31) Nahuatl de Xaltipan (Uto-Aztecan, Sa´nchez personal communication) neh kachi ni-nohnel tein ti-yetok-eh kal-ihtik 1 more 1-small than 1pl-be.there-pl house-loc ‘I am smaller than we [who] are in the house.’ But the transfer of the Spanish comparative construction with mas and a preposition is not very common. A review of Mesoamerican languages shows that the presence of this construction in Zoque, in example (32), is a borrowing of the Spanish construction with mas . . . de respecting the order of the elements of the model language, as in (10). (32) Zoque de Chimalapa (Mixe-Zoquean, Knudson 1980: 134) te ladriyus nea mas pfi m-pa de ka mfiki nas nea dem brick wall more have-strong of dem clay wall ‘The wall made of bricks is stronger than the one made of clay.’ (Lit. ‘The wall made of bricks is stronger of the one made of clay.’)

72

Claudine Chamoreau

It is also possible to find a borrowing of the mas . . . de . . . que construction. In this case the languages, Zoque in (33) and Otomı´ in (34), also respect the order of the elements of the Spanish construction in (11); nevertheless the constructions in these two languages do not possess restrictions like those in Spanish. For example, no verb is attested after the comparative marker. (33) Zoque de Chimalapa (Mixe-Zoquean, Knudson 1980: 135) tep mas de kphi ke cci dem more of tall than 1 ‘He is taller than I am.’ (Lit. ‘He is taller of than I am.’) (34) Otomı´ de Santiago Mexquititla´n (Otopamean, Hekking personal communication) ar Pedro ma¨s ؼar da¨ta¨ di-ge ar Mariya sg Peter more 3pres.npred¼sg tall of-that sg Mary ‘Peter is taller than Mary.’ (Lit. ‘Peter is more the tall of that Mary.’)9 The constructions found in other Mesoamerican languages have resulted from the transfer of the Spanish particle construction (see examples (30) and (31)) and the borrowing of the Spanish construction with ma´s . . . de or ma´s . . . de . . . que that are closer to the model construction than the Purepecha one (in particular because of the respecting of the order of the elements). No construction with entre has been found.

7. The strategy of innovation According to the typology proposed by Thomason (2001), it is clear that the constructions studied in this article, in particular the type B constructions, represent a replacement of older native linguistic features by interference processes. This replacement was created on the model of Spanish constructions in subtype B1 (as defined by Heine and Kuteva), but the strategy was not the same in the case of subtype B2. Another strategy is displayed. Something new has been invented. Speakers of Purepecha have taken the Spanish construction with ma´s . . . de . . . que as a point of departure, but the result diverges from it. They have also innovated using entre di¤erently from its function in Spanish, and the resulting construction in Purepecha is distinct from the comparative constructions in this language. 9. I thank Enrique Palancar for helping me to analyze this example.

Contact-induced change as an innovation

73

Speakers have transferred elements from the model language and attributed new functions to them. This is surprising because entre is not a loan word, and it is only used in this structure in Purepecha. It is di‰cult to understand the original motivation behind the transfer of entre and its use in that structure; it may be due to its locative meaning, which may express a possibility of choice between (at least) two entities. In short, Purepecha displays a use of comparative constructions with entre that deviates from the patterns of comparative construction in Spanish and from the use of entre in Spanish. The transfer of Spanish entre allows Purepecha to innovate in the expression of the comparison of superiority and in the context of the use of this Spanish preposition.

Abbreviations aor appl ass caus centrif centrip dem fem ft impf ind inf inst int it kpos loc masc mid neg npred obj past pl pos

Aorist Applicative Assertive Causative Centrifugal Centripetal Demonstrative Feminine Formative Imperfect Independent Infinitive Instrumental Interrogative Iterative Kinship possessive Locative Masculine Middle Negation Nominal predication Object Past Plural Possessive

74

Claudine Chamoreau

pred pres prog sg transf *

Predicativizer Present Progressive Singular Transfer Ungrammatical

References Andersen, P. 1983. Word Order Typology and Comparative Constructions. Amsterdam: John Benjamins. Bath, D. N. S. 1994. The Adjectival Category. Criteria for Di¤erentiation and Identification. Amsterdam: John Benjamins. Capistra´n, A. 2002. Variaciones de orden de constituyentes en p’orhe´pecha. Topicalizacio´n y focalizacio´n. In Del Cora al Maya Yucateco. Estudios lingu¨´ısticos sobre algunas lenguas indı´genas mexicanas, P. Levy (ed.), 349–402. Me´xico: UNAM. Census INEGI. 2005. http://www.inegi.gob.mx/est/contenidos/espanol/sistemas/ conteo2005/ iter2005/selentcampo.aspx Chamoreau, C. 2007. Grammatical borrowing in Purepecha. In Grammatical Borrowing in Cross-linguistic Perspective, Y. Matras and J. Sakel (eds), 465–480. Berlin: Mouton de Gruyter. Chamoreau, C. 2008. ‘Contribucio´n de algunas lenguas de Mesoame´rica a la tipologı´a de las construcciones de comparacio´n de superioridad.’ Talk at X Encuentro Internacional de Lingu¨´ıstica en el Noreste. Hermosillo: University of Sonora. Chamoreau, C. 2009. Hablemos purepecha, Wante´ juchari anapu. Morelia: Universidad Intercultural Indı´genas de Michoaca´n/IIH-UMSNH/IRD/CCC-IFAL/ Grupo Kw’anı´skuyarhani de Estudiosos del Pueblo Pure´pecha. Chamoreau, C. 2010. Development of analytic constructions in Purepecha. In A New Look at Language Contact in Amerindian Languages, C. Chamoreau, Z. Estrada Ferna´ndez, and Y. Lastra (eds). Munich: Lincom. 69–87. Chamoreau, C. under consideration. The geographical distribution of typologically diverse comparative constructions of superiority in Purepecha. Cuzzolin, P. and Lehmann, C. 2004. Comparison and gradation. Nr. 155: 1212– 1220. Dryer, M. S. 2007. Word Order. In Language Typology and Syntactic Description. Clause Structure, T. Shopen (ed.), 61–131. Cambridge: Cambridge University Press (second edition revisited). Friedrich, P. 1971. The Tarascan Su‰xes of Locative Spaces: Meaning and Morphotactics. Bloomington: Indiana University Press. Galant, M. R. 1998. Comparative constructions in Spanish and San Lucas Quiavini Zapotec. Ph.D. diss. Los Angeles: University of California.

Contact-induced change as an innovation

75

Gilberti, M. 1987 [1558]. Arte de la lengua de Michuacan. Morelia: Fimax. Heine, B. 1994. Areal influence on grammaticalization. In Language Contact and Language Conflict, M. Pu¨tz (ed.), 55–68. Amsterdam: John Benjamins. Heine, B. 1997. Cognitive Foundations of Grammar. Oxford: Oxford University Press. Heine, B. 2006. Contact-induced word order change without word order change. In Working papers in Multilingualism. Arbeiten zur Mehrspra¨chigkeit. 76. Hamburg: University of Hamburg. Heine, B. and Kuteva, T. 2003. On contact-induced grammaticalization. Studies in Language 27.3: 529–572. Heine, B. and Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Knudson, L. 1980. Zoque de Chimalapa. Mexico: Colegio de Me´xico/Centro de Investigacio´n para la Integracio´n Social [Archivo de lenguas indı´genas de Me´xico 6]. Ma´rquez Joaquı´n, P. 2007. Tarascos o P’ure´pecha? Voces sobre antiguas y nuevas discusiones en torno al gentilicio michoacano. Morelia: UMSNH-IIH/El Colegio de Michoaca´n, Gobierno del Estado de Michoaca´n/Universidad Intercultural Indı´gena de Michoaca´n/Grupo Kw’anı´skuyarhani de Estudiosos del Pueblo Purepecha/Fondo Editorial Morevallado. Matras, Y. 1998. Utterance modifiers and universals of grammatical borrowing, Linguistics 36.2: 281–331. Monzo´n, C. 2004. Los morfemas espaciales del p’urhe´pecha. Zamora: El Colegio de Michoaca´n. Nava, F. 2004. La voz media en p’urhepecha. Un estudio de formas y significados. Ph.D. diss. Mexico: FFL-UNAM. Noonan, M. 2007. Complementation. Language Typology and Syntactic Description. Complex Constructions, T. Shopen (ed.), 52–150. Cambridge: Cambridge University Press. Price, S. 1990. Comparative Constructions in Spanish and French Syntax. London: Routledge. Rivara, R. 1990. Le Syste`me de la comparaison: sur la construction du sens dans les langues naturelles. Paris: Les E´ditions de Minuit. Rivara, R. 1995. Pourquoi il n’y a que deux relations de comparaison. Faits de Langues 5: 19–39. Rojas Nieto, C. 1990a. Comparacio´n le´xica, comparacio´n sintagma´tica. In Homenaje a Jorge A. Sua´rez. Lingu¨´ıstica indoamericana e hispa´nica, B. Garza Cuaro´n and P. Levy (eds), 447–457. Mexico: El Colegio de Me´xico. Rojas Nieto, C. 1990b. El te´rmino de las construcciones comparativas de desigualdad: de o que. In Estudios de Lingu¨´ıstica de Espan˜a y Me´xico, V. Demonte and B. Garza Cuaro´n (eds), 225–241. Mexico: El Colegio de Me´xico / UNAM. Stassen, L. 1985. Comparison and Universal Grammar. Oxford: Basil Blackwell.

76

Claudine Chamoreau

Stassen, L. 2001. Comparative constructions. In Language Typology and Language Universals: an International Handbook. M. Haspelmath, E. Ko¨nig, W. Oesterreicher, and W. Raible (eds), 993–997. Berlin: Walter de Gruyter. Stassen, L. 2005. Comparative constructions. In The World Atlas of Language Structures, M. Haspelmath, M. Dryer, D. Gil, and B. Comrie (eds), 490–491. Oxford: Oxford University Press. 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 austrone´sicas y amerindias, K. Zimmerman and T. Stolz (eds), 35–56. Frankfurt/Madrid: Vervuert/Iberoamericana. Thomason, S. G. 2001. Language Contact: An Introduction. Washington: Georgetown University Press. Thomason, S. G. and Kaufman, T. 1988. Language Contact, Creolization and Genetic Linguistics. Berkeley: University of California Press.

Language contact in language obsolescence Alexandra Y. Aikhenvald

1. Preamble The di¤erence between language change in ‘‘healthy’’ and in endangered or obsolescent languages very often lies not in the sorts of change, which tend to be the same (Campbell and Muntzel 1989). It tends to lie in the quantity of change and in the speed with which the obsolescent language changes (see Schmidt 1985: 213; Aikhenvald 2002: 243–264). Language displacement frequently results in reduction of paradigms, simplification and loss of the language’s own features, and, ultimately, language shift and loss. As the obsolescent language is ‘‘retreating, contracting, as it gradually falls into disuse’’ (Dixon 1991a: 199), we expect it to be flooded with an influx of patterns and forms from the dominant language. Contact-induced changes can roughly be divided into three sorts, in terms of their stability. Following Tsitsipis (1998: 34), it appears useful to divide changes into completed, ongoing (or continuous), and discontinuous. Completed changes cover those aspects of the grammatical system of a language which do not show any synchronic variation and which go beyond speakers’ awareness (see the discussion of a Spanish-influenced passive in Purepecha by Chamoreau 2005). Ongoing or continuous changes are those in progress; here the degree of influence of the other language depends on the speaker’s competence and possibly other, sociolinguistic, variables (such as age or degree of participation in community life). Discontinuous changes are one-o¤ deviations characteristic of individual speakers. In the situation of language attrition, these often di¤erentiate fluent speakers from less proficient ones. This classification of changes is particularly important for distinguishing between old and established di¤usional processes – characterized by completed changes – and new, in-coming continuous changes making their way into a speech community. In a situation of language obsolescence, one expects to encounter a multiplicity of sporadic changes which would be considered to be mistakes by fluent speakers (if they existed). Such aberrant individual innovations are tantamount to Tsitsipis’ ad hoc or discontinuous changes. The impact of language shift as seen through

78

Alexandra Y. Aikhenvald

discontinuous changes in the context of displacive language contact is the topic of this article. 2. The various facets of language obsolescence An obsolescent language is no longer actively used or transmitted. We distinguish several kinds of social context in which this occurs.1 Firstly, an obsolescent language is no longer actively spoken by a community, and is not transmitted to the next generation. Its knowledge is often confined to a handful of last fluent speakers – as is the case for Ingrian Finnish in Estonia (Riionheimo 2002), Bare (Aikhenvald 1995), Dyirbal and Yidiny (Dixon 1991a, b), Mawayana (Carlin 2006), and Resı´garo (Allin 1975) – or to a handful of not-very-fluent speakers or semi-speakers, or even rememberers – as in the case of Nivkh, a Paleo-Siberian isolate (Gruzdeva 2002), or Nyulnyul, an Australian language (McGregor 2002; see Hill and Hill 1986, Hill 1985, for a definition of the terms). We will refer to this as ‘‘global’’ language obsolescence. Alternatively, language obsolescence can a¤ect individuals or groups of individuals living away from the language community. This is often the case with speakers of immigrant languages, spoken by groups of varied size whose major language is the dominant language of the country. These varieties are sometimes called ‘‘heritage’’ languages. The existing studies include Heritage Russian (Pereltsvaig 2008, and references there; Kagan and Dillon 2001), Heritage Italian, Heritage Norwegian, Heritage Swedish, and Heritage Czech (see Bettoni 1991; Milani 1996; Hjelde 1996; Klintborg 1999; Henzl 1981). Along similar lines, people who live away from the community where the language is actively spoken also display signs of obsolescence. Obsolescent speakers of many indigenous languages of Papua New Guinea form part of urban communities whose dominant language is overwhelmingly Papua New Guinea English and also Tok Pisin. The domain of their ancestral language is often limited to token symbolic use in speech formulas. And when the speakers attempt to use the language, its make-up is markedly di¤erent from the way it is spoken by the speech community in the original area. I have observed this ‘‘individual’’ or ‘‘localized’’ language obsolescence among Manambu speakers – see Section 3.1 (example (1)). 1. The examples discussed here reflect what Campbell and Muntzel (1989) call ‘‘gradual death’’ of a language. We do not consider instances of ‘‘sudden death’’ or ‘‘radical death’’ of a language, nor of ‘‘bottom-up death.’’

Language contact in language obsolescence

79

Instances of individual or localized language obsolescence may occur within a broader context of a ‘‘global’’ obsolescence of a language. Paumarı´, an Arawa´ language from Southern Amazonia, is gradually falling out of use, and more rapidly so in the communities on the River Ituxı´ than on the River Purus. As a result, speakers from the Ituxı´ communities display more signs of language obsolescence (Aikhenvald 2010; Chapman and Derbyshire 1991). The interest of individual language obsolescence for a student of language change lies in the possibility of comparing the obsolescent or heritage language with the variety still actively spoken in the ‘‘homeland.’’ In the case of ‘‘global’’ obsolescence, we are sometimes fortunate to have access to a description of a pre-obsolescent variety of a language. For instance, Krejnovich’s work gives us access to Nivkh as it used to be before the language stopped being transmitted to the next generation. Numerous descriptions of Ingrian Finnish allow us to trace the obsolescence of this language as it is currently spoken in Estonia (Riionheimo 2002). The grammar of traditional Paumarı´ by Chapman and Derbyshire (1991) allows us to trace the nature of obsolescence in the present-day language. The obsolescent Dyirbal (Schmidt 1985; Dixon 1991a) can be contrasted and compared with the language described by Dixon (1972) when it was still fluently spoken. And the Tariana spoken by traditional representatives of the older generation (nowadays in their late eighties) can be contrasted with the speech of younger people who are gradually relinquishing their ancestral language. A situation of language obsolescence presupposes obsolescent speakers. Their proficiency in the given language may, of course, vary (some may be considered barely ‘‘rememberers,’’ others may conserve a degree of fluency). The di¤erence between obsolescent speakers of obsolescent languages and obsolescent speakers of languages in active use elsewhere may be compared to a well-known di¤erence between societal multilingualism and individual multilingualism. The former is a social phenomenon and is of prime concern to sociolinguists. The latter reflects personal history and is of interest to psychologists more than to sociolinguists. However, we do find that processes of language obsolescence appear to be similar in the context of ‘‘global’’ and of ‘‘local’’ obsolescence (at the level of the individual speaker). This suggests the presence of shared mechanisms which could, and should, be investigated. A word of caution is in order. Even if we do have access to what can be considered a ‘‘pre-obsolescent’’ variety, we cannot always be sure that this variety did not already bear some signs of decay. When R. M. W. Dixon

80

Alexandra Y. Aikhenvald

started his fieldwork on Dyirbal in 1963, the language was actively spoken, in the domestic sphere, by several score people, including children. Over a quarter of a century, Dixon has seen the language decline ‘‘from a state in which there was an abundance of speakers . . . to one in which there is just one good consultant left for each of three dialects, with no one to go to for a second opinion’’ (Dixon 1991a: 183). But even in the good old days of the early sixties, older speakers would comment on the fact that ‘‘words used to be longer’’ in the language as they can recall it spoken in their childhood by those old people who had passed away. That is, the process of language contraction may have started long before the linguist came to the scene, and this ‘‘discourse of nostalgia’’ (Hill 1998) may reflect speakers’ awareness of this. The few older representatives of the Tariana-speaking community – the late Caˆndido Brito, Ame´rico and Jose´ Manoel Brito – can be viewed as keepers of the traditional language. However, by the time they were born (between 1911 and 1920), Tariana communities were already a¤ected by Brazilian influence, and their traditional lives were under destruction. None of the three elders could remember the full version of traditional rituals and the ritual language. We can safely assume that even their Tariana, fluent as it is compared to that of the younger generation, has already su¤ered from a certain amount of loss. We can hypothesize that this could have been accompanied by a shift to a dominant language. Sadly, in many cases the obsolescent variety is the only one which is professionally described. Allin (1975) is based on fieldwork with a handful of last speakers of Resı´garo, a North Arawak language. Carlin (2006) is based on her fieldwork with two last speakers of Mawayana, also Arawak. The same applies to Nyulnyul (McGregor 2002), Araki (Franc¸ois 2002), and quite a few other languages from many parts of the world. In none of these cases do we have access to a full ‘‘pre-obsolescent’’ variety. Most likely, we are dealing with ‘‘a mere remnant of what the language must have been like when many speakers used it as their only means of communication’’ (Haas 1941). Linguistic consequences of language obsolescence – ‘‘global,’’ ‘‘individual,’’ or ‘‘localized’’ – include simplification and reduction of grammar and lexicon. Categories absent from the dominant language are particularly endangered. So the system of numeral classifiers becomes reduced to just a few in Korean as it is spoken by young people in Canberra (Lee 1997) whose major language of communication is English. (This is also known as ‘‘negative borrowing.’’) The obsolescent language often su¤ers from stylistic reduction and dialect mixing, and also speakers’ insecurity (see Campbell and Muntzel 1989; Chamoreau 2000; Grenoble and Whaley

Language contact in language obsolescence

81

1998; Aikhenvald 2002: Chapter 11; Dixon 1991a, b, for Dyirbal and Yidiny; Helimsky 2007: 218, for Selkup). The impact of the increasingly dominant language on the receding, obsolescent language gradually falling into disuse tends to involve a massive influx of non-native forms. The outcome of this influx may result in unusual phenomena, which may include occasional borrowed bound morpheme and mixed paradigms (Section 3). If speakers tend to avoid imported forms, impending language shift may result in a spread and expansion of look-alikes and a massive calquing of structures from the dominant language and accelerated di¤usion of patterns (Section 4).2 Speakers of obsolescent languages vary in their proficiency, from fluent language users to semi-speakers with limited competence (Dorian 1973: 417, 1977). In some cases, evaluation may be possible using internal or external clues. But in many cases we have no information about the level of speakers’ knowledge: if a typologically unusual phenomenon is based on such uncertain sources, the validity of the phenomenon is cast in doubt.

3. Non-native forms in language obsolescence An influx of non-native forms is a typical feature of obsolescent speakers. In Haugen’s (1989: 67) words, ‘‘the adoption of English loans’’ was the ‘‘first great step in the direction of English’’ for immigrant speakers of Norwegian. The adoption of non-native forms often involves lexical items and also grammatical forms. Conjunctions and discourse markers, highly susceptible to borrowing under any circumstances of language contact, are the ‘‘usual suspects.’’ In Section 3.1, we discuss relevant examples of individual language obsolescence in Manambu, comparing a fluent and an obsolescent speaker. We 2. Language contact does not explain all the discontinuous changes in language obsolescence. For instance, terminal speakers of Arvanitika Albanian in Greece sporadically lose gender and number agreement; their entire system of tenseaspect-mood categories is disintegrating – imperfective past forms are not used at all, and the marking of grammatical person is ‘‘morphologically distorted’’ (Tsitsitpis 1998: 44–62). This ‘‘agrammatism’’ cannot be explained by ‘‘negative borrowing,’’ that is, loss of categories not present in the dominant language, Greek, since Greek possesses all the categories now lost in the obsolescent Arvanitika (Sasse 1992b: 69–70). Changes in language obsolescence may be motivated by language-internal processes (see, for instance, Dixon 1991b; also Gruzdeva 2002).

82

Alexandra Y. Aikhenvald

then turn to the obsolescent Bare, a North Arawak language from Venezuela and Brazil, and compare two sources on the language which display varying degree of obsolescence. In these instances we can argue that language contact in the situation of obsolescence does not produce any remarkable results – the e¤ects are the same as may have occurred in language contact of a non-replacive nature. This is consistent with the idea that an increase in the quantity and the speed of change is the major e¤ect of language obsolescence.3 Language shift in the context of language obsolescence may also result in inclusion of some less likely candidates for borrowing – personal pronouns, both free and bound. In Section 3.2 we look at Mawayana, a North Arawak language with its last two speakers in Suriname in Trio and Waiwai-speaking communities, and then turn to Resı´garo, a moribund North Arawak language in northeastern Peru, which has undergone a massive impact of Bora and Witotoan. Can the influx of non-native forms in language obsolescence obscure its genetic a‰liation? This is the topic of Section 3.3. 3.1. Non-native free forms: following a beaten path 3.1.1. Manambu Those speakers of Manambu (a Ndu language from the East Sepik area) who live in urban centers and rarely use the language employ numerous non-native forms. An obsolescent speaker who had spent much of his life in an urban town speaking Tok Pisin (the major lingua franca of Papua New Guinea) produced (1). Later on, a fluent speaker volunteered (2), as something she would have said. The Tok Pisin forms are in italics. The form okey comes from English. It is also widely used by speakers of Tok Pisin. Non-native forms are in bold.

3. I have undertaken extensive fieldwork on Manambu (see, for instance, Aikhenvald 2008b), Tariana (see, for instance, Aikhenvald 2002) and also Bare (I worked with the last fluent speaker of the language). For each language, I have recorded a substantial number of texts and natural conversations. ALL the examples in this paper (as in my other work) come from spontaneous discourse. In my fieldwork, I avoid elicitation as being methodologically flawed (see Aikhenvald 2007, Dixon 2007 for further fundamentals of linguistic fieldwork).

Language contact in language obsolescence

83

Obsolescent speaker: (1) asa:y kiya-d-k aw wuna amay father die-3sg.m-compl.ds conn Iþlkþf.sg mother du-ak ra:l okey man-dat marryþ3f.sg.subj OK ata lukautim-d-dwun tasol a then look.after-3sg.m.subj-1sg.m.o but,only that.f.sg sikul-r yi-dwun school-all go-1sg.m.subj

namba tu number two

taim time

‘After my father died, then my mother married a second man, OK, then he looked after me, only that at that time I went to school.’ Fluent speaker: (2) asa:y kiya-d-k aw wuna amay father die-3sg.m-compl.ds conn Iþlkþf.sg mother nk-d du-ak ra:l ya:kya another-sg.m man-dat marryþ3f.sgsubj all.right ata yakwiya-d-dwun aw a skr then look.after-3sg.m.subj-1sg.m.o conn that.f.sg time sikul-r yi-dwun school-all go-1sg.m.subj ‘After my father died, then my mother married a second man, all right, then he looked after me, only that at that time I went to school’ The two versions share one established loan word, sikul ‘school.’ Nonnative forms are not necessarily restricted to lexical items. Example (1) shows that discourse markers, numerals, and conjunctions are imported from the dominant language.4 Influx of loan forms is a striking feature of ‘‘globally obsolescent’’ languages. Extensive lexical impact of English has been observed in the speech of the last speakers of Nyulnyul, an Australian language (McGregor 2002: 177). Traditional Gooniyandi and Warrwa did not have coordinating conjunctions: the remaining obsolescent speakers use English forms nd (from and ) and  (from or). Traditional Nyulnyul did 4. In a situation of obsolescent speakers whose usage is unstable, the boundary between loans and code-switches is even harder to draw than in other languagecontact situations. This is the reason why I use the term ‘‘import’’ to avoid using either ‘‘borrowing’’ or ‘‘code-switch.’’

84

Alexandra Y. Aikhenvald

have a conjunction agal ‘and’; the two remaining fluent speakers use the English import nd. Borrowing conjunctions and discourse markers in itself is not a symptom of impending language death. Hamp (1989) and Johanson (2002) have shown that allowing a certain number of loan forms by no means endangers the language; the opposite can be true. Many fluent speakers of Manambu in the villages use the English discourse marker okey and some occasionally slip in the Tok Pisin tasol ‘but’ as a replacement for the polysemous aw ‘then, but, or’ (see Aikhenvald 2008b, 2009a). This confirms the general assumption that language obsolescence tends to enhance the tendencies present in a ‘‘healthy’’ language. 3.1.2. Bare A comparison between two di¤erent stages of language obsolescence of the same language points in a similar direction, that of increased influx of nonnative forms. Lopez Sanz’s (1972) brief grammatical description of Bare, a North Arawak language from the Upper Rio Negro area in Venezuela, is based on the analysis of materials (including several texts) collected in the late 1960s from two remaining fluent speakers of the language from Santa Rosa de Amanadona (with a total population of ethnic Bare of 140). There are hardly any loans from Spanish, either lexical or grammatical. Nowadays, people in Venezuela who identify themselves as Bare in Venezuela speak Spanish; the Bare in Brazil speak Portuguese (some also know Nheeˆngatu´, or Lı´ngua Geral, a Tupı´-Guaranı´-based lingua franca of the area). In 1991, I worked with the late Candela´rio da Silva (1921–1992), from the Tiburi community, near Cucui, Amazonas, Brazil. Candela´rio’s family moved in 1912 from Venezuela, fleeing from an uprising. His family maintained links with relatives in Venezuela in the communities of Puerto Ayacucho, San Fernando de Atabapo, and Santa Rosa de Amanadona. According to Candela´rio, all the remaining speakers of Bare in the above mentioned localities in Venezuela were older than himself. He frequently referred to his elderly aunts in Puerto Ayacucho as authorities to consult with on words in Bare he himself could not remember. Candela´rio underwent traditional male initiation (his account of it is in Aikhenvald 1995: 52–55), and insisted that he had grown up with Bare as his first language. He was fluent in Bare, his father’s language. His late mother, herself a speaker of Mandawaka (another extinct North Arawak language of the area), had always spoken Bare to him. After her death about 30 years prior to our encounter, he had kept his ancestral language which he used

Language contact in language obsolescence

85

to talk to himself, especially when he used to get drunk (and this, according to Candela´rio, was not infrequent).5 Candela´rio was quadrilingual: his main home language was Nheeˆngatu´, and he was equally fluent in Spanish and Portuguese. His children spoke Portuguese and Nheeˆngatu´. The variety of Bare recorded by Lopez Sanz in Santa Rosa de Amanadona has a richer morphology than the language of Candela´rio. For instance, it has a variety of aspectual and modal markers (e.g. -phe´i ‘durative’ and -ya ‘dubitative’), and a marker of reported speech -man not attested in the corpus collected from Candela´rio. Verb forms attested in Lopez Sanz (1972) contain up to five su‰xes, whereas Candela´rio never used more than one su‰x on the verb.6 Another major di¤erence between texts and examples in Lopez Sanz (1972) and the corpus recorded from Candela´rio is the abundance of Spanish and Portuguese forms, just as would be expected in the case of advanced language obsolescence. Many of these are not lexical forms. Candela´rio made an e¤ort to avoid Spanish or Portuguese forms: the few consistent exceptions include playa ‘sand’ (from Spanish playa ‘beach’) instead of either khaadi ‘sand, earth’ (Arihini variety) or kadieho (Ihini variety: Natterer 1831), precisa‘need, require’ (from Portuguese precisar ‘need’), and gata- ‘spend, waste’ from Portuguese or Spanish gastar ‘spend’). Spanish subordinating conjunctions occur where speakers of the more traditional variety recorded by Lopez Sanz (1972) would use a sequencing clitic -ka. This morpheme in Bare, just like in many other Arawak languages of the area, has a variety of meanings: it marks adverbial clauses 5. My corpus contains over 150 pages of texts and dialogues, and word-lists. See Aikhenvald (1995) for a grammatical analysis of the material assembled, and a survey of literature on Bare. At the time of my work with Candela´rio, and writing the grammar, I did not have access to Lopez Sanz (1972). Materials in Lopez Sanz reflect some language attrition. Traditionally, Bare had two major varieties – Arihini (‘‘the ones from here’’) and Ihini (‘‘the ones from there’’: see Aikhenvald 1995). Newly available materials collected by Johann Natterer in 1831 demonstrate the existence of lexical di¤erences between the two varieties. In the texts and examples in Lopez Sanz (1972), lexical items from the two varieties appear in free variation. Such dialect mixture, or, in Dixon’s (1991a) words, dialect merging, is typical of language obsolescence. Candela´rio knew of the two varieties, but could not tell them apart. 6. Some di¤erences between the variety of Santa Rosa de Amanadona recorded by Lopez Sanz and the language of Candela´rio may be due to additional dialectal or idiolectal variation: these include Santa Rosa de Amanadona he´in and Candela´rio’s hena for declarative negation. Note that I preserve the transcription given by Lopez Sanz (1972) for the examples from Santa Rosa de Amanadona.

86

Alexandra Y. Aikhenvald

of most types except purposive, and regularly occurs on conditional, temporal, and complement clauses (see Aikhenvald 2006b for Tariana, and discussion there). An example of -ka, from Lopez Sanz (1972: 80), is in (3) (in boldface); note that in a negative construction -ka attaches to the negation (glosses are supplied by me): Bare: Santa Rosa de Amanadona (3) hena´-ka i-kasa he´in i-nika´-waka neg-seq 3gs-arrive neg 3sg.m-eat-neg ‘If he does not come, he does not eat.’ The polysemous -ka also appears in Candela´rio’s texts, as shown in the example below from an autobiographical story (also see Aikhenvald 1995: 48–50). Clauses are in brackets, for ease of reference. Bare: Candela´rio da Silva (4) [nu-min˜a¸i ø-maha niku] [a¸i bi-pa¸ata-ni 1sg-master 3sg.m-say 1sgþfor here 2sg-money-poss kuma¸ehe] [bi-katehesa-ka] [beke´ badaha´naka biku˜ big 2sg-know-seq fut one.day 2sgþfor ahaw bi-wakhid’a-ka] [hena-ka bi-katehesa] with what 2sg-live-seq neg-seq 2sg-know [ phin˜uka˜ bi-pa¸ata-ni] 2sg-throw-decl 2sg-money-poss ‘My master said to me: here is your big money, if you know something, one day you will have what to live with, if you do not know, you will throw away your money.’ The sequencing marker is not used to introduce speech reports: as shown in (4), speech reports are juxtaposed to the verb of speech. Besides the sequencing -ka, Bare had an adverbial form abeuku ‘when, as soon as; then’ used both by the two speakers in Santa Rosa de Amandona and by Candela´rio. If accompanied by the sequencing -ka on the verb, abeuku is a temporal linker ‘when.’ Example (5) comes from a mythical text recorded by Lopez Sanz (1972: 83): Bare: Santa Rosa de Amanadona (5) isı´nka abe´uku ihı´wa-ka Puluna-minali. . . 3sg.mþlike when 3sg.mþgo ?-master. . . ‘It was when Pulunaminali (the master of all animals) went (round). . .’

Language contact in language obsolescence

87

In (6), from an autobiographical story by Candela´rio, abeuku is also accompanied by -ka on the verb (in the last clause). Bare: Candela´rio da Silva (6) da-ya¸aki nu-maha dem-whisky 1sg-say ke nihiwa˜ abeuku that 1sgþgo when

i-ku me-maha ni-ku 3sg.m-for 3pl-say 1sg-for i-ma´khi-ka sa ya¸aki 3sg.m-finish-seq dem whisky

‘Then we drank, we managed to drink all the whisky, I said to him, they said to me that I shall go when the whisky finishes.’ In the few examples of abeuku without an accompanying sequencing -ka in the variety of Santa Rosa de Amanadona, the form means ‘then’ (Lopez Sanz 1972: 84): Bare: Santa Rosa de Amanadona (7) abe´uku humadan then 3f.sgþleave ‘Then she lets (him) go.’ Unlike the two speakers from Santa Rosa de Amanadona, Cande´lario used abeuku as a temporal linker without the accompanying -ka on the verb. In (8), -ka appears in the preceding clause, so its absence in the third clause (introduced with abeuku) could be explained as an instance of ellipsis: Bare: Candela´rio da Silva (8) [me-nika kubati ] abeuku idi-ka 3pl-eat fish when then/there-seq abeuku bed’a-waka me-nika matsuka when nothing 3pl-eat manioc.flour ‘They (dogs) eat fish, when/if it is there, when (there is) nothing, they eat manioc flour.’ However, in other examples like the one in (9) -ka is simply not used, and abeuku is the only linker: (9) bihiwa´ awehe´ntei abeuku i-makhi 2sgþgo hereþelative when 3sg.m-finish ‘You will go away when it (the drink) finishes.’

88

Alexandra Y. Aikhenvald

Candela´rio insisted on translating abeuku as ‘when’ (Portuguese quando). The conjunction occupies the same clause-initial position as quando in Portuguese (or cua´ndo in Spanish). The apparent obsolescence of the sequencing -ka in the presence of abeuku may indicate that Candela´rio was adopting a Spanish-Portuguese strategy for temporal linking. He never used this Spanish-Portuguese form himself. He freely used other Spanish or Spanish-Portuguese conjunctions. Just occasionally, the verb in a subordinate clause introduced by a conjunction would be accompanied by -ka. The temporal mientre ke (from Spanish mientras que) ‘while, whereas’ is accompanied by -ka in (10): (10) mientre-ke nu-naku´da-ka i-ma¸e-d’a kubati while-that 1sg-go-seq 3sg.m-steal-inch fish ‘While I was gone, he (the dog) started stealing the fish.’ The causal purke ‘because’ (from Spanish porque) is used on its own in the penultimate clause of (11). It is accompanied by -ka in the second clause of (12): (11) idi me-ma´ha-ka [ke hena me-yehe´-waka then 3pl-say-decl that neg 3pl-can-neg me-dia-sa-ka nu˜ ] [ purke hena hnuwin˜a-waka 3pl-drink-caus-seq i because neg 1sgþfall-neg ´ ya¸aki ahaw] [hena hnuwin˜a-waka] whisky from neg 1sgþfall-neg ‘Then they said to me that they could not make me drunk, because I do not fall down from whisky, I do not fall down.’ (12) [damakaru-kua nu-¸ehedi ] [ purke nu-¸ehedi nu-yuwahada-ka] jungle-locþlong 1sg-like because 1sg-like 1sg-walk-seq [ pero nu-witi hena-hana yada-ka-na˜ ] but 1sg-eye neg-more 3sg.mþsee-decl-perf ‘I enjoy the jungle, because I like to walk, but my eyes do not see any more.’ The linker ke (from Spanish, Portuguese que) is used to introduce speech reports, as in (11) and (6). About 60 percent of speech reports in the corpus contain ke. This same form occurs in the meaning of ‘so that,’ as in (13).

Language contact in language obsolescence

89

(13) [bihı´wa behe´wa] [kuhu˜ ke id’u´a¸i beke bı˜ ] 2sgþtake 2sgþfrom he that well fut you ‘Take it (poisoned pillow) from you, so that you will be well.’ Candela´rio used other Spanish conjunctions, for instance, the coordinator pero ‘but’ (from Spanish pero ‘but’) shown in (12). None of the SpanishPortuguese conjunctions appear in the Santa Rosa de Amanadona variety. Note that conjunctions occupy the same place as in Spanish. In contrast to the other documented variety of Bare, and to most other Arawak languages, Bare spoken by Candela´rio is losing the sequencing enclitic -ka, a marker which has no equivalent in either Spanish or Portuguese. This is an instance of ‘‘negative borrowing.’’ We can thus conclude that the influx of non-native forms into the speech of the last fluent speaker of Bare is accompanied by leveling of structures. The obsolescent Bare imports Spanish and Portuguese forms, and also becomes more similar to the dominant Spanish and Portuguese in terms of its grammatical structure. Conjunctions – especially free forms – are among the most borrowable elements of the language (the interested reader is advise to consult Stolz and Stolz 1996 with special focus on American Indian languages; Matras 1998, and Aikhenvald 2006a). As stated at the end of §3.1.1, the fact that Spanish and Portuguese conjunctions have been borrowed into Bare should not be considered as a special phenomenon in language obsolescence. What is indicative of Bare as an obsolescent language is the high number of loans from the dominant languages. That is, this contactinduced change in language obsolescence appears to follow a beaten path, albeit at an increased rate. we can recall, from §1 of this paper, that the di¤erence between language change in vital, and in obsolescent language, may lie in the ‘quantity of change’. It is indeed the case here. 3.2. Influx of non-native free forms: unusual patterns We now turn to some rather unusual borrowing patterns in obsolescent languages. In a number of instances, obsolescent languages borrow personal pronouns from the dominant language. In the examples available, pronominal forms which express categories attested in the dominant language but absent from the obsolescent one may get borrowed. 3.2.1. Mawayana Mawayana (Carlin 2006) is a highly endangered North Arawak language spoken by just two elderly people in a village where Trio and Waiwai,

90

Alexandra Y. Aikhenvald

from the Carib family, are the dominant languages. The two remaining speakers of Mawayana have little opportunity of using the language, and are aware that when they go, so will Mawayana. Just like most other Arawak languages, Mawayana originally had first, second, and third person, without distinguishing between first person plural inclusive (I and you) and exclusive (I and a third person, excluding you). In contrast, Waiwai and Trio have di¤erent forms for first person inclusive and for first person exclusive. As a result of influence from Waiwai and Trio as dominant languages with an obligatory distinction between inclusive and exclusive, the two remaining speakers of Mawayana consistently use the Waiwai pronoun amna to express the concept of first person plural exclusive (e.g. Waiwai amna krapan ‘our (excl) bow’). The original first person plural prefix wa- in Mawayana has been reinterpreted as inclusive. (14) amna 1þ3pn

saruuka fishtrap

‘Our (excl) fishtrap.’

(14b) wa-saruuka 1pl.poss-fishtrap ‘Our (incl) fishtrap.’

The borrowed form comes from Waiwai. However, the behavior of the verb bears an impact from Trio: in Trio the first person exclusive pronoun requires a third person prefix on the verb, while in Waiwai the third person singular prefix is often dropped. Example (15) shows that Mawayana follows the Trio pattern of person marking: (15) amna rı¨-me 1þ3pn 3a-say.pres ‘We (excl) say.’ The first person inclusive is marked with the Mawayana prefix wa- (originally first person plural): (16) wa-me 1incl.pl-say.pres ‘We (incl) say.’ Carlin (2006) is the first summary of the grammatical features of the language in the light of language contact (and a full grammar is in progress). The borrowed form amna does not occur in the previous records of the language, which include longish lists of words and phrases in Howard (1986), and materials in Farabee (1918: 283–286) and Schomburgk (1848), all collected when the language was more actively spoken than it is at

Language contact in language obsolescence

91

present. This suggests that borrowing a pronoun – something not unheard of, but rather unusual – could be the result of excessive influx of nonnative forms characteristic of Mawayana as an obsolescent language. 3.2.2. Resı´garo The genetic a‰liation of Resı´garo – a small language spoken in northeastern Peru surrounded by speakers of Bora and Witotoan groups – with the Arawak family was established by Igualada (1940) and Igualada and Castellvı´ (1940); also see Loutkotka (1968: 136). The first extensive materials on the language published by Rivet and de Wavrin (1951), and based on the data collected by de Wavrin in the early 1930s, provided ample evidence in the same direction (see Payne 1985 for a summary). The group itself comprised not more than a thousand people at the time of Whi¤en’s (1915) travels in the area. The first mention of Resı´garo (Recı´garo), by Hardenburg (1910), places it among other Witotoan groups. Tessmann (1930: 583) does not provide linguistic a‰liation, but states that culturally they are close to the Bora-Witoto, and linguistically are ‘‘perhaps close to Bora.’’7 At that time, the language was still actively spoken. Note that there is no evidence of any genetic relationship between Bora-Witotoan and Arawak languages (see Loukotka 1968; Aschmann 1993). In his pioneering salvage grammar of Resı´garo, based on fieldwork with ten remaining speakers whose major language was Bora, Trevor Allin (1975) came to a di¤erent conclusion. The sheer number of Bora, and also Witotoan, forms in Resı´garo indicated to him that the languages were genetically related. He did not deny that Resı´garo belongs to the Arawak family, but suggested that, given the high percentage of shared forms between Bora, Witotoan languages, and Resı´garo, the limits of Arawak should be expanded, and Bora and Witotoan be included. There is, however, no doubt that the impressive number of Bora and Witotoan forms in Resı´garo are due to borrowing (see Payne 1985, and detailed discussion in Aikhenvald 2001). These lexical loans constitute about 24 percent of the vocabulary, and include just a few verbs and numerous nouns, covering body parts plus a few other items such as ‘fish’ and ‘hill.’ The most striking is the fact that ‘‘core’’ lexical items, such as terms for body parts, are shared with Bora or with Witotoan languages. ¨ ber die Ressı´garo ist nichts Na¨heres bekannt. Sie geho¨ren kulturell sicher 7. ‘‘U zu der Uitoto-Boragruppe und sprachlich vielleicht in die Na¨he der Bora.’’

92

Alexandra Y. Aikhenvald

However, the lexical data published by Rivet and de Wavrin (limited as they are) often do not register a loan. A prime example is the word for ‘tooth,’ Resı´garo -ehe´pe´ ‘tooth,’ which is similar to Muinane Witoto iı´pe, Proto-Witoto an *pe (Aschmann 1993). The reflex of the Proto-Arawak form *nene (Aikhenvald 2001) survives in Resı´garo -one´ne´ ‘front teeth’ (Allin 1975). Rivet and de Wavrin (1951: 213) give the form wo˜-ne (1pl-tooth) ‘tooth,’ and no form similar to Bora or to Witotoan. The Resı´garo described by Allin uses borrowed numbers ‘one’ and ‘two’ (see Table 1). This is quite remarkable for an Arawak language, since lower numbers (if they exist at all) generally appear to be rather resistant to borrowing. And the overwhelming majority of Arawak languages preserve the reflexes of Proto-Arawak forms (fourth column in Table 1). Once again, Rivet and de Wavrin (1951) register di¤erent forms, which are clearly Arawak in origin. The form for ‘two’ shows the e¤ects of the phonological process *y > tz found in other cognates with ProtoArawak. Does this imply that pre-obsolescent Resı´garo was more Arawak-like in its lexicon and grammar? In all likelihood, yes.

Table 1. Numbers ‘one’ and ‘two’ in Resı´garo, Bora and Arawak No.

Resı´garo (Allin 1975)

Bora

Resı´garo (Rivet and de Wavrin 1951)

ProtoArawak

one

sa-cl

tsa-cl

‘apa# (ha)pe`ne

*pa

two

migaa-

mı´e´e´/mihaa-cl

‘e(i)tza#m

*yama

Bora influence on Resı´garo grammar goes further than free forms (see Aikhenvald 2001 for a detailed discussion of structural influence of Bora on Resı´garo, and also the discussion of borrowed classifiers). Borrowed bound morphemes include one pronoun, number markers, oblique case markers, and also classifiers. The independent pronouns and cross-referencing prefixes in Resı´garo (where they are mostly used to mark A/Sa and as possessors of inalienably possessed nouns) are compared to Bora in Table 2 (Allin 1975: 116–117; Thiesen 1996: 33). Borrowed morphemes are in boldface.

Language contact in language obsolescence

93

Table 2. Pronouns in Resı´garo and in Bora Resı´garo

Bora

Pronouns

prefixes

Pronouns

Prefixes (poss)

Prefixes (subj)

1sg

no´

no-

oo´

ta-

—

2sg

phu´, pha

p-

uu´

di-

—

3sg m

tsu´, tsa´

gi-

diı´bye i-, a´a´di-

—

3sg f

tso´

do-

diı´lle

1incl du m

fa-musi

mee´

1incl du f

fa-mupi

f-/_h va/_elsewhere

1excl du m

muu-musi

mu´u-

muhtsi

1excl du f

muu-mupi

2du m

ha-musi

2du f

ha-mupi

3du m

na-musi

3du f

na-mupi

1pl incl

fa-a, fu´, fa

f/ua-

mee´

1pl excl

muu-a, muu

—

muu´ha

2pl

ha-a, hu

i- (impv)

a´muu´ha

amua´-

3pl

na-a, hna´

na-

diı´tye, aa´tye´

aa´t hje-

me-

muhp hu´-, i- (impv)

a´-muhtsi

amua´-

a´-muhpi n-/_h na-

diitye´-tsi

meaa´t hje-

diitye´-p me-

—

Unlike most other Arawak languages but similarly to the Bora-Witotoan group, Resı´garo has inclusive versus exclusive opposition in first person non-singular, and also a dual number. The first person plural exclusive pronoun muua was borrowed from Bora, similarly to the way the last speakers of Mawayana introduced a Waiwai form to cover the same meaning. In Resı´garo, it was subsequently reanalyzed as consisting of a

94

Alexandra Y. Aikhenvald

prefix muu- and a particle -a, following the analogy of other non-singular pronouns in the language itself, such as na-a ‘third person plural’ and fa-a ‘first person plural inclusive.’ This shows the linguistic creativity of the last speakers, captured by Sasse’s (2001) colorful metaphor, the ‘‘Phoenix from the ashes’’ (in the spirit of Dorian 1999, and Dal Negro 2004). The Resı´garo dual markers feminine -mupi, masculine -musi (also from Bora: see Table 2) combine with muu- reanalyzed as a bound form. Unlike other pronouns, the first person plural exclusive has no corresponding prefix used with nouns and with verbs, which may point towards its later origin. The Bora forms in Resı´garo are in bold in Table 2. In their comparatively detailed discussion of personal pronouns, free and bound, in Resı´garo, Rivet and de Wavrin (1951: 204–206) do not mention the first person plural exclusive form (the analysis of pronominal markers occupies about a half of their short grammatical summary: 204– 209). They do not mention the number markers on nouns at all. We can hypothesize that the introduction of non-native free and bound pronominal forms by the last speakers of the language is likely to be a result of contact-induced change in the situation of extreme linguistic stress. This is not to say that the Resı´garo described by Rivet and de Wavrin (1951) had no loans from Bora or Witotoan; to the contrary. One example is Resı´garo tee´ ´ı (Allin 1975), tehe( y)hı´ (Rivet and de Wavrin 1951) ‘river,’ Bora thee´- i, Proto-Bora-Muinane *tee´- i. The Proto-Arawak form is *huni ‘water, river.’ A reflex of this form is attested in Resı´garo’s closest genetic relatives Tariana, Baniwa, and Piapoco as uni ‘water, river.’ Further bound morphemes borrowed from Bora into Resı´garo include markers of masculine and feminine dual, oblique cases, and numerous classifiers (see Aikhenvald 2001; Allin 1975; Thiesen 1996). None of these are mentioned by Rivet and de Wavrin (1951): we may hypothesize that the influx of borrowed morphemes into the obsolescent language is a recent phenomenon, but we have no means of definitely proving this. Borrowing a pronoun, free or bound, is not unheard of, but is quite unusual (Gardani 2005). Third person plural pronouns they, their, them in English are considered to be borrowings from a Scandinavian source (Campbell 1997; Baugh 1957: 120). Miskito, a vibrant Misumalpan language, is said to have borrowed first and second person singular pronouns from Northern Sumu (Campbell 1997, based on Ken Hale, p.c.), also Misumalpan. Further examples of borrowing individual free pronominal forms come from Matiso¤ (1990: 113) and Newman (1977, 1979a, b).

Language contact in language obsolescence

95

Campbell (1994) reports that Alsea, an isolate from Oregon, borrowed a whole set of Salishan pronominal su‰xes.8 However, the instances such as Mawayana and Resı´garo should be treated with caution as bona fide examples of borrowing pronouns. The fact that these borrowings were documented at a stage when both Mawayana and Resı´garo are used by just a handful of speakers whose major language is di¤erent alerts us to a potential e¤ect of a massive influx of non-native forms characteristic of the last stages of a language’s life. Can a massive influx of borrowed forms obscure a language’s a‰liation? This takes us to our next section. 3.3. Language obsolescence and language a‰liation It is well known that teasing apart similarities due to genetic inheritance from those due to borrowing of varied kinds is one of the hardest problems in comparative linguistics (cf. the classic controversy between Boas and Sapir: see Swadesh 1951). Ideally, if two languages descend from the same ancestor, the forms and their meanings must be easily relatable, via the application of established rules for phonological change and semantic change. In reality, the distinction between inherited and di¤used similarities may be di‰cult to draw, especially in a situation of prolonged and uninterrupted di¤usion of cultural and linguistic traits across an area; see, for instance, Dixon (1997; 2002), Dench (2001), and Heath (1978), for the Australian area, and further examples in Aikhenvald (2006a). Similarities between languages can be suggestive of a genetic relationship, but not su‰cient to postulate it with full assurance. Murrinh-patha and Ngan.gitjemerri, two languages spoken in the Daly River region of Northern Australia, share just cognate paradigms for portmanteau forms of inflective simple verbs, but scarcely anything else in grammar and almost no lexicon (Dixon 2002: 675). The paradigm of free pronouns is the only fully ‘‘Chadic’’ feature of the Tangale group (Jungraithmayr 1995). Such examples are bound to remain ‘‘fringe’’ puzzles to comparative linguists. The case of Resı´garo is rather instructive in this respect. The influx of Bora and Witotoan forms into this language led Trevor Allin to believe that the language was related to Bora and to Witotoan (Allin 1975). Payne 8. Another frequently given example of a putative borrowing of part of the pronominal paradigm comes from Kambot (or Botin), from the Grass family in New Guinea (Foley 1986: 210–211). A closer look at the paradigm of Kambot pronouns in the original sources (Laycock and Z’graggen 1975; Pryor 1990) shows that this hypothesis is based on misinterpretation of the data (see Aikhenvald 2009b, for a full analysis).

96

Alexandra Y. Aikhenvald

(1985) undertook a careful reconstruction and comparison with the previous stage of the language captured by Rivet and de Wavrin (1951), to prove that the language is not Bora-Witotoan. But what if all we have is a highly obsolescent stage? An almost extreme example of influx of non-native forms into a pronominal paradigm and its restructuring comes from Marrku, the traditional language of Croker Island (Australian area) (Evans et al. 2006; Evans 2007). Like many Australian languages, Marrku has been on the decline for many decades. It was reported that by 1939 there were only five speakers left (Evans et al 2006: 2); by 1991 there were only two semispeakers who were then highly proficient in other indigenous languages of the area (especially Iwaidja). The verb paradigms accessible to Evans (2007) show a curious picture: while there is strong evidence from body-part prefixes (Evans 2000) in favor of an erstwhile genetic relationship between Marrku and other Iwaidjan languages, verbal paradigms in Marrku – collected from obsolescent speakers – contain massive borrowings from Iwaidja and its relative Ilgar. This massive influx, without any previous stage of the language to be compared with, makes exact genetic classification of Marrku an almost impossible task (Evans 2007).

4. Further outcomes of language contact in language obsolescence An influx of foreign forms is not a universal outcome of language obsolescence.9 We saw above that an obsolescent language may tend to rapidly become structurally similar to the dominant one. Almost all the categories present in Bora are expressed in Resı´garo; Mawayana replicates the Trio and Waiwai patterns (without necessarily borrowing the forms). Nivkh, a Paleo-Siberian isolate on the path towards extinction, has undergone massive restructuring of imperative paradigms under the influence of Russian (see Gruzdeva 2002). Similar examples abound.

9. Last speakers often avoid consciously using loan forms, even if they were used in the language. R. M. W. Dixon reports that Dick Moses, one of the very last fluent speakers of Yidiny, made sure his language was free of English intrusions. As Dixon (1977: 29) reported, ‘‘Moses has eliminated what were certainly established English loan words’’; ‘‘in place of mudaga ‘motor car’ and biligan ‘billy can,’ he uses dundalay and gunbu:l which he said were originally the avoidance style forms for these items.’’ Similar examples of purism have been documented for Arizona Tewa (Kroskrity 1993).

Language contact in language obsolescence

97

Intensive language contact in the situation of language obsolescence goes together with enhancement of already existing similarities. Forms in the obsolescent language which are similar to those in the dominant one tend to become more frequent, and to assume the meanings influenced by the dominant language. Ingrian Finnish spoken by a handful of Finns scattered around Estonia is a case in point. Most speakers are undergoing a rapid shift to Estonian. The two languages are closely related and structurally similar; as a result, it is not always possible to distinguish Estonian and Finnish forms. The most striking foreign form recorded in the language of the few remaining speakers is the past tense marker -si- employed instead of the Ingrian Finnish -i- (Riionheimo 2002: 201–202). This past tense marker is highly productive in Estonian; its appearance in Ingrian Finnish can thus be explained by the influence of the dominant language. But there is also a language-internal explanation: there is a subclass of verbs in Ingrian Finnish which requires -si- past rather than -i- past. Similarity in form of the Ingrian Finnish and the Estonian past marker is a strong contributing factor to its increased frequency in the moribund Ingrian Finnish. Other than that, speakers tend to avoid using Estonian forms. In a situation of traditional inhibition against borrowed forms, growing language obsolescence may go hand in hand with expansion of those morphemes that have the same form in the obsolescent and in the dominant language. Tariana is the only Arawak language spoken in the Vaupe´s basin in northwest Amazonia (spanning adjacent areas of Brazil and Colombia). This used to be a well-established linguistic area, characterized by obligatory multilingualism based on the principle of linguistic exogamy: ‘‘those who speak the same language as us are our brothers, and we do not marry our sisters’’ (see Aikhenvald 2002 and references there). Languages spoken in this area traditionally included the East Tucanoan languages Tucano, Wanano, Desano, Piratapuya, Tuyuca (and a few others), and the Arawak language Tariana (now spoken by over 100 speakers in two villages). Speakers of these participate in the exogamous marriage network which ensures obligatory multilingualism. Nowadays, Tariana is no longer spoken by children, and fewer and fewer people use the language even in domestic settings. The growing obsolescence of Tariana and its rapid replacement by now dominant Tucano is accompanied by a rapidly increasing number of calqued forms and constructions from Tucano. The long-term interaction based on institutionalized societal multilingualism between East Tucanoan languages and Tariana has resulted in the rampant di¤usion of grammatical and semantic patterns (though not

98

Alexandra Y. Aikhenvald

so much of forms) and calquing of categories. Comparison of Tariana with closely related Arawak languages (such as Baniwa/Kurripako and Piapoco) helps identify the di¤used and the inherited features in Tariana. A striking feature of the Vaupe´s linguistic area is a strong cultural inhibition against language mixing, viewed in terms of borrowing forms, or inserting bits of other languages, in one’s Tariana. This inhibition operates predominantly in terms of recognizable loan forms. Speakers who use non-native forms are subject to ridicule which may a¤ect their status in the community. What often happens in the language of obsolescent speakers is reinterpreting Tariana morphemes in accordance with the meaning their look-alikes may have in Tucano. Consider the Tariana clitic -ya ‘emphatic.’ This clitic is now increasingly used by obsolescent insecure speakers as a marker of immediate command (17), mirroring the Tucano imperative -ya (18): (17) Tariana pi-n˜ha-ya 2sg-eat-impv ‘Eat!’ (18) Tucano ba’aˆ-ya play-impv ‘Play!’ The -ya imperative in Tariana is frequently used by younger speakers, and hardly ever by the few traditional older speakers, who concur that this is not ‘‘proper Tariana.’’ The morpheme -ya in an imperative construction is condemned as a token of identifiable language-mixing (see Aikhenvald 2008a, for cognates of the emphatic -ya in other Arawak languages, and the imperative marker -ya in Tucanoan languages). Another similar example comes from the increased use of nominalizations marked with -¸i in Tariana commands. This is an alternative to simple imperatives, but with a somewhat di¤erent meaning, ‘make sure you do.’ (19) Tariana pi-n˜ha-¸i! 2sg-eat-nominalization ‘Eat!’ (make sure you eat, lest you go hungry) This usage is restricted to casual speech by younger people for whom Tucano is the main language of day-to-day communication. Tucano, just like most

Language contact in language obsolescence

99

other East Tucanoan languages, has a su‰x -ri used in commands with an overtone of warning, with the meaning of ‘or else’ (see Ramirez 1997, Vol. 1: 146–147). The usage of nominalizations as commands in Tariana has in all likelihood been influenced by the -ri marked imperative in Tucano. That the form in (19) is a nominalization is corroborated by the translations given by traditional speakers of Tariana, who themselves avoid using commands like (19), using an apprehensive construction instead. Traditional Tariana did not use to have any special morpheme for first person plural imperative (or hortative). Nowadays, obsolescent speakers employ a hortative -da/-¸a. Compare Traditional Tariana, in (20a), with (20b), recorded from an obsolescent speaker: (20a) Traditional Tariana wa-i¸a 1pl-drink (20b)

Obsolescent Tariana wa-i¸a-da 1pl-drink-hortative ‘Let’s drink!’

Functionally and formally this morpheme is reminiscent of the Tucano hortative -ra˜ /-da˜ (Ramirez 1997, Vol. I: 145) which is shared with other Tucanoan languages: (21) Tucano sı˜ ’ri-da˜ ! drink-hortative ‘Let’s drink!’ The Tariana hortative is likely to be a recent borrowing from Tucano. Or it could be the result of a reinterpretation of already existing Tariana morpheme -da/-¸a ‘dubitative’ which is sometimes used to express politeness. Traditional speakers of Tariana are aware of the similarity between the Tariana and the Tucano morphemes, and treat the hortative (as in (20b)) as ‘‘incorrect’’ Tariana ‘‘mixed’’ with Tucano. This is typical of Tariana language attitudes: given the general prohibition on mixing languages viewed in terms of lexical loans, the hortative is, not surprisingly, a marginal feature of the language (see Aikhenvald 2002: 213–222 on language awareness in the Vaupe´s area). Or a look-alike can oust another, non-shared morpheme. Tariana has numerous verbal markers to do with extent and type of action, among

100

Alexandra Y. Aikhenvald

them the enclitic -pita ‘repetitive action: do again.’ This enclitic is being replaced by the form -ta ‘repetitive’, shared with related languages, but infrequent in the traditional language. The form -ta is similar to Tucano taha, often reduced to -ta (Ramirez 1997: 343–4). These instances of a semantic extension of a native morpheme under the influence of a look-alike in a contact language (known as grammatical accommodation: see Aikhenvald 2006a) are symptomatic of language shift in language obsolescence. This is an alternative to influx of non-native forms. Obsolescent Tariana o¤ers curious examples of drastic restructuring. Tucanoan languages and Tariana are genetically unrelated, and typologically di¤erent. Like many Arawak languages, Tariana employs prefixes for subject cross-referencing, while Tucanoan languages are predominantly su‰xing. As a result of long-term contact, Tariana has developed numerous un-Arawak features, including cases for core arguments and a complex system of evidentials. (These are instances of completed changes.) Obsolescent Tariana is developing a system of cross-referencing enclitics, as exemplified by (22b), mirroring the Tucanoan pattern. The following example is a typical beginning of a story. It was recorded from a fluent middle-aged speaker who always tried to speak the traditional language. The structural parallelism with Tucano is striking, but not complete. The major di¤erence lies in the person that is marked: Tariana employs a prefix ( just like any Arawak language would), while Tucano employs a su‰x (portmanteau with a tense-evidential marker). The relevant forms are in bold. (22a) Traditional Tariana Tariana Payape-se-nuku

paita

Tucano

Diporo´-pi-re ni’ki long.ago-loc-top.non.a/s oneþcl:human

Tariana

dy-uka-na 3sgnf-arrive-rem.p.vis

Tucano

etaˆ-wı˜ a’to-re´ arrive-3sgnf.rem.p.vis here-top.non.a/s ‘A long time ago a man arrived here.’

nawiki masi person

aı˜-nuku here-top.non.a/s

A similar story told by an obsolescent speaker (now in his early thirties) started in a subtly di¤erent way:

Language contact in language obsolescence

101

(22b) Tariana

Payape-se-nuku

paita

Tucano

Diporo´-pi-re ni’ki long.ago-loc-top.non.a/s oneþcl:human

Tariana

dy-uka-na¼diha aı˜-nuku 3sgnf-arrive-rem.p.vis¼he here-top.non.a/s

Tucano

etaˆ-wı˜ a’to-re´ arrive-3sgnf.rem.p.vis here-top.non.a/s ‘A long time ago a man arrived here.’

nawiki masi person

The Tariana in (22b) is structurally closer to Tucano since the speaker employs an encliticized personal pronoun following the evidential. When (22b) was uttered, no one commented on the language di¤erence. Speakers are more aware of non-native forms than they are of non-native patterns. Nevertheless, when I played (22b) back to a traditional elder, he commented that ¼diha should not have been there. Instances like (22b) demonstrate that Tariana is becoming almost like relexified Tucano. But since language change in language obsolescence is unstable and discontinuous, chances are that this relexified variety will not live beyond the life-span of the last speakers. 5. What can we conclude? A study of contact-induced change in the situation of language obsolescence poses specific problems. Basically, the same or similar issues arise when we investigate the speech behavior of obsolescent speakers of otherwise well-spoken languages, and processes of change in those languages which are on their way out. Independently of whether we are dealing with obsolescent languages or just with obsolescent speakers, the influx of non-native forms tends to be pervasive. This is understandable: language obsolescence is typically associated with word-retrieval problems, and it is easier to just use an item from the dominant language. In other instances of language obsolescence, we encounter instances of influx of non-native forms beyond lexicon. Mawayana and Resı´garo have borrowed pronouns, while Resı´garo has also restructured its crossreferencing system, e¤ectively incorporating a non-native bound form of a pronoun. This is in addition to borrowing numbers ‘one’ and ‘two,’ and

102

Alexandra Y. Aikhenvald

numerous further bound morphemes. These instances of borrowing members of closed classes and even bound forms are typologically unexpected and unusual. However, a question arises: are these really borrowings, or are they just instances of nonce forms? That the last speakers’ usage is unstable and ephemeral is a well-known fact. Typologists and historical linguists need to be wary of that when they encounter unusual patterns of borrowed forms in obsolescent languages. A further, commonly attested, e¤ect of language contact in obsolescence is the enhancement of forms already shared with the dominant language. This often concerns frequently used forms and constructions, such as the expression of commands. In addition, enhanced structural di¤usion may result in one language becoming like a reflection of the other: the obsolescent Tariana may sound like relexified Tucano. This is an extreme – but again, often ephemeral – outcome of language shift. In Johanson’s (2002) words, ‘‘languages do not die of ‘structuritis’’’ – that is, contact-induced change does not result in language extinction. But the processes of language obsolescence may promote structural changes amazing in their extent. Before passing into extinction, an obsolescent language may become a ‘‘carbon copy’’ of the dominant idiom. This excessive copying is hardly surprising. The dominant language is the one used on a day-to-day basis by speakers of an obscolescent language, and so the structures from the dominant language get calqued and transferred into the language falling into disuse. (More discussion and examples can be found in Aikhenvald 2002, Grenoble 2000, and classic work by Hill and Hill 1986, Tsitsipis 1998 and Campbell and Muntzel 1989). Contact-induced changes in the situation of language obsolescence are inherently unstable (as was pointed out by Tsitsipis 1998). Ephemeral as they are, their outcomes may go against generalizations obtained in ‘‘healthy’’ language situations.

Acknowledgments I am deeply grateful to the late Candela´rio da Silva, the last speaker of Bare, to the members of the Brito family who taught me their native Tariana, and my adopted family at Avatip who taught me their native Manambu. Special thanks go to R. M. W. Dixon, who provided invaluable comments on this article. I am also grateful to Claudine Chamoreau for her careful editing.

Language contact in language obsolescence

103

Abbreviations a all caus cl compl.ds conn dat decl dem du exc f fut impv inc inch lk loc m neg nf o perf pl pn poss rem.p.vis seq sg subj top.non.a/s

Transitive subject Allative Causative Classifier Completed di¤erent subject Connective Dative Declarative Demonstrative Dual Exclusive Feminine Future Imperative Inclusive Inchoative Linker Locative Masculine Negative Nonfeminine Object Perfective Plural Pronoun Possessive Remote past visual Sequencing Singular Subject Topical non-subject.

References Aikhenvald, A. Y. 1995. Bare. Munich: Lincom Europa. Aikhenvald, A. Y. 1999. The Arawak language family. In The Amazonian Languages, R. M. W. Dixon and A. Y. Aikhenvald (eds), 65–105. Cambridge: Cambridge University Press.

104

Alexandra Y. Aikhenvald

Aikhenvald, A. Y. 2001. Areal di¤usion, genetic inheritance and problems of subgrouping: a North Arawak case study. In Areal Di¤usion and Genetic Inheritance: Problems in Comparative Linguistics, A. Y. Aikhenvald and R. M. W. Dixon (eds), 167–194. Oxford: Oxford University Press. Aikhenvald, A. Y. 2002. Language Contact in Amazonia. Oxford: Oxford University Press. Aikhenvald, A. Y. 2003. A grammar of tariana, from north-west Amazonia. Cambridge: Cambridge University Press. Aikhenvald, A. Y. 2006a. Grammars in contact: a cross-linguistic perspective. In Grammars in Contact: a Cross-linguistic Typology, A. Y. Aikhenvald and R. M. W. Dixon (eds), 1–66. Oxford: Oxford University Press. Aikhenvald, A. Y. 2006b. Semantics and pragmatics of grammatical relations in the Vaupe´s linguistic area. In Grammars in Contact: a Cross-linguistic Typology, A. Y. Aikhenvald and R. M. W. Dixon (eds), 237–266. Oxford: Oxford University Press. Aikhenvald, A. Y. 2007. Linguistic fieldwork: setting the scene. Sprachtypologie und Universalienforschung 60, 1: 1–11. Aikhenvald, A. Y. 2008a. Multilingual imperatives: the elaboration of a category in north-west Amazonia. International Journal of American Linguistics 74: 189– 225. Aikhenvald, A. Y. 2008b. The Manambu Language of the East Sepik, Papua New Guinea. Oxford: Oxford University Press. Aikhenvald, A. Y. 2009a. Semantics of clause linking in Manambu. In Semantics of Clause Linking: A Cross-linguistic Typology, R. M. W. Dixon and A. Y. Aikhenvald (eds), 118–144. Oxford: Oxford University Press. Aikhenvald, A. Y. 2009b. Language contact along the Sepik River. Anthropological Linguistics 50: 1–66. Aikhenvald, A. Y. 2010. Gender and noun class in Paumarı´ in a typological perspective. In Linguistics and Archeology in the Americas, E. Carlin and S. van de Kerke (eds), 236–252. Leiden: Brill. Aikhenvald, A. Y. and Dixon, R. M. W. (eds). 2001. Areal Di¤usion and Genetic Inheritance: Problems in Comparative Linguistics. Oxford: Oxford University Press. Aikhenvald, A. Y. and Dixon, R. M. W. (eds). 2006. Grammars in Contact: a Cross-Linguistic Typology. Oxford: Oxford University Press. Allin, T. R. 1975. A Grammar of Resı´garo. PhD dissertation, University of St. Andrews. Aschmann, R. P. 1993. Proto-Witotoan. Arlington: Summer Institute of Linguistics and the University of Texas. Austin, P. 1986. Structural change in language obsolescence: some Eastern Australian examples. Australian Journal of Linguistics 6: 201–230. Baugh, A. C. 1957. A History of the English Language. New York: AppletonCentury-Crofts.

Language contact in language obsolescence

105

Bettoni, C. 1991. Language variety among Italians: Anglicisation, attrition and attitudes. In Language in Australia, S. Romaine (ed.), 263–269. Cambridge: Cambridge University Press. Brenziger, M. (ed.). 1992. Language Death. Factual and Theoretical Explorations with Special Reference to East Africa. Berlin: Mouton de Gruyter. Brenziger, M. and Dimmendaal G. J. 1992. Social context of language death. In Language Death. Factual and Theoretical Explorations with Special Reference to East Africa, M. Brenziger (ed.), 1–5. Berlin: Mouton de Gruyter. Campbell, L. 1994. Problems with the pronouns in proposals of remote relationships among native American languages. In Proceedings of the Meeting of the Society for the Study of the Indigenous Languages of the Americas and the Hokan-Penutian Workshop, M. Langdon (ed.), 1–20. Berkeley: University of California Press. Campbell, L. 1997. Amerindian personal pronouns: A second opinion. Language 73: 339–351. Campbell, L. and Muntzel, M. 1989. The structural consequences of language death. In Investigating obsolescence. Studies in Language Contraction and Death, N. C. Dorian (ed.), 181–196. Cambridge: Cambridge University Press. Carlin, E. 2006. Feeling the need: the borrowing of Cariban functional categories into Mawayana (Arawak). In Grammars in Contact: a Cross-linguistic Typology, A. Y. Aikhenvald and R. M. W. Dixon (eds), 313–332. Oxford: Oxford University Press. Chamoreau, C. 2000. Chronique d’une mort annonce´e. Un exemple de re´duction en phurhe´pecha. Actes du XXIe`me Colloque international de la Socie´te´ internationale de linguistique fonctionnelle (SILF). Iasi (Romania), 26 June–2 July 1996, 127–132. Chamoreau, C. 2005. Reorganizacio´n de la voz en pure´pecha: una visio´n dina´mica. In Dina´mica lingu¨´ıstica de las lenguas en contacto, C. Chamoreau and Y. Lastra (eds), 67–86. Sonora: University of Sonora. Chapman, S. and Derbyshire, D. C. 1991. Paumarı´. In Handbook of Amazonian Languages, vol. 3, D. C. Derbyshire and G. K. Pullum (eds), 151–355. Berlin: Mouton de Gruyter. Dal Negro, S. 2004. The Decay of a Language. The Case of a German Dialect in the Italian Alps. Bern: Peter Lang. Dench, A. 2001. Descent and di¤usion: the complexity of the Pilbara situation. In Areal Di¤usion and Genetic Inheritance: Problems in Comparative Linguistics, A. Y. Aikhenvald and R. M. W. Dixon (eds), 105–133. Oxford: Oxford University Press. Dixon, R. M. W. 1972. The Dyirbal Language of North Queensland. Cambridge: Cambridge University Press. Dixon, R. M. W. 1977. A Grammar of Yidiny. Cambridge: Cambridge University Press.

106

Alexandra Y. Aikhenvald

Dixon, R. M. W. 1991a. A changing language situation: the decline of Dyirbal, 1963–1989. Language in Society 20: 183–200. Dixon, R. M. W. 199lb. Reassigning underlying forms in Yidiny – a change during language death. In Language and History, Essays in Honour of Luise A. Hercus, P. Austin, R. M. W. Dixon, T. Dutton, and I. White (eds), 89–99. Canberra: Pacific Linguistics. Dixon, R. M. W. 1997. The Rise and Fall of Languages. Cambridge: Cambridge University Press. Dixon, R. M. W. 2002. Australian Languages: their Nature and Development. Cambridge: Cambridge University Press. Dixon, R. M. W. 2007. Field linguistis: a minor manual. Sprachtypologie und Universalienforschung 60, 1: 12–31. Dorian, N. C. 1973. Grammatical change in a dying dialect. Language 49: 413– 438. Dorian, N. C. 1977. The problem of the semi-speakers in language death. Linguistics 191: 23–32. Dorian, N. C. (ed.). 1989. Investigating Obsolescence. Studies in Language Contraction and Death. Cambridge: Cambridge University Press. Dorian, N. C. 1994. Purism vs. compromise in language revitalisation and language revival. Language in Society 23: 479–494. Dorian, N. C. 1999. The study of language obsolescence: stages, surprises, and challenges. Languages and Linguistics 3: 99–122. Dorian, N. C. 2002. Diglossia and the simplification of linguistic space. International Journal of the Sociology of Language 9: 63–70. Evans, N. 2000. Iwaidjan, a very un-Australian language family. Linguistic Typology 4: 91–112. Evans, N. 2007. More di¤erent than we thought: reassessing the genetic status of Marrku. Seminar presented at RCLT. Evans, N., Malwagag, J. W., and Marrala, K. 2006. Majila Inkawart. Six Stories in Marrku, the Traditional Language of the Kroker Island, translated into Iwaidja and English. Jabiru Town Council: Iwaidja Inyman. Farabee, W. C. 1918. The Central Arawaks. Philadelphia: Anthropological Publications, The University Museum, University of Pennsylvania. Foley, W. A. 1986. The Papuan Languages of New Guinea. Cambridge: Cambridge University Press. Franc¸ois, A. 2002. Araki: a Disappearing Language of Vanuatu. Canberra: Pacific Linguistics. Gardani, F. 2005. Borrowing of inflectional morphemes in language contact. MA thesis, University of Vienna. Grenoble, L. A. 2000. Morphosyntactic change: the impact of Russian on Evenki. In Languages in Contact, D. G. Gilbers, J. Nerbonne, and J. Schaeken (eds), 105–120. Amsterdam: Rodopi.

Language contact in language obsolescence

107

Grenoble, L. A. and Whaley, L. J. 1998. Towards a typology of language endangerment. In Endangered Languages, L. Grenoble and L. J. Whaley (eds), 22–54. Cambridge: Cambridge University Press. Gruzdeva, E. Y. 2002. The linguistic consequences of Nivkh language attrition. SKY Journal of Linguistics 15: 85–103. Haas, M. R. 1941. Tunica (extract from Handbook of American Indian Languages, vol. IV). New York: J. J. Augustin. Hamp, E. 1989. On signs of health and death. In Investigating obsolescence. Studies in Language Contraction and Death, N. C. Dorian (ed.), 197–210. Cambridge: Cambridge University Press. Hardenburg, W. E. 1910. The Indians of the Putumayo, Upper Amazon. Man 10: 134–8. Haugen, E. 1989. The rise and fall of an immigrant language: Norwegian in America. In Investigating Obsolescence. Studies in Language Contraction and Death, N. C. Dorian (ed.), 61–74. Cambridge: Cambridge University Press. Heath, J. 1978. Linguistic di¤usion in Arnhem Land. Canberra: AIAS. Helimski, E. A. 2007. Fonetika i morfonologia enetskogo jazyka v uslovijakh jazykovogo sdviga (Phonetics and morphophonology of Enets in language shift). In Jazykovye izmenenija v uslovijakh jazykovogo sdviga, (Language changes in language shift), N. B. Vakhtin (ed.), 213–224. St. Peterburg: Nestor. Henzl, V. M. 1981. Slavic languages in the new environment. In Language in the USA, C. A. Ferguson and S. B. Heath (eds), 293–321. Cambridge: Cambridge University Press. Hill, J. H. 1985. The grammar of consciousness and the consciousness of grammar. American Ethnologist 12: 725–37. Hill, J. H. 1998. ‘‘Today there is no respect’’: nostalgia, ‘‘respect’’ and oppositional discourse in Mexicano (Nahuatl) language ideology. In Language Ideologies: Practice and Theory, B. B. Schie¤elin, K. A. Woolard, and P. V. Kroskrity (eds), 68–86. New York: Oxford University Press. Hill, J. H. and Hill, K. C. 1986. Speaking Mexicano. Dynamics of Syncretic Language in Central Mexico. Tucson: University of Arizona Press. Hjelde, A. 1996. The gender of English nouns used in American Norwegian. In Language Contact across the North Atlantic, P. S. Ureland and I. Clarkson (eds), 297–312. Tu¨bingen: Max Niemeyer Verlag. Howard, C. V. 1986. Mawayana: Formula´rio dos vocabula´rios padro˜es. Rio de Janeiro: Museu Nacional. Ms. Igualada, F. de. 1940. Informe sobre el Centro de investigaciones lingu¨´ısticas y etnogra´ficas de la Amazonia colombiana (1933–1940). Amazonia colombiana americanista, o´rgano semestral de materiales y estudios para la americanistica del CILEAC de Sibundoy. Pasto, t. 1.2–3: 61–91. Igualada, F. de and Castellvı´, M. de. 1940. Clasificacio´n y estadistica de las lenguas habladas en el Putumayo (Caqueta´) y Amazonas. Amazonia colombiana americanista, o´rgano semestral de materiales y estudios para la americanistica del CILEAC de Sibundoy. Pasto, t. 1.2–3: 92–101.

108

Alexandra Y. Aikhenvald

Johanson, L. 2002. Do languages die of structuritis? On the role of code-copying in language endangerment. Rivista Italiana di Linguistica 14: 249–270. Jungraithmayr, H. 1995. Was ist am Tangale noch tschadisch/hamitosemitisch? In Sprachkulturelle und historische Forschungen in Afrika, A. Fleisch and D. Otten (eds), 197–205. Cologne: Ko¨ppe. Kagan, O. and Dillon, K. 2001. A new perspective on teaching Russian: focus on the heritage learner. The Slavic and East European Journal 45: 507–518. Klintborg, S. 1999. The Transience of American Swedish. Lund: Lund University Press. Kroskrity, P. V. 1993. Language, History and Identity: Ethnolinguistic Studies of the Arizona Tewa. Tucson: University of Arizona Press. Laycock, D. C. and Z’graggen, J. 1975. The Sepik-Ramu phylum. In New Guinea Area Languages and Language Study. Volume 1, Papuan languages and the New Guinea linguistic scene, S. A. Wurm (ed.), 731–764. Canberra: Pacific Linguistics. Lee, Y. 1997. Classifiers in Korean. Honors thesis, Australian National University. Lopez Sanz, R. 1972. El Bare. Estudio Lingu¨´ıstico. Caracas: Faculty of Economics and Social Sciences of the Central University of Venezuela. Loukotka, C. 1968. Classification of South American Indian Languages. Los Angeles: Latin American Studies Center, UCLA. Matiso¤, J. A. 1990. On megalo-comparison: a discussion note. Language 66: 106–120. Matras, Y. 1998. Utterance modifiers and universals of grammatical borrowing. Linguistics 36: 281–331. McGregor, W. 2002. Structural changes in language obsolescence: a Kimberley (Australia) perspective. SKY Journal of Linguistics 15: 145–186. Milani, C. 1996. Language contact among North-American people of Italian origin. In Language Contact across the North Atlantic, P. S. Ureland and I. Clarkson (eds), 479–501. Linguistische Arbeiten 359. Tu¨bingen: University of Tu¨bingen. Natterer, J. 1831. Sprachproben der Indier Bare´. Bewohner von Marabitanas vom Stamm Arihini/Alihini. Nation Arhini. Sprache der Nation Bare´ – Stamm ihini. Lists 30–32. ms. Newman, S. 1977. The Salish independent pronoun system. International Journal of American Linguistics 45: 207–223. Newman, S. 1979a. A history of the Salish possessive and subject forms. International Journal of American Linguistics 45: 207–23. Newman, S. 1979b. The Salish object forms. International Journal of American Linguistics 45: 299–308. Payne, D. L. 1985. The genetic classification of Resigaro. International Journal of American Linguistics 51: 222–231. Payne, D. L. 1991. A classification of Maipuran (Arawakan) languages based on shared lexical retentions. In Handbook of Amazonian Languages, vol. 3, D. C. Derbyshire and G. K. Pullum (eds), 355–499. Berlin: Mouton de Gruyter.

Language contact in language obsolescence

109

Pereltsvaig, A. 2008. Aspect in Russian as grammatical rather than lexical notion. Russian Linguistics 32: 27–42. Pryor, J. 1990. Deixis and participant tracking in Botin. Language and Linguistics in Melanesia 21: 1–29. Ramirez, H. 1997. A fala Tukano dos Yepaˆ-masa. Tomo 1. Grama´tica. Manaus: Inspetoria Salesiana. Riionheimo, H. 2002. How to borrow a bound morpheme? Evaluating the status of structural interference in a contact between closely related languages. SKY Journal of Linguistics 15: 187–218. Rivet, P. and de Wavrin, R. 1951. Un nouveau dialecte Arawak: le Resigaro. Journal de la Socie´te´ des Ame´ricanistes de Paris 40: 203–238. Sasse, H.-J. 1992a. Theory of language death. In Language Death. Factual and Theoretical Explorations with Special Reference to East Africa, M. Brenzinger (ed.), 31–57. Berlin: Mouton de Gruyter. Sasse, H.-J. 1992b. Language decay and contact-induced change: similarities and di¤erences’. In Language Death. Factual and Theoretical Explorations with Special Reference to East Africa, M. Brenzinger (ed.), 58–79. Berlin: Mouton de Gruyter. Sasse, H.-J. 2001. A Phoenix from the ashes: linguistic creativity in an obsolescent speech community. Seminar presented at RCLT. Schomburgk, R. 1848. Comparative vocabulary of 18 languages and dialects of Indian tribes inhabiting Guiana. Report of the 18th British Association for the Advancement of Science, 96–99. Schmidt, A. 1985. Young People’s Dyirbal. An Example of Language Death from Australia. Cambridge: Cambridge University Press. Stolz, C. and Stolz, T. 1996. Funktionswortenentlehnung in Mesoamerika. Spanischamerindischer Sprachkontakt (Hispanoindiana II). STUF 49: 86–123. Swadesh, M. 1951. Di¤usional cumulation and archaic residue as historical explanations. Southwestern Journal of Anthropology 7: 1–21. Tessmann, G. 1930. Die Indianer Nordost-Perus. Grundlegende Forschungen fu¨r eine systematische Kulturkunde. Hamburg: Friederichsen, De Gruyter & Co, M. B. H. Thiesen, W. 1996. Grama´tica del Idioma Bora. Se´rie Lingu¨ı´stica Peruana 38. Pucallpa: Instituto Linguı´stico de Verano. Tsitsipis, L. D. 1998. A Linguistic Anthropology of Praxis and Language Shift. Arvanı´tika (Albanian) and Greek in Contact. Oxford: Clarendon Press. Whi¤en, T. W. 1915. The North-west Amazonas: Notes of some Months Spent among Cannibal Tribes. London: Constable and Company.

The emergence of a marked-nominative system in Tehuelche or Aonek’o a jen: a contact-induced change? Ana Ferna´ndez Garay 1. Objective This article explores the variations and changes that can be observed in the syntactic structure of Aonek’o a jen (commonly known as Tehuelche), a language belonging to the Chon family, when it was documented during the eighties. The presence of an adposition that marks the subject/agent of the clause is optional: apparently, the ergative-absolutive system did not mark the agent in the protolanguage. Instead, the adposition was incorporated later in the evolution of the language, but the change was never fully complete. In other words, Tehuelche experienced a change from an ergative-absolutive system to a marked-nominative system through the extension of the adposition or the ergative marker of the transitive clause to the unique participant of the intransitive. In this article we examine this process and see how it was stimulated by the presence of nominativeaccusative languages in the area. 2. Indigenous groups of Patagonia: historical aspects related to contact situations The ethnic groups that have lived in Patagonia can be reduced to three: the Tehuelche complex, the Fueguinos or Canoeros, and the Mapuches. Their settlements extend southwards from the Buenos Aires-Mendoza line. Escalada (1949) argued for the existence of several subgroups of the Tehuelche people who inhabited the area from the north of Patagonia southwards to Tierra del Fuego. These subgroups are, on the mainland: the Gue´nena ke´ne, who spoke Gue´nena la´jitch, in the north; the Chewachekenk, who spoke Teushen, in the midwest; and the Ao´ni-ke´nk, who speak Aoniko-a´ish, in the south (this group includes the Mecharnue). On Tierra del Fuego they include the Selknam or Onas, and the Haush or Manek’enk. This classification was revised by Casamiquela (1965), who believed

112

Ana Ferna´ndez Garay

that there were only two significant Tehuelche subgroups, the northern and the southern. These two subgroups, according to Casamiquela, were separated by the Chubut River (living respectively on the northern and southern side of this river. The group in the north was further divided into two groups, one speaking Querandı´ and the other speaking Gu¨nu¨na ia´jech, the language of the Gu¨nu¨na ku¨ne; similarly, the southern group was divided into speakers of Teushen and speakers of Aonek’o ajen. Querandı´ is a language about which little is known to date but which Casamiquela (1965: 33–45) believed may have been related to Gu¨nu¨na ia´jech. Apparently this language was spoken in the Rı´o de la Plata region to the east; to the west, it was spoken in Co´rdoba, San Luis, and Mendoza. The language disappeared without trace at the beginning of the twentieth century. Gu¨nu¨na iaje¨ch, the language of the Gu¨nu¨na ku¨ne, was described by Casamiquela (1983). Traditionally, the settlements of this group extended from the south of the province of Buenos Aires to the southeast of La Pampa, and the south of Mendoza, Co´rdoba, and Santa Fe to the north of the province of Chubut. They resisted the Mapuches, whom they considered their enemies, but little by little they were nonetheless influenced by this group. The Gu¨nu¨na iaje¨ch language disappeared in 1960 when its last speaker passed away. The southern Tehuelches, who spoke Teushen, lived between the Chubut River and the Santa Cruz River. Their language disappeared when the Aoenk’o ajen, the language of the Aonek’enk, expanded southwards. This language was never described, but approximately thirty vocabulary lists survive, so it can be studied to some extent. The Aonek’enk, who spoke Aonek’o ajen, inhabited the region between the Santa Cruz River and the Straits of Magellan. Although there are still a few remaining speakers of Aonek’o ajen who can remember the language, it is no longer used for intra-group communication. Within the subgroups Casamiquela also includes the Selknam or Onas and the Haush or Manek’enk. The first of these inhabited almost all of Isla Grande in Tierra del Fuego, especially the northwestern side of the island. Their language died out during the second half of the twentieth century; what we know of it comes from vocabulary lists. The most thorough linguistic description of this language is that of Najlis (1973). The second group inhabited the southeastern side of Isla Grande in Tierra del Fuego. Their language has completely disappeared. Only a few vocabulary lists are available, drafted around the end of the eighteenth century.

The emergence of a marked-nominative system

113

All of the languages mentioned belong to the Chon family (Sua´rez 1988: 79–100), except for Querandı´ (or rather, the lack of information on this language makes it impossible to know whether or not it belongs to this family) and Gu¨nu¨na ia´jech, which Sua´rez did not consider a Chon language (Sua´rez 1988: 87). The Tehuelche subgroups were nomads, hunters, and gatherers. Traditionally, they moved across Patagonia on foot, following the hunting cycles. The first mention of the Tehuelches appears in Pigafetta, who chronicled Ferdinand Magellan’s voyage around the world. Based on his account and this initial contact, the legend of the ‘‘race of giants’’ that so attracted Europeans was born. The introduction of the horse, an animal brought to South America by the Spaniards, and the subsequent spreading of wild horses across the territory of the Pampa, made it even easier for the Tehuelches to travel across Patagonia. The indigenous people in Tierra del Fuego, known as ‘‘canoeros australes’’1 by ethnologists, are divided into three groups: the Ya´manas or Yahganes, the Qawasqar or Alacalufes, and the Chonos. They are mainly located on the southern islands of the Chilean shoreline and on Tierra del Fuego. As they are remote from the mainland, we will not include them in this article. The Araucanos or Mapuches (mapu ‘land’ and che ‘people’), the name which they generally use to refer to themselves, are originally from Chile, specifically from the regions between the Bı´o Bı´o and Tolte´n rivers. Researchers are not sure when they began crossing the Andean border and arriving in the regions of the Pampa and Argentinian Patagonia, although documents from the seventeenth century mention certain toponyms of Mapuche origin in Argentinian Patagonia that evidence the presence of this group in the country. These journeys led to what has been called the process of ‘‘Araucanization’’ (Nardi 1985: 235–264) of the indigenous groups of Argentina. In the early nineteenth century, Mapuche tribes from Chile began inhabiting the plains of the Pampa. The Araucanization, which lasted between 300 and 400 years, led to cultural interaction between the Mapuches and the groups living in the Pampa/Patagonia. The Araucanos were enemies of the Tehuelches (the names of the battles that took place in the province of Chubut are well-known). Tehuelche groups were taken prisoner by the Mapuches and as a result they began speaking the enemy’s language. 1. This group used to live and fish on canoes in the channels of Tierra del Fuego for extensive periods.

114

Ana Ferna´ndez Garay

The increasingly widespread use of Mapudungun (mapu ‘land’ and dungun ‘speak, language’) as well as interethnic marriages leads to the hypothesis that this language circulated throughout the mid-southern region of Argentina as a lingua franca, while Argentinian indigenous languages like Gu¨nu¨na ku¨ne, Teushen, and Aonek’o ajen or Tehuelche were also still spoken. The process of Araucanization intensified during the nineteenth century, wiping out some of the languages and cultures of the groups established in the north and north-central areas of Argentinian Patagonia. The only language that has survived to the present day is Aonek’o ajen. We should bear in mind the constant nomadism of the Tehuelche groups, which led them to make contact with the di¤erent ethnic groups that inhabited the area in order to trade and exchange their products. This nomadism is documented by Musters (1964), who traveled across Patagonia in 1860 from Punta Arenas in the south of Chile to Carmen de Patagones in the northeast of this extensive region.

3. Theoretical aspects It is now well established that after a long and intensive period of contact, elements from the syntactic level can be transferred between languages (Harris and Campbell 1995: 149; Thomason 2001: 85; Heine and Kuteva 2005; Matras and Sakel 2007; Matras 2009). There have also been theoretical developments in the study of linguistic areas. Thomason defines a linguistic area as a geographical region containing a group of three or more languages that share a certain structural feature as a result of contact between the languages. The common feature cannot be an ‘‘accidental’’ result or inherited from a common ancestor (Thomason 2001: 99). With extreme caution, a Patagonian linguistic area can be posited, although some languages of the region (Teushen, Gu¨nu¨na ku¨ne) have been only minimally described or not described at all, making it di‰cult to reach definitive conclusions. During the process of Araucanization, before these Tehuelche languages were replaced by Mapudungun, they may have initially undergone changes that brought them closer to this language and later to Spanish. The only language in the area that survived the contact with Mapudungun was Aonek’o ajen, although the influence of this language can be seen not only at the lexical level but at the morphological and syntactic level as well (see Ferna´ndez Garay 2006: 153–155). Another aspect to consider is the extinction of languages, since the phenomena of obsolescence that this process involves must also be taken into

The emergence of a marked-nominative system

115

account – that is to say, the reduction, simplification, variation, and change that take place during the period in which a language is dying (see Dorian 1981: 114–156; Thomason 2001: 221–239). The case that we are addressing involves a language that was almost extinct when it was described, a language that coexisted for centuries (at least four centuries, though perhaps more) with Mapudungun and also with Spanish in Patagonia. However, it was mainly at the end of the nineteenth century and during the twentieth century that Spanish, the dominant tongue, began to rapidly displace local languages. 4. Syntactic structure of Tehuelche or Aonek’o ajen In Tehuelche or Aonek’o ajen, the syntactic structure described initially was the marked-nominative structure. In this structure, A2 in the transitive clause and S in the intransitive clause are marked by the adpositions sˇ P n P r,3 as long as P is not marked. This adpositional marker only occurs when A/S comes before the verb (examples (3) and (4)). When it is located after the verb, generally the marker disappears (examples (5) and (6)), but we have found some cases where an A/S NP is located after the verb and carries the adposition (example (7)) (Ferna´ndez Garay and Herna´ndez 2006: 121). When A/S comes before the verb, the adposition appears with certain restrictions: it cannot appear if A/S is a dependent pronoun4 and there is no quantifier before it (example (1)) or an adverb of any kind (temporal, locative, dubitative: example (2)). In these cases, the marker is a preposition. If A/S is an independent pronoun (example 2. We use ‘‘S’’ to refer to the unique participant of a prototypical intransitive clause, ‘‘A’’ for the most agent-like participant of a prototypical transitive clause, and ‘‘P’’ to indicate the most patient-like participant of a prototypical transitive clause. 3. The adposition sˇ marks S/A when the predicate is determined by the real mood or when the mood is absent; n is employed when the predicate is determined by the unreal mood; r marks S/A in questions when S/A is unknown or absent. 4. There are two types of personals in this language: dependents are those that are cliticized to nouns, verbs, postpositions, and adverbs (e-, m-, t-, etc.); independents do not need to lean on other lexical units: ja:, ma:, ta:, etc. Dependent personals sometimes cross-reference S NP of intransitive Group 2 verbs and A/P NP of transitive Group 2 verbs. Cross-reference is obligatory only with Group 1 verbs that agree in gender with S/P by means of k- P-/Ø-.

116

Ana Ferna´ndez Garay

(3)) or a nominal phrase (example (4)), the marker comes after it; in other words, in these cases we are dealing with a postposition. This marker, as was mentioned, has restrictions in terms of when it can appear, and even in cases in which the conditions are met it may not appear (examples (8) and (9)), which is considered typical in the optional syntax of a language on its way to extinction (see Ferna´ndez Garay 1998: 460). (1) welom sˇ o-sˇ-k’eto pe-k’ all adp 1-pl-well be-rm ‘We are all well.’ (2) ma sˇ e-t-o:mk’e-sˇ-k’ now adp I-her-know-ps-rm ‘Now I know her.’ (3)

ja: sˇ ko:le-k’ I adp stay-rm ‘I stayed.’

(4)

j-a:nk’o sˇ e-mta:we-k’-e 5 my-father adp me-raise-rm-m ‘My father raised me.’

(5) k’o:me-m-tsˇ tewelcˇe-tsˇ disappear-um-pl Tehuelches-pl ‘The Tehuelches disappeared.’ (6) kaj -aXe-sˇ-k’-n wen ka:rken cloak.n 3n-paint-ps-rm-f this woman ‘This woman paints a cloak.’ (7) t-kawr newr e-me-sˇ-n eja: sˇ 3-like this.way I-do-ps-f I adp ‘I do it this way, like she [does].’

5. In this example, we can see one of the numerous gender agreements that the noun has with various syntactic classes. The real mood -k’ can add the morpheme -n when it agrees with a feminine or neuter noun, or the morpheme -e when it agrees with a masculine noun. The noun with which it agrees can function as S/A, as seen in (4).

The emergence of a marked-nominative system

117

(8) emn ka:rken p’aje-nsˇ that woman get.married-dpt ‘That woman got married.’ (9) osˇ-genk’enk te-wa:w a:wke-sˇ our-community.members he-only hunt.guanaco-ps ‘Our community members hunt only guanaco.’ At the same time, there is an ergative subsystem that is manifested in the gender agreement that Group 1 intransitive and transitive verbs maintain with S/P. The intransitives from Group 1 agree with the unique participant or S (example (10)), and the transitives of the same group agree with P (example (11)), thus generating a typical ergative subsystem in which S and P are indexed in the verbs of this group by the morphemes k- P (k- agrees with a masculine or feminine participant, -/Ø- with a neuter participant), while A is di¤erentiated from both because it is optionally marked. In (11), A is marked by means of the ergative adposition sˇ.6 Group 2 intransitive and transitive verbs show person agreement in examples (12) and (13) with S/P (see note 5). (10) -ajq’e-sˇ-k’-n e-or it-be.snubbed-ps-rm-n my-nose.n ‘I am snub-nosed.’ (11) a:we ma:ger sˇ e-k-e:cˇ ’o-sˇ-ko also Ma:ger.m adp I-him-greet-ps-fti ‘I will also say hello (greet) to Ma:ger.’ (12) am t-xam-k’-n enm but 3-die-rm-f that.f ‘but that one (woman) died’

(intransitive clause)

(transitive clause)

(intransitive clause)

(13) am n e-t-or ar-m-n ten-kot cˇen (transitive clause) but adp 1-3-may.be find-um-f someone-nft subs ‘but maybe I will find some (woman)’

6. In ergative-absolutive systems, it is common for A to be the argument marked by case or adposition that indicates the ergative, while S/P are not marked, which is why they are called absolutive. However, ergative languages have di¤erent types of strategies for marking S/P (see Dixon 1994: 40–49).

118

Ana Ferna´ndez Garay

In addition, we should bear in mind that many verbs currently considered Group 2 verbs, because they do not agree with S/P, could have been Group 1 verbs because they have a k- at the beginning of the verbal morpheme. It is evident that the ergative system has been disappearing, especially among intransitive attributive verbs, because there is great variation among such verbs with respect to the appearance of the agreeing morpheme k- P-. This is the case with the intransitive verbs kajcˇer ‘to get twisted,’ kemsˇe ‘to get repented,’ kotqe ‘to get loose,’ and with transitive verbs kamel ‘to give as a present,’ kamgeme ‘to plant,’ which seem to have lexicalized the k- and at this moment they do not agree with S/P. However, we cannot consider this a split system, because the presence or absence of the ergative marker does not correspond with temporal, aspectual, or personal morphemes, nor does it correspond with semantic components like agency or volition of the agent, or semantic distinctions between Group 1 and Group 2 verbs, which would lead us to think that the system had in fact undergone a split (see Dixon 1994: 70–109). In addition (which is what led us to hypothesize that a change was in progress), the two systems coexist and are connected in that they share the adposition. This can be observed in (11), where the ergative marker is identical to the nominative marker in the nominative-accusative system. We can see the polyfunctionality of the adposition and the coexistence of the two agreement systems in the following examples belonging to Group 1 verbs: (14) em ajk’ sˇ e-k-e:ge-k’-e lam that posp adp I-3.m-leave-rm-m wine.m ‘For that [reason], I left the wine.’ (15) ta: sˇ kaj -a:Xe-sˇ-k’ she adp cloak.n it-paint-ps-rm ‘She paints the cloak.’ 5. Historical change: internal or contact-induced? Our conclusion (see Ferna´ndez Garay 2007b: 114–125) is that the observable variation as well as the coexistence of these systems indicate a process of change from an ergative-absolutive system to a marked-nominative system. If we observe one of the languages genetically related to Tehuelche, Selknam (described by Najlis in 1973), we can observe that, though Najlis does not explicitly say so, the syntactic system of this language is ergativeabsolutive (see Ferna´ndez Garay 2007a) since, as she mentions, almost all

The emergence of a marked-nominative system

119

of the verbs of this language are ‘‘prefixable,’’ that is, they have the morphemes k- P h- (cognates of k- P- in Tehuelche) that agree with S in the intransitive clause and with P in the transitive. Apparently, the few verbs that do not have these ‘‘prefixes,’’ as she calls them, are the result of a process of lexicalization that made verbal agreement with S/P impossible. On the other hand, and this is important, Selknam does not have an agent marker. In other words, the ergative is not marked in this language. Therefore, if both languages present ergative-absolutive systems that are manifested through the agreements among the Group 1 verbs in Tehuelche, or when the ‘‘prefixables’’ in Selknam agree with S in the case of intransitive verbs and with P in the case of transitives, we must conclude that the protolanguage, that is, Proto-Chon, had an ergative-absolutive system with the same features, that is, with S/P indexed on the verb and A with no marker. This leads us to ask how Tehuelche first developed an ergative-absolutive system with a marked ergative case and then, based on that system, a marked-nominative system. In other words, the question is: where did the adposition sˇ P n P r that is documented in Tehuelche as marking the agent come from, and how did it develop first into an ergative marker and then into a nominative marker? It is evident that the adpositions come from postpositions existing in the proto-language. We should bear in mind that there are only postpositions in Selknam, while in Tehuelche postpositions are in the majority, though in certain cases there are invariable postpositions7 in this language that serve as prepositions as well (it is important to note the much more frequent use of postpositional forms). This change, which only occurred in Tehuelche, might be owed first to the influence of Mapudungun (Ferna´ndez Garay 2002b: 19, 2005a, 2005b), a language that was widely spoken throughout Patagonia and has both prepositions and postpositions, and later to the influence of Spanish, a language that only has prepositions. It is possible that sˇ might once have been a postposition marking a circumstantial relational complement (‘‘with respect to’’) in Tehuelche, but was later reanalyzed as an agent marker of the transitive clause, while it was still an ergative-absolutive system, and that then the marker slowly 7. In Tehuelche, there are agreeing and invariable postpositions. The majority are agreeing and they carry the gender prefixes k- (masculine and feminine) or - (neuter) and agree with the noun that comes before them (k-awr Pawr ‘on’; k-asˇ P-asˇ ‘in’). The invariables, which may have once been agreeing postpositions that have been lexicalized, do not vary (ka ‘of,’ go ‘like’).

120

Ana Ferna´ndez Garay

extended to S when the contact with other nominative-accusative languages from the area (first Mapudungun and later Spanish) began to change the linguistic panorama of Patagonia. In Ferna´ndez Garay (2000: 165–177), it was observed that the A/S marker was less frequent in a corpus collected by Sua´rez twenty years earlier in the same area – that is, a corpus taken from Tehuelche spoken among the generation previous to our informants. The most interesting di¤erence was that the frequency of the adpositional marker was higher in the transitive clause than in the intransitive one for both corpora, thus revealing that the marker first grammaticalized as an A marker in the transitive sentence and then extended from A to S. This shows that the system was probably ergative at the beginning and later made the switch to marked-nominative when the influence of the nominative-accusative languages in the area imposed on the structure of Aonek’o a jen. This hypothesis of sˇ P n P r as a postposition marking a circumstantial complement in the past can be supported by the following argument. As we proposed in Ferna´ndez Garay (2007a), Selknam has a postposition s¸ whose meaning is, according to Najlis, ‘depends on, with respect to,’ that is very likely a cognate of the agent marker of Tehuelche. At that time, and only because of the meaning that Najlis gave to s¸, we can posit that this postposition was a marker of a circumstantial relational complement, although unfortunately Najlis o¤ers no examples in which its function can be clearly seen. A postposition s¸ could be reconstructed in the protolanguage that would have served as a marker of an oblique complement and thus have been transmitted to both languages; subsequently this circumstantial marker in Tehuelche could be reanalyzed as the adposition that marks the agent in the ergative-absolutive system and later extends to S to constitute the marked-nominative system that manifests itself clearly when the language is described. The reanalysis of a postposition as a new case is well attested in the literature (see Harris and Campbell 1995: 89–90). With respect to r, which marks A/S in questions, it only appears with the indefinite xem ‘who, someone, no one,’ as can be seen in the following example of our corpus: (16) xem r m-mta:we-sˇ-n who adp you-bring.up-ps-f ‘Who brought you up?’ In Selknam, we find the question forms kejs¸ ‘which, what, where, and in which direction’ and kownes¸ ‘who, to whom.’ These two interrogatives, apparently ‘‘prefixable’’ (see the initial k in both cases) present a su‰x s¸

The emergence of a marked-nominative system

121

that currently seems to have lost its value, but which could come from a postposition that marked an oblique complement in questions and ultimately was lexicalized, losing its original meaning. This postposition would have allowed the r in Tehuelche to become the A/S marker for the interrogative clause, first through a process of reanalysis and later by extension. We should keep in mind that the final /r/ sometimes becomes voiceless in Tehuelche and is often confused with /sˇ /. As to the appearance of n as an A/S marker in Tehuelche, employed when the verb is in unreal mood, we could postulate the existence of a postposition that led to the creation of this marker, in the same way that occurred with sˇ and r. In Selknam, there are several postpositions with a nasal, man, on, oni, ink, enk, am (Tonelli 1926), which could be cognates of the Tehuelche marker. Najlis mentions the existence of nearly 70 postpositions in Selknam but lists only 30, which makes it di‰cult to find the cognate form of the A/S marker in Tehuelche that could have once served as a circumstantial marker and was later reanalyzed as an A marker finally extended to S.

5. Conclusion It is highly probable that Proto-Chon presented an ergative-absolutive system that passed into Tehuelche and Selknam. In Selknam the system did not present an ergative case marker, while in Tehuelche this marker began to develop through an adposition that marked a circumstantial relational complement, supposedly the cognate of an adposition that also appears in Selknam, apparently with the same function. Although we know that these processes, which involve realignment resulting from the reanalysis and/or extension of a particular adposition, can occur as an internal process in a language, it seems probable that in the case of Tehuelche the influence of another language in the area helped to transform an ergative language into a marked-nominative one. The coexistence of Tehuelche with Mapudungun, a nominative-accusative language, led the ergative marker or agent marker of the transitive clause to be extended to S, that is, to the intransitive agent, leading to the transformation of this ergative system into a marked-nominative one. However, the process was never completed, since Tehuelche was replaced by Spanish before this could happen. In a previous work (2005b) we have shown the influence that Mapudungun has exerted over Tehuelche in its lexicon and in its

122

Ana Ferna´ndez Garay

phonological and morphosyntactic levels. One feature taken from Mapudungun is that the su‰x -n of the Tehuelche nominalization has replaced su‰x -j, which still exists today in a few infinitives. Selknam has a nominalization in -j, so it is evident that this change comes from the Mapudungun nominalization in -n, the most frequent of the various nominalizations of this language (Ferna´ndez Garay 2006). Another feature taken from Mapudungun is reduplication of nouns and verbs, an extensive strategy of that language that could have passed to Tehuelche nouns (cˇexcˇex ‘sand,’ ka:mka:m ‘dove,’ k’esˇk’esˇ ‘a kind of bird,’ etc.), and one that has not been observed in Selknam (Ferna´ndez Garay 2005b). A third feature is number morphology. In Tehuelche, dual and plural markers are extended to all persons. But the plural su‰x in Tehuelche can determine human, animate, and some inanimate nouns, whereas the dual su‰x can only be observed in humans. Even if number su‰xes in Tehuelche are di¤erent from those of Mapudungun, the extension of the markers in this language may have expanded it in Tehuelche, given that in Selknam the dual appears only in the first person and the plural only in the first and second persons, and that dual and plural markers do not a¤ect nouns (Ferna´ndez Garay 2007c). Based on the analysis of the marked-nominative system and on the results from previous work on some morphosyntactic features of Tehuelche, and in some cases of Gu¨nu¨na ku¨ne, that seem to be borrowed from Mapudungun (reduplication of certain nouns, dual and plural markers extended to all persons, and a su‰x -n that indicates a nominalization (Casamiquela 1983)), we might posit the existence of a linguistic area; but to do so, further studies and comparative work on the languages of Patagonia are needed.

Abbreviations adp dpt fti f m nft rm n pl posp

Adposition Distant past tense Future tense-intention Feminine Masculine Near future tense Real mood Neuter Plural Postposition

The emergence of a marked-nominative system

ps subs um 1, 2, 3

123

Predicate specifier Substitute Unreal mood First, second and third person

References Campbell, L. 2003. Areal Linguistics: A Closer Scrutiny. Paper presented at the 5th NWCL International Conference: Linguistic Areas, Convergence, and Language Change, 22–23 November 2002, University of Manchester. http:// www.ling.canterbury.ac.nz/documents/areal_linguistics.pdf Casamiquela, R. 1965. Rectificaciones y Ratificaciones: hacia una interpretacio´n definitiva del panorama etnolo´gico de la Patagonia y a´rea septentrional adyacente. Bahı´a Blanca: Cuadernos del Sur, Universidad Nacional del Sur. Casamiquela, R. 1983. Nociones de grama´tica del gu¨nu¨na ku¨ne: pre´sentation de la langue des Tehuelches septentrionaux austraux. Paris: CNRS. Dixon, M. R. W. 1994. Ergativity. Cambridge: Cambridge University Press. Dorian, N. 1981. Language Death: the Life Cycle of a Scottish Gaelic Dialect. Philadelphia: University of Pennsylvania Press. Escalada, F. 1949. El complejo Tehuelche: estudios de etnografı´a patago´nica. Buenos Aires: Coni. Ferna´ndez Garay, A. 1998. El Tehuelche: descripcio´n de una lengua en vı´as de extincio´n [Estudios Filolo´gicos, Anejo N 15]. Valdivia: Universidad Austral de Chile. Ferna´ndez Garay, A. 2000. Consecuencias lingu¨´ısticas de la situacio´n de desgaste del Tehuelche. In Memorias del Quinto Encuentro Internacional de Lingu¨´ıstica del Noroeste, Vol. 2, M. C. Moru´a Leyva and G. Lo´pez Cruz (eds), 165–177. Sonora: UniSon. Ferna´ndez Garay, A. 2002a. Incorporacio´n nominal en Tehuelche o Aonek’o ajen. In Actas del XIII Congreso Internacional de la Asociacio´n de Lingu¨´ıstica y Filologı´a de Ame´rica Latina (ALFAL), V. M. Sa´nchez Corrales (ed.), 553– 560. San Jose´ de Costa Rica: Universidad de Costa Rica. Ferna´ndez Garay, A. 2002b. Testimonios de los u´ltimos ranqueles. Textos originales con traduccio´n y notas lingu¨´ıstico-etnogra´ficas [Coleccio´n Nuestra Ame´rica, Archivo de Lenguas Indoamericanas]. Buenos Aires: Instituto de Lingu¨ı´stica, Facultad de Filosofı´a y Letras, Universidad de Buenos Aires. Ferna´ndez Garay, A. 2005a. Cambios en la estructura sinta´ctica del Tehuelche o Aonek’o ajen. Contacto o desgaste? In Contacto de lenguas en el sur argentino, Y. Hipperdinger (ed.), 11–28. Bahı´a Blanca: Ediuns. Ferna´ndez Garay, A. 2005b. Contacto lingu¨´ıstico en Patagonia. Actas del X Congreso Nacional de Lingu¨´ıstica. Salta: Sociedad Argentina de Lingu¨´ıstica y Universidad Cato´lica de Salta, 5–8 July 2005. ?

124

Ana Ferna´ndez Garay

Ferna´ndez Garay, A. 2006. La nominalizacio´n en lenguas indı´genas de Patagonia. To´picos del Seminario 15: 141–158. Ferna´ndez Garay, A. and Herna´ndez, G. 2006. Textos Tehuelches (aonek’o ajen): homenaje a Jorge Sua´rez. Munich: Lincom Europa. Ferna´ndez Garay, A. 2007a. Ana´lisis comparativo de los sistemas sinta´cticos del aonek’o a jen y del selknam. In Estudios lingu¨´ısticos y sociolingu¨´ısticos de lenguas indı´genas sudamericanas, A. Fernandez Garay and M. Malvestitti (eds), 89–107. Santa Rosa: Universidad Nacional de La Pampa. Ferna´ndez Garay, A. 2007b. Coexistencia de dos sistemas sinta´cticos en Tehuelche. International Journal of American Linguistics 73.1: 114–125. Ferna´ndez Garay, A. 2007c. Contacto tehuelche-mapuche en Patagonia. Actas del I Congreso internacional de sociolingu¨ı´stica y linguı´stica histo´rica, Universidad Nacional del Sur, Bahı´a Blanca (in press). Harris, A. C. and Campbell, L. 1995. Historical Syntax in Cross-linguistic Perspective. Cambridge: Cambridge University Press. Heine, B. and Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Matras, Y. 2009. Language Contact. Cambridge: Cambridge University Press. Matras, Y. and Sakel J. (eds), 2007. Grammatical Borrowing in Cross-linguistic Perspective. Berlin: Mouton de Gruyter. Musters, G. Ch. 1964. Vida entre los patagones. Buenos Aires: Hachette. Najlis, E. 1973. Lengua Selknam [Filologı´a y Lingu¨´ıstica 3]. Buenos Aires: Universidad del Salvador. Nardi, R. 1985. La araucanizacio´n de la Patagonia (sı´ntesis general). In Culturas indı´genas de la Patagonia, 235–264. Madrid: Biblioteca del V Centenario. Sua´rez, J. 1988. Clasificacio´n interna de la familia lingu¨´ıstica Chon. In Estudios sobre Lenguas Indı´genas Sudamericanas, 79–100. Bahı´a Blanca: Universidad Nacional del Sur. Thomason, S. G. 2001. Language Contact. Washington, DC: Georgetown University Press. Tonelli, A. 1926. Grama´tica e glosario della lingua degli ona-sˇelkna´m della Terra del Fuoco. Turin: Societa` Editrice Internazionale.

On polysemy copying and grammaticalization in language contact1 Bernd Heine 1. Introduction The Norman French dialect Guerne´siais of the Channel Island Guernsey distinguished traditionally between the comitative preposition dauve ‘(together) with’ and the instrumental prepositon atou ‘with, by means of.’ Guerne´siais was in contact with English for roughly 800 years and was extensively influenced by English.2 English does not distinguish between the two, using with for both, and contact with English appears to have been a contributing factor for Guerne´siais speakers to extend dauve from comitative to instrumental function, with the e¤ect that the instrumental preposition atou gradually disappeared from the language (Jones 2002: 157). In terms of a language contact analysis, this process can be described thus: Guerne´siais had two di¤erent case categories while English, the other language spoken on Guernsey, had a case polysemy instead, using the same form for both categories. What must have happened in the contact situation on this Channel Island was that Guerne´siais speakers replicated the polysemy they were confronted with in English, thereby creating an equivalence relation between the two languages concerned. 1. The present article was written while I was a visiting professor at the Centre de Recherches Linguistiques sur l’Asie Orientale, E´cole Nationale des Hautes E´tudes in Paris in May 2007. I wish to thank Hilary Chappell, the director of the Centre, and Alain Peyraube for the outstanding hospitality I was able to enjoy while in Paris. Furthermore, I wish to thank Tania Kuteva and Walter Breu for their valuable assistance on parts of the present article, and to Claudine Chamoreau and Regina Martinez Casas for introducing me to the fascinating world of language contact in Mexico. 2. Guerne´siais (or Guernsey) has been spoken on the island of Guernsey in the Channel Islands archipelago for more than a thousand years but is now moribund. After World War Two, when many island inhabitants who had been evacuated to England during the war returned home, English gradually began to replace this Norman dialect, a process that appears to be ongoing (Ramisch 1989; Jones 2002: 164).

126

Bernd Heine

Cases like this tend to be treated in the literature on language contact as instances of calquing, structural borrowing, loan translation, or, more recently, as polysemy copying (Heine and Kuteva 2005, chapter 3), whereby a meaning or combination of meanings of the model language (English in this case) is copied analogically in the replica language (Guerne´siais). Copying of this kind is a commonplace in the literature on both historical linguistics and contact linguistics; the equivalents of nominal compounds such as English skyscraper in the languages of Europe, or of predicative expressions such as take part (German teilnehmen ‘part take,’ Israeli Hebrew laka´xat xe´lek ‘take (a) part,’ and so on; Matras and Sakel 2007) bear witness to this process. Polysemy copying can be described as an abrupt rather than a gradual change, and it tends to be associated with lexical rather than grammatical replication. But rather than in terms of polysemy copying, there is another possible interpretation of this case. Comitatives and instrumentals are semantically and syntactically distinct categories, each requiring a di¤erent kind of verbal argument structure, and roughly two thirds of the languages of the world distinguish the two, using di¤erent grammatical expressions for them, as Guerne´siais formerly did. This situation contrasts with that of Europe, where less than one third of the languages has such a morphological distinction,3 and one salient process of contact-induced change in European languages concerns the extension of comitative markers to also introduce instruments (Heine and Kuteva 2006, chapter 5), thereby giving rise to comitative-instrumental polysemy (for example, English with, French avec). The process that happened in Guerne´siais thus can be interpreted as being yet another example of contact-induced grammaticalization from comitative to instrumental. The question then is which of the two hypotheses is to be preferred: did Guerne´siais speakers simply copy the polysemy of English with, or did they grammaticalize their comitative preposition to also serve as an instrumental? The former option would seem to be intuitively clearly more plausible, the more so since it is supported by massive evidence from lexical replication, where polysemy copying is fairly common. But there are also arguments in favor of grammaticalization, suggesting that lexical and grammatical replication do not always behave the same. One argument concerns 3. According to Stolz (1996: 127–128), 64.7 percent of his 323 languages of worldwide distribution distinguish comitatives from instrumentals, while in his sample of 51 European languages it is only 31.4 percent.

On polysemy copying and grammaticalization in language contact

127

directionality of change:4 there appears to be a cross-linguistically regular process leading from comitative to instrumental marking, Latin cum and Greek meta` being cases in point, while we are not aware of any development in the opposite direction.5 As has been shown in Heine and Kuteva (2006, chapter 5), the same kind of unidirectionality can be observed in language contact. The process that happened in Guerne´siais thus is in accordance with what we argue is a constraint on grammatical replication in particular and contact-induced change in general; in accordance with this constraint, it is fairly unlikely that Guerne´siais speakers would have drawn on their instrumental preposition atou to match the polysemy of English with. Still, on the basis of one single piece of evidence there is reason to question whether the grammaticalization hypothesis can be upheld. In the present article we will look at other cases of contact-induced change in order to explore whether this hypothesis can be defended. These cases all relate to language contact in Europe; they concern articles in Section 2, the possessive perfect in Section 3, and auxiliation in Section 4. In Section 5 we draw some conclusions from the findings made. Our concern in this article will be with grammatical replication, that is, a process where speakers create a new grammatical meaning or structure in language R (the replica language) on the model of some meaning or structure of another language M (the model language). In the framework used here, expounded in Heine and Kuteva (2005), grammatical replication contrasts with lexical replication, and both contrast with borrowing, which concerns phonetic substance, that is, either sounds or form-meaning units such as morphemes, words, or larger entities. Replication and borrowing are the major manifestations of contact-induced transfer or codecopying (Johanson 1992, 2002). This article will deal with contact-induced grammatical replication as a product, for which there is some cross-linguistic evidence, and I will have little to say about the process leading to this product since it is still largely ill understood. The following remarks are meant to provide at least some general understanding of the nature of this process, which has both a sociolinguistic and a linguistic component. At the beginning of the process as a sociolinguistic phenomenon there typically is spontaneous replication in bilingual interaction, where an individual speaker consciously or uncon4. For additional arguments, see Heine and Kuteva (2005), Section 3.2. 5. Even if it should turn out that there are counter-examples to this generalization, they will be rare.

128

Bernd Heine

sciously propagates novel features in the replica language that have been influenced by some other language (or dialect). Spontaneous replication, described with references to notions such as ‘‘speaker innovation’’ (Milroy and Milroy 1985: 15), is highly idiosyncratic and the vast majority of instances of it will have no e¤ect on the language concerned, being judged as what is commonly referred to as ‘‘speech errors.’’ But some instances may catch on: being taken up by other speakers and used regularly, they may become part of the speech habits of a group of speakers (early adopters), and they may spread to other groups of speakers, in exceptional cases even to the entire speech community. Still, this process does not necessarily lead to linguistic change: such innovations may remain restricted to some specific period of time, being abandoned either by the very speakers who introduced them or by the next generation of speakers. It is only if an innovation acquires some stability across time that grammatical replication has taken place. There is still a widespread assumption among linguists that grammatical structure, or syntax, cannot be ‘‘borrowed,’’ that is, transferred from one language to another. This assumption is reflected in a recent survey article by Sanko¤ (2001), who concludes that ‘‘[w]hether or not ‘‘grammar’’ or ‘‘syntax’’ can be borrowed at all is still very much in question . . . many students of language contact are convinced that grammatical or syntactic borrowing is impossible or close to it’’ (Sanko¤ 2001; see also Silva-Corvala´n 2007). We consider this no longer to be an issue, considering that there is by now abundant evidence to demonstrate that both grammar and syntax can be ‘‘borrowed,’’ or, as we will say here, replicated (see for example Ramisch 1989; Ross 1996, 2001; Johanson 1992, 2002; Aikhenvald 2002; Heine and Kuteva 2003, 2005, 2006), and the present article will provide further evidence in support of these observations.

2. Articles In our first example we will look at what Breu (2003a) calls Slavic microlanguages, namely Upper Sorbian and Molisean. Both Lower and Upper Sorbian are spoken in eastern Germany and have been a¤ected by nearly a millennium of contact with German. The present data are taken from non-standard Upper Sorbian as spoken by the Roman Catholic community in the west of the Oberlausitz (Upper Lausitia) (Breu 2003a: 28). Molise Slavic, in short Molisean, is the language of a community of Croatian speakers from the Hercegovinian Neretva Valley who emigrated

On polysemy copying and grammaticalization in language contact

129

around 500 years ago because of the Turkish invasion in the Balkans, settling in areas of southeastern Italy that were sparsely inhabited due to earthquakes and epidemics; today, Molisean is spoken in only two villages, Acquaviva and Montemitro, in the Molise region of Campobasso province. After contact both with the local varieties and with Standard Italian over a period of half a millennium, their language has been massively influenced by this Romance language (for a survey, see Breu 1998; see also Breu 1999, 2003a, 2003b, 2003c, 2004). There are a number of di¤erences between the two micro-languages. First, unlike Upper Sorbian, Molisean does not dispose of a standard form, and second, the model language is German in the case of Upper Sorbian but non-standard varieties of Italian spoken in the Molise region, as well as, over the past 150 years, Standard Italian in the case of Molisean. Otherwise, however, the situations of the two languages are fairly similar. The two model languages are structurally alike with reference to the following discussion, and contact between model and replica languages has in both cases had a long history. Note that there is no evidence whatsoever of any contact between speakers of the two micro-languages, so the changes that the two languages experienced must have happened independently of one another. The data to be discussed below are overwhelmingly from Breu (Breu 1998, 1999, 2003a, 2003b, 2003c, 2004); most of them were not available to us when we were working on Heine and Kuteva (2006). 2.1. Definite articles Work on grammaticalization suggests that a number of stages need to be distinguished in the gradual pragmatic and semantic progression of the evolution of many definite articles; for the purposes of the present study, they are typically the following (cf. Greenberg 1978; Hawkins 2004): 1. An item serves as a nominal modifier (rather than as a pronoun) for both spatial-deictic (for example near vs. far) and for anaphoric reference (demonstrative). 2. The item is no longer associated with spatial reference; its main function is now to refer to entities (objects or situations) mentioned earlier in discourse (definite anaphoric marker). 3. In addition to previous mentions, the item also refers to definite entities that are recoverable via contextually available knowledge (contextdefinite marker). 4. The item is no longer restricted to contextual knowledge; it may refer to any entity that is identifiable via world knowledge, including both

130

Bernd Heine

individual (token) and generic (type) entities (marker of ‘‘semantically definite’’). 5. The item is no longer restricted to definitely identifiable entities; it may in addition refer to specific indefinite entities, that is, entities that are not necessarily identifiable to the hearer. It can simply assert existence (indefinite-specific marker). 6. The item loses its association with referentiality; it no longer has a pragmatic or semantic function, it can occur in any context and with any noun, and it may be exapted for other functions such as noun classification (article loss). Grammaticalization theory would predict that if a language has reached a given stage then it has also passed through all preceding stages; accordingly, this evolutionary scenario can also be mapped onto synchrony and be used essentially as a synchronic implicational scale, with the proviso that language history is complex and that in a particular case there may be other factors interfering with the evolution. Both German and Italian, the two main languages having served as models for the Slavic languages that we are concerned with here, have a definite article, but neither has proceeded beyond stage four. In this sense they, like other definite articles in Europe (Heine and Kuteva 2006, chapter 3), could be called ‘‘fully grammaticalized’’ articles. But our concern is with what happened in the replica languages. Slavic languages are wellknown for their absence of articles; except for Macedonian, Bulgarian, and North Russian no definite articles are said to exist. So did language contact a¤ect this situation? Thanks to two detailed studies by Breu (2003a, 2005), there is now more detailed information on at least one of the microlanguages, namely Upper Sorbian, and the evidence presented leaves hardly any doubt that this question must be answered in the a‰rmative (see also Heine and Kuteva 2006, chapter 3). Upper Sorbian6 has grammaticalized its proximal demonstrative (‘this’) to a definite article, to the extent that the two are now formally distinct; for example, the forms are to´n/te/ta (masculine/feminine/neuter) for the nominative singular of the definite article and to´ne/tene/tane for the demonstrative.7 However, whereas German has developed a semantically definite 6. What we have to say in the following applies exclusively to non-standard varieties of Upper Sorbian, not the standard language (see Breu 2005). 7. The present treatment is based entirely on Breu (2005), which provides a finegrained analysis.

On polysemy copying and grammaticalization in language contact

131

article (stage four), Upper Sorbian has not proceeded beyond the contextdefinite stage three. In the examples below, sentences from Upper Sorbian (US) are given, followed by a German (G) and an English translation (there are no interlinear glosses in Breu’s publication; the markers in question are printed in bold, ø stands for lack of article). Example (1a) shows the deictic stage one, (1b) the anaphoric use (stage two), where the article refers to an object previously mentioned, and (1c) the context-definite (stage three) use, where the referent has not been previously mentioned but its identity is recoverable via context and/or world knowledge. Stage four relates most of all to generic uses where a referent is understood in its type rather its token value. Thus, a definite article is required in (1d) in German but not in Upper Sorbian. Still, it would be possible to use the article in the sense of stage two or three if the identity of the secretary were recoverable from the previous discourse rather than being understood in its type value. That it is type value that takes priority over discourse reference in the placement of the definite article can be shown with (1e), which is in accordance with the discourse setting of stage two, yet no definite article can be used on account of the generic nature of the referent. (1) Upper Sorbian (Breu 2005: 37¤.) a. Stage 1 US Cesˇ ty to´n kniu meˇc´? G Willst du das Buch haben? ‘Do you want to have the book (there)?’ b. Stage 2 US Wo´n sej sˇitko na jenu cedlku napisa. Ha potom wo´n to´n cedlu to´m po´licajej prˇed nosom dz´erzˇi. G Er schreibt sich alles auf einen Zettel. Und dann ha¨lt er den Zettel dem Polizisten vor die Nase. ‘He writes everything on a sheet of paper. And then he presents the sheet to the policeman right under his nose.’ c. Stage 3 US Mo´ smo´ zade jeno Lkweja jeˇli. To´n kur beˇ sˇreklich. G Wir sind hinter einem LKW hergefahren. Der Rauch war schrecklich. ‘We drove behind a truck. The smoke was terrible.’ d. Stage 4 US ø Sekretarka wot sˇule jo zawo´a´a. G Die Schulsekreta¨rin hat angerufen. ‘The secretary of the school has called.’

132

Bernd Heine

e.

US Neˇke budz´emo wot ø muchi pojeˇdac´. ø Mucha ma dweˇ krˇidle. G Jetzt sprechen wir u¨ber die Fliege. Die Fliege hat zwei Flu¨gel. ‘Now we will talk about the fly. The fly has two wings.’

These examples are meant to show three things: First, they correspond to predictions of grammaticalization theory, according to which grammatical changes proceed from less to more grammatical forms and structures. Accordingly, we would be surprised if Upper Sorbian had grammaticalized stage four but not any of the preceding stages. Second, it confirms what has been argued for in Heine and Kuteva (2003, 2005, 2006), namely that grammatical change in language contact is essentially unidirectional, in the same way as it is in situations not involving contact. And third, replica categories are generally less grammaticalized than the corresponding model categories: As Table 1 shows, the Upper Sorbian definite article has not reached the same advanced stage of grammaticalization as the German model category has. Table 1. Degree of grammaticalization from demonstrative to definite article in German and Upper Sorbian (Source: Breu 2003a, 2005). Stage

Function

German

Upper Sorbian

1

Demonstrative

þ

þ

2

Anaphoric-definite

þ

þ

3

Contextually definite

þ

þ

4

Semantically definite

þ

5 6

Indefinite specific Article loss

2.2. Further on definite articles: advanced grammaticalization Grammaticalization is a continuous process, proceeding from one context to another. When describing the process of grammatical replication in the preceding section in terms of stages, we were segmenting the continuum into a series of more salient points. In doing so, we were aware that the primary locus of change is not a given stage but rather a given context. We may illustrate this with an example from a contact situation that we

On polysemy copying and grammaticalization in language contact

133

mentioned earlier in Section 1, namely that between Guerne´siais and English. We saw in Section 2.1 how a language that had no article created a new definite article in a situation of language contact. But language contact may also be instrumental in the further grammaticalization of an already existing article by extending it to a new range of contexts. Even in a language such as English, which has a full-fledged stage four definite article, there are some restrictions on its use, in that certain kinds of clausal participants do not take the article the. Now, on Guernsey, where English has been in contact with the Norman French dialect Guerne´siais (see Section 1), the English definite article has spread to contexts where it would not be used in England but is used in this Norman dialect, in particular before names of languages (2a), adverbials of direction and position (for example, street names) (2b), adverbials of time expressing a regular repetition (2c), plural nouns with generic reference (2d), or nouns for institutions such as school and bus in generic uses (2e). (2) Replication in Guernsey English on the model of Guerne´siais (Ramisch 1989: 113–6; Jones 2002: 146) a. They never did the Guernsey French at school. b. He’s got a chain of h’m shops in the, in the Fountain Street. c. And we go the Saturday evening like – old time dancing. d. As a whole I believe the Guernsey people – are h’m friendly and they work together. e. It was always by the bus we went. Extension of the English definite article as a result of language contact with Celtic languages has also been reported for Irish English (for example, I had a few jars over the Christmas) and English spoken in the Gaelic-speaking area of Scotland (for example have porridge for the dinner; Ramisch 1989: 117). Furthermore, the extension of definite articles to new contexts is an ordinary grammaticalization process that can happen language-internally as well. For these two reasons, the present case need not be due to language contact. That, nevertheless, contact was a contributing factor in the case of Guernsey English is suggested by the fact that there are corresponding article uses in Guerne´siais, that is, English speakers use the article exactly in those contexts where it would be used in Guerne´siais, as can be seen in the examples of (3), corresponding to the English examples in (2).

134

Bernd Heine

(3) Guerne´siais (Ramisch 1989: 113¤.) a. mo˜ p r sav l bwo˜ fra˜se e lo˜gje e l patwa. ‘My father knew (the) good French, and English, and (the) patois.’ b.

nu vþ a la vil pur §pai. ‘We go to (the) town for shopping.’

c. nuzi vþ l samdi o ser. ‘We go there (the) Saturday (at.the) evening.’ d.

lez fa˜ apro˜ vit ˜ lo˜ga‰ . ‘(The) children learn a language quickly.’

e. nuze tu‰ur alai do˜ la bs. ‘We always went by (the) bus.’ What this example may show is that the development of articles, whether in language contact or language-internally, is not a discrete process but rather is gradual, proceeding from one context to another. 2.3. Indefinite articles The grammaticalization of indefinite articles proceeds similarly through a series of contexts and stages. Thus, it is the following stages that mark the gradual pragmatic and semantic evolution of many indefinite articles (Heine 1997b: 70¤.): 1. An item serves as a nominal modifier denoting the numerical value ‘one’ (numeral). 2. The item introduces a new participant presumed to be unknown to the hearer, and this participant is then taken up as definite in subsequent discourse (presentative marker). 3. The item presents a participant known to the speaker but presumed to be unknown to the hearer, irrespective of whether or not the participant is expected to come up as a major discourse participant (specific indefinite marker). 4. The item presents a participant whose referential identity neither the hearer nor the speaker knows (non-specific indefinite marker). 5. The item can be expected to occur in all contexts and on all types of nouns except for a few contexts involving, for instance, definiteness marking, proper nouns, predicative clauses, and so on (generalized indefinite article).

On polysemy copying and grammaticalization in language contact

135

Like the definite article scenario of Section 2.1, grammaticalization theory would predict that if a language has reached a given stage, then it has also passed through all preceding stages, and once again, this evolutionary scale can be used synchronically as an implicational scale, again with the proviso mentioned above. Both German and Italian have indefinite articles of stage four, but neither has stage five. Slavic languages are well-known for their absence of indefinite articles, with the possible exception of Macedonian. Once again, there is some information to suggest that there is more to this, and that language contact played some role, as Breu (2003a) shows convincingly in his analysis of Slavic micro-languages8 (see also Heine and Kuteva 2006, chapter 3). In the examples below, sentences from Upper Sorbian9 (US) are given, followed by a German (G) and an English translation (once again, there are no interlinear glosses in Breu’s publication, and the markers in question are printed in bold, ø ¼ lack of article, and *ø ¼ the article may not be omitted). The US form for jen- ‘one’ is made up of a complex morphophonological paradigm on the basis of distinctions of six cases, two numbers (singular, plural), and three genders (masculine, feminine, and neuter; see Breu 2003a: 36). Like German, US has a fully grammaticalized stage two, three, and four indefinite article that can be traced back to the numeral jen- ‘one.’ As the following examples show, their use is obligatory. Example (4a) illustrates the presentative use (stage two), characteristic of openings in tales, (4b) the specific indefinite stage 3, and (4c) the non-specific indefinite stage four. With abstract and generic referents as well, US shows roughly the same degree of grammaticalization as the German indefinite article does, cf. (4d). (4) Upper Sorbian (Breu 2003a: 37¤.) a. Stage 2 US To beˇsˇe jemo jena stara zˇona. G Es war einmal eine alte Frau. ‘Once upon a time there was an old woman.’ b. Stage 3 US Najmo´le jo jen to´sty muzˇ nutrˇ sˇisˇo´. G Plo¨tzlich kam ein dicker Mann herein. ‘Suddenly a fat man came in.’

*ø *ø *ø *ø

8. We are not able to do justice to the fine-grained analysis presented by Breu (2003a); the reader is referred to this work for many more details. 9. All data analyzed by Breu (2003) are from a rural, spoken variety of Upper Sorbian, which di¤ers considerably from Standard Upper Sorbian, especially with reference to the phenomena looked at here.

136

Bernd Heine

c. Stage 4 US Dysˇ tybe jen po´lcaj s´o´sˇi, to´n tybe zasˇperwe. *ø G Wenn dich ein Polizist ho¨rt, wird er dich einsperren. *ø ‘If a policeman hears you, he’ll arrest you.’ d. US Jen Serb nebcˇi. G Ein Sorbe lu¨gt nicht. ‘A Sorbian never lies.’ Even in many contexts where the model language German has an optional indefinite article, as with some collective and abstract nouns, the same situation obtains in US. (5) Upper Sorbian (Breu 2003a: 42) US Mensˇ, sym ja jen strach meˇ´ ! G Mensch, habe ich eine Angst gehabt! ‘Boy, was I scared!’ (Lit.: ‘Boy, did I have a fear’)

or ø or ø

Thus, there appears to be a high amount of intertranslatability between the two articles, and Breu (2003a: 66) concludes that US has reached the same degree of development as the German model category, which both are stage four articles. We thus seem to be faced with a situation where the process of replication is concluded, and where the model and the replica categories have become nearly identical. However, it would seem that this is not entirely correct. First, the replica category has not been extended to a number of idiomatic expressions where the model language would require the indefinite article. And second, there are a number of contexts, involving in particular generic concepts, where there must be an indefinite article in German while in US the indefinite article is either optional, as in (6a), or is disallowed, as in (6b). (6) Upper Sorbian (Breu 2003a: 44) a. US To´n jo tak sylny kaj jen elefant. G Er ist so stark wie ein Elefant. ‘He is as strong as an elephant.’ b. US Ja sym ´o´dny kaj ø law. G Ich bin hungrig wie ein Lo¨we. ‘I am hungry as a lion.’

or ø *ø *jen ‘a’ *ø

To conclude: in spite of the fact that the US category has become a nearly complete replica of the German model category, having reached the same general stage of grammaticalization, there remain a number of contexts where the replica category is less grammaticalized than the model.

On polysemy copying and grammaticalization in language contact

137

Breu (2003a) describes a parallel case from Molisean, which also must have lacked article-like grammatical forms prior to language contact. Like the Upper Sorbian numeral for ‘one,’ the Molisean ‘one’ disposes of a paradigm of morphophonological distinctions, one di¤erence being that the Molisean forms have long and short forms in addition, for example, je¨na vs. na (nominative masculine singular). Like US, Molisean shows roughly the same degree of grammaticalizaton of the numeral, having developed a stage four indefinite article of the same kind as the model language Italian. The reader is referred to Breu (2003a) for examples; it will su‰ce here to illustrate the more advanced stages. In (7) below, sentences from Molisean (M) are given, followed by an Italian (I) and an English translation (once again, there are no interlinear glosses; the markers in question are printed in bold, ø stands for lack of article). Example (7a) illustrates the use of a stage three article with an abstract noun, while (7b) shows a generic use of stage four, where use vs. non-use of the indefinite article appears to be lexically determined. Note that in both examples the replica and the model languages agree to the extent that both can be used with and without article. (7) Molisean (Breu 2003a: 42) a. M Jo, sa jima na strah! I Ahi, ho avuto una paura! ‘Boy, was I scared!’ b.

or ø or ø

M Ona je na sˇtudentesa. / ø profesoresa. I Lei e` una studentessa. / ø professoressa. ‘She is a student / a professor.’

As these examples show, Molisean speakers, like US speakers, have carried their numeral through all stages of grammaticalization, developing a stage four indefinite article largely equivalent to the Italian model. But here again, the replica category is not entirely identical to the model category; there are some contexts where the replica language does not use the article, or else the article is accepted by some speakers but not by others. Thus, in some generic uses of the non-specific stage four, Italian has an article, while Molisean speakers preferably do not use one. Interestingly, the model language employs not the indefinite but rather the definite article in the following example:

138

Bernd Heine

(8) Molisean (Breu 2003a: 44–5) M Kjikkjarijasˇ kana ø tovar ada prdi. I Parli come l’asino quando spetezza. ‘You are talking like a donkey when it passes wind.’ Table 2 shows two things in particular. First, it is precisely those Slavic languages that have had the most intensive contact with languages having stage four indefinite articles that also have created corresponding articles. At one end there are Upper Sorbian in eastern Germany with a history of nearly a millennium of contact with German, and Molisean, historically a variety of Croatian spoken in southeastern Italy, that has been in contact with Italian for roughly 500 years; at the other end there are the Eastern Slavic languages Ukrainian and Belorussian, both languages with the least amount of contact with article languages. Second, Table 2 also shows that contact-induced grammaticalization proceeds in one direction from one stage to the next, where a new stage is built on the stage immediately preceding it. Synchronically, this fact can once again be described in the form of an implicational scale of the following kind: If a given article has stage X then it also has all preceding stages of use. Table 2. Degree of the grammaticalization from numeral ‘one’ to indefinite article in selected Slavic languages (Sources: Breu 2003a; Heine and Kuteva 2006, chapter 3).10 Stage Function

Upper Molise Mace- Czech, Serbian, Ukrainian, Sorbian Slavic donian Bulga- Croatian, BeloPolish, russian rian Russian

1

Numeral ‘one’

þ

þ

þ

þ

þ

2

Presentative

þ

þ

þ

þ

(þ)

3

Specific indefinite þ

þ

þ

(þ)

4

Non-specific indefinite

þ

þ

þ

10. Note that we are restricted here to non-standard, colloquial, varieties of the languages concerned. As Breu (2003; 2005) has shown for Upper Sorbian, an entirely di¤erent picture would arise if Standard Upper Sorbian were chosen.

On polysemy copying and grammaticalization in language contact

139

To conclude, the same kind of constraint can be observed in definite as in indefinite articles: First, contact-induced grammaticalization proceeds along a largely predictable scale; there is no example where we would find, for example, that a Slavic language has replicated a stage three article but not a stage two article. Second, this suggests that in the replication of articles, speakers proceeded along the scale of stages, hence stage 1 > 2 > 3 > 4. And third, there is no language where in a situation of language contact the replica language underwent a process in the opposite direction, developing a definite article into a demonstrative or an indefinite article into a numeral. With reference to the question raised in Section 1, namely whether grammatical replication can be accounted for best with reference to polysemy copying or to grammaticalization, an answer in favor of the latter is more plausible: Rather than simply copying a German or Italian polysemy pattern, Sorbian and Molisean speakers appear to have chosen a more complex strategy, going through the whole process from demonstrative or numeral to article.

3. Possessive perfects Possessive perfects (‘have’-perfects), where a possessive verb is used both to encode possession (9a) and verbal aspect or tense (9b), can be considered to be a paradigm areal property of European languages: Nearly all languages of western and central Europe have one, while outside Europe their occurrence is extremely rare (see Haspelmath 2001). The following discussion is largely confined to some morphosyntactic properties of the categories concerned. Thus, issues that have figured prominently in the relevant literature, such as the semantic development from possessive via resultative to perfect (anterior) and to past tense meanings (see Heine and Kuteva 2006, chapter 4), or the relationship between ‘have’- and ‘be’-periphrasis, are not considered here (but see for example Pietsch 2004; Cennamo 2005).11

11. We wish to thank Andrii Danylenko, Bridget Drinka, Zygmunt Frajzyngier, Victor Friedman, and Ulrich Obst for helpful comments and insightful suggestions on an earlier version of this section.

140

Bernd Heine

(9) English a. She has a car. b. She has come. As is argued in Heine and Kuteva (2006), possessive perfects evolved in the languages of Europe roughly in the course of the last two millennia as a result of the grammaticalization of possessive constructions, more precisely of constructions for predicative possession of the ‘have’-type (Heine 1997a). On the basis of their evolution and structural characteristics, the following four main stages of evolution can be distinguished. 0. There is a possessive ‘have’-construction, like in (9a), but no possessive perfect. 1. There is now a resultative use pattern where the subject of the possessive verb is no longer conceived as a possessor but rather typically as an agent referentially identical with that of the verb constructed in the past passive participle (PPP), and the construction expresses a state of a¤airs resulting from the completion of the action denoted by the PPP-verb. At this stage, the construction exhibits many or all of the following properties: (a) Only transitive verbs are allowed as main verbs. (b) The PPP-verb still has the structure of a modifier of the patient, agreeing with the patient noun phrase in case, number, and/ or gender (if there are such morphological categories). (c) Nevertheless, the possessive verb tends to be interpreted as an auxiliary and the PPP-verb as the new main verb. (d) Both the possessive and the PPP-verbs tend to be associated with one and the same agent. 2. The main new properties are: (a) Instead of being transitive, the main verb may be intransitive; cf. (9b). (b) A possessive interpretation is now ruled out. (c) Agreement in number and gender between the main verb and the object gradually disappears, that is, the PPP-verb tends to be presented in one invariable form. (d) There is no more ambiguity, that is, there is only one agent, which can no longer be interpreted as a possessor. 3. The possessive perfect is now fully established and no longer subject to constraints: (a) Instead of human agents there may now be inanimate ‘‘agents.’’ (b) There are no or hardly any restrictions on the kinds of verbs serving as main verbs. Perfect (or anterior) categories found in the languages of the world have a limited number of conceptual sources (see Bybee, Perkins, and Pagliuca 1994). Among these sources, possessive constructions are extremely rare;

On polysemy copying and grammaticalization in language contact

141

more importantly, however, possessive perfects conforming to the fourstage model sketched above are found essentially only in Europe. On typological grounds therefore it seems unlikely that such constructions arose independently in di¤erent European languages; rather, the rise of these constructions must have been due to historical factors. Accordingly, Heine and Kuteva (2006) propose the following hypotheses:12 (a) The spread of possessive perfects across Europe is due mainly to language contact. (b) The di¤usion of these constructions across languages did not involve borrowing, that is, a transfer of form-meaning units, but rather the replication of a process whereby a possessive construction was grammaticalized to a construction marking aspect (in some cases later on also to tense). (c) The process was unidirectional, conforming to the four stages sketched above. 3.1. On the rise of possessive perfects Old Church Slavonic (863–950 CE) had a past passive participle formed exclusively from transitive verbs, but it had no possessive perfect (Friedman 1976: 97),13 and there was also no possessive perfect in the earliest forms of Island Celtic, Baltic, and Balto-Finnic languages. According to a widespread view, the ultimate donor of European possessive perfects was Ancient Greek. Thus, Drinka (2003b) argues that a new transitive periphrastic perfect formed with ‘have’ þ active aorist participle is found already in the writings of the fifth-century BC tragedians Sophocles and Euripides as well as in Herodotus. Greek is said to have provided the model for Latin: it was Latin authors thoroughly educated in Greek who replicated the possessive perfect in Latin. In the absence of a Greek-type active aorist or perfect participle, those Latin writers used their own past passive participle (PPP) as a complement for the verb habere ‘have’ (Drinka 2007: 19). The Latin construction subsequently spread across the Roman Empire, including the Greek-speaking areas of the East. 12. For Old Iranian and Old Armenian – both of which are spoken outside Europe – Benveniste (1952) argues that language contact is a highly unlikely factor for the development of the possessive perfect. 13. Friedman (1976: 97) refers to this construction as an analytic one, being ‘‘midway between a true perfect and an adjectival construction,’’ occurring in Bulgarian and Serbian.

142

Bernd Heine

The possessive perfect of modern European languages has its roots in Early Latin. As a result of a gradual process, the possessive perfect emerged in Late Latin as a distinct periphrastic active aspect category of stage one. It denoted current relevance of a past event (¼present anterior), spreading into narrative contexts. It is only after the sixth century that a stage two perfect began to emerge, subsequently spreading to other languages of western Europe. According to Haspelmath (1998: 285), possessive perfects di¤used across Europe at the time of transition between antiquity and the Early Middle Ages. In Iberian languages, habere was superseded by later reflexes of Latin tenere ‘to hold’ as a possessive verb, and possessive perfects based on tenere emerged fairly late. The Spanish tener-perfect gradually rose from the thirteenth century on, and up to now it is confined to transitive verbs, that is, it did not proceed beyond stage one. The Portuguese terperfect on the other hand has reached stage two: it has spread to intransitive verbs (Vincent 1982: 92). That possessive perfects spread via replication from Romance languages to Germanic is a plausible hypothesis, but it is not uncontroversial (see Heine and Kuteva 2006, Section 4.3 for discussion). In English, the rise of the possessive perfect goes back to the earliest stages of Old English, where it was used only in possessive contexts as an early stage one construction associated with resultative uses, while an advanced stage one possessive perfect must have existed in North Germanic from the Runic Scandinavian languages to Edda, and German appears to have turned into a stage two language by around 1000 CE. A historical reconstruction of the spread of the possessive perfect is urgently required; what surfaces from the sketchy information that is available, however, is that language contact must have played quite a role in its di¤usion. The result is that all Romance and Germanic languages are nowadays stage three languages. But this situation contrasts sharply with that to be found in what we will loosely refer to as Europe’s ‘‘linguistic periphery.’’ 3.2. A survey of ‘‘peripheral’’ European languages The situation in the modern Finnic, Slavic, Baltic, and Celtic languages, and in Basque, tends to be portrayed as one where there is essentially no possessive perfect. Table 3 summarizes this situation with reference to the stages distinguished above. In spite of all the research that has been

On polysemy copying and grammaticalization in language contact

143

carried out on the possessive perfect, the situation in many Slavic languages is still far from clear, especially with reference to which stage a given construction has reached; we have come across quite a number of controversial classifications on this issue, and the following generalizations therefore have to be taken with care.

Table 3. Stages of possessive perfects in ‘‘peripheral’’ European languages (main source: Heine and Kuteva 2006, Section 4.4). Stage Language

Family

0

1

2

Finnish

Finnic

þ

Lithuanian

Baltic

þ

Standard Russian

Slavic

þ

Welsh

Celtic

þ

Irish

Celtic

þ

þ

Polish

Slavic

þ

þ

Ukrainian (dial.)

Slavic

þ

þ

Belorussian (dial.)

Slavic

þ

þ

Czech

Slavic

þ

þ

Slovak

Slavic

þ

þ

Upper Sorbian

Slavic

þ

þ

Slovenian

Slavic

þ

þ

Serbian

Slavic

þ

þ

Croatian

Slavic

þ

þ

Bulgarian

Slavic

þ

þ

Breton

Celtic

þ

þ

þ

Southern Thracian Bulgarian

Slavic

þ

þ

þ

North Russian

Slavic

þ

þ

þ

Estonian Southwestern Macedonian

Finnic Slavic

þ þ

þ þ

þ þ

3

þ

144

Bernd Heine

Note that the structure of the possessive perfect is not the same across all the languages. In the Romance and Gemanic languages, predicative possession is built on what is called in Heine (1997a) the action schema [X has Y], relying on a more or less transitive ‘have’-verb. Accordingly, the possessive perfect also has some features of a transitive structure, where for example the agent is encoded as the subject of the clause. In some other languages, di¤erent conceptual schemas have been employed. Thus in the Celtic languages it was the goal schema [Y is to X] and in North Russian and Estonian the location schema [Y is at X] that were recruited, with the e¤ect that the resulting morphosyntactic structures of the perfect in these languages are strikingly di¤erent from those of Romance or Germanic languages, in that the agent is encoded as a locative argument rather than as the subject of the clause. The following example from Estonian may illustrate this situation (for another example from North Russian, see Heine and Kuteva 2004). In the location schema of the Balto-Finnic language Estonian, the possessor is expressed as a locative complement marked with the adessive case (ade) and placed typically clause-initially; nevertheless, it has some properties of a subject (Erelt and Metslang 2006). The possessee on the other hand is marked as the subject which controls agreement. Thus, the possessive stage 0, illustrated in (10a), can be glossed literally as ‘a new car is at me.’ The patient may take a past passive participle verb (ppp), and the construction expresses a resultant state where the ‘‘possessor’’ can be understood to be either a possessor (i) or an agent (ii): the possessor is the owner of the patient referent or the person a¤ected by the resulting state, cf. (10b). In other uses, this construction can only be interpreted meaningfully as a stage one perfect, especially when the formal subject is suppressed, as in (10c). This marks the transition to a stage two perfect, where the verb marked with the ppp is intransitive, as in (10d). But Estonian does not appear to have developed a stage three perfect, where the construction is used with the inanimate locative participant. This example may show that the absence of a ‘have’-verb was apparently no obstacle for Estonian speakers to develop a possessive perfect: they simply grammaticalized their location-based possessive construction into a perfect. (10) Estonian (Lindstro¨m and Tragel 2007) a. Mu-l on uus auto. i-ade be.3.sg new car ‘I have a new car.’

Stage 0

On polysemy copying and grammaticalization in language contact

b.

145

Mu-l on auto pestud. i-ade be.3.sg car wash.ppp i. ‘My car is washed.’ ii. ‘I have washed the/my car.’

c. Mu-l on (so¨o¨k) so¨o¨dud. i-ade be.3.sg dinner eat.ppp ‘I have eaten (my dinner).’

Stage 1

d.

Stage 2

Mu-l on magatud. i-ade be.3.sg sleep.ppp ‘I have slept.’

The reader is referred to Heine and Kuteva (2006) and Kuteva and Heine (2006) for exemplification of the stages presented in Table 3. This table is meant to show three conclusions. First, it is most of all those ‘‘peripheral’’ languages with a history of intense contact with Germanic or Romance languages that have created a more advanced possessive perfect, such as Breton with French, Estonian with German, and North Russian presumably with Scandinavian languages. Second, with the exception of the southwestern dialects of Macedonian,14 none of these languages has developed a stage three perfect as it is generally found in the Romance and Germanic languages. This is in fact to be expected since replicated categories tend to be less grammaticalized than the categories that provided the model (see Section 2). And third, and this is again most relevant for the purposes of the present article, the replication of possessive perfects followed the same sequence of stages as we observed in the case of articles, allowing for implicational predications of the form: If a language has reached stage X then it has also reached all preceding stages. The fact that there is no language that has, say, a stage three perfect but not a stage one perfect suggests that diachronically the sequence of grammaticalization was stage 0 > 1 > 2 > 3. The observation that we made in Section 2 on articles is thus confirmed by what we find in the replication of possessive perfects: Speakers do not simply copy a polysemy pattern but rather choose a more complex solution by grammaticalizing the model category in a step-by-step procedure.

14. Southwestern Macedonian overlaps with Standard Macedonian since the latter is based on western dialects. Note that Macedonian shows an areal patterning of stages, ranging from stage three in the southwest to stage one in the northeast (see Friedman 1976).

146

Bernd Heine

3.3. Advanced grammaticalization When dealing with definite articles we saw in Section 2.2 that a grammatical category that is fairly well established can still be further grammaticalized as a result of language contact, being extended to new contexts. We will now look at another example to show that this applies more generally to contact-induced change. The situation we are concerned with is contact between German and English (see Heine and Kuteva 2006; Section 4.4.3). Both German and English have a well-established stage three possessive perfect, but there is a di¤erence: While the English one is highly grammaticalized, being used with transitive and intransitive verbs, the German possessive perfect is much less so, showing remarkable contextual constraints in that it is largely restricted to transitive verbs. For example, with intransitive verbs such as geschehen ‘to happen,’ it may not be used; instead, the ‘be’-perfect must be used, cf. (11). As a result of their contact with English, speakers of Pennsylvania German have extended the use of the possessive perfect at the expense of the ‘be’-perfect; the former is now used with all transitive and most intransitive verbs, as it is in English. A mechanism that appears to have contributed to this process of context extension is that speakers of Pennsylvania German tend to equate their verbs with corresponding English verbs and to use the possessive perfect whenever the latter is required by the relevant English verb. For example, the intransitive verb form geschehne ‘happened’ takes the possessive perfect since the corresponding English verb form happened does so too, cf. (12). (11) High German Was ist geschehen? what is happened ‘What has happened?’ (12) Pennsylvania German (Enninger 1980: 344) Nau hoeret moll ihr liewe Leute, was geschehne hott zu derre Zeit. now listen once you dear people what happened has at that time ‘Now listen, dear people, what has happened at that time.’ That there is in fact a unidirectional process of extension of the possessive perfect from intransitive to transitive verbs and hence towards a generalization of this aspect category (at the expense of the older ‘be’-perfect) has been observed in a number of situations where German is in contact with English as the dominant language. Such situations include the Sauk County of Wisconsin in the USA and Australia: in both situations, the

On polysemy copying and grammaticalization in language contact

147

German possessive perfect is reported to have been generalized on the model of English, to the extent that the competing ‘be’-perfect was given up (Eichho¤ 1971: 53; Clyne 1972: 76). To conclude, grammaticalization is a fairly open-ended process, and the e¤ect of language contact can be that well-established functional categories, like the definite article in English or the possessive perfect in German, are pushed further along the cline of grammatical evolution.

4. The auxiliation of ‘threaten’-verbs In many languages there are words that behave both like lexical verbs and like functional categories expressing distinctions of tense, aspect, modality, and so on. The grammatical status of such words is frequently controversial; while some authors treat them as belonging to one and the same grammatical category, others assign them to di¤erent categories. The present section is concerned with such a case of ‘‘doublets,’’ a set of four constructions associated with verbs for ‘threaten’ in European languages. The discussion below is based on Heine and Miyashita (2008); the following example of the Portuguese verb ameac¸ar ‘to threaten’ illustrates these constructions, which we will refer to as C1, C2, C3, and C4. (13) Portuguese (Lima 2006) a.

seu irma˜o ameac¸ava destruir os planos de seus sobrinhos. C1 her brother threatened destroy the plans of her nephews ‘Her brother threatened to destroy the plans of her nephews.’

b.

A firma ameac¸a faleˆncia. the firm threatens bankruptcy ‘The company is threatened by bankruptcy.’

c. uma [. . .] melodia de amor [. . .] ameac¸ava na˜o acabar a melody of love threatened never to.finish ‘A melody of love ‘‘threatened’’ to never end.’ d.

C2

C3

um gordo e rubicundo merceeiro [. . .] ameac¸ava estalar C4 a fat and reddish merchant threatened to.tear toˆdas as costuras da farda. all the seams of costume ‘A fat, reddish trader was about to burst the seams of his attire.’

148

Bernd Heine

In the C1 construction of (13a), ameac¸ar functions as a lexical verb whose meaning can be paraphrased as in (14a), while in all remaining constructions there is what we will call functional ameac¸ar. The meaning of the latter, roughly paraphrased in (14b), has been described variously as an epistemic, subjective, modal, semi-modal, evidential, or temporal-aspectual auxiliary. C2 di¤ers from C1 in having an inanimate rather than a human subject, and C4 di¤ers from C3 in having a human rather than an inanimate subject. While C1 and C2 are lexical constructions, C3 and C4 can be described as ‘‘subject-to-subject raising’’ constructions; Table 4 summarizes the main grammatical properties of the four constructions. (14) A paraphrasis of the meaning of lexical (14a) and functional ‘threaten’ (14b) a. ‘Someone points out that s/he intends to do something that is undesirable to someone else.’ b. ‘Something undesirable is about to happen.’ Table 4. Distinguishing properties of the four ‘threaten’ constructions. Construction

The subject referent is human

threaten takes a subject argument

threaten expresses a speech act

Meaning of threaten

C1

þ

þ

þ

Lexical

C2



þ



Functional

C3







Functional

C4

þ





Functional

Portuguese is Europe’s most westerly language but, as Table 5 shows, roughly the same situation is found in other languages across Europe, and it is not restricted to Indo-European languages: it also includes FinnoUgric languages such Hungarian and Estonian. Di¤erences among these languages relate in particular to three points. First, the degree of productivity di¤ers among the languages concerned. While C1 is fully productive in all languages, the remaining constructions may di¤er in the extent to which they can be used productively. On the one hand there are languages such as Dutch, German, Spanish, or Portuguese, where all constructions are fully productive; on the other there are also languages where one of the constructions is severely restricted in its occurrence, to the extent that it has more in common with idiomatic expressions than with regularly used

On polysemy copying and grammaticalization in language contact

149

grammatical constructions. Second, the meaning of functional ‘threaten’ is not exactly the same across languages; in some languages it is more strongly associated with epistemic modality, while in others it is the notion of a proximative aspect (‘be on the verge of doing X’) or of evidentiality that is more pronounced. Third, the morphosyntactic constructions are also not really identical in the languages concerned. While most of the languages present the complement of the ‘threaten’-verb in the C3 and C4 constructions as an infinitival phrase, as can be seen in the Portuguese example of (13), some languages use a finite complement clause instead, as illustrated with the following examples from Hungarian, where there is a complementizer (hogy) and a finite verb in the complement clause. (15) Hungarian (Ferenc Ho¨rcher, personal communication) a. A fal azzal fenyegetett, hogy ledool. the wall with.that threatened that it.falls ‘The wall threatened collapsing.’ b.

Fenyegeto volt, hogy M ¨aria el ¨ajul. threatening it.was that Mary s/he.lose.consciousness ‘Mary threatened fainting.’

C3

C4

And finally, not all languages distinguish all constructions. More generally, it is the most easterly European languages that show the smallest range of constructions; thus, in Russian, Bulgarian, and Greek, only two of the four constructions are found (cf. the overview in Table 5). As shown in Heine and Miyashita (2008), the presence of these constructions across Europe must be the result of language contact, for several reasons. First, we are not aware of any language outside Europe that exhibits the same range of constructions. Second, genetic relationship can be ruled out as a possible explanation: neither Proto-Romance, ProtoGermanic, nor any other early European language distinguished these constructions, but at the same time the constructions are found in language families in Europe that are as far as we know genetically unrelated (Indo-European and Finno-Ugric). And third, the rise and development of these constructions took place roughly around the same general period in the history of European languages (see below). Our knowledge of the diachronic processes leading to the presence of this ‘‘polysemy’’ pattern in European languages, while limited, still allows for a couple of cross-linguistic generalizations. The first concerns chronology: there are a few historical data that make it possible to date the changes

150

Bernd Heine

Table 5. Degree of grammaticalization of ‘threaten’-constructions in European standard languages (Parentheses ¼ use of the construction is either marginally possible or is restricted to certain contexts). Language

C1

C2

C3

C4

Portuguese

þ

þ

þ

þ

Spanish

þ

þ

þ

þ

French

þ

þ

þ

(þ)

Italian

þ

(þ)

þ



Friulian

þ

þ

þ



Romanian

þ

þ

(þ)



English

þ

(þ)

þ

þ

Dutch

þ

þ

þ

þ

German

þ

þ

þ

þ

Danish

þ

þ

þ



Norwegian

þ

þ

þ



Swedish

þ

þ

þ



Estonian

þ

þ

þ



Serbian

þ

þ

(þ)



Bulgarian

þ

þ





Slovak

þ

þ

þ

þ

Russian

þ

þ





Greek Hungarian

þ þ

þ þ

 þ

 þ

that are responsible for the structural diversity characterizing the ‘threaten’constructions in the modern European languages. These data are summarized in Table 6 (for more details, see Heine and Miyashita 2008). What they suggest is that the lexical C1 construction was the first to exist; except for French, it was essentially the only construction to be found in European languages prior to 1500. C2 appears to have been next to arise, to be followed by C3 and, from the eighteenth century onward, by C4. Thus,

On polysemy copying and grammaticalization in language contact

151

there is a diachronic sequence C1 > C2 > C3 > C4 which is largely in accordance with what grammaticalization theory would have predicted.15 This chronology furthermore suggests that the grammaticalization of ‘threaten’-constructions must have originated in French, subsequently being replicated in other languages of western Europe, where it is attested only several centuries later.16 The di¤usion of this grammaticalization process in central and eastern Europe appears to be a more recent development, being weakest in eastern Europe, where the process has not proceeded beyond the C2 construction. As we observed already in Section 3, neither genetic nor typological factors constituted any significant boundary in this di¤usion process, which a¤ected Indo-European languages in much the same way as their Finno-Ugric neighbors Hungarian and Estonian, both sharing a long history of intense contact with German (cf. Table 5).

Table 6. A chronological overview of first attestations of stages in the grammaticalization of ‘threaten’-constructions in European languages. Construction

French menacer

Spanish amenac¸ar

German drohen

Dutch dreigen

English threaten

C1

Before 1100

Before 1500

Before 1500

Before 1500

Before 1500

C2

1200

1495

1560

C3 C4

1200 1751

1494 19th century

1738 ca. 1800

1627 1566

1780

15. As pointed out in Heine and Miyashita (2008), the reconstruction based on grammaticalization theory using synchronic evidence yields the development C1 > C2 > C3/C4. This reconstruction is less specific than the one based on historical records since it does not determine whether C3 preceded or followed C4. 16. This hypothesis can be reconciled with extra-linguistic observations on European history: Paris was in a culturally and intellectually privileged situation around the time between the twelfth and the fourteenth centuries, being a center of cultural di¤usion across much of Europe. Accordingly, there is reason to assume that the development from lexical to functional ‘threaten’ started out in northern France as part of a more general cultural di¤usion process a¤ecting a larger part of Europe.

152

Bernd Heine

The second kind of generalization concerns the grammaticalization process that gave rise to the ‘threaten’-constructions. This process proceeded gradually from lexical to grammatical to even more grammatical structures along the following stages (we will return to this case in Section 5). C1: At the beginning there was only a lexical construction which consisted of ‘threaten’ as a control verb taking an agentive subject acting intentionally. C2: The transition was made possible when the lexical C1 construction was allowed to take inanimate subjects. Inanimate subjects are incompatible with agents acting intentionally and with the semantics and valency of ‘threaten.’ While in the new construction ‘threaten’ still had the morphosyntactic format of a clausal predicate, its lexical semantics was desemanticized, giving way to that of the functional notion ‘something undesirable is about to happen’ (14b). C3: The presence of inanimate subjects and a verb expressing a grammatical function paved the way for the rise of the auxiliary-like ‘‘raising’’ construction, with ‘threaten’ increasingly acquiring the properties of an auxiliary and an infinitival complement assuming the role of the new main verb.17 C4: Finally, the end-point was reached when C3 was no longer restricted to inanimate subjects but could also take human subjects; accordingly, the emerging C4 construction is characterized by lack of the animacy constraint.

This example of ‘threaten’-constructions confirms what we saw in the preceding Sections 2 and 3. First, contact-induced grammatical replication does not take place overnight; rather, it may require centuries to be accomplished. Second, it is clearly structured, proceeding gradually from less grammatical to more grammatical structures. Accordingly, we are dealing once more with a grammaticalization scale of the kind we observed in the preceding sections. Third, this example also shows that contact-induced grammatical change has both a language-internal and an external component. The change is internal since it is in accordance with universal principles of grammaticalization (Heine, Claudi, and Hu¨nnemeyer 1991; Hopper and Traugott 2003) and, hence, could as well have happened without language contact; as the rich literature on grammaticalization shows, similar changes from lexical verb to auxiliary structure without involving language contact are well documented (see for example Bybee, 17. In languages such as Slovak and Hungarian this was not a verbal infinitive complement but rather a finite complement clause (see Heine and Miyashita 2008).

On polysemy copying and grammaticalization in language contact

153

Perkins, and Pagliuca 1994). The external component relates to the fact that the process was propelled by language contact: it is unlikely that a process that took place in a number of European languages about the same time in Europe’s history, but as far as we know nowhere else in the world, could be accounted for in any way other than in terms of language contact. And finally, this example lends further support to the conclusion that polysemy copying is not the mechanism, or the only possible mechanism, leading to grammatical replication. As Table 6 shows, there is once more an implicational scale where presence of construction X (for example, C4) implies that all preceding constructions (C1, C2, and C3) have existed in the language concerned. Accordingly, grammatical replication leading to the rise of ‘threaten’-constructions in many European languages can hardly be reduced to one where speakers in language contact simply replicated a polysemy pattern of another language; rather, these speakers were constrained by what is a possible replication and what is not. For example, developing a C3 construction when a C2 construction already exists is a possible contact-induced change, while developing a C2 construction out of a C3 construction does not seem to be a linguistic change that can occur in language contact.

5. Accounting for the sequencing of stages Perhaps the most surprising observation made in the preceding paragraphs was that there are significant constraints characterizing sequences of development. This raises three questions: (a) Why did speakers replicate one stage after the other rather than all in one go? (b) Why did these speakers not follow any other conceivable order in replication? (c) Is this behavior motivated by the situation that speakers of the replica languages found in the respective model languages? That question (c) must be answered in the negative is suggested by the fact that in most, if not all, of the cases examined above, speakers of the replica languages were exposed to the full range of stages and constructions to be found in the model languages. For example, speakers of Upper Sorbian and Molisean must have been familiar with all the stages of articles in German and Italian respectively; nevertheless, they did not grammaticalize the entire range of stages (see Section 2). In a similar fashion, Breton

154

Bernd Heine

speakers were familiar with the whole range of stages of the French possessive perfect, and Estonian speakers with that of German; nevertheless, they did not carry grammaticalization to completion (Section 3). Accordingly, there is reason to argue that the structure of the model languages cannot be held responsible for the sequencing of stages nor for the fact that the replica categories are less grammaticalized than the corresponding model categories. There is a general answer to questions (a) and (b). The particular sequencing of stages in the development of replica categories is exactly as grammaticalization theory would predict it, proceeding from lexical to grammatical and from grammatical to even more grammatical constructions. But why should this be so? Unfortunately, we still know too little about the historical and sociolinguistic processes concerned to give a satisfactory answer. But among the case studies discussed above there is one that may shed some light on the kind of process concerned. We argued in Section 4 that the grammaticalization of ‘threaten’-verbs to a modal-aspectual auxiliary in the languages of Europe is the result of a historical process of contactinduced linguistic change. As the chronology of first attestations in Table 6 suggests, this process can be assumed to have started in medieval France, subsequently di¤using across Europe (see Heine and Miyashita 2008 for more details). Accordingly, in languages such as Portuguese, German, or Hungarian the process must have been due to grammatical replication on the model of French. In the languages other than French, up until the end of the fifteenth century there was only a lexical construction (C1) which consisted of ‘threaten’ as a control verb taking an agentive subject acting intentionally. What was replicated subsequently was not really a new construction, C2, but rather the use of the lexical construction C1 with inanimate subject referents that were presented metaphorically as agents. Thus, in Martin Luther’s writings in the early sixteenth century, common subject referents in Early New High German were abstract concepts like himmel ‘heaven,’ su¨nde ‘sin,’ urteil ‘judgment’ or gesetz ‘law.’ The following example illustrates the new pattern arising. (16) Early New High German (Hans Sachs, 1494–1576) dergleichen auch ohn-zahlbar sorgen, troen im abendt such also countless sorrows threaten him evening und den morgen. and the morning ‘Countless sorrows of this kind are threatening him evening and morning.’

On polysemy copying and grammaticalization in language contact

155

Since inanimate subjects are incompatible with agents acting intentionally and with the semantics and the valency of ‘threaten,’ the lexical meaning (14a) of ‘threaten’ was desemanticized, giving rise to the aspectual-modal meaning sketched in (14b). This new discourse pattern gained in frequency of use and gradually turned into a new construction, that is, the C2 construction as depicted in Table 4. The next major innovation occurred in the late seventeenth century, when the German C2 construction was no longer restricted to nominal complements but could now be used with infinitival complements, as in (17). (17) New High German (Johann Heinrich Merck, 1741–1791; Briefsammlung 2, 45) . . . die brust droht zu zerspringen. the breast threatened to burst ‘. . . the breast threatened to burst.’ In the same way as the infinitival complement gradually assumed the function of the main verb, the ‘threaten’-verb acquired properties of an auxiliary, and the end product of this process was the grammatical construction C3, characterized by an inanimate subject referent and an infinitival main verb. The final major change occurred in the second half of the eighteenth century, when speakers gave up the restriction of taking only inanimate subjects, thereby making it possible to also have human subject referents, as in (18) – thereby giving rise to the C4 construction. (18) New High German (Goethe, 1749–1832, Hermann und Dorothea, 40, 320) es knackte der fuss, sie drohte zu fallen, . . . it cracked the foot she threatened to fall ‘Her foot cracked, she was about to fall down. . . .’ The main changes portrayed in this brief sketch of the development of German drohen ‘threaten’ suggest that first, rather than constructions or stages, speakers appear to be replicating specific semantic and morphosyntactic properties in certain contexts that they observe in the model language. The resulting sequence of stages thus can be interpreted most appropriately as an epiphenomenal product of discourse manipulation. And second, the changes that we observed are not independent of one another; rather, one builds on the other in the rise of new constructions: The development from human to inanimate subjects is a prerequisite for C2, that from nominal to infinitival complements for C3, and the desemanticization of subjects for C4.

156

Bernd Heine

These observations may have provided a basis for understanding why grammatical replication in language contact is constrained in the way we described it in the preceding sections, proceeding from one stage to another rather than in one go and showing the kind of directionality it does, but more data are required on this issue.

6. Conclusions In the preceding sections we were restricted to a limited spectrum of questions that need to be addressed in understanding grammatical replication. We ignored in particular the question of what exactly speakers take as their model of transfer. Do they replicate the process of grammaticalization that took place earlier in the model language or do they create a new category in the replica language on the basis of universal principles of grammaticalization? In other words, do they use replica or ordinary grammaticalization (Heine and Kuteva 2005)? All evidence available suggests that, at least in the cases examined in this article, it is invariably the latter that must have been involved. What speakers have at their disposal is as a rule spoken or written discourse in the model language which provides them with information on the structures concerned. But in designing the replica categories they are constrained by what already exists in the replica language. For example, they will not replicate a stage three structure unless there already exists a stage two structure in the replica language; in other words, they will ignore more advanced stages of grammaticalization, even though the model language provides them with su‰cient information on the presence of such advanced stages. That such information in fact exists is hardly open to question, considering that in the cases that we were concerned with, such as contact between Upper Sorbian and German, or Molisean and Italian, there has been intense linguistic interaction for centuries. But what remains unclear, in spite of all the work that has been done by students of grammaticalization, is what exactly is responsible for the existence of such constraints; more research is required on this issue. The examples discussed in this article lend support to what has been observed in other cases where we have comparative data on contactinduced grammatical change. Rather than replicating a grammatical category in toto, speakers start out with the replication of a use pattern characterizing the initial stages of grammaticalization, and it requires a situation of long and intense contact for the replica category to attain the same

On polysemy copying and grammaticalization in language contact

157

degree of grammaticalization as the corresponding category of the model language. This constraint on contact-induced grammatical replication suggests that, at least in cases such as the ones discussed in this article, there really is no polysemy copying; rather, what language contact triggers is a gradual process from less to increasingly more grammatical structure, a process that occasionally ends up in a fully equivalent replica category, the case of indefinite articles in Upper Sorbian and Molisean coming close to being complete replicas of their respective German and Italian models. In most of the cases that have been reported on grammatical replication, however, the process does not run its full course; rather, the replica categories remain clearly less grammaticalized than the corresponding model categories. In other words, they are not really complete replicas of their models. Another problem with the term ‘‘polysemy copying,’’ relating to directionality in contact-induced grammatical change, was pointed out in the introductory Section 1. While being a complex process, grammatical replication exhibits one important constraint: it is essentially unidirectional. Accordingly, we find comitative markers assuming the function of instrumental markers, demonstratives developing into definite articles, numerals for ‘one’ into indefinite articles, possessive constructions giving rise to verbal aspect categories, or lexical verbs turning into auxiliaries, but we will not expect to find developments in the opposite direction, where for example an article develops into a demonstrative or numeral, or an auxiliary into a lexical verb (see Heine and Kuteva 2003, 2005, 2006 for more examples). This generalization about contact-induced grammatical change is hard to reconcile with the view that speakers in language contact simply copy a grammatical polysemy pattern. It may well be that the term ‘‘polysemy copying’’ has some relevance in cases where replication is restricted to one single grammatical property, for example when a comitative marker assumes an instrumental function, as we saw in Section 1. But even such a seemingly simple change actually involves a more complex process: it entails that speakers, on the model of some other language M, extend an existing use pattern in language R to a new range of contexts where an interpretation in terms of a comitative notion does not really make sense, including contexts where the participant introduced by the comitative marker is more reasonably interpreted as an instrument rather than a companion. It furthermore entails that, over time, this new context extension acquires some stability of use and eventually comes to be accepted by some community of R speakers, first as a use pattern and eventually perhaps as a new construction.

158

Bernd Heine

To conclude: polysemy copying does exist, as we saw in Section 1, being based on a formula of equivalence like (19), where Mx stands for a category (or structure) of the model language and Rx for a corresponding category of the replica language. Compared to that, grammatical replication presents a more complex structure, as depicted in (20), where the equivalent of Mx is not a category in the replica language but rather a process from a non-equivalent category Ry to an equivalent category Rx. (19) Mx ¼ Rx (20) Mx ¼ [Ry > Rx] Unlike grammatical replication, polysemy copying can be described as an abrupt rather than a gradual process (cf. Matras and Sakel 2007), and it tends to be associated with lexical rather than with grammatical replication. Thus, useful as it is for describing lexical replication, the notion of polysemy copying does not contribute much to understanding what grammatical replication is about.

Abbreviations ade ppp sg 1, 2, 3

Adessive case Past passive participle Singular First, second, third person

References Abraham, W. (ed.). 2004. Focus on Germanic Typology (Studia Typologica, Beihefte zu Sprachtypologie und Universalienforschung [STUF] 6). Berlin: AkademieVerlag. Aikhenvald, A. Y. 1996. Areal di¤usion in northwest Amazonia: the case of Tariana. Anthropological Linguistics 38: 73–116. Aikhenvald, A. Y. 2002. Language Contact in Amazonia. New York: Oxford University Press. Aikhenvald, A. Y. 2003a. Mechanisms of change in areal di¤usion: new morphology and language contact. Journal of Linguistics 39: 1–29. Aikhenvald, A. Y. 2003b. Multilingualism and ethnic stereotypes: the Tariana of northwest Amazonia. Language in Society 32: 1–21.

On polysemy copying and grammaticalization in language contact

159

Aikhenvald, A. Y. and Dixon, R. M. W. (eds). 2001. Areal Di¤usion and Genetic Inheritance: Problems in Comparative Linguistics. Oxford: Oxford University Press. Alanne, E. 1972. Zur Rolle der syntaktischen Interferenz der verwandten und unverwandten Sprachen. Neuphilologische Mitteilungen 73: 568–574. Aranovich, R. (ed.). 2007. Split Auxiliary Systems: a Cross-linguistic Perspective. (Typological Studies in Language, 69) Amsterdam: John Benjamins. Backus, A. 2005. Codeswitching and language change: one thing leads to another? International Journal of Bilingualism 9.3–4: 307–340. Blake, B. J. and Burridge, K. (eds). 2003. Historical linguistics 2001: selected papers from the 15th International Conference on Historical Linguistics, Melbourne, 13–17 August 2001. (Amsterdam Studies in the Theory and History of Linguistic Science, 237) Amsterdam: John Benjamins. Boeder, W., Schroeder, C., Wagner, K. H., and Wildgen, W. (eds). 1998. Sprache in Raum und Zeit: In memoriam Johannes Bechert. Volume 2. Beitra¨ge zur empirischen Sprachwissenschaft. Tu¨bingen: Gunter Narr. Boretzky, N., Enninger, W., and Stolz.T. (eds). 1989. Beitra¨ge zum 5. Essener Kolloquium u¨ber ‘‘Grammatikalisierung: Natu¨rlichkeit und Systemo¨konomie’’ Volume 1. (Essener Beitra¨ge zur Sprachwandelforschung, 7) Bochum: Brockmeyer. Breu, W. (ed.). 1990a. Slavistische Linguistik 1989. Munich: Sagner. Breu, W. 1990b. Sprache und Sprachverhalten in den slavischen Do¨rfern des Molise (Su¨ditalien). In Slavistische Linguistik 1989. W. Breu (ed.), 35–65. Munich: Sagner. Breu, W. 1992. Das italokroatische Verbsystem zwischen slavischem Erbe und kontaktbedingter Entwicklung. In Slavistische Linguistik 1991, T. Reuther (ed.), 93–122. Munich: Sagner. Breu, W. 1994. Der Faktor Sprachkontakt in einer dynamischen Typologie des Slavischen. In Slavistische Linguistik 1993, H. R. Mehlig (ed.), 41–64. Mu¨nchen. ¨ berlegungen zu einer Klassifizierung des grammatischen Wandels Breu, W. 1996. U im Sprachkontakt (am Beispiel slavischer Kontaktfa¨lle). Sprachtypologie und Universalien-forschung 49.1: 21–38. Breu, W. 1998. Romanisches Adstrat im Moliseslavischen. Die Welt der Slaven 43: 339–354. Breu, W. 1999. Die Komparation im Moliseslavischen. In Des racines et des ailes: the´ories, mode`les, expe´riences en linguistique et en didactique, R. Me´trich et al. (eds), 37–63. Nancy: A.N.C.A. Breu, W. 2003a. Der indefinite Artikel in slavischen Mikrosprachen: Grammatikalisierung im totalen Sprachkontakt. In Slavistische Linguistik 2001, H. Kuße (ed.), 27–68. Munich: Sagner. Breu, W. 2003b. Bilingualism and linguistic interference in the Slavic-Romance contact area of Molise (Southern Italy). In Words in Time: Diachronic Semantics from Di¤erent Points of View, R. Eckardt et al. (eds), 351–373. Berlin: Mouton de Gruyter.

160

Bernd Heine

Breu, W. 2003c. Impersonales Neutrum im Moliseslavischen. In Rusistika, Slavistika, Lingvistika: Festschrift fu¨r Werner Lehfeldt, S. Kempgen et al. (eds), 57–71. Munich: Sagner. Breu, W. 2004. Der definite Artikel in der obersorbischen Umgangssprache. In Slavistische Linguistik 2002, M. Krause and C. Sappok (eds), 9–57. Munich: Sagner. Bybee, J. L., Perkins, R. D., and Pagliuca, W. 1994. The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago: University of Chicago Press. Cennamo, M. 2005. The rise and development of analytic perfects in Italo-Romance. Paper presented at the Conference on Linguistic Theory and Grammatical Change, Rosendal (Norway), 31 May 31–4 June, 2005. Chikovani, G. 2005. Linguistic contacts in Central Asia. In Linguistic Convergence and Areal Di¤usion: Case Studies from Iranian, Semitic and Turkic, E. Csato´ et al. (eds), 127–32. London/New York: Routledge Curzon. Clyne, M. 1972. Perspectives on Language Contact Based on a Study of German in Australia. New York: W. S. Heinman. Comrie, B. and Corbett, G. G. (eds). 1993. The Slavonic Languages. London/New York: Routledge. Csato´, E.A., Isaksson, B. and Jahani, C. (eds). 2005. Linguistic Convergence and Areal Di¤usion: Case Studies from Iranian, Semitic and Turkic. London/New York: Routledge Curzon. Danchev, A. 1989. Language change typology and adjectival comparison in contact situations. Folia Linguistica Historica 9.2: 161–174. Dorian, N. 1994. Varieties and variation in a very small place: social homogeneity, prestige norms, and linguistic variation. Language 70: 631–696. Drinka, B. 2003a. Areal factors in the development of the European periphrastic perfect. Word 54.1: 1–38. Drinka, B. 2003b. The formation of periphrastic perfects and passives in Europe: an areal approach. In Historical linguistics 2001: selected papers from the 15th International Conference on Historical Linguistics, Melbourne, 13–17 August 2001, B. J. Blake and K. Burridge (eds), 105–128. Amsterdam: John Benjamins. Drinka, B. 2004. Pra¨teritumschwund: evidence for areal di¤usion. In Focus on Germanic Typology, W. Abraham (ed.), 211–240. Berlin: Akademie-Verlag. Drinka, B. 2007. The development of the HAVE perfect: mutual influences of Greek and Latin. In Split Auxiliary Systems: a Cross-linguistic Perspective, R. Aranovich (ed.), 101–121. Amsterdam: John Benjamins. Durie, M. and Ross. M. (eds). 1996. The Comparative Method Reviewed: Regularity and Irregularity in Language Change. New York: Oxford University Press. Dutkova-Cope, L. 2001. The language of Czech Moravians in Texas: Do you know what Pa´rknu Ka´ru u Hauza means? Southwest Journal of Linguistics 20: 51–84. Eckardt, R., von Heusinger. K., and Schwarze. C. (eds). 2003. Words in Time: Diachronic Semantics from Di¤erent Points of View. Berlin: Mouton de Gruyter.

On polysemy copying and grammaticalization in language contact

161

Eichho¤, J. 1971. German in Wisconsin. In The German Language in America: a Symposium, G. G. Gilbert (ed.), 43–57. Austin: University of Texas Press. Enninger, W. 1980. Syntactic convergence in a stable triglossia plus trilingualism situation in Kent County, Delaware, U.S.A. In Sprachkontakt und Sprachkonflikt, P. H. Nelde (ed.), 343–350. Wiesbaden: Franz Steiner. Erelt, M. and Metslang, H. 2006. Estonian clause patterns – from Finno-Ugric to Standard Average European. Linguistica Uralica 4: 254–266. Filppula, M. 2003. The quest for the most ‘‘parsimonious’’ explanations: endogeny vs. contact revisited. In Motives for Language Change, R. Hickey (ed.), 161– 173. Cambridge: Cambridge University Press. Flores Farfa´n, J. A. 2004. Notes on Nahuatl typological change. Sprachtypologie und Universalienforschung 57: 85–97. Fortescue, M. 1993. Eskimo word order variation and its contact-induced perturbation. Journal of Linguistics 29: 267–289. Friedman, V. A. 1976. Dialectal synchrony and diachronic syntax: the Macedonian perfect. Chicago Linguistic Society, Papers from the Parasession on Diachronic Syntax, 96–104. Gast, V. and van der Auwera, J. 2006. What is ‘contact-induced grammaticalization’? evidence from Mayan and Mixe-Zoquean languages. Typescript, University of Antwerp. Gilbert, G. G. (ed.). 1971. The German Language in America: a Symposium. Austin: University of Texas Press. Gomez-Imbert, E. 1996. When animals become ‘‘rounded’’ and ‘‘feminine’’: conceptual categories and linguistic classification in a multilingual setting. In Rethinking Linguistic Relativity, J. Gumperz and S. Levinson (eds), 438–469. Cambridge: Cambridge University Press. Greenberg, J. H. 1978a. How does a language acquire gender markers? In Universals of Human Language, J. H. Greenberg et al. (eds), 47–82. Stanford: Stanford University Press. Greenberg, J. H., Ferguson, C. A., and Moravcsik. E. (eds). 1978. Universals of Human Language. Volume 3: Word Structure. Stanford: Stanford University Press. Gumperz, J. and Levinson. S. (eds). 1996. Rethinking Linguistic Relativity. (Studies in the Social and Cultural Foundations of Language 17.). Cambridge: Cambridge University Press. Haase, M. 1997. Gascon et basque: bilinguisme et substrat. Sprachtypologie und Universalienforschung 50.3: 189–228. Haig, G. 2001. Linguistic di¤usion in present-day East Anatolia: from top to bottom. In Areal Di¤usion and Genetic Inheritance: Problems in Comparative Linguistics, A. Y. Aikhenvald and R. M. W. Dixon (eds), 195–224. Oxford: Oxford University Press. Harris, A. C. and Campbell, L. 1995. Historical Syntax in Cross-linguistic Perspective. Cambridge: Cambridge University Press.

162

Bernd Heine

Haspelmath, M. 1998. How young is Standard Average European? Language Sciences 20.3: 271–287. Haspelmath, M. 1999. Why is grammaticalization irreversible? Linguistics 37.6: 1043–1068. Haspelmath, M. 2001. The European linguistic area: Standard Average European. In Language Typology and Language Universals: an International Handbook, M. Haspelmath et al. (eds), 1492–1510. New York: Walter de Gruyter. Haspelmath, M., Ko¨nig, E., Oesterreicher, W., and Raible, W. (eds). 2001. Language Typology and Language Universals: an International Handbook, Volume 2 (Handbu¨cher zur Sprach- und Kommunikationswissenschaft 20.2), New York: Walter de Gruyter. Hawkins, J. 2004. E‰ciency and Complexity in Grammars. Oxford: Oxford University Press. Heine, B. 1997a. Possession: Sources, Forces and Grammaticalization. Cambridge: Cambridge University Press. Heine, B. 1997b. Cognitive Foundations of Grammar. Oxford: Oxford University Press. Heine, B., Claudi, U., and Hu¨nnemeyer, F. 1991. Grammaticalization: a Conceptual Framework. Chicago: University of Chicago Press. Heine, B. and Kuteva, T. 2002. World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Heine, B. and Kuteva, T. 2003. Contact-induced grammaticalization. Studies in Language 27: 529–572. Heine, B. and Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Heine, B. and Kuteva, T. 2006. The Changing Languages of Europe. Oxford: Oxford University Press. Heine, B. and Miyashita, H. 2008. Accounting for a functional category: German drohen ‘to threaten’. Language Sciences 30: 53–101. Hickey, R. (ed.). 2003. Motives for Language Change. Cambridge: Cambridge University Press. Hill, J. H. and Hill, K. C. 2004. Word order type change and the penetration of Spanish de in modern Nahuatl. Sprachtypologie und Universalienforschung 57: 23–48. Hopper, P. J. and Traugott, E. C. 2003. Grammaticalization. Cambridge: Cambridge University Press. Hurch, B. 1989. Hispanisierung im Baskischen. In Beitra¨ge zum 5. Essener Kolloquium u¨ber ‘‘Grammatikalisierung: Natu¨rlichkeit und Systemo¨konomie’’, N. Boretzky et al. (eds), 11–35. Bochum: Brockmeyer. Jastrow, O. 2005. Uzbekistan Arbic: A language created by Semitic-IranianTurkic linguistic convergence. In Linguistic Convergence and Areal Di¤usion: Case Studies from Iranian, Semitic and Turkic, E. A. Csato´ et al. (eds), 133– 139. London/New York: Routledge Curzon.

On polysemy copying and grammaticalization in language contact

163

Johanson, L. 1992. Strukturelle Faktoren in tu¨rkischen Sprachkontakten. (Sitzungsberichte der Wissenschaftlichen Gesellschaft an der Johann Wolfgang GoetheUniversita¨t Frankfurt am Main, 29.5, 169–299.) Stuttgart: Franz Steiner. Johanson, L. 2002. Structural Factors in Turkic Language Contacts. London: Curzon. Jones, M. C. 2002. Mette a haout dauve la grippe des Angllais: convergence on the island of Guernsey. In Language Change: the Interplay of Internal, External and Extra-linguistic Factors, M. C. Jones and E. Esch (eds), 143–168. Berlin: Mouton de Gruyter. Jones, M. C. and Esch. E. (eds). 2002. Language Change: the Interplay of Internal, External and Extra-linguistic Factors. (Contributions to the Sociology of Language, 86.) Berlin: Mouton de Gruyter. Kahr, J. C. 1976. The renewal of case morphology: sources and constraints. Working Papers on Language Universals (Stanford) 20: 107–151. Karttunen, F. 1976. Uto-Aztecan and Spanish-type dependent clauses in Nahuatl. (Papers from the Parasession on Diachronic Syntax.) Chicago Linguistic Society, 150–158. Keesing, R. M. 1988. Melanesian Pidgin and the Oceanic substrate. Stanford: Stanford University Press. Keesing, R. M. 1991. Substrates, calquing and grammaticalization in Melanesian Pidgin. In Approaches to Grammaticalization, E. C. Traugott and B. Heine (eds), 315–342. Amsterdam: John Benjamins. Kempgen, S., Schweier, U., and Berger, T. (eds). 2003. Rusistika, Slavistika, Lingvistika: Festschrift fu¨r Werner Lehfeldt. (Die Welt der Slaven, 19.) Munich: Sagner. King, R. 2000. The Lexical Basis of Grammatical Borrowing: a Prince Edward Island French case study. Amsterdam: John Benjamins. Kiral, F. 2005. Modal constructions in Turkic of Iran. In Linguistic Convergence and Areal Di¤usion: Case Studies from Iranian, Semitic and Turkic, E. A. Csato´ et al. (eds), 285–293. London/New York: Routledge Curzon. Krause, M. and Sappok, C. (eds). 2004. Slavistische Linguistik 2002. Munich: Sagner. Kroskrity, P. 1993 [1978]. Aspects of syntactic and semantic variation within the Arizona Tewa speech community. Anthropological Linguistics 35: 250–273. Kuße, H. (ed.). 2003. Slavistische Linguistik 2001. Munich: Sagner. Kuteva, T. and Heine, B. 2004. On the possessive perfect in North Russian. Word 55.1: 37–71. Lass, R. 1997. Historical Linguistics and Language Change. Cambridge: Cambridge University Press. Leisio¨, L. 2000. The word order in genitive constructions in a diaspora Russian. International Journal of Bilingualism 4: 301–325. Leslau, W. 1945. The influence of Cushitic on the Semitic languages of Ethiopia. A problem of substratum. Word 1.1: 59–82.

164

Bernd Heine

Leslau, W. 1952. The influence of Sidamo on the Ethiopic languages of Gurage. Language 28.1: 63–81. Li, C. N. 1983. Language contact in western China. Papers in East Asian Languages 1: 31–51. Li, C. N. 1984. From verb-medial analytic language to verb-final synthetic language: a case of typological change. Berkeley Linguistics Society 10: 307–323. Lindstro¨m, L. and Tragel, I. 2007. Agent, patient and possessor in the Estonian participle construction [ARGADE olema ‘be’ (P) VPAS-PST-PTCL ]. Paper presented at the Spring School on Grammaticalization and Typology, Rakvere, Estonia, 21–24 March, 2007. Matras, Y. 1998. Convergent development, grammaticalization, and the problem of ‘‘mutual isomorphism.’’ In Sprache in Raum und Zeit: In memoriam Johannes Bechert. Volume 2. Beitra¨ge zur empirischen Sprachwissenschaft, W. Boeder et al. (eds), 89–103. Tu¨bingen: Gunter Narr. Matras, Y. and Sakel, J. 2007. Investigating the mechanisms of pattern replication in language convergence. Studies in Language 31.4: 829–865. Menz, A. 1999. Gagausische Syntax: eine Studie zum kontaktinduzierten Sprachwandel. Wiesbaden: Harrassowitz. Me´trich, R., Hudlett, A., and Lu¨ger, H. H. (eds). 1999. Des racines et des ailes: the´ories, mode`les, expe´riences en linguistique et en didactique. Nancy: A.N.C.A. Miller, R. A. 1967. The Japanese Language. Chicago: University of Chicago Press. Milroy, J. 2003. On the role of the speaker in language change. In Motives for Language Change, R. Hickey (ed.), 143–157. Cambridge: Cambridge University Press. Milroy, L. and Milroy, J. 1985. Linguistic change, social network and speaker innovation. Journal of Linguistics 21: 339–384. Mougeon, R. and Beniak, E. 1991. Linguistic Consequences of Language Contact and Restriction: The Case of French in Ontario, Canada. Oxford: Clarendon Press. Nelde, P. H. (ed.). 1980. Sprachkontakt und Sprachkonflikt. (Zeitschrift fu¨r Dialektologie, Beihefte, 32.) Wiesbaden: Franz Steiner. Pietsch, L. 2004. Re-inventing the ‘perfect’ wheel: grammaticalization and the Irish English medial-object perfects. Arbeiten zur Mehrsprachigkeit 65. Hamburg: University of Hamburg. Prince, E. F. 1998. The borrowing of meaning as a cause of internal syntactic change. In Historical Linguistics 1997, M. S. Schmid et al. (eds), 339–362. Amsterdam: John Benjamins. Ramisch, H. 1989. The Variation of English in Guernsey/Channel Islands. (Bamberger Beitra¨ge zur Englischen Sprachwissenschaft, 24). Frankfurt: Peter Lang. Ratcli¤e, R. R. 2005. Bukhara Arabic: a metatypized dialect of Arabic in Central Asia. In Linguistic Convergence and Areal Di¤usion: Case Studies from Iranian, Semitic and Turkic, E. A. Csato´ et al. (eds), 141–159. London/New York: RoutledgeCurzon.

On polysemy copying and grammaticalization in language contact

165

Rayfield, J. R. 1970. The Language of a Bilingual Community. (Janua Linguarum, Series Practica 77.) The Hague: Mouton. Reuther, T. (ed.). 1992. Slavistische Linguistik 1991. Munich: Sagner. Ross, M. D. 1996. Contact-induced change and the comparative method: cases from Papua New Guinea. In The Comparative Method Reviewed: Regularity and Irregularity in Language Change, M. Durie and M. Ross (eds), 180–217. New York: Oxford University Press. Ross, M. 2001. Contact-induced change in Oceanic languages in North-West Melanesia. In Areal Di¤usion and Genetic Inheritance: Problems in Comparative Linguistics, A. Y. Aikhenvald and R. M. W. Dixon (eds), 134–166. Oxford: Oxford University Press. Ross, M. 2003. Diagnosing prehistoric language contact. In Motives for Language Change, R. Hickey (ed.), 174–198. Cambridge: Cambridge University Press. Ross, M. 2006. Calquing and metatypy. Journal of Language Contact 1. http:// cgi.server.uni-frankfurt.de/fb09/ifas/JLCCMS/ Sarhimaa, A. 1999. Syntactic transfer, contact-induced change, and the evolution of bilingual mixed codes: focus on Karelian-Russian language alternation. (Studia Fennica, Linguistica, 9) Helsinki: Finnish Literature Society. Schmid, M. S., Austin, J. R., and Stein, D. (eds). 1998. Historical Linguistics 1997: Selected papers from the 13th International Conference on Historical Linguistics, Du¨sseldorf, 10–17 August, 1997. (Amsterdam Studies in the Theory and History of Linguistic Science, 164) Amsterdam: John Benjamins. Silva-Corvala´n, C. 2007. The limits of convergence in language contact. Paper presented at the symposium ‘‘Language Contact and the Dynamics of Language: Theory and Implications,’’ Max Planck Institute for Evolutionary Anthropology, Leipzig, 10–13 May, 2007. Stolz, T. 1996. Some instruments are really good companions – some are not: on syncretism and the typology of instrumentals and comitatives. Theoretical Linguistics 23.1–2: 113–200. Sussex, R. 1993. Slavonic languages in emigration. In The Slavonic Languages, B. Comrie and G. G. Corbett (eds), 999–1035. London/New York: Routledge. Taeldeman, J. 1978. Franzo¨sisch-fla¨mische Sprachinterferenz in Flandern. In Sprachkontakte im Nordseegebiet: Akten des 1. Symposions u¨ber Sprachkontakt in Europa, Mannheim 1977, P.S. Ureland (ed.), 43–66. Tu¨bingen: Max Niemeyer. ¨ berlegungen zu Ternes, E. 1999. Ist Bretonisch SVO oder VSO? Typologische U einer umstrittenen Frage. In Akten des Zweiten Deutschen Keltologen-Symposiums, S. Zimmer et al. (eds), 236–253. Tu¨bingen: Niemeyer. Thomason, S. G. 2001a. Language Contact: an Introduction. Washington, DC: Georgetown University Press. Thomason, S. G. 2001b. Language Contact. Edinburgh: Edinburgh University Press. Thomason, S. G. 2007. Social and linguistic factors as predictors of contactinduced change. Paper presented at the symposium ‘‘Language Contact and

166

Bernd Heine

the Dynamics of Language: Theory and Implications,’’ Max Planck Institute for Evolutionary Anthropology, Leipzig, 10–13 May, 2007. Tosco, M. 2000. Is there an ‘‘Ethiopian language area’’? Anthropological Linguistics 42.3: 329–365. Trask, R. L. 1998. The typological position of Basque: then and now. Language Sciences 20.3: 313–324. Traugott, E. C. and Heine, B. (eds). 1991. Approaches to Grammaticalization. Volume 1. Amsterdam: John Benjamins. Ureland, P. S. (ed.). 1978. Sprachkontakte im Nordseegebiet: Akten des 1. Symposions u¨ber Sprachkontakt in Europa, Mannheim 1977. (Linguistische Arbeiten, 66). Tu¨bingen: Max Niemeyer. Verschik, A. 2004. Estonian compound nouns and their equivalents in the local variety of Russian. Scando-Slavica 50: 93–109. Vincent, N. 1982. The development of the auxiliaries habere and esse in Romance. In Studies in the Romance Verbs, N. Vincent and M. Harris (eds), 71–96. London: Croom Helm. Vincent, N. and Harris, M. (eds). 1982. Studies in the Romance Verbs. London: Croom Helm. Wehr, B. 1984. Diskurs-Strategien im Romanischen: ein Beitrag zur romanischen Syntax. Tu¨bingen: Narr. Zavala, R. 2002. Calcos sinta´cticos en algunos complejos verbales Mayas y MixeZoques. Pueblos y Fronteras 4: 169–187. Zimmer, S., Ko¨dderitzsch R., and Wigger, A. (eds). 1999. Akten des Zweiten Deutschen Keltologen-Symposiums. Tu¨bingen: Niemeyer.

The attraction of indefinite articles: on the borrowing of Spanish un in Chamorro Thomas Stolz 1. Introduction1 The overt marking of definiteness and/or indefiniteness in noun phrases is a relatively widespread phenomenon among the languages of the world. In the World atlas of language structures (Haspelmath et al. 2005), Dryer (2005a–b) provides two chapters dedicated to definite articles and indefinite articles respectively. The focus of these chapters is on (in)definiteness marking on nouns and thus does not include other marking strategies in noun phrases that do not directly a¤ect the head noun.2 In his surveys, Dryer (2005a: 154) starts from the functions fulfilled by the expressions; this practice justifies the lumping together of bound morphological markers 1. I am grateful to my French colleagues for inviting me to contribute to this collection of articles. I am also indebted to my discussants on the occasion of the workshop on Spanish contact influence on Amerindian languages (Amsterdam, October 4 2008). Jorge Go´mez-Rendo´n, Jeanette Sakel, and Otto Zwartjes deserve special mention for their thought-provoking comments on my talk, on which this article is based. Rafael Rodrı´guez-Ponga kindly drew my attention to some properties of the indefinite article in peninsular Spanish and to the possibility that un in modern Chamorro shares more traits with its Spanish ancestor than I originally assumed. I am also grateful to my anonymous reviewer who made a number of insightful comments on the first draft of this paper. Barbara Dewein kindly provided me with much needed reading matter. As always, I assume the full responsibility for everything that is said in my contribution. 2. In the two Baltic languages and some of the members of the South Slavonic sub-phylum, definiteness is exclusively marked in complex NPs that contain attributes such as adjectives, as in Latvian vec-s vır-s # {old}-{nom.sg.m} {man}-{nom.sg.m} ‘an old man’ vs. vec-ai-s vır-s # {old}-{def}-{nom.sg.m} {man}-{nom.sg.m} ‘the old man’ where the definiteness is marked overtly on the adjective. In the absence of an attribute, however, the NP is ambiguous as to definiteness. Unsurprisingly, other seemingly indirect strategies of definiteness ¨ lku¨ televizyon marking such as the definite object marking in Turkish (e.g. U

168

Thomas Stolz

and free or clitic-like articles. Typical representatives of these categories are given under (1)–(2).3 (1)

Articles

(1.1) Definite article: Hungarian [LPP Hungarian, 9] A ko¨nyvben ez a´llt def.art book:iness dem.dis assume:past ‘In the book, this was stated: . . .’ (1.2) Indefinite article: Turkish [LPP Turkish, 12] Bana bir koyun c¸iz I:dat indef.art sheep draw ‘Draw me a sheep!’ (2)

Bound markers

(2.1) Definiteness marker: Swedish [LPP Swedish, 9] bild-en fo¨resta¨llde en boaorm picture-def.ut represent:imperf indef.art.ut boa ‘The picture represented a boa snake.’ (2.2) Indefiniteness marker: Kurdish (Wurzel 1997: 34) Ez kec¸ık-ek-eˆ dıbinım I girl-indef-obl.f see:1sg ‘I see a girl.’ ¨ lku¨ has bought a TV. – Where has he aldı. – Televizyon-u nereden aldı? ‘U bought the TV?’ with -u marking the definite direct object [Ersen-Rasch 1980: 77]) are not taken into account in Dryer’s (2005a) survey. Thus, the topic of definiteness vs. indefiniteness in language is much more complex than the extant typologies suggest (Lyons 1999). 3. In the examples, I use boldface to highlight articles and occasionally also additional elements further discussed in the ensuing paragraphs. Morpheme boundaries are marked by hyphens only where I deem it indispensable for the understanding of the data. Everywhere else I leave morpheme boundaries unspecified and separate morphemes by in the glosses. Except otherwise stated, all translations are mine. Owing to the various spelling reforms and individual orthographies used by particular authors, I keep the conventions of my Chamorro sources, as any attempt at unification would be too timeconsuming.

The attraction of indefinite articles

169

In (1), the free pre-nominal morphemes Hungarian a ‘the’ and Turkish bir ‘a’ are employed to mark definiteness and indefiniteness, respectively, whereas the definiteness marker Swedish -en ‘the’ and the indefiniteness marker Kurdish -ek ‘a’ in (2) are su‰xes on their nominal hosts. Thus, di¤erent morpheme classes can fulfil similar functions when it comes to marking (in)definiteness. The definitions of the notions of definiteness and indefiniteness employed by Dryer simplify the complex situation in this functional area.4 However, they are handy for a first approach. The term ‘‘definite’’ requires that the noun that is characterized as definite refers to an entity known to the speaker and hearer either because of prior introduction into the discourse or because of common knowledge. Accordingly, the Hungarian a ko¨nyvben ‘in the book’ in (1.1) and the Swedish bilden ‘the picture’ in (2.1) are anaphorically related to the antecedents Hungarian egy ko¨nyvben ‘in a book’ and Swedish i en bok ‘in a book’ in the initial sentence of the same paragraph. In contrast, the indefinite nouns in (1.2) and (2.2) are newly introduced participants that have no antecedents and thus refer to hitherto unknown entities. Disregarding the slightly di¤erent sizes of Dryer’s samples (566 languages checked for definiteness vs. 473 languages checked for indefiniteness), it is clear from his statistics that it is relatively common for human languages to mark definiteness in their noun phrases: 60 percent of the languages of the survey attest definiteness marking either via proper articles (¼75 percent) or by bound morphemes (25 percent). In contrast, indefiniteness is less often made explicit. About 43 percent of the sample languages employ overt strategies to mark indefiniteness. In the bulk of the world’s languages, bare nouns tend to have an indefinite interpretation, as in (3). (3) Maltese [LPP Maltese, 1] Din kienet turi serp boa qieg¡ed jibla’ c˙erv dem.prox.f be:3sg.f 3sg.f:show boa prog 3sg.m:swallow stag ‘This [¼the picture] showed a boa that swallowed a stag.’ In the Maltese example, the new participants serp boa ‘boa’ and c˙erv ‘stag’ are bare nouns and thus indefinite.5 4. According to Lyons (1999: 157–198) various kinds of (in)definiteness have to be distinguished (specificity, referentiality, generic). Heine (1997: 70) mentions five di¤erent stages relevant to the grammatical development of articles, see Section 2 below. 5. In languages that mark neither definiteness nor indefiniteness overtly, bare nouns are ambiguous. The Finnish clause kirjassa sanottiin ‘it was said in

170

Thomas Stolz

Only 11 percent of the languages with indefiniteness marking achieve this goal via a‰xation of an indefiniteness marker. Some 88 percent concur with Turkish in (1.2) in so far as they display proper indefinite articles qua free morphemes. There are about twice as many languages that only mark definiteness (81P14 percent) as there are languages that exclusively mark indefiniteness (41 ¼ 7 percent). All these statistical facts are suggestive of the higher degree of markedness of indefiniteness marking as opposed to the relatively unmarked status of definiteness marking. Dryer’s (2005b: 160–161) map also shows that there are certain hotbeds of indefiniteness marking on the globe, albeit less clearly delimited than in other cases of areal clustering. Indefinite articles abound especially in the better part of Europe, the Near and Middle East, Papua-New Guinea, Mesoamerica, and West Africa, with additional zones of accumulation on the northern border of India as well as stretches of East Africa and Austronesia. In point of fact, indefinite articles occur on each continent. However, in some areas (such as South America and Australia) indefiniteness marking is clearly a minority solution, and throughout Northern Asia the phenomenon is absolutely unknown. In genealogical terms, indefinite articles are not monopolized. Indo-European, Uralic, Austronesian, and other macro-phyla as well are divided as to the marking of indefiniteness. Given that the geographic distribution of a typologically marked phenomenon is uneven, one gets to thinking about the origin of some reported indefinite articles in languages whose next of kin or immediate neighbors do not boast this category. According to Dryer’s map, there are several instances of genetically and/or areally unexpected attestations of indefinite articles in various parts of the world. It is always possible that, at least in some of these cases, indefiniteness marking has di¤used via language contact. In this contribution, I exclusively look at indefinite articles.6 Thus I survey the emergence/replication of the relatively marked category of indefinite articles in situations of language contact (Section 2). A more detailed case study is devoted to the borrowing of Spanish un in the Austronesian language Chamorro (Section 3). The conclusions in Section 4 assign the phenomenon under scrutiny a place in the catalogue of contact-induced structural changes. a/the book’ is a case in point, although word order and, for full-blown (internal) arguments of verbs, also case-marking are ways to mark indefiniteness or definiteness indirectly in Finnish. 6. Since my examples illustrate only articles as such, I henceforth discontinue the use of the abbreviation art in the morpheme glosses: def and indef su‰ce.

The attraction of indefinite articles

171

2. Indefinite articles and language contact Building on prior work by Haspelmath (2001: 1495), the areal spread of indefinite articles throughout Europe is a major issue in Heine and Kuteva (2006: 119–133). These authors observe that especially (but by no means only) in non-standard varieties of languages spoken in eastern Europe (plus Basque and Breton in the west), the cardinal numeral one has acquired properties associated with indefinite articles. This development from one > indefinite article is the most common grammaticalization path of indefinite articles (Givo´n 1981). Among the prerequisites for one to grammaticalize in this way is the use of the erstwhile cardinal in contexts where the exact quantification of the referent of the accompanying noun is irrelevant or backgrounded. In these contexts, one would no longer serve the purpose of quantifying the noun referent. Heine and Kuteva (2006) report numerous cases of incipient and at times also further advanced grammaticalization of one according to the stage it has reached on the ‘‘five grades’’ scale of Heine (1997). The stages come in the following order: Stage 1: Stage 2: Stage 3: Stage 4: Stage 5:

purely numerical value of ONE presentative marker specific-indefinite marker non-specific indefinite marker generalized article

B grammaticalized þ

The map provided by Heine and Kuteva (2006: 133)7 is suggestive of a kind of wave-like di¤usion from a partly SAE-borne center (mostly Romance and Germanic) to the outskirts of the continent (Heine and Kuteva 2006: 120).8 The further away from the center a language is located, the lower the stage its one-morpheme has reached on the scale of grammaticaliza7. In Stolz (2005, 2006), I provide maps that capture the distribution of various definiteness marking strategies in Europe. These maps too show an areally skewed distribution that is largely compatible with the findings by Heine and Kuteva (2006). 8. Incidentally, in the softcover edition of Heine and Kuteva (2006) I have consulted, the maps 3.1 for definite articles and 3.2 for indefinite articles in Europe are absolutely identical as far as the hatching of the individual languages goes. Judging from the accompanying discussion of data by Heine and Kuteva (2006), I doubt that this can be correct. On map 3.2, for instance, Czech is presented as a language on whose article-like use of one there is no

172

Thomas Stolz

tion towards an indefinite article. Stage 1 is typical of the outer periphery in the east, north, and northwest, and the authors emphasize in their conclusions that outside the center of di¤usion, the articles – be they definite or indefinite – have not advanced too far on the scale yet. Most often, the best examples of the rise of definite and indefinite articles can be found in non-standard varieties. However, it is also possible to find examples of newly developed indefinite articles that are tolerated by the norm of the languages, although the origin of the indefinite article is presumably via language contact. Haase (1992: 59–61) assumes that Basque bat ‘one’ has become the ubiquitous indefinite article under the influence of the Romance neighbors of Basque (Heine and Kuteva 2006: 132). Heine and Kuteva (2006: 131–132) argue that the Breton indefinite article un ‘a’ has been remodeled after the French pendent un(e) ‘a.’ It is also very likely that Hungarian egy ‘one’ developed in a parallel fashion because of the contact with German (and partly also with Romance) (Ba´rczi 2001: 184–185). In both cases, we know from the history of the two non-Indo-European languages that on their earliest documented stages, Hungarian and Basque did not employ indefinite articles. The first attestations in Hungarian date back to the sixteenth century, while the first half of the eighteenth century gives testimony of the earliest uses of Basque bat as an indefinite article. Moreover, the closest relatives of Hungarian are not equipped with indefinite articles even today. Both Basque and Hungarian also display definite articles (or rather definiteness markers in the case of Basque: [Heine and Kuteva 2006: 30–1]). For Breton, the historical evidence is less compelling, as the indefinite article is attested already in Middle Breton (Lewis and Piette 1966: 10). What makes these cases especially intriguing from an areallinguistic point of view is the fact that it is next to impossible to pinpoint language contact as the instigator of the grammaticalization processes. information available, whereas in the main body of the text it is depicted as a language whose term for one has reached stage 2 (perhaps even stage 3) on the grammaticalization scale (Heine and Kuteva 2006: 123–124). Since map 3.1 likewise suggests that there are no data confirming or disconfirming the presence of a definite article in Czech, while at the same time the Czech evidence is discussed at length on the previous pages (Heine and Kuteva 2006: 114–115), I assume that something went wrong when the maps were printed. This skepticism is further supported by the fact that Icelandic is correctly located on stage 1 as to the grammaticalization of one on map 3.2. However, it is assigned the same place with reference to the definite article on map 3.1, which is untenable as it must have reached stage 4.

The attraction of indefinite articles

173

This is precisely the argument of Heine and Kuteva (2006), who claim that only those grammaticalization processes can be triggered via language contact that would also be possible without an external stimulus. That is, what happened in Basque and Hungarian may or may not have been the result of influence exerted by neighboring languages. Of course, the likelihood that we are facing contact-induced language change is rather high in both these and other instances discussed by Heine and Kuteva (2006). However, as the assumed replications only involve pattern and not matter, according to the terminology introduced by Sakel (2007),9 absolute certainty about the contact origin of the phenomenon can never be obtained. To underline this problem, a brief discussion of data from insular North Germanic is in order. Icelandic and Faroese are representatives of stage 1 on map 3.2 in Heine and Kuteva (2006: 133), that is, their expression for one has not advanced on the grammaticalization scale as it has purely quantifying functions. This classification holds largely for Icelandic whereas it is certainly wrong for Faroese, cf. (4)–(5). (4) Icelandic [LPP Icelandic, 8] Hu´n ta´knaði kyrkislo¨ngu sem var að She picture:pret:3sg boa:acc rel be.pret.3sg to ´ melta fıl digest:inf elephant ‘It [¼the picture] showed a boa that was digesting an elephant.’ (5) Faroese [LPP Faroese, 10] Eg hevði teknað ein-a kvalarslangu I have:pret:1sg draw:pp indef.art-acc boa:acc sum var um at sodna ein elefant rel be.pret.3sg about to digest:inf indef.art elephant ‘I had drawn a boa that was about to digest an elephant.’ Icelandic and Faroese are close relatives. Nevertheless, their article systems di¤er considerably. Where Icelandic lacks an overt marker of indefiniteness (as the bare nouns kyrkislo¨ngu ‘[a] boa’ and fil ‘[an] elephant’ show), Faroese obligatorily employs a full-blown indefinite article: eina kvalarslangu ‘a boa’ and ein elefant ‘an elephant.’ The Faroese indefinite article 9. This also applies to Breton no matter how closely Breton un and French un(e) resemble each other: usually, it is assumed that Breton un originates from the segmental erosion of the cardinal numeral unan ‘one’ (Lewis and Piette 1966: 21), that is, Breton un is not a direct loan from French.

174

Thomas Stolz

einn ‘a’ is phonologically identical to the cardinal numeral einn ‘one’ with which it shares the ability to inflect for gender, case, and number. The latter ability to have plural forms qualifies the Faroese indefinite article for stage 5 on the grammaticalization scale (Heine and Kuteva 2006: 105). Thra´insson et al. (2004: 91–2) describe the occurrence of the plural forms of the Faroese indefinite article and state that its use is restricted to combinations with pluralia tantum and ‘‘to indicate a pair of something,’’ cf. einir sko´gvar ‘a pair of shoes’ vs. sko´gvar ‘shoes’ where einir is the nominative plural masculine of the indefinite article. Icelandic (Kress 1982: 100) makes similar use of the plural forms of the numeral einn ‘one’ to indicate sets of objects represented by pluralia tantum. Icelandic clearly reflects the historically older stage as there is no trace of indefinite articles in the earliest (mediaeval) sources of Old Icelandic/ Old Norse (Braunmu¨ller 1991). All the mainland Scandinavian languages developed indefinite articles whereas Icelandic remained true to its old solution without an indefinite article. For many centuries, Faroese has been under linguistic pressure from Norwegian and especially from Danish. In the absence of extensive textual documentation of stages of Faroese prior to the early nineteenth century, we do not know when exactly the development one > indefinite article commenced and how fast it proceeded (Thra´insson et al. 2004: 370–372). Thus, it cannot be decided conclusively whether the indefinite article in Faroese arose as a replication of its Danish/Norwegian equivalents en/ett or on an independent (languageinternal) basis. If intra-Scandinavian language contacts are responsible for what happened in Faroese, we are nevertheless dealing with pattern replication. Trivially, the nature of pattern replication is such that there is always at least a slim chance of independent parallel development. Thus, to get a better understanding of the behavioral patterns of indefinite articles in language contact situations, it is important to investigate instances of matter replication or overt borrowing. Section 3 addresses the particularly interesting case of the Spanish indefinite article un and its fate as a grammatical borrowing in Chamorro.

3. Spanish un in Chamorro Chamorro is the indigenous Austronesian language of the Marianas Islands on the western rim of the Pacific. Starting with Magellan’s visit to the islands in 1521, Chamorro remained in contact with Spanish until the end of the nineteenth century. For 230 years (1668–1898/9), the language contact was especially intensive in the then colony of Spain. Direct Spanish

The attraction of indefinite articles

175

influence came to a sudden halt at the turn of the twentieth century. However, modern Chamorro still carries the marks of Hispanization.10 According to some estimates (Rodrı´guez-Ponga 1995), the share of Hispanic elements in the lexicon of present-day Chamorro amounts to approximately 60 percent. Apart from purely lexical Hispanisms,11 there are scores of function words with a Spanish etymology (discourse particles, conjunctions, adpositions, modal verbs: Stolz and Stolz 1997). Furthermore, the comparative construction is partially Hispanized (Stolz and Stolz 2001). For Spanish-derived adjectives, we also find occasional evidence of gender agreement (with human nouns: Stolz 1998, 2002). Sentence (6) from contemporary Chamorro prose illustrates the high incidence of Hispanisms in the language. To facilitate recognition, Hispanisms appear in boldface in this example. (6) Chamorro [Hinengge 27] Guaha unu ni’ gaige gi hiyong i sengsong exi one rel be in outside def village gi sentra˚t na pa˚ tte-n i isla in central link part-link def island taiguini puesto-n˜a ya put i kla˚se-n familia because place-por.3 and for def class-link family ni’ man˜a˚ saga guihi rel subj.pl:red:dwell there kontiempo na humuyong este na estoria. prior_to link af:go_outside this link story ‘There is one [¼house] situated outside the village in the central part of the island [which is special] because of its location and because of the kind of families who were living there before this story started.’ 10. Note that the general typological make-up of Chamorro has remained remarkably una¤ected by Spanish influence (Pagel 2010). It is still a split-ergative language with VSO/SVO word-order, and its rich morphology allows for all kinds of a‰xation practices, including reduplication (Stolz 2003). For a formalist appraisal of Chamorro syntax, I refer the reader to Chung (1998). 11. The massive borrowing from Spanish has resulted in the creation of numerous synonym pairs consisting of an Austronesian lexeme and a semantically identical Spanish-derived equivalent. Modern speakers of Chamorro are not always aware of the non-Austronesian origin of many of their lexical items (Salas Palomo and Stolz 2008).

176

Thomas Stolz

There are altogether eleven Hispanic elements (types and tokens) in the sentence, that is, slightly more than 30 percent of all words. Besides the usual lexical borrowings and function words,12 there is also the numeral unu ‘one’, which ultimately guides us to the discussion of the indefinite article in Chamorro. Before we enter the area of indefiniteness, however, it must be pointed out that Chamorro also employs an inherited, that is Austronesian, set of articles that remotely resemble the Philippine system of focus articles (Topping 1973: 245–253).13 The four instances of i ‘the’ in (6) illustrate the usage of the most generalized of these articles. 3.1. On and about stage 1 For general information on the indefinite article and the numeral one in contemporary Spanish and its overseas varieties, I refer the reader to the reference grammar by Bosque and Demonte (1999). For brevity’s sake, the following statements are made from the perspective of Chamorro only. In (6), Chamorro unu ‘one’ clearly reflects the pronominal use of the long form of the Spanish cardinal numeral uno ‘one.’ The distinction of long and short forms in Spanish is restricted to the masculine gender. The long form uno occurs only when used independently as head of a noun phrase, whereas the short form un has to be used if the numeral serves as a modifier of a head noun ([LPP Spanish, 62] En tu planeta los dı´as duran un minuto. ‘On your planet the days count [ just] one minute.’). A similar distinction is made in Chamorro, as there are two allomorphs of the numeral one, namely unu used pronominally or as head and un used attributively. As in Spanish, the attributive allomorph is segmentally 12. The Spanish etymologies of the borrowed items are transparent: unu < Spanish uno ‘one,’ sentra˚ t < Spanish central ‘central,’ pa˚ tte < Spanish parte ‘part,’ isla < Spanish isla ‘island,’ puesto < Spanish puesto ‘place,’ put [usually pot] < Spanish por ‘for, because of,’ kla˚ se < Spanish clase ‘class,’ familia < Spanish familia ‘family,’ kontiempo < Spanish con ‘with’ þ tiempo ‘time,’ este < Spanish este ‘this,’ estoria < Spanish historia ‘story.’ All segmental di¤erences between source language and target language reflect regular phonological correspondence laws of Spanish loans in Chamorro. 13. The articles are: proper articles si and as, toponym article iya, common articles i, ni, nu (Topping 1973: 130–136). Steve Pagel (personal communication) ponders the idea that the article i, which tends to be used widely beyond the boundaries of its supposed focus-based domain, has been influenced by Spanish el ‘the.’ However, materially, i was already attested as a ‘‘definite article’’ in the earliest sources of Chamorro in the seventeenth century.

The attraction of indefinite articles

177

identical to the indefinite article un. This is the current general consensus among the experts of Chamorro, all of whom acknowledge that Chamorro un stems from Spanish no matter whether it is used as a numeral or as an indefinite article. In Father Sanvitores’ grammar-cum-catechism of 1668 there is no trace of un, nor is there any evidence of an indefinite Austronesian article (Burrus 1954). Since it took almost 200 years before the next written documents appeared in Chamorro – among them the Spanish school grammar by Iba´n˜ez del Carmen (1865)14 – we are left completely in the dark about the intermediate processes that ultimately led to the creation of the indefinite article in Chamorro. Moreover, scholars disagree as to the extent of the domain of the indefinite article in Chamorro. The e¤ects of the Spanish-derived indefinite article on the split-ergative system of Chamorro are described in Stolz (2010). In the only modern reference grammar of Chamorro, Topping (1973: 136–138) describes the position of un in the grammatical system of Chamorro as marginal because he assumes that it is severely restricted in use, such that it ‘‘usually occur[s] in fixed idiomatic expressions’’ (Topping 1973: 136). The examples Topping (1973: 137) provides are set phrases of the type un diha ‘one day’ that consist entirely of Spanish-derived elements. In Stolz and Sabater Fuentes (2002), we demonstrate that already in the earliest translation of the New Testament (1904), un freely combines with Austronesian nouns too. The apparent preference for combinations with Spanish etyma is the incidental e¤ect of the high number of Spanish loan nouns in Chamorro. It is true that collocations like un bi’ahi ‘once’ ( ipitaneta ‘give a name’ (compare Hup hatni- [name-be] ‘give a name’ above). In Baniwa, verbalization is realized via either the prefix ka- or the su‰x -hı´ta; the latter is probably cognate with the Tariana su‰x, but is not productive and occurs with only a few roots (relating to the expression of a physiological state). The derivation of verbal ‘have-N’ constructions via a verbalizing mechanism, often one of noun incorporation, once again appears to be an areal feature of the Vaupe´s region. This is further supported by the presence of parallel, idiosyncratic lexical items resulting from this construction in many of the region’s languages. However, the ultimate source of this strategy is unclear. Comparative data suggest the ‘‘verbalizing’’ function of ni is old in the Nadahup family, and its presence in Hup cannot be easily attributed to Tukanoan contact, especially since Nade¨b is largely out of the Tukanoan sphere of influence. Yet whatever its origin, it is likely that the ni verbalization strategy in Hup has been shaped or restructured by contact with Tukanoan (especially given that influence in the opposite direction is unattested), particularly regarding the calqued lexical constructions. If this is the case, the Hup examples serve as a further indication that grammatical transfer involving ni is not precluded by its formal equivalence to a Tukanoan verb.

On form and function in language contact

219

5. Ni forms and functions: areal influence among Vaupe´s languages Table 1 summarizes the functions served by Hup ni, the presence of similar constructions in other Vaupe´s languages, and the forms that code them (which in many cases resemble ni).10 As the preceding discussion has demonstrated, the verb ni in Hup serves a wide range of functions, ranging from the lexical to the relatively grammatical. All of these occur in constructions that are widely similar across other Vaupe´s languages (some of which also employ variants of ni), and are particularly well established in Tukano and other languages of the East Tukanoan family, which have had a profound influence on Hup’s grammar in many other ways. In contrast, a number of these constructions are apparently not shared by other Nadahup languages, particularly Hup’s more distant sisters Daˆw and Nade¨b: namely, the use of ni (or comparable form) as a copula in equative clauses, as an inferred evidential, as an auxiliary verb relating to progressive aspect, and as a light verb in a predicate chain (although this last function appears to be present in Daˆw).11 It is therefore probable that the variety of functions of Hup ni have been structured at least in part according to an East Tukanoan model, as is consistent with what we already know about Tukanoan > Hup contact. We can also assume the same for a number of 10. Some classifications of the family treat Kubeo and Retuara˜ (Tanimuka) as a separate Central Tukanoan branch (for example Barnes 1999). Sources of data: Hup (my fieldnotes; Epps 2008b); Yuhup (Ospina 2002); Daˆw (Martins 2004); Nade¨b (Weir 1984, 1990, 1994); Tukano (Aikhenvald 2002, Ramirez 1997, West 1980); Desano (Miller 1999); Kotiria (Wanano; Stenzel 2004 and forthcoming); Tatuyo (Gomez-Imbert and Hugh-Jones 2000); Barasana (Jones and Jones 1991, Elsa Gomez-Imbert, personal communication); Tuyuka (Barnes and Malone 2000, Malone 1988); Pisamira (Gonza´lez de Pe´rez 2000); Karapana (Malone 1988, Metzger 2000); Siriano (Criswell and Brandrup 2000); Yuruti (Kinch and Kinch 2000); Makuna (Malone 1988, Smothermon and Smothermon 1993); Kubeo (Thiago Chacon, personal communication; Morse and Maxwell 1999); Retuara˜ (Tanimuka; Strom 1992); Orejon (Velie and Velie 1981); Koreguaje (Cook and Criswell 1993); Siona (Wheeler 1970); Tariana (Alexandra Aikhenvald, personal communication, 2002, 2003); Baniwa (Ramirez 2001a, 2001b). 11. Of course, the fact that a feature is not reported in the description does not necessarily mean that it is not present in the language. The conclusions presented here are contingent on the available information about these languages, and may require revision in the future. 12. Aloisio Cabalzar, personal communication.

220

Patience Epps

East Tukanoan

Nadahup ðMak´uÞ

Table 1. Uses of ni and its counterparts in languages of the Vaupe´s and beyond (Blanks in the table indicate the feature is not reported.) existence/ location: ‘be, exist’

equative copula

inferred evidential

auxiliary, progressive aspect

light verb in predicate chain

noun incorp./ verbalizer

Hup

ni

ni, g’˝ h

-ni verbal su‰x

ni (SVC)

ni

-ni

Yuhup

Pdi

Pdi

Pdi verbal su‰x

Daˆw

nı˜

none

nı˜ (?)

nı˜

Nade¨b

n or na: /n :

none

Tukano

niıˆ

niıˆ

niıˆ constr.

wee ‘do’ (periph.)

wee ‘do’

-ti

Desano

a´rı˜

a´rı˜

a´rı˜ constr.

ii ‘do’ (periph.)

ii ‘do’

-k

Kotiria

hi

hi

hi constr.

Pdi ‘be.prog’ (periph.)

yoa ‘do’

-ti

Tatuyo

Pa´dı´

Pa´dı´

Pa´dı´ (SVC) ‘‘durative’’ ya ‘do,’ bahi ‘be’ (periph); ya˜ ‘be’ (SVC) ‘‘durative’’

yi ‘do’

-kti

n

Barasana

ya˜

ya˜, bahi

Tuyuka

dı˜˜ı, bi 12

dı˜ ˜ı

dı˜˜ı constr.

Pisamira

Pdi

Pdi

Pdi constr.

Karapana

ani

ani

Siriano

a˜a˜rı˜

a˜a˜rı˜

Yuruti

dı˜˜ı

ja˜a˜

dı˜˜ı constr.

Makuna

n˜a, baji

ya˜

ya˜ constr.

Kubeo

k, ba

ba and copular enclitics -be, -bu

ba constr.

Retuara˜

˜ıba˜

˜ıba˜

tii ‘do’ (periph.)

ja- or iri ‘do’ (periph.) tii ‘do’ (periph.)

ba ‘be’ (periph.)

-kti

Arawak

West Tukanoan

On form and function in language contact existence/ location: ‘be, exist’

equative copula

Orejon

biayi

bai, be

Koreguaje

pa i, pani

pa i

pa i ‘be’ (periph.)

Siona

ba i

ba i

ba i ‘be’ (periph.)

Tariana

alia

alia (younger speakers)

Baniwa

ne´eni, niı´ni (loc ad)

none

inferred evidential

nhi-/nih(reanalyzed aspect marker)

auxiliary, progressive aspect

ni ‘do’ (SVC) ‘‘prolonged action’’ -nhi verbal su‰x ‘‘durative’’

light verb in predicate chain

221 noun incorp./ verbalizer

-a , -

-ni ‘do’

-ita, -ta

ka-, -hita

the Yuhup categories, although whether this is due to Tukanoan influence on the two languages independently or on their common ancestor is not known. East Tukanoan influence appears to be likewise responsible for at least some of the Tariana functions of alia ‘be, exist’ and ni ‘do’ (cf. Aikhenvald 2002).13 The available information on other Arawak languages of the region suggests that these have been less closely involved in the exchange of ‘be/do’ features; however, at least some influence of Baniwa and Yukuna (the other Arawak languages in close contact with Tukanoan varieties) is evident in the partial loss of the copula in Kubeo and Retuara˜. The form ni appears to be lacking from the West Tukanoan languages, as are many of the associated functions discussed here. However, the presence of a verb meaning ‘be, exist,’ which has additional functions as a copula in equative constructions and as an aspectual auxiliary, suggests that at least these features may be common to the Tukanoan family as a whole.

13. It is also possible that some of the parallel developments in the Vaupe´s can be attributed to common ‘‘typological poise’’ (cf. Enfield 2003: 5–6); that is, languages may undergo similar but independent developments due to parallel previous developments (which may have involved contact). However, given that most of the Vaupe´s languages in question are currently in contact, the question of independent development may be essentially moot.

222

Patience Epps

That similar constellations of functions for the verb ‘be, exist’ (and, to a lesser extent, ‘do’) are found across a range of unrelated languages is reminiscent of other contact situations. One relevant example is the polyfunctionality of the verb ‘acquire’ in languages of mainland Southeast Asia, as described by Enfield (2003); however, while these languages share the complex range of functions performed by the ‘acquire’ verb (such as a variety of modal/aspectual markers and a marker of descriptive complement constructions), these are in general carried out by distinct etymons. Such a copying of function independently of phonological form is certainly well represented in the Vaupe´s as well, and is consistent with speakers’ avoidance of shared forms generally. What is remarkable about the case of ni, however, is that many functions of the same form are shared from language to language, in spite of the sociolinguistic pressure to avoid formal overlap. In some cases it is probable that shared form has even facilitated the transfer of the new grammatical pattern. In still other cases, particularly in Hup and Tariana, additional uses of ni have apparently been developed or restructured through contact, unimpeded by any perception of it as a shared form. While all of the functions of Hup ni appear to be areal features of the Vaupe´s, and undoubtedly owe at least some of their attributes to influence from East Tukanoan, there are nevertheless some intriguing complications to this picture. In particular, two functions of ni appear to be common to all the Nadahup languages, and therefore cannot be easily attributed to Tukanoan influence: the basic verbal value ‘be, exist,’ and the verbal derivational function. Both of these linguistic features are likewise widespread among the East Tukanoan languages and exist in Arawak Tariana, although the formal means of realizing them are more variable (in particular, the Tukanoan languages and Tariana all use verbalizers unrelated to ni). Whatever their origin, the close functional parallels among these features across the Vaupe´s languages suggest that areal di¤usion has continued to shape them over time. It is likewise di‰cult to pin down the origin of the form ni itself. That both the phonological form of this etymon and its basic lexical function ‘be, exist’ in Hup, Daˆw, and Yuhup closely resemble the corresponding etymons found in many Eastern Tukanoan languages would seem to suggest that the form itself was borrowed among these languages, or between their common ancestors. The phonologically and semantically similar verb n/n :/na: in Nade¨b cannot at this point be simply dismissed as noncognate with the forms in Hup, Yuhup, and Daˆw; similarly, a clearer

On form and function in language contact

223

picture of the history of Tariana alia and ni must await further comparative work within Arawak. In sum, there seem to be at least two strata of ni phenomena in the Vaupe´s languages, particularly in Nadahup and in Tariana. The presence of a basic verb ‘be, exist’ having a form akin to ni appears to be very old. If ni was present in Proto-Nadahup, then independent innovation or ancient areal contact seem to be the most likely explanations for its common presence in Nadahup and East Tukanoan, given that di¤usion from Tukanoan into Nade¨b or into Proto-Nadahup is otherwise unattested, as is di¤usion from Nadahup into Tukanoan or Arawak. Similarly, the verbal derivational function of ni in Nadahup also appears to be old; more information about the phenomenon in Nade¨b will help to determine whether the parallels with the verbalizer -ti in Tukanoan are due to more than a later restructuring of the Hup, Yuhup, and Daˆw counterpart through contact. In contrast, the functions of ni as a copula in equative clauses, an inferred evidential, an aspectual auxiliary, and a light verb in a predicate chaining construction all appear to be relatively recent in Hup, and more easily attributed to East Tukanoan influence; the same certainly applies to at least some of the Tariana counterparts of these constructions. That the lexical function of ni appears to predate most of its more grammatical functions in these languages is consistent with typological tendencies of language contact and language change: lexical material is more likely to develop grammatical functions than the reverse, and is also (cross-linguistically) more likely to be borrowed than grammatical material.

6. Conclusions In a typology of language contact situations, the Vaupe´s region is an unusual case. Local language ideologies promote widespread multilingualism but restrict language mixing, resulting in speakers’ resistance to the borrowing of linguistic elements of which they are more aware, primarily lexical and morphological forms. Yet this situation also leads to speakers’ assimilation of many elements that are less easily identified as foreign, primarily grammatical categories and structures. In this context, the overlapping constellations of similar functions and forms associated with the etymon ni in the Vaupe´s languages are remarkable. While it is not fully clear to what extent the ni forms have a common origin, their unmistakable formal similarity suggests that in many cases the matching of

224

Patience Epps

ni forms to congruent functions throughout Vaupe´s languages is not an accident, and that, at least in some instances, form itself has played a role in facilitating the spread of grammatical structures. For constructions like the inferred evidential and the aspectual auxiliary, described above, it is likely that the formal resemblance to the target construction in the contact language has allowed speakers and listeners to assemble and decode the novel native-language construction in the early stages of language change. Even where the involvement of a shared form in a given construction may not have facilitated structural borrowing, it clearly has not deterred it. Particularly in the Hup case, where ni is virtually identical to its Tukano counterpart, the existence of a shared form, and speakers’ probable awareness of this, has not prevented its expansion to additional functions, making it more ubiquitous in the language. This has occurred in spite of speakers’ general resistance to language mixing and formal borrowing. For Hup speakers, it is likely that the expanding uses of ni – unlike most potentially borrowable lexical and grammatical material – escaped censure because ni was already present in the language as a nativized (or native) lexical item, and because it was embedded within morphologically complex grammatical constructions composed of native material (compare Hup’s preference for borrowing verbs over nouns, which probably has the same motivation). Accordingly, the case of the ni etymon in Hup and other languages of the Vaupe´s indicates that even where speakers actively avoid the sharing of lexical and morphological forms, there are instances in which this resistance may lapse and similarities of form slip ‘‘under the radar.’’ When this occurs, we see that shared form may both precede and enable the di¤usion of grammatical structures among the region’s languages, rather than hindering it. Thus, while the Vaupe´s contact situation appears at first glance to be exceptional, a closer look reveals that it is still structured by many of the same mechanisms that characterize language contact cross-linguistically.

Abbreviations anph appl assert cl com conj

Anaphoric Applicative Assertion Classifier Comitative Conjunction

On form and function in language contact

coop decl dem dep desid dim dir dst:cntr dst:pst dynm emph emph.co f fact flr frust fut imp impfv inch infr ints itg lim loc m neg nmlz nonf nonvis obj obl pfv pl poss pres prog prox pst purp

Cooperative Declarative Demonstrative Dependent Desiderative Diminutive Directional Distant past contrast Distant past Dynamic Emphasis Emphatic coordinator Feminine Factitive Filler Frustrative Future Imperative Imperfective Inchoative Inferred Intensifier Intangible Limiter Locative Masculine Negative Nominalizer Non-feminine Non-visual Object Oblique Perfective Plural Possessive Present Progressive Proximate Past Purpose

225

226

Patience Epps

rec:pst rel:dem rep seq sg sim spcr tel vent vis

Recent past Relative demonstrative pronoun Reported Sequential Singular Simultaneous Spacer Telic Venitive Visual

References Aikhenvald, A. Y. 2001. Language awareness and correct speech among the Tariana of Northwest Amazonia. Anthropological Linguistics 43.4: 411–430. Aikhenvald, A. Y. 2002. Language Contact in Amazonia. Oxford: Oxford University Press. Aikhenvald, A. Y. 2003. A Grammar of Tariana. Cambridge: Cambridge University Press. Barnes, J. 1999. Tucano. In The Amazonian Languages, R. M. W. Dixon and A. Y. Aikhenvald (eds), 207–226. Cambridge: Cambridge University Press. Barnes, J. and Malone, T. 2000. El tuyuca. In Lenguas indı´genas de Colombia: una visio´n descriptiva, M. S. Gonza´lez and M. L. Rodrı´guez (eds), 437–452. Bogota´: Instituto Caro y Cuervo. Cook, D. and Criswell, L. 1993. El idioma Koreguaje (Tucano occidental). Bogota´: SIL. Criswell, L. and Brandrup, B. 2000. Un bosquejo fonolo´gico y gramatical del siriano. In Lenguas indı´genas de Colombia: una visio´n descriptiva, M. S. Gonza´lez and M. L. Rodrı´guez (eds), 395–418. Bogota´: Instituto Caro y Cuervo. Enfield, N. J. 2003. Linguistic Epidemiology: Semantics and Grammar of Language Contact in Mainland Southeast Asia. London: Routledge Curzon. Epps, P. 2005. Areal di¤usion and the development of evidentiality: evidence from Hup. Studies in Language 29.3: 617–650. Epps, P. 2007a. The Vaupe´s melting pot: Tukanoan influence on Hup. In Grammars in contact: a cross-linguistic typology, A. Y. Aikhenvald and R. M. W. Dixon (eds), 267–289. (Explorations in Linguistic Typology 4.) Oxford: Oxford University Press. Epps, P. 2007b. Birth of a noun classification system: the case of Hup. In Language endangerment and endangered languages: linguistic and anthropological studies with special emphasis on the languages and cultures of the Andean-Amazonian border area, L. Wetzels (ed.), 107–128. (Indigenous Languages of Latin America

On form and function in language contact

227

series [ILLA].) Leiden University, The Netherlands: The Research School of Asian, African, and Amerindian Studies (CNWS). Epps, P. 2008a. Grammatical borrowing in Hup. In Grammatical Borrowing: a Cross-Linguistic Survey, Y. Matras and J. Sakel (eds), 551–566. Berlin: Mouton de Gruyter. Epps, P. 2008b. A Grammar of Hup. (Mouton Grammar Library 43) Berlin: Mouton de Gruyter. Epps, P. 2009. Loanwords in Hup, a Nadahup language of Amazonia. In Loanwords in the World’s Languages: a Comparative Handbook, M. Haspelmath and U. Tadmor (eds), 992–1014. Berlin: Mouton de Gruyter. Epps, P. Forthcoming. Language and subsistence patterns in the Amazonian Vaupe´s. In The Languages of Hunter-gatherers: Global and Historical Perspectives, T. Gu¨ldemann, R. Rhodes, and P. McConvell (eds). Cambridge: Cambridge University Press. Facundes, S. da Silva. 2000. The language of the Apurina˜ people of Brazil (Maipure/ Arawak). Ph.D. thesis, State University of New York at Bu¤alo. Givo´n, T. 1979. On Understanding Grammar. New York: Academic Press. Gomes dos Santos, M. 2006. Uma grama´tica do Wapixana (Arua´k) – Aspectos da fonologia, da morfologia e da sintaxe. Ph.D. thesis, University of Campinas. Gomez-Imbert, E. 1996. When animals become ‘rounded’ and ‘feminine’: conceptual categories and linguistic classification in a multilingual setting. In Rethinking Linguistic Relativity, J. Gumperz and S. Levinson (eds), 438–469. Cambridge: Cambridge University Press. Gomez-Imbert, E. and Hugh-Jones, S. 2000. Introduccio´n al estudio de las lenguas del Piraparana´. In Lenguas indı´genas de Colombia: una visio´n descriptiva, M. S. Gonza´lez and M. L. Rodrı´guez (eds), 317–320. Bogota´: Instituto Caro y Cuervo. Gonza´lez de Pe´rez, M. S. 2000. Bases para el estudio de la lengua pisamira. In Lenguas indı´genas de Colombia: una visio´n descriptiva, M. S. Gonza´lez and M. L. Rodrı´guez (eds), 373–394. Bogota´: Instituto Caro y Cuervo. Heine, B. and Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Jackson, J. 1983. The Fish People: Linguistic Exogamy and Tukanoan Identity in Northwest Amazonia. Cambridge: Cambridge University Press. Jones, W. and Jones, P. 1991. Barasano syntax. (Studies in the languages of Colombia.) Arlington: SIL/University of Texas at Arlington. Key, M. R. 2007. Yavitero: IDS list. Available at http://lingweb.eva.mpg.de/cgibin/ids/ids.pl?com=simple_browse&lg_id=268. Kinch, R. and Kinch, P. 2000. El yuruti. In Lenguas indı´genas de Colombia: una visio´n descriptiva, M. S. Gonza´lez and M. L. Rodrı´guez (eds), 469–488. Bogota´: Instituto Caro y Cuervo. Koch-Gru¨nberg, T. 1906. Die Maku´. Anthropos 1: 877–906. Malone, T. 1988. The origin and development of Tuyuca evidentials. International Journal of American Linguistics 54: 119–140.

228

Patience Epps

Martins, S. 2004. Fonologia e grama´tica Daˆw. Ph.D. thesis, Free University, Amsterdam. Amsterdam: LOT. Martins, V. 2005. Reconstruc¸a˜o fonolo´gica do Protomaku Oriental. Ph.D. thesis, Free University, Amsterdam. Amsterdam: LOT. Metzger, R. G. 2000. Marı˜ Yaye Mena Carapana, Diccionario Carapana. Bogota´: Editorial Buena Semilla. Michael, L. 2008. Nanti evidential practice: language, knowledge, and social action in an Amazonian society. Ph.D. thesis, University of Texas at Austin. Miller, M. 1999. Desano grammar. (Studies in the Languages of Colombia 6.) Arlington: SIL/University of Texas at Arlington. Moravcsik, E. 1978. Language contact. In Universals of Human Language: Vol. 1, Method and Theory, C. Ferguson, J. Greenberg, and E. Moravcsik (eds), 93– 122. Stanford: Stanford University Press. Morse, N. and Maxwell, M. 1999. Cubeo grammar. Dallas: SIL. Mosonyi, J. C. 1987. El idioma Yavitero. PhD Dissertation. Universidad Central de Venezuela. Caracas. Ospina Bozzi, A. 2002. Les structures e´le´mentaires du Yuhup Maku, langue de l’Amazonie colombienne: morphologie et syntaxe. Ph.D. thesis, University of Paris 7 – Denis Diderot. Pustet, R. 2003. Copulas - Universals in the Categorization of the Lexicon. (Oxford Studies in Typology and Linguistic Theory.) Oxford: Oxford University Press. Ramirez, H. 1997. A fala Tukano dos Ye’pa-Masa, Vol. 1: Grama´tica. Manaus: Inspetoria Salesiana Missiona´ria da Amazoˆnia, CEDEM. Ramirez, H. 2001a. Diciona´rio da Lı´ngua Baniwa. Manaus: Editora da Universidade do Amazonas. Ramirez, H. 2001b. Lı´nguas Arawak da Amazoˆnia Setentrional: Comparac¸a˜o e Descric¸a˜o. Manaus: Editora da Universidade do Amazonas. Ramirez, H. 2001c. Famı´lia Maku´ ou famı´lia Uaupe´s-Japura? Paper presented at the meeting of ANPOLL, Bele´m, Brazil. Schauer, J. (ed.). 2005. Meke kemaka´naka pura´kaaloji: wapuraako´ chu, eya´ karı´wana chu (Diccionario bilingu¨e: Yukuna – Espan˜ol; Espan˜ol – Yukuna). 1st ed. Bogota´: Editorial Fundacio´n para el Desarollo de los Pueblos Marginados. Smothermon, J. R. and Smothermon, J. H. (eds). 1993. Masa ye, gawa ye ra˜ca a˜mara tuti (Macuna – espan˜ol diccionario de 850 palabras). Bogota´: Editorial Alberto Lleras Camargo. Sorensen, A. 1967. Multilingualism in the Northwest Amazon. American Anthropologist 69: 670–684. Souza, I. de. 2008. A lı´ngua dos ´ındios Kinikinau. Ph.D. thesis, University of Campinas. Stenzel, K. 2004. A reference grammar of Wanano. Ph.D. thesis, University of Colorado. Stenzel, K. Forthcoming. A Reference Grammar of Kotiria (Wanano). Lincoln: University of Nebraska Press.

On form and function in language contact

229

Strom, C. 1992. Retuara˜ Syntax. (Studies in the languages of Colombia 3.) Dallas: SIL. Thomason, S. G. and Kaufman, T. 1988. Language Contact, Creolization, and Genetic Linguistics. Amsterdam: John Benjamins. Velie, D. and Velie, V. 1981. Vocabulario Orejon. Lima: SIL. Weir, H. 1984. A negac¸a˜o e outros to´picos da grama´tica Nade¨b. M.A. thesis, Federal University of Campinas. Weir, H. 1990. Incorporation in Nade¨b. In Amazonian Linguistics, D. L. Payne (ed.), 321–363. Austin: University of Texas Press. Weir, H. 1994. Nade¨b. In Typological Studies in Negation, P. Kahrel and R. van den Berg (eds), 291–323. (Typological Studies in Language, 29.) Amsterdam: John Benjamins. West, B. 1980. Grama´tica popular del tucano. Bogota´: Ministerio de Gobierno. Wheeler, A. 1970. Grammar of the Siona language, Colombia. Ph.D. thesis, University of California/Berkeley.

The Basque articles -a and bat and recent contact theories Julen Manterola1 1. Introduction The aim of this article is to criticize some claims that have been made for Basque definite and indefinite articles within certain theories of contact. It will become clear that contact played a crucial role in the evolution of Basque articles, but not exactly the way proposed by the criticized authors. Haase (1992) and following him Heine and Kuteva (2003, 2005, 2006, 2007) have claimed that the indefinite article bat ‘a’ in Basque has developed diachronically under the influence of this grammatical category in the Romance languages. This issue has not been explicitly discussed by mainstream Basque diachronic linguistics. A contact-induced origin for the definite article -a ‘the’ has been suggested (Michelena [1978] 1987: 366; Trask 1997: 199), although Haase and Heine and Kuteva do not mention this hypothesis. Still, as far as I can see, Haase and Heine and Kuteva are the first researchers who take into consideration current theories on language contact in approaching the question of Basque articles. There is a twofold problem with Haase and with subsequent work by Heine and Kuteva. First, when talking about the Basque indefinite article they neglect one important fact, namely the existence of an ancient plural indefinite article batzu ‘some’; this plural indefinite article is clearly essential for the understanding of the development of the singular indefinite

1. I would like to thank the following people: the audience at the workshop on Language Contact and Morphosyntactic Variation and Change for their attention, Marianne Mithun and students of her Language Contact 2008 winter course at the UCSB for their comments, Bernard Comrie for his support. Joseba Lakarra and Ce´line Mounole deserve special mention, since many of the ideas here have been discussed with them. Finally, I am greatly indebted to two anonymous reviewers’ comments, which really helped me improve this article. The research was made possible thanks to financial aid from the Research Department of the Government of the Basque Country.

232

Julen Manterola

article bat ‘a.’ Second, there is a problem with Heine and Kuteva’s main hypothesis: it predicts that replicated grammatical features are less developed, but that is not at all the case with the Basque definite article -a, which is much more developed in its grammaticalization path than its counterparts in Romance languages. My critique here falls into two parts: it relates to the diachrony of the Basque indefinite and definite articles on the one hand, and on the other hand to how Heine and Kuteva fit the diachronic development of Basque articles into their theory of contact. The extreme grammaticalization degree of the definite article, for instance, seems to be due to a contact process di¤erent from their contact induced grammaticalization hypothesis. The article is organized as follows: in section 2 I explain what definite and indefinite articles are in grammaticalization terms; section 3 is devoted to the relevant facts we know about Basque articles, their diachrony and the role of contact in their development; section 4 focuses on how Basque articles have been dealt with in most recent contact theories; finally in section 5, I summarize the critique developed in the previous sections.

2. What is an article? 2.1. Definite articles Following Himmelmann (2001), who in turn relies on Greenberg’s seminal work (1978), I take a fairly strict diachronic view of what a definite article is; it is only in this way that we can compare Basque and Romance definite articles, since from a synchronic point of view one might think that they do not represent the same functional category. I illustrate this synchronic di¤erence in behavior with some examples below (see section 3.1.1). Briefly, a definite article can be defined as a grammatical category at a certain point on the diachronic continuum that leads from distal demonstratives to definite articles,2 then to specific articles, and finally to noun

2. This deliberately non-concrete synchronic definition of what we call ‘‘definite articles’’ goes together with the wide range of di¤erent uses displayed by elements of di¤erent languages that are assumed to be demonstratives and definite articles. The diachronic and language-specific view we are taking in this article will allow us to avoid these problems.

The Basque articles -a and bat and recent contact theories

233

markers, via grammaticalization, as in Himmelmann’s schema (2001: 832): ‘‘demonstrative ! definite article ! specific article ! noun marker’’.3 As a general term for the grammaticalized elements that derive from a demonstrative, Himmelmann uses the term D-element. From a strict methodological point of view, calling a certain morpheme a definite article is thus simply a convention: we will not find a D-element in one language that behaves exactly the same way as in another. This is why it is so important to keep a diachronic perspective in mind when analyzing such morphemes, especially when, following Heine and Kuteva, I discuss contact-induced grammaticalization. There are at least two observations to be made about this deliberately simple definition. First, although this is not commonly mentioned in the literature, it seems that definite articles may develop from sources other than demonstratives. Frajzyngier (1996) discusses how what he calls ‘‘definite markers’’ have diachronically arisen from items such as va´ ‘hand’ in Gidar and from verbs of saying such as *(V )nV in other Chadic languages; these definite markers seem to display a range of uses analogous to that of ‘‘typical,’’ such as European, definite articles. Although these cases of grammaticalization provide very interesting data about the diachronic evolution of definiteness marking, I do not believe they a¤ect what I have to say here about the Basque definite article, since we can almost certainly take for granted that it belongs to the D-element continuum schema cited above. Second, the line separating demonstratives from articles and noun markers may sometimes be unclear, although we can draw on some criteria for the distinction (see Himmelmann 2001). Good examples of how fuzzy the borderlines between methodologically established phases of the continuum may be include Chinese and Montagnais, an Algonquian language. They are both said to be ‘‘non-article’’ languages. Work by Huang (1999) and Chen (2004: 1148–1156), among others, suggests that some instances of demonstrative use in Chinese can be better understood as article-like use; this might perhaps be understood as an indication of the incipient development of a definite article in Chinese. The case of Montagnais, according to Cyr (1993), may be much more extreme, since we may be dealing with a D-element at a very high degree of grammaticalization: in this 3. Of course, this is a rudimentary schema, as Himmelmann himself admits (2001: 832). We can look at many studies on di¤erent languages to get an idea of the details of this grammaticalization path; good examples are Company (1991) for Spanish and Epstein’s work (1993, 1994, 1995) for French.

234

Julen Manterola

case, the identity of form has somehow concealed the fact that D-elements preposed to the noun phrase are in fact articles, in contrast to postposed demonstratives. These problems do not arise in the tradition built by linguistics scholars who have worked on Basque, since here the definite article is a wellestablished category. On the contrary, problems may arise from the fact that -a ‘the’ seems to be highly grammaticalized, as we see in section 3.1 below. These two observations serve to emphasize the specific perspective from which I approach the diachronic dimension of definite articles. This progressive grammaticalization perspective is the one taken by Heine and Kuteva (2005) in their contact-induced grammaticalization approach to contact issues. Moreover, it is this same diachronic point of view that allows us to talk about a definite article – crucially, a D-element – in Basque, even though its usage does not greatly resemble definite articles in other western European languages.4 Usually terminological problems, such as calling Basque -a an ‘‘individualizer’’ instead of a definite article, are rooted in methodological decisions made when trying to identify the essence of a hypothetical ideal definite article from a strictly synchronic point of view. 2.2. Indefinite articles I also look at indefinite articles in their diachronic dimension, as grammaticalizing items; as mentioned above, this way of looking at the morphemes is one of the basic features of contact-induced grammaticalization, as outlined in Heine and Kuteva’s work (see section 4.2.1 below for a summary of their hypothesis). Even so, in section 4.3 I will criticize some particular claims these authors make regarding the specific grammaticalization path of the Basque indefinite article. It is well known that cross-linguistically the main source for an indefinite article is the numeral ‘one.’ As Heine points out (1997: 71), ‘‘the evolution from lexical to grammatical structure is not discontinuous but proceeds gradually.’’ Looking at the progressive ‘‘contextually defined extensions’’ of the use of the numeral ‘one,’ it is possible to divide these extensions

4. As Milsark reminds us (1977: 5), the term ‘definite article’ ‘‘has been used for generations in the pedagogy and scholarly description of the Indo-European languages’’ and its synchronic formal and semantic characterization has always been performed on the basis of the behavior of these languages.

The Basque articles -a and bat and recent contact theories

235

into descriptively convenient stages. Heine proposes a five-stage model for the diachronic development of the indefinite article. I invoke this model mainly because it is the one on which Heine and Kuteva’s arguments are based (2006: 104–105): ‘‘numeral ! presentative marker ! specific marker ! non-specific marker ! generalized article’’. I will not retrace the details of each of these stages; the interested reader is referred to Heine’s work (1997: 71–76). I will simply mention one of the characteristics Heine gives for the fifth stage, the generalized article: it is in this last stage, he says, where ‘‘the use of the article is no longer restricted to singular nouns but is extended to plural and mass nouns, as in the following example from Spanish’’ (1997: 73). Then he observes how Spanish uno/una ‘a, one’ can be used with plural morphology, unos/unas ‘some’ (the two forms stand for masculine/feminine marking). I understand that he intends this Spanish plural indefinite article to be a characteristic of a final phase of the grammaticalization of an indefinite article. The relevance of this point will become apparent later, where we see that Basque also has an ancient plural indefinite article. Besides this diachronic grammaticalization view, which following Heine I take as a basic approach to indefinite articles, there are some further points worth mentioning, related to certain implicational relationships between definite and indefinite articles. First, it is widely noted in the literature that definite articles develop earlier than indefinite ones, so that there are more languages with a definite article but no indefinite article than vice versa. Heine himself generalizes this observation (1997: 69): ‘‘If a language has a grammaticalized indefinite article, it is likely to also have a definite article, while the reverse does not necessarily hold true. Thus, the presence of an indefinite article is likely to be accompanied by that of a definite article, but not vice versa.’’ Perhaps related to this observation, there is the question of what areal typology could tell us about indefinite articles; I quote Heine (1997: 79) again: Thus, one might expect with a certain degree of probability that a given language will have an indefinite article if the neighboring language or languages also have one. The older Germanic languages did not have a definite or indefinite article, in much the same way as the ancestor of the modern Romance languages did not. On the other hand, most modern European languages across genetic boundaries have both kinds of article.

Interestingly enough, Basque also has both kinds of article, although we do not know, at least as far as the indefinite article is concerned,

236

Julen Manterola

whether it had them prior to contact with indefinite-article languages. I will come back to this point. A second noteworthy cross-linguistic observation is that it seems that, rather than extending the former numeral one to plural nouns (recall Spanish unos/unas ‘some’), languages most frequently use alternative strategies to introduce indefinite articles for plural nouns (Heine 1997: 77). This observation of the rarity of plural articles derived from the numeral one is also made by Himmelmann (2001: 838), using the same example from Spanish. Since languages with plural indefinite articles derived from numerals are not cross-linguistically common, it is highly significant that there is an area where this kind of item can be found across languages that are not even genetically related, like Basque and Romance languages. 3. Basque articles As a brief introduction, I outline the Basque declension. Some of the paradigms here can be found in Hualde and Ortiz de Urbina (2003: 173–174). Here I decline the word etxe ‘house’ in its determinerless form, definite singular/plural, and indefinite singular/plural; only four cases are illustrated. Table 1. Standard inflectional paradigm of etxe, ‘house’ - DET

DEF SG

DEF PL

INDF SG

INDF PL

ABS

etxe

etxe-a

etxe-ak

etxe bat

etxe batzu-k

ERG

etxe-k

etxe-a-k

etxe-ek

etxe bat-ek

etxe batzu-ek

DAT GEN

etxe-ri etxe-ren

etxe-a-ri etxe-a-ren

etxe-ei etxe-en

etxe bat-i etxe bat-en

etxe batzu-ei etxe batzu-en

I prefer here to call the nouns in the first column determinerless or bare, rather than indefinite, which is the term used in the grammar referred to above. I keep ‘indefinite’ for the nouns with an indefinite article in the last two columns. These may be mere terminological divergences; the important thing is to know which are the morphemes under each label. 3.1. -a: Basque definite article Basque does have a definite article, or at least a D-element as defined in 2.1; from a diachronic point of view the so-called definite article in Basque is just one more instance of a grammaticalized distal demonstrative.

The Basque articles -a and bat and recent contact theories

237

The definite article in modern Basque is -a, a bound morpheme attached to the rightmost element of the whole phrase it modifies; for further information about its behavior, see Trask (2003: 118–121) and Hualde (2003: 171–177). 3.1.1. The origin of the Basque definite article There are two main forms of the distal demonstrative in modern Basque, depending on the dialect: in western Basque, the distal demonstrative is a ‘that’ and in central and eastern Basque (h)ura ‘that.’ The central and eastern form is said to be a restructured form of a former distal demonstrative, usually reconstructed as *(h)a(r) ‘that,’ although no convincing explanation has been given for its exact formation. As we can see, the distal demonstrative a ‘that’ in western Basque coincides exactly in form with the definite article -a, as in an example from Azkue (1923: 269). (1) a.

gizon a man that ‘that man’

b. gizon-a man-the ‘the man’

Example (1b) illustrates the use of the definite article in all varieties of Basque; for surface phonetic variants, see Hualde and Gaminde (1998). The whole system of demonstratives is reconstructed as having an initial sound, usually an initial aspiration, which is why in standard Basque their normative form is hau ‘this,’ hori ‘that,’ and hura ‘yonder.’ This is how it is still pronounced in some northeastern varieties of the language. Medieval documents contain some instances of the article that still include the aspiration: Udalha, Adurzaha (Manterola 2006: 674). These aspirated instances of the D-element are in fact very close to what has been reconstructed as *(h)a(r), and confirm the common opinion of its demonstrative origin (Azkue 1923: 269; Michelena [1971] 1987: 146; Trask 1997: 199). In short, the definite article -a in Basque perfectly fits the D-element description. It is in these terms that we can continue to call -a a definite article; it has been so called in traditional Basque linguistics. These, I believe, are terms Heine and Kuteva would agree with. It is important to clarify this point about the origin of the Basque definite article. Heine and Kuteva have written that ‘‘one may argue that -a is not really structurally equivalent to definite articles in SAE languages’’ (2006: 32). It is true that, from a strictly synchronic point of

238

Julen Manterola

view, the definite articles in Basque and in geographically adjacent languages do not share the same morphosyntactic features, nor a common behavior; but when Heine and Kuteva (2003, 2005) talk about model and replica features, they are not thinking, as far as I understand them, in terms of a strict synchronic grammatical or structural equivalence, but rather of an equivalent grammaticalization path. To this extent, inasmuch as it is an instance of the grammaticalization path outlined above, the Basque definite article o¤ers a straightforward parallel to that of the Romance languages. The only di¤erence lies in the broader use Basque speakers make of it; as Trask says, ‘‘[t]he label ’definite article’ is misleading, since this article is of much broader use than the English definite article’’ (Trask 2003: 119). It can even be used in predicates and existential sentences (2), as well as in the citation form of nouns and adjectives. We are thus dealing with an article moving from Greenberg’s stage II toward stage III (Greenberg 1978: 62–74), or toward the rightmost edge of the D-element continuum. (2) a.

ardo-a badago 5 wine-the there.is ‘There is wine.’

b.

ibai irakasle-a da Ibai teacher-the is ‘Ibai is a teacher.’

We can also find this D-element, -a, in adjective predicates. (3) a.

b.

Nerea neska jatorr-a da Nerea girl nice-the is ‘Nerea is a nice girl.’ Nerea eta Maider neska jatorr-ak dira Nerea and Maider girl nice-the.pl are ‘Nerea and Maider are nice girls.’

Later on it will become apparent why I give the plural example in (3b), since some interesting contact-based reasons have been proposed for these plural predicates (Irigoien 1985: 129). I cannot here provide a complete description of the uses of -a; in fact, an exhaustive study of its use across dialects and through history is still lacking. For further information about dialectal and historical variation on the use of this D-element, as well as

5. I will continue to gloss it as the, in order to make explicit once again the parallel diachronic source shared by both morphemes, Basque -a and English the.

The Basque articles -a and bat and recent contact theories

239

for some possible semantic and functional explanations of its spread and ´ lvarez (1977), Artiagoitia (1998, 2002), and behavior, see among others A Etxeberria (2005: 167–250). 3.1.2. Basque definite article -a and contact Thus far, I have made two claims about the definite article -a: first, it has a demonstrative origin, and second, it has gone further along the grammaticalization path for articles than Romance languages have done. Two further points relate to what has been said in traditional Basque linguistics regarding this -a article and contact issues. First, it has usually been assumed to have arisen due to contact with Late Latin and incipient Romance languages (Michelena [1978] 1987: 366). There are at least two noteworthy reasons in support of this hypothesis: languages typologically akin to present-day Basque (agglutinative, SOV, postpositional) do not usually have a definite article (Himmelmann 1998: 350; Plank and Moravcsik 1996: 205), so contact seems an appealing explanation for its presence in Basque. On the other hand, Basque seems to have begun to develop its definite article at the same time as neighboring languages (Lapesa 1961; Epstein 1994); this would mean that Basque is simply one example of a widespread western European phenomenon. The still aspirated instances from the Middle Ages cited above point toward this dating of the emergence of the article in Basque.6 Why the relative order (Basque noun þ article versus Romance article þ noun) is di¤erent has never been addressed. I believe that the contact scenario might be an appropriate one, since the development of definite articles seems to have been an areal event in western Europe during the Middle Ages (Haspelmath 1998); however, we have to distinguish clearly between speculation and empirically or theoretically based certainties. Until now no thorough study has been carried out to compare the parallel development of articles in Basque and Romance languages (and data are not extensively available). 6. Some researchers have noticed (Irigoyen 1986: 86) the interesting existence of a roughly 2000-year-old Latin inscription found close to Ca´ceres (Spain) where the word Ibarra appears. It has been taken as proof of the early existence of the definite article by others (Iglesias 2007), since in present-day Basque the word means ‘the valley,’ analyzed as ibarr-a ‘valley-the.’ As long as this kind of data remains so scanty and isolated, I feel it more prudent not to draw large conclusions from it. Other authors with no connection to this tradition have also suggested the possibly ancient character of the Basque definite article, although for other reasons (Putzu and Ramat 2001: 121), and in a very tentative way.

240

Julen Manterola

Second, in discussions of contact and articles another issue in the Basque linguistics tradition calls for attention: one of the reasons claimed (Irigoien 1985: 129) as explaining the spread of the singular and plural definite articles -a/-ak is the need to make clear a distinction otherwise non-existent in the language, namely the morphological marking of the singular/plural distinction. Since the articles were the only place in which that distinction was overtly encoded, it seems that, by ‘‘forgetting’’ about their definiteness, both the articles (the singular -a and plural -ak) spread following the model of Romance singular/plural overt morphology (cf. Spanish cama/camas ‘bed/beds’).7 It might then be said that the overt singularplural distinction in nouns and adjectives has expanded at the expense of the definite singular and plural articles. While we wait for an exhaustive study of this topic, (3b) above might be an example of this. The Spanish equivalents of the sentences in (3) would be: (4) a.

b.

Nerea es una chica agradable Nerea is a girl nice ‘Nerea is a nice girl.’ Nerea y Maider son chica-s agradable-s Nerea and Maider are girl-pl nice-pl ‘Nerea and Maider are nice girls.’

We can see here that the bare plural predicate chicas agradables ‘nice girls’ possesses a plural marker, the bold -s at the end of both noun and adjective, a marking that modern Basque would have replicated using its phrasal articles. The same holds for Trask’s observation that ‘‘ura may correspond either to ‘water’ or to ‘the water,’ and umeak may correspond either to ‘children’ or to ‘the children’.’’ (2003: 121). Thus, even though the article might have arisen through contact, the path of its expansion has perhaps not strictly followed the typical grammaticalization process usually assumed in the case of D-elements. Here another contact factor may be involved, a factor with no direct relationship to the emergence of a definite article: namely, the need for an overt morphological distinction between singular and plural of the kind already present in the nearby languages. This might well have played a crucial role 7. One could have doubts about the exact nature of the plural definite article; it is most usually related to the toponymy morpheme -aga, and said to be more recent than the singular. This is a discussion I cannot take up here.

The Basque articles -a and bat and recent contact theories

241

in the spread of the article, in that it is only in the articles that singularity and plurality were overtly marked. 3.2. bat: the Basque indefinite article Basque has an indefinite article, bat, which has exactly the same form as the numeral meaning ‘one.’ It is thus commonly supposed that it has its origin in the numeral. As far as I know, there is no extensive study of its modern use, nor of how it has evolved through the centuries and across di¤erent dialects. It is also commonly believed that its use is much more restricted than in Romance languages (Trask 2003: 122). I will o¤er here a single example in order to show briefly how bat ‘a’ functions in contrast to the ‘‘definite’’ -a ‘the.’ These examples, reminiscent of Givo´n’s (1981: 36), are both translated by the English indefinite article a: (5) a.

b.

azeri bat ikusi dute herrian fox a seen have in.town ‘They have seen a (certain) fox in town.’ azeri-a ikusi dute herrian fox-the seen have in.town ‘They have seen a fox in town.’ [not e.g. a wolf ]

We see that bat ‘a’ (5a) is used as a specific marker (Heine 1997: 72–73), exactly as in Givo´n’s street Hebrew -xad. The noun phrase with -a in (5b), given the appropriate context, can be interpreted in terms of kind reference; its street Hebrew counterpart would be a bare noun. If my language intuitions are correct, these are instances of central Basque. There may be (in fact there are) di¤erences across dialects and speakers. However, no thorough study of the di¤erent values of bat has been carried out up to now. In French and Spanish, both sentences would also be translated with un, the indefinite article diachronically resulting from the numeral. It is widely recognized in the literature that the use of bat in Basque is much more restricted than in its Romance counterparts. Looking at data like those in (5), one might wonder whether there are other reasons for this besides those proposed by Heine and Kuteva; I will come back to this question in section 4.2.3.2. With respect to the earliest evidence of the existence of the indefinite article bat in Basque, we can only say that it appears in every text, in di¤erent dialects, in the sixteenth century, and is thus not a recent innova-

242

Julen Manterola

tion. Its presence in every dialect might indicate that it goes back as far as the period of ancient common Basque, around the fourth or fifth century; but we have no examples of this like those we have for the definite article -(h)a in the Middle Ages. The kind of corpus we have at our disposal from the Middle Ages – mostly person and place names inserted in Latin or Romance-language texts – does not make it likely that we will find instances of indefinite articles. We simply cannot know how old bat is in its role as an indefinite article. 3.2.1. batzu: the Basque plural indefinite article Basque crucially has a plural indefinite article batzuk ‘some,’ morphologically based on the numeral/indefinite article bat ‘a, one.’ Leaving aside the final -k, a newer addition to the older batzu, we can dissect it as bat þ zu. Bat has already been discussed in section 3.2, and -zu is a collective su‰x that is no longer productive in modern Basque; indeed, its productivity, as far as we can trace it, was already decreasing in the Middle Ages, as shown by Michelena ([1971] 1987: 147). Batzu is also common to all historical dialects,8 and has been present in the records since the very beginning of the historical period for Basque in 1545. This most probably means that batzu is at least a thousand years old, from a time when plurality was marked in ways other than using the articles, as seen in 3.1.2; one possibility is that it already existed at the time of the ancient Basque koine´ 1500 years ago (Michelena 1981). Of course, another possibility is that it later spread from one dialect to other. Michelena himself thinks of it in terms of a replication of the Spanish unos/unas, a hypothesis that is also possible ([1971] 1987: 148). 8. A reviewer has pointed out to me that in present-day Zuberoan, an eastern dialect, elibat ‘a bunch’ is used instead of batzu; Otsibar’s texts o¤er an example of this (2003). Nevertheless, we can confirm the use of batzu in some ancient texts of that same dialect (Tartas 1666, Egiategi 1785). Interestingly, we may make an observation in line with this reviewer’s doubt about the pandialectal character of batzu: the contiguous eastern dialect, the extinct Roncalese, has another option as well as batzuk, seemingly also based on the numeral bat ‘one’; these forms are banak (absolutive) and banek (ergative), whose exact morphological nature is unclear to me, but significantly seems to have D-element based plural markers. In fact, contrary to what Azkue’s dictionary says (1905–1906: 138), I could not find a single instance of Roncalese batzuk in the texts I consulted (Irigoyen 1957, Pagola 2004). A study of the specific evolution of these eastern forms that could shed light on the diachronic evolution of the use of batzu is lacking.

The Basque articles -a and bat and recent contact theories

243

Again, we simply do not know, and maybe cannot ever know. The important point here for the following discussion is its unquestionable antiquity, as shown by its morphology (the old collective su‰x) and especially by its presence in all dialects. 3.3. Basque articles and contact: summary The best guess is that both articles arose roughly at the same time during the Middle Ages. But there are some caveats here. With respect to the indefinite article, the data we have at hand do not shed any light on its status in the Middle Ages. We simply cannot demonstrate its existence or non-existence prior to the Middle Ages; this is above all an empirical issue for which no data are available. In other words, we have no evidence for a stage of the language in which the indefinite article did not exist; we also know of the intriguing ancient plural indefinite article. The situation for the definite article might seem clearer, since even as late as the eleventh century there are instances of aspirated articles; this could mean that it began to grammaticalize quite ‘‘late’’. An alternative hypothesis, however, which is possible although maybe not probable, is that it began to grammaticalize earlier in the Middle Ages or even before then, and retained the aspiration for a longer time. My own view is that, as long as no strong counter-evidence appears, the medieval character of the definite article is the least extreme hypothesis. It is worth keeping in mind that these hypotheses have at best the status of most probable guesses; in responding to the question ‘‘What was the situation before contact?’’ we should make a clear distinction between what can be considered empirical evidence and what is purely hypothetical. A final note on determinerless nouns – the first column in Table 1, or what Haase calls transnumerals: it bears recalling that these forms were much more widely used 500 years ago in every dialect. The most straightforward guess is that the old Basque noun phrase had no overt morphological mark for plurality (except probably for some collective particles) nor for definiteness (Lafon 1954). Eastern dialects, especially Zuberoan and Roncalese, exhibit an interestingly archaic character in this regard. 4. Basque articles and recent literature on contact 4.1. Haase, contact, and Basque articles Haase’s (1992) analysis of Basque articles bears rereading, since it is the first study intended as a general survey of Basque from the point of view

244

Julen Manterola

of modern contact theories. This rereading turns out to be unavoidable given that important studies on contact, such as Heine and Kuteva’s, rely almost exclusively on Haase as far as Basque is concerned. I first sum up in subsection 4.1.1 his position on the definite article, and then in subsection 4.1.2 what he says about the indefinite article. 4.1.1. The definite article in Basque: Haase’s view Haase devotes some 4–5 pages (1992: 53–58) to what I am calling here the definite article. He limits himself to a brief description of its use and nonuse, a description that we can find in Lafitte (1944). He discusses instances of the definite article in predicative sentences, and, following Iturrioz (1985), states that the Basque definite article -a is a kind of individualizer, not really an article. Iturrioz’ analysis of the -a morpheme, as far as I can follow his main argument (1985: 176–181), is just a synchronic account of the amazingly wide range of uses of the -a morpheme. The inaccuracy of Haase’s description heavily biases Heine and Kuteva’s view of the Basque definite article. There are some points I feel are lacking in Haase’s analysis. First and most important, he says nothing about the origin of the article; as noted in section 3.1.1, its characterization as a Delement is widely accepted in the literature. Second, he says nothing about the hypothesis according to which the Basque -a morpheme arose through contact. This hypothesis is often mentioned and widely accepted in Basque linguistics (see 3.1.2). He also says nothing about the possible role of the singular/plural overt distinction in Romance languages in the spread of the Basque definite article. There are two other points which are more minor, but essential when one is talking about contact over centuries. Haase says nothing about dialect variation in the use of -a. It may be worth mentioning this since the immediately contiguous dialect to the one he analyzes, Zuberoan, exhibits extensive absence of the article in contexts where most dialects would use it; interestingly, the behavior of this contiguous dialect has been attributed either to archaism or to French contact (Azkue 1923: ´ lvarez 1977). Moreover, he does not deal at all with historical varia265; A tion, but many observations have been made about the gradual extension of -a (Lafon 1954; Michelena [1970] 1987: 293, 1978). Along with Haase’s failure to mention these facts about the Basque definite article in his 1992 work, there are incorrect analyses and methodological gaps in his treatment of the definite article -a. First, he analyzes the noun phrase of his sentence in (134) (1992: 55), here in (6), as if it

The Basque articles -a and bat and recent contact theories

245

were an instance of what he calls a transnumeral. I give my own English glosses. (6) Hemen badira jende xahar bat-zu here there.are people old one-pl ‘Here there are some old people.’ What Haase, following Iturrioz, calls transnumerals would correspond to the determinerless or bare nouns in the first column of my Table 1. Since the phrase at stake is jende xahar batzu ‘some old people,’ and it clearly includes a plural indefinite article batzu modifying the noun or adjective phrase, it is wrong to label it a transnumeral. Furthermore, Basque grammars never mix up the transnumeral declension with the indefinite articles of the noun declension (see Hualde and Ortiz de Urbina (2003: 118–136) for a recent example). Second, Haase says nothing about the history and development of definite articles in Romance languages, a debate with a large literature; this gap is to some extent understandable, since he is not ¨ bersetzungsa¨quivalent of aware that Basque -a is what he would call an U definite articles in Romance languages. 4.1.2. The indefinite article in Basque: Haase’s view Haase devotes 1–2 pages to the indefinite article bat ‘one’ (1992: 59–61, 71). He aims to demonstrate that it arose due to contact, but in my opinion he does not provide enough empirical support for his hypothesis (Haase 1992: 59).9 Der baskischen Transnumeral-Singular-Plural-Opposition steht in den romanischen Kontaktsprachen die Definit-Indefinit-Opposition gegenu¨ber. Hierbei entspricht der indefinite Artikel dem Zahlwort ‘eins.’ Im Sprachkontakt wird das baskische Zahlwort ‘eins’ ebenfalls zum unbestimmten Artikel (Emphasis mine – JM) . . . Anders ausgedru¨ckt: bat und frz./gask. un sind im Bereich der Zahlwo¨rter ¨ bersetzungsa¨quivalente. Wie in anderen Fa¨llen . . . kann sich nun der U Funktionsbereich von bat auf alle die Fa¨lle ausbreiten, in denen in den Modellsprachen un gebraucht wird, also auch auf die Signalisierung von Indefinitheit. The Basque transnumeral-singular-plural opposition contrasts with the definite-indefinite opposition of Romance contact languages. In that sense the indefinite article corresponds to the numeral ‘one.’ In a language contact situation the Basque numeral ‘one’ likewise becomes an indefinite article (Emphasis mine – JM) . . . 9. I would like to thank Max Hofheinz and Ursula Laarmann for their help with the exact understanding of these texts.

246

Julen Manterola

In other words: bat and French/Gascon un are translation equivalents in the domain of the numerals. As in other cases . . . the range of functions of bat can be extended to all the cases in which un would be used in the model languages, including therefore also the marking of indefiniteness.

After having stated this, he gives two more examples to show that bat, the numeral ‘one’ moving towards an indefinite article, has been extending its semantic meaning. He gives a sentence from a 1782 work and another from the first printed book in Basque (1545), here in (7) (his examples (161–162), (1992: 60)). (7) balia dikezit senhar gaixto bat can be.for you husband bad a ‘I can be a bad husband for you.’ He adds a comment on the use of the indefinite article bat: Der Gebrauch von bat ko¨nnte durch das Verb baliatu ausgelo¨st worden sein. Im Keim zeigt sich aber schon die im Sprachkontakt katalysierte Entwicklung. The use of bat could have been triggered by the verb baliatu. However, at its core it appears to be a development catalyzed by language contact.

I am not really convinced by a single example from a single language that this bat use was triggered by language contact. I do not mean that language contact plays no role (in fact, I believe it may have played a determining role), but I would expect a much deeper analysis to support this claim, with examples of as many old texts as possible, comparing them to data from other dialects, periods, and model languages. Even then, after having ‘‘squeezed’’ our data as much as we can, we sometimes have to admit we can not go any further. In any case, a thorough knowledge of old texts and dialects always comes first. These are our tools, and we cannot neglect them. In short, two examples, dating from 1545 and 1782, are not enough support for the claim that Basque bat and French and Gascon un are translation equivalents, nor for a direct inference about the direction of an alleged contact-induced change. Besides these poorly supported statements, there is a crucial silence about another aspect of the indefinite article: its plural batzu ‘some.’ Since Haase simply omits this article, we cannot know whether he would also attribute its existence to contact. We may recall that batzu has to be quite ancient (see section 3.2.1), or at least older than Haase’s description together with Heine’s grammaticalization scale for indefinite articles would lead us to think.

The Basque articles -a and bat and recent contact theories

247

One further claim by Haase deserves comment (Haase 1992: 61), his final statement before he proceeds to discuss case and postposition systems. Das baskische Determinationssystem, das auf der Opposition zwischen Transnumeral, Singular und Plural beruht, ist – wie wir gesehen haben – destabilisiert worden. Zum einen wird die Unterscheidung von nichtindividualisiertem und individualisiertem Pra¨dikatsnomen aufgegeben, zum anderen wird das Zahlwort fu¨r ‘eins’ nach romanischen Vorbild zum indefiniten Artikel, der anstelle des Individualisierers eintreten kann. The Basque system of determination, which rests on the opposition between transnumeral, singular, and plural, has been – as we have seen – destabilized. On the one hand, the di¤erence between non-individualized and individualized noun predicates is abandoned, on the other hand the numeral for ‘one’ becomes, following the Romance model, an indefinite article, which can take the place of the individualizer.

It is di‰cult to understand which time period Haase is taking as a basis at any one point: sometimes it seems he is talking about recent changes in Basque. An example of this might be the sentence in (166) from his own fieldwork, given just before the sentence just quoted; I repeat it in (8) with my own glosses. (8) tokero bat zen driver a was ‘(S)he was a cattle-driver.’ This might be a good example of how bat has extended its use on the model of Romance languages, since in the dialect from which he takes examples this profession noun predicate would usually bear no determiner, while in western dialects it would take -a. But Haase’s data do not tell us how ancient this instance of bat is, nor can we determine this based on the sole example he provides in his analysis. Nor do we know which Romance model he has in mind. In his defense, it does not seem that Haase intends this sentence to be more than an example of current contact-induced use of bat. If he is taking this kind of example as evidence for a recent destabilization of what he calls Basque transnumeral-singular-plural opposition, I might perhaps agree with him. At other times, still with reference to the last quoted example, one has the impression that Haase is talking about changes from long ago: when he says that bat became an indefinite article following the model of Romance languages, one can assume that he is aware of the relative antiquity of bat in indefinite article uses. If this is so, then there is a problem

248

Julen Manterola

when he treats all contact-induced changes and analysis of language systems/oppositions from di¤erent periods simultaneously. The Transnumeral-Singular-Plural opposition he takes as the ancient and original determination system in Basque, the one represented by the three leftmost columns in Table 1, is no such thing. As discussed in 3.1.2, this opposition in present-day Basque cannot really be the ‘‘original’’:10 what Haase calls singular and plural in fact possess a demonstrative-based definite article. We have also seen that they arose in the Middle Ages at the same time as in some other western European languages, most probably in an areal configuration. The problem is, again, that Haase does not treat his ‘‘individualizer’’ as an instance of the D-element grammaticalization path (section 3.1.1). Suppose that at roughly the same time, in the Middle Ages, an indefinite article had appeared (recall our ignorance about its exact date of arrival, section 3.3). In fact the alleged appearance of bat in the Basque system might well have happened simultaneously with the appearance of the singular and plural articles and the configuration of bare-noun vs. definite articled-nouns in modern Basque (Haase’s transnumeral-singular-plural opposition). In that case the indefinite article bat would not have destabilized any former transnumeral-singular-plural opposition; this latter opposition would surely have been developing together with the further grammaticalization of the indefinite article bat. All these observations, I believe, make it much more di‰cult to understand what Haase means when he treats the Basque transnumeral-singularplural system as in opposition to the Romance definite-indefinite. Moreover, the singular-plural opposition might have developed by ‘‘parasitizing’’ the definite-indefinite one, as suggested in section 3.1.2. As a final comment on the quotation above regarding the transnumeralsingular-plural opposition that Haase takes as originally Basque, recall the points made in section 3.3. The ‘‘original’’ Basque, of perhaps 1600 years ago, certainly had no Romance-influenced overt morphological marking of singular vs. plural, or any overt morphological definiteness marking. Whatever we might think about how Haase deals with contact and the indefinite article bat, I believe some of the spread of the indefinite article could be accounted for in terms of contact; we should, however, start by 10. It always depends, of course, on what we mean by ‘‘original.’’ Here I refer (and I believe Haase wanted to talk in these same terms) to the possible system of Basque before contact with Latin and subsequent Romance languages (bearing in mind that we have no data available).

The Basque articles -a and bat and recent contact theories

249

locating it in its correct chronology relative to the development of articles in Romance languages. This is a basic task that remains to be done. Basque linguistics should some day construct a description of the uses of the indefinite article in historical data from 1545 onward, across di¤erent dialects. Claims about change have to be supported by as many examples as possible, coming from di¤erent dialects and historical periods. Unfortunately, Haase’s contribution, discussing only four or five examples of bat, is not helpful for the accomplishment of this task. As I have tried to show here, some points in his reasoning should be taken cautiously, while others could be better understood with a wider knowledge of Basque diachrony. 4.2. Heine and Kuteva’s model of contact and Basque articles In this section I first o¤er a summary of some of the generalizations Heine and Kuteva make about language contact situations (section 4.2.1), then focus on how they have treated Basque indefinite and definite articles (4.2.2). 4.2.1. Generalizations about contact-induced grammaticalization One of the basic features of contact-induced grammaticalization as explained by Heine and Kuteva is that change is gradual rather than abrupt. Speakers of the replica language activate a pattern in their own language, the one corresponding most closely to the model, thus developing a structure that is equivalent to the one in the model language. This pattern eventually grammaticalizes into a new fully-fledged grammatical category, similar to that of the model language (Heine and Kuteva 2005: 121). Thus, although initially lacking a category structurally equivalent to that of the model language, a pattern in the replica language may follow a grammaticalization path analogous to the one the model language may previously have followed. To this extent, it is legitimate to suppose that the similarity between replica and model language resides especially in the fact that they share the same grammaticalization path for their parallel structure. Other main points Heine and Kuteva make about contact-induced grammaticalization are summed up well in this passage (2005: 101): [W]herever there is su‰cient evidence, it turns out that the replica construction is less grammaticalized than the corresponding model construction . . . in the initial stage of grammaticalization, the new category tends to be ambiguous between its literal and its grammaticalized meaning, it tends to be confined to few contexts, and its use is optional. . . . Such properties are commonly encountered in replicated categories.

250

Julen Manterola

Thus, according to them, when a language ‘‘copies’’ a certain feature or category, it does so gradually, somehow beginning a grammaticalization process of its own that parallels the one in the model language. This hypothesis is a very appealing one, since it allows us to reconstruct, on the basis of the degree of grammaticalization of two features, which language has been the model and which one the replica (Heine and Kuteva 2005: 120): ‘‘[I]t seems possible to determine in a situation where no diachronic information is available which is the model and which is the replica category.’’ (Emphasis mine – JM) Of course, the authors themselves are aware of the limits of this reconstructive technique, and admit that if the contact situation lasts long enough, both categories, model and replica, may eventually ‘‘become structurally indistinguishable’’ (2005: 120). There is a risk here, as I see it, of falling into a circular argument when determining the contact relationship of two languages on the basis of this hypothesis; an in-depth knowledge of the diachrony of the languages concerned should always come first. These indications of Heine and Kuteva’s approach to contact-induced grammaticalization will su‰ce to make sense of the discussion in the next sections. The following section is a reminder of how cautious we have to be when we seemingly lack diachronic information. Data from Basque, which at first sight seemed to fit the above hypothesis, turn out to be contrary to it if we look at them with no theoretical bias of any kind. 4.2.2. The Basque definite article in Heine and Kuteva’s work As far as I can see, Heine and Kuteva explicitly discuss the definite article in only one book (2006: 32). Here they write: [T]he primary function of the ‘definite article’ -a is to individualize referents, and these referents can be, and not uncommonly are, indefinite or even nonspecific . . . . . . that Basque has a definite article can be justified on the grounds that -a is more likely to mark definite than indefinite reference. However, one may argue that -a is not really structurally equivalent to definite articles in SAE languages. In this case, a taxonomic conclusion that one could draw from the observations made is that, rather than having a definite but no indefinite article, Basque has an indefinite but no definite article – hence, quite the opposite of what a discrete-categorization approach of the kind employed by the typologists cited suggests. (Emphasis mine – JM)

The Basque articles -a and bat and recent contact theories

251

Although they make no strong claims about the definite article in Basque, it is clear that relying exclusively on Haase’s work has heavily biased their view of what has been happening over the last thousand years. It is true, as they point out, that -a can be used for non-specific reference (see (4b)), and to that extent one could argue it is something other than a definite article; again, this depends on what we understand by ‘definite article’. I do not believe that being structurally equivalent, or not, to other languages is at all relevant to the point Heine and Kuteva want to make. Indeed, it is contrary to the terms they themselves are proposing for their own approach. What has to be equivalent in the languages we compare is the grammaticalization path of the relevant feature in each language, as suggested in section 4.2.1. Moreover, at least where definite articles are concerned we can barely find such a structural equivalence across languages, depending of course on how we define ‘structurally equivalent’. As already seen in previous sections (3.1.1), the so-called definite article in Basque perfectly fits the grammaticalization path that leads from demonstratives to articles; articles in Romance languages do so as well, inasmuch as many of them have a Latin ille origin. To this extent I believe that Heine and Kuteva would have to admit that we can compare definite articles in Basque and in Romance languages. Moreover, they have already been compared before in the literature, and it has been claimed that the Basque definite article arose in a contact situation (section 3.1.2). However, a problem immediately arises for Heine and Kuteva’s reasoning: Basque -a is used in a more extended way than its Romance counterparts. This relationship between the degree of grammaticalization of articles in Basque and in Romance languages does not fit their expectations: their working hypothesis is that in Basque, as the replica language, the replica feature should be much less grammaticalized than in the model languages. As far as I can see, there are two logically possible solutions to this situation if we want to retain Heine and Kuteva’s hypothesis: either (A) Basque is the model language and Romance languages are the replica languages, or (B) the contact situation has lasted so long that the former relative degrees of grammaticalization between features of the replica and the model have been blurred by time. But each of these possibilities has its problems. In case (A), can we say that Basque is the model language for the definite article, but the replica language for the indefinite one? As we will see below, and as already suggested, according to Heine and Kuteva the development of the Basque indefinite article fits perfectly with its characterization as a replica feature. Is this a problem if we also want to claim

252

Julen Manterola

that the Basque definite article is a model feature for Romance languages? The most straightforward answer to this is yes. If we adopted this explanation, we would be deciding which is the replica and which the model language not on the basis of real sociolinguistic data (which we lack), but on the basis of what fits our hypothesis. In order to avoid this and other methodological problems I propose we disregard this possibility. What the discussion of this logically possible solution indirectly suggests is that perhaps we cannot decide which one is the replica and which one the model on the basis of the relative degree of grammaticalization of the relevant features. This should lead us to reconsider the position of Heine and Kuteva regarding the Basque indefinite article bat; we will come back to this issue in the next section (4.2.3.1). Solution (B) is indeed one that Heine and Kuteva take into consideration, although not specifically in the Basque case. They write (2005: 265): ‘‘[O]ne caveat with regard to this generalization: given enough time, replica categories can develop in the same way as their models. . .’’. Again, if we are to adopt this solution, a problem mirroring the one we sketched for the first solution arises: if the contact situation has lasted so long that the replica category (definite article -a in our case) has developed to the same degree (and beyond, in this case) as its models, what are we to say about the indefinite article bat? Is not the contact period equally long for both definite and indefinite articles? As I suggested above (section 3.2), the indefinite article in Basque may be at least as old as the definite one. Another possibility, of course, is that in our alleged model language(s) the grammaticalization of definite and indefinite articles happened at di¤erent speeds or times. There is a deeper question hovering over these considerations, that of the relative speed at which language change happens, but I will not take this up here. As I have briefly sketched out, some problems arise when we take the development of the definite article as something to be explained in the same way Heine and Kuteva propose for other features. Furthermore, when we analyze it together with the indefinite article bat new problems surface for their hypothesis. In the next section I discuss the indefinite article in more detail. 4.2.3. The Basque indefinite article in Heine and Kuteva’s work The Basque indefinite article bat is discussed or mentioned in four works by Heine and Kuteva (2003: 556–557; 2005: 101, 247; 2006: 30, 132, 246;

The Basque articles -a and bat and recent contact theories

253

2007: 327). Here I will only o¤er some relevant quotations, since the ideas presented in each work do not di¤er substantially. 4.2.3.1. The indefinite article bat as a replica feature Heine and Kuteva take for granted that the indefinite article was acquired via contact, although we have no evidence of a time when Basque lacked such a category. ‘‘As a result of this contact, Basque speakers introduced a category which they did not have previously, namely an indefinite article.’’ (Heine and Kuteva 2003: 556; emphasis mine – JM) We can find other claims in the same vein in their work (Heine and Kuteva 2005: 247). In their most recent work (Heine and Kuteva 2007: 327) they write: [I]n the earlier history of the Basque language there was no indefinite article, while the surrounding Romance languages Spanish, French, and Gascon had indefinite articles. As a result of centuries of close contact with these Romance languages, speakers of Basque grammaticalized their numeral for ‘one,’ bat, to an indefinite article. . . . as Haase (1992) demonstrates, it was only one out of a large number of instances of grammatical replication that Basque speakers introduced on the model of their dominant Romance neighbor languages. . .

This explicitly denies the existence of an indefinite article in the earlier history of Basque. Of course it depends on how we understand the term ‘‘history,’’ but as I have shown in sections 3.2 and 3.3, they are strictly speaking not correct: bat as an indefinite article appears in all Basque historical records. There are no extensive records of Basque for the time when it allegedly lacked an indefinite article; again, we simply do not know when it emerged in Basque. As a possibility, as plausible as any other, we should conjecture that Basque had an indefinite article prior to contact with Romance languages. Of course, we know (see section 2.2) that languages do not tend to have only the indefinite article, so one hypothesis based on cross-linguistic tendencies is that Basque did not have an indefinite article before the definite one emerged, allegedly in the Middle Ages. But we must distinguish between what we conclude on the basis of our theoretical assumptions from what we know for sure on the basis of actual data. We should also keep in mind that languages like Turkish are an exception to this tendency, especially since Turkish seems to be close to Basque in typological terms (Comrie 2008).

254

Julen Manterola

In short, we cannot draw any conclusions as to whether Basque had an indefinite article before the Middle Ages by asking which possibility best fits a given hypothesis. This is first of all an empirical problem, one of the lack of relevant data, or rather of how to make good use of the data at our disposal. 4.2.3.2. The gradual (and ‘‘delayed’’) grammaticalization process in a replica language Heine and Kuteva also focus on the lesser degree of grammaticalization that Basque bat shows as compared to Romance un ‘a.’ That Basque bat is less grammaticalized fits with their expectations about the relative chronology of model and replica features. As they write (2003: 556–557): The grammaticalization of indefinite articles normally proceeds along the following main stages. . . . While the French indefinite article has gone essentially through all these stages, the Basque indefinite article has not. . . . While there are incipient uses as a non-specific marker as early as 1545, the grammaticalization as a non-specific article is clearly a recent innovation of Basque. . . . it has not reached the same degree of grammaticalization as e.g. the corresponding French article. . .

The only relevant passage from their 2005 work expresses essentially the same idea (Heine and Kuteva 2005: 101): . . . and the indefinite article of Basque, replicated on the model of Romance languages, exhibit properties of categories in the early stages of grammaticalization. . . . They thus di¤er from the corresponding categories in the model languages, which both are fully grammaticalized articles.

I do not believe that the reason the Basque indefinite article is less developed than in the alleged model languages is exclusively that it is a replica category. Consider this example of present-day central Basque: (9) a.

Eneko gizon on-a da. Eneko man good-the is ‘Eneko is a good man.’

b.

??Eneko gizon on bat da. Eneko man good one is

In its Romance equivalents we find most typically Eneko est un bon homme or Eneko es un buen hombre, with the un indefinite article. It becomes clear that there may be other issues at stake: how did the spread of -a a¤ect the use and further grammaticalization of bat? The role that the remarkable spread of the definite article may have played in preventing the use of the indefinite bat should also be taken into consideration.

The Basque articles -a and bat and recent contact theories

255

At the same time, a mirroring phenomenon should be noted: bat can be used in certain constructions in some texts and varieties (my impression is that this mostly a¤ects eastern dialects). Sentences like Fantosma bat da ‘It is a phantom’ by Leizarraga (1571: 334, Mat. 14: 26) are good examples; however, it is also possible to find examples like the western Basque Lecu on bat da Escocia ‘Scotland is a good place’ in Pe´rez de Lazarraga (c. 1564: 1204r).11 The question is then whether this broader use of bat is related or not to the lesser use of -a in these dialects. This is a further point that lacks detailed study, and these conjectures all call for empirical testing. These questions are raised by internal facts about definiteness marking and the behavior of determiners in Basque; I believe that a general theory should also o¤er solutions to this kind of language-specific problem. In support of Heine and Kuteva I can refer back to their suggestion that a long contact period may have blurred the relative degree of grammaticalization of model and replica languages; but in that case we need to come back to the problems I sketched in 4.2.2. No clear solution can be found by retaining their arguments, at least for Basque. At the same time, we may ask what it means to not be strongly grammaticalized (in Heine’s 1997 terms), since Basque has a plural indefinite article batzu. The next section is devoted to this topic. 4.2.3.3. The plural indefinite article batzu and the theories of Heine and Kuteva In their 2006 work Heine and Kuteva mention the indefinite article bat three times (2006: 29, 132, 246); they discuss it in the same terms as before, 11. I owe these specific data to a reviewer. I would also like to note that example (9b) may indeed be a correct one in present-day central Basque given the appropriate context, and was most surely built on the basis of a particular Spanish model construction. As a first approach to the data, I would say that the phrase with -a in (9a) is not a referential one but rather some sort of kind reference; its most direct Spanish counterpart would be Eneko es buen hombre, not such a ‘‘good’’ sentence to me, especially when compared to Eneko es buen chico ‘Eneko is a good boy.’ The (9b) example would ideally stand for Spanish Eneko es un buen hombre (or maybe for Eneko es un hombre bueno?), a noun phrase with presumably a higher degree of referentiality and probably of emphatic expressiveness. The borderline between di¤erent readings is often fuzzy; the di¤erent readings of these constructions, together with the e¤ect of model constructions’ readings on replica constructions, are interesting aspects of a multi-faceted discussion I cannot address here.

256

Julen Manterola

still relying on Haase’s (1992) work. The bulk of my criticism of Haase in section 4.1 is relevant here and will not be repeated; however, one passage (2006: 246) calls for two further comments: ‘‘[B]ut the grammaticalization as a non-specific article is clearly a recent innovation of Basque. While the Basque article exhibits a high degree of grammaticalization, it is still less grammaticalized than its equivalents in the Romance model languages.’’ First, the data Heine and Kuteva o¤er to show that the grammaticalization of a non-specific article is a recent innovation are not conclusive. I do not mean that this is definitely not the case, but that Haase’s few examples are not enough. This is another topic for which a detailed analysis, based on as many texts as we have at our disposal, is still lacking. Second, what would it mean in Heine’s (1997) terms that the article in Basque is still less grammaticalized? I have repeatedly noted the existence of an ancient plural indefinite article batzu (section 3.2.1). I see two logical possibilities if we seek to retain Heine and Kuteva’s (2003, 2005, 2006, 2007) and Heine’s (1997) views on this issue. If we follow Heine’s grammaticalization scale for the indefinite article, especially what he says about the final stage, we have to admit that the Basque indefinite article bat was highly grammaticalized at an early period. If this were the case, the basis of Heine and Kuteva’s work would be weakened, as it would seem that the usual grammaticalization path does not apply in this specific case. The second logical possibility is that Heine’s scale is not correct, since the Basque indefinite article bat has not advanced very far along its grammaticalization path, yet it has an ancient plural batzu. What is striking, of course, is that bat as an indefinite article seems to be grammaticalized and not grammaticalized at the same time. If we feel free to reject these authors’ theoretical proposals, we may decide that the grammaticalization of the indefinite article, triggered or not by a contact situation, has not followed the typical path proposed by Heine. In fact, as we have seen already (section 3.2.1), plural batzu could indeed have been formed on the model of Romance languages (at least Spanish); the fact that plural indefinite articles based on the numeral are rare (section 2.2), also makes us think that this is an areal feature. Many questions come to mind that will inevitably remain open. What do intensive contact situations mean for grammaticalization scales? We could perhaps answer that the model indefinite plural unos ‘some’ was so powerful that it made Basque bat ‘a’ skip over some stages in its grammaticalization path. We would then not need to reject Heine’s grammaticalization path, and we would open up a new line of research into

The Basque articles -a and bat and recent contact theories

257

the interaction between ‘‘natural’’ grammaticalization paths and contacta¤ected ones. This option could perfectly well be complementary to Heine and Kuteva’s proposals for grammaticalization and contact issues. From a more local perspective, we might also wonder about the relationship between the Spanish romances and present-day French Basque dialects, or what the internal relationship between western and eastern dialects has been, or what is the complementary relationship between definite -a and indefinite bat. General theories ought also to have something to say about these seemingly less important issues. 4.2.4. Concluding remarks on Heine and Kuteva’s hypothesis for Basque I have tried to show that Basque data as used by Heine and Kuteva do not o¤er solid support for any of their contact hypotheses. Nevertheless, I would like to make clear that contact has surely played a determining role in configuring the character and range of uses of articles in Basque. On the one hand, although I have not focused on this issue, the remarkable extension of the definite article -a could be explained by contact – not simply by the ‘‘typical’’ contact-induced grammaticalization of demonstratives, but rather by another e¤ect of contact, the spread of the overt marking of singular / plural morphology. On the other hand, it is true that the use of the indefinite article bat has been extended on the model of Romance languages, as Trask (2003: 122) notes: ‘‘The quantifier batzuk ‘some, several’ . . . is formally a plural of this bat. Among some younger speakers, there is a tendency to extend the use of bat to calque the much broader use of the Spanish article un(a).’’ This extension may in fact be a measurable contact-induced change, but recent work by Heine and Kuteva (with Haase as their basis) does not address this phenomenon. The fault is not entirely theirs, since Basque linguistics in general still has no in-depth study of the topic. Many interesting issues arise from this probably contact-induced extension; here I propose guesses about two of them. First, as mentioned above, the indefinite article’s extension might have been a¤ected by the widespread use of the definite article; this definite article is the one whose use has most evidently been spreading during the last four or five centuries when Basque as a whole is considered. Second, it might seem that the indefinite article on the model of Romance languages is more widespread in eastern than in central and western varieties of Basque. One could speculate that this is due to the lesser use that eastern varieties make of the definite article -a.

258

Julen Manterola

These possibilities, as I said, are merely guesses; further study is needed to corroborate or dismiss them. They are intended simply as a sample of the line of research we might follow, since they are not addressed by Haase.

5. Summary and concluding remarks Thus far, it has become clear that linguistic reality presents a much more complex, variegated situation than Haase’s and Heine and Kuteva’s proposals would lead us to imagine. I have shown that Heine and Kuteva’s hypotheses cannot hold for both Basque articles, definite and indefinite, together (section 4.2.2). Their basic problem comes from the fact that as far as Basque is concerned they rely exclusively on Haase’s work. There are many other sources against which Haase’s seminal work on contact should be checked in order to get a reliable analysis of diachronic, dialectal, and contact issues in Basque. Dialectal and historical data cannot be neglected; proposals arising from theoretical insights should help in understanding the history, dialectal variation, and distribution of language-specific features. But Heine and Kuteva leave many issues unaddressed and unexplained. 1. Their approach does not solve specific problems already identified in Basque diachronic linguistics, such as the early existence of the indefinite plural batzu (section 3.2.1). 2. It says nothing about contact issues present in the literature on Basque, such as the contact-induced emergence of the definite article -a. 3. There is no mention of the remarkable spread of the definite article -a, which seemingly could have been due to contact. 4. Interesting questions, such as the relationship between definite and indefinite articles in a contact situation, are not raised. This entirely new view of the issues relative to articles in Basque forces us to review Heine and Kuteva’s theoretical claims (section 4.2.1), insofar as facts about Basque were supposed to support them. 1. The relative degree of grammaticalization of parallel categories is not to be taken as the first approach to contact issues between two languages. If we want to establish their diachronic relationship, other questions have to be answered first.

The Basque articles -a and bat and recent contact theories

259

2. When diachronic information is not available we should be much more cautious; the example of Basque has shown that when we do have more and better information than Heine and Kuteva (considering both articles together), their predictions turn out to be incorrect. 3. Thus, their caveat in the passage I quoted in section 4.2.2 points in an interesting direction for further studies, focusing on the time dimension of these developments. 4. New questions are prompted by their hypotheses, such as how di¤erent contact e¤ects can each conceal the other’s typical diachronic paths (see section 3.1.2, on the role of the overt distinction of singular/plural marking). The question of how intensive contact a¤ects the replica language should also be a concern of the theory (see sections 3.2.1 and 4.2.3.3 on the possible contact origin of batzu). All these considerations, rather than completely ruling their hypotheses out, might help to improve and update them with what we empirically know about Basque. They are also a call for caution, given that we are dealing with a language with no decisive data in some of its aspects, and a reminder of how important it is to acquire a good knowledge of the history of the languages involved. At this point we may recall some of Thomason’s (2007) proposals for basic steps to be taken before a claim of contact-induced change can be considered firmly established. Her fourth and fifth steps (Thomason 2007: 11–12) are directly linked to my point about the importance of a good knowledge of the history of the languages involved. She notes that we need to prove that the proposed interference features did not exist in the receiving or replica language before it came into contact with the source language, which can be done either by inspecting documents showing earlier stages of the language or by examining related languages which can give us clues to the ancient mother language. And, of course, we need to prove that the relevant transferred features were already present in the source or model language by the time it came into contact with the receiving language. As far as I can judge, neither of these basic steps has been accomplished by Haase for Basque definite and indefinite articles, nor by Heine and Kuteva; ancient texts are not well analyzed, and variation between dialects (replacing related languages in the case of isolate Basque) has not been studied. These methodological gaps diminish the accuracy of Heine and Kuteva’s analyses of Basque contact issues. Furthermore, one would expect their

260

Julen Manterola

hypotheses to shed some light on language-specific problems, and theory and fact to strengthen each other. This has not been the case; Basque has been shown not to be a good starting point for their hypotheses as far as articles are concerned. Nonetheless, Heine and Kuteva’s hypotheses have been seen to be testable against new data; their theory may be strengthened as these data are gradually fed into it.

References ´ lvarez, J. L. 1977. L’emploi de l’inde´fini en souletin. Fontes Linguae Vasconum A 25: 29–54. Artiagoitia, X. 1998. Determinatzaile sintagmaren hipotesia euskal gramatikan. Uztaro 27: 33–61. [Available at http://www.uztaro.com/] Artiagoitia, X. 2002. The functional structure of the Basque noun phrase. In Erramu Boneta: Festschrift for Rudolf P. G. de Rijk [ASJUren Gehigarriak XLIV], X. Artiagoitia, P. Goenaga and Lakarra J. A. (eds), 73–90. Bilbao: EHU. Azkue, R. M. 1905–1906. Diccionario vasco-espan˜ol-france´s. Bilbao: La Gran Enciclopedia Vasca. [Reprinted 1969] Azkue, R. M. 1923. Morfologı´a Vasca. Bilbao: La Gran Enciclopedia Vasca. [Reprinted 1969] [Available at http://www.euskaltzaindia.com/euskera] Chen, P. 2004. Identifiability and definiteness in Chinese. Linguistics 42.6: 1129– 1184. Company, C. 1991. La extensio´n del artı´culo en el espan˜ol medieval. Romance Philology XLIV. 4: 402–424. Comrie, B. 2008. Basque, Romance, and areal typology: what do we learn from the World Atlas of Language Structures? In Lenguas en dia´logo. El iberorromance y su diversidad lingu¨´ıstica y literaria. Ensayos en homenaje a Georg Bossong, H.-J. Do¨hla, R. Montero, and F. Ba´ez de Aguilar (eds), 55–63. Madrid: Iberoamericana. Cyr, D. 1993. Cross-linguistic quantification: definite articles vs. demonstratives. Language Sciences 15.3: 195–229. Egiategi, J. 1785. Lehen liburia edo filosofo huskaldunaren ekheia [Euskararen Lekukoak 6]. Bilbao: Euskaltzaindia. [Reprinted 1983] Epstein, R. 1993. The later stages in the development of the definite article: evidence from French. In Historical linguistics [Current Issues in Linguistic Theory 124], H. Andersen (ed.), 159–175. Amsterdam: John Benjamins. Epstein, R. 1994. The development of the definite article in French. In Perspectives on Grammaticalization [Current Issues in Linguistic Theory 109], W. Pagliuca (ed.), 63–80. Amsterdam: John Benjamins. Epstein, R. 1995. L’article de´fini en ancien franc¸ais: l’expression de la subjectivite´. Langue Franc¸aise 107: 58–71.

The Basque articles -a and bat and recent contact theories

261

Etxeberria, U. 2005. Quantification and domain restriction in Basque. PhD Dissertation, University of the Basque Country. [Available at www.ehu.es/hitt/ etxeberria.htm] Frajzyngier, Z. 1996. On sources of demonstratives and anaphors. In Studies in Anaphora [Typological Studies in Language 33], B. Fox (ed.), 167–203. Amsterdam: John Benjamins. Givo´n, T. 1981. On the development of the numeral ‘‘one’’ as an indefinite marker. Folia Linguistica Historica 2.1: 35–53. Greenberg, J. H. 1978. How does a language acquire gender markers? In Universals of Human Languages, J. H. Greenberg, J. Ferguson, and E. Moravcsik (eds), 47–82. Stanford: Stanford University Press. Haase, M. 1992. Sprachkontakt und Sprachwandel im Baskenland: Die Einflu¨sse des Gaskognischen und Franzo¨sischen auf das Baskische. Hamburg: Helmut Buske Verlag. Haspelmath, M. 1998. How young is standard average European? Language Sciences 20.3: 271–287. Heine, B. 1997. Cognitive Foundations of Grammar. New York/Oxford: Oxford University Press. Heine, B. and Kuteva, T. 2003. On contact-induced grammaticalization. Studies in Language 27.3: 529–572. Heine, B. and Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Heine, B. and Kuteva, T. 2006. The Changing Languages of Europe. Oxford: Oxford University Press. Heine, B. and Kuteva, T. 2007. The Genesis of Grammar. A Reconstruction [Studies in the evolution of language 9]. Oxford: Oxford University Press. Himmelmann, N. P. 1998. Regularity in irregularity: article use in adpositional phrases. Linguistic Typology 2: 315–353. Himmelmann, N. P. 2001. Articles. In Language Typology and Language Universals: An International Handbook [Handbooks of Linguistics and Communication Science 20.1], M. Haspelmath, E. Ko¨nig, W. Oesterreicher, and W. Raible (eds), 831–841. Berlin/New York: Walter de Gruyter. Hualde, J. I. 2003. Case and number inflection of noun phrases. In A grammar of Basque, J. I. Hualde and J. Ortiz de Urbina (eds), 171–186. Berlin: Mouton de Gruyter. Hualde, J. I. and Gaminde, I. 1998. Vowel interaction in Basque: a nearly exhaustive catalogue. Studies in the Linguistic Sciences 28.1: 41–77. [Available at https://netfiles.uiuc.edu/jihualde/www/V_sequences.pdf ] Hualde, J. I. and Ortiz de Urbina, J. (eds). 2003. A grammar of Basque. Berlin: Mouton de Gruyter. Huang, S. 1999. The emergence of a grammatical category definite article in spoken Chinese. Journal of Pragmatics 31: 77–94.

262

Julen Manterola

Iglesias, H. 2007. Observations concernant les re´centes critiques et omissions de Joseba Lakarra a` propos des recherches d’He´ctor Iglesias sur la proble´matique ‘‘basco-ibe´rique’’ suivies d’une hypothe`se ine´dite concernant l’inscription de Liria. [Available at http://artxiker.ccsd.cnrs.fr/] Irigoien, A. 1985. Euskarazko izen sintagma mugatzailerik gabekoez. Euskera 30: 129–139. [Available at http://www.euskaltzaindia.net/euskera] Irigoyen, A. 1957. Cartas de Mariano Mendigacha a D. Resurreccio´n Marı´a de Azkue. Euskera 2: 119–170. [Available at http://www.euskaltzaindia.com/ euskera] Irigoyen, A. 1986. En torno a la toponimia vasca y circumpirenaica. Bilbao: Universidad de Deusto. Iturrioz, J. L. 1985. La funcio´n de -a y de -ta a la luz de la dimensio´n de individuacio´n. Euskera 30: 175–213. [Available at http://www.euskaltzaindia.net/euskera] Lafitte, P. 1944. Grammaire Basque (Navarro-Labourdin litte´raire). Donostia, Bayonne: Elkar. Lafon, R. 1954. Le nombre dans la de´clinaison basque. Via Domitia 1: 111–121. [Reprinted in Lafon 1999: 209–217] Lafon, R. 1999. Vasconiana [IKER 11]. Bilbao: Euskaltzaindia. Lapesa, R. 1961. Del demostrativo al artı´culo. Nueva Revista de Filologı´a Hispa´nica 15: 23–44. Leizarraga, I. 1571. I. Leic¸arragas Baskische Bu¨cher von 1571, herausgegeben von Th. Linschmann und H. Schuchardt. Bilbao: Euskaltzaindia. [Reprinted 1990] Manterola, J. 2006. -a euskal artikulua definituaren gainean zenbait ohar. In Studies in Basque and Historical Linguistics in Memory of R. L. Trask [Anuario del Seminario de Filologı´a Vasca ‘‘Julio de Urquijo’’ XL], J. A. Lakarra and Hualde J. I. (eds), 651–676. Donostia, Bilbao: Gipuzkoako Foru Aldundia. Michelena, L. 1970. Nombre y verbo en la etimologı´a vasca. Fontes Linguae Vasconum 2: 67–93. [Reprinted in Michelena 1987: 283–309] Michelena, L. 1971. Toponimia, le´xico y grama´tica. Fontes Linguae Vasconum 9: 241–267. [Reprinted in Michelena 1987: 141–167] Michelena, L. 1978. Miscela´nea filolo´gica vasca. Fontes Linguae Vasconum 29: 205–228. [Reprinted in Michelena 1987: 363–385] Michelena, L. 1981. Lengua comu´n y dialectos vascos. Anuario del Seminario de Filologı´a Vasca ‘‘Julio de Urquijo’’ 15: 291–313. [Reprinted in Michelena 1987: 35–55] Michelena, L. 1987. Palabras y textos. Bilbao: EHU. Milsark, G. L. 1977. Toward an explanation of certain peculiarities of the existential construction in English. Linguistic Analysis 3.1: 1–29. Otsibar, K. 2003. Lehenago Gamere. Maule: Su¨ Azia. [Available at www.suazia. com] Pagola, R. M. (ed.). 2004. Bonaparte Ondareko eskuizkribuak. Bilduma osoaren Edizio Digitala. Bilbao: Deustuko Unibertsitatea. [Available at http://tesitek.com]

The Basque articles -a and bat and recent contact theories

263

Pe´rez de Lazarraga, J. C. 1564. Manuscript. [http://lazarraga.gipuzkoakultura.net] Plank, F. and Moravcsik, E. A. 1996. The Maltese article: language-particulars and universals. Rivista di Linguistica 8.1: 183–212. Putzu, I. and Ramat, P. 2001. Articles and quantifiers in the Mediterranean languages: a typological-diachronic analysis. In Aspects of Typology and Universals [Studia Typologica 1], W. Bisang (ed.), 99–132. Berlin: Akademie Verlag. Tartas, I. 1666. Onsa hilceco bidia. Orthez: Iacques Rovyer. [Available at http:// klasikoak.armiarma.com/] Thomason, S. G. 2007. Social and linguistic factors as predictors of contactinduced change. Paper presented at the Symposium on Language Contact and the Dynamics of Language: Theory and Implications, Leipzig, 10–13 May 2007. Trask, R. L. 1997. The history of Basque. London: Routledge. Trask, R. L. 2003. The Noun Phrase: nouns, determiners and modifiers; pronouns and names. In A grammar of Basque, J. I. Hualde and J. Ortiz de Urbina (eds), 113–170. Berlin: Mouton de Gruyter.

Contact phenomena/code copying in Indian Ocean Creoles: the post-abolition period1 Sibylle Kriegel In this article, I will show that some minor di¤erences in the syntax of closely related Mauritian and Seychelles Creole clearly can be interpreted as contact-induced phenomena due to di¤erent contact situation after the abolition of slavery in 1835. In order to explain these contact induced phenomena, I will work with the concept of code copying developed by Johanson (2002).

1. Socio-historical background Mauritian and Seychelles Creole are mutually intelligible despite minor di¤erences between the two languages. To explain this fact, I begin this article with a brief socio-historical survey. I do not retrace here the debate between Chaudenson (1974, 2003, etc.) and Baker and Corne (1982, 1987; see also Baker 2007) with respect to the exact conditions of creolization in the Indian Ocean; I only briefly summarize the comparable migration movements up to the abolition of slavery in 1835, before focusing on the di¤erent contact situations after 1835. Following a series of settlement disputes, Mauritius was ultimately settled by the French in 1721. According to Baker (1982), the population groups imported to Mauritius up to 1740 came from Madagascar, India, Benin, 1. Most parts of this article were written during a stay at the Freiburg Institute for Advanced Studies (FRIAS) where I was invited as an external fellow in November 2008. I am very grateful to the colleagues of the FRIAS and the University of Freiburg, Peter Auer, Daniel Jacob, Stefan Pfa¨nder, and Wolfgang Raible for discussion of the main points of this article. I am also very grateful to Paula Prescod, Ralph Ludwig, and Fabiola Henri. Ralph Ludwig and Fabiola Henri are my co-authors on two articles dealing with language contact between Mauritian Creole and Bhojpuri (see references). Some of the data discussed in these articles as well as a yet to be published corpus collected during fieldwork in 2005 with Ralph Ludwig and Fabiola Henri largely inspired section 3 of this article.

266

Sibylle Kriegel

Senegambia, and Mozambique. Between 1740 and 1835, Baker (1982: 51) states, the vast majority of arrivals were slaves from two regions, Madagascar and East Africa, in spite of the great diversity of places from which Mauritius was settled. As far back as 1773, the existence of a Creole language is attested: Un jeune Ne´grillon Mozambique, nomme´ Favori, aˆge´ de 13 ans, appartenant au Sr. Pierre Maheas, habitant a` la Montagne Longue, a disparu depuis le 31 Janvier. Comme ce jeune noir sest probablement e´gare´ & qu’il n’entend pas la langue cre´ole, il n’aura pu dire le nom de son maıˆtre ni retrouver sa maison. (1773, ‘‘Annonces, a‰ches et avis divers pour les colonies des isles de France’’) A young male slave from Mozambique called Favori, aged 13 and belonging to Pierre Maheas, a planter at Long Mountain, has disappeared since January 31st. As this young slave is probably lost and does not understand the Creole language, he will not be able to give the name of his owner nor find his way home . . . (translation by Baker and Mu¨hlha¨usler 2007: 85).

The uninhabited islands of the Seychelles were settled in 1770 by the French, mainly from Mauritius, but also from Reunion (Bolle´e 2007b). The settlers and their slaves imported what was already considered a Creole language into this new subcolony, which continued to be ruled from Mauritius. Until the abolition of slavery in 1835 both territories evolved in almost the same contact situation, although they both came under the rule of Britain during the Napoleonic Wars in 1814. However, after abolition in 1835, the demographic situations in the two territories developed di¤erently (Kriegel 2008). In Mauritius, in order to compensate for the lack of labor for the sugar industry, indentured laborers from the Indian subcontinent were imported and quickly became the dominant population group. According to Baker (1982), as early as 1871 they formed 68 percent of the population. This trade lasted until the beginning of the twentieth century, and today the descendants of Indian indentured workers still form the majority of the population in Mauritius. These indentured laborers were mostly speakers of Bhojpuri2 or related Indic languages (Neerputh 1986: 9¤., see also 2. We will use the name Bhojpuri because the speakers of this variety themselves call their language by this name, even if, according to linguistic and geographic criteria, this does not seem to be entirely justified. In this vein, Baker and Ramnah (1988: 67) state: ‘‘In view of our findings that Magahi was a major contributor to MB [Mauritian Bhojpuri, SK], ‘‘Bhojpuri’’ would seem to be less than ideal as the choice of name. A more suitable alternative, on both linguistic and geographic grounds, might be ‘‘Mauritian Bihari’’ (Baker and Ramnah 1988: 67).

Contact phenomena/code copying in Indian Ocean Creoles

267

Mesthrie 1991: 26). Bhojpuri is still spoken in Mauritius: according to an o‰cial census in 2000 it is the first home language of 12.07 percent of the total population (Kriegel et al. 2008), and less recent figures indicate that approximately a third of the population speaks Bhojpuri (alongside Creole). The Seychelles received ‘‘rescued’’ slaves in the second half of the nineteenth century (see Bolle´e 1977; Chaudenson 1974; Nwulia 1981), when the English liberated people found on ships in the Indian Ocean destined for the illegal slave trade. According to Baker they were predominantly speakers of Bantu languages and formed a third of the population in the Seychelles in the late nineteenth century (Baker 1993: 130). However, the use of their languages has been lost in the Seychelles, where Creole is today the main spoken language. Today, both territories are independent countries belonging to the British Commonwealth. Mauritius became independent in 1968 and was declared a republic in 1992. Even though English is the o‰cial language, its use is very restricted. Mauritius is multilingual, with Mauritian Creole as the main language spoken at home followed by Bhojpuri (see above; for a detailed account see Kriegel et al. 2009). In the Seychelles, three languages were made o‰cial in 1978, two years after independence. Since 1981, Creole (Kreol Seselwa) has been the first o‰cial language, followed by English and French. Creole is the native language of about 95 percent of the population (Michaelis 2008). Nowadays, even if in both countries we are witnessing an increased use of Creole in formal contexts as a consequence of independence, it must be stressed that the use of Creole in formal contexts is much more common in the Seychelles than in Mauritius (see Kriegel 2008). Recent research (Bolle´e 2007a and b; Kriegel 2008; Michaelis 2008) tends to present Seychelles Creole as a continuation of stable varieties of Mauritian Creole. The varieties are extremely close, but there are some minor di¤erences between them, above all in the lexicon (see e.g. Chaudenson 1974: 448). More recent research has also focused on slight di¤erences in the field of morphosyntax (see e.g. Bolle´e 2004; Kriegel 1996). For instance, the di¤erent encoding of the passive voice in both varieties has been interpreted elsewhere as a direct consequence of the increased written use of Seychelles Creole during the post-independence years. This article examines two grammatical features which will be interpreted as a consequence of the di¤erent language contact situations in which the two varieties have evolved since the abolition of slavery in 1835.

268

Sibylle Kriegel

2. The tools of contact linguistics Creoles play a predominant role in contact linguistics because they are considered the outcome of extreme contact in special sociolinguistic settings. Like pidgins and bilingual mixed languages, they are examples of language genesis situations. Although Creoles are the object of my analysis, I will not address the issue of genesis; rather, I am interested in the evolution of already existing Creole languages in new contact situations. In addition to situations of language genesis, Thomason and Kaufman (1988; see also Thomason 2001, etc. and Winford 2003) subdivide contact patterns into two further types: language maintenance, where a language is maintained and influenced by another language, and language shift, concerned with the death of a language that only leaves traces in another language. The cases I will analyze are situations of language maintenance, where extant Creole languages are influenced by other languages. In the case of Mauritius this influence came from the Indic language Bhojpuri, and in the case of the Seychelles from Bantu languages, di‰cult though this is to prove. Situations of language maintenance are typically related to processes of borrowing. On Thomason and Kaufman’s influential borrowing scale, one end is characterized by casual contact, where only non-basic vocabulary is borrowed, whereas with growing intensity of contact structural elements are also borrowed. Here I am interested in some function words between lexicon and grammar that the French linguist Meillet called ‘‘des petits mots a` valeur grammaticale.’’ They are the adposition depi of IndoMauritian Creole varieties and the complementizer pourdir of Seychelles Creole. According to Thomason and Kaufman (1988: 74), function words are borrowed in situations of ‘‘more intense contact’’ (Phase 3 of their scale). Given the heterogeneous terminology in the English and French literature for the definition of terms like borrowing and calquing, I will base my considerations on the concept of code copying developed by Johanson (e.g. 2002). Johanson distinguishes between selective and global copying. Other authors make similar distinctions, for instance Sakel (2007; see also Matras and Sakel 2007) speaks of MAT borrowing (relating to matter) and PAT borrowing (relating to pattern). PAT describes the case in which only a pattern from one language is replicated in another language. In MAT borrowing, morphological material and its phonological shape are replicated. In most cases the function of the borrowed element is also adopted, that is, MAT and PAT are combined. Heine and Kuteva (2005) are

Contact phenomena/code copying in Indian Ocean Creoles

269

almost exclusively interested in PAT borrowing, which they call ‘‘grammatical replication’’ (see also Chamoreau 2007). Stolz and Stolz (1996) make a similar distinction when they speak of overt copies (overte Kopien) where phonological material is concerned (this would correspond to combined MAT and PAT borrowing) and of covert copies (koverte Kopien) where the material side is not concerned (corresponding to PAT). The examples discussed in this paper refer to PAT borrowing. Following Stolz and Stolz (1996), I will speak of covert copies.

3. The data Our corpus contains diachronic texts as well as synchronic spoken and written data from Mauritian and Seychelles Creoles. Particular attention is paid to modern corpus data, which are sometimes complemented by elicited examples. The first written texts in Mauritian Creole date from the late eighteenth century and, to a greater extent, the beginning of the nineteenth century (Baker and Fon Sing (eds) 2007, Furlong and Ramharai 2006, Chaudenson 1981). The first written data for Seychelles Creole are from the late nineteenth century (Stein 2007) and the first text of significant length dates from the first third of the twentieth century (Young 1983). In this paper, I analyze some examples from the old texts Philip Baker circulated in an electronic version to the contributors to Baker and Fon Sing (2007). The spoken data in my corpus of modern Mauritian Creole were collected in fieldwork and contain the spoken, spontaneous texts in Kriegel (1996) and Ludwig et al. (2001) as well as the unpublished corpus by Kriegel, Ludwig, and Henri (2005). The spoken data from Seychelles Creole mainly draw on Bolle´e and Rosalie (1994). The modern written data imclude a wide variety of written texts in di¤erent genres. I will examine two function words: some uses of the French-derived preposition depi in Indo-Mauritian Creole varieties (Section 3.1) and the complementizer pourdir in Seychelles Creole (Section 3.2). I argue that in both cases we are dealing with phenomena involving code copying. 3.1. Depi: Path marking in Indo-Mauritian Creole varieties Depi is derived from the French preposition depuis. Its temporal uses, attested in all varieties of Mauritian Creole, can easily be explained by French where depuis has exactly the same temporal uses.

270

Sibylle Kriegel

(1) Mauritian Creole ou ’nn konn enn sanzman ki ’nn, ki ’nn 2sg compl know indf change rel compl rel compl koul pei la net depi lindepandans drown country def completely since independence ‘[It’s as if ] you had realized a change which, which completely ruined the country since Independence.’ (unpublished corpus Kriegel, Ludwig, and Henri 2005) (2) Mauritian Creole li ’nn ale li ’nn ale depi lontan wi 3sg compl go 3sg compl go since long time yes ‘He went away, he went away a long time ago.’ (Kriegel 1996) Alongside those temporal uses we find, in Indo-Mauritian Creole varieties, depi in path marking (Talmy 1985), more precisely in ablative marking (motion from) as in (3). Although local uses of depuis in French are attested (Kriegel et al. 2008), the much more productive use of depi in local contexts of Mauritian Creole cannot be explained by French semantics. (3) Indo-Mauritian Creole mo papa sort depi Sesel, li ’nn vini poss father come.from abl seychelles 3sg compl come pou travay dan Moris for work Loc Mauritius ‘My father comes from the Seychelles, he came to work in Mauritius. . . .’ (Bord la Mer 1980) Before analyzing the data in more detail, a closer look at the theoretical background on which I base my considerations may be useful (see Kriegel et al. 2008). According to Lehmann (1992), a local situation presents the following structure: a moving or located object (the Figure in Gestalt psychology), is involved in a situation and locally related or oriented with respect to a local region of a reference object, the Ground. An important distinction is made between the local region or place of the Ground and the orientation or path (Talmy 1985). The following types of path can be distinguished: essive (at rest, ‘‘to be at’’), allative (motion to, ‘‘to go to’’) and ablative (motion from, ‘‘to come from’’).3 In this article we will 3. Like Michaelis (2008: 238), I understand the notions of path as semantic categories and not as morpho-syntactic language-specific cases. Therefore they are represented in small capitals.

Contact phenomena/code copying in Indian Ocean Creoles

271

be concerned with the coding of the ablative (and allative) relation in intransitive movement.4 In most varieties of Mauritian Creole we have the construction type common to a lot of Creole languages, including Seychelles Creole, in which ablative and allative are not marked di¤erently (Michaelis 2008: 239). Example (4) from Seychelles Creole is an exact match with example (5) from Mauritian Creole. (4) Seychelles Creole Ablative mon sorti dan lafore 1sg.sbj come.from loc.in forest ‘I come out of the forest.’ Allative mon al dan lafore 1sg.sbj go loc.in forest ‘I go into the forest.’ (5) Mauritian Creole Ablative mo sorti dan lafore 1sg.sbj come.from loc.in forest ‘I come out of the forest.’ (Kriegel et al. 2008: 175) Allative mo al dan lafore 1sg.sbj go loc.in forest ‘I go into the forest.’ The unmarked expression of ablative and allative in Mauritian Creole is exactly the same as in Seychelles Creole. Michaelis (2008) argues convincingly that the pattern of ablative coding, which seems strange from a European perspective, ‘‘clearly mirrors the Eastern Bantu pattern in that allative and ablative are not marked di¤erently.’’ The element dan refers to the ‘‘local region’’ or ‘‘place’’ of the Ground and not to path. Path, ‘‘the orientation with respect to,’’ is typically coded by a preposition or by case in European languages. More specifically, French uses de for ablative as opposed to the unmarked coding of allative. In Seychelles Creole and most varieties of Mauritian Creole, path is exclusively coded in the semantics of the verb sortir. 4. Also see Kriegel et al. (2008) for Mauritian Creole. Michaelis (2008, section 6) analyzes data from Seychelles Creole and Eastern Bantu languages and gives a visual presentation of the structure of a local situation following Lehmann (1992) and Jackendo¤ (1983: 161¤.).

272

Sibylle Kriegel

However, here I will deal with the marked expression type of ablative, attested in Indo-Mauritian Creole varieties. Alongside examples of type (3) drawn from a spoken register, we also find examples in written registers of Mauritian Creole. The extensive use of depi in the political writings of the political party Lalit (7)–(9) is striking and could be interpreted as an indication of its propagation to other varieties of Mauritian Creole. (6) Mauritian Creole . . . ti ena enn vie diksioner ek enn vie liv gramer pst have Indf old dictionary and indf old book grammar Angle ki li ti amen ar li depi lot-pey. English rel 3sg pst bring with 3SG.OBJ abl other country ‘There was an old dictionary and an old English grammar he brought with him from another country.’ (TIZISTWAR 1, Dev Virahsawmy, http://pages.intnet.mu/develog/) (7) Mauritian Creole Fode pa zot gayn sa kas-la depi dan pos Modal neg 3.pl.sbj get dem cash abl loc.in pocket klas capitalist class capitalist ‘They should not get this money from the pocket of the capitalist class.’ (Lalit 24 May 2008, Akimilasyon capital, article by Rosa Luxemburg, translated into Mauritian Creole, http://lalitmauritius.org) Depi in this use, marking the ablative, is not attested in closely related Seychelles Creole nor in any other French-based Creole. The combined use of depi with dan in (7) clearly shows that depi is used in the path expression while dan marks the local region or Ground. Like ablative markers of other languages, depi may also be used in a range of non-concrete spatial (or temporal) functions referring to source in a more abstract sense, as in (8)–(9). (8) Mauritian Creole Li pa kapav tini enn sanglo 3sg.sbj neg can prevent indf sob ki sorti depi profonder so nam. rel come.out abl depth poss soul ‘She couldn’t help but let out a sob from the depth of her soul.’ (TIZISTWAR 1, Dev Virahsawmy, http://pages.intnet.mu/develog/)

Contact phenomena/code copying in Indian Ocean Creoles

273

(9) Mauritian Creole Li sibir presyon depi institisyon kuma FMI, kuma 3sg.sbj su¤er pressure abl institution like FMI like Labank Mondyal, WTO. world bank WTO (FAS A KRIZ SISTEMIK, FAS A POLITIK BURZWA KI STRATEZI? by Diskur Ram Seegobin, Jean-Claude Bibi, Oupa Lehulere, Lalit 27/07/2007, http://lalitmauritius.org) ‘He is subjected to the pressure of institutions like FMI, like the World Bank, WTO.’ These more abstract uses seem to be limited to written registers. Given the high token frequency of this construction type, for instance in the texts published by Lalit, it is reasonable to claim that the use of depi by some writers is a conscious strategy to copy the preposition de coding ablative in French, which was lost during creolization. But this is certainly not the case for the concrete spatial contexts in which we find uses of depi, as in (3) or (6)–(7). These uses of depi in concrete contexts of ablative coding are already attested in old texts. The first attestations are from the 1880s, the ‘‘critical’’ period when the majority of the population became of Indian origin. (10) Old Mauritian Creole Lher la foul conne c¸a, zot sivre´ li a pie´ when def crowd know this 3pl follow 3sg.obj by foot dipi tou zot la vil. abl all 3pl.poss town ‘The people heard about it, they followed him on foot from the towns.’ (Matthew, 14: 13, translation by Anderson 1885) (11) Old Mauritian Creole soley va vine noar, la line na pa va sun fut become black moon neg fut donne so clarte´, e´ ze´toal va tombe´ dipi dan le´ ciel . . . give poss clearness and star fut fall abl loc heaven ‘Soon after the trouble of those days, the sun will grow dark, the moon will no longer shine, the stars will fall from heaven. . .’ (Matthew 24: 29, translation by Anderson 1885, electronic corpus Baker)

274

Sibylle Kriegel

Another interesting phenomenon is the use of depi to encode not only the ablative but also the allative in patterns where local points of departure and arrival are expressed. Here depi is used instead of ziska to mark the endpoint of a movement. The first attestation is from 1880. (12) Old Mauritian Creole Mais so cloisons lac¸ambe la` napas dibois napas plances: but poss partition room def neg wood neg planks dipis en haut, dipis en bas toute loison ne`que e´ne grand abl top all bottom all partition just indf big grand laglace meˆme. big mirror ‘But the partition in his/her room is not made of wood or boards: from top to bottom the partition is a big, big mirror.’ (Baissac 1880: 56) This confusion between depi and ziska is still rather common in IndoMauritian Creole varieties, and Baissac makes the following comment on this rare phenomenon in his 1880 grammar: Depuis, dipis. Depuis ici jusque-la`, Dipis ici zousqua`-la`; mais le cre´ole disait avant qu’il connuˆt zousqua` ou zisqua`, jusque, au lieu de: J’ai saute´ depuis ici jusque la`, Mo te´ saute dipis la`, dipis-la`, ce qui e´tait plus original. (Baissac 1880: 78) From, dipis. From here to there, Dipis ici zousqua`-la`; but the Creole would say before he knew zousqu’a` or zisqua`, ‘to’, instead of: I jumped from here to there, Mo te´ saute´ dipis la`, dipis-la`, which was in fact more original.

In the modern Indo-Mauritian Creole variety, we also have examples of the following type: (13) Indo-Mauritian Creole Depi sannmars depi lagar ena trafik abl champs.de.mars all station aux tra‰c ‘From the Champs de Mars up to the station, the tra‰c is jammed.’ (unpublished corpus Kriegel, Ludwig, and Henri 2005) (14) Indo-Mauritian Creole Depi lao depi anba ena bokou pou marse abl top all bottom aux much to walk ‘[To go] from top to bottom, there’s a lot of walking to do.’ (unpublished corpus Kriegel, Ludwig, and Henri 2005)

Contact phenomena/code copying in Indian Ocean Creoles

275

As in (12), in example (13) and (14) not only is the ablative marked by depi but so too is the allative, a construction type which is impossible in all French varieties. In French and also in the Creole variety spoken by speakers without Indo-Mauritian background, the allative (motion to) is normally marked by ziska ( jusqu’a`) as is the case in (15) and (16). (15) Mauritian Creole Depi Vakwa ziska Maybour Pol inn dormi dan loto abl vacoas all Mahe´bourg Paul compl dormir loc voiture ‘Paul has been sleeping in the car from Vacoas to Mahe´bourg.’ (unpublished corpus Kriegel, Ludwig, and Henri 2005) (16) Mauritian Creole Li ’nn get mwa depi lao ziska anba. 3sg compl look.at 1sg.obj abl top all bottom ‘He looked at me from top to bottom.’ (¼from head to toe) (unpublished corpus Kriegel, Ludwig, and Henri 2005) The confusion between depi and ziska in Mauritius is associated with ‘‘people who come from homes where Bhojpuri was the principal language during their childhood’’ (Baker 1996: 48), a statement which our informants confirmed during fieldwork carried out in 2005 with Ralph Ludwig and Fabiola Henri. The ‘‘construction material’’ is from French, but depi cannot be explained by French in most of its local uses, especially when combined with dan or when encoding ablative and allative. Instead an examination of Bhojpuri may be of help. We argue that the generalization of depi as an ablative and sometimes even as an allative marker may be explained by influence from Bhojpuri (see also Kriegel et al. 2009). In all Bhojpuri varieties the highly frequent and polyvalent postposition se is attested (see Baker and Ramnah 1988, Mesthrie 1991: 262, Shukla 1981: 161). The similarity between the two elements can be observed in example (17). (17) a.

b.

Mauritian Bhojpuri Ham ghar se awa thain 1sg house abl come aux.asp.1sg Mauritian Creole Mo pe vini depi lakaz 1sg asp come abl house ‘I’m coming from the house.’ (unpublished corpus Kriegel, Ludwig, and Henri 2005)

276

Sibylle Kriegel

In cases expressing the local beginning and endpoint we have the following examples in Bhojpuri: (18) Mauritian Bhojpuri Ronpwen se lagar (le) trafik ba roundabout abl station (all) tra‰c aux ‘From the roundabout up to the station, the tra‰c is jammed.’ (unpublished corpus Kriegel, Ludwig, and Henri 2005) The use of le is optional and according to our informants, it is often confused with se. Se also has temporal uses, exactly like depi. I believe that depi is a copy from se. Furthermore, I posit that it is a covert copy, or an instance of PAT borrowing, following the term used by Sakel (2007). The phonological shape does not come from Bhojpuri but from Creole, the copying code. The copy is also ‘‘selective’’ (Johanson 2002) in another way: in Bhojpuri se is a postposition, but it appears as a preposition following the word order rules of the copying code. But the fact that it is indeed a copy is supported by sociolinguistic evidence. 3.2. Pourdir: complementizing in Seychelles Creole My second example comes from Seychelles Creole and is based on corpus analysis. In some corpora of Seychelles Creole, especially in the oral texts in Bolle´e and Rosalie (1994) and a corpus of folk tales published in Kriegel and Neumann-Holzschuh (2007), we often find, alongside ø marking and ki from French que, the complementizer pourdir, a lexicalization of the French preposition pour in combination with the infinitive of the verb dire. Examples (19)–(22) illustrate the three types of coding. Ø-complementation is highly frequent in spoken registers. (19) Seychelles Creole Dizef kot nou pa gannyen nou dir nou a manz tou le zour egg where 1pl neg have 1pl say ø 1pl fut eat everyday ‘When people do not have any eggs they say they could eat eggs every day.’ (Bolle´e and Rosalie 1994: 202) Complementizing by French derived ki is more frequent in written registers, but also attested in spoken data.

Contact phenomena/code copying in Indian Ocean Creoles

277

(20) Seychelles Creole Ou ’n deza dir mwan ki ou deza travay lo 2sg compl already say 1sg.obj that 2sg already work on en serten zil indf certain island ‘You have already told me that you worked on some island or another.’ (Bolle´e and Rosalie 1994: 216–7) In this article we will focus on the third and least frequent technique of complementizing, the use of pourdir.5 (21) Seychelles Creole Me Zozef-Fou ti konmans rakont bann serviter but Joseph-Fou pst start tell pl servant pourdir kaptenn ki ’n zet Tizan dan delo comp captain rel compl throw Tijean in water ‘But Joseph-Fou began to tell the servants that it was the captain who had thrown Tijean into the water.’ (Kriegel and Neumann-Holzschuh 2007) (22) Seychelles Creole Dimoun lontan ti per, ti kwar pourdir i people long time pst fear pst believe compl 3sg annan en bonnfanm ki apel Bonnfanm San Tet . . . have indf woman that call woman-without-head . . . ‘A long time ago people were afraid, they believed that there was a woman called Woman-without-head.’ (Bolle´e and Rosalie 1994: 266/7) According to the corpus data and based on the interviews conducted during fieldwork in the Seychelles in 2003, only elderly people make a wider use of pourdir as a complementizer ((21)–(22)), especially in spoken discourse. It appears with certain verb groups (verbs of utterance, knowing, believing, and perception), and especially in the corpus of spoken texts by Bolle´e and Rosalie (1994) in contexts where the speaker is doubting the factual nature of the utterance (see Kriegel 2004, 2008). This nuance seems to get 5. While Bolle´e (1977: 84) is the first linguist to note the existence of pourdir in a complementizing function, Chaudenson (2003: 380¤.) still questions the existence of pourdir as a complementizer in Seychelles Creole.

278

Sibylle Kriegel

lost in more recent written data: semantic bleaching seems to have taken place (for a discussion in the context of grammaticalization theory, see Kriegel 2008). Interestingly, pourdir in Seychelles Creole has another use: it also serves as a downtoning particle. This use corresponds to French pour ainsi dire (so to speak/as such). An analysis of the Estrie corpus revealed that in overseas varieties of French we even find pour dire without ainsi in a modalizing function (see also Be´lisle 1979: 754). (23) Quebecan French on fait absolument rien / les feˆtes pour nous / c¸a veut pas dire grand#chose pour dire ‘We do not do anything at all, celebrations for us, they do not mean much, as such.’ (Estrie corpus) This use of pourdir in a modalizing function is attested in the dictionary of De St. Jorre and Lionnet (1999: 240) and in Chaudenson (2003: 383). We can also find some examples in corpora of spoken language, as in (24). (24) Seychelles Creole me pour dir fer en louvraz metye non but modalizer do indf work job neg ‘But you cannot really say that I had a job as such.’ (Ludwig et al. 2001: 259) Such uses as a modalizer or downtoning particle are also attested in other French-based Creole languages, as for instance in Dominican and Guadeloupean French Creole (see Ludwig et al. 2002, poudi ‘c’est le cas de dire,’ so to speak/as such). (25) Dominican Creole Mem si mon pa ni twavay poudi mon ka twavay even if 1sg neg have work modalizer 1sg ipfv work ‘Even though I don’t have a job which could really be called a job as such. . .’ (elicited from Shelly-Ann Meade) However, pourdir as a complementizer is only possible in Seychelles Creole. In this function it does not exist in any other Indian Ocean French Creole, nor in the other French Creoles. Whereas an explanation of pourdir as modalizer by French influence seems su‰cient, it is not plausible to explain the complementizer function of pourdir as a result of French influence.

Contact phenomena/code copying in Indian Ocean Creoles

279

The grammaticalization of verbs of saying is a very common grammaticalization path in languages worldwide (Heine and Kuteva 2002; Ebert 1991). In Creole Studies, the complementizers derived from verbs of saying (above all in English-based Creoles) are a very popular example for substrate influence from West African languages (see e.g. Bruyn 2003 for Sranan, Parkvall 2000: 66 for a list of varieties, Mufwene 2008 for a discussion). After closer examination, we realize that in these languages we are concerned with a finite form of say, whereas in Seychelles Creole we are dealing with a lexicalization of the French preposition pour and the infinitive of the verb dire. Grammaticalizations of this type are also known from di¤erent language families, including di¤erent Bantu languages. Gilman (1993) was the first linguist to point out the possible relationship between the use of pourdir as a complementizer in Seychelles Creole and Bantu languages. I will take an example from Swahili, one of the languages that may have been spoken by the late nineteenth-century immigrants to the Seychelles. Note, however, that the pattern is also attested in other Eastern Bantu languages (see Kriegel 2008; Gu¨ldemann 2002). (26) Swahili a-li-sem-a kw-amb-a a-ta-kw-end-a. he-past-say-fin to-say-fin he-irr-to-go-fin ‘He said he would go.’ (Perrot 1951: 166, cited in Gilman 1993: 52) Kw (or ku before consonants) is an infinitive prefix also used to mark finality. So a possible explanation for pourdir in a complementizing function lies in the fact that during the second half of the nineteenth century only the Seychelles received large numbers of speakers of Bantu languages (see section 1), even if from a methodological point of view this is very di‰cult or impossible to prove. Pourdir could also be a much older copy dating from the period of creolization: it could have come into Seychelles Creole via Mauritian Creole. However, pourdir as a complementizer is not attested in Mauritian Creole, either old or modern varieties. Of course, this can be due to coincidence, the element simply not being attested in the relatively sparse data. Nonetheless, I think that it is safer to assume that pourdir is a copy which came into Seychelles Creole in the second half of the nineteenth century as a consequence of the immigration of speakers of Bantu languages. Leaving aside the discussion surrounding the likely period when the complementizer pourdir was copied into Creole, I think that we have, as for depi, a covert copy into Creole. The evolution

280

Sibylle Kriegel

of pourdir in Seychelles Creole could be interpreted as a case of convergence (Bolle´e 1982; Kriegel 2003) between a French pragmatic strategy of modalizing and a Bantu complementizer.

4. Conclusion and perspectives Let us briefly summarize the main points: Mauritian and Seychelles Creole are two closely related French-based Creoles. The varieties are mutually intelligible. The extant Mauritian Creole was exported to the Seychelles before the end of the eighteenth century. The minor di¤erences between the two languages concern, above all, the lexicon and some aspects of syntax on which research has not really focused until now. I have tried to show some contact-induced phenomena concerning function words: I have examined the ablative marking adposition depi in Mauritian Creole, which I interpret as a covert copy from Bhojpuri, as well as the complementizer pourdir in Seychelles Creole, which I interpret as a covert copy from Bantu languages. The concept of code copying, and the analysis of copies which do not come from the base language that provided the majority of the ‘‘construction materials’’ in the respective Creoles, enable us to understand the challenges posed by the notion of linguistic change in Creole languages. When we deal with elements that come from the base languages we are very often unable to judge if the existence of certain phenomena dates from the period of creolization or if we are seeing more recent copies. The absence of data from the period of constitution of Creoles and the sparse data in their later evolution do not permit of a definite answer. This problem has been discussed at length in Kriegel (2008). A good candidate to illustrate the problem of determining when code copying occurred is pourdir, because we have to distinguish between two periods of contact with Bantu languages. Oral varieties of French as well as Bantu languages have contributed to the constitution of Mauritian and Seychelles Creole (roughly before 1773); after the abolition of slavery in 1835, the Seychelles (though not Mauritius) received significant numbers of Bantu-speaking immigrants. It is therefore di‰cult to determine at which moment of the evolution of Seychelles Creole the element in question, in our case pourdir, was copied into Creole. The analysis of data from Mauritian Creole, a Creole that is evolving in a new contact situation with a language that did not contribute to its constitution (Bhojpuri), sheds new light on an often neglected topic in Creole studies: the indeterminacy related to dating

Contact phenomena/code copying in Indian Ocean Creoles

281

the copy of elements into the Creoles. This problem is even more pronounced when we are dealing with elements from the base language.

References Baissac, C. 1880. Etude sur le patois cre´ole mauricien. Baker, P. 2007. Elements of a sociolinguistic history of Mauritius and its Creole (to 1968). In The Making of Mauritian Creole. Analyses diachroniques a` partir des textes anciens, P. Baker and G. Fon Sing (eds), 133–156. London: Westminster Creolistics series, vol. 9, Battlebridge Publications. Baker, P. 1996. On the development of certain prepositional forms in Mauritian and other French Creoles. In Mate´riaux pour l’e´tude des classes grammaticales dans les langues creoles, D. Ve´ronique (ed.), 41–59. Aix-en-Provence: PUP. Baker, P. 1993. Assessing the African Contribution to French-Based Creoles. In Africanisms in Afro-American Language Varieties, S. Mufwene (ed.), 123–155. New York: The University of Georgia Press. Baker, P. 1982. The contribution of non-Francophone immigrants to the lexicon of Mauritian Creole. PhD dissertation, School of Oriental and African Studies, University of London. Baker, P. and Corne, C. 1987. Histoire sociale et cre´olisation a` La Re´union et a` Maurice. Revue que´be´coise de linguistique the´orique et applique´e 6.2: 71–87. Baker, P. and Corne, C. 1982. Isle de France Creole: A‰nities and Origins. Ann Arbor: Karoma. Baker, P. and Fon Sing, G. (eds). 2007. The Making of Mauritian Creole. Analyses diachroniques a` partir des textes anciens. London: Westminster Creolistics series, vol. 9, Battlebridge Publications. Baker, P. and Mu¨hlha¨usler, P. 2007. Creole Linguistics from its Beginnings, through Schuchardt to the Present Day. In Creolization. History, Ethnography, Theory, C. Stewart (ed.), 84–107. Walnut Creek, CA: Left Coast Press. Baker, P. and Ramnah, A. 1988. Recognizing Mauritian Bhojpuri. In Language Transplanted. The Development of Overseas Hindi, R. K. Barz and J. Siegel (eds), 45–65. Wiesbaden: Otto Harrassowitz. Be´lisle, L. A. 1989. Dictionnaire nord-ame´ricain de la langue franc¸aise, Laval, Quebec: Beauchemin. Bolle´e, A. 2007a. Beitra¨ge zur Kreolistik. Hamburg: Kreolische Bibliothek Band 21, Helmut Buske Verlag. Bolle´e, A. 2007b. Les textes religieux de Philippe Caulier C. M. et l’histoire des cre´oles de l’oce´an Indien. In P. Brasseur and D. Ve´ronique (eds), Mondes cre´oles et francophones. Me´langes o¤erts a` Robert Chaudenson, 61–72. Paris: L’Harmattan. Bolle´e, A. 2004. Le de´veloppement du de´monstratif dans les cre´oles de l’oce´an Indien. Creolica. http://www.creolica.net/bollee2.pdf. Bolle´e, A. 1982. Die Rolle der Konvergenz bei der Kreolisierung. In Die Leistung der Strataforschung und der Kreolistik. Typologische Aspekte der Sprachkon-

282

Sibylle Kriegel

takte. Akten des 5. Symposions u¨ber Sprachkontakt in Europa, Mannheim 1982, P. S. Ureland (ed.), 391–405. Tu¨bingen: Niemeyer. Bolle´e, A. 1977. Le cre´ole franc¸ais des Seychelles. Esquisse d’une grammaire – textes – vocabulaire. Tu¨bingen: Niemeyer. Bruyn, A. 2003. Grammaticalisation, re´analyse et influence substratique: quelques cas du sranan. In Grammaticalisation et re´analyse: Approches de la variation cre´ole et franc¸aise, S. Kriegel (ed.), 25–48. Paris: CNRS Editions, Langage. Chamoreau, C. 2007. Grammatical borrowing in Purepecha. In Grammatical Borrowing in Cross-linguistic Perspective, Y. Matras and J. Sakel (eds), 465– 480. Berlin: Mouton de Gruyter. Chaudenson, R. 2003. La cre´olisation: the´orie, applications, implications. Paris: Langues et De´veloppement, L’Harmattan. Chaudenson, R. 1981. Textes cre´oles anciens (La Re´union et Ile Maurice). Hamburg: Buske, Kreolische Reihe. Chaudenson, R. 1974. Le lexique du parler cre´ole de la Re´union, 2 volumes. Paris: Champion. De St. Jorre, D. and Lionnet, G. 1999. Diksyonner kreol – franse. Dictionnaire cre´ole seychellois – franc¸ais. Bamberg/Mahe´: Imprimerie Difo-Druck. Ebert, K. H. 1991. Vom Verbum dicendi zur Konjunktion – Ein Kapitel der Grammatikentwicklung. In Von Europa bis Ozeanien – von der Antonymie zum Relativsatz, W. Bisang and P. Rinderknecht (eds), 77–95. Zu¨rich: University of Zu¨rich. Furlong, R. and Ramharai, V. (eds). 2006. Panorama de la litte´rature mauricienne. La production cre´olophone. Volume 1: des origines a` l’inde´pendance. Mauritius: TIMAM. Gilman, C. 1993. A Bantu model for the seychellois ‘pour-dire’ complementizer. In Atlantic meets Pacific, F. Byrne and J. Holm (eds), 49–55. Amsterdam: John Benjamins. Gu¨ldemann, T. 2002. When ‘say’ is not say. The functional versatility of the Bantu quotative marker ti with special reference to Shona. In Reported Discourse. A Meeting Ground for Di¤erent Linguistic Domains, T. Gu¨ldemann and M. von Roncador (eds), 253–288. Amsterdam: John Benjamins. Haspelmath, M., Dryer, M. S., Gil, D., and Comrie, B. (eds). 2005. The World Atlas of Language Structures. Oxford: Oxford University Press. Heine, B. and Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Heine, B. and Kuteva, T. 2002. World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Jackendo¤, R. 1983. Semantics and Cognition. Cambridge, MA: MIT Press. Johanson, L. 2002. Contact-induced change in a code-copying framework. In Language Change. The Interplay of Internal, External and Extra-Linguistic Factors, M. C. Jones and E. Esch (eds), 285–313. Berlin/New York: Mouton de Gruyter. Kriegel, S. 2008. Cre´oles entre lexique et syntaxe: contact et changement. Document de synthe`se pour l’obtention de l’Habilitation a` diriger des Recherches, Aix-en-Provence.

Contact phenomena/code copying in Indian Ocean Creoles

283

Kriegel, S. 2004. Juste pour dire que pourdir existe toujours. Creolica, 16 December 2004, http://www.creolica.net. Kriegel, S. (ed.). 2003. Grammaticalisation et re´analyse: Approches de la variation cre´ole et franc¸aise. Paris: CNRS Editions, Langage. Kriegel, S. 2003. Introduction: Vers une interpre´tation multicausale du changement linguistique. In Grammaticalisation et re´analyse: Approches de la variation cre´ole et franc¸aise, S. Kriegel (ed.), 7–21. Paris: CNRS Editions, Langage. Kriegel, S. 1996. Diathesen im Mauritius – und Seychellenkreol. Tu¨bingen: Gunter Narr, ScriptOralia. Kriegel, S., Ludwig, R., and Henri, F. 2009. Les rapports entre cre´ole et bhojpouri a` Maurice: contact de langues et actes identitaires. In Multiple Identities in Action. Mauritius and Some Antillean Parallelisms, V. Hookoomsing, R. Ludwig and B. Schnepel (eds), 203–252. Frankfurt am Main: Peter Lang. Kriegel, S., Ludwig, R., and Henri, F. 2008. Encoding PATH in Mauritian Creole and Bhojpuri: problems concerning language contact. In Roots of Creole Structures. Weighing the Contribution of Substrates and Superstrates, S. Michaelis (ed.), 169–196. Amsterdam: John Benjamins. Lehmann, C. 1992. Yukatekische lokale Relatoren in typologischer Perspektive. Zeitschrift fu¨r Phonetik, Sprachwissenschaft und Kommunikationsforschung 45: 626–41. Ludwig, R., Montbrand, D., Poullet, H., and Telchid, S. (1990/ 2002). Dictionnaire cre´ole, 2, revised edition. Paris: SERVEDIT–Maisonneuve and Larose– E´ditions Jasor. Ludwig, R., Telchid, S., and Bruneau-Ludwig, F. (in collaboration with Pfa¨nder, S. and de Robillard, D.) (eds). 2001. Corpus cre´ole. Textes oraux dominicais, guadeloupe´ens, guyanais, haı¨tiens, mauriciens et seychellois. Hamburg: Buske, Kreolische Bibliothek. Matras, Y. and Sakel, J. (eds). 2007. Grammatical Borrowing in Cross-Linguistic Perspective. Berlin: Mouton de Gruyter. Meillet, A. [1921] (1982). Linguistique historique et linguistique ge´ne´rale. Geneva/ Paris: Slatkine/Champion. Mesthrie, R. 1991. Language in Indenture. A Sociolinguistic History of BhojpuriHindi in South Africa. Johannesburg: Witwatersrand University Press. Michaelis, S. (ed.). 2008. Roots of Creole Structures. Weighing the Contribution of Substrates and Superstrates. Amsterdam: John Benjamins. Michaelis, S. 2008. Valency patterns in Seychelles Creole: Where do they come from? In Roots of Creole Structures. Weighing the Contribution of Substrates and Superstrates. S. Michaelis (ed.), 225–251. Amsterdam: John Benjamins. Mufwene, S. S. 2008. Language Evolution. Contact, Competition and Change. London/New York: Continuum International Publishing Group. Neerputh, N. C. 1986. Le syste`me verbal du Bhojpuri de L’Ile Maurice. Paris: L’Harmattan. Nwulia, M. D. E. 1981. The History of Slavery in Mauritius and the Seychelles, 1810–1875. Madison, NJ: Fairleigh Dickinson University Press. Parkvall, M. 2000. Out of Africa. African influences in Atlantic Creoles. London: Battlebridge Publications.

284

Sibylle Kriegel

Sakel, J. 2007. Types of loan: matter and pattern. In Grammatical Borrowing in Cross-Linguistic Perspective, Y. Matras and J. Sakel (eds), 15–29. Berlin: Mouton de Gruyter. Shukla, S. 1981. Bhojpuri Grammar. Washington, DC: Georgetown University Press. Stein, P. 2007. Le cre´ole Seychellois en 1882: les textes re´colte´s par Hugo Schuchardt. In Mondes cre´oles et francophones. Me´langes o¤erts a` Robert Chaudenson, P. Brasseur and D. Ve´ronique (eds), 129–140. Paris: L’Harmattan. Stolz, C. and Stolz, T. 1996. Funktionswortentlehnung in Mesoamerika: spanischamerindianischer Sprachkontakt (Hispanoindiana II). Sprachtypologie und Universalien-forschung (STUF) 49.1: 86–123. Talmy, L. 1985. Lexicalization patterns: semantic structure in lexical forms. In Grammatical Categories and the Lexicon, Volume 3, T. Shopen (ed.), 57–149. Cambridge: Cambridge University Press. Thomason, S. G. 2001. Language Contact. An Introduction. Edinburgh: Edinburgh University Press. Thomason, S. G. and Kaufman, T. 1988. Language Contact, Creolization, and Genetic Linguistics. Los Angeles: University of California Press. Winford, D. 2003. An Introduction to Contact Linguistics. Oxford: Blackwell Publishing.

Corpora6 Baissac, C. 1880. Etude sur le patois cre´ole mauricien. Nancy: Imp. Berger-Levrault. Bord la Mer. 1980. Port Louis: Port Louis Harbor and Dock Worker Union. Bolle´e, A. and Rosalie, M. 1994. Parol ek memwar. Re´cits de vie des Seychelles. Hamburg: Buske. Corpus Estrie: Martel, P. and Beauchemin, N. Echantillon de textes libres, Sherbrooke: Recherches sociolinguistiques dans la re´gion de Sherbrooke. Kriegel, S. 1996. Diathesen im Mauritius und im Seychellenkreol. Tu¨bingen: Gunter Narr Verlag, ScriptOralia. Kriegel, S., Ludwig, R., and Henri, F. 2005. Six hours recorded in Mauritius. Kriegel, S. and Neumann-Holzschuh, I. 2007. Siranndann! Zanbaget! Contes cre´oles des Seychelles. Creolica, 52 p., http://www.creolica.net/. Virahsawmy, D. http://pages.intnet.mu/develog/. Young, R. 1983. Fables de La Fontaine traduites en cre´ole seychellois. Hamburg: Buske, Kreolische Bibliothek.

6. Some of the corpora are also cited in the references, where we refer to theoretical information and to examples.

Grammaticalization of modal auxiliary verbs in Pima Bajo: an internal or a contact-induced change? 1 Zarina Estrada-Ferna´ndez 1. Introduction Second position clitics are traditionally taken to be auxiliaries responsible for encoding information related to verbs, such as tense-aspect-mood, or information related to the topic (subject) of a sentence, that is, person and number or agreement (Steele 1999). The second position clitics have been widely discussed in the linguistic literature as the Wackernagel’s position, since such clitics usually appear after the first word or constituent in a sentence. Moreover, for some languages, such as Serbo-Croatian, second position clitics are considered to be verbs. The linguistic discussions relative to this kind of element deal mostly with the morphological aspects, although syntacticians also consider them; among them, Anderson (2005: 9) proposes that many auxiliary verb constructions have their origins in complex predicate constructions. Takic and Tepiman Uto-Aztecan languages from the southwest US and northwest Mexico, including Cupen˜o (Hill 2005), Luisen˜o (Steele 1981), and O’odham (Saxton 1982), are recognized as having second position clitics, most of them functioning as auxiliary verbs.2 A closer comparative approach to the study of second position clitics in Tepiman languages (Estrada 2005, 2007) shows that for some languages of this Uto-Aztecan branch (such as O’odham) second position clitics are obligatory, while in 1. I gratefully acknowledge the valuable comments of Søren Wichmann, Marianne Mithun, Daniel Hintz, and Frank Seifart on various versions of this manuscript. Of course, the responsibility for any errors is entirely mine. Financial support for the research reported here was provided by Conacyt. 2. According to Steele (1999: 49), ‘‘auxiliary is a term that names elements that bear resemblance to verbs in both their morphology and position.’’ Anderson (2005: 4) considers an auxiliary verb ‘‘to be an item on the lexical verb-functional a‰x continuum, which tends to be at least somewhat semantically bleached, and grammaticalized to express one or more of a range of salient verbal categories, most typically aspectual and modal categories.’’ See Heine (1993) for a similar definition.

286

Zarina Estrada-Ferna´ndez

others (Northern and Southern Tepehuan as well as Nevome, an extinct variety of Pima Bajo) they are optional. The scenario for Pima Bajo is quite distinct: on the one hand, the second position clitics in the language are restricted to mood, that is, the imperative markers, ¼in ‘sg.imperative’ and ¼var ‘pl.imperative,’ or to person and number pronouns in dependent clauses only, but on the other hand, the language has modal auxiliary verbs. In this respect Pima Bajo is quite di¤erent from other languages of the Tepiman branch, although, as I will show, some evidence of auxiliary verbs is sporadically observed in other modern varieties of Uto-Aztecan languages – Yaqui, Tarahumara, and Nahuatl. This article addresses the topic of modal auxiliary verbs in Pima Bajo as well as the possible trajectories of change that have given rise to the appearance of these verbs in the grammar of this language. The analysis presented here suggests two hypotheses: that internal reconstruction is central to the analysis of the evolution of a particular element or category within a language, and that language contact must be considered as a possible influence which might contribute to making a particular change viable. A necessary background for the analysis of modal auxiliary verbs in Pima Bajo is the description of inter-clausal relations, in particular of verbal complements, since it is in this domain that auxiliary verbs appear in this language.3 The study of verbal complements among the Uto-Aztecan languages of northwestern Mexico helps us to understand two possible directions in the evolution of auxiliary verbs. One occurs in languages with a strong tendency to be polysynthetic, as I will argue is the case for Yaqui, and another occurs in languages that show a strong tendency to be analytical, which seems to be the case in Pima Bajo. The study of verbal complements, as I will demonstrate, is relevant to our understanding of the development of modal auxiliary verbs as the result of a typological change in languages such as Pima Bajo, but in this case the typological change may have been influenced by language contact. The phenomenon is not a direct case of language change caused by contact, but it may be an instance of the indirect typological change described by Heine and Kuteva (2008: 218) as ‘‘language-contact phenomena working in conspiracy with grammaticalization.’’ Similar cases are observed when languages with particular kinds of morphosyntactic properties are in contact with languages 3. According to Givo´n (2001), modal auxiliary verbs must be considered part of a semantic continuum which also includes manipulative, cognitive, and utterance verbs.

Grammaticalization of modal auxiliary verbs in Pima Bajo

287

with distinct morphosyntactic profiles. The phenomenon is complex, since it involves not only the emergence of auxiliary verbs but also the unstable occurrence of what has been described as the second position clitic or auxiliaries in some Uto-Aztecan languages from the Takic and Tepiman branches (cf. Steele 1981; Hill 2005). For the purposes of the present article, I adopt Anderson’s (2005) typology of Auxiliary Complex Constructions (ACC) and his suggestion that some Uto-Aztecan languages are doubly inflected – aux-headed and lex-headed – while others are lex-headed only.4 Based on Anderson’s typology, I will show that the information conveyed by some verbal complement constructions, and in particular the morphological information related to the predicate (tense-aspect-mood and person-number), has followed two distinct directions or pathways of grammaticalization.5 In certain Uto-Aztecan languages, including Cupen˜o (Hill 2005), Yaqui (Estrada and Buitimea 2009; Guerrero 2004), and Pima Bajo, both of those pathways – the aux-headed and the lex-headed – are present, whereas the lex-headed pattern is the only one available for some other Uto-Aztecan languages, such as Ute (Givo´n 1990). In the lex-headed pattern all the tense-aspect-mood information is encoded within the boundaries of the verb (su‰xed to it), while in the aux-headed pattern the inflectional morphology is only encoded in the auxiliary verb. In languages where both types of auxiliary complex constructions have developed, the auxheaded and the lex-headed, the influence of language contact with some Yuman languages in California and Arizona should be considered; later, contact with Spanish may have led to the development of bare modal auxiliary verbs such as those in Pima Bajo (see below). This article is organized as follows: Section 2 provides a description of verbal complements in Pima Bajo against the backdrop of other UtoAztecan languages. In Section 3 the analysis of verbal complements in 4. Anderson (2005: 23¤ ) defines five distinct types of auxiliary constructions or ‘‘macro-patterns’’: (i) the so called ‘aux-headed’ constructions, where only the auxiliary verb contains inflectional morphology; (ii) the ‘doubled’ auxiliary construction, where both the lexical verb and the auxiliary verb contain it, (iii) the ‘lex-headed’ auxiliary construction, where only the lexical verb is inflected; (iv) the ‘split’ auxiliary construction, where the information conveyed by the auxiliary and the lexical verb is divided among the lexical head and the auxiliary; and (v) the ‘split/doubled’ construction, which combines patterns (iii) and (iv). 5. Anderson (2005: 111) considers Tu¨batulabal and Serrano to follow the auxheaded pattern.

288

Zarina Estrada-Ferna´ndez

Nevome, an extinct variant of Pima Bajo, is addressed. Section 4 discusses the grammaticalization pathways for the modal auxiliary verbs in Pima Bajo and section 5 introduces a contact-induced change hypothesis. 2. Verbal complements in Pima Bajo and Uto-Aztecan Uto-Aztecan languages – in particular, Northern Uto-Aztecan languages like Ute, Cupen˜o, and Kawaiisu – have been characterized as languages in which most if not all of the subordinate clauses are nominalized (Givo´n 1990, 2006; Hill 2005; Zigmond et al. 1990). Observe in (1) an example of a verbal complement from Ute where the nominalization is accomplished by means of two nominal su‰xes, -na- and -’ay, and a dependent genitive subject, ’a´apa-ci ‘boy-gen.’ Both the control verb and the nominalized verb are inflected with the aspectual anterior su‰x -qa or -kaa, which could be an argument for a not fully nominalized verb. (1) Ute (Givo´n 1990: 288)6 mama´-ci ’u pucu´cugwa-qa woman-nom that/sbj know-ant ’a´apa-ci ’uwa´-y picu´˛-kaa-na-’a˛y boy-gen that/gen arrive-ant-nmlz-obj ‘The woman knew (anterior) that the boy had arrived.’ Examples from Cupen˜o in (2a–b) (Jane H. Hill personal communication), however, show a two-verb construction: a finite inflected control verb tul ‘to finish’ followed by half-nominalized verbal complements, where both of them, qine ‘to plough’ in (2a) and wal ‘to dig’ in (2b), bear person, number, and case inflections, and at the same time a nominalizer su‰x -’a in (2a) and -a in (2b). The auxiliary verb root -tul- ‘to finish’ in both (2a–b) is fully inflected. (2) a.

pe-tul-qali pe-qini-’a-y, me¼m¼pe 3sg-finish-ds.sg 3sg-plough-nmlz-obj, and¼3pl¼irr maan-pem-ngiy-pi leave-3pl-go.away-irr.sub ‘When he has finished plowing, they will let him go.’ (that is, no longer employ him)

6. For Ute, Cupen˜o, Kawaiisu, Tumbisa, Yaqui, O’odham (Papago), Tarahumara, Nahuatl, Nevome, and Huichol, I follow the glosses and translations provided by the authors.

Grammaticalization of modal auxiliary verbs in Pima Bajo

b.

289

tul-qa¼ne wal-ne-n-a-y temal finish-prs¼1sg.erg dig-1sg-in-nmlz-obj dirt ‘I finished digging.’7

The example in (3) from Kawaiisu (Zigmond et al. 1990), illustrates an instance where the main predicate is nominalized, with the control verb mee ‘to say,’ but the verb  uskwee ‘to go’ in the complement clause is not. The overall construction corresponds to a direct quotation. (3) n  mee-g-ka-d ¼m  ukkwee-n¼bn I say-ben-r-nmlz-¼you go-mom-ex ‘I said to you, Go!’ Nominalizations like those illustrated in (1–3) have been observed in Yaqui, a southern Uto-Aztecan language where all verbal complements, except for the auxiliary verb aa ‘to be able,’ as I will show later, are expressed by either morphologically complex verb constructions or by nominalized clauses. Nominalized clauses in Yaqui are characterized by having at least one nominal su‰x attached to the complement verb: the plural -m or the case su‰x -ta (both in 4a), a dependent subject encoded as an accusative (or genitive) pronoun, enchi ‘2sg.acc’ in (4a–b), and one of the subordinator su‰xes -’u or -m(e) respectively in (4b) and (4c). (4) a.

b.

Yaqui (Guerrero 2004) inepo Maria-ta enchi kuna-m-ta bicha-k 1sg.nom Maria-acc 2sg.acc marry-nmlz-acc see-pfv ‘I saw that Maria married you.’ Yaqui (Guerrero 2004) Peo kaba’i-m enchi jinu-ka-’u suale-n Peter.nom horse-pl 2sg.acc buy-pfv-comp believe-past ‘Peter believed that you bought the horses.’

c.

itepo wa’ame lu´’ute-m-me bicha-k 3sg.nom dem.acc finish-nmlz-nmlz see-pfv ‘We saw that they have finished.’

7. The abbreviation in is provided for a thematic su‰x, which according to Hill (2005) appears on most transitive verbs.

290

Zarina Estrada-Ferna´ndez

The examples in (5), by contrast, show morphologically complex verbs in Yaqui with verbal su‰xes, for example -tua causative, -pea desiderative, -i’a prospective.8 (5) a.

b.

aapo kutam a’abo¼nee tohi-tua-k 3sg.nom wood.pl loc ¼1sg.dat bring-caus-pfv ‘He made me bring wood.’ inepo kot-pea 1sg.nom sleep-des ‘I want to sleep.’

c. im¼ne enchi tawa-ı´’a here¼1sg.nom 2sg.acc remain-prosp ‘I want you to continue here.’ Nominalizations such as those illustrated in (1–4) are rarely observed in Pima Bajo, where verbs requiring a verbal complement are encoded with three distinct types of constructions: (i) a morphologically complex verb, as in (6), (ii) an analytical construction, as in (7), or (iii) an uninflected modal auxiliary, as in (8–10).9 (6) Hoan in-daad si’ a’as-tar John 1sg.nsbj-mother int laugh.pfv-caus ‘John made my mother laugh.’ In analytical constructions, the verbal complement is treated as an adjunct or peripheral argument. This is illustrated for a verb of perception in (7a), a mental predicate in (7b), and a causative verb in (7c). In all the examples in (7) the verbal complement is introduced by the subordinator ko followed by a person and number subject clitic: ¼(a)p ‘2sg.sbj’ in (7a), ¼at ‘1pl.sbj’ in (7b), or zero for the third singular person in (7c). 8. Unless otherwise noted, the data from Yaqui come from my own field work or from Estrada and Buitimea (2009). 9. As an exception, Pima Bajo also has a nominalized construction that is restricted to the verb lid ‘to think, to want, to like.’ The construction is somehow frozen or fossilized, since it is only associated to one single lexical item: aan oob no’ok in-lid 1sg.sbj Pima speak.pfv 1sg.nsbj-want.non.fin ‘I want to speak Pima’ / ‘My wanting is to speak Pima.’

Grammaticalization of modal auxiliary verbs in Pima Bajo

(7) a.

aan im vagmad ko¼p tud-an 1sg.sbj neg like.prs sub ¼2sg.sbj dance-irr ‘I don’t like you to dance.’

b.

Peier mat k¼at kav mua Peter know.pfv sub ¼3pl.sbj horse kill-pfv ‘Peter knew that we killed the horse.’

291

c. Peier tiah ko n’ir Peter make.pfv sub.3sg.sbj sing.pfv ‘Peter makes him sing.’ Verbal complements in (8–10) follow uninflected modal auxiliary verbs: apod ‘can,’ vutag ‘begin,’ tum ‘try.’ Following Anderson (2005), I consider auxiliary verbs such as apod ‘can,’ vutag ‘begin,’ tum ‘try,’ to be elements that contribute some grammatical or functional content to the construction, and the lexical verbs, as tuda ‘dance,’ mua’a ‘kill,’ or n’id ‘see,’ to be those that contribute the lexical content. Such auxiliary verbs in Pima Bajo, for example in (8–10), correspond to the class known as modal, equi, or subject control verbs. In Pima Bajo such verbs have di¤erent degrees of grammaticalization: some, as in (8), are now invariable uninflected modal verbs (that is, they bear no tam morphology and do not have their own set of arguments). (8) a.

b.

aan apod da’ad-a 1sg.sbj can fly-fut ‘I can fly.’ Huan vutag tkpan-ia John begin work-prob ‘John begins to work.’

c. li oob tum koi am kav-tam dim person try sleep.pfv loc horse-loc ‘The boy tried to sleep on the horse.’ Constructions with another type of modal auxiliary verb are provided in (9). Verbs such as sontag ‘to start,’ vuus ‘to finish,’ have been grammaticalized from adverbials or adjectives, but coexist in the grammar as both. The diachronic origin of verbs in (9) is as follows: sontag ‘to start’ (on a daily basis) < adv. sontag ‘early,’ in (9a); vuus ‘to finish’ < adv. vs ‘all,’ in (9b), and vutag ‘begin to’ (for the first time), ‘start to’ < vtag ‘new,’ in (9c).

292

Zarina Estrada-Ferna´ndez

(9) a.

b.

kkl sontag tkpan-a serrus-tam rdp.man start.st work-fut sawmill-loc ‘The men are starting to work at the sawmill.’ aan a¼vuus ni-va10 1sg.sbj unsp.obj ¼finish sing.compl ‘I just finished singing it.’

c. Mari iskueel-tam vutag dah Maria school-loc begin.st be.pfv ‘Maria begins to be at school.’ A third type of auxiliary verb is maat ‘to know’; this verb occurs in the language as both a modal auxiliary verb, in (10), and an independent main verb, in (11). (10) Huaan maat n’i John know sing.prs ‘John knows how to sing.’ (11) Marii in¼maat te’op-tam Maria 1sg.nsbj ¼know.pfv church-loc ‘Maria knew me at the church.’ Modal auxiliary verbs in Pima Bajo, like those illustrated in (8) to (10), have not yet received any attention in the linguistic literature about UtoAztecan languages. This class of verbs are considered here to be modal auxiliaries since they exhibit four characteristic properties: they constitute a closed class of verbs, that is, only a few of them have been attested; they express a modal meaning, since they convey some information concerning the mood of the main verb; they are subject-control verbs (equi or raising verbs); and they correspond to a class of verbs that, according to Heine (1993: 9) as well as Givo´n (1984), has been characterized as midway

10. In this construction, the clitic pronoun a¼ attaches to the left of the modal auxiliary verb, but is an argument of the lexical verb nia ‘sing.’ The clitic cannot intervene in the verbal sequence vuus niva.

Grammaticalization of modal auxiliary verbs in Pima Bajo

293

between auxiliary and main verbs.11 Heine (1993: 17) considers modal auxiliary verbs to be polysemous by nature, since usually they can function as both main and auxiliary verbs.12 Documentation of this class of verbs in southern Uto-Aztecan languages is scarce. In Yaqui, as mentioned earlier, only one modal auxiliary verb, aa ‘to be able,’ has been attested (Estrada et al. 2004; Guerrero 2004; Dedrick and Casad 1999). (12) jamuchim tajo’o-ta aa baksia women.nom cloth-acc be able wash.prs ‘Women can (know how to) wash the clothes.’ In other southern Uto-Aztecan languages modal auxiliary verbs are rarely described or even mentioned. An example from Tarahumara, in (13), is provided by Caballero (2006: 61–62), where the verb i¢ire ‘make,’ can have a modal interpretation, ‘being able to,’ and is attested as an uninflected verb. (13) a.

Tarahumara (Caballero, 2000: 61–62): Juane i¢ire me’a-re rowi Juan make kill-pfv rabbit ‘John was able to kill the rabbit.’

In a similar way, examples from Nahuatl in (14) containing the verb tamik are provided by Peralta (2005), who claims that this class of verbs has evolved in the language as a result of language contact with MixeZoquean languages. Observe that in this language, the lexical verbs gocˇi ‘to sleep’ and pa:k ‘to wash’ are, following Anderson (2005), lex-headed inflected verb forms, since they have person and number prefixes attached to them. (14) a.

Nahuatl (Peralta 2005) tamik ni-gocˇi-k v.aux 1sbj-sleep-pret ‘I just slept.’

11. For a full discussion of the list of properties of auxiliary verbs, see Heine (1993: 22–23). 12. Marchese (1986: 96) points this out for Kru languages: ‘‘when a verb takes on auxiliary characteristics, the verb from which it is derived does not cease to exist.’’

294

Zarina Estrada-Ferna´ndez

b.

tamik ni-k-pa:k mo-goton v.aux 1sbj-3obj-wash.pret 2poss-shirt ‘I just washed your shirt.’

Examples from Pima Bajo (8–10), Yaqui (12), Tarahumara (13), and Nahuatl (14) contrast with the one provided in (15) from Huichol (Go´mez 1999), a southern Uto-Aztecan language of the Cora-Chol branch. In this construction, the verb yua ‘to begin’ is inflected for mood, as in Ute example (1), but not for person and number, as in Cupen˜o example (2). (15) Huichol (Go´mez 1999: 41) muwa niu ta m-e-ta-yua he-i-kwa-ne-t loc quot part as-inv-am-begin inv-3sg.obj-eat-pgr-simss he-i’-in-ne-t inv-3sg.obj-cut-pgr-simss ‘It is said that (he) began to eat them (the fruits) cutting (them) from the tree.’ For other northern Uto-Aztecan, as for example Timbisha (McLaughlin 2006) and Kawaiisu (Zigmond et al. 1990), there is almost no documentation of auxiliary verbs, since modality is provided in those languages by su‰xes on the main verb. Examples of morphologically complex verbs from Timbisha (Shoshoni) and Kawaiisu are given in (16) and (17) respectively; the examples show the modal verb su‰xed to the main verb. (16) a.

Timbisha (McLaughlin 2006: 42) sutn tkka-h/ttki 3sg.dem.nom eat-start ‘He started to eat.’

b.

Timbisha (McLaughlin 2006: 43) sutn nukkwi-sua 3sg.dem.nom run-want ‘He wants to run.’

(17) a.

Kawaiisu (Zigmond et al. 1991: 97) ka-ga a-na¼ina  vi rdp-eat-start¼his13 now ‘He is starting to eat now.’

13. Zigmond et al. (1991) gloss the su‰x -na- which translates as ‘start’ as cmp, that is, as a complementizer.

Grammaticalization of modal auxiliary verbs in Pima Bajo

b.

295

Kawaiisu (Zigmond et al. 1991: 97) ta nipizi ka a-sbi-ga-di wanipi-a man eat-want14-exht-nmlz mush-acc ‘The man wants to eat the pin˜on mush.’

(16) and (17) illustrate morphologically complex verbs such as those provided earlier for Yaqui in (5) and for Pima Bajo in (6). Aspectual or modal verbs illustrated so far for Yaqui, Pima Bajo, Timbisha (Shoshoni), and Kawaiisu constitute a morphologically complex class of verbs. These elements di¤er from the auxiliary verbs in Pima Bajo provided in (8–10) and (12–14); the latter occur in analytical constructions and correspond to auxiliary verbs. In this section I have shown that Uto-Aztecan languages have distinct possibilities for encoding verbal complements: (i) nominalized constructions for Ute, Cupen˜o, and Kawaiisu, examples (1–3), and Yaqui (4); (ii) morphologically complex predicates in Yaqui, example (5), Pima Bajo (6), Timbisha (16), and Kawaiisu (17); (iii) analytical periphrastic constructions, that is, modal auxiliary verbs, for Pima Bajo, examples (8–10), Yaqui (12), and Tarahumara (13). Verbal complements may also be encoded as in (iv), the least integrated biclausal construction or adjunct-like construction, as was illustrated for Pima Bajo, example (7). No claim is made here for Huichol and Nahuatl, since at least for the latter language the analysis of such constructions will only be possible after considering the MixeZoquean languages. However, a finer distinction must be provided for Cupen˜o (2), a language that marks person and number agreement on both the control or auxiliary verb and the lexical verb. Anderson (2005) uses this property to distinguish between Uto-Aztecan languages such as Pipil, which has double subject-marking, example (18), since the inflectional morphology occurs in both the auxiliary and the lexical verb, and other languages like Tu¨batulabal (19) and Serrano (20), which he considers to be aux-headed rather than lex-headed. Examples provided by Anderson in support of this claim are given below. (18) Pipil (Anderson 2005: 158 [Campbell 1985: 138]) ti-yu-t ti-yawi-t ti-pa:xa:lua-t ne:pa ka ku:htan 1pl-aux-pl 1pl-go-pl 1pl-walk-pl there in woods ‘We are going to go take a walk there in the woods.’ 14. Zigmond et al. (1991) gloss the su‰x -s bi which translates as ‘want’ as irr, that is, as irrealis.

296

Zarina Estrada-Ferna´ndez

(19) a.

b.

Tu¨batulabal (Uto-Aztecan, USA) (Anderson 2005: 111 [Voegelin 1935: 128–129]) ta’naha˙’-gilu˙’ts tı¨ ’ tı¨ ’k opt-1pl part eat ‘would we were eating.’ ih-ma’-ts tı¨ ’k here-hort-3 eat ‘let him eat here.’

(20) Serrano (Uto-Aztecan, USA) (Anderson 2005: 111 [Langacker 1977: 36]) kw’¼n k wa’a pot-1 eat ‘could I eat it.’ The typology of verbal complements described so far can only be explained, in accordance with Givo´n (2001, 2006), as di¤erent stages on the grammaticalization pathways available in these languages for verbal complements. However, two questions in connection with these data arise: Why are bare modal auxiliary verbs, as illustrated above for Yaqui, Pima Bajo, and Tarahumara, only attested in languages that have been in contact with Spanish over the last three hundred years? And why are languages from the Takic branch, in particular Serrano and Tu¨batulabal, the only aux-headed languages in the group? These questions will be considered in the next section after a brief presentation of the relevant data from Nevome, an older variety of Pima Bajo. The data from Nevome are particularly useful for our analysis because of the historical information they may provide.

3. Nevome as the ancestor of Pima Bajo The study of modal auxiliary verbs in Nevome, an extinct variety of Pima Bajo documented in the seventeenth century, is facilitated by the grammar published by Smith (1862), as well as later studies by Shaul (1982, 1986). The available data show that in Nevome most, if not all, verbal complements were encoded as morphologically complex predicates. This seems to suggest that we may be faced with languages with di¤erent morphological profiles, Nevome being more polysynthetic and Pima Bajo more analytic.

Grammaticalization of modal auxiliary verbs in Pima Bajo

297

The author of the seventeenth-century grammar of Nevome describes distinct infinitival verbs in this language (pp. 25–27), among them complementtaking verbs like muta ‘to want’ (a same subject verb) and orida ‘to want’ (a di¤erent subject verb); verbs of manipulation such as tani ‘to cause,’ or tuhanu ‘to make’; mental verbs such as urha ‘to think,’ ‘to imagine,’ or simatu ‘to know,’ and finally utterance verbs, such as aaga ‘to tell.’ Among these verbs only muta ‘to want’ has a modal interpretation, even though all of them share the same morphosyntactic properties.15 One such property is related to word order position: the examples in (21) show that the verbs are adjacent to each other, resembling a morphologically complex predicate rather than an auxiliary verb construction, where the control verb appears on the rightmost edge of the construction and the complement precedes the main verb. (21) a.

b.

mumu an’ igui cauari s’ haquiard’ ori-da16 2pl.nsbj 1sg.sbj as eggs st count want-ds ‘I want you to count the eggs.’ am’ an’ igui s’ himi muta-da, posa pare pima loc 1sg as st go want-ds but priest neg ‘I want to go there, but the priest does not.’ (Shaul 1986: ex. 47)

c. Pare Tonich vusa ni buy n’ himi tani Padre Tonichi dir 1sg.nsbj to 1sg.nsbj go make ‘The priest made me go to Tonichi.’ Although the constructions in (21a–c) look quite similar, in that both verbs appear in a sequence where the control verb, ori ‘to want,’ muta ‘to want,’ or tani ‘to make,’ is the first verb on the right and the dependent verb is ordered to the left of it and no other element intervenes between the two verb roots, the constructions di¤er: (21a–b) illustrate cases of bi15. One of the reasons that modal verbs have not yet received good documentation in both old and new grammars is that modal verbs provide information related to pragmatics rather than semantics (see Levinson 1983, Hopper and Traugott 1993: 7–98); since semantics takes care of stable meanings and pragmatics deals with beliefs and inferences of the participants, it is common that this kind of information is underrepresented in many grammars. 16. For the purposes of this article I have omitted Shaul’s segmentation marks between words and morphemes and follow the original presentation of the data by Loaysa, the author of the original grammar, but I have preserved Shaul’s translations and glosses.

298

Zarina Estrada-Ferna´ndez

clausal constructions and the switch-reference su‰x -da for the di¤erent subject confirms the analysis. Conversely, (21c) illustrates a more integrated or monoclausal construction, since the causative verb tani ‘to make’ is responsible for changing the case marking of the agent of the verb himi ‘to go,’ which is encoded as a non-subject, with the non-nominative pronoun n’ ‘1sg.nsbj.’ Constructions in (22) show the same distributional properties of the control and dependent verb; hakiarida mut’ ‘want to count,’ in (22a), and ohana simat ‘know how to write,’ in (22b). However, in such constructions a second position clitic, which encodes person and number as well as TAM values, is present: in (22a), the second position clitic an’ igui ‘1sg.sbj e,’ appears after the verb sequence hakiarida mut’ ‘want to count,’ and in (22b) there appear two distinct clitics, one for each clause: cad’ am igui ‘impf loc e’17 after the verbal sequence ohana simat ‘know how to write,’ and an’ t’ igui ‘1sg.sbj pfv e’ preceding the verb hukibuo ‘to forget,’ in the second clause. (22) a.

b.

humatcama s’ hakiarida mut’ an’ igui people int count.appl want 1sg.sbj as ‘I want to count the people.’ ohana simat cad’ am igui, posa vusi an’ t’ igui hukibuo write know impf loc as but all 1sg.sbj pfv as forget ‘I knew how to write, but I have forgotten everything.’

The examples in (23) show that the second position clitics can break up the verbal sequences by appearing in the middle between the two verb roots, that is, in the penultimate position, or second position starting from the end of the sentence. (23a) illustrates the clitic t’ io ‘pfv.fut’ appearing after the verb sicoana, and (19b), the clitic t’ ‘pfv’ in a medial position between the verbal sequence ohana urha. (23) a.

b.

pare oi aspi ti gaga sicoana t’ io ti tuhanu priest soon likely our fields weed pfv fut us order ‘The priest is likely to order us to weed our fields soon.’ haitu an’ igui ohana t’ urha thing 1sg.sbj as write pfv think ‘I think that I wrote something.’

17. Although Shaul considers this particle an ‘irrealis marker,’ throughout his examples he represents it with the abbreviation E. I consider it to be an assertive particle (as) with its diachronic origin in a demonstrative.

Grammaticalization of modal auxiliary verbs in Pima Bajo

299

The diachronic scenario observed for Nevome shows no auxiliary verbs in the language. All verbal complements in Nevome appear either encoded by morphologically complex constructions, such as (21–22), which resemble the one provided for Yaqui in (5), or those from Pima Bajo in (6), Timbisha (Shoshoni) in (15), and Kawaiisu in (16), or by means of constructions such as those illustrated in (23) where the second position clitic – that is, the encoding of the future tense io in (23a), or the perfective aspect t’ in (23b) – appears between the two verbs, the main or control and the dependent verb. I will return to such cases in the following section.

4. Grammaticalization of auxiliary verbs in Pima Bajo We now turn to the analysis of the development of modal auxiliary verbs. Cross-linguistically, Anderson (2005), Frajzyngier (1996), Givo´n (2001), Heine (1993), and Traugott and Heine (1991) have observed that the grammaticalization of complex constructions or sentences is the result of the di¤erent constructions which are available for the encoding of a particular syntactic domain. Thus, the grammaticalization of complex constructions observed in a given language is the direct result of the di¤erent principles that operate in the language. In addition, Givo´n (2006) has pointed out that the di¤erent types of verbal complement must be considered the result of the distinct grammaticalization pathways of such constructions. The development of complex constructions or predicates, however, may also be the result of language contact (Bowern 2006). With regard to the genesis of modal auxiliary verbs in Pima Bajo, we may consider two di¤erent modes of explanation, the internal hypothesis and the contact hypothesis. These are discussed in turn below. The development of auxiliary verbs, in particular when seen from the perspective of clause linkage, clause combining, or clause union, has been richly discussed in the linguistic literature. Thus, Haiman (1985: 212) believes that an auxiliary verb is formed as a result of an ‘‘extreme’’ conceptual fusion among two verbs where one of them loses some of its properties. Foley and Van Valin (1985) and Van Valin (1993) view auxiliary verb constructions as cases of nuclear cosubordination, a type of clause union or linkage where one of the verbs loses its argument requirements. In his discussion of the typology of clause linkage, Lehmann (1988) calls attention to the di¤erent stages in the desentialization of subordinate clauses, and proposes a continuum in which auxiliary verbs are ordered between serialized verbs and verbal derivation, that is, morphologically complex

300

Zarina Estrada-Ferna´ndez

verbs. Recently, Anderson (2005: 303¤.) has also discussed the historical development of auxiliaries in terms of a continuum of monoclausal verb combinations, that is, serialized constructions or verb combinations. In particular, Anderson distinguishes between the lexical head (the semantic head or element that determines the number and semantic role of its arguments) and the auxiliary verb, or element that conveys some ‘‘auxiliary’’ information, such as tense, aspect, mood, polarity, and so on. As mentioned earlier, all of the Uto-Aztecan languages discussed in this paper possess more than one type of construction for verbal complements. As a starting point for my analysis, I have proposed four groups of languages (although a finer distinction might be necessary for languages like Cupen˜o, as suggested in Section 2). Recall that languages can be grouped according to four types of construction: (i) nominalized constructions such as those observed in Ute, Cupen˜o, Kawaiisu, and Yaqui, examples (1–4); (ii) morphologically complex predicates observed in Yaqui (5), Pima Bajo (6), Timbisha (15), Kawaiisu (16), and Nevome (21–22); (iii) analytical periphrastic constructions where the auxiliary verb occurs without any inflectional morphology, that is, modal auxiliary verbs, such as those shown above for Pima Bajo (8–10), Yaqui (12), and Tarahumara (13). Finally, (iv) the least integrated biclausal construction or adjunct-like construction, as was illustrated for Pima Bajo (7). The special status of languages like Cupen˜o is mainly due to the occurrence of the second position clitic or clitic complex (Hill 2005). Data from Nevome with this type of element, (23), also enters into the discussion, since this clitic may be creating the syntactic conditions for the grammaticalization of modal auxiliary verbs. Steele (1990) has characterized the aux or second position clitic as an element that may contain all the relevant grammatical information for an utterance: the person and number of the subject as well as the tense, mood, voice, and aspect of the verb. The examples in (24–28) illustrate the second position clitic, or clitic complex, in some modern Uto-Aztecan languages, some from the Takic branch (Luisen˜o and Cupen˜o) and others from the Tepiman branch (O’odham, Northern Tepehuan, and Southern Tepehuan). (24) Luisen˜o (Steele 1999: 6, 129) a. notaax nil chaqalaqiqus 1sg.refl aux.1sg tickle.past.cont ‘I was tickling myself.’

Grammaticalization of modal auxiliary verbs in Pima Bajo

b.

301

chaqalaqi-wun pum hengeemali tickle.pl-prs aux.3pl boy.obj ‘They were tickling the boy.’

In Luisen˜o, in (24), the clitics nil or pum encode the person and number of the subject, and appear in a second position either preposed (24a) or postposed (24b) to the verb. In Cupen˜o, the clitics encode only mood (25a) or person, number, and case of the subject as well as aspect and mood (25b–c). (25) Cupen˜o (Hill 2005: 86, 72) a. me aya¼’ep hay-pe-ya-qal and then¼r finish-3sg-yax-pis ‘and then it was finished.’ b.

Ne¼’ep¼ne ersaar-qa 1sg¼r¼1sg.erg pray-prs ‘I was praying.’

c. Hi-sh¼qwe¼me aya pu’u’uy? What-npn¼noni ¼3pl.erg then eat.hab ‘What can they eat then?’ The examples from the Tepiman branch, O’odham (26), Northern Tepehuan (27), and Southern Tepehuan (27), confirm the inflectional possibilities of the second position clitic or complex. In all these languages, the clitic encodes the person and number of the subject as well as the tense-aspect or mood of the verb. (26) O’odham (Saxton, 1982: 128) am a-t-ki ˘ uu loc 3sg-t/a-mod rain.pfv ‘It rained there.’ (27) Northern Tepehuan (Bascom, 1982: 281) im-na-p-sa go-pot-2sg-quot ‘He said that you must go.’ (28) Southern Tepehuan (Willett, 1979) ya’ n-pix ca-vaqui-a’ loc 1sg-t/a while-enter-fut ‘I will be inside there in a while.’

302

Zarina Estrada-Ferna´ndez

In addition, in examples from Nevome provided in (23) and in languages from the Tepiman branch, the second position clitic expresses tense. The second position clitic must be analyzed as a suitable position for the encoding of subject agreement markers, that is, person and number, as well as for the temporal values/operators of the clause – tense, aspect, and mood. This scenario predicts that in any combination of verbs, that is, periphrastic or serialized constructions, the languages have two alternatives for encoding this information: either a bare uninflected root or stem, as in Serrano (20) (aux-headed according to Anderson 2005: 111), or more than one element encoding the inflectional values, as in Cupen˜o (2) (double-subject marking according to Anderson 2005). A third possibility is the one observed in Pima Bajo, where modal auxiliary verbs appear; in fact, Anderson (2005: 130) considers Pipil a lex-headed language, that is, a language where the modal auxiliary verb occurs without any inflection and all the relevant information is encoded on the lexical head. (29) Pipil (Anderson 2005: 130 [Campbell 1985: 139]) weli ni-nehnemi wehka mod 1-walk far ‘I can walk far.’

5. Contact-induced change hypothesis The Pipil example in (29), as well as the auxiliary verbs provided for Pima Bajo, Tarahumara, and Yaqui, resemble periphrastic constructions with modal auxiliary verbs from Spanish. (30) Yo puedo caminar Pedro quiso cantar El hombre debio´ venir ayer La mujer empezo´ limpiando la casa

‘I can walk’ ‘Peter wants to sing’ ‘The man should have come yesterday’ ‘The woman started cleaning the house’

Such periphrastic or multiple-verb constructions from Spanish contain a first or aux-verb conveying tense-aspect-modality as well as subject person and number; the second verb, the lex-verb in Anderson’s (2005) terms, can be either an infinitive or a gerund, that is, a non-finite verb form. If we consider that the Uto-Aztecan languages under discussion have been in contact with Spanish for at least the last three hundred years, then it is

Grammaticalization of modal auxiliary verbs in Pima Bajo

303

possible that the modal auxiliary verbs in languages like Pima Bajo are an instance of structural language change due to the influence of Spanish. However, how can we demonstrate that language contact is the preferred analysis for the development of auxiliary verbs in Pima Bajo? Or else, how can we argue in favor of an external explanation rather than a genetic internal change? Authors like Thomason and Kaufman (1998: 36) have argued for a theory where the sociolinguistic history of the speakers of a language and not only the structural facts or constructions must be considered as the most important factor for determining language contact. Moreover, Thomason and Kaufman claim (1998: 38) that minor structural borrowing, and probably calques, should be expected in situations where speakers are able to show minor phonological interference, i.e. adoption or incorporation of some of the phonemes of the contact language. The Pima Bajo’s speakers as well as the Tarahumara and other indigenous languages originally from Mexico, have been in contact for more than three hundred years with speakers of Spanish. Its common day life turns out to be reduced to family members only, but for any need the speakers of Pima Bajo must deal with speakers of Spanish. The Pima Bajo language has been influenced by Spanish with a huge amount of lexical borrowings, most of all used for naming artifacts, instruments, meals, domestic animals, etc. Structurally, and as a result of language contact, the language have also developed a middle marker pronoun a-, which is now used in almost the same contexts as the se middle marker from Spanish (Estrada 2005: 288). (31) a.

b.

pueert a-kuup door 3numntr.nsbj-close.pfv ‘The door closed’ (Sp. ‘La puerta se cerro´’) ko’okil vg a-nat’ia chilies red 3numntr.nsbj-become.fut ‘The chilies will become red’ (Sp. ‘Los chiles se enrojecieron’)

c. tmis-kar rbadag a-hain tortilla-ins middle 3numntr.nsbj-break.pfv ‘The comal [tortilla grill] broke in the middle’ (Sp. ‘El comal se quebro´ en medio’) The adoption of clause connectives or conjunctions from Spanish, e.g. porque, para que, hasta, is also common on the everyday use of the language (Go´mez Rendo´n 2008: for borrowing of conjunctions and prepositions in Quichua, Guaranı´ and Otomı´ ).

304

Zarina Estrada-Ferna´ndez

However, the borrowing of grammatical elements like the auxiliary verbs is not common. In his study about language borrowings in the American languages, Quechua, Guaranı´ and Otomı´, Go´mez Rendo´n (2008: 413) found out that borrowing auxiliary verbs was only observed in Otomı´ (where the percentage of borrowing in contrast with other grammatical categories was of only 0.6%). The few examples of Otomı´ of the borrowed auxiliaries show similar properties than those that I have pointed out for examples from Pima Bajo illustrated in (8–9): the verbs occur in a periphrastic or serializing construction, and in at least two examples, illustrated in (32), the verb borrowed from Spanish occur as bare forms, i.e. without TAM morphology or person and number agreement markers: (32) a.

nesesita da. . . nuya ja¨’ui da¼hn˜unta need FUT.3 DEM.PL person FUT.3¼get.together pa da¼hoku ’nar¼mehe for FUT.3¼build INDEF.S¼well ‘These people need to get together in order to build a well’

b.

ya mi¼pwede nda¼mats’I j¼ar ’batha. already IMPF.3¼be.able FUT.3¼help LOC¼DEF.S field ‘They could already help in the field’ (Go´mez Rendo´n 2008: 407–408)

As Go´mez Rendo´n (2008: 512) had mentioned ‘‘linguistic borrowing is an adaptation to discursive and communicative needs imposed by the dominant language,’’ Calque is a step ahead of such complex behavior. In the case of Pima Bajo, no more empirical data are available by now. The hypothesis that I have explored for the auxiliary verbs in Pima Bajo shall remain descriptive rather than explanatory. The need of more written and oral materials in this language may provide empirical evidences to demonstrate the viable contact origin of the auxiliary verbs that are emerging in this language. The documentation of those materials is still waiting to be collected. 6. Conclusion The linguistic facts described in this paper have two possible hypothetical explanations. One is internal to the language, where the cross-linguistic comparative data allow us to propose di¤erent diachronic stages in the

Grammaticalization of modal auxiliary verbs in Pima Bajo

305

grammaticalization of verbal complements, while the other is contactinduced change. The comparison of the data from Nevome and modern Pima Bajo might lead us to reject unidirectionality as an explanation of the grammaticalization of modal auxiliary verbs. This would be based on a scenario in which the Nevome data were taken to be the historically previous stage; but according to the internal reconstruction method this is not necessarily the case. As Givo´n (2006) says, the synchronic di¤erences observed in a language must be analyzed as ‘‘mere syntactic consequences of the di¤erent diachronic pathways.’’ The same situation may apply to the di¤erences among languages. Nevome does not necessarily represent the oldest stage of structural development. Within this context, modal verbs in Pima Bajo, may be viewed as the result of structural borrowing from Spanish.

Abbreviations 1, 2, 3 acc am ant as aux ben cap caus comp compl cont dat dem des dim dir ds emph erg ex fut gen

First, second, third person Accusative Actional mood Anterior Assertive Auxiliary Benefactive Capabilitive Causative Complementizer Completive Continuous Dative Demonstrative Desiderative Diminutive Directional Di¤erent subject Emphatic Ergative case Exhortative Future Genitive

306

Zarina Estrada-Ferna´ndez

hab imp impf in ins int inv irr loc mod mom neg nmlz nom nsbj numntr obj opt pasc pass pfv pgr pl pot prob prosp prs pst quot rdp refl sbj sg simss ss st sub unsp.obj v yax

Habitual Imperative Imperfective Thematic su‰x (Hill 2005) Instrument Intensive Invisible Irrealis Locative Modal Momentaneous Negative Nominalizer Nominative Non-subject Number neutral Object Optative Past continuous Passive Perfective Progressive Plural Potential Probability Prospective Present Past Quotative Reduplication Reflexive Subject Singular Simultaneity same subject Same subject Stative Subordinator Unspecified object Verb Theme class su‰x.

Grammaticalization of modal auxiliary verbs in Pima Bajo

307

References Aikhenvald, A. Y. and Dixon, R. M. W. 1998. Dependencies between grammatical systems. In Language 74.1: 56–80. Anderson, G. D. S. 2005. Auxiliary Verb Constructions. Oxford: Oxford University Press. Bascom, B. 1982. Northern Tepehuan. In Studies in Uto-Aztecan Grammar 3. UtoAztecan Grammatical Sketches, R.W. Langacker (ed.), 267–393. Dallas: The Summer Institute of Linguistics/The University of Texas at Arlington. Bisang, W. 2004. Grammaticalization without coevolution of form and meaning: the case of tense-aspect-modality in East and mainland Southeast Asia. In What Makes Grammaticalization? A Look from its Fringes and its Components, W. Bisang, N. P. Himmelmann, and B. Wiemer (eds), 109–138. Berlin: Mouton de Gruyter. Bowern, C. The historical linguistics of complex predication. http://www.ruf. rice.edu/~bowern/Mypapers/Intro.Cpreds.pdf Caballero, G. 2002. Mecanismos de transitividad del Rara´muri. B.A. tesis. University of Sonora. Campbell, L. 1985. The Pipil Language of El Salvador. Berlin: Mouton de Gruyter. Dedrick, J. M. and Casad, E. H. 1999. Sonora Yaqui Language Structures. Tucson: University of Arizona Press. Estrada Ferna´ndez, Z. 2005. Gramaticalizacio´n de los conectivos en construcciones complejas en pima bajo. In Trace 47: 18–29. Estrada Ferna´ndez, Z. 2005. The Pronominal Form -a as a Middle Marker in Pima Bajo. International Journal of American Linguistics. 71.3: 277–302. Estrada Ferna´ndez, Z. 2007. Cambio lingu¨´ıstico y contacto entre lenguas: gramaticalizacio´n de verbos auxiliares en pima bajo. In Revista UniverSOS. Revista de Lenguas Indı´genas y Universos Culturales 4: 91–114. Estrada Ferna´ndez, Z. and Buitimea Valenzuela, C. 2009. Yaqui de Sonora. Archivo de lenguas indı´genas. Mexico: El Colegio de Me´xico. Estrada Ferna´ndez, Z., Buitimea Valenzuela, C., Gurrola Camacho, A. E., Castillo Celaya, M. E., Carlo´n, A. 2004. Diccionario yaqui-espan˜ol. Obra de preservacio´n lingu¨´ıstica. Mexico: Editorial Plaza y Valde´s. Foley, W. A. and van Valin Jr., R. D. 1984. Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Frajzyngier, Z. 1996. Grammaticalization of the Complex Sentence. A Case Study in Chadic. Amsterdam: John Benjamins. Givo´n, T. 1981. Typology and functional domains. In Studies in Language 5.2: 163–193. Givo´n, T. 1984. Syntax. A Functional-typological Introduction 1. Amsterdam: John Benjamins.

308

Zarina Estrada-Ferna´ndez

Givo´n, T. 1990. Ute reflexives, complementation and clause integration. In Development and Diversity. Linguistic Variation across Time and Space. A Festschrift for Charles-James N. Bailey, J. A. Edmondson, C. Feagin, P. Mu¨hlha¨usler (eds), 287–296. Dallas: The Summer Institute of Linguistics/The University of Texas at Arlington. Givo´n, T. 2001. Syntax. An introduction 1. Amsterdam: John Benjamins. Givo´n, T. 2009. Multiple routes of clause union: the diachronic of syntactic complexity. In The Genesis of Syntactic Complexity, T. Givo´n, 61–96. Amsterdam: John Benjamins. Go´mez Lo´pez, P. 1999. Huichol de San Andre´s Cohamiata, Jalisco. Archivo de lenguas indı´genas de Me´xico. Mexico: El Colegio de Me´xico. Go´mez Rendo´n, J. 2008. Typological and social constraints on language contact: Amerindian languages in contact with Spanish. Volume 2. Netherlands: Graduate School of Linguistics. Guerrero, L. 2004. The syntax-semantic interface in Yaqui complex sentences, a role and reference grammar analysis. PhD Dissertation. The State University of New York at Bu¤alo. Haiman, J. 1985. Natural Syntax. Cambridge: Cambridge University Press. Heine, B. 1993. Auxiliaries, Cognitive, Forces, and Grammaticalization. Oxford: Oxford University Press. Heine, B. and Kuteva, T. 2002. World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Heine, B. and Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Heine, B. and Kuteva, T. 2008. The explanatory value of grammaticalization. In Linguistic Universals and Language Change, J. Good (ed.), 215–230. Oxford: Oxford University Press. Hill, J. H. 2005. A grammar of Cupen˜o. University of California Publications in Linguistics 136. Berkeley: University of California Press. Hopper, P. J. and Traugott, E. C. 1993. Grammaticalization. Cambridge: Cambridge University Press. Langacker, R. W. 1977. Studies in Uto-Aztecan Grammar vol 1. An Overview of Uto-Aztecan Grammar. Dallas: The Summer Institute of Linguistics/The University of Texas at Arlington. Lehmann, C. 1998. Towards a typology of clause linkage. In Clause Combining in Grammar and Discourse, J. Haiman and S. Thompson (eds), 181–225. Amsterdam: John Benjamins. Levinson, S. C. 1983. Pragmatics. Cambridge: Cambridge University Press. Marchese, L. 1986. Tense/Aspect and the Development of Auxiliaries in Kru Languages. Dallas: The Summer Institute of Linguistics/The University of Texas at Arlington. McLaughlin, J. E. 2006. Timbisha (Panamint). Munich: LINCOM-Europa.

Grammaticalization of modal auxiliary verbs in Pima Bajo

309

Peralta Ramı´rez, V. 2005. El nawat de la costa del Golfo. Algunas semejanzas y diferencias estructurales con el na´huatl central. Proceedings of the Conference on Indigenous Languages of Latin America II. 27–29 October 2005. http:// www.ailla.utexas.org/site/cilla2/Peralta_CILLA2_nawat.pdf Saxton, D. 1982. Papago. In Studies in Uto-Aztecan Grammar 3. Uto-Aztecan Grammatical Sketches, R. W. Langacker (ed.), 93–266. Dallas: The Summer Institute of Linguistics/The University of Texas at Arlington. Shaul, D. L. 1982. A grammar of Nevome. Ph. D. Dissertation. University of California, Berkeley. Shaul, D. L. 1986. Topics in Nevome Syntax. Publications in Linguistics 109. Berkeley: University of California Press. Smith, B. 1862. Arte de la lengua ne´vome que se dice pima de Sonora, propia de Sonora; con la doctrina Christiana y confesionario an˜adidos. San Agustı´n de la Florida. New York: Cramoisy Press. [1970. New York: Shea’s Library of American Linguistics. AMS Press, Inc.] Steele, S. 1990. Agreement and Anti-agreement. A Syntax of Luisen˜o. Dordrecht/ Boston/London: Kluwer Academic Publishers. Steele, S. 1999. Auxiliaries. In Concise Encyclopedia of Grammatical Categories. K. Brown and J. Miller (eds), 49–56. Oxford: Elsevier. Steele, S. 1981. An Encyclopedia of AUX: a Study in Cross-linguistic Equivalence. Cambridge: MIT Press. Thomason, S. G. and Kaufman T. 1991. Language Contact. Creolization and Genetic Linguistics. Berkeley-Los Angeles: University of California. Traugott, E. C. and Heine, B. (eds). 1991. Grammaticalization. Amsterdam: John Benjamins. Van Valin, Jr., R. D. (ed.). 1993. Advances in Role and Reference Grammar. Amsterdam: John Benjamins. Voegelin, C. F. 1935. Tu¨batulabal Grammar. Berkeley: University of California Press. Willett, E. 1979. Reduplication and accent in Southeastern Tepehuan. In Work Papers of the Summer Institute of Linguistics, University of North Dakota, 23: 47–65. Zigmond, M. L., Booth, C. G., and Munro, P. 1990. Kawaiisu. A Grammar and Dictionary with Texts. University of California Publications in Linguistics 119. Berkeley: University of California Press.

Contact, convergence, and conjunctions: a cross-linguistic study of borrowing correlations among certain kinds of discourse, phasal adverbial, and dependent clause markers1 Anthony P. Grant 1. Introduction Certain discourse markers and conjunctions that head many types of dependent or subordinate clauses are among the first structural or ‘‘grammatical’’ features speakers of a less dominant language are likely to borrow from the language of a more dominant or prestigious group. This claim has been put forward in Matras (1998 and subsequent work) and his observations have helped inform the statements which I intend to make in this contribution. In some cases, such as the Eskimo-Aleut language Siberian Yupik, these kinds of borrowings (discourse markers, adverbials, and other function words, which have been taken from Chukchi) account for more than half the total of loans into the less prestigious language from the more prestigious one, with considerable consequences for the shape and flexibility of the syntactic structure of this language at clause level and above (as shown in de Reuse 1994). Following the method of comparison introduced by Matras (1998), in this article I discuss the degree to which certain kinds of discourse markers, phasal adverbs, coordinating and especially subordinating conjunctions (the latter as used in some major and frequently occurring kinds of dependent clauses) have been borrowed in a wide range of languages (some 22 in all), which themselves have had frequent recourse to the replacement of inherited elements by means of borrowing, or in some cases the reinforce-

1. I would like to thank an anonymous referee, Alexandra Aikhenvald, Lameen Souag, Graham Thurgood, and Miriam van Staden for otherwise unavailable data, and a second anonymous referee, Alexandra Aikhenvald, Dik Bakker, Paul Heggarty, and Paz Buenaventura Naylor for suggestions which have found their way into this article, though they are in no way responsible for any use I may have made of them.

312

Anthony P. Grant

ment of inherited modes of indicating clausal coordination and subordination by means of borrowed elements. One might call such languages ‘‘heavy borrowers.’’ I am interested in seeing the extent to which it is possible, if at all, to establish an implicational hierarchy of conjunction borrowing, given that it has long been recognized that markers of dependent clauses in the form of conjunctions are quite widely borrowed in many of the world’s languages. For example, if a conjunction with a sense X is borrowed from another language, can we assume that a conjunction with a sense Y is also going to be borrowed? Examination of such works as Heine and Kuteva (2002) shows how the meanings of discourse particles, phasal adverbs, coordinating conjunctions, and even subordinating conjunctions interact with one another, and also how the sense of these items can change over time. In the case of borrowed elements, we can sometimes see how the senses of such items can change between their use in the donor language and their incorporation in the recipient language. The categories listed above are more porous than initial inspection may suggest, as can be seen from the tracking of the meanings of loan items. Thus asta in Cochabamba Quechua, borrowed from Spanish (and itself taken from Arabic ¡atta) means ‘until’ as in Spanish, and also ‘even’; it is used in Cochabamba Quechua as both a subordinating conjunction and a phasal adverb. A form deriving from Spanish mas que, Portuguese mas que ‘but, on the other hand,’ and found in languages as widespread as Afrikaans and Tok Pisin, is a discourse particle in some languages (such as Tok Pisin maski ‘it doesn’t matter’; Alexandra Aikhenvald, personal communication) and an adversative subordinating conjunction in others (such as Tagalog maski ‘although’). Similarly the Chamorro subordinating conjunction sinembatgo ‘although’ derives from the Spanish adversative discourse particle sin embargo ‘however.’ As I point out in Section 8, Finnish ja ‘and’ is of Germanic origin; its etymon is reflected by Gothic jah ‘indeed’ and German ja ‘yes.’ Here a discourse particle has changed over time to a coordinating conjunction. Contact-induced syntactic change and the development of subordinating conjunctions from other sources (including forms themselves deriving from other subordinating conjunctions) have long been attested in languages. It is also the case that various kinds of function words may develop from words belonging to other form classes in the course of the history of a language: Posner (1966: 220–227) demonstrates this for several kinds of function words in a number of Romance languages. Furthermore, di¤erent forms

Contact, convergence, and conjunctions

313

within the same form class, such as subordinating conjunctions, can change or expand their meanings over time. Deutscher (2000) discusses the rise of a complementizer kıˆma in Akkadian which derived from a form with the primary meaning of ‘because.’ A certain amount of work has thus already been done on this general subject. Much of this has focused on borrowing such items into Aztecan and Mayan languages and on other languages of Mesoamerica which have been influenced by Spanish. Already in 1930 Franz Boas was drawing examples of syntactic borrowing in a corpus of modern Nahuatl texts he had gathered at Milpa Alta, near Mexico City, and was discussing the degree to which Milpa Alta Nahuatl was borrowing these kinds of forms and also absorbing more purely lexical and acculturational elements from Mexican Spanish (Boas 1930). This kind of borrowing in Mesoamerican languages is not confined to items from Spanish. Macri and Looper (2005) point out the borrowing of Nahuatl i:wa:n ‘and then’ as a discourse particle into the Mayan language of ancient Ch’olan inscriptions, where it is recorded from as early as 702 CE. More recently Brody (1987, 1993) discussed the use of discourse markers in a number of Mayan languages where such discourse markers have been borrowed from Spanish. The borrowed items are used to link segments of discourse of various sizes (phrase, clause, sentence, or paragraph), and many of them have the same illocutionary e¤ect as conjunctions have. An important article by Matras (1998), inspired in part by the work on bilingual discourse particles in Salmons (1990), examines the borrowing of fillers, tags, interjections, some phasal adverbs, and some coordinating conjunctions and certain other items, which he categorises as ‘utterance modifiers’ because of their detachability from clauses or because they constrain the contextual relevance of their host utterance. These are elements against which the borrowing of subordinating conjunctions should be seen, and indeed Matras provides a set of purported universals of borrowing of utterance modifiers, among which he includes the discourse markers and coordinating conjunctions mentioned above. Such purported universals can therefore be analyzed and tested for their universality against a di¤use data set of languages from around the world. Matras’s own work here focuses on findings from varieties of Romani, Russian influence on the German of speakers born in Russia, and on languages influenced by Hebrew or Arabic, such as Swahili or Domari (there styled Na´wari). Matras (2000a) suggests that this borrowing of discourse particles in particular (and the borrowing of utterance modifiers in general) occurs because of bilingual speakers’ overwhelming need to monitor and direct their dis-

314

Anthony P. Grant

course for the benefit of their hearers; he terms this feature Fusion. Matras further classifies and grades these forms according to three criteria: degree of pragmatic detachability (turn-related forms are borrowed before content-related forms), their place on the semantic scale (lexical and deittic forms are borrowed last), and their place on the category-senstitive scale, where forms expressing restriction, change or contrast are borrowed before forms which express continuatiom elaboration or addition. The work which has been carried out at various times by Boas, Brody, and Campbell has been copiously supported and amplified by subsequent work on the structural e¤ects of Spanish on indigenous languages in Mesoamerica and beyond by Thomas Stolz, often in conjunction with Christel Stolz (notably Stolz and Stolz 1997, and Stolz 2002).2 Torres (2006) continues this work and discusses consequences of the borrowing of Spanish discourse markers into some languages of Latin America. Turkic languages have also received much attention in regard to their borrowing of conjunctions from other languages (mostly from Arabic and Persian, but in languages spoken in the former Soviet Union also from Russian). Baran (2000), discussing the borrowing of Russian discourse particles into the Uzbek of Tashkent, and works by Lars Johanson, such as Johanson (1993; 2002), discussing the borrowing of causal and other conjunctions into several Turkic languages from Russian, Farsi, and other sources, are of major note here. A later work on this topic is Matras and Sakel (2007). My approach to the syntactic analysis of subordinating conjunctions and their use in dependent clauses is informed by the work of Thompson, Longacre, and Hwang (2007). That chapter (happily for us) includes inter alia a short discussion of borrowing among dependent clauses, drawing examples from two indigenous Mexican languages, namely the Uto-Aztecan language Yaqui and the Otomanguean language Isthmus Zapotec. In both cases the borrowed subordinating conjunctions discussed are taken from Spanish, though the two languages have not borrowed the same conjunctions in every case. The e¤ects of this borrowing are those we find in all the languages surveyed in this study: borrowing conjunctions aids convergence insofar as it makes details of the clausal syntax of the borrowing or recipient language resemble the respective details of the donor language syntax, often more than the syntax of the recipient language previously did. 2. One may see also studies of the impact of Spanish on Nahuatl by Sua´rez (1977), Hill and Hill (1987) and Field (2002), while Karttunen and Lockhart (1976) examine the impact of Spanish in Nahuatl in earlier centuries. Sua´rez (1983) looks at this issue cross-linguistically within Mesoamerica.

Contact, convergence, and conjunctions

315

It is essential from the start for us to recognize that, at the level of a construction, the act of borrowing from one language to another involves the use of one or both of two strategies that are certainly not mutually exclusive in their operations and often co-occur. The first of these is ‘‘transfer of pattern’’ (Grant 2002). In this strategy, structural features of a construction in one language (the donor language) are replicated in another by means of morphemes already existing in the language that took them over, the recipient language. (This term was first mentioned, though referred to as ‘‘pattern transfer’’ without a specific definition being provided, in Heath 1984.) The other strategy is ‘‘transfer of fabric,’’ also discussed by Grant (2002). In this operation, morphemes used to express the structural concept under examination are taken over from the language from which the construction is itself taken, the donor language. The two types of transfer are not mutually incompatible. In this article I also intend to determine the degree to which morphemic transfer correlates with pattern transfer, and to see what light these findings shed on an enhanced theory of the role of contact influences in actuating and reinforcing morphosyntactic change. 2. A crosslinguistic perspective on the borrowing of conjunctions and phasal adverbs 2.1. The languages surveyed The languages whose dependent clause linking strategies are surveyed in this article were chosen because they all exhibited a high degree of borrowing or replacement of everyday lexicon (and often also replacement of Swadesh list-style basic lexicon) from other languages. All of them exhibit some borrowing of certain dependent clause markers. (English itself has borrowed the forms of until and although and the first part of the bipartite coordinating conjunction both . . . and . . . from Norse, and the second morpheme of because from Latin by way of French.) In order to ensure that a fuller sense of the degree of such borrowing could be ascertained, and to make sure that enough information was available on the nature and mode of expression of a wide variety of clause and discourse-linking techniques, I only included languages in which at least 12 of the 18 discourse markers, adverbs and conjunctions discussed in this study were represented in the material available to me. In this way I hoped to ensure that the borrowed items which have been chosen were set more fully

316

Anthony P. Grant

within the wider context of the di¤erent kinds of conjunctions and other items being surveyed. Setting aside Present-Day Standard English, for which my source has been the OED online, the 21 chosen languages are listed below. The major data source has been asterisked (other data sources are used only if relevant forms are missing from the major data source): – Eastern Yiddish (*Jacobs 2005) – Kalderash Romani, a diasporic language that spread from Romania in the late nineteenth century (*Gjerdman and Ljungberg 1964; Boretzky 1992) – Kildin Saami of the Kola Peninsula, Russia (*Kert 1968, 1971; Riessler 2007) – Livonian of Kurzeme or ‘‘Curonia,’’ coastal Latvia, in the form of the Kolka dialect (de Sivers 2001; Kettunen 1938; Va¨a¨ri 1968; *Moseley 2002; Sˇuvca#ne and Ernsˇtrieite 1999), and formerly also (in the form known as the Salis dialect) spoken in Livland (as described in *Sjo¨gren 1861, 1861b) – Uluag˘ac¸ Cappadocian Greek (*Dawkins 1916; Kesisoglou 1951) – Standard Turkish (*Kornfilt 1997) – Turoyo (properly T¸u#royo) Eastern Neo-Aramaic of southeastern Turkey, a variety widely used in the ‘‘Assyrian’’ diaspora (*Jastrow 1992; see also Matras 2000b for etymologies) – Urdu (*Schmidt 1999; Shahani n.d.) 3 – Acehnese of northern Sumatra (Cowan 1981; *Daud and Durie 1998) – Tsat (Hainan Cham) of Hainan Island, China (*Zheng 1997; personal information from Graham Thurgood) – Tagalog of the Philippines (*Rubino 1998) – Tidore of Tidore Island, Halmahera, Indonesia (van Staden 1999) – Tetum Dili of Timor Leste (*Williams-van Klinken, Hajek, and Nordlinger 2002) – Chamorro of Guam and the Northern Mariana Islands (*Topping with Dungca 1973)

3. Many of the forms of Farsi or Arabic origin found in Urdu are also recorded for Hindi, which also avails itself of some forms of Sanskrit origin, such as parantu ‘but,’ which are not used in Urdu. It is salutary to note both the paucity of inherited conjunctions and the like from Sanskrit to be found in Urdu and Romani and also the lack of shared inherited forms common to Kalderash Romani and Urdu.

Contact, convergence, and conjunctions

317

– Ifira-Mele of Mele village and Fila Island, Vanuatu (Capell 1942; *Clark 1998, 2002) – Siwi Berber of Siwa Oasis, Egypt (*Laoust 1931; Vycichl 2005; Souag 2010) – KiUnguja, a southern Swahili dialect that is the basis of Standard Swahili (Wald 2001; *Perrott, revised Russell 2002) – Siberian Yupik (St. Lawrence Island, Alaska; de Reuse 1994; a similar variety also containing many Chukchi loans is also spoken in Siberia and is known as Chaplinski after Chaplino, the main village where it is used) – Pipil of El Salvador (Campbell 1985, 1987) – Garifuna of Belize (Taylor 1958, 1977; Cayetano 1993) – Bolivian Quechua, especially Cochabamba Quechua (*Lastra 1968 documents this diasporic form of Southern Peruvian Quechua, though much of the syntax is undiscussed; see also Muysken 2001). Notes on a few of these languages may not be unwelcome. I have selected the variety of Greek spoken until the early 1920s at Uluag˘ac¸ in Cappadocia because this variety of Cappadocian Greek, now apparently extinct, was probably the one most heavily influenced by Turkish. I regret that a dearth of suitable descriptive material precluded me from including a similar study of a dialect of Arvanitika or of Arbe¨resht (varieties of emigrant Southern Tosk Albanian, as spoken in mainland Greece or in pockets of central and southern Italy respectively), or one of Italiot Greek, Molise Croatian, or of Istro-Romanian. For its part, Siwi is probably the most heavily Arabized variety of Berber in existence, and by virtue of being spoken hundreds of miles from the nearest Berber varieties, located in Libya, it has developed in general isolation from other Berber varieties in North Africa. English, Yiddish, Kildin Saami, and KiUnguja Swahili are also among the 40 or so languages being examined in the Loanword Typology Project hosted by the Max Planck Institute for Evolutionary Anthropology, Leipzig, a project that looks also at other varieties of Berber, Romani, and Quechua (Tarifiyt, Hungarian Rumungro, and Imbabura Ecuadorian Quechua, respectively) from those which are discussed here. The di¤erent kinds of material available to me for each language vary widely; I had access to some textual material for all of them, and to lexical material (of widely varying degrees of comprehensiveness and copiousness) and sentential material for most of them (though my data on Tidore, Siberian Yupik, and Turoyo are probably the sparsest in this regard). Information taken from dictionaries is the kind of material most readily

318

Anthony P. Grant

available to me for many of these languages, and this study is a little more ‘‘dictionary-driven’’ than some might like; the authors of other studies on bilingual discourse markers, such as those by Matras, Boas and Weilbacher, Stolz, and Salmons, argue from the findings in their text corpora. But for all of them, the information on the nature and approximate number of borrowed conjunctions is su‰cient for me to be able to develop the present study, and the conclusions are borne out by textual material on these languages. Certain unwelcome external constraints upon this study need to be pointed out and certain assumptions need to be taken on trust. To take one example, this study relies largely on contemporary data sources. We must avoid relying on argumenta a silentio when positing ideas about the syntactic history of a language; simply because we do not have any evidence in modern materials on a language for a pre-contact system of dependent clause markers, this does not mean that the language in question never had them. We must not assume that languages acquired subordinating conjunctions solely through borrowing, and that prior to borrowing these conjunctions the speakers of such languages had no means of indicating hypotactic relations in speech, so that they strung clauses together without connecting elements. However, we cannot always be sure of the status or even of the presence of overt conjunctions, as opposed to other clausecombining techniques in earlier stages of the languages being surveyed. This is because of the paucity or absence of documentation (especially earlier stages of documentation) of some of the languages described. Indeed, some potentially promising contributors to this study had to be excluded because the minimum of 12 elements I felt were needed for inclusion in this study were not present in the accumulated literature on the language in question available to me. In many cases we simply do not possess information on the means of forming some of the kinds of dependent clauses being examined here as they might have been available to speakers of these languages, say 300 years ago, or at least in the period before the language(s) that provided subordinating conjunctions came into contact with the language in question. There is no textual material, for example, of Ifira-Mele that contains dependent clauses and also dates from before the 1940s; Capell (1942),4 is 4. This article contained a text in the form of a letter from a schoolboy who was being educated in Suva, Fiji, but who came from Fila, and who spoke a dialect which was somewhat less strongly influenced by the locally once dominant South Efate language than Mele is (though still strongly so).

Contact, convergence, and conjunctions

319

the first textual example of any use for this language in this study. Premodern textual material that may show strategies for discourse marking that do not use borrowed items, and for which evidence from closely related languages is not available, is also elusive for Tsat, Cochabamba Quechua, Kildin Saami, and Tidore. Dictionary or other lexical material (often provided with extensive and reliable etymological information) is plentiful for most of these languages, structural material is available and often abundant, but textual material of the kind most useful for this sort of study is often sparse, and even exemplary sentential material which demonstrates the use of these conjunctions in relevant dependent clauses is not always available. Unfortunately the material available to me regarding borrowed conjunctions in some of these languages was a bare listing of conjunctions and of the etymological sources of these forms. Examples of some of the conjunctions or of the types of dependent clauses expressed by such conjunctions are not always available to me in the technical literature (insofar as it exists) for some of the languages being examined. The result of these omissions from the sources is that there are gaps in the published linguistic record on the coverage of such clauses. Indeed, this study has focused on the kinds of dependent clause markers most frequent in occurrence: the most common manifestations of those are associated with conditional, causal, temporal, purposive, and concessive dependent clauses, and complement and relative clauses. Among the temporal markers I have preferred to look for instances of borrowed ‘before’ and ‘until’ rather than those of borrowed ‘while’ and ‘after,’ for two reasons. Firstly, ‘after’-clauses are often subsumed in ‘when’-clauses, which if showing identical agents in main and dependent clause themselves may be expressed or implied in some languages in non-finite dependent clauses relating to a main clause whose verb refers to an action that occurred subsequent to that of the dependent clause. Secondly, ‘while’-clauses express the coterminous nature of a verb’s action, which results in their being expressed by non-finite verb-forms in many languages. WHEN is the cardinal member of any class of temporal adverbial clause markers by its very nature, so that data on this have been collected and used in this study. But even then one must take what one can find. For instance, one should not be surprised to find that it is easier when looking through this material to find attested examples of conditional clauses formed with equivalents of IF in the materials than examples of conditional clauses formed with equivalents of EVEN THOUGH or with EXCEPT (FOR THE FACT) THAT, because IF is more frequently used than these other conjunctions (the website for Leech, Rayson and Wilson 2001 has 59

320

Anthony P. Grant

occurrences of EVEN THOUGH, 14 for EXCEPT THAT, 41 for EXCEPT as a conjunction and 2369 occurrences for IF). It is also easier to find instances of conditional IF than of the IF that in English introduces indirect yes-no questions and alternates with WHETHER. This is an interesting kind of dependent clause marker, one that is less frequently illustrated in materials on many of the examined languages than others, which I have thus sadly had to omit from the study. I have also concentrated on modes of expression of subordination in verb clauses where the subject of the main clause and that of the dependent clause are not the same, as many languages (Hindi/Urdu for instance) have non-finite construction strategies available for use when the subject or agent of main and dependent clause is identical. 2.2. The kinds of markers under study The result of this paucity of data is that even though some of the material available to me (for instance the rather copious data on Chamorro, Yiddish, Urdu, and Kalderash Romani, to say nothing of English) suggests that these less frequently occurring dependent clause markers are more likely to be encoded by borrowed items than the more common equivalents, I have reluctantly decided not to include instances of equivalents of these dependent clause markers of rarer occurrence in the cross-linguistic study. This is because it has not always been possible to find equivalents for these glosses in material available on other languages in the sample. As stated above, I have only included languages in my sample if at least 12 of the 18 elements surveyed can be found in the materials available to me, with at least one example each of the discourse markers (or ‘‘Wackernagel particles’’), the adverbial conditional and causal dependent clause markers, complementizers, relativizers, and purpose and temporal clause markers attested before the language is included in the sample. Sometimes not even extensive grammars of the language can provide us with enough of the required information. Information on certain kinds of temporal adverbial dependent clauses is often especially hard to gather from published descriptions for a number of languages. Some considerations should be brought to attention from the first, however, as they may have a bearing upon the hierarchy of borrowability of the coordinating conjunctions in question. The first caveat is that many languages use more than one form that can be translated as ‘and,’ since they use one conjunctional form (not necessarily a free form) to connect noun groups and another, di¤erent form (often reinforced with a temporal

Contact, convergence, and conjunctions

321

adverb or other element in order to express the sense of ‘and then’) to connect verb groups. In a number of languages, the form used to link noun groups also has a meaning (sometimes a primary meaning) of ‘with,’ so that this form for ‘with’ may be inherited (or innovated), with ‘and’ as a secondary meaning. The second consideration is that both AND and OR, but not BUT (by its very nature and meaning as an adversative which does not need a preceding contrasting conjunction), may in many languages be used in paired forms, as a means of forming multiple conjunctions, sometimes involving other grammatical elements (with senses such as EITHER . . . OR . . . , NEITHER . . . NOR . . . , and BOTH . . . AND . . . ). This usage has been attested in a number of languages throughout time and space. For instance, in Latin we find et . . . et . . . (‘both . . . and . . . ,’ literally ‘and . . . and . . .’), aut . . . aut . . . , sive . . . sive . . . , seu . . . seu . . . , all three pairs meaning ‘either . . . or . . .’ (literally: ‘or . . . or . . .’), and nec . . . nec . . . (‘neither . . . nor . . . ,’ literally ‘and not . . . and not . . .’). Similar situations involving the use of doubled conjunctions can be found elsewhere, for instance in Attic Greek and in Classical Arabic. In fact, doubling of use of these forms in such constructions may serve as a means of reinforcing or extending the use of some conjunctions that might otherwise be open to replacement, for instance because of their phonological brevity. If BUT is paired with or opposed to anything it is generally the second half of a pair of which the first is a negator. (Neither of these points was exploited much in Matras 1998, which surveys borrowing of simplexes for AND, OR, BUT.) Doubling of such forms is the case, for instance, in Standard Yiddish, which uses paired conjunctions, some of them inherited from German, such as say . . . say . . . ‘either . . . or . . . .’ (etymologically identical with modern Yiddish zay ‘be!’ and with modern German sei . . . sei . . . . ‘be it . . . be it . . . ,’ and which Jacobs 2005: 16 indicates as being remarkable as the only Yiddish form of German origin in which original initial s- is still realized as an unvoiced sibilant in Yiddish), others of them from Hebrew or Aramaic (hen . . . hen . . . ‘either . . . or . . . ,’ in rabbinical and formal usage) or from Slavic (for example i . . . i . . . ‘both . . . and . . . ,’ an item probably of Polish origin, or the variant found in Ukrainian Yiddish, namely to . . . to . . . ‘both . . . and . . . ’; Eastern Yiddish data are from Jacobs 2005, and have been retranscribed into the YIVO orthography which is generally used for Romanizing Standard Yiddish). BUT does not (and cannot) be part of such a pair of conjunctions, since it either contrasts with a negator or with nothing at all. The ‘‘simple’’ conjunctions BUT (and here nayert is used after negatives, otherwise ober), AND (un), OR (oder) in Yiddish are

322

Anthony P. Grant

all of Middle High German origin; however, this is not the case with forms for BOTH . . . AND . . . , and one of the forms used for EITHER . . . OR . . . . is also borrowed. But the equivalent forms in Modern German (sowohl X wie Y ‘both X and Y,’ entweder X oder Y ‘either X or Y,’ weder X noch Y ‘neither X nor Y’) are not used in modern Standard Yiddish, which uses the Slavic forms listed above. Again, Tagalog uses a conjunction at from Kapampangan (the language once spoken in what is now the heart of Tagalog territory in southern Luzon) for ‘and,’ as I mention below, but for ‘both . . . and . . . .’ it uses i . . . i . . . from Spanish, while o ‘or’ and o . . . o . . . ‘either . . . or . . . ’ (also Tagalog ni . . . ni . . . ‘neither . . . nor . . . .’) are borrowed from Spanish, which itself uses these repeated conjunctions for ‘either . . . or . . .’ and ‘neither . . . nor . . . .’ The following sentences are English examples of the kinds of sentences whose use of conjunctions interests me; I have capitalized the conjunctions in question. I regret not having had the opportunity of eliciting such sentences (or of collecting texts containing clause subordinators) from speakers of the languages in question. My debt in regard to the delineation of sample sentences to the contents of the Tense-Aspect-Mood questionnaire in Dahl (1985: 198–206) will be obvious. (1) ALTHOUGH it is raining, I will walk into the village (concessive). (2) IF the teacher comes, we will go to the beach. (3) IF the teacher came here, he would be lonely (1–2 are conditional clauses). (4) UNTIL the dog finds the bone, the rabbit will be scared. (5) BEFORE the nurse began to work in the hospital, the doctor emptied her o‰ce. (6) AS the night became wet, we ran into the barn. (7) WHILE it rained we stayed in the barn. (8) SINCE the house is empty, I shall sing and dance (3–8 are temporal clauses). (9) He cannot swim BECAUSE his arm is broken (causal clause). (10) I came to the city IN ORDER TO find work. (11) The king built a wall SO THAT/IN ORDER THAT the soldiers would be safe (10–11 are purposive or purpose clauses).

Contact, convergence, and conjunctions

323

(12) They say THAT you are happy. (13) I know THAT you are happy. (14) I hope THAT you are happy (12–14 are complement clauses, 13 expressing a realized complement while 12 and 14 expressed unrealized complements). (15) The farmer WHO/THAT lives by the river grows wheat. (16) The woman WHO/WHOM/THAT we saw lives near the forest. (17) The cat WHICH/THAT I saw yesterday had a long tail. (18) The car WHICH/THAT rolled down the hill broke the gate. (19) The child TO WHOM I gave the book likes toys a lot (or: The child THAT I gave the book to likes toys a lot). (20) The woman WHOSE umbrella I borrowed stayed inside when it rained. (21) WHOEVER/They WHO ate that bread will be sick. (Examples 15–21 are relative clauses; 21 is a so-called headless relative clause). Here we see the overt use of dependent clause markers (capitalized) in the sample of English sentences. However, in English and many other languages (including several of those illustrated here) it is possible to construct certain kinds of dependent clauses without being compelled to introduce them with an overt dependent clause marker, just so long as the main clause is well-formed. Present-Day Standard English is able to omit THAT when it introduces post-verbal complement clauses, and it can also omit the forms THAT (which is more colloquial in terms of register in UK English than the other English relativizers are), WHICH, and WHO or WHOM in cases when the antecedent in the main clause is the direct or indirect object of the relative clause. Additionally, speakers of English sometimes replace these with non-finite dependent clauses or with nominalizations. I have represented the ‘‘missing’’ complementizers and relativizers (and in the case of sentences 26 and 27, also the missing copular verbs) with Ø: (22) They think Ø it’s all over. (23) They say Ø there is no hope. (24) I wish Ø the rain would stop. (25) I hope Ø the bus will come soon.

324

Anthony P. Grant

(26) I saw the man Ø running away from the blazing building. (27) The train Ø now standing at Platform 5 does not stop at Warrington. (28) I saw the film Ø I had read about last week. (29) They have met the people Ø they wish to hire. Similarly, some temporal clauses in English only contain an implied temporal element, by which the sequence of events is implicit within the discourse structure of the sentence; in such cases English can employ non-finite dependent clauses (which are also available for other kinds of dependent clauses). This is possible (but not compulsory) if the subject of the two clauses is the same. (30) They went through the park. They saw a red kite. (31) Having gone through the park, they saw a red kite. (32) Subsequent to going through the park, they saw a red kite. (33) After going through the park, they saw a red kite. (34) After they had gone through the park, they saw a red kite. All these expressions are generally equivalent.5 Although English uses the same marker, in this case THAT, as the complementizer for both realized and unrealized complement clauses, many languages do not do so. They use distinctive forms (for instance Croatian sˇto for realized and da for unrealized complements), and if one of the members of this set of complementizers is borrowed, it appears from the evidence we have that it is most often the one used with unrealized complement clauses. In some of the languages in this sample, the use of dependent clauses not specifically introduced with (or followed by) dependent clause markers represents the most common pattern of formation of such clauses. The use of kinds of nominalization constructions is especially characteristic of the formation of restrictive relative and complement clauses in such languages. This is also the case with some kinds of dependent temporal clauses in which the time of the action of the dependent clauses is either synchronous with that of the main clause or else precedes it immediately (the VERB 5. As are other constructions, such as They went through the park, (and) After this, they saw a red kite.

Contact, convergence, and conjunctions

325

STEM þ kar clause in Hindi/Urdu is an instance of this). This being the case, we should not be surprised if we fail to find a large number of instances of borrowed markers for relative clauses, or a great number of borrowed complementizers among the sampled languages. It should finally be noted that in most of the languages surveyed (Quechua being an exception), dependent clause markers usually precede the body of the clauses with which they are connected. Reasons for the relative placement of main and dependent clauses within the sentence are discussed in Diessel (2001, 2005), while Diessel (2004) points out that children acquiring complex sentences in English as their first language make great use of the strategic positioning of main and dependent clauses within a sentence.

3. A caveat: sources of dependent clause marker constructions other than borrowing Although borrowing is a widespread process, we have to keep the degree of borrowing of dependent clause markers into a language in its proper perspective. It would be foolish for us to assume from the outset that the only way in which speakers of a language which has hitherto made little or no use of dependent clause markers can equip themselves with such elements is for them to borrow such markers in large quantities from languages of greater prestige with which they are in contact. Although borrowing is one of the means of acquiring such forms, it is in no sense the only means, any more than zero-marking such relations is. Processes of grammaticalization that expand the meaning of a morpheme already being used in another sense provide one pathway for developing dependent clause markers. For instance, English OR is historically related to OTHER, from which it has developed, and has been semantically bleached in the course of the history of English (www.oed.com). Similarly, as I pointed out in Section 2.2, the use by speakers of Yiddish of what were originally subjunctive forms of the verb ‘to be’ to express ‘either . . . or . . .’ is an example of (secondary) grammaticalization of a preexisting form for a new purpose (this is a case of grammaticalization which is also found in German, though the subjunctive mood has been lost in Yiddish). Such a construction is not exclusive to Germanic languages; it can be found, for example, in Spanish sea . . . sea . . . ‘either . . . or . . . ,’ literally ‘that it be . . . that it be. . . .’ Kortmann (2001) contains a wealth of information about the rise, development, and expansion of adverbial conjunctions in European languages. The secondary use of

326

Anthony P. Grant

formerly spatial adverbs to express temporal relations is also very frequent in these cases.6 Borrowing in the form of transfer of fabric should be seen as just one resource for linguistic expansion that will provide subordinating conjunctions. Transfer of pattern has also long played a part in the creation of new subordinators. One notes, for instance, that French parce que ‘because’ and Old English for þœ´m þe are exactly congruent: both literally mean ‘for this that.’7 The use of former discourse markers or pronouns, or even members of other form-classes, as conjunctions is also well attested in studies of grammaticalization. English THAT and the cluster of forms comprising WHICH, WHO(M), WHOSE were originally demonstrative or interrogative pronouns, and they still serve these roles, just as the word ho, he#, to which had started out as the definite article in Attic Greek, also (and secondarily) served as a subject relative pronoun. English BUT comes from Old English be-u#tan ‘outside,’ which completely supplanted inherited ac. ALTHOUGH is both a borrowing and a form deriving from a discourse particle, while UNTIL is borrowed from a Norse noun phrase und till ‘up to the end (of ).’ OR derives from the same stem which gives OTHER. As to temporal adverbial forms such as AFTER and BEFORE, which in English as in many other languages are secondary developments from spatial adpositions, Leech, Rayson, and Wilson (2001), drawing on the massive British National Corpus, point out that BEFORE occurs 434 times per million words as a preposition and 305 times per million words as a conjunction. For AFTER the proportions are 927 and 233 occurrences per million words respectively, though SINCE’s figures are 295 occurrences per million words as a conjunction and only 178 as a preposition. It would also be erroneous for us to assume ab initio that all languages will use dependent clause markers to the same extent and with the same frequency in their equivalents of the same prompt sentences or clauses, and that they continue to operate according to this principle to this day. An analysis of comparable texts in some of these languages, which perforce are translations (for instance of admittedly very formal texts such as the Paternoster and the United Nations Universal Declaration of Human 6. An interesting counterexample to this is the English compound preposition in front of, in which front is a borrowing from French. The earlier preposition before can still be used spatially, but its use as a temporal adverbial conjunction is its most frequent one nowadays. 7. So also does Russian potomu cˇto, which has been borrowed bodily into Kildin Saami (Rießler to appear).

Contact, convergence, and conjunctions

327

Rights, translations of both of which I have examined for several of the languages included here) makes clear that this is not the case. See www. christusrex.org for the Lord’s Prayer, and www.un.org/en/documents/ udhr/ for the Universal Declaration of Human Rights. Indeed, some languages use conjunctions in their version of this text much more than others do, these others making more use of asyndeton or of clause-combining clitics. This last situation is very much the case with Turkish, in which the use of borrowed conjunctions of Farsi origin is well attested (it also makes much use of asyndeton). Nonetheless in Turkish the di¤erent kinds of dependent clauses are frequently formed using bound morphs incorporated in dependent verb forms. In fact, an examination and comparison of the 1000 commonest Turkish and English words, based on a Wikipedia frequency count for the Turkish data and Leech, Rayson, and Wilson (2001) for written English data, shows that the English words AND, THAT (when it is lemmatized by Leech, Rayson, and Wilson as a complementizer), BUT, IF, OR, and BECAUSE occur at 3rd, 12th, 28th, 33rd, 49th, and 117th places in terms of frequency in English, while their Turkish equivalents ve, ki, amma, eg˘er, (ve)ya, and c¸u¨nku¨, all of them borrowings from Farsi (some of them being loans into Farsi from Arabic), occur at 2nd, 21st, 39th, 238th, 43rd (97th in the case of veya), and at 112th places respectively. Indeed, only the relatively poor showing of eg˘er for ‘if,’ coming in at 238th place, is cross-linguistically somewhat surprising. But its rather low ranking when compared with other borrowed conjunctions becomes clear when one understands that Turkish has a complete conditional mood in its verbal system for the expression of ‘if ’-clauses, and that (as Lewis 1967: 270 points out) the use of eg˘er, which itself requires use of the conditional mood of the verb, at the beginning of a Turkish sentence serves largely to indicate to the listener that a sentence containing a conditional clause or two is about to begin. For its part Swahili similarly expresses conditionals with an infixed conditional verbal mood with or without the clause-initial use of kama ‘if.’

4. The data for the study Data on borrowed conjunctions have been taken from English and 21 other languages from around the world. Table 1 gives some basic information on the languages being examined. For each language I supply the name of the language, its genealogical a‰nity, and the language(s) that

328

Anthony P. Grant

Table 1. The languages used in the sample Language

Genealogical a‰liation

Language(s) of influence (italicized languages furnish conjunction)

Are modern speakers bilingual in the language of influence?

Borrowed items on Swadesh 207item list?

English

Indo-European: West Germanic

French, Norse, Latin

No, most never were

16%

Kalderash Romani

Indo-European: Indic

Romanian, Greek, South Slavic, Iranian, Armenian, contiguous languages

Not since late nineteenth century in the case of many Kalderara with Romanian

25%

Eastern Yiddish

Indo-European: West Germanic

Hebrew-Aramaic, West Slavic

Not completely; much knowledge of Hebrew

5%

Kildin Saami

Uralic: Finnic: Saami

Russian, older forms from Germanic

Yes, in Russian

Not known but at least 20%; mostly from Germanic

Livonian

Uralic: Finnic: Balto-Finnic

Latvian; Low German

Yes, in Latvian