Grammatical Borrowing in Cross-Linguistic Perspective (Empirical Approaches to Language Typology)

Grammatical Borrowing in Cross-Linguistic Perspective

≥

Empirical Approaches to Language Typology 38

Editors Georg Bossong Bernard Comrie Yaron Matras

Mouton de Gruyter Berlin · New York

Grammatical Borrowing in Cross-Linguistic Perspective

Edited by Yaron Matras Jeanette Sakel

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

앝 Printed on acid-free paper which falls within the guidelines of the 앪 ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication Data Grammatical borrowing in cross-linguistic perspective / edited by Yaron Matras, Jeanette Sakel. p. cm. ⫺ (Empirical approaches to language typology ; 38) Includes bibliographical references and index. ISBN 978-3-11-019628-3 (cloth : alk. paper) 1. Language and languages ⫺ Foreign elements. 2. Grammar, Comparative and general. I. Matras, Yaron, 1963⫺ II. Sakel, Jeanette, 1973⫺ P324.G73 2007 410⫺dc22 2007042917

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at http://dnb.d-nb.de.

ISBN 978-3-11-019628-3 ISSN 0933-761X © Copyright 2007 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Printed in Germany.

Contents List of contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Yaron Matras and Jeanette Sakel Types of loan: Matter and pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Jeanette Sakel The borrowability of structural categories . . . . . . . . . . . . . . . . . . . . . . . . . . 31 Yaron Matras Grammatical borrowing in Tasawaq . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 Maarten Kossmann Grammatical borrowing in K’abeena . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 Joachim Crass Grammatical borrowing in Likpe (Sɛkpɛlé) . . . . . . . . . . . . . . . . . . . . . . . 107 Felix K. Ameka Grammatical borrowing in Katanga Swahili. . . . . . . . . . . . . . . . . . . . . . . 123 Vincent A. de Rooij Grammatical borrowing in Khuzistani Arabic . . . . . . . . . . . . . . . . . . . . . 137 Yaron Matras and Maryam Shabibi Grammatical borrowing in Domari. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 Yaron Matras Grammatical borrowing in Kurdish (Northern Group) . . . . . . . . . . . . . . . 165 Geoffrey Haig Arabic grammatical borrowing in Western Neo-Aramaic. . . . . . . . . . . . . 185 Werner Arnold

vi

Contents

Grammatical borrowing in North-eastern Neo-Aramaic . . . . . . . . . . . . . 197 Geoffrey Khan Grammatical borrowing in Macedonian Turkish . . . . . . . . . . . . . . . . . . . 215 Yaron Matras and Şirin Tufan Grammatical borrowing in Kildin Saami . . . . . . . . . . . . . . . . . . . . . . . . . 229 Michael Rießler Grammatical borrowing in Yiddish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245 Gertrud Reershemius Grammatical borrowing in Hungarian Rumungro . . . . . . . . . . . . . . . . . . 261 Viktor Elšík Grammatical borrowing in Manange . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283 Kristine A. Hildebrandt Grammatical borrowing in Indonesian . . . . . . . . . . . . . . . . . . . . . . . . . . . 301 Uri Tadmor Grammatical borrowing in Biak . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329 Wilco van den Heuvel Sino-Vietnamese grammatical borrowing: An overview. . . . . . . . . . . . . . 343 Mark J. Alves Recent grammatical borrowing into an Australian Aboriginal language: The case of Jaminjung and Kriol. . . . . . . . . . . . . . . 363 Eva Schultze-Berndt Grammatical borrowing in Rapanui . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387 Steven Roger Fischer Grammatical borrowing in Nahuatl. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403 Una Canger and Anne Jensen Grammatical borrowing in Yaqui . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419 Zarina Estrada Fernández and Lilián Guerrero

Contents

vii

The case of Otomi: A contribution to grammatical borrowing in cross-linguistic perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435 Ewald Hekking and Dik Bakker Grammatical borrowing in Purepecha. . . . . . . . . . . . . . . . . . . . . . . . . . . . 465 Claudine Chamoreau Grammatical borrowing in Imbabura Quichua (Ecuador) . . . . . . . . . . . . 481 Jorge Gómez-Rendón Grammatical borrowing in Paraguayan Guaraní. . . . . . . . . . . . . . . . . . . . 523 Jorge Gómez-Rendón Grammatical borrowing in Hup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 551 Patience Epps Mosetén borrowing from Spanish. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 567 Jeanette Sakel Index of subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581 Index of authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594

List of contributors

Mark Alves Montgomery College

Wilco van den Heuvel University of Manchester

Felix Ameka Leiden University

Kristine Hildebrandt University of Manchester

Werner Arnold University of Heidelberg

Anne Jensen University of Copenhagen

Dik Bakker University of Lancaster

Geoffrey Khan University of Cambridge

Una Canger University of Copenhagen

Maarten Kossmann Leiden University

Claudine Chamoreau CELIA (CNRS-IRD-INALCOPARIS VII)/CIESAS-Mexico

Yaron Matras University of Manchester

Joachim Crass University of Mainz Viktor Elšík Charles University, Prague Patience Epps University of Texas at Austin Zarina Estrada Fernández Universidad de Sonora Steven Roger Fischer Auckland, NZ Jorge Gómez-Rendón University of Amsterdam Lilián Guerrero Universidad de Sonora

Gertrud Reershemius Aston University, Birmingham Michael Rießler Humboldt-Universität zu Berlin Vincent de Rooij Amsterdam School for Social Science Research, University of Amsterdam Jeanette Sakel University of the West of England Eva Schultze-Berndt University of Manchester Maryam Shabibi University of Manchester

Geoffrey Haig University of Kiel

Uri Tadmor MPI for Evolutionary Anthropology/ Jakarta field station

Ewald Hekking Universidad de Querétaro

Şirin Tufan University of Manchester

Introduction Yaron Matras and Jeanette Sakel

1. Borrowing in cross-linguistic perspective1 Like any metaphor, the term “borrowing” has its drawbacks. We have decided to ignore possible reservations about the term, both in the title of this collection and in the advice on the use of terminology which we have given to the contributors. Whether the “borrowed” substance is perceived as belonging or as alien, whether its source is described as a “donor” and the language into which it is integrated as the “recipient”, “copier”, or “replica”, seems immaterial as long as clarity prevails as to the kind of phenomena that we are addressing when talking about contact-induced change. We use the term “borrowing” as a cover-term for the adoption of a structural feature into a language as a result of some level of bilingualism in the history of the relevant speech community. This collection is about the structural effects of language contact. We have asked each contributor (or pair of contributors) to focus on the diachronic impact that language contact has had on the structure of a particular language. Accompanying these descriptions are comments on societal multilingualism, the roles that are assigned to various languages in the community, patterns of language mixing, and issues of language policy and language education, which are dealt with in relation to each case study in the introductory sections of each chapter. The purpose of the compilation is to be able to compare the effects of different kinds of contact on different kinds of languages, and so to help forward our understanding of universal effects of language contact.

2. Sampling in contact linguistics Linguistic typology tries to make generalizations about human languages. For this purpose, typologists rely on sampling methods. Language samples make it possible to make generalizations without studying each and every individual language, which would be a costly and time-consuming endeavour. Since Greenberg (1966) it has been accepted that samples should try and reflect at least the present-day diversity of languages in order to be truly

2

Yaron Matras and Jeanette Sakel

representative of human language. Most researchers have therefore made an effort to avoid areal or genetic biases when compiling a sample, though basic typological parameters and extralinguistic factors have played less of a role (cf. Comrie 1981, Stassen 1985, Dryer 1989; Rijkhoff et al. 1993). At times samples have been used by a group of studies in a coordinated fashion, to study the distribution of several different phenomena across the same set of languages; this was partly the case in the EUROTYP project; and in the World Atlas of Linguistic Structures – see Comrie, Dryer, Gil and Haspelmath 2005). Nevertheless, it is fair to say that on the whole linguistic samples have been used in order to study a particular structural phenomenon or structural category. The cross-linguistic study of structural borrowing is a challenge at a different level. Firstly, borrowing can affect many different categories. A comprehensive, comparative study of borrowing must therefore take into account both the “horizontal” diversity of sample languages, and the “vertical” diversity of structural categories on which contact can have an impact. This already makes the sampling of borrowing phenomena a much more complex task than the comparative study of any particular domain of linguistic structure. Second, borrowing is a historical dimension, which can only be identified and assessed if diachronic information on the relevant language(s) is available. This factor seriously disadvantages the consideration of entire areas of the world from which we lack secure and reliable information on linguistic diachrony, and so in effect it counteracts the need to maintain areal diversity in a representative sample. Finally, there is general agreement that the outcomes of language contact (or, to be more precise: of widespread bilingualism in a community) depend not just on structural factors, but to a great extent on extralinguistic factors. The duration and intensity of cultural contact, the roles and status of the participating languages, the degree of institutional support awarded to languages in various stages of their history (e.g. the presence of literacy or use in the public, acrolectal domain), and speakers’ attitudes toward their own and their neighbours’ forms of speech – all these play a vital role in determining the direction of change and so in shaping the structural outcome of language contact constellations. In order to investigate the universal possibilities of contact-induced change, one needs to take sociolinguistic factors into account. The ideal sample for the investigation of contact is therefore one that is, like other samples, stratified to take into account various language-genetic groupings, structural types, and regions of the world; but in addition it must also be informed about diachronic depth and allow the author or user to control factors that are external to language.

Introduction

3

Given these difficulties of sampling it is not surprising that most attempts to make generalizations about contact-induced change have been based on casual observations, rather than on systematic comparative studies. This is true of Moravcsik’s (1978) discussion of borrowing universals even within the context of the Greenbergian project, as well as of Thomason and Kaufman’s (1988) frequently-cited borrowing scale. Some generalizations about borrowing have been proposed with reference to a case study of just one single contact situation (cf. Haugen 1950, van Hout and Muysken 1994, Ross 2001, Field 2002), while some have concentrated on identifying counter-examples to generalizations proposed by others (cf. Cambpell 1993, cf. also Thomason 2001). To the extent that samples have been used in contact linguistics, they have tended to control one of the key factors in the contact situation, such as the donor or the recipient language, or even the type of category affected. Stolz and Stolz (1996, 1998), for example, discuss the borrowing of Spanish function words into a diverse set of languages in Central America and the Pacific. Johanson (2002) discusses the contact behaviour of Turkic languages, and Matras (2002) and Elšík and Matras (2006) evaluate structural borrowing from a diverse set of European contact languages in the dialects of Romani. It is noteworthy that in these samples the extra-linguistic parameters are also kept constant. Thus, Spanish is the colonial language in the Pacific and in Central America, Romani is an oral language of dispersed, socially marginalized, bilingual ethnic minorites, and Persian has played a similar role in the history of various Turkic languages. A wealth of data for comparison from various contact situations can be found in a number of collections devoted to case studies of language contact (e.g. Gilbers, Nerbonne, and Schaeken 2000, Aikhenvald and Dixon 2001, Matras, McMahon and Vincent 2005). Aikhenvald and Dixon (2006) especially contains contributions that cover a wide range of languages, regions, and contact phenomena. These are accompanied by important summary observations on general factors and constraints that operate in language contact situations (cf. Aikhenvald 2006). Put together, these and other excellent contributions to the study of language contact have taken us a significant step forward toward a typology of contact-induced language change. Still missing, however, from the body of work produced in recent years is an attempt at a systematic comparison of the behaviour of grammatical categories across a sample of languages in contact. Of interest is the question whether some grammatical categories are universally more susceptible to contact-induced change than others. A further question is whether there is any recurring correlation between the borrowing of structures in one cat-

4

Yaron Matras and Jeanette Sakel

egory, and those belonging to another. Both these issues can be expressed in terms of hierarchies of borrowing. These in turn may contain either implicational statements (if X is borrowed, then Y is also borrowed), or just plain frequency statements (X is borrowed more frequently in the sample than Y, and hence it can be said to be more prone to borrowing than Y). Equally of interest is the correlation between a category and the type of contact-related change that is more likely to affect it: a shift in meaning or in the distribution of existing structures (which we term “pattern replication” below), or the actual adoption of a structure from another language for circulation in the recipient system (“matter replication”). Finally, we are interested in the interaction between the contact behaviour of a category, and other factors that condition the nature of the contact situation, including both language internal features (such as the typological parameters of the languages involved) and extra-linguistic features (such as the type of bilingualism and the roles played by the respective languages in various domains of communicative interaction). The purpose of this collection is to facilitate a discussion of questions of this kind, and to provide information on the basis of which these questions can be addressed.

3. The data compilation tool A major difficulty in sampling for the purposes of contact linguistic studies is the accessibility of relevant information. Unlike synchronic structural facts about language, the description of borrowing requires diachronic information. Even if such information is available in principle for some languages, it is not always the case that it is included into grammatical descriptions. Grammar books do not generally tend to highlight borrowings at all. Identifying borrowed structures requires a high degree of expertise and specialization in the language and its history, and familiarity, at the very least, with the languages with which it has been in contact. Relevant extralinguistic information is often missing from grammatical descriptions, too. For these reasons, it is hardly feasible for a lone researcher to survey published descriptions of various languages in order to compile a representative sample corpus of grammatical borrowing. Sampling in this field is best achieved through team effort, with experts contributing first-hand information on contact-induced phenomena. Underlying the team effort on which this volume is based is a uniform questionnaire, formatted as a user-friendly database (in FilemakerPro 67).

Introduction

5

The aim of the questionnaire was to obtain a representative and comparable sample of data on contact-induced change in a variety of languages. For this purpose it was distributed to the contributors as a detailed reference grid. The questionnaire can of course continue to constitute a description standard for language contact phenomena, serving as a checklist for information to be covered in an exhaustive description of borrowing into any given language. The questionnaire opens with information on relevant metadata (source of information, affiliation to sub-samples, date of input, and so on), and continues to cover extralinguistic information about the language and the speech community. The remaining chapters cover all principal domains of structure: Phonology, Typology (a characterization of principal typological traits), Nominal structures, Verbal structures, Other parts of speech (e.g. quantifiers, indefinites, phasal adverbs, discourse markers and connectors), Constituent order, Syntax (clause combining), and Lexicon (general information on the presence of lexical loans in various semantic domains, as well as specific

Figure 1. Information page of the Language Convergence database (entry: Domari) showing open chapter menu in the top left corner

6

Yaron Matras and Jeanette Sakel

Figure 2. Encoding the sociolinguistic situation (Mosetén)

questions on expressions of time and space). Using the “Layout” function in FilemakerPro, each chapter is displayed on a separate page, accessible through a menu box (Figure 1). Individual records, each representing a language in contact, can be tagged for different kinds of contact constellations or sub-samples. We can distinguish, for instance, borrowing situations where just two languages are in contact, from observations on a more widespread regional distribution of shared phenomena, or “linguistic areas”. Special attention is given to the coding of a series of extra-linguistic indicators (Figure 2), allowing the user to assess the correlations between sociolinguistic factors and the contact behaviour of a language. Depth of contact is taken into account by distinguishing, where applicable, several different layers of contact (see also Matras 1998): The Current contact language (i.e. the object of widespread bilingualism), a Recent contact language (that may still be spoken by an older generation of speakers), and an Old contact language that has made an impact on the

Introduction

7

Figure 3. Coding of contact languages (Manange)

language in the past, but has little or no contemporary role in the speech community (Figure 3). In order to be able to investigate the precise effects of contact on structural compositions, a distinction is maintained throughout the questionnaire between the replication of linguistic matter (MAT) consisting of actual phonological segments, and the replication of patterns (PAT), which pertains to the semantic and grammatical meaning and the distribution of a construction or structure (see Matras and Sakel 2007). This distinction is encoded alongside every relevant description of a contact phenomenon in an individual category (Figure 4). The advantages of working with the questionnaire database are obvious: While the checklist ensures uniform and comprehensive coverage of the same phenomena, and so comparability throughout the sample, the database allows to filter and to query the results, to retrieve examples of the various kinds of contact phenomena, and to view correlations among the data sets (see e.g. Figure 5).

8

Yaron Matras and Jeanette Sakel

Figure 4. Encoding MAT (matter) and PAT (pattern) replications (Domari)

A sample comparison – piloting just two languages, Kelderash Romani and Mosetén – was already presented, based on the database questionnaire, in Sakel and Matras (2007). The present collection features a case-by-case assessment of borrowing. A preliminary assessment of some salient, common patterns is provided in the two evaluation chapters, by Matras and by Sakel. Our intention at this point is to continue to expand the data sample, and eventually to make the data accessible to users online.

4. Coverage of phenomena and languages In assessing the diachronic impact of contact, many of the contributors faced the dilemma of how to tell apart the ongoing effects of current bilingualism

Introduction

9

within a speech community, and structural changes that may be regarded as permanent. The distinction is often referred to as one between “borrowing” and “code-switching”, but since there is widespread agreement that alternational code-switching is unlikely to fossilize into a permanent pattern (see Backus 2003, 2005), it is in fact the distinction between insertional switches and borrowing that requires clarification. Our recommendation to the contributors was generally to regard as borrowings those cases where a speaker’s selection of a particular structure is not facilitated by a choice between two alternatives, and is not triggered by the speech situation or by the topic of conversation, nor by a need for a special conversation effect. We thus consider as borrowings all structures that appear with a certain regularity and are used by different speakers in a range of different situations. Problem cases are otherwise highlighted in the individual chapters.

Figure 5. Comparing results on the presence/absence of contact phenomena for the following categories: Connectors (O_PA_co), Focus particles (O_PA_fo), Numerals and quantifiers (O_nu), and Personal pronouns (O_P_pp)

10

Yaron Matras and Jeanette Sakel

Even when drawing on first-hand data and the expertise of field linguists, gaps in the coverage are inevitable; in our case this is due mainly to limitations of space, resources, and the constraints of the production schedule. The sample nevertheless grants representation to most areas of the world: Saharan and sub-Saharan Africa (Tasawaq, K’abeena, Likpe, Katanga Swahili), the Middle East (Khuzistani Arabic, Domari, Kurmanji-Kurdish, Western and North-eastern Neo-Aramaic), the Balkans (Macedonian Turkish), Europe (Kildin Saami, Yiddish, Hungarian Rumungro), the Himalaya (Manange), South Asia (Indonesian, Biak), East Asia (Vietnamese), Australia (Jaminjung), the Pacific (Rapanui), Central America (Nahuatl, Yaqui, Otomi, Purepecha), and South America (Imbabura Quichua, Guaraní, Hup, Mosetén). It contains languages with a tradition of native literacy (Vietnamese, Indonesian, Arabic, Turkish, Swahili, and arguably also Neo-Aramaic) and others without one; languages of ethnic minorities, those that are or were majority languages, and regional languages in post-colonial settings; languages with a single contemporary contact language as well as those spoken in either a multilingual setting or a linguistic area. Lacking representation in our sample are languages of North America, Central Asia, and Siberia. Pidgins, creoles, and mixed languages were not considered either, since their borrowing characteristics are potentially different in principle than those of other languages that are not themselves the product of recent contacts (but see Thomason 1997, Matras and Bakker 2003). The present volume appears alongside a two-part publication devoted to lexical borrowing, which is the result of the Loanword Typology project based at the Max Planck Institute for Evolutionary Anthropology in Leipzig. We hope that together, both publications will help shed new light on an ancient process that will have been as instrumental in shaping the development of the world’s languages as human contacts has been in shaping general human cultural experience. Note 1. We gratefully acknowledge support from the Arts and Humanities Research Council for a three-year research project on “Language Convergence and Linguistic Areas” (award number RG/AN4725/APN16320), during which the questionnaire and database tools used for this collection were developed, as well as support from the School of Languages, Linguistics and Cultures for a workshop on “Grammatical Borrowing in Cross-Linguistic Perspective” in Manchester in September 2005 and for the production of the present volume. We also wish to

Introduction

11

thank Georg Bossong for helpful comments, Peter Kahrel for typesetting and copy-editing the manuscript, and Ursula Kleinhenz for her support during the production process.

References Aikhenvald, Alexandra Y., and R. M. W. Dixon (eds.) 2001 Areal Diffusion and Genetic inheritance. Oxford: Oxford University Press. 2006 Grammars in Contact: A Cross-Linguistic Typology. Oxford: Oxford University Press. Aikhenvald, Alexandra Y. 2006 Grammars in contact: A cross-linguistic perspective. In: Alexandra Y. Aikhenvald and R. M. W. Dixon (eds.), Grammars in Contact: A Cross-Linguistic Typology, 166. Oxford: Oxford University Press. Backus, Ad 2003 Can a mixed language be conventionalized alternational codeswitching? In: Y. Matras and P. Bakker (eds.), The Mixed Language Debate: Theoretical and Empirical Advances, 237270. Berlin: Mouton de Gruyter. 2005 Codeswitching and language change: One thing leads to another? International Journal of Bilingualism 9 (3/4): 307340. Campbell, Lyle 1993 On proposed universals of grammatical borrowing. In: H. Aertsen and R. Jeffers (eds), Historical Linguistics 1989: Papers from the 9th International Conference on Historical Linguistics, 91109. Amsterdam: John Benjamins. Comrie, Bernard 1989 Language Universals and Linguistic Typology: Syntax and Morphology. Oxford: Blackwell. Dryer, Matthew S. 1989 Large linguistic areas and language sampling. Studies in Language 13: 257292. Elšík, Viktor, and Yaron Matras 2006 Markedness and Language Change: The Romani Sample. Berlin: Mouton de Gruyter. Field, Fredric 2002 Linguistic Borrowing in Bilingual Contexts. Amsterdam: Benjamins. Gilbers, D. G., J. Nerbonne, and J. Schaeken (eds.) 2000 Languages in Contact. Amsterdam: Rodopi. Greenberg, Joseph H. (ed.) 1966 Universals of Language. Cambridge, MA: MIT Press.

12

Yaron Matras and Jeanette Sakel

Haugen, Einar 1950 The analysis of linguistic borrowing. Language 26 (2): 210231. Haspelmath, Martin, Matthew Dryer, David Gil, and Bernard Comrie (eds.) 2005 The World Atlas of Language Structures. Oxford: Oxford University Press. Johanson, Lars 2002 Structural Factors in Turkic Language Contacts. London: Curzon. Matras, Yaron 1998 Utterance modifiers and universals of grammatical borrowing. Linguistics 36 (2): 281331. 2002 Romani: A Linguistic Introduction. Cambridge: Cambridge University Press. Matras Yaron, and Peter Bakker (eds.) 2003 The Mixed Language Debate. Berlin: Mouton de Gruyter. Matras, Yaron, and Jeanette Sakel 2007 Investigating the mechanisms of pattern replication in language convergence. Studies in Language 31 (4): 829865. Matras, Yaron, April McMahon, and Nigel Vincent (eds.) 2006 Linguistic Areas: Convergence in Historical and Typological Perspective. Houndmills: Palgrave. Moravcsik, Edith 1978 Language contact. In: Joseph H. Greenberg, Charles A. Ferguson, and Edith A. Moravscik (eds.), Universals of Human Language, Volume 1, 93122. Stanford: Stanford University Press. Rijkhoff, Jan, Dik Bakker, Kees Hengeveld, and Peter Kahrel 1993 A method of language sampling. Studies in Language 17: 169203. Ross, Malcolm 2001 Contact-induced change in Oceanic languages in north-west Melanesia. In: Alexandra Aikhenvald and R. M. W. Dixon (eds.), Areal Diffusion and Genetic Inheritance: Problems in Comparative Linguistics, 134166. Oxford: Oxford University Press. Sakel, Jeanette, and Yaron Matras 2007 Modelling contact-induced change in grammar. In: Thomas Stolz et al. (eds.), Aspects of Language Contact: New Theoretical, Methodological and Empirical Findings With Special Focus on Romanisation Processes. Berlin: Mouton de Gruyter. Stassen, Leon 1985 Comparison and Universal Grammar. Oxford: Blackwell. Stolz, Christel, and Thomas Stolz 1996 Funktionswortentlehnung in Mesoamerika, Spanisch–Amerindischer Sprachkontakt. In Sprachtypologie und Universalienforschung 49 (1): 86123.

Introduction 1997

13

Universelle Hispanismen? Von Manila über Lima bis Mexiko und zurück: Muster bei der Entlehnung spanischer Funktionswörter in die indigenen Sprachen Amerikas und Austronesiens. Orbis 39 (1) [1996– 1997]: 177. Thomason, Sarah G. (ed.) 1997 Contact Languages: A Wider Perspective. Amsterdam: Benjamins. Thomason, Sarah G. 2001 Language Contact: An Introduction. Edinburgh: Edinburgh University Press. Thomason, Sarah, and Terrence Kaufman 1988 Language Contact, Creolization and Genetic Linguistics. Berkeley: University of California Press. van Hout, Roeland, and Pieter Muysken 1994 Modelling lexical borrowability. Language Variation and Change 6: 3962. Weinreich, Uriel 1966 Languages in Contact: Findings and Problems. The Hague: Mouton (first publ. 1953).

Types of loan: Matter and pattern Jeanette Sakel

1. Introduction1 A central concern of contact linguistics has long been to categorize the ways in which elements are borrowed from one language into another. For this purpose Matras and Sakel (2004) introduced the terms matter (MAT) and pattern (PAT) in the questionnaire on which the sample of contact situations in this book is based (cf. also subsequent publications on the issue, such as Matras and Sakel 2007). In the present chapter I will re-visit the definition of MAT and PAT, as well as address what this distinction could mean in phonology. I will furthermore give an overview of the overall distribution of MAT/PAT in the languages of the sample in order to address the validity of a MAT/PAT distinction in the categorization of contact situations.

2. Definitions MAT and PAT denote the two basic ways in which elements can be borrowed from one language into another. We speak of MAT-borrowing when morphological material and its phonological shape from one language is replicated in another language. PAT describes the case where only the patterns of the other language are replicated, i.e. the organization, distribution and mapping of grammatical or semantic meaning, while the form itself is not borrowed. In many cases of MAT-borrowing, also the function of the borrowed element is taken over, that is MAT and PAT are combined.2 In other instances, MAT and/ or PAT are borrowed, but deviate considerably in their form or function from their original source. In some categories, making a distinction between MAT and PAT does not make much sense. For example, word-order changes will invariably be PAT. In other areas, such as phonology, the MAT/PAT distinction applies only in a restricted way, as MAT and PAT are primarily defined as functioning above the morpheme level. The concept behind MAT and PAT is well-grounded in the literature, but only rarely figures in the categorization of contact situations. One exception is Heath (1984: 367), who bases his approach to language contact on this opposition, distinguishing between

16

Jeanette Sakel

“direct transfer of forms from the other language” and “structural convergence”. Other approaches to language contact mention similar distinctions, often with very different terminology, such as Haugen’s (1950) “importation” for outright borrowing and “substitution” for loanshifts or calques. Likewise, Weinreich ([1953] 1966: 7) speaks about “transfer of elements” and “interference without outright transfer”, which re-appears in Weinreich’s distinction between “source” and “recipient language” for MAT-borrowings, and “model” and “replica language” for PAT-borrowings. The distinction is mentioned in many subsequent approaches to language contact, such as Gołąb’s (1956, 1959) “form” versus “substance”, Johanson’s (1992) “global copying” and “partial copying”, Nau’s (1995) “material borrowing” and “loanmeaning, loan-translation” and Treffers-Daller and Mougeon’s (2005: 95) “borrowing, code-switching” versus “transfer”. Many other studies of language contact pay only minor attention to this matter, such as Thomason and Kaufman (1988), who do not include this classification in their influential borrowing scale (1988: 74). Some approaches to contact focus almost exclusively on either MAT or PAT. In this manner, approaches to substrate influence and contact-induced grammaticalization (Siegel 1997; Keesing 1991; Heine and Kuteva 2005) focus mainly on PAT-type loans. Likewise, linguistic areas have often been described as zones in which PAT-borrowing appears. On the other hand, much of the early literature on code-switching primarily addresses contact phenomena of the MAT type, while more recently PAT has been integrated into frameworks of code-switching (cf. Savić 1995, Bolonyai 1998 and MyersScotton 2002: 2122). There are reasons why some approaches seem to favor either MAT or PAT: when studying linguistic areas or substrate influence, focus is often on PAT because a major part of the loans in these situations are of this type and indeed areas and substrate influence are often defined as displaying mainly pattern-loans. On the other hand, when stuyding situations of code-switching, MAT is often very prominent.

3. Integration of MAT and PAT loans Let us now look at how grammatical MAT/PAT-loans are integrated into the recipient language. In many cases of MAT or PAT-borrowing, not the entire function or form is taken over, but the borrowed elements differ from their original source. Take for example the way in which Domari copies Arabic aspect marking (Matras, this volume); Domari borrows Arabic auxiliaries as

Types of loan: Matter and pattern

17

MAT, but it does not simultaneously make use of a subjunctive verb form as in Arabic because it has its own subjunctive. Hence, not the entire construction of aspect marking is taken over, but only parts of it. In Otomi the functions of a borrowed form are extended: the shortened form ko from Spanish como is used in a number of constructions, also expressing ‘made of’, a function it does not have in Spanish (Hekking and Bakker, this volume). Hence, when MAT-elements are borrowed, their functions are not necessarily the same as in the source language; sometimes only parts of the function are borrowed, sometimes the functions are extended and the loans are rarely mere copies of those of their counterparts in the source language. Also the forms of loans are frequently adjusted, for example by phonological integration of MAT-loans into the recipient language, which in some cases makes them difficult to identify as loans without a careful analysis. In the same way, PAT-loans inherently involve a process of grammaticalization, often leading to different patterns as those in the source language (cf. Heine and Kuteva 2005). PAT-borrowing is facilitated by a pivot common to both languages (Matras and Sakel 2007). Such cases of PAT-adjustment in the sample include the shift in grammatical meaning of native Yiddish elements to correspond to Slavic aspectual markers (Reershemius, this volume).

4. MAT and PAT in phonology How could the distinction between MAT and PAT-loans be employed in phonology? In most contact situations, MAT-loans involve phonological changes that can go in two directions: (1) MAT-borrowed element is phonologically integrated into the recipient language; e.g. Mosetén ishkweera for Spanish escuela ‘school’. (2) MAT-borrowed element is not integrated and may introduce new phonemes into the recipient language; e.g. in Jaminjung, where loan-phonemes have risen to phoneme status within the language and are now used with native words (Schultze-Berndt, this volume). Only the latter strategy, (2), involves phonological borrowing in which elements from the source language are transferred into the recipient language. One could argue that loans in (2) are MAT-loans if they introduce a new phone, and PAT-loans if they introduce a new phoneme. The loss of certain phonological distinctions would possibly be a case of PAT, since no new material

18

Jeanette Sakel

is introduced, and a change in the stress patterns could be counted as PAT as well, since it involves the overall patterning rather than actual forms. A change in tone, on the other hand, could be MAT-borrowing if it occurred in isolated loanwords and PAT if it affected the whole language. A cautious attempt to classify the phonological contact phenomena found in the sample is the following: – Borrowing of individual phonemes that are also used in native elements is found in Vietnamese, Indonesian, Jaminjung and Paraguayan Guaraní. This type of loan would probably best be classified as MAT, since it involves borrowing of phonological material without a major disruption of the phonological system. – Borrowing of a series of distinctions is common. K’abeena, Rumungro and Kildin Saami have borrowed a palatal series. K’abeena has furthermore borrowed ejectives, while Tasawaq gained pharyngealization and Vietnamese had a retroflex series introduced by contact. These would be MAT-loans in the sense that elements are borrowed and PAT-loans in the sense that a whole series of phonemes within the phonological system is affected by the contact. Adaptation of stress, syllable structure, prosody or tone systems would classify as PAT-borrowing in phonology since they affect the system, rather than individual elements. Stress patterns are adjusted in Western Neo-Aramaic, Yiddish, Hup, Manange, North-eastern Neo-Aramaic and Paraguayan Guaraní. The syllable structure is adapted due to contact in Yiddish and Indonesian, allowing consonant clusters. Prosody is affected in Hup and Jaminjung, while tone has undergone contact-induced changes in Vietnamese and Manange. Hence, MAT and PAT could be used to classify contact phenomena in the phonology, even though the strength in having this MAT/PAT distinction makes most sense in loans from the morpheme-level upwards.3

5. The MAT/PAT distinction in grammar: findings from the sample We will discuss the data from the sample by looking at the following cases: (1) situations with overall MAT-borrowing; (2) situations with overall PATborrowing; and (3) hierarchical relations between contact languages. When talking of “overall” contact phenomena, we deliberately do not quantify loans

Types of loan: Matter and pattern

19

in contact situations to avoid sampling problems, but rather only look at cases where the overwhelming majority of the loans are of either type.

5.1. Situations with overall MAT-borrowing In the sample, Jaminjung, Biak and Vietnamese stand out in having predominantly MAT-loans. Jaminjung is in contact with Kriol, a creole which is likely to have been influenced by the native languages of the area (cf. Schultze-Berndt, this volume). The patterns of Kriol are in many ways similar to those of Jaminjung possibly due to substrate influence from the languages of the area on Kriol during its development. For this reason, Schultze-Berndt chose not to include PAT-loans in her overview to avoid circularity and in this particular case the explanation for the predominance of MAT-loans is hence merely methodological. Biak, on the other hand seems to have a preference for MAT-loans. Biak word order is highly fixed, which means that in many cases MAT-loans are easier to integrate than a change in word order or re-modelling of existing material (cf. van den Heuvel, this volume). For example due to Indonesian pressure negation is expressed by a sentence-initial adverb. A native element in Biak could have been adjusted as a PAT-loan, but the pressure against the change in word order this would involve was higher than that of borrowing new material. Vietnamese has had substantial contact influence from Chinese. While most other situations involve bilingualism and language contact in the spoken language, Chinese influence on Vietnamese has predominantly been through written materials. The types of loans encountered are mostly MAT. This can be attributed to the type of contact situation, as without oral bilingualism pattern copying is difficult, and MAT-loans prevail. This rather trivial fact makes sense in the current discussion since it is directly related to the types of loans encountered. MAT-loans can appear even in cases of monolingualism, such as in the Muran language Pirahã (not discussed further in this volume), which has exclusively MAT-loans from Portuguese due to very rudimentary bilingualism. Concluding, there are both structural reasons for MAT-loans, as in Biak, where PAT is not preferred as it would lead to changes of a highly fixed word order, as well as reasons found in the type of contact situation linked to the degree of oral bilingualism, as in the case of Vietnamese.

20

Jeanette Sakel

5.2. Situations with overall PAT-borrowing Likewise, a number of languages in the sample show an overall majority of PAT-borrowing. These are K’abeena, Hup, Macedonian Turkish and Khuzistani Arabic, most of which are also part of well-established linguistic areas. K’abeena is discussed for its participation in the Gurage linguistic area (cf. Crass, this volume) and exhibits areal patterns of varying sources. Not surprisingly, most of the contact phenomena found in K’abeena are PATloans. K’abeena has a number of MAT-loans as well, part from lexical elements, it has borrowed a few markers of temporal and causal clauses, discourse markers and interjections. These belong to the category of function words, comprising conjunctions, discourse markers and other elements detached from the main proposition of the clause (cf. Matras 1998). Hup is part of the Vaupés are in the North-Western Amazon. Epps (this volume) focuses on the contact situation between two of the languages from this area, Hup and Tukano. Hup borrows a vast number of PAT-features from Tukano, while there are only few MAT-loans. The reasons for this are strong cultural restrictions in the Vaupés region against MAT-loans. These can nonetheless be overridden, as is the case in MAT-loans from Tukano, including an adverbial particle ‘until’, a negative emphasis marker and a disjunction marker. Again, these belong to the category of function words and the only MAT-loans outside this category in Hup are numerals.4 Macedonian Turkish, which is spoken in the Balkan linguistic area and displays a large number of areal features that can be established when comparing the language to its neighbourng Balkan languages and Standard Turkish. While most loans are PAT, some of the discourse markers, as well as the phrasal conjunctions ama ‘or’ and i ‘and’ are MAT-loans from Macedonian. Matras and Tufan (this volume) draw the conclusions for Macedonian Turkish that MAT structures are probably new additions since the change in hierarchical structures between the languages, while PAT-borrowing must have occurred earlier as a means of “utterance-organizing strategies”. A similar picture emerges in Khuzistani Arabic, which has a majority of PAT-loans, and again the only MAT-loans are found in categories such as discourse markers, fillers, tags and focus particles where they have an interaction-qualifying function (cf. Matras and Shabibi, this volume). The reason why Khuzistani Arabic has many PAT-loans could be its status in the contact situation, which is similar to that of Macedonian Turkish in that both languages are spoken across the border from areas where they are majority languages. The variants discussed here are dominated, but the speakers are

Types of loan: Matter and pattern

21

frequently exposed to the standard languages, for example through television and visits. Speakers may have a wish to keep the two languages apart as they attribute equally high status to them in the different contexts of use. This restricts MAT-loans and leads to mainly PAT-loans, while in the category of function words the two systems collide, very similar to what we found in Hup and K’abeena. The fact MAT-loans of function words are frequent in these languages appears to confirm Matras’s (1998) hypothesis that bilingual speakers will have greater difficulties keeping their linguistic systems apart around such markers than for other domains of functions.

5.3. Relations between languages and MAT/PAT borrowing Let us now look at the data from another perspective, namely whether the relations between the languages in contact have an impact on the types of loans encountered. In the sample, most borrowing appears from hierarchically higher – or dominant – languages into lower, dominated languages. Dominance is here broadly defined: a language is dominant when used for administration, as a lingua franca, and when it has to be learnt by the speakers of the dominated language, which in return is usually not used for any of the above or which is used in less official environments. A language can be dominant in one contact situation, while dominated in another. For example, Katanga Swahili is dominant in being the lingua franca of the region but dominated by French, which is the official and administrative language of the country. Apart from one-to-one borrowing situations, some of the languages in the sample belong to well-established linguistic areas. Table 1 summarizes the contact situations treated by authors of the chapters. Many of the languages in the sample are also part of linguistic areas, but where this is not immediately discussed in the chapters I left it out of the current discussion. X marks the main focus of the contribution to this volume, Y marks another contact situation which is not the main focus of the chapter. We can usually see layers of different types of contact in languages that belong to more than one contact situation. This is visible in the first group of languages in table one, i.e. languages that are dominant in some situations but dominated in others. In this way, Katanga Swahili has only few MAT-loans, most of which are of French numbers and discourse markers. The loans from the substrate languages are PAT, apart from three MAT noun-class markers. If we would expect substrate-influence only leading to

22

Jeanette Sakel

Table 1. Types of contact situation and hierarchical status in contact situation dominated 1 Indonesian, Katanga Swahili 2 K’abeena 3 Biak, Domari, Imbabura Quichua, Khuzistani Arabic, Kurmanji Kurdish, Manange, Mosetén, Northeastern Neo-Aramaic, Otomi, Paraguayan Guaraní, Purepecha, Rapanui, Saami, Tasawaq, Western Neo-Aramaic, Vietnamese, Yaqui, Yiddish 4 Hup, Jaminjung, Likpe, Nahuatl, Rumungro, Macedonian Turkish

areas (treated here)

X

dominant Y

X X

X

Y

PAT-borrowing, this would be exceptional. We can see reasons for why these markers have been borrowed in this way by looking at the system. De Rooij (this volume) shows that many of the Swahili noun class markers correspond in form and function to those of the substrate languages. The differences – in particular the three markers in question – were borrowed to assimilate Katanga Swahili’s system of noun classification to that of the substrate languages. We are therefore not dealing with a mere MAT-loan of three markers in isolation, but with a general adjustment of two systems that are already largely identical. In the same way we find MAT-loans from current substrate languages in Indonesian. Indonesian is the dominant language in contact with other languages spoken in Indonesia, such as Javanese but it has also been dominated by a range of other languages, in particular Sanskrit, Chinese and some European languages. The loans from the dominant languages are both MAT and PAT and have, for example, led to a number of changes in the sound system of Indonesian. The influence from substrate languages is astonishingly likewise MAT and PAT. MAT-loans from Javanese include collective particles, a third-person plural pronoun, some interrogative markers, a focus marker and modal particles. If Javanese was a substrate language at the time of borrowing, these MAT-loans would be exceptional in the light of general preconceptions that substrate-influence is mainly PAT. However, we can explain there MAT-loans from Javanese: being a substrate language

Types of loan: Matter and pattern

23

today, it used to have a different status in the past, described by Tadmor as a “quasi-symbiosis” between Indonesian and Javanese. The MAT-loans may have entered Indonesian during this time. The only language in the second group, K’abeena, has been discussed in the previous section for its prevalence of PAT-loans. When looking at the few attested MAT-loans, these are also from languages that take part in the area: (Ethio-) Semitic lexical elements are used as adverbial clause markers (time, reason, cause), an adjectival suffix is borrowed form Amharic and a number of interjections have the same forms as in other languages of the area. Apart from the adjective suffix these fall under the category of function words. Even though this is only one example, it shows that MAT-loans in the category of function words do appear in linguistic areas.5 The majority of the languages in the sample are in a contact situation with a dominant language. These are listed as group 3 in Table 1. In these languages both MAT and PAT loans are common in most categories, while there is heavy borrowing of function words, which are overwhelmingly borrowed as MAT. The contact phenomena encountered in these languages are in general very similar, independent of whether the languages are spoken by small minorities or whether they are major national languages. In this way, the major South American indigenous languages Paraguayan Guaraní and Imbabura Quichua (cf. Rendon, this volume) borrow heavily from Spanish and have comparable contact phenomena to the small indigenous languages Mosetén (cf. Sakel, this volume) and Purepecha (cf. Chamoreau, this volume). Stolz and Stolz (1996, 1997) come to similar conclusions comparing a range of languages in contact with Spanish. They argue that the similarities in contact phenomena are found within the same categories. Take the following example: most languages in contact with Spanish borrow an element out of the category of temporal adverbial clause markers, such as hasta ‘until’, desde ‘from’, cuando ‘when’, but which ones of these markers are borrowed, and which functions they assume in the recipient language, is language-specific (cf. Stolz and Stolz 1996, cf. also Sakel’s 2007 description for Mosetén). Similarities in contact phenomena have also been attested for languages in contact with Russian as the dominant language. Rießler (this volume) discusses how Russian contact phenomena in Kildin Saami resemble those in other contact situations with Russian as the dominant language, referring to Majtinskaja (19789). The reasons for this have to do with universals of borrowing, rather than individual languages. Comparing the contact phenomena in the sample, we find that similar principles apply for all contact situations of the same type, independent of the source language. Indeed, when compar-

24

Jeanette Sakel

ing the situations where Spanish is dominant to those where Russian or other languages are dominant, elements within the same categories are taken over, and often these are MAT-loans of function words. This also includes dominant languages that are not Indo-European, such as Arabic in contact with Western Neo-Aramaic. These situations are similar due to general principles of contact, rather than language-specific ones.6 The built-up of the system and availability of certain structures in Spanish and Russian may nonetheless play a role in a few cases, since they have similar typological profiles. The final group of languages in Table 1 comprises Hup, Jaminjung, Likpe, Nahuatl, Otomi, Rumungro and Macedonian Turkish. For these languages the authors of the individual chapters discuss different layers of contact in both one-to-one borrowing situations and linguistic areas. The general results are that the languages display PAT-loans from the areas they are in, as well as MAT/PAT loans from their particular one-to-one borrowing situations. In some cases area and borrowing situations overlap, as in the case of Hup in contact with Tukano, both of which are part of the Vaupés area. In other cases, there are clear differences between the contact influence from the area and the borrowing situation, as in Nahuatl, which is part of the Meso-American area (as proposed by Campbell et al. 1986), as well as being part of a one-to-one borrowing situation with Spanish. Comparing the contact phenomena in the latter case between a one-to-one borrowing situation and an area, these are very different: the areal phenomena are predominantly PAT, while Spanish loans are very similar to other situations in which indigenous languages of the Americas are in contact with Spanish as the dominant language and include many MAT-loans. Concluding, MAT-loans appear in dominated languages in one-to-one borrowing situations, but also PAT-loans are frequent in this cases. While linguistic areas and situations of substrate influence display mainly PAT-loans, MAT-loans are very frequent in the category of function words and indeed appear in all languages of the sample (cf. also Matras’s chapter on borrowability of categories, this volume). This suggests that function words are borrowed easily and relatively early on in contact situations. We can see this in layers of contact, for example, in the variant of North-eastern Neo-Aramaic (cf. Khan, this volume) whose speakers have all immigrated to Israel. The original contact situation with Kurdish as the dominant language has been replaced by the new situation with Hebrew as the new dominant language. Indeed, the loans from Hebrew are primarily found in the lexicon and function words. More evidence for this comes from Romani (Elšík, this volume, cf. also Elšík and Matras 2006). Also other sample languages show early borrowing of function

Types of loan: Matter and pattern

25

words: all of the sample languages show some degree of MAT-borrowing, even in those languages where all or most other loans are PAT, such as Hup and K’abeena.7 In many cases MAT-loans of function words have been on a long journey, being borrowed from one language into another. For example, some Kurdish discourse markers and connectors are of Arabic origin and have entered the language through contact with Turkish (Haig, this volume). Likewise, some Hup function words of Portuguese origin have most likely entered the language via Tukano (Epps, this volume).

6. Conclusion We have shown that the distinction between MAT and PAT is very useful for classifying contact phenomena. For contact theory in general this means that the distinction between MAT and PAT-borrowing should be included in attempts to categorize contact situations. We can conclude the following for the regularities behind MAT/PAT borrowing: We have found that hierarchical relations between languages have an impact on the types of loans encountered. For example Arabic is a dominant language in some contact situations and a dominated language in others: it is dominant in contact with Domari and Western Neo-Aramaic, while Khuzistani-Arabic is dominated by Persian. In the first case, the dominant languages incorporate many MAT-loans from Arabic, while in the second case, MATloans are from Persian.8 Hence, the direction of types of contact phenomena (in this case MAT-loans) depends on the hierarchy between the languages. Furthermore, when a dominant language has high status, MAT-loans into a dominated language are often easily accepted. For example many MAT-loans from Spanish appear in Purepecha, Nahuatl and Yaqui, for which Spanish is a highly dominant and high-status language. In the same way, MAT-loans can be shunned from contact situations, such as in areas with social constraints against pattern-replication in the Vaupés, as discussed for Hup above, and also in other areas such as parts of Papua New Guinea (Ross 1996). Also the degree of bilingualism plays a role in the way elements are borrowed. Without bilingualism, patterns are usually not copied and MAT is only borrowed in a restricted sense. The type of contact influence is likewise closely related to degree of (oral) bilingualism. For example Vietnamese has experienced massive contact influence from Chinese, though this was mainly through written materials and rarely through oral transmission involving bilingualism, leading to a majority of MAT-loans.

26

Jeanette Sakel

All languages of the sample display MAT-loans of function words, even those that have a majority of PAT-loans. This suggests that reasons for borrowing function words as MAT are stronger than other constraints in contact situations. Such reasons include the detached nature of function words, which makes them easily borrowable, as well as their function as discourse structuring devices (cf. Matras 1998). It could furthermore point at a general change of contact situations across the world towards situations with one dominant language that is used in administration, television and general communication in a highly mobile, globalized world. This does not mean that dominant languages did not exist in the past, cf. the influence of Chinese on Vietnamese or, for instance, Sanskrit influence on Indonesian. In those cases, however, contact was prevailing in some communities – among the elite, for example, while in other communities or at other times contact was less intensive. In today’s globalized world, however, some languages, often former colonizers’ languages, have impact on entire speech communities. Television, schooling, trade, mobility, indigenous organizations, health, developmental organizations and other ways of communicating with the “outside” world have lead to the rise of these already highly dominant, major languages through increased bilingualism. Many contact situations in the sample show this tendency. For example in Hup, area-internal restrictions such as taboos against MAT-borrowing have been overridden by overwhelming influence of Tukano and eventually Portuguese as administrative languages (cf. also discussion by Aikhenvald 2002).

Notes 1. I would like to thank Yaron Matras and Kristine Hildebrandt for comments on earlier versions of this chapter. 2. MAT-borrowing without any PAT will not be discussed in this chapter as it is very rare and mainly occurs in the lexicon; i.e. usually MAT is taken over with at least part of its original PAT. An example of MAT-only borrowing in the lexicon is the noun handy ‘mobile phone’ in German, which does not have this meaning in the source language English. 3. Yaron Matras made the useful comment that just like the shape of morphemes can combine with a certain meaning and appear in a certain organization pattern, so can a phone acquire meaningfulness as a phoneme, and combine with certain patterns of prosody, tone, or permissible combinations of sounds. Rather than define a perfect match for MAT and PAT in phonology, we can simply remain conscious of the layered structure of phonological representation and the

Types of loan: Matter and pattern

27

fact that contact-related change may affect one level without affecting another. 4. Numerals are very frequently borrowed as MAT in many contact situations. 5. One could speculate that the reasons for this could be temporal dominance of the source language. 6. These principle are anchored in the role of categories in processing discourse, and the way that operating in a bilingual setting influences language processing; see contribution by Matras. 7. We find similar loans in Indonesian, where some function words are taken from dominant languages in the history, e.g. Sanskrit, Arabic, Creole Portuguese and Persian. Other function words in Indonesian were borrowed as content words and then grammaticalized, such as ‘and’ from the Sanskrit word for ‘company’. 8. There are mainly PAT-loans in Khuzistani Arabic for reasons discussed above. Also, Arabic is still a dominant language in some religious contexts in Iran, which skews the picture somewhat.

References Aikhenvald, Alexandra Y. 2002 Language contact in Amazonia. Oxford: Oxford University Press. Bolonyai, Agnes 1998 In-between languages: Language shift/maintenance in childhood bilingualism. International Journal of Bilingualism 2 (1): 2143. Campbell, Lyle, Terrence Kaufman, and Thomas C. Smith-Stark 1986 Meso-America as a linguistic area. Language 62 (3): 530570. Elšík, Viktor, and Yaron Matras 2006 Markedness and Language Change: The Romani sample. Berlin: Mouton de Gruyter. Gołąb, Zbigniew 1956 The concept of isogrammatism. Buletin Polskiego Towarzystwa Jezykoznawczego 15: 112. 1959 Some Arumanian–Macedonian isogrammatisms and the social background of their development. Word 15 (3): 415435. Haugen, Einar 1950 The analysis of linguistic borrowing. Language 26 (2): 210231. Heath, Jeffrey 1984 Language contact and language change. Annual Review of Anthropology 13: 367384. Heine, Bernd, and Tania Kuteva 2005 Language Contact and Grammatical Change. Cambridge: Cambridge University Press.

28

Jeanette Sakel

Johanson, Lars 1992 Strukturelle Faktoren in türkischen Sprachkontakten. Stuttgart: Franz Steiner Verlag. Keesing, Roger M. 1991 Substrates, calquing and grammaticalization in Melanesian Pidgin. In: Elizabeth C. Traugott and Bernd Heine (eds.), Approaches to Grammaticalization Vol. 1: Focus on theoretical and methodological issues, 315342. Amsterdam: John Benjamins. Majtinskaja, K. E. 19789 Zaimstvovannye elementy, ispol’zuemye v finno-ugorskih jazykah pri obrazovanii form naklonenij [Borrowed elements, used in inflectional forms in Finno-Ugric languages]. Études Finno-Ougriennes 15: 227– 231. Matras, Yaron 1998 Utterance modifiers and universals of grammatical borrowing. Linguistics 36 (2): 281331. 2000 Fusion and the cognitive basis for bilingual discourse markers. International Journal of Bilingualism 4 (4): 505528. Matras, Yaron, and Jeanette Sakel 2007 Investigating the mechanisms of pattern replication in language convergence. Studies in Language 31 (4): 829865. Myers-Scotton, Carol 2002 Contact Linguistics: Bilingual Encounters and Grammatical Outcomes. Oxford: Oxford University Press. Nau, Nicole 1995 Möglichkeiten und Mechanismen kontaktbewegten Sprachwandels – unter besonderer Berücksichtigung des Finnischen. Munich: Lincom. Ross, Malcolm D. 1988 Proto Oceanic and the Austronesian Languages of Western Melanesia. Canberra: Pacific Linguistics. Sakel, Jeanette, Yaron Matras 2004 Database of Convergence and Borrowing. Manchester: University of Manchester. 2007 Language contact between Spanish and Mosetén: A study of grammatical integration. International Journal of Bilingualism 11 (1): 2553. Savić, Jelena M. 1995 Structural convergence and language change: Evidence from Serbian– English code-switching. Language in Society 24: 475492. Siegel, Jeff 1997 Mixing, leveling, and pidgin/creole development. In Arthur K. Spears and Donald Winford (eds.), The Structure and Status of Pidgins and Creoles, 111149. Amsterdam: John Benjamins.

Types of loan: Matter and pattern

29

Stolz, Christel, and Thomas Stolz 1996 Funktionswortentlehnung in Mesoamerika, Spanisch–Amerindischer Sprachkontakt. Sprachtypologie und Universalienforschung 49 (1): 86123. 1997 Universelle Hispanismen? Von Manila über Lima bis Mexiko und zurück: Muster bei der Entlehnung spanischer Funktionswörter in die indigenen Sprachen Amerikas und Austronesiens. Orbis 39 (1) [1996– 1997]: 177. Thomason, Sarah, and Terrence Kaufman 1988 Language Contact, Creolization and Genetic Linguistics. Berkeley: University of California Press. Treffers-Daller, Jeanine, and Raymond Mougeon 2005 The role of transfer in language variation and change: Evidence from contact varieties of French. Bilingualism: Language and Cognition 8: 9398. Weinreich, Uriel 1966 Languages in Contact: Findings and Problems. The Hague: Mouton (first publ. 1953).

The borrowability of structural categories Yaron Matras

1. Introduction The question of the “borrowability” of categories has often been equated with the presence or absence of constraints that rule out the borrowing of certain kinds of structures (cf. Campbell 1993, Thomason 2001 and elsewhere). I use the term here in a different sense. Borrowability is taken to mean the likelihood of a structural category to be affected by contact-induced change of some kind or other (whether matter- or pattern-replication; see Matras and Sakel 2007). From a strictly structure-oriented point of view, one might interpret this as the “ease” with which a category can be re-shaped through contact. I am not quite happy with this formulation, either, since it leaves open the source of the process and its motivation. Nor is the issue resolved by re-stating the obvious, namely by claiming that there is a link between the sociolinguistic norms of a speech community, the intensity of cultural contacts, and the outcomes of structural processes of change (cf. Thomason 2001, following Thomason and Kaufman 1988); for such a statement does not account for the fact that the borrowing of some categories requires more intensive contact than that of others. In other words, it fails to explain the hierarchical relationship between individual positions on the borrowing continuum. When we speak of “ease” of borrowing, we are referring implicitly at least to the communicative behaviour of speakers in a bilingual setting and to changes in that behaviour that have a long-lasting effect on the shape of the language that they use. What interests us in this connection is the likelihood that, in respect of a particular structure which serves a particular function in language processing, speakers might give up the separation of two sub-components within their linguistic repertoire – the two “languages” – and begin to employ the structure in question regardless of the choice of language. Bilingual speakers of English and German, for example, take for granted that the concepts computer, download, and internet are common to both sets of communicative interactions in which they normally engage: those where the chosen language of conversation is English, and those where it is German. Bilingual speakers of Domari and Arabic (see Matras,

32

Yaron Matras

this volume) are fully at ease with the fact that the entire system of clause combining and connectors is shared by their two languages; a shift in the interaction setting will lead them to switch into another “language”,1 and this will affect the selection of various structures – vocabulary, inflections, anaphora and deixis, and so on – but it will not affect strategies of clausecombining, which remain the same in all settings (i.e. for both languages). And for speakers of Macedonian and the local Turkish dialects spoken in that country (see Matras and Tufan, this volume) the way of organizing information in copula sentences is identical regardless of the language that is being spoken. This, in essence, is the core of the diachronic process that we call “borrowing”. With “borrowability”, then, we mean the likelihood that speakers will give up the separation between their “languages” – the mental demarcation line that divides their overall repertoire of linguistic structures – in respect of a particular function-bearing structure (a “category”).2

2. Borrowing hierarchies Essentially two kinds of generalizations have been proposed concerning the borrowing of grammatical categories. Those of the first kind relate to the frequency with which a category may be affected by contact-induced change. Generalizations of the second kind suggest 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. The majority of observations on grammatical borrowing belong to the first group (cf. Haugen 1950, Heath 1984, Thomason and Kaufman 1988, van Hout and Muysken 1994, Stolz and Stolz 1996 and 1997, Winford 2003, Aikhenvald 2006). Some statements are based on casual impressions only, while others report the results of counting exercises performed on a corpus. An issue that merits attention is the distinction between the counting of tokens and the counting of types (cf. also van Hout and Muysken 1994: 4243): can we consider nouns to be more borrowable than, for instance, conditional particles, simply because nominal tokens occur in a corpus more frequently than conditional particles? Surely, token frequency will tell us how often a borrowed form is used, but it will not necessarily reveal how likely it is to be borrowed? Counting types, in turn, raises problems of its own: Can we conclude on the basis of type-frequency that adjectives, for instance, are easier to borrow than conditional particles, considering that the first constitute an open class of nearly unlimited types (for the purposes of any practical

The borrowability of structural categories

33

comparison), while a language is likely to have just a very restricted inventory of conditional particles (if at all more than just one)? Such issues make it difficult to compare frequency-based hierarchies drawn from conversational corpora.3 This kind of dilemma does not present itself when comparing the grammatical (and lexical) systems of different languages in a sample (whether a structured one, or a casual one). Here, we are interested in the number of languages within the sample in which category X has been re-shaped as a result of contact. The more languages show borrowing affecting a certain category, the higher the frequency of borrowing for that particular category in the sample. We might then say that this category is “more likely” to be borrowed, relative to other categories. Implicational hierarchies entail frequency hierarchies,4 but go a significant step further in suggesting a constraint on the occurrence of borrowing with any lower-ranking category. The usual format of the statement is: Y is not borrowed unless X is borrowed as well (cf. Moravcsik 1978, Stolz 1996, Matras 1998 and 2002, Field 2002, Elšík and Matras 2006). The postulation of implicational borrowing hierarchies thus goes beyond the assumption that categories have different susceptibility to contact-induced change. It suggests that the process of contact-induced change follows, to some extent at least, a predictable pathway, with one stage leading as a pre-requisite to another. Moravcsik (1978) had attempted in this way to link the borrowing of non-nouns to nouns (“no non-nouns are borrowed unless nouns are also borrowed”), inflectional to derivational morphology (“no inflectional morphology is borrowed unless derivational morphology is also borrowed”), and so on, resulting in a web of inter-dependencies among various structural types (cf. Field 2002 for a statement on agglutinative > inflectional morphology). Matras (1998, 2002) and Elšík and Matras (2006) concentrate their observations on inter-dependencies of what they consider “values” of the same category, i.e. members of a shared structural paradigm. Matras (1998) postulates a borrowing hierarchy ‘but’ > ‘or’ > ‘and’ with respect to coordinating conjunctions, and Matras (2002) and Elšík and Matras (2006) suggest ‘necessity’ > ‘ability’ > ‘volition’ with respect to expressions of modality, and many more.5 Frequency-based hierarchies and implicational hierarchies may complement one another. Stolz and Stolz (1996) and Ross (2001), both relying on frequency-oriented observations rather than strict implicational hierarchies, conclude that contact-induced change begins at the level of the organization of discourse, proceeds to the organization of the paragraph, utterance, and

34

Yaron Matras

sentence levels, and only then reaches the levels of the phrase and word. The postulation of this kind of implicational hierarchy rests in such cases on accumulated observations of a series of frequency hierarchies. Borrowing hierarchies thus provide us with an opportunity to gain insights into the factors that prompt speakers to allow their language systems to converge around a particular structure. Explanations of borrowing generally take one of three directions: (1) The degree of borrowing is related to the intensity of exposure to the contact language, (2) The outcome of language contact is a product of the structural similarities and differences (congruence) among the languages concerned, and (3) Borrowability is a product of inherent semantic-pragmatic or structural properties of the affected categories. Issues such as prestige and domain-specialization of the languages typically fall under 1, while conjectures about functional “gaps” as motivating factors fall under type 2. Our interest in the present context is in explanations of type 3. This interest derives from the realization that structures and paradigm values often behave in an asymmetric manner when it comes to contact-related change. Under the “prestige” or “intensity of contact” effect there is no a priori reason why ‘but’ should be more vulnerable and prone to borrowing than ‘and’; in many cases structural congruence does not provide an answer to this hierarchical relationship, either (cf. Matras 1998). Where category or paradigm values consistently show unequal or asymmetrical behaviour in contact situations, factors promoting borrowing must be sought in the inherent properties that they possess. In trying to explain the borrowability of categories we must therefore return to our initial assumption that ease of borrowing reflects the ease with which speakers are willing to give up the separation of two “language systems” and allow them to converge or to fuse around a particular linguistic function. The question that we ask is therefore: What is it that makes one category (or category value) a more attractive candidate for “system conflation” than another? Elšík and Matras (2006: 370ff.), following Matras (1998), argue that borrowing is motivated by cognitive pressure on the speaker to reduce the mental processing load by allowing the structural manifestation of certain mental processing operations in the two languages to merge. The need to do so arises especially around operations that gauge the presentation of propositional content to hearer expectations, for example connectivity and modality. In these domains, merger of the structures targets in the first instance those conceptual domains where the speaker’s epistemic authority is in question, and the potential for tension at the interaction level is therefore greatest. This

The borrowability of structural categories

35

occurs for example around the expression of condition, contrast, participantexternal force, or other, more general conceptual complexity. This accounts for the direction of the hierarchy, which prioritizes category values such as “contrast” and “external force”. We should at this stage clarify the notion of “category”. One of the factors impeding straightforward comparability among hierarchies postulated in the literature is the vagueness with which category labels are used. Some researchers have, for example, identified a category of “function words”; others speak of “particles”, and others still make reference to a class of “pronouns”. As “function words” we can classify anything from interjections and fillers on to definite articles and demonstratives – categories that show enormous variation in respect of their contact behaviour. “Particles” can include markers of modality, connectivity, aspect, and more, while “pronouns” are used with reference to such functionally diverse entities as anaphoric or thirdperson pronouns, indefinites, and participant deixis. If we suspect that there is a link between category status and borrowing, then we must assume that “category” represents a functional notion, rather than just a constituent slot or a wholesale cover-term. Categories are understood here to be operational devices that trigger mental processing activities in communicative interaction: nouns name objects, interjections direct attention to emotive evaluations of the speech situation, connectors establish links between the processing of individual propositions, word order serves as a map to organize information at the utterance level, and so on. Their concrete representation in a given language is through a structural form, which may or may not have cross-linguistic equivalents. Our agenda is to accommodate the borrowing behaviour of categories within an explanatory model, one that accounts for the link between the processing function which the category triggers, and the degree to which speakers allow its structural representation to converge or fuse among the two (or more) components of their linguistic repertoire. To be sure, different explanatory models may be appropriate for different structural components of language. There is no doubt that the borrowing of institutional terminology from a language that is dominant in the public or acrolectal domain is not a result of mere tension at the level of processing the speech interaction, but rather an attempt to extend the referential world of the “dominant” language into interactions in which the “minority” or “weaker” language is used. The borrowing of phonemes may, in turn, be simply instrumental in serving the authentic integration of loanwords without “distorting” them, by adjusting the phoneme system to accommodate them. Nevertheless,

36

Yaron Matras

these too are conversation-functional factors that motivate speakers to allow their “linguistic systems” to conflate around certain structures and categories. Finally, we must briefly comment on the relevance of exceptions to postulated hierarchies. In a discussion that focuses on absolute constraints on borrowing and sets out to test their validity, the discovery of counterexamples can have a sensational effect in dismantling earlier claims. We aim here at taking what Aikhenvald (2006: 26) describes as the more “positive” route: understanding a variety of factors and preferences that facilitate structural diffusion among languages. In this spirit, it would be naïve and counterproductive to ignore tendencies that are followed by a substantial group of languages within the sample only because they are not followed by all, or indeed because they might be contradicted in one or two instances. Where there appears to be a motivation behind trends, one that is beyond pure coincidence, then these trends deserve our attention. Quite often, it is the counterexample that can be explained as resulting from a local, language-particular constraint that impedes the realization of common patterns in a particular instance. The following sections mirror the organization of the language-oriented chapters in this collection, and are devoted to an evaluative overview of selected patterns arising from the discussion of the 27 sample languages.

3. Phonology Phonology in particular is an area in which borrowings are traditionally considered to fill so-called “structural gaps”, facilitated especially when borrowing does not entail changes to the actual phonemic system but merely to allophonic distribution (cf. Winford 2003: 5556). The notion of “gap” is vague, given that languages have long been considered in descriptive linguistics to constitute autonomous, functional systems. We should therefore perhaps amend the definition to focus on bilingual speakers’ quest for harmony among the two (or more) systems that constitute their linguistic repertoire; absence of harmony as a result of absence of a phoneme in one of those systems is presumably what is meant by a “gap”. There is a functional motivation favouring consistency in the types and points of articulation as well as the distribution rules of allophonic variation, regardless of the speech situation in which language users find themselves, and hence pressure toward convergence of the two phonological “systems”. At the same time, social norms and

The borrowability of structural categories

37

awareness of identity and loyalty toward the group associated with the home language will counteract levelling. The process of phonological borrowing is the outcome of compromises between these two pressures. Our sample shows three different types of change: (1) Incorporation of phonemes from a contact language in loanwords, (2) adjustment in the articulation of a phoneme following the model of the contact language, and (3) incorporation of a borrowed phoneme into the system of inherited words (substituting an inherited phoneme in some words, though not necessarily in the system as a whole). Changes of the second type may lead to simplification of the system, or to its enrichment through new distinctions, or they may simply alter the nature of certain phonemes, leaving the complexity of the system as a whole intact. In the sample, changes of the first type typically add to the phoneme inventory, as do in most cases changes of the third type. On the whole, then, our observation is that language contact in the cases under scrutiny here typically leads to an enrichment of the phonological system. Another general observation is that contact-related change is more likely to affect consonants than vowels; indeed, we may even be able to postulate an implicational hierarchy of contact-related change: (1)

adoption of new consonants > adoption of new vowels

The reason behind this hierarchy is, however, likely to be rather trivial: It is a product of the fact that consonant inventories are generally larger, and so the potential for lack of overlap between consonant systems in contact is higher, resulting in greater pressure to adjust the consonant system. In fact, the hierarchy under (1) need not at all suggest that contact induced change in phonology begins with consonants, and it is not impossible that vowels are as prone, or even more prone to change in situations where there is no significant difference among the languages in the inventory of consonants. Almost all languages in the sample incorporate loanwords along with at least some of their original phonemes, which are new to the recipient system. Examples are Macedonian Turkish /ts/ with Macedonian loans, the Vietnamese sounds /ʆ/, /f/. /v/, and /z/ with Chinese loans, the Domari pharyngeals /ħ/ and /ʕ/ used in Arabic loans, and Imbabura Quichua /b/, /d/, /g/, /ʋ/, /ʒ/, and vowels /e/ and /o/ in Spanish loans. This indicates that for speakers of the languages in question, the integration of lexical loans in an “authentic” manner, i.e. one that closely replicates their original use in the contact language, takes precedence over the preservation of the coherent phoneme structures of the recipient language. The system of the recipient language is adjusted in order

38

Yaron Matras

to accommodate loans in an unmodified form. Of the sample languages, only Mosetén and Biak appear not to extend their phoneme system to accommodate loanwords, while the overlap between the phoneme systems of Kriol and Jaminjung does not require Jaminjung speakers to make any special effort in order to accommodate Kriol loans. The second type of process is the convergence of articulation modes and positions, which is often, as Winford (2006) suggests, a process affecting allophonic variation. There are numerous cases involving the introduction of allophonic variation, among them the interchange of /dʒ/ to /ʒ/ in Domari based on the Arabic model, of /q/ and /ɣ/ in Khuzistani Arabic based on the Persian model, and of /l/ and /r/ in certain positions in North-eastern Neo-Aramaic, based on the Kurdish model. In some cases, variation leads to a shift in articulation, as in the weakening of /h/ in Macedonian Turkish or the replacement of /h/ by /ɦ/ in some varieties of Yiddish, of pretone /o/ by /a/ in others, and a more general shift in this language from dental /l/ to velarized /ɫ/. From a language user’s perspective, all these are instances of harmonization of articulatory patterns, aiming to ease the burden of having to maintain complete separation of two distinct systems in different settings of conversational interaction. In some cases, phonemic distinctions in the recipient language are even given up in order to enable harmonization. Some varieties of Purepecha for instance lose the opposition between retroflex /ɽ/ and flap /r/, as well as between central /ï/ and front /i/, resulting in a vowel system that matches that of Spanish. In other cases the system becomes more complex, as with the introduction of pharyngealized consonants into Domari from Arabic and into Tasawaq from Tuareg, of long vowels into Rumungro from Hungarian, and of palatalization of stops into Kildin Saami and Yiddish from Slavic contact languages. Informal observations lead us to believe that prosody is a domain of phonology that is particularly prone to contact. This can be the result of two interconnected factors. The first is the peripheral role that prosody has in conveying meaning, and the fact that it is a form of expression of emotive modes, operating at the speech act and utterance level, rather than the word level. This allows speakers to mentally disconnect prosody more easily from the matter or shape of words associated with a particular language, making it prone to change and modification in contact situations. The second factor may be the proven neurophysiological separation between prosody and other aspects of speech production, making prosody more difficult to control. Both factors may contribute to the fact that foreign “accents” are most persistent in the area of prosody.

The borrowability of structural categories

39

We have in our sample little data on prosody, however. Several pairs of languages are reported to share prosodic features with their current contact languages, confirming the above hypothesis: They include Domari, Nahuatl, Rapanui, Rumungro, and Indonesian, and to some extent at least also Kurdish, some varieties of Yiddish, Hup, and Kildin Saami. This is already more than can be said about any other area of phonology, where we find borrowings and convergent tendencies, but no wholesale convergence.6 Hence, we might carefully postulate the following frequency-based hierarchy of likelihood of complete convergence in the phonological system: (2)

prosodic features > segmental phonological features

Once again we need to emphasize that this hierarchy does not suggest that segmental phonological features are unlikely to be borrowed unless prosodic features are also borrowed; it merely reflects the tendencies toward full-scale convergence of the systems. In fact, it does not seem possible at this stage to point to any position within the phonological system (e.g. certain articulatory modes or positions, marked features, etc.) as being particularly prone to contact-induced change. It seems that the details of phonological change are entirely a product of the relations among the two systems – or congruence – and any statistics of change are likely to simply reflect the mere likelihood of the two phoneme systems in contact to share certain phonemes, and to differ with respect of others. The one additional generalization that we can make is that the borrowing of phonemes begins with the borrowing of lexical items that contain them: (3)

phonological features in loanwords > independent phonological features

Concluding this section, it seems that there are two alternative strategies that multilinguals can pursue in respect of phonology, taking for granted that language contact will lead at least to a transfer of lexical items from one language to another. The first is to maintain the complete integrity of the recipient language system by adjusting the phonology of any borrowed word to match that of the recipient system. It would appear that this strategy would be facilitated by widespread monolingualism in the recipient language, and the confinement of bilingualism to just a small or peripheral group of intermediaries. It is also possible that this strategy can be maintained for a while in situations where widespread bilingualism is a relatively new phenomenon,

40

Yaron Matras

or where speakers of the recipient language have no need to appear to have native-like or even good command of the contact (donor) language. In our sample, only Mosetén and Biak appear to adopt this kind of strategy. The alternative, which is the route taken by most of our sample languages, is to maintain the authenticity of donor language items by adjusting the phonological system of the recipient language to accommodate phonological features of the donor language. This would seem to be facilitated by widespread bilingualism and the need for speakers of the recipient language to gain the approval of the donor language community. Authenticity in the pronunciation of loanwords is a token of the social value attributed to the donor language, and is emblematic of the social immersion into the donor language culture. The result is the incorporation of phonological features from the donor language into the recipient language. These will at first accompany loanwords from the donor language. With increased multilingualism and the need to operate regularly in two linguistic environments, it is advantageous for speakers of the recipient language – we assume that this is normally the language that occupies the weaker socio-political position, or a minority language – to allow major components of their phonological system to converge with that of the dominant, donor language, and so to rid themselves of the burden to maintain a separation of their two speech modes. This tendency of course competes with the sense of loyalty toward the group-language, which may favour maintaining an old system and limit innovation. The outcome is often a compromise in the form of an adjustment of certain aspects of the phonological system in favour of common patterns, or the adoption of some features but not others from the donor language, as we saw above.

4. Typology A number of languages show signs of movement between morphological types: There are changes from polysynthetic to less polysynthetic structuring in Nahuatl and Imbabura Quichua, Otomi, and Guaraní, from an agglutinative type to a more isolating type in Indonesian, and from an agglutinative to a more analytic type in Purepecha. An increase in reliance on reduplication is found in (agglutinative) Likpe under the influence of (isolating) Ewe. On the other hand there is some acquisition of agglutination in Hup, and in some traits perhaps also in Kurmanji Kurdish as well as in Rumungro: In Kurmanji, the agglutination of case markers benefits from the presence of

The borrowability of structural categories

41

inherited enclitic case markers, which historically form part of a circumposition.7 In Rumungro, the adoption of agglutinative prefixes is a by-product of the almost wholesale adoption of indefinite markers, superlative markers, and a few other morpheme classes that are high on the “relevance” scale and so easily borrowable on functional grounds. It is noteworthy that none of these developments seems to follow any predictable structural path, and the only common denominator is an accommodation to the patterns of a socially dominant contact language. In all cases, the drift begins in individual constructions such as adjective comparison or case marking, and it is yet to be seen whether it will continue to spread. Frequencies and the evaluation of general trends are not applicable to our sample in the domain of typology, as some language pairs happen to belong to similar types (consider Domari and Arabic, Vietnamese and Chinese, Yiddish and Slavic languages), while others, among them those named above, show typological clashes with their contact languages. It is interesting that Macedonian Turkish maintains strong morphological agglutination despite considerable re-structuring in the domain of clause organization; indeed, the loss of converbal morphology in favour of grammaticalized conjunctions might be considered a small, yet not insignificant step in the direction of a typological drift. On this basis, we might conclude that typological drift begins at the clause level. In Khuzistani Arabic, the reinterpretation of the construct state ending and definite article as equivalents of the Persian ezāfe attribution marker does not constitute a drift in morphological type as such, but it does expose the path taken toward morphological re-analysis, here too in a possessive construction at the phrase level. Perhaps one of the more outstanding typological shifts reported in the sample is the change in alignment in North-eastern Neo-Aramaic, modelled on the Kurdish ergative construction. An interesting aspect of the construction is that it does not mirror the complete ergative formation – in the absence of, for example, nominal case in Aramaic – but chooses instead only a number of pivotal features which it reconstructs with inherited means (see also Matras and Sakel 2007). Alignment is contact-sensitive elsewhere, too: We find an expansion of ergativity in Urban Manange in contact with Nepali, and the incipient loss of ergativity in Kurmanji in contact with Turkish. In conclusion, although we do not have cases in our sample that display far-reaching changes in overall morphological typology, it is very evident that morphological type is certainly not immune from contact-induced change. It would seem fair to state, at least cautiously, that there is by and large an opportunist motivation for typological drift: It is subordinate to pressure to-

42

Yaron Matras

ward convergence in a particular salient construction, such as the possessive construction, clause linking, or indefiniteness marking, or else it is triggered by accidental similarities in the shape and position of markers with similar meaning.

5. Nominal structures This is a diverse and complex domain, containing many different sub-categories, and it is perhaps not a surprise that only two languages in the sample, Jamingjung and Biak, are reported to show no contact influence at all under this heading. A prominent sub-domain of nominal structure is case, but here it is noteworthy that no borrowing of bound case markers is attested in the sample. The closest evidence of contact influence on case markers is the reliance in some contact varieties of Kurmanji on postposed markers as enclitics rather than as components of circumpositions. There are also indications of meaning extension of case markers, such as the ablative becoming genitive under Dutch influence in Indonesian, or the loss of a distinction between comitative and instrumental in Imbabura Quichua. Adpositions on the other hand do show matter-replication; attested cases include Indonesian sama ‘with’ and guna ‘for the purpose of’ from Sanskrit, Spanish de in Guaraní, numerous Arabic prepositions in Domari, Spanish por and para in Purepecha, and Tuareg ámmàs ‘inside’ and àláqqàm ‘behind’ in Tasawaq. The preposition indicating ‘between’ is the most frequently borrowed, examples being Indonesian antara from Sanskrit, Spanish entre in Guaraní and other languages, and Arabic bēn in Domari. This gives some vague evidence in support of the hierarchy proposed by Elšík and Matras (2006) for the borrowing of local relations expressions in Romani dialects: (4)

peripheral local relations > core local relations

“Core” relations (‘in’, ‘at’, ‘on’) are borrowed less frequently than “peripheral” relations (‘between’, ‘around’, ‘opposite’), and this finds some support in the appearance of ‘between’ as the most frequent borrowing in the sample. Developments affecting gender marking include a shift in unmarked gender from feminine to masculine in Mosetén, the loss of neuter gender in NE Yiddish (as in the contact languages Lithuanian and Latvian), and the incipient system of nominal classifiers in Hup (classifying inanimates by shape, and animates by gender), adopted from Tukano. Definitely the most extensive

The borrowability of structural categories

43

development in this domain is the borrowing of Chinese classifiers into Vietnamese. There are, in addition, some marginal phenomena such as the loss of gender in pronouns (in Rumungro as well as in North-eastern Neo-Aramaic). Our sample gives us the impression that gender in the narrow sense (of a two- or three-gender system) is more stable in contact situations than more differentiated classifier systems, where influence might be more extensive. Nominal possession is a domain in which contact phenomena are fairly widespread. The most common change to possessive constructions is a modification of word order, often drawing on existing flexibility and enhancing the frequency of a more peripheral pattern to match that of the contact language. Examples of contact phenomena in possession are found in Domari, Macedonian Turkish, Rumungro, Khuzistani Arabic, Guaraní, and North-eastern Neo-Aramaic, concerning mainly the order, and in some cases the distribution of morphs and their meaning; incipient cliticization of the postposed pronominal possessor in Kurmanji; frequent postpositive possessors in Rapanui; use of a preposition bearing the same meaning as in the contact language in Indonesian; and use of a borrowed preposition in Guaraní. Definiteness is known for its areal diffusion, but in the sample we have only few cases of contact developments in this domain. Rapanui illustrates that re-arrangement of definiteness rules may occur when definite markers exist in both systems prior to contact. Khuzistani Arabic shows selective retreat of definiteness marking in some constructions, in contact with a language with no overt marking of definiteness. Interestingly, North-eastern Neo-Aramaic borrows the Kurdish definite article -ak-. Nominal morphology is most frequently replicated in the case of plural markers, which are often maintained in loan nouns, either as productive markers of plurality, as in Yiddish, Rumungro, and Tasawaq, or in conjunction with a native expression of plurality, as in Domari and Quichua. In Likpe we find replication of a specific plurality marking pattern. Plural markers may be said to occupy a position in between derivation and inflection markers. On the one hand, they are potentially linked to the expression of plural agreement elsewhere in the sentence, and so they operate at the level of the sentence rather than just the word. On the other hand, they indicate clear semantic opposition to singulars at the word level. Morphological plural marking thus meets the criteria for semantic transparency which is so often noted as a factor facilitating morphological borrowing (Moravcsik 1978, Matras 1998, Field 2002, Winford 2003). The direct borrowing of derivational morphemes is attested throughout the sample. Macedonian Turkish, Yiddish, Quichua, Purepecha, and Rumungro all borrow diminutive suffixes (with Kildin

44

Yaron Matras

Saami re-organizing its diminutive derivation based on a Russian model), and Quichua and Yiddish borrow agentive suffixes (Quichua -dur from Spanish, Yiddish -nik from Russian). The overall pattern leaves us with a picture that is not incongruent with that reported on in contact linguistics so far. The most widespread changes are in the possessive construction. They affect the nominal phrase at the syntactic or morphosyntactic level, having to do primarily with the position of the possessor and possessed object, and partly with the arrangement of possessive morphology. This is in line with predictions that phrase-level borrowing will be more intense than word-level borrowing. Borrowing of bound markers favours in particular plural markers, diminutive and agentive derivational markers, and classifiers (but not gender markers), confirming that semantic transparency facilitates borrowing. Adpositions are more borrowable than bound case markers (borrowing of which is not attested in the sample), with between being the most borrowable in our sample, confirming the tendency of borrowing to favour peripheral relations, and so for the process of convergence to begin with remote, cognitively less accessible or conceptually more complex domains. In other domains, such as the distribution of case, definiteness, or gender assignment, languages may develop similarities, often by extending or limiting distributional rules. However, bound case and gender markers remain on the whole among the most stable features in the nominal domain, resisting especially direct replication of matter.

6. Verbal structures Little attention has been granted in the literature to borrowing of features belonging to the domain of verbs (on the borrowing of lexical verbs see below); reports on the borrowing of TMA markers are quite rare. It is useful to consider the categories one by one. In the domain of tense, we see contact-induced similarities in the organization of the future tense in several languages: It is lacking in Domari and Arabic, it is suffixed in Hup and Tukanoan, and it shows a similar periphrastic structure in Kildin Saami and Russian. To this we might add the similar organization of the prospective aspect in K’abeena and Amharic. Contact phenomena appear to be somewhat more frequent in aspect and aktionsart, where we find matter replication as well as shared patterns. Domari uses an Arabic habitual auxiliary kān, and Nahuatl introduces a progressive based on the Spanish model. We find aspectual use of the borrowed comple-

The borrowability of structural categories

45

tive ya from Spanish in Guaraní, and sudah from Sanskrit in Indonesian, as well as similar expressions of experiential perfect in K’abeena and Amharic, while Likpe adopts a periphrastic present progressive similar to Ewe, and North-eastern Neo-Aramaic adopts a present progressive particle-turnedprefix based on a Kurdish model.8 Yiddish re-organizes its verbal prefixes to replicate so-called Slavic Aspect distinctions (essentially, grammaticalized aktionsart), and Rumungro shows iteratives that are calqued on Hungarian. Mood is similarly somewhat richer in contact developments, though differences in the structural organization of mood and modality make an exhaustive comparison somewhat difficult. Noteworthy are similarities in the use of the subjunctive in North-eastern Neo-Aramaic and Kurdish, and Domari and Arabic (cf. Matras and Sakel 2007: 843). Vietnamese prohibitive and conditional markers are borrowed from Chinese, while Kurmanji borrows the Turkish conditional marker -ise. Modality shows the most widespread contact phenomena, especially as regards matter replication. Almost half of the sample languages show matter replication of modality markers, such as Turkish gerek ‘must’ in Kurmanji, Arabic lāzim ‘must’ in Domari, Spanish tiene que ‘must’ in Rapanui, Arabic mungkin ‘can’ in Indonesian, Spanish pudi ‘can’ in Quichua, Malay harus ‘must’ in Biak, and more. The most common are markers of obligation (i.e. expressing external forces), followed in turn by necessity, possibility, ability, and desire. This hierarchy is almost always implicational, the only exception being Domari (which borrows markers for all meanings except ability):9 (5)

obligation > necessity > possibility > ability > desire

The hierarchy proceeds from the most intensive external force, to the most participant-internal dimension. It is identical to the hierarchy identified by Elšík and Matras (2006) for the borrowing of modality markers in Romani dialects: necessity > ability > (inability) > volition. The more abstract theme in this hierarchy might be described as the degree of “speaker control”, low speaker control correlating with high borrowability. As far as Domari is concerned, its minor deviation from the implicational hierarchy can be explained by the fact that all the modality markers that it borrows from Arabic are impersonal expressions, or non-verbs. Even bidd‘want’, is a nominal, and its person inflection in Arabic follows the paradigm reserved for nominals, i.e “my-wish” etc. (and this inflection is carried over into Domari as well). The Arabic expression for ability, however, ‘a-qdar-, is an inflected verb, and although Arabic verbs are borrowable in Domari,

46

Yaron Matras

it competes with an inherited Domari verb sak-, which prevails. Thus, it is the formal inconsistency in the system of the donor language which in this case imposes a constraint that breaks the hierarchy. Returning now to a general view of TMA and modality, we have seen the high density of (matter) borrowing in the domain of modality, in some cases also in mood, frequent matter and pattern replications in the area of aspect and aktionsart, and few cases of pattern replication in tense, all involving the future. This picture lends itself to an interpretation in terms of the hierarchy in (6), which depicts the likelihood of the respective categories to be affected by contact: (6)

modality > aspect/aktiosnart > future tense > (other tenses)

By and large, this hierarchy reflects both frequency, and implicational relationships. There is one case in the sample – Kildin Saami – where there appears to be contact influence in the arrangement of tense, but not in modality. Yiddish might be considered a case for contact influence on aktionsart, but it similarly lacks borrowed modal verbs from its Slavic contact languages, though to some extent this might be explained by the presence in this domain of Hebrew loans acquired through literary tradition. The internal rationale of the hierarchy – which, once again resembles the findings for Romani dialects (cf. Matras 2002) – leads us to postulate again that external circumstances that limit the degree of speaker control – mood and modality in general – are the most contact-sensitive. They are followed by a qualification of the internal structure of the event – aspect and aktionsart – these too being beyond the immediate control of the speaker. Only then do we find contact influence in tense, the most intimate relationship between the event and the speaker’s own perspective, though it is noteworthy that in our sample this is limited to the future tense, which identifies the event as being least stable and secure from the speaker’s perspective. The overall theme is therefore once again the speaker’s epistemic authority; its absence or weakening correlates with high borrowability. Existential and possession verbs are affected by contact in several of the sample languages. In Domari, the Arabic copula is adopted in its function as a periphrastic expression of the habitual aspect, and it co-exists with the Domari enclitic copula. But Rapanui uses Tahitan and Spanish forms as copula, and both Indonesian and North-eastern Neo-Aramaic are reported to have developed copula forms through imitation of forms in their contact languages. The borrowing of ‘have’ is on the whole marginal. Spanish tengo is

The borrowability of structural categories

47

employed in Rapanui, while in some other languages, such as Hup, we find calquing of constructions. Contact phenomena in the area of voice and valency are almost exclusively pattern-oriented, and usually involve an increase in frequency distribution of an existing option: analytic reflexives in Quichua, the periphrastic passive in Purepecha, a causative marker derived from the verb ‘to do’ in Manange (copying the function of the Nepali affix -aau), and productive use of the inherited morphological causative in Rumungro. Vietnamese is exceptional in directly borrowing reflexive and passive markers from Chinese. Recent contact-induced grammaticalizations lead to the emergence of causative, passive, and reflexive markers in Hup, and to a reflexive in Likpe. It is difficult to further sub-divide this domain into category components, or to draw clearcut connections with other components of the verb, other than to say that derivation generally appears much more contact-susceptible than tense, and perhaps at the same level as aspect. Both are in a sense statements about the internal organization of events, not directly connected to the speaker’s position, but with no explicit evaluative statement concerning the truth-relevance or factuality of the event (as in mood or modality), either. Moravcsik (1975) had drawn attention to the frequent use of incorporation strategies to accommodate borrowed verbs, and this typological discussion of verb borrowability has recently been revived (cf. Wichmann and Wohlgemuth, forthc.). There appears to be a near-consensus view that the borrowing of verbs is not, of course, impossible, but made more cumbersome in some languages due to the widespread tendency of verbs to be morphologically more complex (see Winford 2003: 52). In our sample, direct borrowing of verbs, without any formal adaptation, is found in Vietnamese, where there is no morphology in either the recipient or the donor language Chinese; in Likpe, where isolating Ewe contributes verb roots into an agglutinative structure; but also in many other languages, including Tasawaq, Quichua, Otomi, Guaraní, Hup, and K’abeena. There is thus obviously no universal constraint on the integration of borrowed verbs.10 Nonetheless, several languages in the sample prefer to apply an overt, morphological accommodation strategy when incorporating verbs of foreign origin into the lexicon. A favourite strategy is the use of so-called “light” verbs. Macedonian Turkish, Kurmanji, Domari, and Khuzistani Arabic belong to a larger isogloss covering the Caucasian–Mideastern–South Asian area, where mainly two light verbs are used, each combining with a root of nominal form of the borrowed verb. The distinction between the two light verbs is usually one of valency, and they usually derive from or are identical

48

Yaron Matras

to the lexical verbs for ‘to make/ to do’ and ‘to be/ to become’. In Domari, the carrier verbs are semi-grammaticalized, and full forms -kar- ‘to do’ and ‘-hr‘to become’ co-exist with the abbreviated integration markers -k- and -Ø-. A somewhat similar strategy is found in Mosetén, where one of the verb adaptation markers is also a valency-augment, in Yiddish, where Hebrew-derived verbs are accompanied by zayn ‘to be’, and in Jaminjung, where loans have coverb status and are always used in combination with a native inflected verb. A series of other languages use a verbalizing augment which is otherwise employed to derive verbs from non-verbs as an integration marker: Mosetén, Nahuatl, Indonesian, Guaraní, Biak, Purepecha, Manange, Rumungro, and Yaqui. The borrowed verb itself usually appears in either the root form, an infinitive form, or an unmarked inflected form, quite often – for Spanish verbs, especially – the third-person singular present. While no constraint on the borrowability of verbs can be upheld, it is nevertheless evident that a large number of languages require greater grammatical effort in integrating verbs than for the integration of nouns. The borrowing hierarchy (7)

nouns > verbs

expresses the grammatical “ease” or simplicity with which elements belonging to these two word classes can be integrated. Why is it that verbs require greater morphological integration effort, and what does this greater effort represent? On the one hand we find a pair of morphologically isolating languages like Vietnamese and Chinese, with no morphological complexity surrounding verbs in either the recipient or the donor language, and where verbs are integrated in a straightforward manner, just like nouns and other parts of speech. It is difficult, however, to attribute the need for explicit loan-verb adaptation markers even in other languages to the morphological complexity of verbs alone. Recall that most languages tend to integrate simple forms of the verb, such as the root, the infinitive, or an unmarked form. There does not seem to be, in those languages, any difficulty in stripping the target verb to its bare lexical essentials, before transposing it into the host morphology.11 I suggest instead that the difficulty lies in the conceptual complexity of the verb, and the fact that when borrowed and integrated, the verb is expected to perform two operations: The first is to serve as a referential lexical item – a content word, not dissimilar to a noun, adjective, or descriptive adverb. The second is to initiate the predication and so to serve as the principal anchor point for the entire proposition of the utterance. This latter function consti-

The borrowability of structural categories

49

tutes its “verbness”. It appears that borrowing of verbs is motivated by a similar need for modifying the inventory of lexical-referential expressions as the borrowing of nouns (and no doubt various specific semantic motivations could be postulated for groups of lexical content words). Speakers thus allow the lexical component of the verb to “cross” the mental demarcation boundary between languages, i.e. they license themselves to employ the same action/event signifier in any speech interaction. The bare lexical stem, however, is not always sufficient in order to assume the role of predication-initiator. A great number of languages therefore require this additional, crucial function to be explicitly marked out in the verbal expression; in other words, they need to transform the strictly “lexical” depiction of an action/event into a predicate. This is achieved through explicit marking of its “verbness”. A more detailed study is required in order to ascertain the conditions under which languages require some form of loan-verb integration. A first, banal observation is that a pre-requisite for the employment of loan-verb adaptation markers is the availability in the recipient language of a morphological procedure to derive verbs from non-verbs. Whereas an isolating language like Vietnamese may rely to a considerable extent on the pragmatics of morpheme juxtaposition as a way of (indirectly) marking out word classes, flectional languages will often require an additional means of identifying derived verbs. This, of course, only explains a part of the story. It is clear that the similarities in loan-verb adaptation strategies found among the languages of the Caucasian–Mideastern–South Asian area are as much areally motivated (i.e. through contact and imitation among the languages) as they are functionally motivated by the respective morphological structures of these diverse languages. Moreover, not all languages that possess verb derivation strategies employ them with loan-verbs. And finally, languages are known to have changed their loan-verb adaptation strategy over time, without adopting any significant changes to their morphological typology. Thus, Romani appears to have shared the “light verb”, valency-marking strategy of loan-verb adaptation during its early, Byzantine period, with languages of the Caucasus–Anatolia–South Asia “area” (see Matras 2002). It then transferred the function of marking out loan-verbs to a set of Greek-derived aspectual markers. Finally, in some contemporary dialects of Romani, such as Sinti, Kaale, and Vlax (spoken in or around Germany, Finland, and Romania/Hungary, respectively), loan-verb integration markers are being reduced altogether. For the time being, our principal conclusion can be that the hierarchy depicted under (7) applies for those cases where integration of a lexical item requires morphological support through derivational means.

50

Yaron Matras

7. Other parts of speech The present section gives an overview of those elements often summarized together as “function words” or “unbound grammatical lexemes”. Lumping them together in one section is a matter of convenience, and follows, as did the previous sections, the structure of the Language Convergence and Borrowing questionnaire, which forms the basis of the descriptive chapter. By discussing them under a shared heading, I am not suggesting at this stage that they share properties that motivate borrowing, nor that their behaviour in contact situations should, for any reason, be uniform or even similar at all.

7.1. Numerals Numerals, in fact, are often considered low on the borrowing scale. This may derive from an assumption that all languages have some form of quantification. Although it is now known that not all languages possess systems for counting discrete entities, it is not necessarily the clash of systems of quantification that provides the motivation for the borrowing of numerals. Several types of borrowing involving numerals can be identified. Pattern replication appears in some languages: A decimal system is reported to have been adopted as a result of contact in Mosetén, Indonesian adopts a Javanese tag-lexeme indicating “teens” and re-organizes its earlier system of numeral juxtaposition above 10 accordingly, Hup adopts the Tukano quintenary system for numerals between 5 and 20, and a combination lexemes (‘ten-and-one’) replace single lexemes (‘eleven’) in some varieties of Kurmanji, replicating the Turkish arrangement. This rather small group suggests the following implicational hierarchy representing the likelihood of pattern-replication in numerals, which is yet to be confirmed by a larger sample: (8) over 10 > below 10 More than two-thirds of the sample languages show some form of direct matter-replication of numerals. This includes most of the languages that reorganize their pattern of numerals, and which often employ borrowed numerals alongside the re-modelled “internal” or inherited system. In some cases, numeral replication is subject to sociolinguistic constraints, with contact-language numerals used as the preferred system for formal purposes such as citing dates and addresses and performing even simple mathematical tasks such

The borrowability of structural categories

51

as counting (as opposed to the casual use of numerals as attributes), in transactions involving money, in names for coins of banknotes, or in some cases in the citation of grades. Such contextual splits are described for the use of Turkish numerals in Kurmanji, for the use of Hokkien numerals in colloquial Indonesian, for dates in Rumungro, and for Chinese numerals in Vietnamese, leading us to postulate a sociolinguistic hierarchy for the likelihood of use of borrowed numerals: (9) more formal contexts > less formal contexts This hierarchy reflects the fact that numerals enter languages through the dominance of the second language in formal and business transactions, and through education and other forms of institutional discourse. In many of the sample languages, especially those in post-colonial contexts, knowledge of the indigenous system of numerals is reported to be in decline, and the younger generation shows a clear preference for borrowed numerals. The adoption of borrowings in such situations clearly favours higher numerals over lower numerals, allowing us to postulate the following implicational hierarchy for the borrowing of cardinal numerals: (10) higher numerals 1000, 100 > above 20 > above 10 > above 5 > below 5 This hierarchy appears related to some of the hierarchies postulated above, where borrowing is facilitated around conceptual complexity and inaccessibility. At the same time, higher frequency in casual language use of lower numerals is clearly a factor supporting the retention of native forms. Languages for which speakers are reported to be using native forms for lower numerals under 5 – either primarily, or alongside borrowings – but mainly or exclusively borrowings for numerals above 5 include Domari, some Kurdish speakers, Jaminjung (which has no native numerals above 3), Tasawaq, Otomi, Guaraní, Purepecha, Yaqui, Kildin Saami, and we might add Rumungro, where 79 are Greek loans into Early Romani, and 6 has been argued to be an early loan from Dardic (Indo-Iranian frontier languages). For some languages, such as Mosetén, Quichua, Nahuatl, and Biak, the cut-off point tends to be 10, while Hup shows a split at 20. Higher numerals show an independent susceptibility to borrowing. Tasawaq for instance borrows its lower numerals from Arabic, but its word for 100 from Tuareg; Rumungro generally shows Greek borrowings, but 1000 is Hungarian; and Vietnamese uses a Chinese word for 10,000.

52

Yaron Matras

Somewhat paradoxically, “0” ranks closer to the higher numerals 100, 1000, and so on: the only K’abeena borrowing is zeeruta, from Italian via Amharic, while in Rumungro it derives, like the higher numerals, from Hungarian. This is not surprising, and shows that cognitive complexity in the counting system operates in respect of the ability to easily identify and appreciate a quantity. This is hindered the greater the quantity, but it is similarly hindered in the absence of any quantity at all. An additional factor that no doubt plays a role is the relative formality of the term “zero”, which is associated with mathematical and other formal notations and transactions, but not with everyday, casual expression of “nothingness”. Sample languages that do not borrow numerals are Macedonian Turkish, Khuzistani Arabic, Yiddish, and Manange (though incipient influence of Nepali on the numeral system is reported). In all but the latter, we can attribute the stability of numerals to a firm tradition of native-language education, media, and literacy, if not widespread among all speakers, then at least firmly anchored in the community and its history. This confirms once again that the borrowing of numerals is motivated not necessarily by “gaps” in the system of counting, but by a much more general accommodation to the language of formal institutions and the public domain in the way of conceptualizing and expressing formal transactions surrounding quantification. The hierarchies presented in (9) and (10) are fully in line with the observations described for Romani dialects by Matras (2002) and by Elšík and Matras (2006), which lends support to their validity as universal indicators. A tentative case can be made for the following hierarchy of the likelihood of borrowing of ordinal numbers: (11) lower ordinals > higher ordinals This hierarchy is presented by Elšík and Matras for Romani. In the present sample it is confirmed by Kildin Saami and Rumungro, which use borrowed ordinals for ‘first’,12 and Western Neo-Aramaic, which uses Arabic ordinals for ‘110’, while Domari, Otomi and Purepecha generally rely on borrowed ordinals. Note that English (not part of the sample) is an exception to the hierarchy, having borrowed second but not first from Romance. The ordinal ‘first’ is often a separate word, quite often suppletive to the rest of the ordinal paradigm. In some languages, this is also true of ‘second’. This structural conspicuousness could be a factor promoting borrowing. In the assessment of Elšík and Matras (2006) the high borrowability of lower ordinals is a direct factor of this universal tendency to prioritize the ordinal ‘first’

The borrowability of structural categories

53

through lexical suppletion, which in turn is an expression of its cognitive saliency. Borrowing therefore simply follows the same path; in other words, the search for a renewable (=suppletive) item as an outstanding marker of the pragmatic saliency of “firstness” exploits the bilingual situation in recruiting an item from the contact language. We might therefore, in respect of ordinal numbers, postulate the following hierarchy of borrowability (12) exclusivity > inclusivity where “exclusivity” is taken to mean the separation of a single concept, entity, topic or state of affairs from a larger set – here in relation to the order of prominence, such as temporal sequence, or the attention granted to the object. Conceptually, this hierarchy is well in line with the contact-susceptibility of such properties as condition (see above), privative ‘without, instead of, except for’ (cf. Elšík and Matras 2006), contrast, phasal change (‘already’), restriction (‘only’), and the superlative (see below), all of which denote a broken chain of expectations, singling out our delimiting one entity from a presuppositional set.

7.2. Pronominal forms The borrowing behaviour of so-called “pronouns” illustrates how limited a wholesale structural approach to category borrowing can be, and how it is the functionality of categories that motivates borrowing. Only Indonesian shows borrowing of personal pronouns, from Sanskrit and Javanese, into a system of highly differentiated, lexicalized terms of address and reference. Other contact-developments in pronouns are limited to the organization of the system of reference: Imbabura Quecha is reported to have developed a polite form of the second-person pronoun kikin on the basis of Spanish Usted, and in some heavily Hispanicized varieties of Guaraní, the inherited distinction between inclusive and exclusive is dissolved. Borrowing of other deictic and anaphoric forms is limited. Spanish la is used as an anaphor in Guaraní, the Arabic resumptive pronoun iyyā- is used in Domari in relative clauses, and Rumungro borrows the Hungarian deictic prefixes am- and ugyan- which are combined with Romani deictic stems. Reflexive pronouns are borrowed in Tasawaq, Western Neo-Aramaic, and Rumungro. Yiddish shows an extension of reflexivity based on a Slavic model, and Indonesian calques a reflexive apparently on a Sanskrit model.

54

Yaron Matras

Reciprocal pronouns are borrowed in Domari and Western Neo-Aramaic, and Rumungro calques a Hungarian model. What motivates the borrowing of reflexive and reciprocal forms? Unlike straightforward deictic and anaphoric reference devices, reflexives and reciprocals may be said to constitute an extension of the derivational system of the verb, contributing to the layout of actors involved in and affected by an event. They are thus part of a construction that revolves around the verb’s “actionality”. Interrogatives are borrowed into several languages. Those that stand out as more highly borrowable are the interrogatives for quantity (‘how much’), borrowed into Domari, Otomi, and Manange, and time (‘when’), borrowed into Domari, Indonesian, Quichua, and Rumungro. Borrowing also affects indefinites. Domari, Otomi, and Rumungro borrow all or almost all of their indefinite expressions, while Tasawaq and Purepecha borrow time indefinities, Guaraní the indefinites for person and thing, and Yiddish borrows indefinite markers. Although there is no direct, predictable link to other categories, borrowing in the domains of interrogatives and indefinites appears to be a more “advanced” stage of borrowing among Other Parts of Speech.

7.3. Connectors/conjunctions The grammatical category that is by far the most susceptible to borrowing is that of connectors (see already Matras 1998). All languages in the sample borrow connectors, and the general picture confirms the implicational hierarchy postulated as universal in Matras (1998), and confirmed by Elšík and Matras (2006) for Romani (and recently by Stolz 2007 for a number of languages in contact with Italian): (13) but > or > and Sample languages that borrow all three connectors include Domari, Mosetén, Nahuatl, Kurmanji, Rapanui, Indonesian, Quichua, Otomi, Guaraní, Kildin Saami, and Western Neo-Aramaic; languages that borrow only ‘but’ and ‘or’ are Tasawaq, Purepecha, Vietnamese, Rumungro, K’abeena, and Likpe. No languages borrow ‘and’ without also borrowing ‘but’ and ‘or’. There are, however, a few languages that deviate slightly from the expected pattern. Macedonian Turkish borrows Macedonian i ‘and’ as well as ili ‘or’ and a ‘or, whereas’, but retains Turkish ama ‘but’; however, the latter is identical to Macedonian ama, which is a Turkish borrowing (cf. Matras 2004).

The borrowability of structural categories

55

Jaminjung uses the borrowed contrastive marker Kriol ani ‘only’ alongside its native bugu, while only borrowed forms are used for addition and disjunction; but this is due to the absence of any native connectors for addition or disjunction to compete with the borrowings. In Manange, Nepali ani ‘and then’ can also be used for clause coordination (cf. Stolz’s 2007 discussion of Italian allora). Neither of these cases necessarily contradicts the hierarchy in (13). Biak, however, is reported to use the Indonesian disjunction marker atau ‘or’ and less frequently the addition marker dan ‘and’, but no mention is made of a borrowed contrastive marker. Hup borrows ou ‘or’ from Tukano. The source is ultimately Portuguese, and is reported to have diffused widely in the area. Aikhenvald (2002), too, reports on Tariana ou ‘or’, with no borrowing of other Portuguese connectors. Similarly, Yaqui appears to borrow only Spanish o ‘or’. The fact that counterexamples can be found does not invalidate the overall observation that contrast is a semantic-pragmatic feature that facilitates borrowing, nor of course that clause-combining is an operational domain that is prone to contact-related change. Most likely, certain constraints of a structural and perhaps also a cultural nature (conventions on structuring discourse and expressing overt contrast) override the universal tendency in some cases. Noteworthy is the cluster of Amazonian languages within which Portuguese ou diffuses, often via secondary sources only. Borrowing in the domain of coordinating conjunctions is missing only in Yiddish and Khuzistani Arabic. Subordinating conjunctions are similarly a frequent target of borrowings. Among the complementizers, borrowing is almost entirely restricted to those that introduce factual clauses, which are borrowed in Domari, Khuzistani Arabic, Rumungro, Western Neo-Aramaic, and Likpe. Although modality has been shown to be contact-prone, at the level of the organization of the complex sentence factual complements show greater event independence and so greater effort is needed in order to process the connection between the two clauses (see already Givón 1990, Dixon 1995, 2006). Factuality at this level is thus quite in line, as a factor promoting borrowing, with contrast, limitation, exemption and other properties that we have so far encountered at the top of borrowability hiererachies. Also well in line with these tendencies is the high presence, among borrowed conjunctions that introduce adverbial clauses, of those that mark concessive relations (borrowed in Yiddish, Tasawaq, Indonesian, Quechua, Guaraní, Domari, and Likpe), causal relations (Domari, Mosetén, Nahuatl, Kurmanji, Rapanui, Jaminjung, Tasawaq, and numerous others that calque causal subordinators), and purpose (e.g. Domari, Mosotén, Nahuatl). High on the borrowing scale are also conditional subordinators

56

Yaron Matras

(borrowed in Domari, Mosetén, Indonesian, Quichua, and Guaraní), while the borrowing of temporal subordinators is often linked to that of conjunctions expressing purpose and cause. We can therefore postulate the following tentative borrowing hierarchies: (14) concessive, conditional, causal, purpose > other subordinators (15) factual complementizers > non-factual complementizers Concession, of course, is tightly linked to contrast and unexpectedness. Condition is an expression of modality. Cause and purpose are both efforts to link independent events, as are factual complementizers, while cause constitutes in addition an explicit argumentation effort at the interactional level. The hierarchies (14)–(15) thus supply us with a series of semantic-pragmatic properties that are borrowing-prone.

7.4. Particles Not many languages in the sample borrow phasal adverbs, but those that do show a clear implicational hierarchy (16) yet, already > still > (no longer) confirming that observed in Matras (1998) for Romani, as well as in van der Auwera (1998) for a sample of European languages. While Rumungro and Domari show borrowings in all positions, Jaminjung borrows ‘yet’, Guaraní only borrows ‘already’, while Otomi also has ‘still’. The semantic opposition involved is one of change vs. continuation, in the first instance. While ‘no longer’ essentially expresses change, its position on the hierarchy is partly influenced by its tendency to be composed of several structural elements. It is therefore the first two positions on the (left of the) hierarchy that are the most meaningful, and which continue the theme of contact-susceptibility of contrast and discontinuity of pre-suppositional expectations. Another particle that shows frequent borrowing is ‘again’ (Domari, Mosetén, Kurdish, Jaminjung, Indonesian, Otomi), expressing an unexpected repetition of events. Half the languages in the sample borrow focus particles, giving the implicational hierarchy (17) only > too > (even)

The borrowability of structural categories

57

once again in line with observations on Romani and other languages in Matras (1998). In fact, the particles ‘only’ and ‘too’ usually go together, but ‘only’ can be considered higher on the hierarchy, since Indonesian and Western Neo-Aramaic have borrowed ‘only’ but not ‘too’. Khuzistani Arabic borrows Persian ‘too’ (hem), but since Persian ‘only’ is itself an Arabic loan (faqat) it is not identifiable as a borrowing in Khuzistani Arabic. The hierarchy in (17) indicates that restriction facilitates borrowing, while the proneness of focus particles (and indeed of phasal adverbs and repetition adverbs) to borrowing shows the vulnerability of the system of processing states of affairs and attitudes that are high on the relevance scale and that assess information in direct relation to existing hearer-sided presuppositions and expectations. Fully consistent with this observation is the overwhelming tendency of the languages in the sample to borrow discourse markers (once again, cf. Matras 1998). There are only two languages that do not show borrowing of discourse markers: Biak and Vietnamese. There is no obvious explanation for the absence of borrowed discourse markers in these languages, except perhaps the fact that using native intonation and modal particles is considered a central characteristic of talking Biak and so an important identity marker,13 and that Chinese influence on Vietnamese was transmitted to a considerable extent via the formal and literary language, rather than via oral discourse. We also find less extensive borrowing of discourse markers in languages with a tradition of native literacy: Yiddish, Khuzistani Arabic, Macedonian Turkish, and Indonesian. Limited borrowing in this domain is also typical of Manange and Hup, while Domari, Mosetén, Jaminjung, Guaraní, and Purepecha show the most extensive use of borrowed discourse markers. It is noteworthy that Tasawaq has its fillers and discourse markers from Hausa, a contemporary “pragmatically dominant” language (cf. Matras 1998). On the whole, the following hierarchy (from Matras 1998) of borrowability, both frequency-based and at least in most cases implicational, could be upheld: (18) discourse markers > other particles Question and answer particles must also be considered in this connection. The former are not a universal phenomenon anyway, and it is not surprising that they are limited, in our case, to Macedonian Turkish, which borrows its question particle from Albanian, interestingly replacing a native Turkish question particle. The borrowing of the positive answer particle ‘yes’ is more common. It is often employed as a turn-taking particle rather than just as a signal of agreement with content, and it perhaps for this reason that it is more

58

Yaron Matras

vulnerable than its negative counterpart ‘no’. Rumungro and Mosetén borrow ‘yes’ (from Hungarian and Spanish respectively), while both ‘yes’ and ‘no’ are borrowed from Tuareg in Tasawaq, from Arabic in Domari, and from Spanish in Guaraní. The small sub-sample suggests an implicational hierarchy for the borrowing of answer particles which agrees with that postulated for Romani dialects by Elšík and Matras (2006: 343): (19) positive > negative

7.5. Grammatical vocabulary A notable gap in the borrowing inventory surrounds expressions of place deixis, which in our sample appears fully resistant to borrowing. Borrowed expressions of place are largely limited to the place indefinites ‘nowhere’ and ‘anywhere’, in which the borrowed component may either be the entire expression or just the indefinite marker (Domari, Otomi, Rumungro). The Yiddish presentative ot from Slavic is used in pointing only, and although a deixis of sorts, it is arguably more a verbalized gesture and hence more closely related to discourse markers than a member of a deictic paradigm. Borrowed time expressions encompass both indefinites (‘always’, ‘never’) and deixis (‘now’, ‘then’). Here, both classes are subject to borrowing and their behaviour appears to be linked in the following implicational hierarchy: (20) always > never > now, then In line with the high frequency of borrowing around indefinites, ‘always’ is the most frequently borrowed, followed by ‘never’. The relevance properties of indefinites – as operators that process hearer-sided presuppositions – explains their higher position on the hierarchy (at least, this is well in line with other hierarchies discussed so far). The deictic expressions ‘now’ and ‘then’ are usually linked, and therefore occupy a single position on the hierarchy; occasionally they are borrowed independently of one another: Tasawaq borrows just ‘now’ from Tuareg, while Rumungro borrows only ‘then’ from Hungarian. Note that the Romani sample (Matras 2002) shows a clear hierarchy which favours the borrowing of ‘then’, beginning most often in its sequential rather than remote-deictic function, while ‘now’ is rarely borrowed.

The borrowability of structural categories

59

Terms for days of the week are generally borrowed from the language of education, or the formal-official language: We find borrowings in Nahuatl, Jaminjung, Indonesian (ultimately from Arabic), Quichua, Otomi Guaraní, Biak, Purepecha, Hup (from Portuguese), Rumungro, K’abeena, Kildin Saami, and Western Neo-Aramaic. Borrowing of expressions for times of day (‘morning’, ‘noon’, etc.) is usually linked to days of the week, the sample showing an implicational hierarchy: (21) days of week > times of day This hierarchy can be nicely connected to the role of institutional administration and the language of commerce: Between the two categories, it is days of the week which tend to be the property of the public domain, governing the schedule of activities of individuals in relation to institutions, more so than times of the day, which have a greater autonomous role within the private domain. Once again we find that languages that do not have a strong tradition of relying on external languages for affairs of the public domain – Yiddish, Macedonian Turkish, Khuzistani Arabic – show no borrowings in these fields. Manange once again occupies a similar position, showing here too resistance toward borrowing and behaving much like a language with a tradition of literacy and institutional discourse. For adjective comparison, our sample clearly confirms the hierarchy postulated by Elšík and Matras (2006) for Romani dialects (omitting the least borrowable value “positive”, which is the default form of the adjective and is not usually accompanied by any overt derivational marker): (22) superlative > comparative Formal means of constructing both the superlative and the comparative are borrowed in Domari, Otomi, and Vietnamese, as well as Western Neo-Aramaic. Languages in which borrowing is limited to the superlative include Indonesian, with a Javanese particle paling, Rumungro, which borrows the Hungarian superlative prefix leg, and Kildin Saami, which has the Russian the superlative marker same. Yiddish, also confirming the hierarchy, shows an interesting case of hierarchical distribution of matter- and pattern-replication: While the comparative shows pattern-replication – greyser fun mir ‘bigger from me’ – the superlative shows matter borrowing, replicating the Russian marker same: same groys ‘biggest’.

60

Yaron Matras

8. Constituent order and syntax Contact-induced change in word order is generally not common in our sample. Evidence for change in basic word order is found in Rapanui, with a tendency toward change in Quichua and Otomi and sometimes in Khuzistani Arabic (here we are dealing mainly with relaxation of discourse constraints). It is plausible that in their earlier history Hup, Rumungro, and Domari showed different word-order patterns, though concrete evidence is lacking. The most common change in word order appears to affect possessive constructions. Examples are Domari, Macedonian Turkish, Rapanui, Quichua, Rumungro, and Likpe (see above, under Nominal structures). This is understandable, given the fact that a change in the position of possessor and possessed does not affect the position of the verb and so it leaves the organization of the predication intact. The position of adjectives is affected by contact in Domari, sometimes in Quichua, and in urban Manange. Relative clauses change their position relative to the head in Macedonian Turkish and in Nahuatl, as well as, arguably, in Domari, if we compare the language with attested Indo-Aryan languages. The position of the copula appears more vulnerable to change than the position of lexical verbs, as seen in Macedonian Turkish, Rumungro (tendency only), and North-eastern Neo-Aramaic. From this, we might propose the following tentative hierarchy, based partly on frequency and partly on pure prediction, for the likelihood of word order to be affected by contact-related change: (23) nominal constituents (possessor, adjective) > copula predications > verbal predications Note that the hierarchy is sensitive to the presence of a lexical verb as initiator of the predication – a factor which impedes borrowing. Thus the most borrowable are structures that do not involve a full predication, or at least not a verbal one; these are followed by non-lexical predications, while predications that contain full lexical verbs appear last. In the area of clause structure, one of the frequent changes observed is the emergence of copula clauses: Otomi borrows the Spanish copula ta, Hup uses a possible Tukano loan as a copula, and Indonesian creates copulas on a Sanskrit model. This makes sense, if one considers that non-universality of (present-tense) copula predications, and the fact that a clash of systems, and so pressure toward convergence, is more likely to occur here than in other clause types.

The borrowability of structural categories

61

As far as syntax-relevant grammatical vocabulary is concerned, negation particles are borrowed in Domari, Quichua, Guaraní, Biak, Hup. Connectors and conjunctions were dealt with above (under Other Parts of Speech), and it was seen that most languages in the sample show some kind of contact influence on clause combining strategies. This is true especially of coordination, where almost all languages are affected, some by mere borrowing of connectors, others by changing connectivity strategy, as in the case of Rapanui (from serialization to connectors), or Macedonian Turkish (from converbs to connectors introducing finite clauses). A new type of adverbial subordination is attested only in Macedonian Turkish, while relative clauses are re-structured in Macedonian Turkish, Domari, Yiddish (to some extent), and Tasawaq. Like connectors, relative particles appear to be borrowing-prone, and we find loans in Nahuatl, Rapanui, Domari, Otomi, Guaraní, Indonesian, and Kildin Saami. On the whole, then, the sample languages do not offer an extreme wealth of data on contact-induced change in clause or sentence structure. In particular, colonial languages and languages of administration do not seem to have the effect on syntactic structures as they may have on other domains of structure, in particular grammatical vocabulary. Those cases where we do find farreaching changes in syntactic typology tend to be languages in a prolonged situation of stable multilingualism, as in the case of Macedonian Turkish or Domari, confirming Thomason and Kaufman’s (1988) prediction on the link between prolonged and intense cultural contact, and significant typological disruption.

9. Lexicon All languages borrow lexical elements. Based on authors’ responses to the questionnaire, we can make a statement about the likelihood of a certain word-class element to be affected by contact. This hierarchy is not implicational, as there is no evidence to suggest that borrowing in a lower-ranking category necessarily entails borrowing in higher-ranking categories. Rather, it is based on frequency; (24) shows how frequently selected word classes occur among the list of word classes affected by contact in the sample: (24) nouns, conjunctions > verbs > discourse markers > adjectives > interjections > adverbs > other particles, adpositions > numerals > pronouns > derivational affixes > inflectional affixes

62

Yaron Matras

In some salient features, this hierarchy resembles other hierarchies proposed in the literature.14 Thus, nouns appear at the top of the list, unbound grammatical vocabulary is rather high on the list, and bound morphology is low, with derivational morphology outranking inflectional morphology. Note however some differences to the hierarchy proposed by Muysken (1981) and others: Conjunctions and discourse markers occupy a high position, and outrank some of the lexical categories. Numerals outrank pronouns and derivational morphology in this particular sample, much of it composed of languages in contact with a colonial language. No differentiation is made between coordinating and subordinating conjunctions in our discussion, but it is clear that the position even of subordinating conjunctions is far higher on the hierarchy that that assigned to it in Muysken’s data. We must, however, treat the meaningfulness of such a full hierarchy with caution. We have seen evidence that the presence of numerals on the list can be biased by the type of contact situations that are selected, and the presence or absence of a tradition of literacy in the recipient languages. The presence of pronouns on the list is largely a product of the function of terms of address and their degree of lexicalization, or alternatively of the clash of systems that distinguish exclusive/inclusive reference, and those that don’t. These and more are coincidental circumstances that influence the contact situation around particular categories, and which will promote or demote certain categories on the hierarchy, depending on the degree of presence of languages in the sample that answer to certain sociolinguistic and structural criteria. Thus, the more reliable hierarchies are those that provide a picture of the susceptibility of category values to borrowing, while the comparison among categories is not entirely free of arbitrary factors.

10. Concluding remarks Noteworthy is the extent of borrowing across the different languages of the sample. If we take as an indicator 36 prominent categories representing various aspects of structure – from phonology, through to morphology, unbound grammatical vocabulary, lexicon, and syntax15 – and assign scores to languages based on the number of categories that show some kind of contact influence, then the scores range from 31 (Rumungro, Guaraní), to just 6 (Yaqui) and 7 (Biak) (see Table 1). Table 1 shows the overall “borrowing score” for each language. It also shows the borrowing score among 11 categories representing Other Parts of

The borrowability of structural categories

63

Speech (OPS),16 an indicator of mostly unbound grammatical lexemes, and the proportion of OPS among the categories affected by contact. The majority of languages range within a 0.05 distance from the proportional 0.305 share of OPS among the category total, indicating that there is, on the whole, a rather predictable share of OPS among borrowed categories. Note however that some languages show a disproportionately high level of borrowing of OPS: Yaqui, Biak, Likpe, Mosetén, Jaminjung score between 0.40.5, meaning that OPS account for up to 50 percent of borrowed categories. Note that all these languages tend to have overall borrowing scores of between 615, that is, on the lower side of the borrowing range. This might be interpreted as indirect Table 1. So-called “borrowing scores”, and the proportion of Other Parts of Speech among categories affected by contact. Language

Total score

OPS score

OPS/Total ratio

Sampled Yaqui Biak Manange K’abeena Likpe Mosetén Mac. Turkish Rapanui Khuz. Arabic Kildin Saami Jaminjung Vietnamese Nahuatl Tasawaq Purepecha Western Neo-Aramaic Hup Kurmanji Kurdish Yiddish Domari Otomi Quichua Indonesian Guaraní Rumungro

36 6 7 10 10 10 11 12 13 13 15 15 17 18 18 19 20 21 21 22 24 25 26 26 31 31

11 3 3 3 3 4 5 3 3 4 5 7 6 5 7 7 6 5 6 7 9 9 8 9 10 10

0.305 0.500 0.428 0.333 0.333 0.400 0.455 0.250 0.230 0.307 0.333 0.467 0.352 0.278 0.389 0.368 0.300 0.238 0.286 0.318 0.375 0.360 0.307 0.346 0.323 0.323

64

Yaron Matras

evidence that borrowing begins with OPS, before continuing to other categories. The lowest OPS scores, for Rapanui, Hup, and Macedonian Turkish (0.230.25), represent languages that undergo structural changes as a result of contact but in which borrowed OPS are under-represented. These figures generally confirm the predictions and observations that unbound grammatical morphemes are high on the borrowability scale compared to other categories; though they also allow us to conclude that contact influence is rarely limited to them, and that it is not impossible for a language to display even a certain amount of resistance toward borrowing of OPS. It is also interesting to note that a number of categories occupy an entirely peripheral position in the borrowing behaviour of languages in this sample. They include bound case markers, bound tense markers, bound person markers as well as in most cases unbound person markers (deixis and anaphora; exceptions being reciprocal and reflexive pronouns, and “lexicalized” pronouns as in Indonesian), demonstratives and expressions of place deixis. In the following discussion it will hopefully become clear that the absence of borrowing in these domains is not taken to mean that constraints exclude them from being borrowed. Rather, our focus is on those categories that do show a more salient and frequent tendency to be affected by borrowing, and our agenda is to explain why speakers are motivated to borrow forms and structures in those categories. The absence of borrowing among other categories may be left to be interpreted as just that: the absence of any particular motivation to converge the two systems around these particular categories. Typological features are neither excluded nor even rarely affected by borrowing in the sample. A number of languages undergo considerable typological convergence: Macedonian Turkish, Domari, North-eastern NeoAramaic, and Rumungro. Ongoing shifts in morphological typology can be detected in a number of other languages, too. Although statistically, unbound grammatical morphemes are more likely to be borrowed than typological features, there is no direct interdependency between any specific value or category that falls within these respective groups of structures. Implicational hierarchies of the kind postulated above only apply among the values of the same category. But since the likelihood of borrowing is different for different categories, there may be a quasi-implicational relationships across categories in different structural domains. Thus, since connectors are frequently borrowed, but re-structuring within the TMA domain is rare, we might expect a language that shows contact-induced re-structuring in the domain of TMA to show borrowed connectors as well. Such an expectation is based on the higher borrowing frequency of connectors.

The borrowability of structural categories

65

Despite the lack of any direct functional link between the borrowing of connectors and the restructuring of TMA categories (or any other example of contact influence), the challenging question remains, why certain categories are more susceptible to change in situations of language contact, than others. It is here that borrowing hierarchies, especially the implicational relations among paradigm values of the same structural category, can shed some light. The fact that borrowing within such categories often follows a non-arbitrary, predictable course, suggests that semantic-pragmatic features that distinguish among category values participate in motivating borrowing. The relations among borrowed values can thus help us illuminate the motivation behind borrowing, and so help us make sense of the different degrees of susceptibility of categories to the borrowing process. Let us, for this purpose, review the hierarchies. A first set of hierarchies might be grouped together based on a general notion of frequency, referential meaning, and usage context of the borrowed structural material. This group is rather diverse. The more frequent adoption of new consonants over new vowels (1) is conditioned by the mere diversity of consonants and the fact that they tend to outnumber vowels in each of the sample languages. The greater likelihood that phonemes be adopted as part of loanwords than as independent phonological features (3), is similarly a practical issue relating to the need to accommodate loanwords. Borrowing as a utilitarian enrichment of means of expression also belongs here. The greater likelihood of borrowing of nouns over other parts of speech (as expressed in 24) is a product of the likelihood of nouns to express new concepts and to name objects and institutions (see already Weinreich 1953: 37). The high borrowability of lower ordinals (12) is connected to the fact that they mark exclusivity by assigning lower figures a special lexical item rather than a grammatical derivational procedure, with borrowing supplementing the procedure of lexical creativity. Certain usage contexts may favour borrowing, if there is a close association with the contact language in certain domains. Thus, borrowed numerals are more likely to be used in formal than informal contexts (9), higher numerals as well as mathematical “zero” are more likely candidates for borrowing, being reserved primarily to more formal-institutional contexts (10), and names of days of the week are more likely to be borrowed than times of the day (21). An additional theme, which groups together another bundle of hierarchies, may be defined as accessibility, cognitive complexity, and expectedness. Low accessibility and/or high complexity correlate with the borrowing susceptibility of peripheral as opposed to core local relations (4), of higher

66

Yaron Matras

numerals (10) (reinforcing the usage-based motivation cited above), of independent (factual) embedded events over dependent (non-factual) ones (15), and of linked independent events (purpose clauses, causal clauses) over linked dependent ones (adverbial subordinations) (14). Low expectedness can be brought in connection with some of those, and in particular with the contact-susceptible properties of contrast (13), concessive subordination (14), phasal change (16), restrictive focus (17), and superlative (22). In all of these cases, the speaker’s assertive authority is potentially reduced as a result of the speaker venturing into propositional domains involving a degree of uncertainty or unexpectedness. Not unrelated are the properties around which external circumstances reduce the speaker’s confidence and control even more overtly. These include the high borrowing susceptibility of conditionality over other subordinations (14), of participant-external modality over participant-internal modals (5), of modality itself over aspect and tense, as well as of aspect/aktionsart (the internal structure of the event, independent of the speaker’s perspective) over tense, and of the future over other tenses (6), as well as of indefinites (which rely on a presuppositional domain) over deictics (which rely on the speaker’s own orientation perspective)17 (20). Finally, we find a set of hierarchies that operate at the level of the interaction, where those structures are more borrowing-prone that are more tightly connected to the emotive level of the discourse or speech act rather than to the content level of the word or phrase. Such is the case with prosodic features over segmental phonological features (2), discourse markers over other particles (18), connectors over other parts of speech (as in 24), causal argumentation over other forms of subordination (14), and even the positive answer particle over the general (i.e. also content-bound) negative particle (19). What do these three themes – accessibility/expectedness, external dependency, and interaction-level operations – have in common, and why are they especially susceptible to contact-related change? In order to answer this question, we must return to our hypothesis about what constitutes “borrowing” in the first place (cf. Section 1). Borrowing, we had said, is a strategic compromise which bilinguals adopted in conversation and which has become socially acceptable. Social acceptability is a pre-condition for change, since language is the collective, socio-cultural product and asset of a community. But there is no reason to assume that social attitudes should in any way prejudice contrastive connectors over additive connectors, or temporal indefinites over time deixis. The reason for the hierarchical arrangement of

The borrowability of structural categories

67

categories in respect of their borrowing susceptibility has to do with the first part of our definition of borrowing, namely the part that describes borrowing as a strategic compromise adopted in conversation. It is here that speakers are naturally inclined to prioritize when handling the control mechanism that selects certain (“language-particular”) structures in certain sets of interactions. Maintaining the demarcation boundary between repertoire components (or “languages”) is a burden on the mental processing of language in conversation, and yet it is a social requirement. Compromise is sought when the tension assumes its most extreme forms: when the burden of controlling the language selection mechanism coincides with other sources of tension in the interaction itself. Such tension emerges when the speaker’s assertive authority is at stake and a special effort is needed in order to win over the hearer’s confidence: When expressing unexpected chains of arguments, when contradicting or challenging presuppositions, when assuming responsibility for propositional content that lies beyond the domain of secure knowledge, or when directly intervening with hearer-sided processing by monitoring and directing turns and speech acts (e.g. through prosody or discourse markers). Since the conversational tension around such processing tasks cannot itself be reduced, bilingual speakers’ only alternative is to eliminate the need to distinguish between sub-components of their linguistic repertoire – or “languages” – and to unify the structures that trigger the appropriate processing operations. The result is a fusion of the two systems of structures around the relevant functions (see Matras 1998). The trigger for borrowing around these kinds of structures – those that cluster around the themes defined above as “accessibility”, “expectedness”, “interaction-level”, and so on – is thus neither social acceptability, nor prestige, nor gaps in the recipient language. Rather, it is the need to reduce the cognitive load when handling a complex linguistic repertoire. Social acceptability is merely an accompanying condition for spontaneous innovations to become anchored in the long-term speech behaviour of the community. In this respect, the susceptibility of a great number of grammatical categories to borrowing is pre-determined by their language processing function, and therefore universal. One of the most striking findings of the present investigation is the fact that so many hierarchies that were identified for the cross-linguistic sample presented in this volume, were a perfect or near-perfect match to those identified by Matras (2002) and by Elšík and Matras (2006) for the sample of Romani dialects in contact with a variety of different languages, and, to the extent that material for comparison was available,

68

Yaron Matras

also with hierarchies proposed for other samples. This clearly supports their universal predictive power. Moreover, the fact that a sample consisting of multiple recipient languages shows virtually the same results as a sample with a controlled recipient language (Romani) suggests that the structure of the recipient language plays only a secondary, perhaps even just a peripheral role in determining preferences of borrowing. The primary role is played by the functionality of the categories and the extent of bilingual pressure, i.e. the extent to which bilinguals need to make frequent decisions on language choice. Note that this sharpens the focus of what Thomason and Kaufman (1988) had referred to somewhat more bluntly as the intensity of cultural contact, helping us to move toward a more specific characterization of relevant patterns of communicative interaction. To be sure, the structures of the languages involved, especially the recipient language, may play a certain role in the borrowing process. But this role must be seen primarily as an imposition of constraints on what is essentially a universal process, motivated by cognitive features of language processing. Such constraints might include the presence of a competing structure on one side of the paradigm (as with the Jaminjung contrastive marker); or the availability of literacy as a factor strengthening the coherence of the recipient system and thus reinforcing demarcation boundaries and helping to resist borrowing; or indeed the presence of social attitudes that block language mixing. On the other hand, the fact that languages like Malay employ a plethora of lexical means, and not just deictic and anaphoric expressions, to refer to participants creates a motivation for renewal of this inventory of expressions and so also for borrowing in the domain of (socalled) personal pronouns, which is not typically found in languages that rely on participant deixis and anaphora. The principal conclusion that must be drawn from the above observations is that different borrowing motivations apply to different functional categories. With some, the motivation is lexical enrichment. With others, it is the fusion of elements of formal discourse with the language that dominates formal discourse, while in a series of categories the motivation is a reduction in the tension surrounding certain language processing tasks. Though neither gaps nor social prestige are primary motivators for borrowing, both are indirectly involved, as the process of “borrowing” can be defined as a license to speakers to dismantle the mental demarcation boundaries that separate their individual “languages” and, around a particular selection of categories, to make full use of their entire repertoire of linguistic structures and forms irrespective of the setting of the communicative interaction.

The borrowability of structural categories

69

Acknowledgement This chapter was written during a research visit to the Research Centre for Linguistic Typology at La Trobe University, Melbourne, made possible through a Distinguished Fellowship at the Institute for Advanced Study, La Trobe University, and an International Linkage Fellowship of the Australian Research Council.

Notes 1. I use the term “language” in quotes in this context since it is not obvious that multilingual speakers process language in the form of separate systems; it is safer to assume that multilingual speakers have an overall repertoire of linguistic forms, to which constraints are attached concerning the situations and conversational constellations in which those forms may be used, for various purposes. The notion of a separation of “linguistic systems” on the part of the language user is therefore somewhat of an abstraction. 2. We assume that borrowing always begins with at least some degree of bilingualism, however rudimentary, or at least with an exposure to settings of communicative interaction that require the selection of a separate inventory of forms and structures. Once a certain behaviour pattern is adopted by those speakers who interact in a variety of settings – and so have access to two (or more) “languages” – new forms and structures may diffuse into the speech patterns of monolinguals as well, or may survive the historical decline of widespread bilingualism. Such latter process may strengthen our ability to identify borrowings, but it is not a pre-requisite for borrowing. 3. On the problem of establishing “borrowability” on the basis of corpus frequency, see already Weinreich’s (1953: 3536) critical remarks. 4. Since the occurrence in a corpus of a low-ranking category presupposes that of the higher-ranking category, occurrences of the higher-ranking category will always outnumber those of the lower one. 5. Throughout I use the notation “greater than” (>) to denote the value that is more likely to be affected by contact induced change (in a frequency-based hierarchy), and which in an implicational hierarchy constitutes a pre-requisite for the borrowing of any item specified to the right of it and marked “lesser than”. 6. The one exception being Kriol, which has a phonological system that is very similar to Jaminjung. 7. In both Romani and Domari, genetically related material (deriving from IndoIranian postposed adverbial specifiers) undergoes a similar development toward agglutinative case markers (cf. Matras 2002). 8. The model is in fact areal, and is also shared by Persian and Western Armenian, and to some extent by Levantine Arabic as well.

70

Yaron Matras

9. Rapanui uses Spanish tiene que which seems to express both necessity and obligation; Imbabura Quichua has pudi- “can”, which could well cover both ability and possibility. 10. We also have no evidence to uphold the (frequency) hierarchy proposed by Wichmann and Wohlgemuth (forthc.) (note that prominence of strategies is arranged from left to right): light verbs < indirect insertions < direct insertion < paradigm transfer. But we have no grounds on which to challenge this hierarchy, either. 11. See Bakker (1997), however, on constraints that prevent the isolation of the Algonkian verb to a bare stem, in the context of Cree/French contact (albeit in connection with the formation of the mixed language Michif, not with borrowing in the conventional sense). 12. Like all Romani dialects, Rumungro too uses Greek derivation markers for form ordinals from cardinals, but the hierarchy applies to the borrowing of ordinal word forms. 13. Wilco van den Heuvel, p.c. 14. Compare with integrated hierarchy presented by Muysken (1981), repeated by Winford (2003): 51: nouns > adjectives > verbs > prepositions > coordinating conjunctions > quantifiers > determiners > free pronouns > clitic pronouns > subordinating conjunctions 15. The full list is: Consonants, vowels, morphological typology, alignment type, local relations, classifiers/gender, possession, plurality, definiteness, diminution/ augmentation, nominalization, case marking, tense categories, tense marking, aspect categories, aspect marking, aktionsart categories, aktionsart marking, mood categories, modal verbs, voice and valency, numerals, personal pronouns, demonstratives, indefinites, interrogatives, connectors, subordinating conjunctions, phasal adverbs, focus particles, discourse markers, time deixis, adjective comparison, constituent order, syntax, basic cultural vocabulary. 16. These are: Numerals, personal pronouns, demonstratives, indefinites, interrogatives, connectors, subordinating conjunctions, phasal adverbs, focus particles, discourse markers, time deixis, adjective comparison. 17. More precisely, indefinites can be said to engage the hearer more actively in supplementing an imaginary knowledge domain in which the missing context can be situated: consider an indefinite expression such as ‘anywhere’, where it is up to the hearer to construct an image of possible locations that satisfy vague contextual criteria. With deixis, on the other hand, the speaker is confident that speaker and hearer share a very particular perspective. Thus, ‘here’ leaves no room for ambiguity, or for hearer-sided creativity.

The borrowability of structural categories

71

References Aikhenvald, Alexandra Y. and R. M. W. Dixon (eds.) 2001 Areal Diffusion and Genetic Inheritance. Oxford: Oxford University Press. 2006 Grammars in Contact. A Cross-linguistic Typology. Oxford: Oxford University Press. Aikhenvald, Alexandra Y. 2002 Language Contact in Amazonia. Oxford: Oxford University Press. 2006 Grammars in contact: A cross-linguistic perspective. In: Alexandra Y. Aikhenvald and R. M. W. Dixon (eds.), Grammars in Contact. A Cross-Linguistic Typology, 166. Oxford: Oxford University Press. Bakker, Peter 1997 A Language of Our Own: The Genesis of Michif, the Mixed Cree– French Language of the Canadian Métis. Oxford: Oxford Univer sity Press. Campbell, Lyle 1993 On proposed universals of grammatical borrowing. In: Henk Aertsen and Robert Jeffers (eds.), Historical Linguistics 1989: Papers from the 9th International Conference on Historical Linguistics, 91109. Amsterdam: John Benjamins. Dixon, R. M. W. 1995 Complement clauses and complementation strategies. In: F. R. Palmer (ed.), Grammar and Meaning: Essays in Honour of Sir John Lyons, 175220. Cambridge: Cambridge University Press. 2006 Complement clause types and complementation strategies in typological perspective. In: R. M. W. Dixon and Alexandra Aikhenvald (eds.) Complementation. A Cross-Linguistic Typology, 147. Oxford: Oxford University Press. Elšík, Viktor, and Yaron Matras 2006 Markedness and Language Change: The Romani Sample. Berlin: Mouton de Gruyter. Field, Fredric 2002 Linguistic Borrowing in Bilingual Contexts. Amsterdam: Benjamins. Givón, T. 1990 Syntax. A Functional-Typological Introduction. Amsterdam: Benjamins. Haugen, Einar 1950 The analysis of linguistic borrowing. In Language 26 (2): 210231. Heath, Jeffrey 1984 Language contact and language change. In Annual Review of Anthropology 13: 367384.

72

Yaron Matras

Matras, Yaron 1998 Utterance modifiers and universals of grammatical borrowing. Linguistics 36 (2): 281331. 2002 Romani: A Linguistic Introduction. Cambridge: Cambridge University Press. 2004 Layers of convergent syntax in Macedonian Turkish. Mediterranean Language Review 15: 6386. Matras, Yaron and Jeanette Sakel 2007 Investigating the mechanisms of pattern-replication in language convergence. Studies in Language 31 (4): 829865. Moravcsik, Edith 1975 Verb borrowing. Wiener Linguistische Gazette 8: 330. 1978 Language contact. In: Joseph H. Greenberg, Charles A. Ferguson and Edith A. Moravscik (eds.), Universals of Human Language, Vol. 1: 93122. Stanford: Stanford University Press. Muysken, Pieter 1981 Halfway between Quechua and Spanish: The case for relexification. In: Arnold Highfield and Albert Valdman (eds.), Historicity and Variation in Creole Studies, 5278. Ann Arbor: Karoma. Ross, Malcolm 2001 Contact-induced change in Oceanic languages in north-west Melanesia. In: Alexandra Aikhenvald and R. M. W. Dixon (eds.), Areal Diffusion and Genetic Inheritance: Problems in Comparative Linguistics, 134166. Oxford: Oxford University Press Stolz, Thomas 1996 Grammatical Hispanisms in Amerindian and Austronesian languages. The other kind of Transpacific isoglosses. Amerindia 21: 137160. 2007 Allora. On the recurrence of function-word borrowing in contact situations with Italian as donor language. In: Jochen Rehbein, Christiane Hohenstein and Luaks Pietsch (eds.), Connectivity in grammar and discourse, 7599. Amsterdam: Benjamins. Stolz, Christel, and Thomas Stolz 1996 Funktionswortentlehnung in Mesoamerika, Spanisch-Amerindischer Sprachkontakt. In Sprachtypologie und Universalienforschung 49 (1): 86123. 1997 Universelle Hispanismen? Von Manila über Lima bis Mexiko und zurück: Muster bei der Entlehnung spanischer Funktionswörter in die indigenen Sprachen Amerikas und Austronesiens. Orbis 39 (1): 177. Thomason, Sarah G. 2001 Language Contact: An Introduction. Edinburgh: Edinburgh University Press.

The borrowability of structural categories

73

Thomason, Sarah G., and Terrence Kaufman 1988 Language Contact, Creolization and Genetic Linguistics. Berkeley: University of California Press van der Auwera, Johan 1998 Phasal adverbials in the languages of Europe. In: Johan van der Auwera with Dónall P. Ó Baoill (ed.), Adverbial Constructions in the Languages of Europe, 25145. Berlin: Mouton de Gruyter. van Hout, Roeland, and Pieter Muysken 1994 Modelling lexical borrowability. Language Variation and Change 6: 3962. Weinreich, Uriel 1953 [1968] Languages in contact. The Hague: Mouton. Wichmann, Søren, and Jan Wohlgemuth Forthc. Loan verbs in a typological perspective. In: Thomas Stolz, Dik Bakker and Rosa Salas Palomo (eds.), Aspects of Language Contact. Berlin/ New York: Mouton de Gruyter. Winford, Donald 2003 An Introduction to Contact Linguistics. Oxford: Blackwell.

Grammatical borrowing in Tasawaq Maarten Kossmann

1. Background1 Tasawaq (tásàwàq) is the main language of the date palm oasis of In-Gall (íngàl), about 100 km west of Agadez in the desert of Niger (Western Africa). It is sometimes stated that Tasawaq is also spoken in Tegidda-n-Tesemt, an important salt extraction site in the region. This is true, but only in the sense that salt exploitation is seasonal labor (Bernus and Bernus 1972: 23, 30), and that in the rest of the year the great majority of salt miners stay in their homes in In-Gall. The number of speakers is unknown, but probably lies between 2,000 and 10,000. Since 1991, Tasawaq has been recognized as an official language of Niger (Sidibé 2002: 186). This status has no practical consequences. Tasawaq is a Northern Songhay language. Songhay is a close-knit language group, which is commonly regarded to be part of the Nilo-Saharan language phylum (e.g. Bender 1997). Other affiliations have been proposed, and it constitutes one of the better candidates in Africa for an isolated family. All Northern Songhay languages have been heavily affected by language contact. In the case of Tasawaq, the main language of influence is southern Tuareg (also called Tamajeq), the language of the main nomadic group in the desert around In-Gall. Tuareg is a Berber language belonging to the Afroasiatic language phylum. Languages of minor influence on Tasawaq are Arabic (Semitic, Afroasiatic), the language of religion and in earlier times also of long-distance trade, and Hausa (Chadic, Afroasiatic), the lingua franca in this part of Niger, and the language of the only urban center in the region, Agadez. Fulfulde, spoken by another nomadic group in the desert around In-Gall, does not seem to have had any influence on Tasawaq, neither grammatically, nor lexically. According to speakers of the language most Tasawaq speakers also speak Tuareg and Hausa (Sidibé 2002: 194). On the other hand, non-native speakers of Tasawaq seem to be extremely rare. Tuareg has influenced Tasawaq, both at the lexical and the grammatical level. Lexical influence includes the introduction of many items which are considered to be ‘basic’ according to most researchers who have an opinion on this, e.g. body-part terms such as ‘finger’, ‘heart’, ‘tongue’, ‘knee’ and ‘tooth’.

76

Maarten Kossmann

Northern Songhay languages in general, and Tasawaq in particular, have been analyzed as “mixed” languages, with Songhay and Tuareg components (Nicolaï 1990, Wolff and Alidou 2001). This is based on a number of arguments. In the first place, the high degree of lexical influence, in the second place the existence of separate systems for elements of different origin in some sub-systems of the morphology of Tasawaq, and finally a number of important structural features which would have a Tuareg background, e.g. the genitive adposition Hn` and SVO word order. Neither of these features are decisive for ranking the language under the “mixed” languages. For example, Northern Berber languages such as Riffian Berber – which nobody would call “mixed” in the sense of Michif or Ma’á – have similar percentages of loanwords in basic lexicon and have separate morphological sub-systems on the basis of etymological origin. The remaining part of the argument, the structural features, are not decisive, as this type of structural borrowing is attested elsewhere in non-mixed languages, and also because the adduced evidence for borrowing is not always very strong. Therefore, I will consider Tasawaq a Songhay language which was strongly influenced by Tuareg, rather than a mixture of the two languages for which no basic language can be identified.

2. Phonology The most conspicuous phonological influence of Tuareg on Tasawaq is the introduction of pharyngealized consonants, both in Tuareg and in Songhay lexemes. In Proto-Songhay pharyngealization was certainly not present. In Songhay lexemes, pharyngealized consonants mainly occur in the vicinity of a and o; in lexemes borrowed from Tuareg, there are no restrictions on their occurrence. It should be noted, however, that according to Robert Nicolaï (p.c.), who worked with a number of different informants, in most Tasawaq idiolects (among others the one described by Nicolaï 1979), consonantal pharyngealization is absent. Further influence of Tuareg and other languages mainly concerns the distribution of certain consonants. Thus, in the Songhay part of the lexicon q and γ are in near complementary distribution, q being found in word-initial position before o and in word final position after a and o, while γ appears elsewhere. Due to the introduction of especially Arabic lexemes, in which both sounds are phonemic, this distribution has become blurred. Another conspicuous effect of language contact is the change in relative frequency of certain types of consonant cluster. In the Songhay part, most consonant clus-

Tasawaq

77

ters start with a nasal or r. Other clusters are possible, but rare. The influx of foreign lexemes from languages which are less restrictive in their clustering procedures has raised the relative frequency of the other consonant clusters considerably. There exist interesting similarities between Tasawaq and Hausa phonology. In the first place, they share a phonological rule by which the short mid vowels e and o are lowered to a in all contexts except before pause (cf. Newman 2000: 399). The underlying vowel quality reappears in contexts where the vowel in question is lengthened, e.g. Tasawaq: (1)

γáy dàr ‘I stretched out’ γáy dààr-á ‘I stretched it out’ γáy dáb ‘I closed’ γáy dééb-à ‘I closed it’ γáy dàs ‘I touched’ γáy dòòs-á ‘I touched it’

In the second place, Hausa and Tasawaq have similar tonal features. Both languages have three tones, High, Low and Falling. Different from other Songhay languages such as Zarma, but similar to Hausa, Rising tone does not exist in Tasawaq. Moreover, both languages have a strong dislike for all-Low contours in polysyllabic words. In Hausa, such contours are rare and mainly occur in loanwords (Newman 2000: 606). In Tasawaq, underlying all-Low contours (which are frequent) are automatically changed to sequences with an initial Falling tone in polysyllabic words. Because of phonotactic restrictions on the occurrence of Falling tones, it is simplified to a High tone in certain syllable types. The basic all-Low contour reappears in bound syntactic contexts, e.g. in nouns when they are followed by a numeral: (2)

bângù ‘well’ bàngù hínká ‘two wells’

It is difficult to decide whether these similarities between Hausa and Tasawaq phonology are due to Hausa influence on the language, or whether they constitute independent developments.

3. Noun morphology In its noun morphology, Tasawaq makes a strict distinction between native elements and foreign elements. This is most conspicuous in plural formation. Native Songhay words have no lexical plurals; plurality is obligatorily

78

Maarten Kossmann

marked by means of an element H-yo, which occurs in noun-phrase final position.2 This is illustrated in the following examples: (3)

a. dábdè ‘piece of clothing clothing’ (underlying form: dàbdè) b. dàbdá-yo ‘clothes’ clothes-pl c. dàbdè sídày-yo ‘red clothes’ clothes red-pl

With borrowed words, on the other hand, plurality is marked on the noun rather than on the noun phrase. These plurals cannot immediately be followed by the suffix H-yo. This suffix reappears when the noun is not the last constituent of the noun phrase, e.g. (4)

a. tákàrdè ‘(sheet of) paper’ (< Tuareg) paper b. sìkárdààwà n ‘sheets of paper’ (< Tuareg) papers c. sìkárdààwà n sídày-yo ‘red sheets (of paper)’ papers red-pl

The plural formations found in borrowed elements reflect Tuareg morphology. They can be of two types. In the first type, there are both changes in the initial syllable of the noun, and changes elsewhere – either vowel changes in the stem or suffixes. This is illustrated by the following examples: (5)

singular plural singular plural àsáágù ìsúúgà < Tuareg əsegu isuga ‘comb’ táágày sígàyà n < Tuareg tagăyt šigăyyen ‘palm frond’

The second type is characterized by the suffix -(t)à n. As in Tuareg, this type is found with a small set of Tuareg nouns, but especially constitutes the canonical way of integrating loanwords from other languages than Tuareg, e.g. (6)

singular àládày àlkìtáb

plural àládààyà n àlkìtáábà n

< Hausa àládèè ‘pig’ < Arabic al-kitaab ‘book’

Tasawaq

79

The etymological split between a class of nouns of Songhay origin, which exhibits no lexical number marking, and a class of borrowed nouns with singular–plural distinctions is relatively strong. Only eleven Songhay nouns were found, which receive plurals according to the Tuareg fashion of treating loanwords (i.e. by suffixing -(t)à n), e.g. (7)

singular plural gwánsì gwánsìtàn < Songhay ‘snake’

A certain number of borrowings from Arabic and Hausa are treated as if they were Songhay nouns, e.g. (8) singular plural àssàbí àssàbí-yo < Arabic ṣabiiy ‘child’ bàṛqòòní bàṛqòòní-yo < Hausa bàrkòònóó ‘pepper’ Another interesting influence from Tuareg is found in the marking of natural gender. In the Songhay part of the language, the only non-phrasal way of expressing natural gender is by means of suppletive stems. This is only found in a few cases; many items which would be gendered in other languages are left unmarked for gemder in Tasawaq. Examples: (9) àlzírày ízè báàbà áàrù báynà

‘male or female in-law’ ‘son, daughter’ ‘father’ náànà ‘mother’ ‘man’ wây ‘woman’ ‘male slave’ ṭààmú ‘female slave’

< Songhay < Songhay < Songhay < Songhay < Songhay

Tuareg has a derivative gender system, in which masculine denotes male animates and larger (in relation to the corresponding feminine form) inanimates, while feminine denotes female animates and smaller inanimates. In Tasawaq, this system is reflected in a consistent manner in borrowings denoting animates. The size difference with inanimate concrete nouns, on the other hand, is only reflected in a few lexicalized pairs, and cannot be considered a feature of Tasawaq grammar. The natural gender difference is illustrated by the following pairs: (10) àbóóbàz àgéélìm ááràb

‘male cousin’ tàbóóbàz ‘female cousin’ < Tuareg ‘male orphan’ tàgéélìm ‘female orphan’ < Tuareg ‘Arab man’ tááràb ‘Arab woman’ < Tuareg

80

Maarten Kossmann

4. Adjectives Songhay languages have a special class of words, which are only used as nominal modifiers. I will refer to them as “adjectives”. Many of them are derived from verbs by special morphological devices. Tuareg, on the other hand, has no class of adjectives. The type of modification carried out by adjectives in Songhay, is provided by relative clauses in Tuareg. In subject relatives, Tuareg uses a special form of the verb, the so-called participle. With stative verbs, which are very common in Tuareg adjective-like relative clauses, the participle (m.sg.) is marked by a suffix -ăn. Tasawaq has retained the Songhay system of modification with adjectives, most of which are morphologically derived from verbs. There are several regular formations of adjectives. Adjectives based on monosyllabic verbs (always with the shape CVC) are formed by lengthening the (underlying) vowel of the verb and adding a suffix -o. The lexical tone is replaced by a L-H tone pattern, e.g. (11) nàq fár qwáṣ zìr

‘to press’ ‘to open’ ‘to cut’ ‘to wipe’

nààqó fèèró qòòsó zììró

‘pressed’ ‘opened’ ‘cut (adj.)’ ‘wiped’

A similar morphological device is found with disyllabic verbs ending in a vowel. In these verbs, the final vowel is substituted by -o, and the lexical vowel pattern is replaced by a L-H contour, e.g. (12) síírí ‘to be crooked’ sììró ‘crooked’ fúmbú ‘to stink’ fùmbó ‘stinking’ qwàrnó ‘warm’ qwárnò ‘to be warm’ These two related morphological devices reflect Songhay patterns. All verbs to which they apply have a Songhay origin. The other verb types have a different morphology, which consists of the suffixing of -à n, under some circumstances accompanied by vowel lengthening in the preceding syllable. The lexical tone pattern of the verb is retained. Verbs of this class include both original Songhay verbs and verbs borrowed from Tuareg, e.g.

Tasawaq

(13) bààráy kàkáy fáṛàṭ yízmàm gílìllìt fùsús

‘to change’ ‘to build’ ‘to sweep’ ‘to squeeze’ ‘to be round’ ‘be light’

bààráyà n kàkááyà n fáṛàṭà n yízmàmà n gìlíllìtà n fùsúúsà n

‘changed’ ‘built’ ‘swept’ ‘squeezed’ ‘round’ ‘light’

81

< Songhay < Songhay < Tuareg < Tuareg < Tuareg < Tuareg

The suffix -à n reflects the masculine singular suffix of the Tuareg ‘participle’ (subject relative verb form). It should be noted that Songhay verbs which have the required structure for this type of adjective formation are quite rare, while on the other hand all Tuareg loan-verbs belong to this class. Thus, although the distribution of adjectival formations is ruled by structure and not by etymology, there is a clear etymological scission between the class of -o final adjectives (all of which have a Songhay background) and the -à n final adjectives, most of which are based on Tuareg verbs.

5. Verbs The verb in Tasawaq is entirely Songhay in its structure. Different from Tuareg, mood, aspect and negation are not marked in the verb stem. As in other Songhay languages, these categories are expressed by porte-manteau morphemes, which immediately precede the verb. The positive Perfective is unmarked. (14) γá b-sì ṭàkááfùṛ. 1sg imperfective-speak French ‘I speak French.’ Unlike other Northern Songhay languages, the subject is not obligatorily expressed by a pronoun when a lexical subject is present. The Tasawaq construction probably reflects proto-Songhay at this point, e.g. (15) sáy í-n nààná-yo sìní … well 3pl-of mother-pl say ‘And their mothers said …’ Phrases with a lexical subject followed by a subject pronoun are quite frequent, e.g.

82

Maarten Kossmann

(16) àžéémùr à ṇáṣ calf 3sg be.fat ‘The calf is fat.’ This may represent influence from Tuareg, which has obligatory subject inflection, or Hausa, where the subject is obligatorily expressed in a portemanteau morpheme combining pronominal and aspectual information. Tuareg verbs are borrowed according to two strategies. In about half of the borrowed verbs, a form without any reflex of Tuareg person–number affixes is used. This class includes all verbs with more than two syllables as well as a number of disyllabic verbs, e.g. (17) gílìllìt ‘to be round’ < Tuareg gələllət ‘be round’ (Short Imperfective)3 In the other half, the Tuareg 3sg:m prefix y(ə)- is taken over as y(i)- preceding the verb form. This is only found in disyllabic verbs. In Tasawaq the initial y does not refer to person, but is part of the verb stem, e.g. (18) γáy yízmàm < Tuareg y-əẓmăm 1sg press 3sg:m-press:perfective ‘I pressed.’ ‘He pressed.’ In Tuareg, aspect is marked by different vowel patterns in the verb stem. This provides us with the opportunity of deciding which Tuareg aspectual form was used as the basis of the Tasawaq verb. With verbs without the y- prefix, this turns out to be a difficult question, as all kinds of vowel patterns are found and no specific form can be discerned as the basis of borrowing. With y- initial verbs, on the other hand, one finds an interesting distribution (see Kossmann fc.). Tasawaq y- initial verbs which refer to actions reflect the (Positive) Perfective aspectual forms of Tuareg. They all share similar vowel patterns (mainly i ~ a, reflecting the Tuareg Perfective scheme ə ~ ă) and a High–Low tone pattern (reflecting Tuareg penultimate stress), e.g. (19) yígmàm ‘to chew tobacco’ < Tuareg ‘y-əgmăm 3sg:m Perfective yílmàq ‘to swim’ < Tuareg ‘y-əlmăγ 3sg:m Perfective yínḍàb ‘to shoot’ < Tuareg ‘y-ənḍăb 3sg:m Perfective Tasawaq stative y-initial verbs reflect another Tuareg aspectual form, the Resultative. This is shown by the vowel pattern (mainly i ~ a, reflecting the Tua-

Tasawaq

83

reg Resultative scheme ə ~ a), and by the Low–High tone pattern, reflecting Tuareg final stress, e.g. (20) yìgdá ‘to be right’ < Tuareg yəg’da 3sg:m Resultative yìggád ‘to be shy’ < Tuareg yəg’gad 3sg:m Resultative yìláz ‘to be ugly’ < Tuareg yə’laz 3sg:m Resultative One remarks that in Tuareg every verb may appear in any of the aspectual stems. In Tasawaq, this aspectual difference has been lexicalized as a difference in tone class. Both Tuareg and Songhay use affixation for verb derivation. In Tasawaq only one Songhay-based verb derivation exists, the causative suffix -n`dá, for example, (21) káání ‘to sleep’ káán-ìndá ‘to put to sleep’ (< Songhay) In the course of borrowing, many originally derived Tuareg verbs have been borrowed into Tasawaq. In a few cases both a derived verb and an underived verb have been borrowed. This is relatively rare, and far from systematic. Tasawaq does not have the intricate interaction between underived Songhay-based verbs and derived Tuareg-based verbs as found in some other Northern Songhay languages, such as Tadaksahak (Christiansen and Christiansen 2002).

6. Verbal nouns The morphology of Tasawaq verbal noun formation is split according to the etymological origin of the verb. Most Songhay verbs derive their verbal noun either by zero derivation (in one specific class tonal change), or by the suffixation of a suffix -yo (tone uncertain), e.g. (22) verb ḅáq ḍà n bá n

verbal noun ḅáq ḍâ n bá n-yo

origin ‘to break / the breaking’ Songhay ‘to sing / song’ Songhay ‘to finish / end’ Songhay

Tuareg verbs are taken over together with the corresponding Tuareg verbal noun. The great deal of variation and irregularity in Tuareg verbal noun formation is reflected in Tasawaq, e.g.

84

Maarten Kossmann

(23) verb fáṛàṭ kírìnzìt yìlkám zìrgín

verbal noun àfáṛàṭ àkírìnzì àlákàm tàzárgàn

‘to sweep/sweeping’ ‘to claw/clawing’ ‘to follow/following’ ‘to be dirty/dirtyness’

origin Tuareg Tuareg Tuareg Tuareg

7. Other word classes The only obvious influence of Tuareg on the pronominal system of Tasawaq is the introduction of the reflexive element ímà n ‘self’ (< Tuareg iṃan ‘self, spirit’, constructed with a genitival phrase (‘his self’), e.g. (24) γá b-ṣíṛìnkìṭ γá nn ímà n-sí. 1sg imperfective-comb 1sg of self-to ‘I am combing myself.’ The corresponding constructions in Tuareg and in Songhay are similar, both using a genitival phrase. Songhay languages normally have a noun meaning ‘head’ where Tasawaq uses ímà n. Numerals 4 to 10, as well as the decades, are borrowed from Arabic; the numerals 100 and 1000 are borrowings from Tuareg. There is an interesting syntactic difference between the cardinal numerals from 1 to 19 and the cardinal numerals from 20 onward. In the first group, the numeral follows the head noun, thereby reflecting Songhay syntactic patterns, e.g. (25) bàngù hínká bàngù sábàγà

‘two wells’ ‘seven wells’

In the second group, the numeral precedes the head noun and is linked to it by means of a particle Hn`, e.g. (26) γàssìrín in ṭáṛṛày ‘twenty roads’ xàmsín ìn bàngù ‘fifty wells’ téémàdá n bàngù ‘a hundred wells’ This closely reflects the Tuareg pattern, in which the genitive preposition n occurs between the higher numeral and the noun.

Tasawaq

85

Different from Tuareg practice, in both constructions the quantified noun has the singular form (thus ṭáṛṛày ‘road’ rather than ṭáṛṛààyà n ‘roads’). For most other quantifiers no certain loan origin could be discerned, the main exception being àlkúl ‘every’, which was borrowed from Arabic. Simple postpositions all derive from Songhay. A possible exception is the genitival adposition Hn, which in Tasawaq relates a possessor (in first position) to a possessed (in last position), e.g. (27) hááwí n gí cow of grease ‘grease of a cow’ Other Songhay languages mostly use Possessor–Possessed constructions without a linking adposition. As Tuareg has a genitival preposition n, it is generally assumed that this was borrowed into Tasawaq (as well as into the other Northern Songhay languages). In the opinion of the present writer, this is far from evident. In the first place, the Tuareg construction has the inverse structure, i.e. Possessed–Possessor. In the second place, the tone pattern of the Tasawaq adposition remains unexplained if one considers it a borrowing. I consider a Songhay origin (maybe somehow related to the old Songhay genitival pronoun wánè ‘that of’) a serious alternative. Many spatial relationships are expressed by means of complexes containing a preposition and a nominal element. This nominal element is often a borrowing from Tuareg, e.g. (28) kùsú nn ámmàs < Tuareg aṃṃas ‘the inside’ pot of inside ‘inside the pot’ A number of coordinating and subordinating particles have been borrowed. These include most coordinators: kó ‘or’ (< Hausa), mé ‘or’ (< Tuareg), àmmá ‘but’ (< Hausa < Arabic); the preposition n`dá ‘and, with’ (mainly used in NP coordination), however, has a Songhay background. Subordinating particles less commonly have a foreign origin; I have noted wàlá ‘even if’ (< Tuareg < Arabic) and compound tún gá ‘because’ (from Hausa tún ‘because’ and Songhay gá ‘on’).

86

Maarten Kossmann

8. Syntax All Northern Songhay languages have relatively strict S–Aux–V–O word order, while most Songhay languages have an alternation of S–Aux–O–V word order with less frequent S–Aux–V–O word order, mainly found with low-transitivity two-place verbs (Heath 1999: 161). Because of this difference in word order it has sometimes been claimed that Northern Songhay changed its word order under the influence of Tuareg. This claim is difficult to substantiate. In the first place, Tuareg is a VSO language, so the influence would have to be indirect. In the second place, S–Aux–V–O is also attested in Songhay outside the Northern group, in Timbuktu Songhay (Koyra Chiini; Heath 1999b). One part of syntax where Tuareg influence is conspicuous is the formation of relative clauses. In Songhay languages outside Northern Songhay, relative clauses are marked by means of a relative particle kaŋ. This marker is quite similar to a relative pronoun: in adpositional relatives the postposition follows kaŋ, and in most Songhay languages there is no (other) pronominal reference to the head noun in the relative clause. In Tuareg, a different system is found. The relative clause is marked by the use of certain verbal forms (the ‘participle’ with subject relatives) and certain syntactic features (esp. clitic fronting). There is no pronominal reference to the head of the relative clause, except with subject relatives: the ‘participle’ is marked for number and gender. There exists a three-way opposition between relative clauses neutral to definiteness, overtly definite relative clauses, and overtly indefinite relative clauses (Galand 1974). The neutral relative clauses have no overt linking to the head noun, while the definite and indefinite relatives use different pronominal markers as a linking device. In Tasawaq there exist two relative constructions, a construction without relative linker, and a construction which uses a pronominal element à-γó, lit. ‘this one’. In the construction without linker, the head noun may bear the normal deictic clitic -γo, ‘this’ when it is definite; when the head noun is indefinite, no elements come in between the noun and the relative clause. The head of an à-γó relativization is always indefinite. The following examples of subject relatives illustrate this: (29) ààrù-γó [gáw àssáγàl né bí] M à sí. man-proximal [work work here yesterday] PN 3sg be ‘The man [that worked here yesterday] is M.’ (definite head; no relative linker)

Tasawaq

87

(30) γáy gùn(á) bàrá-fó [b-gáw àssáγàl]. 1sg see man-one [imperfective-work work] ‘I saw a man [who was working].’ (indefinite head; no relative linker] (31) áfàẓò hún-kàt [àγó fìrízì]. reed grow-ventive [relative green] ‘Green reed grew.’ (indefinite head; relative linker] The constructions and differentiations are quite similar to Tuareg patterns. Tuareg patterns are also reflected in the way Tasawaq treats extraction from a postpositional phrase (for Tuareg cf. Heath 2005: 633ff.). Both Tuareg and Tasawaq in such cases put the adposition (without proniminal reference) on the left edge of the relative clause. Depending on the type of relative clause, this adposition follows a relative particle or stands alone, for example, (32) γá b-ní n àššááhì [kúná súúkàr à ssí]. 1sg imperfective-drink tea [on sugar 3sg be.not] ‘I drink tea without sugar.’ (Lit. I drink tea [on which there is no sugar]) (33) γáy gùn(á) tùgúzì-fóó [àγó gá àssàbí bárà]. 1sg see tree-one [relative on child be] ‘I saw a tree [on which there was a child].’ 9. Conclusion Tasawaq is strongly influenced by Tuareg and, to a lesser extent, Arabic and Hausa, both in its matter and in its patterns. In lexical borrowing, one finds an interesting feature, which consists of the introduction of foreign morphology through borrowing. Thus in nominal plural formation and in verbal noun formation, elements of different etymological origin have completely different morphologies, reflecting their original morphology. Something similar has probably been the background of the distribution of two adjective formations, one using Songhay morphology, the other using borrowed Tuareg morphology. Pattern borrowing is also found a lot. As the reconstruction of Songhay syntax is a difficult matter, it is often difficult to decide whether certain patterns derive from Songhay (with changes due to internal factors), or whether they have been taken over from Tuareg.

88

Maarten Kossmann

Notes 1. All Tasawaq data in this chapter were collected by the author during fieldwork in Niger in Fall 2003. The fieldwork was carried out with one single informant, Mrs. Ibrahim, born Nana Mariama Aweïssou, a school teacher in her twenties, originary from In-Gall, now resident in Agadez. I wish to thank her for her time and patience. Mrs. Ibrahim has fluent command of Tasawaq, Hausa and French, but does not speak Tuareg. Her idiolect is unusual, it seems, in a number of respects, esp. the presence of phrayngealized consonants, which most speakers of Tasawaq seem to lack (Robert Nicolaï, p.c.). The chapter was written in the framework of the NWO (Netherlands Organization for Scientific Research) research project “Tuareg and the Central Sahelian Languages. A History of Language Contact”. Other sources on Tasawaq are Alidou (1988), Nicolaï (1979, 197984, 1980), Wolff and Alidou (2001). I thank Robert Nicolaï, who gave me the opportunity to listen to his Tasawaq recordings, which helped me a lot in clarifying a number of questions. Transcription follows general practice in the field of African linguistics, with the exception of the following. Nasalization of a vowel is indicated by a superscript n following the vowel. Nasalization is in most cases an allophone of the nasal n, but is unsystematic in word-final position. Pharyngealization is indicated by a subscript dot. Superscript H indicates a floating High tone. In Tuareg transcriptions ă indicates a short low central vowel. Tone is consistently marked; absence of tone marking reflects the author’s uncertainty about the transcription. 2. The tone of the final syllable is difficult to hear, and has been left unmarked in the examples. The application of the preceding floating High tone, on the other hand, is well established. 3. For ease of reference, I adopt the terminology in Heath (2005) for the different aspectual stems. Other researchers on Tuareg use different terms.

References Alidou, Ousseïna 1988 Tasawaq d’In-Gall. Esquisse linguistique d’une langue dite “mixte”. Mémoire d’Études et de Recherches sous la direction de Prof. Dr. Ekkehard Wolff, Université de Niamey. Bender, M. Lionel 1997 The Nilo-Saharan Languages. A Comparative Essay. Munich: Lincom. Bernus, Edmond, and Suzanne Bernus 1972 Du sel et des dattes. Introduction à l’étude de la communauté d’In Gall et de Tegidda-n-tesemt. Études nigériennes 31. Niamey: Centre Nigérien de Recherches en Sciences Humaines.

Tasawaq

89

Christiansen, Niels and Regula Christiansen 2002 Some verb morphology features of Tadaksahak. SIL Electronic Working Papers. (http://www.sil.org/silewp/abstract. asp?ref=2002005). Galand, Lionel 1974 Défini, indéfini, non-défini: Les supports de détermination en touareg. Bulletin de la Société de Linguistique de Paris 69 (1): 205224. Heath, Jeffrey 1999a A Grammar of Koyraboro (Koroboro) Senni. Cologne: Rüdiger Köppe. 1999b A Grammar of Koyra Chiini. Berlin/New York: Mouton de Gruyter. 2005 Tamasheq (Tuareg of Mali). Berlin/New York: Mouton de Gruyter. Kossmann, Maarten Forthc. The borrowing of aspect as lexical tone: y-initial Tuareg verbs in Tasawaq (Northern Songhay). To appear in Studies in African Linguistics. Newman, Paul 2000 The Hausa Language: An Encyclopedic Reference Grammar. New Haven/London: Yale University Press. Nicolaï, Robert 1979 Le songhay septentrional (études phonématiques). Bulletin de l’IFAN, 41, série B; part 1: 304370; 539567; 829866. 197984 Sur la phonologie des langues “mixtes” du songhay septentrional. Comptes rendus du GLECS, 2428: 395412. 1980 Le songhay septentrional (études prosodiques). In: Itinérances … en pays peul et ailleurs. Mélanges réunis à la mémoire de Pierre François Lacroix, I, 261289. Paris: Société des Africanistes. 1990 Parentés linguistiques (à propos du songhay). Paris: Éditions du CNRS. Prasse, Karl-G., Ghoubeïd Alojaly, and Ghabdouane Mohamed 2003 Dictionnaire touareg-français (Niger). Copenhagen: Museum Tusculanum Press. Sidibé, Alimata 2002 Analyse critique de quelques opinions sur l’idiome des isawaghan: Le tasawaq. Mu ƙara sani. Revue de l’Institut de Recherches en Sciences Humaines (Université Abdou Moumouni, Niamey) 10/12: 185197. Wolff, H. Ekkehard, and Ousseïna Alidou 2001 On the non-linear ancestry of Tasawaq (Niger). Or: how “mixed” can a language be? In: Derek Nurse (ed.), Historical Language Contact in Africa, 523574. (Special volume of Sprache und Geschichte in Afrika 16/17).

Grammatical borrowing in K’abeena Joachim Crass

1. Background K’abeena is a Highand-East Cushitic (HEC) language spoken by some 35,000 people living in and around the town of Wolkite. This small town is situated some 160 km south west of Addis Ababa, the capital of Ethiopia, at the northwestern edge of the Gurage region. The Gurage region, a settling of people speaking South-Ethiosemitic languages, is surrounded by areas that are inhabited by people speaking Cushitic languages. These Cushitic languages are – apart from K’abeena – Alaaba, Hadiyya, Libido, and Oromo. The closest related languages to K’abeena are Alaaba, Kambaata and T’imbaaro. These four languages form one subgroup of HEC. The other four subgroups are (1) Burji, (2) Gedeo, (3) Sidaama, and (4) Hadiyya and Libido. K’abeena is a member of the Ethiopian linguistic area (ELA). It is in contact mainly with the South-Ethiosemitic languages Amharic, Chaha, Ezha, Muher, Wolane and with the Lowland East Cushitic language Oromo. Oromo is spoken to the north, Chaha to the south, Ezha to the southeast and Muher and Wolane to the east of the K’abeena speaking area. However, this is only an approximation. In fact, the main town Wolkite and the surrounding rural area are multi-ethnic, i.e. no clear boundaries can be drawn between settling areas of various groups. Almost all K’abeena speak Amharic as a second language, which is the lingua franca of Ethiopian towns (Meyer and Richter 2003) and of the Gurage region. The other contact languages are known by individuals who live or grew up in the respective contact area. Many K’abeena speak three or four languages. The South-Ethiosemitic languages belong to different subgroups of SouthWest-Semitic. Amharic belongs to one branch of Transversal South Ethiopic, Wolane to the second branch of Transversal South Ethiopic, called East Gurage. Chaha, Ezha and Muher belong to different groups of Outer South Ethiopic (Faber 1997: 6). HEC and the Gurage languages form a sub-area of the ELA, which we name HEC-Gurage sub-area. The existence of this sub-area is proposed by Zaborski (1991), who calls it Gurage-Sidamo sub-area. Unfortunately Zaborski does not discuss its features (for recent discussion cf. Bisang 2006, Crass and Bisang 2004).

92

Joachim Crass

The K’abeena reached their actual settling area in the second half of the nineteenth century, after splitting of from the Alaaba (Braukämper 1980). Culturally the K’abeena do not differ considerably from the people speaking Ethiosemitic languages. The staple food is Ensete edulis, known as “false banana”. Whereas the K’abeena are exclusively Muslims, the Ethiosemitic peoples are mainly but not exclusively either Christians or Muslims. This chapter deals with features found in K’abeena and other languages of the area. However, in most of the cases it is difficult to define the source of these features. Several features represent grammaticalization processes or pattern-replication. However, this does not mean, that these features should be excluded as contact-induced features. Especially in the case of rare or unattested grammaticalizations, contact-induced change is one possible way of explaining the similarities (cf. Bisang 1996, Heine 1994, Heine and Kuteva 2003). Furthermore it is not possible to decide whether these features belong to the ELA or to the HEC-Gurage sub-area of the ELA. The other languages dealt with in this chapter are Amharic, Libido, Oromo, Muher, Wolane, Gumär (Outer South Ethiopic) and Zay (East Gurage). Gumär and Zay are not direct contact languages of K’abeena but they are members of the HEC-Gurage subarea. The examples to illustrate the features are all taken from K’abeena. The data of all languages is an outcome of research conducted by the author of this chapter and by Ronny Meyer.1 The research was initiated by Crass in undertaking a comparison between K’abeena and Amharic. It later was extended to other languages of the HEC-Gurage sub-area. Crass provided data on Libido, Meyer provided data on Gumär, Muher, Wolane, Zay, and Oromo. The findings of this research will be published in Crass and Meyer (2007).

2. Phonology Within the sound system system of K’abeena /p/ is most probably introduced due to contact with Ethiosemitic languages via loanwords, to which this marginal morpheme seems to be restricted. The same is true for [ä] which I do not consider to be a phoneme in K’abeena; rather, it is a phonetic variant of /a/. The distribution is not easy to formulate (for a discussion cf. Crass 2005: 25f.). A phonological feature of the ELA is the presence of ejectives (for extensive discussion cf. Crass 2002). K’abeena possesses four ejectives, namely the ejective plosives /p’/, /t’/ and /k’/ and the ejective affricate /c’/. For Proto-HEC only /k’/ is reconstructed by Hudson (1989: 11). On the basis of this finding, one can suppose that ejectives were introduced into the consonant inventory

K’abeena

93

in course of time, possibly due to contact. However, there is no evidence that ejectives were introduced in K’abeena in recent time. If Hudson’s reconstruction holds true, the introduction of ejectives must me very old. Palatalization of alveolar consonants to post alveolar or palatal consonants as a morphophonological process is another areal feature, which might have spread due to contact. In K’abeena the following palatalizations occur: alveolar plosive /t/ to post-alveolar affricate /c/, alveolar plosive /d/ to post-alveolar affricate /j/, alveolar ejective plosive /t’/ to post-alveolar ejective affricate /c’/, alveolar fricative /s/ to post-alveolar fricative /sh/, alveolar fricative /z/ to postalveolar fricative /zh/, and the alveolar nasal /n/ to the palatal nasal /ñ/. In Amharic, in addition, the alveolar lateral /l/ is palatalized to the approximant /y/.

3. Nominal structures The ablative case marker, which itself is not considered to be a result of language contact, is used with verbs to form ‘since’-temporal and real conditional clauses in all languages except Oromo (for the grammaticalization of an ablative case marker to a ‘since’-temporal clause marker, see Haspelmath 1997: 66ff., Heine and Kuteva 2002: 35). In Muher and Gumär the use of the ablative case marker is restricted to ‘since’-temporal clauses. In K’abeena the ablative case marker is the suffix -VVcci. Example (1) includes the function as ablative case marker and as marker for ‘since’-temporal clauses, example (2) the function as real conditional clause marker. (1)

jarman-íicci ’ameeccoomm-íicci kabare. Germany-abl come.prv.1s-abl today.gen ’agana saaminta ’ikko. month.acc week.acc be.prv.3s.m ‘It is five weeks ago since I came from Germany.’

(2)

c’aata k’ama’yoomm-íicci ’óssuti ’affaa-’e-ba. khat.acc chew.prv.1s-abl sleep.nom hold.ipv.3s.f-1s-neg ‘If I chew khat I cannot sleep.’

4. Verbal structures The past tense marker is used in K’abeena and in all other investigated languages of the area to mark the apodosis of irreal or counterfactual conditional

94

Joachim Crass

clauses. This is remarkable because a past marker mostly occur in the protasis (Fleischman 1989: f.) In K’abeena the past tense marker is the suffix -kk’i. Example (3) shows the function as past tense marker, example (4) and (5) the function as a marker of the apodosis. (3)

’anni bokkóoni ’agáre-’e yiye-he father.gen house.loc wait.imp.s-1s say.cnv.1s-2s ’oro’yoommi-kk’i-ba-indo? go.prv.1s-pst-neg-q ‘Didn’t I leave by saying to you: “Wait in the house of my father!”?’

(4)

beréta t’eenoo ’ubbo-ba-’ikkáani t’aafaa yesterday rain.nom fall.prv.3s.m-neg-cnd Tef.acc ’udunnáammi-kk’i. thresh.ipv.1s-pst ‘If it had not rained yesterday we would have threshed Tef.’

When no adverb indicates tense, a past or a non-past interpretation is possible. (5)

t’eenoo ’ubbo-ba-ikkáani t’uma-ha-kk’i. rain.nom fall.prv.3s.m-neg-cnd good-cop-pst ‘It would have been good if it did not rain.’ ‘It would be good if it did not rain.’

The prospective aspect (cf. Comrie 1976: 64f.) is marked in K’abeena by a verbal noun in the dative (cf. example (6)) and in Ethiosemitic languages by a verbal noun with possessive suffixes. In both cases the verbal noun is followed by a copula. In Ethiosemitic languages two morpheme orders occur. Either the copula precedes the possessive suffix (e.g. Wolane) or it follows the possessive suffix (e.g. Muher). (6)

’áni timhirtíta shuuliiháa-ti. 1s.nom study.acc finish.vn.dat-cop ‘I am about to finish my studies.’

The prospective aspect is distinguished formally by the intentional. The latter is expressed with a subordinate clause followed by a copula. In K’abeena, the

K’abeena

95

predicate of the subordinate clause is a converb expressing purpose (example (7)). In Ethiosemitic languages the non-independent imperfective is marked either with a purpose or a locative marker. (7)

hokkoppaati ’intotáa-ti. afternoon.snack.acc eat.cnv.purp.1p-cop ‘We intend to eat our afternoon snack.’

A formal distinction between prospective aspect and intentional is found in K’abeena, Amharic, Gumär, Muher, Wolane, and Zay but not in Libido and Oromo. In the latter two languages the prospective aspect and the intentional are not differentiated morphosyntactically. Libido uses a converb plus a copula for both categories, i.e. the same construction used in K’abeena only for the prospective aspect. In Oromo, however, a construction consisting of a verbal noun in the dative or a verbal noun marked with a possessive morpheme followed by a copula is used. The experiential perfect is expressed by a construction including the respective verb ‘know’ in the main clause and its complement expressed by a converb. The latter expresses the event experienced by the subject. (8)

’ameerikáani ’orootéeni kasseenta-’i? ’ee, ameerikáani America.loc go.cnv.2p know.prv.2p-q yes America.loc ’oróo’ni kansóommi. go.cnv.1p know.prv.1p ‘Have you ever been in America?’ ‘Yes, we have [the experience to have] been in America.’

All languages except Oromo lack a verb ‘have’. To express possession the respective verb ‘to exist, to be present’ is used. The possessor is marked with the dative case, the possessum with the nominative. (9)

kii bíkku c’úulu yóo-’e. 2s.gen size child.nom exist.3-1s.obj ‘I have a child of the same age/size as yours.’

This construction is used in all languages except Libido to express obligation, too. In this case the subject is a verbal noun. Libido cannot make use of this construction, because it does not have object agreement markers.

96

Joachim Crass

(10) ’oró’u yóo-’e. go.vn.nom exist.3-1s.obj ‘I have to go.’ Converbs occur in many languages of the ELA. Typical is that converbs usually cross-reference the subject. According to Tosco, the presence of the converb in most Ethiopic Semitic languages is probably the result of old Cushitic influence … Within Cushitic, the converb is typically found in the languages of the highlands … It is likewise found in the Omotic languages of the highlands, … and can therefore be considered a genuine areal feature. (Tosco 2000: 345)

The number of converbs differs considerably in the languages. Whereas in Ethiosemitic languages generally one or two converbs occur, the number of converbs in HEC languages is much higher. In K’abeena two mainly sequential converbs, one progressive converb, one negative converb, one affirmative and one negative purpose converb occur. Furthermore, K’abeena marks objects on the verb according to the primary/secondary object pattern, i.e. in the case of ditransitive verbs the recipient is marked on the verb, not the theme. This pattern is attested also for Ethiosemitic languages. (Ethio-)Semitic verbs are integrated into K’abeena on the basis of the 3s.m of the perfective, which is the citation form of verbs in Semitic languages. This is remarkable because the 3s.m is the citation form only according to scientific tradition, not in the language use. Normally, the citation form is the verbal noun. However, the 3s.m seems to have a similar function in the language use. The final vowel in (Ethio-)Semitic verbs, which marks the 3s.m, is replaced by the suffix -u which marks verbal nouns, the citation form of verbs in K’abeena. Example: From the Amharic verb kättäbä ’he vaccinated’ the final vowel -ä is replaced by the suffix -u. Therefore, the citation form in K’abeena is kattabu ‘vaccinate’. Other examples are K’AB ’anabbabu ‘read’ from AMH anäbbäbä, K’AB tamaaru ‘lern’ from AMH tämarä, K’AB t’aafu ‘write’ from AMH ts’afä or t’afä. When the second radical of a three-consonantal verb is not geminated in K’abeena, the verb is definitely not borrowed from Amharic, because in this language the second consonant is always geminated in the perfective. The source can only be one of the Gurage languages. However, in these languages, the situation is not homogeneous. Consonant

K’abeena

97

gemination vs. non-gemination is a feature which yields different verb classes. Furthermore, in some cases, languages can be excluded as source, because sound shift took place. An example is the K’abeena verb faradu ‘judge’. In Dobbi, Muher, Mäsqan and Soddo (like in Amharic) the verb is färrädä. In Chaha the cognate verb is fänädä, in Ezha fännädä and in Endegegn, Silt’e, Wolane and Zay färädä (Leslau 1979: 241). Since Wolane is the only contact language of K’abeena, the verb most probably is borrowed from Wolane. The source of verbs of the religious domain is often unclear. They may be borrowed either directly from Arabic or via Amharic or another contact language. An example is saggadu ‘pray’, borrowed from Amharic säggädä or another language with gemination of this verb in the perfective. Here Arabic can be excluded as direct contact language because the verb is realized as sadžada. i.e. without gemination of the second consonant, which in addition is a palatal affricate and not a velar plosive. In other cases the source is not clear. The K’abeena verbs katabu ‘write’ and k’ara’u ‘read’ might be borrowed directly from the Arabic verb kataba und qara’a. The two verbs katabu and k’ara’u are used mainly in religious contexts or by religious people, the verbs t’aafu ‘write’ and ’anabbabu ‘read’, borrowed from Amharic, in nonreligious contexts.

5. Other parts of speech Complementizers are grammaticalized out of a similative marker, which is not considered to be the result of language contact. Primarily they mark complement clauses (example (11)) but they may be used to mark affirmative and negative purpose clauses as well (example (12) and (13)). This feature is found in K’abeena, Libido, Oromo, Amharic, Gumär, Muher, Wolane, and Zay. However, in all languages it is not the main type to mark purpose clauses. (11) moggó-gga híilu ríccu yoo-ba. theft-simil bad thing.nom exist.3-neg ‘There is nothing being as bad as theft.’ (12) shukúru ga’áta ’ameetánu-gga dagáammi. Shukur.nom tomorrow come.ipv.3s.m-simil know.ipv.1s ‘I know that Shukur will come tomorrow.’

98

Joachim Crass

(13) ’áti k’ama’áanti-gga k’aawwáanka-si buuru 2s.nom drink.ipv.2s-simil coffee.loc-3s.m butter.acc wartó-si. put.prv.3s.f-3s.m ‘She put butter into the coffee in order that you may drink it.’ According to Heine and Kuteva (2002: 91) the “directionality proposed here [i.e. the grammaticalization of a complementizer to a marker of purpose clauses, J.C.] has not yet been established beyond reasonable doubt. More data to substantiate this hypothesis are required”. The fact that the grammaticalization of a complementizer to a marker of purpose clauses occurs in the languages of the HEC-Gurage sub-area shoes that this grammaticalization is more frequent than has been considered to be. However, the overall rarity makes it reasonable to consider the occurrence in this area to be due to language contact. The (Ethio-) Semitic noun wäk’t/wak’t ‘time’ is used in K’abeena as the head of a relative clause to form a temporal clause. (14) kesa da’iyoommi wak’ti tassh yiyo-’e. 2s.acc meet.prv.1s.rel time.acc ideo say.prv.3s.m-1s.obj ‘I was happy when I met you.’ The Ethio-Semitic noun məknəyat ‘reason’ is used as the head of a relative clause to form a causal clause. (15) ’ámru dunkiyo-’i mikiññaati daggoom-ba. Amru.nom be.late.prv.3s.m-rel reason.acc know.prv.1s-neg ‘I don’t know why Amru was late.’ The (Ethio-)Semitic noun säbäb/sabab ‘reason’ is used as the head of a relative clause to form a causal clause. The head noun of the relative clause is the subject of a complement clause. (16) c’úulu-’i harafáta ’ameetu’náani child.nom-1s.pss eid.al.adha come.neg.cnv.3s.m fakk’oo sabábati ma ’ikkó-gga stay.away.prv.3s.m.rel reason.nom what.acc be.prv.3s.f-cmpl zaaññiyóommi. not.know.prv.1s

K’abeena

99

‘I don’t know what was the reason why my child did not come to the Eid al-Adha.’ In one case the Amharic denominal derivational suffix -äñña producing adjectives (and nomina agentis) is borrowed in K’abeena. Suffixed to the K’abeena verb ’it-u ‘eat’ it yields ’itañña ‘glutton, stodger’. K’abeena and most of the other languages extensively use ideophones. These ideophones are verbalized by the respective verb ‘say’. Partly, the inventory of ideophones is identical in form and meaning in K’abeena and its contact languages. In the following, I list a few ideophones of K’abeena and Amharic, including the respective citation form of the verb ‘say’ which is in K’abeena the verbal noun yu and in Amharic – according to scientific tradition, not in the language use – the 3s.m of the perfective alä. K’AB ’illimm yu, AMH əlləmm alä ‘disappear suddenly, vanish, disappear from sight’; K’AB bogg yu, AMH bogg alä ‘flare, blaze, appear suddenly (light)’; K’AB sillimm yu, AMH səlləmm alä ‘fall into a swoon, be in a coma’, K’AB will yu, AMH wəll alä ‘desire or crave something (food, drink, smoking), have a momentary strong desire for something’ (the English translations are from Kane 1990).

6. Constituent order SOV word order is an areal feature of the ELA. Most scholars consider Ethiosemitic languages to have shifted their word order because of intensive contact with Cushitic, especially central Cushitic languages.

7. Syntax Another feature found widespread in the languages of the area is the fact that copulas differ in main and subordinate clauses. In K’abeena two types of main clause copulas occur (cf. Crass 2003). One type is gender inflected, namely -ha for masculine and -ta for feminine. The other type of copula, namely -ti, is uninflected. In subordinate clauses (relative clauses, complement clauses, adverbial clauses) the fully inflected verb ’ihu ‘be’ occurs. (17) ’ísu rosisaanco-ha. 3s.m teacher.acc-cop.m ‘He is a teacher.’

100

Joachim Crass

(18) gu’mára halaale ’ikko-’i riccu hasaawwiyye! always true be.prv.3s.m-rel thing.acc talk.imp.p ‘Tell always the truth!’ [Tell always thing, which are true!]

8. Lexicon The ELA is characterized by several interesting types of lexicalizations. Hayward (1991) distinguishes three categories of lexicalizations, which he exemplifies with data on Amharic, Oromo and Gamo, an Omotic language cluster spoken in South West Ethiopia. According to Hayward (1991: 140) these lexicalizations reinforce “the very real cultural unity of Ethiopia” (cf. Hayward 2000). The three categories are (1) single-sense lexicalizations, (2) lexicalizations with two or more distinct senses and (3) lexicalizations involving similar derivations. The first category comprises “single-sense lexicalizations of typically indigenous concepts”, the second category lexicalizations “showing inter-linguistic matching across the three languages” and the third category lexicalizations with a “similar (parallel) ‘derivational pathway’.” To the first category belong mainly nouns such as lexical items for seasons of the year, categories of terrain, categories of dung/excrement, super-categories for birds, types of borrowing and skin colour classification of people of the region. Furthermore, this category includes the suppletive imperative of the verb ‘to come’ (Ferguson’s feature G17) and particles with the meaning ‘Take this!’ which have no obvious etymological relationship to a verb. The second category, lexicalizations with two or more distinct senses, is predominantly comprised of verbs and some nouns. Examples: the respective verbs have the basic meaning ‘hold, catch’ and the secondary meaning ‘start, begin’. The respective verbs with the basic meaning ‘play’ have the secondary meaning ‘chat’. The third category includes verbal derivations, compound verbs, i.e. ideophones verbalized by the respective verb ‘say’, possessive constructions including two NPs, and ideomatic expressions. Examples for verbal derivations are the causative of the respective verb ‘want’ having the meaning ‘need’, the causative of the respective verb ‘enter’ having the meaning ‘marry’ and the causative of the respective verb ‘pass the night’ having the meaning ‘administer’. Compound verbs are ‘become silent’, ‘hurry up’ and ‘jump up suddenly’. Possessive constructions including two NPs have a word by word meaning and a metaphorical meaning. Examples are ‘son of man/people’ having the meaning ‘mankind, human being’ and ‘land of man/people’ with the meaning ‘foreign country’. Idiomatic expressions are ‘regain/recover control,

K’abeena

101

take courage’ being composed of the noun ‘heart’ and the verb ‘return (intr.)’, and ‘catch cold’, of which the noun ‘cold’ is the subject and the experiencer the object of the verb ‘catch’. K’abeena shares at least several of these lexicalizations. I deal only with some lexicalizations of Hayward’s second and third category, which I consider especially interesting. The verb ’afu has the basic meaning ‘hold, catch’ and the secondary meaning ‘begin start’, the verb ’alapp’u ‘play’ the secondary meaning ‘chat’. Hayward’s examples of verbal derivations as part of the third category are attested also in K’abeena. The causative of verb hasu ‘want’ is hasisu ‘need, be necessary’, the causative of the verb ’a’yu ‘enter’ is ’a’isu ‘marry’ and the causative of the verb garu ‘pass the night’ is gasshu ‘administer’. Most of the borrowed nouns are expressions of cultural goods of different kinds (food, cloths, topics related to Islam) and some abstract nouns. Examples for abstract nouns are K’AB suusi from AMH sus ‘mania, passion, rage’ (dt. Sucht) K’AB ’umuri from ARAB cum(u)r ‘age’, K’AB keerti / hayraati from ARAB khayr (engl. ‘good, benefit’), K’AB haali from ARAB ḥaal ‘situation’ (Crass 2005: 5260). Examples for nouns of mainly typical muslim cultural heritage: K’AB maskiida ‘mosque’ from AMH mäsgid, K’AB sheet’aani ‘devil’ from AMH säyt’an or ARAB shayt’an, K’AB t’uuri ‘punishment by God which befalls a wrongdoer’ from AMH t’ur, K’AB kitaaba ‘book’ from ARAB kitaab. Examples for vegetation are K’AB ababa ‘flower’ from AMH abäba, K’AB baarzaafi ‘eucalyptus tree’ from AMH bahər zaf (Remark: AMH bahr zaf literally means ‘tree of the sea/ocean’. The lexeme bahər in compounds often designates foreign origin of an item (Kane 1990: 855). An example for geography is K’AB alamíta ‘world’ from AMH aläm. Examples for tools are K’AB akaafa ‘shovel’ from AMH akafa, K’AB billaawa ‘(kind of) knife’ from AMH billawa, K’AB faasa ‘axe’ from AMH fas. An example for minerals is K’AB work’a ‘gold’ from AMH wärk’. Examples for body parts (and related abstract nouns) are K’AB angooli ‘brain’ from AMH ang(w)äli [ango:li], K’AB k’albi ‘heart, mind, intelligence, reason’ from AMH k’älb, K’AB nafséeta/nabséeta/nafsíta/nabsíta (-éeta and -íta are different flexion classes of the noun) ‘soul, life’ from AMH näfs, K’AB sabri ‘patience’ from ARAB sabr. An example for cultural goods of different kinds is K’AB birati ‘metal’ from AMH bərät. Furthermore, there are contact phenomena in areal cultural vocabulary, especially in relation with the Ensete plant (cf. Crass and Meyer 2005). Other word classes are interjections, e.g. K’AB ciff, AMH cəff, an interjection to chase away cats and fillers, e.g. K’AB bal, AMH bäl ‘well’. In the case

102

Joachim Crass

of interjections to call or chase away animals a remarkable correspondence is attested in many languages of the area. An interesting case is the K’abeena abstract noun ma-ricc-oomáta ‘essence, nature’. It has the following structure: The question word ma ‘what’ is followed by the noun ricc-u ‘thing’ yielding “what-thing”. To this the abstract noun suffix -oomáta is added. In Amharic, the structure of the respective abstract noun mənənnät is similar. Here the question word mən ‘what’ is derived by a suffix for abstract nouns, namely -(ə)nnät. Greetings are expressed by identical patterns, namely by questions. In order to greet somebody in the morning, one says in K’abeena garee gal-ti ‘Did you pass the night well?’, i.e. garee ‘well’ precedes the verb gal-u ‘pass the night’, inflected for the second-person singular. The equivalent expression in Amharic is dähna addär-k, i.e. dähna ‘well’ precedes addär-ä ‘he passed the night’, in this example inflected for the second-person singular masculine. In the evening one asks in K’abeena garee hos-si ‘Did you pass the day well?’. The equivalent in Amharic is dähna wal-k.

9. Conclusion K’abeena shares a lot of phonological, grammatical and lexical features with other languages spoken in the highlands of Ethiopia. This fact allows it to refer to this area as the ELA. The core area comprises the languages of the highlands of Ethiopia. The more distant a given language is situated from this core area the fewer features it has. However, the areal status of individual features is not accepted generally (for the discussion cf. Crass 2002, Crass and Bisang 2004, Tosco 2000, Zaborski 1991). Furthermore, the existence of the ELA is denied by Tosco (2000). However, since only a relatively small number of languages are described adequately, these findings must be considered preliminary.

Abbreviations AMH acc abl ARAB cmpl

amharic accusative ablative Arabic complementizer

cnd cnv cop ELA f

conditional converb copula Ethiopian linguistic area feminine

K’abeena gen HEC ideo imp ipv K’AB loc m neg nom obj p

genitive Highland East Cushitic ideophone imperative imperfective K’abeena locative masculine negative nominative object agreement plural

prv pss pst purp q rel s simil vn 1 2 3

103

perfective possessive past purpose question relative clause singular similative verbal noun first person second person third person

Note 1. This research was undertaken in the scope of the Collaborative Research Center 295 Cultural and linguistic contacts: Processes of change in North Eastern Africa and West Asia (Sonderforschungsbereich 295 Kulturelle und sprachliche Kontakte: Prozesse des Wandels in historischen Spannungsfeldern Nordostafrikas/Westasiens).

References Bisang, Walter 1996 Areal typology and grammaticalization: Processes of grammaticalization based on nouns and verbs in east and mainland south east Asian languages. Studies in Language 20 (3): 519597. 2006 Linguistic areas, language contact and typology: Some implications from the case of Ethiopia as a linguistic area. In: Yaron Matras, April McMahon and Nigel Vincent (eds.), Linguistic areas: Convergence in historical and typological convergence, 7598. Houndmills: Palgrave Macmillan. Braukämper, Ulrich 1980 Geschichte der Hadiya Süd-Äthiopiens (Studien zur Kulturkunde 50). Wiesbaden: Steiner. Crass, Joachim 2002 Ejectives and pharyngeal fricatives: Two features of the Ethiopian language area. In: Baye Yimam, Richard Pankhurst, David Chapple, Yonas Admasu, Alula Pankhurst, and Birhanu Teferra (eds.), Ethiopian studies at the end of the second millennium. Proceedings of the 14th

104

Joachim Crass

International Conference of Ethiopian Studies, 611 Nov. 2000, 1679– 1691. Addis Ababa. Addis Ababa: Institute of Ethiopian Studies. 2003 The copulas of K’abeena: Form, function, and origin. Afrika und Übersee 86: 2342. 2005 Das K’abeena. Deskriptive Grammatik einer hochlandostkuschitischen Sprache (Cushitic Language Studies 23). Köln: Köppe. Crass, Joachim, and Walter Bisang 2004 Einige Bemerkungen zum äthiopischen Sprachbund und ihre Relevanz für die Areallinguistik. In: Walter Bisang, Thomas Bierschenk, Detlev Kreikenbom, and Ursula Verhoeven (eds.), Kultur, Sprache, Kontakt, 169198. Würzburg: Ergon. Crass, Joachim, and Ronny Meyer 2005 Die Komplexität kultureller und sprachlicher Kontakte am Beispiel der Nomenklatur zur Ensete-Pflanze. In: Walter Bisang, Detlev Kreikenbom, and Ursula Verhoeven (eds.) Prozesse des Wandels in historischen Spannungsfeldern Nordostafrikas/Westasiens. Akten zum 2. Symposium des SFB 295, 15.10.–17.10.2001, 411427. Würzburg: Ergon. 2007 Ethiopia. In: Bernd Heine and Derek Nurse (eds.), A Linguistic Geography of Africa. Cambridge: Cambridge University Press. 228249. Faber, Alice 1997 The genetic subgrouping of the Semitic languages. In: Robert Hetzron (ed.), The Semitic Languages. 315. London, New York: Routledge. Fleischman, Suzanne 1989 Temporal distance: A basic linguistic metaphor. Studies in Language 13: 150. Haspelmath, Martin 1997 From Space to Time: Temporal Adverbials in the World’s Languages. München, Newcastle: Lincom. Hayward, Richard 1991 À propos patterns of lexicalization in the Ethiopian language area. In: D. Mendel and U. Claudi (eds.), Ägypten im afro-asiatischen Kontext. Aufsätze zur Archäologie, Geschichte und Sprache eines unbegrenzten Raumes. Gedenkschrift Peter Behrens, 139156. (Afrikanistische Arbeitspapiere. Sondernummer 1991). Köln: Institut für Afrikanistik. 2000 Is there a metric for convergence? In: Colin Renfrew, April McMahon, and Larry Trask (eds.), Time Depth in Historical Linguistics Vol.1, 621640. Cambridge: McDonald Institute for Archeological Research. Heine, Bernd 1994 Areal influence on grammaticalization. In: Martin Pütz (ed.), Language contact and language conflict, 5568. Amsterdam/Philadelphia: John Benjamins.

K’abeena

105

Heine, Bernd, and Tania Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. 2003 On contact-induced grammaticalization. Studies in Language 27 (3): 529572. Kane, Thomas Leiper 1990 Amharic-English dictionary. Wiesbaden: Harrassowitz. Leslau, Wolf 1979 Etymological Dictionary of Gurage (Ethiopic). Volume III. Etymological Section. Wiesbaden: Harrassowitz. Meyer, Ronny, and Renate Richter 2003 Language Use in Ethiopia from a Network Perspective (Research in African Studies 7). Frankfurt: Lang. Tosco, Mauro 2000 Is there an “Ethiopian language area”? Anthropological Linguistics 42 (3): 329365. Zaborski, Andrzej 1991 Ethiopian language subareas. In: Stanislaw Pilaszewicz and Eugeniusz Rzewuski (eds.), Unwritten Testimonies of the African Past. Proceedings of the International Symposium Held in Ojrzanów N. Warsaw on 78 November 1989, 123134 (Orientalia Varsoviensia 2). Warsaw: Wydawnictwa Universytetu Warszawskiego.

Grammatical borrowing in Likpe (Sɛkpɛlé) Felix K. Ameka

1. Background Sɛkpɛlé is the auto-denomination of the language spoken in the area known as Likpe which is to the east and north-east of Hohoe (the district capital and an Ewe (Gbe) speaking town) as far as the Togo border in the northern part of the Volta Region of Ghana (see Map 1). There are approximately 23,000 residents in the area who speak the language (District Assembly Representative 1998 figures). A small percentage, living in the two modern migrant villages, speaks the language as a second language. If we add the other native speakers in the diaspora, there may well be more than 30,000 speakers of the language today. Sɛkpɛlé is one of the fourteen languages most recently characterized as Ghana–Togo–Mountain (GTM) languages (Ring 1995) that were first recognized as a group and referred to as Togorestsprachen by Struck (1912) and, in English, as “Togo Remnant languages”,1 for instance by Westermann and Bryan (1952: 96). Their genetic classification and cultural history have remained controversial (see e.g. Nugent 1997, 2005). The distinctive typological features of these languages that separate them from the surrounding languages such as Ewe and Akan include (i) the active noun class system similar to that of the Bantu languages which they have inherited, undoubtedly, from Proto-Niger-Congo, (ii) head marking at the clause level through subject cross-referencing on the verb, and (iii) their highly agglutinative nature especially their derivational verb morphology. Some scholars wonder whether the group of languages is not a socio-cultural or a typological grouping masquerading as a genetic unit (Blench 2001, Maddieson 1998). Nevertheless, they are classified as belonging to Kwa (Niger-Congo). It is difficult, however, to establish the Ghana-Togo-Mountain (GTM) languages as a group in relation to Kwa. Heine (1968) subclassified the GTM languages into Ka-Togo and Na-Togo subgroups. The current view is that the two subgroups branch out individually from Proto-Kwa as in Figure 1 (Williamson and Blench 2000). Sɛkpɛlé is a Na-Togo language and has two major dialect divisions, namely, Sɛkpɛlé and Sekwa.2 Table 1 shows the dialects and the villages where they are spoken.

108

Felix K. Ameka Ega Avikam Alladian Ajukru Abidji Abbey Attié Ebrie

Potou

Mbatto Krobu Potou-Tano

Abure, Eotilé

West Tano

Akan Nzema-Ahanta

Tano Central Tano

Bia

Anyi, Baule, Anufɔ Efutu-Awutu

Guan

South

Larteh-Cherepong-Anum Northern Guang Ga

Proto-Kwa

Dangme Lelemi-Lefana Akpafu-Lolobi Lipke, Santrokofi Na-Togo

Logba Basila, Adele Avatime Nyangbo-Tafi

Ka-Togo

Kposo, Ahlo, Bowiri Kebu, Animere Ewe

Gbe

Gen, Aja Fon-Phla-Phera

Figure 1. Proto-Kwa (from Williamson and Blench 2000: 29)

A large part of the Likpe-speaking population is multilingual in Sɛkpɛlé, Ewe – the dominant lingua franca – and Akan, not to mention (Ghanaian) English (cf. Ring 1981). Ewe, the dominant lingua franca, co-exists with Likpe in all spheres of life in the home, the community, in the church, in the market,

Likpe (Sɛkpɛlé)

109

Table 1. Sɛkpɛlé dialects and their geographical distribution Language

Sɛkpɛlé

Dialects

Sekwá

Sub-dialects Villages

Sɛkpɛlé L2 communities Situnkpa

Bakwa Todome Nkwanta

Alavanyo Wudome

Semate

Avedzime Mate Agbozume Abrani Koforidua

Sela Bala Kukurantumi

© Likpe Local Committee

Map 1. Likpe Traditional area

110

Felix K. Ameka

at school and other public ceremonies such as funerals and marriages. Akan is less prominent, although it is the lingua franca in a neighboring community. Likpe was first written in about 1933 using the Ewe orthography (see Dogli 1933). The next time there was writing was with the new wave of literacy and language development under the auspices of the Ghana Institute of Linguistics, Literacy and Bible Translation (GILLBT) in 1987. Today there are pamphlets of stories, proverbs and literacy materials as well as a pamphlet containing a translation of the letters of Paul (New Testament). There are also cassettes containing some drama and text in Likpe. Likpe is used partially as medium of instruction and is taught outside the regular curriculum in some villages after school hours to both Primary and Junior Secondary School pupils. The focus of the present chapter is on the grammatical changes in Sɛkpɛlé due to the influence of Ewe, the dominant lingua franca, and areal convergence. That is, the impact of the neighbouring languages on Likpe grammar. Likpe seems to have innovated constructions such as the present progressive and an Undergoer Voice construction due to Ewe influence. At the same time there are several patterns that are found in Likpe due to pressures of areal adaptation, for instance, postpositions and arguably, serial verb constructions have emerged in the language through this mechanism. There are also grammatical items such as intensifiers and particles and interjections which are also shared among the neighbouring languages, but Likpe has borrowed some connectives and a complementizer directly from Ewe (see also Ameka 2007).

2. Typology Likpe is an agglutinative language with some head marking at the clause level and dependent marking in the NP (except for qualifiers), properties presumably inherited from Proto Niger-Congo. Features of nouns, including class and number, are marked by prefixes on nouns. Ewe, on the other hand, is predominantly isolating with agglutinative features. Plural marking in Ewe is by an enclitic which attaches to the element in the NP that occurs before the Intensifier. Likpe marks plural number for a small set of nouns by a suffix. The use of such a structure in Sɛkpɛlé is due to its contact with Ewe (see Section 3). The functional load of the morphological process of Reduplication seems to have increased in Likpe due to its contact with Ewe. Some adjectival modifiers in Sɛkpɛlé are formed by verbal reduplication similar to the pattern found in Ewe. For instance,

Likpe (Sɛkpɛlé)

(1)

111

a. Reduplication of a property verb to form a qualifier ná ‘become.black, dirty’ ná-ná ‘dirty’ bù ‘become.wet, rotten’ bùbù ‘wet, rotten’ b. Reduplication of action verbs to form result-state qualifiers là ‘cut’ là-là ‘torn’ f ‘to arrive newly’ f-f ‘new’

Gerund formation of verb and complement NP structures involves the reduplication of the verb with the NP preposed to it. Such a process is available in Ewe and could have influenced the process. In fact it appears that the gerund involving reduplication is in competition with the ‘older’ form of gerund formation involving the use of the class marker for deverbal nominals. Compare both types of gerund formation shown in (2): (2)

a. bi-sí-tk-tk Compare Ewe te-ʃa-ʃa cm-yam-red-be.on yam-red-plant ‘yam planting’ ‘yam planting’ b. bi-sí bu-tkə cm-yam cm-be.on ‘yam planting’

The gerunds formed by reduplicating the verb part seem to be entering the language through translation, especially of texts mediated through Ewe.

3. Nominal structures Likpe noun stems tend to participate in sg/pl class pairing system for number marking. Some kin terms do not have plural counterpart classes. Kin terms belonging to ego’s parents’ generation and above and proper names are suffixed with the form -m ‘pl’ to signal their plurality. This form is heterosemically related to the third-person plural pronoun. (3)

a. Kofi kú Áma-mə´ ə-diə. name com name-pl scr-quarrel ‘Kofi and Ama and co quarrelled.’ b. ambe ‘mother’ ambe-mə´ ‘mother-pl’ éwú ‘grandmother’ éwú-mə´ ‘grandmother-pl’

112

Felix K. Ameka

This Likpe structure is a replication of the Ewe pattern where plural number is marked by a clitic =wó ‘pl’, which is attached to the last element in the NP before the intensifier, and which is also in a heterosemic relation to the ‘3pl’ pronominal form wó. The use of the Ewe form as an associative plural, that is, N=wó means ‘N and co’, e.g. Áma=wó [Ama=pl] ‘Ama and co’, could have served as the model for the copy.

4. Verbal structure Likpe, like its closest genetic relatives, marks tense, aspect and mood categories by prefixes on the verb (including for example past progressive). However, it has developed a present progressive periphrastic construction similar to the one found in Ewe which has the form: Subject -l ‘hold’ (NP) Gerund. (4a) below is a clause in the aorist with no segmental marker for the category, while (4b) is an instantiation of the progressive construction in relation to the state of affairs represented also in (4a) (see Ameka 2002). (4)

a. o-té ka-m. 3sg-sell cm-rice ‘She sold rice.’ b. ɔ-l ka-m bo-té. 3sg-hold cm-rice cm-sell ‘She is selling rice.’

Compare the Ewe equivalent of (4b) in (4c) in terms of the structure and order of the elements. The phonetic similarity between the operator verb in the two languages could have facilitated the adaptation of the structure into Likpe. (4)

c. Ewe é-le mlu dzráx-ḿ. 3sg-be.at:pres rice sell-prog ‘She is selling rice.’

An operator for the expression of necessity has also been adopted into the language. This may be due to areal rather than Ewe specific influence. Moreover, the construction is one of the structures in which Likpe uses a complementizer that is borrowed from Ewe, namely b ‘quot’, as illustrated in (5).

Likpe (Sɛkpɛlé)

(5)

113

é-hiɛ˜´ bə´ u-tsyi wə ú-su u-bíkə. impers-need quot 3sg-carry 3sg 3sg-go 3sg-bury ‘It was necessary that he (Skunk) should take her (his mother) to go and bury.’

There is no special marking of verbs that are borrowed. They are only phonologically adapted; for example, the verb form that is used in example (5) above has the form hia ‘need’ in both Ewe and Akan, but is adapted in Sɛkpɛlé as hiɛ ‘need’. Likpe uses several patterns for marking voice/valency that conform to areal patterns: the reflexive is formed by the use of a pronoun plus a grammaticalized form of the word for ‘body’, i.e. əsúə (similar to the structure one gets in Akan, but not in Ewe). The use of 3pl to express general subjects and impersonal passive meanings is also an areal pattern. There is an Undergoer Voice construction expressed periphrastically where the Undergoer-like argument is linked to the subject position of the clause and of an operator verb nɔ ‘hear’.3 The operator verb takes a nominalized verb complement with the Actor-like argument, if expressed, functioning as its first object (and like the Goal argument in a double object construction). The semantics of the construction belongs to the semantics space of the so-called ‘potential passives’. There is a similar periphrastic construction in Ewe but its operator verb is a modal verb nyá, grammaticalized from the verb ‘know’. The Actor-like argument in the main event, if expressed, is marked as a dative (experiential) object in the Ewe construction, as illustrated in (6). (6)

Ewe nynu-a nyá kp-ná (ná-m). woman-def mod see-hab dat-1sg Lit: ‘The woman is see-able (to me).’ i.e. ‘The woman is beautiful (to me).’

(7)

Likpe u-sió -m á-nɔ (mɛ) bó-be. cm-woman agr-det scr-hear 1sg cm-look Lit: ‘The woman hears (me) looking.’ i.e. ‘The woman is beautiful (to me).’

114

Felix K. Ameka

(8)

Likpe n-t á-nɔ bú-nə í-tə´ be-tsyúé. cm-alcohol scr-hear cm-drink 3sg:impers-give cm-some Lit: ‘Alcohol hears drinking give some.’ i.e. ‘Alcohol-drinking is enjoyable to some (people)’, ‘Some people like drinking alcohol.’

One argument in support of the Ewe construction potentially influencing the Likpe construction comes from the periphrastic nature of the construction in Likpe. One would have expected an affixal marking of such a meaning on the verb in Sɛkpɛlé. Secondly, if the Actor is expressed it takes the form of a Goal argument similar to the dative marking in Ewe. Thirdly, there is a variation on the expression of the Actor-like argument in the Sɛkpɛlé construction which replicates the Ewe pattern, as illustrated in (8) (see Ameka 2005a for further details on the construction in both Ewe and Likpe).

5. Other parts of speech Likpe has borrowed a few grammatical items from Ewe and also makes use of several forms that are found throughout the Lower Volta Basin area, i.e., the area in which Kwa languages are spoken. It has borrowed the contrast connector gaké ‘but’ from Ewe and adapted it as appropriate as kaké ‘but’ in the Sɛkpɛlé dialect since unlike Sekwa, there are no [−anterior] [+voice] consonants in the Sɛkpɛlé dialect. Languages in the area tend to have two or three disjunction markers, one of which tends to be used in interrogative contexts. Likpe has two disjunction markers: nyé ‘this or that, it does not matter which’ and lee ‘this or that, I don’t know which’. It appears the latter form is influenced by one of the disjunction markers in Ewe, the form lóó ‘this or that, I don’t know which’. Because of their ignorative feature they tend to be used in interrogative contexts and can link phrases or clauses. A hint that this may be the case is that Tuwuli, another GTM language, has a cognate disjunctive marker nye ‘or’ used in general contexts while in interrogative contexts the other form mbɔe is used (Harley 2005). Likpe seems thus to have adapted the Ewe form for use in interrogative contexts. Likpe has also borrowed a complementizer b ‘comp, quot’ from Ewe which it uses in addition to its own complementizer ŋkə ‘quot’. Sometimes the Ewe and the Likpe complementizers are doubled. Compare the use of

Likpe (Sɛkpɛlé)

115

both forms in similar contexts in (9) and (10) both taken from a Likpe settlement history narrative. (9) sé ɔfu kɔdzó -m le-te ŋkə məə-tsyá when name name agr-det dep-know quot 3pl-too a-slé eto be-tídi bé-ni ko ŋkə oo, atúu … cm-church poss cmpl-person 3pl-cop int quot interj welcome ‘When Ofu Kwadzo got to know that they too were church people, he said oo welcome, (he and them will work together).’ (10) nyã bəə-b bə-tu m nyã bə´ə oo ka-sɔ kpé. and 3pl-come 3pl-meet 3pl and quot interj cm-land be.in ‘And they came to meet them and they said “oh there is land”.’ While the relativizer itself is not borrowed, Likpe seems to be developing the use of the definiteness determiner as an optional relative-clause-final marker. This is an areal pattern. The interesting thing is that in the other languages like Ewe, Akan, Ga etc. the same form that is used in this relative clause context is used to mark all other background information constituents such as left-dislocated topic NPs and preposed adverbial clauses such as conditionals and temporals. In Likpe, as we shall see below, there are different forms for marking these, one of which is a borrowed form from Ewe. (11) kɔsídá kó l-yɛ (mə´) week agr dep-pass det ‘the week which passed, the past week’ In addition, two adverbial connectors are borrowed from Ewe: álé bé ‘so that’ (example 12) and tógb bé ‘although’. (12) álé-bé ŋko ni kasé min-yi ba-kpɛlé eto akokosa n. thus-quot this cop how 1sg-know cmpl-Likpe poss history emph ‘So this is how I know the history of Likpe (to be).’ Various intensifying words (focus particles) found in Likpe are forms that have diffused through the West African littoral area. However, the Likpe forms resemble more closely the forms used in Ewe than the others. Thus words such as ko ‘only, just’, boŋ ‘rather’, tsyá ‘also’ are shared with Ewe. Similarly, interjections such as ah㈠‘now I know’ and fillers like oo, an

116

Felix K. Ameka

utterance-initial vocative particle, o …, answer particles like ee ‘yes’, o ‘no’; agreement-signalling particles like yoo ‘OK’ and the palatal click with nasal release, both as an agreement marker and signalling continuation in the sense of ‘I understand’, are all shared by various languages in the area. Additionally utterance final particles for expressing attitudinal meanings; e.g. ló ‘I advise you’, are widespread in the area. Some of these forms give evidence of convergence among the languages in the area and their occurrence in Likpe can be attributed to areal adaptation. A form ma- used for ‘privative’ derivations in Ewe has occurred in some derived words in Likpe (see example 13). But it is unclear whether it is copied from Ewe or it is retained from some ancestor language since other GTM languages, e.g. Tuwuli (Harley 2005) seem to have similar forms. (But of course they could also have taken it from Ewe.) (13) a. Likpe lə-fə n bə-n-sí ko kasɔ´-ma-nɔ-ma-nɔ cm-time agr 3pl:hab-lig-sit int under-priv-hear-priv-hear ə-b-lu-f m bə-tsyú l ntí. scr-vent-leave-dir 3pl cmpl-neighbour loc midst ‘The time they stayed there then misunderstanding emerged among their neighbours.’ b. Ewe nú-gɔme-ma-se-ma-se thing-under-priv-hear-priv-hear ‘misunderstanding’ Likpe has also adopted the background information marking particle lá ‘tp’ from Ewe (although it is an item that is consciously recognized and edited out of texts). (14) kasé mi-nɔ nyã ní b bó ba-kpɛlé lá how 1sg-hear 3sg cop quot 1pl cmpl-Likpe tp bo kə-síə-kɔ fefe ka búu-siə ní atébubu. 1pl cm-sit-place last agr 1pl:past-sit cop name ‘How I heard it is that we the Likpe people, our last place of settlement where we stayed was Atebubu.’ The background information particle in Ewe is in a heterosemic relation to the definiteness marker and is used at the end of left dislocated NPs, connectors as well as preposed adverbial phrases and clauses and relative clauses. In

Likpe (Sɛkpɛlé)

117

Likpe, the Ewe form la´ occurs in all these contexts. The Likpe indigenous form of marking background information in these contexts is to lengthen the phrase-final vowel.

7. Syntax One of the features of Likpe grammar that could have emerged due to areal pressure from the surrounding languages like Ewe and Akan is verb serialization in a single clause. Dimmendaal (2001: 386) claims that the spread of serial verb constructions (SVC) to the GTM languages could account for the reduction in verb derivational morphology that is in progress in these languages (cf. Hyman 2004). Perhaps an indication that this may be so is that the features of SVCs in Sɛkpɛlé share some features with Akan SVCs and other features with Ewe SVCs, and there are other features that are common to all three languages. For instance, the shared subject argument is expressed with each verb in the SVC in both Likpe and Akan, but expressed only once in an Ewe SVC. Negation, on the other hand, is expressed only once in an SVC in both Likpe and Ewe but recapitulated with each verb in Akan. For all the languages the verbs in an SVC should be marked for aspect and modality values that are semantically compatible (see Ameka 2005b). Likpe uses locative verbs in the expression of predicative possession. The verbs are kpé ‘be.in’, t ‘be.at’ and tk ‘be.on’. These are used to express both general location and ‘have’ possession (see Ameka 2007b for a discussion of their use and semantics). With each of these verbs there are two constituent orders for the possessive use: one in which the Possessor is linked to the Subject position and the possessed phrase linked to the object function. The second order involving Figure–Ground reversal has the possessed linked to the subject position and the possessor to the post-verbal object position. The two orders are illustrated for the verb kpé ‘be.in’ below. (15) a. Possessor–verb–possessed o-kpé a-fokpá 3sg-be.in cmpl-footwear ‘He has shoes.’ b. Possessed–verb–possessor a-fokpá kpé wə cmpl-footwear be.in 3sg ‘He has shoes.’

118

Felix K. Ameka

It is possible that the second order in (15b) is due to Ewe influence. Ewe uses a general locative verb le ‘be.at:pres’ in a periphrastic construction to express predicative possession: ‘have’. The structure has the form possessed–verb– possessor NP + possessive postposition si ‘hand’. This Ewe structure is the only order possible for expressing predicative possession in Ewe using the locative schema.

8. Lexicon A number of constructions that one finds in Likpe grammar are based on semantic formulas that are available in other languages in the area. For instance, a verb–noun collocation which literally translates as ‘see/look way/road’ is used to express the idea of ‘hope’ as illustrated for Likpe in (16). (16) ó-be ku-sú ŋkəə mba uuka-wuuns-ko é-bu-b 3sg-look cm-way quot those 3sg:phab-help-assoc 3pl:fut-come ba-wuuns-ko w u-bik wo ambé -m. 3pl-help-assoc 3sg 3sg-bury 3sg mother agr-det ‘He hoped that those who he used to help would come to help him bury his mother.’ Similarly various verb–verb collocations in serial verb constructions form semantic formulas for the expression of particular meanings in the serializing languages. One such formula is for the expression of the sense of ‘believe’. It is composed of two verbs the first of which is invariably a ‘receive/get’ verb and the second an ingestion or imbibing verb such as ‘eat’ or ‘hear’. Likpe makes use of the ‘receive hear’ pattern which is the same pattern that we find in Ewe. The Likpe form is exemplified in (17). (17) n-fo n-nɔ míə yɔɔ-lkɛ. 1sg-get 1sg-hear 1sg:quot 3sg:fut-be.good ‘I believe it will be good.’ A formula for expressing the notion of ‘begin’ relates to making contact with the bottom of the situation that is began. Such expressions occur in several languages in the area.

Likpe (Sɛkpɛlé)

119

9. Conclusion Various changes have taken place and continue to take place in Likpe grammar due to its contact with Ewe on the one hand, and due to pressures of areal adaptation, on the other. The present progressive construction, plural marking on kinship terms, Undergoer Voice construction, and constituent order for predicative possession are constructional patterns that have been directly replicated from Ewe. Patterns involving serial verb constructions and postpositions, intensifiers and various semantic formulas occur in Likpe due to their presence in the convergence area. Furthermore, Likpe has borrowed lexical as well as grammatical items from Ewe replacing indigenous terms in some cases. It is significant that the grammatical items that have been borrowed have discourse structuring or organizational functions such as a contrast marking conjunction, a disjunction marker, and reason expressing connectives. Nouns and verbs are also borrowed. Borrowed verbs do not receive any special treatment. Nouns borrowed into Likpe are integrated on the basis of form and meaning into the noun class system. Thus a noun like agbeli ‘cassava’ which in Ewe is made up of a prefix a- and the stem and which is singular or collective is borrowed into Likpe and analysed as a plural noun to fit the a- plural class and a singular form is formed as le-gbeli ‘one tuber of cassava’. It is in this domain that the borrowing of nouns has an effect on the grammar of Sɛkpɛlé.

Abbreviations agr assoc cm com comp cop dat dem det dep dir emph

agreement marker associative verb extension noun class marker comitative complementizer copula dative (preposition) demonstrative determiner pragmatically dependent subject cross-reference directional emphatic particle

fut impers int interj lig loc neg past phab pl poss pot pres

Future impersonal intensifier interjection ligature locative negative marker past tense past habitual plural possessive marker potential present

120

Felix K. Ameka

priv prog q red

privative progressive propositional question reduplicative

scr subject cross-reference marker sg singular tp background topic marker

Notes 1. They have also been referred to as “Central Togo” (Dakubu and Ford 1988). 2. The findings reported here are based on my own field investigations that I have been conducting in various Likpe communities intermittently over the past decade or so. I am very grateful to my consultants, especially E. K. Okyerefo, Justina Owusu, Cephas Somevi, Comfort Atsyor and the late A. K. Avadu, for patiently teachinjg me their language. 3. I use the terms Actor and Undergoer following their usage in Role and Reference Grammar (RRG) see Van Valin and La Polla (1997: 141147). To characterize the objects in a double object construction, I use the terms “Theme” and “Goal”. The latter is a cover role for the object with the semantic relation of recipient, beneficiary, maleficiary etc. or the dative more generally.

References Ameka, Felix K. 2002 The progressive aspect in Likpe: Implications for aspect and word order in Kwa. In: Felix K. Ameka and E. Kweku Osam (eds.), New directions in Ghanaian Linguistics, 85111. Accra: Black Mask 2005a “The woman is seeable” and “The woman perceives seeing”: Undergoer voice constructions in Ewe and Likpe. In: M. E. Kropp Dakubu and E. Kweku Osam (eds.), Studies in the Languages of the Volta Basin, Volume 3, 4362. Legon: Department of Linguistics, University of Ghana. 2005b Multiverb constructions on the West African littoral: Microvariation and areal typology. In: M. Vulchanova and T. A. Åfarli (eds.), Grammar and Beyond: Essays in Honour of Lars Hellan, 1542. Oslo: Novus Press. 2007a Grammars in contact in the Volta Basin (West Africa): On contact induced grammatical change in Likpe. In: A. Y. Aikhenvald and R. M. W. Dixon (eds) Grammars in Contact: A Cross-Linguistic Typology, 114– 142. Oxford: Oxford University Press.

Likpe (Sɛkpɛlé) 2007b

121

The coding of topological relations in verbs: The case of Likpe (Sɛkpɛlé). In: Felix K. Ameka and Stephen C. Levinson (eds.), The Typology and Semantics of Locative Predication: Posturals, Positionals and Other Beasts, 10651103 (Linguistics 45 (5/6), special issue). Blench, Roger M. 2001 Comparative Central Togo: What have we learnt since Heine? Paper presented at the 32nd Annual Conference on African Linguistics: Berkeley, 2325 March 2001. Dakubu, M. E. Kropp, and Kevin C. Ford 1988 The Central-Togo languages. In M. E. Kropp Dakubu (ed.), The languages of Ghana, 119154. London: Kegan Paul. Dimmendaal, Gerrit J. 2001 Areal diffusion and genetic inheritance: An African perspective. In: A. Y. Aikhenvald and R. M. W. Dixon (eds.), Areal diffusion and genetic inheritance, 358392. Oxford: Oxford University Press. Dogli, A. (Rev.) 1933 Likpe Catechism No 1 and 2. Keta: Catholic Printing Office Heine, Berndt 1968 Die Verbreitung und Gliederung der Togorestsprachen. Berlin: Dietrich Reimer. Harley, Matthew 2005 A descriptive grammar of Tuwuli: A Kwa language of Ghana. Ph.D. thesis, SOAS. Hyman, Larry 2004 How to become a Kwa verb. Journal of West African Languages 30 (2): 6988. Maddieson, Ian 1998 Collapsong vowel harmony and doubly articulated fricatives: Two myths about the phonology of Avatime. In: Ian Maddieson and Thomas Hinnebusch (eds.), Language History and Linguistic Description in Africa, 155166. Trenton: Africa World Press. Nugent, Paul 1997 Myths of Origin and Origins of Myth: Politics and the Uses of History in Ghana’s Volta Region. Berlin: Das Arabisch Buch. 2005 A regional melting pot: The Ewe and their neighbours in the GhanaTogo borderlands. In: Benjamin Lawrence (ed.), The Ewe of Togo and Benin, 2945. Accra: Woeli. Ring, Andrew J. 1981 Ewe as a Second Language: A Sociolinguistic Survey of Ghana’s Central Volta Region. Legon: Institute of African Studies. 1995 Lɛlɛmi tone. Papers from GILLBT’s seminar week 30 January–3 February 1995, Tamale, 1995: 1626.Tamale: GILLBT Press.

122

Felix K. Ameka

Struck, R. 1912

Einige Sudan-Wortstämme. Zeitschrift für Kolonialsprachen 2: 233– 253, 309323. Van Valin, Robert D. Jr, and Randy J. La Polla 1997 Syntax: Structure, Meaning and Function. Cambridge: Cambridge University Press Westermann, Dietrich, and Margaret A. Bryan 1952 Languages of West Africa. Handbook of African Languages, Volume 2. London: Oxford University Press, for the International African Institute. Williamson, Kay, and Roger Blench 2000 Niger-Congo. In: Bernd Heine and Derek Nurse (eds.), African Languages: An Introduction, 1142. Cambridge: Cambridge University Press.

Grammatical borrowing in Katanga Swahili Vincent A. de Rooij

1. Background1 Katanga Swahili, also known as Shaba Swahili, is a contact variety of Swahili spoken in the urban centers of Southern Katanga, dr Congo. Katanga Swahili resulted out of contacts between speakers of closely related Bantu languages.2 Despite the strong structural similarities of these languages, the phonology, tma system, and morpho-syntax of Katanga Swahili have been restructured considerably. Elsewhere (de Rooij 1997), I argued that, apart from adult second language learning, nativization, i.e. first language acquisition by locally born children, may have played a role in the genesis and development of Katanga Swahili. I also argued that selective simplification (Kapanga 1993) cannot have been the sole restructuring process since even features which are characteristic of both the ‘lexifier’ language, the East African Coast variety of Swahili, and the genetically related ad/substrate languages, such as agglutinative verbal morphology and the noun-class agreement system, have been lost to varying degrees in Katanga Swahili. Katanga Swahili is widely used as a first language in Southern Katanga and is spoken by an estimated number of at least 2 million people.3 In the cities of the Southern Katanga Copperbelt, multi-lingualism is widespread. Apart from Katanga Swahili, many speak French, Congo’s official language, and some ‘ethnic’ language (Kabamba 1979). Katanga Swahili is used in all informal settings: it is used in the domestic sphere but also in informal public settings (public transport, markets, shops). Informal notes and letters are often written in Swahili but since Katanga Swahili has no standard orthography, writing is often done in idiosyncratic ways (cf. Blommaert 1999, 2004; Fabian 1990). Books and newspapers in Swahili are widely available but the variety used in these publications is very similar to the Standard Swahili of Tanzania and is almost like a foreign language to speakers of Katanga Swahili. According to Fabian (1986) Katanga Swahili was well established as a language distinct from the East African variety of Swahili by 1940. Katanga Swahili in its present form, however, cannot be taken as representing the language as spoken around 1940, although there are probably no dramatic

124

Vincent A. de Rooij

differences between the two stages. Between 1940 and the present day, the urban centers of Southern Katanga and Elisabethville (the colonial name of present-day Lubumbashi) in particular, have absorbed large numbers of migrants, mainly from the neighbouring Kasai provinces (Fetter 1976: 173– 176). The massive influx of Luba-Kasai, Songye, Kanyok and Kete speakers from Kasai, all of whom have/had to learn Swahili, does/did of course have an impact on the language. It should also be noted that the majority of these migrants’ children acquire/d Katanga Swahili as their first language instead of their parents’ native language. Therefore, Katanga Swahili as it is spoken today is to be seen as the outcome of processes of second language learning and nativization that have been going on since the 1940s.

2. Phonology Katanga Swahili has a symmetrical five-vowel system (see Table 1), also found in ad/substrate languages. Just like East Coast Swahili, it does not have phonemic vowel lengthening nor does it have grammatical, apart from one exception, or lexical tone which are all prominent in the ad/substrate languages. Phonetic values of /e/ and /o/ range, depending on the environments they occur in, from [e] to [ɛ] and from [o] to [ɔ ] respectively. East Coast Swahili dental and velar fricatives do not occur in Katanga Swahili. In East Coast Swahili, they are found exclusively in Arabic borrowings. Many of these words denote Islamic concepts and therefore play no role in Katanga where Islamic influence is virtually non-existent. In the few generally used words of Arabic origin, dental fricatives /ð/ and /θ/ become /z/ and /s/ respectively. The velar fricative /γ/ occurs in Katanga Swahili as /k/. ECS glottal fricative has also been lost. Several other ECS phonemes have not been fully retained in Katanga Swahili either, most probably due to ad/sub-

Table 1. Katanga Swahili vowels

close close-mid open

front

back

i

u e

o a

Katanga Swahili

125

Table 2. Katanga Swahili consonants

plosives nasals fricatives affricates trills tap lateral approximant glides prenasalized consonants

labiobilabial dental

postalveolar alveolar palatal

pb m

t n s

f

velar

glottal

k tʃ

l w mp mb

j mf mv

nt nd ns nz

nʃ ndʒ

nk ng

Note: indicates ECS phoneme with weak phonemic status in Katanga Swahili

strate influence from Bemba and Luba-Kasai, where they do not function as phonemes (cf. Kashoki 1968 and Burssens 1939). In Luba-Kasai and Bemba we do not find: voiced palatal affricate /ʤ/, voiced velar plosive /g/, glottal fricative /h/, and alveolar vibrant /r/. In Katanga Swahili, these are commonly replaced by /j/, /k/, ø, and /l/ respectively. Furthermore, in Bemba we do not find: voiced alveolar plosive /d/, voiced labio-dental /v/ and alveolar fricative /z/. These are often replaced by /l/, /f/, and /s/ respectively. /s/, /z/, /t/ are often palatalized, especially by speakers with a Luba-Kasai background, to [ʃ], [ʒ], and [c~tʃ] respectively when followed by /i/. Furthermore, the ECS preverbal tense affix -li- is normally pronounced as [ri] or[ɾi]. Katanga Swahili has a range of prenasalized consonants that are also found in substrate languages (cf. Bostoen 1997: 91). The status of some of these prenasalized consonants as phonemes remains to be settled (de Rooij 1997: 335). It seems clear that contact with ad/substrate languages has resulted in a phonological system that has drifted away from the East Coast Swahili system and has become more similar to the systems of ad/substrate languages. Phonemes that do not occur in ad/substrate languages have been lost, partly or completely. The clearest example of this process is /h/ which is almost categorically left unarticulated. Other phonemes have blended into one, where the phoneme that is absent in one or more ad/substrate languages has been lost or weakened. An example is the blending of East Coast Swahili velar plosives /g/ and /k/ into /k/.

126

Vincent A. de Rooij

3. Nominal structures Nominal structures in Katanga Swahili are more analytic than those in ECS and ad/substrate languages. Katanga Swahili has retained the typically Bantu nounclass agreement system, but has done so in a reduced and simplified form. Table 3 lists the noun class prefixes of ECS and Katanga Swahili. Noun classes 1 through 10 are arranged pairwise where the even numbered class prefix denotes plurals and the odd numbered class prefix singulars (e.g. class 7 ki-tabu ‘book’ versus class 8 bi-tabu ‘books’). The number of noun class prefixes in Katanga Swahili has increased in comparison to ECS: although it has lost one (Cl. 10 collapses with Cl.6) it has added three (Cl. 11, 12, 13) which have been borrowed from Luba-Kasaï and Bemba. The differences in morphophonemic shapes of noun class prefixes 1, 2, 3, 5, 8, 11, and 14 can also be attributed to ad/substrate influence.

Table 3. Noun class prefixes in Katanga Swahili, ECS, Luba-Kasaï and Bemba (some variants of prefixes not shown for reasons of clarity) Noun class

ECS

Katanga Swahili

Luba-Kasaï

Bemba

1 2 3 4 5 6 7 8 9 10 10 → 6 11 12 13 14 15 16 17 18

mwammi-

mubamumiji-/ri-/∅makibiN-/∅-

mubamumidimatbiNN-

mubamumilimatfiniNniN-

lukatubukuɸakumu-

lukatubukupakumu-

makivi-

u-

ukupakumu-

malukatubukupakumu-

Katanga Swahili

127

It should be noted that infinitives are morphologically marked as nouns by noun class 15 prefix ku-. The locative classes 16, 17, and 18 have a different status than they have in ECS. In Katanga Swahili, pa-, ku-, and mu-, occur as pre-prefixes and may function as prepositions, as shown in (13) where mu is followed by an NP consisting of a demonstrative (ile) and a plural noun (mashiku ‘days’). (1)

(…) mais u-na-kufwa mu ile ma-shiku. but you-tma-die loc dem 6-day ‘(…) but you will die during that period.’ (Félicien/VDK1: 8/38)

In ECS, on the other hand, locative phrases are formed by suffixing a general locative affix -ni to a noun, while pa-, ku-, and mu- can only be affixed to noun modifiers. The use of locative prefixes as pre-prefixes in Katanga Swahili must be attributed to ad/substrate influence, since it occurs in all Central Bantu languages (Grégoire 1975: 17). The semantics of Katanga Swahili pa-, ku-, and mu- is the same as in the ad/substrate languages and ECS where, roughly speaking, pa- expresses a general locative meaning (at), ku- expresses direction (toward), and mu- expresses being inside of (in). Noun–Adjective agreement has been simplified radically: most adjectives have only one generalized form that is used with nouns from different classes. Agreement is marked most strongly in classes 1,2,7,8,12,13,14, especially in the plural classes 2 and 8 among these, but ultimately depends on the strength of the generalized form. This phenomenon does not occur in neighboring languages and can, therefore, not be the result of borrowing. Reduction and simplification is also found in subject and object agreement on the verb. In Katanga Swahili object concord markers co-indexing non-human objects are very seldomly used, except for classes 7/8. The generalized, but not categorical (Bostoen 1997: 106), use of i-, as a subject marker in classes 3 through 10 is striking. This restructuring cannot be explained by invoking ad/substrate influence, because the ad/substrate languages make use of the same agreement system as ECS, where markers have roughly the same morphophonemic shape as the prefixes of the nouns they refer to. The use of marker i- seems to correlate strongly with the feature [−human]: it does not occur with nouns belonging to classes 1/2, denoting human beings while its use is favored in all other classes except classes 7/8 and 11 through 14. Noun class prefixes 11 through 14 stand apart from the others in that they are used productively to derive nouns with very specific meanings (e.g. diminutives).

128

Vincent A. de Rooij

4. Verbal structures Katanga Swahili has the preverbal tma affixes that are typical of Bantu languages. According to Schicho (1988,1990) the ECS preverbal tense affixes that have survived in Katanga Swahili have lost much of their meaning as realizations of tense and may in many cases be regarded as a kind of dummyelements that have to be realized for morpho-syntactic reasons. The most frequently used preverbal tense affixes in Katanga Swahili are: -na- marking present tense, -li- or -ri- marking past tense, and -ta- marking future tense. According to Schicho (1988: 568), in a narrative sequence time reference needs to be marked only once by a tense affix on a verb, by sentenceinitial adverbs, or may even be left unexpressed if time reference can easily be inferred from contextual information. Schicho claims that the following distinctions provide the basis for what he calls the Aspect-dominated tma system of Katanga Swahili: 1. [+anterior] (including [perfect/resultative]) 2. [−anterior, −posterior] 3. [posterior/irrealis] 4. [progressive] (including [habitual], [intensive], [durative], [iterative]) tma is often expressed by auxiliary verbs. [+anterior] with perfective/resultative aspect can be expressed by using -toka (mu-) ‘leave, quit (loc)’, and -isha ‘finish’, also shortened to -sha as shown in (2). -isha is also used in this way in ECS but -toka (mu) is not. (2)

mi-na-isha ku-pakala vernis. I-tma-finish inf-apply varnish ‘I have already varnished it.’ (Schicho 1988: 569)

[posterior/irrealis] can be expressed by -tafuta ‘look for’ and by -enda ‘go’ as in (3). (3)

(h)a-ba-ta-enda ku-ra nani, ku-ra nkuku. neg-they-tma-go inf-eat filler inf-eat chicken ‘they will not eat ehm, eat chicken.’ (Michel/M1: 5/09)

Katanga Swahili

129

[progressive] with habitual, intensive, durative, or iterative aspects can be expressed by the verb -anza ‘start, begin’ as in (4) or by the copulative element -ko- used as a preverbal affix as shown in (5). (4)

a-na-anza ku-fan(ya) ma-bêtise. he-tma-start inf-do 6-stupidity ‘He started (and went on) doing foolish things.’ (Dédé m./M1: 7/08)

(5)

disons be-ko-na< be-ko-na-soigner eh? let’s say they-cop-tma< they-cop-tma-care for tag ‘Let’s say they provide medical care, don’t they?’ (Félicien/VDK1: 7/17)

In ECS -isha is used to mark perfective/resultative aspect but the use of -anza ‘start’ to mark habitual/durative aspect and -enda ‘go’ or -tafuta ‘look for’ to express inchoative aspect can be connected to semantically similar verbs in Luba-Kasai (Bostoen 1997: 121). In ECS and the ad/substrate languages, negation is expressed by a verbal prefix. In Katanga Swahili this strategy is still used with one important difference: in most cases, a sentence-final second negation element is added (cf. Schicho 1992: 84). In (6) we find (h)apana, originally a negative locative copula in ECS, but nowadays used almost exclusively as a negative answer to a yes/no question meaning. (6)

mais zamani (h)a-ba-ku-anz-ak-e ku-rima vile but long ago neg-they-tma-start-int-fin inf-cultivate thus (h)apana. neg ‘But long ago they didn’t cultivate (the fields) like that.’ (Papa Tshibangu/VT3: 1/04)

Periphrastic constructions, however, do also occur making use of Swahili verbs with an inherently negative meaning as in (7) and (8). This use of negative verbs does not occur in ECS but has been attested in the following ad/ substrate languages: Bemba, Luba-Katanga, Luba-Kasai, Lala, and Lamba (Kamba Muzenga 1981: 56). (7)

kama u-na-kosa kw-enda ku-bar hii. if you-tma-refuse inf-go loc-bar dem ‘If you don’t go to that bar.’ (Schicho 1990: 478)

130

Vincent A. de Rooij

(8) a-na-acha ku-tumika. he-tma-discontinue inf-work ‘He doesn’t work any more/longer.’ (Schicho 1990: 478) Another loss of verbal morphology in favor of periphrastic constructions is evident in relativization. In ECS, two relativization strategies occur: a relativizer may be infixed or suffixed on the verbal complex, or a relative pronoun with a relative concord suffixed to it may be used as in (9). This construction alternates with the one in (10) where the relative concord is infixed in the verbal complex: (9) mtu amba-ye a-li-mw-ona simba a-na-ogopa sana. man pron-rel he-past-it-see lion he-prs-fear much ‘The man who saw the lion is very frightened.’ (Vitale 1981: 90) (10) mtu a-li-ye-mw-ona simba a-na-ogopa sana. man he-past-rel-it-see lion he-prs-fear much ‘The man who saw the lion is very frightened.’ (Vitale 1981: 90) In Katanga Swahili we find a demonstrative, agreeing with the noun it modifies, which functions as a relativizer. It may be placed before or after the noun. Examples are given in (11)–(12). (11) u-ko d’accord na (h)ii mambo e-ko-na-sema? you-cop in.agreement with dem things she-cop-tma-say? ‘Do you agree with the things she’s saying?’ (Dédé m./M1: 2/10) (12) (h)aba ba-filles tu-ko-na-ona ba-na-zunguluka (h)umu. dem 2-girls we-cop-tma-see they-tma-walk around here ‘these girls we see walking around here.’ (Fidélie/M1: 4/10) The development of relativizers from deictic elements is a wide-spread phenomenon in contact-induced languages (Romaine 1988: 250). In the case of Katanga Swahili, this common trend has undoubtedly been reinforced by ad/ substrate pressure from Bemba (de Rooij 1997: 330).

5. Other parts of speech Numerals in dates and years are mostly in French. Discourse markers, including items that are traditionally classified as conjunctions, constitute another

Katanga Swahili

131

group of items that occur almost exclusively in French (De Rooij 2000). Frequently occurring are bon as a marker of topic development and transition, non as a quotation marker, mais as a marker of contrast, puisque and parce que as markers of causal relation and alors, donc and et puis as markers of conclusion and succession. Examples of donc, et puis, are given in (13); mais occurs here as an element initiating a switch to French. In (14), mais is used repeatedly in a monolingual Swahili fragment. (13) njo eh Mungu, ni Mungu w-a richesse, top eh God cop God 1-conn wealth ‘God, is a god of riches, (0.5) (0.5) ni Mungu w-a or, ni Mungu cop God 1-conn gold cop God is a God of gold, is a God w-a argent. (0.5) ↓donc (h)ii richesse 1-conn silver (0.5) so dem wealth of silver. (0.5) therefore, all this riches yote (h)ii i-na-tu-appartenir shi all dem it-tma-us-belong to we belongs to us, we who are ba-toto yake. (1.0) et puis, Mungu 2-children poss.3sg (1.0) furthermore God His children. (1.0) furthermore, God a-shi-na Mungu w-a bu-chafu mais he-neg-cop God 1-conn filth but is not a God of squalor mais (but) c’est un Dieu de la propreté. it’s det God of det cleanliness it’s a God of cleanliness.’ (Fidélie/DM13) (De Rooij 2000: 461 [glosses and translation slightly adapted, VDR])

132

Vincent A. de Rooij

(14) njo ku-sema: lu-fu ya bamalaika: c’est que top inf-say 11-death conn 2-angel it.means.that ‘It means that it was the death of angels: it means that mu-ntu: a-ju-e ma-neno: mais a-na-kufa tuu 1-man he-know-neg 6-word but he-tma-die just/merely a human being who does not understand what is happening but still bule. mm? mu-ntu yee hapana ku-jua ma-neno: for.nothing tag 1-man he neg inf-know 6-word he dies a senseless death. see? a man who does not understand what mais yee a-na-kufa: bule. but he he-tma-die for.nothing is happening but he dies a senseless death.’ (Fabian 2000: #5) [translation slightly adapted, VDR]) Question words, demonstratives, connectors, and free pronouns, all remain Swahili. In De Rooij (2000) it is suggested that quasi wholesale borrowing of French markers has pragmatic and syntactic explanations. The pragmatic explanation holds that French markers stand out better in Swahili environments and, hence, are more salient and effective as discourse structuring devices. Using Fench markers in Swahili is, furthermore, not a problem from a syntactic perspective since markers are not or hardly integrated in the morphosyntactic frame of sentences.

6. Conclusion It is clear that features of several Bantu languages have been borrowed into Katanga Swahili: auxiliary verbs to express aspect, locative noun prefixes (that are similar to prepositions), negative verbs to express negation, and an analytic relativization strategy. It is not always possible to identify one language as the source language because of the strong structural similarities between potential source languages. It is clear, however, that those elements that have been borrowed into the language can all be described as analytic. The change from an agglutinative morphosyntax to a more analytic structure is all the more striking since all of the languages involved are related Bantu languages, all of them agglutinative in structure. This particular change may

Katanga Swahili

133

be explained as a result of strategies promoting semantic transparency in first language acquisition and second language learning under less than ideal learning conditions. The other major instance of borrowing, the use of French discourse markers, fits a general pattern in language contact, well-documented in many studies (see e.g. the contributions to Maschler, ed. 2000).

Abbreviations and transcription conventions (0.5) ↓ 2,6-[noun] cop dem fin inf int

pause (in seconds) sharp low-falling pitch contour class 2 or 6 noun prefix copula demonstrative final vowel of verb form infinitive intensive aspect

loc neg past prs rel tag tma top

locative negation past tense present tense relativizer tag tense/mood/aspect topicalizer

Notes 1. This chapter is based in part on de Rooij 1995, 1996, 1997. I am grateful to those who commented on these earlier publications and to the editors of this volume for their comments. All usual disclaimers apply. Unless indicated otherwise, all examples in this chapter were selected from the author’s fieldwork data. Fieldwork was carried out in Lubumbashi, dr Congo, in 1991 (June–October) and 1992 (June–December) with grants from the Institute for Functional Research into Language and Language Use (IFOTT) and the Netherlands Foundation for the Advancement of Tropical Research (WOTRO). The financial support of both institutions is hereby gratefully acknowledged. 2. In 1971, Katanga was renamed into Shaba as part of Mobutu’s policy of ‘zaïrianization’. After Mobutu’s removal from power in 1997, this renaming was undone and the province was given back its former name, Katanga. 3. Due to the fact that nearly all research on Swahili as spoken in Katanga has been carried out in Lubumbashi, the capital of Katanga, some authors refer to the language as ‘Lubumbashi Swahili’ (see e.g. Gysels 1992, Polomé 1968, Schicho 1982). How and to what degree Swahili as spoken on the Copperbelt resembles varieties of Swahili spoken elsewhere in Katanga remains an open question.

134

Vincent A. de Rooij

References Blommaert, Jan 1999 Reconstructing the sociolinguistic image of Africa: Grassroots writing in Shaba (Congo). Text 19: 175200. 2004 Writing as a problem: African grassroots writing, economies of literacy, and globalization. Language in Society 33 (5): 643671. Bostoen, Koen c. 1997 Het Shaba-Swahili: Geschiedenis en bronnen [Shaba Swahili: History and sources]. MA Thesis, Ghent University. Burssens, Amaat 1939 Tonologische Schets van het Tshiluba (Kasayi, Belgisch Kongo). Antwerpen: De Sikkel. Fabian, Johannes 1986 Language and Colonial Power: The Appropriation of Swahili in the Former Belgian Congo, 18801938. Cambridge: Cambridge University Press. 1990 History from Below: The ‘Vocabulaire of Elisabethville’ by André Yav (Texts, translations, and interpretive essay). Amsterdam: John Benjamins. 2000 The history of Zaire as told and painted by Tshibumba Kanda Matulu in conversation with Johannes Fabian, Third Session, Part 2. Archives of Popular Swahili 2 (7). http://www2.fmg.uva.nl/lpca/aps/tshibumba3b.html (3 August 2007). Fetter, Bruce 1976 The Creation of Elisabethville 19101940. Stanford: Hoover Institution Press. Grégoire, Claire 1975 Les locatifs en bantu. (Annales du Musée de l’Afrique Centrale, Série in-8º, Sciences Humaines, no 83.) Tervuren: Musée de l’Afrique Centrale. Gysels, Marjolein 1992 French in urban Lubumbashi Swahili: Codeswitching, borrowing, or both? Journal of Multilingual and Multicultural Development 13 (1/2): 4155. Kabamba, Mbikay 1979 Stratigraphie des languaes et communications à Lubumbashi. Problèmes sociaux zaïrois 124125: 4774. Kamba Muzenga 1981 Les formes verbales négatives dans les langues bantoues. (Annales du Musée de l’Afrique Centrale, Sciences Humaines 106.) Tervuren: Musée de l’Afrique Centrale.

Katanga Swahili

135

Kapanga, André Mwamba 1993 Shaba Swahili and the processes of linguistic contact. In: Francis Byrne and John Holm (eds.), Atlantic Meets Pacific: A Global View of Pidginization and Creolization, 441458. Amsterdam: John Benjamins. Kashoki, Mubanga E. 1968 A Phonemic Analysis of Bemba: A Presentation of Bemba Syllable Structure, Phonemic Contrasts and Their Distribution. (Zambian Papers 3.) Manchester: Manchester University Press. Maschler, Yael (ed.) 2000 Discourse Markers in Bilingual Conversation. Special issue of International Journal of Bilingualism 4 (4). Möhlig, Wilhelm J. G. 1981 Die Bantusprachen im engeren Sinn. In Bernd Heine, Thilo C. Schadeberg, und Ekkehard Wolff (eds.), Die Sprachen Afrikas, 77116. Hamburg: Helmut Buske Verlag. Polomé, Edgar C. 1968 Lubumbashi Swahili. Journal of African Languages 7 (1): 1425. Romaine, Suzanne 1988 Pidgin and Creole Languages. London: Longman. De Rooij, Vincent A. 1995 Shaba Swahili. In: Jacques Arends, Pieter Muysken, and Norval Smith (eds.), Pidgins and Creoles: An Introduction, 179190. Amsterdam/ Philadelphia: John Benjamins. 1996 Cohesion through Contrast: Discourse Structure in Shaba Swahili/ French Conversations. Amsterdam: Ifott. 1997 Shaba Swahili: Partial creolization due to second language learning and substrate pressure. In: Arthur K. Spears and Donald Winford (eds.), The Structure and Status of Pidgins and Creoles, 309339. Amsterdam/Philadelphia: John Benjamins. 2000 French discourse markers in Shaba Swahili conversations. International Journal of Bilingualism 4 (4): 447467. Schicho, Walter 1982 Syntax des Swahili von Lubumbashi. Vienna: Afro-Pub. 1988 Tense vs. aspect in Sango and Swahili of Lubumbashi. In: Mohammad Ali Jazayery and Werner Winter (eds.), Languages and Cultures: Studies in Honor of Edgar C. Polomé, 565579. Berlin: Mouton de Gruyter. 1990 AUX, Creole und Swahili von Lubumbashi. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 43 (4): 476483. 1992 Non-acceptance and negation in the Swahili of Lubumbashi. African Languages and Cultures 5 (1): 7589. Vitale, Anthony J. 1981 Swahili Syntax. Dordrecht: Foris.

Grammatical borrowing in Khuzistani Arabic Yaron Matras and Maryam Shabibi

1. Background The Khuzistani dialect of Arabic (henceforth Kh. Arabic) is spoken natively by over 3 million people, who constitute roughly 7 percent of the population of the province of Khuzistan in south-western Iran. The dialect is the easternmost representative of the continuum of Mesopotamian dialects of Arabic, which cover the river lands of southern Iraq in the west (Ingham 1982: 14). Arab settlement in the area is believed to go back to the beginning of the Christian era. In the centuries following the advent of Islam, the Arabic language enjoyed the status of the literary language of religion, scholarship, and administration, as well as being the primary language of everyday communication in the province. This changed with the coming to power of Reza Shah Pahlavi in 1926 and the introduction of an intensive campaign favouring Persian as the only official state language. The policy included the settlement of a Persian-speaking population in the province. Unlike other dialects of Arabic, Kh. Arabic has not attracted much attention within the linguistic community. Ingham (1997) devotes a chapter to the dialect, focusing, however, on an introductory discussion of phonology and vocabulary only; and Shabibi (2006) provides an overview of the structures of the dialect along with an analysis of contact-induced developments in morphosyntax. Persian is now the only language of education, local media and newspapers, administration, and most urban commerce in the province of Khuzistan. Arabic is the language of the family and Arabic-speaking neighbourhoods, though even as an informal language it is now in decline, and Persian is the preferred language of the younger generation born since the 1970s. All educated adult speakers of Arabic are bilingual, and Arabic monolingualism is limited to the uneducated older generation, and to the older generation in rural communities. Arabic literacy is limited by and large to reading the Qur’an, and to a very basic level of instruction in Modern Standard Arabic, though even most educated Arabs have no active command of Modern Standard Arabic. There is, however, considerable exposure to Arabic-language satellite media, and so to the broadcast (oral) version of Modern Standard Arabic. In some

138

Yaron Matras and Maryam Shabibi

cases, Khuzistani Arabic speakers are able to read modern Arabic by drawing on their exposure to these media, combined with their basic familiarity with the Arabic script and with Classical Arabic (Qur’an).

2. Phonology The only apparent phonological contact phenomenon in Kh. Arabic is the interchange of /ɣ/ and /q/, in words such as /ɣarīb, qarīb/ ‘close’ (cf. Modern Standard Arabic /qarīb/ ‘close’, /ɣarīb/ ‘strange’). This matches the realization in Persian of etymological /q/ as /ɣ/. Phonemes that are otherwise absent from the Arabic system, most notably /p/, are retained in Persian loanwords: panjara ‘window’,

3. Morphological typology A major change under Persian influence is the levelling of the status of attributes. In Arabic, adjectival attributes follow the head noun, and agree with the head noun in gender, number, as well as in definiteness: (1)

Standard Arabic (and other dialects) a. walad kabīr boy big.m ‘a big boy’ b. l-walad l-kabīr def-boy def-big.m ‘the big boy’

Nominal attributes, by contrast, are conjoined by means of the attributive Iḍāfa-construction, whereby only the dependent (genitive) noun is overtly marked for definiteness: (2)

Standard Arabic (and other dialects) walad l-mudīr boy def-director ‘the director’s son’

Khuzistani Arabic

139

In Persian, both types of attributes are treated in the same way: The attribute (whether adjectival or nominal) follows the head, and an attributive particle (the Ezāfe marker) mediates between the two: (3)

Persian a. pesar-e bozorg boy-ez big ‘the big boy’ b. pesar-e modīr boy-ez director ‘the director’s son’ c. xūne-ye sefīd house-ez white ‘the white house’ d. moʔallem-e madrese teacher-ez school ‘the school teacher’

The pattern in Kh. Arabic matches the Persian arrangement (note that, as in other dialects of Arabic, the definite article l- assimilates to dental consonants, resulting in gemination of that consonant): (4)

Khuzistani Arabic a. walad č-čibīr boy def-big.m ‘the big boy’ b. walad l-modīr boy def-director ‘the director’s son’ c. bīәt l-abyaḍ house def-white ‘the white house’ d. moallәm-at l-madrәsa teacher-f.cons def-school ‘the school teacher’

Note that in the adjectival attributive construction in (4a and 4c) overt definiteness agreement between noun and adjective is lacking, just like in the

140

Yaron Matras and Maryam Shabibi

genitive attribute construction in (4b and 4d). Based on the Persian model, Kh. Arabic has reanalysed the definite article in such constructions as a marker of attribution, which matches the Persian (definite) Ezāfe marker -(y)e. Its distribution now resembles that of the Persian Ezāfe attributive marker: It appears, like Persian -(y)e, between the two constituents of the attribution, and it is used to link both adjectival, and nominal attributes. Further evidence that the functions of the Persian construction are mapped onto Arabic structures is provided by the position of the feminine Construct State or Iḍāfa-marker -at, seen in (4d) in a position that is not untypical of Arabic as a whole. In Arabic, the Construct State marker (still recognisable in the vernaculars only in the feminine singular) is reserved for nominal attribution, as in (4d). But in Kh. Arabic we find it in adjectival attributive constructions as well, as in (5a–b); it even attaches directly to adjectives, as in aly-at ‘high.f’ in (5b): (5)

Khuzistani Arabic a. jazīr-at l-xaḍra island-f.cons def-green ‘the green island’ b. ṭōf-at aly-at l-bīәt wall-f.cons high-f.cons def-house ‘the high wall of the house’

This matches again the distribution of the Persian Ezāfe marker -(y)e (6): (6)

Persian a. jazīre-ye sabz island-ez green ‘the green island’ b. dīvār-e boland-e xūne wall-ez tall-ez house ‘the high wall of the house’

Note that in the ‘mixed’ type, in (5b), involving both an adjectival-attribute (‘high wall’) and a genitive attribute (‘wall of the house’), the first (adjectival) attribution relies exclusively on the Construct State marker, while the second (nominal) relies on the combination of the Construct State marker with the following definite article. In fact, Kh. Arabic allows for variation in such cases, and the Construct State marker may be accompanied by a definite art-

Khuzistani Arabic

141

icle in both positions. Consider example (7), where the nouns are masculine, and there is no option of using an overt Construct State marker: (7)

a. Khuzistani Arabic walad č-čibīr l-modīr boy def-big def-director ‘the director’s big/eldest son’ b. Standard Arabic (and other dialects) walad l-mudīr l-kabīr boy def-director def-big ‘the director’s big/eldest son’ c. Persian pesar-e bozorg-e modīr boy-ez big-ez director ‘the director’s big/eldest son’

The crucial aspect of the Kh. Arabic construction is (1) to have a marker of attribution mediating between the head and its attribute, (2) to place the attribute in a position immediately following its head, and (3) to avoid any overt marking of definiteness in the adjectival attribution. In all this, Kh. Arabic copies precisely the Persian attributive construction. Contrasting with Persian, it retains a distinct marking of attribution with feminine singulars, but allows this marking to assimilate into the generic function of the attributive marker. The outcome of the process is (1) the loss of the distinction between nominal and adjectival attribution, (2) the loss of overt marking of definiteness in attributive constructions, (3) a change in the word order in complex (‘mixed’) attributive constructions (as in 5b and 7a), and, finally, (4) gender variation in the marking of the attributive construction, with optional use of the definite article to accompany the Construct State in feminine singulars in complex attributions.

4. Nominal structures The most notable contact-induced change in Kh. Arabic nominal structures is the status of the Iḍāfa-construction alluded to above. The replication of a construction type that is similar to the Persian Ezāfe leads, as discussed above, to the abandonment of definiteness agreement. The decline of overt definiteness marking can also be observed in other constructions in the language,

142

Yaron Matras and Maryam Shabibi

notably in the absence of a definite article with definite head nouns of relative clauses. This too follows a Persian model (where definiteness generally remains unmarked): (8)

Khuzistani Arabic mara lli šift-ū-ha xābar-at. woman rel saw-2pl.m-3sg.f called-3sg.f ‘The woman that you saw called.’

5. Verbal structures In the derivation of verbs, the tendency to paraphrase inchoative and causative verbs drawing on an analytic construction rather than on derivational morphology, although found in other dialects of Arabic, appears to be reinforced by Persian. Thus we find: (9)

Khuzistani Arabic š-šijra z-zɣīr-a šwayye šwayy tṣīr čibīr-a. def-tree def-small-f little little become.3sg.f big-f ‘The small tree gradually grows.’

Loan-verbs appear to be limited to the replication of Persian compound verbs consisting of a nominal stem (masdar) and a verbalizing element or ‘light verb’ (Persian kardan ‘to do’ or šodan ‘to become’). The nominal stem, often itself an Arabic loan into Persian, is replicated directly in Kh. Arabic, while as corresponding native light verb ṣaww- ‘to do’ is employed for Persian kardan, and ṣār- ‘to become’ for Persian šodan, thus: Persian taaɣībeš kard ‘he followed him’ (follow-3sg did.3sg) is rendered ṣawwā-h taɣīb (did.3sg-3sg follow). An additional change to the verb system, brought about through Persian influence, concerns the tense system. Persian has both a simple past tense, which is expressed by the person-inflected past stem of the verb, and a composite past tense, which consists of a past participle and an auxiliary. The auxiliary, based on the existential verb, may inflect for person as well as tense; the present-tense auxiliary is used to form the perfect, the past-tense auxiliary forms the pluperfect. Arabic, by contrast, has only one, simple past tense, though combinations of the past-tense existential verb with the lexical verb (usually in the imperfect or present-future) are also possible, usually expressing habitual aspect of conditional mood. Kh. Arabic copies the Per-

Khuzistani Arabic

143

sian composite past tense, drawing on inherited resources. The only available participle form in Arabic is the present participle, which inflects for gender and number (but not for person), and it is this form that serves as the basis for composite past tense in Kh. Arabic. Since the Arabic existential verb does not have a present-tense form, the only available auxiliary is a past-tense auxiliary; the construction thus matches the Persian pluperfect: (10) a. Khuzistani Arabic mәn rәħ-әt. lә-l-bīet, huwwa mā-rāyәħ čān when went-1sg to-def-home he neg-going.sg.m was.3sg.m ‘When I went home he had not gone away.’ b. Persian vaɣti raft-am xūne, ūn na-rafte būd. when went-1sg home he neg-gone was.3sg.m ‘When I went home he had not gone away.’ (11) a. Khuzistani Arabic mәn gabul šāyfat-ha čәnәt. from past seeing.sg.f-3sg.f was.1sg ‘I had seen her before.’ b. Persian az ɣabl ūn-o dīde būd-am. from past 3sg-acc seen was-1sg ‘I had seen her before.’

6. Other parts of speech A series of Persian discourse markers, fillers, tags, and focus particles are used in Kh. Arabic. Most of these elements are well integrated into Kh. Arabic and are not perceived by speakers as foreign. The category that is most obviously influenced by Persian is that of discourse markers with a primarily interaction-qualifying rather than syntactic-semantic function: xō/xōb/xōš ‘well’, xōlāse ‘in sum’, albate ‘of course’, hič ‘at all, altogether’, ham ‘indeed, well’: (12) xōb w-hāy sabab ham l-laði gabal čān … dm and-this reason dm rel once was ‘Well, and the reason that indeed once existed for this …’

144

Yaron Matras and Maryam Shabibi

(13) xolāse hīč mā-rəħ-na madrasa. dm dm neg-went-1pl school ‘After all, we didn’t go to school at all.’ (14) albate čān-an ham b-ðīč z-zamān banāt č-čān-an dm were-3pl.f dm in-that def-time girls rel-were-3pl.f yarħ-an. go-3pl.f ‘Of course there were indeed girls at the time who used to go [to school].’ These are accompanied by Persian-derived focus particles: ham ‘too’ and ham … ham ‘both … and’. (15) ðīč ənd-ha θnīən frūx ana ham ənd-ī θnīən. that.f poss-f two children I too poss-2sg two ‘She has two children, and I have two children, too.’ (16) ham ana w ham alī rəħ-na l-əl-pārk. both I and also Ali went-1pl to-def-park ‘Both Ali and I went to the park.’ (17) umm-ī ham ɣəsl-at lə-mmāīn ham naḍḍəf-at mother-1sg both washed-3sg.f def-dishes and cleaned-3sg.f l-bīət. def-house ‘My mother [both] washed the dishes and cleaned the house.’ Optional, occurring in variation alongside various Arabic-derived counterparts such as ħatta ‘even’ or lākin ‘but’, is the contrastive correlative balke ‘but [… also]’: (18) huwwa mū bass bāhūš balke šujjā ham. he neg only clever but brave too ‘He is not only clever but also brave.’ Further Persian borrowings that are generalized in Kh. Arabic are the concessive subordinating conjunctions agarče and bā īnke, both 'although/ even though’, and the factual complementizer ke ‘that’:

Khuzistani Arabic

145

(19) huwwa rāħ lwaħda l-əl-pārk agarče umm-a he went.3sg.m alone to-def-park although mother-3sg.m gall-at l-a lā-yrūħ. said-3sg.f to-3sg.m neg-go.3sg.m ‘He went to the park alone, even though his mother told him not to go.’ (20) rayyāl-na bə-l-yōm xəṭab, bā īnke θaləθtaš sana man-1pl in-def-day proposed.3sg.m although thirteen year umr-ī sawwūm rāhnamāī ubū-y qəbal b-ī age-1sg third secondary school, father-1sg accepted for-3sg.m ‘When my husband proposed, although I was [just] thirteen years old, third year of secondary school, my father agreed.’ (21) tədr-īn ke rayl-əč ala kəl-šī čaððab. know-2sg.f comp husband-2sg.f on everything lied.3sg.m ‘You know that your husband lied about everything.’ The latter, the Persian complementizer and relativizer ke, does not appear in non-factual (subjunctive) complements, where instead we find the Arabic (historical) relativizer l-laði or illi, which also continues to cover the function of a relativizer. Nonetheless, occasionally Persian ke is also found in the position of the relativizer: (22) əbən uxū ɣāzī ke huwwa w mart-a hnā … son brother Ghazi rel he and wife-3sg.m here ‘Ghazi’s nephew, who is here with his wife …’ From this we might assume a gradual process of convergence in steps, as follows: in stage 1, the Persian model of having an identical marker for complement clauses and relative clauses (ke) is copied into Kh. Arabic, with the effect of generalizing the relativizer l-laði/illi (at the expense of the historical Arabic complementizer ’inn-) to cover the function of complementizer. The result is a convergence of patterns among the two languages. In stage 2, the actual Persian marker ke is adopted into Kh. Arabic in factual complement clauses, as seen in (21). The result is a split within Kh. Arabic between factual and non-factual complements, whereas the same marker is used in both languages to introduce factual complements. Finally, in stage 3, the beginnings of which are attested in the contemporary language, Persian ke infiltrates Kh. Arabic relative clauses as well, as seen in (22).

146

Yaron Matras and Maryam Shabibi

7. Constituent order One change in constituent order has already been mentioned above, in Section 2: it concerns the shift in ‘complex’ attributive constructions, away from the Arabic norm, which allows an adjectival modifier to be separated from its head (by a nominal modifier of the complex noun phrase), toward the Persian-type constituent order, whereby each attribute must immediately follow its head. We repeat example (7) here: (7)

a. Khuzistani Arabic walad č-čibīr l-modīr boy def-big def-director ‘the director’s big/eldest son.’ b. Standard Arabic (and other dialects) walad l-mudīr l-kabīr boy def-director def-big ‘the director’s big/eldest son’ (also: ‘the big director’s son’) c. Persian: pesar-e bozorg-e modīr boy-ez big-ez director ‘the director’s big/eldest son’

A further issue related to the order of constituents in Kh. Arabic concerns the position of the copula-auxiliary /čān/, which, in the composite past tense (pluperfect), follows the lexical verb: mā-rāyәħ čān ‘he had not gone away’ (Persian: na-rafte būd) (see examples 10 and 11). Noteworthy is also the flexible position of the causal conjunction čīe ‘because’. Like its Persian counterpart čon, it can also occupy the final position in the adverbial clause expressing cause: (23) a. Khuzistani Arabic līeš mā-reħ-tī l-әl-madrasa? why neg-went-2sg.f to-def-school čān edd-ī xuṭṭār čīe. was.3sg with-1sg guests because ‘Why didn’t you go to school?’ ‘Because I had guests.’

Khuzistani Arabic

147

b. Persian čerā be madrese na-raft-ī? why to school neg-went-2sg mehmūn dāšt-am čon. guest had-1sg because ‘Why didn’t you go to school?’ ‘Because I had guests.’ Finally, we must consider what appears to be the beginning of a shift in word order, extending the contexts in which Object–Verb order is favoured to comply more frequently with the Persian type. Object–Verb order in Arabic is generally highly marked and is employed as a means to topicalize the direct object. Kh. Arabic makes use of such strategies, which include – unlike Persian, where OV prevails – the pronominal resumption of the object in a position following the lexical verb. Nevertheless, such constructions in Kh. Arabic do not necessarily express the topicalization of the object: (24) lə-bnayya d-dār naḍḍəf-at-ha. def-little.girl def-room cleaned-3sg.f-3sg.f ‘The little girl cleaned [it] the room.’ (25) haðan xālāt-ī līsāns-hən kaẓẓ-ann-a. these aunts-1sg degree-3pl.f gained-3pl.f-3sg.m ‘My aunts received [it] their degree.’

8. Lexicon The presence of numerous Persian lexical borrowings is a distinguishing feature of Kh. Arabic, setting it apart from other neighboring dialects of Arabic. Nevertheless, there is considerable sociolinguistic stratification in the use of Persian vocabulary among different groups of speakers (cf. Shabibi 1998). As the principal language of the public sphere, Persian supplies numerous lexical items in the domains of trade, institutions, tools, and other aspects of public and technical life (e.g. xarīd-o-furūš ‘trade’, pīč guštī ‘screwdriver’, lebās šū’ī ‘washing machine’, etc.). In everyday vocabulary, Persian idioms are commonly calqued in Kh. Arabic, facilitated by the fact that those idioms themselves are often based on Arabic loan vocabulary in Persian, and so even more easily replicable in Kh. Arabic: Consider Kh. Arabic wāyәd mamnūn, lit. ‘very grate-

148

Yaron Matras and Maryam Shabibi

ful’, in the sense of ‘thank you very much’, based on Persian xeyli mamnūn, or Kh. Arabic yarreti zaħma, lit. ‘you have taken trouble [on my behalf]’, also an expression of gratitude, from Persian zahmat kešīdī. Here, the fact that the languages already share a large part of their vocabulary (as a result of earlier, historical influence of Arabic on Persian), makes replication of lexical Matter redundant, and promotes in turn replication of idiomatic Patterns surrounding a pivotal word in the idiom that is already shared by both languages.

9. Conclusion Matter replication of Persian material is found in Kh. Arabic primarily in the domain of lexical vocabulary, and in part in grammatical vocabulary, covering discourse markers that operate strictly on the interaction level (i.e. not conjunctions), focus particles, a correlative particle, a complementizer and relative particle, and concessive subordinating conjunctions. Pattern replication is most notable in the emerging change of constraints on word order (extension of marked word-order patterns), the favouring of analytic constructions and emergence of a new analytic past tense (pluperfect), and the reduction of overt marking of definiteness. Perhaps the most remarkable contact-induced change, one which strongly affects the typology of attribution in the language, is the identification of Kh. Arabic grammatical morphemes in attributive constructions – the Construct State marker (visible in the feminine singular only) and the definite article that appears between head and attribute – with the Persian attributive particle, and the consequent merger of two historically distinct attributive constructions – adjectival and nominal – into a single type, replicating the state of affairs in Persian.

Abbreviations acc comp cons cop def dm ez f

accusative complementizer construct state marker copula definite article discourse marker Persian Ezāfe attributive marker feminine

indef m neg past pl poss rel

indefinite article masculine negation simple past (perfective) plural possessive expression relative particle

sg

singular

Khuzistani Arabic

149

References Ingham, Bruce 1982 North East Arabian Dialects. London: Kegan Paul International. 1997 Arabian Diversions: Studies on the Dialects of Arabia. Reading: Ithaca Press. Shabibi, Maryam 1998 Variation in the use of Persian loan words among Iranian Arabic speakers. Unpublished Msc. Dissertation. University of Edinburgh. 2006 Contact-induced grammatical changes in Khuzistani Arabic. Ph.D. thesis. University of Manchester.

Grammatical borrowing in Domari Yaron Matras

1. Background Domari (also Domi, sometimes also Qurbati) is the Indo-Aryan language spoken by a population of commercial nomads in the Middle East. The language retains some archaic Indo-Aryan features, such as the Middle Indo-Aryan present-tense conjugation, but also shows some radical re-structuring in the past-tense conjugation and in syntactic typology. In this respect it resembles Romani, also an Indo-Aryan diaspora language of originally peripatetic groups. The self-appellations rom and dom are also related, both deriving from Indic ḍom. However, some isoglosses separating the two language appear to be rather ancient, and it is highly unlikely that they both split from the same ancestor language after leaving India. The present chapter deals with the only dialect that has been extensively documented – that of the Palestinian Doms of Jerusalem (see Matras 1999; Macalister 1914). The language is confined strictly to oral use within the family and, to a limited extent, with members of other Dom communities. The precise history and date of arrival of the group in the region remain unknown. The language shows a layer of Kurdish influence, and the community has a sense of affinity to another, Arabic-speaking population of commercial nomads who are referred to as “Kurds”. Arabic has been the principal contact language for many centuries. The Jerusalem community began shifting to Arabic in the 1960s, and individuals who were raised since this period are largely monolingual in Arabic, with only passive exposure to Domari. It is estimated that out of a total number of between 6001000 community members, only around 10 percent speak Domari fluently; the language is thus endangered or even moribund. Whatever information is available suggests that this is also the situation at least in urban Dom communities elsewhere in the Middle East.

2. Phonology The Domari sound system strongly resembles that of its contact language, Palestinian Arabic, though it is not always obvious that this is due to borrowing

152

Yaron Matras

or convergence. All consonants with the exception of /p/, /v/ (interchangeable with /w/), /tš/ and /g/ are shared with Arabic. Uvular /q/ appears also in the pre-Arabic component and may be the outcome of Iranian (Kurdish) influence (e.g. qišṭoṭa, alongside kišṭoṭa ‘small’). The glottal stop /ʔ/ is distinctive only in the postposed negation marker -éʔ: (1)

n-mang-am-éʔ neg-want-1sg-neg ‘I don’t want.’

The pharyngeals [ħ] and [ʕ] appear to be restricted to Arabic-derived lexical loans, but pharyngealization of dentals /t, d, s, z/ is transferred to the preArabic component as well: [wɑ:ṭ] ‘stone’, [ḍɑnḍ] ‘tooth’. Consonant gemination also appears independently of the Arabic component: [tilla] ‘big’. Under Arabic influence, the affricates /tš, dž/ are being reduced to sibilants /š, ž/. In the case of the voiced affricate, a similar change has fairly recently taken place in the Arabic dialect as well, and some variation is still observable. All vowel sounds with the exception of [ɔ] and [ʌ], both rather infrequent, are shared with Arabic. As in Arabic, there is variation in the realization of the short vowel phonemes /i/ [i, š, ɪ] /a/ [a, æ, ɑ] and /u/ [u, ʉ, ʊ], and of long /ā/ [a: æ:, ɑ:], with back vowels preferred in the vicinity of uvular and pharyngealized consonants, as in [ṭɑ:ṭ] ‘Arab’, but [tat] ‘heat’. As in Arabic, the vowels /u/ and /i/ are often interchangeable: džuwir/džiwir ‘woman’ (also a feature of Kurdish). Prothetic and epenthetic vocalization around initial consonant clusters can also be regarded as a general regional phenomenon. Prosody and intonation are largely shared with Arabic.

3. Morphological typology As a New Indo-Aryan language, Domari will have undergone a re-structuring of its past tense formation leading presumably at some stage to the emergence of split morphological ergativity. Evidence to this effect is the use of the originally oblique form of the pronoun for ama ‘I’, and the construction of the past-tense conjugation, based on the attachment of what once were oblique personal clitics (kard-o-m ‘I did’ < *karda-o-me ’done-by-me’) (cf. Matras 2002: 145151). The language is, however, no longer ergative, though this might be attributed to the general drift away from ergativity in northwestern Indo-Aryan frontier languages as well as in Iranian, and there is no con-

Domari

153

crete evidence linking it with the influence of Arabic, which is not ergative, either.

4. Nominal structures The overall similarity in word-order rules results in a similar positioning of nominal objects in the sentence. The most extensive Arabic influence on nominal structures is in the domain of local relations, more specifically the almost wholesale borrowing of Arabic prepositions. We can assume that Domari lacked prepositions altogether before contact. Local relations expressions that are not borrowed are expressed either by case suffixes, or by genitivepossessive location expressions which consist of a location adverb inflected for (oblique) possession and the Locative case, preceding a head in the Ablative. There is only a very small set of such inherited location adverbs, all expressing strict spatial relations: mandža ‘in’, bara ‘out’, paš ‘behind’, agir ‘in front’, atun ‘above’, axār ‘below’, and čanč- ‘next to’: (2)

čanč-is-ma kury-a-kī next.to-3sg.obl-loc house-obl.f-abl ‘next to the house’ (lit. ‘in its-side from-the-house’).

Arabic-derived prepositions may be integrated into this format, as long as the meaning is stative: (3)

žamb-is-ma lāč-a-ki next.to-3sg.obl-loc girl-obl.f-abl ‘next to the girl’ (< Arabic žamb ‘next to’).

Temporal and more specified spatial relations are generally expressed through Arabic prepositions, the nouns appearing in the Ablative case (serving as a general prepositional case): (4)

baʕd wars-ak-ki after year-indef-abl ‘after a year’ (< Arabic bad ‘after’)

The Arabic prepositions ma ‘with’, min ‘from’, la ‘to’, fī ‘in’and ind ‘at’ compete with the synthetic cases Associative, Ablative, Dative, and Locative

154

Yaron Matras

respectively. Occasional doubling of case may be observed, e.g. Arabic fī ‘in’ + Locative case -ma (see example 6), though in general these prepositions too trigger the default Ablative/Prepositional case on the noun (example 5): (5)

maʕ bɔy-im-ki with father-1sg.obl-abl ‘with my father’ (< Arabic ma ‘with’)

(6)

fī šare-ma in street-loc ‘on the street’ ( burǧ > qāḍi >

Aramaic séʿra ‘price’ (but pl siʿrō´ ) bórža ‘tower’ (but pl buržō´ ) ´ ḳō_̣ya ‘judge’ (but pl ḳa_̣yō´ )

In some loans the Arabic imāla (i-umlaut) ā [a:] → ē [e:] is attested (wētya < wādi ‘valley’) as in many Arabic dialects in the area (Arnold and Behnstedt 1993: 98105).

2.3. Stress and syllable structure Aramaic words stress only the final syllable or the penultima. In analogy to the Arabic geminated stems a vowel a is inserted after the geminated radical in Baxʿa and Jubbʿadīn, so that the stress is now on the antepenultima: (6)

Baxʿa and Jubbʿadīn záppani€ Maʿlūla zappni€

‘I sold’

3. Nominal structures (the integration of loans) Masculine singular nouns of Arabic origin receive the ending -a, feminine singular nouns the ending -€a or -ča. Masculine plural nouns have the ending -ō(ya), the feminine plural-ending is -(y)ō€a. After numerals, an enumeration plural is formed: (7)

Arabic qism

Aramaic (Maʿlūla) ḳesma ‘part’ ḳismō ‘parts’ i€ᵊr qism ‘two parts’ samaka samᵊk€a ‘fish’ (sg) samkō€a ‘fish’ (pl) €arč samkan ‘two fish’

Some Arabic loans receive the original Aramaic diminutive endings -ōna (m) and -(a)nī€a (f), which express diminution no longer:

Western Neo-Aramaic

(8) Arabic muʿallim muʿallima

189

Aramaic mʿallmōna ‘teacher’ (m) mʿallmanī€a ‘teacher’ (f)

As Arabic and Aramaic are two closely related Semitic languages, they share most of their nominal patterns. Only a few forms are borrowed from Arabic as miCCaC/muCCaC and maCCūC: (9) Aramaic < Arabic mufčḥa < miftāḥ ‘key’ makčūba < maktūb ‘letter’ Gender distinction is lost in Baxʿa and among some speakers in Maʿlūla and Jubbʿadīn too, in all plural forms of the adjectives and of the independent or suffixed pronouns. This can be explained only by the influence of the neighbouring Arabic villages, which do not know gender distinction in these forms at all.1

4. Verbal structures The two old Semitic tenses perfect and imperfect are preserved2 probably under the influence of Arabic (Correll 1978: 153). They are used to express preterite tense and subjunctive exactly as in the Arabic dialects of Syria. Different from Arabic are the two new tenses, the present tense and the perfect that have developed from the old participles. Future tense or optative is expressed in Maʿlūla and Baxʿa with the auxiliary verb batt- (< Arabic badd-)3 with Aramaic person inflection and following subjunctive just as in Arabic: (10) Aramaic: batt-e yīxul Arabic: badd-o yākol will/want-3sg eat-3sg.subj ‘He will/wants to eat.’ To express a progressive action the Arabic prefix ʿam- (< ʿammāl) with present tense is used:

190

Werner Arnold

(11) Aramaic: ʿamšō€ mōya Arabic: ʿambišrab mayy ‘He is drinking water.’ Verbs of Arabic origin are treated like Aramaic verbs. They are supplied with the same inflectional affixes as Aramaic and carry the same object suffixes. In fact only the radicals of the Arabic root are borrowed and treated as an Aramaic root. Under the influence of the Arabic dialects in the surrounding villages Baxʿa has lost gender distinction in all plural forms of the verb. The masculine form was generally adopted. The following paradigm gives the preterite tense of the basic stem of the verb i_̣ḥek ‘to laugh’: (12) Coll. Arabic _̣aḥak _̣aḥak-at _̣aḥak-u _̣aḥak-t _̣aḥak-ti _̣aḥak-tum _̣aḥak-tin _̣aḥak-t _̣aḥak-na

Aramaic (Maʿlūla) i_̣ḥek _̣iḥk-a€ i_̣ḥek _̣iḥk-ič _̣iḥk-iš _̣iḥk-ičxun _̣iḥk-ičxen _̣iḥk-i€ _̣iḥk-innaḥ

‘he laughed’ ‘she laughed’ ‘they laughed’ ‘you (m sg) laughed’ ‘you (f sg) laughed’ ‘you (m pl) laughed’ ‘you (f pl) laughed’ ‘I laughed ‘we laughed

The Arabic verbal roots are integrated into the WNA system of verbal stems exactly like Aramaic roots within the traditional Aramaic system of verbal stems. This is the case with the following Arabic stems: (13) I stem II stem IV stem V stem

Arabic loan iḍḥeḳ ‘he laughed’ ḥammel ‘he loaded’ aġreḳ ‘he fell asleep’ čḥammal ‘he endured’

Aramaic word išmeʿ ‘he heard’ baššel ‘he cooked’ arkeš ‘he woke up’ čzappan ‘he was sold’

All Arabic stems can be incorporated into the Aramaic stem system. Arabic stems, which do not correspond to an Aramaic stem are converted in the following way:

Western Neo-Aramaic

(14) III stem VI stem VII stem VIII stem X stem

Arabic šāraṭ tarāfaq infaǧar iftaham istaqbal

WNA > šōreṭ > črōfeḳ > inᵊfžar > if ᵊčham > sčaḳbel

191

‘to bet’ ‘to accompany’ ‘to explode’ ‘to understand’ ‘to accept’

The Arabic VII. and VIII. stems can be formed also from Aramaic roots to express the passive to the Aramaic I. stem and have replaced with some few exceptions the old Aramaic passive stem E§pʿel: (15) Aramaic —f€ḥ

I stem VII stem Aramaic —nġb I stem VIII stem

if€aḥ inᵊf€aḥ inġab inᵊčġab

‘to open’ ‘to be opened’ ‘to steal’ ‘to be stolen’

On the other hand the Aramaic passive stem ččaCCaC (< ettaCCaC) can also be formed from Arabic roots: (16) Aramaic —wrx awrex ččawrax

Arabic —wqf awḳef ‘to prolong’, ‘to erect’ ččawḳaf ‘to be prolonged’, ’to be erected’

Many borrowings of the Arabic I stem appear in Western Neo-Aramaic in the IV stem. The reason for this change is unclear, but it should be mentioned that the IV stem in Western Neo-Aramaic is very productive whereas in the Arabic dialects of Syria the IV stem has disappeared nearly completely: (17) Arabic I stem bada > Aramaic IV stem abᵊt

to begin

5. Other parts of speech All ordinal numerals are borrowed from Arabic and have with the exception of awwal ‘first’ the Arabic imāla ā > ē (€ēni ‘second’, €ēle€ ‘third’, rēbeʿ ‘fourth’, etc.). Western Neo-Aramaic borrowed from Arabic the reflexive ḥōl- (< ḥāl) which is used beside Aramaic nefš-. Furthermore the Arabic reciprocal baʿ_̣-

192

Werner Arnold

is borrowed from Arabic. All of them are used with the Aramaic pronominal suffixes. The dativus ethicus is attested in Western Neo-Aramaic and in the Arabic dialects of the area and may be a result of language contact: (18) Aramaic _mex-le šaʿ€a

Arabic nām-lo sāʿa ‘He slept himself for an hour.’

The Arabic coordinating conjuctions fa ‘and so’, walla ‘or’ and lakin (also lakan, lakinni, lakōn; in Jubbʿadīn lačin, ličin, līčin) or bass for ‘but’ are borrowed from Arabic. The subordinating conjunction _ōb ‘if’ is a calque of Arabic inkān in which in is translated by Aramaic _ and kān by the Aramaic equivalent ōb (Spitaler 1938: 117). The Arabic word is also used in the form nkōn. Other borrowings are iza and law ‘if’, innu (also inne and inni) ‘that’ and lamma ‘when’ (beside Aramaic mi_), illa ‘except’ and others. For the times of the day only _̣ahw€a (< ḍaḥwa) ‘late morning’ and ʿaṣᵊr ‘afternoon’ are of Arabic origin. The days of the week are preserved in Maʿlūla and Jubbʿadīn, but they are replaced by the Arabic names in Baxʿa. Adjectives with the Arabic prefix m-, the infix -č- (< Arabic -t-) and the suffix -ōnay (< -āni) expressing affiliation (amrikōnay ‘American’) were incorporated into the Aramaic vocabulary and among young speakers have sometimes replaced the inherited Aramaic form: (19) Aramaic adjective: išmeʿ ‘audible’ Young speakers: mašmuʿ (Arabic masmūʿ) Old Aramaic has no pattern to express elative terms so that WNA was forced to realize comparative and superlative forms by adoption of the Arabic morphological pattern ʾaCCaC: (20) Arabic ṣaġīr aṣġar minn-o aṣġar wāḥid

WNA izʿur ‘small’ azʿar menn-e ‘smaller than he’ azʿar aḥḥa_ ‘the smallest’

6. Syntax The relative particle ti (also či, Baxʿa ći, all < *dī) is Aramaic but the way of subordinating relative clauses is fully compatible with Arabic. In syndetic

Western Neo-Aramaic

193

relative clauses the antecedent of the relative clause is determined and followed by the relative particle while asyndetic relative clauses have no relative particle and the antecedent is indetermined. In old Aramaic such asyndetic relative clauses are unattested (Correll 1978: 117) and must be considered as borrowed from Arabic. (21) a. Asyndetic wō€ rōʿya ʿamraʿēl ʿizzōye. ep-pret shepherd-indet herd-3sg.pres.pm goat-pl-sf.3sg.m ‘There was a shepherd, (who was) herding his goats.’ b. Syndetic hanna ġamla ti ṭʿille. dp.3sg.m camel rp carry-perf.3sg.m-sf.3sg.m ‘This camel, which has carried him.’ The coordination of circumstantial clauses (Arabic ḥāl) by means of the conjunction w- (and) is not attested in old Aramaic but occurs frequently in WNA and in the modern Arabic dialects of the area (Grotzfeld 1965: 101). Correll (1978: 147) argues that this construction is borrowed from the Arabic dialects as no other Aramaic dialect has similar circumstantial clauses. (22) w hū ʿammallex 4 willa €ōle sōblᵊ cp ip-3sg.m walk-3sg.m.pres-pm lo! come-3sg.m-pret mayor-cs blōta. village ‘While he was walking, lo! – the mayor of the village came.’

7. Conclusion Aramaic and Arabic are two closely related Semitic languages with a long common history. The fact that morphemes in the Semitic languages normally consist of three radicals facilitates the mutual adoption and assimilation of loans, but makes the identification of borrowings from one to the other very difficult. The Arabic dialects spoken in the surroundings of the WNA speech island themselves are very much influenced by Aramaic and sometimes preserve Aramaic words which no longer occur in WNA. For many centuries this type of aramaicized Arabic was the only contact language of WNA, and Correll (1978) believes that the conservatism of Arabic is the reason for the

194

Werner Arnold

archaic structure of WNA in comparison to the eastern Aramaic dialects, where under the influence of Turkish and Iranian languages the old verbal system collapsed. WNA has enriched its vocabulary with thousands of words from Arabic and has incorporated the Arabic system of verbal stems, but the integration of Arabic borrowings was performed by a total adaptation to the Aramaic phonological and morphological system. In general, WNA has preserved its linguistic heritage or has developed independent from Arabic.

Abbreviations c cp cs det dp ep f indet ip m

consonant (radical) coordination particle construct state determined demonstrative pronoun existential particle feminine indetermined independent personal pronoun masculine

perf pl pm pres pret rp sf sg subj WNA

perfect tense plural progressive modifier present tense preterite relative pronoun suffix singular subjunctive Western Neo-Aramaic

Notes 1. In the inflexion of the verb, gender distinction in plural forms is also lost. 2. This is not the case with eastern Aramaic dialects. 3. In Jubbʿadīn the auxiliary verb bēl- of unclear origin is used.

References Arnold, Werner 1990 Das Neuwestaramäische. V. Grammatik (Semitica Viva 4/V), Wiesbaden. 2000 The Arabic dialects in the Turkish province of Hatay and the Aramaic dialects in the Syrian mountains of Qalamūn: Two minority languages compared. In: Jonathan Owens (ed.), Arabic as a Minority Language, 347370. Berlin, New York. 2002 Zur Geschichte der arabischen Lehnwörter im Neuwestaramäischen. In: Norbert Nebes (ed.), Neue Beiträge zur Semitistik. Erstes Arbeitstreffen der Arbeitsgemeinschaft Semitistik in der Deutschen. Morgen-

Western Neo-Aramaic

195

ländischen Gesellschaft vom 11. bis 13. September 2000 an der Friedrich-Schiller-Universität Jena. Ed. by (Jenaer Beiträge zum Vorderen Orient 5), Wiesbaden, 511. Arnold, Werner, and Peter Behnstedt 1993 Arabisch-Aramäische Sprachbeziehungen im Qalamūn (Syrien). Eine dialektgeographische Untersuchung mit einer wirtschafts- und sozialgeographischen Einführung von Anton Escher (Semitica Viva 8), Wiesbaden. Correll, Christoph 1978 Untersuchungen zur Syntax der neuwestaramäischen Dialekte des Libanon (Maʿlūla, Baḫʿa, Ǧubbʿadīn); mit besonderer Berücksichtigung der Auswirkungen arabischen Adstrateinfllusses; nebst zwei Anhängen zum neuaramäischen Dialekt von Ǧubbʿadīn (= Abhandlungen für die Kunde des Morgenlandes XLIV, 4), Wiesbaden. Grotzfeld, Heinz 1965 Syrisch-arabische Grammatik (Dialekt von Damaskus) (Porta linguarum orientalium, Neue Serie VIII). Wiesbaden. Spitaler, Anton 1938 Grammatik des neuaramäischen Dialekts von Maʿlūla (Antilibanon). (Abhandlungen für die Kunde des Morgenlandes XV, 4). Leipzig.

Grammatical borrowing in North-eastern Neo-Aramaic Geoffrey Khan

1. Background Aramaic, which was one of the major Semitic languages in the pre-Islamic Middle East, still survives today in various vernacular dialects. These NeoAramaic dialects can be divided into four main subfamilies, which include (1) the Western group spoken in Maʿlūla and various other villages in the region of Damascus, (2) the Ṭuroyo group, spoken in Ṭūr ʿAbdīn in southeastern Turkey and in the village of Mlaḥsō in southern Turkey, (3) Mandaic, spoken in the city of Ahwāz, Iran, and the surrounding region, and (4) the north-eastern group. North-Eastern Neo-Aramaic (NENA)1 contains a far greater diversity of dialects than any of the other groups. These were spoken across a wide area encompassing northern Iraq, north-western Iran, south-eastern Turkey, Armenia and Georgia. A large proportion of the speakers of these dialects, however, have been supplanted from their original places of residence due to political events during the twentieth century and now live in a diaspora of émigré communities in various parts of the world. On account of this, many of the dialects are now facing extinction. The NENA group includes dialects spoken by Jews and Christians. The Jewish dialects in all cases differ from the Christian dialects, even where the Jews and Christians lived in the same town or region. There are considerable differences, for example, between the Jewish dialect and Christian dialect in the towns of Urmi, Salamas, Sanandaj and Sulemaniyya, in which the two communities lived side by side. In other geographical areas, such as Zakho and the surrounding region, the differences between the dialects of the two communities are of a lesser degree.2 These dialectal cleavages between confessional communities appears to have evolved not only through social divisions but also through different migration histories. Where Jewish and Christian communities existed side by side in towns such as those mentioned above, in some cases it can be established that the settlement of the Jews in the town was earlier than that of the Christians, who were more recent immigrants from the villages in the surrounding countryside.

198

Geoffrey Khan

The NENA dialects exhibit extensive grammatical borrowing from the non-Semitic languages with which they have been in contact for many centuries. The main source of borrowing is Kurdish, an Iranian language that was spoken in numerous dialects across the NENA region. Some features have been borrowed also from Arabic and Turkic languages. Grammatical borrowing is very diverse across the NENA group and much of it still remains unstudied. For this reason I shall here take as a case study one particular dialect of the group that exhibits widespread borrowing and has been fully described, viz. the Jewish dialect of Sulemaniyya (North West Iraq).3 There was a large Aramaic speaking Jewish community in Sulemaniyya since the foundation of the town in 1784 by Ibrāhīm Pāshā Bābān. The majority of the Jews came from the village of Qaradax, situated twenty miles to the south. Between 1950 and 1952 the entire community, which consisted of approximately 500 households, migrated to Israel. Today the dialect is still spoken by only a few elderly immigrants in Israel and is likely to be totally extinct within the next few years. The Muslims of Sulemaniyya are almost entirely Kurdish speaking, with a small minority of inhabitants whose first language is Turkmen. The Jews of the town spoke the local Kurdish dialect in order to communicate with the Muslim inhabitants. The source of the grammatical borrowing in the Aramaic dialect can in some cases be identified as the Sulemaniyya Kurdish dialect. In other cases, however, it appears to have arisen by contact with other Kurdish dialects or other Iranian languages in more remote regions. Sometimes the source is Arabic or a Turkic language. Some of these features may have been transferred to the Sulemaniyya Jewish Aramaic dialect indirectly through other Jewish Aramaic dialects. A few elements have been borrowed also from Hebrew, the traditional language of Jewish education in the community.

2. Phonology A number of phonological changes in the Aramaic dialect appear to be contact-induced. In the Aramaic dialect, for example, the original interdental s € and _ shift to the lateral l:4 (1)

*be€a > bela ‘house’ *ʾida > ʾila ‘hand’

North-eastern Neo-Aramaic

199

This shift is found in all Jewish NENA dialects that were spoken in communities east of the Zab river (the so-called ‘trans-Zab group’).5 The shift of *€ to the lateral /l/ seems to have been preceded historically by a shift to the voiced stop /d/. This is shown by the fact that in some dialects in which the shift to /l/ has taken place the reflex of *€ has remained as /d/ in some words. For instance, in the Jewish dialects of Urmi, Ruwanduz and Rustaqa the root ‘to come’ is ʾdy (< *ʾ€y), e.g. ʾidyele ‘he came’. The process resulting in the lateral /l/ described above, therefore, can be regarded as resulting from the weakening of the articulation of /d/. Interdentals are absent in Sulemaniyya Kurdish and post-vocalic /d/ is to an interdental approximant /đ/ (MacKenzie 1961: 3, 8) but there is no shift of dentals to the lateral l. Such a shift, however, can be identified in the Mukri dialect of Kurdish spoken in north-western Iran (Kapeliuk 1997). The phoneme /w/ often has a labio-dental realization [v] in the Aramaic dialect but the corresponding phoneme in the local Kurdish is always realized as a bilabial. The labio-dental realization is a development found in Farsi and also in the Aramaic dialects of western Iran. In the Aramaic dialect an /l/ in the environment of pharyngalized consonants sometimes shifts to /r/: (2)

parṭixwa < palṭixwa ‘We used to go out.’

This has been found only in the speech of women informants. A parallel phonological shift is attested in the Kurdish speech of some women from Sulemaniyya and regularly in the Kurdish dialects spoken in Arbel, Koy Sanjaq and Ruwanduz (MacKenzie 1961: 4, 28). After a vowel, /d/ and /z/ may freely alternate with each another. This process appears to be restricted to specific lexical items and is best treated as an alternation of phonemes rather than as allophonic alternation: (3)

ʾidyo ~ ʾizyo ‘today’ qadome ~ qazome ‘tomorrow’ guda ~ guza ‘wall’

In the speech of the closely related Aramaic dialect of the Jews of Ḥalabja the three-way alternation d ~ _ ~ z is sometimes heard, with the intermediate interdental:

200

Geoffrey Khan

(4)

xădir ~ xă_ir ~ xăzir ‘he becomes’ guda ~ gu_a ~ guza ‘wall’

The weakening of the stop /d/ to an interdental /¦/ or sibilant /z/ may have been stimulated by contact with the Kurdish dialects of the region, in which the articulation of /d/ is weakened to an interdental approximant in postvocalic position. The shift of etymological /z/ to /d/ would, therefore, have to be regarded as a back-formation. It is also relevant to note that the alternation d ~ _ ~ z is found in the Jewish Persian dialects of western Iran.6 The stress patterns of the Aramaic dialect are similar to those that are found in the Sulemaniyya Kurdish dialect. In both dialects stress is generally placed at the end of a word. Exceptions to this tend to be found in the same categories of words: (5)

Vocative Aramaic táta ‘father!’ Kurdish mā´mwastā ‘teacher!’ Past verbs Aramaic híyen ‘They came.’ Kurdish hátin ‘They came.’ Stress groups combining two or more words Aramaic tré-yome ‘two days’ Kurdish dé-rož ‘two days’

3. Nominal structures The Jewish Aramaic dialect disinguishes two genders (masculine and feminine). This contrasts with the Kurdish dialect of Sulemaniyya, which makes no morphological distinction in gender. Under the influence of the Kurdish dialect, however, the original gender distinction in the third-person singular pronoun of the Aramaic dialect has been lost: (6)

Aramaic Kurdish ʾo aw ‘he/she/it’

The demonstrative adjectives in Aramaic have also lost the distinction between singular and plural in imitation of Kurdish:

North-eastern Neo-Aramaic

(7)

201

Aramaic Kurdish ʾay am ‘this/these’ ʾo aw ‘that/those’

Many nouns have been borrowed by the Aramaic dialect from Kurdish. Since the Kurdish dialect makes no gender distinctions, they have been assigned a gender in the Aramaic dialect. The grammatical gender of all loanwords that refer to human beings corresponds to the sex of the referent. The majority of loanwords that refer to inanimate objects or small animals are construed as feminine in gender, e.g. šāx (f.) ‘mountain’, qali ‘carpet’ (f.), jiji (f.) ‘hedgehog’ The same applies to loans from Arabic which have entered the dialect through Kurdish, e.g. kĭteb (f.) ‘book’, majlis (f.) ‘meeting’. There is a sizeable residue of inanimate loans that are construed as masculine in gender. The gender assignment of these appears to have a semantic basis, in that most of the nouns in question either denote (i) a long, thin entity, e.g. qamīš (m.) ‘cane’, top (m.) ‘gun’ or (ii) a collective or non-solid entity, .e.g. xaḷūz (m.) ‘coal’, čay (m.) ‘tea’. Many borrowed nouns are adapted to Aramaic morphology by adding an Aramaic nominal ending. In most cases the Aramaic masculine ending -a is added, irrespective of the gender assignment: (8)

boqa (f.) ‘frog’ < Kurdish boq lăša (m.) ‘body’ < Kurdish laš

In a few cases the Aramaic feminine ending -ta is added to Kurdish loans that have been assigned feminine gender: (9)

masita (f.) ‘fish’ < Kurd. masi dargušta (f.) < Kurd. darguš

The Aramaic dialect uses a definite article suffix -ăke that has been borrowed from Kurdish. In the local Sulemaniyya Kurdish dialect the form of the article is -aka. The form -ăke is likely to have its origin in a form of the particle with an oblique case marker -aka-y, which is found in Kurdish dialects lying to the North and North East of Sulemaniyya (MacKenzie 1961: 5758). Furthermore, the morphological behaviour of the particle in the Aramaic dialect is different from that of the particle in Kurdish. In the Aramaic dialect it cannot take any further suffixes. In Kurdish, by contrast, it may take plural (-ān) and pronominal suffixes:

202

Geoffrey Khan

(10) Aramaic barux-ăke barux-awal-ăke barux-an

Kurdish dost-aka ‘the friend’ dost-akān ‘the friends’ dost-aka-mān ‘our friend’

Possessive constructions in which a head noun is qualified by a determiner noun are normally formed by juxtaposing the two. The Aramaic genitive particle d-, which regularly occurs in such constructions in some NENA dialects, is rarely used. The construction without the d- corresponds to a functionally equivalent construction in Sulemaniyya Kurdish whereby a head noun is linked to a following determiner noun by means of a compound vowel -a (MacKenzie 1961: 64): (11) Aramaic Kurdish brona mălik kur¯-a pāšā

‘the son of the king’

The final -a of the Aramaic noun is the general nominal ending, but in this environment it is identified with the Kurdish compound vowel -a. The Iranian Ezāfe particle is occasionally used in genitive constructions when the head noun is a loanword: (12) maktab-i hulaye ‘school of the Jews’

4. Verbal structures In both the Aramaic and Kurdish dialects the copula verb ‘to be’ is expressed by an enclitic on the predicate that is inflected for person and number like a verb. Although there are signs of the emergence of such a enclitic in earlier Eastern Aramaic, its full development and acquisition of verbal inflection are apparently due to the influence of Kurdish: (13) Aramaic Kurdish ʾo jwan-ya aw jwān-a ‘She is beautiful’ ʾat jwan-yat to jwān-ī ‘You are beautiful’ The simple past tense of the Aramaic dialect has acquired an ergative type of inflection by contact with Kurdish. The intransitive verbs have active subject inflection, which resembles the inflection of present tense verbs, whereas

North-eastern Neo-Aramaic

203

transitive verbs have a passive type of inflection, with the verb agreeing with the patient of the action and the agent expressed by an oblique agentive pronominal affix that does not serve as the grammatical subject of the verb. The oblique pronominal affix is identical to the pronominal object affix of the present form of the verb. These structures correspond to what is found in the Sulemaniyya Kurdish dialect: (14) Second person Aramaic Simple past intransitive mīl-et ‘You (ms.) died’ mīl-at ‘You (fs.) died’ mīl-etun ‘You (pl.) died’ Simple past transitive qṭil-lox ‘You (ms.) killed him’ qṭil-lax ‘You (fs.) killed him’ qṭil-laxun ‘You (pl.) killed him’ Kurdish Simple past intransitive mird-ī ‘You (s.) died’ mird-in ‘You (pl.) died’ Simple past transitive kušt-it ‘You (s.) killed him’ kušt-tān ‘You (pl.) killed him’

Present mel-et ‘You (ms.) die’ mel-at ‘You (fs.) die’ mel-etun ‘You (pl.) die’ Present + pronominal object qăṭil-lox ‘He kills you (ms.)’ qăṭil-lax ‘He kills you (fs.)’ qăṭil-laxun ‘He kills you (pl.)’ Present a-mir-ī ‘You (s.) die’ a-mir-in ‘You (pl.) die’ Present + pronominal object a-t-kužē ‘He kills you (s.)’ a-tān-kužē ‘He kills you (pl.)’

In addition to this simple past tense, consisting of a past stem and inflectional endings, the Aramaic dialect also has a compound perfect consisting of a passive participle and an enclitic form of the verb ‘to be’. The same two types of past conjugation are found in the Kurdish dialect: (15)

Aramaic Simple past: mīl ‘He died’ Compound past: mila-y ‘He has died’

Kurdish mird ‘He died’ mirduw-a ‘He has died’

Although there are clear contact-induced resemblances between the Aramaic and Kurdish dialects, a closer look reveals several differences. In the Jewish

204

Geoffrey Khan

Aramaic dialect, for example, the agentive pronominal suffix on the simple past transitive verb is bonded to the verb. In Kurdish, on the other hand, this agentive suffix is moveable and is generally attached to a word that occurs earlier in the clause. The transitive predication ‘the man killed the dog’, for example, is expressed in the two dialects as follows: (16) Aramaic gor-ăke kalb-ăke qṭil-le. man-the dog-the killed-past:3ms-3ms:obl. Kurdish pyāw-aka sag-aka-y kušt. man-the dog-the-3ms:obl killed-past:3ms The compound past tense in the Kurdish dialect has an ergative form of inflection whereas the corresponding conjugation in the Jewish Aramaic dialect has an active type of inflection in both the intransitive and transitive verbs: (17) Aramaic gor-ăke mila-y. man-the died-cop:3ms ‘The man has died.’ gor-ăke kalb-ăke qiṭl-u-y. man:the dogs.the killed-3pl:obl-cop:3ms ‘The man has killed the dogs’ Kurdish pyāw-aka mirduw-a. man-the died-cop:3ms ‘The man has died.’ pyāw-aka sag-akān-ī kuštuw-in. man:the dogs.the-3ms:obl. killed-cop:3pl ‘The man has killed the dogs.’ It is worth noting that in a few Jewish NENA dialects in neighbouring western Iran, the compound past verbal form is ergative in inflection. This is the case, for example, in the Jewish dialects of Sanandaj and Kerend, e.g.

North-eastern Neo-Aramaic

205

(18) Jewish Sanandaj gor-ăke kalb-ăke qiṭl-en. man:the dogs.the killed-cop:3pl ‘The man has killed the dogs.’ The Kurdish dialect of Sulemaniyya does not, in fact, have a pure ergative system regarding case-marking and verb agreement in either of the past tenses, since it is lacking an ergative case marker on the agent noun. The agent is not marked with an oblique case but is extraposed and resumed by an oblique pronominal element. This feature is shared by the Jewish Aramaic dialect of Sulemaniyya, but the Aramaic dialect has moved even further away from the ergative system than the Kurdish one. In the Aramaic dialect not only is the compound past conjugation inflected actively but even the simple past conjugation is beginning to be treated as an active form. One reflection of this is that the patient of the action is sometimes given oblique inflection. This is regularly the case, for example, when the object is a first- or second-person pronoun: (19) qṭil-le ʾillox. killed-3ms:obl 2ms:obl. ‘He killed you.’ One linguistic innovation that is found in the verbal system of the Jewish Aramaic dialect is the expression of the present progressive by an infinitive combined with a enclitic form of the verb ‘to be’: (20) šaqola-yet. taking-cop:2ms ‘You are taking.’ This appears to have developed from a construction in which the infinitive is combined with the locative preposition b-, which is still preserved in some NENA dialects, e.g. Christian Urmi: (21) bi-šqala-vit. in-taking-cop:2ms ‘You are taking.’

206

Geoffrey Khan

No direct parallels to this infinitive construction are found in Sulemaniyya Kurdish, or indeed in other Kurdish dialects but there are close parallels to the Aramaic construction in languages spoken further north, such as Turkish, Eastern Armenian (Chyet 1995: 246) and some Iranian dialects belonging to the Tati group spoken in north west Iran (e.g. Chali, cf. Yar-Shater 1969: 225), e.g. Turkish almak-ta-sın ‘You are taking’ (taking-loc-cop:2S.); Chali xordan-u-ind ‘They are eating’ (eating- loc-cop.3pl).

5. Other parts of speech A large number of other grammatical words have been borrowed from Kurdish. These include a variety of particles and adverbials, including subordinating clausal conjunctions such as ʾagar ‘if’, taku ‘in order that’, nakun ‘lest’ and the relative particle ga, connectives such as ʾinja ‘then’, ham ‘also’, yan ‘or’, čunga ‘because’, modal particles such as the volitive particles ba and mar expressing deontic modality in verbs, the phasal particle heštan ‘still’, ‘yet’, the comparative particle biš ‘more’, the negative modifier of nouns hič ‘no’, and a variety of adverbials, e.g. čannakaw ‘suddenly’, dubara ‘again’, har ‘always’. The Kurdish post-verbal particle awa (after a vowel wa) is widely used in the Aramaic dialect after Aramaic verbs. It often expresses the sense of ‘returning back’, ‘restoring’, ‘repetition’ or ‘completion’: (22) hiye-wa. come:past:3ms-particle ‘He came back.’ (23) qadome der-awa. tomorrow return:fut:3ms-particle ‘He will return tomorrow.’ The Kurdish inclusive particle -iš has been borrowed. In the Aramaic dialect, however, the particle is not as integrated into the morphology as it is in Kurdish. This is reflected by the fact that it is always an external affix in word-final position in the Aramaic dialect, whereas in Kurdish it precedes a pronominal suffix (MacKenzie 1961: 128): (24) Aramaic nošew-iš

Kurdish xō-š-i ‘also himself’

North-eastern Neo-Aramaic

207

Some particles in the Aramaic dialect are borrowed from Kurdish but do not correspond exactly to what is found in Sulemaniyya Kurdish. On some occasions we seem to be dealing with doublets. This applies, for example, to the particles gal ‘with’ and laga ‘at the home of’ in the Aramaic dialect, both of which appear to be related to the particle lagaḷ in the Sulemaniyya Kurdish dialect. A particle of Aramaic original may imitate the use of an Aramaic particle. This applies, for example, to the Aramaic particle k- which is attached to present base verbs in the indicative mood. In the present state of the dialect the particle has, in fact, been lost by phonetic attrition, but is still preserved in some verbs. The etymology of the particle is clearly Aramaic (cf. Babylonian Talmudic Aramaic qā- < qāʾem ‘rising up’), but its use is likely to be an imitation of the use preverbal present indicative particles in Kurdish, in the case of Sulemaniyya Kurdish this is a-.

6. Syntax The basic word order of the Jewish Aramaic dialect is SOV when the verbal arguments are free-standing nominals, which corresponds to that of the Kurdish dialect (see examples 16 and 17). Various borrowed subordinating particles are used to introduce subordinate clauses in the Aramaic dialect. The Kurdish subordinating particle ka-, which is used in the Kurdish dialect of Sulemaniyya and the surrounding region (MacKenzie 1961: 131), has been borrowed by the Aramaic dialect (pronounced either ka- or ga-) to introduce various types of subordinate clause. Attributive relative clauses are often introduced by this Kurdish particle. The native Aramaic relative particle d- has been completely lost in relative clauses after head nouns. As in Kurdish, the relative clause follows its head: (25) ʾo-baxta ka-xăzitta ga-doka šwawt-i-ya. that-woman who-see:2ms in-there neighbour-1s:obl-cop:3fs ‘The woman whom you see there is my neighbour.’ The Kurdish particle ka- is sometimes used in the Aramaic dialect as a subordinating temporal conjunction with the sense of ‘when’: (26) ga-wīš-wa, ʾanten-wa-le. when-dried:3ms-pastp took:3pl:imp-pastp-3ms:obl ‘When it had dried, they took it.’ (R:40)

208

Geoffrey Khan

The particle ka- is also used as a complementizer to introduce factive complement clauses, mainly after the verb ‘to know’: (27) kăyen-wa ga-ʾo brata ʾil-d-o-brona know:imp:3pl-pastp comp-that(dem) girl objm-that(dem)-boy gba. loves:imp:3fs ‘They would know that the girl loves the boy.’ The protasis of a conditional constructions in Aramaic is generally introduced by the Kurdish particle ʾagar ‘if’: (28) ʾagar la-gbitta,ʾana kun-naw ta-naš xet. if neg-love:imp:3ms:subj-3fs obl to-person other ‘If you do not love her, I shall give her to some other person.’ (R:165) Temporal ‘when’ clauses are frequently introduced by the Kurdish nominal waxta ‘time’: (29) waxta ʾana menna, ʾay tre-brone ruwwe, băxew la-koli time I die-1ms dem two-sons big care neg-do:fut:3pl l-yal-ax. objm-children-2fs.obl ‘When I die, these two older boys will not look after your children.’ (R:17) The Arabic particle ḥatta, or its variant form hatta with a laryngeal, is used as a clausal conjunction, generally with the sense of ‘until’ and introduces an event that marks the endpoint of the activity or situation denoted by the main clause: (30) rĭqa-le bari ba-ṣiwa ḥatta ran:past-3ms.obl after-1s.obl with-stick until di-le-lli. hit:past-3ms.obl-1s.obl ‘He ran after me with a stick until he beat me.’ Syntactically subordinate clauses that have a concessive sense are generally introduced by the Hebrew particle afillu:

North-eastern Neo-Aramaic

209

(31) ʾafillu ʾo barux-i-ye, ʾana yaridew although he friend-1s.obl-cop.3ms. I help-3ms.obl la-kunna. neg-do:fut.1ms ‘Although he is my friend, I shall not help him.’

7. Lexicon There is a high degree of lexical borrowing in the Aramaic, especially in the category of nouns. On the basis of a sample corpus of material, the broad assessment of the proportion of loanwords in the lexicon of the various categories of words is as follows: nouns, 67%, adjectives, 48%, particles, 53%, verbs, 15%. The vast majority come from Kurdish. A few originate in other languages, such as Arabic and Turkish, though most of these are likely to have been transmitted through Kurdish. The Arabic loans sometimes exhibit features that are distinctive of Arabic lexemes in Kurdish, such as the pronunciation of the tā’ marbūṭa as -at (e.g. ḥukmat ‘government’, sa‘at ‘hour’) and the use of broken plurals with a singular sense (e.g. tujār ‘merchant’). The Aramaic dialect exhibits also many calques of Kurdish expressions. Verbs constitute the most resistant category to loaning, as is commonly the case in language contact situations. The greater facility with which nouns are loaned is clearly demonstrated by the so-called ‘phrasal verbs’, which consist of an inflected verbal form combined with a noun or particle as its complement. These are nearly all based on a Kurdish model. In most cases, however, only the noun or particle complement is a direct loan of a Kurdish word. The verb is an Aramaic calque of the Kurdish: (32) Aramaic rek lpl swal ʾwl sayr ʾwl gargurke ʾwl xafad ʾxl gorane ʾmr wa dmy

Kurdish rek kawtin swal kirdin sayr kirdin gargurke kirdin xafat xuwardin gorani wittin wā zānīn

‘to agree’ ‘to beg’ ‘to look at’ ‘to crawl’ ‘to be distressed’ ‘to sing’ ‘to think’

Borrowed verbs, by contrast, are inflected fully with the Aramaic verbal inflection. The existence of a rich inflectional morphology in verbs is no doubt

210

Geoffrey Khan

one reason why the verbal section of the lexicon has been more resistant than nouns to borrowing. One of the few verbs in the Aramaic dialect that have been loaned from Kurdish is dyy (< Kurd. dān). This is used in various phrasal verbs that are based on Kurdish models, e.g. (33) Aramaic bāz dyy čirike dyy čapḷe dyy

Kurdish bāz dān ‘to jump’ čirike dān ‘to shout’ čapḷe dān ‘to clap’, ‘applaud’

The Kurdish verb dān has a wide range of meanings, including ‘to give’, ‘to hit’, ‘to put’. The corresponding Aramaic form, dyy, regularly occurs in the Aramaic phrasal verbs that have Kurdish models with dān. The distribution of dyy in the Aramaic dialect outside of this context, however, is not as wide as that of Kurdish dān. When used independently of phrasal expressions, the verb dyy in Aramaic most commonly has the sense of ‘to hit’. It is not used in the meaning of ‘to give’, which is one of the basic senses of the Kurdish source word dān. The Aramaic dialect retains the native verb to express this meaning (present indicative: kul, past: hiwle) and so has resisted a complete lexical transfer. The verb borrowed from Kurdish in some cases has taken on the full basic meaning of the Aramaic equivalent but nevertheless the dialect retains the Aramaic verb and uses it in a slightly different meaning. This applies, for example, to the fate of the Aramaic verb prx. In earlier Aramaic *prḥ had the sense of ‘to fly’ and, indeed, this sense is still retained by the verb in some NENA dialects. In the Jewish dialect of Sulemaniyya, however, the meaning ‘to fly’ is expressed by the Kurdish loanword fry (< Kurd. firīn). The native verb prx, nevertheless, is still retained in the sense of ‘to jump over’, which is connected conceptually with the notion of ‘flying’. On some occasions the contact with Kurdish has not brought about a direct borrowing of lexical items but rather as given rise to the development of a phonetic resemblance between morphological forms in the two languages. In the Aramaic dialect, for example, the verb ‘to come’ (< *ʾ€y) undergoes irregular phonetic contraction and loses its original middle radical *€ completely. As a result, the base of the present conjugation of the verb resembles phonetically the corresponding Kurdish verb: (34) Aramaic k-e (3ms indicative) k-en (3pl indicative)

he (3ms subjunctive) hen (3pl subjunctive)

North-eastern Neo-Aramaic

Kurdish e (3s indicative) en (3pl indicative)

211

b-e (3s subjunctive) b-en (3pl subjunctive)

One may perhaps include here the irregular loss of the final /m/ in the word ʾidyo ‘today’ (< *ʾid-yom) under the influence of the vocalic pattern of the Kurdish equivalent imro. Another case of phonetic convergence is the Aramaic form gbe, which functions as an invariable verb form expressing necessity. The form is derived from the Aramaic verb ‘to want’. Its function is identical to that of the Kurdish invariable verb abe (MacKenzie 1961: 106), which it resembles phonetically, e.g. (35) Aramaic gbe hezex

Kurdish abe biçīn

‘We must go’

Some lexical items in the Aramaic dialect are non-Semitic loanwords yet do not correspond to what is found in the local Kurdish dialect of Sulemaniyya. This is the case, for example, with the two basic kinship terms in the Jewish Aramaic dialect tata ‘father’ and lala ‘maternal uncle’. They are not found in the local Kurdish dialect, but parallels can be found in several other languages of the region, the nearest being Hawrami, spoken in the Hawraman mountains. Similarly a loanword in the Aramaic dialect may be identified in the Kurdish dialect of Sulemaniyya but with a different meaning, whereas it is used with the same meaning as is found in the Aramaic dialect in a more remote linguistic source. This applies to the word baba, which, in the Aramaic dialect of Sulemaniyya, usually means ‘grandfather’ rather than ‘father’. In the Kurdish dialect of Sulemaniyya the cognate word has the sense of ‘father’ (bāb, bāw), but in Hawrami it is used with the sense of ‘grandfather’ as in the Aramaic dialect: (36) Jewish Sul. Ar. tata lala baba

Sulemaniyya Kurdish bāw, bāwk xal ba-pīr

Hawrami tata ‘father’ lalo ‘maternal uncle’ baba ‘grandfather’

On some occasions a Kurdish loanword in the Aramaic dialect has a different range of meaning from what it has in the source language. The word čolăka, for example, is used in the Aramaic dialect with the meaning of ‘bird’. In

212

Geoffrey Khan

Kurdish it has the specific meaning of ‘sparrow’, the general term for bird’ being mal.7 The loanword mayīn is used in the Aramaic dialect as a general word for ‘horse’, whereas in the Kurdish source language it refers specifically to a ‘mare’. Most surviving speakers of the Aramaic dialect who have been resident in Israel since the 1950s use Hebrew words in their Aramaic speech. A large proportion of these words are taken from Modern Hebrew. Particularly common are the connective particles ʾaz ‘then’ and ʾaval ‘but’. As is the case with lexical transfer from Kurdish, Hebrew verbs are not so freely borrowed. Speakers prefer to form verbal phrases containing a Hebrew nominal element and an Aramaic verb: (37) mazmīn koli-wa-le. invite-heb.participle do-imp.3pl-pastp.3ms.obl ‘They would invite him.’ Certain Hebrew words that occur in the informants’ speech, however, existed in the Aramaic dialect before the immigration of the Jewish community to Israel. These can usually be distinguished by phonetic features that are not characteristic of Modern Hebrew. In such words, for example, vocalic šewa is pronounced /ă/, as in băraxa ‘blessing’, nădaba ‘charity’, băli ‘without’. Consonantal gemination is pronounced, as in ʾafillu ‘even if’, keʾillu ‘as if’, sukka ‘booth’. Beth is pronounced as a stop where it is a fricative in Modern Hebrew, as in tob ‘good’, kabod ‘respect’. In some cases a Hebrew word has undergone a phonetic process under the influence of Kurdish, which demonstrates that it is a heritage from the Aramaic dialect as it was spoken in Kurdistan, e.g. mira < mila ‘circumcision’.

Abbreviations comp cop dem fut imp loc neg

complementizer copula demonstrative future imperfective locative negator

obj objm obl pastp pres subj

object object marker oblique past particle present subject

North-eastern Neo-Aramaic

213

Notes 1. The term was coined by Hoberman (1988: 557) to replace ‘Eastern Neo-Aramaic of earlier classifications (cf. Socin 1882: v; Duval 1896: 125; Tsereteli 1977, 1978). This was necessary in order to distinguish the north-eastern dialects from modern Mandaic, which is as distant typologically from them as the western Neo-Aramaic dialects. 2. Cf. Hopkins (1993: 65). 3. The dialect is described in Khan (2004). All the data in this chapter are taken from this description, which is based on extensive fieldwork undertaken in Israel with Jewish immigrants from Sulemaniyya. 4. The transcription follows that adopted in Khan (2004). Vowel length is largely predictable from the syllable structure. As a general rule, in open syllables a vowel is long and in closed syllables it is short. Vowels following this principle are not marked by diacritics. The breve and macron signs are used only when the vowel is short in an open syllable or long in a closed syllable respectively. 5. For the main characteristics of the trans-Zab group, see Mutzafi (2004: 910). 6. I am grateful to Don Stilo (personal communication) for drawing my attention to these phenomena in the Kurdish and Jewish Iranian dialects. 7. This is similar to the semantic relationship between Arabic ʿuṣfūr ‘sparrow’ and Hebrew ṣippor ‘bird’, which are cognates.

References Chyet, M. L. 1995 Neo-Aramaic and Kurdish: An interdisciplinary consideration of their influence on each other. Israel Oriental Studies 15: 219249. Duval, R. 1896 Notice sur les dialectes néo-araméens. Mémoires de la société de linguistique de Paris 9: 125135. Hobermann, R. D. 1988 The history of the Modern Aramaic pronouns and pronominal suffixes. Journal of the American Oriental Society 104: 221231. Hopkins, S. 1993 The Jews of Kurdistan in Eretz Israel and their language. Peʿamim. Studies in Oriental Jewry 56: 5074. Khan, G. 2004 The Jewish Neo-Aramaic Dialect of Sulemaniyya and Halabja. Leiden: Brill. MacKenzie, D. N. 1961 Kurdish Dialect Studies, Volume 1. London: Oxford University Press.

214

Geoffrey Khan

Mutzafi, H. 2004 Socin, A. 1882

The Jewish Neo-Aramaic Dialect of Koy Sanjaq. Wiesbaden: Harrassowitz.

Die neu-aramäischen Dialekte von Urmia bis Mosul. Texte aund Übersetzungen. Tübingen: Laupp. Tsereteli, K. G. 1977 Zur Frage der Klassifikation der neuaramäischen Dialekte. Zeitschrift der Deutschen Morgenländischen Gesellschaft 127: 244253. 1978 The Modern Assyrian Language. Moscow: Nauka. Yar-shater, E. 1969 A Grammar of Southern Tati dialects. The Hague: Mouton.

Grammatical borrowing in Macedonian Turkish Yaron Matras and Şirin Tufan

1. Background The variety described here is representative of the Turkish dialects spoken in the Republic of Macedonia, especially those in the west of the country, and to a considerable extent also of Rumelian or Balkan Turkish as a whole (cf. e.g. Matras 1998, 2004; Friedman 2003). The Balkan or Rumelian dialects of Turkish descend directly from Ottoman Turkish and are generally considered mutually comprehensible with Standard Turkish (henceforth ‘Tk.’); there are even direct historical links with Anatolian Turkish (cf. Caferoğlu 1964). We draw here primarily on data from the dialect of Gostivar, a city in the western part of the Republic of Macedonia – henceforth GT for ‘Gostivar Turkish’ (for a comprehensive description see Tufan 2007).1 Turkish is the native language of the Turkish ethnic minority in the various Balkan countries. It is the first language of many Muslim Romani communities, and it is also spoken by some Albanians, Macedonians, and other ethnicities as a second or third language.2 As the official language of the Ottoman Empire, Turkish was a lingua franca and the language of administration and trade in the Balkans for more than half a millennium (between the fourteenth and early twentieth century). With the final collapse of the Ottoman Empire (1912), Turkish became a minority language. In Macedonia, it was not until the 1950s that its status became regulated and Turkish-language education, cultural institutions, and media received state backing. The form of Turkish taught at school was Standard Turkish, while the vernacular continued to be used in the private domain. Turkish speakers in the region are generally biand often trilingual, speaking, in western Macedonia, alongside the state language, also Albanian. Over the past century, and especially since the 1950s, the importance of the state language and its relevance to career progression, education, and mobility has grown immensely, and this is reflected in the amount and the nature of Macedonian lexicon that has found its way into the local varieties of Turkish. In today’s Republic of Macedonia, Turkish speakers have direct contact with Standard Turkish not only through schooling, but also through satellite television and the internet, which are present in almost every Turkish household. Successive waves of emigration to Turkey in recent

216

Yaron Matras and Şirin Tufan

decades have further fortified personal ties with Turkey, and visits to Turkey are frequent, resulting in even greater exposure to Standard and Anatolian Turkish.

2. Phonology Among the consonants, we find the dental-alveolar affricate /ts/, which has its source in Macedonian and Albanian. It is found not only in loanwords (Albanian-derived tsapo ‘goat’, Macedonian-derived tsevka ‘pipe’) and in borrowed affixes (Macedonian feminine-agentive -itsa), but it is also transferred occasionally into native Turkic words: tsıs ‘shut up’ (cf. Tk. sus). Initial consonant clusters are permitted in GT which do not appear in Tk.: GT (also Macedonian and Albanian) Stambol ‘Istanbul’, Tk. İstanbul. There are, on the other hand, also cases of simplification. The surrounding non-Turkic languages simplify Turkish geminates in Turkish borrowings (cf. Friedman 2003: 58), and this trend is also found in GT: /dükan/ ‘shop’, Tk. /dükkan/ ‘shop’; /akıli/ 'clever', Tk. /akıllı/. As in the neighbouring languages, there is a weakening of /h/, though the origins of this development in Western Rumelian Turkish are thought to be in the features carried by immigrants from northeast Anatolia (Németh 1956: 21): GT /ayvan/ ‘animal’, Tk. /hayvan/; GT /paali/ ‘expensive’, Tk. /pahalı/; GT /saba/ ‘morning’, Tk. /sabah/. Recent contact with Standard Turkish appears to have triggered the re-introduction of /h/, and variation is commonly found, especially in grammatical function words such as /em, hem/ ‘and’, /er, her/ ‘every’, or /ep, hep/ ‘all’. In line with the absence of vowel-length distinctions in both Macedonian and Albanian, there is a tendency in GT to shorten ‘double’ or ‘lengthened’ vowels, which appear in Turkish in loans of Persian and Arabic origin: thus /galiba/ ‘probably’ (Tk. /ga»liba/), /hala/ ‘yet’ (Tk. /ha»la/). The loss of /ö/ – which does not exist in the contact languages – may also be a contact-induced phenomenon. In GT historical /ö/ is usually realized as /ü/ or as /o/: GT ürenci ‘student’, Tk. öğrenci; GT dort 'four', Tk. dört.

3. Nominal structures The feminine derivational markers -ka and -(i)tsa are borrowed from Macedonian, and are productive with Turkish word stems: arkadaş ‘friend’ (gender-neutral, and by default masculine), arkadaş-ka ‘female friend’; koyşi

Macedonian Turkish

217

‘neighbour’, koyşi-ka ‘female neighbour’; yalanci ‘liar’, yalanci-tsa ‘female liar’. The suffix -(i)tsa is further extended to denote a female affiliated with an identified male, thus: dayo ‘maternal uncle’, day-tsa ‘maternal uncle’s wife’; Muzafer-itsa ‘Muzaffer’s wife’. The extended distribution of the inherited diminutive suffix -çe appears to be influenced by the presence of a similar form in the neighbouring languages: kış-çe ‘little girl’, Macedonian devoj-če. The case of the dependent in possessive constructions is also affected by contact. The possessor often appears in the ablative case, still accompanied, as in Tk., by possessive inflection on the object of possession (the head), while in Tk. the possessor appears in the genitive: (1)

a. Gostivar Turkish kıskardeş-i güvegi-den sister-3sg.poss groom-abl ‘the groom’s sister’ b. Standard Turkish damad-ın kız kardeş-i groom-gen sister-3sg.poss ‘the groom’s sister’

The construction seems to copy the propositional marking of the possessor in Macedonian, which appears either in the ablative or dative: (2)

Macedonian a. sestra-ta na zet-ot sister-def to groom-def b. od zet-ot sestra-ta from sister-def groom-def ‘the groom’s sister’

4. Verbal structures The copula in GT appears, like in Macedonian, as an independent verb, and not, as in Tk., in an enclitic form. Since this concerns issues of constituent order, the position of the copula will be discussed further in Section 6. A characteristic feature of the verb in Rumelian Turkish is the loss of the modal infinitive, and the reduction of converbal forms in general. As a strategy of clause linkage, this issue is discussed in Section 7 on ‘Syntax’.

218

Yaron Matras and Şirin Tufan

Gostivar Turkish continues the general Turkish pattern of forming new verbs by incorporating lexical nouns from the contact language, and integrating them with a light verb which differentiates valency. Both et- ‘do’ and yap- ‘make’ are employed with transitives, and ol- ‘become’ with intransitives: yaparsın komparatsiya 'you compare', privatizir oldi ‘it was privatized’. Idiomatic structures are often copied as loan-blends, involving Matter replication of a Macedonian noun, accompanied by a translation of the Macedonian verb: rutina alayim ‘I shall get into the habit’, lit. ‘I shall take a routine’, Macedonian da zemam rutina.

5. Other parts of speech A number of conjunctions and particles are borrowed from Macedonian and Albanian. Matras (2004) notes for the Turkish dialects of eastern Macedonia that the Macedonian additive conjunction i is regularly used when conjoining phrases, while Turkish ve is limited to conjoining constituents (as in example 3). Note that the adversative conjunction ama is identical in Turkish and Macedonian, Macedonian having borrowed it from Turkish. The Slavic contrastive-addition marker a indicates opposition between two phrases: (3)

İlk-okul-i ve orta-okul-i bitır-dı-m Türkçe first-school-acc and middle-school-acc finish-past-1sg Turkish dil-ın-de, a fakulted-i bitır-dı-m language-poss-loc and/however university-acc finish-past-1sg Makedonce dil-ın-de. Macedonian language-poss-loc ‘I finished primary and secondary school in Turkish, but university in Macedonian.’

Another use of a is for disjunction: (4)

Amerika a Alman yatırım-i dır. America or German investment-poss is ‘It is an American or German investment.’

It is possible that this function results from a blend between the Macedonian contrastive-additive a, and the Albanian-derived question particle a, which is also borrowed into GT:

Macedonian Turkish

(5)

219

A git-tı-n Stambol-a? q go-past-2sg Istanbul-dat ‘Have you been/ did you go to Istanbul?’

The Albanian requestive particle lu(te)m is also borrowed: (6)

Gel benım-le lum. come me-inst req ‘Please come with me.’

Subordinating conjunctions are mainly grammaticalized interrogatives and thus of Turkish origin, but the presence of kose ‘as if’ seems to indicate a contamination of Macedonian kako ‘as if’ and Albanian kinse ‘as if’, possibly reinforced by the similarity to the Turkish conditional verbal augment -se. Possibly, an Albanian model sepse ‘because’ is also behind the use of se as a subordinator of cause (‘because’).

6. Constituent order Although on the whole still an SOV language, flexibility of word order in Turkish is exploited in GT to extend pragmatically restricted variants to wider contexts, thereby increasing harmony between GT and its contact languages in the organization of utterance structures. Word order shift has acquired different degrees of stability with different constructions. In the possessive construction, the order head–modifier has become the preferred order in GT, mirroring the order in the Macedonian and Albanian constructions: (7)

a. Gostivar Turkish ruba-lar-i damad-ın clothes-pl-3sg.poss groom-gen b. Macedonian ališta-ta na zet-ot clothes-def to groom-def c. Albanian teshat e dhandrit clothes att groom

220

Yaron Matras and Şirin Tufan

d. Standard Turkish damad-ın eşya-lar-ı groom-gen clothes-pl-3sg.poss ‘the groom’s clothes’ The object of comparison is expressed in GT with the help of a preposition neka ‘like’, grammaticalized from the interrogative ne kadar ‘how much’, copying the Macedonian preposition kolku ‘as much’. It is positioned, as in Macedonian, between the attribute and the object of comparison: (8)

a. Gostivar Turkish güzel neka Meryem beautiful like Meryem b. Macedonian ubava kolku Merjem beautiful like Meryem c. Standard Turkish Merye kadar güzel Meryem as.much beautiful ‘as beautiful as Meryem’

This is the only obvious indication of a shift, in any construction, from the postpositional structure of Turkish, to prepositions. In verb phrases, the most stable case of word-order convergence with the neighbouring languages concerns the position of the copula. Whereas the Turkish copula is enclitic, GT tends to preserve a more conservative independent copula stem in i-, which, however, occupies the position between the subject and the predicate noun, as in the contact languages: (9)

a. Gostivar Turkish Sen (i)-sın küçük bir kış-çe. you cop-2sg small indef girl-dim b. Macedonian Ti si edno malo devoj-če. you cop.2sg indef small girl-dim c. Albanian Ti je nji vajz ë vogël. you cop.2sg indef small att girl

Macedonian Turkish

221

d. Standard Turkish Sen küçük bir kız-sın. you small indef girl2sg ‘You are a small girl.’ This is the general rule in the copula construction, irrespective of the word class or case of the predicate (e.g. adjective, locative noun, etc.): (10) Siz i-dı-nız ev-de. you cop-past-2pl house-loc ‘You were at home.’ In other constructions, deviation from verb-final order is much less pragmatically marked, and much more frequent, than in colloquial Tk., indicating a drive toward harmonization of the utterance planning procedures with those of the contact languages. Consider the following sentences, in which direct and indirect objects follow the verb without any inference of de-focusing or de-topicalization (which would be the reading accompanying such constructions in Tk.): (11) Ben gür-dü-m korkuli rüya. I see-past-1sg scary dream ‘I saw a scary dream.’ (12) Ben ver-dı-m bikate ekmek sizın dort tene I give-past-1sg little bread 2pl.poss four item beygir-ınız-e. horse-2pl.poss-dat ‘I gave your four horses some bread.’ (13) Onlar gid-ecek-ler dügün-e benım-le. they go-fut-3.pl wedding-dat me-inst ‘They will go with me to the wedding.’ The default position for objects that constitute new topical information in lexical predications remains, however, the pre-verbal position:

222

Yaron Matras and Şirin Tufan

(14) Parlament-ın-de var-dır iki dil. parliament-3sg-loc exist-cop.3sg two language Arnaut-lar Arnautçe konuş-ur. Albanian-pl Albanian speak-aor Makedon-lar Makedonce konuş-ur. Macedonian-pl Macedonian speak-aor Azınlık-lar Makedonce konuş-ur. minority-pl Macedonian speak-aor Bir tek Arnaut-lar Arnautçe konuş-ur. one only Albanian-pl Albanian speak-aor ‘In Parliament, there are two languages. The Albanians speak Albanian. The Macedonians speak Macedonian. The minorities speak Macedonian. Only the Albanians speak Albanian.’

7. Syntax Some of the most remarkable changes that have affected Rumelian Turkish – a characteristic feature of this group of Turkish dialects – is the adoption of clause combining strategies that are similar to those employed in the surrounding Indo-European languages. Essentially, these are based on the juxtaposition of finite clauses, linked through independent semantic markers that introduce the subordinate clause (subordinating conjunctions). This system replaces almost entirely the Turkic system of converbs and nominal embedding. Modal complements are not introduced by a conjunction, but make use of the historical optative, which, now expressing dependency on the main verb, serves as a subjunctive, with the complement clause generally following the main clause (see also Matras 1998, 2004): (15) a. Gostivar Turkish Yarın ist-er-ım oyna-(ya)-im dügün-de. tomorrow want-aor-1sg play-subj.1sg wedding-loc b. Macedonian Utre saka-m da igra-m na svadba-ta. tomorrow want-1sg comp play-1sg at wedding-def

Macedonian Turkish

223

c. Albanian Nesër dua të luj në darsëm. tomorrow want.1sg comp play.1sg in wedding d. Standard Turkish Yarın düğün-de oyna-mak isti-yor-um. tomorrow wedding-loc play-inf want-prog-1sg ‘I want to dance at the wedding tomorrow.’ The finite embedded predicate in the subjunctive replaces the historical Turkish infinitive. The same type of construction is used in manipulation clauses (modal complements with different subjects): (16) Daa çok sev-er-ım anlat-ır-sın kimse. more much like-aor-1.sg tell-aor-subj.3sg somebody ‘I prefer somebody to narrate it [to me].’ Factual or epistemic complements, which in Tk. may be expressed through either finite clauses, or nominalizations, always appear as postposed finite clauses, introduced by the subordinator ki, which is also common in Tk.: (17) Hised-ıl-mes ki vardır sonbaar. feel-pass-neg.aor comp exist.cop.3sg autumn ‘It does not feel like autum.’ In this manner, GT aligns itself with the other Balkan languages also in respect of the distinction between factual and non-factual complements. While the other languages have complements that specialize for factual/epistemic and non-factual/subjunctive (e.g. Macedonian deka vs. da, Greek oti vs. na, Bulgarian če vs. da, Romani kaj vs. te, and so on), in GT the opposition is expressed by using the inflected subjunctive on the verb in modal complements, and the ki complementizer (and indicative mood) in epistemic complements. Relative clauses also undergo re-structuring in Rumelian Turkish. Like the other Rumelian Turkish dialects, GT shows a relativizer ne, derived from the interrogative ‘what’, which mediates between the head noun and the finite, postposed relative clause (see Matras 1998, 2004). This replaces both the Turkish gerundial relative clause, and its finite counterpart in ki. The formation once again matches that of the principal contact language Macedonian, where the relativizer is equally derived from the interrogative ‘what’:

224

Yaron Matras and Şirin Tufan

(18) a. Gostivar Turkish O kış-çe ne gel-di biz-de şimdi yaşa-r that girl-dim rel come-past 1pl-loc now live-aor.3sg Stambol-da. Istanbul-loc b. Macedonian Devoj-če-to što dojde kaj nas sega živee vo İstanbul. girl-dim-def rel came at us now live.3sg in Istanbul c. Standard Turkish Biz-e gel-en kız şimdi İstanbul-da yaşı-yor. 1pl.dat come-ger girl now Istanbul-loc live-prog.3sg ‘The girl that came to (visit) us now lives in Istanbul.’ Like relative clauses, embedded clauses in GT are finite, usually postposed to the main clause, and introduced by an interrogative, functioning as a conjunction; Turkish-type nominalizations of embedded propositions are not found. Adverbial clauses show a mixed pattern in relation to convergence tendencies. One type of adverbial clause shows an overwhelming tendency to copy the Indo-European subordination type: postposed finite subordinate clauses introduced by a conjunctions. To this end, a series of grammaticalization processes take place giving rise to new subordinating conjunctions. The semantic relations involved in clause combinations of this type are those of time (introduced by açin ‘when’ in GT, or by ne zaman ‘when’ in other dialects of Macedonian Turkish), location (introduced by nerde ne ‘where’ < lit. ‘where what’, cf. Macedonian kade što lit. ‘where what’), reason (introduced by niçin ‘because’ < ‘what-for’, cf. Macedonian zošto lit. ‘for-what’), manner (introduced by kose ‘as if’, possibly a contamination of Macdeonian kako ‘how’, Albanian kinse ‘as if’, and Turkish -se ‘if’; see above), and comparison (introduced by neka ne ‘as much as’ < ne kadar ne ‘how much what’, cf. Macedonian kolku što). Purpose clauses and final clauses are equally finite, and show the verb of the subordinated clauses in the subjunctive. They are introduced respectively by the complementizer ki, directly reinforcing the subjunctive (cf. Macedonian prepositional reinforcer za da), and the conjunction çaki ‘until’. A second type of clause linkage remains largely unaffected by contactinduced restructuring. This involves conditional clauses (‘If I pass my exam my dad will buy me a bicycle’), and concessive clauses (‘Although I want to go to Antalya, I won’t be able to go’). Both are marked by the conditional marker -se (on its own for conditional clauses, with addition of de or hem ‘too’ for concessive clauses), which is added to a finite subordinated clause.

Macedonian Turkish

225

Thus, where Turkish already operates with finite subordinations, there appears to be no motivation to re-organize the structure of clause linking.

8. Lexicon Despite multilingualism in the region where GT is spoken, lexical borrowing is predominantly from Macedonian, reflecting the growing importance of the state language in the past two to three generations especially. Lexical borrowing from Macedonian naturally affects in the first instance semantic areas belonging to the public domain, such as names for institutions (teatar ‘theatre’, fakultet ‘faculty’, univerzitet ‘university’, vodovod ‘water board’, kanalizatsiya ‘infrastructure’, militsiya ‘police’, armiya ‘army’, ordinatsiya ‘dental surgery’, klinika ‘clinic’, autobuska stanitsa ‘bus station’), terms for practitioners and professionals (elektriçar ‘electrician’, stomatolog ‘dentist’, sestra ‘nurse’, apotekarka ‘pharmacist’, direktor ‘director’, student ‘student’, privatnik ‘having business in the private sector’), academic subjects and professions (meditsina ‘medicine’, farmatsiya ‘pharmacy’, stomatologiya ‘dentistry’, ispit ‘exam’, praktiçno ‘practical exam’, poen ‘mark’, matura ‘graduation’), construction and technology (kuyna ‘kitchen’, patos ‘flooring’, şifunyer ‘cabinet’, parno ‘central heating’, radiator ‘radiator’, garaja ‘garage’, satelitska ‘satellite’, kaseta ‘cassette’, elektrika ‘electricity’), as well as miscellaneous domains (maçka ‘cat’, şatka ‘duck’, sok ‘fruit juice’, zvuçnost ‘sound’, spetsiyalizatsiya ‘specialization’, tragediya ‘tragedy’, etc.).

9. Conclusion It is interesting to note once again that Turkish has only been a minority language in Macedonia for some three to four generations now. The fact that Matter borrowing is limited to a rather small number of discourse particles and conjunctions, may be a reflection of this recent retreat of Turkish from public life, and its replacement, to a considerable degree, by Macedonian. The lexicon, of course, reflects the recent dominance of Macedonian-speaking society in the public domain, employment, technology, and so on. Nevertheless, the restructuring of clause combining strategies based largely on a Macedonian model constitutes a radical departure from the Turkic syntactic type, and it is most certainly much older than the retreat of Turkish as the language of the public domain. Rather, the changes in this domain reflect

226

Yaron Matras and Şirin Tufan

century-old multilingualism. It appears that in daily communication, speakers were under pressure to organize complex utterances in a compatible way across the various languages that constituted their linguistic repertoire. What is essentially an economy-driven motivation – reducing multiple pattern types across the linguistic repertoire to just one – might be understood as a harmonization of utterance-organization strategies (Matras 2004). The two areas that are most obviously affected are clause combining strategies, and to a somewhat lesser extent, word order. With the former, it is the packaging of supplementary information through finite subordinations that prevails, and to this end a series of grammaticalization processes are triggered, exploiting elements of the inherited lexicon, often following the Macedonian model (‘replica grammaticalization’ in the terms of Heine and Kuteva 2005). The latter, word order, involves harmonization of strategies of mapping information status at the level of the linear organization of the utterance. Here, some constructions, such as possessive noun phrases and existential (copula) predications, appear more vulnerable to the pressure toward harmonization than others. Nevertheless, even word order in the basic verb phrase shows a partial relaxation of the pragmatic constraints on the appearance of postverbal objects. This in turn provides an extended scope to employ such constructions, which resemble the word order rules of the contact language. We may speculate that it was possible for pattern-replication of this type to emerge in the vernacular language long before Turkish retreated as the official language of the public sphere: it exploited constructions that pre-existed, to some extent at least, in colloquial usage, such as semi-embedded finite optative constructions (see discussion in Matras 1998 and 2004), or finite subordinations introduced with ki, or pragmatically-marked constructions involving de-topicalization of direct and indirect objects (in post-verbal position). Pattern replication was thus a kind of compromise, allowing speakers to maintain language loyalty while assisting the levelling of certain language processing strategies within the multilingual repertoire. We suggest that this latter aspect is a crucial component of the history of linguistic areas, for which the Balkans have long served as a prototype example.

Abbreviations abl acc aor

ablative accusative aorist

att cop dat

attributive marker copula dative

Macedonian Turkish def dim fut gen indef inf inst loc neg

definite article diminutive future genitive indefinite marker infinitive instrumental locative negation

past pl poss prog q rel req sg subj

227

past tense marker plural possessive progressive interrogative particle relative particle requestive singular subjunctive

Notes 1. Examples are taken from Tufan’s fieldwork in Gostivar; observations are based partly on fieldwork data collected by Matras among speakers from Stip. 2. Figures or even estimates of numbers of speakers in the entire region are difficult to obtain. Ethnic Turks in the Republic of Macedonia itself number around 70,000.

References Caferoğlu, Ahmet 1964 Anadolu ve Rumeli ağızları ünlü değişmeleri. TDAYB. 133. Friedman, Victor A. 2003 Turkish in Macedonia and Beyond. Wiesbaden: Harrassowitz Heine, Berndt, and Tania Kuteva 2005. Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Matras, Yaron 1998 Convergent development, grammaticalization, and the problem of ‘mutual isomorphism’. In: Winfried Boeder, Christoph Schroeder, and Karl-Heinz Wagner (eds.) Sprache in Raum und Zeit, 89103. Tübingen: Narr. 2004 Layers of convergent syntax in Macedonian Turkish. Mediterranean Language Review 15: 6386. Németh, Gyula 1956. Zur Einteilung der Turkischen Mundarten Bulgariens. Sofia: Bulgarische Akademie der Wissenschaften. Tufan, Şirin 2007 Language convergence in Gostivar Turkish (Macedonia). PhD thesis, University of Manchester.

Grammatical borrowing in Kildin Saami Michael Rießler

1. General The present chapter deals with borrowings in Kildin-Saami.1 The principal contact language for Kildin is Russian, and has been so at least since the end of the Middle Ages. There are no grammatical borrowings detected from any languages other than Russian. However, many of the contact phenomena dealt with may also be traced in the other Kola Saami languages Akkala, Skolt, and Ter, which are facing the same contact-linguistic environment or have even been exposed to stronger assimilation pressure. But since Kildin is the best documented and the most accessible of the Kola Saami languages this investigation will be restricted to observations on this language. Most of the data used for this investigation have been taken from existing descriptions of Kildin. Examples given without reference come from my own field notes and have been cross-checked with native speakers. The orthographic representation of examples follows the standards set in the dictionary written by Kuruč, Afanasjeva, and Mečkina (1985).

1.1. Linguistic background Saami is a branch of the Uralic language family. All Saami languages are fairly similar in grammatical structure and lexicon. They form a dialect chain stretching from central and northern Scandinavia to the eastern tip of the Kola Peninsula. Kildin belongs to the group of East Saami languages. The other subgroups of Saami are Central Saami (spoken in the northern parts of Finland, Sweden and Norway) and South Saami (spoken in Central Scandinavia), each of which includes several languages. One characteristic of the phonology of most Saami languages is the occurrence of preaspirated voiceless stops and affricates [p, t, k, ¿, ʧ]. Negation in Saami is expressed by means of an inflected negation auxiliary followed by the non-finite main verb in a special connegative form. Phrase structure in Saami is for the most part head-final, including the predominant occurrence of postpositions instead of prepositions and strict head-finality in

230

Michael Rießler

noun phrases with noun, adjective, and pronoun modifiers. Relative clauses, however follow the noun they modify. In the verb phrase a shift from SOV to SVO word order seems to be taking place. The change in the order of verb and direct object as well as the introduction of prepositions and relative clause constructions are probably contact-induced. However, these changes go back to Common Saami tendencies and are subsequently not dealt with in the present investigation. One Kola Saami characteristic – as compared to the western Saami languages – is the relatively large consonant inventory, which is mostly due to the fact that almost all consonants have a phonologically distinct palatalized counterpart. As for the nasal and lateral dentals /n/ and /l/ there is an opposition not only to the respective palatalized phonemes, but also to the nasal and lateral palatals /ɲ/ and /ʎ/; consider the minimal triples mānn / maÃnÃ/ ‘moon; month’ – mannҍ /manÃ/ ‘egg’ – mann' /maɲÃ/ ‘daughter in law’ and pāll /paÃlÃ/ ‘ball’ – māll' /maÃlÃ/ ‘juice’ – māll'j /maÃʎÃ/ ‘rust’. The existence of a phonological opposition between palatal and palatalized consonants seems to be uncommon cross-linguistically (cf. Stadnik 2002: 31, elsewhere).

1.2. Sociolinguistics and geography Kildin is currently spoken on the Kola Peninsula in the northwestern-most part of the Russian Federation by no more than 700 people. The language is endangered due to language shift to Russian and is hardly ever heard in public life nowadays. Only elder Saami use their mother tongue in conversation with family members, relatives or friends. Among the younger generation, there is a strong decline in active language competence due to the lack of a vibrant speech community and the lack of any social motivation for learning and using Saami (an overview on the current socio-political situation of the Saami in Russia is given in Scheller 2006). The integration of the Kola Saami into the Russian Empire, their adaptation to Russian culture, and their conversion to Orthodox Christianity began as early as the fifteenth century. Nevertheless, despite the longstanding assimilation pressure, the territorial communities of the Saami were able to preserve their social, economic and cultural identity – at least in the central and northern areas – until the end of the nineteenth century. By the first half of the twentieth century, the Saami culture was on the verge of destruction. The tapping of mineral resources and the military armament of the region

Kildin Saami

231

were connected with an immense influx of manpower from Russia and other republics of the Soviet Union. The dissolution of the traditional Saami communities – as the result of forced integration of Saami reindeer herders into large new agricultural co-operatives – and the resettlement of the Saami for socio-political, economic and military reasons led to a dispersion of the original speech communities. The former compact Saami settlements and coherent local speech communities were replaced by mixed communities of Saami speaking different local varieties, together with non-Saami (above all Komi and Russians). As a result, within a few decades, the indigenous Saami people became a tiny, scattered minority without any great influence on political decisions (for an overview on Kola Saami history and further references see Kulonen, et al. 2005: 261265).

1.3. Contact induced change in Saami Along with the changes in Kola Saami culture and society, we witness a number of contact-induced linguistic changes. These changes – almost exclusively of Russian origin – concern nearly all domains of the grammar but are especially strong in the lexicon. Still, contact-induced change in Kola Saami has not yet been the subject of a systematic investigation. Certain contactinduced features of Russian origin are mentioned in works on Kola Saami, for example in Kert's (1971) grammar of Kildin. Russian influence on Kola Saami is also the subject of a published conference abstract by Klaus (1977) dealing with borrowed adverbs and the use of Russian numerals in Saami speech. The most considerable listing of contact-induced features in Kildin is found in a paper by Kert (1994). But even here it is mostly lexical borrowings which are dealt with. Only a few grammatical features are listed briefly by the author, such as borrowed phonemes in loanwords, the borrowed superlative particle, and borrowed function words (Kert 1994: 112). All these features are dealt with in more detail in the respective sections below. Another phonological feature which Kert believes to be borrowed from Russian is palatalization (Kert 1994: 111). This idea is shared by Stadnik (2002: 34, 165; without reference to Kert). Neither of the two authors, however, offer any explanations for the contact-linguistic mechanisms behind the proposed development. Instead of being caused by borrowing I find it much more reasonable to assume that Kildin Saami palatalization is triggered by another, probably language-internal development, namely the apocope of the reduced stem-final vowels.

232

Michael Rießler

2. Phonology Kildin has borrowed some phonemes along with Russian loanwords. In most cases, however, the phonological distinctiveness of these phonemes is weak since they only occur in loanwords and there are almost no real minimal pairs available. Consider, for example, the innovative voicing opposition in wordinitial plosives, sibilants, and the labio-dental fricative, i.e. (original) /p, t, k, ʃ, s, v/ – (innovative) /b, d, g, ʒ, z, f/, cf. purrk /purÃk/ ‘lower part of the reindeer's antlers’ – būrka /buÃrka/