Bilingualism and Migration 9783110807820, 9783110163698

Citation preview

Bilingualism and Migration

1749

I

1999

S

Studies on Language Acquisition 14

Editor

Peter Jordens

Mouton de Gruyter Berlin · New York

Guus Extra and Ludo Verhoeven (Editors)

Bilingualism and Migration

W DE

G

Mouton de Gruyter Berlin · New York

1999

Mouton de Gruyter (formerly M o u t o n , T h e Hague) is a Division of Walter de Gruyter & Co., Berlin.

The series Studies on Language Acquisition was formerly published by Foris Publications, Holland.

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

Library of Congress

Cataloging-in-Publication

Data

Bilingualism and migration / edited by Guus Extra, Ludo Verhoeven. p. cm. — (Studies on language acquisition ; 14) ISBN 3-11-016369-1 (alk. paper). - ISBN 3-11-016370-5 (pbk. alk. paper) 1. Bilingualism. 2. Immigrants—Language. 3. Second language acquisition. 4. Language maintenance. 5. Language attrition. I. Extra, Guus. II. Verhoeven, Ludo T h . III. Series. P115.B5455 1998 404'.2-dc21 98-45107 CIP

Die Deutsche Bibliothek

— Cataloging-in-Publication

Data

Bilingualism and migration / ed. by Guus Extra ; Ludo Verhoeven. — Berlin ; New York : Mouton de Gruyter, 1998 (Studies on language acquisition ; 14) ISBN 3-11-016369-1 gb. ISBN 3-11-016370-5 brosch.

© Copyright 1998 by Walter de Gruyter G m b H &C Co., 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. Printing: WB-Druck, Rieden /Allgäu. Binding: Lüderitz & Bauer, Berlin. Printed in Germany.

Acknowledgements

This volume is the follow-up of a conference on bilingualism and migration, held in February 1996 at Tilburg University, the Netherlands, in the context of a Dutch Science Foundation research program on processes of language change in a migration context. We want to thank the DSF/NWO (Nederlandse Organisatie voor Wetenschappelijk Onderzoek) for its long-term financial support of this research program, and Carine Zebedee and Gwen Perret (Tilburg University) for their editorial support of this volume. Guus Extra and Ludo Verhoeven

List of contributors

Jeroen Aarssen, Tilburg University, Research Group on Language and Minorities, PO Box 90153, NL-5000 LE Tilburg, the Netherlands Ruth A. Berman, Tel Aviv University, Department of Linguistics, Ramat Aviv 69978, Israel Ad Backus, Tilburg University, Research Group on Language and Minorities, PO Box 90153, NL-5000 LE Tilburg, the Netherlands Petra Bos, Tilburg University, Research Group on Language and Minorities, PO Box 90153, NL-5000 LE Tilburg, the Netherlands Kees de Bot, Nijmegen University, Research Group on Language and Communication, PO Box 9103, NL-6500 HD Nijmegen, the Netherlands Louis Boumans, University of Leiden, Centre for Non-Western Studies, PO Box 9515, NL-2300 RA Leiden, the Netherlands Sally Boyd, University of Göteborg, Department of Linguistics, Renströmsparken, S-41298 Göteborg, Sweden Annick De Houwer, University of Antwerpen, Department of Politics and Social Sciences, Universiteitsplein 1, B-2610 Antwerpen, Belgium Abder El Aissati, Tilburg University, Research Group on Language and Minorities, PO Box 90153, NL-5000 LE Tilburg, the Netherlands Guus Extra, Tilburg University, Research Group on Language and Minorities, PO Box 90153, NL-5000 LE Tilburg, the Netherlands Hanneke van der Heijden, Centre for Educational Services, Section PO, PO Box 8639, NL-3009 AP Rotterdam, the Netherlands Lars Johanson, Johannes Gutenberg-Universität, Seminar für Orientkunde, Turkologie, Postfach 2980, D-6500 Mainz, Germany Sirkku Latomaa, University of Tampere, Department of Finnish Language and General Linguistics, PO Box 607, F-33101 Tampere, Finland Pieter Muysken, University of Leiden, Department of Linguistics, PO Box 9515, NL-2300 RA Leiden, the Netherlands Carol Myers-Scotton, University of South Carolina, Program in Linguistics, Columbia, SC 29208, USA Carol W. Pfajf, Freie Universität Berlin, JFK Institut, Lansstrasse 5-9, D14195 Berlin, Germany Suzanne Romaine, University of Oxford, Merton College, UK-Oxford, OX1 4JD, United Kingdom Anneli Schaufeli, School of Education, Groenewoudseweg 1, NL-6524 TM Nijmegen, the Netherlands

viii

List of contributors

Arturo Tosi, University of London, Department of Italian, Royal Holloway, Engham, UK-Surray TW20 OEX, United Kingdom Ludo Verhoeven, Nijmegen University, Research Group on Special Education, PO Box 9104, NL-6500 HE Nijmegen, the Netherlands Ake Viberg, Lund University, Department of Linguistics and Phonetics, Helgonabacken 12, S-22362 Lund, Sweden

Contents

Part 1: Introduction Guus Extra and Ludo Verhoeven Immigrant minority groups and immigrant minority languages in Europe Guus Extra and Ludo Verhoeven Processes of language change in a migration context: the case of the Netherlands

3

29

Part 2: Early bilingual development Suzanne Romaine Early bilingual development: from elite to folk

61

Annick De Houwer Environmental factors in early bilingual development: the role of parental beliefs and attitudes

75

Carol W. Pfqff Changing patterns of language mixing in a bilingual child

97

Hanneke van der Heijden Word formation processes in young bilingual children

123

Part 3: Bilingual development at school age Jeroen Aarssen Bosdevelopment Cohesive devicesand in Petra bilingual

143

Ake Viberg Lexical development in a second language

165

Ruth A. Berman Bilingual proficiency / proficient bilingualism: insights from narrative texts

187

χ

Contents

Part 4: Codeswitching and borrowing Carol Myers-Scotton Compromise structural strategies in codeswitching

211

Pieter Muysken Three processes of borrowing: borrowability revisited

229

Lars Johanson Frame-changing code-copying in immigrant varieties

247

Ad Backus The intergenerational codeswitching continuum in an immigrant community

261

Louis Boumans Codeswitching and the organisation of the mental lexicon

281

Part 5: Language maintenance and language loss Sally Boyd theory and Sirkku Latomaa Fishman's of diglossia and bilingualism in the light of language maintenance and shift in the Nordic Region

303

Arturo Tosi The notion of 'community' in language maintenance

325

Kees de Bot The psycholinguistics of language loss

345

Abder El Aissati and Anneli Schaufeli Language maintenance and loss: evidence from language perception and production

363

Index

379

Part 1 Introduction

Immigrant minority groups and immigrant minority languages in Europe Guus Extra and Ludo Verhoeven

After an outline of major keywords in the European public discourse on immigration is presented, the theme of immigrant minority (IM) groups and IM languages is addressed from demographic and linguistic perspectives. The demographic part deals with both trends and criteria regarding immigration and minorization processes in the European Union, illustrated by a case study on the Netherlands. The linguistic part goes into major and minor research questions in the domain of bilingualism and migration, again with the focus on the Netherlands. The chapter ends with an outline of the present volume.

How 'they' hit the headlines Imagine a European citizen who has never been abroad and who travels to San Francisco for the first time in life, walks around downtown for a week, gets an impression of the Chinese community and food, happens to be invited for dinner by a Chinese family, and asks the host at the dinner table: 'How many foreigners live in San Francisco?', in this way referring to the many Asian, Latin, and other non-Anglo Americans (s)he has seen during that week. Now, two things might happen: if the guest's English is poor, the Chinese host might leave this European reference to ethnocultural diversity unnoticed and go on with the conversation; if the guest's English is good, however, the Chinese host might interrupt the dinner and charge his guest with discrimination. In the European public discourse on IM groups, two major characteristics emerge: IM groups are referred to as foreigners (etrangers, Ausländer) and they are referred to as being in need of integration. First of all, it is common practice to refer to IM groups in terms of non-national residents and to their languages in terms of non-territorial or non-indigenous languages. At the national level, IM groups in Great Britain are often referred to as non-English speaking residents and in the Netherlands even more curtly as 'anderstaligen' ('those who speak other languages'). The conceptual exclusion rather than inclusion in the European public discourse derives from a restrictive interpretation of the notions of citizenship and nationality. Such notions in

4

Guus Extra and Ludo Verhoeven

Europe are commonly shaped by a constitutional ius sanguinis (law of the blood) in terms of which nationality derives from parental origins, in contrast to ius solis (law of the ground) in terms of which nationality derives from the country of birth. When European emigrants left their continent in the past and colonized countries abroad, they legitimized their claim to citizenship by spelling out ius solis in the constitutions of these countries of settlement. Good examples of this are English-dominant immigration countries like the USA, Canada, Australia, and South Africa. In establishing the constitutions of these (sub)continents, no consultation took place with native inhabitants, such as Indians, Eskimos, Aboriginals, and Zulus respectively. At home, however, Europeans predominantly upheld ius sanguinis in their constitutions and perceptions of nationality and citizenship, in spite of the growing numbers of IM groups who strive for an equal status as citizens in a new multicultural European context. A second major characteristic of the European public discourse on IM groups is the focus on integration. This notion is both popular and vague, and it may actually refer to a whole spectrum of underlying concepts that vary over space and time (cf. Kruyt and Niessen, 1997, for a comparative study of the notion of integration in five EU countries since the early seventies). The extremes of the spectrum range from assimilation to multiculturalism. The concept of assimilation is based on the premise that cultural differences between IM groups and established majority groups should and will disappear over time in a society which is proclaimed to be culturally homogeneous. On the other side of the spectrum, the concept of multiculturalism is based on the premise that such differences are an asset to a pluralist society which actually should promote cultural diversity in terms of new resources and opportunities. While the concept of assimilation focuses on unilateral tasks of newcomers, the concept of multiculturalism focuses on multilateral tasks for all inhabitants in demographically changing societies. In practice, established majority groups often make strong demands on IM groups for integration in terms of assimilation and are commonly very reluctant to promote or even accept the notion of cultural diversity as a determining characteristic of an increasingly multicultural environment. It is interesting to compare the underlying assumptions of integration in the European public discourse on IM groups at the national level with assumptions at the level of cross-national cooperation and legislation. In the latter context, European politicians are eager to stress the importance of a proper balance between the loss and maintenance of 'national' norms and values. A prime concern in the public debate on such norms and values is cultural and linguistic diversity. In this context, the national languages of EU

Immigrant minority groups and immigrant minority languages in Europe

5

countries are considered to be core values of cultural identity. It is a paradoxical phenomenon that in the public discourse IM languages and cultures are commonly conceived as sources of problems and deficits and as obstacles to integration, while national languages and cultures in an expanding EU are regarded as sources of enrichment and as prerequisites for integration. The public discourse on integration of IM groups in terms of assimilation versus multiculturalism can also be noticed in the domain of education. Due to a growing influx of IM pupils, schools are faced with the challenge of adapting their curricula to this trend. The pattern of modification may be inspired by a strong emphasis on learning (in) the language of the majority of society, given its significance for success in school and on the labour market, or by the awareness that the response to emerging multicultural school populations cannot be reduced to monolingual education programming. In the former case, the focus will be on learning (in) the national language as a second language only, in the latter case on offering more than one language in the school curriculum. For a critical discussion of the concepts of non-nationals and integration in the public discourse on IM groups, we refer to Cohn-Bendit and Schmid (1992), who focus on Germany in a changing multicultural European context. Gogolin (1994) addresses the monolingual habitus of multilingual schools in Germany. Eldering (1996), Broeder and Extra (1998), and Extra and Valien (1998) focus on the consequences of multiculturalism for education in the Netherlands. These studies show that the emergence of multicultural societies in Europe has implications for all citizens, not just for 'newcomers'.

Demographic trends and criteria As a consequence of socio-economically or politically determined processes of migration, the traditional patterns of language variation across Western Europe have changed considerably over the past several decades. The first pattern of migration started in the sixties and early seventies, and it was mainly economically motivated. In the case of Mediterranean groups, migration initially involved contract workers who expected - and were expected - to stay for a limited period of time. As the period of their stay gradually became longer, this pattern of economic migration was followed by a second pattern of social migration as their families joined them. Subsequently, a second generation was born in the immigrant countries, while their parents often remained uncertain or ambivalent about whether to stay or

6

Guus Extra and Ludo Verhoeven

to return to the country of origin. These demographic shifts over time have also been accompanied by shifts of designation for the groups under consideration - 'migrant workers,' 'immigrant families,' and 'ethnic minorities,' respectively. As a result, many industrialized Western European countries have a growing number of immigrant populations which differ widely, both from a cultural and from a linguistic point of view, from the mainstream indigenous population. In spite of more stringent immigration policies in most European Union (EU) countries, the prognosis is that immigrant populations will continue to grow as a consequence of the increasing number of political refugees, the opening of the internal European borders, and political and economic developments in Central and Eastern Europe and in other regions of the world. It has been estimated that by the year 2000, about one third of the population under the age of 35 in urbanized Western Europe will have an immigration background. Within the various EU countries, four major immigrant groups can be distinguished: people from Mediterranean EU countries, from Mediterranean non-EU countries, from former colonial countries, and political refugees (cf. Extra and Verhoeven, 1993a, 1993b). Comparative information on population figures in EU member states can be obtained from the Statistical Office of the EU in Luxembourg (EuroStat). An overall decrease of the indigenous population has been observed in all EU countries over the last decade; at the same time, there has been an increase in the immigration figures. Although free movement of migrants between EU member states is legally permitted and promoted, most immigrants in EU countries originate from non-EU countries. According to EuroStat (1996), in January 1993, the EU had a population of 368 million, 4.8% of whom (almost 18 million people) were not citizens of the country in which they lived. The increase in the non-national population since 1985 is mainly due to an influx of non-EU nationals, whose numbers rose from 9 to 12 million between 1985 and 1992. The largest numbers of immigrants have been observed in France, Germany, and Great Britain. For various reasons, however, reliable demographic information on immigrant groups in EU countries is difficult to obtain. For some groups or countries, no updated information is available or no such data have ever been collected at all. Moreover, official statistics only reflect immigrant groups with legal resident status. Another source of disparity is the different data collection systems being used, ranging from nationwide census data to more or less representative surveys. Most importantly, however, the most widely used criteria for immigrant status - nationality and/or country of birth - have

Immigrant minority groups and immigrant minority languages in Europe

7

become less valid over time because of an increasing trend toward naturalization and births within the countries of residence. In addition, most residents from former colonies already have the nationality of their country of immigration. There are large differences among EU countries as regards the size and composition of immigrant population groups. Owing to labour market mechanisms, such groups are found mainly in the northern industrialized EU countries, whereas their presence in Mediterranean countries like Greece, Italy, Portugal, and Spain is rather limited. Mediterranean groups immigrate mainly to France or Germany. Portuguese, Spanish, and Maghreb residents concentrate in France, whereas Italian, Greek, former Yugoslavian, and Turkish residents concentrate in Germany. The largest immigrant groups in EU countries are Turkish and Maghreb residents; the latter originate from Morocco, Algeria, or Tunisia. Table 1 gives estimates of their size in twelve EU countries in January 1994. Table 1: Estimated numbers of inhabitants of Maghreb and Turkish origin in twelve EU countries, January 1994, based on the nationality criterion (EuroStat, 1997)

Maghreb countries Morocco Algeria Tunisia Belgium Denmark Germany Greece Spain France Italy Netherlands Portugal Finland Sweden Gr. Britain Total

Total Maghreb

Turkey

145.363 3.180 82.803 333 61.303 572.652 77.180 164.567 221 560 1.533 3.000

10.177 368 23.082 180 3.259 614.207 3.177 905 53 208 599 2.000

6.048 404 28.060 314 378 206.336 35.318 2.415 28 142 1.152 2.000

161.588 3.952 133.945 827 64.940 1.393.165 115.675 167.887 302 910 3.284 7.000

88.302 34.658 1.918.395 3.066 301 197.712 3.656 202.618 65 995 23.649 41.000

1.112.695

658.215

282.595

2.053.505

2.514.417

8

Guus Extra and Ludo Verhoeven

According to EuroStat (1997) and based on the conservative nationality criterion, in 1993 the largest Turkish and Maghreb communities could be found in Germany (almost 2 million) and France (almost 1.4 million), respectively. Within the EU, the Netherlands is in second place as the country of immigration for Turkish and Moroccan residents. Given the decreasing significance of nationality and birth-country criteria, collecting reliable information about the composition of immigrant population groups in EU countries is one of the most challenging tasks facing demographers. Complementary or alternative criteria have been suggested in various countries with a longer immigration history, and, for this reason, a history of collecting census data on multicultural population groups. In English-dominant countries such as the USA, Canada, and Australia, census questions have been phrased in terms of self-categorization ('To which ethnic group do you consider yourself to belong?') and home language use. In Table 2, the four criteria mentioned are discussed in terms of their major advantages and disadvantages (see also Extra, 1996; Broeder and Extra, 1998). As Table 2 makes clear, there is no single royal road to a solution of the identification problem. Different criteria may complement and strengthen each other. Given the decreasing significance of nationality and birth-country criteria in the European context, the combined criterion of self-categorization and home language use is a potentially promising long-term alternative.

Immigrant minority groups and immigrant minority languages in Europe Table 2:

9

Criteria for the definition and identification of population groups in a multicultural society (P/F/M = person/father/mother)

Criterion

Advantages

Disadvantages

Nationality (NAT) (P/F/M)

• objective • relatively easy to establish

• (intergenerational) erosion through naturalization or double NAT • NAT not always indicative of ethnicity/identity • some (e.g., ex-colonial) groups have NAT of immigration country

Birth-country (BC) (P/F/M)

• objective • relatively easy to establish

• intergenerational erosion through births in immigration country • BC not always indicative of ethnicity/identity • invariable/deterministic: does not take account of dynamics in society (in contrast to all other criteria)

Selfcategorization (SC)

• touches the heart of the matter • emancipatory: SC takes account of person's own conception of ethnicity/identity

• subjective by definition: also determined by language/ ethnicity of interviewer and by the spirit of times • multiple SC possible • historically charged, especially by World War II experiences

Home language (HL)

• HL is most significant criterion of ethnicity in communication processes • HL data are cornerstones of government policy in areas such as public information or education

• complex criterion: who speaks what language to whom and when? • language not always core value of ethnicity/identity • useless in one-person households

10

Guus Extra and Ludo Verhoeven

The Netherlands as demographic case study For the reasons mentioned before, it is not easy to give a complete and reliable overview of the actual size of the immigrant population in any of the European Union countries. Typical for the Netherlands, in contrast to neighbouring countries like Germany and Belgium, some immigrant groups have had Dutch nationality since birth (cf. Extra and Valien, 1997). These include all Antilleans (> 90.000) and most of the Surinamese (> 260.000) who came to the Netherlands in the last few decades from former Dutch colonies in the Caribbean, and the so-called repatriates from the former Dutch East Indies (the present Republic of Indonesia; > 280.000), who arrived in the Netherlands after Indonesia's independence. Bearing in mind the biassed character of all available nationality statistics, some longitudinal trends in the size and growth of indigenous and non-indigenous population groups during the last decade can nevertheless be derived from recent data published by the Dutch Central Bureau of Statistics (CBS, 1995). In the period 1985-1994, about 25% of the increase in the population of the Netherlands (890.000) was the result of an influx of immigrants, even though the immigrants as a group constitute less than 5% of the total population (> 15 million). Within the immigrant population, the strongest proportional growth stems from the Turkish, Moroccan, Surinamese, and (former) Yugoslavian communities, and from people belonging to the CBS category other non-European countries (mainly refugees). The latter group and the Turks and Moroccans also show the most substantial increase in terms of absolute figures. The Dutch Ministry of the Interior has attempted to reduce the increasing erosion of statistics pertaining to immigrant minorities in the Netherlands by proposing the following three ethnic determinants in all municipal population statistics (cf. Fernandes Mendes, 1991): (1) birth-country of person, father, and mother; (2) nationality of person, father, and mother; (3) selfcategorization. Obviously, the combined birth-country criterion only suffices for first- and second-generation groups. Furthermore, the (combined) nationality criterion has limited value because many IM groups have or will obtain Dutch nationality. The third criterion, self-categorization, led to many objections being raised by both minority and majority groups in the Netherlands because of its subjective loading, the possibility of multiple selfcategorization, and the potential misuse of the data collected. Ultimately, parliamentary support was given to the Ministry of the Interior for a gradual introduction of the combined birth-country criterion in all municipal population statistics, although it was recognized that this criterion would lead to a diminishing identification of IM groups over time (cf. Dales, 1992:17).

Immigrant minority groups and immigrant minority languages in Europe

11

It was also recognized that other criteria could be relevant for specific purposes or domains. Explicit reference in this context was made to the relevance of the home language criterion in the domain of education (cf. Dales, 1992:12). Table 3 gives an overview of population groups in the Netherlands on January 1, 1992, based on several birth-country criteria and the nationality criterion, and derived from Martens, Roijen, and Veenman (1994). Table 3: Population of the Netherlands based on different identification criteria (BCPMF = combination of birth-country person, mother and father; NAT = nationality; BCP = birth-country person; BCM = birth-country mother; BCF = birth-country father) on January 1, 1992 (Martens et al., 1994) Absolute figures Groups

BCPMF

Dutch

Index (column 1 = 100) NAT

BCP

BCM

BCF

12.764.767

113

108

104

105

Turks Moroccans Surinamese Antilleans Greeks Italians Former Yugoslavs Portuguese Spaniards Cape Verdians Tunisians Chinese Vietnamese Other groups

240.810 195.536 262.839 90.650 10.369 32.818 27.117 12.587 29.046 14.330 5.631 39.762 10.435 1.392.435

89 84 8 50 52 56 69 58 19 46 17 46 18

66 67 65 71 59 46 63 68 61 65 56 61 83 47

96 95 87 69 62 51 86 81 75 99 66 91 99 69

99 99 86 63 84 88 82 88 81 96 94 97 96 58

Total non-Dutch

2.364.383

31

54

77

71

Total Netherlands

15.129.150

100

100

100

100

-

Table 3 shows the significant differences which result when different identification criteria are applied. The combined birth-country criterion in the second column shows a proportion of citizens of foreign origin of more than 15% (2.364.383 out of 15.129.150 persons). When the absolute figures in the

12

Guus Extra and Ludo Verhoeven

second column are substituted by an index value of 100, it becomes clear that the 'Dutch' group is overrepresented on the basis of the nationality criterion and the BCP/BCM/BCF criteria, whereas all 'non-Dutch' groups are underrepresented on the basis of these criteria. For groups other than those mentioned in Table 3, only global estimates can be given and it is not clear what criteria are used for these estimates. Roelandt, Roijen, and Veenman (1991:31) give estimates from 1987/1988 for Moluccans (35.000), refugees (18.000), political refugees seeking asylum (8.350) and gypsies (3.700). From the more than 197.000 requests for asylum, about 57.000 were granted in the period 1985-1994 (CBS, 1995). Table 4 gives an overview of population groups in the Netherlands on January 1, 1996, based on the combined birth-country criterion (BCPMF) versus the nationality criterion, and derived from recent CBS statistics (CBS, 1997). Table 4:

Population of the Netherlands (χ J000) based on the combined birth-country criterion (BCPMF) and the nationality criterion on January 1, 1996 (CBS, 1997)

Groups(xlOOO) Dutch Turks Moroccans Surinamese Antilleans Greeks Italians Former Yugoslavs Portuguese Spaniards Cape Verdians Tunisians Other groups Total

BCPMF

Nationality

Abs. diff.

12.872

14.768

1.896

272 225 282 94 11 32 56 13 29 17 6 1.585

154 150 15 5 17 34 9 17 2 2 331

118 75 267 94 6 15 22 4 12 15 4 1.254

15.494

15.494

-

-

Compared to Table 3 and based on the combined birth-country criterion, Table 4 first of all shows an increase of the five largest IM groups over time (i.e., Turks, Moroccans, Surinamese, Antilleans and former Yugoslavs) and a stabilization of most other groups. Moreover, Table 4 again shows strong

Immigrant minority groups and immigrant minority languages in Europe

13

criterion effects of birth-country versus nationality. All IM groups are in fact strongly underrepresented in nationality-based statistics. As mentioned before, however, the combined birth-country criterion does not solve the identification problem. The use of this criterion leads to nonidentification in at least the following cases: • an increasing group of third and further generations (cf. the Moluccan and Chinese communities in the Netherlands); • different ethnocultural groups from the same country of origin (cf. Turks versus Kurds from Turkey); • the same ethnocultural group from different countries of origin (cf. Chinese from China versus Vietnam); • ethnocultural groups without territorial status (cf. gypsies). Verweij (1997) made a short tour d'horizon in four European Union countries (i.e., Belgium, Germany, France, Great Britain) and in the USA in order to study criteria utilized in the national population statistics of these countries. In Belgium, Germany, and France, such statistics have traditionally been based on the nationality criterion; only in Belgium has additional experience been gained with the combined birth-country criterion of persons, parents, and even grandparents. For various reasons, identification on the basis of the grandparents' birth-country is very problematic: four additional sources of evidence are needed (with multiple types of outcomes) and the chances of non-response are rather high. Verweij (1997) also discussed the experiences with the utilization of ethnic self-categorization in Great Britain and the USA, leaving the home language criterion out of consideration. Given the increasing identification problems with the combined birth-country criterion, Verweij, on the basis of Anglo-Saxon experiences, suggested including the selfcategorization criterion in future Dutch population statistics as the 'secondbest' middle- and long-term alternative in those cases where the combined birth-country criterion would not suffice. Moreover, he proposed carrying out small-scale experimental studies on the validity and social acceptance of the self-categorization criterion, given its subjective and historically charged character, respectively (see also Table 2), before this criterion would be introduced on a nationwide scale.

Linguistic perspectives As a consequence of processes of immigration and minorization, new languages have come into contact in Western Europe. These new conditions

14

Guus Extra and Ludo Verhoeven

of language contact have led to an increase in research on language contact that differs from traditional European studies in terms of typological distance and dynamic variability of the languages that are taken into account. In contrast to traditional European contact studies on Germanic or Romanic languages, there is first of all often a large typological distance between the languages of the countries of emigration and immigration. Turkish and Arabic, which are both spoken by millions of immigrants in Western Europe, do not belong to the Indo-European group of languages, but to the Altaic and Hamo-Semitic language families respectively. Secondly, the language contact situation of IM groups is commonly very unstable and likely to change considerably over time, both within and between successive generations. IM groups are confronted with the task of communicating in the dominant language of majority speakers in order to cope with daily life, and this language is often learnt as a second language (L2). At the same time, language varieties of the countries of origin are often learnt as a first language (LI) in the process of primary socialization, and they are used for in-group communication. These languages may also have an important or even core value as symbols of ethnic identity (cf. Smolicz, 1992). From a linguistic point of view, however, IM groups are often viewed as 'L2 learners.' For various reasons, this conceptualization leads to a reduced view of reality (cf. Extra and Verhoeven, 1993b): • not all members of IM groups acquire the dominant language of majority speakers successfully; in fact, L2 acquisition may come to a halt and fossilize at a stage that is far removed from near-native competence; • LI is commonly taken into account as a potential source of (un)successful transfer in L2 acquisition, rather than as a language variety that deserves attention on its own intrinsic grounds; • the concept of 'L2 learners' refers to individuals, whereas group membership is also an important explanatory factor in the language behaviour of IM groups. Given the fact that many members of IM groups make variable use of dominant L2 varieties and dominated LI varieties all their lives, sociolinguistic and psycholinguistic dimensions of both varieties should be taken into account. A good example of crossnational research on L2 varieties at the European level is the European Science Foundation project on initial L2 acquisition by adult immigrants in Great Britain, Germany, the Netherlands, France, and Sweden. Perdue (1993) offers a comprehensive survey of the goals, methods and outcomes of this ESF project. In each of the participating countries multiple case studies were carried out via a pairwise combination of the following source and target languages:

immigrant minority groups and immigrant minority languages in Europe

L2:

Swedish

/\ LI:

Finnish

French

Dutch

/\

/\

Spanish

Arabic

German

/\ Turkish

15

English

/\ Italian

Punjabi

The chosen source languages belong to the major immigrant languages in the participating target countries. Four out of five target languages are Germanic languages and three out of six source languages are non-Indo-European languages. The chosen research domains focus on L2 production, in particular on word order principles at the levels of utterances and noun phrases, and on personal, spatial and temporal reference in the target languages under consideration: moreover, L2 understanding is taken into account by studying misunderstanding and feedback signals. Follow-up studies on the acquisition of L2 Dutch have been carried out by, e.g., Broeder (1991), Deen (1997), and Schenning (1998). The European focus on L2 acquisition and L2 use is an accurate reflection of the vast American literature on bilingualism. The number of American studies on the acquisition and use of Spanish (the most prominent minority language in the US) is extremely limited compared to the existing literature on the acquisition and use of L2 English. Another explanatory factor is the traditionally philological orientation of European research on so-called 'nonWestern' languages such as Turkish, Arabic, or South Asian languages. The fact that in a European context these languages are increasingly spoken 'here and now' has unfortunately not resulted in a significant modification of historically-oriented research priorities. The philologists' lack of scientific involvement most probably stems from a lack of affinity with suitable methods in sociolinguistic and psycholinguistic research. However, the commitment of experts in these languages is indispensable for the development of research on first language varieties of IM groups. Given the social and scientific emphasis on L2 acquisition and L2 use of IM groups in Europe, the following basic questions have received only scant attention: • What immigrant language varieties are actually used in oral or written communication? • Who are the interlocutors, and what are the social settings and the topics of communication? • How are these language varieties acquired in a dominant L2 environment?

16

Guus Extra and Ludo Verhoeven

•

What interactions in terms of codeswitching and language transfer can be observed between dominant and dominated language varieties? • What intergenerational processes of language shift can be observed over time? • What language attitudes towards these issues are manifest within dominant and dominated groups? • What crosslinguistic similarities and differences in each of these domains can be observed for different minority groups in one particular country and, alternately, for one particular minority group in different European countries? Over the last decade, some changes have taken place across Europe to counterbalance the biassed perspective on the language use of IM groups. A landmark in terms of both scope and size was the Linguistic Minorities Project (1985), more recently followed-up by a study by Alladina and Edwards (1991) on languages other than English in Great Britain. Similar studies have been published by Boyd (1985) on Sweden, Vermes (1988) on France, and Extra and Verhoeven (1993b) on the Netherlands.

The Netherlands as linguistic case study In contrast to the growing number of studies on Dutch as a second language (see Spliethoff, 1996, for a recent overview), far less empirical research has been done on the status and use of IM languages in the Netherlands. This holds, in particular, for lesser used languages of relatively small communities. De Ruiter (1991) and Extra and Verhoeven (1993b) cover a variety of the most widely used IM languages in the Netherlands. The majority of research has gone into the status and use of Turkish and Arabic, these being languages used by the Turkish and Moroccan communities in the Netherlands. In most Turkish families, Turkish is the dominant home language. Other languages, however, Kurdish in particular, are also used (Boeschoten et al., 1993). Morocco is a country with intricate patterns of language variation and language choice (De Ruiter, 1989). Classical Arabic and Modern Standard Arabic, the languages of religion and the mass media, respectively, are never learnt as primary language varieties. Primary socialization commonly takes place in Moroccan Arabic or in one of three regionally distinct Berber varieties (Tashelhit, Tamazigt, or Tarifit). Whereas speakers of Berber generally acquire Moroccan Arabic as a lingua franca, Arabophone speakers tend not to learn Berber. Given the complex pattern of language variation in Morocco, however, it is a daily experience for many inhabitants to encounter,

Immigrant minority groups and immigrant minority languages in Europe

17

switch to, and borrow from, different languages. At least part of this complex pattern is clearly represented in the Moroccan community in the Netherlands. With respect to Turkish in the Netherlands, studies have been done on the acquisition of Turkish by preschool or elementary school children (Boeschoten, 1990; Schaufeli, 1991; Aarts, 1994; Aarssen, 1996), on the vitality of Turkish in secondary schooling (Özgüzel, 1994), and on processes of codeswitching amongst first and second generation Turks (Backus, 1996). With respect to (Moroccan) Arabic in the Netherlands, the focus has been on its status in primary education (Van de Wetering, 1990), on codeswitching (Boumans, 1998; Nortier, 1989), and on processes of language maintenance and language loss (El Aissati, 1997). Moreover, language proficiency tests have been developed for Turkish and Arabic for use at the initial stages (Verhoeven et al., 1995) and at the final stages (Aarts and De Ruiter, 1995) of elementary education. Apart from studies on Turkish and Arabic, few studies have dealt with lesser used IM languages in the Netherlands. Examples of such studies are Kook (1994) and Narain (1995) on Papiamentu, Charry et al. (1983) and Van der Avoird (1998) on Surinamese languages, Tahitu (1989) and Rinsampessy (1992) on (Moluccan) Malay, Tinnemans (1991) and Michielsens (1992) on Italian, Dialektopoulos (1998) on Greek, the Research Group Kurds (1996) on Kurdish, and Hendriks (1986) on Chinese. Surinamese languages (in particular Sranan Tongo and Hindi/Hindustani), Papiamentu, and (Moluccan) Malay are spoken as home languages primarily by the Surinamese, Antillean, and Moluccan communities in the Netherlands, respectively. Home language use has earlier been referred to as a complementary or alternative criterion of ethnic identity. Although the Dutch Ministry of the Interior has recognised the potential value of this criterion for population statistics and in particular school statistics, it has never been utilised in practice. Given the relevance of the home language criterion in an educational context, Breeder and Extra (1995, 1998) carried out a Home Language Survey in the school year 1993/1994, with the support of the Ministry of Education, amongst 34.451 elementary school pupils in five medium-sized cities. For each child, a short questionnaire was administered orally and individually by the teacher. After some background questions the following screening question was asked: Is any language other than Dutch ever used in your home? Only those children who gave an affirmative answer to this question were asked additional questions relating to the following five language profile dimensions:

18

Guus Extra and Ludo

Language variety: Language proficiency: Language choice:

Language dominance: Language preference:

Verhoeven

What other language(s) is/are used in your home ? Can you understand/speak/read/write this language? Do you speak this language with your mother/ father/ elder brother(s) or sister(s)/ younger brother(s) or sister(s)? What language do you speak best? What language do you like to speak most?

The first important outcome of the survey was that in the homes of 7.179 out of 34.451 elementary school pupils (21%) another language was in use instead of or apart from Dutch. This percentage is supposed to be even much higher in the four largest cities of the country. All in all, 66 different home languages, originating from all continents, could be traced and classified. The classification process was made possible with the support of the detailed ILEA Catalogue of Languages (1990). The relative share of the ten most frequently mentioned home languages is presented in Figure 1. The figure shows that Turkish, Arabic, and Berber hold strong positions in the top ten of languages.

Turkish 2082

Arabic 803

Hindi 1Θ8

Spanish 211 Malay 250 Chinese 277 Papiamento 487 English 411

Surinamese/Sranan 381

Figure 1: Ten largest language groups in the Home Language

Survey

Immigrant minority groups and immigrant minority languages in Europe

19

On the basis of the language profile dimensions mentioned before, a language vitality index has been composed for the ten most frequently mentioned home languages in the Home Language Survey. The relevant parameters for the index are: • language competition: the language does not need to compete with other languages, with the exception of Dutch; • language proficiency: the language is understood by the children; • language choice: the language is often/always used with the mother, • language dominance: the language is best spoken by the youngest children (grades 1/2); • language preference: the language is most preferred to be spoken by the youngest children (grades 1/2). The results (with scores in %) are presented in Table 5. Table 5: Language vitality index derived from five language profile dimensions Language group Turkish Chinese Berber Arabic Papiamentu Spanish English Sur./Sranan Hind(ustan)i Malay

Language Language Language Language Language Language vitality dominance preference competition proficiency choice 99 96 72 73 87 80 68 88 89 93

98 98 96 94 91 89 90 81 89 68

80 77 76 60 41 46 33 15 24 13

88 74 76 61 33 32 28 26 14 15

78 68 71 52 35 30 29 32 18 18

886 826 782 680 574 554 496 484 468 414

The resulting language vitality index is obviously arbitrary in the sense that different language dimensions are equally weighted. None of the home languages need to compete strongly with languages other than Dutch. Arabic and Berber are relatively often used together as home languages within Moroccan families. English functions relatively often as a lingua franca in a context in which other languages are also used at home. Typically, literacy is a skill that is acquired and enhanced within the school context. It has a relatively weak status compared to oral skills in the home languages. The highest percentages of children who can read in their home language are for Turkish (60%), Spanish (46%), English (45%) and Arabic (40%). Scores

20

Guus Extra and Ludo Verhoeven

below 40% emerge for Papiamentu (32%), Chinese (31%), Surinamese/ Sranan (29%), Malay (23%) and Hindustani/Hindi (10%). Given its noncodified status, Berber is left out of consideration. The pattern of majority language choice (Dutch) of children in interaction with other family members is represented in Table 6. Table 6:

Choice of Dutch by children (in %) in interaction with other family members

Language group Turkish Chinese Berber Arabic Papiamentu Spanish English Surinamese/Sranan Hind(ustan)i Malay

Mother

Father

Younger siblings

12 11 10 18 55 44 62 80 73 82

22 13 20 21 55 46 61 78 71 85

35 52 60 66 61 66 74 85 82 92

Older siblings 48 39 68 54 65 78 76 85 81 92

The data presented in Table 6 are a mirrorlike affirmation of a classical pattern of language shift. The mother emerges as the strongest gatekeeper of language maintenance (thus supporting the widely used concept of 'mother tongue'), whereas the father takes a strong second position. Shift of choice towards the majority language (Dutch) increasingly runs via younger and older siblings, respectively. Table 6 makes clear that there is strong variation in the vitality of IM languages in the Netherlands. This variation, both within and across different communities, needs further exploration from a crosslinguistic and crosscultural perspective. Moreover, reliable data on home language use should be considered as prerequisites for answering basic policy questions about home language instruction. Periodically collected home language data amongst school children would provide the basis for a dynamic language policy and for intergenerational trend-studies on processes of language maintenance and language shift (see also Aarssen et al., 1998).

Immigrant minority groups and immigrant minority languages in Europe

21

The present volume The present volume is an overview of recent research on processes of language change among IM groups. Apart from the two introductory chapters in Part 1, the book consists of four distinct parts. The focus of Part 2 is on processes of early bilingual development. Part 3 goes into processes of bilingual development at school age. In Part 4, the constraints in processes of codes witching and borrowing are dealt with. Part 5 discusses the issues related to processes of language maintenance and language loss. Taken together, the volume offers a comprehensive state of the art of the expanding field of bilingualism and migration. In order to allow for crosscultural and crosslinguistic comparisons, different language varieties in various societies are taken into account in Parts 2-5. The opening chapter in Part 2 is by Suzanne Romaine. She presents an outline for a typology of childhood bilingualism. Moreover, she discusses the educational implications for early bilingual development among IM children. She criticizes the fact that uninformed officials continue to advise immigrant parents to switch from their home language in order to aid their children's acquisition of the majority language. She argues for bilingual education as a program for all students and not just as a special instructional program for disadvantaged IM groups. In the next chapter, Annick De Houwer addresses the question of why it is that some children who are regularly exposed to two languages from a very young age actually start to speak and continue to use these languages, while other children in similar circumstances do not. Her study focusses on parents' impact belief concerning the language involved and their being bilingual. According to De Houwer, parent beliefs and attitudes can be seen to lie at the basis of parent language behaviour towards their children, which in turn acts as a powerful determinant of children's language use. In the following chapter, Carol Pfaff analyses the development of language mixing and codeswitching of a Turkish/German bilingual boy who was followed from 1 to 8 years old. Initially, the boy did not use any German while speaking Turkish. He then went through a stage of insertion of German nouns and verbs syntactically and morphologically integrated into Turkish structure. Finally, he showed complete alternation to German, triggered by the use of key lexical items in German-dominant domains. In the final chapter of this part, Hanneke van der Heijden reports on a study on compounding and derivation in the simultaneous acquisition of Turkish and Dutch by four Turkish children born in the Netherlands. Data

22

Guus Extra and Ludo Verhoeven

were collected when the children were between 2;0 and 3;6 years of age. As a monolingual reference group, she used two monolingual Dutch children and two monolingual Turkish children. For both monolingual and bilingual children, she found that, in Dutch, derivation precedes compounding, whereas, in Turkish, compounding precedes derivation. She concluded that the bilingual child extends its word formation repertoire along different lines in the two respective languages, both lines conforming to the monolingual pattern of acquisition. Jeroen Aarssen and Petra Bos open Part 3 of the book. They present two experimental studies on the comprehension of cohesive devices in Turkish/Dutch and Moroccan/Dutch bilingual children and in monolingual control groups. One study is on the use of bound and free anaphors in an anaphoric reference task. The other study goes into the comprehension of four types of relative clause constructions. The role of both underlying factors and surface structure determinants are investigated. In the subsequent chapter, Äke Viberg presents a study on the lexical development of 20 children with L2 Swedish between 6-10 years of age. The group was heterogenous with respect to their L I . The process of L2 development was compared with that of 11 LI Swedish control children. Viberg found that verbal semantics fields tend to be organised around nuclear verbs. These verbs are typologically unmarked, highly frequent, and with the same basic/general meaning in a wide range of languages. Examples are go, make, put, and say. They turn out to be overextended and overrepresented in L2. Their use seems to reflect universal organizational principles of the verb lexicon in human languages. In the next chapter, Ruth Berman describes a study on narrative text production by three groups of Israeli adults: monolingual speakers of Hebrew, bilingual speakers of Hebrew and English (as a second language), and bilingual speakers of American English and (immersed) Hebrew. The informants were adults with the same level of education and similar cultural background. For English as a second language, she observed more grammatical and lexical deviations, e.g., the use of dative preposition to express the semantic relation of 'malefactee' which is typical of colloquial Hebrew. Similarly, specific verbs such as grab and snatch were seldom used. There were also differences with respect to the use of rhetorical devices. Subordination was more frequent among the Israelis, whereas juxtaposition was more frequent with the Americans. In story telling in Hebrew as a second language, an inappropriate register with literary archaic locutions was often used. It should be remembered that Hebrew shows a very strong diglossia of

Immigrant minority groups and immigrant minority languages in Europe

23

a colloquial standard and a formal academic register. In English as a first and second language, however, subjects chose the appropriate colloquial level. Part 4 of the book on codeswitching and borrowing opens with a chapter by Carol Myers-Scotton. She discusses three compromise strategies of codeswitching, two of which are used by speakers who are proficient in two or more languages. A third strategy, referred to as Matrix Language, is used when one language is problematic, i.e., in situations of convergence, attrition, second language acquisition, and pidgin/creole formation. Her conclusion is that a composite Matrix Language with specific predictions about the source of its lexical structure best explains language changes in progress involving two or more linguistic systems. In the following chapter, Pieter Muysken distinguishes between three processes of borrowing. The first is the insertion of elements from one language into a structure from another language. The second strategy concerns the alternation between constituents from language A and language B. The final borrowing strategy is congruent lexicalization in which words from languages A and Β are inserted more or less randomly into a shared grammatical structure. Muysken also shows that there is no watertight division between codeswitching and borrowing, since any mixed type of item that can be interpreted as a lexical unit could potentially be borrowed. In the next chapter, Lars Johanson deals with the way in which, in asymmetric contact situations, a socially dominated language A is influenced by a socially dominant language B. In particular, the processes underlying insertion of copies from Β into A are investigated from a developmental point of view. It is argued that a shift of the basic code is more likely to be a social process rather than a gradual development, owing to the use of an increasing number of copied elements. The following chapter by Ad Backus focuses on the intergenerational codeswitching continuum in an immigrant community. From his study on Turkish-Dutch codeswitching among Turkish adolescents in the Netherlands, it seems relevant to discriminate between three different lects. Speakers of these lects belong to three distinct generations withing the IM community. A general shift from insertional to alternational codeswitching and from Dutch as Embedded Language to Dutch as Matrix Language can be observed. Embedded language elements are described in terms of devices of morphosyntactic integration. He also investigated why particular elements of the embedded language are selected. In first-generation speech, embedded Dutch content words are primarily used if they have highly specific meanings. In

24

Guus Extra and Ludo Verhoeven

later generations, discourse means, such as the use of conjunctions and emphatic pronouns, are used to achieve effects of high awareness. In the final chapter of Part 4, Louis Boumans deals with codeswitching and the organization of the mental lexicon. He argues that codeswitching is relatively easy with uninflected verb stems and uninflected noun stems of the Embedded Language, and that Embedded Language function morphemes occur in accordance with Matrix Language grammar. This is supposed to constrain the type of Embedded Language morphemes that may occur in a Matrix Language frame. According to Boumans, the embedding of inflected forms reflects the unit status of these complex forms in the mental lexicon. He claims that irregular inflected forms are embedded more often than regular forms, and that affixes are more likely to be embedded together with content word heads than free form function morphemes. Part 5, on language maintenance and language loss, opens with a chapter by Sally Boyd and Sirkku Latomaa. They apply Joshua Fishman' concepts of diglossia and bilingualism to the results of a research project on IM languages in the Nordic region. A stable level of diglossia across generations, along with a decrease of bilingualism from the older to the younger generation, is seen as an indication of language shift. Two cases of minority language maintenance are observed. The first is language maintenance by choice which is often referred to as 'elite bilingualism'. The other one is language maintenance by exclusion due to discrimination and segregation. The focus of the next chapter by Arturo Tosi is on the notion of community in language maintenance. According to Tosi, the use of an IM language within a community can have both positive and negative effects on the individual. He argues that language maintenance can either prevent socioeconomic integration or serve as an instrument for cultural identification. Increasing isolation of IM groups tends to promote individual disorientation rather than group aggregation. Tosi wonders what institution could function in the absence of a family network and group solidarity as the alternative social ingredient that could facilitate a less traumatic transition from the old identity to a new European identity. In the following chapter, Kees de Bot links language attrition and language shift to a psycholinguistic model of language processing. In this model, the distinction between declarative and procedural knowledge is seen as relevant for the explanation of maintenance and loss. Procedural knowledge is rulegoverned and part of a larger linguistic system. Therefore, it is less subject to loss. Declarative knowledge which involves isolated phenomena may get lost much easier.

Immigrant minority groups and immigrant minority languages in Europe

25

In the final chapter, Abder El Aissati and Anneli Schaufeli compare the first language use and proficiency of two groups of second-generation minority language speakers in the Netherlands: Turkish and Moroccan adolescents. The Turkish adolescents report higher proficiency in their LI than their Moroccan Arabic counterparts. Furthermore, Dutch-based strategies in sentence processing are used more prominently by speakers of Moroccan Arabic. These differences are explained in terms of the similarity between the Dutch and the Turkish alphabets, the distance between standard and colloquial languages, and the relative transparency of grammar.

References Aarssen, J. (1996). Relating events in two languages. Acquisition of cohesive devices by Turkish-Dutch bilingual children at school age. Tilburg: Tilburg University

Press. Studies of Multilingualism 2. Aarssen, J., P. Broeder and G. Extra (1998). Allochtone talen in het voortgezet onderwijs. Bouwstenen voor lokaal taalbeleid. Den Haag: VNG Uitgeverij. Aarts, R. (1994). Functionele geletterdheid van Turkse kinderen in Turkije en in

Nederland. De Lier: Academisch Boeken Centrum. Aarts, R. and J. de Ruiter (1995). Toets Turkse Taal, Toets Arabische Taal. Arnhem:

CITO. Alladina, S. and V. Edwards (Eds.) (1991). Multilingualism in the British Isles. Vol. 1: The older mother tongues and Europe. Vol. 2: Africa, Asia, and the Middle

East. London: Longman. Backus, A. (1996). Two in one. Bilingual speech of Turkish immigrants in the

Netherlands. Tilburg: Tilburg University Press. Studies in Multilingualism 1. Boeschoten, H. (1990). Acquisition of Turkish by immigrant children. A multiple case study of Turkish children in the Netherlands aged 4 to 6. Wiesbaden: Harrasowitz.

Boeschoten, Η., M. Dorleijn and M. Leezenberg (1993). Turkish, Kurdish and other languages of Turkey. In: G. Extra and L. Verhoeven (Eds.), Community languages in the Netherlands. Amsterdam: Swets & Zeitlinger. Boumans, L. (1998). The syntax of codeswitching. Analysing Moroccan

Arabic/Dutch

conversations. Tilburg: Tilburg University Press. Studies in Multilingualism 12. Boyd, S. (1985). Language survival: A study of language contact, language shift and

language choice in Sweden. PhD thesis. Göteborg: University of Göteborg. Broeder, P. (1991). Talking about people. A multiple case study on adult language

acquisition. Amsterdam/Lisse: Swets & Zeitlinger. Broeder, P. and G. Extra (1995). Minderheidsgroepen en minderheidstalen. Den Haag: VNG Uitgeverij.

26

Guus Extra and Ludo Verhoeven

Broeder, P. and G. Extra (1998). Language, ethnicity, and education. Case studies on immigrant minority groups and immigrant minority languages. Clevedon: Multilingual Matters. CBS (1995). Allochtonen in Nederland 1995. Voorburg/Heerlen: Centraal Bureau voor de Statistiek CBS (1997). Allochtonen in Nederland 1997. Voorburg/Heerlen: Centraal Bureau voor de Statistiek. Charry, E., G. Koefoed and P. Muysken (red.) (1983). De talen van Suriname. Achtergronden en ontwikkelingen. Muiderberg: Coutinho. Cohn-Bendit, D. and Th. Schmid (1992). Heimat Babylon. Das Wagnis der multikulturellen Demokratie. Hamburg: Hoffmann & Campe. Dales, I. (1992). Registratie en rapportage minderhedenbeleid. Den Haag: Staatsuitgeverij. Deen, J. (1997). Dealing with problems in intercultural communication. Tilburg: Tilburg University Press. Studies in Multilingualism 5. De Ruiter, J. (1989). Young Moroccans in the Netherlands. An integral approach to their language situation and acquisition of Dutch. PhD thesis. Utrecht: University of Utrecht. De Ruiter, J. (1991). Talen in Nederland. Een beschrijving van de taalsituatie van negen etnische groepen. Groningen: Wolters-Noordhoff. Dialektopoulos, T. (1998). Grieks onderwijs in eigen taal en cultuur in Nederland. Utrecht: LIZE. El Aissati, A. (1997). Language loss among native speakers of Moroccan Arabic in the Netherlands. Tilburg: Tilburg University Press. Studies in Multilingualism 6. Eldering, L. (1996). Multiculturalism and multicultural education in an international perspectieve. Anthropology & Education Quarterly, 27(3), 315-330. EuroStat (1996). Statistics in focus. Population and social conditions 2. Luxembourg: EuroS tat. EuroStat (1997). Migration statistics 1996. Statistical document 3A. Luxembourg: EuroStat. Extra, G. (1996). De multiculturele samenleving in ontwikkeling: feiten, beeldvorming en beleid. Tilburg: Tilburg University Press. Extra, G. and T. Valien (1997). Migration and multilingualism in Western Europe. A case study on the Netherlands. In: W. Grabe (Ed.), Multilingualism and multilingual communities. American Review of Applied Linguistics 17. Extra, G. and T. Valien (1998). Dutch as a second language in the Netherland and Flanders. In: G. Extra and J. Maartens (Eds.), Multilingualism in a multicultural context. Case studies on South Africa and Western Europe. Tilburg: Tilburg University Press. Studies in Multilingualism 10. Extra, G. and L. Verhoeven (Eds.) (1993a). Immigrant languages in Europe. Clevedon: Multilingual Matters. Extra, G. and L. Verhoeven (Eds.) (1993b). Community languages in the Netherlands. Amsterdam: Swets & Zeitlinger.

Immigrant minority groups and immigrant minority languages in Europe

27

Fernandes Mendes, H. (1991). Concept-nota registratie en rapportage minderhedenbeleid. Den Haag: Staatsuitgeverij. Gogolin, I. (1994). Der monolinguale Habitus der multilingualen Schule. Münster/ New York: Waxmann. Hendriks, H. (1986). De tweetaligheid van Chinese hinderen in Nederland. Master thesis. Leiden: Dept. of Sinology. ILEA (1990 [4th ed.]). Catalogue of Languages spoken by ILEA school pupils. London: ILEA Research and Statistics Branch. Kook, H. (1994). Leren lezen en schrijven in een tweetalige context. Antilliaanse en Arubaanse kinder en in Nederland. PhD thesis. Amsterdam: University of Amsterdam. Kruyt, A. and J. Niessen (1997). Integration. In: H. Vermeulen (Ed.), Immigrant policy for a multicultural society. Brussels: Migration Policy Group. Linguistic Minorities Project (1985). The other languages of England. Londen: Routledge & Kegan Paul. Martens, E., J. Roijen and J. Veenman (1994). Minderheden in Nederland. Statistisch Vademecum 1993/1994. Den Haag: SDU/CBS. Michielsens, M. (1992). Italia mia. Televisie kijken in ballingschap. Amersfoort/ Leuven: Acco. Narain, G. (1995). Taaltalent in ontwikkeling. Een studie naar het Papiamentu en het Nederlands in de kleuterperiode op Curagao en in Nederland. Tilburg: Tilburg University Press. Nortier, J. (1989). Dutch and Moroccan Arabic in contact. Code switching among Moroccans in the Netherlands. Dordrecht: Foris. Özgüzel, S. (1994). De vitaliteit van het Turks in Nederland. PhD thesis. Tilburg: Tilburg University. Perdue, C. (Ed.) (1993). Adult language acquisition: Cross-linguistic perspectives. Volume I: Field methods, Volume II: The results. Cambridge: Cambridge University Press. Research Group Kurds (1996). Koerden in Nederland: moedertaal en identiteit. Amsterdam: Dept. of Languages and Cultures of the Near East. Rinsampessy, E. (1992). Saudara Bersaudara. Molukse identiteit in processen van culturele verandering. Assen: Van Gorcum. Roelandt, Th., J. Roijen and J. Veenman (1991). Minderheden in Nederland. Statistisch Vademecum 1991. Den Haag: Staatsuitgeverij. Schaufeli, A. (1991). Turkish in an immigrant setting. A comparative study of the first language of monolingual and bilingual Turkish children. PhD thesis. Amsterdam: University of Amsterdam. Schenning, S. (1998). Learning to talk about space. The acquisition of Dutch as a second language by Moroccan and Turkish adults. Tilburg: Tilburg University Press. Studies in Multilingualism 11.

28

Guus Extra and Ludo Verhoeven

Smolicz, J. (1992). Minority languages as core values of ethnic cultures; A study of maintenance and erosion of Polish, Welsh, and Chinese languages in Australia. In: W. Fase, Κ. Jaspaert and S. Kroon (Eds.), Maintenance and loss of minority languages. Amsterdam/Philadelphia: John Benjamins. Spliethoff, F. (Ed.) (1996). Second language acquisition in Europe. 's-Hertogenbosch: KPC. Tahitu, E. (1989). Melaju Sini. HetMaleis van Molukse jongeren in Nederland. PhD thesis. Leiden: Rijksuniversiteit Leiden. Tinnemans, W. (1991). L'ltalianitä. De Italiaanse gemeenschap in Nederland. Amsterdam: Het Spinhuis. Van der Avoird, T. (1998). Hindi in the United Kingdom and the Netherlands. In: G. Extra and J. Maartens (Eds.), Multilingualism in a multicultural context. Case studies on South Africa and Western Europe. Tilburg: Tilburg University Press. Studies in Multilingualism 10. Van de Wetering, S. (1990). Het onderwijs in eigen taal en cultuur aan Marokkaanse kinderen in Nederland. PhD thesis. Utrecht: University of Utrecht. Verhoeven, L., G. Narain, G. Extra, Ö. Konak and R. Zerrouk (1995). Toets tweetaligheid Turks, Marokkaans-Arabisch, Papiamentu, en Nederlands. Amhem: CITO. Vermes, G. (Ed.) (1988). Vingt-cinq communautes linguistiques de la France: les langues immigrees. Paris: Editions l'Hrmattan. Verweij, A. (1997). Vaststelling van etnische herkomst in Nederland. De BiZamethode nader bekeken. Rotterdam: ISEO.

Processes of language change in a migration context: the case of the Netherlands Guus Extra and Ludo Verhoeven

This chapter gives an overview of the four research domains which are outlined in Parts 2-5 of this volume, i.e., early bilingual development, bilingual development at school age, codeswitching and borrowing, and language maintenance and language loss. In discussing each of these research domains, we will focus on the studies that were carried out in the Netherlands as part of the Dutch Science Foundation program on Language and Minorities. The status of this program is outlined in the first part of this chapter.

The Dutch Science Foundation program on Language and Minorities During a period of ten years (1984-1994), the Dutch Science Foundation (DSF or NWO: Nederlandse Organisatie voor Wetenschappelijk Onderzoek) developed and maintained a priority program for research on ethnic minorities in the Netherlands. The program consisted of four major themes. Each theme was explored in a variety of research projects which were carried out in interuniversity cooperation. Table 1 gives an overview of the research themes and the major universities involved (cf. Heeren et al., 1994). Heeren et al. (1996) present state-of-the-art reports for each of these domains, accompanied by a presentation of five of the ongoing research projects. Extra and Verhoeven (1996) give an overview of research regarding the second theme, and Backus and Boumans (1996) and Aarssen and Bos (1996) report on their research projects in the context of this theme. Other accounts of the Language and Minorities research program can be found in Extra and Verhoeven (1993, 1994).

30

Guus Extra and Ludo Verhoeven

Table 1: Overview of research in the DSF priority program on ethnic minorities in the Netherlands Themes

Universities involved

1. Religions of ethnic minorities (with a focus on Islam)

Leiden (RUL) and Amsterdam (VU)

2. Language change in a migration context

Tilburg (KUB) and Nijmegen (KUN)

3. Legal issues in a multi-ethnic society

Nijmegen (KUN) and Utrecht (UU)

4. Interrelation of ethnocultural and socioeconomic status

Utrecht (UU) and Amsterdam (UvA)

Processes of emigration lead by definition to diminished or even broken contact with the original language community and to variable contact (in terms of both intensity and quality) with a new and dominant language community. A consideration of language contact in this context first of all requires a dynamic rather than a static perspective on bilingualism (see Romaine, 1995; Appel and Muysken, 1987, and Grosjean, 1982 for overview studies). The type and degree of bilingualism of IM groups is relatively strongly dependent on processes of language change in such a situation. In studying processes of language change in a migration context, a distinction can be made between intragenerational and intergenerational processes (cf. Extra, 1993). Intragenerational processes can provide insights into the structure and pace of first and second language acquisition. A major question with respect to structural determinants of bilingual development is to what degree intralingual versus interlingual operating principles play a role. In the former case, such principles may have a universal base; in the latter case, transfer may occur from LI to L2 or the reverse. Both intralingual and interlingual principles may eventually lead to the emergence of new language varieties. A major question with respect to temporal determinants is to what degree and in what period of time native-like proficiency in LI and L2 can be achieved. When minority children are overexposed to majority language use and underexposed to minority language use, impaired minority language learning or stagnation may be the result (cf. Mougean et al., 1985 on the French of Canadian youngsters, or Ytsma, 1995 on the Frisian of Frisian

Processes

of language change in a migration

context

31

children). Impaired acquisition of minority languages may lead to considerable language change in successive generations. The common pattern is that the minority language recedes under the influence of an increasing utilization of the majority language because the latter has a higher utility and prestige. The focus of the Dutch Science Foundation program on Language and Minorities was on processes of language change amongst first- and secondgeneration Turkish and Moroccan Arabic inhabitants of the Netherlands (see the first chapter of this volume for the demographic rationale of this focus). The selection of these two language groups offers innovative crosslinguistic perspectives. From a typological point of view, there is a large distance between Turkish and Moroccan Arabic, on the one hand, and between each of these languages and Dutch, on the other. The processes of language change within the chosen target groups relate to four domains of research: 1. early bilingual development; 2. bilingual development at school age; 3. codeswitching among young adults; 4. language shift and language loss among adolescents. Table 2 shows how these processes of language change among speakers of Turkish or Moroccan Arabic and Dutch are taken into account in the research program. The first two research domains focus on processes of first and second language acquisition within the second generation. Longitudinal research designs were used to study bilingual development at preschool (2-4 years) and school age (4-12 years). The last two research domains focus on intergenerational processes of language change. Cross-sectional research designs were used to study the processes of codeswitching and language loss in successive generations (Gl and G2). Below, these four research domains are briefly characterized and related to the outline presented in Table 2.

32

Guus Extra and Ludo Verhoeven

Table 2:

Intra- and intergenerational processes of language change in a migration context (G = generation; Τ = Turkish, MA = Moroccan Arabic, D = Dutch)

Domains of research

Gl >

1

G2 2-4 years

4-12 years

+

-

-

>13 years

Researchers Hanneke van der Heijden

Target language pairs T-D

Jeroen Aarssen and Petra Bos

T-D MA-D

+

Ad Backus and Louis Boumans

T-D MA-D

+

Abderrahman El Aissati and Anneli Schaufeli

MA-D T-D

2

-

+

-

3

+

-

4

+

-

Early bilingual development Background

Dependent on the starting point of language input, two languages can be acquired simultaneously or successively (Extra and Verhoeven, 1994). Simultaneous acquisition occurs when children are confronted with two languages from birth; successive acquisition occurs when children are confronted with a second language at a later age. Most research on children's language acquisition is based on monolingual rather than bilingual development and the most studied target language is English. The study of early bilingual development has, however, a rich tradition. Reviews by McLaughlin (1985, 1994), Romaine (1995), and Taeschner (1983) make clear that traditional studies were mainly single or multiple case studies on two typologically related European languages, one language being the native language of one of the parents and the other language being the language of the wider environment, within high socio-economic status families in which one of the parents acted as researcher. Among the wellknown examples are Leopold's landmark study (1970) of his daughter's simultaneous acquisition of German (father's language) and English (environmental language) and Taeschner's (1983) study of her two

Processes of language change in a migration context

33

daughters' simultaneous acquisition of German (mother's language) and Italian (environmental language). Previous studies of early bilingualism give no reason to believe that the process of language acquisition in bilingual and monolingual children is different in its basic features. The essential difference is that bilingual children are confronted with two sets of input and that they have the additional task of distinguishing between two language systems. In most of the studies carried out so far on early bilingual development, three stages are generally distinguished in which a gradual differentiation takes place from a single unitarian language system towards two distinct language systems. This process of differentiation is the major question in studies like Taeschner (1983), Fantini (1985), Meisel (1990), and De Houwer (1990). In the first stage (0-2 years), children can handle one-word utterances which, if put in a particular order, can also express relations. Two-word utterances also occur, in which conceptual relations are expressed by word accentuation and accompanying gestures rather than by morphosyntactic rules. In this first stage, only one lexical system emerges which contains elements from both languages (equivalents of words in the two languages hardly occur). In the second stage (2-3 years), the lexicon becomes differentiated for the two languages, and children start using equivalent words in the two languages. From a morphosyntactic point of view, a gradual system separation takes place. Children learn to produce utterances on the basis of an early grammatical system. In the third stage (3-4 years), children become more and more aware of the existence of two distinct codes. Based on their implicit knowledge of grammatical rules within each of the languages, they enlarge the complexity of generated sentences. An important question in the study of early bilingual development relates to the role of interference between the two languages under consideration. The traditional point of view is that interference plays an important role (Romaine, 1995:216-225). Recent studies like Meisel (1990) and De Houwer (1990, 1994), however, make clear that the syntactic development of bilingual children takes places separately in the two languages and follows universal operating principles. The degree of interference appears to be dependent on both the degree of asymmetry in language input and the degree of typological distance between the two languages under consideration. There is a great chance of interference phenomena, in particular at the levels of lexicon and phonology, if the input of the two languages is unbalanced and if the typological distance between the two languages is small.

34

Guus Extra and Ludo Verhoeven

Young bilingual children are soon aware of their handling of two language systems and start to codeswitch (Fantini, 1985; Bentahila and Davies, 1994). The reasons these children codeswitch may be lack of competence in the other language or a more or less conscious one-person one-language strategy. Design and results of the DSF project Most of the studies that have been conducted so far were limited in their scope, given the fact that the languages under consideration were highly related. The analysis of children's speech data in two typologically unrelated languages offers new perspectives on the role of the structural properties of these languages in the process of acquisition. In the present project, a report is given on the results of a study on bilingual development of Turkish children born in the Netherlands. These children form part of a second generation of immigrants who originally moved from rural sites in Turkey to industrialized areas in the Netherlands. Unlike most of the work that has been done so far, the acquisition of these two typologically distinct languages has been studied by outside researchers in children of a low socio-economic background, LI being the ethnic community language spoken by both parents, and L2 being the dominant language of the environment. The process of bilingual development has been studied from an interdisciplinary point of view, combining insights from linguistic theory and developmental theory. From a formal linguistic point of view, bilingual development is often defined as an 'instantaneous' process in an ideal context in which the child has at his disposal all of the principles and parameters of universal grammar and two sets of input data necessary to fix those parameters. Given the obvious fact that languages are not acquired instantaneously, developmental theory must explain the various stages which characterize both first and second language development and the apparent difficulties children encounter in separating the two languages. In the project, the following research questions were addressed: How are the first and second language systems of children built up in the course of time, and under what conditions and with what lexical and syntactic rules do processes of transfer and codeswitching occur? A multiple-case study was conducted with a longitudinal design, covering three groups of informants:

Processes

•

of language change in a migration

context

35

a core group of four Turkish girls, born in the Netherlands and growing up with two languages, i.e., Turkish and Dutch; • a comparison group of two monolingual Dutch girls from a low socioeconomic background; • a comparison group of two monolingual Turkish girls from a low socioeconomic background, living in Turkey. Twice every month an audio-recording of one hour was made of the speech of each child in the group of core informants: one aimed at the elicitation of Turkish speech, and one at the elicitation of Dutch speech. In order not to provoke codeswitching from the child as an artifact, recordings were made by a native speaker of the target language. In the two monolingual conditions, the same amount of data was collected in one language. The period of data collection covered the informants' age range from 2;0 to 3;6 years. During the recording sessions, the informants wore a jacket with a small microphone and a wireless transmitter. The informants' free conversations with peers and caretakers as well as their monologues were included in the recordings. Interactive storybook reading initiated by the researcher was one of the standard activities in each session. All recordings were transcribed and stored in the computer according to CHAT conventions (MacWhinney, 1991). Van der Heijden and Verhoeven (1994) presented preliminary results on the development of clause structure, lexicon, and reference to time and space during the first twelve months of data collection of one of the core informants, Berrin, who attended an international day-care center for five days a week. The data show that Berrin's dominant language was Turkish. In all domains analyzed, her level of proficiency in Turkish was higher than that in Dutch. The verb inflection system forms a case in point. In Dutch, she only realized infinitives and present tense forms in the third person singular, whereas in Turkish, she used standard devices for reference to present, past and future events with certain aspectual properties involved. Moreover, the rate of development in Turkish turned out to be higher than in Dutch. Van der Heijden (1997) provided evidence on the development of the class of verbs in Turkish by monolingual and bilingual children. The growth of the verbal lexicon was analyzed from both a quantitative and a qualitative point of view. It was found that in rate of acquisition, the monolingual children were somewhat ahead of the bilingual children. At every moment of measurement, the index of lexical richness was higher for the monolinguals. As regards the structure of acquisition of verbs, the two

36

Guus Extra and Ludo Verhoeven

groups of informants showed no differences. They applied the same types of over- and undermarkings. These parallels suggest that for the bilingual children, there was no evidence of transfer from Dutch to Turkish. This finding is in line with the analyses on nominal word formation in the development of both Turkish and Dutch (see Part 2 of this volume). In her doctoral thesis, Van der Heijden elaborated on the structural and functional aspects of the codeswitching behaviour of the four bilingual core informants. The informants proved to vary considerably in the amount of codeswitching they used. Nevertheless, clear parallels between the informants' codeswitching patterns could be established. As far as the structure of codeswitching in the data base was concerned, intersentential codeswitching was the most frequently used type by all informants. Furthermore, for all informants, Turkish was the preferred matrix language for all types of switching. In intrasentential codeswitching, most switches consisted of single nouns. Structural aspects of these informants' codeswitching patterns were compared with Turkish/Dutch youngsters' codeswitching patterns (Backus and Van der Heijden, 1998). Functional aspects of codeswitching, on the other hand, were analysed in a cursory fashion. Notwithstanding the informants' young age, different types of both situationally and stylistically motivated codeswitching, mentioned in previous studies, were found in the data base. In addition, codeswitching was also used as a strategy to express communicative needs when proficiency in the contextually required language was not sufficient.

Bilingual development at school age Background Recent studies of first language acquisition have made clear that, by the age of 4, children are in command of many of the grammatical principles and rules governing their native language (cf. Meisel, 1995). However, several studies have provided evidence that, both at the level of competence rules and performance preferences, language development goes on into the school years (Bowerman, 1979, 1982; Karmiloff-Smith, 1979, 1986; Hickman, 1995). Later language development in children can be characterized by a growing command of discourse principles. Discourse organization can be seen as the most significant domain of later language acquisition. Around the age of 5, developmental shifts take place from intra- to intersentential devices, and from basic structures to additional functions.

Processes of language change in a migration context

37

There are syntactic structures which children learn to understand only at a later age. A typical example is anaphoric reference. With respect to the difference between bound anaphora (e.g., himself) and free anaphora (e.g., him), Köster (1988, 1993) found that the understanding of bound anaphora comes soon and first, and that it takes more time to come to understand the referential function of free anaphora. Another late device is relativization. A variety of factors determines the degree of difficulty of interpreting relative clauses (Hakuta, 1981; MacWhinney and Pleh, 1988). The grammatical roles of the heads of relative clauses play an important role. Children who learn an Indo-European language appear to find subject-subject sentences relatively easy and subject-object sentences relatively difficult to understand, whereas object-object sentences and object-subject sentences take an intermediate position. A third type of construction that is understood relatively late by children is compound sentences. Clancy et al. (1976) found that coordination was understood before subordination. Subordinate sentences are particularly difficult to understand when the order of mention does not coincide with the order of events. An important shift in the process of language development at school age is the utilization of decontextualized language by children (Olson, 1977, 1994). This holds in particular for the production and interpretation of monologues such as personal narratives and formal texts. Olson described how monologues are marked by hierarchically built representations of ideas and logical connections between such ideas; in dialogues, speaker and interlocutor usually have access to a broad spectrum of contextual cues which are often completely absent in monologues. Heath (1983) showed how different text structures have different complexities for children. Descriptions of the here-and-now are relatively easy in a given situational context, e.g., when children are asked to give a description of a concrete event or experience. More difficult are narratives which refer to an experience which is only partially shared by the speaker and the interlocutor. The child has to take into account the foreknowledge of his interlocutor, the difference between matters of major versus minor importance, and cause-and-effect relations. Even more complex are narrative and informative texts which do not rely on the child's personal experiences. Karmiloff-Smith (1979, 1986) demonstrated which cognitive processes underlie the development of specific textual skills. Children have to learn that pronominal forms (such as personal or possessive pronouns) not only refer to concrete persons or objects, but are also used as cohesive devices in running texts. Nominal and pronominal devices are put to use in alternate

38

Guus Extra and Ludo Verhoeven

ways. Their internal order is dependent on the topical structure of a text (presentation of major versus minor characters) and on demands of the local text level, related to textual distance and avoidance of interference. If a character in a story is not referred to during a series of utterances, nominal constructions are needed for (re-)identification. The temporal framework also determines the cohesion of a story. Temporal oppositions do not only locate events in time, but they also contribute to the organization of the narrative structure. With respect to bilingual development at school age, empirical evidence on children's text (re)construction strategies is scarce. Little evidence is available on what kind of operating principles children use in the production and interpretation of texts, and on the role of language transfer and codeswitching at this later age. Design and results of the DSF project Part 2 of the DSF research program deals with the process of first and second language acquisition by Turkish and Moroccan children in the Netherlands at school age (cf. Verhoeven, 1993; Aarssen and Bos, 1996). The analysis of language development focuses on the children's organizing principles over spans of connected utterances. The narrative, in its broader sense, is taken as the relevant form of extended discourse. The following main questions were addressed: How are children's grammatical systems in first and second language use elaborated in the age range between 4 and 12, and how do children in this age range learn to master cohesive devices for anchoring discourse structure in narrative production? To answer these questions, first and second language data on children aged 4 to 12 years were collected. The former study on early bilingual development was confined to oral language data. Given the fact that, starting at age 7, the school curriculum is in large part devoted to reading and writing, written language data was explored as well. In this study on the bilingual development of Turkish and Moroccan children, the following linguistic domains have been taken into account: anaphoric reference, topic continuity, temporality, clause linking, and relativization. These domains have proved to be highly significant in a large number of crosslinguistic studies on language acquisition in both children (cf. Roeper and Williams, 1987; Slobin, 1985, 1989; Karmiloff-Smith, 1979) and adults (cf. Givon, 1983). Crosslinguistic attention is given to typological differences between Turkish, Arabic, and Dutch in each of these domains.

Processes of language change in a migration context

39

The present study has a pseudolongitudinal design, based on first and second language data collected with two groups of 50 Turkish, 50 Moroccan, and 50 Dutch children, who were 4 and 8 years old at the start of data collection, respectively. All the Turkish and Moroccan children attended a Dutch primary school and had been living in the Netherlands for at least two years. The Turkish children came from families in which Turkish is the preferred language, whereas the Moroccan children originated from primarily Moroccan Arabic-speaking families. For purposes of comparison of the processes of monolingual and bilingual development, additional data were collected among small groups of monolingual Turkish and Moroccan Arabic-speaking children in Turkey and Morocco, respectively. For the sake of control, different measures were taken. The socioeconomic and educational level of the parents of all the informants was low. The parents of the Turkish and Moroccan children in the Netherlands were part of the generation of immigrants who immigrated into the Netherlands during the seventies. Moreover, they originated from the same areas as the parents of the children in Turkey and Morocco. Each group of informants was tested four times with one-year intervals. The Frog Story (cf. Mayer, 1969; Berman and Slobin, 1994) was used to elicit narrative production. All story telling by the children was recorded and transcribed for further analysis. For the transcription of speech data, the procedures and conventions of the coding system of the Child Language Data Exchange System (CHILDES) were followed (MacWhinney, 1991). In addition oral language data, written samples based on series of pictures were collected among the two oldest age groups. The analysis of the narrative data started with a quantitative account of the distribution of referential devices in the children's narratives. In addition, a more qualitative analysis of micro-developmental changes in strategies for narrative cohesion was carried out. With respect to topic continuity, how children represented characters when these were introduced, reintroduced, or subsequently discussed was also examined. With regard to temporality, the tense and aspect variation in subsequent spans of utterances were explored. In the analysis of temporality, foreground and background elements were distinguished. As regards foreground elements, topic information was distinguished from focus information. In addition to the spontaneous oral and written language data, two experimental tasks were constructed. First of all, a comprehension task for anaphoric reference was developed. The children were presented with a one sentence/four-picture task, in which the sentence consisted of a possessive

40

Guus Extra and Ludo Verhoeven

head-modifier construction (e.g., The mother of the girl), combined with a verb and either a free anaphor (e.g., dresses her), or a bound anaphor {dresses herself). A similar task was developed by Deutsch et al. (1986). The second experimental task involved relativization. In this task, children were asked to act out the action described in a stimulus sentence. Four types of stimulus sentences were distinguished: subject-subject, subject-object, object-subject and object-object sentences. The results on the experimental data involving anaphora and relativization are presented in Aarssen et al. (1994), Aarssen (1996), and Bos (1997). Analysis of the results shows that Turkish and Moroccan children do not fall behind their monolingual Dutch peers. It seems that the pace of development of receptive skills in complex syntax of Dutch is more or less similar in first and second language learners. From a structural point of view, there is evidence that the two groups of learners rely on highly comparable intralingual strategies. With respect to anaphora, there was no significant difference in scores referring to free versus bound anaphora. As regards relativization, it was found that subject-subject and object-subject sentences were relatively easy to understand for all children, whereas object-object and subject-object sentences were relatively difficult to understand. This outcome corresponds with findings from earlier studies across Indo-European languages. The results on narrative development in a bilingual context are presented in Aarssen (1996), Bos and Aarssen (1996), and Bos (1997). With respect to topic continuity in narrative production, it was found that there was no significant difference in development over time for bilingual and monolingual children. The developmental strategies of the informants could be characterized as universal rather than dependent on language background or on whether the child was monolingual or bilingual. In all cases, children learn to establish new characters in narratives by means of the least presupposing device which in all languages under consideration is an indefinite NP. Given the condition that characters have already been mentioned, subsequent references are made by less explicit devices, such as zero anaphors (in Turkish) or pronouns (in Arabic and Dutch) for maintaining the character's identity in a chain of foregrounded activities. Reference shift by means of nominal forms occurred when a character was temporarily out of focus. The general conclusion was that the attribution of referential functions to characters in narratives is governed by pragmatic motivations and is not highly dependent on syntactic skills. With regard to temporality, the results were also comparable from a pragmatic point of view, although the devices used were rather different,

Processes of language change in a migration context

41

owing to language-specific characteristics. For the Turkish part of the study, a clear developmental trend was found towards the use of one and the same tense for foregrounding events in the narrative (past tense in Turkish, present tense in Dutch). Tense mixing was found to decrease over time. Across age groups, the bilingual children used an expanding set of forms to express simultaneous relations between events. Devices used for the connection of simultaneous episodes were not found in the data. Simultaneity was only expressed at the local level. The function of expressing simultaneity differed over the years. Four- and five-year-old children relied on deictic means for giving descriptions, whereas older children used more advanced devices of simultaneity, such as interrelating foregrounded events or relating foregrounded events to simultaneously backgrounded events. For the Moroccan part of the study, a similar developmental trend was found for time anchoring. The younger children tended to place their stories in the context of the here-and-now by using the present tense in Dutch and the inaccompli in Arabic. The older children, on the other hand, used the past tense in Dutch and the accompli in Arabic. Furthermore, there was a difference between bilingual and monolingual children in the use of temporal adverbials. This story-telling device was used less by the bilingual children. The adverbials they used were at the beginning of a main clause without bringing about a change in the basic structure of the sentence. The monolingual Moroccan Arabic and Dutch children were more inclined to use sophisticated adverbials, conjunctions, and particles that require knowledge of rules and consequences concerning subordination and inversion of basic word order than their bilingual peers. In this sense, the register of the bilingual children tended to be more restricted. In conclusion, the study on bilingual development at school age shows that, although bilingual children tend to be slower in the early stages than monolingual children, the processes of later language development tend to be highly comparable both from a structural and temporal point of view. However, a tendency to have a more restricted register in narrative production was evidenced on the part of the Arabic-Dutch bilingual children.

Codeswitching Background The phenomenon of codeswitching (henceforward CS) has already been referred to in the context of early bilingual development. CS is the alternate

42

Guus Extra and Ludo Verhoeven

use of two (or more) languages by the same speaker within the same conversation. It is a phenomenon that has been observed and documented in many of the bilingual or multilingual communities that have been studied by linguists (see Weinreich, 1953 for an early account). Poplack (1980) proposed the following typology of CS: • intersentential CS: at the boundaries of coordinate or subordinate sentences of the same or different speakers (the latter, e.g., in question/ answer sequences or in quoted speech); • intrasentential CS: within sentences of the same speaker; • extrasentential or emblematic CS: interjections like yes, no, I see, you know, this type of CS is hardly subject to syntactic constraints and can thus easily and frequently appear without violating syntactic rules. According to various studies of CS in different language contact situations, intrasentential CS occurs more often than intersentential CS (Berk-Seligson, 1986; Nortier, 1989). Moreover, intrasentential CS relates most commonly to single words, in particular content words and more in particular nouns. Nouns appear to be less subject to syntactic constraints than other word categories, and they have a clear link to cultural contents. A difficult and controversial issue is the distinction between CS and borrowing. Traditionally, CS has been distinguished from borrowing by referring in the latter case to the 'established integration' of elements from one language in another one. The problem with this definition is that there is no dichotomy between the presence or absence of integration, but a fluid continuum of different degrees of integration. Nortier (1989:184) proposed the following combination of criteria for the qualification of borrowing: borrowing refers to (1) single words or short frozen idiomatic expressions (2) which recur regularly, (3) which are generally accepted in the pertinent language community, and (4) which are commonly grammatically integrated in the language of this community. Borrowing often originates from the prestige of a particular language and commonly refers to sociocultural entities or activities which are typical for the speakers of that particular language. An important area of research in the domain of intrasentential CS relates to the question of linguistic constraints on CS, and, if so, to what degree such constraints are language-specific or have a universal base. Undoubtedly, CS emerges as systematic. Gumperz and Hernandez-Chavez (1975) concluded on the basis of the alternate use of English and Spanish by Hispanic Americans that CS does appear in particular contexts and not in others. Favourable sites for the appearance of CS were between a head noun

Processes of language change in a migration context

43

and a postponed relative clause, or between a subject and predicate in copula constructions. On the other hand, CS proved to be improbable between an auxiliary and main verb, between a main verb and a subject or object pronoun, between an article and a noun, and within a prepositional phrase. On the basis of empirical research in widely different language contact situations, a whole series of linguistic constraints on CS have subsequently been proposed which were thought to be universal (see Nortier, 1989:155-180 for a well-documented survey). It is, however, an open question to what degree the proposed constraints are independent of the typological distance between two languages. Boeschoten and Verhoeven (1987) and Nortier (1989) found numerous violations of proposed universal constraints in Turkish-Dutch and Arabic-Dutch language contact situations, respectively. In both conditions, the typological distance between the studied language pairs is much greater than between, e.g., English and Spanish. The question how CS occurs has led to a classification and quantification of CS typologies and to the formulation of language-specific and universal CS constraints. Complementary to this line of research, the question why CS occurs has led to distinguishing different functions of CS. Bilingual speakers and researchers, both often consider CS as a sign of language shift or language loss, but it can also derive from a consciously chosen conversational strategy (Gumperz, 1982; Romaine, 1995:142-161). Appel and Muysken (1987:118-121) distinguished six different CS functions: • a referential function: lack of knowledge of one language or lack of facility in that language on a certain subject; • a directive function: exclusion or inclusion of one or more interlocutors; • an expressive function: emphasis on a mixed ethnocultural identity; • a metaphorical function: highlighting the information conveyed; • a metalinguistic function: commenting directly or indirectly on the languages involved; • a poetic function: referring to the literary register of a text. These functions play different and partially overlapping roles in various contexts of CS. A final area of CS research is attitudes towards CS (Romaine, 1995:241187). Although bilingual speakers are not always aware of their CS behaviour, CS patterns are subject to conscious evaluations by both speakers and hearers in the form of negatively stigmatizing stereotypes (e.g., as a sign of lack of competence, education, or even manners). In Poplack's typology (1980), referred to earlier, intrasentential CS in particular leads to negative evaluations (Amuda, 1986).

44

Guus Extra and Ludo Verhoeven

Design and results of the DSF project The emphasis in the DSF project was on CS in contact between typologically distinct languages, i.e., between Turkish and Dutch, and between (Moroccan) Arabic and Dutch (see Backus and Boumans, 1996). The main question was: What patterns of CS can be observed in young adult informants from different generations? The data consisted of the spontaneous, informal interaction of 23 Turkish and 23 Moroccan informants, respectively, in the age range of 16-29 years and with a different migration history. The informants were not selected on an individual base, but as members of social networks. This was done so the informants would be more at ease and thus more likely to speak their vernacular. Moreover, they spoke to one another, as they did daily, rather than with outsiders, c.q. outside researchers. The instructions given to the informants was to try and audio-record conversations with a few people of the network who were present. The Turkish informants belonged to seven different networks and were distinguished in terms of first generation (arrived in the Netherlands at the age of 13 years or older), intermediate generation (arrived at an age between 5 and 12 years), and second generation (born in the Netherlands or under 5 at the time of arrival). All informants were between 16 and 25 years old at the time of recording; there were 14 female and 9 male informants. A subsample of 15 Moroccan informants was used for the Arabic-Dutch CS data analysis. Most of the informants were between 17 and 29 years at the time of recording, with the exception of three (two were younger and one was older). Apart from one informant, the informants were distinguished in terms of second generation and younger first generation. Depending on the informants' age at the time of recording, the number of years spent in the Netherlands varied considerably, i.e., between 2 and 19 years. There were 9 male and 6 female informants, distributed over 10 conversations of approximately 9 hours in total. Fragments of CS from the collected data base, were transcribed and coded. Coding was done in terms of direction (from Turkish/Arabic to Dutch or reverse), site (at the beginning of a speech turn, a new sentence or a clause, or at some other place), type (inter-, extra- or intrasentential CS), and element (single versus multiple-word units plus their specifications). The major results on Turkish-Dutch and (Moroccan) Arabic CS patterns are presented in Backus (1996) and Boumans (1998), respectively. Backus' data analysis is a follow-up of an earlier study (Backus, 1992) and based on a descriptive model, derived from two recent proposals in the literature, i.e.,

Processes of language change in a migration context

45

the Matrix Language Frame model developed by Myers-Scotton (1993) and the Code Copying model proposed by Johanson (1992, 1993) (see also Part 4 of this volume). The major dimension of Backus' model is a bipartite description in terms of total versus partial morphosyntactic integration. In the latter case of CS, certain integrating elements may be missing and/or may have been taken over from the other language. One and the same CS can also belong to both categories because the marking by means of Dutch elements may co-occur with integration in a Turkish constituent. It appears that different generations show different types of CS. First-generation informants commonly make use of isolated, single Dutch words which are prototypically inserted into Turkish sentences; these insertions are also prototypical from a semantic point of view: highly specific rather than general or abstract words dominate in the Dutch data. Intermediategeneration informants make also use of other CS types: apart from 'ordinary' insertions of, for example, single Dutch nouns, there are many examples of constituent insertions and CS at sentence and clause boundaries. Moreover, apart from highly specific word insertions, less prototypical and less predictable word insertions also appear. The same holds for second-generation informants. These informants are the only ones who show evidence of insertion in the opposite direction, i.e., by using Turkish words in Dutch utterances; these words are generally predictable in terms of specificity. One frequent type of insertional CS is highlighted, i.e., a bilingual construction based on a Dutch infinitive plus the Turkish auxiliary yapmak (= make, do). Backus' study shows the dynamic patterns of CS of a specific (Turkish) community in a specific (Dutch) immigration context. The first generation speaks mainly Turkish, with some highly specific Dutch words. The intermediate generation switches continuously between Turkish and Dutch, but shows only intrasentential CS when the matrix language is Turkish. The second generation makes use of multiple CS at sentence and clause boundaries. Intrasentential CS does not occur frequently in their language use, but if it occurs, it may go from Turkish to Dutch or vice versa. Generally speaking, the chance of a Dutch word in Turkish sentences is much greater than the reverse. Boumans (1998) studied the patterns of CS in Dutch-Moroccan Arabic conversations. The focus of his study was on the definition and operationalization of the concepts of 'Matrix Language' and 'Embedded Language' in insertional CS models. Derived from earlier proposals, Boumans opts for a Monolingual Structure Approach (MSA), in which the matrix language is exclusively defined on the basis of morphosyntactic properties of the matrix

46

Guus Extra and Ludo Verhoeven

structure, which may consist of a finite sentence or a sentence constituent. The constituent which contains one finite verb is referred to as the finite sentence or finite clause. The MSA is based on the assumption that all instances of CS can be described in terms of insertion, or, to put it differently, that, in all cases, a matrix language can be identified which can explain the distribution of elements in the matrix structure. Similar to Backus' findings and those of Nortier (1989), Boumans reports asymmetry in Dutch/Arabic insertion patterns in the sense that most insertions are Dutch content words in Arabic sentences. The overview presented (Boumans, 1998:334-336) clearly shows that there is an overwhelming majority of Dutch insertions in Moroccan Arabic sentences, in contrast with the reverse. There are hundreds of occurrences of Dutch nouns and verbs in Moroccan Arabic sentences versus 7 and 0 occurrences of the reverse, respectively. Compared to the insertion of Dutch content words in Moroccan Arabic sentences, the insertion of more complex constituents is rare. Arabic nominal constituents in Dutch sentences, however, appear to be equally frequent as the reverse. Boumans also goes into possible explanations for the observed asymmetry between CS patterns and discusses three types of factors: (1) differences in the sociolinguistic status of the languages under consideration (Moroccan Arabic being an IM language and Dutch being the superimposed language), (2) differences in the linguistic competence of the informants, and (3) differences in the language preference of the informants. The value of the studies by Backus (1996) and Boumans (1998) is that they focus on CS patterns in a dynamic context of immigration instead of a static condition of languages in contact, and that they offer an extensive description and evaluation of a wealth of data. Through these studies it becomes clear how different generations of IM groups make use of what is often unidimensionally referred to as CS.

Language shift and language loss Background Much of the research on language shift and language loss derives from social contexts in which the earliest acquired or first language is a dominated language and in which the later acquired or second language has the status of dominant language. Most commonly, the concepts of dominated and dominant language are equivalent with the concepts of

Processes

of language change in a migration context

47

minority and majority language, respectively, although such equality is not a universal phenomenon (see, e.g., Extra and Maartens, 1998). Native speakers of majority languages often do not speak or understand minority languages and, as a consequence, intercultural communication usually takes place in the majority language. Although the concepts of language shift and language loss are often used in the same breath, they actually refer to different dimensions of language change. Language shift is, in fact, the redistribution of different language varieties over different language domains, in interaction with different interlocutors. This process shows directionality in the sense that the functional redistribution of a first and second language occurs most commonly at the cost of the former. Language loss refers to an individual's diminishment of language proficiency. This process is also referred to as language regression or language erosion. By implication, there must have been a stage of fuller proficiency, which distinguishes the concept of language loss from impaired language learning by, for instance, IM children in a submersion context. Processes of language shift in IM groups are generally not the result of the necessity to understand or to be understood by the interlocutor, but the result of processes of change in sociocultural orientation. Fishman (1966, 1985, 1989) is one of the pioneers in this domain of research. The most important areas of research are the domains and the determinants of language shift (cf. Appel and Muysken, 1987:32-45). In the former case, the emphasis is on where language shift takes place (i.e., in which conditions of language use); in the latter case, on why such shift takes place. A wellknown example of domain analysis is Gal's study (1979) on the use of German and Hungarian in the Austrian border area Oberwart. Using a crosssectional design, Gal searched for patterns of language choice by speakers and interlocutors of different age, socio-economic status, and mutual familiarity. A substantial amount of our knowledge of processes of language shift is based on the outcomes of successive nationwide censuses in which questions have been asked about (home) language use (see Broeder and Extra, 1998, for an international comparison, and De Vries, 1989, for a methodological discussion of such demolinguistic studies). Intergenerational patterns of language shift have been extensively studied in English-dominant immigration countries, such as the USA, Canada, and Australia. A multitude of factors is held responsible for processes of language maintenance versus shift (cf. Fase et al., 1992; Clyne, 1991; Boyd, 1985).

48

Guus Extra and Ludo Verhoeven

Two types of approaches to these processes can be illustrated with earlier and recent studies by Howard Giles and his colleagues. Giles et al. (1977) introduced the concept of 'ethnolinguistic vitality' and distinguished three types of vitality factors, i.e., status factors, demographic factors, and institutional support factors. Status factors derive from the degree of economic, social, historical, and linguistic status of an ethnic community; demographic factors derive from group size, group spread, and the degree of interethnic marriages; and institutional support factors derive from the degree of support by mass media, religion, public services, and education. This list of factors, however, is neither exhaustive nor mutually exclusive. Other factors, such as gender and cultural distance between majority and minority groups, have been shown to play a role as well. On the basis of 1986 census data in Australia, Clyne (1991:85) found that processes of language maintenance versus shift were predominantly determined by group spread, gender, and marriage pattern. Western Europeans in Australia showed a faster shift towards English than Mediterranean Muslims or Eastern Orthodox Christians. According to Smolicz (1980, 1992), each ethnic community has certain values which are fundamental to its existence and self-definition. Language seems to be more important as a core value of cultural identity to some groups than to others. On the basis of this corevalue approach, Smolicz tried to explain why the Greeks in Australia maintain their language most and the Dutch least. The main criticism leveled at the former approach is that language use does not always coincide with the concept of a uniform and cohesive ethnic group. Giles et al. (1990) and Dow (1987, 1988) argued for studying variation in language shift factors between individuals rather than between groups and for focusing on the subjective perception of the shift factors mentioned earlier rather than on the objective status of these factors as such. The crucial question in their approach is to what degree objective vitality factors such as socio-economic status, group size, group spread, and institutional support are recognized by individual members of ethnic communities. With Finland as a notable exception, there is no long-standing tradition of periodically collecting and analyzing census data on (home) language use in Western Europe (see also the first chapter of this volume). If census data were collected at all, some countries never introduced language-related issues in their questionnaires, while others included such questions only with reference to specific indigenous minority languages in selected regions (e.g., France, Italy, Britain) or they dropped such questions for political reasons (e.g., Belgium). In spite of the increasing size of IM groups and

Processes of language change in a migration context

49

their descendants in Western Europe, most countries in this part of the world are only beginning to embark on larger-scale trend studies in the domain of language shift (cf. Boyd, 1985; Fase et al., 1992). Whereas the focus of studies on language shift is on determinants of the pace of shift in successive generations, the focus of studies on language loss is on the structure of regression or erosion in individual language users. The latter type of research has been given far less attention than the former (see Van Els, 1986, for a typology and overview of European language loss research). Compared to real-time studies on language acquisition, real-time studies on language loss are faced with two specific methodological problems related to the need for a point of reference. First of all, decrease in language proficiency generally takes more time than increase. It is easier to track language change with 'zero' proficiency as the point of reference than language change with 'full' proficiency as the point of reference. Moreover, it is easier to define 'zero' proficiency than to define 'full' proficiency. Jaspaert et al. (1986) discussed these and related methodological issues in operationalizing and measuring the nature and extent of processes of LI loss. There is a growing interest in studying processes of language loss (cf. Yagmur, 1997; Schoenmakers-Klein Gunnewiek, 1998). The focus in recent studies is not only on what is being lost, but also on how and why this happens. Although the traditional position that first language loss takes place in the reverse order of second language acquisition has generally been abandoned, similar questions remain for both types of language change: To what degree are these two processes characterized by language-specific versus universal characteristics, and what is the role of interlingual versus intralingual factors in these processes? Language loss is often qualified in terms of reduction or simplification of linguistic properties. Such process terms may relate to different levels of the linguistic system. Lexical reduction may occur in terms of relexification, a process by which lexemes of the original language are substituted by lexemes of the new language. Morphological reduction may occur in terms of simplification or overgeneratlization of morphological rules. And register reduction may occur in terms of a diminished variation in styles of language use in different situations. Studies of such phenomena have been carried out by Clyne (1991) and Dorian (1981) regarding Dutch and German in Australia and regarding Gaelic in Scotland, respectively.

50

Guus Extra and Ludo Verhoeven

Design and results of the DSF project In the Dutch Science Foundation project, Moroccan and Turkish adolescents in the Netherlands were taken into account. In this chapter, we will focus on the Moroccan informants (cf. El Aissati, 1997). For a comparison of the outcomes between Moroccan and Turkish informants, we refer to El Aissati and Schaufeli (this volume). The following question was addressed: What patterns of language loss can be observed in adolescent informants from different generations, in comparison with a reference group in the source country? The study was based on a cross-sectional design, with a kernel group of 25 Moroccan informants in the Netherlands and a reference group of 30 Moroccan informants in Morocco. The kernel group consisted of participants born in Morocco and in the Netherlands. There were 14 female and 11 male informants. All of them had Moroccan Arabic as a home language. The group was selected on the basis of duration of stay in the Netherlands (at least 5 years), level of education (6-9 years of schooling), and age (13-17 years). The reference group consisted of 19 male and 11 female informants. Their region of origin and level of education was similar to the kernel group. The following means were used to elicit data: questionnaires on language choice and language proficiency, and more or less experimental tasks on language perception and language production. The experimental tasks consisted of: • plural formation of singular nouns, with a focus on broken plurals and sound plurals in Arabic; • sentence interpretation, with a focus on four types of cues (NVN/NNV/ VNN word order, +/- animacy of nouns, +/- noun-verb agreement, +/noun stress); • perception and production of phonological oppositions, based on lists of same/different word pairs and word translation tasks, respectively; • perception and production of relative clauses, based on a comprehension task for anaphoric reference (cf. Deutsch et al., 1986) and a narrative production task (Mayer's 1969 Frog Story). The two tasks of perceiving and producing relative clauses were similar to the tasks used in the earlier reported study of bilingual development at school age. The resulting narratives were audio-recorded, transcribed, and coded for analysis. The questionnaires led to reported patterns of language choice and language proficiency. The Moroccan adolescents reported they used more Dutch than Moroccan Arabic in their everyday interactions, except when

Processes

of language change in a migration

context

51

their parents were involved. The reported language proficiency scores for both speaking and listening were higher for Dutch than for Moroccan Arabic. These results indicate that the informants are more likely than not to undergo a process of language shift in the Dutch immigration context. With respect to the experimental tasks, the following outcomes emerged. The plural noun formation task led to various overgeneralization strategies in the kernel group of informants, in terms of regularization of morphological rules. The cue validity experiment on sentence interpretation showed differences between the language processing strategies of the kernel group of Moroccans in the Netherlands and the reference group of Moroccans in Morocco; the main differences were detected at the level of noun animacy and the order of importance of the different word orders in agent assignment: for the kernel group, this order was SVO>SOV>VSO; for the reference group, SVO>VSO>SOV. Perception of phonological oppositions in Moroccan Arabic was more difficult for the kernel group than for the reference group in particular cases (/h/ versus /H/, /q/ versus fkl). Production of phonological oppositions led to sound substitutions in the kernel group: /q/ was substituted by /k/, ΙχΙ by /r/, /s/ by /s/, and 111 by /z/. The observed sound substitutions may be explained in terms of intralingual and/or interlingual effects of language loss. Perception of relative clauses in the comprehension task for anaphoric reference did not lead to significant differences between the kernel and reference groups, and the differences that were found between the two groups could not be interpreted in terms of language loss evidence. Production of relative clauses was studied on the basis of the Frog Story data; few relative clauses, however, emerged in this data, and their structure was rather simple. In sum, although the reported patterns of language choice and proficiency of the Moroccan informants in the Netherlands point in the direction of language loss, little evidence for this process could be found in the most experimental tasks. It remains an open question whether adolescents in a migration context actually lose any features of their language or proceed in their language acquisition at a slower pace than do their peers in the country of origin. In the latter case, as was mentioned before, 'language stagnation' would be a better process qualification than 'language loss'.

52

Guus Extra and Ludo Verhoeven

Note Apart from a variety of interim and forthcoming publications, the Dutch Science Foundation (NWO) program on Language and Minorities has led so far to five dissertations in the reported domains of research on processes of language change in a migration context (i.e., Aarssen, 1996; Backus, 1996; Bos, 1997; Boumans, 1998, and El Aissati, 1997). All dissertations have been published by Tilburg University Press in the series Studies in Multilingualism. All domains of research will be reported on in the following parts of this volume. The studies carried out in the Netherlands will be outlined in more detail by Van der Heijden (Part 2), Aarssen and Bos (Part 3), Backus and Boumans (Part 4), and El Aissati and Schaufeli (Part 5).

References Aarssen, J. (1996). Relating events in two languages. Acquisition of cohesive devices by Turkish-Dutch bilingual children at school age. Tilburg: Tilburg University Press. Studies in Multilingualism 2. Aarssen, J. and P. Bos (1996). Verhalen in twee talen: hoe Turkse en Marokkaanse kinderen verwijzen naar personages. In: H. Heeren, P. Vogel and H. Werdmölder (red.), Etnisch minderheden en wetenschappelijk onderzoek. Amsterdam/Meppel: Boom. Aarssen, J., P. Bos and L. Verhoeven (1994). Acquisition of complex syntax in Turkish and Moroccan children in the Netherlands. In: G. Extra and L. Verhoeven (Eds.), The cross-linguistic study of bilingual development. Amsterdam: North Holland. Amuda, A. (1986). Yoruba/English code-switching in Nigeria. Aspects of its functions and form. PhD thesis. University of Reading. Appel, R. and P. Muysken (1987). Language contact and bilingualism. London: Edward Arnold. Backus, A. (1992). Patterns of language mixing. A study in Turkish-Dutch bilingualism. Wiesbaden: Harrasowitz. Backus, A. (1996). Two in one. Bilingual speech of Turkish immigrants in the Netherlands. Tilburg: Tilburg University Press. Studies in Multilingualism 1. Backus, A. and L. Boumans (1996). Codewisseling en taalvariatie bij Turkse en Marokkaanse jongeren. In: H. Heeren, P. Vogel and H. Werdmölder (red.), Etnisch minderheden en wetenschappelijk onderzoek. Amsterdam/Meppel: Boom.

Processes of language change in a migration context

53

Backus, A. and H. van der Heijden (1998). Life and birth of a bilingual: the mixed code of bilingual children and adults in the Turkish community in The Netherlands. In: L. Johanson, in cooperation with E. Csato et al. (Eds.), Proceedings of the 7th International Conference on Turkish Linguistics (Mainz, augustus 1994). Wiesbaden: Harrassowitz. Turcologica 32. Bentahila, A. and E. Davies (1994). Two languages, three varieties: code-switching patterns of bilingual children. In: G. Extra and L. Verhoeven (Eds.), The crosslinguistic study of bilingual development. Amsterdam: North Holland. Berk-Seligson, S. (1986). Linguistic constraints on intra-sentential code-switching. A study of Spanish/Hebrew bilingualism. Language in Society, 15, 313-348. Berman, R. and D. Slobin ( 1994). Different ways of relating events in narratives. Hillsdale, NJ.: LEA. Boeschoten, H. and L. Verhoeven (1987). Language mixing in children's speech. Dutch language use in Turkish discourse. Language Learning, 37, 191-215. Bos, P. (1997). Development of bilingualism. A study on Moroccan children living in the Netherlands. Tilburg: Tilburg University Press. Studies in Multilingualism 8. Bos, P. and J. Aarssen (1996). Turkish and Moroccan children acquiring complex syntax. In: K. Sajavaara and C. Fairweather (Eds.), Approaches to second language acquisition. Jyväskylä: Jyväskylä University. Boumans, L. (1998). The syntax of codeswitching. Analysing Moroccan Arabic/Dutch conversations. Tilburg: Tilburg University Press. Studies in Multilingualism 12. Bowerman, M. (1979). The acquisition of complex sentences. In: P. Fletscher and M. Garman (Eds.), Language acquisition. Cambridge: Cambridge University Press. Bowerman, M. (1982). Starting to talk worse: Clues to language acquisition from children's late speech errors. In: S. Strauss (Ed.), U-shaped behavioral growth. New York: Academic Press. Boyd, S. (1985). Language survival: A study of language contact, language shift and language choice in Sweden. PhD thesis. Göteborg: University of Göteborg. Broeder, P. and G. Extra (1998). Allochtone talen in het voortgezet onderwijs. Bouwstenen voor lokaal taalbeleid. Den Haag: VNG Uitgeverij. Clancy, P., T. Jacobson and M. Silva (1976). The acquisition of conjunction. Papers on Research in Child Development, 12, 71-80. Clyne, M. (1991). Community languages. The Australian experience. Cambridge University Press. De Houwer, A. (1990). The acquisition of two languages from birth: a case study. Cambridge University Press. De Houwer, A. (1994). The separate development hypothesis: method and implications. In: G. Extra and L. Verhoeven (Eds.), The cross-linguistic study of bilingual development. Amsterdam: North Holland.

54

Guus Extra and Ludo Verhoeven

Deutsch, W., C. Koster and J. Koster (1986). What can we learn from children's errors in understanding anaphora? Linguistics, 24, 203-225. De Vries, J. (1989). On coming to our census. A layman's guide to demolinguistics. In: D. Gorter et al. (Eds.), Fourth International Conference on Minority Languages, Vol 1. Clevedon: Multilingual Matters. Dow, J. (Ed.) (1987). New perspectives on language maintenance and language shift I. International Journal of the Sociology of Language, 68. Dow, J. (Ed.) (1988). New perspectives on language maintenance and language shift II. International Journal of the Sociology of Language, 69. El Aissati, A. (1997). Language loss among native speakers of Moroccan Arabic in the Netherlands. Tilburg: Tilburg University Press. Studies in Multilingualism 6.

Extra, G. (1993). Language acquisition, shift, and loss of immigrant minority groups in Europe. In: B. Kettemann and W. Wieden (Eds.), Current issues in European second language acquisition research. Tübingen: Narr. Extra, G. and J. Maartens (Eds.) (1998). Multilingualism in a multicultural context. Case studies on South Africa and Western Europe. Tilburg: Tilburg University Press. Studies in Multilingualims 10. Extra, G. and L. Verhoeven (Eds.) (1993). Community languages in the Netherlands. Amsterdam: Swets & Zeitlinger. Extra, G. and L. Verhoeven (Eds.) (1994). The cross-linguistic study of bilingual development. Amsterdam: North Holland. Extra, G. and L. Verhoeven (1996). Processen van taalverandering bij etnische minderheidsgroepen. In: H. Heeren, P. Vogel and H. Werdmölder (red.), Etnische minderheden en wetenschappelijk onderzoek. Amsterdam/Meppel: Boom. Fantini, A. (1985). Language acquisition of a bilingual child. Clevedon: Multilingual Matters. Fase, W., K. Jaspaert and S. Kroon (Eds.) (1992). Maintenance and loss of minority languages. Amsterdam: Benjamins. Fishman, J. (1966). Language loyalty in the United States. The Hague: Mouton. Fishman, J. (1989). Language and ethnicity in minority sociolinguistic perspective. Clevedon: Multilingual Matters. Fishman, J. et al. (1985). The rise and fall of the ethnic revival. Berlin: Mouton de Gruyter. Gal, S. (1979). Language shift. Social determinants of linguistic change in bilingual Austria. London: Academic Press. Giles, H., R. Bourhis and D. Taylor (1977). Towards a theory of language in ethnic group relations. In: H. Giles (Ed.), Language, ethnicity, and intergroup relations. London: Academic Press. Giles, H., L. Leets and N. Coupland (1990). Minority language group status. A theoretical conspectus. Journal of Multilingual and Multicultural Development, 11, 37-55.

Processes of language change in a migration context

55

Givon, T. (1983). Topic continuity in discourse: A quantitative cross-linguistic study. Amsterdam: Benjamins. Grosjean, F. (1982). Life with two languages. An introduction to bilingualism. Cambridge (Mass.): Harvard University Press. Gumperz, J. (1982). Discourse strategies. Cambridge University Press. Gumperz, J. and E. Hernändez-Chavez (1975). Cognitive aspects of bilingual communication. In: E. Hernändez-Chavez, A. Cohen and A. Beltramo (Eds.), El lenguaje de los Chicanos. Arlington. Hakuta, K. (1981). Grammatical description versus configurational arrangement in language acquisition. Cognition, 9, 197-236. Heath, S. (1983). Ways with words: language, life and work in communities and classrooms. Cambridge (Mass.): Harvard University Press. Heeren, Η., Η. Entzinger, G. Extra, C. Groenendijk, L. Leertouwer and W. Shadid (1994). Prioriteitsprogramma etnische minderheden. Den Haag: NWO. Heeren, Η., P. Vogel and Η. Werdmölder (red.) (1996). Etnische minderheden en wetenschappelijk onderzoek. Amsterdam/Meppel: Boom. Hickman, M. (1995). Discourse organization and the development of refernce to person, space and time. In: P. Fletcher and B. MacWhinney (Eds.), The handbook of child language. Oxford: Basil Blackwell. Jaspaert, K., S. Kroon and R. van Hout (1986). Points of reference in first-language loss research. In: B. Weltens et al. (Eds.), Language attrition in progress. Dordrecht: Foris. Johanson, L. (1992). Strukturelle Faktoren in Türkischen Sprachkontakten. Stuttgart: Franz Steiner Verlag. Johanson, L. (1993). Code-copying in immigrant Turkish. In: G. Extra and L. Verhoeven (Eds.), Immigrant languages in Europe. Clevedon: Multilingual Matters. Karmiloff-Smith, A. (1979). A functional approach to child language. Cambridge: Cambridge University Press. Karmiloff-Smith, A. (1986). From meta-processes to conscious access: Evidence from children's metalinguistic and repair data. Cognition, 23, 95-147. Koster, C. (1988). An across experiments analysis of children's anaphor errors. In: G. de Haan and W. Zonneveld (Eds.), Formal parameters of generative grammar. Dordrecht: Foris. Koster, C. (1993). Errors in anaphora acquisition. PhD thesis. University of Utrecht (OTS). Leopold, W. (1970). Speech development of a bilingual child. Vols. 1-4. New York: AMS Press. MacWhinney, B. (1991). The Childes Project. Tools for analyzing talk. Hillsdale: Erlbaum. MacWhinney, B. and C. Pleh (1988). Relative clauses in Hungarian. Cognition, 29, 95-141. Mayer, M. (1969). Frog where are you? New York: Dial Press.

56

Guus Extra and Ludo Verhoeven

McLaughlin, Β. (1985). Second language acquisition in childhood. Hillsdale, N.J.: Erlbaum. McLaughlin, B. (1994). Retrospective on the study of early bilingualism. In: G. Extra and L. Verhoeven (Eds.), The cross-linguistic study of bilingual development. Amsterdam: North Holland. Meisel, J. (1990). Two first languages. Dordrecht: Foris. Meisel, J. (1995). Parameters in acquisition. In: P. Fletcher and B. MacWhinney (Eds.), The handbook of child language. Oxford: Basil Blackwell. Mougeon, R., E. Beniak and D. Valois (1985). Issues in the study of language contact. Evidence from Ontario French. Toronto: Ontario Institute of Education. Myers-Scotton, C. (1993). Duelling languages. Grammatical structure in codeswitching. Oxford: Clarendon. Nortier, J. (1989). Dutch and Moroccan Arabic in contact. Code switching among Moroccans in the Netherlands. Dordrecht: Foris. Olson, D. (1977). Oral and written language and the cognitive processes of children. Journal of Communication, 27, 10-26. Olson, D. (1994). Literacy and the making of the Western mind. In: L. Verhoeven (Ed.), Functional literacy: Theoretical issues and educational implications. Amsterdam: Benjamins. Poplack, S. (1980). Sometimes I'll start a sentence in English y termino en Espanol. Toward a typology of code-switching. Linguistics, 18, 581-616. Roeper, T. and E. Williams (1987). Parameter setting. Dordrecht: Reidel. Romaine, S. (1995). Bilingualism. Oxford: Blackwell. Schoenmakers-Klein Gunnewiek, M. (1998). Taalverlies door taalcontact? Een onderzoek bij Portugese migranten in Nederland en Frankrijk. Tilburg: Tilburg University Press. Slobin, D. (1985). The cross-linguistic study of language acquisition. Hillsdale: Erlbaum. Smolicz, J. (1980). Language as a core value of culture. Journal of Applied Linguistics, 11, 1-13. Smolicz, J. (1992). Minority languages as core values of ethnic cultures; A study of maintenance and erosion of Polish, Welsh, and Chinese languages in Australia. In: W. Fase, Κ. Jaspaert and S. Kroon (Eds.), Maintenance and loss of minority languages. Amsterdam/Philadelphia: John Benjamins. Taeschner, T. (1983). The sun is feminine. A study on language acquisition in bilingual children. Berlin: Springer. Van Els, T. (1986). An overview of European research on language attrition. In: B. Weltens et al. (Eds.), Language attrition in progress. Dordrecht: Foris. Van der Heijden, H. (1997). Verbs in mono- and bilingual acquisition of Turkish. In: K. Imer and N. Uzun (Eds.), Proceedings of 8th International Conference on Turkish Linguistics. Ankara, August 7-9, 1996.

Processes of language change in a migration context

57

Van der Heijden, H. and L. Verhoeven (1994). Bilingual development of Turkish preschool children in the Netherlands. In: G. Extra and L. Verhoeven (Eds.), The cross-linguistic study of bilingual development. Amsterdam: North Holland. Verhoeven, L. (1993). Acquisition of narrative skills in a bilingual context. In: B. Ketteman and W. Wieden (Eds.), Current issues in European second language acquisition research. Tübingen: Narr. Weinreich, U. (1953). Languages in contact. Findings and problems. Den Haag: Mouton. Yagmur, K. (1997). First language attrition among Turkish speakers in Sydney. Tilburg: Tilburg University Press. Studies in Multilingualism 7. Ytsma, J. (1995). Frisian as first and second language. Sociolinguistic and sociopsychological aspects of the acquisition of Frisian among Frisian and Dutch primary school children. Leeuwarden: Fryske Akademy.

Part 2 Early bilingual development

Early bilingual development: from elite to folk Suzanne Romaine

In principle, the acquisition of bilingual competence is no different from monolingual acquisition. Despite the predominance of bilingualism in the world's population, most of the research on children's language acquisition has been concerned with monolinguals rather than bilinguals. Moreover, most of it deals with the acquisition of English (see however Slobin, 1985, 1992), and is largely biassed towards middle class children. The study of early childhood bilingualism has been similarly biassed towards the study of 'elite' rather than 'folk' bilingualism. Many children have had no choice in becoming bilingual. For them, questions which have motivated most of the scholarly research into childhood bilingualism among relatively privileged populations, e.g., at what age and in what manner is it best to introduce a second language, are purely academic. The available literature on children's bilingualism is still fraught with methodological problems and does not yet provide a solid basis for answering decisively many of the questions which have guided research into bilingualism in this century, e.g., to what extent are the bilingual's two languages differentiated both at the conceptual and linguistic levels?, to what extent does bilingual acquisition parallel monolingual acquisition?, is it the case that a feature or category acquired from one language acts as a booster to its acquisition in the other?, and is there a cognitive advantage to bilingualism? In this chapter I discuss how narrow conceptions of bilingualism resulting from this research bias on elite bilingualism have dictated a number of important research questions relating to children's linguistic development. They have also had ramifications for bilingual education policies, as I will show later. Not surprisingly, in Europe, the US and many other parts of the world the educational achievement of (im)migrant children has generally lagged considerably behind that of indigenous children. This has gone hand in hand with the way in which bilingualism and bilingual education have been differently conceived of in majority and minority populations. Moreover, most of the linguistic research dealing with (im)migrants has looked at the acquisition of the majority (usually European) language of the respective host country, particularly among the adult population, rather than at migrant children's early linguistic development.

62

Suzanne Romaine

Given the diversity of the contexts in which children become bilingual, it has not been easy to extrapolate from the available studies, what the normal sequence of development might be for a child growing up in very different circumstances from those which have been studied. It is therefore not yet clear what constitutes delay or abnormality as far as bilingual acquisition is concerned. Investigators face the difficult issue of how to decide whether utterances deviating from adult monolingual norms reflect errors or are simply transitional stages in the normal development of bilingual competence. This is particularly the case in contexts of migration/immigration, where rapid processes of language change are affecting the maintenance of a language among the adult population and the transmission of it to the younger generation is thereby disrupted. There is already clear evidence in the Netherlands and elsewhere to indicate that children do not reach native-like proficiency in their parents' language(s). Yet it is by no means clear exactly what this means in both the short and long term, or what can or should be done about it.

Definitions of bilingualism and types of childhood bilingualism Bilingualism has been often been defined, described and measured in terms of categories, scales and dichotomies such as elite versus folk, ideal versus partial bilingualism, coordinate versus compound bilingualism, which are related to factors such as proficiency, function or context of acquisition. I use the term bilingual broadly here to include cases involving the acquisition of two as well as more than two languages/varieties. I also regard bilingualism as a relative concept (Romaine, 1995:1.3). No doubt influenced by the concentration of mainstream linguistic theory on the abstraction of the ideal speaker-listener, in a completely homogeneous speech-community, who knows its language perfectly (Chomsky, 1965:3), the study of bilingualism has been plagued with a tendency to elevate the ideal bilingual to some special theoretical status as a person who has native-like control of two languages, as Bloomfield (1933:56), for example, put it. Other concepts, such as balanced bilingualism, have followed from the assumption that bilingual persons are composed of two monolinguals. Fishman (1971:560), however, has cautioned against the notion of balanced bilingualism in more general terms. He says that bilinguals are rarely equally fluent in both languages about all possible topics. This reflects the fact that the allocation of functions of the languages in society is normally imbalanced and in complementary distribution. Any society which produced functionally balanced bilinguals who used

Early bilingual development: from elite to folk

63

both languages equally well in all contexts would soon cease to be bilingual because no society needs two languages for the same set of functions. Because of the inherent connection between proficiency and function, it is doubtful whether bilingualism per se can be measured apart from the situation in which it functions for a particular individual. Yet much of the research concerning the advantages and disadvantages of bilingualism and bilingual education has tried to ignore the social context in which bilingualism develops and changes over time. Related concepts such as dominance also have to seen in context. Cooper (1971), for instance, found that Spanish/ English bilinguals had different scores on word naming tasks depending on whether the domain was family, neighbourhood, school etc. In some domains they would have been rated as balanced bilinguals, while in others they would not. The age at which acquisition takes place can of course have consequences for the level and kind of skills that develop. Thus, it has been customary within the literature on children's early bilingual development to make a distinction between simultaneous and sequential acquisition. In the former case researchers have referred to the acquisition of 'bilingualism as a first language' (Swain, 1972) or 'bilingual first language acquisition' (Meisel, 1990). McLaughlin (1978) suggests that acquisition of more than one language up to age three should be considered simultaneous, but he readily admits this criterion is arbitrary. Padilla and Lindholm (1984) argue that we should speak of simultaneous acquisition of two languages only when a child has been exposed to two languages from birth onwards. There are at least two arguments for maintaining such a strict criterion. One is that anything a child learns in one language might have a subsequent effect on the language learned later. The second is to ensure adequate comparisons with monolingual children. Otherwise, we will not be able to disentangle two variables from one another, i.e., age of first exposure and exposure to two languages. Unfortunately, studies often do not describe their subjects' linguistic history or input and exposure patterns precisely enough to decide whether we are dealing with a strict case of simultaneous acquisition or McLaughlin's looser sense. It is not clear how helpful this distinction is in the (im)migration context, where for various reasons there may be a continuum of types ranging from simultaneous to sequential acquisition taking place even within the same family. Studies of immigrant bilingualism have often noted how older children's proficiency in the parents' language is greater than that of the younger children. Children's place of birth is an important factor too in determining patterns of bilingual acquisition. I will have little to say about consecutive or successive bilingualism here. Generally, this is considered to

64

Suzanne

Romaine

be the subject matter of the field of second language acquisition, though the boundaries between it and the study of bilingualism are arbitrary. In my book I attempted to classify the main types of early childhood bilingualism which have been studied into six categories, depending on factors such as the native language of the parents, the language of the community at large and the parents' strategy in speaking to the child (Romaine, 1995:5.1). I gave each of the six types a brief descriptive name based on some aspect of the strategy employed by the parents, e.g., one person-one-language or one context-one-language. My motivation was to organize the available studies into a framework that might yield some generalizations, despite the fact that most of the families whose children's bilingualism was studied in detail were living in monolingual rather than bilingual societies. The sixth strategy which I called mixed languages leads (arguably) to more mixing and interference than the other types, but it is nevertheless probably the most frequently occurring context for natural bilingual acquisition in multilingual societies. Unfortunately, it has also been the least systematically studied and most researchers treat mixing as a stage in the child's development which must be overcome if the child is to be a 'true' bilingual. Weinreich's (1968:73) view of the ideal bilingual who "switches from one language to another according to appropriate changes in the speech situation (interlocutors, topics, etc.), but not in an unchanged speech situation" does not readily accommodate the norms of many bilingual communities where intrasentential codeswitching is frequent, yet do not conform to this ideal. If we accept Weinreich's view, we would have to conclude that the majority of bilingual speakers are not ideal bilinguals and that they have less than nativelike control of both languages. Both within and outside the communities concerned there has been a tendency to regard language mixing and codeswitching as signs of imperfect competence rather than as legitimate modes of communication in their own right. These narrow definitions of bilingualism and of ideal bilingual behaviour have also had a number of consequences, particularly in the field of bilingual education, but they have also dictated to some extent a number of important research questions relating to children's bilingual development. One example is the issue of whether the bilingual child's languages are separate or merged at particular stages of development. While many researchers agree that the child has only one system in the early stages, others have claimed that differentiation is present from the early stages. This presupposes that there are two separate languages to be acquired. In some cases it is not clear whether there are in fact two separate languages to be acquired because the adults to

Early bilingual development: from elite to folk

65

whom the child is exposed always use a code-mixed variety. This is especially the case in (im)migration contexts, where contact with the dominant language has brought about changes in the (im)migrant language. For the most part this has not been an issue in the study of elite bilingualism, where the prevailing research paradigm has been that of the linguist parent studying the bilingual acquisition of his or her own children in situations where the languages being acquired are clearly separate at both the individual and community level. Leopold, for instance, whose four volume study (1939-49) of his daughter Hildegard's acquisition of German and English in the US set the trend for future studies, did not regard Hildegard as really bilingual during the first two years because she did not show awareness of the two languages as separate. In examining the role of the mixed utterance in the child's early bilingual acquisition, I have been reminded of what Lavandera (1978) some time ago referred to as the variable component of bilingual performance. Commenting on the communicative competence of speakers whose ordinary mode of conversing involves the use of two or more codes, plus the ability to mix or switch between them, Lavandera (1978:391) observes that the structure of each code taken separately is usually reduced in some dimension. This is of course a familiar enough finding. There has been a tendency to regard bilingual competence as the sum of the acquisition of monolingual competence in each of the two languages rather than as a unitary system which allows the pooling of resources across two languages. Researchers have often noted that bilingual children lag behind monolinguals in terms of vocabulary development. It is more revealing, however, to look at the bilingual's total vocabulary across the two languages since a bilingual's total vocabulary may exceed that of a monolingual child. It does not make much sense to assess bilinguals as if they were two monolinguals since it is unlikely that a bilingual will have the same experiences in both languages. The bilingual child's capacity to produce new words is split between the two languages. The child deals with this by giving priority to new words at the expense of equivalents. Thus, by using the total lexicon available at any given time, s/he is able to speak both languages and to denote the same number of new concepts as the monolingual child. If such a speaker's ability is evaluated in a situation where she is forced to stay within a single code, e.g., in contacts with members of a monolingual community, this speaker's communicative competence will see less rich than it actually is. Speakers' repertoires are fully exploited in those multilingual settings where they can call on resources from each of the available codes and on strategies for switching between them. Thus, codeswitching is a mode of bilingual performance which allows the bilingual to display his or her full

66

Suzanne

Romaine

communicative competence. As a case in point, Lavandera discusses what happens when Italian Argentines have to communicate with monolingual Spanish speakers in an imperfectly acquired variety of Spanish. Even though their total communicative repertoire includes much more than this variety, they are not able to draw on the full stylistic resources and choices which monolinguals have available to them. Lanza's (1997) research, in particular, bears out Lavandera's claims. In her study of the bilingual acquisition of Norwegian and English in Norway she measured MLU (mean length of utterance) over time in English, Norwegian as well as in mixed utterances used by Siri in conversations with her mother and father. Her data show that in conversations with her mother Siri's MLU in English ranged from a low of 1.05 at stage one to 1.92 at the last stage measured, with the highest value reached of 2.06. In Norwegian the values ranged from a low of 1.0 to a high of 2.0. Siri's mixed utterances, however, were at each stage of development more complex than her monolingual utterances in either Norwegian or English. They ranged from 2.07 to 4.43. Similarly, in conversations with her father, Siri's mixed utterances, were more complex than her monolingual ones in either language. The pattern of development for her monolingual utterances is the mirror image of that found in conversations with her mother; namely, those in Norwegian are more complex than those in English, although both languages develop over time. In English Siri's MLU values range from 1.00 to 2.50, and in Norwegian from 1.08 to 2.93. Other researchers too such as De Houwer (1990), Döpke (1992), Gawlitzek-Maiwald and Tracy (1996) have recognized the importance of looking at mixed utterances as well as the parents' strategies in addressing the child. Even though we have known for some time that input is a critical variable in the child's linguistic development, in my typology of childhood bilingualism I did not treat as a separate issue the actual rather than the stated strategy used by the parents in addressing the child, simply because there was little discussion of it in the literature. Both Döpke (1992) and Lanza (1997) show that the strategy used by the minority speaking parent is crucial for the development of the language least likely to receive support otherwise. While the children studied by Döpke and Lanza were all middle class, there were considerable differences in the extent to which the parents adhered strictly to the one person-one-language principle. These were reflected in differing degrees of development of the non-dominant language. Unfortunately, uninformed officials in schools and health care continue to dispense advice to (im)migrant parents that they should switch from their home language in order to aid their children's acquisition of the majority language. Many

Early bilingual development: from elite to folk

67

professionals such as speech therapists view normal language mixing as harmful and are therefore liable to give advice to parents which is not in keeping with the realities of normal bilingual development in bilingual communities elsewhere. Beliefs about bilingualism causing stuttering and delayed onset of language are also widespread, despite lack of evidence for them.

Educational implications for early bilingual development among migrant children It is a commonplace that most European countries have been slow in recognizing the changing face of Europe in the latter decades of this century. The prevailing educational strategy has been one of assimilation in the belief that problems arising from cultural and linguistic differences would be shortlived. In so far as support for mother tongue teaching was envisaged, it was seen as a prerequisite for repatriation. From an early stage, however, the Council of Europe took an interest in the education of children of so-called migrant workers. The 1977 Directive of the Council of the European Community on the education of the children of migrant workers (Brussels 77/486/EEC) instructed member states to "take appropriate measures to promote the teaching of the mother tongue and of the culture of the country of origin of the children of migrant workers, and also as part of compulsory free education to teach one or more of the official languages of the host state." Like other policy statements relying on notions like mother tongue, home language or native language, the issue of who has a mother tongue or home language is fraught with political problems. International recommendations such as that of UNESCO in 1953 suggested that use of the mother tongue be extended to as late a stage in education as possible. If the term is interpreted strictly as the language of the home, very few children in any part of the world, monolinguals included, are instructed in their home language or mother tongue. When various minority groups campaign for provision of so-called mother tongue teaching, the question of what one's mother tongue or home language is designated to be can be crucial because it determines who has a right to education in a particular language. In the US at least one state (Massachusetts) has had to develop programs of bilingual instruction for creole speaking children, whose 'home language', Cape Verdean Creole, is not recognized in their country of origin. In the Cape Verde Islands the

68

Suzanne

Romaine

official language of education is Portuguese. Like most pidgins and Creole, Cape Verdean Creole has not been standardized. The measures taken within Europe to comply with the directive varied considerably from country to country, as did the philosophy underlying them and their manner of implementation, reflecting in part the differing political statuses of migrants. In most continental European countries the children are still for the most part technically foreigners, while in the UK and Sweden, for instance, they have citizenship. In 1986 the Commission of the European Communities set up various pilot projects on the education of migrant workers' children. The various pilot projects ranged from the development of teaching materials for Turkish in Berlin to pre-school education of immigrant children and their mothers in Denmark (see Reid and Reich, 1992, for a comparative evaluation and critical appraisal). Teunissen's (1992) report on equality of educational opportunity shows how ethnic minority children drop out earlier, obtain lower qualifications, and are overrepresented in special education programs. Even now schools are actually widening the gap in educational results between different social and ethnic groups. Less effective schools tend to be found in areas with high concentrations of ethnic minority children and children from lower socio-economic background. The economic returns from schooling are in general much greater for those who are advantaged to begin with. Few educational systems in either the US or Europe accept the child's bilingualism as a starting point to build on and as a resource across the school curriculum. Bilingualism is rarely an explicit educational goal in either of these places. None of the early pilot projects undertaken by the EC members states contributed substantially to a reduction of the status differences between the languages of host and sending countries. This, however, in my view is a prerequisite for bilingual education across the curriculum. It is extraordinarily wasteful to spend resources on teaching the languages of ethnic minorities on a limited basis only to those of a particular minority background when it is widely recognized from research findings that bilingualism can be an intellectually enriching experience under the right learning circumstances. This illustrates yet another irony of the conflict over bilingual education. Opposition to it has occurred side by side with increasing concern over lack of competence in foreign languages. Bilingual education has been equated with ethnicity, poverty and compensatory education, while foreign language instruction in mainstream schools in the world's major languages has been seen as valuable, both economically and culturally. Bilingual education needs to be reconceptualized as a strategy for all students and not as a special instructional program for disadvantaged minorities. There

Early bilingual development: from elite to folk

69

is a need for bilingual immersion programs which simultaneously meet the language education needs of majority and minority students. At the moment though, due to status differences, there is little enthusiasm for the languages of ethnic minorities, even when the language concerned is a world language like Spanish, as is the case in the US, or Arabic in the Netherlands. There is often little enthusiasm among minorities themselves, a point which the majority uses as effective propaganda to justify not teaching these languages. Because delays in second language acquisition are often associated with bilingual development, some teachers are convinced that bilingualism itself is actually the cause of poor educational achievement. Often children are caught in a vicious circle. Because the school fails to support the home language, skills in it are often poor. At the same time they do not progress in the new language at school, and are labelled semilingual, also a politically loaded term (see Romaine, 1995:6.4). Thus, the failure of the school to let children develop further in their mother tongue is often used to legitimize further oppression of it (see Skutnabb-Kangas, 1984:19). The term semilingualism, as I have argued elsewhere, is a politically loaded concept based on questionable assumptions about language proficiency and how it is measured (see Martin-Jones and Romaine, 1985). It is no accident that the term has emerged in the discussion of populations which are disadvantaged to begin with. If all we mean by semilingualism is that a bilingual's languages are unequally developed, then the term is not very helpful. The search for the true balanced bilingual with native-like fluency depicted in some of the literature on bilingualism is elusive. The notion of balanced bilingualism is an ideal one, which is largely an artefact of a theoretical perspective which takes the monolingual as its point of reference. Studies purporting to show that bilingualism is associated with lower IQ, poorer achievement in language tests, etc. have been obtained largely in connection with subtractive bilingualism in submersion type programs, where the populations concerned are undergoing language shift. Immersion programs of the Canadian type usually result in additive bilingualism. The child's native language is intact and develops, even though the child has not had the same amount of instruction as its monolingual peers in majority language schools. It adds a second language without threatening the first. Practically all the cases of childhood bilingualism studied in detail have been of this type, and most of the claims for positive results of bilingualism, e.g., greater metalinguistic awareness, cognitive flexibility, etc., have been obtained from this kind of acquisitional context. Indeed, Paulston (1994) has recently identified context as a major determinant of the results of bilingual education, in particular, whether the

70

Suzanne Romaine

instructional program takes place in a situation of language shift or maintenance. Nevertheless, in an earlier report to the Swedish National Board of Education (Paulston, 1982), she noted that although home language instruction was a positive experience for immigrant children, it was strictly speaking not necessary since the situation in which it took place was one of very rapid shift to Swedish. No other country has taken upon itself home language instruction to immigrant children on such a massive scale as Sweden did. In recent years, however, home language instruction has been greatly reduced. Paulston (1994:31) then goes on to state that once it has become clear "whom the children of migrant workers in Europe will marry, the setting of educational policy will be much facilitated. If they commonly marry nationals of the host culture, there will be no need of special or different educational policies for their children. If, however, they marry exclusively within their own ethnic group, learn the national language poorly and show other trends of strong cultural maintenance (arranged marriages with partners from the home country, vacations in the home country, etc.), then a strong case can be argued for ... bilingual education." There is no time to wait and see who will marry whom. Within Europe all of these factors must be weighed up in light of recent events such as the acceptance of the Maastricht treaty which came into effect in November 1993, thus ensuring mobility of workers at all social levels. This means that the social character of migration will change and all member states will be potential receivers and senders of migrants. At the same time the developed technological economy requires an increasingly highly educated work force. The discourse in which migration and its 'problems' are discussed must change to reflect these new social realities. An index of the present state of affairs can be seen in the differing labels attached to various educational programs, e.g., intercultural, multicultural, multiracial education, pluralism or the German Ausländerpädagogik ('pedagogy for foreigners') and of course to the people at whom such efforts are aimed, e.g., migrant workers, immigrants, foreigners, guest workers, etc. The term 'migrant' itself (and the stigma attached to it) will need to be abandoned or the concept and name must be revalorized. Future educational policies cannot be based on a distinction between foreigners and natives. Those involved in planning and implementing educational policies of course stress the necessity of evaluating the economic costs involved in different types of programs relative to their aims and outcomes. However, it is well known that policies which might seem divergent can in fact have the same outcome. Moreover, what looks like the same educational policy can

Early bilingual development: from elite to folk

71

have very different outcomes, depending on the context. Some good examples can be seen in the US and the Netherlands. In the US, for instance, Ramirez, Yuen and Ramey (1991) found that Spanish-speaking students could be provided with substantial amounts of Spanish instruction without impeding their acquisition of English. However, they also found little difference between early exit and English immersion students. In the Netherlands Valien and Stijnen (1987:119) observe that most nonindigenous pupils perform poorly at almost all levels in all types of strongly monolingual Dutch-oriented programs. However, they also argue that children participating in assimilation, transition or maintenance programs show no substantial differences in the level of Dutch proficiency attained. Only occasionally do better results in Dutch occur in bilingual programs. Although there is no educational argument in favour of the bilingual programs as far as acquisition of Dutch is concerned, they conclude (1987:122) that there is an economic justification for bilingual instruction since more or less the same levels of proficiency can be obtained, even though less time is spent teaching Dutch. Moreover, students have a chance to attain some level of competence in their own language. Nevertheless, one could also use the economic argument to justify teaching only through Dutch since provision for minority languages generally involves additional expense. Research findings such as these can be taken to support two very different positions: one is that instruction in a minority language does not retard progress in the majority language. The other is that similar levels of attainment in English and Dutch can be achieved by ignoring the children's first language. Disregarding the child's language is often easier since this is already what happens in many cases anyway. This conclusion, however, ignores the less easily measurable social costs of various policies. If we calculate the long-range social and economic cost of continuing the present pattern of under-educating minority children in Europe and elsewhere (the US being a good case in point), the results are enormous. The projections for increasing diversity in the United States in the next century indicate that Hispanics alone may constitute over 30% of the total population. It is these children who will become the majority and upon whom the economic burden will fall of caring for the next generation of children and the previous generation soon to be retirees. Already in the EC 10% of the school age population have a culture and language different from that of the majority of the country in which they reside. This figure obscures wide variation among member states. In the Netherlands, for instance, Extra and Verhoeven (1993:72) say that the influx of ethnic minority children in elementary schools

72

Suzanne Romaine

in the four largest cities is presently about 40% and will increase to more than 50% in the year 2000. As events of the past few years and the present in Eastern Europe show, the Europe of the future must be truly multicultural or it will not exist at all.

References Bloomfield, L. (1933). Language. New York: Holt. Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, Mass: MIT Press. Cooper, R.L. (1971). Degree of bilingualism. In: J.A. Fishman, R.L. Cooper and R. Ma, Bilingualism in the Barrio. Bloomington: Indiana University Press. De Houwer, A. (1990). The acquisition of two languages from birth: A case study. Cambridge: Cambridge University Press. Döpke, S. (1992). One Parent, One Language. An Interactional Approach. Amsterdam: John Benjamins. Extra, G. and L. Verhoeven (1993). A bilingual perspective on Turkish and Moroccan children and adults in the Netherlands. In: G. Extra and L. Verhoeven (Eds.), Immigrant Languages in Europe. Clevedon, Avon: Multilingual Matters. Fishman, J. A. (Ed.) (1971). Advances in the Sociology of Language. The Hague: Mouton. Gawlitzek-Maiwald, I. and R. Tracy (1996). Bilingual Bootstrapping. Linguistics, 34 (5), 901-927. Lanza, E. (1997). Language Mixing in Infant Bilingualism. Oxford: Oxford University Press. Lavandera, B.R. (1978). The variable component in bilingual performance. In: J.E. Alatis (Ed.), International Dimensions of Bilingual Education. Washington, DC: Georgetown University Press. Leopold, W. (1939). Speech Development of a Bilingual Child: A Linguist's Record. Vol.1: Vocabulary growth in the first two years. Evanston, II.: Northwestern University Press. Leopold, W. (1947). Speech Development of a Bilingual Child: A Linguist's Record. Vol.11: Sound learning in the first two years. Evanston, II.: Northwestern University Press. Leopold, W. (1949a). Speech Development of a Bilingual Child: A Linguist's Record. Vol.111: Grammar and General Problems. Evanston, II.: Northwestern University Press. Leopold, W. (1949b). Speech Development of a Bilingual Child: A Linguist's Record. Vol.IV: Diary from Age 2. Evanston, II.: Northwestern University Press. Martin-Jones, M. and S. Romaine (1985). Semilingualism: A half-baked theory of communicative competence. Applied Linguistics, 6, 105-117. McLaughlin, B. (1978). Second-language acquisition in childhood. Hillsdale, NJ: Lawrence Erlbaum Associates.

Early bilingual development: from elite to folk

73

Meisel, J. (1990). Grammatical development in the simultaneous acquisition of two first languages. In: J. Meisel (Ed.), Two First Languages - Early Grammatical Development in Bilingual Children. Dordrecht: Foris. Padilla, A.M. and K. Lindholm (1984). Child bilingualism: the same old issues revisited. In: J.L. Martinez Jr. and R.H. Mendoza (Eds.), Chicano Psychology. Orlando: Academic Press. Paulston, C.B. (1982). Swedish research and debate about Bilingualism. A critical review of the Swedish research and debate about bilingualism and bilingual education in Sweden from an international perspective. Stockholm: National Swedish Board of Education. Paulston, C.B. (1994). Linguistic Minorities in Multilingual Settings. Implications for Language Policies. Amsterdam: John Benjamins. Ramirez, J.D., S.D. Yuen and D.R. Ramey (1991). Final Report: Longitudinal study of structured English immersion strategy, early-exit and late-exit programs for language-minority children. Report submitted to the US Department of Education. San Mateo, CA: Aguirre International. Reid, E. and H. Reich (Eds.) (1992). Breaking the Boundaries. Migrant Workers' Children in the EC. Clevedon, Avon: Multilingual Matters. Romaine, S. (1995 [2nd ed.]). Bilingualism. Oxford: Blackwell. Skutnabb-Kangas, T. (1984). Bilingualism or Not: The Education of Minorities. Clevedon, Avon: Multilingual Matters. Slobin, D.I. (Ed.) (1985/1992). The Crosslinguistic Study of Language Acquisition. 3 Vols. Hillsdale, NJ: Lawrence Erlbaum Associates. Swain, M. (1972). Bilingualism as a First Language. PhD thesis. University of California, Irvine. Teunissen, F. (1992). Equality of educational opportunity for children of ethnic minorities. In: E. Reid and H. Reich (Eds.), Breaking the Boundaries. Migrant Workers' Children in the EC. Clevedon, Avon: Multilingual Matters. Valien, Τ. and S. Stijnen (1987). Language and educational success of indigenous and non-indigenous minority students in the Netherlands. Language and Education, 7, 109-124. Weinreich, U. (1968). Languages in Contact. The Hague: Mouton [1st edition 1953, New York: Linguistic Circle of New York Publication No. 2],

Environmental factors in early bilingual development: the role of parental beliefs and attitudes Annick De Houwer1

Parents of preschool children who are exposed to two languages from an early age are often baffled by the fact that their children do not actually use two languages in everyday interaction, but limit themselves mainly to one, although they do seem to understand two languages perfectly. Very little is known in the literature about these young passive bilinguals. Studies of early bilingual acquisition have mainly recruited and reported on young children who actually speak two languages (for a recent overview, see De Houwer, 1995). Most of these studies have been more psycholinguistic in their aims and objectives, and in this context it is easy to see why children are chosen that actually speak two languages. There has been very little research, however, that is specifically geared to answering the more sociolinguistic question of why it is that some children regularly exposed to two languages from a very young age actually start to speak and continue to actively use these languages, and other children in similar circumstances do not (but see Döpke, 1992). Yet an understanding of the mechanisms that lead to the active use of language in early bilingualism is important for a greater understanding of some of the processes that lead to language change and language shift. After all, children who are being raised bilingually, but do not in fact produce two languages, will not likely speak two languages to their offspring once they have become parents themselves. If this pattern is in evidence for a large group of people within more or less the same community at the same time, then on a more global scale one may get a clear pattern of language shift. On the other hand, children from linguistically mixed marriages who become early active bilinguals are in a position to forge relationships with other active bilinguals using the same languages. As a result, they can help create new bilingual communities (cf. also Heller and Levy, 1992), thus once more contributing to processes of language change and shift. This chapter aims to explore some possible environmental factors that help determine whether very young children will grow up to speak two languages from a very tender age rather than just one. In this exploration, I will be focusing particularly on the possible role of parental beliefs and attitudes. As

76

Annick De Houwer

I will try to explain, such beliefs and attitudes can be seen to lie at the basis of parents' language behaviour towards their children, which in turn is a powerful contributive factor in children's patterns of language use. I have chosen to focus specifically on parental attitudes, beliefs and language behaviour because of the particularly significant role of parents in young children's lives (see Sigel, 1985b). While I do not wish to underestimate the role of other people besides parents in a child's linguistic development, I start from the assumption that what happens to children in infancy and in later childhood ultimately depends on their parents. In my opinion this is the case even in societies where parents soon leave the care of their offspring to third parties, since the fact that parents see fit to do so other than in emergency situations is in itself part of parents' belief system about their own roles as caretakers. In what follows, I will to a large extent make abstraction from the world outside the parents-child unit, although it is obvious that such an abstraction does not correspond to the full complexity of reality. In order to better understand that complexity, however, it is necessary to focus attention on its components.

Parents' linguistic behaviour and children's degrees of bilingualism Young children may exhibit varying degrees of active versus passive bilingualism. Pupier (1982), for instance, notes that for any of their languages, children growing up in a bilingual input situation may show the following usage patterns: • children may simply actively refuse to speak a particular language, whether in response to it or in initiations; • children may use avoidance strategies, and may disassociate themselves from a particular situation in which they might be called upon to use a particular language by leaving a room, or starting to play, or by resorting to non-verbal communication; • children may severely limit their use of a particular language by resorting to single-word responses, imitations, or using only backchannels; • children may also interrupt conversations in a particular language by initiating a new topic in the other language (another avoidance strategy), or; • children may respond in a particular language, initiate conversations in it, and be quite verbose in it.

Environmental factors in early bilingual development

77

While fully recognizing that there is a wide range of variation, I will speak of early active bilingualism when pre-school-age children regularly respond and initiate conversations in either of two languages (on any specific occasion, even highly active bilingual children may of course exhibit all of the above mentioned discourse strategies for any particular language). Early passive bilingualism I take to be the state in which a child under the age of 6 appears to understand two languages but produces only one. Early monolingualism is the state in which a child under the age of 6 does not appear to understand two languages, and produces only one. In individual cases it might not always be easy to clearly draw the line between either of the three categories (early active bilingualism, early passive bilingualism, and early monolingualism). On a most obvious level, if within the parents-child unit only one language is ever used, children will not have the chance of becoming bilingual within that unit. The moment that within the parents-child unit two languages are used, children do have a chance of becoming bilingual. Whether children will actually use two languages in their interactions or will limit themselves mainly to one can be seen to depend on many different aspects of parents' linguistic behaviour. These different aspects can for the most part be categorized under two main headings: parents' linguistic choices and parents' interaction strategies. There are a number of ways in which parents can make use of two languages in addressing their children. Some of the main possibilities for these linguistic choices are shown in Table 1, which is constructed around four dimensions: • the fact whether parents use one or two languages in speaking to their children, and thus whether they project themselves as a monolingual versus a bilingual speaker when addressing their children; compare, e.g., Situation #1 and Situation #6: in #1 both parents project themselves as monolingual speakers - the famous 'one person, one language' principle, first proposed by the linguist Grammont as quoted in Ronjat (1913); in #6 both parents project themselves as bilingual speakers; • the degree to which parents share a particular language; this dimension is related to the first; compare Situation #1 and #5, where in #1 there is no 'shared language space' between the parents, and where in #5 the 'shared language space' is the same for both parents; • the absolute frequency with which parents use a particular language in addressing their children; see Situation #3, in which the question can be asked whether the amount of input for language Y is sufficient for language learning to take place, and;

78 •

Annick De Houwer the relative frequency with which parents use a particular language in addressing their children rather than another one; compare, e.g., Situation #4 and #7, where, given the - admittedly unrealistic - assumption that both parents spend an equal amount of time with their children, Situation #4 shows an imbalance across the two languages with much more frequent input for language X, but much less input for language Y, and where Situation #7 shows an equal frequency for the input in the two languages.

Table 1: Some ways in which parents can make use of two languages in addressing their children Parent A

Parent Β

Situation #1 uses language X uses language Y

only

-

-

only

Situation #2 uses language X uses language Y

only

Situation #3 uses language X uses language Y

only

-

sometimes mainly

-

mainly sometimes

Situation #4 uses language X uses language Y

mainly sometimes

mainly sometimes

Situation #5 uses language X uses language Y

sometimes mainly

mainly sometimes

Situation #6 uses language X uses language Y

half of the time half of the time

mainly sometimes

Situation #7 uses language X uses language Y

half of the time half of the time

half of the time half of the time

To my knowledge, there have been no studies so far that have specifically investigated the combined role of the four dimensions outlined above. At the same time there are various sources of information, both published and based

Environmental factors in early bilingual development

79

on personal observation, suggesting that each of the four dimensions above may indeed be a contributing factor towards determining whether children will actually speak two languages from an early age or not. There have been claims that the first dimension, viz. the monolingual or bilingual persona of parents, leads to different paths in bilingual acquisition (cf. a review of this issue in De Houwer, 1995), but empirical evidence providing support for this claim is currently lacking. To what extent the monolingual or bilingual persona of parents, and any changes in this persona as Goodz (1994) has shown to occur, make a difference in the development of early active bilingualism remains to be investigated. Nevertheless, it appears to me to be a potentially influential factor. The second dimension I am proposing, i.e., the degree to which parents' language choices overlap, has to my knowledge not received much attention in the literature. Yet, from children's point of view it may well be significant whether their parents project a similar or a widely different linguistic identity. Evidence from second language acquisition has led Wong Fillmore (1989:321) to the finding that "language learning requires frequent and continuing contact between target language speakers and learners", a finding that I believe holds for early bilingual acquisition as well, and that can help explain very many cases in which children who hear just a bit of a second language do not in fact use that language actively. Hence the dimension 'absolute frequency of input'. Although there have been some guesses at what minimum amount of input is required in order for young children to be able to learn to speak a language, the issue is far from settled. Pearson et al. (1997) do furnish evidence showing a strong relation between the quantity of input in each of two languages and the amount of vocabulary learning in either language by bilingual children in the second year of life. The fourth dimension that I have indicated, the relative frequency or the 'balance' in the input of the two languages concerned, is also a factor that needs systematic investigation, given the often quite major differences in the relative frequencies with which children hear two languages from their parents. 'Relative frequency' may turn out to be a factor that is actually not important by and in itself, but that is largely dependent on the third dimension (absolute frequency). Preliminary results of a study by Pearson et al. (1993) do seem to indicate that relative frequency differences between the two languages that children between the ages of 10 and 30 months are exposed to have an effect on children's language use. Although I have presented the four dimensions as steady-state phenomena, it must be taken into account that all four of them are subject to constant flux.

80

Annick De Houwer

As mentioned above, not only parents' linguistic choices but also the way they respond to children's utterances, can be seen to be of major relevance in the development of children's early bilingualism. A first indication of this relevance is shown in the work of Elizabeth Lanza (1990, 1992). Lanza has been able to find differences in parental interaction strategies that according to her are a possible cause for specific linguistic behaviours of children growing up in a bilingual environment. She has proposed that within a bilingual situation adult bilingual speakers may use either bilingual or monolingual interaction strategies. A bilingual interaction strategy allows the use of two languages within a particular discourse situation, whereas a monolingual interaction strategy disallows or discourages it and creates the necessity of adhering to one language only. Bilingual and monolingual interaction strategies can play a role at all levels of discourse, including the utterance level. Thus, use of a bilingual versus a monolingual interaction strategy also determines whether the use of mixed utterances, i.e., utterances containing morphemes from two languages, is condoned or not (Lanza, 1990; Goodz, 1994). As Lanza (1990) points out, there exists a continuum between a strongly bilingual interaction strategy on the one hand, and a strictly monolingual interaction strategy on the other. Stated in extreme terms, the use of a bilingual versus a monolingual interaction strategy determines whether it is acceptable for a child to answer a bilingual speaker (parent) using a different language than the one that that speaker used. In other words, the type of interaction strategy determines whether there is a need for the child to use two languages or not. A similar suggestion has been made by Döpke (1992). Because of the general lack of studies systematically investigating the links between bilingual children's language development and the environments they grow up in (cf. also Schieffelin, 1993 and De Houwer, 1994), it is at present not sufficiently clear just in what ways young children's actual language usage patterns are determined by parental linguistic choices and by parental interaction strategies. It is well established, though, that the range of variation in bilingual parental interaction strategies and linguistic choices is quite large (cf. Arnberg, 1979; Döpke, 1992; Garcia, 1983 and Goodz, 1994). An analysis of the possible causes of this variation can contribute to a greater understanding of the factors leading to the development of early active bilingualism.

Environmental factors in early bilingual development

81

Parents' linguistic behaviour and their beliefs and attitudes The wide range of variation that exists in bilingual parents' linguistic behaviour towards their children can be seen as a rather direct result of parental beliefs and attitudes, i.e., parental beliefs about how children acquire language and parents' own roles in the acquisition process, and parental attitudes to a particular language, to bilingualism in general, and to particular patterns of language choice. I will leave out of consideration attitudes to the particular ways in which children speak, although such attitudes may influence how adults interact with young children as well (see Bloom and Lo, 1990). I will also not attempt to offer a precise definition of 'attitude' and 'belief. It is not clear just how these two notions need to be defined and how they relate to each other (cf. McGuire, 1954 and Sigel, 1985b). Whatever definitional problems there may be with the notions of 'attitude' and 'belief, it is well accepted in social and developmental psychology that attitudes and beliefs may help determine behaviour in crucial ways (see McGuire, 1954; Ninio, 1988; Sigel, 1985b). Although it is not clear just how attitudes and beliefs give rise to particular behaviours (see McGillicuddy-De Lisi, 1985; Sigel, 1985a), it is quite well established that the range of variation in parental beliefs and attitudes concerning children's development is considerable (Goodnow, 1985; Ninio, 1988). I will work from the assumption that (a) parental beliefs and attitudes regarding children's language development are part of parents' more general belief and value systems regarding children's overall development, and that (b) just like parents' belief systems concerning children's overall development, parental beliefs and attitudes regarding children's language development show a lot of variation. In the remainder of this section I will try to outline some possible relationships between language-related parental beliefs and attitudes on the one hand and parental linguistic practices on the other. At a most mundane but quite significant level, negative attitudes to a particular language will keep some bilingual speakers from using that language with their offspring. Burton (1994), for instance, notes that bilingual women may quite deliberately decide not to use one of their languages with their young children. Their reason for this decision often is that they associate the language in question with a way of life that they do not wish to perpetuate. Positive attitudes to a particular language, on the other hand, may determine not only whether a particular language is used at all, but also how it is used. An example of this can be found in a report by Evans (1987) on her own family situation. Evans describes how her Welsh-speaking husband's use of 'accommodating errors' in Welsh can be attributed to his positive attitude

82

Annick De Houwer

towards the fact that his English wife is using Welsh in the house as a L2 learner. It is not only parents' attitudes towards a particular language that help determine their linguistic choices in interactions with their children, but also parents' attitudes towards bilingualism in general and child bilingualism in particular. This is exemplified, for instance, by Zierer (1977), who shows how a negative attitude to early simultaneous bilingualism resulted in a conscious decision on Zierer's and his wife's part to initially limit their son's exposure to only German, although Zierer's wife's first language was Spanish. In contrast, in many of the 45 English and German-speaking families that Behne (1994) interviewed in Berlin, parents held a neutral or positive attitude towards early simultaneous bilingualism. This neutral or positive attitude can be seen to be at least partially responsible for the fact that reportedly parents decided to use two languages in the home from the moment their baby was born. A third possible strand to the language attitude part is that parents can have a positive, neutral or negative attitude to particular types of language choice. They may frown upon the use of mixed utterances, or they may accept them without comment. They may have a tolerance for the alternate use of singlelanguage utterances in a conversation or they may not. Parents' attitudes towards language choice can be seen to play a role in their own use of mixed utterances and in their use of inter-sentential codes witching. Parents' attitudes towards language choice can also be seen as determining their interactional strategies when they are talking to their children. An example of this is given by Goodz (1994): parents in her study who used only or mainly language Y with their child were willing to repeat a mixed child utterance or a child utterance in language X depending on whether the parent in question had a more flexible versus a less tolerant attitude towards the use of language X or mixed utterances. Behne's (1994) interviews show that parents can hold very wide-ranging overt views on the use of mixed utterances and inter-sentential codeswitching. These interviews also suggest that as a consequence of these wide-ranging views, parents adopted quite diverse strategies both in their own linguistic practices and in response to their children's use of mixed utterances and inter-sentential codeswitching. The exact relationships are unclear, however, since Behne's findings are based only on self- and other-report, which, as for instance Schieffelin, 1993 and Hakuta and Pease-Alvarez, 1994 have pointed out, cannot necessarily be equated with actual speech practices. Parents' attitudes to particular languages, (early child) bilingualism, and aspects of language choice will of course interact in complex ways that are not always transparent, and that may in fact result in ambivalent situations

Environmental factors in early bilingual development

83

which do not necessarily correspond to parents' educational goals. Parents may hold a positive attitude to each of two languages and to early child bilingualism. Thus they would want their child to grow up speaking two languages. If in this situation parents have a tolerant attitude towards mixed utterances and inter-sentential codeswitching, parents will tend to allow the use of mixed utterances and inter-sentential codeswitching by their children, thereby relaxing the need for their children to use a particular language and possibly contributing to their children's gradually declining use of one of the languages involved. Clearly, attitudes towards particular linguistic practices here are closely related to beliefs about the possible relationship between such linguistic practices and children's language learning. Parental beliefs about how children acquire language and about parents' own roles in this process can be seen to have a substantial effect on parents' linguistic behaviour towards their children. Mertz (1989), for instance, found that bilingual parents on Cape Breton deliberately stopped speaking Gaelic to their children because they thought that knowledge of Gaelic would hinder children's acquisition of English, which was deemed necessary for their upward mobility. Zierer (1977) reports how he and his wife forbade their son's monolingual Spanish-speaking grandmother to speak to him in Spanish until around the child's third birthday, and how they deliberately kept their son from having playmates in the Spanish-speaking environment of Peru that the family lived in, since Zierer believed that direct exposure to Spanish in the first three years of life would greatly hinder the child's acquisition and maintenance of German, the language used by the parents in talking to the child. On a more subtle level, the 'accommodating errors' in Welsh that Evans' Welsh-speaking husband (Evans, 1987) made cannot only be related to his positive attitude towards Welsh but can also be linked to his apparent belief that by incorporating some of these errors in his own speech he was providing support for Welsh in the family. Evans' husband also never corrected his son's errors in Welsh due to his belief that any correction of these errors would threaten his son's developing bilingualism. Deliberate decisions not to use a particular language in speaking to children in order to reach a particular educational goal fit in with what I call an 'impact belief. By using the term 'impact belief I am referring to the parental belief that parents can exercise some sort of control over their children's linguistic functioning. An impact belief can be very strong, and may include the notion that the parent has an important exemplary function to fulfill, and that thus the parent's own language use has a direct influence on what the child will learn to say. A strong impact belief may in addition include the conviction that children's language use can be influenced by

84

Annick De Houwer

telling the child to use a particular linguistic form, or by punishing or praising the child for using a particular linguistic form. An impact belief can be fairly weak as well, and may consist of just the loosely held conviction that in general children will pick up language from the environment. Parents may also believe that whatever they say makes no difference to the child's learning of language, and that the child will gradually learn to talk through some mechanism or other that has nothing to do with the parents. As Heath (1989:337) notes: "In not every society do adults see infants of their community as conversational partners. In not every society do adults believe they have to teach directly or model speech for their children" (in a similar vein, see Goodnow, 1985 and Ninio, 1988 for overviews of studies showing large cultural differences between parents in their beliefs about children's overall development). A 'no impact' belief appears to be common in quite a few non-Western societies (see Lieven, 1994; Ochs and Schieffelin, 1995), but can be seen to be present in parts of Western industrial society as well (cf. Behne, 1994). Finally, it should be pointed out that attitudes and beliefs as expressed in interviews are not necessarily the attitudes and beliefs that people really hold. Nor do these overtly expressed attitudes and beliefs necessarily coincide with covert or less consciously held attitudes and beliefs. As McGuire (1954) already noted, people's verbal reports of their attitudes generally have a low correlation with their actual behaviour towards the object of the attitude. Overtly expressed beliefs may show just as little correlation with actual behaviour. As Goodz (1994) has found, despite the prevalent and confident overtly expressed belief that in a bilingual family it is best that each parent sticks to one language only, parents do occasionally use more than one language in addressing their children. The mismatch between overtly expressed beliefs (and attitudes) and actual behaviour can be used to call into question the assumed link between beliefs and attitudes on the one hand and behaviour on the other. Alternatively, the mismatch can be explained as showing that it is not so much overtly expressed beliefs and attitudes as primarily covert or less consciously held beliefs and attitudes that lie at the basis of behaviour. Such covert or less consciously held beliefs and attitudes can be investigated using various indirect techniques (see De Houwer and Wölck, 1997), but to my knowledge such techniques have not so far been used to explore parental beliefs and attitudes in bilingual family situations. Figure 1 summarizes the main points made in this section.

Environmental factors in early bilingual development

85

ATTITUDES (continuum: negative/neutral/positive) • towards a particular language • towards (early child) bilingualism • towards (aspects of) language choice BELIEFS • impact belief (continuum: strong/weak) belief that parents can exercise some control over their children's linguistic functioning • no impact belief (continuum: strong/weak) belief that parents cannot exercise any control over their children's linguistic functioning Figure 1: Parental beliefs and attitudes that are relevant to parental choices and interaction strategies

linguistic

Parental beliefs and attitudes, parents' linguistic behaviour and children's language development In developmental psychology it is suggested that parents' ideas underlie parents' practices, which in turn are seen to be strong determinants of children's development (see Johnson and Martin, 1985). Similarly, I propose a three-tiered framework to help explain the development of early active versus early passive bilingualism, and indeed early monolingualism in what is potentially a bilingual input condition (see Figure 2): parents' beliefs about how children learn language and their attitudes to particular styles of language use, particular linguistic varieties and particular languages affect whether parents will use a particular style, variety or language in addressing their children and how they will speak to their children. Parents' choices in this regard ultimately affect what styles, varieties and languages children will initially acquire, and, even more dramatically, whether they will acquire any language at all.

86

Annick De Houwer

PARENTAL BELIEFS AND ATTITUDES

PARENTAL LINGUISTIC CHOICES AND INTERACTION STRATEGIES

CHILDREN'S LANGUAGE DEVELOPMENT

Figure 2: The relationship

between parental language development in a potentially

beliefs/attitudes bilingual input

and children's condition

I will not attempt to explain here how the arrows in Figure 2 need to be elaborated. Suffice it to say that the relationships between the three levels are highly complex, and that the arrows should probably be bidirectional (cf. the bidirectional feedback model proposed by Sigel, 1985a, to account for the nature of parental beliefs regarding children's overall development; see also Kulick, 1992:201-209, who shows how parents' interpretations of very young children's vocalizations influence parents' subsequent linguistic choices in interactions with children). A very dramatic and sad example of how important parental beliefs and attitudes can be in determining parents' own linguistic behaviour, and, as a result, their children's language learning, is reported by Clerebaut (1993), who gives an account of 12 families in Brussels in which the parents are entirely deaf. In these families, there were hearing pre-school children, usually the second or third child in the family. These hearing pre-school children had by age three acquired no language at all: their signing parents had not signed to them, and their parents had not undertaken any specific steps to ensure that from infancy onwards these children would have the chance to interact with hearing individuals. The deliberate decision not to use signing with their hearing children was based on the parents' firm belief that their hearing children had no need for signing. According to the parents, their hearing children would automatically start to speak at around age two and become the deaf parents' links to the hearing world. The parents saw signing very much as a potentially negative thing that would stand in the way of their children's expected use of spoken language. Here, in this potentially bilingual situation, not even monolingualism was attained as a result of parental attitudes and beliefs. 2

Environmental factors in early bilingual development

87

Unfortunately, such reports of negative effects of parental beliefs on children's linguistic development are confirmed in a variety of other publications. Schiff and Ventry (1976), for instance, discuss the case of several hearing children of deaf mothers who would not take their hearing and speaking children to speech therapists although such had been urgently recommended by the investigators on the basis of their assessment of the children's spoken language skills. It appears that the children's mothers, who themselves had a total lack of knowledge about spoken language, considered the fact that their children spoke at all such an achievement that they could not believe that there might in fact be a problem. As a result, the children's spoken language skills remained very poor. Evidence for the three-tiered framework as proposed in Figure 2 from situations involving two or more spoken languages is hard to come by. A major reason is that very few studies offer information about all three tiers for one particular family. At best, there is explicit information available about only two of the three layers involved. In cases where there is information about parental linguistic choices and interaction strategies on the one hand, and developmental patterns in children on the other, though, it may be possible to infer the nature of particular beliefs and attitudes (one must take care not to get into a circular argument, however): for instance, the fact that the bilingual mother of a Norwegian/English-speaking child "frequently feigned the role of a monolingual [English speaker] by questioning her child's [Norwegian] utterance said in what she saw as the inappropriate language" (Lanza, 1992:647) probably relates to the mother's belief that her linguistic actions can have an effect on her daughter's linguistic behaviour, and also to a positive attitude towards English. Given the paucity of empirical studies that have systematically and centrally focused on the possible role of the environment's beliefs and attitudes in early bilingual development, any hypotheses about possible relationships are in danger of having a short life. Yet I think there is some evidence that makes it possible to formulate a hypothesis concerning the relationship between parental attitudes and beliefs on the one hand, and the development of early active bilingualism on the other. In order for early active bilingualism to develop, a minimal condition is that parents have a positive attitude to both languages involved and to early child bilingualism. This is not enough, however: parents must also hold an impact belief concerning their own roles in the language acquisition process. In all other situations, I believe, there is insufficient support for the development of early active bilingualism (see Figure 3, which makes abstraction from the second tier in Figure 2, namely parental linguistic choices and interaction strategies).

88

Annick De Houwer Parental attitudes to the languages involved and to early child bilingualism

Impact belief No impact belief

positive

negative

yes no

no no

Figure 3: +/- Supportive environment for the development of early active bilingualism Before I present some empirical evidence from the literature to support my proposal it needs to be explained why attitudes to aspects of language choice (cf. Figure 1 previously) do not feature in my proposal here. The reason is not that these attitudes are not important, but rather that I have found no relevant evidence in the literature that was sufficiently compelling as a basis for any hypothesis-formation. It may well be the case that, more so than with attitudes to the languages involved and to bilingualism, the significance of parental attitudes towards aspects of language choice is related to language use outside the parents-child unit, and dependent on the degree to which the family is an 'isolated' bilingual family or, alternatively, can be seen to be part of a bilingual community. These issues need to be further explored. As I mentioned before, there is hardly any information on the three tiers in Figure 2 for any one study, that is, (1), parental beliefs and attitudes in a bilingual situation, (2), parents' linguistic choices and interaction strategies, and (3), children's language development. Two major exceptions are the 1988 book by George Saunders, and the work of Don Kulick (1992). These studies lend support for two of the four belief/attitude constellations in Figure 3: case 1, where positive attitudes combine with an impact belief (Saunders, 1988), and case 2, where positive attitudes combine with a no impact belief (Kulick, 1992). George Saunders (1988) clearly makes explicit his and his wife's reasons for wishing to raise bilingual children in a monolingual environment. He also explains some of his reasons for the particular linguistic choices that were made by him and his wife. Finally, there are abundant data available about the developing active bilingualism of the Saunders' three children. It is very clear from all the explicit information that Saunders gives that he and his wife have a very positive attitude to bilingualism, as well as to the two particular languages involved, English and German. From the many examples of his own and his wife's reactions to the children's language use it is also very clear that the Saunders firmly believe that they can influence their children's linguistic functioning. Without a doubt, the Saunders case

Environmental factors in early bilingual development

89

represents the one case in which active bilingualism is sure to develop, namely when there are positive attitudes present as well as an impact belief. A caveat is necessary here, though: as I suggested in Figure 1, an impact belief can be very strong. It can be so strong that parents use non-linguistic means to control their children's linguistic behaviour, and even punish their children for using the 'wrong' language. As Fredman (1995) has reported, for instance, one mother in her English/Hebrew-speaking sample said she had actually withheld food from her child as punishment for the child's use of Hebrew when the mother wanted the child to use English instead. I would venture to say that such extreme measures do not in fact create a supportive environment for the development of early active bilingualism, and are actually quite counterproductive in the long run. It seems, then, that the single occurrence of 'yes' in Figure 3 should really be read as a 'yes, but...'. In Saunders' case, the attitudinal and the belief component have together resulted in three German-English-speaking children who spoke two languages from the beginning of recognizable speech. The attitudinal and the belief component, however, can be at odds with each other, in that the belief component may get in the way of the attitudinal component. This appears to be the case in the Papua New Guinean village of Gapun as studied by Don Kulick (1992). Parents in this village have overt positive attitudes to both Tok Pisin and Taiap and would like their children to speak both languages. Parents speak both languages to their children, but use Tok Pisin more often than Taiap. At the same time, however, Gapun villagers strongly believe that they can have no impact whatsoever on their children's language development (see also Stroud, 1990). As a result, they hardly ever talk to very young children, leaving that task to older children who speak Tok Pisin to the young ones. Also, if parents do talk to their children and their children speak Tok Pisin to them, they will not make the slightest effort to change children's language choice. They thus use bilingual interaction strategies (cf. Lanza, 1990). The result of all this is that no child under the age of 10 speaks Taiap in the village, although children may in fact understand the language. Parents are unhappy about this, but feel they cannot change the situation. Very clearly, Kulick's (1992) study shows how overtly positive attitudes can clash with a no impact belief, resulting in young children's passive rather than active bilingualism.3 According to the hypothesis as outlined in Figure 3, if there are negative parental attitudes towards one of the two languages in a bilingual situation and towards early child bilingualism, there will be no basis for the development of active bilingualism, regardless of parents' beliefs about their own role in the language learning process. Negative parental attitudes will in

90

Annick De Houwer

many cases have the obvious effect that the child is exposed to one language within the parents-child unit, and as such will not become an early active bilingual within that unit. A case in point is Zierer's son (Zierer, 1977), who was initially a German-speaking monolingual despite a potentially bilingual input condition within the family (cf. earlier discussions) and who only started to learn Spanish once his parents had decided that exposure to Spanish outside the parents-child unit would not hinder his continued acquisition of German. Matters are less obvious when attitudes are quite negative, but not negative enough to keep a parent from using a particular language at least some of the time. In this case either one or both of the two no's in the right column of Figure 3 may need to be changed into hesitant yeah's but so far I have not come across any material to warrant such a change. As mentioned before, relevant attitudes concern attitudes to at least the two languages involved and to early child bilingualism (and bilingualism in general). I would venture to say that parental attitudes towards both of these objects need to be positive for parents to create a supportive environment for the early active bilingual development. Take the case of a French/Englishspeaking Ontarian woman, Annette, who was married to an English speaker (Heller and Levy, 1992). As became apparent from interviews with Annette, she had a fairly positive attitude towards French, and started out using French with her third son. However, Annette reports that she switched entirely to English on the advice of a medical doctor who claimed that continuing to speak French to the child would hinder his development. Annette readily accepted this advice, thus changing to or adding a negative attitude towards early child bilingualism. This negative attitude clearly superseded any positive attitude towards French that she had before. Her switch to another language with a particular result in mind does indicate, however, that Annette has an impact belief as far as the relationship between her own linguistic behaviour and that of her children is concerned (as such, just like Zierer's (1977), Annette's may be seen as an example of case 3 in Figure 3, which combines a negative attitude with an impact belief). We find further evidence for the possible links between parental attitudes to a particular language and bilingualism, parental linguistic choices and children's language use in Heller and Levy (1994). In this study, a group of 11 French/English bilingual mothers in Ontario with highly positive attitudes towards French claimed to have always spoken French to their children, and in fact had children who spoke French when they were very young. The children of a group of 7 French/English bilingual mothers in Ontario with less positive attitudes towards French, on the other hand, tended to use English at

Environmental factors in early bilingual development

91

home at a young age, even though their mothers might have used mainly French with them (some mothers in this group, contrary to the 'positive' group, used both English and French in speaking to their young children). The mothers' attitudes towards French were determined on the basis of in-depth interviews regarding the interviewees' language use patterns in various situations, their involvement with francophone organizations, their education and employment, and so on. In other words, the mothers' attitudes were determined independently from the linguistic choices they made with regard to their children. Although Heller and Levy's study does not offer information about actual linguistic practices in the home and the children's linguistic development obtained through direct observation, their study does suggest that the investigation of the potential links between bilingual parents' attitudes and beliefs, parental practices and children's language development is a promising avenue for research.

Conclusions and discussion This chapter is rooted in the underlying assumption that the development of language in small children crucially depends on the language learning environment, regardless of how many languages or what kinds of languages are involved. In a monolingual environment, however, the central role of environmental factors in language development is less easy to distinguish. Within a monolingual context, it is only when there is hardly any language environment available to the young child that the prominent role of the environment is made obvious. In a bilingual context however, environmental factors become much more immediately noticeable. Although the details still need to be filled in, it is obvious that in order for active bilingualism to develop from an early age and to be sustained as the child matures, children must be regularly and frequently exposed to two or more languages, and must be growing up in a situation in which active use of these languages by the child is necessary for socio-communicative purposes. In both monolingual and bilingual contexts children's linguistic environments are to a large degree shaped by the beliefs and attitudes of the people who constitute these environments. These beliefs and attitudes partly determine how adults will interact with children and what language(s) they will be addressing children in. These interaction patterns in turn affect children's language use, thus ultimately contributing to more macrosociological processes of language shift.

92

Annick De Houwer

One of the hypotheses put forward is that the only family situation in which there is sufficient support for early active bilingualism to come about is the situation where a parent or a pair of parents have what I have called an impact belief concerning their own possible role in the language acquisition process, and where there is a general positive attitude towards the languages involved and to being bilingual. In other situations, the prospects for early active bilingualism are not so good. Due to the general lack of empirical studies investigating the contributory role of bilingual parents' attitudes and beliefs in children's bilingual development, however, many of the ideas put forward here await investigation. At present, most relevant information is available by chance rather than by design. Some central questions to be answered are: to what extent do covert attitudes and beliefs contribute to linguistic practices?; do continuums in the ranges of attitudes and beliefs correspond to continuums in types of supportive behaviour?; do continuums in types of supportive behaviour relate to degrees of active bilingualism in young children? Questions such as these need to be answered on the basis of well-designed, multi-level, and microinteractional empirical studies of bilingual parents and their children. Once we have some answers, we shall be closer to a fuller understanding of the dynamics underlying processes of language shift and language change.

Notes 1. The work presented here was carried out while the author was a researcher with the Belgian National Science Foundation (N.F.W.O.) and while she was a V.N.C.fellow at NIAS (Netherlands Institute for Advanced Study), Wassenaar, the Netherlands, in 1996.1 would like to thank the participants of the International Symposium on Processes of Language Change in an Immigration Context (Tilburg University, 1996) where the main ideas for this chapter first were presented for their valuable comments and suggestions. I am also grateful to Pieter Muysken, Ted Wachs and Wolfgang Wölck for helping me find useful literature sources and to Marion Fredman for insightful comments. 2. Yet there is evidence that hearing children of deaf parents can and do develop 'normally' (see Bonvillian et al., 1990), and that if deaf parents offer their hearing children the possibility of learning to sign from early on, their children will learn to interact using a symbolic communication system (in this case, sign language), and will at least have the opportunity of becoming monolingual, if not bilingual.

Environmental factors in early bilingual development

93

3. Unfortunately, lack of space does not allow a more detailed discussion of the Gapun case, and my discussion here does not do full justice to the rich and insightful analyses of Gapun's linguistic situation that Kulick presents in his outstanding book.

References Arnberg, L. (1979). Language strategies in mixed nationality families. Scandinavian journal of psychology, 20, 105-112. Behne, P. (1994). Simultaneous acquisition of German and English: experiences and restrospective views of bilingualism by parents of bilinguals in Berlin. Unpublished MA thesis. Freie Universität Berlin, Berlin. Bloom, K. and E. Lo (1990). Adult perceptions of vocalizing infants. Infant behavior and development, 13, 209-219. Bonvillian, J., M. Orlansky and R. Folven (1990). Early sign language acquisition: implications for theories of language acquisition. In: V. Volterra and C. Erting (Eds.), From gesture to language in hearing and deaf children. Berlin: Springer Verlag. Burton, P. (1994). Women and second-language use: an introduction. In: P. Burton, K. Kushari Dyson and S. Ardener (Eds.), Bilingual women: anthropological approaches to second language use. Exeter: Short Run Press. Clerebaut, N. (1993). Enfants entendants de parents sourds. Nineteenth study-day, Belgian association for audiophonology, Brussels, March 1993. De Houwer, A. (1994). Review of One parent one language: an interactional approach by S. Döpke. Journal of Child Language, 21, 745-752. De Houwer, A. (1995). Bilingual language acquisition. In: P. Fletcher and B. MacWhinney (Eds.), Handbook of Child Language. Oxford: Blackwell. De Houwer, A. and W. Wölck (1997). An ethnographic method for studying attitudes towards child language. In: Pütz (Ed.), Language choices? Conditions, constraints, and consequences. Amsterdam: John Benjamins. Döpke, S. (1992). One parent one language: an interactional approach. Amsterdam: John Benjamins. Evans, M. (1987). Linguistic accommodation in a bilingual family: one perspective on the language acquisition of a bilingual child being raised in a monolingual community. Journal of multilingual and multicultural development, 8, 231 -235. Fredman, M. (1995). Maintaining English as the home language in pre-school bilingual children in Israel. Paper presented at the International Symposium on Communication Disorders in Bilingual Populations, Haifa, Israel, August 1-4, 1995. Garcia, E. (1983). Early childhood bilingualism. Albuquerque: University of New Mexico Press.

94

Annick De Houwer

Goodnow, J. (1985). Change and variation in ideas about childhood and parenting. In: I. Sigel (Ed.), Parental belief systems. The psychological consequences for children. Hillsdale, NJ: Lawrence Erlbaum Associates. Goodz, N. (1994). Interactions between parents and children in bilingual families. In: F. Genesee (Ed.), Educating second language children. Cambridge: Cambridge University Press. Hakuta, K. and L. Pease-Alvarez (1994). Proficiency, choice and attitudes in bilingual Mexican-American children. In: G. Extra and L. Verhoeven (Eds.), The crosslinguistic study of bilingual development. Amsterdam: North-Holland. Heath, S.B. (1989). The learner as cultural member. In: M. Rice and R. Schiefelbusch (Eds.), The teachability of language. Baltimore: Paul H. Brookes. Heller, M. and L. Levy (1992). La femme franco-ontarienne en situation de mariage mixte: feminite et ethnicite. Recherches feministes, 5, 59-82. Heller, M. and L. Levy (1994). Les contradictions des mariages linguistiquement mixtes: strategies des femmes franco-ontariennes. Langage et societe, 67, 53-88. Johnson, J. and C. Martin (1985). 'Parents' beliefs and home learning environments: effects on cognitive development. In: I. Sigel (Ed.), Parental belief systems. The psychological consequences for children. Hillsdale, NJ: Lawrence Erlbaum Associates. Kulick, D. (1992). Language shift and cultural reproduction. Socialization, self, and syncretism in a Papua New Guinean village. Cambridge: Cambridge University Press. Lanza, E. (1990). Language mixing in infant bilingualism: a sociolinguistic perspective. Unpublished PhD thesis. Georgetown University, Washington. Lanza, E. (1992). Can bilingual two-year-olds code-switch? Journal of child language, 19, 633-658. Lieven, E. (1994). Crosslinguistic and crosscultural aspects of language addressed to children. In: C. Gallaway and B. Richards (Eds.), Input and interaction in language acquisition. Cambridge: Cambridge University Press. McGillicuddy-De Lisi, A. (1985). The relationship between parental beliefs and children's cognitive level. In: I. Sigel (Ed.), Parental belief systems. The psychological consequences for children. Hillsdale, NJ: Lawrence Erlbaum Associates. McGuire, W. (1954). The nature of attitudes and attitude change. In: L. Gardner and E. Aronson (Eds.), The handbook of social psychology III. Rowley, MA: Addison Wesley. Mertz, E. (1989). Sociolinguistic creativity: Cape Breton Gaelic's linguistic 'tip'. In: N. Dorian (Ed.), Language obsolescence. New York. Ninio, A. (1988). The effects of cultural background, sex, and parenthood on beliefs about the timetable of cognitive development in infancy. Merrill-Palmer Quarterly, 34, 369-388.

Environmental factors in early bilingual development

95

Ochs, E. and B. Schieffelin (1995). The impact of language socialization on grammatical development. In: P. Fletcher and B. MacWhinney (Eds.), Handbook of Child Language. Oxford: Blackwell. Pearson, B., S. Fernandez, V. Lewedag and D. Kimbrough Oiler (1993). Bilingual environment and lexical learning in bilingual infants (10-30 months). Paper presented at the 14th Symposium on Spanish and Portuguese Bilingualism, Fairfield, Connecticut, November 12-13, 1993. Pearson, B., S. Fernandez, V. Lewedag and D. Kimbrough Oiler (1997). The relation of input factors to lexical learning by bilingual infants (ages 10 to 30 months). Applied psycholinguistics, 18, 41-58. Pupier, P. (1982). L'usage du fran?ais et de l'anglais et le probleme de la dominance linguistique chez des petits enfants bilingues. In: P. Pupier, K. Connors, K. Lappin, W. Greene and L. Nuckle (Eds.), L'acquisition dufrangais et de l'anglais chez des petits enfants de Montreal. Quebec: Gouvernement du Quebec. Ronjat, J. (1913). Le developpement du langage observe chez un enfant bilingue. Paris: Champion. Saunders, G. (1988). Bilingual children: from birth to teens. Clevedon: Multilingual Matters. Schieffelin, B. (1993). Code-switching and language socialization. Some probable relationships. In: J. Deuchan, L. Hewitt and R. Sonnemeier (Eds.), Pragmatics: from theory to practice. Englewood Cliffs, NJ: Prentice Hall. Schiff, Ν. and I. Ventry (1976). Communication problems in hearing children of deaf parents. Journal of speech and hearing disorders, 41, 348-358. Sigel, I. (1985a). A conceptual analysis of beliefs. In: I. Sigel (Ed.), Parental belief systems. The psychological consequences for children. Hillsdale, NJ: Lawrence Erlbaum Associates. Sigel, I. (Ed.) (1985b). Parental belief systems. The psychological consequences for children. Hillsdale, NJ: Lawrence Erlbaum Associates. Stroud, C. (1990). A cultural organization of codeswitching. In: Papers for the workshop on constraints, conditions and models. London: ESF Network on CodeSwitching and Language Contact. Wong Fillmore, L. (1989). Teachability and second language acquisition. In: M. Rice and R. Schiefelbusch (Eds.), The teachability of language. Baltimore: Paul H. Brookes. Zierer, E. (1977). Experiences in the bilingual education of a child of pre-school age. IRAL, 15, 143-148.

Changing patterns of language mixing in a bilingual child1 Carol W. Pfaff

There is a steadily growing body of research on the language of bilinguals in countries with large immigrant or indigenous minorities. In the USA, Hispanic groups have received much attention in recent years and in Europe, migrants from Mediterranean countries or from former colonies and their children have been the focus of many projects. At the center of much of this bilingual research are language-contact phenomena: language mixing and codeswitching (CS), fusion and convergence. These phenomena are the focus of this chapter in which I explore the development of patterns of explicit language mixing and CS in the developing competence of one bilingual boy, a Turkish/German bilingual in Berlin who attended a bilingual day care center (Kita) in which we conducted a longitudinal study2. This child, like most of his age cohort in the second generation, is a successive bilingual, with Turkish the language learned first and German acquired later. We followed his development in both languages from age 1;06 to 8;00. Specifically, this chapter examines instances of overt language mixing and codeswitching, Johanson's global code copying (1993, this volume), to address several psycholinguistic and socio-historical questions concerning the development of two languages in the individual and in migrant communities: • How does mixing type relate to proficiency? • Is there selective code-copying at mixing sites suggestive of grammatical fusion or linguistic convergence? • How do interlocutors respond to mixing and what effect does this have on the child's speech? • Does the progression of mixing types in the child's speech parallel that found in successive generations of adult immigrants? The findings discussed here are based on analysis of selected transcripts from a corpus of 59 recordings with the child, which were made separately in each language by the adult researchers. The earlier recordings were made at the Kita, where other children and caretakers were sometimes also present. Later recordings were made at the home of the child where sometimes parents or siblings were also present. A few recordings, part of a

98

Carol W. Pfaff

follow-up study by Sava§ (1994), were made at her home, where she was sometimes present during the German language interviews as well as at the Turkish interviews she conducted. The presence of several interlocutors at some of the interviews allowed us to record the target child in both what Grosjean (1985) terms the 'bilingual mode' and the more 'monolingual mode'. The latter was used especially in the more directed conversation with the researchers, who, it should be noted, had some knowledge of the other language. This is true not only of the Turkish interviewers, first generation immigrants from Turkey who had lived in Berlin for several years and spoke German, but also for the German interviewers, who acquired some knowledge of Turkish, particularly of lexical items frequently used by the children. Because language mixing is the subject of the present chapter, it is important to comment on the policy followed by the project staff as they functioned as interviewers. At the outset of the project, we decided neither to 'correct' the children if they used a word from the other language nor to pretend not to understand. Instead, our decision was to accept the utterance and to offer expansions or further utterances with the translation equivalent in subsequent turns. How this worked out in practice will be seen below. Background of the target child The following information on the background and linguistic development of our target child, the boy we designate here as Orhan ( ' Τ 2 Γ in earlier publications), is based on our observations, conversations with his mother and his own reports (see Sava§, 1994, for further details). Turkish is the primary language of communication between Orhan's parents. Orhan's mother was born in Gaziantep, where she attended primary school until the second grade when her family migrated to Germany in 1975. She attended a 'foreigner class' with other non-German children, mostly also from Turkey, until third grade, after which she attended classes with German peers. Since completing the ninth grade, she has worked in a factory where she speaks both Turkish and German. She understands but does not use Kurdish. Orhan's father was born and raised in Gaziantep and came to Berlin after marriage to Orhan's mother in 1985. He attended a German course for a few months only, and currently works in a factory. He sometimes speaks Kurdish with his family and friends but not with his wife or children, with whom he uses Turkish. Turkish in the family is reinforced by summer vacations in Turkey each year.

Changing patterns of language mixing in a bilingual child

99

Orhan, the first child, was born in Berlin in 1986. He entered the Kita at age 1 ;06 and continued there throughout his preschool years. There he was exposed to adult native varieties of Turkish and German but also sometimes to the interlanguage varieties of Turkish caretakers when they occasionally addressed German or German-dominant Turkish children in German or spoke to German colleagues. Orhan started primary school at one of the schools in Berlin which has a bilingual literacy program in Turkish and German but left at the end of the first grade because the family moved to another part of the city. Since then he has attended 'regular' classes together with both migrant and German peers. At home, Orhan usually speaks Turkish with his father, but with his mother he prefers to speak German, though both often switch languages during a conversation. Orhan uses both languages when he plays with his younger sister, who now attends the same bilingual Kita that he did. During the period of our study, Orhan develops from strongly Turkishdominant to Turkish/German bilingual with increasing preference for German. In the following sections, I show how this is reflected by changes in his patterns of language mixing and codeswitching. I also examine how the reactions of adult interlocutors to his language mixing contribute to the development of the monolingual mode in German but the bilingual mode in Turkish. Frequency of language

mixing

Recordings with Orhan began at age 1;06. His early insertions and alternations of the two languages starting with the earliest instances at 1 ;07 are discussed below in qualitative terms. However, the quantitative analysis presented in the following tables starts with age 2;06 because prior to that time there is no mixing of German in Turkish recordings at all and very little intelligible German in the German recordings either. Tables 1 and 2, which present the type/token frequencies of Turkish words in the German and Turkish recordings, give a first quantitative overview of the asymmetrical development of mixing in the two language contexts.

100

Carol W. Pfaff

Table 1:

Word types and tokens in German recordings Types Turkish/German + Turkish

Tokens Turkish/German + Turkish

Age

η

% Turkish

η

% Turkish

2;06 2;08 3;00 3;03 3;06 3;07 3;08 4;02 5;08 7;02 8;00

11/58 7/68 11/53 17/95 4/88 30/109 2/113 6/177 0/232 8/276 6/305

19.0 10.3 20.8 17.9 4.5 27.5 1.8 3.4 0.0 2.9 2.0

14/160 8/144 18/103 44/300 5/287 63/234 2/420 7/633 0/857 8/922 7/761

8.8 5.6 17.5 14.7 1.7 26.9 0.5 1.1 0.0 0.9 0.9

Table 2:

Word types and tokens in Turkish recordings Types German/Turkish + German

Age

η

% German

2;06 2;07 2; 10 2; 11 3;06 4;03 7;02 8;00

0/91 0/114 2/343 0/223 6/350 17/560 12/440 103/306

0.0 0.0 0.6 0.0 1.7 3.0 2.7 33.7

Tokens German/Turkish + German η 0/154 0/194 4/648 0/421 8/584 25/1353 19/932 149/495

% German 0.0 0.0 0.6 0.0 1.4 1.8 2.0 30.1

As shown in Tables 1 and 2, the amount of Turkish in German declines over time while the amount of German in Turkish increases. These results reflect two aspects of language development in the migrant community. First, they reflect the individual child's changing proficiency. Orhan begins without any competence in German and gradually acquires the language, albeit a non-standard variety. Secondly, they reflect that this child, like adults and other children in immigrant settings, is developing toward a 'monolingual

Changing patterns of language mixing in a bilingual child

101

mode' in German and a 'bilingual mode' in Turkish (see Boeschoten and Verhoeven, 1985; Backus, 1996 and Pfaff, 1991, 1994). Patterns

of language

mixing: alternation,

insertion,

naming

While Tables 1 and 2 show the extent of mixing, they reveal nothing of its nature or structure. In Tables 3 and 4, the instances of mixing are classified into three categories: alternation, insertion, and naming. The tables indicate the frequency of each type at each point of investigation as well as the integration, if any, of the mixed elements into the matrix languages. For purposes of the present analysis, alternation is a complete switch to language Β in a language A context. Insertion includes single content words, formulaic expressions, tags, and discourse markers in language Β within a given utterance, basically in language A. The third category, 'naming', is a functional rather than structural category typical of the early child-adult interactions, designates single lexical items, usually language Β nouns in a language A context. Table 3:

Age 2;06 2;08 3;00 3;03 3;06 3;07 3;08 4;02 5;08 7;02 8;00

Alternations, insertions, naming in German recordings (" = to German speaker,b = to Turkish speaker)

Alternations to Turkish Γ 0 0 0 0 a l +4b 0 1 0 2b 2b

(instances)

Insertions of Turkish

Naming in Turkish

12 0 0 0 0 0 0 2 0 0 4

0 7 16 43 5 12 2 0 0 0 0

102

Carol W. Pfaff

Table 4: Alternations, insertions, naming in Turkish recordings (instances) (" = to German speaker,b = to Turkish speaker)

Age

Alternations to German

Insertions of German

Naming in German

2;06 2;07 2;10 2;11 3;06 4;03 7;02 8;00

0 0 0 0 2a 0 0 15b

1 0 0 0 4 25 19 13

0 0 2 0 0 0 0 0

Tables 3 and 4 show that the patterns of mixing of Turkish in German recordings differ qualitatively as well as quantitatively from the patterns of German in Turkish recordings. These tables indicate further that the patterns in both contexts change markedly over time, as discussed and exemplified below.

Turkish in German: decreasing use of Turkish with increasing proficiency in German The earliest transcripts of Orhan's interactions with the German interviewers show that due to his limited competence in German, he relies extensively on Turkish. In fact, Orhan's speech consists essentially of Turkish with a few alternations to German. These alternations consist primarily of formulaic expressions, deictics and repetitions of the interviewers' utterances, naming objects or animals. This limited lexical repertoire, however, allows Orhan to participate in the interactions, if minimally. Formulaic expressions The first use of German in the context of 'German' recording sessions occurs at 1 ;07, when Orhan answers questions with ja but otherwise says little. At 1;08 he uses several German words and phrases, including: ja 'yes' and nein 'no', Kind 'child', the deictic expressions, da 'there' and hier 'here' and the formulaic expression geh weg 'go away'. He uses the Turkish imperative bak\ 'look' to attract the interviewers' attention to the toys and

Changing patterns of language mixing in a bilingual child

103

pictures. At 1; 10, he uses ich Τ and guck 'look' to draw the interviewers attention. At 1; 11, he repeats the interviewers' use of so and ο lala and spontaneously uses the formulaic expressions: hallo 'hello', ich auch 'me too', [gwga/]=?guck mal 'look' to get the interviewers' attention. Early predications Within a few sessions, however, he develops a wider repertoire of formulaic expressions and elementary syntactic structures for naming and predication which allow him to participate more actively in interactions in German. At 2;01, we note what seem like the first predications in German, albeit with missing and/or unintelligible elements, as in [1] - [3] and the attentiongetting imperative in [4]: [1]

2;01

xx das

'xxthat' 3

[2]

2;01

Auto da

'car there'

[3]

2;01

nein xxx

'no xxx'

[4]

2;01

guck xx

'lookxx'

At 2;04, Orhan alternates languages appropriately for addressees, speaking to a Turkish caretaker in Turkish and responding in German to the German interviewer, using predicates such as das meine\ 'that's mine' and weg 'away'. At 2;05, he attends and nods when interviewers talk to him, repeats some names of animals in German, Elefant 'elephant', Katze 'cat' and names some items in Turkish ekmek 'bread'. The fact that he restricts his use of Turkish to simple naming rather than whole sentences is further evidence that he is attempting to use the appropriate language for the Germanspeaking interlocutors. German predications with Turkish naming While the earliest transcripts show almost exclusive use of Turkish apart from the 'emblematic' formulaic uses of German, gradually the use of Turkish diminishes. When a syntactic frame for predication is acquired, the use of Turkish subsides to 'naming', primarily of 'things' but, as more competence in German develops, also to verbal predicates. Orhan fixes on a frame, das 'that', which allows naming in Turkish to be syntactically

104

Carol W. Pfaff

integrated into German. At first the structure is incomplete, lacking copula and articles and with the new information at the beginning of the utterance, as in [5] - [6]: [5]

2;01

xx das

'xx that'

[6]

2;06

köpek das

'dog that'

By 3;03 such utterances have the form of German copular sentences but without an explicit copula, with the new information, the name in final position as in [7]: [7]

3;03

das auch balik

'that's also a fish'

Shortly thereafter, this structure develops to a form with an explicit copula, is(t) as in [8], and becomes a grammatical German syntactic frame for insertion of Turkish nouns: [8]

3;06

das is ςαία [=qatai]

'that is (a) fork'

German predications with Turkish verbs Subsequently, Orhan extends the das is X frame to structures which permit the insertion of Turkish verbs in German contexts, as in [9]: [9]

4;02

das is auch ör+

'that is also knit+' 'and she's knitting'

The example in [9] appears to be a truncated instance of the is+V construction, found, more typically with German infinitive verb forms, in the Turkish of several of the Turkish/German bilingual children discussed in Pfaff (1991, 1992, 1994) and in Kuhberg (1987). It has also been reported in the speech of adult first generation immigrants (Von Stutterheim, 1986). It seems to correspond to the Turkish present progressive construction, which has no parallel in German syntax, and represents a type of what Johanson (1993) terms 'selective code copying', and which von Stutterheim has called 'conceptual transfer'. This structure for incorporating Turkish verbs in German, though it has the same superficial structure as the nominal frame, occurs in Orhan's German only after he has developed German auxiliary use of the verbs haben and sein at age 3;7, as reported in Pfaff (1992).

Changing patterns of language mixing in a bilingual child Alternation to Turkish

105

participant

At this stage of development, Orhan has established a variety of German which allows for insertion of Turkish lexical items in both nominal and verbal structures. Hereafter, the use of alternations to complete sentences in Turkish during German conversations declines, with the exception of utterances in which the child addresses a Turkish-speaking participant or bystander, as found in [10] at 7;02 and [11] at 8;00. Note that the Turkish adult researcher is also present during the recording of the German conversation - and that she participates in the German conversations, speaking in German. In [10], Orhan is speaking German with Karin (the German interviewer) and Tülay (the Turkish interviewer) about what his family did at the festival of sacrifice. Orhan switches to address Tülay in Turkish when he runs into vocabulary difficulties about a Turkish dish containing meat [10-9], responding to her teasing question in [10-8] which refers to their earlier conversation in Turkish where he claimed that eating meat was an essential part of the celebration (which it is not). The passage is quoted at length because it displays relevant syntactic variation in addition to lexical mixing, as discussed below. One of the rare examples of an intrasentential switch longer than one word is found in [10-9], apparently an alternation triggered by the vocabulary gap. It begins in the middle of the sentence between the German word, fleisch 'meat' and the Turkish word fasulye 'beans'. This is a case of ragged switching and in which the object noun 'meat' is doubled, once in each language, et and fleisch. Perhaps this is a production strategy which minimizes the non-congruence of the two languages in verbal morphosyntax and word order. Alternatively, Orhan simply starts over in Turkish after breaking down lexically in German. I suggest, however, that the switch here actually begins syntactically before the occurrence of the Turkish lexical item. This analysis is based on the fact that in this utterance, Orhan atypically omits the German auxiliary from the perfective construction so that, while the second part of the utterance is a complete Turkish sentence, the structure of the first part is not standard German, lacking the auxiliary verb haben. Such omissions are not characteristic of Orhan's interlanguage grammar at this stage. In fact, the sequence in [10] happens to contain several relevant instances of the perfective structure, in sentences 2,1, 11 and 16-18. In all of these instances, the auxiliary is not only present, it occurs in standard second position, even when the preposed object requires

106

Carol W. Pfaff

subject verb inversion: wir haben nicht gefeiert, ich hab nicht gesagt, wir haben+ and Bohnen und Fleisch /mhm/ haben wir geeßt.5 [10]

7;02

Conversation about the Festival of Sacrifice

1

Karin:

2 3 4 5 6 7

Orhan: Karin: Tiilay: Karin: Orhan: Orhan:

8

Tiilay:

9

Orhan:

10 11 12

Tiilay: Orhan: Orhan:

was habt ihr zuhause gemacht dann? wir haben nicht gefeiert nee? ach schade und warum nicht? weiss ich nich ich hab nich gesagt [=? niemand hat mir gesagt] und Fleisch auch nicht gegessen? doch wir Fleisch fasulye yedik etlen4 [laugh] wir haben [/] Deutsch heisst wie wie wie?

13 14

Orhan: Orhan:

neydi ya? neydi almancasifasulyenin ?

15 16 17 18

Tiilay: Orhan: Karin: Orhan:

Bohnen Bohnen und Fleisch mhm haben wir geeßt [=gegessen]

'what did you do at home then?' 'we didn't celebrate' 'no?' 'oh, too bad' 'and why not?' Ί don't know' Ί didn't say' [=? they didn't tell me] 'and didn't eat meat either?' 'yes, we meat we ate beans with meaf [laugh] 'we have' '(in) German it's called what what what?' 'what then?' ' what do you call beans in German?' 'beans' 'beans and meat' 'mhm' 'we eated'

A further example of alternation to Turkish within a German conversation is found in [11] at 8;00. Again, the Turkish interviewer is also present. Orhan is discussing what he did on vacation in Turkey with Tina and Tiilay and Orhan switches to Turkish, addressing Tiilay:

Changing patterns of language mixing in a bilingual

[11] 1 2

8;00

Conversation about vacation in Turkey

Orhan: Orhan:

3 4 5 6 7 8

Orhan: Tina: Tülay: Orhan: Tülay: Orhan:

da bin ich immer/ da gehen wir immer zu schwimm dann gibs draußen so/ wohin? schwimmen? ja?

child

107

was?

'there I always/' 'we always go there to swim' 'then outside there's such a/' 'where?' 'swimming? yeah?' 'ach uh at (the) sea' 'what?'

denizde

'at the sea'

ach aaa an deniz j a

The same Turkish lexical item, deniz 'sea', is involved in both instances of CS, but the grammatical system or matrix language changes. The first utterance is in German, with case marking with a preposition (though without the necessary article), while the second utterance addressed to Tiilay's question (in German) switches entirely to Turkish, with appropriate postpositional marking. Embedded language islands / convergent

lexicalization

Though we find considerable alternation and insertion, there are relatively few instances of 'embedded language islands' in Orhan's speech at any stage. One instance was found in the ragged switch in [10-9] discussed above. Another example is [12], which occurs at 4;02 during the stage of maximal use of Turkish integrated in German. Here he seems to be attempting both simultaneous morphological and syntactic integration of Turkish and German. The utterance occurs in the context of a German recording, talking about the picture book, Lady and the Tramp, telling how Lady, the dog, wants to sleep in her master's big bed rather than her own small one. Orhan begins in German, breaks off, restarts with a mixed sentence and ends with a mixed German/Turkish NP: [12] 1 2

4;02

Picture book ein große [grein] groß [is] klein großi großirin yataga var

'a big [?] big is little' 'there's (a) big bed like this,

böyle, ein großinin yataga

a big bed'

As discussed more extensively in Schmidt (1995), the first part [12-1] clearly begins as a German noun phrase with the indefinite article ein and the

108

Carol W. Pfaff

German adjective groß. However, Orhan does not (cannot?) continue with the German word Bett and the rest of the sentence is unclear, syntactically and semantically, though apparently involving a contrast between the large and small beds. He begins again, in [12-2], using the Turkish noun for 'bed', yataga, combined syntactically with the German adjective groß to form an EL+ML constituent NP as an apparent possessive (with nonstandard morphology) in a Turkish existential construction. The NP is then repeated, postposed, this time a mixed Turkish/German NP in which the article and adjective are lexically German; the adjective is inflected in Turkish and the noun is Turkish. Thus, this NP has both system morphemes as well as content morphemes from both languages. It appears from the above examples that in this intermediate period, Orhan has the competence to resolve discongruities which arise in mixing and, in these cases, rather than triggering 'embedded language islands', he does so by combining system morphemes of both languages in [12] and later by sacrificing German morphosyntax while maintaining Turkish, as in [10]. This topic and more generally the issue of 'selective code copying' in utterances with and also without lexical mixing deserves much closer investigation. Changing patterns of mixing Turkish into German To recapitulate, we have found that Orhan's pattern of mixing Turkish into German develops as follows: First German occurs as limited alternations of formulaic utterances within sequences in Turkish. As German competence grows, German becomes the base language and Turkish lexical items, especially nouns, are inserted into (developing) German syntax. Alternation to Turkish - except for addressing Turkish interlocutors who happen to be present - is very limited, seemingly triggered by lexical gaps. These few instances result in ragged switching rather than full alternation or embedded language islands. The trend thus is toward monolingual German, using as much German structure as possible and, after the naming phase, keeping as much as possible to German lexical items. The lack of lexical incursion of Turkish into German can be linked to the behaviour of interlocutors who consistently suggest German words even if they do not insist on the child's producing them. Whether these words are internalized as exact translation equivalents, however, remains to be seen. Below, I present some evidence that in mixing in the opposite direction, German in Turkish conversation, the words are not initially taken as translation equivalents.

Changing patterns of language mixing in a bilingual

child

109

German in Turkish: increasing use of German with decreasing use of Turkish Turning now to German in Turkish, we find a pattern which is almost the opposite of the development of mixing into German. The earliest Turkish recordings contain no German at all. As shown in Table 2, the overall type/ token ratios of German in Turkish rise, with a considerable jump at the last session, when Orhan is 8;00. Table 4 shows instances of insertion and naming and the younger period, with insertion becoming increasingly frequent until, in the latest recordings, complete alternation becomes more characteristic. To examine the specifics of this pattern, consider a few examples from early, middle and later conversations. Formulaic expressions The first use of German elements in Turkish does not occur until 2;06, almost a year later than the first use of German in interactions with German interlocutors. It is noteworthy that the first German element found in this context is a formulaic rather than referential expression, [13] where Orhan comments on the fact that his two-year old Turkish playmate has hurt herself. She exclaims aua! 'ouch', using the German child language expression; Orhan comments using a syntactically integrated insertion, substituting the German aua for the equivalent child Turkish expression, uf oldu 'got hurt': [13]

2;06

Aua oldu Berranin

'Berran got hurt'

Insertion of bare forms as naming Thereafter, Orhan begins with insertions of German lexical items in Turkish. The first instances occur at 2; 10. The two types are the animal name Elefant 'elephant' and the sports-domain term Fußball 'football or soccer(ball)'. Words in these domains are characteristically German for many children and adults accept them. In the utterance with Elefant, for instance, the adult even repeats the German word and does not mention the Turkish equivalent fil. In [14] the word Fußball occurs as naming and is also syntactically integrated into Turkish as a direct object, though not marked as definite accusative. The interviewer repeats Fußball but suggests top, apparently not because she wants to suggest a 'translation equivalent' but because she

110

Carol W. Ρ f a f f

rejects the identification of the pictured ball as a football. The ensuing interchange is about how one plays with balls of various types and eventually, Orhan capitulates and uses the Turkish top. 2; 10

Picture book

1 2 3

Tiilay: Orhan: Tiilay:

4

Orhan:

5

Filiz:

oyuncaksi mi? Fußball Fußball Fußball, Fußball [=!laugh] top o, top, top top degil, Fußball atiyo böyle Fußball atiyo gocuk böyle

6

Tiilay:

Fußball atiyo 90cuk böyle

7 8 9

Tiilay: Tiilay: Orhan:

Fußball atmiyor top oynuyor 90cuk top oynuyor

[14]

'is he going to play?' 'football, footbalV \football, football [laugh] that's a ball, ball, ball' 'it's not a ball football he kicks like this' 'football the boy throws like this' 'football the boy throws like this' 'he's not kicking a football' 'the boy is playing ball' 'he's playing ball'

At 3;02, in [15], the issue of terminology arises again in the conversation, this time with a different outcome. Orhan uses the German Brötchen 'roll' and when the interviewer suggests ekmek 'bread' as a translation equivalent, he rejects it consistently. In this case, it is the interviewer who gives in, adopting the German word, which indeed is a more appropriate term for the type of food being eaten by the German family pictured in the book. This type of example illustrates the role of 'specificity' in determining CS and language mixing discussed by Backus (1996). 3;02

Picture book

1 2 3 4 5 6 7

Filiz: Orhan: Filiz: Filiz: Orhan: Orhan: Filiz:

8 9 10 11

Orhan: Filiz: Orhan: Filiz:

... ne yapiyolar? Brötchen yiyolar 'Brötchen' yiyorlar, ha? Ekmek yiyorlar, ekmek Ekmek yemiyolar Brötchen yiyolar Ε 'Brötchen' ekmek degil mi? Ekmek digil Ne ο peki? Brötchen Burgin, peki...

[15]

'... what are they doing?' 'they're eating rolls' 'they're eating 'rolls', ha?' 'they're eating bread, bread' 'they're not eating bread' 'they're eating rolls' Έ 'rolls' are bread aren't they?' 'not bread' 'what is it then?' 'rolls' 'rolls, ok...'

Changing patterns of language mixing in a bilingual child

111

Note the two spellings for the same lexical item in the transcription, which was made by the interviewer herself. She represents Orhan's more German phonology - and her own when she is discussing the lexical item with him in standard German orthography, Brötchen, but uses a more Turkish orthographic representation, biirgin, to indicate that she uses a more Turkish phonology for her acceptance of the term. This example captures an instance of an interactive process which surely plays an important role in the relexification of an immigrant language: a first generation immigrant speaker who undoubtedly has passive knowledge of the lexical item in question (and may even use it in conversation with members of the host community) adopts it as a 'loan word' into her ethnic language as a result of interaction with the second generation speaker, for whom it may be the core lexical item. Alternation to participant Orhan next acquires control of language alternation in response to participants. In example [16], at 3;08, he is playing 'telephone' with another Turkish child: [16] 1 2 3

3;06

Playing 'telephone'

Orhan: Orhan: Orhan:

Konu§um bi Almanca xx Wie gehts aber? Ich kann xx Nein. Oooo, ich kann nicht. Manuu. [=Nanu1] Aber

Til speak a (little) German xx' 'but how 's it going ? I can xx' l No. Oooo, 1 can't. [=oh] But'

This role-play episode in which Orhan pretends to be speaking to a German interlocutor on the telephone shows the same type of formulaic expressions that occur in his early German conversations, recorded with actual adult interlocutors. Note that they are complete switches to German, not mixed with Turkish, but with very little lexical content. Insertion with inflectional

morphology

By age 4;03, Orhan has acquired a considerable vocabulary in German. In [17-18], we find instances of insertion of German lexical items integrated into Turkish morphosyntactically. In [17], the noun Vorschule 'preschool' occurs with the Turkish dative suffix, the pattern typical of adult usage as well:

112 [17]

Carol W.Pfaff 4;03

Vorschule'ye

'to preschool'

Integration of German lexical items as Turkish verbs is seen in example [18] at 4;03, using the pattern noun (or bare verb stem) + auxiliary yapmak, a pattern he retains as late as 7;02 in [19]. In [20], at 8;00, we find the first occurrence of the more frequently reported form with the infinitive: [ 18]

4;03

hex yapiyo

's/he is casting a spell'

[19]

7;02

grill yapti

4

[20]

8;00

gratulieren yapiyo

'(she) congratulates'

(s/)he grilled'

The use of this light verb yapmak 'do, make' as an auxiliary has been discussed extensively in the literature on Turkish in contact (e.g., Boeschoten, 1990; Pfaff, 1991; Backus, 1996) and represents a pattern found in numerous other languages as well (Romaine, 1995; Myers-Scotton, 1993; Muysken, 1997). Note that the earlier examples of verb mixing depart from the typical adult form with a German infinitive+yapmak, but they do not employ the device of combining a German stem with Turkish verbal inflections, which Backus and Boeschoten (1996:3) suggested is a widespread pattern in early bilingual acquisition. This pattern was not found at all in Orhan's production. 6 Insertion with derivational

morphology

Some particularly interesting examples of morphosyntactic integration with derivational morphology are illustrated by [21] about the fall of the Berlin wall. This historic event impinged directly on the children's lives since one of the Kita buildings was located immediately adjacent to the wall. When the wall was literally broken through, many people from both sides came to participate in the ritual (and business) activity of tearing it down to save or sell pieces as souvenirs. Orhan is not too clear about the political significance, but he is definitely aware of the talk and activities at that time.

Changing patterns of language mixing in a bilingual child 4;03

Conversation about Berlin wall

1

Filiz:

ne, nasil bir duvar o?

2

Orhan:

Ost'lar var

3

Orhan:

Ost/ Ost vardi ya.

4

Filiz:

mhm. Ost vardi

6

Filiz:

niye ο duvara vuruyolar?

7

Filiz:

dursaydi öyle

8

Orhan:

9 10 11

Filiz: Orhan: Tiilay:

12 13

Orhan: Tiilay:

14 15 16

Orhan: Tiilay: Orhan:

ama yapamiyolar Ost Ost'gular Ost'gularl Ost'a gitmek istiyolar [=!laugh] ee ne yapcaklar oraya gidip? duvari kircek peki sen Ost'gularl gördiin+mü? gördüm nasil onlar? böyle [acts out hitting the wall]

[21]

113

'what, what kind of a wall is 11 'there are Easts [or Easterners]' 'East/ there was (an) East you know?' 'mhm, there was (an) East'

c J

'why are they hitting that wall?' 'it could have stayed the way it was' 'but they can't do it, the East Easterners' 'Easterners?' 'they want to go to the East' '[laugh] so what will they do there?' 'they will break the wall' 'well, have you seen the Easterners?' 'I've seen (them)' 'what are they like?' 'like this [acts out hitting wall]'

In addition to the regular case inflection in Ost'a 'to the East', Orhan uses two terms to designate people from East Germany, Ost'lar 'Easts' and Ost'cular 'Easterners'. The first parallels the German term Ossis or Ostler 'people from East Germany'. Given the phonological similarity of the Turkish Ost'lar and the German Ostler, this instance could also be analyzed as an insertion of the morphologically complex German lexical item. Orhan's second term, Ost'cular, uses the Turkish derivational morpheme cu which normally designates someone who habitually or professionally does something. The usual derivational suffix which designates place of origin is -II, so the regular formation from this German noun would be Ost'lular. There is a form current in migrant Turkish speech which is parallel to the one Orhan uses, Almancilar 'Germans', which designates Turks who have lived in Germany. It is impossible to know whether Orhan has heard anyone using this form - or perhaps even the form Ost'cular itself, or whether this is a neologism he himself has coined. It is noteworthy

114

Carol W.Pfajf

that although the interviewer suggests the alternative Ost Berlin 'e 'to East Berlin' for Orhan's form Ost'a, here she simply adopts his form Ost'cular without attempting to substitute a translation equivalent or even a standard derivational form. The segments of conversation in [22] and [23] at 7;02 about what Orhan does at the sports center provide further examples of morphological and syntactic integration of nouns in the domain of sports, with nouns such as Halle 'hall, gym,' and Tischtennis 'table tennis'. Note that the interviewer accepts all of these forms without question and in fact uses them herself. Note the use of the plural of the proper name, Boris Becker in Boris Becka'lar gibi 'like Boris Beckers', likely a coinage rather than a form which he has learned as a lexical item. Orhan's vowel harmony is consistent with the colloquial German r-less pronunciation with a schwa, while the interviewer's repetition Boris Becker'ler uses the front vowel allomorph consistent with the spelling or, as Boeschoten (p.c.) suggests, an instance of vowel harmony with the preceding syllable. [22]

7;02

Conversation about sports center

1

Orhan:

2 3 4 5

Tülay: Orhan: Tülay: Orhan:

oynuyoz Halle 'ye gidiyoz orda oynuyoz Halle'ye gidiyosunuz? Halle neresi bu Halle Halle/ orda/ orda futbol da oynayabilirsin

7;02

Conversation about tennis

1

Orhan:

2 3

Tülay: Orhan:

4

Tülay:

basketbol oynuyoz Tischtennis oynuyoz orda hmhm tenis oynuyoz ondan sonra §ey oynuyoz nasil ne tenisi oynuyosunuz?

5 6

Orhan: Tülay:

eh + eh Boris Becka'lar gibi Boris Becker'ler gibi ya?

[23]

'we play, we go there to the gym' 'you go to the gym?' 'gym' 'where is this gymT 'gym/ there/ you can play football there'

'we play basketball and table tennis there' 'hmhm' 'we play tennis then we play what-do-you call it' 'how what kind of tennis do you play?' 'eh + eh like Boris Beckers' 'like Boris Beckers yeah?'

Changing patterns of language mixing in a bilingual child Triggered

115

alternation

Finally, at age 8;00, in the last recording, Orhan's use of sports terminology in the context of discussion of sports at school now appears to trigger complete alternation to German while conversing with the Turkish interviewer only. In [24], he uses the word Kegel'lari 'pins' as an insertion with Turkish morphology. It is likely that he does not know the Turkish equivalent term. The alternation to German in the next utterance, however, is very unlikely to be due to a lexical gap, in light of his appropriate Turkish formulation of the same content in the first utterance of the passage. [24]

8;00

Conversation about bowling

1 2

Orhan: Tiilay:

3 4 5

Orhan: Tiilay: Orhan:

ben hepsini dü§ürdüm neleri dü§ürüyorsun attigin zaman? Kegel'lari mhm die müssen alle runter gehen

Ί knocked them all down' 'what do you knock down when you throw (it)?' 'the pins' 'mhm' ' they all must go down'

The final passage to be cited here, [25] also at 8;00, shows a complex variety of triggered alternation and insertions of types discussed above and, additionally, illustrates Orhan's control of rules of both Turkish and German pluralization in a structure where these are not congruent. The German lexical item Runde 'lap, round' occurs both as an insertion and as part of an alternation and in both languages it occurs with the plural quantifier 'ten', German zehn and Turkish on. Orhan conforms to Turkish rules in Turkish stretches and to German rules in German stretches. Thus plural morphology occurs on the noun in German, zehn Runden, but is absent in Turkish on Runde, on Runde'yi, in accordance with the Turkish rule which specifies that plural is not marked on nouns in NPs with plural quantifiers.

116

Carol W. Pfaff

8;00

Conversation about class 'Athlete of the Day'

1

Tülay:

siz de var mi öyle bi§ey?

'do yo also have such a thing?'

2 3

Orhan: Tülay:

nein sie gratuliert

'no, she congratulates'

ne kim? anlamadim ama türk^e ko\

'what? who? I don't understand, but sp- Turkish'

4

Orhan:

bei Laufen! bei Lau eh:

'when we run! when ru- uh: ii

ko§uda on Runde 'yi

IS]

5 6 7 8

Tülay: Orhan: Tülay: Orhan:

mhm gratulieren yapiyo mhm kim yapiyo eh ögretmen on

a race to whoever do-/ does ten lap' 'mhm' 'she congratulates (him/her)' 'mhm who does?' 'uh the teacher ten

9

Orhan:

Runde! weil wir zehn Runden gemacht haben

lapsΓ 'because we have done ten laps'

10

Tülay:

11

Orhan:

anlamadim anlamadim bi daha on Run! on/ on defa ko§tuk

Ί don't understand, I don't understand, (say it) again' 'ten la- ten1 ten times we ran

Runde

lap'

In this passage, as in the earlier ones, the interviewer does not react negatively to Orhan's use of German words in Turkish. However, she finally does protest when the sentences become entirely German, using the excuse that she doesn't understand. We cannot know here whether these alternations would have been maintained over a long stretch if she had not intervened, or whether he would have produced longer 'embedded language islands' in this case. Although further evidence would be required for confirmation, my hunch is that a complete switch to German would have been likely since the alternations here and in the preceding example are not limited to syntactic constituents and formulaic expressions. Changing patterns

of mixing German into

Turkish

The above examples have illustrated Orhan's development from using no German in Turkish, then increasing insertion of German nouns and verbs syntactically and morphologically integrated into Turkish structure, and finally complete alternation to German, triggered by use of key lexical items in German-dominant domains. They also demonstrate how lexical incursion of German into Turkish is facilitated by the interlocutors, who frequently do

Changing patterns of language mixing in a bilingual

child

117

not suggest translation equivalents and who themselves use the same items, sometimes apparently adopting them from him.

Conclusions and discussion The findings reported above show changes in the patterns of overt language mixing over time, reflecting the developing linguistic and sociolinguistic competence of the Turkish/German bilingual child. The type of language mixing and codeswitching (insertion or alternation) reflects his increasing proficiency in the two languages as well as his changing ideas about the linguistic competence of his German and Turkish interlocutors and his notions of communicative norms in the community. The patterns diverge sharply in the two languages, clearly showing his development toward a monolingual mode in German and toward a bilingual mode in Turkish. They also illustrate his tendency toward a complete language shift to German, at least in certain domains. In particular, it is the proficiency in his second language, German, which appears to determine the changing patterns of mixing in both German and Turkish contexts. In interactions with the German interlocutors, he initially uses Turkish almost exclusively, with a few German discourse markers and formulaic expressions. Subsequently, he masters minimal structures for German predications in which he names objects in Turkish. In an intermediate phase, we find a few instances of utterances of mixed Turkish/ German syntax. Finally, his utterances in German conversations are completely German, except for alternation to Turkish to address a Turkish interlocutor, typically when confronted with a vocabulary gap. In general, Turkish words are not adopted into German. This behaviour is typical of the monolingual mode. In contrast, in interactions with the Turkish interlocutors, almost the opposite pattern of development is found. First there is a considerable period of monolingual Turkish. Insertion of German formulaic phrases and lexical items begins well after he has used them in German contexts. First nouns and then verbs are morphosyntactically integrated into Turkish structures. Finally, the insertions trigger language alternations, complete switches to German, even when talking to the Turkish interviewer. This behaviour is typical of the bilingual mode. In addition to his own proficiency, the changing patterns reflect the child's changing perceptions of the proficiency of his interlocutors. His language choice and mixing behaviour reveal early sensitivity to the preferred language of the interlocutor, whose perceived language preference

118

Carol W. Ρ faff

he attempts to match. Thus, as early as 1;07, it appears that Orhan is aware of the role of language in establishing intersubjectivity. At the early stages when he has almost no proficiency in German, he attends to the German interviewers and responds with his minimal bits of formulaic German. Similarly, Orhan's side comments to Turkish participants, at the earlier stages, appear to follow his notions about the participants' preferred language, Turkish, even if the speaker had in fact been using German to him. These turn-initial alternations provide evidence that he has a sense that a 'one person - one language' strategy is appropriate. In contrast, in the few instances of alternations to Turkish in German conversations when Orhan is older, the 'one person - one language' strategy is no longer employed. In these conversations Orhan continues to speak German - even in utterances directed to the Turkish interviewer - until he is confronted with a vocabulary gap in conversation about a Turkish domain, which eventually triggers a switch. In the later Turkish conversations, Orhan alternates to German in the supposedly monolingual interactions with the researchers. This is evidence that he realizes that they are also bilingual and share at least passive knowledge of the lexical items and structures of German that he uses. His patterns now conform to the language mixing and CS patterns which have become established in his wider speech community. It is noteworthy that the pattern of development of CS in this child's speech parallels the successive patterns characteristic of three generations of Turkish/Dutch bilinguals by Backus (1996, this volume). For these adults, insertion of Dutch lexical items in Turkish was characteristic of the first generation immigrants. A mixed system with maximal switching of all types was found in his 'intermediate generation' born in Turkey but raised in the Netherlands, and alternation to Dutch in the second generation born in the Netherlands. For both children and adults, the mixing patterns reflect their increasing proficiency in the host language syntax and lexicon and their declining use of Turkish with peers and others in school (or work) domains. The parallel developments in the speech of one child studied longitudinally and in adults studied cross-sectionally over different generations reinforce the theoretical interest in studying development in both childhood and history, not only in the structural terms suggested by Jakobson (1941), Slobin (1977) and others but also in terms of proficiency and use. In addition to their parallel progression, child and adult patterns of language mixing are also more overtly linked. The child's developing patterns of mixing are fostered by the language behaviour of the adult interlocutors. In particular, relexification of Turkish is facilitated by

Changing patterns of language mixing in a bilingual

child

119

members of the Turkish/German community who accept German terms which have no Turkish equivalents or are specific to German contexts rather than encouraging the child to use Turkish translation equivalents. Most of the discussions between the child and his adult interlocutors about appropriate vocabulary hinge on nuances of meaning rather than making an issue of using Turkish rather than German. Their bilingual behaviour is reminiscent of the findings of research on input and feedback to monolingual first language acquisition, which has shown that parents and caretakers correct children on points of fact rather than on grammatical well-formedness. Only a very limited amount of 'selective code copying' or convergence was found in the environment of overt language mixing in structures which are syntactically or semantically non-congruent. Rather, as illustrated by our findings with respect to pluralization, word order and the presence of auxiliaries in verb phrases, the systems of Turkish and German seem to be maintained separately in the child's grammars in such contexts. Although as would be expected in the context of diachronic change in a migration setting, there is evidence of convergence found in this child's grammar and in the grammars of some of the other children (see Pfaff, 1992, 1994), evidence of it shows up more in the monolingual passages than in conjunction with overt language mixing. The progress of grammatical convergence requires much more investigation and will be treated elsewhere.

Notes 1. This chapter was prepared while I was a 'Fellow in Residence' at the Netherlands Institute for Advanced Study in the Humanities and Social Sciences (NIAS) 1995/96. I am grateful to NIAS and to members of the bilingualism theme group, Pieter Muysken, Annick De Houwer and particularly Rik Boeschoten for discussion. 2. The KITA study entitled 'Natürlicher bilingualer Spracherwerb von KitaKindern: Vom Krippenalter bis zu den ersten Grundschuljahren' (Natural bilingual language acquisition of day-care children: from nursery to the first primary school years) was funded by the Deutsche Forschungsgemeinschaft 1987-1992 under grant #Pf-201/l-5. The sample comprised 22 Turkish, 5 German and 4 children of mixed marriages. I am indebted to the children, their parents and the staff of the VAK Kita as well as to the members of the project staff and students: Filiz Kardam, Tülay Sava§, Ogiin £akarcan and Jutta Voß, Heike MacKerron, Martina Thiel, Karin Schmidt, who conducted and/or analyzed Turkish and German interviews, respectively.

120

3.

4. 5.

6.

Carol W.Pfaff

A number of publications present various aspects of the study. Kardam and Pfaff (1993) discuss the general social and neighbourhood setting and Pfaff et al. (1987) have details of the Kita, the design and methodology. Some findings on the language mixing of several children in this and an earlier study are reported in Pfaff (1991). Early German second language, Turkish first language and individual differences are discussed in Pfaff (1992, 1994). Examples are given in modified CHAT transcriptions, using standard German and Turkish orthography, with the addition of apostrophes between German stems and Turkish suffixes. Boeschoten (1990) notes that forms like etlen are typically Anatolian, with -len as the realization of the standard commitative suffix -le/la. Note that in other respects Orhan's grammar is still nonstandard, as seen in overgeneralized past participle of an irregular verb geeßt 'eated' in [10-18] and, more strikingly, in the nonstandard semantics in [10-7]. In fact only one child, the least proficient Turkish speaker in our sample, used such a form (see Pfaff, 1991).

References Backus, A. (1996). Two in One. Bilingual Speech of Turkish Immigrants in the Netherlands. PhD thesis. Tilburg University. Tilburg: Tilburg University Press. Studies in Multilingualism 1. Backus, A. (this volume). The intergenerational codeswitching continuum in an immigrant community. Backus, A. and H. Boeschoten (1996). Turkish-Dutch codeswitching: levels of lexical structure. Paper presented at the annual meeting of the Linguistic Society of America, San Diego. Boeschoten, H. (1990). Acquisition of Turkish by Immigrant Children. PhD thesis. Tilburg University, Tilburg. Boeschoten, H. and L. Verhoeven (1985). Integration niederländischer lexikalischer Elemente ins Türkische: Sprachmischung bei Immigranten der ersten und zweiten Generation. Linguistische Berichte, 98, 347-364. Grosjean, F. (1985). The bilingual as a competent but specific speaker-hearer. Journal of Multilingual and Multicultural Development, 6, 467-477. Jakobson, R. (1941). Kindersprache, Aphasie und allgemeine Lautgesetze. Uppsala: Sprä kvetenskapliga Sällskapets Förhandlingar. Johanson, L. (1993). Code-copying in immigrant Turkish. In: G. Extra and L. Verhoeven (Eds.), Immigrant Languages in Europe. Clevedon: Multilingual Matters. Johanson, L. (this volume). Frame-changing code-copying in immigrant varieties.

Changing patterns of language mixing in a bilingual child

121

Kardam, F. and C. Pfaff (1993). Issues in educational policy and language development. In: S. Kroon, D. Pagel and T. Valien (Eds.), Multiethnische Gesellschaft und Schule in Berlin. Münster/New York: Waxmann. Kuhberg, H. (1987). Der Erwerb der Temporalität des Deutschen durch zwei elfjähriger Kinder mit Ausgangssprache Türkisch und Polnisch: Eine Longitudinal-untersuchung. Frankfurt/M.: Peter Lang. Muysken, P. (1997). Codeswitching processes: alternation, insertion, congruent lexicalization. In: M. Pütz (Ed.), Language Choices: Conditions, Constraints and Consequences. Amsterdam: Benjamins. Myers-Scotton, C. (1993). Duelling Languages: Grammatical Structure in Codeswitching. Oxford: Oxford University Press. Pfaff, C. (1991). Mixing and linguistic convergence in migrant speech communities: linguistic constraints, social conditions and models of acquisition. In: Code-switching and language contact: Constraints, conditions and models. Strasbourg: European Science Foundation Scientific Networks. Pfaff, C. (1992). The issue of grammaticalization in early German second language. Studies in Second Language Acquisition, 14, 273-296. Pfaff, C. (1994). Early bilingual development of Turkish children in Berlin. In: G. Extra and L. Verhoeven (Eds.), The Cross-Linguistic Study of Bilingual Development. Amsterdam: Royal Netherlands Academy of Arts and Sciences. Romaine, S. (1995 [2nd ed.]). Bilingualism. Oxford: Blackwell. Sava§, T. (1994). Kommunikationsstrategien im Erst und Zweitsprachgebrauch am Beispiel zweier türkischer Kinder in Kreuzberg. Unpublished MA thesis. Freie Universität Berlin, Berlin. Schmidt, K. (1995). Die grammatische Kategorie des Adjektivs im Erst- und Zweitspracherwerb. Unpublished MA thesis. Freie Universität Berlin, Berlin. Slobin, D. (1977). Language change in childhood and history. In: J. Macnamara (Ed.), Language Learning and Thought. New York: Academic Press. Von Stutterheim, C. (1986). Temporalität in der Zweitsprache. Berlin: de Gruyter.

Word formation processes in young bilingual children1 Hanneke van der Heijden

The intriguing language acquisition as a whole may be, acquisition of a language's lexicon in particular has an aura of infeasibility around it for those who finally managed to master the system. When two languages are involved lexical acquisition seems even more amazing. Part of the amazement about language learners' achievements in this field probably emanates from the lexicon's seemingly idiosyncrasy. But even if this impression is correct for part of the lexicon, another part consists of lexemes which are the outcome of word formation. These devices of word formation can relieve the task of lexical acquisition considerably. Broeder et al. (1994:114) distinguish four devices: • creation of new roots; • compounding: combining of two or more lexemes; • derivation: combining of a lexeme with one or more bound morphemes; • conversion of existing lexemes into other word classes. In this chapter an analysis will be made of the role of two of these devices in language acquisition: compounding and derivation. A restriction will be made to nominal word formation, i.e., compounds and derivations with nominal heads or stems. Acquisition of these devices will be studied in young children involved in the simultaneous acquisition of Turkish and Dutch, two typologically different languages. In the domain of word formation several studies on first and second language acquisition have been carried out. Simultaneous acquisition of two languages, however, has hardly been the subject of research in this area. In studies by Clark (1981, 1983) it was shown that one of the leading principles in the structure of acquisition of word formation devices in a first language is transparency. Transparent word formation devices are claimed to be acquired earlier than opaque ones. In general, derivation is considered to be a more opaque word formation device than compounding. Arguments in favour of this view are: (1) bound morphemes, used in derivation, never occur as independent words, (2) they are usually unstressed, and (3) they have rather subtle and sometimes also ambiguous meanings. Compounding is therefore believed to appear in language learner data before derivation does. A crosslinguistic study on second language acquisition by adults,

124

Hanneke van der Heijden

among whom Turks acquiring Dutch, supported this hypothesis (Broeder et al., 1994). These adult learners of Dutch used a great deal of compounds in referring to entities whereas derivations only made up a very small proportion of the word formations. Following such results a first hypothesis on the relative order of acquisition of both devices can be formulated. Hypothesis I: Compounding precedes derivation in the acquisition of word formation devices This hypothesis only concerns the relative order of acquisition of both devices within one language. An interesting case can be presented by the simultaneous acquisition of two structurally different languages. In such an acquisitional context structural characteristics of each of the languages and its preferences for certain devices over others might conflict with the other language's typology. Thus, acquisitional data on word formation devices might give indications on the general applicability of these hypotheses and on the interrelatedness of two language systems in acquisition. From a structural point of view, Turkish is a highly agglutinative language (e.g., Lewis, 1985). It has a wide stock of word formation suffixes with few allomorphs. In compounding, lexemes are ordered in a modifierhead structure, the head being inflected with a possessive ending. Also, but to a smaller extent, linearity is possible in compounds if both elements express an identity of the person or object denoted. Examples of these three devices are given in [1]. [1]

boyaci boya fabrikasi kadm polis

paint-AGENT paint factory-POSS woman policeman

'painter' ' paint factory' 'police woman'

Dutch, on the other hand, is a synthetic language (Donaldson, 1983; Geerts et al., 1984). Derivation by means of pre- and suffixes does occur, but these morphemes can not always be isolated easily from the root. Besides, there is a rather large range of allomorphs. In noun-noun compounding, modifierhead order is obligatory. Examples are given in [2], [2]

fietser

bicycle-AGENT

'cyclist'

fietsband

bicycle tyre

'bicycle tyre'

Considering Clark's principle of transparency in language acquisition and given the fact that word formation suffixes in Turkish are more transparent

Word formation processes in young bilingual children

125

in form and function than Dutch ones, a second hypothesis can be formulated concerning the relative order of appearance of derivation in the acquisition of the two languages under consideration. Hypothesis II: Derivational devices are applied in Turkish before they are applied in Dutch Summarizing, both in Dutch and in Turkish we expect compounding to appear first in language development. Next, derivational devices are expected to be used by the language learner, but these will appear earlier in Turkish than in Dutch. In this chapter both hypotheses will be tested on acquisitional data from Turkish/Dutch bilingual children growing up in the Netherlands, as compared to acquisition in monolingual children. As will be clear, both hypotheses concern the order of acquisition. Apart from this a qualitative analysis will be presented of form and functions of word formation devices in language acquisition.

Informants and data base Data derive from a research project on early bilingual acquisition of Turkish children in the Netherlands 2 . This project is designed as a multiple case study and is carried out among four Turkish children born in the Netherlands. In order to vary the relative amount of Turkish and Dutch in the children's language input (and thus have a reflection in the group of informants of language input variation existing in the Turkish community in the Netherlands) two children were selected who visited a day care centre during five days a week; two other informants were being brought up at home. When data collection started, the children were 2;0 years of age. Data were collected by native speakers of the target languages; once every month a native speaker of Turkish collected data and once a month the same was done for Dutch by a native speaker of this language. Data collection continued until the children were aged 3;6. Most of the resulting base consists of spontaneous interactions between an informant and an adult (being the researcher, mother, caretaker or a visitor) and/or between the informant and other children (either siblings or playmates). The same procedure was followed for a monolingual reference group as well, consisting of two monolingual Dutch children living in the Netherlands and two monolingual Turkish girls living in Turkey 3 . The

126

Hanneke van der Heijden

reference group was matched with the core informants on several relevant variables such as the children's age, gender and their parents' educational background. The data base thus collected consists of 94 transcripts of sessions with Turkish as a target language, and 87 sessions with Dutch as a target language. Data were transcribed according to the Childes-format (MacWhinney, 1991) and stored as computer files. For the analyses presented in this chapter three sessions of four informants were selected, as is shown in Table 1. For every informant the whole period of data collection was covered by choosing one transcript from the beginning of the period of data collection, one from the middle and one from the end, resulting in a total of 9 transcripts which had Turkish as the target language and 9 which were aimed at eliciting Dutch. Table 1: Ages of the informants at three selected sessions of data recording with Dutch or Turkish being the target language of elicitation

Session 1 Informant

Dutch Turkish

Session 2 Dutch Turkish

Session 3 Dutch Turkish

Monolingual Dutch Astrid

2;2 -

2;11 -

3;6 -

2;1 2;2 2;1 2;2

2;9 2; 10 2; 10 2; 10

3;6 3;6 3;6 3;4

- 2;2

- 2; 10

- 3;6

Bilingual Turkish/Dutch Filiz Selma

Monolingual Turkish Jlknur

Concordances were run on the selected transcripts with the Oxford Concordance Program. From the resulting lists all derivations consisting of a nominal base and one or more bound morphemes were extracted. Nominal compounds were defined as those compositions which exist of two or more lexemes, at least one of them being a noun. Since the free lexemes which make up a compound may be written separately, compounds could not be detected from the concordances. Instead, all transcripts were searched for compounds by hand. In the analyses only those derivations and compounds were included which were used by the informant in the target language of the recording session.

Word formation processes in young bilingual children

127

Results Order of appearance of compounding and derivation In order to test the two hypotheses concerning the relative order of appearance of compounds and derivations in the Dutch and Turkish data base, an overview of the use of both devices by the informants is presented in Table 2. Table 2: Frequency distribution of compounds and derivations in types (tokens) in the data base Dutch Informant

Compounding Derivation

Monolingual Dutch Astrid session 1 session 2 session 3 Bilingual Turkish/Dutch Filiz session 1 session 2 session 3 Selma session 1 session 2 session 3 Monolingual Turkish Aknur session 1 session 2 session 3

Turkish Compounding Derivation

1 (1) 21 (34) 7(15) 25 (49) 20 (32) 32 (74)

3 (3)

3 (6) 1 (2) 1 (2)

1(7) 5 (5) 4(6)

4(19) 20(33) 10(19)

2 (7) 1 (2)

1 (1) 1 (2)

2(2) 1(1)

1 (1) 2 (5) 7 (8)

1 (2) 6 (6) 1 (1)

When the frequency distributions of compounds and derivations for every single session are compared the following observations can be made. First, in the Dutch data base derivations outnumber compounds by far. This is true for both the monolingual and the bilingual informants, the only exception being Filiz in session 3. Furthermore, these derivations are used

128

Hanneke van der Heijden

from an earlier time onwards than compounds are. In the Turkish sessions, the number of word formations is much lower than is the case for Dutch. Keeping this draw back in mind, it can be stated that in the Turkish data base compounds are used more frequently than derivations are. Moreover, these compounds emerge at an earlier point in time than derivations do. Again, this trend can be observed in the case of both monolingual and bilingual informants. Thus, the Turkish data base provides supporting evidence for the first hypothesis, stating that compounding precedes derivation in the acquisition of word formation devices, but the same principle does not hold for the Dutch data base. As far as the relative order of appearance of the word formation devices in the two languages is concerned, the data collected from the two bilingual informants Filiz and Selma show that Dutch derivations appear earlier than Turkish ones do. Especially in Selma's case there is a definite lag between the use of both devices: whereas she does not use any Turkish derivations whatsoever, the number of occurrences of Dutch derivations in her data is quite high. Derivational devices seem to be applied earlier in Dutch than in Turkish. Thus, the second hypothesis, claiming that derivational principles are applied in Turkish before they are applied in Dutch, is not supported by the data. In the following a qualitative analysis of forms and functions of the word formation devices will be presented. First, derivational devices will be studied, followed by an analysis of compounding devices. Forms andfiinctions of derivations In the Dutch data base the diminutive suffix accounts for nearly all derivations. In standard Dutch this suffix has five allomorphs: -je, -tje, -etje, -pje and -kje, which are selected according to phonological and prosodic characteristics of the noun the suffix is attached to. In addition, the diminutive has regional variants such as -ie, which is used in the western part of the Netherlands (Geerts et al., 1984:79-85). The variant -je is the most frequent one in the data base. The forms -pje and -kje were not used at all. Only in two cases non-standard variants of the suffixes were used, as is shown in the sequences under [3].

Word formation processes in young bilingual children [3]

129

Astrid (session 3) Astrid: standard:

jij kauwom/ye. kauwgomme//'e

'you chewing gum-DlM.'

Selma (session 1) Caretaker:

wil je nog brood?

xx4 un /kapje. Child: xx. Child: Interviewer: ja das guusje. guus. Child: Selma: fkappetje.

'do you want some more bread?' xx a crust-DlM.' XX.'

yes that's guusje.' guus.' crust-DlM.'

T a b l e 3 gives an overview of the types and frequencies of derivational suffixes used by the informants in Dutch and Turkish. Table 3:

Frequency distribution of types of nominal derivational suffixes in the data base (DIM stands for diminutive) Dutch

Informant Monolingual Dutch Astrid session 1 session 2

DIM DIM

-er session 3

DIM

-er -aar

Bilingual Turkish/Dutch Filiz session 1 session 2 session 3 Selma session 1 session 2 session 3

Turkish

34 48 1

64 4 6

-in

1

DIM

6 2 2

DIM DIM

DIM DIM DIM

19 33 19

-cigim -cik

1 2

130

Hanneke van der Heijden

Table 3 cont. Monolingual Turkish Ilknur session 1 session 2

session 3

-lik -lik -gil -o§ -gil

2 2 3 1 1

It might be suggested that the diminutive forms turning up in the data base are simple root forms, not reflecting any productive process on the learner's part. Counter evidence for this is presented in [3] above and in [4]. In [3] Selma uses the non-standard form of kappetje although the correct form kapje was used only a few utterances before by another child. In [4] Selma plays with the lexeme konijn 'rabbit' both in its inflected and uninflected form. [4]

Selma (session 2, talking to herself) ' there baby-DIM baby-DIM babydaar babietje babietje babietje. DIM.' die is niet # wak [= wakker] hoor. 'that's not # awake you know.' dieze slaapt. 'this one sleeps.' ma bed. 'my bed.' wakker? 'awake?' st is niet slaapt [=? slaap] he? 'sht is not sleep(s) is it?' ## konijnty'e konijn konijn konijn '## rabbit-DlM rabbit rabbit rabbit konijn. rabbit.'

As regards the function of the Dutch diminutive, this suffix is in adult language not only used to express small size but can also fulfil an emotive function, such as indicating affectivity (Geerts et al., 1984:84-85). In the data base examples of both functions are present. An illustration is given in [5].

Word formation processes in young bilingual [5]

children

131

Selma (session 3) Interviewer: Selma: Interviewer: Interviewer: Interviewer:

kijk wat heeft ze op dur hoofd? hoedje. un hoedje he? en hier naast haar? wat Staat hier nou?

Slema: taartje. Interviewer: ja un hele grote taart. Interviewer: gaat ze die helemaal alleen ope ten? Selma: nee. Interviewer: nee? Interviewer: wie gaat er dan meehelpen? Interviewer: ooh al die +/. Selma: kindjes. Interviewer: ja al die kindjes krijgen ook un stukje he? Selma: Interviewer:

ja. en wat hebben die kindjes dan meegenomen?

Selma:

kadoo tje.

'look, what does she have on her head?' 'hat-DiM: 'a hat-DIM, isn't it?' 'and here next to her?' 'what is standing over here?' 'cake-DIM.' 'yes a very big cake.' 'is she going to eat it all by herself?' 'no.' 'no?' 'who is going to help then?' 'ooh all those +/.' 'children-DlM.' 'yes all those children-DIM get a piece-DlM as well, don't they?' 'yes.' 'and what did these children-DIM bring with them?' 'present-DlM.'

Sequence [5] above shows a large amount of diminutive forms. From the use of the inflected noun taartje 'cake' we can derive that these diminutives are not only used for size but also for endearment: the interviewer clearly states that there is a big cake at issue. Supporting evidence is also found in cases where the adjective klein 'small, little' is used together with a noun in diminutive form; two of these cases are shown in [6]. [6]

Astrid (session 3) eerst even die kleine boekje. zo kleine oortjes.

'first that small book-DlM.' 'little ear-DIM-s like this.'

Although, when compared to other languages, the diminutive is in Dutch rather frequent anyway, this form is in particular abundantly used in baby talk. Thus the difference in frequency of occurrence between Astrid and Selma on the one hand, and Filiz on the other (see Table 3) can be interpreted as a reflection of language input. Astrid and Selma both attend a

132

Hanneke van der Heijden

day care centre, where there is a large amount of Dutch baby talk, whereas Filiz is brought up mostly at home by her Turkish mother. As derivation in Dutch is concerned, Astrid is the only one of the three informants who also uses other inflections than the diminutive alone, although their number of occurrences is marginal (see [7]). The noun veger 'broom' is derived from the verb stem veeg 'to sweep' with an agentive suffix. [7]

Astrid (session 3) kijk dit is un veger.

Hook, this is a broom.'

Astrid (session 3)

en alletwee vriendinnetjes.

'and both

ofthemfriend-FEM-DlM-s.'

Neither Filiz nor Selma uses derivations including one of these suffixes. Nevertheless, Selma may be aware of the morphological value of the agentive suffix -er (cf. wak for wakker 'awake' in [4] above and tijger/tijg 'tiger' in sequence [8]). [8]

Selma (session 3)

Selma:

hallo:::!

Selma:

Ik ben un tijger.

Child:

wat?

Selma:

ik ben un tijg.

'hello:;:/' 7am a tiger.' 'what?' Ί am a tig.'

Turkish disposes of a diminutive suffix as well, but this hardly occurs in the data base. Again this finding reflects language input, since in Turkish the diminutive is not very frequently used. The suffix -o§, however, placed after proper names to indicate endearment, can be said to share this function with the Dutch diminutive. The suffix -gil, a regional suffix used after relative names to denote 'the house or family of...' (Lewis, 1985:65) is used several times, but only in the monolingual informant's data base. Although in Boeschoten's (1990) study among Turkish 4-6 year olds in the Netherlands this word formation suffix proved to be one of the most frequently used, not one instance of it was found in the data base of our bilinguals. The suffix lik can be attached to such different word classes as nouns, adjectives, numerals and adverbs, and can accordingly fulfil various functions, such as adding a meaning of 'intended for or suitable for ...' (Lewis, 1984:62-64). It is especially in this sense that Ilknur uses the -lik suffix, for example in the derivation gecelik (night-lik) 'nightgown'.

Word formation processes in young bilingual children Forms and functions

of

133

compounds

Table 4 gives an overview of Dutch and Turkish compounds in the data base. According to the word class to which the separate lexemes of the compound belong, three types are distinguished: noun-noun, verb-noun and adjective-noun compounds. Table 4:

Frequency distribution of compound types in the data base (NN = nounnoun, VN = verb-noun, AN = adjective-noun compound)

Turkish

Dutch

Informant Monolingual Dutch Astrid

session 1 session 2

session 3

NN NN VN AN NN VN

1 10 3 2 28 4

Bilingual Turkish/Dutch Filiz

session 1 session 2 session 3

NN NN

9 2

NN

2

NN

1

NN NN AN NN VN

1 4 1 6 2

Selma

session 1 session 2 session 3

NN NN VN NN VN

7 4 1 4 2

Monolingual Turkish Aknur

session 1 session 2 session 3

134

Hanneke van der Heijden

One example of each of these types is given in [9], [9] NN

Dutch fotoboek

photograph book 'book of photographs'

VN

zwembad

swim pool

'swimming pool'

AN

roodkapje

red cap-DiM

'Little Red Riding Hood'

Selma, (session 2) Filiz, (session 3) Astrid, (session 2)

Turkish birth day

'birthday'

NN

dogum gtinti

VN

dolma yemegi filling food

'stuffed food'

AN

yalniyak

'barefoot'

bare foot

Filiz, (session 2) Ilknur, (session 3) Ilknur, (session 2)

Both in the Dutch and the Turkish data base noun-noun compounds are the most frequently used type. With the exception of Filiz' Dutch data base, this type is at the same time the first to appear in the data base of each individual informant. Broeder et al. (1994) also found noun-noun compounding by far the most productive word-formation type in L2 Dutch, especially in the case of the Turkish learners. All compounds in the data base, both Turkish and Dutch, are of the modifier-head type. Apart from few realisations of babaanne (father mother) 'paternal grandmother', linearly structured compounds do not occur in the data at all. In standard Dutch the two parts of a modifier-head compound may be joined without the use of any binding phoneme, or by including either sjwa (spelled as -e-) or -s- as a binding vocal. In the Dutch data base, in by far the most cases of noun-noun compounds no binding vocal was realised, mostly resulting in compounds matching the standard adult form (see Table 5). The -s phoneme was never used as a binding vocal.

Word formation processes in young bilingual children Table 5:

135

Overview of the use of binding vocals in compounds in the Dutch data (* indicates a case which deviates from the target language norm) Binding phoneme zero

Monolingual Dutch Astrid session 1 session 2 session 3

1 14 27 (+1*)

Bilingual Turkish/Dutch Filiz session 1 session 2 session 3 Selma session 1 session 2 session 3

-e(n)-

1 4

-

-

-

-

1 7 4 6

-s-

1*

-

1 -

Only two non-standard forms were found. In one case no binding vocal was realised whereas a binding sjwa is the standard form. In the other case a binding vocal was used while the standard Dutch form has a zero-marker. Both instances are presented in [10]. These occurrences can be taken to be evidence for the claim that at least not all compounds are realised as unanalysed lexical units. [ 10]

Astrid (session 3) Interviewer:

nee kijk wat is dit? wat heb ik hier getoverd?

Astrid: Interviewer: [...] Interviewer: Interviewer: Astrid:

puntslijper. en dit nou?

'no look what is this?' ' what did I make appear here?' 'pencil sharpener.' 'and this

heb ik dat niet getoverd? jawel hoor. un punteslijper.

'didn 't I make appear this ?' 'oh I did: 'a pencil sharpener.'

136

Hanneke van der Heijden

Filiz (session 3) Filiz: ### [?] # glijnebaanl

'### [?] # slider

In Turkish two nouns together denoting one (unspecific) concept can be combined either by using a possessive suffix on the last noun or by simple juxtaposition, examples of which were shown in [1] above. Table 6:

Overview of compounding devices in Turkish compounds (* indicates a case which deviates from the target language norm) Compounding type

Informant

NN

NN-poss

Monolingual Turkish ilknur session 1 session 2 session 3

1 2(+l*)

Bilingual Turkish/Dutch Filiz session 1 session 2 session 3

2*

9

1

2

Selma session 1 session 2 session 3

4 3

Both types were used by all three informants with a preference for the N-N-poss type; one example of each construction is given in [11]. [11]

Filiz (session 3) Interviewer: §imdi ben an kovanmi bulacam. Filiz: an kovan # an kovan Ilknur (session 2) Ilknur: §imdi sana gorba yemegi.

'now I'll look for the bee hive 'the bee hive, bee hive.'

'and now a soup meal for you.'

Word formation processes in young bilingual children

137

Filiz was the only one to produce an N-N construction deviating from the target language norm, in that no possessive suffix was realised (see [11]). This case is especially remarkable because the standard form with a possessive suffix was used immediately before Filiz' utterance. Boeschoten (1990:71) reports some of these underrepresentations of the possessive suffix as well, but also in his data occurrences are infrequent.

Conclusions and discussion Quantitative and qualitative analyses of the data base on word formation devices taking nouns as a base bring to light both language system intrinsic developments and evidence for an influence of language input. As language input is concerned, in the Dutch data diminutive nouns were much more frequent in the data base of the two informants attending a day care centre than in the data of Filiz, who was brought up at home by her mother. On the other hand, it is striking that lexemes which are used particularly in their diminutive form in adult language, were by Filiz only realised as bare nouns. Thus, in Dutch the lexeme ijs 'ice cream' is a material noun whereas its diminutive form ijsje only refers to a portion of ice cream. Nearly the same can be said for the pair of koek and koekje, the first lexeme being a material noun, the second denoting a single biscuit (although koek may sometimes be used as an object name for large sized biscuits). Filiz, however, only used the bare word forms ijs and koek in the object name meaning; their diminutive forms did not occur in her data base. Two explanations can be put forward to account for this phenomenon. First, it may be thought of as a reflection of language input characteristics of Dutch in the Turkish community. For this explanation no evidence is available yet. Second, it may be the result of an analysis by the informant of this type of derivations in stem and bound morphemes, in the same way as Selma both uses konijn and konijntje (see [4] above). The latter interpretation would at the same time be evidence against those who state that derivations and compounds realised by the informants are only the result of rote-learning, and should certainly not taken to be an indication that the informant acquired the underlying word formation device. Evidence supporting the view that informants are in fact engaged in analysing those complex words comes from new formations deviant from the target language norm as in kappetje, puntslijper and an kovan (see [3], [10] and [11] above). Moreover, it can be argued that, even if several derivations and compounds are at first used as the result of rote-learning, they will in a later

138

Hanneke van der Heijden

stage push the language learner to analyse and deduce the device lying underneath. Without any instance showing its appearance, the device itself certainly has no chance of making itself clear to the language learner. With respect to the acquisition of word formation devices in Turkish and Dutch the data show us tendencies in both the rate and in the structure of acquisition. The rate of acquisition in each of the bilingual's languages is slower than the monolinguals' rate. At age 3;6 Dutch monolingual Astrid uses word formation suffixes on nouns other than the diminutive alone and she is also able to extend the range of applications of the derivation and compound devices (as illustrated in [12]). She uses two derivational suffixes on one noun (vriendinnetje), or a derivational suffix on a compound (kikkerbadje), and creates a compound with a diminutive noun as a modifier (bloemetjesdeken). None of these types of applications were found in the bilinguals' data base. [12]

vriendinnetje kikkerbadje bloemetjesdeken

friend-FEM-DiM frog pool-DlM flower-DiM-PLUR blanket

'girl friend-DIM' 'paddling pool' 'flower-patterned blanket'

A similar observation can be made for the Turkish data base. Whereas the bilingual informants Selma and Filiz hardly use any derivational suffixes in Turkish at all, the monolingual Ilknur has several different suffixes at her disposal. Regarding the structure of acquisition in this domain, two hypotheses were tested. The first one stated that compounding precedes derivation in the acquisition of word formation devices. The Turkish data base supported this hypothesis, but it did not match with findings from the Dutch data base. In Dutch, both the monolingual and the bilingual informants first realised derivations; compounds only turned up in a later stage. The second hypothesis, stating that derivational principles are applied in Turkish before they are applied in Dutch did not hold for the data base under consideration. In both a quantitative and qualitative comparison of the two bilingual informants on the one hand and the monolingual informants on the other, the bilinguals show developments in their respective languages which are parallel to the developments the monolingual informants of these languages show. The bilinguals' preference in Dutch for derivation over compounding conformed to the monolingual Dutch informant's. In the same way, both the monolingual Turkish and the bilingual informants more often used compounding than derivation in Turkish. These parallels between bilinguals

Word formation processes in young bilingual children

139

and monolinguals were also found in their devices for binding in compounds. Since developmental preferences in Dutch are different from those in Turkish, this finding implies that the bilingual child extends her word formation repertoire along different lines in its two respective languages, both lines conforming to the monolingual way of doing the job. It should be recalled that the analyses presented in this chapter only concern a very limited domain of lexical acquisition. A study of the acquisition of word formation devices regarding verbal roots, in which Turkish is particularly productive, might yield interesting results. Also, a concluding example from Selma's data base (see [13]) shows us that even if both language systems are developing along separate lines, this does not exclude the existence of a mutual influence between the two languages in idiosyncratic lexemes. [13]

Selma (session 3) ha: oyuncak bathes i mi? standard: ijocuk bah£esi innovation: oyuncak balvjesi Dutch: speeltuin

'ah (the)

playground?'

'child garden-POSS' 'toy garden-POSS' 'play garden'

The standard Turkish word for playground is gocuk bahgesi 'children's garden 1 in literal translation; oyuncak bahgesi on the other hand literally stands for 'toy garden'. The noun oyuncak is morphologically related to the verb oyna = 'to play'. Although Selma used the standard lexeme gocuk bahgesi in the second session, she replaced it by oyuncak bahgesi in the third session of recording. Selma's new formation can be conceived of as a translation of the Dutch equivalent term speeltuin 'play garden'. This compound is made up of a verbal root speel = 'to play' and a noun tuin 'garden', the only difference being that instead of a nominalized verb root which is used in the Dutch lexeme Selma uses a semantically related noun as the first lexeme of her innovative compound.

Notes 1. I would like to thank Peter Broeder (Tilburg University) for his thorough comments on earlier versions of this chapter. 2. This project was supported by the Linguistic Research Foundation (grant 300174-001), which is funded by the Netherlands Organisation for Scientific Research, NWO, and by the Research Group on Language and Minorities, Faculty of Arts, Tilburg University.

140

Hanneke van der Heijden

3. The Dutch data from the bilingual and monolingual informants were collected by myself. Turkish data from the bilingual informants in the Netherlands were collected and partly transcribed by Sibel Köklü, Turkish data in Turkey were collected and transcribed by Hatice Sofu, under the supervision of Prof. Ö. Ekmek^i (£ukurova Üniversitesi, Adana, Turkey), to whom I am all very grateful. All transcripts were finally checked by myself. In the fragments below taken from the transcripts, the person collecting the data is indicated as 'interviewer'. 4. In the Childes conventions (MacWhinney, 1991) 'xx' stands for an unintelligible fragment of speech, '#' stand for a pause, '+/.' stands for an interruption, '/' indicates emphasis on the following syllable.

References Boeschoten, H. (1990). Acquisition of Turkish by immigrant children: A multiple case study of Turkish children in the Netherlands aged 4 to 6. PhD thesis. Tilburg University, Tilburg. Wiesbaden: Otto Harrassowitz. Turkologica 6. Broeder, P., G. Extra and R. van Hout (1994). First language influence on second language acquisition: Turkish and Moroccan adults in the Netherlands. In: G. Extra and L. Verhoeven (Eds.), The cross-linguistic study of bilingual development. Amsterdam: Royal Netherlands Academy of Arts and Sciences. Clark, E. (1981). Lexical innovations: how children learn to create new words. In: W. Deutsch (Ed.), The child's construction of language. London: Academic Press. Clark, E. (1983). Convention and contrast in acquiring the lexicon. In: Th. Seiler and W. Wannenmacher (Eds.), Concept development and the development of word meaning. Berlin: Springer Verlag. Donaldson, B. (1983). A linguistic history of Holland and Belgium. Leiden: Nijhoff. Geerts, G., W. Haeseryn, J. de Rooij and M. van den Toorn (1984). Algemene Nederlandse Spraakkunst. Groningen: Wolters-Noordhoff. Lewis, B. (1985). Turkish grammar. Oxford: Clarendon. MacWhinney, B. (1991). The Childes Project: Tools for analyzing talk. Hillsdale: Erlbaum.

Part 3 Bilingual development at school age

Cohesive devices in bilingual development Jeroen Aarssen and Petra Bos

The development of cohesive devices in narratives has become a major theme in language acquisition research over the last twenty years. Main topic of study were not single, isolated sentences, but longer, connected pieces of discourse (e.g., Karmiloff-Smith, 1979, 1986; Bamberg, 1987; Hickmann, 1991; Berman and Slobin, 1994). Many studies focused on one particular type of discourse, i.e., narratives based on picture story retellings. Cohesive means are grammatical devices (e.g., relativisation, subordination, ellipsis), lexical devices (e.g., adverbs, particles, prefixed verbs) and rhetorical strategies (e.g., questions, repetitions) used by a speaker for organising a number of propositions in order to express a connection (e.g., temporal or causal) between these propositions. The use of a range of expressions for reference to entities (persons or objects) in discourse is another form of cohesion. The major part of studies on development of cohesion involved first language acquisition. Only a few studies took the acquisition of cohesive devices in a second language into account. Studies which treated the simultaneous acquisition of two languages are even more exceptional. In this chapter we describe the design and outcomes of a research project on the process of bilingual language acquisition of Turkish and Moroccan children in the Netherlands. Two groups of bilingual children are involved: Turkish-Dutch bilinguals and Moroccan Arabic-Dutch bilinguals. The basic research questions are as follows: • How are the grammatical systems in the first and second language of these bilingual children in the age range of 4 to 10 extended over time? • How do children in this age range acquire cohesive devices in LI and L2, that is learn to comprehend and use these devices in order to produce fully cohesive narratives? • Under what conditions and with what grammatical rules do processes of transfer occur? In fact, the focus is to study the development of cohesive devices in three typologically unrelated languages, i.e., Turkish, Moroccan Arabic and Dutch. All three languages come from different language families. Dutch is an Indo-European language, whereas Turkish belongs to the Turkic

144

Jeroen Aarssen and Petra Bos

language family. Moroccan Arabic is a language from the Semitic branch of the Afro-Asiatic language family. As many theories make claims about universal properties of languages, it is interesting to describe features of three typologically distinct languages. The research condition becomes even more interesting when two unrelated languages are acquired by the same speaker or group of speakers. The assumption is made that linguistic proficiency can be operationalised in terms of grammatical and pragmatic proficiency. Therefore, development of both syntax and discourse are investigated within the research project. On the level of discourse, the acquisition of linguistic devices in bilingual children is investigated by means of analysing the production of connected discourse, and more specifically narratives. The assumption is made that children get more competent over time in telling stories, that is they get less dependent from the non-linguistic context. The linguistic forms they use in narrative discourse will, in this respect, become more sophisticated over time. Two conceptual domains are taken into account in the analyses of the narrative data: the acquisition of reference to characters or topic continuity and the acquisition of temporality. The linguistic forms that are used to express the concepts of topic and time have an important function in the construction of narratives. There is no narrative without characters or without some reference to a time axis. On the level of syntax, comprehension of sentence-internal anaphors and relative clause constructions is investigated. For both domains, experimental tasks were designed. For the analysis of the acquisition of sentence-internal anaphora the Government and Binding theory is used as a framework. For the acquisition of relativisation, the role of both underlying factors and surface structure determinants are investigated. Here we will restrict ourselves to presenting the experimental data. Analyses of narratives will not be presented (see, however, Aarssen, 1996 and Bos, 1997).

Informants Informants in the project are 50 Turkish and 50 Moroccan children living in the Netherlands. The age range of these bilingual informants is from 4 (from the first months of school contact onwards) until age 10. The design of the study is pseudo-longitudinal, i.e., two consecutive cohorts of 25 informants have been followed during a longer period of time: the younger cohort (T1 and M l ) in four rounds (from age 4 to age 7) and the older

Cohesive devices in bilingual development

145

cohort (T2 and M2) in three rounds (from 8 to 10). An overview of data collected is presented in Table 1. Table 1: Numbers of bilingual informants and monolingual control group informants participating in the project Informants

Language(s)

Ν

Data collected at age

Turkish-Dutch bilinguals cohort T1 Turkish and Dutch cohort T2 Turkish and Dutch

25 25

4

Moroccan Arabic-Dutch bilinguals cohort Ml Mor. Ar. and Dutch cohort M2 Mor. Ar. and Dutch

25 25

4

Dutch monolinguals cohort D1 Dutch cohort D2 Dutch

25 25

4

Turkish monolinguals cohort T3 Turkish cohort T4 Turkish cohort T5 Turkish

25 25 25

5

Moroccan monolinguals cohort M3 Moroccan Arabic cohort M4 Moroccan Arabic cohort M5 Moroccan Arabic

25 25 25

5

5

5

5

6

6

6

7 8

9

10

8

9

10

8

9

10

7

7

7 9

7 9

The interval between subsequent periods of data collection was one year. The subsequent age groups of both cohorts taken together correspond to the first 6 (out of 8) years of the Dutch primary school system (in which kindergarten is included). There are several control groups of monolingual informants as points of reference. The first control group consists of 50 monolingual Dutch children. Again, data were collected pseudolongitudinally, in a design corresponding to the one used with the bilingual informants. Furthermore, control data were gathered in Turkey and Morocco from three different age groups (5, 7 and 9) of monolingual speakers of Turkish and Moroccan Arabic respectively. All bilingual informants were born in the Netherlands or immigrated at a young age (well before the beginning of school). Their parents originate from small cities and rural areas in Turkey and Morocco. Most families

146

Jeroen Aarssen and Petra Bos

can be regarded as to have a low socio-economic status. The monolingual informants have a comparable SES.

Experiment 1: Anaphoric reference The first experiment is concerned with anaphor resolution. The experiment aims at investigating differences in the development of reflexives and pronouns. In the linguistic domain of anaphoric reference, a distinction is commonly made between reflexive or 'bound' anaphors such as 'himself', and pronominal or 'free' anaphors such as 'him'. Within the framework of UG, the Binding Theory (Chomsky, 1981) states the conditions under which possible antecedents for both types of anaphors can be determined. Principle A of the Binding Theory concerns bound anaphors, such as reflexives. It states that bound anaphors are bound within their governing category (= co-indexed with a c-commanding argument). Reflexives must refer to an entity within the same sentence. Compare the following sentences: [1]

The friendj of John washes himself,

[2]

*The friendi of John washes himselfj

[1] is grammatical, while [2] is ill-formed: himself refers to the friend and not to John. On the other hand, Principle Β concerns free anaphors, such as non-reflexive pronouns. It states that free anaphors must be free in their governing category. They cannot have a formal relation through coindexing with a c-commanding argument within the same governing category. This means that [3] can be a grammatical structure, while [4] is ill-formed: [3]

The friend of Joh^ washes himj

[4]

*The friend of Johnj washes Μιη

A correct understanding of both types of anaphors involves different kinds of knowledge. In the case of reflexives, a language user needs to have knowledge of a deterministic linguistic rule (expressed by Principle A) for the correct resolution of the anaphor. On the other hand, the linguistic knowledge expressed in Principle B, provides only a negative rule for the resolution of pronouns. It merely limits the choice of potential antecedents.

Cohesive devices in bilingual development

147

In the friend of John washes him, the correct antecedent for the anaphor him cannot be the friend. A possible antecedent might be John, but in a context as in [5] him most probably refers to the NP a little baby boy in the previous sentence. [5]

Mary and John have a little baby boy. The friend of John washes Ιιίιη.

On the basis of this difference in knowledge needed for correct resolution, predictions have been made about the order and rate of acquisition of bound and free anaphors in child language acquisition. As soon as children learn to master the linguistic rule expressed in Principle A, a rapid increase of correct interpretations of sentences with reflexives is expected (Deutsch and Köster, 1982; Deutsch, Koster and Koster, 1986; Koster, 1993). Bound anaphor resolution is primarily based on syntactic knowledge: there is only one possible antecedent. On the other hand, the relation between a free anaphor and its antecedent is not deterministic. It will therefore be more difficult to find a possible antecedent. In fact, a number of experiments on LI acquisition confirm these predictions (Wexler and Chien, 1985; Deutsch et al„ 1986; Solan, 1987). The use of experimental tasks for examining the development of anaphoric reference has turned out to be effective in earlier studies. For the study of anaphors in the three languages under consideration, i.e., Turkish, Moroccan Arabic and Dutch, three versions of the same anaphoric reference task were designed. It is a one sentence/four pictures multiple choice task, consisting of 24 short sentences in each language. Three factors were varied systematically in the test sentences: [1] type of anaphor (reflexive versus pronoun, see Table 2), [2] actions expressed in the verb (Table 3) and [3] antecedents. Table 2: Bound and free anaphors used in the anaphoric reference task Anaphor

Turkish

Moroccan Arabic

Dutch

English

bound/reflexive

a) kendini b) suffix -(t)n-

ras-u

zieh

himself

free/pronoun

onu

suffix -u

hem

him

Each sentence contained either a bound or a free anaphor. In Dutch, the second language of the bilingual children, the reflexive lexeme zieh 'himself and the personal pronoun hem 'him' are used for these purposes

148

Jeroen Aarssen and Petra Bos

respectively. Turkish uses the pronoun onu (him/her) and has two linguistic devices for reflexivisation. The first device is based on the use of a verb combined with the free morpheme kendini (himself) in direct object position, quite similar to 'himself in English and 'zieh' in Dutch. The second device is the reflexive suffix -(J)n- added to the verb stem. In reflexive test sentences, both possibilities are equally distributed. In Moroccan Arabic, the noun ras (literally: 'head') plus the possessive suffix indicates a reflexive action. In the case of third person singular masculine the form becomes ras-u. The suffix for direct object third person singular masculine, similar in meaning to the English personal pronoun him, is -u. In every sentence one of the selected verbs in Table 3 expressed the action. These verbs were chosen because any of these verbs a) is transitive, b) can have an animate direct object, and c) can be made reflexive in Dutch, Moroccan Arabic and Turkish. All actions were performed by two characters: two friends named Martijn (a Dutch boy's name) and Kerim/ Karim (a Turkish/Arabic boy's name). Table 3: Set of stimulus verbs used in the anaphoric reference task Turkish

Moroccan Arabic

Dutch

English

yikamak/yikanmak ka§imak/ka§inmak korumak/korunmak (jimdiklemek baglamak kurtarmak

ka-yegsel ka-yxebbes ka-ydafeccla ka-yeqres ka-yerbet ka-yfekk

wassen krabben verdedigen knijpen vastbinden bevrijden

to wash to scratch to defend to pinch to tie up to release

Before the experiment started, the characters (the two friends Martijn and Kerim) were introduced and described according to outward appearance (Martijn was fair-haired, while Kerim was black-haired; Martijn wore a striped shirt, Kerim a white one). The children had to learn who was who, before the first stimulus item was read. Each test sentence contained two possible antecedents in one and the same complex NP: a proper noun {Martijn or Kerim) and the relationship term the friend of. [6] and [7] are examples of test sentences.

Cohesive devices in bilingual development

[6]

Turkish: Moroccan Arabic: Dutch: English equivalent:

Kerim'in arkada§i kendini (jimdikliyor saheb Karim ka-yeqres ras-u de vriend van Kerim knijpt zieh the friend of Kerim pinches himself

[7]

Turkish: Moroccan Arabic: Dutch: English equivalent:

Kerim'in arkada§i onu ^imdikliyor saheb Karim ka-yqers-u de vriend van Kerim knijpt hem the friend of Kerim pinches him

149

The test sentences were read out loud, while 4 pictures were shown at the same time. Only one picture matched the situation described in the test sentence. The child was asked to point to the picture that matched the sentence (one possibility). The other three pictures showed actions that differed systematically from the 'right' picture: one picture showed the right actor but the wrong action (anaphoric error), another one showed the wrong actor but the right action (antecedent error), and the third one showed both the wrong actor and the wrong action (double error). Results Mean correct scores on the first experiment were computed. In the next paragraphs these scores are presented per subgroup of informants. Turkish bilinguals Table 4 presents the mean correct scores in percentages in both languages of the Turkish-Dutch bilingual informants. In Turkish, the bilingual informants generally obtain higher scores on items with pronouns than on items with reflexives. For the children in cohort T1 (from age 4 up to age 7) the difference is significant (F(l,23)=58.39, pc.001).

150

Jeroen Aarssen and Petra Bos

Table 4:

Mean correct scores (percentages) of bilingual informants on items with reflexive or pronominal anaphors in Turkish and Dutch

Turkish anaphor

Dutch

reflexive

pronoun

reflexive

pronoun

age

4 5 6 7

27 30 33 34

53 44 45 46

21 20 21 25

29 35 43 46

age

8 9 10

74 71 87

74 75 88

50 76 91

52 80 89

Within cohort 1 (from age 4 to 7) there is hardly any progress on both types of anaphors in Turkish. A big leap between cohort T1 and T2 in scores can be witnessed. This particularly applies to bound anaphor acquisition. Within cohort T2 (from age 8 to 10) there is a significant development over time (F(2,48)= 11.40, pc.001). Bound and free anaphors develop in the same rate. The Turkish bilingual informants show roughly the same pattern of anaphor acquisition in Dutch as in Turkish: in all groups up to age 7 (= cohort Tl) they perform significantly better on items with a free anaphor (F(l,24)=31.29, pc.001). From age 8 (= cohort T2) scores seem to run parallel. No significant difference between free and bound anaphors is found. In general, the Turkish children obtain higher scores on the Turkish version of the experiment than on the Dutch version (F( 1,23)= 17.89, pc.001 for cohort T l and F(l,24)=4.83, pc.05 for cohort T2). The interaction between Time and Language in cohort T2 indicates that the development in one language (Dutch) is faster than in the other (Turkish). Moroccan

bilinguals

In Moroccan Arabic, the Moroccan Arabic bilingual informants first obtain higher scores on pronouns (until age 7) and at later ages perform better on reflexives. The major difference between age groups is between age 5 and 6, which may be a positive effect of emergent literacy. The same holds for Dutch: young Moroccan bilingual informants perform better on Dutch pronouns than on Dutch reflexives (F(l,24)=24.32, pc.001). After age 7

Cohesive devices in bilingual development

151

they improve their scores on both types and scores seem to run parallel. No significant difference between free and bound anaphors is found. Table 5: Mean correct scores (percentages) of bilingual informants on items with reflexive or pronominal anaphors in Moroccan Arabic and Dutch Moroccan Arabic anaphor

Dutch

reflexive

pronoun

reflexive

pronoun

age

4 5 6 7

27 29 52 52

32 38 51 59

16 20 32 42

35 40 38 49

age

8 9 10

72 80 94

65 63 82

71 88 93

70 76 89

The factor Language turns out to be significant for both age groups: the informants in the youngest age group perform best on Moroccan Arabic (F(l,24)=30.78, pc.001) and for the informants in the older age group this difference vanishes for items on bound anaphors and becomes the opposite for items with a free anaphor (significantly better scores in Dutch than in Moroccan Arabic: (F(l,24)=9.95, p {OS, 00}. This only holds for the sentences in Moroccan Arabic that are in SVO-order. For the

Cohesive devices in bilingual development

155

ones in ovs-order, the following applies: OS, 00 > {55, 50}. The situation for Turkish is more complicated. First of all, Turkish makes use of nonfinite verb forms in relative clauses which are available as cues for parsing the sentence. Since word order in Turkish is relatively free, it may not be a reliable cue for processing. Instead, Turkish sentences provide other clues, for instance case markings, for a correct resolution of subjects and objects of clauses. Therefore, it is difficult to foresee how the conjoined clause analysis would work for Turkish. Slobin's (1973) universal operating principle to avoid interruption or rearrangement of linguistic units has led to the Embeddedness Hypothesis (Kuno, 1974). It states that relativised clauses in an embedded position within the matrix sentence will be more difficult to comprehend than those at the beginning or the end of a sentence. For Dutch, this would mean 05, 0 0 > 55, so. For Turkish and Moroccan Arabic, the relative difficulty depends on word order. The predicted order in Turkish sov and SVO sentences is 55, 50 > 05, 00; in sentences with OSV and OVS word order it is OS, 00 > 55, 50. For Moroccan Arabic the following order of predicted difficulty can be derived: in SVO order 05, o o > 55, 50 and in OVS order 55, 50 > OS, 00.

According to Hakuta (1981), a combination of both surface structure and role determinants plays a decisive role in the processing of relative clauses. He claims that in languages where the head noun (HN) is on the left of the relative clause (RC), subject focus will be easier than object focus, whereas in languages where the head noun is on the right of the relative clause, object focus will be easier, all other things being equal: • •

If [HN] [RC], then ss, OS > so, oo (e.g., Dutch, Moroccan Arabic) If [RC] [HN], then so, oo > ss, os (e.g., Turkish)

Finally, the way in which a particular language makes use of grammatical markers as cues in processing, may be an important factor in the comprehension of relatives. For instance, Turkish has clear marking of accusative case on definite object nouns, whereas Dutch has no case marking. Following earlier acting-out experiments on relativisation, the current experiment takes into account four types of relative clause construction: SS, SO, OS and OO. The task sentences were constructed by attaching a relative clause to a main clause. Turkish relative clauses are prenominal. Two suffixes can be attached to the verb stem, in order to make non-finite verb forms. The first possibility is the suffix -An, indicating a present participle

156

Jeroen Aarssen and Petra Bos

(SS and OS sentences). The second one is the suffix -DIG-

for object

relative clauses (SO and 0 0 sentences). Four word order types are possible in Turkish: SOV (the unmarked order), SVO, OVS and OSV. Moroccan Arabic has postnominal relative clauses. Although the unmarked word order in complex sentences (such as relative clauses) in Moroccan Arabic is SVO, OVS word order is also a possibility. OVS word order, however, requires a prestated object and a coreferential pronoun suffixed to the verb (in the main clause). In relative clauses in Dutch, a postnominal pronoun precedes the verb, and there is one single word order in the main clause: SVO. The following factors were systematically varied in the experiment: •

role of the head noun in main clause: Subject or Object;

•

role of the head noun in relative clause: Subject or Object;

•

word order in main clause: SOV, SVO, OVS or OSV in Turkish, and SVO or OVS in Moroccan Arabic.

T h e actors and undergoers in the test sentences were toy animals (a monkey, lion, bear, dog, cat and mouse). The possible actions were represented by simple transitive clauses, using the following set of verbs: the Dutch, Turkish and Moroccan Arabic equivalents of 'to stroke', 'to hit', 'to pinch' and 'to kiss'. The experimenter read out the test sentences one by one. Subsequently, the subjects were asked to act out these sentences, using the toy animals. Each acting-out response was noted on a score form; of each action acted out by the informants, the actor and the undergoer were registered. Special reference should be made here to Dutch SO and OO sentences, for they are at risk to be ambiguous. For instance [12] has both an SS and SO interpretation, although the unmarked interpretation is SS. [12] main clause • relative clause • relative clause

de beer die de aap kust, aait de leeuw [ s [ 5 [ Ο

ν Ο S

ν ] ν ]

ο

] SS interpretation SS interpretation

'the bear that the monkey KISS-3-sg STR0KE-3sg the lion' Test sentences of this kind were not used. In order to include unambiguous SS and SO sentences, the factor number agreement was introduced. In this way there can be no confusion on actors and undergoers in sentences [13] and [14]:

Cohesive devices in bilingual development [13]

SS

de beren die de aap küssen, aaien de leeuw 'the bears that kiss the monkey, stroke the lion'

[14]

SO

de beren die de aap kust, aaien de leeuw 'the bears that the monkey kisses, stroke the lion'

157

Results

Mean correct scores on the relative clause experiment were computed. In the next paragraphs, scores are presented per sentence type and per subgroup of informants. Turkish

bilinguals

For the youngest group of Turkish bilingual children (age 4-7, see Table 8) no clear pattern can be distinguished, neither in the Turkish nor in the Dutch data. Scores on all four sentence types in Turkish stay low until age 6, whereas from age 5 on the scores on Dutch SS and o o sentences gradually improve. Both in Turkish (F(3,72)= 10.67, pc.001) and in Dutch (F(3,72)=14.87, p