First Language Attrition
Studies in Bilingualism (SiBil)
Editors Kees de Bot University of Groningen
Thom Huebner S...
44 downloads
1149 Views
2MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
First Language Attrition
Studies in Bilingualism (SiBil)
Editors Kees de Bot University of Groningen
Thom Huebner San José State University
Editorial Board Michael Clyne, University of Melbourne Kathryn Davis, University of Hawaii at Manoa Joshua Fishman, Yeshiva University François Grosjean, Université de Neuchâtel Wolfgang Klein, Max Planck Institut für Psycholinguistik Georges Lüdi, University of Basel Christina Bratt Paulston, University of Pittsburgh Suzanne Romaine, Merton College, Oxford Merrill Swain, Ontario Institute for Studies in Education Richard Tucker, Carnegie Mellon University
Volume 28 First Language Attrition: Interdisciplinary perspectives on methodological issues Edited by Monika S. Schmid, Barbara Köpke, Merel Keijzer and Lina Weilemar
First Language Attrition Interdisciplinary perspectives on methodological issues
Edited by
Monika S. Schmid Free University Amsterdam
Barbara Köpke University of Toulouse - Le Mirail
Merel Keijzer Free University Amsterdam
Lina Weilemar Free University Amsterdam
John Benjamins Publishing Company Amsterdam/Philadelphia
8
TM
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data International Conference on Language Attrition: Interdisciplinary Perspectives on Methodological Issues (2002 : Vrije Universiteit Amsterdam) First language attrition : interdisciplinary perspectives on methodological issues / edited by Monika S. Schmid … [et al.]. p. cm. (Studies in Bilingualism, issn 0928–1533 ; v. 28) Includes bibliographical references and index. 1. Language attrition--Congresses. 2. Bilingualism--Congresses. I. Schmid, Monika S. II. Title. III. Series. P40.5.L28 I58 2002 306.44’6-dc22 isbn 90 272 4139 2 (Eur.) / 1 58811 549 6 (US) (Hb; alk. paper)
2004053124
© 2004 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
For Bert Weltens
Table of contents
Acknowledgment Language attrition: The next phase Barbara Köpke and Monika S. Schmid
ix 1
Part I Theoretical models and methodological aspects L2 influence and L1 attrition in adult bilingualism Aneta Pavlenko A sociocultural approach to language attrition Antonio F. Jiménez Jiménez Perceived language dominance and language preference for emotional speech: The implications for attrition research Jean-Marc Dewaele The role of grammaticality judgments in investigating first language attrition: A cross-disciplinary perspective Evelyn P. Altenberg and Robert M. Vago
47 61
81
105
Part II Attrition in progress – observations and descriptions Issues in finding the appropriate methodology in language attrition research Kutlay Ya˘gmur Language contact and attrition: The spoken French of Israeli Francophones Miriam Ben-Rafael Is there a natural process of decay? A longitudinal study of language attrition Matthias Hutz In search of the lost language: The case of adopted Koreans in France Valérie A. G. Ventureyra and Christophe Pallier
133
165
189 207
Table of contents
Part III How the study of attrition can contribute to the understanding of language Attrition in L1 competence: The case of Turkish Ay¸se Gürel Methodological aspects of a generative-based attrition study Bede McCormack Convergent outcomes in L2 acquisition and L1 loss Silvina Montrul A modest proposal: Explaining language attrition in the context of contact linguistics Steven Gross No more reductions! – To the problem of evaluation of language attrition data Elena Schmitt Language attrition research: An annotated bibliography Monika S. Schmid A new blueprint for language attrition research Monika S. Schmid Author index Subject index
225 243 259
281
299
317 349
365 373
Acknowledgment
The present collection represents a rigorously reviewed selection of papers presented at the International Conference on First Language Attrition: Interdisciplinary Perspectives on Methodological Issues, held at the Vrije Universiteit Amsterdam, Aug. 24–26th 2002. The great success of this conference was made possible with the kind and generous financial help of the Dutch National Science Foundation NWO, the Dutch Royal Scientific Academy KNAW, the Dutch Society of Applied Linguistics Anéla, the Faculty of Arts of the Vrije Universiteit Amsterdam and the Laboratoire Jacques Lordat of the Université de Toulouse – Le Mirail. To ensure the success of a conference, financial support is indispensible, but at least as important is the help, enthusiasm and support of many people. We are particularly grateful to Marijke Blom of the office of the managing director of the Faculty of Arts of the VU, who volunteered her assistance at the conference desk; to Ingrid van de Vegte of the VU Congress Desk, who was endlessly kind, helpful and (above all) patient during the months leading up to the conference; to Yvonne Wesseling of the VU catering staff, who made sure that everyone was comfortable and well looked-after during the conference; to Sonja van der Oord of the financial administration of the Faculty of Arts of the VU, whose patience in the face of mountains of reimbursement forms can only be described as superhuman; and to Julian Orton and Paul Bossenbroek who not only helped with the complex infrastructure of getting copies of handouts, posters, abstract volumes and so forth ready when they were needed and of providing audiovisual equipment and making sure it worked, but also swept up the occasional conference organiser suffering from a nervous breakdown. We are furthermore grateful to the enthusiastic student helpers at the conference desk: Martine de Boer, Chrissy Hossea and Paula Jansen. In order to ensure the high scientific standard of both the conference and the present volume, we had to rely on the expertise and help of many colleagues within the international scientific community. Unfortunately the anonymous refereeing process of both abstracts and conference volume submissions makes it impossible for us to thank them by name here: they will know that they are meant, and how grateful we are to them.
Language attrition The next phase Barbara Köpke and Monika S. Schmid Université de Toulouse – Le Mirail / Vrije Universiteit Amsterdam
Introduction The study of first language attrition is currently entering its third decade. However, after twenty years of diligent investigation resulting in numerous theoretical and empirical papers the questions on this topic still greatly outweigh the answers. Findings from individual studies seem to indicate that it cannot even be said with any certainty whether a first language in which a certain level of proficiency has been reached can ever undergo significant attrition, let alone how or why it might. There are many reasons for this failure to arrive at a more integrated and explanatory framework. Some are theoretical, some methodological, and some are linked to communication or lack thereof between individual researchers. These were the problems that our conference, this volume, and subsequent work have been trying to address. A further problem is, of course, the extraordinary complexity and multi-facetedness of the phenomenon of language attrition. It is our belief, however, that this last issue cannot even begin to be tackled unless the first three have been solved. This chapter will attempt to identify and clarify these issues, in order to establish the starting ground from which the individual papers will proceed. The first section will give a chronological overview of how the field of language attrition has developed over the past two decades, try to identify the concerns and forces that have shaped it, and point out developments and changes. The second section introduces those extralinguistic factors that have been shown to play important roles in attrition. The third section presents an overview of theoretical frameworks within which language attrition studies have been
Barbara Köpke and Monika S. Schmid
conducted, while the fourth section focusses on questions of research design. The final section is an outline of the structure and contents of this volume.
.
A historical overview
The topic ‘language attrition’ seems to exercise a strange fascination. The question of whether an individual can really forget a first or second language once learned, and how and why this ‘loss’ might proceed, is an intriguing one. One could even go so far as to say that most people would intuitively describe themselves as ‘attriters’ – “I took Spanish in school for three years, but I’ve forgotten it all”. Especially fascinating are questions about the nature of memory and the possibility of reactivating a forgotten language system, e.g. under hypnosis. This fascination was also at the root of the beginning interest in language attrition in the early 1980s – Richard D. Lambert, one of the organisers of the ‘inaugural’ UPenn conference “The Loss of Language Skills”, and co-editor of the subsequent conference volume (Lambert & Freed 1982), describes his interest as “both personal and professional”. Having studied a variety of foreign languages, ranging from a number of standard Indo-European ones to Urdu, Bengali, Sanskrit and Marathi, and having achieved fluency in at least some of these, he found that his proficiency was decreasing through the course of the decades (Lambert p.c.).1 Such reports of a feeling, based on introspection, of erosion in a first, second or foreign language seem very frequent among language attrition researchers. On the professional level, too, interest in language attrition often seems to have been initiated by experiences with FL students who had devoted a substantial amount of time trying to achieve proficiency which they subsequently lost: To a considerable extent, the large amount of time spent learning an Indian language was wasted. And this problem was not limited to students of Indian languages, but was more widespread throughout the field of area studies. I had conducted two major national surveys of university-based language and area studies [...]. The survey showed that the problem of language skill loss was endemic to African, East European, Far East, Middle East, South Asian, and South East Asian studies as well. (Lambert, p.c.)
This observation, coupled with a paucity of published literature, motivated Lambert to organise a national stocktaking conference which “might turn up more information and perhaps help establish an area for future research in the field.” (p.c.) The study of language attrition was thus inaugurated in a very
Language attrition
teleological and conscious way, certainly a rather unusual procedure for a research topic. Unusual it might have been; effective it no doubt was. The Loss of Language Skills (Lambert & Freed 1982) was a benchmark publication that, two decades later, has lost none of its importance and relevance to current research. It not only covers the issue of (first and second) language loss from every conceivable perspective, taking into account issues such as language shift/death, pathological language loss, social and political factors, but it also contains, in a methodological section, the groundbreaking papers by Andersen, Oxford and Clark. Of these, Andersen’s paper, in particular, is worthy of far more consideration than it has, so far, received in language attrition research, since it provides an outline of a set of preliminary assumptions and testable hypotheses and thus draws up a blueprint for research. Both the conference and the publication thus took it upon themselves to provide an emerging field with theoretical and methodological frameworks as well as background information from neighbouring disciplines. At this point, the study of language loss research branched out across disciplines and countries. In the USA, the Language Skills Attrition Project (LSAP, see Ginsberg 1986) was founded in an attempt to compile an archive of foreign language attrition tests and data in the less commonly taught languages Chinese, Japanese and Arabic. The aim of this project was (...) to measure the language competency at the end of their training of a cohort of students in the AOLC’s [Advanced Overseas Language Centers, BK & MSS] in Cairo, Taipei, and Tokyo, then measuring their proficiency again at yearly intervals after their return to the United States. (Richard Lambert, p.c.)
The methodological problems involved in developing a roughly equivalent set of tests across these languages, investigating features that were assumed to be vulnerable to attrition, were immense. Unfortunately, very little has been published on how these tests were designed, and what the outcome of the pilot studies were (the only available reports are Lambert & Moore 1984, 1986). Lambert himself describes them as “not substantial enough to help much in the development of the new field of language skill attrition. Nor, unfortunately, were the students’ competencies measured after a lapse of time as originally planned.” (Richard Lambert, p.c.) Meanwhile, the Pennsylvania conference had inspired interest in Europe, and specifically in The Netherlands. A personal link developed between Richard Lambert and Theo van Els, leading to the organisation of a two-day workshop with participants from the US and The Netherlands at Nijmegen University in October 1982. A number of European projects were thus set
Barbara Köpke and Monika S. Schmid
in motion (van Els 1986: 3). The ‘Dutch contingent’ introduced several new concepts to this developing field. Researchers had begun to feel that the microlinguistic approach put forward so far was lacking in theoretical depth, and that language attrition had to be seen not merely as a linguistic but also as a cognitive process, and thus an integral component of language acquisition (Bert Weltens, p.c.). A further important consideration was that the research strategies and models which had, so far, been used in large-scale projects on language shift had to be incorporated into the typically small studies of individual loss. In other words, it was hypothesised that attrition could not be studied in isolation from its social context. It was via this angle of interest that questions about linguistic minorities, migrants, and the attitudes of the surrounding dominant society entered the field (Sjaak Kroon, p.c.). These efforts eventually led to the Language Loss Symposium held in Kerkrade (near Maastricht, NL) in 1986. The proceedings of this conference (Weltens, de Bot & van Els 1986) are discernibly different from Lambert & Freed. Where the earlier volume had contained a collection of papers that were largely if not exclusively theoretical-methodological, the 1986 volume presents only a few methodological papers – which, furthermore, are clearly the outcome of preliminary reflections on a very specific research project to be undertaken – augmented by a number of reports on pilot studies or work in progress. The 1986 volume furthermore attempts to span the gap between language shift, dialect death and second language attrition, contributions on individual first language attrition still being notably absent. These early years witnessed a mere handful of investigations of first language attrition which – in contrast to the studies that emerged from the dense network described above – were apparently conducted and published in relative isolation, such as Sharwood Smith (1983), Hiller-Foti (1985), Bettoni (1986) and Py (1986). The next big surge in publications took place in 1989, when two journals devoted a special issue to language attrition: Studies in Second Language Acquisition 11:2 (ed. by Andrew Cohen & Bert Weltens) and Review of Applied Linguistics (I.T.L.) 83–84, comprising papers presented at the Language Loss Symposium held at the World Congress of Applied Linguistics at the University of Sydney, 1988. Both collections witness a further shift in balance from exclusively theoretical to applied papers and towards L1 attrition in an L2 environment. These two volumes are a good indication of the type of research that was being conducted at this time. Most papers are short preliminary, inprogress, or concluding summaries of projects, pilot studies or PhD-studies (some of these projects, unfortunately, were abandoned before completion).
Language attrition
The next, and – thus far – last collection of papers that explicitly limits itself to language attrition, in this case even narrowed down to L1 attrition, is Seliger & Vago (1991). This is the first collection that was not the outcome of some kind of meeting of researchers. The editors – both of whom had also become interested in the topic through personal experience – contacted potential contributors directly (Robert Vago, p.c.), which is a testimony to the fact that the field was beginning to establish itself. Rather interestingly, although this volume is the first collection of papers to proudly carry the term ‘attrition’ in its title, few of the reported studies actually deal with this phenomenon in the strict meaning of the term – the non-pathological decrease in proficiency in a language that had previously been acquired by an individual, i.e. intragenerational loss. Apart from a number of theoretical papers, Seliger & Vago contains several papers on language shift and death, as well as one on aphasia. In addition to that, there are five case studies on the attrition of one (in one case two) subjects, and only three larger studies – with between 6 and 30 subjects – are reported on. This volume ushered in the nineties, a decade that witnessed the disciplines of L1 and L2 attrition drifting apart. L1 attrition, by and large, seemed to be content to be a subfield of the larger ‘Maintenance and Loss of Minority Languages’ area, on which a series of conferences were held in The Netherlands between 1988 and 1998 (see below). It is symptomatic that in 1998, Kees de Bot pointed out in exasperation – and doubtlessly not for the first time – that he and Bert Weltens were the two only ‘knights’ in pursuit of the shining goal of terminological distinction between Shift (intergenerational) and Attrition (intragenerational) as hyponyms of the more general Loss.2 L2 attrition, on the other hand, became more and more firmly established within the field of foreign language teaching. This division in ‘affiliation’ impacted very heavily on the methods of data elicitation, testing, and analysis used by the two fields, which eventually made comparisons between individual studies virtually impossible. The loss of close links was not, however, confined to the two branches of the discipline. The personal ties between most researchers working on the narrowly confined field of language attrition that had characterised and shaped the eighties also ‘attrited’. If we take Seliger & Vago to be the transitional point between the decades, it is very noticeable that hardly any of those names that had been so prolific in the eighties recur in the nineties (except, as e.g. in the case of Bert Weltens, as PhD supervisor – i.e. not themselves actively involved in research on the topic). The major exception is Kees de Bot, who can be de-
Barbara Köpke and Monika S. Schmid 30 25 20
PhD
15
Article
10 5 0 L1
L2
1982–1986
L1
L2
1987–1991
L1
L2
1992–1996
L1
L2
1997–2002
Figure 1. Types of studies published
scribed as the only constant in this quickly changing field – and the single force that held it together during the next decade. Where L1 attrition is concerned, the nineties were the decade of large but solitary studies. While the eighties had posed interesting questions, a great proliferation of short papers hoping to find quick and easy answers, towards the end of the decade a sobering effect had set in, as researchers realised that it might all be far more messy than they had previously thought. On the other hand, a large enough body of theoretical work was now available to make the topic attractive for PhD studies. Figure 1 illustrates the development of articles and PhD dissertations in relation to each other.3 It is evident how the past decade, in particular the past five years, has witnessed an astonishing increase in dissertations while over the same span of time, the number of articles published has remained relatively stable (many of these latter publications, however, are exclusively theoretical treatises in handbooks or encyclopaedias). Language attrition research was thus characterised by the solitary existence of the PhD-student, spending a large part of her time doing fieldwork and, typically, having little opportunity and less money to go to conferences and make contacts in the international scientific community. Unfortunately, many of these PhD-dissertations are not easily available,4 which means that, while much work – and good work – was being done on language attrition in the 90s, many researchers could not draw on each other’s work, and
Language attrition
the field lost most of the overall coherence and collective impetus it had begun to enjoy in the decade before. In addition to that, many researchers who were doing a PhD during this period abandoned the topic (if not the profession) after completing their dissertation. It is hard to say whether this ‘individualisation’ of language attrition research was the cause or the outcome of the fact that the frequency of conferences or workshops specifically dedicated to language attrition also decreased sharply within this period. The political climate, which made it possible to obtain very generous funding for activities on language loss, necessitated the link between research on loss of indigenous and non-indigenous languages – upon which the field experienced what can only be called media hype. The first International Conference on Maintenance and Loss of Minority Languages (de Bot & Fase 1991), organised by researchers from the University of Tilburg in Noordwijkerhout (The Netherlands) was reported on in national and international TV programs and newspapers in 1988 (Kroon, p.c.). This great public interest diminished over the next decade, during which two more very successful conferences were held (in 1992 and 1998, see Fase, Jaspaert & Kroon 1995; Klatter-Folmer & van Avermaet 2001; Ammerlaan, Hulsen, Strating & Ya˘gmur 2001). At all of these conferences, L1 attrition was also represented, but the focus on a diminishing proficiency across one lifespan that had commanded such interest in the decade before became fuzzier (see also the account of one of the organisers of the 3rd and thus far last such conference, Kutlay Ya˘gmur, this volume). It was only after the turn of the millennium that the field started to consolidate itself again. Workshops on L1 attrition were organised at international conferences – at the Second Language Research Forum 2000 in Madison, Wisconsin by Dorit Kaufman (see some of the papers in Bonch-Bruevich et al. 2001) and at the 3rd International Symposium on Bilingualism in Bristol, 2001, by Agnes Bolonyai. In this situation we felt that it would be beneficial to the field of L1 attrition studies to provide a platform for researchers to exchange their findings, reflect on methodological issues, and engage in an intensive discussion of theoretical and practical aspects of language attrition research, in order to try and re-establish something approximating the close network of the eighties. The following sections will attempt to clarify some of the issues that we found especially in need of attention in such a network.
Barbara Köpke and Monika S. Schmid
. Some basic issues in attrition research Researchers beginning work on language attrition often seem disconcerted that, contradicting all their intuitions on the ‘undoubted’ and unquestioned existence of the phenomenon we label attrition, twenty year’s work has not yet been able to turn up unambiguous answers. This, however, is hardly surprising. The question “Can people forget a language?” bears, at least in one respect, a close resemblance to the Ultimate Question of Life, the Universe and Everything invoked in Douglas Adams’ Hitchhiker’s Guide to the Galaxy. Adams describes a race of “hyperintelligent pandimensional beings” who build a supercomputer whose task is to calculate the answer to this Ultimate Question – a task which takes the computer 7.5 million years to accomplish. Eventually, it comes up with the answer “Forty-Two”. The problem, as Adams points out, was that it had never been precisely specified what the question was in the first place. In the same way, the question “Can a speaker forget a language?” is imprecise, and therefore unanswerable as it stands, in various ways: there is, as yet, no workable and testable definition of language attrition – or of what would constitute evidence for its presence. However, the growing body of work provides converging evidence for at least some of the assumptions pertaining to what attrition might be, and how and why it might happen. . Extralinguistic aspects Many attempts were made during the first and largely methodological decade to provide a precise terminological and taxonomical framework within which language attrition research could be conducted. The best-known (and most extensively quoted) of these is the division of the types of language attrition according to what language is lost (L1 or L2) and in what environment (L1 or L2) this loss takes place (first made by de Bot and Weltens 1985, but usually ascribed to van Els 1986). This split results in four possible types of attrition (see Fig. 2), of which two – reversion and dialect loss – have received relatively marginal attention.5 This is understandable, since including these two areas under ‘attrition’ causes a variety of problems of methodology, terminology, and data analysis. Language reversion is very difficult if not impossible to separate from the kinds of linguistic problems that are also encountered in the monolingual elderly (see Goral 2004). Furthermore, the informants under analysis in this type of L2 attrition most often are migrants who have never received systematic instruction in their L2. The problem of the point of reference – what was the L2
Language attrition Linguistic environment Language lost L1
L1 Dialect loss
L2 L1 attrition
L2
L2 attrition
Language reversion (in the elderly)
Figure 2. The ‘van Els taxonomy’
competence of the informant before psychosocial and/or neurolinguistic reasons connected with aging led to L2 attrition and reversion to the L1? – thus becomes impossible to deal with. The inclusion of dialect loss in a taxonomy of language attrition, too, causes many problems. Firstly, it has to be assumed that dialect loss typically is more similar to language shift than to attrition, i.e. it is something that happens inter- rather than intragenerationally due to incomplete acquisition (Kroon p.c.). Secondly, the distinction between an L1 and L2 environment seems, in this case, to hinge critically on the rather arbitrary and messy classification of whether two linguistic systems are two ‘different’ languages or ‘merely’ two different varieties of the same language. Nevertheless, the four-way taxonomy is invoked in virtually every study on attrition – perhaps because it is such a clear milestone in the development of the field, and such a (seemingly) neat and discrete way of carving it up, rather than because of its actual long-term impact in establishing subfields of ‘attrition’. Given the overwhelming focus of language attrition studies on L1 attrition in an L2 environment and L2 attrition in an L1 environment (as well as the fact that the significance of the very distinction between these two has recently been called into question, see de Bot 2002), we would propose, with respect, that this taxonomy may have outlived its usefulness. That notwithstanding, there are other methodological distinctions that are clearly necessary, if an overall picture of attrition is ever to be gained. These have not, so far, been clearly established or respected. The most obvious of these concerns the difference between language attrition in children – i.e. the attrition of a linguistic system which has not yet stabilised – vs. attrition in adult speakers. Clearly related to this is the distinction of attrition among simultaneous, early and late bilinguals. The results from previous studies indicate that it is necessary to take into account both age at the onset of bilingualism (see e.g. the findings in Montrul 2002) and age at the onset of attrition. There is converging evidence suggesting that an L1 system can indeed be eroded to a quite dramatic degree if the attrition process sets in well before puberty (e.g. Bode 1996; Isurin 2000a; Isurin 2000b; Kaufman & Aronoff 1991; Nicoladis & Grabois 2002;
Barbara Köpke and Monika S. Schmid
Schmitt 2001, this volume; Seliger 1989, 1991; Turian & Altenberg 1991; Vago 1991). Similar findings were obtained for L2 attrition among children (Berman & Olshtain 1983; Kuhberg 1992; Olshtain 1986). Preliminary results from a recent study even point towards the L1 being lost to the extent that psycho- and neurolinguistic methods can detect no trace of it any more (Ventureyra & Pallier, this volume), while no study researching attrition among speakers who were above the age of twelve when input from the L1 was reduced has found anything remotely similar to such a drastic picture. Where such speakers are concerned, the prevailing impression seems to be that the amount of attrition present in the data is usually surprisingly low, even after many decades spent in an L2 environment. For those few studies which looked at both types of subjects (e.g. Ammerlaan 1996; Pelc 2001), age turned out to be the most important factor for the prediction of attrition. It has been suggested that the younger the child is when the language of her environment changes, the faster and deeper she will attrite; yet neither of these studies allows one to specify a precise age limit up to which attrition is more likely to occur.6 Whether this difference is merely quantitative, or whether clear qualitative differences also exist between child and adult L1 attrition has yet to be established, until which time extreme caution should be advised in transferring general observations from one area to the other (see also below, Section 3.2). Less categorical, but none the less important, distinctions will also have to inform language attrition research to a far larger degree than they have previously done. Among these is the level of education – an aspect that has been largely neglected so far, probably partly because it is very difficult to define in a discrete way in a multilingual and multicultural context. A further methodological problem associated with this factor is how to determine whether what is influenced here is language attrition, or linguistic performance in general. In the first case, the education effect should be present only in the attriting group, whereas in the second case, similar effects would be observable in the control group as well, the groups with more education being better equipped to solve specific tasks. However, the fact that most studies which included education level among their extralinguistic variables determine variable effects across tasks suggests that the issue may be even more complex than that. Jaspaert & Kroon’s (1989) pilot study on attrition among 30 Italian immigrants in The Netherlands, measured by scores on various language tests (vocabulary, text editing, sentence correction, and general comprehension as measured by a picture-sentence matching task), for example, showed education level to be the most strongly significant extralinguistic factor in the tests where evidence for attrition was found, namely the text editing and vocabulary test. Education
Language attrition
level also influenced verbal fluency in the L1 in the findings reported in Waas (1996), while it was not a significant factor for her other analyses. However, Ya˘gmur (1997), who treats education level as one of the main independent variables, found that performance on the same verbal fluency task was influenced by education level in both attriters and controls, with less educated subjects achieving lower performance levels across groups. Similarly, his highly educated informants (both attriters and monolinguals) performed much better on his relativisation task, although a weak effect of education level on the attrition of this skill was also detected (Ya˘gmur 1997:91).7 This finding, suggesting that education level can influence the capacity to carry out a sentence generation task, was corroborated by the results of such a task in Köpke (1999), on which more educated subjects in both the experimental and the control group produced the target structure more often.8 Education level also influenced the judgement of ungrammatical L1 sentences among her bilingual subjects and proved to be one of the most important factors influencing attrition in this task, while on the other hand it had no effect on general discourse measures in a picture description task. A slightly different approach is chosen by Pelc (2001), who measures education level on the basis of number of years of L1 (Greek) education prior to emigration and years of education in the US.9 The results show that the number of years of education in Greece correlates with linguistic performance – a finding which is probably related to the age effect she reports and may indicate a complex relationship between age, literacy and attrition (see below, Section 3.2). On the whole, these findings indicate that the role of education level needs to be explored further, and certainly is an important factor which has to be controlled for, especially in attrition studies based on predominantly metalinguistic tasks (such as correction, editing, sentence generation, grammaticality judgments etc.). A third factor generally assumed to play an important role in attrition is the length of time since the onset of attrition (cf. also Hutz, this volume). Preliminary psycholinguistic studies suggest that immersion in a different linguistic environment leads to a change in language dominance, access to L1 gradually becoming slower than access to L2, without L2 competence necessarily being native-like. Empirical evidence suggests that such a reversal of language dominance may occur after a comparatively short period: answers in L2 were found to become faster than in L1 on a picture naming test after around 7 years (Lachman & Mistler-Lachman 1976; Mägiste 1979), while a lexical decision task produced a similar effect after as little as three years (Frenck-Mestre 1993). It has
Barbara Köpke and Monika S. Schmid
to be stressed that such a change in language dominance is not identical with attrition: differences in processing speed as measured in experimental tasks are hardly noticeable in normal speech situations, and the complexity of the relationship between two or more languages in the mind may well entail other consequences on language processing that are most likely to be fluctuating as a consequence of the bilingual’s language use of the moment. Nonetheless, even if a reversal in language dominance is not necessarily followed by attrition, it is most likely that attrition is preceded by such a reversal and depends on time in a similar way. However, attrition studies investigating subjects with a period of residence of more than ten years generally find no or very little time effect (Ammerlaan 1996:209; Bouba et al. 2002; Brown 2001:31; de Bot & Clyne 1994:17; Gürel 2002:170), and it has further been suggested that length of residence may only have an effect when there is little or no contact with the L1 (de Bot et al. 1991:94; Soesman 1997:190). Köpke (1999) did find an effect of immigration length with respect to grammaticality judgments and picture description, but notes that this effect might be due to differences in the immigration length of the two experimental groups (English vs. French L2). Whether this implies that the question of whether the subject will attrite or not is determined in the period between roughly 5 to 10 years after immigration (as has been suggested in e.g. de Bot & Clyne 1994: 27) will, however, need further investigation. For the moment, it appears that the role of immigration length for attrition is not as important as generally supposed. This is also what is suggested by research conducted with very long-term immigrants, indicating that the L1 system can remain surprisingly stable even after many decades in an L2 setting (e.g. Hutz this volume; Jaspaert & Kroon 1992; Schmid 2002). . Language attitudes, motivations, and use Far more crucial than the length of time appears to be the influence of language use and attitude. The latter feature in particular has proved one of the most slippery and difficult to measure factors in attrition research (see Schmid forthc. a). Attitudes, motivation and other affective factors have been shown to have a strong impact on language learning (e.g. Gardner & Lambert 1972) and, at least in some cases, possibly also on recovery patterns in polyglot aphasic patients (Minkowski 1928). Such factors have also been shown to play a role in multilingual community settings, so that it seems reasonable to assume that they might influence attrition in some way. The work of Pavlenko (2002) and Dewaele (this volume) further suggests that the bilingual’s different languages
Language attrition
may fulfil different emotional functions and that in some cases affective factors might influence the process of attrition. One of the earlier studies looking for a link was Waas (1996), who examined the influence of ethnic affiliation (via active membership in German clubs or churches, endogamous relationships, or use of the language at work) and citizenship (permanent resident vs. naturalised citizen) in German immigrants in Australia as attitudinal factors. Waas’ study is largely based on self-report data, for which her classification yields the expected results – however, where she investigates actual linguistic performance, she detects no difference. This suggests that attitude, as measured by her study, influences self-perception but not actual proficiency, i.e. language attrition vs. retention. Further attempts to operationalise this factor have been made within the framework of Ethnolinguistic Vitality Theory (EVT), a framework attempting to identify “that which makes a group likely to behave as a distinctive and active collective entity in inter-group situations” (Giles et al. 1977: 308). In other words, EVT comprises the factors responsible for the strength of the group’s identity feeling: status, demography, institutional support and control (see Hulsen 2000 and Ya˘gmur 1997 for a more detailed discussion of EVT in the context of attrition). Since language has a great symbolic value in a group’s identity, it can be supposed that strong ethnolinguistic vitality would prevent attrition. The findings from the two studies investigating the role of these factors with respect to L1 attrition are, however, not conclusive. Ya˘gmur (1997, see also this volume), in his study on Turkish immigrants in Australia, found no correlation between subjective EV measures and linguistic performance, and neither did Hulsen’s (2000) study on Dutch immigrants in New Zealand. Köpke (2000) attributed differences in the linguistic performance of two groups of German immigrants in Canada and France to differences in the sociocultural environment of the subjects, creating different attitudes and motivations with respect to bilingualism and language maintenance. Nevertheless, the design of the study (Köpke 1999) does not allow validation of this hypothesis and the question needs to be further investigated. However, despite the failure of sociolinguistic frameworks to account for attrition effects, the assumption of a link between attitudes and attrition still stands. Supporting evidence has recently been reported by Schmid (2002), who showed that exceptional settings (such as persecution) might generate emotional factors which influence attrition more strongly than any other extralinguistic circumstance. It is not clear yet to which extent adopted children may suffer from similar traumata, creating emotional factors eventually favouring the attrition of the L1 (e.g. Bode 1996; Isurin 2000; Nicoladis & Grabois
Barbara Köpke and Monika S. Schmid
2002; Ventureyra & Pallier this volume). Attrition studies have to date merely scratched the surface of this complex and elusive phenomenon. The factor amount of contact with the L1 is a complex one in this respect, since it is often impossible to say to what degree it is influenced by attitude, or simply outside the speaker’s control for sociogeographical reasons, if there is no L1 speech community in the country of emigration and financial means do not allow travel home (see Jaspaert & Kroon 1989). The hypothesis that attrition can often be ascribed to lack of contact – more so than to the actual length of time spent in an L2 setting – has been corroborated repeatedly by direct (de Bot et al. 1991; Köpke 1999) and indirect evidence from studies that find less attrition than expected and ascribe this to the fact that their subjects still had substantial contact with L1 (e.g. Schoenmakers 1989). However, Jaspaert & Kroon (1989), who also found “much less language loss than we had expected” failed to find a correlation with language use (Jaspaert & Kroon 1989: 94f.). Contact thus plays an ambivalent role in previous studies, which may have to do with the methodological problems involved. More than with any other factor, measuring the frequency of use depends on self-report data which may be determined more by “how a person wishes to view herself than by an accurate assessment of her linguistic behavior” (Schmid 2002: 23). Furthermore, ‘contact’ is such a varied phenomenon which can take place at varied levels – reading, email, telephone conversations etc. – that a unified research design investigating its impact seems a formidable challenge. Attempts to assess this factor within a more rigorous framework were made on the basis of Social Network Theory (Milroy 1987), attempting to establish a link between attrition on the one hand and the multiplexity and density of the relationships an individual has contracted with other L1 speakers on the other (Hulsen 2001; Kipp 1999). Here again the findings suggest a complex picture. In Hulsen’s study, a larger number of L1 contacts in the neighbourhood even had a negative effect on the number of correct responses in the picture naming experiment (see also Köpke 2001, for an attempt to explain this type of observation). . Conclusion Where extralinguistic factors in language attrition are concerned, we would thus propose that, in order to arrive at a stage where different L1 attrition studies can inform each other, it is necessary to take into account the following factors:
Language attrition
– – – – – –
age at onset of L2 acquisition (simultaneous/early/late bilingual) age at onset of L1 attrition (pre-/post puberty) time since onset of attrition level of education attitudes frequency, amount and settings of use of the attriting language
The first two factors are easily established and discrete. The third one – time since onset of attrition – can also be considered relatively unproblematic. For the other three factors, however, no cross-cultural framework of classification and taxonomy exists as yet. We would suggest that these are pressing issues of extralinguistic factors that language attrition studies still have to address over the next years.
. Explanatory frameworks . Linguistic issues As with the predictor variables outlined above, research on the role of linguistic variables has so far suffered from the lack of dialogue between theoretical frameworks and applied research. It has been generally assumed that attrition is a selective phenomenon, a “rule-governed process” as pointed out by Seliger (1991: 28), which is why it is such an interesting field within linguistics at large. The search for underlying linguistic principles to explain the phenomena witnessed in language attrition must, however, be described as inconsistent and lacking coherence – even though there are individual studies that have applied specific frameworks successfully to their data. Schmid (2002: 11) has identified four linguistic models that have so far influenced the interpretation of language attrition data: – – – –
Regression Simplification Interlanguage Universal Grammar
The earliest frameworks which played a role in language attrition studies were not so much linguistic as cognitive-psychological in nature, dealing with the nature of processes of learning and forgetting. The earliest, most debated and most controversial one of these is doubtlessly the regression hypothesis, which
Barbara Köpke and Monika S. Schmid
plays quite a prominent role in the earlier speculative texts on attrition. This theory is based on the assumption that language loss in aphasia mirrors language development in the child. It rapidly became evident that this hypothesis does not hold with respect to aphasia, since this disorder is generally not progressive in nature and affects only parts of linguistic competence, depending not on internal linguistic factors, but on external factors, related to the kind of brain injury.10 However, as de Bot & Weltens (1991) point out, attrition is a gradual process and might thus be more likely than aphasia to follow the regression hypothesis. The idea that attrition might be governed by a process that has its roots in the overall cognitive nature of memory, and only indirectly in the linguistic system (insofar as the order of acquisition of this system is also governed by the same principles), apparently was very compelling (studies within this framework are e.g. Brewer Bomar 1982; Jordens et al. 1986; Schmid 2002 (L1 attrition); Hedgcock 1991; Kuhberg 1992; Olshtain 1989 (L2 attrition)), especially at a time when research on attrition was very heavy on methodology and light on linguistic theory. Different versions of this theory have been put forward – one that is based on chronology (‘that which is learned last is lost first’) and one that is based on reinforcement (‘that which is learned best, i.e. most often used/reinforced, is preserved longest’). However, although there is quite strong evidence from L2 attrition research for the regression process (Hansen 1999: 9), it has hardly ever been thoroughly tested in L1 attrition studies – instead, the mere idea has often been rejected out of hand. What little evidence there is seems to be conflicting (for an overview see Schmid 2002: 13f., see also ibid: 147f. about an interpretation of Schmid’s own findings in the light of this hypothesis). In all fairness it would have to be said that the jury is still out on the regression hypothesis. A study that is being conducted at the moment (Keijzer in prep.) may eventually shed some more light on this. The second ‘framework’ on the list is one that has very often been used in explanation of language attrition data – despite the fact that it cannot, in reality, explain anything. Within linguistic theory at large, the notion of ‘simplification’ is, in itself, meaningless and only gains its substance within the confines of another theoretical framework, since the idea that the loss of a particular structure or linguistic feature is determined by that item’s complexity is valid only if an ‘enforceable’ definition is given. A random enumeration of what ‘complexity’ could refer to (and has been used to refer to in linguistic theory) includes synthetic as opposed to analytic structures, structures that are acquired late vs. structures that are acquired early, structures that take longer to process (or produce), structures that require many transformations etc. As Schmitt (this
Language attrition
volume) points out, models will have to be developed and applied to attrition data that are more in-depth linguistic in nature – ‘simplification’ or ‘reduction’ is merely a label with no explanatory power.11 Schmitt (and other recent studies) have attempted to operationalise this approach within the 4M-model described by Myers Scotton and Jake (2000). This is a very recent innovation in language attrition research (see also Bolonyai 2002; Gross 2000, this volume) with rather interesting explanatory potential. A variety of the ‘simplification’ approach is one that would assume attrition to follow the same principles that guide language change. An interesting finding in this respect is pointed out by Raidt (1997), who found features in L1 Dutch attrition in South Africa that were reminiscent of early stages of Afrikaans (Raidt 1997: 232). According to Schmid (2002: 13), such processes might be expected in the following domains: reduction in registers due to reduction in functions; reduction of the lexicon affecting more specifically low-frequency items; and reduction in morphological complexity resulting in a more analytical language structure. What is essential is that language change – albeit observed in language contact settings – is language internal. It has to be noted that this framework is more often applied to morphology (Bettoni 1991) or morphophonology (Vago 1991) than to other linguistic domains, and that it is frequently used in studies that are concerned with intergenerational language shift or change (e.g. Bettoni 1991; or the studies reviewed in Maher 1991). Attrition is, however, generally not the consequence of lack of L1 use alone – as it would occur in the hypothetical desert island situation mentioned by Sharwood Smith & Van Buren (1991: 22) and de Bot (2001: 68). On the contrary, L1 is generally replaced by another language, and this language is often assumed to influence the process of L1 attrition. Sharwood Smith’s (1983) hypothesis that transfer is one of the most important processes determining attrition – the interlanguage or cross-linguistic influence (CLI) hypothesis – has been adopted by a large majority of researchers in the field (Altenberg 1991; Grosjean & Py 1991; Gürel 2002; Isurin 2000; Leisiö 2001; Kaufman & Aronoff 1991; Köpke 1999; Major 1992; Pavlenko this volume; Pelc 2001; Seliger 1991, among others). It has been proposed (Seliger 1991: 237) that, in the absence of L1 input, L2 could be taken as a source of “indirect positive evidence”, and that as a result of this, “L2 rules which are formally less complex and have a wider linguistic distribution” will replace more complex and more narrowly distributed L1 rules serving the same semantic function (Seliger 1989: 173). Transfer from L2 undeniably plays some role in L1 attrition (see Pavlenko 2000 for a review), however, it surely is not the only source of linguistic changes in L1 attrition (e.g. Köpke 1999) and it remains to be elucidated in which cases the
Barbara Köpke and Monika S. Schmid
CLI hypothesis applies to attrition, and in which cases it does not. By analogy to what has been suggested by Seliger, Altenberg (1991: 203) claims that “L1 and L2 similarity is a necessary condition for transfer”. Experimental data increasingly point in that direction (e.g. Altenberg 1991; Köpke 1999; Pelc 2001) but it is not clear whether specific linguistic domains of L1, like the lexical-semantic domain for instance, are more susceptible to influence by the L2 than others, such as morphology. The impression that L2 influence has been more often documented in the lexical-semantic domain might be due to the fact that these aspects of attrition simply have been investigated more often. Additionally, in other domains, such as syntax, it might be more difficult to conclude whether attrition is more influenced by external (L2) or internal (simplification) factors (Schmid 2002: 13), since most attrition studies investigate an analytical language (English) as L2, making it difficult or impossible to determine whether attrition follows internal or external factors. Given the strong predominance of theoretical approaches to L1 and L2 acquisition within the framework of Universal Grammar (UG), it seems somewhat surprising that generative models have only relatively recently been applied to language attrition – all the more so since these models, due to their rigor, allow for hypotheses that are far more specific and testable than anything possible within the frameworks discussed above. Such predictions are generally based on the assumption of parameters which are fixed in the language acquisition process, and the questions that have been asked in L2 acquisition with respect to these models (can parameters be reset in L2 acquisition, and what other features – e.g. markedness – play a role in this?, see Clahsen & Muysken 1989) inform the approaches to L1 attrition within a generative framework. The role of markedness is ambiguous in this context, since from a theoretical point of view it has been predicted that ‘marked’ parameters in L1 might be reset to an unmarked value – on the other hand, the validity of this prediction is in doubt, since the setting of parameters is generally assumed to be conditioned by evidence from input, the lack of which is precisely what characterises the attrition process (Sharwood Smith & van Buren 1991: 26). The assumption that a marked parameter may eventually revert to its unmarked setting, if evidence is absent, would thus suggest the need for ‘maintenance’ through reinforcement, which goes somewhat against the grain of parameter setting models. Among the linguistic features investigated within this perspective are the use of null versus overt pronouns in Italian or Greek versus English (Bouba et al. 2002; Sorace 2000), in Turkish versus English (Gürel 2002, this volume) and Japanese versus English (McCormack, this volume) or the compounding parameter in Spanish versus English (Cuza 2002). In these cases, L1
Language attrition
attrition attributed to L2 influence has been documented. More recently, based on findings which fail to detect such influence from L2 (e.g. Håkansson 1995), approaches within the Minimalist Program have speculated on the possible influence of another factor, namely that of [± interpretable], on language attrition (Montrul 2002, this volume; Platzack 1996; Sorace 2000). Such studies predict that features at the interface between syntax and semantics (+interpretable) will be affected by attrition, while ‘mere’ morpho-syntactic features (–interpretable) will remain stable. At the present stage of research, it seems likely that each of the linguistic hypotheses discussed above plays some role in L1 attrition. Kaufman (2001: 186) claims the coexistence of “(. . .) autonomous (language internal) processes [that] trigger regression to earlier developmental forms and integrative processes [that] generate forms that result from contact and interplay between the native language and the dominant language of the environment.” In her study on L1 attrition in Israeli children in the US, Kaufman identifies several areas prone to attrition that can be qualified as an autonomous process, while other areas appear more susceptible to integrative processes.12 . Psycholinguistic aspects From a psycholinguistic point of view, it is hypothesised that insights into the internal reasons for language attrition might help gain a better understanding of the mechanisms underlying language processing and their interaction with respect to two languages that can be considered to be in competition.13 These aspects of L1 attrition have received little attention until now, despite the fact that the evidence available suggests that attrition may be psycholinguistic in nature (de Bot 2002). One of the most important factors to be considered from a psycholinguistic point of view is age. As Kaufman (2001: 185) points out: “Attrition of L1 among older children and adults differs from the L1 attrition process among pre-puberty children.” If there is one thing for which there seems to be ample converging evidence, it is that L1 attrition in children is much more severe than in adults (see above Section 2.1). To explain the influence of this variable, two possibilities must be considered. The first of these is that attrition might be influenced by the same factors which have led to the postulation of the critical period hypothesis, which assumes that, due to brain maturation constraints, L2 learning becomes more difficult past a certain age. With respect to L2 learning there still is no consensus on the existence of this critical period (see Harley & Wang 1997; Long 1993),
Barbara Köpke and Monika S. Schmid
the end of which Lenneberg (1967) initially assumed to coincide with the beginning of puberty (around age 12). It has since repeatedly been proposed that the line should be drawn earlier at least with respect to some linguistic levels. For phonology the critical period is generally assumed to end first, then for syntax and morphology, whereas the lexicon appears least prone to these restrictions. Since no consensus has been reached so far with respect to the existence and age limit of the critical period, Harley & Wang (1997) suggest that a more appropriate concept would be a “sensitive period” which declines gradually after the age of 6 or 7 years. With respect to attrition, the sensitive period hypothesis would imply that the easier it is for the child to learn L2, the more likely is it that she will forget her L1. The findings of empirical research so far strongly suggest that it takes a certain number of years for the L1 to be completely established in the human mind/brain, and that before this moment, the L1 can be easily replaced by another language (e.g. Nicoladis & Grabois 2002; Pallier et al. 2003; Ventureyra & Pallier this volume). It is not clear at what age the L1 can be considered to be fixed. Studies which comprise subjects for whom the onset of attrition can be situated before and after puberty (Ammerlaan 1996; Hakuta & d’Andrea 1992; Pelc 2001) have found age to be one of the most important factors, whereas other studies that include only subjects who immigrated beyond puberty (e.g. Jaspaert & Kroon 1989; Köpke 1999) do not find age effects (see Section 2.1 above). The second possibility is that literacy plays some role in the resistance to attrition, as has been suggested by Olshtain (1986), possibly in interaction with age. The age limit up to which a very severe attrition process can set in for children – until approximately 8 or 9 years – coincides with the time at which most children can be considered to have acquired reading and writing skills in at least one language (L1 or L2). Not only do these skills allow further contact with the language in case of migration (via books, written correspondence, etc.), it is also possible that they facilitate ‘fixing’ the language in the brain, adding another network (see Köpke 2004 for a more detailed discussion). Indeed, psycholinguistic models of the lexicon generally assume the written code to constitute a supplementary set of linguistic representations in the lexicon, be it in the form of independent modules (Caramazza 1988; Caramazza & Miceli 1990; Morton 1969) or in the form of new connections added to the existing oral representations (e.g. Grainger 1993). In addition to these complexities, it seems likely that the ‘fixing’ of the language in the brain is not purely dependent on biological factors like brain maturation and a sensitive period (if such a period could be established), but is
Language attrition
also determined in interaction with factors of identification and emotion (see Section 2.2 above). Another question to be asked within the psycholinguistic perspective is whether a language (and a fortiori an L1) can be truly lost (i.e. erased from memory), or only become more and more inaccessible (cf. Sharwood Smith & van Buren 1991: 17). Fromm (1970) reports the case of a 20-year old man who rediscovered a forgotten childhood language under hypnosis – and was no longer able to understand what he just said in that language once awake.14 Such cases would suggest that there is no true forgetting (in the sense of the disappearance of knowledge): things (and languages) only become inaccessible once they are no longer needed.15 The first scholar to look at attrition from that point of view was Sharwood Smith, who drew attention to the necessity of taking into account the distinction between competence and performance with respect to attrition data as early as 1983 (Sharwood Smith 1983a). Attrition at the level of competence concerns underlying linguistic competence and entails a restructuring of what is known about the language (cf. Seliger 1989; Grosjean & Py 1991). At the level of performance (or processes), attrition results in difficulties in control of that knowledge (which may remain intact) (Ammerlaan 1996: 18). Performance attrition can be supposed to yield two types of phenomena. The most common manifestations are word finding difficulties,16 documented extensively by virtually any attrition study addressing this question (Ammerlaan 1996; Köpke 1999; Olshtain & Barzilay 1991; Waas 1996; Ya˘gmur 1997). The clearest demonstrations of this ‘disorder’ can be found in Ammerlaan (1996) and Hulsen (2000) whose studies are based on a psycholinguistic design allowing a differentiation between productive and receptive language skills with respect to the lexicon: both use a picture naming task to test lexical retrieval (active retrieval of a lexical item from memory) and a picture-wordmatching task for testing lexical access (‘passive’ activation of an item by an external stimulation). The results show that accessibility of the lexicon is clearly reduced, as evidenced by difficulties in the picture naming task (which are quite extensive in the case of the Ammerlaan study), whereas receptive skills measured in the matching task are less impaired (or even perfect in the case of the Hulsen study with respect to first generation immigrants). However, lexical retrieval problems are not the only manifestation of performance attrition. Processing difficulties can be expected in close relation with the cognitive demands of the tasks used for data collection (cf. Dussias 2002). Due to changing dominance patterns between the two languages, subjects may find it more and more difficult to inhibit the L2 while using the L1 in order
Barbara Köpke and Monika S. Schmid
to prevent interference. This inhibition will then consume resources (Green 1986) which will be withdrawn from other levels of processing. At this point, nothing is known about this kind of mechanism in relation to attrition, but the few studies that do allow a comparison of performance across tasks find variability in performance across all linguistic levels. For lexical retrieval, such variability was found in the studies described above (Ammerlaan 1996; Hulsen 2000), but similar findings have been reported with respect to phonetic measures (Major 1992) and morpho-syntactic processing (Köpke 1999). Köpke’s findings on this last point indicate that structures which are prone to errors in a sentence generation task generate even more errors in a grammaticality judgement task. The issue of performance variability is in obvious need of further investigation, variability in performance suggesting strongly that L1 attrition among adults is, in most cases, an issue of processing difficulties (see also de Bot 2002). Sharwood Smith further suggests that the competence/performance distinction he assumes for attrition may evidence itself in three distinct stages (Sharwood Smith 1983b: 51). The first stage would be characterised by systematic deviations in performance while competence remains stable. Stage II is described as a transitional period where the bilingual is in possession of a new (contact) variety, but preserves the ability to switch back to the old variety when required by the circumstances. Stage III would involve the emergence of a new competence – possibly a reduced one, characterised by a decrease in structures available to the speaker, but conceivably also an enriched one, where the repertoire has been changed in line with universal grammar. While there is as yet no evidence for such a restructured linguistic system among adult speakers, Stage II can be illustrated by data from Major (1992), the only attrition study focussing on phonetics. Major reports that in spontaneous speech, some subjects in this multiple case study produce Voice Onset Times (VOTs) in L1 (English) which are no longer native-like and clearly influenced by L2 (Brazilian). However, in a more formal read-aloud task, the same subjects produce more native-like VOTs. Further studies combining formal and casual speech are clearly called for to establish whether such a co-existence of varieties is confined to phonetics, or whether it can be observed in a similarly systematic way on other linguistic levels. A third psycholinguistic framework that attrition research has to take into account concerns the notions of activation and inhibition, which are directly related to the frequency of use or reinforcement of a linguistic system or of a particular linguistic item/structure. An early proposal (Andersen 1982) was that low-frequency items would be more vulnerable to attrition, and this as-
Language attrition
sumption has been validated by experimental data since, at least with respect to lexical retrieval (cf. Ammerlaan 1996; Hulsen 2000; Schmid forthc. b). From a psycholinguistic point of view, this phenomenon can be explained within the concept of activation, which has become very popular with the development of connectionist models since the beginning of the eighties. Activation and inhibition mechanisms, on one view, appear to account for the control of multiple languages in the brain (Green 1986; Paradis 1993) as well as for changing dominance patterns. The Activation Threshold Hypothesis assumes that items (or languages) that are more frequently activated need less stimulation to be reactivated than items (or languages) that are less frequently activated (Paradis 1985, 1993, 2001). Within this framework, attrition is thus predicted in the form of reduced accessibility as a natural consequence of lack of language use. However, it is only recently that the Activation Threshold Hypothesis has been applied to language attrition data (Köpke 1999, 2002; Gürel 2004). Further psycholinguistic approaches to language attrition are clearly called for, in order to shed some light upon important methodological issues, such as a more comprehensive understanding of the implication of age at attrition onset,17 a better knowledge of task demands and their implications for the study of attrition (see below, Section 4), or a better understanding of frequency effects in the context of attrition. It is nevertheless striking that all sociolinguistic factors which have so far been established as bearing on the attrition process seem to have psycholinguistic consequences: a. Education level: A higher education level can be supposed to lead to more explicit knowledge of a language, which will make interference with implicit competence less likely.18 b. Age of departure: In accordance with the sensitive period hypothesis, attrition data suggest that the younger a child is when she changes her linguistic environment, the higher the probability that the L2 will replace the L1. c. Contact with L1: the amount of contact has direct consequences on frequency of activation of L1, and thus influences the psycholinguistic balance between the two languages. . Conclusion The outline of the predominant models applied in language attrition studies presented above suggests a recursiveness in theoretical considerations. These started from cognitive aspects with the regression framework, then moved to-
Barbara Köpke and Monika S. Schmid 100 80
UG Simplification
60
4M Regression
40
Psycholing Interlanguage
20 0 1982–1986
1987–1991
1992–1996
1997–2002
Figure 3. Theoretical frameworks
wards increasingly more rigorous assumptions based on various aspects of the linguistic systems themselves, only to eventually arrive back at cognition and psycholinguistic aspects of language processing. This development in theoretical approaches is also evident in the interest curves associated with these various frameworks within applied language attrition studies. The picture presented by this overview is a heartening one, since it suggests a growing tendency to apply theoretically more rigorous models (evident, for example, in the move from ‘simplification’ to the 4M-model), and also an interest in psycholinguistic aspects. Based on this, our prediction is that language attrition will be an interesting field to watch as it develops over the coming decade(s).
. Practical concerns: Research designs An important explanatory factor in the context of the wide variety of results reported in language attrition studies is the nature of the data on the basis of which ‘loss’ was assessed. This is a factor with two-fold importance, one concerning the way in which the data was elicited and the other the phenomena within these data which were selected for analysis. A clear differentiation between elicitation designs seems all the more important for attrition research, since the evidence found in research on pathological populations – among which the deficits can be expected to be much
Language attrition
more severe – showed that performance is dependent on task demands (e.g. Waters, Rochon & Caplan 1998). There is abundant evidence that the way a specific linguistic structure is elicited may impact very heavily on the amount of ‘errors’ an individual produces, particularly in linguistic situations that are in some way ‘problematic’ (aphasia, agrammatism, dementia, etc., cf. Penke 1998; Waters et al. 1998). However, none of the existing studies on language attrition have discussed the methods of elicitation used or considered the suitability of the task to detect problems in language production which are the outcome of the attritional process. There furthermore is a great amount of variation in the research designs and elicitation techniques. This variation is to a large degree conditioned by the discipline within which the individual research is conducted. This split is most evident in the differences between L1 and L2 attrition research which, as was noted above, drifted apart in the 1990s. Since that time, L2 attrition research has predominantly focussed on tests typically used for L2 acquisition studies, whereas L1 attrition has often tended towards the broader survey techniques as applied by sociolinguistic approaches on multilingualism and minority languages. There appear to be three larger categories of data elicitation techniques in L1 attrition research. The first one is to invite self-assessments (typically CanDo-Scales), the second involves ‘formal’ tasks targeted at eliciting the linguistic structure or phenomenon of interest, and the third attempts to stimulate data that are as close as possible to ‘natural’, i.e. unobserved, spoken data. Self-assessments have to be considered a rather slippery factor in measuring language attrition (for a discussion see Hulsen 2000: 24), although there does appear to be increasing evidence that they can be a relatively good predictor of L1 attrition as measured by more formal tasks. Various studies, using Can-Do scales similar to the ones originally devised by Clark (1981) find that self-assessments correlate with the linguistic scores achieved (Hulsen 2000; Ya˘gmur 199719 ) – it has to be added, however, that self-assessments have also been found to be influenced by attitudinal factors (Hakuta & D’Andrea 1992: 90). We would maintain that the usefulness of this type of data for studies investigating the (psycho)linguistic processes governing language attrition is doubtful – the insights (if any) to be gained from this kind of test being purely sociolinguistic. In the category of formal elicitation techniques we find a wide variety of tasks, the most frequent type being some kind of judgment task on semantic or morphosyntactic features (Altenberg 1991; de Bot et al. 1991; Gürel 2002; Grosjean & Py 1991; Jordens et al. 1989; Köpke 1999; McCormack this volume; Münstermann & Hagen 1986; Pelc 1998; Polinsky 1994; Seliger 1991). The cor-
Barbara Köpke and Monika S. Schmid
rection tasks employed in some studies (Jaspaert & Kroon 1989; Schoenmakers 1998) can be considered a variant of the judgment task, since subjects have to judge the correctness of the linguistic material in order to correct errors. More recently, truth-value judgments (described further in Gürel, this volume, and Montrul, this volume) have been used within the generative approach. In order to test general comprehension, editing tests (Jaspaert & Kroon 1989; de Bot et al 1991) as well as sentence matching tests (Jaspaert & Kroon 1989) have been employed. Other tasks attempt to tap more specifically into the production skills on some linguistic feature, testing e.g. lexical retrieval via naming tasks to elicit isolated words (Ammerlaan 1996; Dugas 2000; Isurin 2000; Schoenmakers 1998; Hulsen et al. 1999) or words in context (Olshtain & Barzilay 1991), via matching tasks (Ammerlaan 1996; Hulsen et al. 1999; Schoenmakers 1998), or via fluency in controlled association tasks (Waas 1996; Ya˘gmur et al. 1999). Productive morphosyntactic skills are generally tested through the instruction of inflecting a given word (or providing the relevant free morpheme) (Altenberg 1991; Bode 1996) or generating a certain kind of syntactic structure (Köpke 1999; Ya˘gmur 1997). This great variety makes overall observations about language attrition almost impossible, since findings from other fields of applied linguistics suggest variance for at least some tests. While, for example, the validity of fluency tests in assessing verbal retrieval difficulties (a test used frequently in attrition studies) has been demonstrated in adult aphasics (Goodglass & Kaplan 1983) and healthy elderly speakers (see Goral 2004), these tests were shown not to be an adequate tool to assess such difficulties in normal vs. language-disordered children by Hall & Jordan (1987), who conclude that “word-finding problems may be symptoms of a variety of language problems and therefore elude any single identification technique” (Hall & Jordan 1987: 109). Verbal fluency tests can thus not a priori be assumed to be a valid instrument in assessing adult non-pathological language loss. It is furthermore important to remember that this test is a performance measure (not allowing for conclusions about non-pathological language loss, only about accessibility) restricted to specific lexical domains. As such, it certainly is a valid measure for lexical retrieval abilities; but it has not been demonstrated whether the results of this and other tests correlate with difficulties the attriter encounters in the production of unguided discourse. L1 acquisition studies suggest that elicited data often contain a number of errors which dramatically exceeds that found in free discourse (e.g. Clahsen et al. 1990: 112; Penke 1998). The most controversial task in this context is certainly translation – a paradigm often used to assess language skills in shift/death studies (Dorian
Language attrition
1982; Dal Negro 2001; De Kadt 2001; Elordui 2001; Smits 1996) but recently also applied in language attrition (Ben Rafael 2001; Cuza 2002; Hirvonen 1995; Macevichius 2001; Polinsky 1994; Schoenmakers 1998). It is unclear what skills such tests attempt to measure, the activity of translation being a highly specialised one and not necessarily comparable to other kinds of linguistic activities (see also Ya˘gmur, this volume). Furthermore, there are three practical reasons why we consider translation tasks at the very least highly problematic: –
–
–
translations have been shown to contain far more variation, and a far higher number of unacceptable structures than spontaneous conversation (Smits 2001: 315) the nature of the task makes it difficult to compare findings against an unattrited control group, since the distinction between attriters and active bilinguals is a complex one translation tasks oblige the subject to activate the two languages simultaneously which may lead to interference (cf. also Grosjean 1998, for a discussion of the role of monolingual and bilingual modes on bilingual processing)
We would therefore suggest that controversies and debates on the validity of specific research designs have to be pointed out and reflected in any study using them. Finally, a large number of attrition studies have attempted to stimulate their subjects into producing data that resemble spontaneous speech as closely as possible. This has been done by interviews of varying degrees of formality, or by eliciting a narrative on the basis of some kind of visual stimulus (pictures or films). This method of data elicitation seems compelling, since it attempts to measure attrition at the level at which it actually occurs, as against the largely abstract skills that formal tests aim at tapping into. However, the methodological concerns in working with free data are formidable, as the question of how to quantify any of the findings in such a setting remains yet to be settled (Schmid forthc. b), as does the question of how to address the issue of avoidance strategies. Obviously the manner of data elicitation can only ever be evaluated alongside a fair consideration of the goals pursued, since they permit very different conclusions. Naturalistic methods are ideal for global approaches, aimed at measuring all aspects of performance, or for preliminary research in order to identify sensitive areas. More formal tasks, on the other hand, allow for better control of linguistic and psycholinguistic factors and should thus be applied in the investigation of specific domains. It should nevertheless be kept in mind
Barbara Köpke and Monika S. Schmid 100 80
Can-Do scales/self reports Lexical task
60
Grammatical task Naturalistic data
40 20 0 1982–1986
1987–1991
1992–1996
1997–2002
Figure 4. Elicitation techniques
that the means of data elicitation should be carefully chosen in accordance with the hypothesis to be tested. That said, the application of the different tasks over the past two decades appears to have remained relatively constant, with one exception: recently, selfreports have been all but abandoned (Fig. 4). The second issue raised above with respect to research designs concerns the features of the data under observation which are actually used in the analysis. The study of language attrition has always been the study of deviance. The underlying and unquestioned assumption in practically any given study is that to measure ‘mistakes’ in attrited language – mistakes of production, mistaken grammaticality judgments, mistakes of grammar, mistakes of naming – is to measure the degree of an individual speaker’s attrition. Furthermore, in many cases no unattrited control group is used, so that the baseline for comparison is the tacit assumption of error-free homogenous native speech. The issue of ‘mistakes’ is discussed in detail in Schmid (forthc. b, this volume). Suffice it to say here that – methodologically or even fundamentally – this is a very problematic concept for (at least) three reasons: – – –
establishing if a certain structure is a mistake or not is highly difficult – ‘right’ or ‘wrong’ being anything but discrete establishing a base-line for comparison in unattrited speech is equally problematic the distribution of mistakes has to be correlated with the obligatory contexts, a comparison of the number of mistakes on two features is, in itself, meaningless
Language attrition
We would thus strongly advocate that future studies on language attrition combine the analysis of ‘what went wrong’ with data on ‘what went right’ – while using data from all three categories described above in order to make correlations possible.
Conclusion and outlook All of these considerations motivated us to attempt to initiate a ‘next phase’ in language attrition research. We therefore published a call for papers for a threeday symposium, inviting participants to explicitly address one or more of the following questions: Concerning data: 1. What is a ‘mistake’? How many judges do you need in order to establish whether something is a mistake or not? 2. What are the differences in the occurrence of ‘mistakes’ in elicited vs. spontaneous data? What methods are there available to generate elicited data that is more like spontaneous data? 3. Can a study of language attrition be confined to analysing interference data, or is it necessary to also include an analysis of ‘proficiency data’ (e.g. type-token frequencies, syntactic and morphological complexity etc.)? Concerning data collection: 4. What are the possibilities and limitations of different tests, i.e. what are the processing demands involved in a specific test and what kind of observations does it allow? 5. How does the procedure of data collection influence the results? (language choice, bilingual vs. unilingual interviewers, codeswitching during the test, etc.) 6. Is a control group necessary, and, if so, how do we establish one? (methodological differences here include the fact that a control group should, ideally, be unilingual – but that very often, choosing a unilingual group of subjects will produce an average level of education that is much lower than that of the group of attriters) Concerning sociolinguistic factors: 7. What extralinguistic factors (age, education, gender etc.) do we have to take into account?
Barbara Köpke and Monika S. Schmid
8. To what degree do we have to allow for dialectal and sociolectal variation among our attriters, and how do we control for that? 9. How do we account for variation in the socio-cultural environment and/or history of the subjects and of its consequences on attitudes or motivation? And finally: 10. If I could do the research all over again from the start, what would I do differently? These were the central issues dominating the structure of the conference. The first day focused on issues concerned with theoretical frameworks, the second day on issues of data collection, and the last day on data analysis. A large number of issues and problems in connection with these overall topics were raised, while at the same time many researchers had the chance to present their research and findings to the community for the first time. The purposes that we intended to achieve with this congress can be summed up as follows: 1. To arrive at a clearer, and more explicit, picture of the theoretical models that have been used to account for the findings in language attrition studies than is usually presented 2. To achieve a better communication between researchers engaged in language attrition studies, in order to make it unnecessary for everyone to ‘re-invent the wheel’ at the outset of their study 3. To achieve a level of consciousness about the benefits and disadvantages of particular research designs and data elicitation methods 4. To begin forming an international network of attrition researchers 5. To achieve communication between ‘established’ scientists and young researchers who are starting out in language attrition studies We are pleased to state that, on all of these topics, great progress appears to have been made during the symposium. Specifically, 1. In the course of the presentations, issues in connection with theoretical models such as the regression hypothesis and the Basic Variety in L2 acquisition (Jordens), Universal Grammar (Montrul), language contact (Muysken), dialectology (van Hout), psycholinguistics (Ventureyra), sociolinguistics (Jääskeläinen), social networks (Kipp) – among others – were raised.
Language attrition
2. The presentations and discussions, both formal and informal, as well as the connections formed and friendships struck at the conference gave all of the participants a good overview of the state-of-the-art in current research. 3. Both presentations that explicitly focused on the merits and disadvantages of specific elicitation methods (Altenberg & Vago, Schmid) and presentations and discussions about individual research projects (Hutz, Latkowska, among others) greatly contributed to the overall understanding of problems concerned with specific research designs. 4. It was unanimously considered desirable to have follow-up events, such as further conferences, an internet discussion list (which has since been installed, and to which many of the conference delegates as well as researchers who were unable to attend have already subscribed), and a language attrition website (in development on the server of the Vrije Universiteit). 5. As a follow-up, a graduate workshop took place at the Vrije Universiteit in January 2003, under the auspices of Kees de Bot and Monika S. Schmid. Overall, the strong presence of young researchers working on an MA (Leppanen, Weilemar), just starting a Ph.D. (Dostert, Jääskeläinen, Keijzer, Opitz, Prescher, Thuss), or just having finished a Ph.D. (Gross, Gürel, Pelc, Schmitt) as well as that of established scientists (de Bot, Dewaele, Dussias, Jordens, van Hout, Muysken, Pavlenko, Vago) was one of the strongest and most beneficial aspects of the conference. The papers included in this volume fall into several groups. The first part focuses on theoretical and methodological issues of data collection and analysis, beginning with Pavlenko’s contribution on the difference between L2 influence on L1 on the one hand and L1 attrition on the other. This issue is a crucial one, since it is generally assumed in attrition studies that all cases of interference are due to attrition, without taking into account that interference may occur in ‘normal’ bilinguals as well. Jiménez Jiménez and Dewaele both present innovative approaches to the study and analysis of attrition from the point of view of attitudes, Jiménez Jiménez’ approach being located within a theoretical framework (Vygotskian Sociocultural Theory), while Dewaele chooses a corpus-based approach. The methodological perspective is also reflected in Altenberg & Vago’s article on the advantages and disadvantages of (and conclusions to be drawn from) grammaticality judgments. This reflection is important, as grammaticality judgments are still one of the most popular data elicitation methods. The second section contains a number of studies observing and describing the process of attrition in a range of languages and settings. It is introduced
Barbara Köpke and Monika S. Schmid
by Ya˘gmur’s reflection and reassessment of his 1997 study on Turkish in Sydney, followed by Ben Rafael’s large-scale study of ‘Franbreu’, the spoken French of Israeli Francophones, presenting a qualitative analysis of language contact phenomena (codeswitching, borrowing, interference) in different discourse situations among Francophones in Israel. The interest of this approach is that attrition is not seen as the only reason for modifications in language performance. Instead, it attempts to investigate the difference between ‘contributions and enrichments’ of Franbreu due to the language contact situation on the one hand and what should be considered attrition on the other. Especially interesting within this respect is the fact that the author takes her examples from different informants involved in different communication situations varying with respect to degree of formality. A rather different research design is chosen by Hutz, who presents a detailed investigation of a written corpus spanning six decades: a collection of letters written by a German speaker in emigration in the US, the findings of which suggest that even after such a long emigration span, attrition phenomena can be very minor indeed. The very opposite case is documented in the last study in this section: Ventureyra & Pallier present an investigation of the ‘lost language’ of Korean adoptees in France from a neurolinguistic perspective, suggesting far more radical patterns of attrition than documented by any other study so far. The last section represents studies which interpret their data within the range of a clearly defined theoretical framework, with a view to finding evidence substantiating the model. First of all, there is a cluster of papers reporting on studies carried out from the point of view of Universal Grammar, with very suggestive findings. Gürel’s data on the L1 attrition of Turkish in an L2 English context suggest that, as had been speculated (see above), similarity of languages is a “necessary condition for transfer” (Altenberg 1991: 203); i.e. that L1 forms that have no analogue in L2 are more easily preserved than those which do have an L2 equivalent. McCormack looks at a linguistic feature that is very similar to that analysed by Gürel – constraints on reflexive binding – but investigates this in the context of the L2 attrition of English in a Japanese context, concluding that the attriting L2 system is still sanctioned by UG. Based on a very similar approach, Montrul completes these findings with data from a study comparing L2 learning and L1 attrition in the case of Spanish heritage speakers in the USA who are furthermore confronted with the problem of incomplete learning. Two further papers in this section report on studies carried out within Myers Scotton & Jake’s Abstract Level and 4-M models. Gross investigates the attrition of L1 German in late bilinguals, while Schmitt gives an account of the attrition of L1 Russian in children (both studies are carried out in an L2 English
Language attrition
environment). These papers suggest that attrition might not be a process the effect of which is apparent across the entire morphological system, but that it is selective, affecting some levels of abstract lexical structure before others. The conclusions drawn from these studies can only be tentative, but the theoretical framework presented here does look promising. This selection of recent research and reflections on L1 attrition thus constitutes a broad overview of the current state of the art and shows that attrition research is indeed entering a new phase. While the definition of attrition is becoming more concrete and viable, aspects from various other disciplines are permitting a broader view of the subject in all of its fascinating facets. At this point, there is one thing that we can predict with absolute certainty for the future of the discipline: the coming ten years will be exciting.
Acknowledgment We would like to thank the following people for their useful comments on earlier versions of this article, as well as for sharing their insights and experiences with us during the writing process: Kees de Bot, Barbara Freed, Ay¸se Gürel, Merel Keijzer, Sjaak Kroon, Richard D. Lambert, Chris McCully, Aneta Pavlenko, Robert Vago, Theo van Els, Bert Weltens.
Notes . A large amount of the information on the background history of language attrition research presented in this section comes from personal conversations and emails exchanged with researchers who played ‘key roles’ in the first decade or so of the research. We are very grateful to Kees de Bot, Sjaak Kroon, Richard D. Lambert, Robert Vago, Theo van Els, and Bert Weltens for sharing their recollections. . This remark was made in a keynote address at the 3rd International Conference on maintenance and loss of minority languages, Veldhoven, The Netherlands, Nov. 1998. . These observations are based on an analysis of some 160 publications on attrition, i.e. on non-pathological, intragenerational language loss (see also the annotated bibliography, this volume). Where more than one publication exists about the same study, only the most extensive one was counted (except in the case of pilot studies with a clear difference in the data), leaving a body of some 100 publications. We tried to be very thorough in our investigation, but in this type of overview, the possibility of overlooking some studies can, of course, never be totally excluded, so the figures presented here are to be considered approximations, not the gospel truth.
Barbara Köpke and Monika S. Schmid . This problem is aggravated by the fact that Tilburg University Press, which published many PhD studies and other works on language attrition – such as El Aissati 1997, KlatterFolmer & Kroon (eds.) 1997; Ya˘gmur 1997 – went out of business, making it nearly impossible to obtain copies of these important studies and collections. . At least within the field of attrition research. . In Ammerlaan’s study, age at departure varied from 6 to 29 years, and Pelc’s subjects were between 8 and 32 years old when they left Greece, with 50 % of the subjects having been between 12 and 17 years of age. Note however that Köpke (1999) found no age effect among her subjects, all of whom were beyond puberty when they emigrated (14–36 years, mean = 24 years), and neither did Schmid (2002) between her younger (11–16 years, mean 13.17) and her older (17–29 years, mean 22.53) group of emigrants. . Note that there was a correlation between level of education and length of stay in Ya˘gmur’s attriting groups, which may have distorted findings here (Ya˘gmur 1997: 89). . Note that in this study, it could not be prevented that the control group had a lower education level than the experimental groups. This might well be the case for other attrition studies too, since ‘unilinguals’ frequently have a lower education level than bilinguals or polyglots. . This approach is convincing, since most of her subjects were children or teenagers at the moment of immigration and completed schooling in L2. . Note that Jakobson modified his initial (1940) position later on (see Nespoulous 1997, for more details). . Unless embedded in a larger explanatory framework, as in the case of Seliger’s ‘Redundancy Reduction Principle’ (see below). . It is not clear from this research whether the coexistence of the two types of processes is specific to attrition in children or if it can be observed to the same extent in adults. . “The languages spoken by the bilingual may be said, metaphorically, to coexist in a state of competition for a finite amount of memory and processing space in the mind of the speaker.” (Seliger & Vago 1991: 4) . Note, however, that this finding has not been confirmed by another hypnosis study with a similar case (As 1963). . For a discussion of psychological theories of forgetting and their implications for attrition, see Ammerlaan (1996: 10–19). . Note that word finding difficulties are also the most common disorder in all types of aphasia, in healthy aging (see Goral 2004), as well as one of the most frequent performance problems observed in young healthy unilinguals (particularly in case of fatigue, stress, etc.), suggesting that lexical retrieval is one of the most vulnerable aspects of language processing. . Note that there are no attrition studies of early bilinguals for whom the onset of attrition is situated in adulthood after having passed through a period of balanced bilingualism. If such a case exists, it should be added to the puzzle! . For a discussion of explicit vs. implicit memory for bilingualism and second language learning, see Paradis (1994, 1997).
Language attrition
. We consider the self-report scale used by Ammerlaan (1996), which only allows for the answers ‘yes’ and ‘no’ to questions such as “Can you speak Dutch?” or “Can you write in Dutch?” to be too unspecific to yield interpretable results, and will not discuss it here.
References Altenberg, E. P. (1991). “Assessing first language vulnerability to attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 174–188). Cambridge: CUP. Ammerlaan, T. (1996). “You Get a bit Wobbly...” – Exploring bilingual lexical retrieval processes in the context of first language attrition. Unpublished Doctoral Dissertation, Katholieke Universiteit Nijmegen. Ammerlaan, T., Hulsen, M., Strating, H., & Ya˘gmur, K. (Eds.). (2001). Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages. Münster: Waxmann. Andersen, R. W. (1982). “Determining the linguistic attributes of language attrition”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 83–117). Rowley, MA: Newbury House. Armon Lotem, S. (2000). “Language attrition: Why are resumptive pronouns so susceptible?” In S. C. Howell, S. Fish & K. Thea (Eds.), Proceedings of the 24th Annual Boston University Conference on Language Development (pp. 58–67). Somerville, MA: Cascadilla. As, A. (1963). “The recovery of a forgotten language knowledge through hypnotic age regression: A case report”. American Journal of Clinical Hypnosis, 5, 24–29. Ben-Rafael, M. (2001). Contact de Langues: le français parlé des francophones israéliens. Unpublished Doctoral Dissertation, University of Tel Aviv. Berman, R. A., & Olshtain, E. (1983). “Features of first language transfer in second language attrition”. Applied Linguistics, 4(3), 222–234. Bettoni, C. (1986). “Italian language attrition in Sydney: The role of birth order”. In C. Bettoni, T. de Mauro, P. Crino, & J. J. Kinder (Eds.), Italian Abroad: Studies on language contact in English-speaking countries (pp. 61–85). Sydney: Camilla Frederick May Fndtn. for It. Studies. Bettoni, C. (1989). “Language loss among Italians in Australia: A summary of current research”. I.T.L. Review of Applied Linguistics, 83/84, 37–50. Bode, S. de (1996). First Language Attrition: Productive Morphology Disintegration and Neurobiological Support. A case study. Unpublished MA Thesis, Ponoma, CA: California State Polytechnic University. Bolonyai, A. (1999). The Hidden Dimension of Language Contact: The case of HungarianEnglish bilingual children. Unpublished Doctoral Dissertation, Columbia, SC: University of South Carolina. Bonch-Bruevich, X., Crawford, W. J., Hellermann, J., Higgins, C., & Nguyen, H. (Eds.). (2001). The Past, Present, and Future of Second Language Research: Selected Proceedings of the 2000 Second Language Research Forum. Somerville, MA: Cascadilla.
Barbara Köpke and Monika S. Schmid
Bot, K. de (2001). “Language use as an interface between sociolinguistic and psycholinguistic processes in language attrition and language shift”. In J. Klatter-Folmer & P. Van Avermaet (Eds.), Theories on Maintenance and Loss of Minority Languages. Towards a more integrated explanatory framework (pp. 65–81). Münster: Waxmann. Bot, K. de (2002). “Language attrition – where are we and where are we going?” International Conference on First Language Attrition. Amsterdam: 22–24 August 2002. Bot, K. de, & Clyne, M. (1989). “Language reversion revisited”. Studies in Second Language Acquistion, 11, 167–177. Bot, K. de, & Clyne, M. (1994). “A 16-year longitudinal study of language attrition in Dutch immigrants in Australia”. Journal of Multilingual and Multicultural Development, 15 (1), 17–28. Bot, K. de, & Fase, W. (Eds.). (1991). Migrant languages in Europe (=International Journal of the Sociology of Language, 90). Berlin: Mouton de Gruyter. Bot, K. de, Gommans, P., & Rossing, C. (1991). “L1 loss in an L2 environment: Dutch immigrants in France”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 87–98). Cambridge: CUP. Bot, K. de, & Weltens, B. (1985). “Taalverlies: Beschrijven versus verklaren”. Handelingen van het 38e Nederlands Filologencongres, 51–61. Bot, K. de, & Weltens, B. (1991). “Regression, recapitulation, and language loss”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 31–52). Cambridge: CUP. Bouba, M., Filiaci, F., Heycock, C., Sorace, A., & Tsimpli, I. (2002). “Syntactic attrition in Greek and Italian near-native speakers of English”. International Conference on First Language Attrition. Amsterdam, 22–24 August 2002. Brewer Bomar, K. E. (1982). “Second language lexical and syntactical interference on the first language of two four year-old Spanish speakers”. Dissertation Abstracts International (DAI). Ann Arbor, MI, 42(12), 5105A. Brons-Albert, R. (1992). “Verlust der Muttersprache in fremdsprachiger Umgebung”. Info DaF, 19 (3), 315–325. Brons-Albert, R. (1994). “Interferenzfehler in der Muttersprache von in den Niederlanden lebenden Deutschen”. In B. Spillner (Ed.), Nachbarsprachen in Europa (pp. 96–104). Frankfurt: Peter Lang. Brown, J. C. (2001). L1 Attrition Among Native Speakers of English Residents Abroad: A sociolinguistic case study from Foggia, Italy. Unpublished MA Thesis: University of Surrey. Caramazza, A. (1988). “Some aspects of language processing revealed through the analysis of acquired aphasia: The lexical system”. Annual Review of Neuroscience, 11, 395–521. Caramazza, A., & Miceli, G. (1990). “The structure of graphemic representations”. Cognition, 27, 243–297. Clahsen, H., & Muysken, P. (1989). “The UG paradox in L2 acquisition”. Second Language Research, 5, (1), 1–29. Clahsen, H., Rothweiler, M., & Woest, A. (1990). “Lexikalische Ebenen und morphologische Entwicklung: Eine Untersuchung zum Erwerb des deutschen Pluralsystems im Rahmen der Lexikalischen Morphologie.” Linguistische Berichte (Supp. 3), 105–126.
Language attrition
Clark, J. L. D. (1981). “Language”. In T. S. Barrows (Ed.), College Students’ Knowledge and Beliefs: A survey of global understanding (pp. 25–35; 87–100). New Rochelle, NY: Change Magazine Press. Clark, J. L. D. (1982). “Measurement considerations in language attrition research”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 138–152). Rowley, MA: Newbury House. Cuza, A. (2002). “Attrition and the ‘compounding parameter’: L1 Spanish in a contact situation with English”. International Conference on First Language Attrition, Amsterdam: 22–24 August 2002. Dal Negro, S. (2001). “Language shift and change in a Walser dialect in Italy”. In T. Ammerlaan, M. Hulsen, H. Strating, & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 47–60). Münster: Waxmann. Dorian, N. C. (1982). “Language loss and maintenance in language contact situations”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 44–59). Rowley, MA: Newbury House. Dugas, L. G. (2000). “Attrition of pronunciation accuracy among advanced American learners of French”. Dissertation Abstracts International, Section A: The Humanities and Social Sciences (DAIA), 60(11), 3986. Dussias, P. E. (2002). “What you see is not always what you get: Cognitive load disguised as L1 attrition”. International Conference on First Language Attrition. Amsterdam, 22–24 August 2002. Elordui, A. (2001). “Processes of language shift and loss: Evidence from Basque”. In T. Ammerlaan, M. Hulsen, H. Strating, & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 77–92). Münster: Waxmann. Els, T. van (1986). “An overview of European research on language attrition”. In B. Weltens, K. de Bot & T. van Els (Eds.), Language Attrition in Progress (pp. 3–18). Dordrecht: Foris. Fase, W., Jaspaert, K., & Kroon, S. (Eds.). (1992). Maintenance and Loss of Minority Languages. Amsterdam and Philadelphia: John Benjamins. Fase, W., Jaspaert, K., & Kroon, S. (Eds.). (1995). The State of Minority Languages. International perspectives on survival and decline (= European Studies on Multilingualism 5). Lisse: Swets & Zeitlinger. Frenck-Mestre, C. (1993). “Use of orthographic redundancies and word identification speed in bilinguals”. Journal of Psycholinguistic Research, 22 (4), 397–410. Fromm, E. (1970). “Age regression with unexpected reappearence of a repressed childhood language”. International Journal of Clinical and Experimental Hypnosis, 18, 79–88. Fujita, M. (2002). “Second language English attrition of Japanese bilingual children”. Dissertation Abstracts International, Section A: The Humanities and Social Sciences (DAIA), 63 (1), 59–60. Gardner, R. C., & Lambert, W. E. (1972). Attitudes and Motivation in Second Language Learning. Rowley, MA: Newbury House.
Barbara Köpke and Monika S. Schmid
Giles, H., Bourhis, R. Y., & Taylor, D. M. (1977). “Towards a theory of language in ethnic group relations”. In H. Giles (Ed.), Language, Ethnicity and Intergroup Relations (pp. 307–348). New York, NY: Academic Press. Ginsberg, R. B. (1986). “Issues in the analysis of language loss: Methdolology of the Language Skills Attrition Project”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 19–36). Dordrecht: Foris. Goodglass, H., & Kaplan, E. (1983). The Assessment of Aphasia and Related Disorders. Philadelphia, PE: Lea & Febiger. Goral, M. (2004). “First-language decline in healthy aging: Implications for attrition in bilingualism”. Journal of Neurolinguistics, 17 (1), 31–52. Grainger, J. (1993). “Visual word recognition in bilinguals”. In R. Schreuder & B. Weltens (Eds.), The Bilingual Lexicon (pp. 11–25). Amsterdam and Philadelphia: John Benjamins. Green, D. W. (1986). “Control, activation, and resource: A framework and a model for the control of speech in bilinguals”. Brain and Language, 27, 210–223. Grosjean, F. (1998). “Studying bilinguals: Methodological and conceptual issues”. Bilingualism: Language and Cognition, 1, 131–149. Grosjean, F., & Py, B. (1991). “La restructuration d’une première langue: l’intégration de variantes de contact dans la compétence de migrants bilingues”. La Linguistique, 27, 35–60. Gross, S. (2000). The Role of Abstract Lexical Structure in First Language Attrition: Germans in America. Unpublished Doctoral Dissertation, Columbia SC: University of South Carolina. Gürel, A. (2002). Linguistic Characteristics of Second Language Acquisition and First Language Attrition. Turkish overt versus null pronouns. Unpublished Doctoral Dissertation, Montreal: McGill University. Gürel, A. (2004). “Selectivity in L2-induced L1 attrition: A psycholinguistic account.” Journal of Neurolinguistics, 17 (1), 53–78. Håkansson, G. (1995). “Syntax and morphology in language attrition: A study of five bilingual expatriate Swedes”. International Journal of Applied Linguistics, 5 (2), 153–171. Hakuta, K., & d’Andrea, D. (1992). “Some properties of bilingual maintenance and loss in Mexican background high-school students”. Applied Linguistics, 13 (1), 72–99. Hall, P. K., & Jordan, L. S. (1987). “An assessment of a controlled association task to identify word-finding problems in children”. Language, Speech, and Hearing Services in Schools, 18, 99–111. Hansen, L. (Ed.). (1999). Second Language Attrition in Japanese Contexts. New York, NY: Oxford University Press. Harley, B., & Wang, W. (1997). “The critical period hypothesis: Where are we now?” In A. M. B. de Groot & J. F. Kroll (Eds.), Tutorials in Bilingualism. Psycholinguistic Perspectives (pp. 19–52). Mahwah, NJ: Lawrence Erlbaum Associates. Hedgcock, J. (1991). “Foreign language retention and attrition: A study of regression models”. Foreign Language Annals, 24, 43–55. Hiller-Foti, I. (1985). “Erhebung über den Verlust der Muttersprache Deutscher Immigranten in Sizilien). Rassegna Italiana di Linguistica Applicata, 17 (1), 107–110.
Language attrition
Hirvonen, P. A. (1995). “Phonological and morphological aspects of Finnish language attrition in the United States”. In W. Viereck (Ed.), Soziolinguistische Variation (pp. 181– 193). Stuttgart: Steiner. Huls, E., & van der Mond, A. (1992). “Some aspects of language attrition in Turkish families in The Netherlands”. In W. Fase, K. Jaspaert, & S. Kroon (Eds.), Maintenance and Loss of Minority Languages (pp. 99–115). Amsterdam and Philadelphia: John Benjamins. Hulsen, M. (2000). Language Loss and Language Processing. Three generations of Dutch migrants in New Zealand. Unpublished Doctoral Dissertation, Nijmegen: Katholieke Universiteit Nijmegen. Hulsen, M., de Bot, K., & Weltens, B. (1999). “Language shift, language loss and language processing: An investigation of three generations of Dutch immigrants in New Zealand”. In E. Huls & B. Weltens (Eds.), Artikelen van de Derde Sociolinguïstische Conferentie (pp. 221–232). Delft: Eburon. Isurin, L. (2000a). “Deserted islands or a child’s first language forgetting”. Bilingualism: Language and Cognition, 3, 151–166. Isurin, L. (2000b). “Mechanisms of First Language Forgetting”. Dissertation Abstracts International, Section A: The Humanities and Social Sciences (DAIA) 60 (9), 3342. Jakobson, R. (1941). Kindersprache, Aphasie und allgemeine Lautgesetze. Uppsala: Almqvist & Wiksell. Jaspaert, K., & Kroon, S. (1989). “Social determinants of language loss”. I.T.L.: Review of Applied Linguistics, 83/84, 75–98. Jaspaert, K., & Kroon, S. (1992). “From the typewriter of A.L.: A case study in language loss”. In W. Fase, K. Jaspaert, & S. Kroon (Eds.), Maintenance and Loss of Minority Languages (pp. 137–147). Amsterdam and Philadelphia: John Benjamins. Jaspaert, K., Kroon, S., & van Hout, R. (1986). “Points of reference in first-language loss research”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 37–49). Dordrecht: Foris. Jordens, P., de Bot, K., van Os, C., & Schumans, J. (1986). “Regression in German case marking”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 159–176). Dordrecht: Foris. Jordens, P., de Bot, K., & Trapman, H. (1989). “Linguistic aspects of regression in German case marking”. Studies of Second Language Acquisition, 11, 179–204. Kadt, E. de. (2001). “You still speak German? Teenage language skills in a Germanspeaking community”. In T. Ammerlaan, M. Hulsen, H. Strating. & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 61–76). Münster: Waxmann. Kaufman, D. (1998). “Children’s assimilatory patterns and L1 attrition”. In A. Greenhill, M. Hugehes, H. Littlefield, & H. Walsh (Eds.), Proceedings of the 22nd Annual Boston University Conference on Language Development, I, II (pp. 409–420). Somerville, MA: Cascadilla. Kaufman, D. (2001). “Tales of L1 attrition – Evidence from pre-puberty children”. In T. Ammerlaan, M. Hulsen, H. Strating, & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 185– 202). Münster: Waxmann.
Barbara Köpke and Monika S. Schmid
Kaufman, D., & Aronoff, M. (1991). “Morphological disintegration and reconstruction in first language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 175–188). Cambridge: CUP. Keijzer, M. (In preparation). First Language Attrition: A Crosslinguistic Investigation of Jakobson’s regression hypothesis. Doctoral Dissertation, Vrije Universiteit Amsterdam. Kipp, S. (1999). “Networks and language use in an historical context”. Monash University Linguistic Papers, 2 (1), 25–32. Klatter-Folmer, J., & van Avermaet, P. (Eds.). (2001). Theories on Maintenance and Loss of Minority Languages. Towards a more integrated explanatory framework. Münster: Waxmann. Köpke, B. (1999). L’attrition de la première langue chez le bilingue tardif: Implications pour l’étude psycholinguistique du bilinguisme. Unpublished Doctoral Dissertation, Toulouse: Université de Toulouse-Le Mirail. Köpke, B. (2000). “Effet du pays d’accueil sur le maintien de la langue. Le cas des immigrés d’origine allemande”. Éducation et Sociétés Plurilingues/Educazione e Societa’ Plurilingue, 9, 59–65. Köpke, B. (2001). “Quels changements linguistiques dans l’attrition de la L1 chez le bilingue tardif?” TRANEL, 34/35, 355–368. Köpke, B. (2002). “Activation thresholds and non-pathological L1 attrition”. In F. Fabbro (Ed.), Advances in the Neurolinguistics of Bilingualism. Essays in honor of Michel Paradis (pp. 119–142). Undine: Forum. Köpke, B. (2004). “Neurolinguistic aspects of attrition” Journal of Neurolinguistics, 17 (1), 3–30. Kuhberg, H. (1992). “Longitudinal L2-attrition versus L2-acquisition, in three Turkish children – empirical findings”. Second Language Research, 8, 138–153. Lachman, R., & Mistler-Lachman, J. (1976). “Dominance lexicale chez le bilingue”. Bulletin de Psychologie, 15, 281–288. Lambert, R. D., & Freed, B. (1982). The Loss of Language Skills. Rowley, MA: Newbury House. Lambert R. D., & Moore, S. J. (1984). “Recent research on language skill attrition”. ERIC/CLL News Bulletin, 8, 1–8. Lambert, R. D., & Moore, S. J. (1986). “Problem areas in the study of language attrition”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 177–186). Dordrecht: Foris. Leisiö, L. (2001). Morphosyntactic Convergence and Integration in Finland Russian. Unpublished Doctoral Dissertation, University of Tampere. Lenneberg, E. H. (1967). Biological Foundations of Language. New York, NY: Wiley. Long, M. H. (1993). “Second language acquisition as a function of age: Research findings and methodological issues”. In K. Hyltenstam & A. Viberg (Eds.), Progression and Regression in Language: Sociocultural, neuropsychological and linguistic perspectives (pp. 196–221). Cambridge: CUP. Macevichius, J. (2001). “The beginnings of language loss in discourse: A case study of referentiality in American Lithuanian”. In T. Ammerlaan, M. Hulsen, H. Strating, & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 235–248). Münster: Waxmann.
Language attrition
Mägiste, E. (1979). “The competing language systems of the multilingual: A developmental study of decoding and encoding processes”. Journal of Verbal Learning and Verbal Behavior, 18, 79–89. Maher, J. (1991). “A crosslinguistic study of language contact and language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 66–86). Cambridge: CUP. Major, R. C. (1992). “Losing English as a first language”. The Modern Language Journal, 76 (2), 190–208. McCormack, B. (2001). The Maintenance and Loss of Reflexive Anaphors in L2 English. Unpublished Doctoral Dissertation, University of Durham, UK. Milroy, L. (1987). Language and Social Networks. Oxford: Basil Blackwell. Minkowski, M. (1928). “Sur un cas d’aphasie chez un polyglotte”. Revue Neurologique, 49, 361–366. Montrul, S. (2002). “Incomplete acquisition and attrition of Spanish tense/aspect distinctions in adult bilinguals”. Bilingualism: Language and Cognition, 5, 39–68. Morton, J. (1969). “Interaction of information in word recognition”. Psychological Review, 76, 165–178. Münstermann, H., & Hagen, T. (1986). “Functional and structural aspects of dialect loss: A research plan and some first results”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 75–96). Dordrecht: Foris. Myers-Scotton, C., & Jake, J. (2000). “Four types of morphemes: Evidence from aphasia, codeswitching, and second language acquisition”. Linguistics, 8 (6), 1053–1100. Nespoulous, J.-L. (1997). “L’aphasie progressive: un modèle pathologique ideal (?) pour tester la validité de la théorie linguistique ou ‘si Jakobson avait connu l’aphasie progressive”’. Revue de Neuropsychologie, 7 (1), 95–100. Nicoladis, E., & Grabois, H. (2002). “Learning English and losing Chinese: A case study of a child adopted from China”. International Journal of Bilingualism, 6 (4), 441–454. Olshtain, E. (1986). “The attrition of English as a second language with speakers of Hebrew”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 187– 204). Dordrecht: Foris. Olshtain, E. (1989). “Is second language attrition the reversal of second language acquisition?” Studies in Second Language Acquisition, 11 (2), 151–165. Olshtain, E., & Barzilay, M. (1991). “Lexical retrieval difficulties in adult language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 139–150). Cambridge: CUP. Oxford, R. (1982). “Research on language loss – A review with implications for foreignlanguage teaching”. Modern Language Journal, 66, 160–169. Pallier, C., Dehaene, S., Poline, J.-B., LeBihan, D., Argenti, A.-M., Dupoux, E., & Mehler, J. (2003). “Brain imaging of language plasticity in adopted adults: Can a second language replace the first?” Cerebral Cortex, 13, 155–161. Paradis, M. (1985). “On the representation of two languages in the brain”. Language Sciences, 7 (1), 1–39. Paradis, M. (1993). “Linguistic, psycholinguistic, and neurolinguistic aspects of ‘interference’ in bilingual speakers: The Activation Threshold Hypothesis”. International Journal of Psycholinguistics, 9 (2), 133–145.
Barbara Köpke and Monika S. Schmid
Paradis, M. (1994). “Neurolinguistic aspects of implicit and explicit memory: Implications for bilingualism and SLA”. In N. Ellis (Ed.), Implicit and Explicit Learning of Languages (pp. 393–419). London: Academic Press. Paradis, M. (1997). “The cognitive neuropsychology of bilingualism”. In A. M. B. de Groot & J. F. Kroll (Eds.), Tutorials in Bilingualism: Psycholinguistic perspectives (pp. 331–354). Hillsdale, NJ: Lawrence Earlbaum Associates. Paradis, M. (2001). “An integrated neurolinguistic theory of bilingualism (1976–2000)”. In R. M. Brend, A. K. Melby, & A. R. Lommel (Eds.), LACUS Forum XXVII. Speaking and Comprehending (pp. 5–15). Fullerton, CA: LACUS. Pavlenko, A. (2000). “L2 influence on L1 in late bilingualism”. Issues in Applied Linguistics, 11 (2), 175–205. Pavlenko, A. (2002). “Bilingualism and Emotions”. Multilingua, 21 (1), 45–78. Pelc, L. (2001). L1 Lexical, Morphological and Morphosyntactic Attrition in Greek-English Bilinguals. Unpublished Doctoral Dissertation, New York: CUNY. Penke, M. (1998). Die Grammatik des Agrammatismus: eine linguistische Untersuchung zu Wortstellung und Flexion bei Broca-Aphasie. Tübingen: Niemeyer. Platzack, C. (1998). “The initial hypothesis of syntax: A minimalist perspective on language acquisition and attrition”. In H. Clahsen (Ed.), Generative Perspectives on Language Acquisition (pp. 369–414). Amsterdam/Philadelphia: John Benjamins. Polinsky, M. (1994). “Strucutral dimensions of first language loss”. Chicago Linguistic Society, 30, 257–276. Py, B. (1986). “Native language attrition amongst migrant workers: Towards an extension of the concept of interlanguage”. In E. Kellerman & M. Sharwood Smith (Eds.), Crosslinguistic Influence in Second Language Acquisition (pp. 163–172). New York, NY: Pergamon. Raidt, E. (1997). “Interference, shift and loss of Dutch in South Africa”. In J. Klatter-Folmer & S. Kroon (Eds.), Dutch Overseas: Studies in maintenance and loss of Dutch as an immigrant language (pp. 215–236). Tilburg: Tilburg University Press. Schmid, M. S. (2002). First Language Attrition, Use, and Maintenance. The case of German Jews in Anglophone countries. Amsterdam/Philadelphia: John Benjamins. Schmid, M. S. (Forthcoming a). “Identity and attrition: A historical approach”. To appear in A. Pavlenko & J.-M. Dewaele (Eds.), Bilingualism and Emotions. Special issue of Estudios de Sociolinguistica. Schmid, M. S. (Forthcoming b). “Measuring language attrition in free speech: What’s in a mistake?”. International Journal of Bilingualism. Schmitt, E. (2001). Beneath the Surface: Signs of language attrition in immigrant children from Russia. Unpublished Doctoral Dissertation, University of South Carolina. Schoenmakers-Klein Gunnewiek, M. (1998). Taalverlies door Taalcontact? Een onderzoek bij Portugese migranten. Tilburg: Tilburg University Press. Seliger, H. W. (1989). “Deterioration and creativity in childhood bilingualism”. In K. Hyltenstam & L. K. Obler (Eds.), Bilingualism Across the Lifespan (pp. 173–184). Cambridge: CUP. Seliger, H. W. (1991). “Language attrition, reduced redundancy, and creativity”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 227–240). Cambridge: CUP. Seliger, H. W., & Vago, R. M. (Eds.). (1991). First Language Attrition. Cambridge: CUP.
Language attrition
Sharwood Smith, M. A. (1983a). “On explaining language loss”. In S. Felix & H. Wode (Eds.), Language Development at the Crossroads (pp. 49–69). Tübingen: Gunter Narr. Sharwood Smith, M. A. (1983b). “On first language loss in the second language acquirer: Problems of transfer”. In S. Gass & L. Selinker (Eds.), Language Transfer in Language Learning (pp. 222–231). Rowley, MA: Newbury House. Sharwood Smith, M. A., & van Buren, P. (1991). “First language attrition and the parameter setting model”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition, (pp. 17– 30). Cambridge: CUP. Smits, C. (1996). Disintegration of Inflection: The case of Iowa Dutch. Unpublished Doctoral Dissertation. Amsterdam: Vrije Universiteit. Soesman, A. (1997). “An experimental study on native language attrition in Dutch adult immigrants in Israel”. In J. Klatter-Folmer & S. Kroon (Eds.), Dutch Overseas: Studies in maintenance and loss of Dutch as an immigrant language (pp. 181–194). Tilburg: Tilburg University Press. Sorace, A. (2000). “Differential effects of attrition in the L1-syntax of near-native L2 speakers”. In C. Howell, S. Fish, & T. Keith-Lucas (Eds.), Proceedings of the 24th Boston University Conference on Language Development (pp. 719–725). Somerville, MA: Cascadilla Press. Toribio, A. J. (2000). “Code Switching and minority language attrition: Papers from the 1999 Conference on the L1 and L2 acquisition of Spanish and Portuguese”. In R. P. Leow & C. Sanz (Eds.), Spanish Applied Linguistics at the Turn of the Millennium (pp. 174–193). Somerville, MA: Cascadilla. Turian, D., & Altenberg, E. P. (1991). “Compensatory strategies of child first language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 207–226). Cambridge: CUP. Vago, R. M. (1991). “Paradigmatic regularity in first language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 241–252). Cambridge: CUP. Waas, M. (1996). Language Attrition Downunder. Frankfurt: Peter Lang. Waters, G. S., Rochon, E., & Caplan, D. (1998). “Task demands and sentence comprehension in patients with dementia of the Alzheimer’s type”. Brain and Language, 62, 361–397. Weltens, B., de Bot, K., & van Els, T. (Eds.). (1986). Language Attrition in Progress. Dordrecht: Foris. Weinreich, U. (1953). Languages in Contact: Findings and problems. New York, NY: Linguistic Circle of New York. Ya˘gmur, K. (1997). First language Attrition Among Turkish Speakers in Sydney. Tilburg: Tilburg University Press. Ya˘gmur, K., de Bot, K., & Korzilius, H. (1999. “Language attrition, language shift and ethnolinguistic vitality of Turkish in Australia”. Journal of Multilingual and Multicultural Development, 20 (1), 51–69. Yoshitomi, A. (1994). “The attrition of English as a second language of Japanese returnee children”. Dissertation Abstracts International (DAI) 55(3), 553A.
P I
Theoretical models and methodological aspects
L2 influence and L1 attrition in adult bilingualism Aneta Pavlenko Temple University
Introduction The goal of this paper is to examine the relationship between first language (L1) attrition and second language (L2) influence on the first in individual adults. I will argue that L2 influence on L1 is a phenomenon in its own right and cannot always be taken as evidence of L1 attrition (for an up-to-date treatment see Cook 2003). My discussion will be organised around a classificatory Crosslinguistic Influence (CI) framework, proposed earlier (Pavlenko 2000). This framework differentiates between five processes in the interaction between two languages: 1. borrowing, or addition of L2 elements to the L1 (e.g., lexical borrowing); 2. restructuring, or deletion or incorporation of L2 elements into L1 resulting in some changes, substitutions, or simplifications (e.g., syntactic restructuring or semantic extensions); 3. convergence, or creation of a unitary system, distinct from both L1 and L2 (e.g., production of consonants that are situated at a mid point between the L1 and L2 values); 4. shift, or a move away from L1 structures or values to approximate L2 structures or values (e.g., a shift in typicality, category boundaries, or consonant values); 5. attrition, i.e. loss of some L1 elements, seen in inability to produce, perceive, or recognise particular rules, lexical items, concepts, or categorical distinctions due to L2 influence.
Aneta Pavlenko
This list does not aim to be final nor exhaustive, rather I propose this framework in the spirit of scholarly discussion and discovery. In illustrating the processes, I will draw on my research with Russian L2 users of English (Pavlenko 2002, 2003a, b), on other scholarly investigations, and on personal narratives of adult bilinguals. The discussion below will be limited to L2 influence on L1 in adults and will leave out childhood bilinguals, as well as interactions between three or more languages. It is theoretically grounded in Grosjean’s (1998) and Cook’s (1991) view of bilinguals or L2 users as people with unique linguistic competence, which Cook (1991) terms multicompetence. This perspective emphasises the dynamic nature of bilingual competence whereby speakers’ use of their linguistic and conceptual repertoires is influenced both by their interlocutors and the linguistic and social context of the interaction.
.
Borrowing
The first process in the CI framework is borrowing of new elements, which is most visible in the lexicon, but can also surface in pragmatics or in rhetoric. Oftentimes, lexical borrowing is used as one of the indicators of L1 attrition. Sometimes this indeed could be the case, yet I want to bring up two caveats with regard to borrowing as a sign of L1 attrition. Firstly, research indicates that some – although clearly not all – lexical borrowing in the L2 context stems from the need to name new objects and new types of relationships, or recognise new conceptual distinctions (cf. Andrews 1999). For instance, Russian L2 users of English in the US use phonologically and morphologically assimilated lexical borrowings, such as intruzivnost’ ‘intrusiveness’, dauntaun ‘downtown’, lendlord ‘landlord’, apointment ‘appointment’, or boifrend ‘boyfriend’, all of which correspond to concepts non-existent in Russian culture (Andrews 1999; Pavlenko 2003a; Pavlenko & Jarvis 2002). While this is not the only kind of borrowing possible, I argue that if the lexical item does not have an L1 translation equivalent, we cannot take the use of the item as a sign of L1 attrition – rather, from a multicompetence perspective, it signifies an enrichment of the bilingual’s linguistic and conceptual repertoire. This enrichment is also visible in cases of loan translation, or literal translation of compound words, idioms, and lexical collocations from the L2, such as ones produced by the Russian L2 users of English in my studies (Pavlenko 2003a, b):
L2 influence and L1 attrition in adult bilingualism
(1) a.
он... мм... вторгается в её одиночество ‘ he... uhm... invades her solitude’
b. он вторгается в её эмоции, чувства ‘he invades her emotions, feelings’
Russian lacks the notion of privacy, and thus in order to describe what they see as an invasion of privacy the participants have to appeal to its closest equivalents in Russian, such as odinochestvo ‘solitude’, and at times violate semantic and morphosyntactic constraints of standard Russian. The participants’ selfcorrections, hesitations, pauses, and metalinguistic comments indicate that they are aware of lexical retrieval difficulties in on-line speech. Some explicitly admit difficulties in expressing their thoughts in Russian when particular concepts, such as privacy, are involved. The participants are also aware of the fact that in cases like this it is the conceptual differences between their two speech communities, rather than L1 attrition, that lead to difficulties in word choice and self-expression in the first language: (2) a.
Или, например, privacy... какая privacy?.. по-русски этого нету, я не могу сказать по-русски, знаешь, ну я могу сказать ‘‘Я хочу побыть одна’’, но это звучит слишком драматично, да?.. когда ты говоришь по-английски “I need my privacy” это более как ежедневная вещь и никто, никого это не волнует... ‘Or take, for instance, privacy... what privacy?.. in Russian this doesn’t exist, I cannot say in Russian, you know, well, I can say “I want to be alone”, but this sounds too dramatic, yes?.. when you say in English “I need my privacy” this is more like an everyday thing and no one, it doesn’t bother anyone...
b. мне пришло в голову понятие, но я не нашел слов в русском языке, чтобы это описать... в общем, пришлось бы достаточно долго и неточно как-то описывать ‘A notion came to me, but I didn’t find Russian words to describe it... I would have had to describe it for a long time and inexactly’
Secondly, even if there appears to be an L1 translation equivalent, lexical borrowing may take place to emphasise particular conceptual distinctions and the two lexical items may co-exist to signify distinct conceptual entities. In studies conducted with Cuban immigrants in the US, Otheguy and Garcia (1988, 1993) found that these Spanish-English bilinguals do not always find traditional Spanish terms congruent with their new circumstances. For instance, the informants indicated that the rather informal and easy-going solicitud ‘job
Aneta Pavlenko
application’ is not really a translation equivalent of the bureaucratic ‘application’ and so they adopted the form aplicacion. The researchers have found that translation equivalents, such as ‘Easter’ and Semana Santa, el lunchroom and el comedor escolar, el bildin and el edificio, involve different visual images and conceptualisations for the study participants. They showed that both sets of words are active in the bilinguals’ repertoires and their use is entirely determined by the context of the interaction. Consequently, I argue that the use of lexical borrowing should not necessarily be taken as a sign of L1 attrition without a close examination of the type of borrowing and the reasons for which particular lexical items have been borrowed. Lexical borrowing is evidence of L1 attrition only in cases where an exact L1 equivalent exists but is no longer available to the speaker in either production or comprehension, even when tested in a monolingual L1 context (or at least, when a retrieval of such an item causes significant difficulty and delay). In all other cases, the use of a borrowed lexical item – especially in an L2 context or with a bilingual interlocutor – should not be taken as a sign of L1 attrition, since it is quite possible that the same bilingual may appeal to the L1 translation equivalent in a different setting. When L1 translation equivalents do not exist, as in the case of privacy in Russian, the disfluency in the L2 users’ speech is not a sign of L1 attrition but evidence of conceptual/cultural differences between the two speech communities, while lexical borrowings are evidence of expansion of L2 users’ conceptual repertoires. Jarvis (2003) makes a similar argument about incorporation of morphosyntactic rules suggesting that in the speech of a Finnish L2 user of English in his study, L1 rules and structures have remained intact, and were augmented rather than replaced by L2 rules, structures, and meanings.
. Restructuring The next process, restructuring, typically involves a deletion of certain L1 elements and/or addition of L2 elements into L1. It often occurs in morphosyntax, whereby the L2 rules may be incorporated – perhaps temporarily – into the L1 grammar. To date, morphosyntactic restructuring has been documented in the areas of case-, gender-, and number-marking (Håkansson 1995; Schmid 2002), preposition choice (Jarvis 2003; Py 1986; Schmid 2002), and word-order rules and subcategorisation (Boyd & Andersson 1991; de Bot & Clyne 1994; Py 1986; Schmid 2002; Seliger & Vago 1991; Waas 1996).
L2 influence and L1 attrition in adult bilingualism
Restructuring has also been documented in the study of the lexicon, where, on the one hand, meanings of L2 translation equivalents may be incorporated into the L1 word in semantic extension, and, on the other, certain L1 meanings may be subject to attrition in semantic narrowing. An example of semantic extension is the use of the verb correr ‘to run’ in the speech of Cuban immigrants in the US, where, under the influence of English, the verb has acquired the metaphoric meaning of running for office, e.g. correr para gobernador ‘to run for governor’ (Otheguy & Garcia 1988). An example of semantic narrowing is seen in the use of emotion adjectives by Russian L2 users of English in my studies (Pavlenko 2002, 2003a): (3) a.
но ещё была очень как бы... зла на кого-то ‘and also [she] was very somehow... angry at someone’ (an appropriate lexical choice would be сердита ‘cross, angry at the moment’)
b. она видит, что её дочь не очень такая... счастливая ‘she sees that her daughter is not that... happy’ (an appropriate lexical choice would be недовольная ‘dissatisfied’ or грустная ‘sad’)
Russian has three translation equivalents of angry – serdityi ‘cross’, ‘angry at the moment’, zloi ‘malicious’, ‘very angry’, ‘mean’ (typically used as a personality characteristic), and gnevnyi ‘irate’, ‘in wrath’ – each more intense than the preceding one. An analysis of stories elicited by the same stimuli from Russian monolinguals and Russian L2 users of English showed that in descriptions of the same woman, monolinguals favoured the term serditaia ‘cross/Fem’, while some bilinguals inappropriately used the short adjective zla ‘malicious’, ‘mean/ Fem’ (the closest formal translation equivalent of ‘angry’). In the second example, the participant similarly misused the adjective (ne)schastlivaia ‘unhappy’ which would have been appropriate in English but not in Russian. While in English ‘unhappy’ could mean ‘temporarily displeased’ or ‘dissatisfied’, the Russian translation equivalent typically refers to a more permanent state of unhappiness, sadness, or continuous lack of luck. In both cases, the speakers appear to have collapsed the distinctions between two Russian translation equivalents of an English word and used literal translations of the English terms, inappropriate semantically in these contexts. Intriguing evidence of restructuring in collocational knowledge is offered by Laufer (2003). In 40% of the cases her study participants, Russian immigrants in Israel, were unable to identify collocations based on the L2 patterns as incorrect. For instance, they now linked the words telefon ‘telephone’ or televizor ‘TV’ to the word ‘close’, since in Hebrew one ‘closes the telephone or TV’
Aneta Pavlenko
while in Russian you put the receiver down (polozhit’ trubku) and turn off the TV (vykluchit’ televizor). To date restructuring has been documented mostly in on-line production and, in a few cases, in metalinguistic judgments. Jarvis’ (2003) study offers strong evidence in favour of triangulation of elicitation methods. The Finnish L2 user of English in his study was found to judge some of the structures and uses as inappropriate for conventional Finnish – yet she was heard using the same structures later the same day. Consequently, I suggest that to count as evidence of L1 attrition, restructuring has to be demonstrated in both production and perception/comprehension in at least two ways: (a) through grammaticality or semantic judgments distinct from those offered by monolingual L1 speakers; (b) through processing delays in production and comprehension difficulties. It is also important to examine whether restructuring, and in particular semantic narrowing and extension, are limited to bilingual and L2 contexts or if the L2 users would resort to these rules and items in monolingual L1 contexts as well.
. Convergence The third process identified in the CI framework is convergence. Convergence of L1 and L2 values, distinctions, or boundaries, has been best documented in the areas of phonology and conceptual representation. The research paradigm most commonly employed in the studies of bilinguals’ phonology is perception and production of stop voice onset time (VOT), which is a sufficient acoustic cue for distinguishing between initial stop consonants in many languages. Even though VOT is just one feature characterising one’s production and perception, several studies have confirmed that VOT values are closely correlated with overall judgments on the ‘nativeness’ or ‘accentedness’ of one’s speech (Flege & Eefting 1987; Major 1992). Convergence of L1 and L2 VOT values was documented by Major (1992, 1993) in the speech of English-Portuguese bilinguals and by Flege (1987) in the speech of French-English bilinguals. Where concepts are concerned, convergence entails a creation of a unitary concept, domain, or system, distinct from both the L1- and L2-based. For instance, Ervin-Tripp (1961) found that colour categories used by NavahoEnglish bilinguals differed systematically from the monolingual norms in the respective languages. Her findings were confirmed in a study by CaskeySirmons and Hickerson (1977) which showed that speakers of several languages who learned English in adulthood mapped larger total colour areas,
L2 influence and L1 attrition in adult bilingualism
had less stable colour category boundaries and more variable category foci than monolingual speakers. At present, more information is needed to decide whether convergence is a temporary or permanent phenomenon, exhibited only in L2 contexts, or also in the L1 environment. If the effects are stable and exhibited systematically in a variety of contexts, they may indeed be evidence of L1 attrition.
. Shift The fourth process specified by the CI framework is a shift which is seen as a move away from L1 structures, boundaries or values toward those specified by the L2. This phenomenon is most often evidenced in phonology and in the mental lexicon. In the study of phonology it has been documented by Williams (1979, 1980) who found L2 influence on L1 in both production and perception. In her studies Spanish teenagers shifted from a Spanish-like to an English-like manner in producing both English and Spanish word-initial tokens. What is particularly interesting and important here is that these teenagers also exhibited a gradual shift from a Spanish-like to an English-like pattern in perception, evident in their labelling the VOT series: all discrimination peaks were found close to the area of the English contrast. In the mental lexicon, studies found evidence of shift in typicality judgments of Spanish L2 users of Hebrew (Shimron & Chernitsky 1995), lexical associations of English L2 users of Spanish (Grabois 1999), and linguistic framing of emotions by Russian L2 users of English (Pavlenko 2002, 2003a; Pavlenko & Jarvis 2002). And in the study of pragmatics, Tao and Thompson (1991) documented a shift in the backchanneling behaviour of Mandarin L2 users of English who made extensive use of American English backchannel strategies in conversations with Mandarin-speaking interlocutors. Once again, however, these studies were conducted in the L2 context. To decide whether they offer evidence of a temporary shift or of L1 attrition, we need to examine metalinguistic judgments of the bilingual participants and their verbal behaviour in a monolingual L1 context.
. Attrition Since I have argued that none of the four processes above could be unproblematically considered evidence of L1 attrition, it is reasonable to ask: what
Aneta Pavlenko
could be taken as evidence? In the view adopted here, L2 influence on L1 is seen as an either temporary or permanent phenomenon which is exhibited predominantly in the L2 or bilingual contexts or in the presence of bilingual interlocutors, but does not signify that corresponding L1 lexical items, rules, or distinctions have been permanently lost. In turn, L1 attrition involves a more or less permanent restructuring, convergence, or loss of previously available phonological and morphosyntactic rules, lexical items, concepts, classification schemas, categorical distinctions, and conversational and narrative conventions, exhibited not only in the L2 but also in a monolingual L1 context, and not only in production but also in perception and comprehension. In phonology, this means, first of all, that the L2 user would no longer be perceived as a native speaker of his or her L1 when interacting with monolingual speakers of the L1. Anecdotal evidence suggests that this may indeed be the case for some adult bilinguals. For instance, in Latomaa’s (1998) study of English L2 users of Finnish, one of the American informants complained: “After five years here in Finland I went back to the States and the neighbours asked which country I am from” (p. 65). Similarly, in my study (Pavlenko 2003a), an 18 year old Russian L2 user of English who had arrived in the US four years previously, mentioned that during his annual summer visits to Moscow people tease him about his Russian: “they say that I have an accent in Russian (laughs)... no, I don’t make mistakes, but... I have a bit of an accent...” Hypothetically, attrition in phonology should also imply that some bilinguals – just like foreign language learners – will exhibit perception/decoding delays and problems interpreting L1 intonation patterns. Since L1 attrition of morphosyntactic competence is widely addressed elsewhere in this volume, I will only mention it briefly, stating once again that convincing evidence of such attrition should include studies of L2 users’ spontaneous and elicited production in monolingual L1 contexts and studies of L2 users’ perception, comprehension, and metalinguistic judgments of L1 rules and constraints. In the mental lexicon, evidence of attrition, as mentioned earlier, would encompass L2-influenced use of L1 words and expressions in monolingual L1 contexts (given availability of L1 translation equivalents) and difficulties in perception, comprehension, and metalinguistic judgments of the use of L1 lexical items. Finally, in pragmatics, L1 attrition would involve a loss of pragmatic competence, that is, the inability to interpret particular illocutionary intentions. Anecdotal evidence suggests that this may be the case for some bilinguals. For instance, Kyoko Mori (1997), a Japanese woman who has lived in the US for 20 years since the age of 20, complains that while in the Midwest she can per-
L2 influence and L1 attrition in adult bilingualism
fectly well distinguish symbolic invitations from the real ones, this is not the case in Japan: (4) In Japan, there are no clear-cut signs to tell me which invitations are real and which are not. People can give all kinds of details and still not expect me to show up at their door or call them from the train station. I cannot tell when I am about to make a fool of myself or hurt someone’s feelings by taking them at their word or by failing to do so. (p. 10)
Another area of loss in pragmatic competence, identified by Mori (1997), concerns the level of politeness in conversation: (5) I can only fall silent because thirty seconds into the conversation, I have already failed at an important task: while I was bowing and saying hello, I was supposed to have been calculating the other person’s age, rank, and position in order to determine how polite I should be for the rest of the conversation. (p. 11)
Similar personal insights are available on the subject of L2 influence on L1 in rhetoric. These insights come from two distinguished linguists, Ulla Connor (1999), a renowned expert on L1 transfer in L2 rhetoric, and Suresh Canagarajah (2001), a well-known scholar working on acquisition of second language and literacy. Connor (1999) tells how she went back to Finland, her native country, after having spent several years in the United States, writing in English. In Finland, she was supposed to co-author a Finnish language manual on writing grant proposals in English. She found the process of collaboration extremely difficult, as her views on organisation of the material differed from those of her Finnish co-authors. When the first draft of the manual appeared incoherent to her, Connor attempted to argue with her colleagues that the main point should be in the beginning of the paragraph, followed by examples supporting the point. At the end, however, the researcher realised the extent of her own Americanisation: the changes she suggested were inappropriate for Finnish rhetoric style. Not surprisingly, no major changes based on her suggestions were included in the final version of the Finnish booklet. Similarly, Canagarajah (2001) recalls that when he returned to Sri Lanka and attempted to write articles in the local vernacular Tamil, his native language, his papers, structured in the manner of an American expository essay – announcing the thesis, delineating the steps of his argument, and proving the points – disappointed and angered his colleagues. Even some of his own students told him that “the introductory paragraph had sounded a bit too pompous and over-confident” (Canagarajah 2001: 31) and reminded him that
Aneta Pavlenko
in the vernacular tradition one opens the paper with a brief confession of one’s limitations and a praise for the knowledge of the audience. Finally, my own studies documented conceptual attrition that is slowly taking place in the mental lexicon of Russian-English bilinguals (Pavlenko 2002, 2003a, b). In the interview below, Julia, a 25 year old female study participant who arrived in the US 5 years earlier, complains that her Russian is no longer adequate to describe some of her thoughts and feelings: (6) Что я хочу сказать? Что я чувствую себя все более и более clumsy, неловко, когда я говорю по-русски, мне трудно подбирать слова, и очень часто мне кажется, что мне легче выразить это по-английски, даже не потому что я не знаю слов по-русски, а потому что выражение, которое я хочу употребить в данной ситуации, оно настолько английское, и оно настолько... например... что-то очень трудно сделать, да?.. и я не знаю получится у меня или нет, я говорю себе сама по-английски “I can make it”... почему – потому что по-русски это не звучит, это не то же самое, и нету такой... уверенности, что ты, да, ты сможешь это сделать, нет такой бравады, нет такого спокойствия, ещё чего-то такого, нет вот этого, нету... ‘What do I want to say? That I feel more and more clumsy, uncomfortable, when I speak Russian, it is difficult for me to choose words, and very often it seems to me that it is easier for me to express something in English, not even because I do not know Russian words, but because the expression I want to use in a particular situation, it is so English, and it is so... for example... something is very difficult to accomplish, yes? ... and I do not know if I will be able to succeed or not, and I say to myself in English “I can make it”... why – because in Russian it doesn’t sound right, it is not the same, and there is no such... confidence, that you, yes, you will be able to do this, there is no such bravado, no such calm, and something else, it is not there, no... ’
For me, this quote functions as a reminder that the inability to express particular meanings in one’s native language is not purely a psycholinguistic phenomenon, rooted in the disuse of the L1 or the overuse of the L2. It is not attrition that prevents Julia from expressing herself in Russian, but the fact that certain nuances and shades of meaning are not translatable. This neoWhorfian view of bilingualism and thought (Pavlenko, in press) sees the interaction between two or more languages as a complex phenomenon where linguistic and conceptual enrichment and transformation may be present side by side with at-
L2 influence and L1 attrition in adult bilingualism
trition. It is up to future inquiry to see if the two phenomena can be sufficiently disambiguated.
References Andrews, D. (1999). Sociocultural Perspectives on Language Change in Diaspora: Soviet Immigrants in the United States. Amsterdam and Philadelphia: John Benjamins. Bot, K. de, & Clyne, M. (1994). “A 16-year longitudinal study of language attrition in Dutch immigrants in Australia”. Journal of Multilingual and Multicultural Development, 15 (1), 17–28. Boyd, S., & Andersson, P. (1991). “Linguistic change among bilingual speakers of Finnish and American English in Sweden: Background and some tentative findings”. International Journal of the Sociology of Language, 90, 13–35. Canagarajah, S. (2001). “The fortunate traveler: Shuttling between communities and literacies by economy class”. In D. Belcher & U. Connor (Eds.), Reflections on Multiliterate Lives (pp. 23–37). Clevedon: Multilingual Matters. Caskey-Sirmons, L., & Hickerson, N. (1977). “Semantic shift and bilingualism: Variation in color terms of five languages”. Anthropological Linguistics, 19, 358–367. Connor, U. (1999). “Learning to write academic prose in a second language: A literacy autobiography”. In G. Braine (Ed.), Non-native Educators in English Language Teaching (pp. 29–42). Mahwah, NJ: Lawrence Erlbaum. Cook, V. (1991). “The poverty of the stimulus argument and multicompetence”. Second Language Research, 7, 103–117. Cook, V. (Ed.). (2003). Effects of the Second Language on the First. Clevedon: Multilingual Matters. Ervin-Tripp, S. (1961). “Semantic shift in bilingualism.” American Journal of Psychology, 74, 233–241. Flege, J. (1987). “The production of “new” and “similar” phones in a foreign language: Evidence for the effect of equivalence classification.” Journal of Phonetics, 15, 47–65. Flege, J., & Eefting, W. (1987). “Cross-language switching in stop consonant perception and production by Dutch speakers of English.” Speech Communication, 6, 185–202. Grabois, H. (1999). “The convergence of sociocultural theory and cognitive linguistics: Lexical semantics and the L2 acquisition of love, fear and happiness”. In G. Palmer & D. Occhi (Eds.), Languages of Sentiment: Cultural constructions of emotional substrates (pp. 201–233). Amsterdam and Philadelphia: John Benjamins. Grosjean, F. (1998). “Studying bilinguals: Methodological and conceptual issues.” Bilingualism: Language and Cognition, 1, 131–149. Håkansson, G. (1995). “Syntax and morphology in language attrition: A study of five bilingual expatriate Swedes.” International Journal of Applied Linguistics, 5 (2), 153–171. Jarvis, S. (2003). “Probing the effects of the L2 on the L1: A case study”. In V. Cook (Ed.), Effects of the Second Language on the First (pp. 81–102). Clevedon: Multilingual Matters.
Aneta Pavlenko
Latomaa, S. (1998). “English in contact with “the most difficult language in the world”: The linguistic situation of Americans living in Finland.” International Journal of the Sociology of Language, 133, 51–71. Laufer, B. (2003). “The influence of L2 on L1 collocational knowledge and on L1 lexical diversity in free written expression”. In V. Cook (Ed.), Effects of the Second Language on the First (pp. 19–31). Clevedon: Multilingual Matters. Major, R. (1992). “Losing English as a first language.” The Modern Language Journal, 76, 190–208. Major, R. (1993). “Sociolinguistic factors in loss and acquisition of phonology”. In K. Hyltenstam & A. Viberg (Eds.), Progression and Regression in Language: Sociocultural, neuropsychological and linguistic perspectives (pp. 463–478). Cambridge: CUP. Mori, K. (1997). Polite Lies: On being a woman caught between two cultures. New York, NY: Henry Holt. Otheguy, R., & Garcia, O. (1988). “Diffusion of lexical innovations in the Spanish of Cuban Americans”. In J. L. Ornstein-Galicia, G. K. Green, & D. Bixler-Marquez (Eds.), Research Issues and Problems in U.S. Spanish: Latin American and Southwestern varieties (pp. 203– 242). Brownsville, TX: University of Texas. Otheguy, R., & Garcia, O. (1993). “Convergent conceptualisations as predictors of degree of contact in U.S. Spanish”. In A. Roca & J. Lipski (Eds.), Spanish in the U.S.: Linguistic contact and diversity (pp. 135–154). Berlin: Mouton de Gruyter. Pavlenko, A. (2000). “L2 influence on L1 in late bilingualism.” Issues in Applied Linguistics, 11 (2), 175–205. Pavlenko, A. (2002). “Bilingualism and emotions.” Multilingua, 21 (1), 45–78. Pavlenko, A. (2003a). “I feel clumsy speaking Russian”: L2 influence on L1 in narratives of Russian L2 users of English”. In V. Cook (Ed.), Effects of the Second Language on the First (pp. 32–61). Clevedon: Multilingual Matters. Pavlenko, A. (2003b). “Eyewitness memory in late bilinguals: Evidence for discursive relativity.” International Journal of Bilingualism, 7 (3), 257–287. Pavlenko, A. (In press). “Bilingualism and thought.” In A. De Groot & J. Kroll (Eds.), Handbook of Bilingualism: Psycholinguistic Approaches. Oxford: Oxford University Press. Pavlenko, A., & Jarvis, S. (2002). “Bidirectional transfer.” Applied Linguistics, 23 (2), 190– 214. Py, B. (1986). “Competence and attrition in the native language of immigrants”. In E. Kellerman & M. Sharwood Smith (Eds.), Crosslinguistic Influence in Second Language Acquisition (pp. 163–172). Oxford: Pergamon Press. Schmid, M. S. (2002). First Language Attrition, Use, and Maintenance: The case of German Jews in Anglophone countries. Amsterdam and Philadelphia: John Benjamins. Seliger, H. W., & Vago, R. M. (1991). “The study of first language attrition: An overview”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 3–15). Cambridge: CUP. Shimron, J., & Chernitsky, R. (1995). “Typicality shift in semantic categories as a result of cultural transition: Evidence from Jewish Argentine immigrants in Israel”. Discourse Processes, 19, 265–278.
L2 influence and L1 attrition in adult bilingualism
Tao, H., & Thompson, S. (1991). “English backchannels in Mandarin conversations: A case study of superstratum pragmatic ‘interference.’ ” Journal of Pragmatics, 16, 209–223. Waas, M. (1996). Language Attrition Downunder: German Speakers in Australia. Frankfurt: Peter Lang. Williams, L. (1979). “The modification of speech perception and production in secondlanguage learning.” Perception and Psychophysics, 26, 95–104. Williams, L. (1980). “Phonetic variation as a function of second-language learning”. In G. Yeni-Komshian, J. Kavanagh, & C. Ferguson (Eds.), Child Phonology. Volume 2. Perception (pp. 185–215). New York, NY: Academic Press.
A sociocultural approach to language attrition Antonio F. Jiménez Jiménez California State University, Channel Islands
Introduction Traditionally, the field of applied linguistics, borrowing from work in linguistics and psychology, has tended to set up dichotomies as a way of defining its scope of study and its research methodologies. Among the most pervasive are langue/parole, competence/performance, input/output, form/meaning, acquisition/use etc. One of the prevailing dichotomies in the field is constituted by the dualism between the Internal (or the self, the individual, the cognitive. . .) and the External (or the other, the society, the community. . .). Research on language attrition has followed this bicephalous tradition: on the one hand, a large number of studies has concentrated on investigating the internal side of the dichotomy, be it from a psycholinguistic (e.g. de Bot 1999; Hansen 1999; Hedgcock 1991; Olshtain 1989), neurolinguistic (e.g. Berko-Gleason 1993; Obler 1982; Yoshitomi 1992), or Universal Grammar approach (e.g. Håkansson 1995; Platzack 1998; Sharwood Smith & van Buren 1991; Sorace 2000; Toribio 2001). On the other hand, other researchers have focused on examining the external factors, such as Hulsen (2000), Schmid (2002), Waas (1996), Ya˘gmur (1997). This article explores an innovative way of approaching attrition based on Sociocultural Theory (SCT) by linking social participation (the external side of the coin) to linguistic production and psychological functioning (the internal side). Its primary objective is to provide an overview of the main tenets of the theory and how it accounts for the phenomenon of language attrition through its particular lenses. In order to do this, the present paper has been divided into three parts. The first one introduces some of the theoretical foundations of Sociocultural Theory while the second one explores the role of language under
Antonio F. Jiménez Jiménez
this perspective and provides a new interpretation and definition of language attrition based on, and consistent with, its theoretical roots. The third part discusses how language attrition can be assessed following this reinterpretation of the phenomenon. A special emphasis is placed on the analysis of communicative strategies as a way of evaluating language attrition from this theoretical perspective.
.
Sociocultural Theory
Even though Sociocultural Theory (SCT) was not born as a linguistic theory but as a theory of cognitive development, for the last twenty years, an increasing number of linguists such as Ahmed (1994), De Guerrero (1990), Donato (1994), Frawley (1985, 1997), McCafferty (1994, 1998), or Ushakova (1994), to mention just a few, have been applying the tenets of this theory to the study of L2 acquisition. Although it is not the main objective of this paper to provide a thorough characterisation of Sociocultural Theory, a discussion of language attrition under this perspective requires a preliminary understanding of some of the main tenets that SCT is based on. The work of Vygotsky, founder of Sociocultural Theory, started to develop in response to Piaget’s theory of cognitive development. Piaget argued that children possess an innate and inherent ability to progressively construct a more logically complex and abstract cognitive system as they grow older. Conversely, Vygotsky (1978), together with his colleagues A. R. Luria, and A. N. Leont’ev, maintained that this innate component does not itself determine the child’s mental development. Instead, they maintained that higher forms of cognitive development first originate in the social world, in interaction between individuals (interpsychological level) before they are internalised (intrapsychological level). In other words, there is a process of internalisation by which the individual incorporates, in a transformative process, the higher psychological functions that already exist socially into his own developing cognitive system. These higher cognitive functions include voluntary attention, logical memory, conceptual thought, problem solving, learning, voluntary inhibitory and disinhibitory faculties, and the planning, execution, and monitoring of mental processes. Each individual’s functional system is shaped essentially by his experiences and interactions with the surrounding community so it should be expected, then, that different communities provide different functional system to their members.
A sociocultural approach to language attrition Mediating artefact
Subject
Object
Figure 1. Artefact-mediated object-oriented activity system
It is important to specify at this point that the process of internalisation does not imply a direct transfer of these higher cognitive functions from the community to the individual. Instead, this process is mediated by the use of tools. These tools or artefacts establish an indirect relationship between the individual and the social environment. Tools can have a physical or symbolic nature. Physical tools comprise those material artefacts that mediate our relationship with the external physical world and they include hammers, computers, pencils, bulldozers, books, etc. In the same way as human beings make use of these material tools to interact with the physical environment, symbolic tools are employed to mediate and regulate our relationships with others and with ourselves. Among all the symbolic tools, such as numbers, schemes, arithmetic systems, diagrams, graphs, mnemonic devices, music, and art, language is considered as the main mediating artefact through which speakers engage in social interaction. Figure 1 graphically represents Vygotsky’s model of mediated action. Both physical and symbolic tools have been culturally constructed over time. They are passed on from generation to generation, and in this process, they are constantly modified. Change, then, is a social process and our relationship with the external world, with others and with ourselves changes as physical and symbolic tools are transformed over time. Language, as any other tool, also stands between the subject and the reality. Language is appropriated in the process of social interaction and as it is being internalised, it shapes the way we think about the world and our relationship to it. According to Vygotsky, language activity (which may also entail written as well as spoken communication) does not serve the function of transmitting information from one speaker to the other. Instead, it is the tool through which individuals regulate each other and ourselves as we engage in different types of social and psychological activities. Thus, the concept of “activity” becomes pivotal for the understanding of the theory and its notion has been expanded since its original conception by Vygotsky. Initially, activity developed around the idea of mediation (as de-
Antonio F. Jiménez Jiménez
picted in Figure 1), that is, human activity is mediated by culturally constructed artefacts. Vygotky’s interpretation of activity is undertaken by a human agent (subject) who is motivated to intentionally move toward a particular goal or object and does so in conjunction with mediational tools or artefacts. Leont’ev (1981) differentiates between collective and individual activity. In his own words, “only through a relation with other people does man relate to nature itself, which means that labour appears from the very beginning as a process mediated by tools (in the broad sense) and at the same time mediated socially” (Leont’ev 1981: 208). He incorporated the idea of division of labour into the picture, with his famous example of the primeval collective hunt, in which he distinguishes between individual action and a collective activity: A beater, for example, taking part in a primeval collective hunt, was stimulated by a need for food or, perhaps, a need for clothing, which the skin of the dead animal would meet for him. At what, however, was his activity directly aimed? It may have been directed, for example, at frightening a herd of animals and sending them toward other hunters, hiding in an ambush. That, properly speaking, is what should be the result of the activity of this man. And the activity of this individual member of the hunt ends with that. The rest is completed by the other members. This result, i.e. the frightening of game, etc., understandably does not in itself, and may not, lead to satisfaction of the beater’s need for food, or the skin of the animal. What the processes of his activity were directed to did not, consequently, coincide with what stimulated them, i.e., did not coincide with a motive of his activity; the two are divided from one another in this instance. Processes, the object and motive of which do not coincide with one another, we shall call “actions”. We can say, for example, that the beater’s activity is the hunt, and the frightening of the game his action. (Leont’ev 1981: 210)
This excerpt also depicts Leont’ev’s distinction between activity, action and operation, which became the basis of his three-level model of activity. According to this model, activities are based on three levels, the first being the level of motives. This level, also referred to as the level of activity, is driven by an object-related motive and explains why something is done. Individuals engage in different types of activities motivated by either biological or culturally constructed needs. The second level, the level of action, is driven by a conscious goal and accounts for what is done. The third level is the level of operations. It explains how something is done and it comprises the physical and temporal circumstances under which activities take place. These three levels of activity are outlined in Figure 2.
A sociocultural approach to language attrition
Activity level: Motives Action level: Goals Operation level: Conditions Figure 2. Three levels of activity Instruments
Subject
Rules
Object
Community
Outcome
Division of Labor
Figure 3. Extended version of activity system
Leont’ev’s ideas of activity was graphically depicted by Engeström (1987), who, taking Vygotsky’s original model, expanded the triangle in its basis including three new elements, namely the community, the rules, and the division of labour (see Figure 3). Rules regulate actions and interactions within the activity system. The division of labour discusses how tasks are divided horizontally between community members as well as referring to any vertical division of power and status and the community is comprised by one or more people who collaboratively participate in the activity. Most recent interpretations of the concept of ‘activity’ are aiming to incorporate cultural diversity into the picture (Cole 1988). Previous models of activity did not take into account interactions and activities between individuals from different cultures and traditions. However, in recent decades, intercultural interchanges have exploded with new means of communication (satellite systems, email, internet, etc.) and transportation. Thus, the notion of “activity” needs to be expanded again to integrate multiculturalism and new forms of dialogue, cooperation, and interaction into the system.
Antonio F. Jiménez Jiménez
. A Sociocultural account for language attrition . The role of “language” As mentioned before, under Sociocultural Theory, language is the tool through which the individual engages in different types of social and psychological activities. In the theory, language is intrinsically related to the process of thinking. Vygotsky (1986) argues that thought and language have different roots in humankind. As he puts it, “in the speech development of the child, we can with certainty establish a preintellectual stage, and in his thought development, a prelinguistic stage” (1986: 83). That is, thinking and speaking follow their own developmental lines. In Vygotsky’s view, “[p]rogress in thought and progress in speech are not parallel. Their two growth curves cross and recross. They may straighten out and run side by side, even merge for a time, but they always diverge again” (ibid.: 68). However, at a certain moment during the ontogenesis of the child, these lines meet and merge together initiating a new form of behaviour. This is the crucial moment that differentiates apes from humans, i.e., when thought becomes verbal and speech becomes rational. At first, the child uses language in superficial social interaction. However, at some point, social speech moves towards egocentric speech and becomes silent, representing the essence of conscious mental activity, that is, inner speech. Through internalisation children gain increasing control over their own behaviour and mental processes, becoming, in sociocultural terms, selfregulated. As mentioned before, this internalisation process does not occur from the social to the individual plane in a direct way, but it is mediated by artefacts, language (or more specifically speaking activity) being the most important one. Thus, when speaking and thinking start to merge in terms of one providing resource to the other, thought becomes verbal, speech becomes rational, and speech becomes essential in forming thought. Thus, children in the process of acquiring their L1, as well as individuals in the process of acquiring an L2 “gain increasing control over the mediational means made available by [the surrounding culture], including language, for interpersonal (social interaction) and intrapersonal (thinking) purposes” (Lantolf 2000: 6). This process involves a series of stages (applicable to both L1 and L2 learners) that are described in the following paragraph. During the language acquisition process, both L1 and L2 novices experience communicative challenges due to the learners’ incomplete knowledge of the language (i.e., in sociocultural terms, novices are object-regulated by the language). In order to gain control over the language (and over the commu-
A sociocultural approach to language attrition
nicative task they are engaged in), novices may rely on the help of others (i.e., becoming other-regulated). This assistance needs to take place within the learners’ zone of proximal development. This is a key element in the theory and it is defined as the “distance between the actual developmental level as determined by independent problem solving and the level of potential development as determined through problem solving under adult guidance or in collaboration with more capable peers” (Vygotsky 1978: 86). Through interaction, novices start to develop into more knowledgeable and independent users of the language (i.e. achieving self-regulation). Thus, acquiring a language means to move from object- and other-regulation to self-regulation. Self-regulation is a relative phenomenon and its full understanding needs some sharpening. On the one hand, self-regulation is activity-dependent, that is, a novice learner can achieve self-regulation in a certain type of task and this does not necessarily mean that the learner has gained self-regulation in all types of tasks (Lantolf & Appel 1994: 12). On the other hand, the fact that a learner has achieved self-regulation in a certain task does not imply that selfregulation can be maintained in the task at all times. At certain points, if a task becomes particularly difficult, even experts may turn to previous stages of development and become other-regulated (seeking help from other more knowledgeable individuals) or even object-regulated (by using other artefacts available in the culture, such as a computer, a dictionary, a calculator, a book, etc.). This quality of mental activity has been coined by Frawley and Lantolf (1985) as the principle of continuous access.1 In summary, it can be concluded that, in sociocultural terms, language acquisition is based on the idea of control, which is about movement from objectand other-regulation to self-regulation. Self-regulation then is the ultimate objective of the individual engaged in a given communicative task. However, if the task becomes sufficiently difficult, self-regulation can be lost and the speaker will need to go back to previous stages of development (that is, becoming other- or object-regulated). . Language attrition Consequently, taking into account the above information, language attrition, under Sociocultural Theory, entails the loss of the once attained level of selfregulation in a language (L1 or L2) in a particular activity and the momentary return to a previous stage of object- and/or other-regulation. As in the case of language acquisition, language attrition should be also regarded as task specific. Thus, language attrition occurs when a speaker experiences an increasing
Antonio F. Jiménez Jiménez
loss of self-regulation in a given task over time. Or, in other words, the speaker experiences an increasing need to reaccess previous levels of development (object- or other-regulation). There are only two studies, to the best of my knowledge, that address the attrition phenomenon from a sociocultural perspective. One of these studies (Coughlan 1995) examines the acquisition and the attrition of a child’s L2 within the framework of Activity Theory. The study examines a number of phone calls between a child, named Vovó, and his grandmother in Portuguese, the child’s L2. During these interactions, the grandmother scaffolds the child, enabling him to express himself in a way he would not be able to do alone. During the time these phone calls were taking place, Vovó experienced linguistic development in terms of lengthened conversations and less frequent use of English (the child’s L1). However, this trend was interrupted for a month, during which period the grandmother left the country to visit Portugal. After this break, when the grandmother returned, the phone calls continued, but both Vovó and the grandmother abandoned the use of Portuguese and, with the exception of greetings and leave-takings, the calls were conducted in English. This change in the activity system led to L2 attrition. However, at the same time, this change had a dramatic impact on Vovó’s role in the conversations, as he now could engage more independently in the communicative activity, that is, without the assistance of the parents or the grandmother, taking more responsibility in the activity. Coughlan maintains that the loss of Portuguese is beneficial for the child as it allows him to participate more meaningfully in the conversations with his grandmother and, similarly, it benefits the grandmother as well as they can discuss other substantive issues without paying that much attention to the language barriers. The author maintains that changes in learners’ production over time may be manifestations of changes in the activity systems. These changes at the linguistic level may not find an explanation if the focus of the study is L2 interaction alone. Lastly, Coughlan concludes that language acquisition and loss should be regarded as by-products of participation in a given social group’s range of activities. Only by investigating these activities will we understand how language change operates. The other study that looks at language attrition from a Vygotskyan perspective is McCreary (1985) who examines both the L2 acquisition and the attrition of Japanese in two children in terms of regulation before, during, and after a two-month stay in Japan. The author examines the use of otherregulation and object-regulation when the children do not know or cannot express their own ‘self ’ in Japanese. During the acquisition period, that is, when the children were in Japan, they improved their ability to speak Japanese, which
A sociocultural approach to language attrition
is explained by the change of context and the children’s use of (other-, object-, and self-regulatory) strategies in the process of gaining control over their new linguistic environment. During the incubation period, that is, upon the children’s return to their home environment in the United States, the children’s control over Japanese decreased as they had to adjust to the new social and strategic environment. The author maintains that “self-regulation is lost first because it requires cognitive ability in the language that calls for a syntactic access to many forms, questions, commands, cohesive markers, and a variety of verbs and predicates, to name just a few” (1985: 168). Then, this loss of selfregulations is followed by a loss of object-regulation (i.e. names of objects) and other regulatory forms that are rarely needed in the USA. However, the author notices that there are certain remains of both object- and other-regulated speech that are maintained and that probably will not be lost.2
. Assessing language attrition under SCT Language attrition, in this paper, has been defined as the loss of the once attained level of self-regulation of a language (L1 or L2) in a particular activity and the momentary return to a previous stage of object- and/or otherregulation. However, how can we assess this shift of regulation? The following paragraphs outline some basic elements needed in order to be able to assess language attrition in terms of regulation across time. First of all, the attainment of self-regulation is task dependent. That is, a speaker who achieves a high level of control of a certain communicative activity does not necessarily translate into a control of other types of activities. Thus, a study that assesses change at the regulatory level needs to have a longitudinal pretest-posttest research design, in which the task is constant across time. The type of task that is especially appropriate for this type of analysis is the elicitation of free speech in the framework of a meaningful and goal-oriented task. If we want to access the control that a certain individual has about the language in question, we have to think in terms of activity and design a task that can relate to the types of activities that this individual usually undertakes in his interactions with the community. These tasks should be as concrete and meaningful to the participant as possible. Some examples of tasks may include the writing of a letter to a relative, friend, boss, etc., a role-play activity imitating a situation that the individual may have lived when immersed in the community, leaving a message on a friend’s answering-machine, a picture/video-based story telling, a reading recall, a listening recall, etc. The use of the same type of task over time
Antonio F. Jiménez Jiménez
will provide a comparative paradigm that allows the researcher to establish the level of self-regulation achieved by each participant across time. However, how can it be determined that there has been a loss of self-regulation in a given task? Loss of self-regulation can be clearly observed in the analysis of the communicative breakdowns experienced by any given speaker and his use of compensatory strategies in order to solve these communicative problems. Breakdowns may be motivated either by the individual’s novelty in completing the task (or, in other terms, because the task is too difficulty for the speaker), by a problem with the language used, or by a conceptual misunderstanding/error in the speaking activity. For example, a novice speaker defending a paper at a conference in his L1 may lose self-regulation due to reasons such as lack of knowledge about the appropriate script followed in this type of medium, nervousness, uncertainty in the exposition of ideas, etc. This lack of control over the task can surface in his speech through hesitations, unfinished sentences, repetitions, rephrasings, etc. On the other hand, the same speaker presenting a paper in his non-dominant (L2) language may not only encounter problems with the difficulty of the task, but also with his incomplete knowledge of the language of communication. Consequently, the frequency of use of strategies in order to solve these communicative problems is likely to increase. Thus, the study of strategies as indicators of language attrition should take into account the origin of the communicative breakdown. As mentioned, there are three possibilities: task difficulty, language problem, or conceptual problem. If the focus is on the linguistic difficulties of the speaker during the communicative activity, the researcher has to control that task difficulty and conceptual misunderstandings are not going to contribute to a higher production of compensatory strategies. In other words, a study on language attrition under this perspective has to make sure that task difficulty remains constant in the different measurements over time and that the speaker completely understands the conceptual representations involved in the completion of the task. If the researcher keeps the difficulty of the task stable and makes sure that the speaker has a full conceptual understanding of the communicative task, the communicative production of the speaker can be compared across time being certain that any change observed in the speaker’s use of communicative strategies will be due to problems at the linguistic level. In language attrition studies, a higher frequency in the use of communicative strategies indicates an increase in communicative breakdowns. The use of communicative strategies is an indication of the cognitive processes that are taking place in the speaker’s head and they have been largely studied as such indicators in L1 and L2 acquisition literature (Faerch & Kasper
A sociocultural approach to language attrition
1983; Kasper & Kellerman 1997; Poulisse 1999; Tarone 1977, 1980). There has been extensive debate on the topic of communication strategies, especially on its definition and on the taxonomies created of its multiple manifestations (for a critical review of CS see Lantolf & Frawley 1985). A large number of strategies has been identified as being used by speakers when trying to overcome a communication problem and equally large is the number of taxonomies that have been created in the process of categorising these strategies. The constant reformulation of categories has resulted in a clear divergence in the way researchers approach data analysis. I believe that this permanent recategorisation of communicative strategies shows that strategies are not rigid identities that pertain to a single and exclusive category. Instead, they are fluid manifestations that surface in the speech differently depending not only on linguistic factors (i.e. inability to produce the desired linguistic elements) but also on other contextual concomitant factors (such as the type and the difficulty of the task, the kind of relationship created between the interlocutors, the rules that govern the task, etc.). Lantolf and Frawley (1985) propose a functional typology of communicative strategies in reaction to more orthodox taxonomies. This typology is based on the notion of regulation and on the idea that speakers tend to achieve linguistically-based self-regulation at all times during the communicative act. However, this is not always achievable in an efficient and timely manner (taking into account the rules of communication). Thus, a speaker, faced with a communicative problem, can choose three different forms of regulatory strategies in order to regain control of the activity. These strategies are presented in the following paragraphs. It should be remarked here that even though strategies have been associated with a main regulatory function, these associations are by no means rigid. They solely indicate that these strategies usually serve the regulatory function to which they have been associated. Strategies constitute fluid identities, and the use of one of these strategies may have indeed different regulatory functions. Self-regulatory strategies are attempts to maintain the self during the communicative breakdown or, in other words, “when maintaining self in communicative act regulates choice of strategy” (Lantolf & Frawley 1985: 153). That is, the speaker employs intrapsychological strategies intended to solve the communicative breakdown. Within this category, we find three types of selfregulatory strategies. Firstly, there are certain strategies that indicate that the speaker is engaged in intrapersonal activity in search for self-regulation. They do not surface in fully extended linguistic appearance during speech activity, but instead in the form of filled (1), unfilled (or silent) pauses (2), as well as in
Antonio F. Jiménez Jiménez
the segmental lengthening of some words (3).3 These strategies result in a slowing down of the speech rate that reveals an increase of metacognitive activity and processing time. The first three strategies have been frequently nominated in the literature as stalling or buying-time strategies, while the fourth could be considered more as a consequence of the use of the other three. (1) Um al principio um um, los amigos um empieza whistle (laugh) um y um cerrar los las cartas para um mandar y um Phoebe tiene un papercut (laugh) y um Joey um le da un un copa de agua sucio. ‘Um at the beginning um um the friends um start whistle (laugh) um and um close the the letters to um send and um Phoebe has a papercut (laugh) and um Joey um gives her a a glass of dirty water) (2) Ross perdió el (2,64 seconds) (laugh) el (laugh) (4,68 secs.) um ◦ yo me sé◦ (3,50 secs.) um Ross necesitó (5,63 secs.) um um shoot! ‘Ross lost the (2,64 seconds) (laugh) the (laugh) (4,68 secs) um ◦ I know it◦ (3,50 secs.) um Ross needed (5,63 secs) um um shoot!’ (3) Los chicos um no están contentos porque la comida necesita tiene necesita tener nombres sssssimplimo- simples no nombres muy largos ‘The guys um are not happy because the food needs to needs to have sssssimp- simple names not long names’
The second form of self-regulatory strategies consists of linguistically shaped (overt) externalisations of the problem-solving thinking process, such as discourse fillers (4), and self-repetitions (5). These strategies serve at least two different functions: on the one hand, they have the private purpose of keeping focused attention on the problem, and on the other, they possess the social function of maintaining the floor in the speech event. (4) Y luego ellos están en la clase de de las mujeres embarazadas y: pero la la chica que está embarazada no está allí. ‘And then they are at Lamaze class and but the the girl who is pregnant is not there’ (5) Y uh el fin del del del juego del juego Ra- Rachel no tiene la moneda uh para paga ‘And at the end of of of the game of the game Ra- Rachel does not have the coin uh to pay’
The last cohort of self-regulatory strategies is comprised by self-corrections (6), self-recast (7), and self-retrieval (8) procedures, and they represent concrete solutions to the communicative problem for which the speaker has to lean exclusively on himself. These strategies also have a double function: first,
A sociocultural approach to language attrition
they fulfill the communicative need created by the social environment as they represent articulated elements that are conveyed in order to continue the communicative activity; and second, they also serve a private function as they are ways that help the individual to plan and restructure their intrapsychological activity. It is important to note that the use of self-regulatory strategies does not necessarily lead to a successful (non-attriting native speaker’s preferred) outcome. Instead, it generates the best alternative possible that the speaker can come up with at that particular moment, given the time and social constraints established by the communicative activity. (6) Y es una situación un poco incómodo incómoda ‘And is a situation a little bit uncomfortable’ (7) Ella resumen es muy mal or no es muy mal pero hay un un mistakes ‘Her resume is very bad or it’s not bad but there is a a mistake’ (8) Pero um hizo un equi- ¿equívico? mistake con su resume. ‘But um made a mis- mistake? Mistake with her resume’
Analysis of these manifestations in relation to the speaker dominant or nondominant language can offer very valuable insights into the speaker’s cognitive functional system and the extent to which the speaker has appropriated and internalised the language, not only to communicate with others, but to communicate with the self. An individual may carry out intrapersonal activity in his L1 while his interpersonal communication may be grammatically appropriate in his L2.
Other-regulatory strategies. When self-regulation cannot be maintained and self-regulatory strategies are not available at the time of the breakdown, the speaker can rely on other-regulatory strategies to regain control over the task. Other-regulation takes place when the “presence of other(s) in the communicative act regulates choice of strategy” (Lantolf & Frawley 1985: 153). Otherregulatory strategies are based on collaborative interaction with the interlocutor/s. These types of strategies clearly bring upfront the type of problem the speaker is facing at that moment. Speakers engaged in a meaningful communicative activity stop the argumentative line of conversation to momentarily focus on metalinguistic activity. As soon as a common agreement has been reached, they renew the course of the communication. Other-regulatory strategies comprise repetition requests (9), clarification requests (10), confirmation requests (11), comprehension checks (12), and expressions of non- or mis-understanding (13).
Antonio F. Jiménez Jiménez
(9) ¿Puedes repetir, por favor? ‘Can you repeat, please?’ (10) ¿Cómo se dice. . . en español? ‘How do you say. . . in Spanish?’ (11) Puedo cambiar mi carta para tuya y ¿es carta? Creo que sí. ‘I can change my car with yours and is it car? I think so’ (12) ¿Dices que te gusta o no te gusta? ‘Are you saying that you like it or that you don’t like it?’ (13) No comprendo. ‘I don’t understand’
It is important to note here that, even though the speaker appeals to the interlocutor in order to overcome the communicative problem, the final objective of this strategy is to regain self-regulation and continue the discourse. Also, the rules and the conditions of the communicative activity have to allow collaborative interaction so that the speaker can use these interactional strategies. In some situations, appeal for help is not an option, especially in asynchronous activities, that is when the interlocutor is not present and is not readily available to provide the needed assistance, such as in e-mail interchanges, leaving a message on an answering machine, etc.
Object-regulatory strategies. These types of strategies are employed when the speaker can not rely on himself or on the interlocutor to solve the communicative problem or, in other terms, “when interlanguage as object regulates choice of strategy” (Lantolf & Frawley 1985: 153). When this happens, two outcomes are possible: formulate a compensatory strategy that is not the original intended solution to the problem but that satisfies, to some extent, the communicative need, or avoid the problem altogether, restructuring the speaker’s course of action, and choosing an alternative perspective on the topic. In the first scenario, we find circumlocutions (14), approximations/all-purpose words (15), word-coinage (16), literal translations (17), onomatopoeias (18) and, code-switching (19). In the second scenario, the dominant strategy is message abandonment (20). (14) Máquina que tú haces para ap- que tú haces para cambiar el canal en la televisión (= remote control) ‘machine that you do to tu- that you do to change channels in the television’
A sociocultural approach to language attrition
(15) Mónica viene con comida como salmon y cosas así ‘Monica comes with food like salmon and things like this’ (16) Rachel se relició* se que hay algo no es correcto4 ‘Rachel realised that there is something that is not correct’ (17) Rachel está buscando para un empleo ‘Rachel is looking for a job= “Rachel está buscando un empleo”’ (18) Los chicos están bailando y es ha ha ha ha ha (laugh) ‘The guys are dancing and it’s ha ha ha ha ha (laugh)’ (19) The setting is in el café y um Jennifer quiere uh saber ‘The setting is in the café and um Jennifer wants um to know’ (20) El mono de Ross es- um está un poquito mal y (laugh) el el mono está um took um (laugh) el mono uh hay una canción ‘the monkey of Ross is um is a little bad and (laugh) the the monkey is um took um the monkey uh there is a song’
Lastly, I would like to make special mention of the use of metacomments during the communication breakdown. Metacomments may have all three regulatory functions. Metacomments may be addressed to the self (about one’s own ability or inability to solve the situation) (21), to the other (about the other’s influence on his own in/ability to overcome the breakdown) (22), or to the object (i.e. the difficulty of the task, or the difficulty to control the language) (23). (21) Y mientras uh Marcel tiene el remote, no sé, la la cosa de televisión ‘And in the meantime uh Marcel has the remote, I don’t know, the the thing for the television’ (22) Lo siento ‘I’m sorry’ (23) Pero Ross quiere quedarse y Susan tembién (with English pronunciation). Tembién (repeating the English pronunciation) that´s good! ‘But Ross wants to stay and Susan also (Engl. Pronunciation). Also (repeating the English pronunciation) That’s good!’
Metacomments can also be interpreted differently according to the context. A sympathetic listener may catch a self-addressed metacomment as an appeal for help. On the other hand, an other-oriented metacomment may trigger sufficient cognitive assistance for the speaker to regain self-regulation before the interlocutor’s participation. Again, all possible factors should be taken into consideration in order to carry out the analysis of regulatory strategies and metacomments.
Antonio F. Jiménez Jiménez
. Conclusion As derived from the discussion in this paper, the analysis of regulatory strategies needs to take into account three important elements: 1. the origin of the communicative breakdown (be it due to task difficulty, to a problem with the language of communication, or to a conceptual misunderstanding); 2. the activity system in which communication takes place (considering the characteristics of the elements that take part in the activity, such as the subject, the object, the mediational artefact, the rules of communication, etc.); and 3. the particular regulatory function that strategies fulfil in the communicative activity.5 Based on these elements, an alternative framework for the study of communicative strategies in relation to language attrition is presented in Figure 4 below. This framework is intended for both L1 and L2 attrition and any study on language development in general. The association of particular strategies with a main regulatory function is just indicative of the function they have been found to usually fulfil. However, strategies are fluid entities and the same strategy can serve different regulatory functions depending on the intention of the speaker. Language attrition reveals itself to be a very complex phenomenon that can take place under very different circumstances and at different levels. The type of analysis proposed in this paper aims to offer a more comprehensive account for language attrition by not only focusing on the linguistic elements, but also bringing upfront the individual and the activity as they can not be separated from the speaking activity. The language, the individual, and the activity form a system that makes sense only if contemplated as a whole. For this reason, the omission of one of these elements will necessarily provide a partial and, maybe, misleading picture of the language attrition phenomenon. A significant aspect that emerges from the present discussion is the possibility of attriting psychologically (at the regulatory level) but not linguistically. In fact, this is what may be happening in some studies on language attrition where no ‘apparent’ sign of attrition is observed (linguistically speaking) but in which a general slowing down of the communicative process is observed (see e.g. Verkaik & van der Wijst 1986; Cohen 1989). This, in psycholinguistic terms, may be explained by indicating that there is an accessibility difficulty that requires, consequently, more processing time but that eventually
A sociocultural approach to language attrition
COMMUNICATIVE PROBLEM DUE TO: Task difficulty/Conceptual problem Problem with the language
1
in relation to a particular ACTIVITY SYSTEM Instruments
2
Subject
Rules
Outcome
Object
Community
Division of labor
it influences regulatory strategy selection
3
- – filled pauses – unfilled pauses – segmental lengthnening – slower speech rate – discourse fillers – self-repetitions – self-recasts – self-corrections – self-retrieval – private speech
-
-
– repetitions – clarifications – confirmations – comprehension requests – expressions of non-or-mis understanding
– circumlocutions – approximations – word-coinage – literal translations – code-switching/ code-mixing – onomatopoeias – message abandonment
Figure 4. A framework for communicative strategies in relation to language attrition
may result in the production of a normatively correct linguistic form. However, through the lenses of Sociocultural Theory, this ‘processing time’ may be the temporal manifestation of the search of regulatory strategies in which the speaker engages in order to maintain self-regulation in the activity. Even though the application of this cognitive theory in the field of applied linguistics is relatively recent, studies under this framework are making very valuable contributions to the field. I anticipate that this theory will also be able to uncover important new information on the attrition phenomenon
Antonio F. Jiménez Jiménez
and I hope that the theoretical and practical considerations portrayed in this paper trigger future research on language attrition from this sociocultural perspective.
Notes . The principle of continuous access has been supported by Appel (1986), De Guerrero (1990), and McCafferty (1992). . The author comments that there are certain lexical items that the children always produce in Japanese, such as rice and beans (natto gohan) and please (chodai). . The examples shown for the strategies provided in this article have been selected from a video-based recall protocol for which native speakers of English had to retell in their L2, that is, in Spanish, the story conveyed in a video-clip extracted from a very famous TV series. Even though these strategies have been produced by L2 learners, they are also very commonly found in L1 speakers. Translations in English are provided for all the examples. However, it should be noticed that these translations try to maintain the structure of the Spanish example without considering their grammatical correctness in English. . Relició (= to realise) instead of darse cuenta . It is argued, as in Lantolf and Frawley (1985), that speakers may manifest any of the three regulatory strategies in different and unpredictable ways for which it becomes fundamental to analyse strategy use taking into consideration these three elements.
References Ahmed, M. K. (1994). “Speaking as cognitive regulation: A Vygotskian perspective on dialogic communication”. In J. P. Lantolf & G. Appel (Eds.), Vygotskian Approaches to Second Language Research (pp. 157–171). Norwood, NJ: Ablex Press. Appel, G. (1986). L1 and L2 Narrative and Expository Discourse Production: A Vygotskian analysis. Unpublished Doctoral Dissertation, University of Delaware, Newark, DE. Berko-Gleason, J. (1993). “Neurolinguistic aspects of first language acquisition and loss”. In K. Hylstenstam & A. Viberg (Eds.), Progression and Regression in Language: Sociocultural, neuropsychological, and linguistic perspectives (pp. 147–177). Cambridge: CUP. Bot, K. de (1999). “The psycholinguistics of language loss”. In G. Extra & L. Verhoeven (Eds.), Bilingualism and Migration (pp. 345–361). Berlin: Mouton de Gruyter. Cohen, A. D. (1989). “Attrition in the productive lexicon of two Portuguese third language speakers”. Studies in Second Language Acquisition, 11, 135–149. Cole, M. (1988). “Cross-cultural research in the sociohistorical tradition”. Human Development, 31, 137–151.
A sociocultural approach to language attrition
Coughlan, P. J. (1995). “Conversations with Vovó: A case study of child second language acquisition and loss”. Issues in Applied Linguistics, 6 (2), 123–136. De Guerrero, M. C. M. (1990). Nature of Inner Speech in Mental Rehearsal of the Second Language. Unpublished Doctoral Dissertation, Inter-American University of Puerto Rico, San German. Donato, R. (1994). “Collective scaffolding in second language learning”. In J. P. Lantolf & G. Appel (Eds.), Vygotskian Approaches to Second Language Research (pp. 33–56). Norwood, NJ: Ablex Press. Engeström, Y. (1987). Learning by Expanding: An activity-theoretical approach to developmental research. Helsinki: Orienta-Konsultit. Faerch, C., & Kasper, G. (Eds.). (1983). Strategies in Interlanguage Communication. London: Longman. Frawley, W., & Lantolf, J. P. (1985). “Second language discourse: A Vygotskian perspective”. Applied Linguistics, 6, 19–44. Håkansson, G. (1995). “Syntax and morphology in language attrition: A study of five bilingual expatriate Swedes”. International Journal of Applied Linguistics, 5 (2), 153–171. Hansen, L. (1999). “Not a total loss: The attrition of Japanese negation over three decades”. In L. Hansen (Ed.), Second Language Attrition in Japanese Contexts (pp. 142–153). New York, NY: Oxford University Press. Hedgcock, J. (1991). “Foreign language retention and attrition: A study of regression models”. Foreign Language Annuals, 24, 43–55. Hulsen, M. (2000). Language Loss and Language Processing. Three generations of Dutch migrants in New Zealand. Unpublished Doctoral Dissertation, Nijmegen: Katholieke Universiteit Nijmegen. Kasper, G., & Kellerman, E. (Eds.). (1997). Communication Strategies: Psycholinguistic and sociolinguistic perspectives. London: Longman. Lantolf, J. P. (1994). “Sociocultural theory and second language learning: Introduction to the special issue”. The Modern Language Journal, 78, 418–420. Lantolf, J. P. (2000). “Introducing sociocultural theory”. In J. P. Lantolf (Ed.), Sociocultural Theory and Second Language Learning (pp. 1–26). Oxford: Oxford University Press. Lantolf, J. P., & Appel, G. (1994). “Theoretical framework: An introduction to Vygotskian approaches to second language research”. In J. P. Lantolf & G. Appel (Eds.), Vygotskian Approaches to Second Language Research (pp. 1–32). Norwood, NJ: Ablex Press. Lantolf, J. P., & Frawley, W. (1985). “On communication strategies: A functional perspective”. Rassegna Italiana di Linguistica Applicata, 17 (2–3), 143–157. Leont’ev, A. N. (1981). Problems of the Development of the Mind. Moscow: Progress Publishers. McCafferty, S. (1992). “The use of private speech by adult second language learners: A crosscultural study”. Modern Language Journal, 76, 179–189. McCafferty, S. (1994). “Adult second language learners’ use of private speech: A review of studies”. The Modern Language Journal, 78, 421–436. McCafferty, S. (1998). “Nonverbal expression and L2 private speech”. Applied Linguistics, 19, 73–96. McCreary, D. R. (1985). “A Vygotskyan psycholinguistic perspective on the acquisition and loss of Japanese”. Rassegna Italiana di Linguistica Applicata, 17, 159–171.
Antonio F. Jiménez Jiménez
Obler, L. K. (1982). “Neurolinguistic aspects of language loss as they pertain to second language attrition”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 60–79). Rowley, MA: Newbury House. Olshtain, E. (1989). “Is second language attrition the reversal of second language acquisition?” Second Language Acquisition, 11, 151–165. Platzack, C. (1998). “The initial hypothesis of syntax: A minimalist perspective on language acquisition and attrition”. In H. Clahsen (Ed.), Generative Perspectives on Language Acquisition (pp. 369–414). Amsterdam and Philadelphia: John Benjamins. Poulisse, N. (1999). Slips of the Tongue: Speech errors in first and second language production. Amsterdam and Philadelphia: John Benjamins. Schmid, M. S. (2002). First Language Attrition, Use and Maintenance: The case of German Jews in Anglophone countries. Amsterdam and Philadelphia: John Benjamins. Sharwood Smith, M., & van Buren, P. (1991). “First language attrition and the parameter setting model”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 17– 30). Cambridge: CUP. Sorace, A. (2000). “Differential effects of attrition in the L1 syntax of near-native L2 speakers”. In C. Howell, S. Fish, & T. Ketih-Lucas (Eds.), Proceedings of the 24th Boston University Conference on Language Development (pp. 719–725). Somerville, MA: Cascadilla Press. Tarone, E. (1977). “Conscious communication strategies in interlanguage”. In H. D. Bown, C. A. Yorio, & R. C. Crymes (Eds.), On TESOL ’77 (pp. 194–203). Washington, DC: TESOL. Tarone, E. (1980). “Communication strategies, foreigner talk and repair in interlanguage”. Language Learning, 30, 417–431. Toribio, A. J. (2001). “On Spanish language decline”. In A. H.-J. Do, L. Domínguez, & A. Johansen (Eds.), Proceedings of the 25th Boston University Conference on Language Development (pp. 768–779). Somerville, MA: Cascadilla Press. Ushakova, T. (1994). “Inner speech and second language acquisition: An experimentaltheoretical approach”. In J. P. Lantolf & G. Appel (Eds.), Vygotskian Approaches to Second Language Research (pp. 135–156). Norwood, NJ: Ablex Press. Verkaik, P., & van der Wijst, P. (1986). Language Attrition and Word Recognition in French as a Foreign Language. Unpublished MA Thesis, Nijmegen: University of Nijmegen. Vygotsky, L. S. (1978). Mind in Society: The development of higher psychological processes. Cambridge, MA: Harvard University Press. Vygotsky, L. S. (1986). Thought and Language. (Alex Kozulin, Rev. and ed.). Cambridge, MA: MIT Press. Waas, M. (1996). Language Attrition Downunder. Frankfurt: Peter Lang. Ya˘gmur, K. (1997). First Language Attrition among Turkish Speakers in Sydney. Tilburg: Tilburg University Press. Yoshitomi, A. (1992). “Towards a model of language attrition: Neurobiological and psychological contributions”. Issues in Applied Linguistics, 3 (2), 293–318.
Perceived language dominance and language preference for emotional speech The implications for attrition research Jean-Marc Dewaele Birkbeck College, University of London
Introduction Ana Söter, born in Austria but transplanted to Australia at the age of 6, notes in her contribution to the book Reflections on Multiliterate Lives (Belcher & Connor 2002), that her native attrited German occasionally flashes back to her when writing about emotions: Although my second language has long been the sole language of my professional life, I find that I still occasionally resort to my L1 when seeking a phrase that aptly describes an emotion I want to capture in my creative writing (prose or poetry). I may not use the L1 term for that emotion or experience but discover a better English alternative during the process. (Söter 2002: 72)
Söter’s activation of her L1 shows that bilinguals may resort to a weaker or partially attrited language to express emotion. This phenomenon has been amply documented in previous research on the language choice of bilinguals. The L1 is often preferred to express personal involvement while the L2 is the language of distance and detachment (Bond & Lai 1986; Javier & Marcos 1989; Gonzalez-Reigosa 1976). Bilinguals’ language choices have been analysed using very different approaches and methodologies. Linguistic autobiographies and narratives of multilinguals represent one particularly rich source of evidence (Pavlenko 1998, 2001, 2002b; Besemeres 2003). Multilingual writers and linguists with a keen meta-linguistic awareness provide valuable insights on the subjective and psychological dimensions of their language use. Julia Alvarez, for example, a Dominican-American writer, describes her Spanish self as the
Jean-Marc Dewaele
core of her being although she uses the language infrequently. A single word with a Spanish pronunciation “resuscitates” her Spanish persona: Spanish certainly was the language of storytelling, the language of the body and the senses and of the emotional wiring of the child, so that still, when someone addresses me as “Hoolia” (Spanish pronunciation of Julia), I feel my emotional self come to the fore. I answer Si, and lean forward to kiss a cheek rather than answer Yes, and extend my hand for a handshake. Some deeper or first Julia is being summoned. (Alvarez, in Novakovich & Shapard 2000: 218)
The originality of Söter’s and Alvarez’ testimonies resides in the fact that even for an L1 attriter the first words that spring to mind to describe emotions belong to the L1 and that, inversely, one word in the L1 can trigger an emotional reaction and a switch in body language. L1 attrition and the resulting language choices are an extremely variable phenomenon. Not only does L1 attrition and L1 use vary between groups according to the type of experience each group had in the L1 (cf. Schmid 2002; forthc.), there is also within-group variation (de Bot & Hulsen 2002), and even a certain amount of conscious or unconscious intra-individual variation. Some bilingual authors consciously decided not to write in their L1 (which does not necessarily mean that their L1 had become attrited). This is the case for the above mentioned Julia Alvarez, and also for the Polish American Eva Hofmann, the Israeli Arab Anton Shammas, the émigré Spaniard Felipe Alfau, the Mexican American Ilan Stavans and the Irish prodigy Samuel Beckett (Kellman 2000). The case of Samuel Beckett is particularly interesting. Many critics see the reason for his decision, after World War II, to write exclusively in French, the language he learned as a schoolboy and studied at university, as a way to avoid the pathos and emotionality he associated with his L1 English and gain a greater simplicity and objectivity. Writing in French, he was able to “restrain his native verbal profligacy” (Kellman 2000: 28). Knowlson (1996) adds that for Beckett: “Writing poetry in French allowed him to get away (...) from the dense allusiveness, wide erudition and ‘intimate at arms length’ quality of his English poems” (pp. 293–294). Using French enabled him “to cut away the excess, to strip away the colour” (p. 357). The Croatian author Josip Novakovich advances similar argument to explain his decision to write in English: In my own case, English words didn’t carry the political and emotional baggage of repressive upbringing, so I could say whatever I wanted without pro-
Language dominance and language preference for emotional speech
voking childhood demons, to which Croatian words were still chained, to tug at me and to make me cringe. (Josip Novakovich, in Novakovich & Shapard 2000: 16)
The decision not to write in one’s L1 can also be dictated by more pragmatic concerns, namely the intended readership. Ilan Stavans (2001) relates how his grandmother, Bobbe Bela, a Polish Jew who emigrated to Mexico to escape persecution, sent him her diary “Mi Diario (...) in broken Spanish” (Stavans 2001: 50). She dedicated it to him in Spanish: ‘Dedico mi (diario) (o un chico relato) de mi juventud a mis queridos nietos, Ilan, Alison y Josh, como recuerdo de Bobe [sic] Bela.’1 (. . .) And not only is the dedication in Spanish but also the entire narrative. This, in fact, is the most urgent question I have, the one I would love to discuss with Bela. Why not in Yiddish, her mother tongue, the tongue in which she sobs and screams? I know the answer of course: she wants to be read, understood and appreciated; she is eager to reach not only me but my wife and, at some point in the future, my children as well. (2001: 51)
Cognitive psychologists have developed a completely different – and yet complimentary – approach to the issue of emotional resonance of multiple languages. Their aim is to match bilingual’s reports of feeling stronger emotions when speaking and hearing the L1 with psychophysiological assessments in a laboratory setting. The work of Cathy Harris and her collaborators at Boston University is particularly relevant for the present study because the bilinguals involved in the experiments also included L1 attriters. The researchers monitored autonomic arousal via fingertip electrodes while participants read or heard words and phrases in their L1 or L2 (defined by order of acquisition). Participants in a first study (Harris et al. 2003) included Turkish-English bilinguals which were late learners of English. Participants in a second study included a group of Spanish-English bilinguals who had acquired English early (average age: 3.7) and a group of Spanish L1 speakers who had learned English later (average age: 7.9 (Harris forthc.). The former group reported feeling more proficient in English. Stimuli in the two studies included taboo words, childhood reprimands (Don’t do that!), insults (you suck!), terms of endearment (I love you) and single words varying in emotional valence (cancer, joy, table). The researchers found that emotional expressions heard in the L1 elicited larger skin conductance amplitudes than comparable expressions in the L2 for the Turkish-English bilinguals (Harris et al. 2003). In the second study, electrodermal responsiveness in English was smaller than in Spanish for those who had learned English in middle childhood, while no such difference was obtained
Jean-Marc Dewaele
for those who learned English in early childhood (Harris forthc.). It thus seems that when two languages are learnt early, emotional expressions activate the autonomic system equally. Harris points out that her data are inconsistent with the strong variant of the “L1 is more emotional” thesis: “the first language is more emotional when it is the more proficient language” (p. 3). In other words, the emotional resonance of the L1 is deeply anchored within the individual, as the testimonies collected by Novakovich and Shapard (2000) and Belcher and Connor (2002) suggested, but it is mediated by proficiency. In sum, what this short review suggests is that the relationship between mother tongue use and emotion is a very complex one. It seems clear that the L1 retains very strong emotional connotations even if that language is not used regularly. The present paper will address these issues by adopting a bird’s-eye perspective on data obtained from a large group of adult multilinguals.
.
Method
. Rationale for the present study Language learning memoirs, linguistic autobiographies and laboratory experiments provide valuable insights in the processes of language loss and language acquisition. Patterns emerge, but limited sample sizes make it difficult to draw more general conclusions. Ideally one needs a sufficient amount of comparable data (by putting the same questions to many people) that could be analysed both quantitatively and qualitatively. The use of a web questionnaire allows the collection of self-reported data from a very large sample of multilinguals – and L1 attriters – from all possible linguistic backgrounds. The combination of quantitative data through Likert-scale type responses and qualitative data through open questions makes it possible to draw a fairly detailed picture of multilinguals’ self-reported speech behaviour and perceptions. The questionnaire was not designed with attrition in mind but this potential drawback turned out to be an unexpected bonus as it allowed a comparison between multilinguals who reported to be still dominant in their L1 and those who reported shared dominance of the L1 and one or more other languages (aptly described by Stavans (2001) as the “mother versus father tongue”2 ) and finally those individuals who had become dominant in another language. The latter category consists – by implication – of people whose L1 is more likely to be attrited.
Language dominance and language preference for emotional speech
. Research questions In the present study we will firstly investigate which of the four skills in the L1 (speaking, understanding, reading, writing) suffers most from perceived attrition. Secondly, we will look at the effect of perceived attrition on perceptions of the L1 (useful, colourful, rich, poetic, emotional). Thirdly, we will analyse how perceived L1 attrition affects the use of the L1 for expression of anger and feelings; for inner speech and mental calculation; and finally for swearing in the L1 and perceived emotional force of swearwords. Finally, we will check for inter- and intra-group differences (especially within the group of perceived L1 attriters). . Participants A total of 1039 multilinguals contributed to the database (731 females, 308 males). The participants spoke a total of 75 different L1s. English native speakers represent the largest group: n = 303; followed by native speakers of Spanish: n = 123; French: n = 101; German: n = 97; Dutch: n = 76; Italian: n = 52; Catalan: n = 32; Russian: n = 29; Finnish n = 28; Portuguese: n = 20; Greek: n = 15; Swedish: n = 15; Japanese: n = 11; Welsh: n = 10... The 127 remaining participants share another 61 L1s among themselves. Our population could be described as highly polyglot with 144 bilinguals, 269 trilinguals, 289 quadrilinguals and 337 pentalinguals. The L2 was defined as the second language to have been acquired by the individual, the L3 as the third language etc. A closer look at the ages of onset for learning the L2, L3 revealed that 157 L2 users are in fact “bilingual L1” users, having learned the L2 from birth. This represents 15% of the L2 group. Similarly, 19 L3 users are “trilingual L1” users (representing 1.8% of the L3 group). There are no “quadrilingual L1” users. Our participants are generally highly educated with 115 having a high school diploma or less, 273 with a Bachelors degree, 308 with a Master, and 338 with a Ph D. Age ranged from 16 to 70 (Mean: 35.6; SD: 11.3). A majority (n = 837) reported working in a language-related area, a minority (n = 200) had professions that were not linked to languages. The strong proportion of highly educated female participants means that it is impossible to claim that the sample is representative of the general population. This potential pitfall (cf. Dörnyei 2003: 75) is inevitable with web-based questionnaires and it needs to be kept in mind when interpreting the patterns, as results might be different for a sample of – for example – mainly uneducated males. To partially remedy this problem data were collected through a printed version of the questionnaire from about 50 multilinguals in
Jean-Marc Dewaele
the London area who did not finish high school. Statistical analysis revealed no significant differences between this group and the rest of the sample for the dependent variables under consideration in the present study.3 . Research instrument and methodological considerations The following sociobiographical information was collected: sex, age, education level, ethnic group, occupation, languages known to the participant, dominant language(s), chronological order of language acquisition, context of acquisition, age of onset of learning a language, frequency of use and typical interlocutors. Self-rated proficiency scores for speaking, comprehending, reading and writing in the different languages were obtained. The first part of the questionnaire consists of closed questions with 5-point Likert scales, the second consists of open questions where the participants were asked to write a response. Language choice was determined for self- and other-directed speech, for emotional and non-emotional speech. For the latter we enquired about mental calculation and inner speech, for the former we asked about the expression of anger and feelings to different interlocutors including the self, as well as the expression of emotion in letters. We also enquired about language choice for swearing and perceived emotional weight of swear words (see also Dewaele forthc. a, b). Attitudes towards the different languages were measured with 5-point Likert scales. We focussed on perceived usefulness, colourfulness, richness, poetic and emotional character of the languages. The open questions enquired about the emotional weight of the sentence I love you in the different languages, about their emotional significance, about the preferred language for writing a personal diary, for recalling bad or difficult memories, for arguments. We also asked whether using a different language made one feel a different person, and whether it was easier or more difficult to talk about emotions in the different languages. A comparison of answers given on the L1 to have been acquired, and the dominant language at the time of filling out the questionnaire, made it possible to distinguish 3 categories of participants: Firstly, those who were still dominant in the L1 (n = 561); secondly those who reported dominance in two or more languages including the L1 (n = 373); and thirdly those reported dominance in language(s) not including the L1 (n = 105). The latter category could contain perceived L1 attriters.4 Schmid (2002) pointed out that overcoming methodological problems was one of the major challenges in the field of L1 attrition. Different elicitation techniques yield different results, and the results themselves are drawn
Language dominance and language preference for emotional speech
from different theoretical approaches, considering various domains of linguistic competence such as accuracy in phonology, morphology, syntax, lexis, as well as concepts that are harder to define such as fluency. The fragmentation of the results and the language-specific concepts make ulterior comparisons very difficult. While the problem in itself is probably unsolvable, it is possible, as was done in the study to be discussed, to choose a very general level of analysis, such as frequency of choice of the L1 for specific speech acts and perception of the L1 on a number of specific dimensions, and to compare these results with the data for the same variable in other languages known to the participant. Another problem in attrition studies concerns the selection of the experimental group and the control group, or, as Jaspaert, Kroon and van Hout (1986) put it: what is the point of reference? There is a danger for example that a comparison of potentially attrited multilinguals and monolingual controls becomes in fact a comparison between monolinguals and multilinguals, rather than a comparison between attriters and non-attriters. Moreover, it has been pointed out regularly in the SLA literature that it is not satisfactory to compare a monolingual’s competence in a language with a bilingual’s competence in that language (cf. Cook 2002). One could therefore argue that the only valid comparisons in attrition research would be between multilinguals with similar languages of which one may be attrited. The present study attempts at overcoming this problem by doing just that, i.e. by only considering multilinguals. The other problem concerns research designs with only two groups (i.e. experimental versus control group). The researcher will obviously use a set of strict criteria to assign a participant to either group and then carry out a comparison between group A and B. But is such a dichotomy theoretically acceptable? Researchers report wide variation within groups of attriters, so much so that a researcher might have wished after an analysis of the data to exclude some of the participants in the experimental group. There is no reason either to believe that a (multilingual) control group would be completely homogeneous. Levels of activation of languages in the multilingual mind fluctuate (cf. Green 1986). In other words, the nature of attrition is continuous rather than dichotomic. A lower level of activation of a language could be interpreted as the onset of an attrition process. People who spend two weeks immersed in the L2 might show their first signs of L1 attrition upon their return to their home. In the present study this problem was largely avoided. Firstly, because the words “language loss” or “attrition” were absent in the questions, with only one of many questions enquiring about dominant language(s). This formulation does not have the dramatic negative connotations of attrition, perceived
Jean-Marc Dewaele
as non-desirable and possibly skewing the responses. Secondly, the categorisation was carried out post-hoc by the researchers, after a comparison of the L1(s) and the reported dominant language(s). The choice of three levels (dominance in L1, L1+LX and LX) meant that more sophisticated statistical analyses (M/ANOVA) could be used rather than simple t-tests which are typically used for comparisons between two levels. The advantage with analyses of variance is that that effect sizes can be calculated and that post-hoc tests can be carried out to see the ranking order of the three groups, including reports on possible significant differences between them. As it happened, it turned out that, despite considerable intra-group variation, the rank order remained constant. This finding has important theoretical implications about the nature of perceived L1 attrition. The final methodological point concerns the size of the samples and the variety of languages under consideration. Any research based on production data is extremely time-consuming, researchers therefore consider that a group of 50 participants constitutes a very good sample. The question remains however to what extent results on such a sample for one specific language are generalisable? One way to overcome this problem is by concentrating not on production data but on self-reported data. We created an on-line web questionnaire on multilingualism and emotions (Dewaele and Pavlenko 2001) that allowed us to gather data from a very large sample of learners and long-time users of multiple languages from across the world and from a wide age range, i.e. not only the 18–22 years-old which are predominantly used in empirical research in applied linguistics and psychology. This approach is not without its own methodological limitations (cf. Pavlenko 2002a). The problem of respondent self-selection has been mentioned before. To fill out the questionnaire, participants needed access to internet and a certain degree of metalinguistic awareness. Other methodological objections could be raised, for example, how strongly are self-evaluations about L1 loss correlated with objective measures? There seems to be some variation but overall researchers did find significant positive relations (Hulsen 2000; Waas 1996; Ya˘gmur 1997). Doubts about the validity of our research instrument can easily be assuaged as questionnaires with Likert scales responses have been tried and tested extensively in sociopsychological research (cf. Dörnyei 2003). These quantitative data can provide an excellent baseline, provided they are backed up by different types of data. We argued in favour of triangulation in bilingualism research in Dewaele and Pavlenko (2002). This is the reason why we also gathered short narratives through our web questionnaire, using open questions inquiring into emotion and communicative behaviour. The use of self-reports through web question-
Language dominance and language preference for emotional speech
naires may be an innovative and useful complement to existing methodologies in attrition research but we are certainly not claiming that it could substitute actual production and comprehension data. . Dependent variables Dependent variables are primarily of a quantitative nature (scores on 5-point Likert scales). Qualitative data from the open questions have been used to illustrate how the perceived L1 attriters felt about their L1. The closed questions considered in the present study were formulated as follows: On the scale from 1 (least proficient) to 5 (fully fluent) how do you rate yourself in speaking, understanding, reading, writing?
Self-perceived proficiency reflects the individuals’ perception of their competence in a language. Similar self-report measures have been proved to correlate highly with performance measures of proficiency (e.g. Dufour & Kroll 1995; Kroll, Michael, Tokowicz & Dufour 2002). Dewaele (2002) found that selfperceived competence in French L2 and English L3 from 100 Dutch L1 learners had a strong inverted relation with communicative anxiety in these languages. Self-perceived competence in English appeared to be linked to the degree of extraversion which confirmed earlier findings, i.e. that introvert and anxious foreign language students tend to underestimate their competence relative to more extravert and less anxious students, who tend to overestimate their competence. This has no consequences for the present study, as it can be assumed that the population is normally distributed in terms of extraversion scores. The second series of closed questions were stated as follows: Here are some subjective statements about the languages you know. Please mark to what extent they correspond to your own perceptions. There are no right/wrong answers. (My L1/L2/L3/L4/L5 is....useful, colorful, rich, poetic, emotional).
The third series of questions sought to elicit information about self- and otherdirected speech, both emotional and non-emotional. Participants were asked to circle the appropriate number. –
If you are angry, what language do you typically use to express your anger? (never=1, rarely=2, sometimes=3, frequently=4, all the time=5)
Jean-Marc Dewaele
This question was repeated for the expression of anger when alone, in letters and e-mail, when talking to friends, family or partners, and strangers. –
What language do you express your deepest feelings in?
This question was also repeated for the expression of feelings when alone, in letters and e-mail, when talking to friends and family or partners. The last four questions were formulated as follows: –
If you perform mental calculation, what language do you typically use?
–
If you form sentences silently (inner speech), what language do you typically use?
–
If you swear in general, what language do you typically swear in?
–
Do swear and taboo words in your different languages have the same emotional weight for you? (1=does not feel strong, 2=fairly weak, 3=fairly strong, 4=strong, 5=very strong).
One might object that this type of questionnaire cannot capture the richness, the inherent variability and the complexity of the phenomenon under consideration. Indeed some participants commented: “it depends!”. It is true that questionnaires are but one of many possible methods to focus on variation in non/emotional communication. Questionnaires may not be the linguists’ preferred method of data collection but it is the most popular instrument in personality and social psychology. Participants fill out questionnaires with closed questions relating to their usual behaviour in prototypical situations. One of the most robust questionnaires in psychology, the Eysenck Personality Inventory (Eysenck & Eysenck 1968), contains questions like “Would you describe yourself as a nervous person?” where the participant is asked to tick either YES or NO. These answers constitute the first level of abstraction. Factor analyses allow researchers to compare factor scores for these questions on various dimensions, allowing them to pick out the questions that have the best predictive power in order to locate an individual on a personality dimension (cf. Digman 1990). Our multi-item scales on language choice for specific speech acts and perception of languages constitute a second level of abstraction. . Research design Our main independent variable is the type of language dominance (either L1 only, shared L1+LX, or LX only). Internal consistency, which “refers to the homogeneity of the items making up the various multi-item scales within the
Language dominance and language preference for emotional speech
questionnaire” (Dörnyei 2003) was measured by the Cronbach alpha coefficient.5 Multivariate analyses of variance (MANOVA) and Scheffé post-hoc tests were used to check for inter-group differences. The number of participants may vary across the analyses because some did not provide data for all the in/dependent variables and were therefore excluded from a specific calculation. Paired t-tests allowed us to check for intra-group differences between frequency of use of L1 for emotional (expression of anger) and non-emotional speech (mental calculation).
. Results The analyses revealed that type of language dominance had a significant effect overall, although there was some variation in the between-subjects effects. Results almost always appeared in the same order: the L1 dominant group obtaining the highest scores, followed by the L1+LX group, and finally the LX dominant group. . Self-rated proficiency scores in the L1 Internal consistency reliability was extremely high (alpha = .92). A MANOVA with language dominance as main independent effect and proficiency in the four language skills as dependent variable shows a significant effect (Wilks lambda = .80, F (12, 2725) = 20.4, p < .0001, eta2 = .07). Not surprisingly the LX dominant group reported significantly lower proficiency for the four skills in their L1 compared to the L1 and the L1+LX groups. Post-hoc tests showed that the LX dominants reported significantly lower proficiency for the four skills than the 2 other groups. The difference between the L1 and the L1+LX groups is not significant however. The analysis of between-subjects effects sugTable 1. Between-subjects effects for self-rated proficiency scores in the L1 (df = 3,1036) Skill
F
eta2
Scheffé post-hoc tests
Writing Reading Speaking Comprehension
66.2*** 29.4*** 26.8*** 7.6***
.16 .08 .07 .02
LX/L1***, LX/LX+L1***, L1/L1+LX LX/L1***, LX/LX+L1***, L1/L1+LX LX/L1***, LX/LX+L1***, L1/L1+LX LX/L1***, LX/LX+L1***, L1/L1+LX
*p < .05, **p < .001, ***p < .0001
Jean-Marc Dewaele (L1: n=558; L1+LX: n=371; LX: n=107) 5
Proficiency score
4.5 L1 L1+LX LX
4 3.5 3 Speak
Comprehend
Read
Write
Figure 1. Self-rated proficiency in the L1 for the L1 dominant group, the L1+LX dominant group and the LX dominant group
gests that the effect of language dominance was strongest for writing, followed by reading, speaking and finally comprehension (see Table 1). It thus seems that self-reported proficiency in writing the L1 suffers more from attrition than oral L1, and that reception is less affected than production. The LX group thus presents clear indications of L1 attrition, according to de Bot and Hulsen’s (2002) definition, “language attrition reflects a decrease in language proficiency at the individual level” (p. 255). Means for the three groups are presented in Figure 1. . Expression of feelings in the L1 (frequency of use) The Cronbach alpha coefficient revealed a low degree of internal consistency (alpha = .46) because of the presence of the item “communication of feelings in letters and email”. When this item was removed, the Cronbach alpha coeffiTable 2. Between-subjects effects for expression of feelings in the L1 (frequency of use) (df = 2,970) Expression of feelings
F
eta2
Scheffé post-hoc analyses
When alone To friends To parents
98.3*** 77.5*** 42.3***
.17 .14 .08
LX/L1***, LX/L1+LX***, L1/L1+LX*** LX/L1***, LX/L1+LX***, L1/L1+LX** LX/L1***, LX/L1+LX***, L1/L1+LX
*p < .05, **p < .001, ***p < .0001
Language dominance and language preference for emotional speech (L1: n=528; L1+LX: n=337; LX: n=93) 5
Frequency of use
4.5 4 L1 L1+LX LX
3.5 3 2.5 2 Alone
Friends
Family
Figure 2. The expression of feelings in the L1 for the L1 dominant group, the L1+LX dominant group and the LX dominant group
cient increased for the three remaining items (alpha = .83). This suggests that there is a clear effect of modality for language choice in expression of feelings. The item “communication of feelings in letters and email” was left out of the subsequent analyses. The overall effect of type of language dominance is highly significant (Wilks lambda = .82, F(6, 1936) = 33.9, p < .0001, eta2 = .095). All the betweensubjects effects are highly significant. The effect is strongest for the expression of feelings alone, followed by expression of feelings to friends and finally expression of feelings to parents (see Table 2). A Scheffé post-hoc test revealed that the L1 dominant group use their L1 significantly more to express their feelings, the L1+LX group does so significantly less and the LX group uses the L1 less still to talk about feelings. The only exception was for the expression of emotion to parents where the difference between the L1 and the L1+LX group was non-significant (see Figure 2). . Expression of anger in the L1 (frequency of use) The Cronbach alpha coefficient revealed a high degree of internal consistency for this 6-item scale (alpha = .87). The item “communication of anger in letters and email” was therefore included in the analyses.
Jean-Marc Dewaele
The MANOVA shows the same pattern as for the expression of feelings. The overall effect of type of language dominance is highly significant (Wilks Lambda = .77, F (12, 1772) = 20.1, p < .0001, eta2 = .12). All the betweensubjects effects are highly significant. In other words, L1 attrition is linked to a similar reduction in the use of the L1 to swear, to express anger alone, to friends, to family, and to strangers. The effect is strongest for the expression of anger in letters / emails and to friends. It decreased gradually for the expression of anger alone, to strangers, to parents and for swearing (see Table 3).
Table 3. Between-subjects effects for expression of anger in the L1 (frequency of use) (df = 2,891) Expression of anger
F
eta2
Scheffé post-hoc analyses
In letters/ emails To friends When alone To strangers To parents Swearing
92.8*** 85.1*** 78.8*** 60.3*** 46.1*** 39.8***
.17 .16 .15 .12 .09 .08
LX/L1***, LX/L1+LX***, L1/LX/L1*** LX/L1***, LX/L1+LX***, L1/LX/L1*** LX/L1***, LX/L1+LX***, L1/LX/L1*** LX/L1***, LX/L1+LX***, L1/LX/L1*** LX/L1***, LX/L1+LX***, L1/LX/L1* LX/L1***, LX/L1+LX***, L1/LX/L1***
*p < .05, **p < .001, ***p < .0001
(L1: n=517; L1+LX: n=348; LX: n=98) 5
Strength of perception
4.5 4 3.5 3 Useful
Colorful
Rich
Poetic
Emotional Swearwords
L1 L1+LX LX
Figure 3. The expression of anger in the L1 for the L1 dominant group, the L1+LX dominant group and the LX dominant group
Language dominance and language preference for emotional speech
A Scheffé post-hoc test (see Table 3) revealed that the L1 dominant group use their L1 significantly more to express anger in the different situations, the L1+LX group does so significantly less and the LX group comes out at the bottom (see Figure 3). . Inner speech and mental calculation in L1 (frequency of use) The Cronbach alpha coefficient revealed a relatively low degree of internal consistency (alpha = .65), suggesting that these two items may, to a certain degree, reflect different aspects of silent speech. The overall effect of language dominance is highly significant (Wilks Lambda = .75, F(4, 1982) = 76.3, p < .0001, eta2 = .13). Between-subjects effects are highly significant: the effect is strongest for mental calculation (F(2, 992) = 145.6, p < .0001, eta2 = .227) followed by inner speech (F(2, 992) = 72.1, p < .0001, eta2 = .127). A Scheffé post-hoc test showed significant differences (p < .0001) between the 3 groups for both variables. This suggests that language dominance also strongly affects self-directed, silent speech, and especially mental calculation. (L1: n=544; L1+LX: n=354; LX: n=97)
Frequency of use
5
4.5 L1 L1+LX LX
4
3.5
3 Inner speech
Mental calculations
Figure 4. Inner speech and mental calculation in L1 (frequency of use) for the L1 dominant group, the L1+LX dominant group and the LX dominant group
Jean-Marc Dewaele
. Perceptions of the L1 The Cronbach alpha coefficient revealed a relatively high degree of internal consistency (alpha = .76). Here again a highly significant overall effect of language dominance emerges (Wilks lambda = .92, F(12, 1868) = 7.7, p < .0001, eta2 = .047). The analysis of between-subjects effects confirms the existence of strong differences between the dependent variables: the effect of language dominance is strongest for perceived usefulness, followed by perceived richness and perceived colourTable 4. Between-subjects effects and post-hoc results for perception in the L1 (frequency of use) (df = 2,939) Perception of the language
F
eta2
Scheffé post-hoc tests
Useful Rich Colorful Emotional Poetic Perception of emotional force of swearwords
20.7*** 5.1*** 4.9** 1.8 1.2 3.3*
.08 .02 .01 0 0 .01
LX/L1***, LX/L1+LX***, L1/L1+LX* LX/L1***, LX/L1+LX**, L1/L1+LX LX/L1**, LX/L1+LX*, L1/L1+LX LX/L1, LX/L1+LX, L1/L1+LX LX/L1, LX/L1+LX, L1/L1+LX LX/L1, LX/L1+LX*, L1/L1+LX
*p < .05, **p < .001, ***p < .0001
(L1: n=517; L1+LX: n=348; LX: n=98) 5
Strength of perception
4.5 4 3.5 3 Useful
Colorful
Rich
Poetic
Emotional Swearwords
L1 L1+LX LX
Figure 5. Perceptions of the L1, including swearwords, for the L1 dominant group, the L1+LX dominant group and the LX dominant group
Language dominance and language preference for emotional speech
fulness. The effect sizes are very small however. The effect is not significant for perceived emotionality of the language nor for its poetic character. The effect is significant for perception of emotional force of swearwords but the effect size is very limited (see Table 4). The Scheffé post-hoc test revealed that the differences between the three groups are not consistently significant for the different variables: significant inter-group differences emerged for perceived usefulness, perceived richness, perceived colourfulness and perception of the emotional/locutionary force of swearwords but none appeared for perceived poetic character and emotionality of the language. . Intra-group differences The last part of the quantitative analysis focuses on intra-group differences in L1 use, in other words, do people belonging to one of the three groups (L1 dominant, L1+LX or LX) differ significantly in their L1 use according to the emotional content of the speech act? Paired t-tests suggest that there is no significant difference between L1 use for the expression of anger alone and mental calculation for the LX group: (t = –1.3, p = ns). The LX group thus uses the L1 equally for the expression of anger alone and mental calculation. There were highly significant differences however within the L1+LX group: (t = 7.1, p < .0001) and the L1 group: (t = 7.6, p < .0001). In these two groups, L1 was used much more often for mental calculation that for the expression of anger. . Testimonies from LX dominant participants Reading through the comments written by the LX dominant participants, one realises that their relation to their L1 varies enormously. Some feel completely detached from the L1, as is the case for Gertrud (German L1, English L2, French L3, Spanish L4, Italian L5): English is the language in which I can express my emotions most directly. French the language I enjoy playing about with. German is the language from which I probably feel most distant (apart from Spanish and Italian which I do not speak well enough to make any difference).
Others like David (Sindhi L1, English L2, Malay L3) express a certain nostalgia when talking about their L1: Malay I see as my working language used as medium of instruction and with colleagues in code switched utterances to signify friendship English as a prac-
Jean-Marc Dewaele
tical language –to indicate status. Sindhi as the language of my forefathers and I wish I had more people to speak it with, I miss the language.
Virhe (Estonian L1, English L2) also associates his L1 with a mythical past: Perhaps my association of Estonian as a precious language of a lost world has more to do with my view of my father as a tragic exiled figure than my own associations with any lost land but those feeling must have been very profoundly passed on to me and my brother for the language to carry the sort of significance it does: I believe it goes beyond the history and the culture it is attached to.
Another sentiment expressed by participants is that the L1 no longer fits their personality. Eric (German L1, English L2) notes that his L1 belongs to some distant former self: When speaking German I feel either immature or unprofessional to a certain extent. I feel like my German has been frozen at the age of 18. When speaking German I feel as if I should be expressing myself more maturely though I’m not sure exactly in what way. I feel less confident when using German than English now never having had to function professionally completely in German.
Others, like Martina (German L1, English L2, French L3), observe that they assume different roles or identities when speaking the L1: When I use my L1 (German) I am often the daughter, the friend and until recently ‘the teacher’. French is nowadays a ‘social’ language used for chat, gossip and light-hearted discussion. Against this English is now my ‘language of habitual use’.
The environment might also hinder or favour the use of the L1. Kristina (German L1, English L2) notices that when German expatriates meet where she lives in the UK they often speak English, which could be an indication of a gradual shift (cf. de Bot & Hulsen 2002): I am intrigued that when meeting other German L1 speakers the conversation will often continue in English.
Some participants resent standing out linguistically and avoid the use of the L1 in places where they can be overheard. This is a well-documented phenomenon among multilingual schoolchildren (cf. Yamamoto 2001). The following participant, Nathan (French L1, Greek L2, English L3) is an adult however, with a similar social uneasiness: I do not feel comfortable when my wife speaks L1 to me in public places, in order not to attract attention I answer in L2.
Language dominance and language preference for emotional speech
One participant, who belongs to the L1+LX group, observes that she is no longer able to be angry in her L1 after spending several years in her L3 environment (Johanna: English L1, French L2, Italian L3, Spanish L4): I’m more likely to express anger in Italian. Mainly because I’ve only really learned how to in the last few years and since I’ve spent my young adulthood here I’ve gotten more practice raging at the government or the landlord in my adopted language. I still end up feeling ridiculous when I get worked up about things in English.
. Discussion One of the striking results in the present study is the emergence of the same order for the three groups, and this for every dependent variable: the L1 dominant group almost always comes first, followed by the L1+LX dominant group, with the LX dominant group coming in last. Two observations can be drawn from this pattern. Firstly, it is a clear indication that self-reported L1 attrition similarly affects self-reported proficiency, self-reported emotional as well as non-emotional L1 use, and self-reported silent as well as articulated speech in the L1. Secondly, it suggests that shared L1+LX dominance corresponds to weak perceived L1 attrition for emotional as well as non-emotional L1 use. However, shared L1+LX dominance does not affect self-rated proficiency in the L1 nor perceptions of the L1. Perceived LX dominance was also found to affect self-reported oral and written skills in the L1 to varying degrees. Written proficiency appears to be affected more strongly than oral proficiency, and writing more so than reading. Oral production suffers more than auditory reception. It thus seems that the LX dominant participants feel they retained their capacity to decode the L1 better than the capacity to encode it. A closer look at the perceptions of different types of emotional and nonemotional speech revealed that both are similarly affected by LX dominance. L1 use for silent speech (inner speech and mental calculations) is affected as much by perceived L1 attrition as articulated speech. Perceived L1 attrition has a strong effect on perceived usefulness, less on perceived colourfulness, perceived richness, perception of the emotional force of swearwords and no effect on poetic and emotional properties of the L1. This is interesting because it suggests that the emotional resonance of the L1 remains largely unaffected by perceived L1 attrition. Just as L1 swearwords tend
Jean-Marc Dewaele 45 40
% of respondants
35 30 25 20 15 10 5 0
never inner speech
rarely
sometimes
mental calculation
frequently anger alone
all the time feelings alone
Figure 6. Distribution of use of the L1 in the LX dominant group for inner speech, mental calculations, expression of anger and feelings alone (percentage of respondants)
to retain their force, despite being rarely – if ever – used, the whole L1 does retain emotional connotations. Söter’s testimony which opened the present chapter offers another intriguing perspective on the role of the L1. When writing in the dominant L2, she reports that emotions activate emotion words in the L1, which in turn activate a further lexical search in the L2. This means that no L1 words survive in the end-product, despite the fact that the L1 did play a significant role in the psycholinguistic genesis of the production. It is possible however that the L1 did affect lexical choice in the L2 (cf. Pavlenko 2002b). Our analysis showed that compared to the L1 dominant and the L1+LX dominant groups, the LX dominant group used the L1 significantly less for inner speech, but even a lower average shows that some LX dominant participants do use their L1 regularly. A closer look at the distribution (Figure 6) reveals that only 14% of LX dominant participants declare to never or rarely use their L1 for inner speech, a majority uses the L1 sometimes, frequently or all the time. A larger proportion of LX dominant participants (30%) never or rarely uses the L1 for mental calculation. Feelings and anger are never or rarely expressed in the L1 by respectively 32% and 23% of L1 attriters. What this demonstrates is that a majority of LX dominant participants in the sample do use their L1 at least occasionally, and most frequently so for inner speech.
Language dominance and language preference for emotional speech
. Conclusion The results of the quantitative part of study that emerged through multivariate analyses show that language dominance / perceived L1 attrition has a significant effect on self-related proficiency in the L1, on frequency of use of the L1 for the expression of feelings, anger and swearing to different interlocutors and in different modalities, for silent and articulated speech. Language dominance also affects perceptions of certain characteristics of the L1, but not the perception of its emotional and poetic character, and the perception of emotional force of swearwords. These patterns were confirmed by the narrative evidence provided by the participants and strengthened the views expressed by bilingual writers (cf. Besemeres forthc.; Kinginger forthc.; Pavlenko 2002b). The attrited L1 retains powerful emotional connotations, which can be either positive or negative. One possible explanation is that because of the number of neural connections established in early and middle childhood, the whole L1 is linked to systems for emotional arousal (Harris et al. 2003). L1 attriters adopt new languages to express themselves in and to project their adult personalities. The L1 recedes to the deepest layers of the self (cf. Alvarez: “Some deeper or first Julia”; in Novakovich and Shapard 2000: 218), and maintains an unseen / unheard presence in the inner speech of the individual. Metaphorically one could compare the use of the attrited native tongue as an activity, like cycling, that hasn’t been performed for a long time. The memory of what the performance feels like remains intact, the emotional image hasn’t deteriorated, but the actual procedures needed to execute the action are dormant and, at least temporarily, inaccessible because of lack of exposure (cf. de Bot 1999). This study shows that the bird’s-eye perspective on quantitative and qualitative data from a large sample of multilinguals can provide useful and novel information about perception and self-reported use of the L1. The comparison of the (self-reported) L1 dominant group with the L1+LX dominant group and, finally with the LX dominant group (potential L1 attriters) shows the dynamic nature of language dominance and L1 attrition. It raises a question about the exact definition of attrition: from what point on the continuum can one be labelled an “attriter”? A person’s self-perception is a subjective statement and it cannot be verified without production data. One of the most interesting findings is that the mere presence of a second dominant language is linked to an almost imperceptible decrease in self-reported proficiency in the L1. As frequency of use of the L1 diminishes so does the perception of usefulness, colourfulness, richness, poetic and emotional character of the L1. These patterns are merely more pronounced among the participants who did no longer
Jean-Marc Dewaele
include the L1 among their dominant languages. The latter did report significantly lower proficiency levels in the L1 and lower levels of use of the L1, suggesting that they belong in the category of perceived “L1 attriters”, but even within that category there is a large amount of inter-individual variation. In sum, we are certainly not claiming that self-reported data obtained via questionnaires suffice for the analysis of the complex phenomenon of attrition. It merely offers a possible third triangulation point, in addition to production data and in-depth face-to-face interviews.
Acknowledgment We would like to thank Aneta Pavlenko, Catherine Harris and the four anonymous reviewers for their excellent feedback on a previous version of this chapter.
Notes . “I dedicate the diary of my youth (or a brief tale) to my beloved nephews Ilan, Alison and Josh as a souvenir from Bobe Bela” [my translation]. . “Yiddish, for me, was truly the mother tongue, whereas Spanish, the street language, the one I most often used, was the father tongue” (Stavans 2001: 83). . The danger of respondent self-selection is more acute if the dependent variable under consideration has a perceived “desirable” end as it might discourage a sizable proportion of potential participants. Dörnyei notes: “volunteers may be different from non-volunteers in their aptitude, motivation, or some other basic characteristic, and drop-outs may also share some common features that will be underrepresented in the sample with their departure (e.g., drop-outs may be more unmotivated than their peers and therefore their departure might make the remaining participants’ general level of motivation unnaturally high)” (2003: 75). As the main dependent variables in the present study are relatively valueneutral: ”how often do you use language X to express Y...”, it seems less likely that respondent self-selection would unduly skew the results. . I will use the term “LX dominants” and “perceived L1 attriters” interchangeably, with the understanding that this is a subjective, self-reported fact, and impossible to verify without production data. . Dörnyei observes that: “L2 researchers typically want to measure many different areas in one questionnaire, and therefore cannot use very long scales (...). This means that somewhat lower Cronbach Alpha coefficients are to be expected, but even with short scales of 3–4 items we should aim at reliability coefficients in excess of .70; if the Cronbach alpha of a scale does not reach .60, this should sound warning bells” (2003: 112).
Language dominance and language preference for emotional speech
References Belcher, D., & Connor, U. (2001). Reflections on Multiliterate Lives. Clevedon: Multilingual Matters. Besemeres, M. (Forthcoming). “Different languages, different emotions? Perspectives from autobiographical literature”. Special issue of the Journal of Multilingual and Multicultural Development (2004). Bond, M., & Lai, T.-M. (1986). “Embarrassment and code-switching into a second language”. The Journal of Social Psychology, 126, 179–186. Bot, K. de, & Hulsen, M. (2002). “Language Attrition: Tests, self-assessments and perceptions”. In V. Cook (Ed.), Portraits of the L2 User (pp. 253–274). Clevedon: Multilingual Matters. Bot, K. de (1998). “The psycholinguistics of language loss”. In G. Extra & L. Verhoeven (Eds.), Bilingualism and Migration (pp. 345–361). Berlin: Mouton de Gruyter. Cook, V. (2002). “Background to the L2 user”. In V. Cook (Ed.), Portraits of the L2 User (pp. 1–28). Clevedon: Multilingual Matters. Dewaele, J.-M. (2002). “Psychological and sociodemographic correlates of communicative anxiety in L2 and L3 production”. The International Journal of Bilingualism, 6 (1), 23– 39. Dewaele, J.-M. (Forthcoming a). “Blistering Barnacles! What language do multilinguals swear in?! ” Estudios de Sociolinguistica, 5 (2004). Dewaele, J.-M. (Forthcoming b). “The emotional force of swearwords and taboo words in the speech of multilinguals”. Special issue of the Journal of Multilingual and Multicultural Development (2004). Dewaele, J.-M., & Pavlenko, A. (2001). Web questionnaire Bilingualism and Emotions (http://www.bbk.ac.uk/llc/biling+emotions/). Dewaele, J.-M., & Pavlenko, A. (2002). “Emotion vocabulary in interlanguage”. Language Learning, 52 (2), 265–324. Digman, J. M. (1990). “Personality Structure: Emergence of the Five-factor Model”. Annual Review of Psychology, 41, 417–440. Dörnyei, Z. (2003). Questionnaires in Second Language Research: Construction, administration, and processing. Mahwah, NJ: Lawrence Erlbaum. Dufour, R., & Kroll, J. (1995). “Matching words to concepts in two languages: A test of the concept mediation model of bilingual representation”. Memory and Cognition, 23, 166–180. Eysenck, H. J., & Eysenck, S. B. G. (1968). Manual for the Eysenck Personality Inventory. San Diego, CA: Educational and Industrial Testing Service. Green, D. W. (1986). “Control, activation and resource: A framework and a model for the control of speech in bilinguals”. Brain and Language, 27, 210–223. Gonzalez-Reigosa, F. (1976). “The anxiety arousing effect of taboo words in bilinguals”. In C. D. Spielberger & R. Diaz-Guerrero (Eds.), Cross-cultural Anxiety (pp. 89–105). Washington, D.C.: Hemisphere. Harris, C. L. (Forthcoming). “Bilingual speakers in the lab: Psychophysiological measures of emotional reactivity”. Special issue of the Journal of Multilingual and Multicultural Development (2004).
Jean-Marc Dewaele
Harris, C. L., Ayçiçegi, A., & Berko-Gleason, J. (2003). “Taboo words and reprimands elicit greater autonomic reactivity in a first than in a second language”. Applied Psycholinguistics, 24, 561–579. Hulsen, M. (2000). Language Loss and Language Processing: Three generations of Dutch migrants in New Zealand. Unpublished Doctoral Dissertation: University of Nijmegen. Jaspaert, K., Kroon, S., & van Hout, R. (1986). “Points of reference in first-language loss research”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 37–49). Dordrecht: Foris. Javier, R., & Marcos, L. (1989). “The role of stress on the language-independence and codeswitching phenomena”. Journal of Psycholinguistic Research, 18, 449–472. Kellman, S. (2000). The Translingual Imagination. Lincoln, NB: University of Nebraska Press. Kinginger, C. (Forthcoming). “Bilingualism and emotion in the autobiographical works of Nancy Huston”. Special issue of the Journal of Multilingual and Multicultural Development (2004). Knowlson, J. (1996). Damned to Fame. The life of Samuel Beckett. London: Bloomsbury. Kroll, J., Michael, E., Tokowicz, N., & Dufour, R. (2002). “The development of lexical fluency in a second language”. Second Language Research, 18, 141–175. Novakovich, J., & Shapard, R. (Eds.). (2000). Stories in the Stepmother Tongue. Buffalo, NY: White Pine Press. Pavlenko, A. (1998). “Second language learning by adults: Testimonies of bilingual writers”. Issues in Applied Linguistics, 9 (1), 3–19. Pavlenko, A. (2001). “Language learning memoirs as a gendered genre”. Applied Linguistics, 22 (2), 213–240. Pavlenko, A. (2002a). “Poststructuralist approaches to the study of social factors in L2”. In V. Cook (Ed.), Portraits of the L2 User (pp. 277–302). Clevedon: Multilingual Matters. Pavlenko, A. (2002b). “Bilingualism and emotions”. Multilingua, 21 (1), 45–78. Schmid, M. S. (2002). First Language Attrition, Use and Maintenance: The case of German Jews in Anglophone countries. Amsterdam and Philadelphia: John Benjamins. Schmid, M. S. (Forthcoming). “Identity and first language attrition: A historical approach”. Special issue of Estudios de Sociolinguistica, 5 (2004). Söter, A. (2001). “Straddling three worlds”. In D. Belcher & U. Connor (Eds.), Reflections on Multiliterate Lives (pp. 67–73). Clevedon: Multilingual Matters. Stavans, I. (2001). On Borrowed Words. A memoir of language. New York, NY: Viking Penguin. Waas, M. (1996). Language Attrition Downunder. Frankfurt: Peter Lang. Ya˘gmur, K. (1997). First Language Attrition among Turkish Immigrants in Sydney. Tilburg: Tilburg University Press. Yamamoto, M. (2001). Language Use in Interlingual Families: A Japanese-English sociolinguistic study. Clevedon: Multilingual Matters.
The role of grammaticality judgments in investigating first language attrition A cross-disciplinary perspective Evelyn P. Altenberg and Robert M. Vago Hofstra University / Queens College and The Graduate Center, City University of New York
Introduction The grammaticality judgment task has been widely used in a variety of related areas of inquiry, including theoretical linguistics, first language (L1) acquisition, second language (L2) acquisition, and communicative disorders. More recently, the grammaticality judgment task has been included among the methodologies used in the emerging field of L1 attrition; see Altenberg (1991), de Bot et al. (1991), Grosjean and Py (1991), Köpke (1999), Polinsky (1994), Seliger (1991), and Sharwood Smith (1983). We will use the term L1 attrition (roughly equivalent to primary language attrition; see Seliger 1996) to refer to the loss of language abilities of nondisordered individuals in an L2 environment. The popularity of the grammaticality judgment task is understandable. First, it can access structural attributes which occur infrequently in actual speech production. Second, it can tell researchers not only which structures are possible, but also which ones are not, reaching into a domain that is not overtly observed. As Schmid (2002: 37) points out, “the overwhelming majority of language attrition studies have concentrated on ‘what is lost’ to the exclusion of ‘what is retained”’. The grammaticality judgment task can potentially tell us both. Third, it addresses the problem of avoidance (see Schachter 1974), whereby speakers stay clear of structures whose grammaticality they are not sure of. Finally, it can typically be carried out without special equipment, in a
Evelyn P. Altenberg and Robert M. Vago
short amount of time, and with groups of subjects at once. While the grammaticality judgment task has most commonly been conducted to examine aspects of syntax, it can, and has, on occasion been used in other domains, for example, phonology (e.g., Altenberg & Cairns 1983; Greenberg 1987). The grammaticality judgment task has come under a great deal of scrutiny, with questions raised about both its validity and its methodological implementation. Issues related to the task have been widely discussed and reviewed in the literature. In this regard, we want to bring special attention to the works of, among others, Chaudron (1983), for an extensive review of experimental psycholinguistic studies concerning grammaticality judgments; Birdsong (1989), for examining the role of grammaticality judgments in L2 learning, and Cowart (1996) and Schütze (1996) for detailed examinations of issues relating to the syntactic grammaticality judgments of nondisordered adult L1 speakers. Our goal here is to review the issues related to the grammaticality judgment task, and to determine what we do and do not know about it, in particular with regard to those issues which seem to be most relevant to assessing the potential role the grammaticality judgment task might play in investigating L1 attrition. Our discussion is organised around four central questions: (1) What is the relationship between grammaticality judgments and competence? (2) Do grammaticality judgments rely on knowledge and processes that are different from those used in normal sentence processing? (3) What is the significance of possible inter-subject and intra-subject inconsistency? (4) How can one best set up a grammaticality judgment task? We will conclude that grammaticality judgment tasks are worth pursuing in L1 attrition studies, subject to significant controls.
.
The relationship between grammaticality judgments and competence
An issue which has received attention in the L1 attrition literature is that of whether L1 attrition is a loss of competence, or performance, or both, with researchers taking a variety of positions (see for example Köpke & Schmid, this volume; Polinsky 1997; Seliger 1996; Schmid 2002; Sharwood Smith 1983). The relationship between grammaticality judgments and competence is therefore of particular relevance here. Much of the early research using grammaticality judgment tasks, for example Arthur (1980) and Kellerman (1986), was conducted on the assumption that such tasks directly tap into competence; for a more recent exposition of this view, see Seliger (1996). However, there is generally agreement today that
Grammaticality judgments in first language attrition
grammaticality judgment tasks do not provide a direct window into an individual’s competence alone, but rather other factors play a role as well. We present here just a few quotations out of a potentially much longer list. Chomsky (1986: 36): “In general, informant judgments do not reflect the structure of the language directly; judgments of acceptability, for example, may fail to provide direct evidence as to grammatical status because of the intrusions of numerous other factors”. Chaudron (1983: 345, Footnote 2): “Judgments must be recognised as a secondary form of access to underlying competence, primarily because of the mediation of possibly conflicting knowledge, attitude, or performance variables in the expression of a judgment response”. Gass (1994: 306): grammaticality judgments “are not a direct reflection of competence for competence is an abstraction”. Han (2000: 180): “grammaticality judgment tasks, like any other data collection instrument, provide a kind of performance data. They are no longer assumed to have undisputed validity as an index of linguistic competence, as was claimed in the past”. Finally, consider Schachter and Yip (1990: 381): grammaticality judgment tasks “reflect a complex interaction between competence and performance variables inherent in the judgment task”. There are, in fact, good reasons to assume that judgments involve performance. After all, in order to assess a sentence, one presumably has to first process it, or attempt to process it; thus, at least some of the performance factors involved in normal sentence processing ought to play a role in grammaticality judgment tasks. If our judgments provided a direct window into competence and only competence, then speakers should consistently be able to recognise all the meanings of an ambiguous sentence that are allowed by their grammar. Chomsky (1965) and others have pointed out that this is not the case: speakers do not necessarily recognise all the meanings of an ambiguous sentence. A similar argument can be made for garden-path sentences. A sentence such as the well-known The horse raced past the barn fell may be judged as ungrammatical even though it is grammatical. However, as Pinker (1994: 211) demonstrates, the very same sentence may be judged grammatical if placed in a different context: “The horse that was walked past the fence proceeded steadily, but the horse raced past the barn fell”. Again, sentences of this sort ought to be consistently judged as grammatical in a grammaticality judgment task if such a task directly accessed competence and only competence. While it would certainly be convenient to have a task that directly tapped competence, it is important to make clear that it is not necessarily our goal to understand only competence, or to somehow value an understanding of
Evelyn P. Altenberg and Robert M. Vago
competence over an understanding of performance. Eckman (1994: 6) points out that “the boundary between linguistic competence and performance is not fixed but instead is adjusted as the field advances”. He argues (ibid.: 9) that “a linguist establishes empirically, not a priori, what is included under that competence”, and suggests (ibid.: 10) that one cannot “give a compelling argument that a given theory must restrict itself to the study of some idealised construct”. Our concern here is simply to know what we are getting when we conduct a grammaticality judgment task, in particular with L1 attriters.
. Do grammaticality judgments rely on knowledge and processes that are different from those used in normal sentence processing? It seems reasonable, then, to proceed on the understanding that grammaticality judgment data are performance data, which may incorporate aspects of competence, rather than solely a direct reflection of competence. The question is, what aspects of performance do they reflect? In particular, do grammaticality judgment tasks rely on knowledge and processes that are different from those used in on-line, normal sentence processing? This question has been raised, for example, by Bever (1970, 1974). Our discussion will focus on four issues that bear significantly on this question. . Similarities between grammaticality judgment tasks and normal language use As we think through the role of grammaticality judgment tasks, it is important to note that events similar to making grammaticality judgments occur in normal language use. When a native speaker hears an utterance and says That sounds weird, or Huh? we need to account for how that response is arrived at. That is, there has to be a mechanism, during on-line processing, for recognising an ungrammatical sentence and also for recognising one that needs to be re-parsed, i.e., that is ambiguous. Meng and Bader (2000: 646), using a timed grammaticality judgment task, claim that “the same mechanisms are involved in response to a local syntactic violation, no matter whether the sentence is ambiguous, i.e., revisable, or ungrammatical”. Are these mechanisms and the knowledge base they utilise the same as, or different from, that used in detecting ungrammaticality in a grammaticality judgment task? Similarly, in production, speakers self-correct as they speak. They must in some way be monitoring their speech to do this. While speakers are clearly
Grammaticality judgments in first language attrition
monitoring for features other than grammaticality (e.g., content and clarity), they are at times correcting grammaticality. Cowart (p.c.) points out that in order for a community to maintain its grammatical coherence in light of the forces of language change, individuals would seem to need some sort of cross check between a representation of current grammatical norms in the community and their own speech. Is monitoring part of on-line processing or does it rely on a separate process and knowledge base? We see then that there are a variety of activities that speakers engage in that require at least a similar knowledge base: the recognition of ungrammaticality and revision of ambiguity during normal perception, monitoring during normal production, and making grammaticality judgments. The most parsimonious system is one in which the knowledge source is the same for each of these activities. In fact, Allen and Seidenberg (1999: 121), operating from a connectionist perspective, suggest that “the capacity to make...judgments emerges out of the ability to process language normally”. Note, however, that Bever (1975) suggests that there is a representation of grammatical knowledge that is separate from production and comprehension mechanisms in adults. The fact that one knowledge base is the most parsimonious proposal does not necessarily make it correct. The point here is simply that the knowledge used in grammaticality judgment tasks may not be far removed from that used in more typical on-line activities. . Explicit and implicit memory A distinction that is crucial to our discussion here is that between explicit and implicit knowledge. According to Berry (1994: 147), implicit memory “is shown on tasks that do not require deliberate recollection of a past event although the event influences performance”. Paradis (1994), linking implicit knowledge and competence, views implicit competence as knowledge that individuals themselves are not aware of. It “is acquired incidentally, is stored in the form of procedural know-how, without conscious knowledge of its contents, and is used automatically” (Paradis 1995: 6). Winter and Reber (1994: 118) say that implicit knowledge “typically remains resistant to conscious inspection”. Explicit knowledge, according to Paradis (1994: 394), refers to “the knowledge of which individuals are aware and that they are capable of representing to themselves and of verbalizing on demand”. He suggests that it is usually learned in school. For Bialystok (1986: 499), explicitness is a matter of degree: “Analysis of linguistic knowledge is the skill component responsible for making explicit those representations that had previously been implicit or intuitive”. Note that
Evelyn P. Altenberg and Robert M. Vago
while Paradis (1994) includes the ability to verbalise knowledge as a typical characteristic of explicit knowledge, not all researchers agree. It is generally assumed that implicit learning leads to implicit knowledge, explicit learning to explicit knowledge. Stanovich (1990: 91) warns that the use of terms like conscious and awareness have led to confusion in the research on reading, and that it would therefore be wise to attempt to avoid such confusion in other disciplines. Durkin (1994: 524–525), points out that “it is difficult to find a robust statement of exactly what implicit memory is or does”, and that “the very definition of the implicit/explicit distinction is by no means settled”. Nonetheless, those of us who have taught an L2 are well aware that there are students who can write a rule out perfectly on an examination, do exercises in class in which the rule is perfectly adhered to, and yet, in conversation, not follow the rule at all. Of course, it is precisely experiences such as these that lend support to Krashen’s (1982) distinction between the acquisition and the learning of an L2, where acquisition is an implicit process and learning is an explicit one. While recognising the different ways that the terms implicit and explicit have been used, we will assume, for the moment, that implicit knowledge is not conscious and is used automatically, and that explicit knowledge of language is conscious knowledge that has been learned, as Paradis suggests, primarily in school. For present purposes, we will assume that implicit and explicit knowledge do not lie on a continuum. Further, we recognise the possibility of specific knowledge being both explicit and implicit. This is the case, for instance, when a native speaker obtains explicit knowledge of a rule, say, through taking a linguistics course. Does a metalinguistic task such as the grammaticality judgment task necessarily rely on explicit knowledge? A metalinguistic task is one that requires subjects to do something to language, rather than to do language; it is offline rather than on-line. Birdsong (1989: 1, Footnote 1) states that “the term [metalinguistic performance] requires a rather vague interpretation, something on the order of attention to linguistic form rather than content”. According to Chaudron (1983: 344), things are metalinguistic that “treat language as an object”. Metalinguistic tasks come in many varieties: for example, tasks in which speakers decide if two sentences mean the same thing, tasks in which speakers segment words into syllables, or they can be grammaticality judgment tasks. There is no way of knowing, a priori, whether or not a metalinguistic task is dependent, either fully or partially, on explicit knowledge. Certain metalin-
Grammaticality judgments in first language attrition
guistic tasks, such as counting the number of syllables in a word or judging the synonymy of two sentences in someone’s L1, may rely largely on information that has never been explicitly learned, while others, such as proofreading for spelling errors, may rely almost entirely on explicit knowledge. Thus, a metalinguistic task may rely on metalinguistic, explicit knowledge, or it may not. This fits in with Birdsong’s (1989: 50) position that metalinguistic ability is unlikely to rely on a single set of skills, and “should not be regarded as a unitary concept”. Ideally, for each metalinguistic task one needs to determine the kinds of knowledge and processes involved. Let us return now to the specific metalinguistic task with which we are concerned here, namely the grammaticality judgment task. Do grammaticality judgment tasks tap explicit knowledge, implicit knowledge, or both? Before addressing this question, we need to make clear that it rests on somewhat shifting sands. First, it assumes that researchers are in agreement on the definition of the terms explicit and implicit. But as we have just seen, this is not the case. And second, it assumes that the implicit/explicit distinction is a valid one. Given the problem with the first assumption, it is difficult to assess the second. Still, there is a large body of literature supporting the notion of these two kinds of memory stores (see Berry 1994, for example, for references). With these concerns in mind, let us proceed to the question we just posed, namely the type of knowledge grammaticality judgment tasks utilise. The answer may vary, depending on the population we are considering. Let us focus, for the moment, on children acquiring their L1. Various versions of grammaticality judgment tasks have been conducted with this group of speakers. Some ask children whether a sentence sounds right or wrong to say; this approach can be used to elicit well-formedness judgments from children as young as two years old (Hiramatsu 2001; McDaniel & Cairns 1996). Early L1 acquisition must be a case of implicit learning. Presumably, it thus leads to implicit knowledge. For example, Sharwood Smith (1994: 33) states: “Little children acquiring their mother tongue clearly do not have recourse to rules in this conscious, explicit sense”. And Paradis (1994: 403) claims that “explicit representations are nonexistent in children”. Therefore, young children, who have primarily implicit knowledge of language, must be using implicit knowledge in their responses in grammaticality judgment tasks. This is not to say that performance factors (some obvious examples being fatigue and inattention) cannot play a role in children’s judgments; clearly, they can. But what we can say is that explicit knowledge cannot be the sole source of children’s judgments. Therefore, if we want to say that they are a source of adults’
Evelyn P. Altenberg and Robert M. Vago
judgments, we need to claim that children use primarily implicit knowledge in making grammaticality judgments, but adults do not. Do adults making judgments in their native languages rely only on implicit knowledge, only on explicit knowledge, or on both? Perhaps the most important argument for the claim that grammaticality judgment tasks cannot rely entirely on explicit knowledge is the fact that native speakers are able to make judgments about structures with no explicit knowledge about them. If grammaticality judgment tasks relied only on explicit information, then native speakers would not be able to make correct judgments about any structure which had not explicitly been taught to them and/or whose structure they did not explicitly understand. Native English speaking college students can tell, for example, that the tag question *John isn’t leaving early, can he? is ungrammatical without being able to say why. Linguists use their intuitions to help them figure out rules; they do not use their rules to tell them what their intuitions are. The very fact that linguists base their analysis of a syntactic rule on native speaker intuitions means that the judgment, or intuition, comes first, before explicit knowledge of the rule. Hence the judgment cannot be based on explicit knowledge. De Bot et al. (1991), examining L1 attrition, compare the loss of metalinguistic skills in a grammaticality judgment task and the loss of skills using an interview task. They find less loss in metalinguistic skills. Their explanation (p. 94) is that “the linguistic skills are lost first because without adequate metalinguistic ‘monitoring’ language use as such is hampered”. That is, their explanation assumes that normal language use requires metalinguistic monitoring. Under this scenario, at least for production, metalinguistic monitoring is a part of the normal process of speech and as such cannot rely exclusively on explicit knowledge. Do adult responses on grammaticality judgment tasks rely only on implicit knowledge, or can they rely at least in part on explicit, particularly, prescriptive notions held by the subjects? Paradis (1994) indicates that one may know an utterance is wrong because of either implicit competence or explicit knowledge. He suggests (p.c.) that in L1 grammaticality judgments people tend to go by prescriptive grammar rather than by what they actually say. Sorace (1988) and Munnich, et al. (1994) note as well the possible influence of prescriptive knowledge on judgments. According to Paradis (p.c.), the same may be true for the L1 attriter, who is, after all, a native speaker. It thus seems reasonable to suggest that grammaticality judgment tasks rely on implicit knowledge which, for adults, though not for young children, can be supplemented by explicit, in particular prescriptive, knowledge. If that is the
Grammaticality judgments in first language attrition
case, how can one interpret the results of grammaticality judgment tasks? For L1 attriters, we need to ask the same sorts of questions as for non-attrited native speakers in order to determine what their responses are based on. At the very least, we need to determine the extent of their education and how likely they are to have been taught the structures we are interested in. We can also check textbooks and curricula from the time and place they went to school to see if the structures of interest are discussed. The extent of explicit knowledge would be expected to vary, depending on the circumstances of L1 use: for instance, if someone has gone back to school to brush up on his or her L1, if someone is extremely uncertain about his or her L1 proficiency, or if someone is, say, teaching his or her L1, in which case he or she would be expected to have more explicit knowledge. One might also have subjects rate sentences about which they are likely to have prescriptive ideas, as well as those about which they probably do not have prescriptive ideas. For example, in English, one could compare judgments for sentences such as Who did you see?, Whom did you see?, and *See you did who?. Techniques such as these would help researchers determine the effect of prescriptive notions on native speakers’ ratings and to tease these apart from implicit knowledge. Along these lines, a number of researchers have conducted think-aloud tasks, in which subjects discuss their judgments as they are making them. R. Ellis (1991: 179), using such a task, reports that “the learners appeared to rely extensively on ‘feel’ in making their judgments”. Similar results were found by Goss et al. (1994). While Cowan and Hatasa (1994) are rightly concerned about the potential subjectivity involved in assessing responses in think-aloud grammaticality judgment tasks, the findings of think-aloud tasks thus far suggest that implicit knowledge is the primary basis for native speaker decisions. Attempts to develop such tasks so that their responses are assessed less subjectively may be worthwhile avenues for future research, although Cowart (p.c.) points out that the relation of these tasks to the ordinary course of language comprehension is another complex issue. . Findings of grammaticality judgment tasks and other tasks If grammaticality judgment tasks tap, at least to some extent, the knowledge and processing used in on-line processing, should we then expect associations between the findings of grammaticality judgment tasks and those of other tasks? A number of studies have found such associations. Rice et al. (1999) performed a grammaticality judgment task with children with specific lan-
Evelyn P. Altenberg and Robert M. Vago
guage impairment (SLI), along with a normal control group. They concluded (p. 959) that “children with SLI, and younger unaffected children, are likely to interpret simple clausal structures, under simple input conditions, according to the same principles that guide their productions”. Blackwell et al. (1996) conducted a word-by-word grammaticality judgment task, a cloze task, and an on-line judgment task with native speakers and found high correlations among the three tasks. They suggest (p. 337) that “the incremental tasks were tapping into the same decision-making process as is found on-line”. Similarly, Greidanus and van der Linden (1986) conclude that the findings of a grammaticality judgment task and a translation task are essentially the same. Finally, consider Munnich et al.’s (1994: 228) conclusion: [The results with L2 learners] indicate a certain convergence and comparability between the elicited imitation and the grammaticality judgment tasks with respect to the linguistic phenomena investigated. This finding suggests that both of these tasks evaluate developing language abilities in a similar manner while reflecting task-specific differences.
Other studies have found dissociations between grammaticality judgment tasks and other tasks. For example, Tomioka (2002) reports, in a study with L2 learners of English, that subjects who did well on a grammaticality judgment task did extremely poorly on a repetition task. And Linebarger et al. (1983) found that agrammatic subjects were able to make highly accurate judgments but could not comprehend sentences. Wulfeck (1988) corroborated this finding independently. In L1 attrition, de Bot et al. (1991) found evidence of attrition using an interview task, but little evidence using a grammaticality judgment task. As discussed by R. Ellis (1991), some studies have found that beginning L2 learners did worse on grammaticality judgment tasks than on other tasks and that differences disappeared with advanced learners. According to R. Ellis (1991: 166): “These studies suggest that for beginning learners, at least, judging the grammaticality of sentences and producing L2 structures constitute different and unrelated types of behaviour”. Note that this difference between beginning and advanced learners is, in fact, in the opposite direction from what one would expect if one assumes that beginning L2 learners rely largely on explicit knowledge in making judgments. What conclusion can we draw from dissociations between grammaticality judgment tasks and other tasks? Unfortunately, there is no way of knowing, a priori, whether the dissociation is due to the fact that the tasks tap into different kinds of knowledge or simply because of factors related to participating
Grammaticality judgments in first language attrition
in the tasks themselves. Some studies, for example, have found dissociations among tasks that all appear to tap implicit memory (e.g., Roediger et al. 1989). Should we conclude that they necessarily access different memory stores? What if we ask subjects to imitate sentences under two conditions: while sitting comfortably on a chair and while being distracted, say by a menacing dog? While performance in the distraction condition may be assumed to be less accurate, this does not mean that the tasks necessarily draw on different sources of knowledge. If we are reluctant to draw conclusions based on dissociations between tasks, should we be equally reluctant to draw conclusions based on associations between tasks? The possibility always exists, if two tasks converge on a pattern, that the tasks use entirely different processes and sources of knowledge, and that their convergence is simply coincidental. However, the likelihood of this happening by chance, especially with multiple tasks, seems slim, and the question is one that may well be amenable, depending on the design, to statistical analysis. Along these lines, a variety of investigators have suggested that if multiple tasks converge on a core of knowledge, we can have some confidence that we have identified knowledge rather than task effects. See, for example, Birdsong (1989), Chaudron (1983), Flynn (1986), Jaspaert et al. (1986), Klein and Martohardjono (1999), Martohardjono (1998), and Schütze (1996). Thus, Klein and Martohardjono (1999: 16) suggest that “performance effects are likely to be different for different tasks, whereas knowledge should remain constant across tasks. By using a multiplicity of tasks on both L2 learners and native speaker controls, it might be possible to tease apart knowledge from task effects”. Schütze (1996: 196) states: ...if the same results show up reliably across additional types of tasks, such as unwarned judgments, ...short-term memory measures, ...sentence completion tasks, or naturalistic observation of speech and writing, then the odds are much higher that the evidence does represent a convergence on fundamental underlying knowledge.
Flynn (1986: 155) cautions, however, that “we need converging evidence from a wide range of experimental tests in which the methodology is precisely controlled and we must be informed about how different tasks relate to each other”. One difficulty is that, in order to understand how tasks work, we need to investigate them; in order to investigate them, we often use the tasks themselves as tools. Nonetheless, these arguments suggest that using a variety of tasks with L1 attriters, including grammaticality judgment tasks, may bring us closer to understanding the attriter’s underlying knowledge.
Evelyn P. Altenberg and Robert M. Vago
. Judgment factors other than grammaticality Are there factors other than grammaticality on which subjects may be basing their decisions? Schütze (1996) proposes a model in which a variety of factors play a role in making grammaticality judgments. These factors are: parsing strategies, general knowledge of the world, production strategies, competence (the same competence used in perception and production), analysed linguistic knowledge (i.e. explicit knowledge), nonlinguistic control strategies, and metalinguistic strategies. The last three are used primarily for making judgments; the first four are also used in on-line production and comprehension. Other researchers have suggested a variety of criteria as well. According to Birdsong (1989: 33), “among children, non-structural considerations (plausibility, truth value) appear to be criteria for judgments of acceptability”. Yet the experiments discussed by McDaniel and Cairns (1996) do not support this. Scholes and Willis (1987) suggest that illiterates use semantic rather than syntactic criteria in judging sentences. However, Schütze (1996), after examining the relevant evidence, concludes that although their explicit grammatical knowledge is affected, illiterate individuals are able to identify ungrammaticality. With regard to L2 learners, Jordens (1991: 201) suggests that “grammaticality judgments are the outcome of a production task in which all kinds of linguistic information may interact”. These include L1 intuitions, L2 knowledge, and the strategy of rejecting unfamiliar structures. And Bard et al. (1996: 33) state: “Impressions of acceptability may be based, for example, on estimated frequency of usage, on conformity to a prescriptive norm or a prestigious register, or on degree of semantic or pragmatic plausibility”. Kellerman (1985) and Birdsong (1989) suggest that we also need to determine which aspect of the stimulus (e.g., vocabulary, pronunciation, syntax) subjects focus on, as well as the locus of their attention. How can researchers handle these issues? Birdsong (1989: 110) states: We have seen that when a researcher is unable to direct subjects’ attention toward certain types of errors, there is the risk of falling into the trap identified by Kellerman, i.e., assuming that subjects have spotted just the deviance the researcher has in mind. On the other hand, if ... the procedure were to be successful in priming respondents to look for this or that type of error, the results would not be reflective of their normal disposition toward the stimuli, but rather of an experimentally induced perceptual set.
Note that other tasks, e.g., elicited imitation, may be subject to the same questions. However, carefully selected stimuli can help to control some of these
Grammaticality judgments in first language attrition
variables, for example, by having pairs of stimuli that differ only in the variable under consideration (see Birdsong 1989; Cowart 1996, for discussions). R. Ellis (1990) suggests that having subjects make corrections to stimuli they judge as ungrammatical can help show what it is that they are focusing on, while Schütze (1996) recommends having a think-aloud component during practice trials to determine if subjects are using the criteria one wants. Along these lines, if a structure is given a relatively low grammaticality rating, does that indicate that it is not a possible structure for that individual, or simply a less preferred structure (Cairns 1999; Lakshmanan & Teranishi 1994)? That is, a sentence might be less preferred, due to, for example, frequency or pragmatic factors. The role of preference can perhaps be investigated by constructing sentences equal in grammaticality but different in preferentiality and testing them with different instructions. The goal would be to identify instructions that elicit grammaticality, but not preference, judgments. For example, for English, one could compare the ratings for sentences such as John bought the book with The book was bought by John, assuming that an objective measure had already determined that actives are preferred to passives. What is the role of parsing? Various relationships between the grammar and parser have been proposed (see for example Pritchet 1992). The most reasonable conclusion at this point is probably, as Schachter and Yip (1990: 380) state: “While it is uncontroversial that grammar and parser must bear some kind of relationship since they have to assign compatible structures, it is not clear exactly how they are related”. Similarly, according to McDaniel and Cairns (1996: 252) “little is actually known about the interaction of parsing and judgments”. Until more is known about parsing, it is difficult to assess its effect on grammaticality judgments, which must remain, at this point, an open question. Still, it would be wise to avoid constructions known to involve parsing difficulty, such as center embeddings, when constructing grammaticality judgment tasks.
. Inter-subject and intra-subject inconsistency We turn now to another issue which has been of concern in investigations of grammaticality judgment tasks. Various researchers (e.g. Birdsong 1989; Ellis 2002; Sorace 1988) have noted the lack of consistency in subjects’ responses on grammaticality judgment tasks, observing both inter-subject and intrasubject variability. Inter-subject variability is perhaps not problematic, since it has never been claimed that all individuals have the identical knowledge base
Evelyn P. Altenberg and Robert M. Vago
for a given language. It is also a fact of psycholinguistic testing in general. And with proper methodology, we may find that inter-subject variability is less of an issue than has been thought. Indeed, Cowart (1996), in a series of experiments, reports statistically stable responses from subjects across sentence types. And Fernández (2000) finds greater variability with bilinguals than with monolinguals, suggesting that one might be able to isolate some factors contributing to inter-subject variability where it does exist. We concern ourselves here with intra-subject variability, the fact that the same subject may respond differently to the same or similar structures at different times. Given that grammaticality judgment tasks do not directly tap into competence, it is not surprising that the judgments of individuals are not necessarily consistent. On the other hand, Cowart (1996) reports on a test-retest in which he found a strong relation between judgments in the two sessions. However, we are still left with the question of how to account for intrasubject inconsistency when it does occur. First, we need to determine whether there are structures for which subjects’ responses are consistent vs. structures for which responses are not consistent. If some structures are more stable than others, it may be that we can use the grammaticality judgment task to learn something about the status of a particular structure in competence. Along these lines, Snyder (2000) induced satiation in his subjects, i.e., getting sentences initially judged as ungrammatical to be judged as grammatical upon repeated exposure; he found that satiation occurred only for some sentence types. De Bot (1999: 348) suggests: linguistic skill or linguistic knowledge is not an either/or phenomenon: some parts of this knowledge are more stable than other parts, and elements may show various shades of knowing rather than a clear distinction between what is known and what not.
In particular, with regard to L2 learners, Alanen (1999), Han (2000), Schachter et al. (1976) and Sorace (1996) suggest that inconsistency may indicate a lack of complete knowledge rather than a problem with grammaticality judgment tasks. In fact, why would people who are in the midst of learning or forgetting something not be uncertain? If this is the case, then L1 attriters’ judgments should vary in consistency, depending on the degree of attrition, and there should be correlations between their consistency of judgments and degree of attrition. It may also be that grammaticality judgment tasks are more reliable and stable in some domains (e.g., morphology, phonology) than in others. De Bot
Grammaticality judgments in first language attrition
(1999: 358) suggests, with regard to phonology, that “the high degree of automatisation of procedures on this level restricts the possibilities of speakers to consciously pay attention to their pronunciation”. Perhaps extending metalinguistic tasks to other domains, such as phonology, will shed some light on the nature of these tasks, as Schütze (1996) suggests, although we will need to be cautious about generalising from one domain to another. If inconsistency is not limited to specific structures, then other factors need to be taken into consideration. Schütze (1996: 98) suggests that intrasubject inconsistencies “are liable to be the result of factors having nothing to do with subjects’ linguistic representations, e.g., whether they are fresh or fatigued, uncooperative, attentive or distracted, etc”. Cowart (1996) makes an interesting comparison with psychophysics which may help give us some perspective on inconsistency. (Psychophysics is concerned, among other things, with responses to flashes of light, brief tones, touches of the skin.) He points out (ibid.: 6) that in this domain, “subjects routinely give variable responses to different presentations of a single stimulus” and therefore subjects are typically asked to “judge dozens or even hundreds of instances of the same stimulus”. We need to be careful, when we retest subjects in order to investigate reliability, that we are not influencing their responses by satiating them. One way to get around this might be to wait a period of time, say a few weeks, between testing sessions. Further, as Schütze (1996) suggests, we need to be certain that their judgment process in general has not changed as a result of their prior experience with the task. Thus, subjects with prior experience might be compared to subjects without prior experience. At this point, the cause of inconsistencies, when they do occur, remains an open, though empirical, question. Perhaps it is most appropriate to conclude this portion of our discussion with the following quote from Birdsong (1989: 69): “Metalinguistic data are like 25-cent hot dogs: They contain meat, but a lot of other ingredients, too”. What, then, do we need to do to get as much “meat” as possible in a grammaticality judgment task? To address this question, we turn to some of the issues that affect the way a grammaticality judgment task is set up.
. Methodological issues in setting up a grammaticality judgment task . The role of time What is the role of time in grammaticality judgment tasks? This is a particularly relevant question in L1 attrition, where de Bot (1990: 64) suggests that
Evelyn P. Altenberg and Robert M. Vago
the L1 attriter may have a “slowed-down system”. Some investigators have assumed that grammaticality judgment tasks in which speed of response is a factor tap on-line processes (Meng & Bader 2000) and/or rely on only implicit knowledge (Han 2000). Others (e.g. Blackwell et al. 1996; Kail & Bassano 1997) use word-by-word judgment tasks, in which subjects see or hear a sentence one word at a time and indicate its grammatical status as soon as they feel they know. The researchers here also assume that these tasks tap on-line processing. However, “limitations on time may detract from the fullest use of knowledge resources”, whatever that knowledge base may be, according to Chaudron (1983: 367). From that perspective, limitations on time may impose performance limitations that interfere with the full utilisation of implicit knowledge, assuming, of course, that one is accessing implicit knowledge in the grammaticality judgment task in the first place. Köpke and Nespoulous (2002) seem to be making the interesting point that errors may be found in areas that are not necessarily problems but from which resources have been withdrawn in order to focus on problem areas during processing. A timed grammaticality judgment task, then, may force subjects to withdraw resources from accessing knowledge to other areas, resulting in a less accurate assessment of knowledge. Giving subjects more time does not necessarily mean they are not responding intuitively, i.e., based on implicit knowledge. As R. Ellis (1991: 180) points out, subjects may be “repeating” a sentence, which “reintroduces the sentence into immediate memory, thus enabling the learner to produce an intuitive response”. But if this is the case, then responses for individual subjects should be the same, regardless of whether the grammaticality judgment task is timed or untimed. While Han (2000) found that responses of L2 learners on timed and untimed versions of the task were not the same overall, there was no native speaker control group. Therefore, we cannot judge whether the same finding would be true for other populations. It may be that a timed grammaticality judgment task is more likely to access implicit knowledge than explicit knowledge, but at the same time increase the likelihood of performance factors interacting with implicit knowledge. So if one had other ways of limiting or identifying explicit knowledge, as discussed earlier, it might be more profitable, though perhaps counter-intuitive, to use untimed grammaticality judgment tasks if one is interested primarily in implicit competence.
Grammaticality judgments in first language attrition
. Magnitude estimation Bard et al. (1996) and Sorace (1996) discuss the advantages of using magnitude estimation with grammaticality judgments. That is, subjects make proportional judgments, for example, by assigning a number to a sentence to indicate how acceptable it seems. This is similar to techniques that have been used to assess subjects’ ratings of factors such as brightness and loudness. . Analysing the results of individuals as well as groups Cairns (1999), R. Ellis (1990), Klein and Martohardjono (1999) and Schachter (1989), all make the point that we need to analyse the results of individuals to see if they are behaving consistently. While group tendencies can be useful in establishing general developmental trends, as Cairns (1999) points out, they can obliterate individual consistencies. This point is particularly relevant with L1 attriters, given the variability of attrition from person to person (see Altenberg 1991, for example). . Response bias As Birdsong (1989) and others have discussed, a number of studies show a response bias on the part of subjects. Response bias can be due to a number of factors, including the instructions given to subjects and the scale used in the task. Some studies (e.g. Bley-Vroman et al. 1988) show a response bias on the part of subjects towards responding not grammatical; in other studies, subjects show a response bias towards responding grammatical (e.g. Birdsong 1989; Han 2000; cf. also Wulfeck 1988; Wulfeck et al. 1991 for agrammatic aphasia patients). Paolillo (2000: 225) suggests that using a logistic regression statistical model allows us to “discriminate accuracy effects from an over-all response bias”. A non-attrited native speaker control group can also be helpful here, depending on the design of the experiment. In fact, having a non-attrited control group is important in many studies utilising the grammaticality judgment task, though defining such a group raises its own questions; see Jaspaert et al. (1986), Schoenmakers-Klein Gunnewiek (1989), and Seliger (1996). But since the problem of an appropriate control group for L1 attriters is true for many studies, not just grammaticality judgment tasks, we will not address it here. Next we turn to the important methodological details, additional factors which should be controlled in order to enhance the soundness of a grammati-
Evelyn P. Altenberg and Robert M. Vago
cality judgment task. While these factors are too numerous to discuss in detail or to list exhaustively, we present a selection here. . Subject characteristics The following subject characteristics are culled from Bradac et al. (1980), de Bot et al. (1991), Köpke and Nespoulous (2002), Schütze (1996), and Sorace (1988). These characteristics, among others, have been hypothesised to play a role in making judgments: field dependence or independence, linguistic background, handedness, age, gender, creativity, working memory capacity, the ability to reason by analogy, education, amount of language experience in both L1 and L2, social background, and attitudes towards bilingualism and language maintenance. We recognise, however, that assessing and controlling these variables is not necessarily a simple and straightforward task; see Cowart (2002). . Additional task characteristics The following non-exhaustive list of task characteristics, in addition to the role of time discussed earlier in Section 4.1, is culled from Birdsong (1989), Cowart (1996), N. C. Ellis (2002), R. Ellis (1991), Murphy (1997), Nagata (1990), and Schütze (1996): the truth of the sentence, sentence order, sentence complexity, the distribution of grammatical vs. ungrammatical sentences, subjective or objective self-awareness, modality (i.e., visual or spoken stimuli), pragmatic context, meaningfulness, position of error in a sentence, frequency of occurrence of the stimulus materials, including the frequency of the lexical items as well as the frequency of the structures themselves, linguistic complexity, nature of response requested, language of instruction and of interaction, imagery, distractor items, and degree of ungrammaticality. Note that many of the factors we have mentioned are relevant to the construction of any task, not just the grammaticality judgment task. Two factors deserve special mention for their significance. One is the kind and number of levels of rating scale used. For a discussion of the pros and cons of various rating scales, see especially Cowart (1996), Schütze (1996) and Sorace (1996). The second factor is the important role of instructions to subjects, an issue to which we have returned a number of times throughout our discussion. Cowart (1996) tested two sets of instructions, one focusing on intuitions, one focusing on prescriptive notions. He found no difference in the pattern of subject responses. Schütze (1996) suggests giving subjects specific instructions
Grammaticality judgments in first language attrition
with examples of good and bad sentences and explanations of why they are good or bad.
. Conclusion It is tempting, in light of all the serious concerns which have been raised regarding grammaticality judgment tasks, to avoid using them. However, one needs to be wary here of throwing the baby out with the bathwater. Grammaticality judgment tasks are showing something; they are measuring some kind of proficiency, i.e., they still provide data that need to be accounted for, but what that something is may be different for different populations, different task designs, and different materials. Schütze (1996: 212) states that “relatively few experiments have shown that the pattern of results is changed by the various manipulations that have been tried. This should increase our confidence that judgments do tell us about something real and important”. And Tarone (1994: 326), referring to off- and on-line methods, notes that “each provides a different kind of information about the learner’s interlanguage”. Certainly, grammaticality judgment tasks can be made more reliable, along the lines outlined here and elsewhere. Perhaps the most productive approach is to further explore what exactly it is that grammaticality judgment tasks do and how to get them to do what we want them to do. This is the conclusion reached by Blackwell et al. (1996) and Schütze (1996), among others. For example, testing different kinds of instructions, as in Cowart’s (1996) study, is crucial, as well as testing different kinds of stimuli with different instructions. A closer examination of native speakers’ response consistency, analysed by individual and by structure, over more than one testing sessions, also seems to be a fruitful avenue to pursue. That is, researchers need to focus on the grammaticality judgment task per se in order to understand this task more fully. Why bother even exploring a task associated with so much controversy? Note in this regard that other tasks, such as the elicited imitation task, are controversial, too. Still others may not be controversial, but only because they have not yet been as closely scrutinised. Birdsong (1989: 208) claims that “our understanding of metalinguistic performance is, at present, limited. What we know is that it is complex. But so too are all other higher-order activities involving a translation of knowledge into behavior”. Or as Schütze (1996: 10) aptly puts it: “While the psychology of grammaticality judgments might hold as many complexities and mysteries as language itself, that is no reason for
Evelyn P. Altenberg and Robert M. Vago
despair or dismissal – it is all the more reason for us to begin the task of unravelling them”. As we consider applying the grammaticality judgment task to the investigation of L1 attrition, there are a number of points to keep in mind. While recognising that grammaticality judgment tasks do not directly reflect only competence, we can reasonably assume that L1 attriters are likely, in most cases, to rely heavily on implicit knowledge in judging grammaticality. Furthermore, we can devise procedures to help us elucidate the role that any explicit knowledge may be playing in L1 attrition judgments. Given that making judgments relies on a knowledge base that may not be far removed from that used in on-line activities, it will be interesting to see, as research continues, where L1 attriters’ judgments diverge from those of non-attrited controls. For L1 attriters, the degree of intra-subject consistency may be particularly significant, as inconsistency may at least partially reflect degree of attrition. Perhaps our most productive approach is to conduct carefully constructed grammaticality judgment tasks and administer them in conjunction with multiple other tasks to L1 attriters, while recognising the limitations in our understanding of the grammaticality judgment task itself. Our discussion highlights the fact that those of us who pursue the use of grammaticality judgments with L1 attriters need to be cautious as we face the rather daunting task of controlling the subject and task variables involved. If we can make grammaticality judgment tasks work for us in investigating L1 attrition, the advantages we mentioned at the outset of our discussion remain. Ultimately, the close scrutiny under which grammaticality judgment tasks have come, and one hopes will continue to come, can only lead to a greater understanding of their usefulness and of what they can tell us about many issues, including the abilities of L1 attriters.
Note Authors’ names are listed in alphabetical order. We are indebted to Ronald L. Bloom, Helen Smith Cairns, Wayne Cowart, James J. Drieves, Eva Fernández, Gita Martohardjono, Monika S. Schmid, and Carolyn Sobel for comments and suggestions.
Grammaticality judgments in first language attrition
References Alanen, R. (1999). “Grammaticality judgments and reaction time measurement: A tool for analyzing the use of second language knowledge”. Dissertation Abstracts International, Section C: Worldwide, 60 (1), Item-20. Allen, J., & Seidenberg, M. S. (1999). “The emergence of grammaticality in connectionist networks”. In B. MacWhinney (Ed.), The Emergence of Language: 28th Carnegie Mellon Symposium on Cognition (pp. 115–151). Hillsdale, NJ: Lawrence Erlbaum. Altenberg, E. P. (1991). “Assessing first language vulnerability to attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 189–206). Cambridge: CUP. Altenberg, E. P., & Cairns, H. S. (1983). “The effects of phonotactic constraints on lexical processing in bilingual and monolingual subjects”. Journal of Verbal Learning and Verbal Behavior, 22, 174–188. Arthur, B. (1980). “Gauging the boundaries of second language boundaries: A study of learner judgments”. Language Learning, 30, 177–194. Bard, E. G., Robertson, D., & Sorace, A. (1996). “A magnitude estimation of linguistic acceptability”. Language, 72, 32–68. Berry, D.C . (1994). “Implicit and explicit learning of complex tasks”. In N. C. Ellis (Ed.), Implicit and Explicit Learning of Languages (pp. 147–164). London: Academic Press. Bever, T. G. (1970). “The cognitive basis for linguistic structures”. In J. R. Hayes (Ed.), Cognition and the Development of Language (pp. 279–362). New York, NY: John Wiley. Bever, T. G. (1974). “The ascent of the specious, or There’s a lot we don’t know about mirrors”. In D. Cohen (Ed.), Explaining Linguistic Phenomena (pp. 173–200). New York, NY: John Wiley. Bever, T. G. (1975). “Psychologically real grammar emerges because of its role in language acquisition”. In C. A. Ferguson & D. P. Dato (Eds.), Sound Patterns in Language Acquisition [Georgetown University Round Table in Languages and Linguistics] (pp. 63–75). Washington, DC: Georgetown University Press. Bialystok, E. (1986). “Factors in the growth of linguistic awareness”. Child Development, 57, 498–510. Birdsong, D. (1989). Metalinguistic Performance and Interlinguistic Competence. Berlin: Springer. Blackwell, A., Bates, E., & Fisher, D. (1996). “The time course of grammaticality judgment”. Language and Cognitive Processes, 11, 337–406. Bley-Vroman, R., Felix, S., & Ioup, G.L. (1988). “The accessibility of universal grammar in adult language learning”. Second Language Research, 4, 1–32. Bot, K. de (1990). “Language attrition: Competence loss or performance loss?”. In B. Spillner (Ed.), Sprache und Politik (pp. 63–65). Frankfurt: Peter Lang. Bot, K. de (1999). “The psycholinguistics of language loss”. In G. Extra & L. Verhoeven (Eds.), Bilingualism and Migration (pp. 345–361). Berlin: Mouton de Gruyter. Bot, K. de, Gommans, P., & Rossing, C. (1991). “L1 loss in an L2 environment: Dutch immigrants in France”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 87–98). Cambridge: CUP. Bradac, J. J., Martin, L. W., Elliott, N. D., & Tardy, C. H. (1980). “On the neglected side of linguistic science: Multivariate studies of sentence judgment”. Linguistics, 18, 967–995.
Evelyn P. Altenberg and Robert M. Vago
Cairns, H. S. (1999). “Common methodological issues in L1 and L2 research”. In E. C. Klein & G. Martohardjono (Eds.), The Development of Second Language Grammars: A generative approach (pp. 37–44). Amsterdam and Philadelphia: John Benjamins. Chaudron, C. (1983). “Research on metalinguistic judgments: A review of theory, methods, and results”. Language Learning, 33, 343–377. Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, N. (1986). Knowledge of Language: Its nature, origin and use. New York, NY: Praeger. Cowan, R., & Hatasa, Y. A. (1994). “Investigating the validity and reliability of native speaker and second-language learner judgments about sentences”. In E. Tarone, S. M. Gass, & A. D. Cohen (Eds.), Research Methodologies in Second-Language Acquisition (pp. 287– 302). Hillsdale, NJ: Lawrence Erlbaum. Cowart, W. (1996). Experimental Syntax: Applying objective methods to sentence judgments. Thousand Oaks, CA: Sage. Cowart, W. (2002). “Dialects in that-trace”. Paper presented at the Conference on Gradedness, University of Potsdam, Germany, October 21–23. Durkin, K. (1994). “The lure and language of implicit memory: A developmental perspective”. In N. C. Ellis (Ed.), Implicit and Explicit Learning of Languages (pp. 523– 548). London: Academic Press. Eckman, F. R. (1994). “The competence-performance issue in second-language acquisition theory: A debate”. In E. Tarone, S. M. Gass, & A. D. Cohen (Eds.), Research Methodologies in Second-Language Acquisition (pp. 3–15). Hillsdale, NJ: Lawrence Erlbaum. Ellis, N. C. (2002). “Frequency effects in language processing: A review with implications for theories of implicit and explicit language acquisition”. Studies in Second Language Acquisition, 24, 143–188. Ellis, R. (1990). “Grammaticality judgments and learner variability”. In H. Burmeister & P. L. Rounds (Eds.), Variability in Second Language Acquisition: Proceedings of the Tenth Meeting of the Second Language Research Forum (pp. 25–60). Eugene, OR: University of Oregon. Ellis, R. (1991). “Grammaticality judgments and second language acquisition”. Studies in Second Language Acquisition, 13, 161–186. Fernández, E. M. (2000). Bilingual Sentence Processing: Relative clause attachment in English and Spanish. Unpublished Doctoral Dissertation, New York: CUNY Graduate Center. Flynn, S. (1986). “Production vs. comprehension: Differences in underlying competence”. Studies in Second Language Acquisition, 8, 135–164. Gass, S. M. (1994). “The reliability of second-language grammaticality judgments”. In E. Tarone, S. M. Gass, & A. D. Cohen (Eds.), Research Methodologies in Second-Language Acquisition (pp. 304–321). Hillsdale, NJ: Lawrence Erlbaum. Goss, N., Ying-Hua, Z., & Lantolf, J. P. (1994). “Two heads may be better than one: Mental activity in second-language grammaticality judgments”. In E. Tarone, S. M. Gass, & A. D. Cohen (Eds.), Research Methodologies in Second-Language Acquisition (pp. 263– 285). Hillsdale, NJ: Lawrence Erlbaum. Greenberg, C. A. (1987). “Metalinguistic awareness and adult literacy”. Dissertation Abstracts International, 47 (12-A, Pt.1), 4376–4377.
Grammaticality judgments in first language attrition
Greidanus, T., & van der Linden, E. (1986). “Relations between FL grammaticality judgments and FL production”. I.T.L.: Review of Applied Linguistics, 73, 51–82. Grosjean, F., & Py, B. (1991). “La restructuration d’une première langue: L’intégration de variantes de contact dans la compétence de migrants bilingues”. La Linguistique, 27, 35–60. Han, Y. (2000). “Grammaticality judgment tests: How reliable and valid are they?” Applied Language Learning, 11, 177–204. Hiramatsu, K. (2001). “Accessing linguistic competence: Evidence from children’s and adults’ acceptability judgments”. Dissertation Abstracts International, Section A: The Humanities and Social Sciences, 61 (10), 3974. Jaspaert, K., Kroon, S., & van Hout, R. (1986). “Points of reference in first-language loss research”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 37–49). Dordrecht: Foris. Jordens, P. (1991). “Linguistic knowledge in second language acquisition”. In L. Eubank (Ed.), Point Counterpoint: Universal Grammar in the second language (pp. 199–218). Amsterdam and Philadelphia: John Benjamins. Kail, M., & Bassano, D. (1997). “Verb agreement processing in French: A study of on-line grammaticality judgments”. Language and Speech, 40, 25–46. Kellerman, E. (1985). “Dative alternation and the analysis of data: A reply to Mazurkewich”. Language Learning, 35, 91–101. Kellerman, E. (1986). “An eye for an eye: Crosslinguistic constraints on the development of the L2 lexicon”. In E. Kellerman & M. Sharwood Smith (Eds.), Crosslinguistic Influence in Second Language Acquisition (pp. 35–48). Oxford: Pergamon. Klein, E. C., & Martohardjono, G. (1999). “Investigating second language grammars: Some conceptual and methodological issues in generative SLA research”. In E. C. Klein & G. Martohardjono (Eds.), The Development of Second Language Grammars: A generative approach (pp. 3–34). Amsterdam and Philadelphia: John Benjamins. Köpke, B. (1999). “L’attrition de la première langue chez deux groupes de bilingues tardifs”. Revue Parole, 11–12, 199–219. Köpke, B., & Nespoulous, J-L. (2002). Task Dependency in First Language Attrition. Ms., Université de Toulouse-Le Mirail, France. Krashen, S. (1982). Principles and Practice in Second Language Acquisition. Oxford: Pergamon. Lakshmanan, U., & Teranishi, K. (1994). “Preferences versus grammaticality judgments: Some methodological issues concerning the governing category parameter in secondlanguage acquisition”. In E. Tarone, S. M. Gass, & A. D. Cohen (Eds.), Research Methodologies in Second-Language Acquisition (pp. 185–206). Hillsdale, NJ: Lawrence Erlbaum. Linebarger, M., Schwartz, M., & Saffran, E. (1983). “Sensitivity to grammatical structure in so-called agrammatic aphasics”. Cognition, 13, 361–392. Martohardjono, G. (1998). “Measuring competence in L2 acquisition: Commentary on Part II”. In S. Flynn, G. Martohardjono, & W. O’Neil (Eds.), The Generative Study of Second Language Acquisition (pp. 151–157). Mahwah, NJ: Lawrence Erlbaum.
Evelyn P. Altenberg and Robert M. Vago
McDaniel, D., & Cairns, H. S. (1996). “Eliciting judgments of grammaticality and reference”. In D. McDaniel, C. McKee, & H. S. Cairns (Eds.), Methods for Assessing Children’s Syntax (pp. 233–254). Cambridge, MA: MIT Press. Meng, M., & Bader, M. (2000). “Ungrammaticality detection and garden path strength: Evidence for serial parsing”. Language and Cognitive Processes, 15, 615–666. Munnich, E., Flynn, S., & Martohardjono, G. (1994). “Elicited imitation and grammaticality judgment tasks: What they measure and how they relate to each other”. In E. Tarone, S. M. Gass, & A. D. Cohen (Eds.), Research Methodologies in Second-Language Acquisition (pp. 227–243). Hillsdale, NJ: Lawrence Erlbaum. Murphy, V. A. (1997). “The effect of modality on a grammaticality judgment task”. Second Language Research, 13, 34–65. Nagata, H. (1990). “On-line judgments of grammaticality of sentences”. Perceptual and Motor Skills, 70 (3, pt.1), 987–994. Paolillo, J. C. (2000). “Asymmetries in universal grammar: The role of method and statistics”. Studies in Second Language Acquisition, 22, 209–228. Paradis, M. (1994). “Neurolinguistic aspects of implicit and explicit memory: Implications for bilingualism and SLA”. In N. C. Ellis (Ed.), Implicit and Explicit Learning of Languages (pp. 393–419). London: Academic Press. Paradis, M. (1995). “Introduction: The need for distinctions”. In M. Paradis (Ed.), Aspects of Bilingual Aphasia (pp. 1–9). Oxford: Pergamon. Pinker, S. (1994). The Language Instinct. New York, NY: Harper Collins. Polinsky, M. (1994). “Structural dimensions of first language loss”. Chicago Linguistic Society, 30, 257–276. Polinsky, M. (1997). “American Russian: Language loss meets language acquisition”. In W. Browne, E. Wornisch, N. Kondrashova, & D. Zec (Eds.), Annual Workshop of Formal Approaches to Slavic Linguistics: The Cornell Meeting 1995 (pp. 370–406). Ann Arbor, MI: Michigan Slavic. Pritchet, B. L. (1992). Grammatical Competence and Parsing Performance. Chicago, IL: University of Chicago Press. Rice, M., Wexler, K., & Redmond, S. M. (1999). “Grammaticality judgments of an extended optional infinitive grammar: Evidence from English-speaking children with specific language impairment”. Journal of Speech, Language, and Hearing Research, 42, 943–961. Roediger, H. L., Srinivas, K., & Weldon, M. S. (1989). “Dissociations between implicit measures of retention”. In S. Lewandowsky, J. C. Dunn, & K. Kirsner (Eds.), Implicit Memory: Theoretical issues (pp. 67–85). Hillsdale, NJ: Erlbaum. Schachter, J. (1974). “An error in error analysis”. Language Learning, 24, 205–214. Schachter, J. (1989). “A new look at an old classic”. Second Language Research, 5, 3–42. Schachter, J., Tyson, A. F., & Diffley, F. J. (1976). “Learners’ intuitions of grammaticality”. Language Learning, 26, 67–76. Schachter, J., & Yip, V. (1990). “Why does anyone object to subject extraction?” Studies in Second Language Acquisition, 12, 379–392. Schmid, M. S. (2002). First Language Attrition, Use and Maintenance: The case of German jews in Anglophone Countries. Amsterdam and Philadelphia: John Benjamins.
Grammaticality judgments in first language attrition
Schoenmakers-Klein Gunnewiek, M. (1989). “Structural aspects of the loss of Portuguese among migrants: A research outline”. I.T.L.: Review of Applied Linguistics, 83–84, 99– 124. Scholes, R. J., & Willis, B. J. (1987). “Language and literacy”. Journal of Literary Semantics, 16, 3–11. Schütze, C. T. (1996). The Empirical Base of Linguistics: Grammaticality judgments and linguistic methodology. Chicago, IL: University of Chicago Press. Seliger, H. W. (1991). “Language attrition, reduced redundancies, and creativity”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 227–240). Cambridge: CUP. Seliger, H. W. (1996). “Primary language attrition in the context of bilingualism”. In W. C. Ritchie & T. K. Bhatia (Eds.), Handbook of Second Language Acquisition (pp. 605–626). San Diego, CA: Academic Press. Sharwood Smith, M. A. (1983). “On first language loss in the second language acquirer”. In S. Gass & L. Selinker (Eds.), Language Transfer in Language Learning, (pp. 222–231). Rowley, MA: Newbury House. Sharwood Smith, M. A. (1994). “The unruly world of language”. In N. C. Ellis (Ed.), Implicit and Explicit Learning of Languages (pp. 33–43). London: Academic Press. Snyder, W. (2000). “An experimental investigation of syntactic satiation effects”. Linguistic Inquiry, 31, 575–582. Sorace, A. (1988). “Linguistic intuitions in interlanguage development: The problem of indeterminacy”. In J. Pankhurst, M. Sharwood Smith, & P. van Buren (Eds.), Learnability in Second Languages: A book of readings (pp. 167–207). Dordrecht: Foris. Sorace, A. (1996). “The use of acceptability judgments in second language acquisition research”. In W. C. Ritchie & T. K. Bhatia (Eds.), Handbook of Second Language Acquisition (pp. 375–409). San Diego, CA: Academic Press. Stanovich, K. E. (1990). “Concepts in developmental theories of reading skill: Cognitive resources, automaticity, and modularity”. Developmental Review, 10, 72–100. Tarone, E. E. (1994). “A summary: Research approaches in studying second-language acquisition or ‘If the shoe fits..’ ” In E. Tarone, S. M. Gass, & A. D. Cohen (Eds.), Research Methodologies in Second Language Acquisition (pp. 323–336). Hillsdale, NJ: Lawrence Erlbaum. Tomioka, N. (2002). “How to distinguish procedural and declarative memory use in L2: Processing with a behavioral test”. Paper presented at the International Linguistics Association 47th Annual Conference, York University, Toronto, Canada. Winter, B., & Reber, A. S. (1994). “Implicit learning and the acquisition of natural languages”. In N. C. Ellis (Ed.), Implicit and Explicit Learning of Languages (pp. 115– 145). London: Academic Press. Wulfeck, B. B. (1988). “Grammaticality judgments and sentence comprehension in agrammatic aphasia”. Journal of Speech and Hearing Research, 31, 72–81. Wulfeck, B., Bates, E., & Capasso, R. (1991). “A cross-linguistic study of grammaticality judgments in Broca’s aphasia”. Brain and Language, 41, 311–336.
P II
Attrition in progress – observations and descriptions
Issues in finding the appropriate methodology in language attrition research Kutlay Ya˘gmur Tilburg University
Introduction This paper on methodological issues in language attrition research will begin by briefly recapitulating the past research in the field. In the second part, the emphasis will be on methodological issues such as sampling, instrumentation, point of reference for testing, and a criticism of the ‘bold’ generalisations commonly made in research on language attrition on the basis of research findings. The question ‘If I could do the research all over again, what would I do differently’ posed in the call for papers for the Amsterdam conference presented me with the opportunity and motivation to re-evaluate my earlier research on language attrition. In line with the objectives of the conference, a critical evaluation of my own research on L1 attrition among Turkish speakers in Australia will be presented. Examined will be the motivations for the study, choosing the appropriate methodology in accordance with the relevant theories, issues involved in subject selection, point of reference, instrumentation, and so on. In the last section of this paper, I will present the changes that I would make to my earlier research design, given the chance.
.
Language attrition: An incomplete puzzle
There have been several major international conferences on the topic of language maintenance, shift and loss. Language attrition has attracted the attention of language researchers in the late 70s and research in this field truly gained momentum after the inaugural conference on ‘Attrition of Language
Kutlay Ya˘gmur
Skills,’ at the University of Pennsylvania in 1980. Before that conference, language loss was considered to be a clinical incident such as aphasia or other language disorders caused by tumours, strokes, or traumas to the head (Smith & Wilson 1979). After the UPenn conference, a number of conferences were organised in the Netherlands (see also Köpke and Schmid, this volume). The first and second Noordwijkerhout conferences (1988, 1992) concentrated mainly on issues of structural and linguistic loss as well as on the sociological and social-psychological explanatory factors of language loss. The third conference in Veldhoven (1998) focused on the sociolinguistic background of ‘why’ languages are lost and ‘how’ a language is lost. In addition to these sociolinguistic and psycholinguistic perspectives, one major aim of this gathering was the attempt to arrive at an integrated framework synthesising the sociolinguistic and psycholinguistic paradigms. The contributions to this last conference, as well as the final workshops and round table discussions on the development of an integrated theoretical framework of negative language development and revitalisation, were expected to serve as a renewed starting point for continued research into language shift and loss in a multidisciplinary context. Whether such a paradigm integration has been achieved or not is to be seen in future research on language attrition, but the plenary addresses by Kees de Bot (2001) and Joel Walters (2001) took a major step towards the intended integration between the sociolinguistic language shift paradigm and the psycholinguistic language loss paradigm. Nevertheless, as expressed by Walters (2001) the contradictory dimensions and entrenched conflicting methodologies of sociolinguistics and psycholinguistics makes the integration of ‘loss’ and ‘shift’ a difficult task. In spite of this difficulty, studies investigating both dimensions are necessary, since language use is never divorced from its social context. It is not possible to investigate language attrition or maintenance on the basis of a number of linguistic tests alone, without a careful description of the social context, the attitudes of speakers, ethnolinguistic vitality of the group and so on. An examination of the earlier conference proceedings and a number of dissertations written on language loss shows that there are more questions than answers concerning the theoretical framework and research methodology of language attrition. Depending on the approach to attrition within which researchers work, different theories and hypotheses concerning language loss have been postulated. Taking a contact linguistics, sociolinguistic, language change or acquisition perspective bring about varying hypotheses about the causes and the effects of language loss. Accordingly, the methodology used in such studies varies as well. Due to this inconsistency there are no commonly
Issues in finding the appropriate methodology
agreed definitions of language attrition, language loss or shift as yet (see also Köpke and Schmid, this volume). This variety of approaches, analyses and findings was also apparent in the papers presented at the Amsterdam conference as is evidenced by the present collection of papers. . Definitions In the 1980 conference, Lambert and Freed (1982) defined language attrition as follows: ... language attrition may refer to the loss of any language or any portion of a language by an individual or a speech community. It may refer to the declining use of mother tongue skills by those in bilingual situations or among ethnic minorities in (some) language contact situations where one language, for political or social reasons, comes to replace another.(Lambert & Freed 1982: 1)
Although this definition recognised the interface between linguistic and sociolinguistic aspects of language attrition, no distinction was made between ‘societal loss’ (which is more commonly identified as language shift rather than language loss), and individual loss (which is identified as language attrition over time in language contact situations). In the same vein, Andersen (1982: 85) suggests that language attrition is the loss of a language by successive generations in immigrant or minority communities, and that individual language attrition refers to forgetting a language due to lack of use. On the other hand, Richards et al. (1985) define language shift as a change from the use of one language to the use of another. Drawing a distinction between language attrition and language shift thus seems to be highly problematic. It can generally be argued, however, that language attrition is an intragenerational phenomenon while language shift is an intergenerational one. The term intragenerational implies that the loss of linguistic skills takes place within individuals over a more or less extended period of time mainly because of a break in the linguistic tradition. With respect to intergenerational and intragenerational language loss, an examination of Gonzo and Saltarelli’s (1983) well-established model is helpful for the differentiation between ‘language attrition’ and ‘language shift’. In relation to the linguistic situation in immigrant families, Gonzo and Saltarelli suggested that, because of non-use of the L1 and interference from the L2, first generation immigrants experience language attrition over the years. The children of these immigrants acquire the already attenuated L1; even this reduced system is, however, not entirely passed on to the second generation for a number of linguistic and sociolinguistic reasons.
Kutlay Ya˘gmur
Following this incomplete L1 acquisition, there is again L1 attrition among second-generation immigrants, and even further ‘reduced’ forms are transmitted to the third generation. This cascade model thus predicts that within three or four generations, immigrant languages in contact with a majority language are doomed to be lost (Gonzo & Saltarelli 1983). In this paper, L1 attrition is defined as the gradual loss of competence in a given language. It is generally agreed that ‘changes in language use’ are identified as language shift, and ‘changes in language proficiency’ as language loss (Fase, Jaspaert, & Kroon 1992: 4) and that, furthermore, language shift is located between generations (intergenerational), while attrition occurs within individuals (intragenerational). In the 1998 Veldhoven language loss conference, Kees de Bot (2001) suggested ‘language loss’ to be a generic term denoting a decline in language skills in both individuals and groups. This approach assumes attrition and shift to be individual and societal aspects of loss, respectively: attrition suggesting the loss of skills in individuals and shift implicating a decline in language use on the community level. It can be argued that in this approach, ‘shift’ becomes a sub-branch of language loss studies; however, in de Bot’s treatment of the matter, ‘shift’ can be intra- or intergenerational, whereas talking about intergenerational attrition is highly problematic due to points of reference in measurement. This issue will be discussed in more detail below in the section on sampling. . Extralinguistic factors It is important to point out that language attrition is not an isolated phenomenon, rather it is a special case of variation in the acquisition and use of language/s and can best be studied, described, documented and explained within a large framework which includes all other phenomena of L1/L2 acquisition, bilingualism, language use/choice, code-switching/mixing, and language attitudes (Andersen 1982: 86; Ya˘gmur 1997). In order to systematically investigate ethnic minority language contexts, various language use typologies have been proposed, among them the well-known and frequently cited works of Ferguson (1975), Giles, Bourhis, and Taylor (1977), Haarmann (1986), Haugen (1972), and Edwards (1992). Each of these frameworks considers some linguistic and social factors to be essential for an accurate description of language contact situations. However, as they have been discussed in detail elsewhere (Ya˘gmur 1997; see also Schmid 2002; Köpke & Schmid, this volume), sociolinguistic factors involved in language attrition will not be further discussed here.
Issues in finding the appropriate methodology
. Methodology in language attrition studies The organisers of this conference have pointed out that many studies on language attrition report to have found substantial loss, but do not present convincing evidence. This argument seems compelling, given the fact that the collections of various conference proceedings, collected volumes and special issues of journals published so far do not present a uniform approach to language attrition research. Divergent methods of data collection, sampling and instrumentation are still major problem areas in this field (Köpke and Schmid, this volume), which will have to be identified in order to prevent exaggerated descriptions of attrition in future studies. This section will present a critical evaluation of methodological issues on the basis of the collection of papers from the Veldhoven conference (Ammerlaan, Hulsen, Strating and Ya˘gmur 2001) and further past research. In order to achieve this, firstly, a short overview of methodological issues in language attrition will be presented. Secondly, some problem areas in such studies will be pointed out on the basis of past research, and finally, a number of proposals with respect to research designs will be made. . What type of design is appropriate? The decision for an appropriate methodology for any given research project is usually taken on the basis of relevant research questions, to which the issue of choosing the appropriate data collection methods, instruments and informants are crucial. A close examination of past language attrition research reveals serious flaws in this respect. The biggest consequence of such methodological problems is large variation in the amount of language attrition reported in different studies. Most of the earlier language attrition conferences contained sections specifically devoted to the research methodology of language attrition. In the Pennsylvania conference, the papers by Andersen (1982), Oxford (1982), and Clark (1982) have set the ground for future research. Andersen called his own paper ‘a tentative blueprint’ for choosing linguistic features to be incorporated in any research design. The strength of Andersen’s hypotheses is that, since they are not language-specific, they can be adapted and modified for a large variety of languages and contexts. In addition, they are mostly testable (see below). Unfortunately, there are as yet not many studies testing Andersen’s hypotheses, which is why we are not in a position to compare findings across studies to identify common patterns observed in language attrition.
Kutlay Ya˘gmur
Oxford’s paper (1982) on designing and conducting research in language skill attrition illustrates the considerable confusion over the types of tests that can be used in language attrition studies. Oxford discusses proficiency and achievement testing, pointing out that the types of testing instruments described can be adapted to language attrition studies. However, the general language proficiency measures she discusses have been developed for student populations and are not suitable for L1 attrition studies, since level of education and familiarity with such tests play an important role in the response level. Past experience with proficiency tests, the feeling of being tested and so on are all factors that can have an effect on the results. Such tests can thus be predicted to be invalid or unreliable for informants who are less educated or less familiar with formal testing procedures. Obtaining a fail or a pass mark from a proficiency test would thus not mean much in the context of language attrition. Furthermore, even students themselves often fail such tests for the reasons given above, which obviously does not license the conclusion that they are suffering from ‘language loss’. As opposed to Oxford, Clark (1982) suggests designing specific diagnostic testing instruments for the measurement of language attrition on the basis of Andersen’s hypotheses. Clark’s suggestion concerning diagnostic feedback testing instruments, specifically designed for language attrition research, is highly relevant, since such tests are constructed on different principles and goals than proficiency testing instruments. Nevertheless, the methods proposed by Clark do not distinguish the investigation L1 versus L2 loss, let alone intra- versus intergenerational loss. Subsequent volumes on language attrition (Weltens, de Bot and van Els 1986) still testify to the preoccupation with methodological issues. Ginsberg’s (1986) sophisticated paper on statistical data analysis procedures provided researchers with insights into the differences between logistic models analysing one parameter and those which allow to investigate two or three parameters, illustrating what tests can reveal possible correlations between predictive and criterion variables. We need to point out, however, that all those so-called diagnostic tests were again designed for student populations (mostly college students), and are thus unsuitable for less educated immigrants. A further drawback is that all of the structural and logistic models outlined in Ginsberg’s paper require a minimum of several hundred informants in order to allow reliable estimates, which again makes them impractical for small-scale studies. That notwithstanding, the quantitative data analysis procedures used by Ginsberg are highly valuable guidelines for the standardisation of language attrition tests.
Issues in finding the appropriate methodology
Before Jaspaert, Kroon and van Hout (1986), no study had explicitly dealt with methodological problems involved in research on L1 loss. Their paper suggests that the most appropriate way of establishing language attrition would be a carefully implemented longitudinal design, in which informants are tested before the onset of attrition (e.g. shortly after arriving in the new country), and again after some years by using a pre- and post-test to identify changes in language proficiency. However, such research is not always possible due to problems of logistics, finance, and feasibility. In order to allow for the time factor in a research design, Jaspaert, Kroon and van Hout (1986) proposed to solve this problem by static group comparison. In a quasi-experimental design, the immigrant group can be compared to a control group living in the country of origin. This is a valid approach in testing first generation informants, but it cannot be applied with second or third generation informants (due to their reduced proficiency according to Gonzo and Saltarelli’s cascade model, see above). Another problem is the issue of language change. In the case of Turkish, for example, the majority of potential L1 attriters emigrated to other countries around 1960, and they thus maintain the variety of Turkish spoken in the 60’s in Turkey. Turkish immigrants in Western Europe have frequent contacts with the homeland, which brings their language use closer to the present-day standard, but speakers who emigrated to more distant countries, such as Australia, have limited exposure to Turkish spoken in Turkey, where the purification movement of language reform has led to considerable change in the Turkish lexicon. Due to frequent contacts, immigrants in Western Europe are thus more up to date with such changes than immigrants in Australia. On a more general level, language change is a natural process for all living languages, and during that process the change experienced in a monolingual speech community will be different from that which takes place in a language contact situation, where two or more languages are in interaction. Therefore, when we measure the differences between the two contexts, what we claim as lost might simply be an artefact of varying degree of change in two different contexts. . Causes of claiming ‘massive’ loss As seen in the previous section, even the best practice is not without problems. In the process of editing the most recent language loss conference proceedings (Ammerlaan, Hulsen, Strating & Ya˘gmur 2001), it was observed that in some papers information on the type of language test used, the types of linguistic features investigated, the nature of the sample, the control group characteris-
Kutlay Ya˘gmur
tics, and so on were missing (Ammerlaan et al. 2001). In addition, there were bold generalisations declaring large losses on the basis of data obtained from single case studies or very young speakers of a language. In most studies on language attrition the relationship between data, research questions, and informant characteristics seem to be rather obscure. It is evident that claims of massive L1 attrition simply on the basis of self-reported data or survey information data is bound to lead to an account that is not representative of the actual situation. Accordingly, simply conducting a number of isolated linguistic tests and claiming loss on the basis of the results without any relevant background information on the speaker characteristics is also problematic. Researchers need to consider that if they wish to make claims about a specific community, the selection of representatives of this community should be randomly chosen; otherwise, the scope of the research is restricted. Such a ‘random choice’ can often be compromised by the methods used to find informants, as in the case of Waas (1993), who had to place advertisements in German-language journals in Sydney. Ethnic group affiliation was one of her predictor variable to account for L1 maintenance or loss. In terms of subject selection procedures, one would assume that those informants who read German papers or journals are already maintenance-oriented speakers, which is why the representativeness of the sample for the given hypothesis might not be well founded. Furthermore, language attrition studies are often characterised by rather broad conclusions drawn about the extent of attrition observed. It should be obvious that generalisations are not permissible across linguistic levels. If the study has been confined to the investigation of a number of lexical features, for example, valid conclusions can only be drawn concerning that domain. Also, if groups are compared in different contexts, it has to be taken into account that aspects other than linguistic ones may account for different patterns in language usage. The two most problematic areas in this respect, as became evident during both the Veldhoven and the Amsterdam conferences, are sampling and instrumentation. Along with empirical designs, there are single-subject case studies such as the one suggesting syntactic loss on the basis of a very young child’s use of pronouns in a given language (Ammerlaan et al. 2001). The problems involved with such a presentation are obvious: it is not possible to come to bold generalisations on the basis of findings from a single informant; and furthermore, it is generally agreed in language attrition studies that for something to be lost, it should have been there in the first place. Given the age factor of young informants, some structures might not have been acquired yet, thus what is shown as ‘lost’ might not have been acquired in the first place,
Issues in finding the appropriate methodology
and studies investigating structural loss or lexical attrition among very young children are not persuasive (Kaufman 2001; De Kadt 2001). These problems are compounded if research on L1 attrition is combined and compared with findings from second or third generation informants, which is an extremely difficult area of study, since it is impossible to determine whether all of the forms were available to the speakers in the acquisition process (See Ya˘gmur 1997 for details). The age factor can be a problematic issue not only in studies involving very young informants, but also in those investigating the elderly, since even native speakers who live in a monolingual environment may experience L1 loss due to old age. In that respect, investigating L1 loss (in an L2 environment) among elderly informants may not provide representative findings either. In order to investigate structural change in the verb inflection system of some Basque dialects, for instance, Elordui (2001) asked 10 Basque-English bilinguals, the youngest of whom was 50 years old, to translate 500 sentences from English into Basque. The validity of this research design is highly questionable, not only because of the age of the informants, but also because of the nature of the task, the latter of which will be more elaborately discussed below. Indeed, data collection instruments seem to be another serious area of concern in language attrition studies. Most of the papers presented at both the Veldhoven and the Amsterdam conference report the use of questionnaires, interviews, linguistic tests, translation tasks and language proficiency tests for their data collection. The concern about the suitability of language proficiency tests in language attrition studies was expressed in the preceding sections (2.1, see also Köpke and Schmid, this volume). In assessing the suitability of an instrument, the research questions and the characteristics of the informants (in particular, the age of the informant) as well as the design turn out to be highly important. None of these dimensions should be evaluated in isolation. However, special attention will be paid here to the suitability of translation tasks in assessing language attrition, since they appear to have become rather frequent (Dal Negro 2001; De Kadt 2001; Elordui 2001; Macevichius 2001; Smits 2001). In all of these studies, considerable language attrition has been reported. Table 1 summarises the tasks, informants, and the age of the informants in these studies. In all of the above studies, varying degrees of attrition are reported. As is evident from the overview presented in Table 1, the generalisability of the findings may be complicated by factors such as the types of informants and their age. However, the most serious methodological problem is the data collection instrument, namely a translation task. Translation is a special skill that requires
Kutlay Ya˘gmur
Table 1. Overview of data collection instruments in some recent language attrition studies Study
Task
No of informants
Age span
Dal Negro 2001
Translation of 21 sentences, Interviews
23 informants,
Ages 6–85
De Kadt 2001
End of school examination, Translation from English to German
25 pupils
Ages 12–13 and 17–18
Elordui 2001
Translation, interviews
16 Basque-Spanish bilinguals and 10 Basque-English bilinguals
Youngest informant is 50 years old (the rest is not specified)
Kaufman 2001
Oral narration task
30 pupils
Ages 6–13
Macevichius 2001
Word translation task, written document analysis
6 second-generation adult informants
Born in the 1950’s
Smits 2001
Translation tasks, free conversations
10 second to fourth generation informants
Average age is 65
training and expertise. Even educated speakers who are proficient bilinguals may not be able to translate or interpret effectively. Language loss generalisations based on translation tasks and proficiency testing can thus be highly questionable. These problems are compounded by the fact that in a translation task, the competence in both the target language and the source language is very important. For example, if I had used translation tasks for my earlier research on language attrition among Turkish speakers in Sydney (Ya˘gmur 1997), massive attrition of Turkish would have been reported on all points of testing, due to a lower degree of English proficiency among the less educated informants. As a matter of fact, the ineffectiveness of the translation task is expressed by Smits (2001) herself, who acknowledges that “from a comparison of the percentage of Standard Dutch forms it becomes clear that inflection for noun, adjective and verb is preserved better in conversations than in translations” (2001: 315). In other words, the translation task did not yield representative results in that particular study. The inappropriateness of the task was, in this case, revealed by a crosscheck against data elicited by other means, namely an interview technique, preventing bold generalisations.
Issues in finding the appropriate methodology
In short, these examples illustrate that reports of ‘massive loss’ are often caused by insufficient consideration of methodological issues. If a research project is not carefully designed, the findings may obviously be inaccurate or distorted. Many studies thus report massive loss, not because they have actually detected substantial attrition, but due to: – – – – – –
Flaws in the research design A lack of relationship between the conceptual framework and the data collected Inappropriate or irrelevant data collection instruments A non- representative choice of informants Use of either very young or very old informants Problems in the data analysis (inappropriate statistical analyses)
All of these factors will produce exaggerated or distorted results. In order to overcome most of the methodological difficulties presented in this section, researchers need to have a clear research focus by means of research questions and hypotheses. The working hypotheses offered by Andersen (1982) are still highly valuable for researching linguistic aspects of language attrition. Naturally, the extralinguistic predictor variables and the relationship between the linguistic (individual) and social factors (group) need to be examined. Collecting data on both aspects in a single research design is always possible. The main issue is the accurate operationalisation of the constructs that we want to measure. In the next section, the research design on language attrition of Turkish immigrants in Sydney (Ya˘gmur 1997) will be presented.
. A research design of language attrition . Preliminary observation The two most basic guiding questions in developing any research design on language attrition are: – –
What type of evidence is needed as proof that attrition has occurred? On what grounds have we as researchers been persuaded that this topic is worthy of investigation?
In order to provide some answers to these questions and to present a methodological framework in which to study them, I will reflect on my earlier research of language attrition among Turkish speakers in Sydney.
Kutlay Ya˘gmur
I first became interested in this phenomenon when, in the late 1980s, I observed a distinctive variety of Turkish spoken in Australia, different in many respects from Turkish as it was spoken elsewhere. This variety was most often used among first generation Turkish immigrants, who usually used their L1 among each other, but frequently inserted English lexical items in their Turkish discourse, as is illustrated by the following examples: Group I examples: (1) Türk ¸sopuna gittim. Turkish shop-dir go-past-1sg ‘I went to the Turkish shop.’ ˙ suransı yokmu¸s klaym yapamadı. (2) In¸ Insurance-gen not-past-infer claim make –mod-neg-past ‘As he didn’t have insurance he couldn’t make a claim.’ çıktı. (3) Kompaya Compensation-dir go out-past ‘He went out to compensation.’ (he applied for compensation) delinmi¸s. (4) Maflırı Muffler-poss-acc pierce-pass-yinfer ‘His (car’s) muffler is (reported to be) pierced.’
Structures such as these are very frequent; words such as “shop – s¸ op” and “compensation – kompa” are commonly used especially across generations, although the amount of code-mixing among second generation and bettereducated Turkish immigrants is much higher than among first generation and less-educated immigrants. On the basis of the code copying exemplified above, could it be claimed that the words, which have the same meaning as “shop, insurance, compensation” and so on, have been lost in Turkish? Can it be seen as an initial stage of language attrition? Or is there simply a new form of Turkish emerging (in which case it would imply language change)? These questions have not yet been answered conclusively. There are equivalent forms of all the above words in Turkish; thus, it is possible that this copying occurs because the speakers cannot access the Turkish forms in their mental lexicon. It may, however, also be the case that these English copies are simply more active in the speakers’ mental lexicon and are just inserted into a Turkish frame. Like many other studies of language attrition, mine was initially conceived on the assumption that this kind of copying could be taken as an indication of language attrition in progress. The most relevant question in this respect was whether these speakers would use the same kind of strategies with monolingual
Issues in finding the appropriate methodology
Turkish speakers. Without such data it would have been premature to suggest language attrition on the basis of code-mixing data. That notwithstanding, a further interesting observation about the lexical copies illustrated above is that in the speech of working class people, the copies are phonologically adjusted to fit the phonological and morphological structure of Turkish, whereas bettereducated bilinguals tend to copy without any phonological transformation. The type of lexical copies presented in (I) are very common, and it is doubtful whether they indeed suggest language attrition; the examples in (II), however, observed among second-generation immigrants, may be considered as a clear indication of attrition. Group II examples: (5) Küpemi giydim. Earring-pos-acc put on-past-1sg ‘I put on my earring.’
(giymek vs. takmak) (wear vs. put on)
ö˘greniyor. (gitar çalmak vs. futbol oynamak) (6) Gitar oynamayı Guitar play-nom-acc learn-prog-3sg ‘He learns how to play the guitar.’ (7) Çok kitaplar aldı. Many book-pl buy-past-3sg ‘He bought many books.’
(Çok kitap aldı )
sen bana anlatın(?) (8) Ö˘gretmen bunu Teacher this-acc you me-dir explain-int-2sg ‘Teacher you explain this to me (?)’
As opposed to the examples in group I above, which simply involve a lexical copy, these utterances are illustrative of processes, which might be interpreted as language attrition. The speakers here apparently apply not only lexical but also semantic copying, with structural consequences for the utterances, rendering them ungrammatical. (5) and (6) involve the violation of semantic restrictions – the English verb ‘to put on’ can be used both with reference to clothes (‘I put on my shirt’, meaning ‘I wore my shirt’) and to jewellery (‘I put on my earring’). In Turkish, however, two different verbs are used: giymek (to wear clothes) and takmak (to put on jewellery). Apparently, the specificity in meaning has been lost for this distinction. Similarly, as seen in example (6), there are two different verbs to express the actions indicated by the verb ‘to play’ in Turkish: ‘to play the guitar’ (gitar çalmak) and ‘to play football’ (futbol oynamak). Ex. (7) illustrates a more grammatical violation: Turkish does not allow double plurals, but especially younger speakers use such structures with
Kutlay Ya˘gmur
English as a model. In English, the noun following ‘many’ requires pluralisation; in Turkish, however, the modifier ‘many’ already announces the plural nature of the noun that follows, which is why no plural suffix is added. Violations of the L1 Turkish norm can also extend to pragmatic features, as seen in (8), where politeness norms of Turkish are not obeyed and prosodic characteristics of English are copied to turn a statement into a question. In Turkish culture, there is a certain hierarchy between a teacher and a student, which is also explicit in address forms. Thus, a student cannot address his teacher with her profession name alone. The first person possessive suffix -im needs to be added to the word ö˘gretmen (resulting form is ö˘gretmenim = my teacher, indicating authority of the addressee) so that it complies with the politeness norms of Turkish. However, these numerous instances of lexical and even semantic codecopying observed are not conclusive evidence for L1 attrition. More evidence is needed in line with a conceptual framework and a research design suitable for investigating language attrition. Moreover, it is equally important to document not only linguistic evidence, but also social, psychological and other extralinguistic factors. . Setting the design As discussed in detail in much of the literature available on language attrition, the major problem area of language attrition research is the methodology involved. This includes designing appropriate tests, finding suitable informants and deciding on the type of design to be adopted. In order to make valid and reliable generalisations, quantitative data and well-defined categories are needed. The research design for the study discussed here was chosen on the basis of claims concerning the regularity of Turkish language structure, ethnolinguistic vitality theory studies, previous findings on the topic, and a number of initial observations. These include the fact that compared to Turkish speakers living in an L1 environment, Turkish immigrants in Australia often speak at a much slower rate when using Turkish. I had also observed that many Turkish immigrants in Australia seem to prefer simple, analytic structures in their spoken Turkish discourse over more complex, synthetic forms. Many immigrants reported that they felt they had ‘lost’ a great deal of their mother tongue. Some of these people were not fluent in their L2, English, either. The aim of the study was to investigate ‘what is lost’ and ‘why it is lost’. In order to identify what is lost, a number of linguistic tests were used. In line with my initial observations
Issues in finding the appropriate methodology
and the relevant findings in the literature, I devised and also adapted already available tests for this purpose. The first aspect, ‘what is lost?’ is looked at within a linguistic framework, whereas the second aspect focuses on the sociolinguistic causes of language attrition within the social psychological framework of Ethnolinguistic Vitality Theory (EVT). As pointed out earlier, language attrition is a gradual process, occurring slowly over a long period of time. Longitudinal studies or pre- and post-test designs therefore do not seem appropriate in this kind of research, as many factors would be unaccounted for. The design that seemed most appropriate for my own research was therefore ‘static group comparison’ (Jaspaert et al. 1986: 38). In this research design, the amount of attrition is established by comparing the attriters’ language test results with those of a monolingual reference group. In Ya˘gmur (1997), informants in Australia were matched with informants in Turkey in terms of age, social background, sex, and education in order to provide the optimum point of reference. The study investigates the loss of Turkish (L1) in an English (L2) environment. The studies previously conducted in an Australian context all investigate L1 attrition among typologically related languages such as Dutch (Ammerlaan 1996; de Bot 1990; de Bot & Clyne 1994) German (Clyne 1967; Harres 1989; Waas 1993), and Italian (Bettoni 1989). There are thus similarities between both the host language (English) and the attriting languages in terms of linguistic structure, socio-historical factors, cultural values, religion, and social customs. In terms of language structure, these languages are all typologically related; Turkish, however, belonging to the Altaic group and being typologically unrelated to English, is a typical SOV language, exhibiting features such as agglutination, vowel harmony and the lack of grammatical gender. The linguistic component of the present investigation spans several levels. In order to trace syntactic attrition, the first question is whether there are any differences between first generation Turkish immigrants and monolingual Turkish speakers living in Turkey in the production of the late acquired complex Turkish forms of relative clauses. The second question is to what extent the immigrant Turkish speakers use analytic versus synthetic structures and what the implications for language maintenance or loss would be. In order to trace lexical attrition, the extent of any gaps in the lexicon of Turkish immigrants is investigated by means of verbal fluency tasks and a narration task. Finally, it is hypothesised that even if there is no extensive use of L2, Turkish immigrants might exhibit lexical and syntactic attrition because of a very limited number of domains for L1 use. These issues of syntactic and lexical loss are investigated
Kutlay Ya˘gmur
through a comparison of the test results of the Turkish-Australian groups to those of the reference groups in Turkey. .. Focus of the investigation Clearly, one of the most important aspects in the selection of an appropriate research design is the formulation of relevant research questions, on the basis of which the appropriate data collection instruments are designed and representative informants are selected. The study reported here was guided by four such research questions, namely: 1. Is there any lexical and syntactic loss evidenced among Turkish migrants in Sydney as compared to Turkish speakers living in Turkey? 2. Do narrative skills in the L1 of Turkish migrants erode over a long period of time in a language contact situation? 3. What are the factors or reasons behind L1 attrition? 4. Is there a relationship between ethnolinguistic vitality perceptions and L1 attrition of Turkish migrants? In order to answer the first question, data were collected by means of tests measuring lexical and syntactic proficiency, namely a lexical naming task and a relativisation production task (RPT). The latter is a special diagnostic test that was developed to assess whether the production of Turkish relative clauses showed any evidence of attrition. These tests were applied both to the immigrant groups in Sydney and to the monolingual Turkish control groups in Turkey. In this way, comparative evidence was obtained on the lexical and syntactic aspects of L1 use of Turkish immigrants in Sydney. The migrant groups were also asked to assess their L1 proficiency by means of a self-rating scale in order to show possible correlations between the results of the linguistic tests and the subjective self-ratings. The correlations between these linguistic tests and the extralinguistic factors obtained using other instruments were examined. These formal test-based data on the maintenance of linguistic skills were compounded by spontaneous speech data collected by means of a retelling task, using the picture-based story book Frog where are you? (Mayer 1969), commonly known as ‘the frog story’. This story has been used in a number of cross-linguistic language development and attrition studies (Berman & Slobin 1994; Cohen 1986, 1989; Olshtain & Barzilay 1991) and has proven to be highly appropriate for linguistic analyses. For the present study, lexical and textual analyses were carried out to see whether any evidence for lexical and/or pragmatic attrition could be found, using Aksu-Koç’s (1994) Turkish data on re-tellings of the ‘frog story’ data as a reference. The two immigrant
Issues in finding the appropriate methodology
groups’ scores on spontaneous speech were compared, particularly focusing on narrative skills and lexical attrition. In order to shed some light on the extralinguistic factors that potentially influence and determine language attrition, data was collected on the language use and – choice of Turkish immigrants in Sydney. For this investigation, a survey instrument, the ‘Language Use-Choice Questionnaire’, was developed in order to gather information concerning background characteristics, language use/choice, and language attitudes. This data was then used to detect possible causes of L1 attrition. Findings obtained on the basis of this instrument were correlated to other findings in order to establish possible relationships between attitudinal and linguistic aspects of L1 attrition. Finally, data was collected on ethnolinguistic vitality perceptions of Turkish speakers living in Sydney by means of a ‘Subjective Ethnolinguistic Vitality Questionnaire’ (Bourhis et al. 1981). In order to identify the relationship between ethnolinguistic vitality perceptions and erosion of L1 skills, the findings derived from the two sets of data – linguistic and extralinguistic – were correlated. An attempt was thus made to relate societal factors impinging upon individuals to the degree of L1 attrition. Ethnolinguistic vitality data from two Turkish migrant groups in Sydney was analysed to see whether variables such as education and social background had any effect on subjective vitality perceptions of the informants. .. Specific hypotheses on the linguistic aspects of language attrition In addition to the general research questions outlined above, a number of specific hypotheses were tested, namely Andersen’s (1982: 83–118) general linguistic hypotheses with regard to syntax and lexicon and Slobin’s (1977) hypotheses with regard to specific Turkish structures. Slobin’s (1977) argument concerning the structure of Turkish lays the foundation for the structural aspect of the study described here. Slobin argues that the Turkish system of agglutinative inflectional morphology is remarkably transparent, or in other words entirely regular. There are almost no exceptions to general rules, and there are no arbitrary sub-classes on the basis of features such as grammatical gender or phonological shape of the stem. Each inflectional morpheme is syllabic and acoustically salient. The nominal inflectional paradigm is so analytic that the entire system is mastered well before the age of two. Slobin also suggests that Indo-European languages tend to collapse and simplify their inflectional systems over time or in bilingual contact situations, whereas the Turkish system
Kutlay Ya˘gmur
of agglutinative inflectional morphology has remained stable across a great range of time and contact situations. (Slobin 1977: 191)
Drawing on evidence from L1 acquisition data, Slobin also argues that complex structures like relative clauses in Turkish are hard to acquire, and difficult to replicate in comparable structures in the English language. He provides evidence from child language acquisition studies, showing that children from Indo-European language groups are able to form relative clauses at around age two, while Turkish children do not produce relative clauses until around age five. The explanation given for this is that in Indo-European languages, when forming a relative clause the underlying semantics can easily be mapped onto the surface structure, so that minimal adjustment is needed. The child simply inserts a relative pronoun between two underlying sentences. In the case of Turkish, however, this procedure is much more complex: the predicate of the embedded sentence must be preposed, turned into a participle and possessed by its agent, making Turkish relative clause formation highly opaque. Slobin (1977) further proposes that late acquired and complex constructions such as relative clauses and verb complement constructions are most vulnerable to change in bilingual contact situations. Andersen’s Morphological Reduction hypothesis (1982: 97) complements Slobin’s argument in claiming that “those grammatical morphemes acquired earliest will be retained the longest by an LA1 and those acquired latest will be lost earliest by the LA”. Thus, it can be hypothesised that the late acquired Turkish relative clauses may be vulnerable to attrition. In the same vein, Slobin (1977: 190–191) proposes that the underlying characteristic of Turkish is its semantic transparency. The complex structures are acquired last and there seems to be a general tendency among Turkish children towards analytic structures rather than complex synthetic ones. Similarly, Andersen’s Syntactic Reduction hypothesis states that “the LA will preserve and overuse syntactic constructions that more transparently reflect the underlying semantic and syntactic relations” (Andersen 1982: 99). Following these hypotheses, it was argued that an investigation of analytic versus synthetic structures in Turkish might yield noteworthy findings with regard to the process of language attrition. It was thus decided to investigate structural loss with regard to these linguistic aspects. Andersen (1982: 92) states that the lexicon depends greatly on the linguistic experience of the speakers of a language and [that] the lexical store of an individual language user depends on
Issues in finding the appropriate methodology
the frequency, utility, and adequacy of the contents of that store for linguistic communication.
Accordingly, it can be hypothesised that even among first generation monolingual Turkish speakers living in Sydney there will be lexical attrition due to the functional restrictions of L1 use. In the same vein, Andersen’s Lexical Reduction hypothesis is highly relevant for the purpose of investigating the extent of attrition in the lexical repertoire of Turkish immigrants: “What lexicon the LA has retained will be of common, highly-frequent, unmarked lexical items; the gaps will be of less-common, low-frequency, highly marked items” (Andersen 1982: 94). In complementing the controlled lexical naming tasks, I thus investigated lexical attrition following Andersen’s Lexical Reduction hypothesis, using a re-telling task involving the ‘frog story’ (Ya˘gmur 1997). According to Andersen, in addition to structural and lexical language attrition, there are also paralinguistic consequences of linguistic erosion. He claims that “an LA will speak at a slower rate with more frequent hesitations, pauses, repairs, false starts, etc. than an LC of the same language” (Hypothesis 12a; 1982: 112). This hypothesis was also tested on the basis of narrations of the ‘frog story’. .. Testing the hypotheses Based on the hypothesis that Turkish immigrants would exhibit lexical and syntactic attrition in their L1, irrespective of the extended use of the L2, two groups of Turkish immigrants in Australia were included in the present study. Group A consisted of informants who had lived in Australia for fifteen or more years, but have a limited proficiency in English; the informants in Group B had also spent fifteen or more years in Australia and were well-educated and proficient in English. In order to match these two groups of immigrants, reference groups with similar background characteristics from Turkey were needed. Prior to migration, the informants of Group A had been living in Central Anatolian towns, whereas Group B had mainly lived in urban areas of Turkey. Applying the ‘static group comparison’ framework to this study, 40 Turkish immigrants in Australia were matched with 40 Turkish informants in Turkey. .. Informants In relation to the design of the study and the research questions posed, particular care was taken with regard to informant selection, based on those variables that in previous literature on language maintenance and loss have been reported to be most relevant. These include place of birth (to check whether
Kutlay Ya˘gmur
they were first generation migrants or not), age (at the time of investigation as well as at the time of migration), period of residence in the L2 environment, gender, education, marriage patterns, prior knowledge of the L2, the reason for migration, and the language variety used (standard L1 or a dialect of L1). Questions concerning these variables on an individual level were included in the questionnaires used in this study. The target groups for this study comprised 40 Turkish speakers living in the Sydney metropolitan area. As the study focuses on intragenerational aspects of L1 attrition, all the subjects were adult speakers of Turkish. In order to control for some of the above variables, all subjects were chosen among first generation Turkish migrants. They were all born in Turkey and all of them emigrated to Australia as adults, where they had been living for at least 15 years at the time of the investigation. Informants who were married all had a spouse of the same ethnolinguistic background. Education has been reported to be a highly significant variable in studies on language maintenance and loss, but has not very often been controlled for. In this study, education was used as a control variable for all aspects measured. On the basis of the educational level of the informants, the subjects were divided into two groups of 20 subjects each, Group A and Group B; Group A comprising less educated speakers while Group B included the better-educated informants. In order to control for the factor ‘gender’, an almost equal number of male and female subjects were included. There were 9 female and 11 male informants in Group A with a mean age of 50.5; in Group B there were 11 female and 9 male informants with a mean age of 42.3. The Turkish community in Australia consists of centralised groups concentrated in certain suburbs of the urban area of Sydney. The subjects of Group A were chosen from two different ‘working-class’ suburbs: Auburn, which is one of the areas with a high concentration of Turkish immigrants, and Mascot. The subjects of Group B came from different suburbs that are known to be ‘middleclass’. The Group A informants were mostly labourers and housewives, whereas the Group B informants had various occupations such as teacher, engineer, clerk, accountant, etc. Thus, the criteria for inclusion in the empirical study were that the subjects should have been in the L2 environment for at least fifteen years; they should be first generation immigrants; and they should all be native speakers of Turkish. The time span was decided upon because it was assumed that after 15 years of residence in the L2 environment, the subjects would feel ‘settled’. It was also assumed that in terms of language behaviour, they would have had certain set choices concerning domains of language use. The informants of Group A had
Issues in finding the appropriate methodology
Table 2. Informants in Ya˘gmur (1997) (2×2×20 = 80) Immigrant groups in Australia
Reference groups in Turkey
Low level of schooling
Group A (n = 20) Sydney Suburbs
RA (n = 20) Kayseri
High level of schooling
Group B (n = 20) Sydney Suburbs
RB (n = 20) Ankara
been in Australia for an average of 23.8 years, the informants in Group B for an average of 18.9 years. Five Group A informants reported that they had not visited their homeland for more than 10 years; one male informant had not been to Turkey for 22 years. In Group B, however, only 10 informants reported that they had not been back to their homeland for more than 10 years. The method of sampling can be identified as ‘snowball sampling’. Suitable persons already known to the researcher were contacted and encouraged to nominate further candidates for participation in the study. In total, at least a hundred persons were contacted; some of whom, however, refused to participate for various reasons. Some female subjects of Group A were reluctant to meet with the male researcher; a trained female assistant then conducted the tests and questionnaires. Finding suitable subjects to be included in Group B was particularly difficult, since few well-educated first generation Turkish immigrants meet the criterion of having spent at least 15 years in Australia. The two reference groups matched the immigrant groups’ educational level and regional background in Turkey. In order to match Group A, reference group informants were chosen from a central Anatolian city, Kayseri. This seemed appropriate since most of the immigrant subjects in Group A had also immigrated from central Anatolian towns and thus had a similar social and cultural background. As Group A immigrant subjects in Sydney mainly worked as labourers, the reference group was accordingly chosen from working class suburbs in Kayseri. In order to match Group B immigrant informants with monolingual speakers, twenty well-educated informants from Ankara were chosen as a reference group. The mean age was 47.6 years for reference group A (RA) and 42.2 years for reference group B (RB). In both Australia and Turkey, education was the main criterion for the grouping of the informants. The subject information presented above is summarised in Table 2. Finally, all other factors reported to be significant in studies on language maintenance and loss, both on the group level and on the individual level, were taken into consideration in the research design.
Kutlay Ya˘gmur
.. Instruments The following tests and questionnaires were used as data collection instruments and presented to the informants in the order as presented here: 1. 2. 3. 4. 5. 6.
Language Use-Choice Questionnaire Subjective Ethnolinguistic Vitality Questionnaire Self-rating Scale (Can-Do Scale) Controlled Lexical Naming Task Relativisation Production Task Picture-based Story Telling Task (‘the frog story’)
For further details on instruments and data collection procedures, see Ya˘gmur 1997. .. Representativeness of the sample In order to meet some key criteria such as the informants’ length of stay in the L2 environment and being first generation immigrants the study by design was not representative but purposive. As a result of the constraints of the sampling, the informants’ selection was limited to the eligibility criteria. The proportion of males and females was representative with almost equal numbers of males and females. The sample was also representative for the distribution of the Turkish-born population in Australia. According to the Bureau of Immigration Multicultural and Population Research (BIMPR) (1995: 4), 96 per cent of the “Turkey-born population was very heavily concentrated within a major urban location”; all the informants in this study came from the Sydney urban area. However, to have obtained results as representative as possible, the informant selection should ideally have been completely random. . Measuring attrition The purpose of the present paper, in keeping with the aims of the conference, is a critical evaluation of methodology in language attrition research, which is why the main focus here is on issues of formulating research questions, stating relevant hypotheses, constructing effective data collection instruments and informant selection. The details of identifying relevant extralinguistic variables will not be further discussed here; neither will I give any background on Ethnolinguistic Vitality Theory and its accompanying instrument, the ‘Subjective Ethnolinguistic Vitality Questionnaire’ or details of collecting discourse data and analyses procedures (for further details, see Ya˘gmur 1997). In the next section, I will instead concentrate on the relativisation production task to ex-
Issues in finding the appropriate methodology
emplify test development, operationalisation and analysis procedures. Finally, I will reflect on the crucial question posed by the conference organisers: ‘If I were to do my research all over again, what would I do differently?’
Measuring syntactic loss In this section, only one aspect of structural attrition will be discussed, namely relative clause production. The focus is on two commonly known types of relative clause used in Turkish, the dık and -en types, traditionally referred to as object and subject clauses (for details see Ya˘gmur 1997). Hypothesis: As a result of the language contact situation, Turkish speakers in an immigration context will experience language attrition in the use of late acquired complex Turkish forms of relative clauses.
The first question in this respect was whether there are any differences between first-generation Turkish immigrants in Australia and monolingual Turkish speakers living in Turkey in the production of relative clauses.
The relativisation production task As outlined earlier, Turkish relative clauses are acquired relatively late and in language contact situations are predicted to be lost relatively soon. In order to test this claim, a relative clause production test was designed, keeping in mind that in a test context it may be very hard to make informants produce the structures one wants to investigate. Even by providing a context or situation that would normally require multiple uses of relative clauses, informants might not produce the desired structures, but instead opt for simpler, analytic structures. However, it was assumed that if the informant was provided with the words of a sentence involving a relative structure in scrambled order and asked to repeat it in its correct form, they would process the words and provide the appropriate form; if they experience attrition concerning this particular structure, however, they would not be able to produce it correctly. On the basis of this assumption, ten Turkish sentences, each including a relative clause, were constructed. Particular attention was paid to issues of vocabulary by not including words used in specialised or technical fields. Words and contexts from everyday communication were selected in order not to cause extra memory strains. The order of words was then changed. In order to eliminate a recency effect, the positions of subject, verb, object, and relative clause were changed to prevent informants from following a pattern. To make memory processing easier, some words were presented in phrases, whereas key words were presented in isolation. In order to do the piloting for this task, the
Kutlay Ya˘gmur
sentences were read in scrambled order to five Turkish immigrants who had recently arrived in Australia and to six immigrants who had been in Australia for more than 17 years. The former group did not have much difficulty producing the sentences in the correct order; the latter group, however, showed considerable difficulties in processing the sentences in which the words had been scrambled. Instead of including the relative clauses, they simplified the structures. During the piloting of the relativisation production task, it was noted that after the sixth sentence, the informants started repeating the words in the order they had been read to them. Therefore, it was decided to include only five sentences in the task. In order to eliminate the memory factor, all words and phrases in the scrambled form were repeated to the informant two or three times and if requested, more than three times. Particular attention was paid to this aspect, since it was not our aim to test memory skill. Before initiating the task, the nature and procedure of the task was explained to the informant. By means of two example sentences, the informant was asked to complete a practice session in order to show that they had fully understood the task. If a subject sounded unsure, a second and – where necessary – third practice sentence was provided. The following is an example of a sentence in its scrambled form. The informant was expected to produce a structurally correct sentence. yedi (1) / babasının (2) / Ali (3) / kıpkırmızı elmaları(4) / getirdi˘gi (5). Ali babasının getirdi˘gi kıpkırmızı elmaları yedi. Ali father-poss-gen bring-part carmine apple-pl-acc eat-past Ali ate the carmine apples which his father brought.
The following five-point-scale was regarded to be methodical in terms of grading the sentences produced by the respondents, providing generalisable ratings for similar studies of relativisation in Turkish. 5. 4. 3. 2. 1.
All components placed correctly, the emphasis is on relativisation Incomplete relativisation as of “reduction” No relativisation but cause-effect relationship / conjoined clauses Coordinate structure (and) No target structure and post-relativisation (as in English)
Procedure Before implementing the task, the procedure was carefully explained to the informants. To save time, a tape recorder was used to record the informants’ utterances to be transcribed after the task. Prior to the actual task, the words and phrases were read to the informants twice or thrice in a mixed order; the
Issues in finding the appropriate methodology
informants were then asked to put them in the order that they are usually spoken. After the first reading, there was an interval of 2–3 seconds; the researcher then repeated the same words and phrases in the same scrambled order. The informants were asked whether they needed another repetition and if so, a third or even a fourth reading was provided in order to eliminate memory effects. The informants then put the words and phrases in a particular order of use. After the production of two, in a few cases three examples, the actual test was carried out. On average, the monolingual informants needed eight minutes to complete the task, whereas the immigrant group needed much longer.
Results of the relativisation task and discussion Before detailing the statistical analyses carried out on the data, I would like to present some observations concerning the sentences produced by the immigrant informants, some of whom concentrated on the meaning of the words presented to them and formed their own sentences. In these cases, they produced rather transparent structures. In other cases, informants opted for incomplete relativisation or the use of ‘ki’, which is a Persian loan indicating relativisation. Among the reference groups, this strategy was only found among the less educated informants in the RA group, none of the better-educated subjects opted for ‘ki’. Five informants out of the 20 RA group subjects chose the ‘ki’-strategy for the first sentence, only one using ‘ki’ and -en together. Leaving out certain words or inserting their own words was a common strategy among the immigrant informants. All reference group subjects, however, rigorously used the words provided to them. On the basis of the five-point scale described above the scores obtained from each group were computed. The mean scores of immigrant and reference groups are presented in Table 3. The overall results of the relativisation production task were turned into a scale resulting in a total score for each group. With respect to the differences between the immigrant groups (A and B) and reference groups (RA and RB), analyses of variances were done. The findings are summarised in Table 4.
Table 3. Mean scores for immigrant and reference groups on relativisation Group
Sentence 1
Sentence 2
Sentence 3
Sentence 4
Sentence 5
Immigrant A Immigrant B Reference A Reference B
3,15 3,90 4,55 4,90
2,72 3,95 4,55 5,00
3,57 3,84 4,65 4,95
3,27 4,25 4,95 4,95
3,52 4,00 4,85 4,70
Kutlay Ya˘gmur
Table 4. Analysis of variance between groups Group
Mean
SD
Immigrant A Immigrant B
15.30 19.75 F = 2.61 p = 0.013
6,56 3,86
Reference A Reference B
23.55 24.50 F = 5.047 p = 0.30
1,63 .95
Table 5. ANOVA results on Australian and Turkish groups (n = 80) Country Australia Turkey
Relativisation Task Mean 17.52 24.02 F = 47.87 p < .001 Eta = .62
SD 5.77 1.40
As is clearly seen from the above figures, the group score of even the less educated reference group A is significantly higher than the better-educated immigrant group B. Therefore, we are not in a position to suggest that the differences are a result of education received in the L1. Furthermore, in order to test our hypothesis, the data from the relativisation production task for the groups in Australia and Turkey were compared. The analysis of variance (ANOVA) between the immigrant Turkish groups A and B in Sydney and the reference groups RA and RB in Turkey indicates clear differences between the groups. Table 5 summarises the ANOVA results of the groups. These findings strongly suggest that the reference groups in Turkey perform much better with respect to the relativisation task. The differences between the groups in Australia and Turkey suggest a marked decrease in language processing skills of Turkish immigrants with respect to the above task. Between the groups in Australia and in Turkey, there are main effects both for group and for country, but there is no two-way interaction between the groups in Australia and Turkey: Groups A and B behave in the same way in both countries. However, with regard to the relativisation task, there is also an inter-
Issues in finding the appropriate methodology
Table 6. ANOVA’s for main effects by groups for the relativisation task Country
54.894
.000
Group Aust-Turkey Group (2-way)
9.472 3.979
.003 .050
action effect. There appears to be an interaction between country and groups. In particular, the difference between group RA (M = 23.55, SD = 1.64) and RB (M = 24.50, SD = 0.95) in Turkey is much smaller than that between their counterparts in Australia (group A M = 15.30, SD = 6.56; group B M = 19.75, SD = 3.86). Furthermore, the standard deviations show that the groups in Turkey are much more homogeneous in the production of correct realisations of relativisation.
Some remarks on the results of the relativisation task These findings clearly support our initial hypothesis that, as a result of the language contact situation, the informants in the migration context would not be able to perform equally well on the production of relative clauses as the informants in Turkey. Most of the immigrant subjects needed more time to produce the sentences, whereas the monolingual subjects in Turkey were generally swift and spontaneous in their productions of these sentences, irrespective of their educational background. It was assumed that the Turkish immigrants in Australia would have experienced L1 processing difficulties due to their extended stay in the L2 environment. Accordingly, it was assumed that irrespective of the extent of their use of L2 English, Turkish immigrants exhibited lexical and structural attrition as compared to Turkish speakers living in Turkey. It seems convenient to suggest that bilingualism of the immigrant groups plays a role in the differences observed between the performance of the immigrant and the reference groups. However, most of the immigrant Group A subjects have a rather low proficiency in English. The only viable explanation seems to be the restriction of domains for L1 use, Turkish being mainly limited to the domestic domain and active use of the mother tongue not being extensive. In line with these assumptions, the results suggest that Turkish immigrants experience language-processing difficulties and language attrition irrespective of their L2 skills. The findings of this study support Slobin’s (1977) claim that Turkish immigrants who have been in an L2 environment for a long period of time have difficulties in producing relative clauses in Turkish. A number of immigrant in-
Kutlay Ya˘gmur
formants opted for transparent structures by combining two dependent clauses in a coordinate structure or in a cause-effect relationship. However, in the collection of spontaneous speech data the same informants’ narrations of the Frog Story contained well-formed relative clauses. This implies processing difficulties rather than the loss of relative clause structures. Thus, L1 processing difficulties are apparent among the immigrant subjects, and this may have implications for language shift observed among the second and third generations. The difficulties in L1 use among the first generation subjects lead to more restrictions in the use of Turkish for second and third generations and a gradual shift towards the dominant language from generation to generation. . Recommendations for future research If I were to re-design my research, I would maintain most of the instruments with some revision and additions. However, concerning the informant selection, I would certainly adopt a random selection structure. Even though there are certain difficulties involved in selecting informants randomly, the study would in this way be rendered highly representative. In terms of research focus and data collection instruments, I would certainly like to add new dimensions and diagnostic testing instruments. In my earlier research, pragmalinguistic and functional aspects of language use have not been taken into account. Future work should concentrate not only on syntactic, but also socio-pragmatic aspects of language attrition. A functional language use perspective might contribute to our understanding of language attrition because of the emphasis on using language effectively and appropriately in different contexts. The study described in this paper was limited to the first generation Turkish immigrants in Sydney that the sample represents. The findings of this study cannot be generalised for language shift in the second generation, but only for the aspects measured within the population represented. With regard to lexical attrition, it is necessary to note that lexical production tests were limited to two categories of naming: animals and fruit and vegetables. In order to obtain further evidence, more lexical dimensions as well as lexical recognition tests should be included. In this study, the results of spontaneous speech data provided evidence for naming certain entities in the process of story telling but inclusion of specifically designed recognition tasks might provide firmer evidence. In this study, relative clause formation in Turkish was tested with only five sentences, for the reasons given above. Even though the results of the rela-
Issues in finding the appropriate methodology
tivisation production task support Slobin’s (1977) claims concerning relative clauses in Turkish, further evidence should be obtained by designing different tests. Considering the present linguistic situation among Turkish speakers in Sydney, it is highly recommended that Gonzo and Saltarelli’s (1983) predictions on inter- as well as intragenerational loss be tested. The findings of the present study clearly suggest that there are lexical retrieval difficulties as well as syntactic loss among Turkish immigrants. Following Gonzo and Saltarelli’s hypothesis, it can be suggested that second generation immigrants have already acquired those attenuated forms. In order to determine the extent of ‘incomplete transfer’, a detailed examination of L1 use among second generations in Australia might yield important findings with respect to language shift phenomena. This study has shown that the synthetic structures of Turkish relative clauses are especially vulnerable to language attrition. Investigation of the use of relative clauses among second-generation Turkish immigrants might yield relevant information concerning incomplete transfer. Finally, compared to Turkish immigrants in a European context, Turkish immigrants in Australia have much less contact with their homeland. A comparative study between these two contexts would shed light on the role of extralinguistic factors that are said to be influential in language attrition and – shift.
Note . LA stands for a person whose competence in a language has eroded as a result of language attrition. On the other hand, LC stands for a linguistically competent individual (Andersen 1982: 83–84).
References Aksu-Koç, A. (1994). “Development of linguistic forms: Turkish”. In R. Berman & D.I. Slobin (Eds.), Relating Events in Narrative: A crosslinguistic developmental study (pp. 329–385). Hillsdale, NJ: Lawrence Erlbaum. Ammerlaan, T., Hulsen, M., Strating, H., & Ya˘gmur, K. (Eds.). (2001). Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages. Münster: Waxmann.
Kutlay Ya˘gmur
Ammerlaan, T., Hulsen, M., Strating, H., & Ya˘gmur, K. (2001). “Language maintenance, shift, and loss: Work in progress”. In T. Ammerlaan, M. Hulsen, H. Strating, & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 1–13). Münster: Waxmann. Ammerlaan, A. (1996). ‘You Get a bit Wobbly...’ exploring bilingual lexical retrieval in the context of first language attrition. Unpublished Doctoral Dissertation, Nijmegen: Katholieke Universiteit Nijmegen. Andersen, R. W. (1982). “Determining the linguistic attributes of language attrition”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 83–118). Rowley, MA: Newbury House. Berman, R. A., & Slobin, D. I. (Eds.). (1994). Relating Events in Narrative: A cross-linguistic developmental study. Hillsdale, NJ: Lawrence Erlbaum. Bettoni, C. (1989). “Language loss among Italians in Australia: A summary of current research”. I.T.L.: Review of Applied Linguistics, 83–84, 37–50. Bureau of Immigration Multicultural and Population Research (BIMPR) (1995). Community Profiles: 1991 Census Turkey Born. Canberra: Australian Government Publishing Service. Bot, K. de (1990). “Metalinguistic awareness in Dutch immigrants in Australia”. Paper presented at the Combined ALAA/ALS Conference, September 1990, Sydney Macquarie University. Bot, K. de (2001). “Language use as an interface between sociolinguistic and psycholinguistic processes in language attrition and language shift”. In J. Klatter-Folmer & P. van Avermaat (Eds.), Theories on Maintenance and Loss of Minority Languages (pp. 65–81). Münster: Waxmann. Bot, K. de, & Clyne, M. (1994). “A 16-year longitudinal study of language attrition in Dutch immigrants in Australia”. Journal of Multilingual and Multicultural Development, 15 (1), 17–28. Bourhis, R. Y., Giles, H., & Rosenthal, D. (1981). “Notes on the Construction of a ‘subjective vitality questionnaire’ for Ethnolinguistic Groups”. Journal of Multilingual and Multicultural Development, 2 (2), 145–155. Clark, J. L. (1982). “Measurement considerations in language attrition research”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 138–152). Rowley, MA: Newbury House. Clyne, M. (1967). Transference and Triggering. The Hague: Martinus Nijhoff. Edwards, J. (1992). “Sociopolitical aspects of language maintenance and loss: Towards a typology of minority language situations”. In Fase et al. (Eds.), Maintenance and Loss of Minority Languages (pp. 37–54). Amsterdam and Philadelphia: John Benjamins. Cohen, A. (1986). “Forgetting foreign-language vocabulary”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 143–158). Dordrecht: Foris Publications. Cohen, A. D. (1989). “Attrition in the productive lexicon of two Portuguese third language speakers”. Studies in Second Language Acquisition, 11, 135–149.
Issues in finding the appropriate methodology
Dal Negro, S. (2001). “Language shift and change in a Walser dialect in Italy”. In T. Ammerlaan, M. Hulsen, H. Strating, & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 47–60). Münster: Waxmann. De Kadt, E. (2001). “You still speak German?” Teenage language skills in a Germanspeaking community. In T. Ammerlaan, M. Hulsen, H. Strating, & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 61–76). Münster: Waxmann. Elordui, A. (2001). “Processes of Language Shift and Loss: Evidence from Basque”. In T. Ammerlaan, M. Hulsen, H. Strating, & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 77–92). Münster: Waxmann. Fase, W., Jaspaert, K., & Kroon, S. (Eds.). (1992). Maintenance and Loss of Minority Languages. Amsterdam and Philadelphia: John Benjamins. Ferguson, C. (1975). “National sociolinguistic profile formulas”. In W. Bright (Ed.), Sociolinguistics (pp. 309–324). The Hague: Mouton. Giles, H., Bourhis, R. Y., & Taylor, D. M. (1977). “Towards a theory of language in ethnic group relations”. In H. Giles (Ed.), Language, Ethnicity and Intergroup Relations (pp. 307–348). London: Academic Press. Ginsberg, R. B. (1982). “Issues in the analysis of language loss methodology of the Language Skills Project”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 19– 36). Rowley, MA: Newbury House. Gonzo, S., & Saltarelli, M. (1983). “Pidginization and linguistic change in emigrant languages”. In R. W. Andersen (Ed.), Pidginization and Creolization as Language Acquisition (pp. 181–197). Rowley, MA: Newbury House. Haarmann, H. (1986). Language in Ethnicity: A view of basic ecological relations. Berlin: Mouton de Gruyter. Harres, A. (1989). “Being a good German: A case study analysis of language retention and loss among German migrants in North Queensland”. Journal of Multilingual and Multicultural Development, 10 (5), 383–399. Haugen, E. (1972). “Language and immigration”. In A. S. Dil (Ed.), The Ecology of Language: Essays by Einar Haugen (pp. 1–36). Stanford, CA: Stanford University Press. Jaspaert, K., Kroon, S., & van Hout, R. (1986). “Points of reference in first language loss research”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 37–49). Dordrecht: Foris. Kaufman, D. (2001). “Tales of L1 attrition – Evidence from pre-puberty children”. In T. Ammerlaan, M. Hulsen, H. Strating, & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 185– 202). Münster: Waxmann. Lambert, R. D., & Freed, B. F. (Eds.). (1982). The Loss of Language Skills. Rowley, MA: Newbury House. Macevichius, J. (2001). “The beginnings of language loss in discourse: A case study of referentiality in American Lithuanian”. In T. Ammerlaan, M. Hulsen, H. Strating, & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 235–247). Münster: Waxmann.
Kutlay Ya˘gmur
Mayer, M. (1969). Frog Where Are You? New York, NY: The Dial Press. Olshtain, E., & Barzilay, M. (1991). “Lexical retrieval difficulties in adult language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 139–150). Cambridge: CUP. Oxford, R. L. (1982). “Technical issues in designing and conducting research on language skills attrition”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 119–137). Rowley, MA: Newbury House. Richards, J. C., Platt, J., & Platt, H. (1985). Longman Dictionary of Applied Linguistics. London: Longman. Schmid, M. S. (2002). First Language Attrition, Use and Maintenance: The case of German Jews in Anglophone countries. Amsterdam and Philadelphia: John Benjamins. Slobin, D. I. (1977). “Language change in childhood and in history”. In J. Macnamara (Ed.), Language Learning and Thought (pp. 185–214). New York, NY: Academic Press. Smith, N., & Wilson, D. (1979). Modern Linguistics: The results of Chomsky’s revolution. Bloomington, IN: Indiana University Press. Smits, C. (2001). “Iowa Dutch inflection: Translations versus conversations”. In T. Ammerlaan, M. Hulsen, H. Strating, & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 299– 318). Münster: Waxmann. Van Els, T. (1986). “An overview of European research on language attrition”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 3–18). Dordrecht: Foris. Waas, M. (1993). Language Attrition among German Speakers in Australia: A sociolinguistic inquiry. Unpublished Doctoral Dissertation, Sydney: Macquarie University. Walters, J. (2001). “Socio-pragmatic processing in bilingual production and attrition”. In J. Klatter-Folmer & P. van Avermaat (Eds.), Theories on Maintenance and Loss of Minority Languages (pp. 83–140). Münster: Waxmann. Weltens, B., de Bot, K., & van Els, T. (Eds.). (1986). Language Attrition in Progress. Dordrecht: Foris. Ya˘gmur, K. (1997). First Language Attrition among Turkish Speakers in Sydney. Tilburg: Tilburg University Press.
Language contact and attrition The spoken French of Israeli Francophones Miriam Ben-Rafael Tel Aviv University
Introduction In recent decades, L1 attrition has been considered one of the typical processes of change that occur among languages in situations of contact. Researchers, however, do not unanimously regard L1 changes as indications of attrition. Basic issues in this context concern when codeswitchings (CS) and borrowings or interferences, confusions, and neologisms should be viewed as fulfilling specific functions and when they should be construed as compensation for forgetting and manifestations of attrition. As Köpke & Schmid show in the introduction to this volume, even after twenty years of “diligent investigation in L1 attrition,” theoretical and methodological questions “still by far outweigh the answers.” Definitions and methods of collecting and analysing data, as well as the parameters that are taken into account, often differ from one study to another. Terms such as ‘change’, ‘shift’, and ‘loss’ may overlap, notwithstanding numerous attempts at clarification (Waas 1996: 17–26). For Hiller-Foti (1985: 108), attrition does not necessarily indicate an inability to retain the use of L1 for communication. Instead, it may be the expression of a loss of L1 spontaneity and of the formation of a gap between active and passive command of the mother tongue. In a similar vein, Seliger (1985: 4) defines attrition as the “erosion of the linguistic performance of a first, or primary language which has been fully acquired and used before the onset of bilingualism.” Other researchers define the process of attrition ex negatio, i.e., by explaining “what language attrition is not” (Pavlenko 2002) or “what it cannot be” (Waas 1996: 26–30). In fact, attrition is a highly complex process that, in the view of some analysts, depends essentially on the particular circumstances
Miriam Ben-Rafael
of each speech community. Moreover, it should be construed differently when observed from intergenerational or intragenerational perspectives. Beyond this diversity, according to Hagège, a general profile of attrition and the way attrition impacts on the various components of the language will have to be developed (Hagège 2000: 109–110). In most cases, he states, the lexicon would be the first area affected, while grammar and phonology would show stronger resistance to erosion. A strong destabilisation of the latter components and the substitution of typical systems pertaining to a given language by other systems would, in extreme cases, announce the decline and loss of the language, if not its death (ibid.: 108). Various studies also illustrate different methods of investigation – experimental and quantitative, non-experimental and qualitative (see overview in Schmid 2003: 39–43). A few use several methods within a single research endeavour, e.g., self-reported data, can-do scales, fluency tests etc. (Waas 1996; Ya˘gmur 1997). Köpke & Schmid, in their comprehensive introduction to this volume, show that each method has its difficulties and focuses on diverse aspects of the attrition process. It is in this context that this chapter uses a number of methods in order to make out the different sides of the attrition process in spoken Israeli French or Franbreu (FrB). This “in-between” language or “interlanguage” has developed among Francophones in Israel, whom we shall call “Franbréophones” in the following. As they acquire Hebrew as an L2 and use it fluently in their daily lives, their French changes under Hebrew influence. Our question is whether these changes should be considered forms of attrition of French in contact with Hebrew. Our point of departure is the assumption that L1 attrition should be viewed not only as a psychological process but also as a sociolinguistic phenomenon, i.e., one that takes place in, and is accounted for by, speech situations. In different contexts speakers may variously “deviate” from standard forms and use L2 elements in various ways. Hence, the attrition of speakers’ linguistic competence in language-contact situations cannot be the only explanation of occurrences such as CS, calques, and innovations, and often may not even be the most important one. It is from this perspective that the study reported below focuses on a diversity of speech situations that differ in their degree of formality, employing different methods of data collection.
Language contact and attrition
.
Methodology
Our main group of subjects consists of Francophones who emigrated to Israel at the age of 18–30 and were between 45 and 60 years of age at the time of data collection. Their cultural and social context is relatively homogeneous: secondary education or higher, and middle socioeconomic class. Pursuant to the goal of the study, we collected our data by means of several methods and different samples. This methodology would elicit a general FrB profile, i.e., a set of speech activities typical of different speech situations. Each informant was investigated only once, and the number of subjects in the various phases totalled 83. The experimental group consisted of 65 Francophones living in Israel, and a control group of 18 non-Hebrew-speaking Francophones living in France.1 Furthermore, in this chapter we focus on qualitative analysis and do not consider statistical correlations. This approach, as a whole, allowed us to study the changes that occurred in the subjects’ French (Fr/L1) from a broad perspective. The sets of data include: 1. 13 Franbréophone narratives of the “Frog-Where-Are-You” story (Mayer 1969) (hereinafter, corpus 1 or C1); this is the less spontaneous phase because the subjects are openly requested to participate in a test of linguistic competence. 2. Recordings of discussions among French teachers at professional meetings (C2) – 5 meetings of 8 to 10 subjects (in total, 10 subjects who participated in at least one meeting). This phase is less formal than (1), since it involves colleagues who interact on an equal footing. However, these meetings took place at the Institut Français in Tel Aviv, in the presence of an official from the French Embassy, a circumstance that may have imposed constraints on the discursive activity. 3. 9 semi-structured interviews of Franbréophones conducted by L1 Hebrewspeakers who happened to be students of French (C3). These interviews were less formal than the previous phase since the subjects were in the position of interviewees. Moreover, since they were interacting with students of French, they felt committed to behave in a somewhat “didactic” manner. 4. 13 semi-structured interviews of Franbréophones by an interviewer who is a Franbréophone herself (C4). In this case, the atmosphere was less tense than in Phase 3, since the interviewer and the subjects belonged to the same milieu. Still, the semi-structured nature of the interviews was itself a factor of constraint.
Miriam Ben-Rafael
5. 9 spontaneous conversations of three or four Franbréophone participants in each case – 20 subjects in all (C5). The discourse at this stage was completely free because it took place among friends. This range of speech settings is widely representative of linguistic activity in general and allows us to sketch a profile of Israeli French as a whole. The various corpora complement each other and allow us to reveal the changing nature of Franbreu as determined both by processes of attrition and by sociolinguistic situations. The French speakers in our control group (CG) were from a similar background to our Israeli subjects but had never had any contact with Hebrew. The data for this group were collected in France. 12 of the 18 subjects in this group were asked to tell the “Frog Story” and 6 were asked to answer questions that paralleled those asked in Israel in Phases 3 and 4. The data elicited in this manner gave us an additional opportunity to assess the impact of the contact with Hebrew, to draw the contours of attrition, and to judge the pertinence of alternate hypotheses of linguistic change. We now present data from our diverse corpora, focusing on selected aspects of codeswitching, lexicon, and syntax. These aspects, elucidated below, sustain the contention that speakers in different speech situations diverge from standard L1 in different ways, revealing thereby that L1 attrition is not a sufficient explanation for occurrences that are characteristic in L1–L2 contact.
. Codeswitchings (CS) CS manifestations have been interpreted in various ways2 (Poplack 1980, 1982, 1988; Poplack et al. 1989; Clyne 1986; Jacobson, ed. 1998, 2001; Field 2002; Myers-Scotton 2002). For some (Dorian 1989; Hamers & Blanc 1983), they reflect attrition or some loss of L1; for others, they make up for a deficiency in L1 and provide lexical and semantic enrichment. Some researchers (Auer 1984, 1995, 1996; Lüdi 1990; Myers-Scotton 1993) consider them new discursive tools at a speaker’s disposal that eventually convey self-identity. For still others, CS and, in particular, borrowings essentially fulfil pragmatic, ludic, or mystic functions (Pergnier 1989). Pavlenko (this volume) emphasises that one should differentiate between L2 influence on L1 and L1 attrition. This issue, as Köpke & Schmid (this volume) report “. . .is a crucial one, since it is generally assumed in attrition studies that all cases of interference are due to attrition,
Language contact and attrition
without taking into account that interference may occur in ‘normal’ bilinguals as well.” What, then, has happened in the case of Franbreu? . CS without L1 Attrition French–Hebrew CS is the most remarkable feature in our data. CS may vary from the insertion of a single Hebrew element to the insertion of entire Hebrew sequences. (1) Mo (I began) le projet de construire une école régionale ‘the project of building a regional school’ M une école élémentaire régionale ‘a regional elementary school’ Mo ‘ad vav ‘up to sixth grade’ M même alef? où elle doit être cette école? ‘even first grade? where it should be this school?’ c’est un truc important ‘it’s important stuff ’ Mo tov . . . proyekt shel shesh milyon . . . ve-bet sefer shel shtem ‘esre kitot ‘well . . . a project of six million . . . and a school of twelve grades’ M (to E.) tu as entendu ce que Mo fait? un truc terrible ‘you heard what Mo does? big stuff ’ oui . . . il crée un bet sefer ‘yes . . . he’s creating a school’ (C5)
The large majority of single-element CS or borrowings involve nouns; they may be collective borrowings as well as personal ones. Many of them replace French equivalents; others compensate for the absence of appropriate French terms (gar’in = Israeli settlement core group) or for terms that exist in French but carry special connotations in their Hebrew versions (aliya vs. immigration). Some borrowings may alternate with their French equivalents but others are fixed and never switch with French terms, lest they lose their inherent meaning. Thus, there is a clear distinction between what is said in French and in Hebrew. Both French “mariage” (wedding) and Hebrew hatuna occur, for example, but a hupa (the Jewish nuptial ceremony) is referred to only in Hebrew. Similarly, both the terms “gouvernement” (government) and memshala occur, but no French equivalent of the knesset (the Israeli parliament) is used.
Miriam Ben-Rafael
Most borrowed nouns relate to the Israeli public sphere, e.g., immigration, education, work, economy, health, religion, or the army: (2) a.
immigration: aliya/immigration, gar’in/settlement core group, hanikhim/trainees or youth-movement members, ‘olim hadashim/new immigrants, shaliah/emissary, ‘ayarat pituah/development town b. education: hinukh/education, morim/teachers, ganenet/kindergarten teacher, bet sefer/school, sifriya/library, yo’ets/adviser, miqtso’ot behira/elective subjects, bagrut/matriculation at the end of high school, mefaqe’ah/school inspector, tsarfatit meduberet/spoken French c. work: avoda/work, parnasa/livelihood, kalkala/economy, qablan/building contractor, shutaf /associate, musakh/garage, misim/taxes (C4)
Most often, Franbréophone speakers prefer these Hebrew terms even when French equivalents are available, since the Hebrew terms better reflect Israeli cultural values that are taken for granted. Therefore, in Franbreu memshala (government) denotes the Israeli government and none other and mitun (economic slump or recession) refers primarily to Israel’s economic difficulties. The same is true for shevita (strike), tsava (army), and mikhlala (college). In professional conversations among Franbréophone French teachers, for instance, the speakers use borrowings relating to education which express their experience. In one of the meetings, for example, a non-Israeli Francophone who has no knowledge of Hebrew (Mut) was present. Even then, the speakers did not refrain from inserting Hebrew terms into their discourse and were oblivious to the fact that Mut had no idea what they meant. Only when Mut asked for the explanation of yo’ets (counsellor) was he given the French equivalent: (3) T1 qu’est-ce-qu’il a comme tafqid? ‘what is his function?’ T2 il est yo’ets tokhniyot limudim ‘he is a curriculum advisor’ T1 non non yo’ets c’est un autre ‘no no advisor this is another one’ T2 alors il est yo’ets de . . . ‘so he’s a counsellor of ’ Mutyo’ets c’est quoi? ‘yo’ets what’s that? T2 yo’ets c’est conseiller ‘advisor it’s an advisor’
An additional example is:
Language contact and attrition
(4) T3 ce garçon il était rosh bet ha-sefer le-hinukh ‘this guy was the head of the school of education’
Then, turning to Mut, T3 translates into French: T3 il était directeur de l’école d’éducation ‘he was the director of the school of education’ (C2)
It turns out that CS fulfils various functions in Franbréophone discourse, especially in spontaneous conversation (Ben-Rafael 2001a). It may point to the development of a discursive sequence or a rhetorical juxtaposition: (5) R
ils parlent très peu le français (in Prague). . . l’hébreu bevaday she-lo ve . . . ça suffit amplement . . . deux jours ‘they speak very little French (in Prague). . . Hebrew of course not and . . .two days . . . is plenty’ (C4)
(6) H il adore le chocolat u-mamash makhur le-ze ‘he loves chocolate he is really addicted to it’ (C4)
Speech shifts or reported speech are nearly always expressed in Hebrew: (7) la dernière fois qu’il est venu il a dit . . . at yoda’at ma . . . ze yofi kakha az amarti lo ze etgar bishvilkha ‘the last time he came he said . . . you know what . . . it’s nice that way so I told him it’s a challenge for you’ (C4)
CS also occurs in the forms of ready-made expressions, idioms of religious nature, greetings, and congratulations: (8) toda la-el ‘thank God’, zikhrono livrakha ‘may his memory be blessed’, barukh ha-shem ‘bless God’, halevay ‘may it only be’, mazal tov ‘good luck’, kol ha-kavod ‘congratulations’ (C3)
In other cases, CS structures the conversation by signalling the end of a sequence and the transition to a new one: (9) M il voulait étudier à l’époque . . . il a fait un petit morceau de mekhina vraiment tout petit . . . après il est revenu . . . Rivka tagidi ma im Pnina? ‘he wanted to study at that time. . .he did a small bit of a preparatory course. . . after he came back. . . Rivka, tell me what’s going on with Pnina?’ R Pnina I nesu’a ‘Pnina is married’
Miriam Ben-Rafael
M ah oui elle s’est mariée! quand? ‘ah yes she got married! when?’ R im amerika’i she-haya po ‘with an American who was here’ (C5)
CS also defines the relationship among different speakers and provides a way of structuring the constellation of the conversation. In example (10), S speaks in French with her two interlocutors, switches to Hebrew when she turns specifically to E, and reverts to French when she addresses M: (10) S
(addressing E and M) j’ai appelé Avi pour analyser avec lui la situation. . . (to E.) hu hakham non? . . . (to M.) tu sais qui c’est Avi? ‘I called Avi to analyse the situation with him . . . (To E.) he’s clever isn’t he? (To M.) you know who Avi is?’ (C4)
In the following example, the same process appears. Hu tells her friends, M and E, what happened to her son when he went to register in the university. Although she speaks almost exclusively in Hebrew, when she addresses only M and asks her to repeat E’s question, which she did not grasp, she codeswitches to French: (11) Hu az hu nirsham la-mekhina ve. . . ‘so he signed up for a preparatory course and...’ M hmm. . . Hu alors en un mois il a fait deux bagruts moins qu’un mois je crois non? hu asa gam tanakh ‘so in a month he did two matriculation exams less than a month I think isn’t that right? he also did Bible’ gam lashon be-pakhot me-hodesh ki hu haya gam be-miluim ‘also grammar in less than a month because he was also called up for reserve duty’ E hmm. . . Hu az hu hitqabel la-universita bli mekhina ve -shum davar ‘so he was accepted at the university without a preparatory course or anything’ E il est content à l’univ? ‘is he happy at the univ(ersity)?’ Hu. (addressing only M) qu’est-ce-qu’il dit, E.? ‘what does he say, E.?’ (C5)
CS is also often used for subjective and expressive support. Segmental CS and borrowings become means of judgmental and personal expression. In (12), for
Language contact and attrition
example, the speaker becomes agitated when speaking about a friend who has just lost her husband: (12) oh bein elle est vaillante . . . c’est lui qui faisait tout et . . . elle est beseder at yoda’at . . . hi mamash beseder . . . ze haya nora ve-ayom . . . on lui dit ton mari est mort . . . at yoda’at ma ze efshar lehishtolel ‘oh yeah she is courageous . . .. he’s the one who did everything and . . . she’s OK you know . . . she’s really OK . . . it was awful and terrible . . .you’re told that your husband is dead . . . you know what it is you can go crazy’ (C5)
Another speaker uses a similar strategy in reference to the present situation in Israel: (13) oh j’en ai assez j’en ai marre mamash nimas li ‘oh I have had enough I’m sick of it I’ve really had enough’ (C4)
In these cases, the switch from one language to the other takes place in a very natural way. Borrowings and segmental CS are generally unflagged and are inserted in a Franbreu discourse that remains fluid and uninterrupted, both in spontaneous and professional conversations as well as in interviews. In all these settings, casual CS can be evaluated according to discursive efficiency and evoke no special reaction from interlocutors (see examples (1), (9), (12)). . CS with L1 attrition On the other hand, there are also the cases where speakers encounter difficulties in retrieving specific French terms. Aware of their problems, they themselves may speak of oblivion and loss of memory, and express embarrassment: (14) y a certains mots que je retrouve pas . . . je commence une phrase et je m’arrête parce que ce mot j’ arrive pas à le retrouver ‘there are words I can’t find. . . . I begin a sentence and I stop because I can’t manage to find this word’ (C4) (15) je cherche mes mots je sens que je perds la langue . . . des mots m’ échappent . . . je sais pas les nouveaux mots ‘I grope for my words I feel that I’m losing the language . . . words escape me . . . I don’t know the new words’ (C4)
In such cases, CS appears as a stratagem of last resort for overcoming lexical inaccessibility. This, however, takes place mostly in regard to items that are rather infrequent in daily life:
Miriam Ben-Rafael
(16) mon père est venu avec plein de . . . un tas de livres . . . avec des choses comment on appelle ça . . . une krikha . . . une krikha en cuir ‘my father came with a lot of . . .plenty of books . . .with things . . . how do you say it . . . a binder . . . a leather binder’ (C4)
One strategy for dealing with an inaccessible French term is to switch to Hebrew and then back to French (17); another strategy is to attempt to express oneself in French before switching to Hebrew and eventually finding the appropriate French word (18). Sometimes, however, the reconstruction remains unsuccessful. In (19), for instance, the speaker feels that the word “dettes” (debts) is not appropriate, but since she cannot remember “emprunt” (loans), she switches to the Hebrew equivalent halva’ot. (17) dans la hevra des français . . . dans la société française ‘in the society of French people . . . in the French society’ (C3) (18) c’est mon pays . . . moledet . . . patrie ‘it’s my country. . . homeland. . . homeland’ (C3) (19) nous avons fait des *dettes . . .halva’ot we have taken *debts . . . loans’ (C3)
Briefly stated, lexical attrition appears when speakers need precise terms. This phenomenon is evident in the interviews, as the previous examples show, and in the narratives, when speakers are asked to tell a story on the basis of a set of pictures.3 In the latter case, respondents manage to retell the story but some of them, finding it difficult to pinpoint key words, use paraphrases or other strategies and, failing these, codeswitch to H. Thus, “ruche” (beehive) becomes kaveret (20), “hibou” (owl) becomes tinshemet (21), and “falaise” (cliff) becomes tsuq (22): (20) qu’est-ce-que c’est . . . kaveret lo yoda’at ‘what’s that . . . a hive . . . I don’t know’ (C1) (21) tinshemet . . . oiseau . . . lo yoda’at ‘an owl . . . a bird . . . I don’t know’ (C1) (22) tsuq . . . lo yoda’at ‘a cliff . . . I don’t know’ (C1)
When narrators grope for terms unsuccessfully, they may prefer to avoid the concept or ignore the specific sequence. In the following example, the narrator, failing to recall both “essaim” (swarm) and “ruche” (hive), goes on with the narrative but avoids these terms, thereby signalling her inability to do so:
Language contact and attrition
(23) le chien a trouvé . . . ze ani lo yoda’at . . . des abeilles ‘the dog found . . . I don’t know this . . . bees’ (C1)
Later, when she does not find the word “hibou” (owl), she uses a generic term, “oiseau” (bird): (24) Philippe s’est effrayé de du oiseau . . . ani lo yoda’at ekh qor’im le-ze ‘Philippe was scared by a a bird . . . I don’t know what it’s called’ (C1)
. Lexicon In the lexicon, too, our corpora show wide deviations from standard French. In particular, we encounter a large number of lexical interferences. Many expressions are Hebrew calques and almost literal translations from Hebrew: (25) hier j’ai fermé vingt-huit kgs /cf. etmol sagarti ‘esrim kilo – meaning: hier je suis arrivée à mon vingtième kg. d’amaigrissement ‘yesterday I lost my twentieth kilo’ (C4)
Confusion, groping for words, and self-questioning about the accuracy of given terms are evident. The speakers also stumble over certain words and distort ready-made expressions: (26) je suis né * le 1937 – instead of: je suis née le (date and month) 1937 or: . . . en 1937 ‘I was born in 1937’ (C3) (27) j’ai mis des prunes * à table – instead of: j’ai mis des prunes sur la table ‘I put prunes on the table’ (C4) (28) les hommes sont plus brutaux . . . *brutals . . . brutaux . . . ‘men are more brutal . . . *brutals . . . brutal . . . oh bein je sais plus oh I don’t know anymore’ (C5)
All-purpose words like “truc,” “machin,” and “chose” (stuff, thing) are substitutes for problematic terms, the verb “faire” (to do) is used frequently, and lexical repetitions are abundant while numerous French lexical variants are ignored: (29) c’est vrai que chacun a ses raisons pour lesquelles il décide d’ aller vivre au kibuts . . . des raisons . . . des raisons . . . chacun a ses raisons . . . des raisons qui sont liées avec des raisons économiques peut-être quelquefois des raisons . . . .
Miriam Ben-Rafael
elles sont * habillées de d’idéal ‘it’s true that everybody has his reasons for deciding to go and live on a kibbutz . . . reasons . . . reasons . . . each one has his reasons . . . reasons that are connected with economic reasons, perhaps sometimes reasons . . . idealistic ones’ (C4)
Lexical confusions, calques, and oblivion mingle and intersect. The Franbréophone speakers make themselves understood but non-Franbréophones would express things differently, as our control group (CG) clearly shows. Members of the control group make richer and more varied lexical choices and exhibit lexical accuracy that Franbréophone narrators fail to attain. When they tell the “Frog story”, for example, CG narrators use the most appropriate verb, “coincer” (30), to describe the dog getting his head stuck in the jar: (30) Snoopy se coince la tête (‘Snoopy jammed its head’) le chien s’est coincé le museau(‘the dog’s muzzle got jammed’) mon chien (. . .) se coinça la tête (‘my dog jammed his head’) le chien resta coincé (‘the dog remained jammed’) en étant coincé il tomba par la fenêtre (‘jammed, he fell through the window’) (CG)
In contrast, in Franbréophone narratives the narrators invoke numerous – more or less precise – lexical variants to describe the same event (31), and only one narrator uses the accurate term, “coincé”: (31) le chien introduit sa tête dans le bocal (‘the dog placed its head in the jar’) le chien a mis la tête dans la jarre (‘the dog put its head in the jar’) le chien a toujours la tête enfoncé dans le bocal (‘the dog still has the head stuck in the jar’) le chien peut pas enlever sa tête du bocal (‘the dog cannot take its head out of the jar’) le chien a sa tête flanquée dans le bocal (‘the dog flung its head into the jar’) (C1)
This recurs when the narrators describe feelings; the lexicon of Francophone narratives is much more differentiated and precise (32) than that in the Franbréophone ones, which mostly express less feelings. Moreover, when the narrators do express emotion, they use a more general and limited vocabulary, e.g., “triste,” “heureux,” and “content” (33): (32) ça le perturbe (‘it upsets him’) le hibou est perturbé (‘the owl is upset’)
Language contact and attrition
Tom le console (‘Tom consoles him’) le chien souffre (‘the dog suffers’) ils partent rassurés (‘they go back reassured’) le chien est épuisé (‘the dog is exhausted’) la chouette est effrayée (‘the owl is afraid’) tout le monde regrette d’avoir entrepris cette course (‘everybody regrets this race’) la grenouille va être bien traitée (‘the frog will be treated well’) (CG) (33) I sont tristes (‘they are sad’) la petite grenouille est heureuse (‘the little frog is happy’) I sont très contents Toto et le chien (‘Toto and the dog are very pleased’) alors tout le monde était très content (‘everybody was very pleased’) il est assez content (‘he was quite pleased’) (C1)
The Franbréophone speakers sometimes struggle to find precise terms: (34) y avait un cerf . . . comment on appelle ça .. un cerf. . . ça s’ appelle pas le bambi . . . le bambi. . . c’est un cerf plutôt ‘there was a deer . . . what do you call that . . . a deer . . .it’s not called a bambi . . . the bambi . . . no it’s a deer’ (C1)
When they do not succeed, they try other strategies such as CS, French substitutes or others, as already mentioned (ex. (22)–(24)). Another typical Franbreu feature in the lexical domain is the tendency to simplify, i.e., to subject French items to a semantic reduction that reflects the semantic scope of Hebrew equivalents: (35) *chanson (‘song’) instead of: poème (‘poem’) / cf. shir (= ‘song, poem’) (C3) (36) *métier (‘profession’) instead of: matière scolaire / cf. miqtso’a (= ‘school subject, profession’) (C3) (37) *apporter (‘to bring something’) instead of: amener (‘to bring somebody’) / cf. lehavi (= ‘to bring somebody or something’) (C4) (38) il me *donne pas à travailler, instead of: il me laisse pas travailler / cf. hu lo noten li la-’avod . . . (= ‘he doesn’t let me work’) (C4) (39) le docteur * invite tous les gens à la même heure, instead of: le docteur convoque . . . / cf. ha-rofe mazmin et kol ha-anashim be-ota ha-sha’a (= ‘the doctor calls all the people [to come for their appointments] at the same time’) (C4)
Miriam Ben-Rafael
In addition to such cases, which are probably evidence of attrition, there is also a tendency to lexical innovation in this context of “oblivion and confusion”, a pattern that applies primarily to verbs: (40) *profaniser (‘to profane’) instead of: profaner; * légitimiser (‘to legitimise’) instead of: légitimer; * activiser (‘to activate’) instead of: activer (C5)
Generally speaking, however, new Franbreu verbs are combinations of Hebrew radicals and French verbal suffixes, with priority given to the French first verbal group -er:4 (41) tsilumer (tsilum + er) (‘to make a photo’); sidurer (sidur + er)(‘to organise’); zazer (zaz+er) (to move); tikhnuner (tikhnun+er) (‘to plan’) (C5) (42) peut-être qu’on a fisfusé le concert ‘we may have missed the concert’ (C5)
Importantly, however, these Franbreu neologisms are not necessarily indications of attrition; they may more simply be new lexical Franbreu variants that are semantically convenient for Franbréophone discourse. This feature is quite common and evidenced in all our corpora.
. Syntax As for the syntax, the language-in-contact literature mentions changes as well as phenomena of attrition (Appel & Muysken 1987; Dorian 1989; Schmid 2002). Regarding Acadian and Ontario Canadian French, for instance, a tendency to omit some verbal forms and to alter the word order is evident (King 1989; Mougeon & Beniak 1989, 1991). In American Spanish, one finds a tendency to simplify the verbal system and to favour morphological innovations (Silva Corvalan 1986). Speakers of “Québécois” French underuse the subjunctive form (Chantefort 1976) and tend to omit que in relative and circumstantial phrases (Martineau 1985). Some scholars, however, focus primarily on lexical changes, claiming that grammatical ones are relatively infrequent. For them, syntax belongs to the hard core of the language and is relatively resistant to the influence of other languages and new social or cultural circumstances (Hagège 1987, 2000), and grammatical changes in general, and particularly in language-contact situations, progress very slowly in any case (Leeman Bouix 1994; Walter 1988).
Language contact and attrition
In Franbreu, grammatical deviations from standard French are frequent indeed, and since they never appeared in our control group, we may consider them as marks of attrition. Most of these deviations result from the contact of Hebrew and French. One example among many is the tendency to simplify the verbal system. Hence, verbal agreement (concordance des temps) is not strictly respected. The distinction between past tenses – passé-composé and imparfait – which does not exist in H is not always retained while the conditional form is expressed by the future tense – also like in Hebrew: (43) si I * seront au tikhon si je * vivrai je leur dirai d’apprendre le français ‘if they * will be at high school if I * will live I will tell them to study French’ (C4) (44) on va voir si la cassette * marchera ‘we will see if the cassette *will work’ (C4)
The subjunctive mode is often disregarded – under the influence of Hebrew, which lacks this verbal form: (45) je pense pas que les gens * sont heureux, instead of: soient ‘I don’ t think people are happy’ (C3) (46) je veux que vous * saviez, instead of: sachiez ‘I want you to know’ (C3)
Verbal valencies differ from standard French and are often calques of Hebrew equivalents: (47) elle aide * à sa mère, cf. I ‘ozeret le-ima, instead of: elle aide ( ) sa mère ‘she helps her mother’ (C3)
Speakers also experience difficulties with the pronominal system. The relative pronoun que, for instance, is used instead of qui and dont: (48) tu me donnes ce * que j’ai besoin, cf. ata noten li ma she-ani tsarikh, instead of: . . .ce dont . . . ‘you give me what I need’ (C5)
This corresponds to the Hebrew grammatical system, in which only one element, ‘she’, expresses the French relative pronouns qui, que, dont, and où. Speakers also omit adverbial pronouns, such as y and en, that have no Hebrew equivalents. Where this occurs, French adverbs such as là and là-bas (“there”) are used as y substitutes:
Miriam Ben-Rafael
(49) je vais souvent là, instead of: j’ y vais souvent, cf.: ani holekhet le’itim qerovot le-sham ‘I often go there’ (C3)
Finally, indefinite determinants are sometimes omitted, again like in Hebrew which does not have indefinite articles: (50) le français est * Ø jolie langue, cf. tsarfatit Ø safa yafa, instead of: le français est une jolie langue ‘French is * Ø( ) nice language’ (C3)
In view of these examples, we can speak of grammatical attrition. The grammatical changes illustrated here may be construed as Hebrew grammatical interferences. Yet, in order to better understand this attrition pattern, it is also necessary to point out that some Franbreu grammatical variants – especially those that concern the conditional and subjunctive modes and relative pronouns (see ex. (43)–(46), (48)) – are also recurrent in some discursive registers of native spoken French. Though, it is to signal that such variants did not occur in the data from our own Francophone control group. They have been identified as typical syntax variation tendencies even outside any contact situation and have often been related to lower education, though, in fact, they were identified by researchers in the French of popular milieus as well as that of cultivated circles (Ager 1990: 138; Blanche-Benveniste 1997: 53–54; Gadet 1989: 148). These features are slips, confusions and variants which belong to the reality of a spoken language (Leeman-Bouix 1994: 101–104) and depend, among other factors, on the variety of discursive situations. BlancheBenveniste (1990, 1997) and Gadet (1989) signal a simplification of the verbal system and a tendency to misuse or omit the subjunctive mode. In the same vein, Frei (1971: 200) speaks of the problematic of the conditional while Leeman-Bouix (1994: 101, 102) reports confusions between the preterit and the compound past (il *disa vs dit) and regularised forms of the present tense (vous *disez vs. dîtes). Numerous researchers (Blanche-Benveniste 1997: 42; Gadet 1989: 147–159; Leeman-Bouix 1994; Walter 1988: 296–297) show that the relative pronoun que tends to replace qui and dont and that the compound relatives (lequel, laquelle etc.) tend to disappear. There is also a tendency to extend and generalise the use of que as a link word (Ager 1990: 138; Gadet 1989: 161–168). These tendencies which are endemic to spoken French and characterise specific discursive situations thus converge with those tendencies that we found in our own corpora. We could then argue that this is the very reason of their presence in Franbreu. However, how could we explain that these tendencies
Language contact and attrition
do not appear among our control group? The hypothesis which I suggest here refers to the convergence of features of spoken French and Hebrew grammar. More precisely, it is when typical features of spoken French converge with equivalent grammatical features in Hebrew that their emergence in Franbreu is accelerated through the contact of French with Hebrew. This hypothesis, if validated, could explain both the presence of those features in Franbreu and their absence from the control group who had no contact with Hebrew. On the other hand, when French grammatical features have no equivalent in Hebrew, they tend to resist its influence – as shown below, for the French negative, the gender of nouns and the prepositional system. a. The French negative can take on two forms in spoken language – the simple form (pas) following the verb, and the compound form (ne . . . pas) on both sides of the verb or the auxiliary (je (ne) parle pas, je (n’) ai pas parlé). Pas is by far the most frequent in informal spoken French, while ne...pas appears in more formal contexts (Coveney 1990, 1996; Gadet 1989; Moreau 1986; Posner 1997; Sanders 1993). The Hebrew negative system, on the other hand, consists of the only pre-verbal negator (lo). Hence, the difference between the French and the Hebrew negative systems concerns both the word order and the number of eventual negators. When it comes to our findings, it appears that French is not influenced by Hebrew. In the spontaneous conversations, like in spoken French, the simple negator pas overcomes the bipartite negation ne..pas, and in the interviews, which are more formal, the two models are used interchangeably. The resistance of the French system is such that even when the negator itself is borrowed, the French negative word-order remains intact: (51) le yayin était lo tov; cf. le vin était pas bon vs. hayayin lo haya tov ‘the wine was not good’ (C5)
A possible explanation for these results is the fact that the French and Hebrew negative systems do not converge, and therefore do not influence each other (see Ben-Rafael 2001b, 2003). b. As for the gender of nouns, both French and Hebrew have two genders – masculine and feminine; however, the gender of individual nouns does not necessarily correspond in the two languages. Our findings show that wherever differences exist between French and Hebrew words, it is the French gender which prevails. The Hebrew gender only influences the French article in borrowed words which are adopted en bloc together with their gender- thus “le fromage” (cheese) or “le parti politique” (political party)
Miriam Ben-Rafael
remain masculine in Franbreu like in French, while the Hebrew borrowings la gvina and la miflaga are feminine like in Hebrew. Another interesting observation is that when French adjectives are paired with borrowed nouns they also often agree with the gender of those Hebrew borrowed terms: (52) il y a toujours des choses que tu peux acheter à des tnayim (m.) spéciaux (m.) vs. il y a toujours des chose que tu peux acheter à des conditions (f.) spéciales (f) ‘there are always things you can buy on special terms’ (C4)
c. The prepositional system is another area where French shows resistance to Hebrew influence in the context of a lack of equivalence. Yet, some typical French forms do not appear in Franbreu (such as: je vais sur Paris) while one may find confusions (53) and Hebrew interferences (54): (53) *à la fin de compte il fait tomber le nid, instead of: en fin de compte. . . ‘at the end he drops the nest’ (54) on voit bien là *dans cette photo, cf. . . .ba-tmuna haz’ot, vs. . . . sur cette photo ‘it is quite clear here in the picture’
All in all, most prepositional changes in Franbreu concern the verbal valencies which are Hebrew calques, as was demonstrated in (47) above. However, despite these Franbreu variations, French prepositions remain relatively stable and even Hebrew borrowings do not lead to incorrect usages of French prepositions (55) (see also Ben-Rafael 2002a): (55) je vais essayer de faire mimuner, cf. je vais essayer de faire financer ‘I will try to get some financial help’
In conclusion, we may say that some attrition is in progress, due to two essential factors: the effect of Hebrew on the one hand, and the acceleration of natural tendencies in French grammar in the Hebrew-French contact situation on the other. The Hebrew syntax influences French mainly when it concurs with syntax variation tendencies of certain discursive registers in spoken French. When other characteristic tendencies of French differ from Hebrew grammar, French shows a stronger ability to resist the influence of Hebrew; when, moreover, a discrepancy appears between the two systems, it is French which generally prevails.
Language contact and attrition
. Conclusion This study investigated Francophone immigrants to Israel who have acquired Hebrew, the country’s principal language. French is now of secondary importance in their daily lives and, as a spoken language, is open to new influences. It has evolved into the vernacular that we call Franbreu. This medium of expression is quite remote from French speech in France today and has strayed far from the preferred expressions and idioms of the mainstream vernacular. In many respects, it reflects a “Hébréophone” mood. Throughout the phases studied – interviews, spontaneous conversations, professional discussions, and narratives – we uncovered a general Franbréophone profile. All these phases have common lexical and grammatical tendencies that fluctuate according to the degree of formality of the discursive situation. In more formal situations – narratives, professional meetings, and interviews conducted by “Hébréophones” – the speakers’ efforts to maintain a coherent French discourse show signs of some lexical attrition. In contrast, in informal contexts – spontaneous conversations and interviews conducted by a Franbréophone – lexical changes are less obvious while codeswitchings and lexical innovations do not seem to be compensatory elements for lexical insufficiencies. They are, instead, new vehicles that, among other functions, (1) enrich discursive resources, (2) characterise specific social realities, and (3) express subjective feelings and self-identities. Grammatical changes – this is stated with respect to all our corpora – may be viewed as reflections of the Hebrew grammatical system and, thus, be interpreted as attrition. However, they are often similar to some other registers of spoken French. From this perspective, these changes may be construed as manifestations of inherent tendencies of French that are accelerated by the contact with Hebrew. Moreover, as a rule, French grammar displays resilience and remains predominant in the structuring of the Franbreu discourse wherever there is a discrepancy between Hebrew and French systems. In conclusion, we maintain that the changes we find in Franbreu should not be viewed entirely as attrition phenomena. Some undoubtedly result from linguistic erosion but others are genuine contributions to French when it comes into contact with Hebrew. This “in-between” language – or interlanguage – varies in its discursive practices according to speech contexts, as different degrees of attrition as well as lexical and grammatical enrichment and innovations build up the French-Hebrew bilingual discourse. Aware of the complexity that characterises the definitions and methods of research on linguistic attrition, we chose in this study to combine several meth-
Miriam Ben-Rafael
Nature of changes
Attrition
Lexicon
Lexical confusions Hebrew interferences Semantic reductions Repetitions All-purpose terms Fr substitutes Lexical voids
Thereof: neologisms Codeswitching
Syntax
Non-Attrition
New variants 1. New discursive means 2. Refer to the Israeli public scene 3. Express feelings, self, and identity
Verbal simplification Pronominal reduction Pronoun and determinant omission Verbal valence variations Grammatical neologisms
Grammatical neologisms connected to lexical innovation H grammatical influence Fr grammatical resistance ↓ ↓ Fr attrition in progress Contributions and enrichment The emergence of Franbreu
Figure 1. Fr/L1 and H/L2 in contact and the emergence of Franbreu
ods to obtain a comprehensive dynamic picture of Franbreu (Figure 1). We hope that by so doing we have made a contribution to the search for an interpretation of the concept of attrition by drawing a distinction between changes that we may define as attrition and others that we may view as contribution. At this stage, we may envisage further work that would add a new dimension of comparison by juxtaposing our Franbreu speakers with their offspring who acquired Hebrew as the language of their environment but grew up hearing their parents speaking Franbreu. The interest of such research may reside in the question of whether or not this second generation retains French – or, more accurately, Franbreu – and carries it in new directions. Such a comparison would presumably help us to define the notion of attrition and its linguistic derivatives with greater precision and in a new empirical context.
Language contact and attrition
Notes . Two-thirds of the subjects were women and one-third men, but this variable is not analysed any further in this work. . See for an overview Ben-Rafael 2001a and 2001b. . These proceedings have been used by numerous researchers, and a major example is Olshtain & Barzilay’s (1991) work with American subjects living in Israel. . The examples of lexical neologism presented here are also examples of lexical morphology; though, we prefer, in the restricted frame of this chapter, to discuss them under the heading of lexical issues in language contact.
References Ager, D. (1990). Sociolinguistics and Contemporary French. Cambridge: CUP. Appel, R., & Muysken, P. (1987). Language Contact and Bilingualism. London: Edward Arnold. Auer, P. (1984). Bilingual Conversation. Amsterdam and Philadelphia: John Benjamins. Auer, P. (1995). Code-Switching in Conversation, Language, Interaction and Identity. London: Routledge. Auer, P. (1996). “Bilingual conversation, dix ans après”. Aile, 7, 9–34. Ben-Rafael, M. (1994). “Contact de langue et simplification”. Cahier du français contemporain, I, 77–87. Ben-Rafael, M. (1999). “Emigration et contact de langues: emprunts et configurations sémantiques en Franbreu.” In F. S. Martinez & C. H. Gonzalez (Eds.), Las lengas en la Europa Communitaria III (pp. 423–438). Amsterdam: Rodopi. Ben-Rafael, M. (2001a). “Codeswitching in the immigrants’ language: The case of Franbreu”. In R. Jacobson (Ed.), Codeswitching Worldwide II (pp. 251–307). Berlin: Mouton de Gruyter. Ben-Rafael, M. (2001b). Contact de Langues: le français parlé des francophones israéliens. Unpublished Doctoral Dissertation, Université de Tel-Aviv. Ben-Rafael, M. (2002a). “The French preposition in contact with Hebrew”. In S. Feigenbaum & D. Kurzon (Eds.), Prepositions in their Syntactic, Semantic and Pragmatic Context (pp. 194–209). Amsterdam and Philadelphia: John Benjamins. Ben-Rafael, M. (2002b). “Le franbreu: les traits essentiels”. In D. Mendelson (Ed.), La culture francophone en Israel (Vol. I, pp. 79–96). Paris: l’Harmattan. Ben-Rafael, M. (2003). “Pratique discursive et mise en place grammaticale”. Marges Linguistiques, 5 (2), 168–180. Blanche-Benveniste, C. (1990). Le français parlé, études grammaticales. Paris: CNRS. Blanche-Benveniste, C. (1997). Approche de la langue parlée en français. Paris: Ophrys. Chantefort, P. (1976). “Diglossie au Québec, limites et tendances actuelles.” Langue Française, 31, 91–104.
Miriam Ben-Rafael
Clyne, M.C. (1986). “Towards a systematisation of language contact dynamics”. In J. A. Fishman et al. (Eds.), The Furgusonian Impact (pp. 483–492). Berlin: Mouton de Gruyter. Coveney, A. (1990). “The omission of ‘ne’ in spoken French”. Francophonie, 1, 38–43. Coveney, A. (1996). Variability in Spoken French: A sociolinguistic study of interrogation and negation. Exeter: Elm Bank. Dorian, N. C. (Ed.). (1989). Investigating Obsolescence – Studies in language contraction and death. Cambridge: CUP. Field, F. W. (2002). Linguistic Borrowings in Bilingual Contexts. Amsterdam and Philadelphia: John Benjamins. Frei, H. (1971). La grammaire des fautes. Genève: Slatkine Reprints (first edition: 1929). Gadet, F. (1989). Le Français ordinaire. Paris: Armand Colin. Hagège, C. (1987). Le Français et les siècles. Paris: Odile Jacob. Hagège, C. (2000). Halte à la mort des langues. Paris: Odile Jacob. Hamers, J. F., & Blanc, M. (1983). Bilingualité et Bilinguisme. Bruxelles: Pierre Margada. Hiller-Foti, I. (1985). “Erhebung über den Verlust der Muttersprache Deutscher Immigranten in Sizilien”. Rassegna Italiana di Linguistica Applicata, 17 (1), 107–110. Jacobson, R. (Ed.). (1998). Codeswitching Worldwide. Berlin: Mouton de Gruyter. Jacobson, R. (Ed.). (2001). Codeswitching Worldwide II. Berlin: Mouton de Gruyter. King, R. (1989). “On the social meaning of linguistic variability in language death situations: Variations in Newfoundland French”. In N. C. Dorian (Ed.), Investigating Obsolescence – Studies in language contraction and death (pp. 139–148). Cambridge: CUP. Leeman Bouix, D. (1994). Les fautes de français existent-elles? Paris: Armand Collin. Lüdi, G. (1990). “Les migrants comme minorité linguistique en Europe”. Sociolinguistica, 4, 113–135. Martineau, F. (1985). “L’élision de variable ‘que’ dans le parler d’ Ottawa-Hull”. Cahiers linguistiques d’ Ottawa, 14, 53–70. Mayer, M. (1960). Frog where are you? New York, NY: The Dogal Press. Moreau, M. L. (1986). “Les séquences préformées: entre les combinations libres et les idiomatismes-le cas de la négation avec ou sans ‘ne’”. Le Français Moderne, 137–160. Mougeon, R., & Beniak, E. (1989). “Language contraction and linguistic change: The case of Welland French”. In N. C. Dorian (Ed.), Investigating Obsolescence – Studies in language contraction and death (pp. 287–313). Cambridge: CUP. Mougeon, R., & Beniak, E. (1991). Linguistic Consequences of Language Contact and Restriction. Oxford: Oxford University Press. Myers-Scotton, C. (1993). Social Motivations for Code-Switching. Oxford: Clarendon. Myers-Scotton, C. (2002). Contact Linguistics – Bilingual encounters and grammatical outcomes. Oxford: Oxford University Press. Olshtain, E., & Barzilay, M. (1991). “Lexical retrieval difficulties in adult language attrition”. In H. W. Seliger and R. M. Vago (Eds.), First Language Attrition (pp. 139–150). Cambridge: CUP. Pavlenko, A. (2002). “‘I feel clumsy speaking Russian’: L2 influence on L1 in adult bilingualism”. Plenary speech held at the International Conference on First Language Attrition, Aug. 22–24 2002. Amsterdam: Vrije Universiteit Amsterdam. Pergnier, M. (1989). Les anglicismes. Paris: PUF.
Language contact and attrition
Poplack, S. (1980). “Sometimes I’ll start a sentence in Spanish y termino en Espagnol: Toward a typology of code-Switching”. Linguistics, 18, 581–618. Poplack, S. (1983). “Intergenerational variation in language use and structure in a bilingual context”. In C. Rivera (Ed.), An Ethnographic Sociolinguistic Approach to Language Proficiency Assessment (pp. 42–70). Clevedon: Multilingual Matters. Poplack, S. (1988). “Conséquences linguistiques du contact de langues, un modèle d’ analyse variationniste”. Langages et société, 43, 23–48. Poplack, S., Wheeler, S., & Westwood, A. (1989). “Distinguishing language contact phenomena: Evidence from Finnish-English bilingualism”. In K. Hyltenstam & L. Obler (Eds.), Bilingualism across the Life-Span: Aspects of acquisition, maturity and loss (pp. 132–154). Cambridge: CUP. Posner, R. (1997). Linguistic Change in French. Oxford: Clarendon Press. Sanders, C. (1993). “Sociosituational variation”. In C. Sanders (Ed.), French Today – Language in its social context (pp. 27–53). Cambridge: CUP. Schmid, M.S. (2002). First Language Attrition, Use and Maintenance – The case of German Jews in Anglophone countries. Amsterdam and Philadelphia: John Benjamins. Seliger, H. W. (1985). Primary Language Attrition in the Context of Other Language Loss and Mixing. New York, NY: Depart of Linguistics, City University, mimeo: 1–53. Silva Corvalan, C. (1986). “Bilingualism and language change: The extension of ‘estar’ in Los Angeles Spanish.” Language, 62 (3), 587–608. Waas, M. (1996). Language Attrition Downunder. Berlin: Peter Lang. Walter, H. (1988). Le Français dans tous les sens. Paris: Edition Robert Laffont. Ya˘gmur, K. (1997). First Language Attrition among Turkish Speakers in Sydney. Tilburg: Tilburg University Press.
Is there a natural process of decay? A longitudinal study of language attrition Matthias Hutz Justus-Liebig-Universität Giessen
.
Longitudinal studies in language attrition
It has often been claimed (e.g. Lambert 1989: 15; Soesman 1997: 19) that studies of language loss ideally should be longitudinal in order to determine the precise development of attrition. However, to this date there are only very few longitudinal studies in the field of language attrition. This fact is hardly astonishing since L1 attrition appears to be a very slow process and informants would have to be tested at various points in time, possibly over a period of several decades. However, most research designs cannot afford to allow for such long intervals as a result of financial constraints or mortality effects among the informants (cf. Jaspaert, Kroon, & van Hout 1986; de Bot & Clyne 1994: 21). Such studies also require a high life-expectancy among the linguists investigating the process of language decline. Thus, most studies in language attrition tend to be cross-sectional in nature, measuring the L1 competence of individuals at one particular point of time and implicitly comparing the speaker’s present competence and the competence this person once had (Lambert 1989: 15; Jaspaert & Kroon 1989: 81). However, it is often extremely difficult to establish a clear point of reference for attrition (de Bot & Clyne 1994: 19) just as it is to reconstruct the original proficiency of a language user before the process of language loss sets in (e.g. Pan & Berko-Gleason 1986: 198; Fase, Jaspaert, & Kroon 1992: 8). Since it is very difficult to incorporate the dimension of time, it is usually necessary – as Clyne has pointed out (2001: 143) – to employ surrogates to compensate for the lack of information about the individual’s former skills, such as cross-sectional or inter-generational comparisons, data from a control group in the country
Matthias Hutz
of origin or a speaker’s retrospective self-evaluation. A further methodological problem might be that the amount of time that has elapsed since emigration often varies among the individuals within the sample (Lambert 1989: 15). Apart from the problem of making tacit assumptions about a language user’s former competence an additional methodological problem lies in the fact that cross-sectional research can just present information about the linguistic skills at one particular point in time, but does not provide an opportunity to study the development of attrition. Longitudinal research designs, on the other hand, have the general advantage of allowing assessments of the development of language attrition by comparing specific assessments of the proficiency of individuals at different moments in time resulting in a more reliable measurement of the process of attrition (Jaspaert & Kroon 1989: 82). In particular, a longitudinal research design may serve to study changes in an individual’s L1 proficiency by examining – – –
changing degrees of attrition in specific linguistic domains (e.g. lexical, morphological or syntactic phenomena), the impact of cross-linguistic influence at various moments in time (e.g. the amount of lexical borrowing or transfer of syntactic structures), intralinguistic developments at various intervals as a result of prolonged periods of non-use (e.g. simplifications, generalisations of structures or the use of communication strategies).
In addition, longitudinal studies allow us to make more objective assumptions about the control of specific structures prior to the onset of attrition by providing us with a reference point with which to compare the speaker’s present competence. Thus, it becomes easier to establish whether the speaker was fully competent in the language. It is evident, however, that longitudinal studies also have several drawbacks apart from the obvious ones, such as mortality effects and financial restrictions. As emigrants are often separated from the natural linguistic evolution of their country of origin for a long period of time, some of the differences between them and control groups in the country of origin may not be the result of attrition after all, but rather the result of linguistic change within the L1 community (cf. Jaspaert & Kroon 1989: 80). A further problem relates to the selection of intervals for the longitudinal measurements. Attrition may not be a process linear in time, i.e. language proficiency may deteriorate or improve based on factors such as maintenance ef-
Is there a natural process of decay?
forts, age or language contact. Points of reference which are randomly selected may therefore not necessarily reflect the true process of attrition. Despite these constraints longitudinal studies can help us to gain insight into the process of language attrition and provide us with an objective measure to establish linguistic competence prior to the onset of attrition and to evaluate control of specific structures.
. The factor ‘time’ and attrition Extralinguistic factors often reported to be crucial predictor variables concerning language attrition include factors such as age, education, time since onset of attrition, identity, amount of L1 contact as well as attitudinal and motivational factors. Longitudinal studies which incorporate several points of reference allow for an investigation of the relationship between the onset of attrition and the degree of attrition. At first glance it may seem obvious that there is an important link between the time that an immigrant has been exposed to another language and the degree of erosion in the L1, eventually leading even to a reversal of language dominance. However, the results of research into the effect of time on attrition have been inconclusive so far. Some studies in the context of emigration have shown that the time elapsed since emigration can have a strong influence on language competence (e.g. Waas 1996; Soesman 1997) while others indicated that this factor may only have a limited effect (e.g. de Bot & Clyne 1994; de Bot, Gommans, & Rossing 1991; Jaspaert & Kroon 1992; Giesbers 1997). In her study on language attrition in Dutch adult immigrants in Israel, Soesman (1997) found that increased loss had occurred after the first 5–13 years of emigration. The study also demonstrated that immigrants who had emigrated a relatively short time ago had shorter response times than those who had done so a long time ago. In their longitudinal study of language attrition in Dutch migrants in Australia, de Bot and Clyne (1994) tested their informants, who had already lived in Australia for a longer period of time, in 1970 and re-tested them in 1987. They found no clear evidence of additional attrition over this 17-year period. The emigrants were reported to still have a very good command of Dutch even 15 to 20 years after their arrival in Australia, leading to the conclusion that language skills may deteriorate in the first decade, but then remain fairly stable after this period.
Matthias Hutz
Schmid (2002) also found a high degree of stability in the L1 system of her informants demonstrating that it is, in fact, possible to maintain a language over an extremely long period of emigration. Her informants, German Jews, had lived in the country of emigration for a minimum of sixty years and had in many cases very limited contact with their native language. Schmid showed that the morphological competence and syntax remained largely unaffected even after a prolonged period of restricted contact with the L1. Jaspaert and Kroon (1992) studied attrition in an 83-year-old Dutch emigrant who had lived in the United States for over sixty years. They analysed the main verbs he used on the basis of 29 letters he wrote to the Netherlands over a period of seven years (1981–1987). They conclude that linguistic erosion, at least at the lexical level, may not be as important as is sometimes suggested, since only 5 per cent of the verbs appeared to be markedly affected by loss. It seems that the variance in the amount of attrition found in these studies is probably due to several factors, including methodological differences (e.g. free vs. elicited data) as well as great variation in the time elapsed since emigration and in the amount of contact the informants still had with other L1 speakers. For those immigrants who manage to maintain a high degree of contact with other L1 speakers in their new environment the time factor may become less relevant (de Bot, Gommans, & Rossing 1991). However, there is a linear relation between ‘time’ and attrition when there are few contacts with the L1 (de Bot, Gommans, & Rossing 1991: 94). A general problem with longitudinal studies so far is that they cover relatively short time spans, but the selection of points of reference is decisive for the interpretation of the amount of language loss (Jaspaert & Kroon 1989: 81). Longitudinal studies stretching over several decades may indicate different developmental patterns from those covering just a few years, especially with regard to changing degrees of attrition in specific linguistic domains.
. A natural process of decay? De Bot (2001: 68) raised the question of whether there is a natural process of linguistic decay, comparing the process of losing a language to the gradual process of deterioration in an abandoned house. At first there may just be a few ‘leaks’ in the competence of a native speaker which are hardly visible, but, as time passes by, more and more of the fundamental structures of the language are affected.
Is there a natural process of decay?
An interesting issue in this context is the question which of those linguistic structures are more susceptible to loss and which linguistic structures show more resistance as a result of a continuing language contact situation (cf. Andersen 1982: 91). Are there some units of discourse that enjoy a particularly invulnerable status when attrition occurs? It has been generally argued that language attrition will first affect the level of the lexicon, and only later does it manifest itself in morphological and syntactic categories as well (e.g. Andersen 1982: 114; Romaine 1995: 64; Schmid 2002: 31; Sharwood Smith & van Buren 1991: 19), since lexical features appear to be less well integrated in the structural system of a language. It has also been assumed that most L2 transfer will be found in the lexicon as a result of lexical adaptation to the L2 conceptual system, whereas the transfer of morphological or syntactic categories seems less likely (Giesbers 1997; Haugen 1953; Romaine 1995: 64). The objective of the following longitudinal case study is to analyse the process of attrition over a long period of time based on written material, in this case personal letters. In particular, the study aims at investigating the relationship between the factor ‘time elapsed since emigration’ and the extent of attrition found in specific linguistic domains based on the numbers of interferences found in these domains. The main focus will be on the question which parts of the linguistic system tend to be more resistant to attrition over a long period and which categories show less resistance to the intrusion of L2 elements. For this purpose, lexical borrowing from L2, semantic transfer, as well as morphological and syntactic interferences will be analysed.
. Set-up of the investigation The case study is based upon a corpus consisting of 75 letters written by a German emigrant to members of her family in Germany between 1939 and 1994 (see Table 1). In 1937, at the age of 33, she emigrated to the US without any knowledge of the English language. 57 years later, at the age of 90, she refers to writing letters in German as being a very strenuous process. While most of the letters are from the late 1970s and the 1980s when the process of language attrition was well advanced, there are a few letters written relatively shortly after she had left Germany which very closely reflect her actual competence in German. There are also several letters from the 1950s which have been saved. Unfortunately, however, there are no letters from the 1960s, thus creating a considerable, but inevitable, gap in the following analysis.
Matthias Hutz
Table 1. Corpus Interval
Period
Number of letters
Number of words
1 2 3 4 5 6
1939 1949–1956 1974–1978 1979–1983 1984–1988 1989–1994
2 5 18 18 20 12
1,709 3,016 12,852 9,189 13,590 5,925
75
46,281
Total
Six intervals were chosen for the analysis based upon the availability of letters. Starting in 1974, five-year-intervals were selected as points of reference for establishing the extent of language attrition based on the consideration that attrition is a very slow and gradual process. However, since there are periods where no data are available (from 1940 to 1948 and from 1957 to 1973) and since the data are not equally distributed, the analysis can only give a rough indication of the extent and development of attrition. Thus the results have to be interpreted very cautiously in terms of linearity. There are two general constraints which have to be taken into account concerning the discussion of the results. First, the corpus consists entirely of written material. In writing, there is obviously much more focus on form and accuracy and therefore a higher degree of monitoring compared to the spoken channel (cf. Hughes 1996; Jaspaert & Kroon 1992: 140). The letters were apparently all proof-read by the sender and, if necessary, corrections were made. In other words, the corpus reflects a true picture of the speaker’s linguistic competence at various stages, but it does not necessarily reflect her competence in speaking. Judging from some remarks in her letters, it is quite clear that she becomes increasingly aware of her linguistic problems. Second, when she wrote her last letters she was almost 90 years old and obviously some of the mistakes she made were due to fatigue or lack of concentration. When analysing data from elderly people, we cannot rule out the possibility that specific retrieval problems might be caused by age-related processing difficulties. However, a clear demarcation between language loss as a result of non-use and language loss as a result of age is doomed to fail (de Bot & Weltens 1991: 41).
Is there a natural process of decay?
. Analysis For the purpose of this study, interferences, i.e. deviations from an assumed norm (Weinreich 1953: 1; Schmid 2002: 79), were categorised and analysed. The analysis included interferences in the lexical domain, semantic transfers as well as morphological and syntactic interferences. The interferences found in the corpus were then related to the time elapsed since emigration. . Lexical borrowing When speakers in language contact situations find that their lexical repertoire is inadequate to convey certain concepts or notions, they often resort to lexical borrowing or to lexical innovations (e.g. Brons-Albert 1994). Andersen (1982: 106) hypothesised that a language attriter will not only borrow lexical items from another language and adapt the lexical item to the phonological, lexical, semantic, and morphosyntactic structure of the borrowing language, but will also be likely to create new lexical items. The analysis of lexical borrowing is based on Haugen’s distinction (1953) between necessary loanwords which fill lexical gaps and unnecessary ones, which are essentially gratuitous because of the existence of an equivalent item. For the purpose of this study, a third category was added, ‘strategic loanwords’, i.e. the conscious use of loanwords as a communicative strategy. a. Necessary loanwords Sometimes lexical items from the L2 are used when the speaker feels that there is no equivalent to express the same concept in the L1 (e.g. terms for specific institutions or certain phenomena which are inextricably linked to the other culture). Therefore, those parts in the letters which are closely related to the American way of life are more likely to contain lexical borrowing from English, for instance, work or leisure activities. However, since speakers have little choice but to use these words, this kind of code-switching can hardly be considered attrition. In fact, it may even be regarded as a form of language gain. (1) Die Kinder sind dieses Jahr alle 3 in der Junior High-School. (1954) ‘All three children are in Junior High School this year.’ (2) Die Expressways sind ja so voll. (1977) ‘The expressways are so crowded.’
Matthias Hutz
b. Strategic Borrowing In this case the attriter also uses the L2 word to fill lexical gaps, but here the borrowing might be considered a lexical communication strategy (Odlin 1989: 146) or lexical ‘repair’ (Jaspaert & Kroon 1992: 139). Attriters often experience lexical access problems, i.e. a discrepancy between what they would like to say, their communicative intention, and what they still know how to say in their L1. In order to bridge the gap between their communicative needs and the limited communicative resources, attriters may make deliberate use of borrowing. In this case the speaker assumes or at least hopes that the listener or reader shares with him the language from which the word is borrowed. Strategic borrowing may also occur when the speaker is aware of the fact that he or she is not familiar with the lexical item, perhaps because the concept or the object simply did not exist at the time of emigration. The fact that borrowing can be an intentional process or a strategy involving some conscious effort becomes clear in the corpus because a specific expression is often used with inverted commas to signal that an English word is being used or there is some lexical indication which makes it clear that the loanword is used strategically (e.g. the loanword is introduced by “so-called” or a phrase like “here people call it”). (3) Als “Senior Citizen” kostet es mich nichts. (1977) ‘As a senior citizen it doesn’t cost me anything.’ (4) Ich benutze jetzt einen “Walker”. (1993) ‘I’m using a walker now.’
c. Transference of single lexical items/code-switching In this case lexical items are borrowed despite the fact that there is a corresponding L1 item, but the attriter uses the L2 item subconsciously either because he cannot access the L1 item or he is not familiar with it (e.g. technological innovations) or because there are structural or phonological similarities (e.g. cognates). In this case a single constituent is inserted into a frame provided for by the matrix language (Muysken 1994: 14). This kind of code-switching occurs, of course, frequently in speech, but it is very interesting that it may also occur in written texts where a speaker has more opportunities to monitor his or her language: (5) die idea der jungen Leute (1982) ‘the idea of the young people’ (6) Das ist 70 Meilen outside of Denver. (1994) ‘That’s 70 miles outside of Denver.’
Is there a natural process of decay? 12
Frequency per 1000 words
10 8 6 4 2 0 1939
1949–1956
Necessary loanwords
1974–1978
1979–1983
1984–1988 1989–1994
Strategic loanwords
Code-switching
Figure 1. Lexical borrowings
From the data analysed (see Figure 1) it becomes clear that there is a great deal of inconsistency in the frequency of necessary loanwords since the use of those lexical items depends highly on the specific context and topic, i.e. their occurrence is relatively independent of the linguistic competence of a speaker. As far as the strategic loanwords are concerned, an increase in frequency can be observed in the first 10–15 years after emigration, followed by a decrease in the mid 1970s. This was a period characterised by an increasing amount of contact between the writer and the L1 culture which manifests itself not only in a high number of letters during this period, but also by a short visit to Germany. Approximately 50 years after emigration there is an increase in loanwords again reflecting a widening gap between the writer’s communicative intentions and her proficiency. More and more often she has to resort to the strategic use of English loanwords to convey particular concepts. This tendency is even more evident in the case of code-switching where a steady rise in English items can be observed culminating in the final period from 1989 to 1994. The total number of lexical borrowings (see Figure 2) increases gradually as time elapses since emigration, interrupted only by a period approximately 37–45 years after emigration during which the number of loanwords decreases again. Since this was a period of frequent L1 contact, this would suggest that
Matthias Hutz
‘amount of contact’ is another key factor in the process of lexical attrition. However, after this period the number of lexical borrowings increases again significantly, indicating that the retrievability of lexical items becomes increasingly difficult as a result of strong L2 impact and a lack of contact with L1 speakers. . Semantic transfer The second domain to be examined was the area of loanshifts (Haugen 1953) or semantic transfer (de Bot & Clyne 1994: 20), i.e. the overextension of word meanings. In this type of interference, the meaning of a word from German is extended so that it corresponds to that of English, which leads to semantic overgeneralisations. (7) Er *geht mit dem Flugzeug zurück. (1984) ‘He goes back by plane.’
In this case the meaning of the German verb gehen (‘to walk’) is extended by analogy to the English verb to cover the concept of ‘flying’ as well. A special kind of transfer occurs when the meaning of an English lexical item which is phonologically similar and which has a partial semantic identity is transferred to a German word. Such a “homophonous borrowing” (cf. Schmid 2002: 80) or “false friend” can be found in the following two examples from the corpus: (8) Ich bin *gültig. (1985) ‘I am guilty.’ (9) Es war aber ein sehr *schmaler Sturm. (1979) ‘But it was a very small storm.’
In the first example, the German word gültig (‘valid’) adopts an additional meaning influenced by the phonological similarity with the English word ‘guilty’. In the second example, schmal (‘narrow’) is used as a result of its similarity with ‘small’. These examples of meaning extensions show that there is a constant and very complex interaction between L1 and L2 use. In some cases semantic transfer extends to longer constructions, yielding what has been called a loan-translation or a calque, i.e. a morpheme-bymorpheme translation. Calques are errors that reflect very closely a native language structure (Odlin 1989: 37), e.g. literal translations of idiomatic expressions.
Is there a natural process of decay?
(10) *Ich renne kurz an Briefpapier. (1978) (instead of: mein Briefpapier wird knapp) ‘I am running short of letter-paper.’ (11) *Das ist feine mit mir. (1988) (instead of: damit bin ich einverstanden) ‘That’s fine with me.’
The results show that semantic transfer occurs very early in the corpus, although there are relatively few examples in the first years (see Figure 2). The number of semantic overextensions increases dramatically during the first three decades after emigration, reaching a climax in the late 1970s. After that, the semantic transfers decrease slightly, but the total number remains on a very high level. Compared to the other linguistic subsystems, these findings suggest that this domain is indeed very susceptible to language attrition, being strongly affected especially in the first decades after the process of attrition has set in. The dramatic increase of semantic transfers may be explained by the fact that the process of semantic transfer can be viewed as a speaker’s gradual process of adaptation of conceptual systems to that of the L2 speakers around him (cf. Jaspaert & Kroon 1992: 144). The reality constructed by the interaction partners is adopted, which becomes very obvious in those areas where lexical gaps exist. If there are phonological or semantic similarities between the word in L1 and in L2, this process of semantic transfer may even be accelerated creating a merged concept in the speaker’s mind. Additionally, word meanings in general are often subjected to language change. . Morphological and morphosyntactic domain Another linguistic subsystem that may be affected by language attrition is the morphological domain. Some of the features that have been cited in studies are a reduction in allomorphic variation, including, for example, plural marking (Schmid 2002). It has also been hypothesised that the loss of morphological features may eventually result in a move from synthetic structures to more regularised or analytical ones (cf. Schmid 2002: 34). In many respects the German morphosyntactic system appears to be more complex than the English one, thus making a simplification or regularisation of certain structures quite likely in the process of language attrition. Some morphological areas which may be affected by morphological reduction are case marking, plural marking and gender assignment.
Matthias Hutz
a. Case marking In comparison with the relatively simple nature of English case marking, German appears to be very complex in this respect since all the constituents within a noun phrase have to be marked for one of the four cases (nominative, genitive, accusative, and dative). In the following example, the dative would be required instead of the accusative: (12) *Deinen Vater habe ich schon geschrieben. (1984) ‘I wrote to your father already.’
b. Plural marking Plural marking may also be affected by attrition as a result of the discrepancy between the highly regular nature of English plural morphology and the great variety of German plural allomorphs depending on grammatical gender or word ending (Altenberg 1991: 200; Schmid 2002: 113). In addition, determiners and adjectives require plural marking as well. The following example from the corpus, which would require the plural morpheme -en in German, has been obviously influenced by English pluralformation: (13) mit Ausnahme *Bananas (1987) ‘except bananas’
c. Gender assignment A third very complex morphological system in German is gender assignment. The assignment of one of the three grammatical genders (masculine, feminine or neuter) to nouns is not only characterised by a high degree of arbitrariness, but also determines the endings of articles, adjectives or participles. In the following example, the noun would require the feminine gender instead of the neuter: (14) *das Sehkraft (1986) ‘the vision’
The analysis also included other morphosyntactic categories, such as negation, passive voice, verb morphology and the deviant use of tenses. In general, the morphosyntactic domain does not seem to be largely affected by language loss compared to lexical borrowing or semantic transfer (see Figure 2). Most of the features analysed appear to be negligible in terms of frequency – only a few examples occurred in the corpus. It is only in case-marking that there seems to be a certain tendency to simplify the morphological system as a result of English influence – 56 per cent of all morphological interferences
Is there a natural process of decay?
were found in this area. All other features analysed do not seem to be particularly vulnerable to L1 attrition. Misapplications in plural marking accounted for only 7 per cent of the morphological interferences and errors in gender assignment only for 5 per cent. Other categories accounted for the remaining 32 per cent. The results largely confirm Schmid’s findings in her study of the L1 attrition among German Jews who emigrated to Anglophone countries. Schmid (2002) showed that interlanguage effects of German morphology under L2 influence of English are hard to establish since there were only a small number of interferences, for instance, in plural allomorphy and agreement. As far as the development of language attrition in the morphological domain is concerned in the corpus under investigation here, there seems to be a slight increase in interferences in relation to the length of time that has elapsed since emigration. The findings suggest that morphological attrition is a very slow process and that L2 influence does not appear to be very strong in this domain. This may be due to the fact that morphological rules are acquired as a finite set of rules and that the process of morphological acquisition is often completed before emigration. In contrast, the lexicon represents an infinite set, i.e. it is not totally acquired, even if a person is fully proficient. Thus, the lexicon appears to be much more vulnerable to language attrition. . Syntax: Word order From a contrastive perspective, word order may appear to be vulnerable to language attrition, too. The assumption that word order in one’s native language is adapted as a result of cross-linguistic influence seems to be quite wide-spread (Odlin 1989: 85). Waas (1996: 165), for instance, reported that her subjects, German immigrants in Australia, frequently showed word order difficulties and often followed the L2 patterns. Deviations from standard German word order might be expected especially in those areas where German and English syntactic structures differ, for example, in the placement of verbs in subordinate clauses (for a detailed discussion of syntactic interferences see Schmid 2002). In the following example, the finite part of the verb is required to appear in final position in subordinate clauses: (15) . . ., dass E. *wollte mit L. in Briefwechsel treten. (1954) ‘that E. wanted to correspond with L.’
Matthias Hutz 16 14
Frequency in 1000 words
12 10 8 6 4 2 0
1939
1949–1956
1974–1978 1979–1983 1984–1988 1989–1994
Total number of lexical borrowings
Semantic transfer
Morphosyntax
Syntax
Figure 2. Overview of results
Schmid (2002: 168) also found a trend towards the overproduction of L2 English SVO structures in German L1 attriters which suggests that word order patterns in L2 may lead to syntactic overgeneralisations in the L1. Such overgeneralisation may, for instance, occur concerning the order of adverbials, where English requires putting place adverbials before time adverbials, e.g. (16) Sie waren hier einige Tage. (1978) ‘They were here for several days.’
The corpus contained relatively few deviations from basic German syntactic patterns (see Figure 2), possibly as a result of the fact that German is rather flexible in its word order, but possibly also because this feature may be more resistant to loss than other domains. This finding, however, does not rule out the overproduction of particular syntactic features, e.g. placing time adverbials behind place adverbials or a higher frequency of SVO word order in main clauses. With regard to the development of syntactic interferences it can be stated that the number of syntactic deviations increases only very slowly as more time
Is there a natural process of decay?
elapses since emigration. In fact, this increase is barely noticeable compared to the other linguistic subsystems analysed.
. Conclusion The study confirms that among all the linguistic levels analysed it is the lexicon which is first and most severely affected by language attrition, whereas morphological and syntactic structures seem to be more resistant to language loss. The factor ‘time since emigration’ appears to have less influence with regard to the latter linguistic levels. The stability found in these two areas 57 years after emigration is remarkable. Attrition in morphology and syntax proves to be a very slow and gradual process, whereas the lexical domain undergoes a much more rapid change. Resuming de Bot’s analogy between the linguistic decay and the process of deterioration of an abandoned house, we may compare the lexical domain to those parts of a house which are more susceptible to decay, like the roof or the windows – syntax and morphology, on the other hand, can be compared to the more fundamental parts of the house, i.e. the walls and the foundation. Based on the data in this case study, a considerable amount of attrition, especially in the lexicon, seems to occur during the first 10 to 15 years. But the data available also suggests that attrition seems to continue, albeit more slowly, in all linguistic domains during the following decades. It is necessary, however, to differentiate between the linguistic domains. With regard to morphological and syntactic categories it might take much longer than just a decade to observe significant erosion of L1 structures. In addition to this, the amount of contact with L1 seems to be another crucial factor to explain the degree of attrition (cf. de Bot, Gommans & Rossing 1991). A temporary increase in L1 contact (in this case, approximately 40 years after immigration) may slow down the process of attrition and serve to stabilise the L1 system. Several limitations of the data set, however, have to be taken into account when interpreting the results of this longitudinal case study. The higher degree of monitoring involved in writing is very likely to lead to a lower frequency of interferences compared with free spoken discourse. Furthermore, it is not always possible to draw a clear line between age-related loss and loss as a result of non-use. The lack of data for the first years after emigration obviously poses some problems concerning the interpretation of the findings since attrition is often assumed to occur very early after emigration (e.g. de Bot & Clyne
Matthias Hutz
1994: 27). Thus, the results found in this study can give only a rough indication of the development of attrition. For a precise analysis of the degree of linearity involved in this case the ideal data set would require an equal distribution of letters. In order to verify whether lexical attrition does indeed outweigh attrition on morphological and syntactic levels, it appears to be necessary to investigate a greater number of informants over a long period of time and to move from the analysis of interferences to a general performance analysis of L1 attriters. Only a few studies (e.g. Schmid 2002) have attempted to present an overall description of an individual’s competence, including not only interferences between L1 and L2, but also linguistic structures which are still intact. Only empirical data about the correct and incorrect use of L1 allow us to evaluate the overall degree of attrition. Despite these restrictions, longitudinal studies, in general, can help us gain insight into the changing degrees of attrition in specific linguistic domains and into the proficiency of an individual at different moments in time. In this way, they can provide us with a more accurate and holistic picture of the language user’s former competence and the impact of cross-linguistic influence and thus contribute to our understanding of the complex process of attrition.
References Altenberg, E. P. (1991). “Assessing first language vulnerability to attrition.” In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 189–206). Cambridge: CUP. Andersen, R. W. (1982). “Determining the linguistic attributes of language attrition”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 83–118). Rowley, MA: Newbury House. Bot, K. de (2001). “Language use as an interface between sociolinguistic and psycholinguistic processes in language attrition and language shift”. In J. Klatter-Folmer & P. van Avermaet (Eds.), Theories on Maintenance and Loss of Minority Languages (pp. 65–81). Münster: Waxmann. Bot, K. de, & Clyne, M. (1994). “A 16 year longitudinal study of language attrition in Dutch immigrants in Australia”. Journal of Multilingual and Multicultural Development, 15 (1), 17–28. Bot, K. de, & Weltens, B. (1991). “Recapitulation, regression, and language loss”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 31–51). Cambridge: CUP. Bot, K. de, Gommans, P., & Rossing, C. (1991). “L1 loss in an L2 environment: Dutch immigrants in France”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 87–98). Cambridge: CUP.
Is there a natural process of decay?
Brons-Albert, R. (1994). “Interferenzfehler in der Muttersprache von in den Niederlanden lebenden Deutschen”. In B. Spillner (Ed.), Nachbarsprachen in Europa [Kongreßbeiträge zur Jahrestagung der Gesellschaft für Angewandte Linguistik 23] (pp. 96–105). Frankfurt: Peter Lang. Clyne, M. (2001). “Paradigm hopping and other feats of Veldhoven”. In J. Klatter-Folmer & P. van Avermaet (Eds.), Theories on Maintenance and Loss of Minority Languages (pp. 141–151). Münster: Waxmann. Fase, W., Jaspaert, K., & Kroon, S. (1992). “Maintenance and loss of minority languages: Introductory remarks”. In W. Fase, K. Jaspaert, & S. Kroon (Eds.), Maintenance and Loss of Minority Languages (pp. 3–13). Amsterdam and Philadelphia: John Benjamins. Giesbers, H. (1997). “Dutch in Indonesia: Language attrition or language contact?” In J. Klatter-Folmer & S. Kroon (Eds.), Dutch Overseas: Studies in maintenance and loss of Dutch as an immigrant language (pp. 163–180). Tilburg: Tilburg University Press. Haugen, E. (1953). The Norwegian Language in America: A study in bilingual behaviour. Philadelphia, PA: University of Pennsylvania Press. Hughes, R. (1996). English in Speech and Writing: Investigating language and literature. London and New York: Routledge. Jaspaert, K., & Kroon, S. (1992). “From the typewriter of A.L.: A case study in language loss”. In W. Fase, K. Jaspaert, & S. Kroon (Eds.), Maintenance and Loss of Minority Languages (pp. 137–147). Amsterdam and Philadelphia: John Benjamins. Jaspaert, K., & Kroon, S. (1989). “Social determinants of language loss”. I.T.L.: Review of Applied Linguistics, 83–84, 75–98. Jaspaert, K., Kroon, S., & van Hout, R. (1986). “Points of reference in first-language loss research”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 37–49). Dordrecht: Foris. Lambert, R. D. (1989). “Language attrition”. Review of Applied Linguistics, 83–84, 1–18. Major, R. C. (1992). “Losing English as a first language”. The Modern Language Journal, 76, 190–208. Muysken, P. (1994). “The typology of code-switching”. In F. Sierra-Martinez, M. PujolBerche, & H. den Boer (Eds.), Las Lenguas en la Europa Comunitaria (pp. 11–21). Amsterdam: Rodopi. Odlin, T. (1989). Language Transfer: Cross-linguistic influence in language learning. Cambridge: CUP. Pan-Alexander, B., & Berko-Gleason, J. (1986). “The study of language loss: Models and hypotheses for an emerging discipline”. Applied Psycholinguistics, 7, 193–206. Romaine, S. (1995). Bilingualism. Oxford: Blackwell. Schmid, M. S. (2002). First Language Attrition, Use and Maintenance: The case of German Jews in Anglophone countries. Amsterdam and Philadelphia: John Benjamins. Sharwood Smith, M. A., & van Buren, P. (1991). “First language attrition and the parameter setting model.” In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition, (pp. 17– 30). Cambridge: CUP. Soesman, A. (1997). “An experimental study on native language attrition in Dutch adult immigrants in Israel”. In J. Klatter-Folmer & S. Kroon (Eds.), Dutch Overseas: Studies in maintenance and loss of Dutch as an immigrant language (pp. 181–194). Tilburg: Tilburg University Press.
Matthias Hutz
Waas, M. (1996). Language Attrition Downunder: German speakers in Australia. Frankfurt: Peter Lang. Weinreich, U. (1953). Languages in Contact. The Hague: Mouton.
In search of the lost language The case of adopted Koreans in France Valérie A. G. Ventureyra and Christophe Pallier Laboratoire de Sciences Cognitives et Psycholinguistique, EHESS-CNRS, Paris / Unité de Neuroimagerie Cognitive, Inserm U562, SHFJ & IFR49, Orsay
Introduction Studies of language attrition have often focused on language loss and interference in a contact situation and hence involve individuals who at some point could have been considered as ‘bilingual’. Under such circumstances it is difficult to determine the degree of ‘attrition’ due to erosion from disuse and that resulting from interference from another language. These confounding factors complicate the understanding of mechanisms leading to language attrition. Many of these studies are also primarily aimed at evaluating morphosyntactic and lexical modifications in the attrited language, excluding phonology, an aspect of language known to be acquired within the first year of life (Werker & Tees 1983), and therefore quite relevant to the study of L1 attrition. The present study addresses these two issues. Firstly, it considers individuals who, at a relatively early age, were extracted from their native linguistic and cultural environment only to be immersed in a radically different milieu, requiring the acquisition of a new language and rapid adaptation, to the detriment of the native language which had become obsolete. These individuals were then not fluent in both languages, or at least not for very long, thus minimising the interference factor. In this particular case, attrition may result from ‘brain plasticity’: the brain’s capacity to adapt to a changing environment, which is believed to decline with age and brain maturation. The decline in the brain’s ability to adapt and to learn is described by the ‘critical period hypothesis’ as it pertains to
Valérie A. G. Ventureyra and Christophe Pallier
language acquisition (Lenneberg 1967). According to this author, a ‘critical period’, or a time-frame during which the learning of a given cognitive ability is optimal, applies to language and extends from age two years to puberty. Numerous studies ranging from brain-damaged / hemispherectomised children (Basser 1962) to the extraordinary cases of ‘wild-children’ deprived of language until relatively late in development (Curtiss 1977), and congenitally deaf individuals having been exposed to sign language at different ages (Mayberry & Eichen 1991; Newport & Supalla 1987), support the claim of maturational mechanisms involved in L1 acquisition. On the other hand, the maturational hypothesis also suggests that exposure to L1 may leave long-lasting traces in the neural circuits involved in language processing, and thereby prevent total L1 loss. Thus, according to this hypothesis, we should expect to find traces of the Korean language at some linguistic level in our population of Korean adoptees, who were exclusively exposed to Korean during the first 3 to 9 years of life, and subsequently isolated from their L1 during 15 to 30 years. Conversely, the absence of remnants of L1 would indicate an overwhelming plasticity and replacement of L1 (Korean) by L2 (French). Secondly, an important part of this study is devoted to phonology, as it is the first linguistic level acquired by infants, and perhaps one of the last to be lost. Several studies aimed at determining the benefits of exposure to a language in childhood, followed by a long period of disuse and then by relearning several years later, have shown a pronounced advantage in perception and production in the attrited language (Au, Knightly, Jun, & Oh 2002; Oh, Jun, Knightly, & Au 2003; Tees & Werker 1984) relative to unexposed learners. The study by Au et al., however, did not reveal any advantage of early exposure versus late exposure in morphosyntax, suggesting that the mechanisms of language loss, as for acquisition, may be different for diverse linguistic aspects, and that phonology may be particularly resistant to attrition.
Part I: The first behavioural tests and the fMRI study The study by Pallier, Dehaene, Poline, LeBihan, Argenti, Dupoux, & Mehler (2003) involved eight individuals of Korean origin adopted by French families between the ages of 3 and 8 (mean age at adoption: 5.5 years; 2 women and 6 men ranging in age from 20 to 32 years, mean = 26.8). They were recruited with the help of an adoption agency in France. These subjects had not been re-exposed to Korean since their arrival in France, and all claimed to have
In search of the lost language
completely forgotten their native language. Eight native monolingual French speakers (2 women, 6 men ranging in age from 22 to 28 years, mean = 23.5) and unexposed to Asian languages, composed the control group. These subjects participated in three behavioural tests. The first was a “language identification” test: sixty sentences from five different languages (Korean, Japanese, Polish, Swedish, Wolof) were presented one at a time and subjects were required to give a confidence score between 1 and 7 indicating their degree of certainty that the sentence was Korean. The second test was a “word recognition” test: on each of 24 trials, the subject was required to choose one of two spoken Korean words as the translation of a given French word. The third test used a “speech fragment detection”: a sentence was presented followed by a 400 ms speech segment and the subject had to decide whether this fragment appeared in the sentence or not. One-hundred and twenty-eight sentences were presented in four different languages: Korean, French, Japanese and Polish. This task was performed while the subjects were scanned with functional MRI (Magnetic Resonance Imaging). The results from the behavioural tests appear on Figure 1. In the Language Identification task, the Korean adoptees did not recognise the Korean sentences. They tended to confound the Korean sentences with those in Japanese, by giving these higher confidence ratings than the Polish, Swedish and Wolof sentences. The French control subjects’ performance on this task followed the same pattern and did not differ significantly from that of the Korean adoptees. In the Word Recognition task, requiring the identification of the Korean word corresponding to a given French word, the performance of the Koreans (56%) and that of the French subjects (52%) did not differ significantly. In the Fragment Detection task, both the Korean and French subjects showed better performance on the French sentences than on those in the other three languages. Once again no significant differences were observed between the two groups. Two main results emerged from the brain imaging data acquired while subjects listened to sentences and performed the Fragment Detection task. The first result concerned the processing of Korean: no activation specific to Korean was found in the Korean adoptees and thus, there was no detectable difference between the activations elicited by the Polish, Japanese or Korean sentences (see Figure 2). The second result indicated that the brain areas activated by French were comparable in both the Korean adoptees and the French control groups. Activations were mainly found in the left superior temporal sulcus (Wernicke’s area) and in the left inferior frontal gyrus (Broca’s area), as is expected for L1 processing.
Valérie A. G. Ventureyra and Christophe Pallier a. Korean sentences identification
Rating as a Korean sentence
7 Korean subjects French subjects
6 5 4 3 2 1 Korean
Japanese
b. Korean word recognition
Polish Stimuli
Swedish
Wolof
c. Speech segment detection (performed during fMRI) 100
80
% correct detections
% correct recognition
100
60 40 20 0
80 60 40 20 0
K F Group
French
Korean
Japanese
Polish
Stimuli
Figure 1. Performance of the Korean adoptees and native French subjects on three behavioural tests.
In general, the results of this first study seem to imply that French has overridden Korean in the brains of the Korean adoptees. However, fMRI technology has its limits and may not be able to reveal subtle traces of Korean left in the microcircuitry of the language processing areas, and the behavioural techniques used in this study may not have been capable of tapping into subthreshold, implicit knowledge that requires other methodologies. The second part of this study is aimed at addressing these issues and further assessing the degree of L1 attrition in our population of Korean adoptees.
In search of the lost language
Figure 2. Brain activation patterns of Korean adoptees while listening to Korean (left), Japanese (middle) and Polish (right). All languages activated the superior temporal gyri and there is no statistical difference according to the language of stimulation.
Part II: The new behavioural tests Five experiments seeking to reveal any possible traces of Korean were designed and are presented here. These behavioural tests concern phonology and lexical knowledge: they include Phonetic Perception, Phonetic Retraining, Sensitivity to Korean Phonotactics, Word and Number Recognition, and Implicit Memory for Series. These experiments are presently being tested and results are reported here only for the Number Series Recognition Task. All experiments took place on a computer which displayed the task instructions, played the stimuli and recorded the responses. The participants in these studies comprised eighteen Korean adoptees, including some who had participated in the fMRI experiment. The new subjects were either recruited for the first study via the adoption agency or via a cultural association for Korean adoptees, Racines Coréennes, assembling adoptees from France, Switzerland and Belgium. Some of the recruits from this latter source have been re-exposed to Korean, either by formal instruction or by a trip or a stay in Korea, or both. There were two control groups. The first group consisted of twelve native Korean speakers who have been residing in France for several months or years and still regularly speak Korean. They were matched for age and educational level to our group of adoptees. The second control group consisted of twelve native French speakers matched for age and educational level to the Korean adoptees.
Valérie A. G. Ventureyra and Christophe Pallier
Experiment 1: Series of numbers in different languages Number series are learned early in life and are memorised by rehearsal. Moreover, number words are quite frequent in everyday language. Thus, numbers may be expected to be more resistant to attrition than other types of words. The following experiment specifically tests for the recognition of Korean number series among series from several languages. Stimuli: Twelve series of numbers (one in each of ten different languages and two in Korean, which has two different counting systems) of the numbers 1 to 10 spoken by native speakers were either recorded or downloaded from the Internet. Korean has one counting system proper used mainly for counting objects. The other system used in Korean is derived from Chinese (Sino-Korean) and is used for years, months, days, minutes, money and telephone numbers. As both of these systems are essential to daily life, we can safely assume that young children acquire both. The other languages used were chosen on the basis of their unfamiliarity for average French speakers (most series were nonIndo-European). Some Asian languages had to be excluded as some of our Korean adoptees had been in contact with these (Japanese, Vietnamese, Mandarin). The other ten series of numbers were in: Greek, Turkish, Hungarian, Finnish, Hebrew, Farsi, Quechua, Tagalog, Thai and Cantonese. Procedure: Each series was presented once for familiarisation. Three blocks, each containing all series in random order, were then presented to the subjects, who were required to rate each series on a continuum from 1 to 7 indicating their certainty that the series in question was or was not in Korean (1 = sure that series was not Korean, 4 = no idea whether series was Korean and 7 = sure that series was Korean. Intermediate ratings were also accepted). The results of 9 adoptees and 10 French speakers who performed this task are shown in Figures 3 and 4. An analysis of variance with the factors Group (Adoptees vs. French controls) and Languages revealed a significant main effect of language (F(11,187) = 14.7; p<0.00001), and an interaction (F(11,187) = 2.03; p = 0.03). Welch two sample t-tests comparing the groups for each language showed no significant difference on the Korean (t(13.7) = 1.3; p = 0.20) and Sino-Korean series (t(16.6) = 0.7; p = 0.48); however the French group gave significantly higher ratings on the Cantonese (t(15.4) = 2.18; p = 0.04) and Thai series (t(16.2) = 2.9; p = 0.009) than the adoptees group. The groups did not differ significantly on any other language. Within the Korean adoptees group, paired t-tests were used to compare pairs of languages; these tests re-
In search of the lost language 7 Fr Kor 6
5
4
3
Turkish
Thai
Tagalog
Sino-Kor
Quechua
Hungarian
Hebrew
Greek
Finnish
Farsi
Korean
1
Cantonese
2
Figure 3. Mean confidence ratings for Number Series in Different Languages (Experiment 5) for the adoptees and for a control group of native French speakers. (1 = sure that series not Korean, 4 = no idea whether Korean, 7 = sure that series is Korean). Error bars indicate s.e.m.
vealed that the ratings on the Korean, Sino-Korean, Quechua, Thai, Cantonese and Farsi series did not differ at the p<.05 significance level. Though the adoptees attributed the highest score to the Korean number series, there is no explicit recognition of the number series by this group (see Figure 4). Yet, the difference between the two groups on the Cantonese and Thai number series suggests that the representation of Korean is not exactly the same in both groups. Interestingly, Cantonese and Thai are both tone languages, contrary to Korean. Hence, the Korean adoptees may have been more aware that Korean does not use tones than naive French speakers. Thus, these data suggest that the adoptees have a somewhat more precise notion of the sound pattern of Korean than native French speakers.
Valérie A. G. Ventureyra and Christophe Pallier Tagalog
Sino-Kor
Thai
Turkish
7 6 5 4 3 2 1 Greek
Hebrew
Hungarian
Quechua 7 6 5 4 3 2 1
Cantonese
Korean
Farsi
Finnish
7 6 5 4 3 2 1 Fr
Kor
Fr
Kor
Fr
Kor
Fr
Kor
Figure 4. Individual confidence ratings for the recognition of number series (Experiment 5).
Experiment 2: Phonetic discrimination The phonetic/phonemic categories of L1 are acquired in the first years of life. It is also known that adults have difficulties perceiving some non-native contrasts. Korean voiceless stop consonants use contrasts which are particularly difficult for monolingual speakers of French. In this experiment, we test the adoptees’ ability to perceive such contrasts. Stimuli: Twenty-four CVCV pseudowords of Korean were recorded by each of two native speakers of the Seoul dialect of Korean (one female and one male). The first syllable always started with a consonant characteristic of Korean: lenis, fortis or aspirated [k], [p], or a lenis or fortis [s], and was followed by [a], [i] or
In search of the lost language
[u]. The second syllable was constant: ‘ma’. (Examples of pseudowords: pima, kama, suma).
60 40 0
20
% correct resp.
80
100
Procedure: The stimuli were presented through headphones and responses made on keys on the keyboard. A training session with twelve pairs of pseudowords of French occurred first. The first member of the pair was always pronounced by a male speaker and the second by a female speaker. Then, the 138 pairs of Korean pseudowords were presented. As in the training session, the male speaker appeared first followed by the female speaker. The second stimulus of the pair appeared 500 ms after the first. The following pair was presented 1000 ms after the subject’s response. Subjects were required to indicate whether the two words of the pair were the same or different by pressing a given button on the keyboard as quickly and accurately as possible. There were five conditions: Same, and four types of Different, according to the consonant or vowel contrast in question. They were distributed as follows: (1) 48 pairs of Same (P); (2) 48 pairs of Different Vowel (DV); (3) 18 pairs of Different Consonant lenis-fortis contrast (DC1); (4) 12 pairs of Different Consonant lenis-aspirated (DC2) and (5) 12 pairs of Different Consonant fortis-aspirated (DC3).
F.DC1 K.DC1 F.DC2 K.DC2 F.DC3 K.DC3 F.DV K.DV
F.P
K.P
Figure 5. Percentage correct responses in the Phonetic Discrimination task (Experiment 1) in two control groups (native Koreans (K), native French (F) speakers) and the different types of contrasts. (P = Same, DV = different vowel, DC1 = lenis-fortis, DC2 = lenis-aspirated, DC3 = fortis-aspirated)
Valérie A. G. Ventureyra and Christophe Pallier
Results of the two control groups: As shown in Figure 5, both the P and DV categories were easily identified by both native Korean speakers (n = 12) and native French speakers (n = 12). All three DC contrasts were well discriminated by the native Korean subjects and poorly discriminated by the native French speakers with no overlap between groups, thus showing that the DC contrasts are indeed difficult for native French speakers with no previous exposure to Korean. The DC1 contrast (lenis-fortis) is apparently the most difficult for the French subjects. Experiment 3: Phonetic retraining Assuming that the results of the previous experiment show that the adoptees are not better than native French speakers at perceiving the Korean contrasts, it may still be possible that the adoptees will learn such contrasts faster than the native French speakers with an intensive phonetic training program. Stimuli: Four minimal triplets of Korean words and four minimal triplets of Korean pseudowords, all commencing with either a lenis, fortis or aspirated [p], were chosen as the stimuli for this experiment. Each of the 8 words/pseudowords was recorded by 6 native speakers of Korean (3 female and 3 male). There were 4 tokens of each stimulus per speaker. Thus, a total of 96 different stimuli were recorded by each speaker (8 words x 3 types x 4 tokens). The vowel following the [p] was variable. Procedure: There are 15 training sessions in total. Each session lasts approximately 30 minutes. The sessions are spread over 3 weeks, totalling 5 sessions per week. During the course of each session subjects are presented with 6 blocks of 64 stimuli each. Each block trains the subject to identify one of two types of [p], presented at the beginning of the block. Hence, only one consonant contrast (lenis-fortis, lenis-aspirated or aspirated-fortis) is presented in a given block in order to facilitate the learning of these different types of [p]. There are thus 3 different blocks per session, each repeated once: a total of six blocks per session. The first six sessions are all single speaker sessions. The following sessions become progressively more difficult by a gradual mixing of stimuli by different speakers: two speakers for Session 7 and all six speakers by Session 15. In each block and session, the subject’s task is the identification of the type of [p] heard, by pressing the corresponding key on the keyboard. Feedback is given after each trial in order to enhance the learning process. Three tests of generalisation to other words using the Korean [p]’s as well as to other consonants sharing the same characteristics are administered after
In search of the lost language
the 15 training sessions. The first test involves new real words of Korean beginning with the different types of [p] and is composed of two blocks. One block is composed of words spoken by one of the speakers heard in the training sessions and the other is composed of words pronounced by a new speaker. The second test is the Phonetic Discrimination described above and the third involves words and pseudowords of lenis/fortis/aspirated [t].
Experiment 4: Sensitivity to phonotactics Consonant clusters (e.g. pr, sl. . .) such as those existing in English or French do not exist in Korean or in Japanese. It has been demonstrated that Japanese speakers perceive an illusory vowel within foreign consonant clusters (Dupoux, Kakehi, Hirose, Pallier, & Mehler 1999). This phenomenon is known as vowel epenthesis. It is due to the fact that the human speech decoding system uses the phonotactic constraints imposed by L1 to decode utterances. The properties of Korean phonology lead us to expect that native speakers of Korean will perceive an illusory vowel ([Y]) within consonant clusters of French words. If such is the case, it would be interesting to see whether the adoptees perceive French consonant clusters like native Koreans or like native French speakers. Stimuli: Four pseudowords of Korean were recorded by two native Korean speakers (one female and one male). Two pseudowords involving a change in vowel (ibaki/ibuki) are used in the control condition and two pseudowords presenting a consonant cluster and the epenthetic vowel (asta/asYta) are used in the experimental condition. Each word was modified giving rise to six tokens varying in pitch and speed. Procedure: Each condition is composed of a training session for familiarisation with each stimulus, presented one at a time, and its corresponding button on the keyboard. Following the training session, series of four consecutive stimuli are presented aurally. There are twenty-four such series in each condition. The subject’s task is the identification of the sequence heard during each trial (ex. asta, asYta, asYta, asta). Feedback is given after each trial in order to encourage learning. Experiment 5: Word recognition The previous study (Pallier et al. 2003) implemented a word recognition test in which the participants had to choose the translation of a given French word among pairs of spoken Korean words. The participants performed at chance level, suggesting a radical loss of lexical knowledge. To test this hypothesis
Valérie A. G. Ventureyra and Christophe Pallier
further, we implemented a new test with more trials and words taken from a corpus of Korean children’s productions (see Appendix). In this new task the subject was presented with one Korean word per trial and had to choose the correct translation among two French words.
Stimuli: Sixty common Korean words, thought to be known by children (52 of which were taken from a corpus compiled by Choi & Gopnik 1995) and recorded by a native speaker of Korean, were presented aurally. One-hundred and twenty common French words, sixty of which were the translations of the Korean words and sixty of which were matched for written lexical frequency (New, Pallier, Ferrand, & Matos 2001) were presented visually on the computer screen. Procedure: In a given trial, subjects first heard a Korean word twice and then were presented with two French words simultaneously on the screen. The task was a forced choice recognition whereby the subjects indicated which of the two French words was the translation of the Korean word. Experiment 6: Implicit memory for series of days of the week, months and numbers Experiment 1 and 5 used recognition tests which assess explicit memory for Korean words. We tried to devise a test that would tap into implicit memory for series learned in early childhood: days of the week, months and numbers. The test is based on the “Mere Exposure Effect” which reveals a preference for familiar items vs. unfamiliar items (Kunst-Wilson & Zajonc 1980). Stimuli: Recordings of days of the week in Korean were made. Each day was isolated and later combined with two other days. Series of three days were created. Half of the series were regular (Monday, Tuesday, Wednesday) and the other half were irregular (Monday, Thursday, Sunday). The same procedure was used for months of the year and numbers in Korean. Procedure: In each experiment, (days, months and numbers each gave rise to one experiment), a regular series was matched with an irregular series. On a given trial, the subject hears a pair of regular-irregular series and indicates which one he/she prefers by pressing on a button on the keyboard. (The regular series appear first on half of the trials and last on the other half.) There are 14 trials for days, 24 trials for months and 14 trials for numbers.
In search of the lost language
Conclusion The study of Korean adoptees having been completely cut-off from their native language and immersed in an L2 environment relatively early in life presents an interesting case to the field of language attrition. A possible mechanism for language attrition due to erosion of an unused language is brain plasticity, and is supported by the experiments described in the first part of this paper. However, fMRI technology has its shortcomings and thus, the behavioural tests presented in the second part of this article, and presently being tested, are aimed at a more detailed assessment of L1 attrition in our population of Korean adoptees. So far, the data obtained on the recognition of the number series suggest that the adoptees have a somewhat more precise notion of the sound pattern of Korean than the native French, but no explicit access to knowledge of Korean lexical items.
References Au, T. K., Knightly, L. M., Jun, S-A., & Oh, J. S. (2002). “Overhearing a language during childhood”. Psychological Science, 13 (3), 238–243. Basser, L. S. (1962). “Hemiplegia of early onset and the faculty of speech with special reference to the effects of hemispherectomy”. Brain, 75, 427–460. Choi, S., & Gopnik, A. (1995). “Early acquisition of verbs in Korean: A cross-linguistic study”. Journal of Child Language, 22, 497–529. Curtiss, S. (1977). Genie: A psycholinguistic study of a modern-day ‘wild child’. New York, NY: Academic Press. Dupoux, E., Kakehi, K., Hirose, Y., Pallier, C., & Mehler, J. (1999). “Epenthetic vowels in Japanese: A perceptual illusion”. Journal of Experimental Psychology: Human Perception and Performance, 5, 1568–1578. Kunst-Wilson, W. R., & Zajonc R. B. (1980). “Affective discrimination of stimuli that cannot be recognized”. Science, 207, 557–558. Lenneberg, E. H. (1967). Biological Foundations of Language. New York, NY: John Wiley. Mayberry, R., & Eichen, E. B. (1991). “The long-lasting advantage of learning sign language in childhood: Another look at the critical period for language acquisition”. Journal of Memory and Language, 30, 486–512. New, B., Pallier, C., Ferrand, L., & Matos, R. (2001). “Une base de données lexicales du français contemporain sur internet: Lexique”. L’Année psychologique, 101, 447–462. Newport, E. L., & Supalla, T. (1987). A Critical Period Effect in the Acquisition of a Primary Language. Unpublished ms. Oh, J., Jun, S-A., Knightly, L., & Au, T. K. (2003). “Holding on to childhood language memory”. Cognition, 86, 53–64.
Valérie A. G. Ventureyra and Christophe Pallier
Pallier, C., Dehaene, S., Poline, J-B., LeBihan, D., Argenti, A-M., Dupoux, E., & Mehler, J. (2003). “Brain imaging of language plasticity in adopted adults: Can a second language replace the first?” Cerebral Cortex, 13, 155–161. Tees, R. C., & Werker, J. F. (1984). “Perceptual flexibility: Maintenance or recovery of the ability to discriminate non-native speech sounds”. Canadian Journal of Psychology, 38, 579–590. Werker, J. K., & Tees, R. C. (1983). “Developmental changes across childhood in the perception of non-native speech sounds.” Canadian Journal of Psychology, 37, 278–286.
Appendix List 1 French translations of the words that were aurally presented in Korean in the Word Recognition Task (Exp. 4) List 2 French words matched in written frequency to the words in List 1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
List 1
List 2
chapeau (hat) lumière (light) eau (water) oeil (eye ) train (train) poisson (fish) médicament (medication) lait (milk) baguettes (chopsticks) biscuit (cookie) c’est bon (it is good) c’est petit (it is small) c’est piquant (it is spicy) c’est lourd (it is heavy) c’est pareil (it is the same) ouvre la porte! (open the door! ) réveille-toi! (wake-up!) descend! (come down!) va-t’en! (go away!) sors! (go out!) j’ai soif (I am thirsty) j’ai faim (I am hungry) école (school) livre (book) oreille (ear) couteau (knife)
fauteuil (couch) matin (morning) jour (day) jeu (game) cheveux (hair) rideau (curtain) fourmi (ant) doigt (finger) chaussettes (socks) sauterelle (grasshopper) c’est nouveau (it is new) c’est vrai (it is true) c’est gelé (it is frozen) c’est triste (it is sad) c’est plat (it is flat) lave tes mains (wash your hands!) emporte-le! (take it!) tourne! (turn around!) laisse-le! (leave it!) devine! (guess!) je suis fier (I am proud) je suis tombé (I fell) famille (family) argent (money) sac (bag) bain (bath)
In search of the lost language
27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
chaussures (shoes) clef (key) nez (nose) horloge (clock) balle (ball) bonbons (candy) c’est bruyant (it is noisy) c’est beau (it is beautiful) c’est beaucoup (it is a lot) c’est n’est pas bon (it is not good) c’est peu (it is not a lot) regarde ici ! (look here!) habille-toi (get dressed!) ferme la porte! (close the door!) range! (put your things away!) pousse! (push!) je joue (I play) j’ai peur (I am scared) je ne veux pas! (I do not want to!) je frappe (I hit) je ne sais pas (I do not know) jette! (throw!) donne! (give!) j’ai mal! (I am hurting!) je bois (I drink) lève-toi! (get up!) viens! (come!) assieds-toi! (sit down!) c’est grand (it is big) jamais (never) toujours (always) j’ai froid (I am cold) j’ai chaud (I am hot) c’est rouge (it is red)
parc (park) fruit (fruit) mur (wall) bol (bowl) manche (sleeve) puzzles (puzzles) c’est fatigant (it is tiring) c’est noir (it is black) je suis seul (I am alone) c’est le premier (it is the first) encore (again) tu peux! (you can!) coiffe-toi (comb your hair!) calme-toi! (calm down!) on va te punir! (we will punish you!) prends-le! (take it!) je veux entrer (I want to come inside) je veux le prix (I want the prize) je l’ai pris! (I took it!) je souris (I smile) je le sens (I feel it) essaie! (try!) silence! (quiet!) je suis le premier! (I am the first!) je compte (I count) tire! (pull!) tiens! (hold!) avale! (swallow!) quelques uns (some) moins (less) et alors (and so) j’ai compris (I understood) c’est rapide (it is quick) c’est plein (it is full)
P III
How the study of attrition can contribute to the understanding of language
Attrition in L1 competence The case of Turkish Ay¸se Gürel Bo˘gaziçi University
Introduction This paper investigates the impact of a dominant L2 English as a possible cause of loss or restructuring in the L1 Turkish grammar of adults who have lived in the L2 setting for an extended period of time. Some methodological issues such as participant selection criteria (e.g. length of stay in an L2 setting, frequency of L1 use and L2 proficiency) and the need for the analysis of individual results along with group results will become relevant in the discussion of the present data.
.
Background
. L1 attrition in L2 settings The scope of L1 attrition in L2 settings is rather wide (see Köpke and Schmid, this volume). However, the present study is limited to L1 attrition in adults. I am interested in individual data from first-generation immigrants, who are assumed to have had a fully developed, mature native grammar, before they first came to an L2 setting. I believe that data from this population will enable us to gain insight into the question of restructuring of the mature L1 grammar under L2 exposure. From this standpoint, the following definition of L1 attrition is adopted: L1 attrition is the restructuring of and incorporation of L2 elements/rules into the
Ay¸se Gürel
L1 grammar as reflected in a speaker’s acceptance of syntactically deviant L1 sentences under the influence of L2 rules and constraints (Pavlenko 2000: 179). . Competence change in language attrition Language attrition can manifest itself at different levels of language competence and performance such as lack of fluency and inability or difficulty in retrieving items from the lexicon, deviation from native pronunciation, and also divergence from native syntax (e.g. Seliger 1996; Sharwood-Smith 1989). It is this last point that I concentrate on in this investigation, i.e. whether or not some aspects of L1 syntactic competence will undergo change (i.e. restructuring) under extensive L2 input. In order to examine this, I focus on transfer effects in the domain of Binding Principles (Chomsky 1981, 1982) in general and binding of overt and null pronominals in particular, because, they represent Universal Grammar (UG)-governed, unconscious knowledge that constitutes an important part of native competence (Thomas 1993; White 1989). Any divergence from native norms in this aspect of grammar would then suggest alterability of the L1 grammar at the syntactic competence level. Another motivation for concentrating on Binding Principles is that certain binding properties are subject to language-specific constraints (cf. Manzini & Wexler 1987), and this is conducive to the study of cross-linguistic transfer effects in language attrition (see McCormack, this volume for an L2 attrition study designed to test a similar phenomenon within the same theoretical approach). . Syntactic property under investigation The specific aim of the present study is to examine possible attrition effects in binding conditions of overt and null pronouns in pro-drop L1 Turkish under the influence of non-pro-drop L2 English. The syntactic focus is Binding Principle B. The theory of Binding formulated in Chomsky (1986) is adopted in this investigation: Binding Principle B: A pronominal is free (not bound) in its governing category (GC). Binding is A-binding, i.e., the c-commanding antecedent of the bound element is in an argument (A) position: a. α is A-bound by β iff α and β are co-indexed, β c-commands α, and β is in an A-position. b. α is A-free iff it is not A-bound.
Attrition in L1 competence
In this version of Binding Theory, the GC is defined as: The GC for a pronoun or an anaphor α is the minimal complete functional complex (CFC) that contains α and a governor of α and in which α s binding condition could, in principle, be satisfied. (Chomsky & Lasnik 1995: 102)
.. Binding of pronouns ... English (1) Briani believes that [hei/j is intelligent] (2) Briani ate [hisi/j cake] (3) Nobodyi believes that [hei/j is intelligent]
As illustrated above, pronouns may take referential (1)–(2) or quantified antecedents (3). In the former case, the pronoun is used in coreference with its referential antecedent (Brian), in the latter, it is interpreted as a variable bound by the quantified subject, Determiner Phrase (DP) (nobody). In both cases, pronominals are subject to Principle B, which requires pronouns be free in their GC. The GC for the pronoun in (1 and 3) is the embedded clause. Thus, the overt embedded subject he can be coreferential (1) or be bound (3) by the matrix subject that is outside this domain. Similarly, in (2), the possessive pronoun his can be coreferential with the matrix subject Brian.1 In these examples, the embedded pronoun can also have a disjoint reference (i.e., pick up a free referent in the discourse). The distinction between bound and disjoint interpretation can be formally explained as follows (consider the example in (3)): Bound variable interpretation: (No x: x is a person) x believes that x is intelligent. Disjoint interpretation: (No x: x is a person) x believes that y is intelligent. ... Turkish Let us now look at binding facts in Turkish. Turkish has two overt pronominal forms: o and kendisi, which can be translated as ‘s/he’ and ‘self ’, respectively. Being a pro-drop language,2 Turkish also has a null pronoun (pro). Examples in (4) and (5) illustrate sentential positions that these pronouns can occupy and binding possibilities they allow.3 In a sentence like (4), the subject position can be occupied by the overt pronoun o, kendisi and pro.4 Similarly, these pronouns
Ay¸se Gürel
can also appear in object positions where the pronoun o strictly obeys Principle B, while pro and kendisi are unconstrained (5). (4) O / kendi-si / pro ev-e git-ti. S/he self-3sg pro house-dat go-pst ‘S/he went home.’ (5) Murati o-nu*i/j kendi-si-nii/j proi/j be˘gen-iyor. Murat s/he-acc self-3sg-acc pro like-prg ‘Murati likes him*i/j / selfi/j / proi/j. ’
The following examples illustrate binding possibilities of overt and null pronouns in embedded subject positions. ol-du˘g-u]-nu dü¸sün-üyor. Murati [o-nun*i/j zeki Murat s/he-gen intelligent be-nom-3sgposs-acc think-prg ol-du˘g-u]-nu b. Murati [kendi-si-nini/j zeki Murat self-3sg-gen intelligent be-nom-3sgposs-acc dü¸sün-üyor. think-prg ol-du˘g-u]-nu dü¸sün-üyor. c. Murati [proi/j zeki Murat pro intelligent be-nom-3sgposs-acc think-prg ‘Murati thinks that he*i/j / selfi/j / proi/j is intelligent.’
(6) a.
ye-di. Murati [o-nun*i/j kek-i]-ni Murat s/he-gen cake-3sgposs-acc eat-pst ye-di. b. Murati [kendi-si-nini/j kek-i]-ni Murat self-3sg-gen cake-3sgposs-acc eat-pst ye-di. c. Murati [proi/j kek-i]-ni Murat pro cake-3sgposs-acc eat-pst ‘Murat ate his*i/j /self ’si/j /proi/j cake.’
(7) a.
ol-du˘g-u]-nu Kimsei [o-nun*i/j zeki Nobody s/he-gen intelligent be-nom-3sgposs-acc dü¸sün-m-üyor. think-neg-prg ol-du˘g-u]-nu b. Kimsei [kendi-si-nini/j zeki Nobody self-3sg-gen intelligent be-nom-3sgposs-acc dü¸sün-m-üyor. think-neg-prg ol-du˘g-u]-nu dü¸sün-m-üyor. c. Kimsei [proi/j zeki Nobody pro intelligent be-nom-3sgposs-acc think-neg-prg ‘Nobodyi thinks (that) s/he*i/j /selfi/j / proi/j is intelligent.’
(8) a.
Attrition in L1 competence
As can be seen from the coindexations above, in Turkish, the overt pronoun o can never be bound by the matrix subject. It only allows a disjoint reading. In contrast, the overt pronoun kendisi and pro can be bound by the matrix subject or have disjoint readings (see b and c examples above). Crucially, these observations hold in both referential (6)–(7) and quantified contexts (8). This observation is important as it suggests that the Overt Pronoun Constraint (OPC), postulated as a universal property of pro-drop languages (Montalbetti 1984), is not exemplified in Turkish (cf. Kanno 1997).5 The examples above demonstrate clear contrasts between Turkish and English pronominal binding. Besides the presence of null pronouns and an overt reflexive pronominal like kendisi in Turkish, the Turkish overt pronoun o does not pattern similarly with its English counterpart s/he in embedded subject positions (compare e.g. (1) to (6)). I suggest that the difference between these pronouns stems from the differences in governing categories in two languages. .. Governing categories in English and Turkish The difference between English and Turkish overt pronoun binding is illustrated once again in the following examples where the embedded subject pronoun can refer to the sentential subject Brian in English but cannot in Turkish: (9) a. Briani said IP [hei/j would come]. söyle-di. b. Briani DP [o-nun*i/j gel-ece˘g-i]-ni Brian he-gen come-nom-acc tell-pst (10) a. Briani ate DP [hisi/j cake]. ye-di]. b. Briani DP [o-nun*i/j kek-i-ni Brian he-gen cake-3sgposs-acc eat-pst
According to the version of Binding Theory adopted here, the embedded inflectional phrase (IP) in (9a) and the bracketed DP in (10a) are GCs in which the pronoun is free. This explains the grammaticality of coindexation between the subject antecedent Brian and he/his in (9a) and (10a). In contrast, the Turkish pronoun o in ((9b) and (10b)) is obligatorily disjoint from the antecedent outside the DP. In order to understand why this would be so, we have to look closely at the structure of complement clauses in Turkish. Most complement clauses are nominalised in Turkish. Nominalised constructions behave exactly like lexical DPs in many respects such as their internal morphology, and case marking (compare (9b) and (10b)) (George & Kornfilt 1981). Thus, unlike English, embedded clauses in Turkish are DPs rather than finite IPs. Embedded nominalised clauses, being like lexical DPs, do not count
Ay¸se Gürel
as GCs in Turkish. In these cases, the GC is the tensed matrix clause that includes the pronoun and a finite I0 . However, in English, besides finite IPs, DPs also constitute a binding domain. Thus, the difference between the binding of English and Turkish overt pronouns in complement clauses (9) or in the possessive DPs (10) is not surprising. I consider this difference to be a consequence of a parametric option permitting DPs as GCs in English but not in Turkish (cf. Manzini & Wexler 1987). Within this background, in this study, I am interested in seeing how much knowledge of the binding properties of overt and null pronouns are maintained in L1 attrition under extensive exposure to the L2 which has no null pronouns and has only one overt pronoun. The research questions are formulated as follows:
. Research questions 1. Does the definition of binding domain in the L1 change under attrition? More specifically, do native Turkish speakers maintain the contrast between the L1 and L2 overt pronouns with respect to their binding properties? 2. Do native Turkish speakers maintain the contrast between the two overt pronominals in L1 Turkish (o and kendisi) or do they reduce the contrast in such a way that both are treated like the L2 overt pronoun with respect to their binding possibilities? 3. Do native Turkish speakers maintain L1 knowledge regarding binding properties of the null pronoun despite extensive exposure from the nonpro-drop L2? Binding properties of overt and null pronouns are integral parts of adult native grammar knowledge (i.e. linguistic competence) and thus it is less likely that they undergo L1 attrition. Therefore, I predict that native Turkish speakers will be able to maintain binding properties of overt and null pronominals even under extensive L2 exposure.
Attrition in L1 competence
. The study . Participants 24 native Turkish speakers who are end-state L2 speakers of English participated in this study. The participants, aged between 29–72 (mean age: 47), had immigrated to North America (Canada or the US) at an adult age (ages between 16–44) (mean age of immigration: 25.5) and had been living in an L2 country for at least 10 years at the time of testing. The length of stay ranged from 10 to 43 years (mean length of stay: 21.5 years). The 10-year-stay in an L2 country was taken as one of the inclusionary criteria because this seems to be a generally accepted baseline reported in attrition studies (e.g. de Bot, Gommans, & Rossing 1991; Jordens, de Bot, & Trapman 1989, Köpke and Schmid, this volume). Following de Bot et al. (1991), length of stay was defined in three levels: 10–19; 20–29; 30 years and above. All participants, with the exception of two, are post-puberty L2 learners.6 Age of first L2 exposure was between 7 and 26 years (mean age of first L2 exposure: 13.2). Participants’ contact with L2 English increased after they moved into the L2 country where some of them had pursued their academic careers in North American universities and some started to work in various Englishspeaking environments. Given these educational and professional backgrounds of participants here, it can readily be assumed that they are all advanced speakers of English. However, in order to ensure that all participants are at a certain English proficiency level, I used a cloze test. According to the results of the cloze test, participants were grouped into two levels: high and high-intermediate. The cloze test was also given to 15 native English speakers to obtain a baseline for comparison. With respect to the frequency of L1 use, some the individuals in this group had English-speaking and some Turkish-speaking partners. This naturally caused some variability among participants with respect to the amount of L1 contact. The difficulty of defining and quantifying the amount of contact with the L1 has already been addressed in other attrition research (e.g. de Bot et al. 1991). Following de Bot et al., frequency of use was defined in two levels: frequent and infrequent. This grouping is based on information gathered from participants’ responses on a questionnaire. It appeared that participants with an English-speaking partner used the L1 Turkish less than those with a Turkish partner. In the area of language attrition, it is important to establish a baseline to which any language change can be compared. To establish this baseline, 30 na-
Ay¸se Gürel
tive Turkish speakers, aged between 20–70 (mean age: 40), were tested. This group was composed of people who have been living in Turkey since birth and who had some knowledge of English. This was necessary as one of the tests involved simple short stories in English. Except for the two English teachers in this group, all of the controls had beginner-level English. The control group was also matched with the experimental group for educational level.7 . Tests The tests included a written interpretation task, a truth-value judgment task and a picture identification-listening task. .. Test 1: Written interpretation task This test was adopted from Kanno (1997). It included 48 items with 24 referential and 24 quantified antecedents, where each category had 12 overt and 12 null pronouns. In this test, participants were given a Turkish sentence like (11) and asked to select possible antecedent(s) (from among the three options given) for the embedded subject pronoun. For example in (11), participants were expected to circle the option (b) (i.e., disjoint reading) as the overt pronoun o cannot be bound by the matrix subject in Turkish. söyle-di. (11) Murati [o-nun*i/j sinema-ya gid-ece˘g-i]-ni Murat s/he-gen cinema-dat go-nom-3sgposs-acc say-pst ‘Murati said (that) [ s/he*i/j would go to the movies].’
Soru (question): Sizce bu cümleye göre kim sinemaya gidecek olabilir? (According to this sentence, who could be the person that would go to the movies?) (a) (b) (c)
Murat Ba¸ska bir ki¸si Hem (a) hem (b)
(some other person) (Both (a) and (b))
Besides these complex clauses, the test also included simple possessive DP constructions such as (12) below: (12) Herkes [elbise-si-nin çok pahalı ol-du˘g-u]-nu Everyone dress-3sgposs-gen very expensive be-nom-3sgposs-acc söyle-di say-pst ‘Everyonei said [proi/j (their/his/her) dress] is expensive.’
Attrition in L1 competence
Soru (question): Sizce bu cümleye göre kimin elbisesi çok pahalı olabilir? (According to this sentence, whose dress could be expensive?) (a) (b) (c)
Herkesin Ba¸ska bir ki¸sinin Hem (a) hem (b)
(Everyone’s) (Some other person’s) (Both (a) and (b))
In the example above, the matrix subject is a quantified DP, the embedded subject is a null pronoun, which can potentially take both ‘everyone’ (bound reading) and ‘some other person’ (disjoint reading) as antecedents. Thus, participants were expected to choose the option (c) here. Testing knowledge of binding in ambiguous contexts such as (12) above is a problem that L2 researchers have recently addressed (e.g. Bruhn-Garavito 1995; White, Bruhn-Garavito, Kawasaki, Pater and Prévost 1997). If participants are given only two options to choose from, they are forced to make a preference between the two possible antecedents. However, accepting one of them does not necessarily mean that participants completely exclude the other option. They might simply be expressing their preference for one over the other. Given this potential problem, I also included another task, a truth-value judgment task that has been developed to deal especially with this particular ambiguity and preference problem. .. Test 2: Truth-value judgment task (Story task) This test was adopted from Dekydtspotter et al. (1997) and White et al. (1997). In this task, participants were given 12 short English stories and asked to judge the following Turkish sentence as true or false according to the context given in the story. All of the sentences were grammatical; their appropriateness depended on the context provided by the stories. The stories were used to make as salient as possible in the minds of participants one of the interpretations (bound or disjoint). The following test item illustrates this point: (13) Story: Mary and Brian went to a restaurant. Mary ordered seafood and Brian ordered a pizza. The bill came to 50 dollars. Brian complained that the bill was high but Mary didn’t agree. Target sentence to be judged: Mary restoran-ı pahalı bul-du˘g-u-nu söyle-di. Mary restaurant-acc expensive find-nom-3sgposs-acc say-pst ‘Maryi said (that) proi/j (s/he) found the restaurant expensive.’
Ay¸se Gürel
In this item, the embedded subject position is occupied by pro. Thus, the sentence is ambiguous as pro has two potential antecedents, Mary and Brian. As mentioned earlier, the null pronoun is likely to have sentential rather than disjoint reference. The story above requires the antecedent to be disjoint. The assumption is that forcing the disjoint reading might override any preference for sentential antecedents. If learners have knowledge that coreferentiality of pro with disjoint reference is possible, they should choose true in this item; in contrast, if they assume that this is not possible; in other words, if they allow the matrix subject as antecedent, they should choose false. In this task, each story appeared three times with a target Turkish sentence following it. Each time, the target sentence included a different pronoun (i.e., the overt pronouns, o, kendisi and the null pronoun). For example, in another test item, the story in (13) appears with the overt pronoun o in the embedded subject position. This time, if participants answer as true, this means that they get the disjoint reading but if they say false, that suggests that they get the bound reading for the overt pronoun o, an option that is not allowed in native Turkish grammar. However, in English, the embedded subject pronoun can be bound by the matrix subject Mary. Thus, any response in that direction might suggest L2 English effects. .. Test 3: Picture identification-listening task This test is also a truth value judgment task and in that sense it is similar to the previous task described above but instead of stories, it involves listening and picture identification. In this test, participants were asked to listen to a series of Turkish sentences, involving pronouns and judge the corresponding colour picture in front of them as true or false according to the sentence they hear. Participants listened to the sentences on a tape-recorder. Each sentence was played only once. Participants responded out loud during this period and their responses were marked by the researcher. This test included 24 sentences (8 different pictures, each repeated three times) in referential contexts only.8 The items included 16 overt and 8 null pronouns. One of the test pictures and the test sentences is given below (the colours are typed in the following illustration): Participants hear: söyle-di. (14) Ahmet Sarıi o-nun*i/j resim çek-ti˘g-i-ni Ahmet yellow s/he-gen picture take-nom-3sgposs-acc say-past ‘Ahmet Sarıi said that s/he*i/j took picture(s).’
Attrition in L1 competence
Mehmet Yesil (Mr. Green)ã
Ahmet SarF (Mr. Yellow)
Figure 1. A picture used in the picture identification-listening task (Test 3)
In comparison to the first two untimed tests, in this task, more ‘on-line’ processing was involved as participants heard target sentences in real time duration and made their judgments in a short period of time.
. Results A one-factor (i.e., group: control, acquisition and attrition) ANOVA was conducted. It should be noted at this point that since this study is part of a larger investigation (Gürel 2002) that includes an L2 acquisition group (n = 28), the total number of subjects on which the original analysis is based is 82.
Ay¸se Gürel
. Summary of group results As can be seen in Table 1, although, in all tests, there was a tendency to have a disjoint reading for the overt pronoun o, the attrition group diverged from the native controls by allowing more bound as well as ambiguous interpretations for o. Planned comparisons conducted within an ANOVA revealed that in the quantified context in Test 1, the attrition group allowed the bound reading for the overt pronoun o at a significantly higher percentage [F(1,79) = 3.98, p = 0.05]. A similar difference was obtained in Test 2 in referential [F(1,79) = 24.13, p < 0.0001] and quantified contexts [F(1,79) = 22.39, p < 0.0001]. The Table 1. Group results for Test 1, 2 and 3 Referential antecedents Overt subjects Null subjects O kendisi pro
Quantified antecedents Overt subjects Null subjects O kendisi pro
TEST 1 Control n = 30 Bound Disjoint Bnd & Dis
1%* 94% 5%
36% 0% 64%
16% 0% 84%
2% 89% 9%
32% 0% 68%
10% 3% 87%
Attrition n = 24 Bound Disjoint Bnd & Dis
1% 83% 15%
81% 3% 16%
64% 2% 34%
6% 82% 12%
63% 7% 30%
43% 14% 43%
TEST 2 Control n = 30 Bound Disjoint
4% 96%
79% 21%
76% 24%
3% 97%
81% 19%
74% 26%
Attrition n = 24 Bound Disjoint
30% 70%
74% 26%
82% 18%
22% 78%
73% 27%
76% 24%
TEST 3 Control n = 30 Bound Disjoint
0% 100%
85% 15%
78% 22%
Attrition n = 24 Bound Disjoint
21% 79%
97% 3%
92% 8%
*Percentages indicate how many times the participants interpreted each pronoun with a particular (e.g., bound or disjoint) interpretation.
Attrition in L1 competence
same results hold in Test 3 in the referential context F(1,79) = 13.72, p < 0.001]. These results suggest that the ‘disjointness’ requirement of the overt pronoun o in Turkish is not strictly obeyed by L1 attriters, possibly due to interference from L2 English. As can be seen in Table 1, none of the groups treated kendisi like o in any of the tasks. Thus, the attrition group was able to preserve the distinction between the two overt pronominals with respect to their binding properties. With respect to pro, the attrition group seemed to have maintained that pro is potentially ambiguous between bound and disjoint readings. In sum, similar to the native controls, for L1 attriters, there are clear interpretational differences between o and pro on the one hand, and between o and kendisi, on the other hand. While the overt pronoun o was most often assigned the disjoint reading, pro and kendisi received mostly bound or ambiguous readings. Recall that participants in the attrition group were grouped according to the number of years they spent in the L2 country. In order to see whether the length of stay factors into the degree of loss in L1 attrition, a separate one-factor ANOVA was conducted. However, results revealed no length of stay main effect in any of the tests for the attrition group. This means that ‘time spent in an L2 country’ was not relevant for the level of performance in pronoun binding. That is, those who stayed in an L2 country longer do not necessarily show more L1 attrition than those who stayed less. This suggests that the length of stay is not an important factor in the L1 attrition process. Probably, as de Bot et al. (1991) suggest, time only becomes relevant when there is very little L1 contact.9 . Individual results It has been suggested that group results may not always reflect the properties of individual grammars in L2 acquisition (Eckman 1994; White et al. 1997). I assume that L1 attrition, like L2 acquisition, is subject to considerable individual variation. Therefore, I also analysed the individual results. In this analysis, individual judgments for the overt pronoun o were examined in order to see the extent and systematicity of L2 transfer in individual grammars (see Table 2).10 As long as similar tendencies are obtained in both group and individual results, we can be confident about the findings. In Test 1, 75% of the participants allowed correct interpretation to the overt pronoun o. In Test 2, this percentage is somewhat lower (67%) and in Test 3, 71% of the attriters responded correctly. If we compare these individual results to those of the controls, we see a difference (compare 90% for the controls to
Ay¸se Gürel
Table 2. Individual results for the overt pronoun o (Test 1, 2 and 3) Test 1 Consist Inconsist.
Test 2 Consist Inconsist.
Test 3 Consist Inconsist.
Control n = 30
27/30 (90%)
3/30 (10%)
30/30 (100%)
0/30 (0%)
30/30 (100%)
0/30 (0%)
Attrition n = 24
18/24 (75%)
6/24 (25%)
16/24 (67%)
8/24 (33%)
17/24 (71%)
7/24 (29%)
Years of stay 1 (10–19 yrs) 2 (20–29 yrs) 3 (30 yrs-) Total
10 6 2 18
3 1 2 6
9 5 2 16
4 2 2 8
11 4 2 17
2 3 2 7
Use of L1 Frequent Infrequent Total
12 6 18
5 1 6
11 5 16
6 2 8
11 6 17
6 1 7
L2 Proficiency High High-Intermediate Total
17 1 18
5 1 6
14 2 16
8 0 8
15 2 17
7 0 7
75% in Test 1, 100% to 67% in Test 2, and again 100% to 71%). This suggests that L1 attriters seem to be different from the native controls at group as well as individual level. With respect to the effects of the length of stay, L1 use and L2 proficiency, we do not see any clear role of the length of stay in the distribution of consistent and inconsistent behaviours. That is, it is not the case that the rate of inconsistent behaviours to consistent behaviours gets smaller as the time spent in an L2 country increases. The use of the L1 seems to have a role here. For example, in Test 1, out a total of 17 frequent L1 users, 12 are found to be consistent, whereas only 5 are inconsistent. This might suggest that the frequent use of the L1 has a ‘blocking effect’ in attrition. However, the amount of inconsistent behaviours is also less than the amount of inconsistent behaviours in the case of infrequent L1 users. Furthermore, the proficiency level in L2 does not seem to have a ‘facilitative effect’ in attrition. For example, in Test 1, out of 22 highly proficient English speakers, 17 made consistent judgments, only 5 were found to be inconsistent. That suggests that the use of L2 English does not necessarily lead to inconsistency in L1 grammar.
Attrition in L1 competence
In sum, it was not possible to clearly determine the source of L1 attrition. A more controlled participant selection and more clear quantification of factors such as the frequency of L1 use (or L1 contact) are necessary. Although the source is not clear, there are obvious effects of L2 transfer that lead to L1 attrition. Individual results are in line with group results in the sense that in both analyses, L1 attriters seem to diverge from the native controls in assigning the correct ‘disjoint’ reading for the overt pronoun o.
. Conclusion Despite the influence of non-pro-drop L2 English, L1 attriters are able to maintain the subtle distinction between the binding properties of overt and null pronouns and between the two overt pronouns in Turkish. However, due to L2 English, the Turkish overt pronoun o is attributed binding properties of the English overt pronoun ‘s/he’, disregarding the language-specific constraints in terms of binding of the overt pronoun o in the subject position. This is a restructuring or reanalysis of L1 Turkish binding options on the model of L2 English. In conclusion, what is documented in this study is restructuring of the L1 grammar at some deep syntactic competence level. In that sense, this suggests that native competence of late L2 acquirers is alterable due to extensive L2 exposure, together with less accessible L1 input. As a final note, this study provides a new set of data from adult L1 attrition in the domain of syntax. I hope that this will bring in new empirical perspective into the study of competence change defined within the generative framework and will also contribute to the understanding of the language attrition phenomenon in general.
Notes . In the rest of the paper, the term ‘bound’ is used instead of ‘coreferential’ for interpretations in referential contexts for the sake of unity with interpretations in quantified contexts. . Turkish has traditionally been analysed as a pro-drop language (e.g. Enç 1986; Kornfilt 1984; Özsoy 1987) and the theoretical assumption on which this study is based follows this traditional view (Gürel 2002). However, there are some recent arguments that Turkish can actually be analysed as a non-pro-drop language (Öztürk 2001).
Ay¸se Gürel . For the sake of clarity, the indices in English translations indicate what is (im)possible in Turkish. . The following abbreviations are used in this paper: acc = Accusative; dat = Dative; gen = Genitive; nom = Nominalisation; neg = Negation; poss = Possessive; prg = Progressive; pst = Past; sg = Singular. . According to the OPC, in pro-drop languages, overt and null pronouns demonstrate different interpretative properties as bound or referential pronouns. The overt pronouns cannot receive a bound variable interpretation in contexts where a null pronoun could occur. . These two participants had their first exposure to English at the age of 7. . Out of 24 people in the attrition group, 17 had a university and 5 had a high school degree. The other 2 people were secondary school graduates. Similarly, the majority of controls had a university degree (25 out of 30). 3 of the other 5 participants were high school and 2 of them were secondary school graduates. . This is due to the fact that illustrating contexts with quantified antecedents (e.g. ‘Everyone says s/he sings well’ or ‘Nobody said he painted the wall’) is very difficult. Even if done, illustrations might not be very clear for the participant who has to judge them in a restricted time period. Therefore, all the actions illustrated in the test were carried out by two characters ‘Mr. Yellow’ and ‘Mr. Green’ (adapted from White et al. 1997). . De Bot et al. (1991) found that the amount of time spent in an L2 setting becomes relevant only when there is not much contact with the L1. Following this, ‘L1 contact’ was also analysed as one of the main effects. However, the ‘length of stay-L1 contact’ interaction could not be analysed as in some cases there were not enough participants on which the analysis could be performed (e.g. there was no one who fell into the ‘level 3’ in terms of length of stay and the ‘infrequent’ level in terms of L1 use). Therefore, an analysis of the ‘frequency of L1 use-length of stay’ interaction effect in conjunction with ‘frequency of L1 use’ main effect could not be done in this study. This was all due to the fact that the main selection criterion in this study was based on ‘length of stay’. Therefore, L1 use could not be controlled for in the initial selection procedure. I would like to note this as one of the factors that future researchers should consider. Yet, given the practical difficulties related to the participant selection/availability, this factor cannot always be controlled for by the researcher. . In this analysis, I wanted to see the consistency of correct ‘disjoint’ responses for the overt pronoun o. In Test 1 and 2, the total number of items involving the overt pronoun o was 12 and in the last test, it was 8. Following Eckman (1994) and White et al. (1997), I took the consistency threshold to be 75%. This corresponds to 9 or more correct responses out of 12 (Test 1 and 2) and 6 or more correct responses out of 8 (Test 3). This means, for example, if a participant responded correctly (i.e. gave ‘disjoint-only’ responses) 9 times out of 12, then s/he was included in the ‘consistent’ category. If, however, the correct number of responses is below 9, s/he was considered in the ‘inconsistent’ category. Then, the percentage of participants that had consistent and inconsistent behaviour with respect to the judgment of the overt pronoun o was calculated.
Attrition in L1 competence
References Bot, K. de, Gommans, P., & Rossing, C. (1991). “L1 loss in an L2 environment: Dutch immigrants in France”. In H. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 31–51). Cambridge: CUP. Bruhn-Garavito, J. (1995). “L2 acquisition of verb complementation and Binding Principle B”. In F. R. Eckman, D. Highland, P. W. Lee, J. Mileham, & R. R. Weber (Eds.), Second Language Acquisition Theory and Pedagogy (pp. 167–188). Mahwah, NJ: Lawrence Erlbaum. Chomsky, N. (1981). Lectures in Government and Binding. Dordrecht: Foris. Chomsky, N. (1982). Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, MA: MIT Press. Chomsky, N. (1986). Knowledge of Language. New York, NY: Praeger. Chomsky, N., & Lasnik, H. (1995). “Principles and parameters theory”. In N. Chomsky (Ed.), The Minimalist Program (pp. 13–127). Cambridge, MA: MIT Press. Dekydtspotter, L., Sprouse, R., & Anderson, B. (1997). “The interpretative interface in L2 acquisition: The process-result distinction in English-French interlanguage grammars”. Language Acquisition, 6 (4), 297–332. Eckman, F. R. (1994). “Local and long-distance anaphora in second language acquisition”. In E. E. Tarone, S. M. Gass, & A. D. Cohen (Eds.), Research Methodology in Second Language Acquisition (pp. 207–225). Hillsdale, NJ: Lawrence Erlbaum. Enç, M. (1986). “Topic switching and pronominal subjects in Turkish”. In D. I. Slobin & K. Zimmer (Eds.), Studies in Turkish Linguistics (pp. 195–208). Amsterdam and Philadelphia: John Benjamins. George, L. M., & Kornfilt, J. (1981). “Finiteness and boundedness in Turkish.” In F. W. Henry (Ed.), Binding and Filtering (pp. 105–127). Cambridge, MA: MIT Press. Gürel, A. (2002)). Linguistic Characteristics of Second Language Acquisition and First Language Attrition: Turkish overt versus null pronouns. Unpublished Doctoral Dissertation, Montreal: McGill University. Jordens, P., Bot, K. de, & Trapman, H. (1989). “Linguistic aspects of regression in German case marking”. Studies in Second Language Acquisition, 11, 179–204. Kanno, K. (1997). “The acquisition of null and overt pronominals in Japanese by English speakers”. Second Language Research, 13 (3), 299–321. Kornfilt, J. (1984). Case Marking, Agreement, and Empty Categories in Turkish. Unpublished Doctoral Dissertation, Cambridge: Harvard University. Manzini, R. M., & Wexler, K. (1987). “Parameters, binding theory, and learnability”. Linguistic Inquiry, 18 (3), 413–444. Montalbetti, M. M. (1984). After Binding. On the interpretation of pronouns. Unpublished Doctoral Dissertation, Cambridge, MA: MIT Press. Özsoy, A. S. (1987). “The null subject parameter and Turkish”. In H. E. Boeschoten & L. T. Verhoeven (Eds.), Studies on Modern Turkish. Proceedings of the Third Conference on Turkish Linguistics (pp. 82–90). Tilburg: Tilburg University Press. Öztürk, B. (2001). “Turkish as a non-pro-drop language”. In E. E. Taylan (Ed.), The Verb in Turkish (pp. 239–259). Amsterdam and Philadelphia: John Benjamins.
Ay¸se Gürel
Pavlenko, A. (2000). “L2 influence on L1 in late bilingualism”. Issues in Applied Linguistics, 11 (2), 175–205. Seliger, H. W. (1996). “Primary language attrition in the context of bilingualism”. In T. J. Bhatia & W. C. Ritchie (Eds.), Handbook of Second Language Acquisition (pp. 605–626). New York, NY: Academic Press. Sharwood-Smith, M. (1989). “Cross-linguistic influence in language loss”. In K. Hyltenstam & L. K. Obler (Eds.), Bilingualism Across the Life Span (pp. 185–201). Cambridge: CUP. Thomas, M. (1993). Knowledge of Reflexives in a Second Language. Amsterdam and Philadelphia: John Benjamins. White, L. (1989). Universal Grammar and Second Language Acquisition. Amsterdam and Philadelphia: John Benjamins. White, L., Bruhn-Garavito, J., Kawasaki, T., Pater, J., & Prévost, P. (1997). “The researcher gave the subject a test about himself: Problems of ambiguity and preference in the investigation of reflexive binding”. Language Learning, 47 (1), 145–172.
Methodological aspects of a generative-based attrition study Bede McCormack Columbia University
Introduction This paper discusses methodological and linguistic issues associated with a generative approach to language attrition research. Although it considers a study of second language attrition, the research methodology used and the choice of linguistic structure to investigate are applicable to first language attrition studies of a generative nature as well. The study described here is based on my 2001 PhD thesis on the loss of principles of reflexive binding in L2 English among L1 Japanese speakers. This was a qualitative, longitudinal, multiple case study which investigated the attrition of the binding principles which govern the relationship between reflexive pronouns such as himself and herself, and their potential antecedents. Since English and Japanese differ in the domains in which reflexives may be bound, with English at one end of the markedness hierarchy first proposed by Wexler and Manzini (1987) and Japanese at the other, the binding patterns which emerge provide an interesting area for attrition research. The results of the study suggest that, despite the apparent loss of control over some aspects of English reflexive binding, the attrited grammar which emerged was neither L1 nor L2-like, yet remained within the bounds of binding patterns allowed by the principles of Government and Binding as proposed by Chomsky (1980a, 1980b, 1981). This suggests, then, that an attrited aspect of language may not only be the result of pressure from the dominant language, but also a result of unconscious, speaker-internal processes. This paper will focus on the following aspects of the study: a. Adopting a theoretical framework to investigate attrition b. Selecting a linguistic feature for analysis
Bede McCormack
c. d. e. f.
Informant selection Test design Presentation of results Discussion of results
.
Adopting a theoretical framework to investigate attrition
As described in the introduction to this volume, attrition research has typically involved two versions of the regression hypothesis, namely the “reverse order”, or “regression” hypothesis which proposes that the order in which attrition occurs is opposite to the order in which the language was acquired, and the “inverse relation hypothesis” which states that what has been learned best will be most resistant to loss (Yoshitomi 1994: 12). Studies using these hypotheses provide valuable descriptions of a broad range of linguistic categories which are lost such as lexical items and morphology, and the social circumstances which surround such loss. In addition to looking at pedagogic and social factors involved in attrition, language attrition researchers also need to consider other factors such as the “frequency of input, perceptual saliency, features of the attriting as well as the replacing language, and considerations of markedness and language universals” (Hansen 1999: 150). Hansen and Chen’s (2001) investigation of the loss of numeral classifiers indeed supports the proposition that the regression of the acquisition sequence may be driven by a universal accessibility hierarchy. Pushing this line of inquiry, we might further narrow our focus and ask how we can account for attrited forms which are neither L1 nor L2 induced, yet which adhere to the constraints found in natural language. Are such forms entirely the result of transfer and forgetting? Most probably not. One possible explanation might be that these attrited forms are in part the result of psycholinguistic processes similar to those which guide L1, and possibly L2, acquisition. One theoretical framework which addresses such psycholinguistic processes is Chomsky’s (1981, 1986) theory of Government and Binding (GB). In adopting this framework for their work on L1 attrition, Sharwood Smith and van Buren (1991: 17) note that “however satisfactorily situated attrition data were within current sociolinguistic models, they could not be properly understood without considering their psycholinguistic status.” Even as far back as 1983, Sharwood Smith noted: “It would be interesting to see to what extent the tendency to restructure grammars in cases of loss was dictated by principles of markedness and core grammar as propounded in Chomsky (1980b)” (p. 229).
Generative-based L2 attrition
In particular, it seems important to know whether an individual has lost or is even able to lose those kinds of underlying mental representations of his or her attriting language that may be identified as Chomskian style Universal Grammar (UG)-based competence. Or perhaps the competence remains, but the ability to access it is lost. This could be thought of as performance attrition (O’Grady 2001). If we accept Sharwood Smith and van Buren’s (1991) claim that attrition researchers need to include an attriter’s psycholinguistic status in their theoretical framework, what new suggestions can be made? One possibility is that, if an attriter does have access to UG, we might surmise that in the absence of exposure to the target language, any UG-constrained aspects of that language such as parameterisation of headedness or constraints on movement might behave differently from knowledge of non-UG-constrained areas such as lexical items or inflectional and derivational morphology. Finding evidence that UG-driven aspects exhibit differing patterns of loss would suggest that there are certain UG constraints on language attrition and would contribute to a more explanatorily powerful hypothesis of language attrition than the different versions of the regression hypothesis alone provide. As descriptive models of both L1 and L2 language attrition increasingly contribute to our understanding of the social and pedagogical influences on loss, it seems appropriate then that there be an accompanying focus which seeks to explain this loss at the psycholinguistic level. Working within such an abstract framework as generative linguistics, researchers need to identify and investigate measurable linguistic features which can inform us of an aspect of a speaker’s knowledge of the language in question.
. Selecting a linguistic feature for analysis As discussed by Gürel (this volume), evidence of changes in an individual’s competence at the syntactic level can be observed in binding phenomena, where deviant forms of anaphoric relationships occur in the attriting language. The study under consideration in this chapter complements Gürel’s investigation by examining the loss of constraints on reflexive binding, specifically Principle A of the binding theory which requires that an anaphor be bound in a local domain. “Bound” here means the anaphoric expression and its antecedent are co-indexed, and that the antecedent constituent-commands (c-commands) the anaphor, as in (1).
Bede McCormack
(1) Alexi knew Jasonj deceived himself*i/j
In (1), himself can refer back to Jason (co-index j), but not to Alex (co-index i). For our purposes, the definition of c-command proposed by Chomsky and Lasnik (1995: 102) and described earlier by Gürel will be used. The focus of this study then, was whether the governing category of the language prone to loss (English) attrites, and if so, how does the attrited form affect a speaker’s reflexive pronoun system. Let us briefly contrast the English and Japanese governing categories. In English, a reflexive must be bound to a “local” antecedent, one within the minimal clause containing a subject. Japanese, on the other hand, does not require the reflexive to be bound to a local antecedent, but allows co-reference with either the local or “long-distance” (matrix) antecedent. In neither language, however, does the finiteness of the subordinate clause play a role in determining grammaticality. Examples (2) through (5) compare English and Japanese and illustrate this point by using asterisked co-indices to identify grammatically unacceptable antecedents in English. (2) a. Alicei blamed herselfi b. Alicei -ga zibuni -o semeta Alice-nom self-acc blame-past (3) a. Alicei said that Suej blamed herself*i/j b. Alicei -wa Suej -ga zibuni/j o semeta-to itta. Alice-top Sue-nom self-acc blame-past-comp say-past (4) a. Alicei told Suej to blame herself*i/j b. Alicei -wa Suej-ga zibuni/j o semete-to itta. Alice-top Sue-nom self-acc blame-comp say-past (5) a. Alicei said that Suej said that Maryk blamed herself*i/*j/k b. Alicei -wa Suej -ga Maryk -ga zibuni/j/k -wo semeta-to Alice-top Sue-nom Mary-nom self-acc blame-past-comp itta-to itta. say-past-comp say-past
In (2a) herself refers to the local antecedent Alice, which also happens to be the matrix subject. The Japanese equivalent of this monoclausal sentence, (2b), corresponds to the English interpretation. In (3a) however, herself refers to Sue, but not to Alice, as Alice is outside the local clause. In the Japanese equivalent of this biclausal sentence, (3b), however, the reflexive zibun is bound by an NP that can appear in any position in the sentence that c-commands zibun. In this way the antecedent of the reflexive is not limited to the nearest clause
Generative-based L2 attrition
and may refer back to either Sue or Alice. Since the antecedent here may be in the matrix sentence subject as well as in the local clause, the Japanese range of potential antecedents is greater than that of English. An attrited English system could show evidence of transfer from the Japanese system in that the informant allows the greater (i.e. Japanese-like) range of antecedents in English. (4a) and (b) illustrate the same phenomena as in (3a) and (b), but here the subordinate clause in (4) includes a nonfinite verb form whereas (3) includes a finite form. Despite the tense difference, the grammatical antecedent of each reflexive in each language respectively remains the same. This illustrates that finiteness does not play a role in English or Japanese. As we will see, however, it does affect reflexive binding in other languages such as Russian in that longdistance binding is allowed out of a clause with a nonfinite verb. (5a) and (b) illustrate a more extreme form of (3) and (4). Despite the multiple NPs, the reflexive herself in English still only refers back to the nearest antecedent, Mary, whereas Japanese, zibun in (5b) can refer back to any of the three possible antecedents – Alice, Sue, or Mary. The above examples illustrate the tension between English and Japanese regarding which clause may function as a governing category. English is more restricted, Japanese freer.
. Informant selection The test subjects for this investigation were six Japanese female college students. Their knowledge of English reflexive binding was tested over various periods of time, ranging from 9 months to more than 16 months, depending on the test subject. Some of the variables which complicated the selection of informants were their age at first exposure to English, the length of exposure, the input type, and their proficiency at cessation of exposure to English (i.e. when they returned to Japan) immediately prior to data collection. Three further complications were that the test subjects’ level of English at the start of data collection had to be high, testing had to begin immediately upon their return from abroad, and the subjects had to be willing to be involved with the study for at least a year. As it turned out, finding test subjects was difficult, and even then, English levels as reflected in their scores on the Test of English as a Foreign Language (TOEFL) were slightly compromised in two cases (see Table 1 below). Nonetheless, a high level of proficiency with respect to English binding principles was observed in all six test subjects upon their return to Japan, and they generously participated in the research project.
Bede McCormack
In this study, three of the test subjects were childhood bilinguals (“returnee” informants R1, R2, and R3), and three had no exposure to English in the L2 environment until they were college aged (“student” informants S1, S2, and S3). The inclusion of three returnees and three non-returnees in this study allowed for the effects of age at first exposure to the L2 on attrition to also be considered. In addition to the test subjects, the results of six native English speakers were also included as a control group. The native speaker controls were given the entire battery of test sessions at once and they responded to all 234 truth value test target items and distractors and 143 grammaticality judgment target items and distractors over the course of one to two weeks. Additional native speaker results were expected, but only six completed all test items. Details of the Japanese test subjects appear in Table 1.
. Test design Although spontaneous data collection techniques can provide a broad range of authentic speech patterns, it is unlikely that such an approach can elicit a significant amount of target forms as specific as those required to investigate binding. Therefore, the present study used both truth value judgment and grammaticality judgment tests, each test attempting to establish each informant’s degree of adherence to the local binding restriction in English biclausal sentences with finite subordinate clauses, and biclausal sentences with nonfinite subordinate clauses. In truth value Type 1a sentences, an associated story makes the local antecedent in the stimulus sentence the pragmatic (and in English the grammatical) choice of the two potential antecedents, and for Type 1b sentences, the story makes the long distance antecedent in the stimulus sentence the pragmatically correct (but for English ungrammatical) choice. Let us look at an example of each type. (6) Truth Value Type 1a: Biclausal sentence, finite verb in embedded clause, pragmatically correct The CIA officer captured a foreign spy who knew many secrets. The spy drank some poison and died. The officer knew the spy killed himself. (Expected judgment: True)
In (6), the two candidate antecedents of himself are the spy and the officer. If a subject maintains her knowledge of UG constraints on binding with respect
US local kindergarten and primary school Compulsory EFL instruction in Japan, including 2 years at university US university
US local primary school Compulsory EFL instruction in Japan including 2 years at university US university
US local primary school Compulsory EFL instruction in Japan including 2 years at university US university
Compulsory EFL instruction in Japan including 2 years at university US Pre-sessional language school and university
Compulsory EFL instruction in Japan including 1.5 years at university US university
Compulsory EFL instruction in Japan including 2 years at university US Pre-sessional language school and university
0;0–8;10 12;11–19;11 20;4–21;4
6;3–10;2 13;1–20;2 20;8–21;7
8;3–11;1 13;0–20;1 20;8–21;8
13;4–20;6 20;11–21;8
12;6–19;5 19;6–20;7
13;1–20;8 21;1–22;3
R1
R2
R3
S1
S2
S3
520
593
610
533
580
630
TOEFL Score
19
19
20
20
19
18
Age at TOEFL
*The TOEFL consists of four sections (listening, reading, structure and writing) and has a maximum combined score of 677 points. A minimum score of 550 is generally required for undergraduate study in the US, 600 for graduate study.
Type of L2 Exposure
Ages
Test Subject
Table 1. Ages of all six test subjects at significant periods of L2 acquisition, including type of exposure and TOEFL* details (McCormack 2001: 133)
Generative-based L2 attrition
Bede McCormack
to English, she should judge the stimulus sentence true and that it accurately reflects the content of the story, namely that the spy killed himself. A judgment of false was interpreted as reflecting a loss of English binding restrictions in favour of the wider ranging Japanese restrictions, i.e., L1 transfer. (7) Truth Value Type 1b: Biclausal sentence, finite verb in embedded clause, pragmatically incorrect One day, John was delivering letters when a big dog ran at him. John tried to stop the dog, but it was too late. The dog grabbed John’s leg with his teeth. John knew the dog bit himself on the leg. (Expected judgment: False)
In (7), the two candidate antecedents of himself are the dog and John. If a subject maintains her knowledge of UG constraints on binding for English, she should believe the stimulus sentence to be false and that it inaccurately reflects the content of the story as himself is restricted to the embedded clause and cannot refer back across the clause boundary to John. A judgment of true, however, may reflect a change in her knowledge of constraints for English binding in that she has broadened the governing category to allow the matrix subject to serve as antecedent. Truth value Type 2a and 2b sentences are structurally similar to TV 1a and 1b sentences, differing in that TV Type 2 sentences have a nonfinite verb in the embedded clause, as in (8) and (9). (8) Truth Value Type 2a: Biclausal sentence, nonfinite verb in embedded clause, pragmatically correct Arnold always asked George what the time was. George got tired of this and told Arnold to buy a watch. George wanted Arnold to buy himself a watch. (Expected judgment: True) (9) Truth Value Type 2b: Biclausal sentence, nonfinite verb in embedded clause, pragmatically incorrect Ed was a customer at a fancy restaurant. He waited a long time for the waiter to take his order. Finally he got up and asked the waiter for service. Ed asked the waiter to serve himself. (Expected judgment: False)
As with the finite Type 1a sentences, maintenance of reflexive binding should result in all TV Type 2a stimulus sentences being judged true, and TV Type 2b sentences judged false. Incorrect judgments here would suggest a subject’s knowledge of the local domain for English had attrited under the influence of her Japanese L1 binding principles.
Generative-based L2 attrition
In addition to the truth-value judgment test, attrition data were also collected by conducting a grammaticality judgment test. The stimulus sentences in this test also reflected the local vs. long-distance distinction, and were rendered either grammatical or ungrammatical by virtue of number discrepancies as in (10), or gender discrepancies, as in (11). (10) GJ Type 1a: Biclausal sentence, finite verb in embedded clause, number agrees His parents knew Ben had asked about himself. (Expected judgment: OK (i.e., grammatical)) (11) GJ Type 1b: Biclausal sentence, finite verb in embedded clause, gender agrees Adam thought his daughter had served herself. (Expected judgment: OK (i.e., grammatical))
A judgment of Wrong on sentences like (10) would suggest that the subject’s knowledge of UG-determined constraints on the governing category for English had attrited as the test subject would presumably be trying to bind back to the long distance antecedent His parents, but the anaphoric relationship is blocked by the wrong number. In sentences such as (11), the gender of the reflexive and the local antecedent agree. As in (10), a judgment of Wrong for (11) would suggest the test subject is attempting to bind long distance back to the matrix subject Adam, but this time the anaphoric relationship is blocked by the gender difference.
. Presentation of results For each test subject, the period of time from her return to Japan to her last test session was referred to as her ‘data collection period’. Although test subjects had varying numbers of test sessions, all six had a data collection period lasting at least 1 year, except test subject S1 whose data collection lasted 9 months. In order to compare test subjects’ results within a meaningful time-frame, each test subject’s entire data collection period was divided into three approximately equal sub-periods of time which are referred to as the initial, the secondary, and the final data collection sub-periods. The exact length of each sub-period is determined by when test sessions took place with respect to time since exposure to English in the L2 environment ceased (i.e. time since returning to Japan), thus causing the same sub-period for each test subject to vary somewhat. The initial sub-period includes the point of re-entry into Japan to whichever test
Bede McCormack
S3
S2
S1
R3
R2
R1 0
5 10 Months Since Return to Japan Initial
Secondary
15
Final
Figure 1. Test subject data collection sub-periods by TSR (Time since re-entry into the L1 environment)
session is closest to 4 months since her re-entry. For example, test subject S2 had her 6th test session on the 4th month and 3rd day after her return to Japan. This test session is referred to as Time Since Re-entry (TSR) 4.1, the number of months, in months and tenths of months, since re-entering the Japanese environment. The secondary sub-period includes the approximately 4-month period subsequent to the initial sub-period. For example in S2’s case, her secondary sub-period lasted from TSR 4.1 to TSR 8.1, 8 months and 3 days after TSR 0, her re-entry date into Japan. These initial, secondary and final data collection sub-periods for each Japanese test subject are represented in Figure 1. Note that as test subject S1’s entire data collection period lasted only 9 months, she has only initial and secondary sub-periods. In analysing the six test subjects’ test performance, these three sub-periods were used to compare the test subjects’ performance to each other. Because
This table includes each test subject’s score (percentage) on each TV test session she took, along with the test session number and corresponding TSR number. The initial, secondary and final data collection sub-periods presented in Figure 1 are indicated by shading.
Table 2. TV scores at time since L1 reentry (TSR)*
Generative-based L2 attrition
Bede McCormack
of space constraints, we will only look at the truth value judgment test results, which were analysed in terms of degree of adherence to the principles of reflexive binding in English as measured against the native English speaker data. These results are represented in Table 2. In Table 2, time is calculated from when a test subject’s exposure to English ceased upon her return to Japan (TSR = 0). That is, all test subjects returned to Japan at “0” months and then participated in the study for various periods of time. For example, with a 16 month data collection period, R1 participated the longest. She also exhibited the most stable results. At 9 months, test subject S1 has the shortest data collection period. However, compared to the other five subjects at the nine-month mark, she showed one of the highest degrees of maintenance of reflexive binding. On the other hand, test subject R3, a returnee with childhood exposure to English in the States, shows a high level of knowledge of reflexive use on her return to Japan, but this knowledge begins to become unstable just five months after returning. By 14 months, R3’s performance on the tests has attrited to a level similar to that of test subjects with no childhood exposure to English (S2, and S3).
. Discussion The general pattern exhibited by all six test subjects initially shows varying but high levels of knowledge of reflexive binding. These levels then become unstable, considerably in some cases, over the course of approximately 9 to 16 months. In particular, reflexives in nonfinite subordinate clauses show a tendency to become bound outside the subordinate clause rather than to the grammatical local antecedents, though local binding is maintained in tensed subordinate clauses. Local binding in tensed clauses thus appears more resistant to attrition than to binding in nonfinite clauses. This result correlates with Thomas’ (1993) findings in her study of the acquisition of L2 English in Japanese and Spanish learners. Thomas’ results suggest that her test subjects acquired the local binding constraint in English non-tensed clauses later than in tensed clauses. My results then, considered in light of the attrition data discussed here, lend support to de Bot and Weltens’ (1991) regression hypothesis. If Thomas is right, and English binding restrictions in sentences such as Type 2a (biclausal nonfinite) are acquired only after those applying to sentences like 1a (biclausal finite) are in place, then the evidence found here of restrictions applying to Type 2a sentences becoming unstable and attriting while restric-
Generative-based L2 attrition
tions applying to Type 1a sentences are maintained would support the “first learned, last lost” claims of the regression hypothesis. One explanation for this may be that these speakers gradually lose the local governing category restriction in English in the face of no L2 input, and come to associate the governing category with their Japanese L1. In this case, L1 influence can be held responsible for the loss of the test subjects’ English binding restrictions and we can attribute attrition to L1 transfer. In particular, this would explain the increasing rejection of the locally bound anaphors, as Japanese does not allow binding to object antecedents. If changes in the test subjects’ grammar were due solely to L1 influence, then the original L2 English settings for reflexive binding may be thought of as having been “deset” to the L1 setting, and the attriting grammar would eventually come to take on more and more of the L1 grammar’s rules. This parameter desetting model would not rely on any UG-like speaker-internal cognitive knowledge or processing mechanism beyond the already instantiated L1 and L2. However, the increasing distinction made by test subjects such as R3, S2 and S3 between tensed and non-tensed complements does point to the attrition process being influenced by more than just L1 transfer. Finer and Broselow (1986) suggest that evidence of this distinction is consistent with the value for the governing category parameter of a possible natural language somewhere between English and Japanese, such as Russian, in which the finite/nonfinite distinction plays a role in determining the governing category. Progovac (1992) illustrates this distinction and shows Russian allowing binding outside an infinitival clause as in (12a), but not outside a finite clause, as in (12b). citat’ svoii/j doklad Professori poprosil assistentaj professor asked assistant-acc to-read self ’s report-acc ‘The professori asked the assistant to read self ’si/j report’ b. Vaniai znaet cto Volodiaj liubit svoi-u*i/j zen-u Vania knows that Volodia loves self ’s-acc wife-acc ‘Vaniai knows that Volodiaj loves self ’s*i/j wife’ (Progovac 1992: 675)
(12) a.
In (12a), we see a nonfinite verb citat’ in the complement, and the pronoun svoi taking Professor as its long distant antecedent, whereas in (12b), the finite verb liubit occurs in the complement, and svoi-u refers back to the local antecedent, Volodia. Despite the fact that neither English nor Japanese makes this finite/nonfinite distinction, test subjects in the present study did increasingly develop a tendency to make the distinction, accepting long distance antecedents in nonfinite biclausal sentences, but preferring the local antecedent in finite sentences, where pragmatically appropriate.
Bede McCormack
Although ungrammatical in English, such long distance antecedents are UG-sanctioned. Were the test subjects to only accept long distance antecedents and reject all local antecedents, a case could be made that by violating the GCP, the test subjects’ grammar was not UG constrained. But this did not happen. Instead, the tendency was for them to regress to an earlier stage of their interlanguage in a UG-constrained way by increasingly accepting long distance antecedents in nonfinite sentences, while maintaining the local binding restriction in finite sentences. Never do any of the test subjects come to uniformly reject all local antecedents. In particular, the robustness of the test subjects’ TV type 1a sentences (biclausal finite) reflects strong adherence to the least marked binding pattern (minimal category must have a subject). Only 7.5% (13 out of 172) Type 1a sentence misjudgements appear in the data, making the TV Type 1a sentences the sentence type most resistant to attrition. This combination of maintenance of high levels of accuracy on the local binding pattern, along with the acceptance of ungrammatical (for English) long-distance antecedents in nonfinite clauses, supports the possibility that, despite the emergence of a deviant grammar as attrition sets in, the emergent grammar nonetheless remains UG constrained. And if we assume those same UG constraints were at least partly responsible for the development of the grammar of binding during the acquisition process, as Thomas (1993) suggests, then it is entirely plausible that we can look to the psycholinguistic processes guiding both the acquisition and attrition to account for corresponding Regression Hypothesis-related phenomena.
. Conclusion Although most test subjects failed to maintain their initial levels of knowledge of the English Principle A-related binding phenomena, by the end of their data collection periods, the attrited patterns identified in this study in no way appeared to constitute a non-UG constrained “rogue” grammar. Neither can these new forms be attributed to L1 transfer alone as they show patterns of anaphoric relationships not found in either Japanese or English. The implications of these results illustrate that adding a psycholinguistic thread to the dominant sociolinguistic and regression-based approaches to attrition research can potentially lead to richer, more multi-dimensional investigative techniques that will further contribute to a unified theory of both first and second language attrition.
Generative-based L2 attrition
References Bot, K. de, & Weltens, B. (1991). “Regression, recapitulation, and language loss”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 31–51). Cambridge: CUP. Chomsky, N. (1980a). “On binding”. Linguistic Inquiry, 11, 1–46. Chomsky, N. (1980b). Rules and Representations. New York, NY: Columbia University Press. Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. (1986). Knowledge of Language. New York, NY: Praeger. Chomsky, N. & Lasnik, H. (1995). “Principles and parameters in syntactic theory”. In N. Chomsky (Ed.), The Minimalist Program (pp. 13–127). Cambridge, MA: MIT Press. Finer, D., & Broselow, E. (1986). “Second language acquisition of reflexive binding”. Proceedings of the Sixteenth Annual Meeting of the North Eastern Linguistic Society, 16, 154–168. Hansen, L. (1999). “Not a total loss: The attrition of Japanese negation over three decades”. In L. Hansen (Ed.), Second Language Attrition in Japanese Contexts (pp. 142–153). New York, NY: Oxford University Press. Hansen, L., & Chen, Y.-L. (2001). “What counts in the acquisition and attrition of numeral classifiers?” JALT Journal, 23 (1), 90–110. McCormack, B. (2001). The Maintenance and Loss of Reflexive Anaphors in L2 English. Unpublished Doctoral Dissertation, University of Durham, UK. O’Grady, W. (2001). “Language Acquisition and Language Loss”. Plenary presentation, Pacific Second Language Research Forum (PacSLRF), Honolulu, October 2001. www.ling.hawaii.edu/faculty/ogrady/langacqlangloss.pdf. Progovac, L. (1992). “Relativized SUBJECT: Long-distance reflexives without movement”. Linguistic Inquiry, 23, 671–680. Sharwood Smith, M. (1983). “On first language loss in the second language acquirer: Problems of transfer”. In S. Gass & L. Selinker, (Eds.), Language Transfer in Language Learning (pp. 222–231). Rowley, MA: Newbury House. Sharwood Smith, M., & van Buren, P. (1991). “First language attrition and the parameter setting model”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition, (pp. 17– 30). Cambridge: CUP. Thomas, M. (1993). Knowledge of Reflexives in a Second Language. Amsterdam and Philadelphia: John Benjamins. Wexler, K., & Manzini, R. (1987). “Parameters and learnability in binding theory”. In T. Roeper & E. Williams (Eds.), Parameter Setting (pp. 41–89). Dordrecht: Reidel. Yoshitomi, A. (1994). The Attrition of English as a Second Language of Japanese Returnee Children. Unpublished Doctoral Dissertation, University of California, Los Angeles, USA.
Convergent outcomes in L2 acquisition and L1 loss Silvina Montrul University of Illinois at Urbana-Champaign
Introduction L1 loss is a general term that covers attrition, incomplete acquisition, language death, and even language change. L1 attrition, in my opinion, should only be used to characterise loss of linguistic ability after an L1 was acquired completely (see also Köpke and Schmid, this volume). The best scenario is the case of first generation immigrants, as those studied by Gürel (2002), Köpke (2002), Major (1992), Sorace (2000), and Schmid (2002), among many others. By contrast, in cases of adult second and third generation immigrants (the simultaneous bilinguals and early child L2 learners that will be reported here), it is hard to establish whether these individuals acquired structural aspects of the family or heritage language completely in early childhood (i.e., they reached a steady state). Most probably they did not, and if they did, they might have lost aspects of it soon afterwards, as recently documented by Silva-Corvalán (2003) in a study of attrition and incompleteness of Spanish verbal morphology in very young children. Moreover, bilingual school-aged children have been shown to be very vulnerable to abrupt shifts in language dominance (Vihman & McLaughlin 1982). Therefore, I refer to these cases as incomplete acquisition/attrition because it is hard to reconstruct their linguistic past. However, for ease of exposition, I will leave out the “/attrition” part of the definition out, and just refer to this group as incomplete learners (see also Polinsky 1997 for this distinction). The potential relationship between L2 acquisition and L1 attrition (1st generation immigrants) within current linguistic theory (Chomsky 1995) has been addressed by Sorace (1999, 2000a), Bouba, Filiaci, Heycock, Sorace, &
Silvina Montrul
Tsimpli (2002), and Gürel (2002). In this chapter, I will attempt to characterise the formal nature of incomplete acquisition by 2nd and 3rd generation bilinguals from the perspective of generative grammar, and to draw some parallels with advanced stages of L2 acquisition, elaborating on conclusions reached in Montrul (2002). L2 learners refer to speakers of an L1 who began learning an L2 after puberty and are still in the process of learning. These can also be considered sequential or late bilinguals. The early bilinguals in this study are Spanish heritage speakers living in the U.S. Heritage speakers are simultaneous bilinguals and child L2 learners whose family language is a minority language and who live immersed in a linguistic environment where the majority language predominates. These 2nd generation immigrants typically display great variability in their proficiency level and in different linguistic skills in the heritage language (Polinsky 1997; Silva Corvalán 1994; Valdés 2000). There are both theoretical and empirical reasons for comparing these two populations. Conceptually, these two situations have in common external and internal factors that appear to play a significant role in language development and underdevelopment. For example, in both L2 acquisition and L1 attrition there is the potential effect of transfer from the dominant language (L1 or L2 depending on the situation) (Cook 2003; Köpke 2002; Pavlenko 2000). There is also the effect of potential universal mechanisms that are not related to the influence of the other language and that are common in language acquisition and diachronic change (mechanisms that are part of Universal Grammar would fall in this category). In L2 acquisition and in L1 attrition reduced input as well as decreased frequency and degree of language use can also contribute significantly to the erosion or lack of linguistic development toward the target grammar (Köpke 2002). Finally, attitudes and motivation play a role in speakers’ willingness to maintain a family language or to continue learning an L2. Based on these factors, my working hypothesis is that the eroded L1 grammars of bilinguals at a given state or even the fossilised endstate resemble the incomplete grammars typical of intermediate and advanced stages in L2 acquisition. A representative schema of the comparison I attempt to make is shown in Figure 1, where an adult monolingual speaker is assumed to have a “stable” degree of (formal) linguistic competence X, an adult L2 learners is assumed to have full competence in the L1 and advanced competence in the L2, and an adult bilingual whose family language (L1) is weaker than the majority language (L2) is assumed to have less competence in some areas of the L1 than in the L2, and perhaps competence in many domains of the L2 is not quite complete either (this is another empirical question). I will also argue that in the case of L2 learners and heritage speakers the stronger language influences
Convergent outcomes in L2 acquisition and L1 loss
L
a. monolingual
L1
L2
b. L2 learner
L1
L2
c. bilingual heritage speaker (incomplete learner)
Figure 1. Patterns of adult bilingual and monolingual competence and language balance
the weaker language and predicts the patterns of incompleteness or loss found at the structural level. This will be evident from the quantitative differences in particular structural areas found between late bilinguals who are learning an L2, and early bilinguals (heritage speakers) who show less proficiency in one of the languages exposed to since birth (i.e., incomplete learners). A second important aim of this chapter is to make a methodological claim: If theories of L2 acquisition (SLA) can contribute to the field of language loss, so can SLA empirical methods. Many studies of language loss have relied on case studies or have looked at very few individuals (Polinsky in press; Silva Corvalán 2003), and this is a valid and suitable methodology to study particular subjects longitudinally and to document how a particular grammar changes. However, the field of language loss can also benefit from applying experimental methods, where a group of participants assumed to be undergoing linguistic change is compared to a control or baseline group of individuals (preferably monolinguals speaking the same dialect) who are assumed to have more “stable” competence. By using a variety of test instruments that have already been successfully employed in SLA studies, namely instruments aimed at revealing linguistic performance in production, recognition, comprehension, judgments (of grammaticality, meaning, truth-value, etc.) and on-line processing of different grammatical phenomena, we can draw interesting comparisons between the two populations and the results can have more validity. This will also allow us to identify and investigate general trends in these populations. The studies to be described in this chapter fulfil this requirement: I consider the combined results of two independent but methodologically identical studies on Tense and Aspect morphology in Spanish that I have conducted with adult Englishspeaking post-puberty L2 classroom learners (Montrul & Slabakova 2003) and with adult Spanish heritage speakers (Montrul 2002) (simultaneous SpanishEnglish bilinguals from Hispanic background living in the United States whose
Silvina Montrul
dominant language is English). Before going into the details of the experiment, in the next section, I will turn to a description of how grammatical aspect is realised in Spanish, followed by a characterisation of the phenomenon in current generative terms.
.
Interpretive properties of the preterit/Imperfect contrast in Spanish
Aspect refers to the internal temporal constituency of a situation, such that a situation can be in progress or finished. Aspect is expressed lexically by the meaning of the verb as well as by other elements in the predicate (direct objects, clitics, prefixes, prepositional phrases, etc.). According to Vendler (1967), predicates are classified into four classes, as in Figure 2, based on eventivity and telicity. States, such as know, are predicates that have no input of energy and are atelic (i.e., they do not have an inherent endpoint). Activities, accomplishments and achievements are eventive predicates, either describing a process, as in activities like run; a result, as in achievements like notice, or both, as in accomplishments like run a mile. Like stative predicates, activities are atelic. By contrast, accomplishments and achievements have endpoints and are telic. In addition to lexical aspect, Spanish also marks grammatical aspect in the past tense. Spanish has two past tense forms – the preterit and the imperfect. The difference between the two has to do with the event being construed as bounded (perfective), as in Paula construyó un castillo de arena ‘Paula built a sand castle’, or unbounded (imperfective) as in Paula construía un castillo de arena ‘Paula was building a sand castle.’ In principle, all four predicate types are compatible with the two forms of the past, depending on context or on what the speaker wants to convey, as shown in (1) through (4). For example, achievements in the imperfect are odd for many speakers if an appropriate context is not provided (hence the symbol #). STATES
know
ACTIVITIES
run
ACCOMPLISHMENTS
run a mile
ACHIEVEMENTS
notice
stative atelic dynamic
Figure 2. Vendler’s classification according to lexical aspect
telic
Convergent outcomes in L2 acquisition and L1 loss
(1) El auto me costó/ costaba $20.000. the car to me cost-pret/ cost-impf $20.000 ‘The car cost me $20,000.’ (2) Juan durmió/ dormía en el sofá. Juan slept-pret/ slept-impf on the sofa. ‘Juan slept on the sofa.’ (3) Juan corrió/ corría 5 kms. John ran-pret/ ran-impf 5 kms. ‘John ran 5 kms.’ (4) Juan alcanzó /#alcanzaba la cima. Juan reached-pret/ reached-impf the top ‘Juan reached the top.’
state
activity
accomplishment
achievement
An important difference between Spanish and English is that English does not have a past tense form equivalent to the Romance imperfect. English expresses the progressive and continuous meaning that the imperfect expresses in Spanish with the past progressive. However, the past progressive is typically not possible with most stative verbs (except for examples like I am liking my math class these days). The English simple past is typically perfective, as the examples in (5) show. With stative verbs, the English simple past can be interpreted as perfective or imperfective, depending on the context, as in (6). (5) a. #Mary ran in the park and is still running. b. #Mary ran 5 miles and is still running 5 miles. c. #The ice melted and is still melting. (6) Mary was sick and she is still sick.
activity accomplishment achievement state
Irrespective of how the Spanish preterit and imperfect combine with lexical predicates, these two tenses have a variety of other meanings or interpretations. The preterit typically expresses that an event culminated in the past, and has a quantified reading that refers to a single event, as shown in (7). (7) Pedro robó algo en el autobus. Pedro robbed-pret something in the bus ‘Pedro robbed something in the bus.’
However, with a small class of stative verbs, use of preterit or imperfect signals an aspectual shift. In most descriptive and prescriptive grammars these verbs are referred to as change of meaning preterits. For example, the verb saber ‘to know’ is stative by default and when it is in the imperfect. However, the preterit
Silvina Montrul
turns it into an eventive verb, more specifically an achievement, emphasising the endpoint. (8) Juan sabía la verdad. Juan know-impf the truth ‘Juan knew the truth’ (9) Juan supo la verdad. Juan know-pret la verdad ‘Juan found out the truth.’
stative
eventive
The imperfect has a variety of other meanings. Like the English progressive, the imperfect can be used to describe an ongoing, progressive activity in the past. The Spanish imperfect progressive can also be used in this case, as in (11). (10) Cuando la directora abrió la puerta la profesora when the headmaster opened the door the teacher entregaba los exámenes. handed-impf in the exams ‘When the headmaster opened the door the teacher was handing in the exams.’ (11) Cuando la directora abrió la puerta la profesora estaba when the headmaster opened the door the teacher was entregando los exámenes. handing in the exams ‘When the headmaster opened the door the teacher was handing in the exams.’
The imperfect can also refer to a habitual activity or event in the past, as in (12). Notice that this particular meaning of the imperfect cannot be translated into the English past progressive, as shown in (13). Rather, auxiliaries like would or used to are used in English to express habituality in the past, as in (14). (12) María practicaba tennis cuando era niña. María practiced-impf when she was-impf a girl ‘María practiced tennis when she was a child.’ (13) *Maria was practising tennis when she was a child. (14) Maria used to/would practice tennis when she was a child.
Lastly, the imperfect also expresses genericity, a meaning that is related to habituality. For example, with impersonal pronouns, the imperfect triggers a specific
Convergent outcomes in L2 acquisition and L1 loss
and inclusive interpretation, as well as a generic interpretation of the subject, as in (15a), where se can refer to people in general or we. (15) a.
Se comía bien en este restaurante. se eat-impf well in this restaurant ‘One/We would eat well in that restaurant.’ se = la gente en general ‘people in general’ = nosotros ‘we’
b. Se comió bien en este restaurante. se eat-pret well in this restaurant ‘We ate well in that restaurant.’ se = # la gente en general ‘people’ = nosostros ‘we’
generic specific
#generic specific
By contrast, when the tense is in the preterit (15b), the impersonal pronoun se gets a specific, inclusive interpretation (we).1 In English, generics are expressed in the simple past or with the subject pronoun (we vs. one), as the translations above show. To summarise, the preterit denotes a completed event in the past, while the imperfect has a variety of other values: progressive, habitual and generic. The subtlest property of the imperfect tense is that with a variety of impersonal constructions, it has both a generic and a specific interpretation, which includes the speaker. By contrast, the preterit typically denotes a specific interpretation.
. Theoretical assumptions Within current grammatical theory (Chomsky 1995) parametric differences between languages are encoded in the lexicon, most specifically in the inventory of functional categories and their associated features and feature values. Following Giorgi and Pianesi (1997) and many others, I assume that aspect is part of Universal Grammar and is encoded in the functional category AspP where the interpretable formal features [± perfective] are checked in the syntax. In Spanish such checking is done overtly through preterit [+perfective] and imperfect [–perfective] morphology (see Montrul & Slabakova 2003 for more details). English also has AspP, but this functional category lacks the feature value [–perfective]. English eventive verbs (accomplishments, achievements and states) are lexically specified as [+perfective] (Giorgi & Pianesi 1997).
Silvina Montrul
In generative grammar adult linguistic (monolingual) competence is assumed to be steady and complete. Once parametric choices are made in early childhood they remain stable in the adult linguistic system. In L2 acquisition, the L1 steady state plays a significant role in shaping, guiding, and constraining interlanguage development. To account for persistence fossilisation and lack of convergence on the target L2 grammar, Hawkins and Chan (1997) proposed the Failed Functional Features Hypothesis, according to which L2 learners only have available the functional categories, features and feature values specified for their L1, and acquisition of different FCs, features or feature values is not possible in the L2 acquisition situation. Therefore, L1 influence limits L2 acquisition and will persist in very advanced L2 learners (see White 2003 for an account of several other positions within this debate). As far as attrition is concerned, Sorace (2000) and Bouba et al. (2002) view syntactic attrition as a change in the L1 grammar, as compared to the grammars of monolingual speakers, but parameters are not lost. Syntactic attrition primarily affects interpretable morpho-syntactic, and I would add semantic, features, especially those concerning aspects of the grammar in which the L1 and the L2 assume different values. This is akin to transfer from the L2 to the L1. Given these theoretical positions, if patterns of “incompleteness” or “change” from monolingual norms are similar in L2 acquisition and in language loss (incomplete acquisition), and if contact with the other language (in this case English) determines to some extent those structural patterns, then we expect to find that both L2 learners and Spanish heritage speakers will have difficulty with verb forms that involve the negative value of the [perfective] feature, since here lies the parametric difference between Spanish and English AspP. Therefore, L2 speakers and heritage speakers should show: 1. less discrimination between bounded (preterit) and unbounded (imperfect) interpretations with stative verbs, and difficulty with those few stative verbs that shift aspectual class in the preterit. 2. more difficulty recognising the habitual, generic and specific meanings of the imperfect tense. . Methodology Participants were 47 adult post-puberty L2 learners and 31 adult Spanish Heritage speakers (simultaneous bilinguals and early child L2 learners) from Mexican-American background living in the US. All these participants were
Convergent outcomes in L2 acquisition and L1 loss
Table 1. Proficiency scores (maximum 50) monolinguals (n = 20)
heritage speakers (n = 31)
advanced L2 speakers (n = 24)
superior L2 speakers (n = 23)
48.35 1.46 45–50
41.70 5.28 28–48
40.12 3.01 33–44
46.69 1.22 45–49
mean sd range
graduate and undergraduate students enrolled in different Spanish language, linguistics, and literature courses at two major research institution. Some of them were also Spanish instructors and language and literature professors. There was also a control group of 20 Spanish speakers from a variety of Spanish-speaking countries who were raised monolingually and had been living in the US for some time, ranging from 6 months to 3 years (for more specific information about the participants refer to Montrul 2002; Montrul & Slabakova 2003). In this chapter I only compare the results of the advanced and superior L2 speakers from Montrul and Slabakova (2003) with those of the simultaneous bilinguals and child L2 learners from Montrul (2002).2 At the outset of these studies, all participants completed parts of a Spanish proficiency test (DELE) consisting of a cloze and a vocabulary section.3 The scores of the comparison groups are listed in Table 1. An ANOVA on the proficiency scores revealed significant differences among the groups (F(3,94) = 62.330, p < .0001). Post hoc comparisons indicated that the heritage speakers, the advanced L2 speakers and the superior L2 speakers were statistically different from the monolinguals. The two groups of L2 speakers were not different from each other. Note that the greatest range of variation in scores (28–48) is observed in the heritage speakers’ group. . Tests The studies included spontaneous oral production, a morphology recognition task and two meaning judgment tasks. For lack of space, I will only refer here to the two instruments testing semantic interpretations. One was a Sentence Conjunction Judgment Task (adapted from Slabakova 2001), and presented sentences containing conjoined clauses. The verb in the first clause was in the preterit or imperfect and the conjoined clause either made the previous clause logical or contradictory. This task tested the interaction of grammatical aspect with lexical aspect (Cf. ex. (1) to (4) above). Participants had to judge how logical or contradictory the two-clause combinations sounded, by marking a five
Silvina Montrul
point scale ranging from –2 (contradictory) to 2 (logical). Twenty-one minimal pairs (preterit vs. imperfect) consisting of 7 states, 7 accomplishments, and 7 achievement predicates were constructed, as well 14 distracters (7 logical, 7 illogical). An example with a stative predicate appears in (16). (16) La clase era a las 10 pero empezó a las10:30. (imperfect-logical) the class was-impf at the 10 but started at the 10:30 2 –2 –1 0 1 La clase fue a las 10 pero empezó a las 10:30. (preterit-contradictory) the class was-pret at the 10 but started at the 10:30 --2 –1 0 1 2
The second instrument was a Truth Value Judgment Task (Crain and Thornton 1998), consisting of stories followed by sentences with verbs in the preterit or imperfect to be judged as True or False depending on the context. The task included three main conditions – change of meaning preterits with stative verbs (6 stories), one time event vs. habitual action in the past (5 stories), and generic/specific interpretation of impersonal subjects with se (6 stories) – plus distracters, for a total of 80 sentence-story combinations (see the appendix for an example of a story from the condition testing one-time events vs. habituals.) . Results
Sentence Conjunction Judgment Task. Results of a factorial ANOVA with repeated measures with predicate (accomplishment, achievement and state), and tense (preterit and imperfect) as within subjects factors, and group (monolingual, heritage speakers, advanced L2 speakers and superior L2 speakers) as between subjects factors, revealed significant main effects for predicate (F(2,93) = 67.101, p < .0001), for tense (F(1,94) = 242.78, p < .0001), and for group (F(3,94) = 8.965, p < .0001). All possible interactions were significant at the .05 level. Since in this task it was important to establish that speakers distinguished between preterit and imperfect forms with each predicate, a series of one-factor ANOVAs and independent sample t-tests were conducted for these contrasts for each group. If speakers do distinguish between logical and contradictory sentences depending on the form of the verb, we expect to find a statistically significant difference between preterit sentences (with mean acceptability scores on the negative side of the scale) and imperfect sentences (with mean acceptability scores on the positive side of the scale). According to
Convergent outcomes in L2 acquisition and L1 loss
the t-tests, for each group there was a statistically significant contrast between preterit and imperfect sentences with the three predicates. This means that the subjects in each group distinguished the overall implications of these two verb forms. In what follows I focus on the comparisons among the four groups. Figure 3 shows the results of accomplishment predicates. A one-factor ANOVA showed no statistically significant differences by group with accomplishments in the preterit (F(3,97) = 2.292, p < .083). As for accomplishments in the imperfect, there were differences only between the heritage speakers and the superior L2 learners, according to Tukey (F(3,97) = 4.141, p < .008). The results of achievements appear in Figure 4. There were significant differences among groups in the responses on achievements in the imperfect (F(3,97) = 4.185, p. < .008), particularly between the monolinguals and the advanced L2 speakers, according to Tukey. Crucially the two L2 speaker and the heritage speakers were not statistically different from each other. Finally, Figure 5 displays the mean acceptability judgments for stative verbs. Here, there were differences among the groups with the mean acceptability judgments of stative verbs in the imperfect (F(3,97) = 11.806, p < .0001). There were no differences between the monolinguals and the heritage speakers, but the two L2 speaker groups were different from the monolinguals and from the heritage speakers. As for stative verbs in the preterit, the differences between the groups were also significant (F(3,97) = 8.161, p < .0001). The heritage speakers scored significantly lower than the monolinguals and the superior L2 speakers, but were no statistically different from the advanced L2 speakers. 2 1.5 1
1.51
0.5
1.42
1.22
–0.78
–1
–0.83
heritage speakers
superior L2 speakers
advanced L2 speakers
1.06
0 –0.5
–1.38
Imperfect Preterit
–1 –1.5 –2
monolingual speakers
Figure 3. Sentence Conjunction Task. Mean acceptability judgments on accomplishments
Silvina Montrul 2 1.5 1 0.5
1.28 0.63
0.56
0 –0.5 –1
–1.78
0.31
–1.39
–1.67
–1.5
heritage speakers
superior L2 speakers
advanced L2 speakers
Imperfect Preterit
–1.5 –2
monolingual speakers
Figure 4. Sentence Conjunction Judgment Task. Mean Acceptability Judgments on Achievements
2 1.5 1
1.55
0.5
0.53
0 –0.5
1.17 0.52 –0.57
–0.66 –1.25
–1.41
Imperfect Preterit
–1 –1.5 –2
monolingual speakers
heritage speakers
superior L2 speakers
advanced L2 speakers
Figure 5. Sentence Conjunction Judgment Task. Mean Acceptability Judgments on States
The Truth Value Judgment Task. According to the results of a factorial ANOVA with repeated measures, there was a main effect for condition (F(2,94) = 29.898, p < .0001), for tense (F(1,94) = 6.117, p < .015), and for group (F(3,94) = 12.733, p < .0001). There was also a condition x tense, and a condition x tense x group interaction. The results reported below focus on differences between groups, for which one-way ANOVAs were conducted on the different variables.
Convergent outcomes in L2 acquisition and L1 loss 6 5.35
5.21
5.2
5
5.45 5 4.86 4.95
4.91
4.73
4.7
4.29 3.92
4
5.3 4.29
4.08
3.25
3 2 1 Preterit (F)
Imperfect (T) Stative
monolingual speakers superior L2 speakers
Preterit (T)
Imperfect (F)
Eventive heritage speakers advanced L2 speakers
Figure 6. Truth Value Judgment Task. Mean scores on change of meaning preterits
The results of condition A, testing stative verbs that change to eventive in the preterit (i.e., change of meaning preterits), are shown in Figure 6. When the story provided a stative context, the imperfect was True (the correct answer) and the preterit was False. All the groups were equally accurate with the correct imperfect Tense response. With the incorrect response (False), however, there were statistically significant differences between the groups (F(3,94) = 10.922, p < .0001). The heritage speakers and the advanced L2 speakers were not different from each other, but these two groups scored statistically lower than the monolinguals and the superior L2 speakers, according to Tukey. When the story presented an eventive context, the preterit was True and the imperfect was False. With the preterit sentences, there were differences among groups (F(3,94) = 7.158, p < .0001) due to the performance of the heritage speakers and the advanced L2 speakers, who were not different from each other but scored lower than the superior L2 speakers and the monolinguals. Finally, the false responses with the imperfect-tense sentences also revealed that the heritage speakers and the advanced L2 speakers scored statistically lower than the monolinguals the superior L2 speakers. This difference was also significant according to ANOVA (F(3,94) = 7.219, p < .0001). Figure 7 displays the mean scores of the condition testing one-time events vs. habitual activities in the past. In habitual stories, sentences in the imperfect were True and the same sentences in the preterit were False. With the imperfect
Silvina Montrul 5
4.85
4.55 4.65
4.95 4.81 4.86 4.95
4.65
4.7 4.26
4.21
4
3.7
3.44
3.41 3.02
3
3.04
2 1 Preterit (F)
Imperfect (T) Habitual
monolingual speakers superior L2 speakers
Preterit (T)
Imperfect (F)
One-time heritage speakers advanced L2 speakers
Figure 7. Truth Value Judgment Task. Mean scores on one-time events vs. habitual actions
sentences overall the heritage speakers (mean 3.45) and the advanced L2 speakers scored alike, and their responses were significantly lower than those of the monolinguals and the superior L2 speakers, according to ANOVA (F(3,94) = 6.395, p < .001). With the preterit sentences, the advanced group scored lower than the three other groups (F(3,94) = 5.182, p < .002), and this time the heritage speakers were like the superior L2 speakers. The results of eventive stories show a similar pattern of performance: while there was no difference among the groups with the sentences in the preterit (F(3, 94) = 1.352, p < .256), with the sentences in the imperfect, the heritage speakers and the advanced L2 learners scored statistically alike but much lower than the monolinguals and the superior L2 speakers. Lastly, Figure 8 displays the mean scores for the condition testing the specific versus generic interpretation of subject pronouns. Recall that for this condition, the preterit form of the verb triggers only a specific interpretation whereas the imperfect form of the verb is compatible both with both a generic and a specific interpretation. Results of generic stories showed that all the groups scored equally high with preterit (F(3,94) = .256, p < .857) and imperfect (F(3,94) = 1.518, p < .215). With the specific stories, all the groups were equally accurate with the preterit sentences (F(3,94) = .222, p < .881). However, the two L2 speaker groups and the heritage speakers showed the exact
Convergent outcomes in L2 acquisition and L1 loss 6 5
5.55 5.14 5.06 5.13
5.8 5.3 5.21 5.2
4.87
5.15
5 4.91 4.4
4 3.4
3.13 3.16
3 2 1 Preterit (F)
Imperfect (T)
Generic monolingual speakers superior L2 speakers
Preterit (T)
Imperfect (T)
Specific heritage speakers advanced L2 speakers
Figure 8. Truth Value Judgment Task. Mean score on Generic vs. Specific subject interpretation
same pattern of response and were statistically significant from the monolingual speakers, according to ANOVA (F(3,94) = 4.489, p < .005).
. Discussion and conclusion The combined results of the two methodologically identical studies compared here show that, in general, there are similarities in the performance of advanced L2 learners of Spanish and Spanish heritage speakers in the semantic interpretations of the preterit-imperfect aspectual opposition (with different predicates and in different contexts), as measured with a sentence conjunction judgment task and a truth value judgment task. Although all speakers showed reliable discrimination of the preterit-imperfect contrast overall, in general advanced L2 learners and heritage speakers differed significantly from monolingual Spanish speakers and even from superior L2 speakers in some areas of semantic interpretations. If the differences found between these two groups of bilinguals and the monolinguals is in some sense related to the influence of English as L1 in the case of the L2 learners, and related to the dominant L1 or L2 in the case of simultaneous bilinguals and early child L2 learners (the heritage speakers), it
Silvina Montrul
was predicted that “incompleteness” or “change” with respect to the grammar of monolinguals would be observed in specific conditions that appear to be regulated by the [–perfective] feature: (1) with stative verbs in the sentence conjunction judgment task and in condition A (change of meaning preterits) in the Truth Value Judgment Task, and (2) with the habitual, generic, and specific meanings of the imperfect, as tested in the Truth Value Judgment Task. As it turned out, stative verbs in the preterit were most problematic for the heritage speakers and the advanced L2 speakers, while stative verbs in the imperfect were most problematic for the two groups of L2 speakers. Another revealing result from the Sentence Conjunction Judgment Task was the performance of the heritage speakers and the two L2 groups on achievements in the imperfect, which were rated much less acceptable as compared with the ratings of the monolinguals, who were more approving of these sentences. As for the Truth Value Judgment Task, changes of meaning preterits were most problematic for the heritage speakers and the advanced L2 learners. While these speakers recognised that these verbs are stative in the imperfect, they had difficulty accepting or rejecting the stative-preterit combination in eventive contexts. Similarly, the same two groups had difficulty with the imperfect sentences in habitual stories. Finally, while all speakers recognised that the imperfect has a generic reading, neither the L2 speakers nor the heritage speakers recognised that the imperfect could also have a specific reading. In short, the combined results of the two studies confirm the predictions in terms of the [–perfective] feature being a vulnerable domain in these grammars. These results suggest that semantic features of functional categories are affected by incomplete acquisition and L2 acquisition, and point to the potential role of the primary or dominant language in the grammar of the “other” language in bilinguals. While the results of the heritage speakers would suggest that this influence is persistent and has led perhaps to a “fossilised” state as in advanced L2 learners, the results of the superior L2 speakers (and the near-native speakers in Montrul & Slabakova 2003) suggest that it is possible for advanced learners to overcome this stage, since the superior speakers were comparable to the monolinguals in all conditions with the exception of one (specific interpretation of the imperfect with impersonal subjects). Thus, the Failed Formal Features Hypothesis (FFFH) (Hawkins & Chan 1997) correctly describes and predicts advanced stages of L2 acquisition and fossilisation as a temporary state or “stage”, but is apparently not suitable to predict end-state outcomes and fossilisation as a permanent state of knowledge. What are the conditions, then, to overcome these “temporary” deficits? The heritage speakers and the advanced L2 learners who participated in these
Convergent outcomes in L2 acquisition and L1 loss
two experiments were all college students taking Spanish language or literature classes. In other words, they were learning Spanish. By contrast, many of the superior L2 speakers were graduate students pursuing studies in Spanish, and others were Spanish professors who had chosen Spanish as a career path. Although they did not live immersed in a Spanish-speaking environment, their commitment to learning and maintaining a very advanced level of Spanish was very strong. This suggests that motivation to seek opportunities to learn and maintain a language results in increased exposure to the language. Obviously, rich input, frequency of exposure, and use lead to parameter resetting and acquisition of functional categories or feature values that are not necessarily represented in the other language. Acquisition of subtleties of semantic interpretations is context dependent and experience-based. Thus, in attempting to pinpoint vulnerable structural areas of grammatical development, it appears that the lexico-semantic and morphosyntax interface are prone to incompleteness in L2 acquisition and early bilingualism (see also Hulk & Müller 2000), although this does not mean that attainment at the native-speaker level is impossible for advanced L2 learners given appropriate exposure, as shown by the superior speakers and the near-native speakers in Montrul and Slabakova (2003). Furthermore, as in different grammatical endstates (Sorace 1993), there is a role for L1 or L2 influence, as shown here, with the meanings of the imperfect tense assumed to be regulated by the feature [–perfective]. Additionally, universal tendencies of lexical and grammatical aspect combinations as proposed by the Lexical Aspect Hypothesis (Andersen 1986) that are typical of early stages of development appear to also be playing a role in shaping the grammatical states of the heritage speakers and the advanced L2 learners in this study. These two groups were more accurate at judging stative verbs in the imperfect and achievement verbs in the preterit, as opposed to the less prototypical uses of stative verbs in the preterit and achievements in the imperfect (although this latter combination was not problematic for the heritage speakers). These grammatical-lexical aspect combinations, which are regulated by pragmatic factors, are the latest to be acquired in L1 and L2 acquisition (Andersen 1986). These results suggest that reduced input during childhood and into the teen years, type of input (written vs. oral, and perhaps also literacy in the case of early bilinguals), and frequency of use, conspire to promote incomplete grammatical states in the heritage language of bilinguals who live immersed in the majority language. As already documented by Schlyter (1993) with bilingual French-Swedish children of unbalanced development, the weaker language (usually the family language) is retarded in relation to the stronger one, and
Silvina Montrul
develops very much like an L2 (see also Kravin 1992 for another case of erosion in early bilingual development). To the extent that the language development histories of the heritage speakers documented in this comparative study could have been similar to the situations reported by Schlyter and Kravin, the results of these adult bilinguals suggest that the weaker language develops but stabilises at an advanced stage, comparable to that advanced level reached by many adult L2 learners. Although both heritage speakers and the L2 learners have quite advanced proficiency, certain areas of language (like tense and aspect interpretations and other semantic-pragmatic areas) may remain different and incomplete with respect to the grammars of monolinguals or near-native speakers of an L2. While the results of Montrul and Slabakova (2003) showed that to attain full linguistic competence by near-natives is possible for L2 learners who started learning the L2 after puberty, what remains an open question for further study is if heritage speakers, who have been exposed to the family language much earlier in life, can also overcome the apparent deficits they show given sufficient exposure and use of their family language. In conclusion, if attrition results when two languages come into contact (Seliger 1996; Sharwood Smith & Van Buren 1991), I have suggested in this chapter that incomplete acquisition in the context of bilingualism is a specific type of language attrition at the individual and probably at the group level, in a specific population of bilinguals (early bilingual heritage speakers). Because the cognitive and linguistic relationship between how the two linguistic systems in this particular population interact and influence each other is very similar to what happens in adult L2 acquisition, I have argued that the field of language attrition in general can benefit significantly, both theoretically and empirically, by adopting methodologies extensively used in mainstream L2 acquisition research. This allows us to investigate issues of overall proficiency, as well as degree of semantic and morphosyntactic knowledge of the target language (L1 or L2), and the limits and possibilities of early and sequential bilingualism, with a variety of testing instruments. Moreover, by applying identical methodologies, studies of this type can address more precisely and directly the potential connection between these two populations.
Notes . The generic interpretation could be pushed if it is the case that restaurant is no longer opened for business, for example.
Convergent outcomes in L2 acquisition and L1 loss . Montrul (2002) addresses the issue of incomplete acquisition versus L1 attrition and included 4 groups: 20 monolinguals, 16 simultaneous bilinguals, 15 early child L2 learners and 8 late child L2 learners. The late child L2 learners would constitute cases of attrition, whereas the simultaneous bilinguals and early child L2 learners are considered cases of incomplete acquisition. Montrul and Slabakova (2003) speak to the longstanding debate on near-native competence in post-puberty adult L2 speakers. The experiment tested 20 monolinguals (the control group), 24 advanced L2 speakers, 23 superior speakers and 17 near-native speakers, as classified according to an oral interview and a proficiency test. The present chapter excluded the results of the late child L2 learners from Montrul (2002) and the near-native speakers from Montrul and Slabakova (2003). These two groups behaved very much like the monolingual comparison group. The results of the simultaneous bilinguals and early child L2 learners are grouped under heritage speakers. . The DELEs (Diplomas of Spanish as a Foreign Language) are the official accreditation of the degree of fluency of the Spanish Language, issued and recognised by the Ministry of Education, Culture and Sport of Spain. For more information visit http://www.dele.org/ dele-eng/index.asp
References Andersen, R. (1986). “El desarrollo de la morfología verbal en el español como segundo idioma.” In J. Meisel (Ed.), Adquisición del lenguaje – Aquisição da linguagem (pp. 115– 137). Frankfurt: Klaus-Dieter Vervuert Verlag. Bouba, M., Filiaci, F., Heycock, C., Sorace, A., & Tsimpli, I. (2002). “Syntactic attrition in Greek and Italian near-native speakers.” Paper presented at the First Conference on Language Attrition. August 21–23, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press. Cook, V. (Ed.). (2003). Effects of the Second Language on the First. Clevedon: Multilingual Matters. Crain, S., & Thornton, R. (1998). Investigations in Universal Grammar. Cambridge, MA: MIT Press. Giorgi, A., & Pianesi, F. (1997). Tense and Aspect: From Semantics to Morphosyntax. Oxford: Oxford University Press. Gürel, A. (2002). Linguistic Characteristics of Second Language Acquisition and First Language Attrition: Turkish overt vs. null pronouns. Unpublished Doctoral Dissertation, McGill University, Canada. Hawkins, R., & Chan, C. (1997). “The partial availability of Universal Grammar in second language acquisition: The failed functional features hypothesis”. Second Language Research, 13, 187–226. Hulk, A., & Müller, N. (2000). “Bilingual first language acquisition at the interface between syntax and pragmatics”. Bilingualism: Language and Cognition, 3, 227–244.
Silvina Montrul
Köpke, B. (2002). “Activation thresholds and non-pathological L1 attrition”. In F. Fabbro (Ed.), Advances in the Neurolinguistics of Bilingualism. Essays in honor of Michel Paradis (pp. 119–142). Undine: Forum. Kravin, H. (1992). “Erosion of a language in bilingual development”. Journal of Multilingual and Multicultural Development, 13, 307–325. Major, R. (1992). “Losing English as a first language.” The Modern Language Journal, 76, 190–208. Montrul, S. (2002). “Incomplete acquisition and attrition of Spanish tense/aspect distinctions in adult bilinguals.” Bilingualism: Language and Cognition, 5, 39–68. Montrul, S., & Slabakova, R. (2003). “Competence similarities between native and nearnative speakers: An investigation of the Preterite/Imperfect contrast in Spanish.” Studies in Second Language Acquisition, 25, 351–398. Schlyter, S. (1993). “The weaker language in bilingual Swedish-French children.” In K. Hyltenstam & Å. Viberg (Eds.), Progression and Regression in Language: Sociocultural, neuropsychological and linguistic perspectives (pp. 289–308). Cambridge: CUP. Schmid, M. S. (2002). First Language Attrition, Use and Maintenance: The case of German Jews in Anglophone countries. Amsterdam and Philadelphia: John Benjamins. Seliger, H. W. (1996). “Primary language attrition in the context of bilingualism.” In W. C. Ritchie & T. J. Bhatia (Eds.), Handbook of Second Language Acquisition (pp. 605–226). New York, NY: Academic Press. Sharwood-Smith, M., & Van Buren, P. (1991). “First language attrition and the parameter setting model.” In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 19– 51). Cambridge: CUP. Silva-Corvalán, C. (1994). Language Contact and Change. Oxford: Oxford University Press. Silva-Corvalán, C. (2003). “Linguistic consequences of reduced input in bilingual first language acquisition.” In S. Montrul & F. Ordóñez (Eds.), Linguistic Theory and Language Development in Hispanic Languages. Papers from the Fifth Hispanic Linguistics Symposium and the 4th Conference on the Acquisition of Spanish and Portuguese (pp. 375–397). Sommerville, MA: Cascadilla Press. Slabakova, R. (2001). Telicity in the Second Language. Amsterdam and Philadelphia: John Benjamins. Sorace, A. (1993). “Unaccusativity and auxiliary choice in non-native grammars of Italian and French: Asymmetries and predictable indeterminacy.” French Language Studies, 3, 71–93. Sorace, A. (1999). “Initial states, end-states and residual optionality in L2 acquisition.” Proceedings of the 23th Boston University Conference on Language Development (pp. 666– 674). Sommerville, MA: Cascadilla Press. Sorace, A. (2000). “Differential effects of attrition in the L1 syntax of near-native L2 speakers.” Proceedings of the 24th Boston University Conference on Language Development (pp. 719–725). Sommerville, MA: Cascadilla Press. Pavlenko, A. (2000). “L2 influence on L1 in late bilingualism.” Issues in Applied Linguistics, 11, 175–205. Polinsky, M. (1997). “American Russian: Language loss meets language acquisition.” In Proceedings of the Annual Workshop on Formal Approaches to Slavic Linguistics: The Cornell Meeting 1995 (pp. 370–406). Ann Arbor, MI: Michigan Slavic Publications.
Convergent outcomes in L2 acquisition and L1 loss
Valdés, G. (2000). Spanish for Native Speakers [AATSP Professional Development Series Handbook for Teachers K–16, Vol. 1]. New York, NY: Harcourt College Publishers. Vendler, Z. (1967). Linguistics in Philosophy. Ithaca, NY: Cornell University Press. Vihman, M., & McLaughlin, B. (1982). “Bilingualism and second language acquisition in preschool children.” In C. Brainerd & M. Pressley (Eds.), Verbal Processes in Children (pp. 35–58). Berlin: Springer. White, L. (2003). Second Language Acquisition and Universal Grammar. Cambridge: CUP.
Appendix Sample story from the Truth Value Judgment Task
Context: One Time Event Panchito era muy tímido y no tenía muchos amiguitos con quien jugar. Panchito pasaba todo el tiempo en su casa con su mamá. Ayer Panchito se reunió por primera vez con sus vecinitos para jugar y pasaron un rato muy agradable. Hoy Panchito se quedó nuevamente con su mamá. Panchito jugaba con sus vecinos. F Panchito jugó con sus vecinos. T “Panchito was (imperf) very shy and he did not have (imperf) friends to play with. Panchito would spend (imperf) time at home with his mother. Yesterday, for the first time, he got together (pret) with his neighbors to play and they had (pret) a very pleasant time. Today Panchito stayed (pret) with his mother again” Panchito played (imperfect) with his neighbors. F Panchito played (preterite) with his neighbors. T
Context: Habitual Laurita tenía muchos amiguitos y después de la escuela pasaba el tiempo en la casa de sus vecinos. Ayer se quedó en casa con su mamá y pasó un rato muy agradable. Laurita jugaba con sus vecinos. T Laurita jugó con sus vecinos. F “Laurita had (imperf) many friends and after school she would spend (imperf) time at his neighbors’ house. Yesterday Laurita stayed (pret) at home with her mother and had (pret) a very good time.” Laurita played (imperfect) with his neighbors. T Laurita played (preterite) with his neighbors. F
A modest proposal Explaining language attrition in the context of contact linguistics Steven Gross East Tennessee State University
Introduction One of the more disconcerting conclusions one arrives at when reviewing reports on L1 attrition is a lack of agreement regarding what aspects of an individual’s grammatical system are most vulnerable to attrition or even whether linguistic competence, once established, can be irretrievably lost (de Bot 1996). As Schmid (2002: 29) notes, this lack of agreement about these very basic issues “can almost certainly be put down to varying notions about [methodology] on the part of researchers”. Regarding methodologies, there are two major considerations that researchers must address. The first involves data collection techniques. What techniques will offer the best evidence for language attrition (for example, free discourse, ‘elicited’ free discourse, grammaticality judgments, experiments, or some combination of these)? The second deals with the analysis of the data. What linguistic features should researchers target in their analyses, and how should these features be organised? While the methodology of data collection is vitally important in language attrition research, how we analyse the data ultimately determines how we interpret our data, what generalisations we can offer, and moreover, how we explain the nature of language attrition. This article focuses on the second methodological consideration: how best to analyse language attrition data.
Steven Gross
.
Defining language attrition
A fundamental question that should concern all language attrition researchers is the question of what exactly language attrition is. To answer this question, Andersen (1982: 91) suggests that L1 attrition is characterised, among other things, by “a lack of adherence” to L1 linguistic norms. In other words, L1 attrition simply involves production ‘errors’ that a fully competent speaker of that language would not make. However, there are a number of complicating factors in assessing what constitutes an error in a given speaker’s linguistic production. First, for many of us who investigate elderly populations, there is the problem of which of these errors we should attribute to attrition and which ones we should attribute to factors related to aging. For example, in a review of the literature on nonpathological forms of language loss, Hupet and Schelstraete (1997: 157) report that individuals over the age of seventy appear to exhibit more difficulties in lexical retrieval and provide fewer and less precise synonyms than younger speakers. In addition, Kynette and Kemper (1986) note that older individuals tend to avoid producing syntactically complex structures (e.g. fewer subordinate clauses and fewer left-branching constructions) but when they do, these elderly individuals exhibit more grammatical errors (e.g. agreement errors or errors in gender assignment) than their younger counterparts. Second, regardless of the age of our subjects, when considering the status of a given form, language attrition researchers need to consider the issue of language variation in the variety they are investigating. For example, in my own data from Bavarian German speakers, the type of construction in (1) is quite common. (1) I muass nach Haus komm-en wei sie ist in ander-e I must to house come-inf because she is in other-pl Umständ. circumstances ‘I had to come home because she was in a family way.’
Note that the verb ist ‘is’ in the subordinate clause wei sie ist in andere Umständ ‘because she was in a family way’ occurs immediately following the subject, and not in clause final position. While German prescriptive grammars maintain that subordinating conjunctions like weil ‘because’ introduce SOV clauses, it is well known that the position of the finite verb in weil-clauses is variable (either following the subject or in clause-final position) for speakers of Bavar-
Attrition and contact linguistics
ian German, as well as for speakers of other varieties of German. This issue of language variation is an important consideration that must be addressed in language attrition research. A third consideration, one that Schmid (2002: 37) rightfully raises, is the question of who should decide what will count as an ‘error’? It seems that this problem is related to the problem of language variation raised in the previous paragraph. I agree with Schmid when she says that there should ideally be several judges who are competent speakers of the language varieties represented in the study employed to assess the status of questionable forms. Yet, the act of judging grammaticality necessarily involves a kind of grammaticality judgment task. Fourth, researchers must address the problem of whether the grammatical changes they observe should best be analysed as internally motivated or externally motivated changes. Distinguishing between these two sources of change is not always a straightforward matter: linguistic change may have multiple causes. For example, linguists have debated, at length, the causes of case mergers in American varieties of German. While most of the evidence seems to be on the side of internally motivated explanations, Salmons (1994: 62) notes that the reduction of case-marking paradigms in Texas German may be expedited because of contact with English. To address some of these difficulties, Seliger and Vago (1991) have catalogued various strategies that may distinguish internally from externally motivated language change. As for the problems of language variation and age, one obvious solution is to collect data from a monolingual control group that closely matches the experimental group in terms of the usual demographic characteristics such as socioeconomic status, education, age, and regional dialect. However, without a methodology that offers a coherent way to interpret the data, a control group can only tell us in what ways its linguistic production differs from the experimental group. In effect, without such a methodology, our observations run the risk of becoming mere laundry lists of divergent constructions. What is important to emphasise about language attrition is that it is one of a number of different language contact outcomes. Like other types of contact phenomena, language attrition involves contact between speakers of different languages, but as Myers-Scotton (2002: 5) points out, what is more important is that “languages are in contact in the sense they are adjacent in their speakers’ mental lexicon and can impinge on each other in production”. Certainly, not all sources of linguistic change in L1 attrition can be attributed to the L2 (see for example Seliger & Vago 1991); yet, crosslinguistic influence has been shown to have a significant effect on L1 behavior (Altenberg 1991; Kaufman & Aronoff
Steven Gross
1991; Pavlenko, this volume; Seliger 1991; Sharwood Smith 1983, among others). Thus, any definition of L1 attrition must acknowledge the role that the L2 plays in restructuring the L1 grammatical system. I offer a definition for L1 attrition that takes into account the role of the L2. The kinds of data that will be presented in this paper reflect such influence. L1 Attrition L1 attrition is the restructuring of the L1 linguistic system according to L2 patterns under the psycholinguistic pressures of bilingualism. However, L1 attrition may also be internally induced in the sense that some changes are motivated by universal strategies based on markedness, for example. L1 attrition is distinguished from language shift in that attrition is intragenerational, whereas language shift is intergenerational.
. General observations on language attrition Despite the apparent lack of agreement on what aspects of grammatical change are the more important indicators of attrition, most researchers would agree that these changes are anything but random. Thus, when we examine the data, the evidence indicates that not all categories of morphemes respond in the same way under the pressures of bilingualism. In fact, when we sift through the attrition literature, we find considerable anecdotal evidence that lexical items dominate as the elements that are vulnerable to attrition. For example, Olshtain and Barzilay (1991: 145) state that all the American Israelis in their study exhibited some reduction in lexical retrieval abilities. In addition to lexical retrieval difficulties, one frequently made generalisation in language attrition studies is that some L1 lexemes can be semantically restructured according to L2 requirements. Examples (2) and (3) illustrate the neutralisation of certain semantic distinctions in German that are not present in English. (2) Neutralisation of wissen ‘know a fact’ and kennen ‘be acquainted with someone’ bie bir her-kumm-a sein ha-bir niemand ge-buss-t when we here-come-prt be have-we nobody prt-know-prt ‘When we came here, we didn’t know anybody.’ (3) Neutralisation of wohnen bei ‘reside with someone temporarily’ and leben mit ‘live with someone in an intimate relationship’
Attrition and contact linguistics
bir sein mit mein Schwäger ein-ge-zong und ham mit jamm we are with my in-laws in-prt-pull and have with them a ganz Jahr ge-leb-t a whole year prt-live-prt ‘We moved in with my in-laws and lived with them for a whole year.’
In (2), the semantic distinction between wissen ‘to know a fact’ and kennen ‘to know/be acquainted with a person’ has been neutralised reflecting the English pattern whereby the verb know is used for all contexts of knowing. This sort of phenomenon has also been reported elsewhere, for example in Fuller (1997) on convergence in Pennsylvania German and in Bolonyai’s research (1999) on Hungarian attrition. Similarly in (3), leben mit ‘to live with someone in an intimate relationship’ is used in a context where wohnen bei ‘to reside at someone’s house temporarily’ would be expected. What is important in examples (2) and (3) is that these apparent attrition effects are manifested at the level of content morphemes. In the domain of morphology, one finds hypotheses that predict, for example, loss of case-marking, analogical levelling, simplification, and outright loss of morphologically marked categories (cf. Andersen 1982; Hagen & de Bot 1990). However, recent studies that offer quantitative results for nontarget-like production of functional morphemes indicate that such disruptions are not a significant aspect of language attrition. For example, in a study of young Croatians living in Australia, Hlavac (2000: 485) shows that 91 percent of all Croatian noun phrases (8,444/9,318) receive target-like case assignments. In addition, Schmid (2002) notes that target-like case, gender, and plural assignment in German noun phrases among her German Jews who immigrated to Anglophone countries is the order of the day. The following section presents a methodology for analysing language change due to attrition that offers a way to integrate these sorts of observations under a single, coherent language production hypothesis. From this hypothesis, I will derive a prediction about language attrition that specifies which categories of morphemes are more or less vulnerable to attrition.
. A methodology for analysing language attrition data The argument developed here begins with the hypothesis that universal aspects of the way in which the mental lexicon is organised determine how different classes of morphemes are activated and projected in language production. This
Steven Gross
Conceptual level
SPEAKERS’ INTENTIONS
Semantic/pragmatic feature bundles
Lemma level
CONTENT MORPHEME LEMMAS (Directly elected)
EARLY SYSTEM MORPHEME LEMMAS (Indirectly elected)
Directions to formulator
Functional level
Positional level
FORMULATOR
LATE SYSTEM MORPHEME LEMMAS (Structurally assigned)
PHONETIC/SURFACE FORMS
Figure 1. A model of language production (Adapted from Myers-Scotton & Jake 2001: 88)
model of lexical organisation presupposes a hierarchy of processing procedures portrayed in Figure 1. The basic outlines of this model follow Levelt (1989) with important modifications developed in Myers-Scotton and Jake (2001). . A language production model Under this model of language production, lemmas that support content morphemes – those morphemes that participate in the thematic grid (nouns and
Attrition and contact linguistics
verbs, for example) – are the only lemmas in the mental lexicon that have a direct link to speakers’ intentional states, to the conceptual level.1 These morphemes are conceptually activated. In addition to conceptually activated content morphemes, there are two other categories of morphemes relevant to this article, early system morphemes and late system morphemes. The first of these, early system morphemes, is also conceptually activated at the lemma level; however, early system morphemes are indirectly elected by content morphemes. According to Bock and Levelt (1994: 953), these morphemes “do not correspond to lexical concepts”; rather, their function is to enrich the semantic and pragmatic content of the content morphemes that elect them. Examples of these types of morphemes are English determiners and plural s. The final category of morphemes is late system morphemes. These late system morphemes are activated at the level of the Formulator (the functional level); they are structurally assigned to fill grammatical slots in the grammatical frame. Levelt (1989: 11) explains that the Formulator’s function is to “translate a conceptual structure into a linguistic structure”. More specifically, lemmas selected by semantic/pragmatic feature bundles send directions to the Formulator to turn on language-specific morphosyntactic procedures for assembling larger constituents. Thus, operations at this level are automatic, blindly carried out by the Formulator, and not subject directly to speakers’ intentions. Late system morphemes are ‘late’ in the sense that they are not fully salient in language production until they are called upon to assemble larger constituents. Included among late system morphemes are subject-verb agreement markers and the case markers we find in languages like German, that is, purely grammatical elements. . Abstract lexical structure According to Myers-Scotton (2002: 194), lemmas contain three levels of abstract lexical structure: (1) lexical-conceptual structure (semantic and pragmatic information); (2) predicate-argument structure (mapping thematic roles to grammatical relations); and (3) morphological realisation patterns (surface realisations of grammatical structure). Now, in certain kinds of language contact situations, e.g. language attrition, convergence, Creole formation, and L2 acquisition, speakers appear to be producing bilingual CPs (projection of the complementiser) in the sense that some CPs include morphosyntactic structure from more than one linguistic source, despite the fact that all the surface forms come from a single language.
Steven Gross
Recall example (2) in which the speaker uses the German verb wissen in an English-like manner to express the notion of ‘knowing a person’. Considering this example in terms of abstract lexical structure, we might say that the occurrence of wissen reflects the influence of English at two levels of abstract structure: lexical-conceptual structure and predicate-argument structure. Both the core meaning of the verb and its semantic selectional properties come from English; yet, the morphological realisation patterns come from German. (Note that the lexical verb occurs in its characteristic clause-final position and has German past participle morphology.) In other words, the lemma entry for the German verb wissen seems to contain aspects of abstract lexical structure from two linguistic sources, German and English. This creation of a bilingual lemma is what Myers-Scotton (1998) calls a composite Matrix Language. So, what are the necessary contributing factors for such reconfigurations of L1 lemma entries? The first occurs when there is a competition between languages for the role of the Matrix Language, as when speakers are shifting from one language to another as their sociolinguistically dominant language (MyersScotton 2002: 22). For example, all but one of the individuals who participated in my study reported that although they were monolingual German speakers at the time they immigrated to the United States, they now feel that English is their stronger language. A second answer to this question has to do with frequent codeswitching. All the bilingual participants in this study are heavy codeswitchers. One widely accepted fact about codeswitching is that content morphemes, mostly nouns and verbs, are the most frequently switched grammatical category. In contrast, Embedded Language late system morphemes are noticeably absent in codeswitching (cf. Myers-Scotton 1993). I want to suggest that codeswitching is the ‘foot in the door’ that makes it possible for aspects of L2 abstract lexical structure to intrude upon L1 lemma entries. . Predicting language attrition outcomes The language production hypothesis presented in Section 3.1 has clear consequences for how we analyse bilingual data. One of the consequences is that it leads to a more precise, more highly articulated morpheme classification model with which to analyse the data. Instead of a large category of functional morphemes, this model provides a heuristic to distinguish different types of functional morphemes based upon the way in which they are projected during language production. More importantly, this hypothesis offers us a way to derive a general, testable prediction regarding bilingual production.
Attrition and contact linguistics
Language attrition prediction content morphemes > early system morphemes > late system morphemes where ‘>’ = are more vulnerable to attrition than
The question at this point is why should content morphemes be more vulnerable to attrition than early and late system morphemes? The answer lies in the architecture of the mental lexicon and the asymmetry in how different types of morphemes are activated. Content morphemes are conceptually activated, and lemma entries that support content morphemes are more richly articulated than system morpheme lemmas; that is, they convey more semantic information. Simply put, when an element is conceptually activated, it is salient as soon as the speaker’s intentions are encoded as language. Early system morphemes are also conceptually activated, but only indirectly by content morphemes. In contrast, the projection of late system morphemes depends on implicit procedural knowledge. This dissociation between the explicit declarative knowledge involved in projecting content morphemes and the implicit procedural knowledge required to project late system morphemes has been supported by a wide range of evidence from psycholinguistic research, neurological cases, and functional imaging studies (Damasio 1992; Squire, Knowlton & Musen 1993; Stromswold et al. 1996; Ullman et al. 1997, among others). Ullman (2001) argues that the lexical memory (i.e. declarative knowledge) involved in learning arbitrary conceptual/semantic relations is relatively accessible to consciousness, whereas procedural knowledge tends to be encapsulated so that it is relatively inaccessible to conscious examination and control. Thus, we would predict that procedural knowledge, once acquired, should be difficult to ‘unlearn’ simply because it is less accessible to conscious control.
. The study The following production data come from a larger study on L1 attrition (Gross 2000a) in six German immigrants, between the ages of sixty-one and seventyfour, who have been living in the United States for forty to fifty years. The six individuals recorded are all speakers of Bavarian German and Schmiedshau German.2 All six speakers were at least sixteen years old when they immigrated to the United States, and their first significant contact with English came after their arrival in the United States. None had received any formal instruction in English, yet today all these individuals use English as their primary means of
Steven Gross
communication in the home and the community. All but one report that they are now English-dominant. . The interviews Initial contact with these six participants was made over the telephone. During the telephone conversation, I indicated that I was interested in interviewing German immigrants in the United States, but I tried to explain as little as possible about the particulars of the study. The participants assumed that my major interest was regarding the circumstances surrounding their immigration and their experiences immediately after arriving in the United States. The resulting interviews, which were conducted entirely in the participants’ regional dialects, generated a body of data that included approximately eight hours of tape-recorded one-on-one conversations in the participants’ homes. In these interviews, I initiated the discourse by asking questions about the individuals’ life histories. Each conversation generally covered the same topics – the participants’ experiences in Europe prior to emigrating, the circumstances surrounding their decision to come to America, and the difficulties they encountered in their newly adopted homeland. Most of the participants vividly remembered their experiences during and shortly after World War II; thus, many of them spent a great deal of time discussing this aspect of their lives. The main agenda following my initiation of the first topic was to allow the participants to control topic selection and the flow of the conversation. All participants seemed eager to tell me about their life histories, and there are many sections in the recordings in which the participants speak uninterrupted for long periods of time. Reflecting on the participants’ engagement with and control of the topic, it appears that they were genuinely unconcerned with the staged research setting. . Data analysis In order to draw valid conclusions about the development of any linguistic variety, one must measure the linguistic data against a reference that represents the input to the system. Therefore, when analysing data from nonstandard dialects, it is entirely inappropriate to use the Standard variety as the basis for comparison. Thus, for this study, the standard of comparison was either Bavarian or Schmiedshau German, not Standard German. After conducting the interviews, the corpus was transcribed and analysed, and all forms that were judged to be deviant in some way were marked and pre-
Attrition and contact linguistics
sented in their appropriate context to one of three native speakers of Bavarian German on temporary assignment in the United States for a German-owned company. If the outside judge agreed that a given form was non-native like, it was included as an inaccurately produced form. Inaccurate production rates were then calculated as a percentage of total contexts for the given form.3
. Testing the prediction To test the first claim in my attrition prediction that content morphemes will be most vulnerable to attrition, I examined the production of various idioms, fixed expressions, and metaphorical phrases. Examples (4)–(5) illustrate the effect of the L2 on these sorts of fixed collocations. In both examples, the speakers produced loan translations, i.e. calques, that are lexified according to patterns established in the L2, English. A quantitative analysis of such fixed expressions shows that 29.7 percent (30/102) exhibit such English-like lexification patterns. (4) Calque of drink like a fish a. Er suaf-t wie an Fisch he drink-pres.3sg like a fish ‘He drinks like a fish.’ b. Expected form: wie an Loch ‘like a hole’ (5) Calque of the fixed expression in a way a. In ein-en Weg war-s vielleicht Gott-es Wille in a-masc.dat way was-it perhaps god-gen will ‘In a way, it was perhaps God’s will.’ b. Expected form: in gewisser Hinsicht ‘in a certain respect’
The second part of the language attrition prediction is that early system morphemes will be less vulnerable to attrition than the content morphemes that elect them. To test this, I focused on the production of reflexive markers in inherently reflexive verbs such as sich etwas vorstellen ‘to imagine something’. Recall that early system morphemes do not correspond to lexical concepts; rather, they add semantic and pragmatic information to their content morpheme heads. In example (6), the verb vorstellen fails to elect the expected reflexive marker sich. In this example, English provides the model in the sense that the equivalent English content morpheme imagine requires no reflexive pronoun. The data analysis of these types of verbs shows that 17.3 percent (18/104) fail to occur with their required reflexive markers.
Steven Gross
(6) Use of sich etwas vorstellen ‘imagine something’ without the required reflexive marker i kinn ø durt net kaan ärg-sten Blatz borste-n as I can there not no annoying-sup place imagine-inf as bie Beiting like Whiting ‘I can’t imagine a more annoying place than Whiting.’
Finally, to test the prediction that late system morphemes will be relatively stable under attrition, I quantified the number of inaccurate case assignments in noun phrases. The results of this analysis support my prediction in that 96.8 percent (2,199/2,271) of all required contexts contain accurate case morphology. Figure 2 summarises the results of the comparative analysis of three types of morphemes in terms of inaccurate production: (1) content morphemes, represented by idioms and other fixed expression; (2) early system morphemes represented by reflexive markers in inherently reflexive verbs; (3) late system morphemes, represented by grammatical case. These results support the prediction that content morphemes are most vulnerable to attrition while late system morphemes are the least likely category to undergo attrition. Furthermore, these results seem to agree with many of the observations that other researchers have made regarding the apparent dissociation between production of content morphemes and functional morphemes in 30.0
% INACCURATE
25.0 20.0 15.0 10.0 5.0 0.0
IDIOMS
REFLEXIVE MARKERS
Figure 2. Summary of production inaccuracy
GRAMMATICAL CASE
Attrition and contact linguistics
L1 attrition as well as in cases of language shift and language death (cf. Dorian 1973; Hlavac 2000; Schmid 2002).
. Language attrition in the context of language contact A final issue that I want to explore is what insights can be gained regarding other types of language contact outcomes using this methodology for analysis. According to the language production hypothesis presented in Section 3, we should expect content morphemes to be implicated to a much greater degree than system morphemes in the development of various bilingual linguistic systems. This is precisely what we find in codeswitching. Recall from Section 3.2 that the Embedded Language is generally free to supply content morphemes to the grammatical frame; however, only the Matrix Language can supply late system morphemes (Myers-Scotton 1993). When we look at the grammatical structure of creoles, especially those that have had limited contact with their lexifier languages since Creole formation, we find that their grammatical apparatus is composed mainly of reanalysed lexifier content morphemes. Late system morphemes from the lexifier are noticeably absent (cf. Gross 2000b for an analysis of Berbice Dutch using this framework). When we look at L2 acquisition, we find that late system morphemes are typically late in being accurately acquired. In fact, Wei (2000) has shown a statistically significant contrast in the acquisition of English late system morphemes versus early system morphemes. For example, he reports that plural s (an early system morpheme) is acquired earlier than 3rd person singular s (a late system morpheme). He also reports that determiners that function as early system morphemes are acquired earlier than those that function as late system morphemes. Even in cases of simple borrowing, it should be no surprise to find that languages readily borrow content morphemes (especially nouns), but that languages rarely borrow late system morphemes. This fact has been recognised in much of the earlier literature on language contact. For example, Haugen (1950: 225) comments that “the instances of new inflections actually introduced into wide use in the language are few”.
Steven Gross
. Conclusion In addition to focusing attention on how attrition researchers should approach the task of data collection, it is vitally important that the methodology of data analysis be considered as well. All too often, data analyses have been content with reporting qualitative rather than quantitative results. Thus, it is not surprising to find that much of the literature on language attrition is filled with anecdotal accounts of which categories are lost and which are retained in attrition settings. One of the challenges facing language attrition researchers is the search for answers, for explanations, that underlie the phenomena we are reporting on. To do this, we should be considering insights gained from psycholinguistic research that seeks to uncover the mechanisms behind language production and comprehension. Furthermore, we need to look beyond our own interests in L1 attrition and consider how our findings intersect with findings from other forms of bilingual speech, such as creoles, interlanguage, convergence, and codeswitching. The language production model presented in this article provides us with a cognitively grounded framework that gives us a road map to navigate through the complexities of our language attrition data. The predictions derived from this model are specific and testable, and the data from this study of GermanEnglish bilinguals appear to support these predictions. In addition, research from other types of bilingual outcomes lends additional support to the basic premises underscoring this production model. But whether the model holds up under the scrutiny of further testing remains to be seen. In conclusion, the challenge that faces language attrition researchers is to transcend the level of description and, instead, to offer explanations that are grounded in universal aspects of linguistic organisation and language production that unify the field of contact linguistics.
Acknowledgment The author wishes to thank the reviewers for their helpful comments on an earlier version of this article.
Attrition and contact linguistics
Notes . Lemmas are not the words themselves; rather, they are a repository of declarative knowledge about the morphemes in a language. Lemma entries contain all the semantic, syntactic, and morphological information relevant for constructing a lexical item’s syntactic environment (Levelt 1989: 6). . Schmiedshau German is a variety of German that was formerly spoken in a larger German speech island called the Hauerland, located northeast of Bratislava in the present-day Republic of Slovakia. This area was settled in the fourteenth century largely by Bavarian and Schwabian speakers from southwestern Germany (Kurbel 1980: 18). After Word War II, these ethnic Germans were expelled from the region and formed part of a larger post-war diaspora that included Germans from the Sudetenland and other Eastern European regions. Although the phonology of Schmiedshau German is distinct from Bavarian, the two are morphosyntactically similar. . The author of this study, who is native to the province of Tirol (Austria) and who acquired Bavarian German as an L1, transcribed and analysed the corpus. One of the possible problems regarding the confirmation of non-native like forms by the outside judges is that their judgments were based on written transcripts that reflected the dialectal pronunciation of the subjects. While it may be argued that such transcription conventions may have negatively influenced the judges’ grammaticality assessments, I was careful to inform these judges about the nature of the transcripts they were to receive and insisted that they disregard orthography in making their judgments. Furthermore, most residents of Austria and Bavarian-speaking Germany are accustomed to seeing their variety of German represented orthographically, on signs and in print, throughout the region. Thus, I would argue that an orthographically realistic representation of speech in this case did not adversely influence the judges; rather, it imparted a sense of authenticity to the written transcript.
References Altenberg, E. P. (1991). “Assessing first language vulnerability to attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 189–206). Cambridge: CUP. Andersen, R. W. (1982). “Determining the linguistic attributes of language attrition”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 83–118). Rowley, MA: Newbury House. Bock, K., & Levelt, W. J. M. (1994). “Language production, grammatical encoding”. In M. A. Gernsbacher (Ed.), Handbook of Psycholinguistics (pp. 945–984). San Diego, CA: Academic Press. Bolonyai, A. (1999). The Hidden Dimension of Language Contact: The case of HungarianEnglish bilingual children. Unpublished Doctoral Dissertation, Columbia, SC: University of South Carolina. Bot, K. de (1996). “Language loss”. In H. Goebl, P. H. Nelde, Z. Stary, & W. Wölck (Eds.), Contact Linguistics: An international handbook of contemporary research, Vol. 1 (pp. 579– 585). Berlin: Walter de Gruyter.
Steven Gross
Damasio, A. R. (1992). “Aphasia”. New England Journal of Medicine, 32 (6), 531–539. Dorian, N. C. (1973). “Grammatical change in a dying dialect”. Language, 49, 413–438. Fuller, J. M. (1997). “Pennsylvania Dutch with a Southern Touch”: A theoretical model of language contact and change. Unpublished Doctoral Dissertation, Columbia, SC: University of South Carolina. Gross, S. (2000a). The Role of Abstract Lexical Structure in First Language Attrition: Germans in America. Unpublished Doctoral Dissertation, Columbia, SC: University of South Carolina. Gross, S. (2000b). “When two become one: Creating a composite grammar in creole formation”. International Journal of Bilingualism, 4, 59–80. Hagen, A. M., & Bot, K. de (1990). “Structural loss and levelling in minority languages and dialects”. Sociolinguistica, 4, 136–149. Haugen, E. (1950). “The analysis of linguistic borrowing”. Language, 26, 210–231. Hlavac, J. (2000). Croatian in Melbourne: Lexicon, switching and morphosyntactic features in the speech of second-generation bilinguals. Unpublished Doctoral Dissertation, Melbourne: Monash University. Hupet, M., & Schelstraete, M-A. (1996). “Les effets du vieillissement sur la compréhension et la production de langage”. Bulletin Suisse de Linguistique Appliquée, 66, 153–166. Kaufman, D., & Aronoff, M. (1991). “Morphological disintegration and reconstruction in first language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 175–188). Cambridge: CUP. Kurbel, R. (1980). Schmiedshau: Ein Deutsches Dorf in der Mittelslowakei – Wie es entstand, war und verging. Stuttgart: Hilfsbund Karpatendeutscher Katholiken. Kynette, D., & Kemper, S. (1986). “Aging and the loss of grammatical forms: A crosssectional study of language performance”. Language and Communication, 6, 65–72. Levelt, W. J. M. (1989). Speaking: From intention to articulation. Cambridge, MA: MIT Press. Myers-Scotton, C. (1993). Duelling Languages: Grammatical structure in codeswitching. Oxford: Oxford University Press. Myers-Scotton, C. (1998). “A way to dusty death: The matrix language turnover hypothesis”. In L. A. Grenoble & L. J. Whaley (Eds.), Endangered Languages: Language loss and community response (pp. 289–316). Cambridge: CUP. Myers-Scotton, C. (2002). Contact Linguistics: Bilingual encounters and grammatical outcomes. Oxford: Oxford University Press. Myers-Scotton, C., & Jake, J. L. (2000). “Explaining aspects of codeswitching and their implications.” In J. Nicol (Ed.), One Mind, Two Languages: Bilingual language processing (pp. 84–116). Oxford: Blackwell. Olshtain, E., & Barzilay, M. (1991). “Lexical retrieval difficulties in adult language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 139–150). Cambridge: CUP. Salmons, J. (1994). “Naturalness and morphological change in Texas German”. In N. Berend & K. J. Mattheier (Eds.), Sprachinselforschung: Eine Gedenkschrift für Hugo Jedig (pp. 59–72). Frankfurt: Peter Lang. Schmid, M. S. (2002). First Language Attrition, Use and Maintenance: The case of German Jews in Anglophone countries. Amsterdam and Philadelphia: John Benjamins.
Attrition and contact linguistics
Seliger, H. W. (1991). “Language attrition, reduced redundancy, and creativity”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 227–240). Cambridge: CUP. Seliger, H. W., & Vago, R. M. (1991). “The study of first language attrition: An overview”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 3–15). Cambridge: CUP. Sharwood Smith, M. (1983). “On first language loss in the second language acquirer: Problems of transfer”. In S. Gass & L. Selinker (Eds.), Language Transfer in Language Learning (pp. 222–231). Rowley, MA: Newbury House. Squire, L.R., Knowlton, B., & Musen, G. (1993). “The structure and organisation of memory”. Annual Review of Psychology, 44, 453–495. Stromswold, K., Caplan, D., Albert, N., & Rausch, S. (1996). “Localisation of syntactic comprehension by positron emission tomography”. Brain and Language, 52, 452–473. Ullman, M. T. (2001). “The declarative/procedural model of lexicon and grammar”. Journal of Psycholinguistic Research, 30, 37–69. Ullman, M. T., Corkin, S., Coppola, M., Hickok, G., Growdon, J. H., Koroshetz, W. J., & Pinker, S. (1997). “A neural dissociation within language: Evidence that the mental dictionary is part of declarative memory, and that grammatical rules are processed by the procedural system”. Journal of Cognitive Neuroscience, 9, 266–276. Wei, L. (2000). “Types of morphemes and their implications for second language morpheme acquisition”. International Journal of Bilingualism, 4, 29–43.
No more reductions! To the problem of evaluation of language attrition data Elena Schmitt Southern Connecticut State University
Introduction Researchers in the field of language attrition have been struggling to identify a unifying explanatory framework to the analysis of language attrition data. However, to date we have not been able to agree on the explanatory value of various approaches, descriptions, and accounts of a great number of data sets collected from individuals and groups of multitude of backgrounds, languages, and social variables. This paper attempts to identify and compare the explanatory power of two approaches to the analysis: the reduction/simplification framework on the one hand, and convergence and codeswitching as mechanisms of language attrition on the other. This paper focuses on two sets of language attrition data produced by Russian immigrant children recorded after five and seven years of immigration to the U.S. The purpose of this study is three-fold: 1. To demonstrate that a view of language attrition as the process of simplification and reduction is not sufficiently explanatory; 2. To identify convergence as the mechanism that underlies simplification/reduction; and 3. To discuss convergence and codeswitching as possible tools for explaining attrition. Previous studies on language loss in adults and children under the circumstances of language contact agree (if only implicitly) that attrition is characterised by reduction and simplification of the available L1 lexicon, morphology,
Elena Schmitt
and syntax. A large number of studies (e.g. Håkansson 1995; Pfaff 1991; Polinsky 1997) point out that attrition leads to significant erosion of the lexicon. Others propose that grammar is also vulnerable during language loss. Kaufman (1995), for example, finds that Israeli immigrant children in the U.S. exhibit simplification in their morphology, while Merino (1983) reports that her bilingual Chicano subjects simplify subjunctive, relatives, and past tense forms of Spanish (their L1) under the influence of L2 (English). All of these and other studies attempt to establish patterns in which reduction takes place, provide taxonomies of reduced/simplified forms, and/or describe the causes that lead to reduction/simplification. However, discussing attrition data in terms of reduction and/or simplification does not seem to provide the desired insights into the nature of interaction between the attriting language (L1) and its dominant counterpart (L2) which necessarily takes place in language loss. In order to gain a better understanding of how L2 affects the structure of L1 under attrition, this study attempts to move beyond the simplification/reduction framework. It accounts for the changes in the structures of Russian produced by immigrant children from the point of view of varying levels of participation of abstract lexical structure of English in the grammatical frame set by Russian, the attriting native language of the subjects. In order to constrain the analysis of the data all the examples are discussed within the framework of the Abstract Level model originally proposed by Myers-Scotton and Jake (1995) (see Section 1). Explanation of non-target production stems from the view that convergence and codeswitching are the mechanisms of language loss (Schmitt 2000).
.
Theoretical framework
The dichotomy of Matrix and Embedded languages, as illustrated in example (1), was first delineated in the Matrix Language Frame (MLF) model (MyersScotton 1993a, 1997). The language that sets the grammatical frame of a complementiser phrase (CP) is referred to as the Matrix language (ML); and the other language, which only fills in some of the content morphemes, is referred to as the Embedded language (EL).1 In example (1) most of the morphemes come from Russian, the content morpheme dog, however, comes from English. Thus, Russian is the ML in this CP and English is the EL.
Evaluation of language attrition data
(1) On dolgo laia-l na dog-ov. He long bark-past/masc/sg on dog-gen/pl. ‘He barked at dogs for a long time.’
It is also necessary to distinguish between four types of morphemes that are activated and selected at different times in the course of production (MyersScotton & Jake 2000): content morphemes, and three types of system morphemes identified as early system morphemes, and two types of late system morphemes – bridge and late system morphemes (for a detailed discussion of morpheme types see Gross, this volume). In example (2) Pepe, like, chew, bone and Mashka are content morphemes, -ing is an early system morpheme, -’s is a bridge morpheme, and –s is a late system morpheme or ‘outsider’. (2) Pepe likes chewing on Mashka’s bone.
Following Rappaport and Levin (1988) among others, Myers-Scotton and Jake introduce the idea that the abstract structure that underlies production is not linear, but consists of several levels. The Abstract Level model identifies three levels of abstract lexical structure: lexical-conceptual, predicate-argument structure, and morphological realisation patterns. Each of these structures plays its distinct role in language production. Thus, at the lexical-conceptual level the speaker’s intentions are mapped onto semantic-pragmatic feature bundles and language specific lemmas (adapted from Levelt 1989) are activated. At the level of predicate-argument structure, arguments are selected and thematic relations are mapped onto grammatical structure. For example, an argument with a thematic role of an Agent is mapped onto the subject position, Benefactor to indirect object position, and Patient to direct object. Finally, at the level of morphological-realisation patterns grammatical relations are mapped onto the surface structure. These include case marking, agreement marking, word order, etc. All three levels of abstract lexical structure are present in mono- and bilingual speech. However, the Abstract Level model suggests that abstract lexical structure is modular, i.e. three levels of abstract structure can be split and recombined depending on the accessibility of the information from the languages identified as the ML and the EL. In other words, some parts of the abstract lexical structure may come from the ML while other parts are provided by the EL. For example, in (3) all morphemes are from the ML (Russian). (3) On igra-et v sport. He-nom/sg/masc play-pres/3rd/sg in sport-acc/sg/masc. ‘He plays sports.’
Elena Schmitt
SR: On zanima-et-sia sport-om. He-nom/sg/masc engage-pres/3rd/sg-refl sport-inst/sg/masc. ‘He plays sports.’
However, the lexical conceptual structure of the verb igraet ‘plays’ clearly does not come from Russian, SR (Standard Russian) requires the verb zanimat’sia ‘to go in for’ when no specific game is indicated. English, on the other hand, does not differentiate between the two: the verb play is used both with general sports and specific games, such as football and tennis. Thus, in order to produce this sentence the speaker mapped the lexical-conceptual structure of English onto the Russian grammatical frame. This type of production is referred to as convergence, defined by MyersScotton (1998) as “the use of morphemes from a single linguistic variety, but with parts of their lexical structure coming from another source” (p. 290). Compared to other definitions of convergence (e.g. Beniak, Mougeon, & Valois 1984; Romaine 1995; Thomason 2000; etc.), this view specifically targets the processes of language attrition (as opposed to creolisation, L2 acquisition etc.) where the more dominant language affects the abstract lexical structure of the attriting language, but no reverse influence is observed. Thus, convergence occurs when speakers do not have complete access to the abstract lexical structure of one language or the influence of the other language is so pervasive that speakers want to switch over to that language to fill in the gaps. The surface of a CP, however, is produced in only one language. Convergence is widely exhibited in many data sets (e.g. Bolonyai 1998, 1999; Jake & Myers-Scotton 1997; Myers-Scotton 1998, 2002; Schmitt 2000) as well as in the data presented here. . Subjects and data . Subjects The data for this longitudinal study have been collected from five Russian boys whose families emigrated from the former Soviet Union in the early 1990s. The boys were brought to the U.S. between the ages of 3;8 and 4;2 years of age (average age of four years). The parents of the subjects report that, despite the fact that all the subjects were brought to this country at a very young age, their Russian at the time of the arrival was well-developed: they had full command of the case system, i.e. they were able to produce all the appropriate case endings required by Russian; they had mastered the verb conjugation system, i.e. they were able to agree the
Evaluation of language attrition data
subject with the verb and produce the appropriate verbal inflection required by such agreement; and they dropped pronominal subjects as required by the pro-drop parameter of Russian. In other words, these boys had mastered the Russian grammatical system and were able to construct a standard Russian grammatical frame for any utterance they wished to produce. These reports are supported by findings on Russian language acquisition by young children (Slobin 1985). Therefore, I believe that treating the non-target forms produced by these children as a sign of language attrition and not as a result of incomplete acquisition is justified. It is important to emphasise that children at this age are not necessarily familiar with such complicated structures as subjunctives, conditionals, etc. They may also find it difficult to correctly conjugate irregular verbs such as khotet’ ‘to want’, plakat’ ‘to cry’ and others. Declension of certain irregular nouns may also be off target in the sense that children have not acquired all allophonic differences at this time. Children also produce mainly simple sentences tending to avoid complex constructions, especially if these require conjunctions. However, what is important for this study is that four-year-old children have full access to most of the grammatical structure of the Russian language and are able to use it in accordance with the requirements in order to be successful in expressing their intentions. Upon their arrival in the United States, all five subjects were immediately exposed to English through day care institutions at the Jewish Community Center. They were not provided with any special instruction in the English language; instead, all their peers were monolingual speakers of English. Therefore, in order to survive in the new environment and make friends these children were forced to acquire English and make it their primary language of interaction outside the family. This intensive exposure to English clearly took its toll on the development of Russian, causing its fossilisation and gradual attrition. . Language policy at home The parents of the boys in the study are concerned about their children’s rapid loss of ability to express their communicative intentions in grammatical Russian. However, this has not been the case from the very beginning of their life in the US. As most new immigrants, the parents of these children were aware of the necessity to learn the language of their host country as soon as possible in order to gain access to economic, professional, and social benefits. They also encouraged their children to acquire the target language speedily. Some of
Elena Schmitt
the parents stated during interviews with the researcher that at the beginning of their American life they were so eager to acquire English and so anxious to demonstrate their early success to other members of the family that they started using newly learned English words in their Russian sentences, i.e. they started to codeswitch. It is important to note that they still spoke Russian, i.e. the grammatical frame of all their utterances came from Russian, but English provided some of the content morphemes. While parental codeswitching in itself did not interfere with the parents’ ability to speak standard Russian, it gave their children a signal that using English at home was not only appropriate but also to be emulated. This might further have stimulated the subjects’ desire to acquire English as fast as possible and replace Russian with it. Once the parents of the subjects situated themselves in the new environment and became more secure financially and professionally, they started noticing that their children were not communicating in Russian even when spoken to in their mother tongue. This observation proved rather disturbing, especially to the mothers of these boys because they wished for their children to have full command of both languages, English and Russian. This led to a rather conscious introduction of a Russian-only language policy at home, an approach that was supposed to serve as a method of maintenance of Russian in an adverse linguistic environment, i.e. in the environment where English language and American culture dominate the life of the children. The parents thus tried to speak only Russian to their children and required that the boys also speak only Russian. However, the boys were not always able or willing to respond to their parents in Russian. English was the language of their first choice. After all, it was English that brought them success in finding friends, establishing ties in the American society, and even elevating them in the eyes of the parents who sometimes had to use them as interpreters. These factors along with the lack of motivation to speak Russian led the children to resist using Russian for interactions at home. The parents, however, report being adamant about not speaking English at home and insisting that their children should use Russian. In any case, the parents used only Russian in all their conversations with the children, as well as with other family members and Russian-speaking friends and visitors. The children gradually adapted to this strategy and started using more and more Russian. The parents claim that the best motivation for speaking Russian is the children’s grandparents, whose English remains minimal and who cannot understand and/or communicate with the grandchildren in English. Thus, the combination of factors, which include parental insistence on speaking Russian, grandparents’ inability to communicate in English, meet-
Evaluation of language attrition data
ing other Russian immigrant children who immigrated recently and have not yet acquired English, encourages the subjects to speak Russian at home. However, speaking Russian, i.e. responding to their parents in Russian (the children rarely or never initiate exchanges in Russian) does not imply that the subjects produce standard Russian: they often incorporate English into their speech both overtly and covertly. Overt incorporation of English results in codeswitching; covert use of English results in convergence. In other words, while building a Russian grammatical frame, the children resort to English for various types of ‘aides’: when codeswitching, they employ English content morphemes to assist with the lexical items that are not easily available to them from Russian; when converging, they use some English abstract lexical structures in order to fill in the gaps left by Russian. Their parents remain rather tolerant to this ‘creative’ use of the two languages. They believe that using some form of Russian is better than using none at all. . Data collection The data for this study were collected at two different times. The first collection took place in the summer of 1997 when the subjects’ median age was nine and a half years. The second collection was carried out in the fall of 1999 when their median age was 11;5. By the time of the first data collection all the subjects had lived in the United States for a period between four and five years. The subjects were tape-recorded in their homes while they were playing and/or had conversations with their parents, siblings and/or family friends and other Russian acquaintances. The researcher, who was also a friend of all the families, asked permission to tape-record the children herself. The parents of the subjects gladly agreed, and I visited each family and spent about two to three hours talking to the children in Russian, playing games, reading books, discussing movies, playing computer games. Whenever the children started having a good time playing and talking, I turned on the tape-recorder and recorded their speech. This resulted in 50 to 75 minutes per session of tape-recorded conversations between the children and the researcher. The conversations were natural, in the sense that the children were not forced to answer any specific questions; they were not interviewed in any kind of formal setting. On the contrary, they were in their home environment playing games that they normally played and talking about the subjects that interested them, such as favourite movies, school, friends, collecting various items, etc. Therefore, the tape-recorded conversations can be considered naturally occurring despite the fact that they were recorded by the researcher. After all the recordings
Elena Schmitt
were made, they were transcribed, transliterated into Roman script, glossed for morphological composition, and translated. This material consisting of 889 Complementiser Phrases (CPs), serves as the data for the primary analysis of the children’s speech in Russian. The second collection of the data followed the same procedures and took place in the fall of 1999 in the same families with the same subjects. The resulting corpus yielded 1240 CPs, which were transcribed, transliterated, glossed for morphological structure, and translated. Both sets of data were analysed for the presence of codeswitching, convergence and other signs of language attrition. The non-target forms and structures that were qualified as resulting from convergence were then accounted for on the basis of the Abstract Level model. The results and analyses are presented in the following section.
. Analysis This section of the paper presents an overview of the data, discusses the major changes exhibited by the informants from the first set of data to the second, and compares the explanatory power of reduction/simplification framework to the analysis of non-target forms within the framework of the Abstract Level model. Table 1 compares the amount of bilingual and monolingual speech across the two data sets. Even though the data were not analysed statistically, it is evident from Table 1 that the amount of convergence has shown tendency towards an increase by 7.34% – convergence constituted 25.85% (N = 227/878) of all the data in Table 1. Comparison of bilingual and monolingual CPs across the two data sets Recording
Total Bilingual CPs Monolingual CPs number Convergence Codeswitching Standard Russian Standard od CPs English (tokens) Total (%) Total (%) Total (%) Total (%)
1st recording
878
25.85% (227/878)
16.63% (146/878)
47.03% (413/878)
10.47% (92/878)
2nd recording
1204
33.19% (399/1204)
18.43% (222/1204)
44.93% (541/1204)
4.78% (42/878)
+ 7.34%
+ 1.8%
–2.1 %
–5.69%
Percent change from the 1st to the 2nd recording
Evaluation of language attrition data
1997, but rose to 33.19% (N = 339/1204) in 1999. The amount of codeswitching also showed a tendency to increase at the time of the second recording by 1.8%. These increases were expected and support the view of the data as resulting from language attrition. Interestingly, the percentage of CPs produced in Standard Russian increased by 2% in the second recording and the amount of Standard English went down from 10% in the first recording to 4% in the second recording two years later. These changes point to the fact that in the second recording children have become more skilful in producing seemingly all-Russian speech which has no English on the surface; they also “learned” to incorporate English covertly, i.e. through convergence. Therefore, it seems useful to discuss these data from the point of view of the underlying mechanisms of bilingual production, i.e. convergence and codeswitching, which will provide us with an explanation and even predictions. While both sets of data contain examples of convergence and codeswitching, I would like to focus only on those samples of children’s speech that can be perceived as examples of reduction and/or simplification and be treated within the Abstract Level model. Let us start with the analysis of examples (4), (5), and (6). (4) A chto ty bud-esh’ dela-t’ v And what you-nom/sg/2nd will-2nd/fut/sg do-inf in park-0? park-acc (or nom?)/masc/sg. (SR) A chto ty bud-esh’ dela-t’ v And what you-nom/sg/2nd will-2nd/fut/sg do-inf in park-e? park-prep/masc/sg. ‘And what will you do in the park?’ (Data set one) dlia CIA. (5) On rabota-l He-nom/sg/masc worked-past/sg/masc for CIA. na CIA. (SR) On rabota-l He-nom/sg/masc worked-past/sg/masc for CIA. ‘He worked for the CIA.’ (Data set two) (6) Ia govori-u __russk-iy I-nom/1st/sg speak-1st/pres/sg __Russian-acc/masc/sg iazyk. language-nom (or acc?)/masc/sg.
Elena Schmitt
(SR) Ya govor-iu na russk-om I-nom/1st/sg speak-1st/pres/sg in Russian-prep/masc/sg iazyk-e. language-prep/masc/sg. ‘I speak the Russian language.’ (Data set two).
In example (4) it is evident that the subject is using a non-target case, replacing the intended prepositional case with the nominative or accusative. In (5) the subject uses a non-target preposition in the verb phrase. In (6) preposition is omitted and the case is used in a non-target fashion. The data contain a great deal of such non-target uses. An analysis within simplification/reduction framework invites the conclusion that the grammatical system of Russian used by our subjects’ has undergone simplification and reduction. However, let us look at Table 2 where overall and target-like case usage is compared. It is evident here that speakers tend to use the majority of cases in a targetlike fashion – 91.42% (N = 714/781) in the first recording and 86.97% (N = 908/1041) in the second recording. We can also see that the subjects still possess a full-fledged six-case system of the Russian language and that replacements arise only in a limited number of CPs. This distribution indicates that even after eight to nine years of immigration the informants preserve a relatively stable case system: they still have a command of all six case markers and use them appropriately most of the time in both the first and second recordings. Table 3 provides further evidence for the fact that all cases are still in use. In fact the second recording boasts a slight increase in the proportion of target-
Table 2. Distribution of cases across the two sets of data 1st Recording
2nd Recording
Case
Total Required Contexts
Total Target
% Target
Case
Nominative Genitive Dative Accusative Instrumental Prepositional
403 93 44 170 24 47
401 79 36 153 10 35
99.5 84.95 81.82 90.00 41.67 74.47
Nominative Genitive Dative Accusative Instrumental Prepositional
TOTAL
781
714
91.42
TOTAL
Total Required Contexts
Total Target
% Target
564 151 51 199 35 41
558 102 37 168 16 27
98.94 67.54 72.55 84.42 45.71 65.85
1041
908
87.22
Evaluation of language attrition data
Table 3. Proportional contexts and target-like use of cases 1st Recording
2nd Recording
Case
TOTAL % of all participating cases
% Target Case of all participating cases
TOTAL % of all participating cases
% Target of all participating cases
Nominative
51.60 (403/781)
56.16 (401/714)
Nominative
54.17 (564/1041)
61.45 (558/908)
Genitive
11.91 (93/781)
11.06 (79/714)
Genitive
14.50 (151/1041)
11.23 (102/908)
Dative
5.63 (44/781)
5.04 (36/714)
Dative
4.89 (51/1041)
4.07 (37/908)
Accusative
21.77 (170/781)
21.43 (153/714)
Accusative
19.11 (199/1041)
18.50 (168/908)
Instrumental
3.07 (24/781)
1.40 (10/714)
Instrumental
3.36 (35/1041)
1.76 (16/908)
Prepositional 6.02 (47/781)
4.90 (35/714)
Prepositional 3.93 (41/1041)
2.97 (27/908)
like uses of Nominative, Genitive, and Instrumental cases. Moreover, all six cases are used for replacement. The combination of these facts points to the possibility that morphological case markers do not erode, rather it is the contexts where these markers are required become less clear to the informants. Therefore, reduction and simplification accounts will not provide us with the desired insight into the processes of attrition that are taking place in the language used by the subjects. On the other hand, if we analyse case loss in terms of convergence, we will have to conclude that Russian largely continues to build the grammatical frame and to supply all case markers. But English also participates in the creation of the grammatical structure by projecting some slots for case marking. To make things clear, consider example (7). (7) Ona bud’-et uchitel’nits-a. She-nom/3rd/sg will be-3rd/fut teacher-nom/sg/fem. SR: Ona bud’-et uchitel’nits-ei. She-nom/3rd/sg will be-3rd/fut teacher-instr/sg/fem. ‘She will be a teacher.’
Elena Schmitt
Example (7) illustrates the replacement of the instrumental case by the nominative case to mark the copula complement. In Standard Russian the verbs byt’ ‘be’, stanovit’sja ‘become’ and others still assign the instrumental case when used in the past or future tenses (Timberlake 1993; Comrie et al. 1995). The subjects in this study, however, consistently replace the required instrumental with the nominative case in copula clauses.2 In order to explain why this happens we need to determine how the abstract lexical structures of English and Russian interact at the level of morphological realisation patterns and predicate-argument structure. If the whole CP is built by Russian alone then the slot for the case marker should be projected by the Russian morphological realisation patterns. However, if the structure of the CP is built by both Russian and English, then it is possible that the slot for case is projected by English and therefore cannot be filled by Russian. This explanation would suffice if the noun complement were not marked at all. But the noun uchitel’nitsa ‘teacher’ has a feminine nominative ending. This can be accounted for if we first suggest that it is English which projects the slot for number marking. Since Russian number only occurs as part of a multimorphemic unit that combines number, gender, and case it is possible that -a is not nominative, but rather singular feminine ending. But since number and gender marking never occur in Russian separately from case, I suggest another explanation that may be more plausible. Since we never see any nouns occurring in the data without multimorphemic units that mark case, even if they are non-target like, we can suppose that the Russian morphological realisation patterns are ‘programmed’ to project some slot for a case for each NP. When the environment is familiar to the speaker, a slot for a specific case is projected. However, example (7) indicates that English may affect the type of the slot that is projected onto the morphological realisation patterns due to the lack of overt case system in English. The result of English influence on Russian morphological realisation patterns is uncertainty about case use. Thus, when the context is not clear-cut, a default slot is projected by the Russian morphological realisation patterns. This default slot may be filled with the nominative case, which is one of the most frequent cases used in Russian (see Table 3 above). The default slot is a result of the influence of English abstract-lexical structure at the level of morphological realisation patterns.3 This explanation is also valid for examples (8) and (9) below. In both examples the target case is replaced with a non-target case. In (8) we notice that the abstract lexical structure of the CP is projected by a composite ML that consists of both Russian and English. It is English that projects the lexical-conceptual structure of the preposition v ‘to’ instead of the Standard Russian preposition
Evaluation of language attrition data
k ‘to’ used with animate objects. In both English and Russian prepositions are case-assignors (Vergnaud 1985), though in English case is not assigned overtly. I propose, then, that the slot for case marking in this CP was projected by the combined abstract lexical structure of Russian and English: the ability of English to assign the abstract accusative case and the requirement of Russian to project a slot resulted in the overt non-target accusative marking of the noun podrug-u ‘female friend’. (8) Ona idj-ot v podrug-u. She-nom/sg/3rd goes-3rd/sg/pres to friend-acc/fem/sg. SR: Ona idj-ot k podrug-e. She-nom/sg/3rd goes-3rd/sg/pres to friend-dat/fem/sg. ‘She is going to a friend’s.’
In (9) the marker -u on the word shkolu ‘school’ is accusative instead of the target prepositional. The overt presence of this non-target case marker indicates that an accusative slot was projected by the Russian morphological realisation patterns. However, Russian requires the use of the prepositional case of location in this CP. So why does it not project the slot for the prepositional case? Once again English may have influenced the choice of the case slot. Since English does not specify which case needs to be marked in each particular NP, speakers whose Russian is undergoing attrition select a case based on the involved preposition. Preposition v has two meanings ‘at’ and ‘to’. Preposition v, meaning ‘at’, assigns the prepositional case of location to the following NP. The preposition v, meaning ‘to’, assigns the accusative case of direction to the following NP. If the speaker does not differentiate between the two meanings of the preposition, the case assignment distinctions may also be indistinguishable. If so, the accusative case is selected. eto (9) Samoe plokhoye shtuka, chto u menia est’ v shkol-u, The worst thing, that I have at school-acc/fem/sg, is math. math. eto SR: Samoe plokhoe shtuka, chto u menia est’ v shkol-e, The worst thing, that I have at school-prep/fem/sg, is math. math. ‘The worst thing that I have in school is math.’
Elena Schmitt
So, instead of saying that the number of cases is reduced and/or that the case system is simplified, we actually are being precise in pointing to case replacement strategies and we attempt to demonstrate how the two languages interact within one CP that is superficially monolingual. The theoretical advantage of this position is that it does not only point to the fact that some endings are omitted, but it explains why such replacements take place. Now let us turn to example (10) below. (10) Nu, est’ baseball v heaven? Well, is baseball in heaven? SR:
Nu, est’ baseball v heaven-e? Well, is baseball in heaven-prep/sg/masc. ‘Well, is there baseball in heaven?’
At a first glance, we can simply note that heaven is a codeswitched form and that it lacks the Russian prepositional case ending. This first observation will probably point to inaccessibility or loss of lexicon, and the second observation will drive us to propose a reduction in case system. From the reduction/simplification point of view there is really not much else to say or explain. However, if we treat this example within the framework of the Abstract Level model, we will be able to answer why there is no overt case marker on the switched form. Two analyses are possible. One way is to propose that zero case marking is a result of transferring English morphological realisation patterns onto Russian by the speaker. If so, the lack of a grammatical marker indicates the presence of English zero morpheme. However, the projection of any English (EL) late system morpheme, even if it is a zero case marker, onto the surface of a Russian (ML) frame would violate the MLF model (Myers-Scotton 1997) which prohibits the use of late system morphemes from the EL in mixed constituents. An approach based on the composite grammatical frame provides a more satisfactory result. The lack of the system morphemes in the intended ML in the above examples may result from the lack of the necessary slot for such a morpheme in Russian, but not from the use of the English system of case marking in a Russian grammatical frame by the subjects, as proposed earlier. The question is, then, why is the slot lacking? The answer lies in the ability of the abstract lexical structure to be split and recombined, as proposed by the Abstract Level model. When the speaker brings the English content morpheme heaven into the Russian frame, he also brings along parts of its morphological realisation patterns. These patterns do not contain the slot for case marking.
Evaluation of language attrition data
Under overt CS that does not involve convergence, the abstract structure of the ML (i.e. Russian) would block the participation of abstract structure of the EL. In this case, however, it is the English content morpheme that brings with it slots for late system morphemes. Russian, being the intended ML, fills these slots with the late system morphemes when the sentence is ready to be assembled. But since English projects all of the abstract lexical structure for heaven which does not provide the slot for case marking, Russian case markers cannot surface. This results in zero marking and bare forms. One can link the surface structure of this sentence with the underlying abstract structure: the late system morpheme is activated only when the grammatical frame of the CP is ready to accept it; otherwise, the morpheme does not surface. Therefore, the speaker activates the English lemma for heaven early on. This activated lemma projects the lexical-conceptual structure, predicateargument structure, and morphological realisation patterns. The result is that the speaker is unable to select the appropriate Russian late system morpheme precisely because the presence of English morphological realisation patterns on the noun prevents the Russian morphemes from appearing on the surface. On the other hand, Russian as the ML in this CP prevents English from projecting its late system morphemes. Thus, a composite ML is produced and a bare form results. Thus, the analysis of these examples in terms of convergence provides us with an explanation of what happens when L1 and L2 are in contact. The reduction/simplification framework does not only lack explanation, but is often misleading as far as the processes exhibited in the attriting language are concerned. Furthermore, the reduction and simplification framework does not allow us to uniformly account for instances of language change that involve an increase in number and complexity of Russian conjunctions used by the subjects, an increased the number of complex sentences in the second recording, and a large increase in the number of sentences that do not contain pro-drop. Conversely, an approach to language attrition via convergence allows us to view the majority of the developments in an attriting language systematically. Therefore, I propose that in order to account for the grammatical structures or features that occur in non-target fashion in the speech of our subjects, we need to evaluate how convergence as a mechanism of bilingual production is used. The levels of convergence and types of codeswitching that subjects utilise for expressing their communicative intentions indicate how much or how little attrition is present in a language.
Elena Schmitt
. Conclusion Based on the results of the analysis of these data, I propose that explaining the processes of language attrition involves identification of the mechanisms present, i.e. codeswitching and convergence, and quantification of the structures produced under the guidance of these mechanisms. Discussing simplification and reduction does not provide an insight into the underlying processes of bilingual production since both are only surface manifestations of language loss. Neither simplification nor reduction happen to all inflections, so citing simplification and/or reduction does not explain why they take place in attrition. Moreover, it is difficult to delineate any pattern with which simplification/reduction occurs. On the other hand, a convergence account provides an internally consistent and context specific argument. Under this we clearly see that target late system morphemes are replaced with non-target ones from the same language in specific circumstances, where English affects the context of the Russian target form, making it unclear to the speaker. Russian, then, projects a generic slot and fills it with a default late system morpheme. Overall, when we use a model which treats change under attrition as substitution or replacement and when we identify a mechanism that produces substitution/replacement (i.e. convergence) and point to a source of this substitution/replacement (i.e. dominant L2), then we have a tool that allows us to explain some of the processes occurring in language loss. This approach to accounting for language attrition allows for the following definitions of reduction and simplification. Reduction is one of the manifestations of language attrition that does not affect all inflections equally. Crucially, it is not a major mechanism of language attrition, but rather a possible observable result of convergence at different levels of abstract lexical structure. Depending on the level of the abstract lexical structure where convergence takes place, reduction may be observed in the number of content morphemes, in the number of grammatical markers, in the syntactic system, and in the use of discourse devices. As the example of the Russian case system demonstrates, reduction may be caused by the lack of clear contextual cues where certain cases are required, rather than by the loss of morphological markers themselves. If speech samples containing this type of ‘pseudo reduction’ are not discussed and explained further, the whole interpretation of language attrition data may be skewed. Simplification is another manifestation of language attrition which, like reduction, does not happen to all inflections. Like reduction, it is not a mech-
Evaluation of language attrition data
anism of language loss but a possible result of the interaction between convergence and codeswitching that leads to the production of bare or uninflected forms and other types of non-target forms. Finally, I would like to emphasise that language loss is not the process of reduction of the grammatical system of a language. Rather, language loss appears to be a process of gradual replacement of one or more levels of abstract lexical structure of the L1 with the abstract lexical structure of the L2. This replacement results in the projection initially of some and gradually most slots of the grammatical frame from L2 onto L1. These slots may be filled with morphemes from L1 or L2. I hypothesise that, once syntactically active late system morphemes of L2 start filling the projected position in the grammatical frame of a CP, the ML turnover has taken place and language attrition is evident.
Notes . The use of instrumental case increases in the second recording (see Table 2) in various contexts, but it is consistently replaced with the nominative case in copula clauses. . An additional factor in the nominative case assignment could be that the predicateargument structure of the Russian copula may have also been mapped from English, where the verb ‘to be’ takes an object with the nominative case. . Note that the MLF model refers only to setting the abstract properties of the grammatical frame. The types of morphemes that fill the slots in the grammatical frame are not specified by the MLF model. Therefore, it is necessary to rely on the Abstract Level model, which specifies how languages interact and produce surface level morphemes.
References Beniak, E., Mougeon, R., & Valois, D. (1984). “Sociolinguistic evidence of a possible case of syntactic convergence in Ontarian French.” Journal of the Atlantic Provinces Linguistic Association, 6 (7), 73–88. Bolonyai, A. (1998). “In-between languages: Language shift/maintenance in childhood bilingualism”. International Journal of Bilingualism, 2 (1), 21–43. Bolonyai, A. (1999). The Hidden Dimensions of Language Contact: The case of Hungarian-English bilingual children. Unpublished Doctoral Dissertation, Columbia, SC: University of South Carolina. Comrie, B., Stone, G., & Polinsky, M. (1995). The Russian Language in the Twentieth Century. Oxford: Oxford University Press. Håkansson, G. (1995). “Syntax and morphology in language attrition: A study of five bilingual Swedes.” International Journal of Applied Linguistics, 5 (2), 153–171.
Elena Schmitt
Jake, J. L., & Myers-Scotton, C. (1997). “Codeswitching and compromise strategies: Implications for lexical structure”. International Journal of Bilingualism, 1, 25–39. Kaufman, D. (1995). “Where have all the verbs gone? Autonomy and interaction in attrition.” Southwest Journal of Linguistics, 14 (1–2), 43–66. Levelt, W. (1989). Speaking: From intention to articulation. Cambridge, MA: MIT Press. Merino, B. (1983). “Language loss in bilingual Chicano children”. Journal of Applied Developmental Psychology, 4, 277–294. Myers-Scotton, C. (1993). Duelling Languages: Grammatical structure in codeswitching. Oxford: Clarendon Press. Myers-Scotton, C. (1998). “A way to dusty death: The matrix language turnover hypothesis”. In L. Grenoble & L. Whaley (Eds.), Endangered Languages (pp. 289–316). Cambridge: CUP. Myers-Scotton, C. (2002). Contact Linguistics: Bilingual encounters and grammatical outcomes. Oxford: Oxford University Press. Myers-Scotton, C., & Jake, J. (1995). “Matching lemmas in a bilingual language competence and production model: Evidence from intrasentential code switching”. Linguistics, 33, 981–1024. Myers-Scotton, C., & Jake, J. (1998). “Codeswitching and the nature of lexical entries”. Pluralinguismes, 14, 219–246. Myers-Scotton, C., & Jake, J. (2000). “Four types of morphemes: Evidence from aphasia, codeswitching, and second language acquisition”. Linguistics, 8 (6), 1053–1100. Pfaff, C. (1991). “Turkish in contact with German: Language maintenance and loss among immigrant children in Berlin (West)”. International Journal of the Sociology of Language, 90, 97–129. Polinsky, M. (1997). “American Russian: Language loss meets language acquisition”. In W. Browne, E. Dornisch, N. Khondrashova, & D. Zec (Eds.), Annual Workshop on Formal Approaches to Slavic Linguistics: The Cornell Meeting (pp. 370–406). Ann Harbor, MI: Michigan Slavic. Rappaport, M., & Levin, B. (1988). “What to do with θ Roles”. In W. Wilkins (Ed.), Syntax and Semantics: Thematic relations (pp. 7–36). New York, NY: Academic Press. Romaine, S. (1995). Bilingualism. Oxford: Blackwell. Schmitt, E. (2000). “Overt and covert codeswitching in immigrant children from Russia”. International Journal of Bilingualism, 4 (1), 9–27. Slobin, D. (1985). The Crosslinguistic Study of Language Acquisition. Hillsdale, NJ: Lawrence Erlbaum. Thomason, S. (2000). Contact Languages: An introduction. Washington, D.C.: Georgetown University Press. Timberlake, A. (1993). “Russian.” In B. Comrie & G. Gorbett (Eds.), The Slavonic Languages (pp. 827–886). London: Routledge. Vergnaud, J. R. (1985). Dependences et Niveaux de Representations en Syntaxe. Amsterdam and Philadelphia: John Benjamins.
Language attrition research An annotated bibliography Monika S. Schmid Vrije Universiteit Amsterdam
Introduction This bibliography is based on a total of 242 titles on language attrition (207 articles, 3 MA theses and 32 PhD theses). The selection criteria were rather strict: only studies or publications which dealt with non-pathological intragenerational L1 or L2 language attrition were included, leaving out the large body of work on minority languages, language shift, language death, aphasia etc. Where several publications were written on the basis of essentially the same data and findings, the widest-reaching publication was selected for the overviews presented below, leaving a total of 137 publications (see Table 1). However, all publications are listed in the reference section, including those which were eliminated from the analysis proper. This bibliography first gives an overview of the studies and articles analysed herein, providing a short summary sentence in each case (Table 2 for L1 attrition, Table 3 for L2 attrition, Table 4 for theoretical articles). It then lists the first and second language contexts in which attrition has been investigated (Tables 5 and 6). Further breakdowns are presented by the linguistic level chosen for the analysis in each case (Table 7), the theoretical framework within which the study was conducted (Table 8) and the data elicitation technique (Table 9). It should be noted that this bibliography is a revision of my own notes as I have kept them over the past years. While I have tried to be as thorough as possible in revising it for the current volume, and to make it as helpful as I could to other researchers, it seems inevitable in this kind of overview that some publications will be missed. I present my apologies to anyone I may have overlooked or mis-classified – no offence was intended.
Monika S. Schmid
Table 1. Theoretical and applied studies on L1 and L2 attrition Attriting language
Applied
Theoretical
Total
L1 L2 not specified
73 24
20 14 7
93 37 7
Total
96
41
137
1 6
English
Russian
Hebrew
French Italian
Russian
Hungarian English
Andrews 1998
Armon-Lotem 2000
Ben-Rafael 2001 Bettoni 1991
Bode 1996
Bolonyai & Dutkova-Cope 2002 Bolonyai 1999 de Bot & Clyne 1994 de Bot, Gommans & Rossing 1991 Brewer Bomar 1982
German
English
Brons-Albert 1993
Brown 2001
Italian
Dutch
English
Spanish
5
20
2
English 6 English 45 French 30
Czech Dutch Dutch
Hebrew 56 English 20
English 13
English 12
English 2 English 76 English 2
German Dutch Spanish
n
Altenberg 1991 Ammerlaan 1996 Anderson 2001
L2
L1
Study
n
n
n
n n n
n
n
y n
y
n
n y n
No evidence found for the regression model, on the contrary. “Perceptual saliency and sociolinguistic factors ultimately determined the direction and extent of the L2 language intrusions.” Errors are restricted to the lexicon (interference from L2), no such phenomena in morphosyntax (p. 101). There was no correlation between L2 proficiency and L1 attrition or between time and attrition, but attitude did correlate with attrition.
Attrition is governed by the structure of the mental lexicon. No further attrition since initial tests in 1972 (p. 17). Amount of contact and time matter, but time matters only when there is few contact (p. 94).
Attrition is most likely to occur where L1 and L2 are similar (p. 204). Lexical access problems are temporal (p. 210). Semantic relevancy and frequency of use do impact the relative strength of lexemes and morphological markers. In particular, more formal language features, such as person and number, will be more susceptible to loss (p. 398). Russian emigrants are rapidly assimilating into English society, prospects for language maintenance [. . .] appear quite slim (p. 158). After 2 years in the USA, at the final stages of bilingual development when L2 English becomes dominant, there is a loss of resumptive pronouns and an unacceptable use of stranded prepositions in attrited L1 Hebrew. This loss depends on what the constituent resumptive pronoun refers to (p. 64). Franbreu develops by both attrition and enrichment through contact. Attrition of nominal and pronominal morphology is dependent on initial levels of proficiency, but verbal morphology is attrited across all speakers (p. 380). Attrition of childhood language proficiency is neurobiologically conditioned; acquisition per se does not prevent attrition, knowledge has to be reorganised and stabilised, “a process which seems to take time and which ceases around the age of 12” (p. 50). Attrition is governed by the structure of the mental lexicon.
CG Findings
Table 2. Applied studies of L1 attrition in an L2 context (CG = Control group)
Language attrition bibliography
French
English
Spanish
German
Turkish
Grosjean & Py 1991 Gross 2000
Gürel 2002a
Turkish
Dutch
Russian
Finnish
Huls & van de Mond 1992
Hulsen 2000
Isurin 2000a
Jarvis 2003
Hakuta & Spanish d’Andrea 1992 Hiller-Foti 1985 German Hirvonen 1995 Finnish
Håkansson 1995 Swedish
Giesbers 1997
Moroccan Dutch 26 Arabic Dutch Indonesian 1
English
English
English
Dutch
1
1
30
11
n
n
n
n
n n
Italian English
12 4
n
n
5 English/ French English 308
English
n
n
y
n
Distinct pattern of convergence vs. preservation that cannot be convincingly explained or understood in terms of models which privilege and thus preserve lexically contrastive information over what might be called low level of allophonic detail. Perception is generally unaffected by attrition, but production skills show some difference to control group, e.g. levelling of allomorphic variation. This case study found surprisingly little overall variation, most of what was there was on the lexical level (p. 172). Most attrition effects seem due to language contact (p. 176). Strikingly, most phonological deviations were on the suprasegmental level (p. 169). Some contact structures are judged as acceptable by attriters, but Standard Spanish structures have higher acceptability (p. 59). Attrition is conditioned by matrix language model, first steps towards matrix language turnover are visible in data. Evidence for the binding properties of Turkish overt pronouns being replaced by those of English in L1 attrition (p. 179), no length of stay effects (p. 170). Syntax (verb second) is unaffected by attrition, morphology (gender marking) is strongly affected (p. 174). The proficiency in Spanish of high school students of Mexican descent is related to the age at which they first started speaking English (p. 81). This is a report on work in progress (no follow-up has been published, however). Language attrition is systematic, there is very little loss in phonology, but some in morphology (case marking) (p. 192). Turkish was found to be the dominant language in the family with a shorter duration of residence in the Netherlands (5 years), whereas Dutch was the dominant language in the family with a longer duration of residence (15 years) (p. 113). Productive skills decrease within generations, receptive skills are largely unchanged. Difficulties encountered are retrieval problems, not ‘loss’ as in ‘erased from memory’ (p. 188). L1 loss is determined by L2 interference, semantic overlap is a mechanism of L1 loss and is noticeable with high amounts of L2 learning (p. 164). L2 effects in certain limited areas of grammar, and more broadly in areas of lexicosemantics and general idiom. Crucially, all the L2-influenced deviant structures seem to be item-specific (i.e. occurring with only a few specific morphemes, words and phrases), and do not appear to affect whole systems of grammar (p. 99).
CG Findings
6 y (+3 CG) 24 y
15
2
El Aissati 1997
English
French
n
Bullock & Gerfen (forthc.)
L2
L1
Study
Monika S. Schmid
English
Dutch
Yiddish
English Portuguese 5 Spanish English 39
English
Major 1992 Montrul 2002
Olshtain & Barzilay 1991
Hebrew
English
15
13
159
Levine 2000
Finnish
Russian
30
Leisiö 2001
Hebrew
Russian
y
y y
n
n
y
n
y
60
German English/ French Finnish English
1
Lexicon more affected by attrition than morphosyntax. Similarity between languages determining for attrition. Variation within and between groups due to sociocultural factors. The speech samples show no increase or decrease in Finnish language development over the course of 10 months. Several areas of Finnish grammar were partially mastered but, overall, the findings indicated that the language development in the child’s Finnish stagnated (p. 307). Incorrect L1 collocations that were modelled on L2 were not recognised by the immigrants in about 40% of the cases (p. 25). Lexical richness in free written speech declines over years of residence in L2 context (p. 29). Attrition is determined by social factors and linguistic similarities between the systems, similarity is conductive to change (p. 248). Attempts to set out a framework which can distinguish between attrition and incomplete acquisition. Mutual interaction of L1 and L2 can affect L1 phonetics (p. 204). The earlier the age of onset of bilingualism and the more intense the exposure to the sociolinguistically dominant language, the more incomplete the adult grammar may turn out to be, suggesting that an early age of onset might be a necessary but not a sufficient condition for convergent language acquisition (p. 61). Attrition leads to problems in vocabulary retrieval (p. 150).
n n
English
57 1
Hebrew Hebrew
Hebrew
English English
Language contact does not seem to influence loss (p. 94), but education appears to be an important factor (p. 95). Overall surprisingly little attrition found. Very little loss found (p. 146).
No correlation between attitudes and language choice (p. 170).
Findings
L2 loss confirms regression hypothesis, L1 loss does not. Little overall attrition found (de Bot, p.c.). Attrition of L1 is an interplay between two languages which results in the creation of a productive interlingual system that produces hybrid morphological innovations. L1 proficiency declines, resulting in a paucity of lexical diversity (p. 415). Idiosyncratic system emerges (p. 187).
n
n
n
CG
not n spec. 1 n
1
30
250
n
Laufer 2003
Kravin 1992
Kaufman 1998 Kaufman & Aronoff 1991 Köpke 1999
Dutch
Italian
German Dutch
Dutch
Italian
Jaspaert & Kroon 1988 Jaspaert & Kroon 1989 Jaspaert & Kroon 1992 Jordens et al. 1986 Kaufman 1992
L2
L1
Study
Language attrition bibliography
Greek
Russian English English Spanish
English Spanish
Spanish Swiss French Dutch English
Pelc 2001
Polinsky 1994 Porte 1999
Porte 2003
Py 1986
Portuguese 2
35 45 5
Schmid 2002 German English Schmid (forthc.) German English Schmitt 2001 Russian English
Dutch
1 97
English English
Polish Dutch
SchoenmakersKlein Gunnewiek 1997
n
n n
y
n
n
n
n y n
n n
n
L2 influence on L1 takes place in the lexicon, semantics and morphosyntax of Russian L2 users of English who learned their L2 in late childhood or adulthood (p. 57). The high acceptance of ungrammatical constructions suggests that alteration in the monolingual configuration has taken place (p. 623). Attriting group performs significantly different from control group on most tasks, age at arrival being one of the most significant factors, but length of stay is also significant (p. 128). Attrition phenomena on different levels of the linguistic system correlate with each other (p. 274). The self-evaluations suggest that the resident native-speaker teacher’s L1 is not a stable system, but rather a changeable one (p. 33). Despite instances of code-manipulation (particularly code-mixing) there is no evidence that the L1 has suffered significant attrition (p.117). The internal restructuring of the system is due in part to input from other migrants, which also contains numerous and varied deviances from L1 norm (p. 171). Amazingly quick shift from Dutch to Afrikaans or English, with patterns of linguistic interference and language change reminiscent of the early stages in the development of Afrikaans (p. 232). Morphological processing modes partly determine whether an inflectional pattern is liable to attrite (p. 264). Overview of methodological and practical issues in designing research. Attempts to develop a comprehensive model for analysing code-switching and borrowing (p. 159). Persecution most important extralinguistic factor (p. 189). Proficiency data can reveal a more accurate picture of attrition than errors. Language loss is a gradual process of replacing one or more levels of abstract lexical structure of the first language with the abstract lexical structure of the second language. This replacement results in the projection initially of some and gradually most slots of the grammatical frame from L2 onto L1 (p. 243). Relatively little loss found, where conceptualisations were adapted it is not always clear whether this is in the direction of the L2. Compared with Jaspaert & Kroon’s findings on L1 Dutch/L2 English attrition, there seems to be even less L1 loss among Dutch immigrants in Brazil than in the USA although there is less typological distance between Dutch and English than between Dutch and Portuguese (p. 117).
CG Findings
not n spec. 260 n
3
30 52
57
9
English
8
30
n
Finnish Estonian
Riionheimo 1998 Ronowicz 1999 Schatz 1989
Raidt 1997
Greek
Pelc 1998
English
Russian English
Pavlenko 2003
L2
L1
Study
Monika S. Schmid
Dutch
English
English
Spanish
1
1 86
English
English
Spanish
Russian
Spanish
Hungarian Hebrew German English
Turkish
Toribio 2001
Turian & Altenberg 1991 Uribe de Kellett 2002 Vago 1991 Waas 1996
Ya˘gmur 1997
English
English
English
Spanish
40
1
2
8
89
English
Danish
Søndergaard 1996 Toribio 2000
45
Dutch
6
Soesman 1997
Hebrew
1 2
Hebrew Dutch
English English
21
40
Portuguese Dutch/ French
n
SchoenmakersKlein Gunnewiek 1998 Seliger 1991 Sharwood Smith 1983a Sharwood Smith 1983b Silva-Corvalán 1991
L2
L1
Study
y
n n
n
n
n
n
n
n
n
y
n n
y
Documents a very successful case of reactivation of the structure of an attrited language in a Spanish child within one month. Attrition is governed by rule simplification, reordering, loss and lexical restructuring (p. 249). Sociodemographic variables have significant influence, but L1 attrition in an L2 environment appears inevitable overall (p. 171). EVT does not provide a sufficiently explanatory framework for attrition (p. 100f.).
Code-switching should not be considered a necessary precursor or resultant condition for minority language attrition (p. 185). Differences between proficient and attrited speech productions may lie outside the scope of a computational system, target-deviant forms may not accurately reflect non-native-like competence (p. 776). The same compensatory strategies are used in L1 attrition and L2 acquisition (p. 216).
Attrition is not unlike acquisition (p. 239). Acquisition could be called convergent competence change and loss divergent competence change. Attrition is governed by relative simplicity of processing of specific features/structures (p. 225f.). Markedness hierarchy proposed by Muysken seems to be valid as a predictor of order of disappearance in subordinate contact languages, such that the most marked tense disappears first (p. 165). Confirms the findings from de Bot et al. (1991) on interaction of time and contact factors. “This study revealed that immigrants who use their L1 extensively had faster response times than immigrants who do not, and that immigrants who had emigrated a relatively short time ago responded faster than those who had done so a long time ago.”(p. 190). Attrition is governed by attitudes, a feeling of ‘Danishness’ being the main factor (p. 554).
All tasks suggest that there is little or no loss (p. 163).
CG Findings
Language attrition bibliography
English
French Japanese
Gonzalez Mon- Spanish cada 1996
Grendel 1993 Hansen 1999
n
200 n 30 n
?
89 n
English
Moorcroft & Gardner 1987
French
6 y
Japanese English
McCormack 2001
Hayashi 1999
Japanese Micronesian 25 n languages Hedgcock 1991 English Spanish 22 n Kuhberg 1992 Turkish German 2 n
Dutch English
French
English
Gardner 1985
79 n
40 n 36 n
French English
Significantly more speakers decreased than increased in accuracy. Age of return strongest factor contributing to L2 maintenance, participants who were 9 to 10 when returned showed greatest individual differences. There may be a critical threshold for language attrition around the age of 9. Those who try to maintain their L2 skills are relatively successful, whereas those who do not do so experience considerable language loss.(p. 537). EFL teachers working in public schools report decrease in language skills, EFL teachers in private language centers report increase. Perceived level of proficiency showed correlation with standardised tests, confirming the reliability of self-assessment among adults. Orthographical and semantic knowledge are unaffected (p. 128ff.). The longer the exposure to L2, the better the retention (p. 147). The findings show an attrition sequence that is congruent with the predictions of the regression theory (p. 149). The findings are consistent with those reported by research on the progression and regression of Japanese negation (p. 166). Predictions made on the basis of the regression hypothesis are not borne out (p. 50). A great deal of attrition features found on all linguistic levels, code-switching is the developmentally most relevant feature in the attrition process (p. 148). Most test subjects failed to maintain their initial level of knowledge of the English reflexive binding system, but attrition data reflect a deviant knowledge of grammar that cannot be accounted for by L1 transfer alone. Oral proficiency declines over a short period of time, but decline is to be found more on the grammatical level than on fluency or vocabulary production (p. 338f.)
A very large amount of information can survive with minimal reharsal (p. 114). Loss is largely dependent on age and literacy, younger children may reach a ‘freezing point’ (p. 232). Significant decrease in the total number of words produced after 9 months (p. 138).
CG Findings
773 y 30 n
n
2 n
Spanish English
L2
Portuguese
L1
Bahrick 1984 English Berman & Ol- Hebrew shtain 1983 Cohen 1989 English/ Hebrew Dugas 2000 English Fujita 2002 Japanese
Study
Table 3. Applied studies of L2 attrition in an L1 attrition context
Monika S. Schmid
Irish
Murtagh 2003
Japanese English
French
Russell 1999 English Tomiyama 1999 Japanese
Weltens 1988
Yoshitomi 1999 Japanese
English
English
Japanese
Dutch
French
ReetzKurashige 1999
7 n
4 n
150 n
20 n 1 n
18 n
60 n
Reversal does exist but is limited in nature, most likely to occur in cases where the criterion variable is a linguistic feature that is highly marked in the attrited language, in contexts where there is a severe reduction in the use of the second language, and in cases where the attriters may lack stable competence of that feature prior to attrition onset (p. 164). Subjects failed to improve upon the communicative skills which they had acquired prior to and during their immersion experience in the target culture, even though they continued to take courses in French during the time interval. Instead most of their communicative skills began to erode during this period. There is a pattern of morpheme loss occurring before reduction in complexity, which in turn occurs before syntactic errors. Decline in vocabulary diversity, reduced vocabulary, ‘downshift’ in verb tense, selection of simpler forms particularly in forming or avoiding negation. Initial proficiency scores are the best predictor of retention. Vocabulary size decreased, less discourse produced, higher ratio of English items (p. 125). Attrition first manifested as lexical retrieval difficulty. Some indications of attrition in morphology and syntax were present, but phonology and receptive lexicon remained quite robust (p. 74). Substantial attrition found only in self-assessments and morpho-syntactic tests, there is no interaction between training level and non-use and attrition: subjects lose a fixed amount of knowledge independent of original level (p. 92). General receptive proficiency in French is clearly not subject to attrition after four years of non-use, whereas grammar – and to some degree also vocabulary – clearly is (p. 95). Overall accuracy declined significantly implying that, while each linguistic sub-skill erodes only slightly, the returnees’ overall command of English does regress. Language attrition seemed to have taken place while the returnees had little or no opportunities to interact with a native speaker on an individual basis (p. 90).
While L2 speakers of Irish report a self-perceived decline in spoken Irish 18 months after leaving school, tests found no significant change over this period (p. 158f.). There is a critical threshold for preventing L2 attrition at the Advanced High level of the Oral Proficiency Interview scale (p. 188). Non-users of Spanish suffered attrition, measured by gaps, pauses, repetitions etc.
CG Findings
95 n
n
Spanish 4 y English/ Ghanaian lg. Hebrew English 11– n 18
Japanese
English
L2
Raffaldini 1988 English
Olshtain 1989
Nakuma 1997
Nagasawa 1996 English
L1
Study
Language attrition bibliography
Summary
Theoretical outline of how attrition research should be conducted. Regression: Language acquisition as model for language loss. Neurolinguistics, aphasia in children and adults. Outline of research program in the Netherlands, first instance of taxonomy of attrition according to attriting language and language of the environment. de Bot & Weltens 1991 Overview article discussing Jakobson’s regression hypothesis and different types of language problems in the elderly. de Bot 1996 Handbook article on language loss. de Bot 1998 Outline of language loss problems from the framework of psycholinguistic language production models. Clark 1982 Theoretical outline of what instruments to measure language attrition should look like. Clyne & Pauwels 1997 Sociodemographic overview of the situation of Dutch in Australia. Clyne 1992 Linguistic and sociolinguistic aspects of language contact, maintenance and loss: towards a multifacet theory. Cohen 1986 Principles governing the forgetting of foreign-language vocabulary. de Vries & de Vries 1997 Sociodemographic overview of the situation of Dutch in Canada, based on census data: Dutch immigrants to Canada have very low levels of language maintenance in comparison with other immigrant groups (p. 135). van Els 1986 Overview of research, usually taken to be first instance of v. Els taxonomy (s. Köpke & Schmid, this volume); this however was first proposed in de Bot & Weltens 1985. Freed 1982 Establishes background for language loss studies, contains overview of findings from related areas. Gardner 1982 Discussion of social factors for language attrition, strong focus on attitudes and motivation. Ginsberg 1986 Methodology of the Language Skills Attrition Project (LSAP), containing mathematical models for the calculation of attrition rates. Grendel, Weltens & de Overview of studies and findings on lexical L2 attrition. Bot 1993 Hagen & de Bot 1990 Loss of minority languages and dialects, focus on structural aspects and mechanisms of interlanguage – transference, compensation etc. – in language contact situations. Short overview of L2 attrition research, introduction to volume on L2 attrition in Japanese contexts. Hansen & ReetzKurashige 1999 Hansen 2001 Review of psycholinguistic literature on language attrition. Hinskens 1986 Discussion of how linguistic variables should be selected in dialect change and loss, criticism of how this has been done in research. Focus on justification of the selection. Hyltenstam & Stroud 1996 Handbook article on language maintenance and shift, overview of studies. Jaspaert & Kroon 1987 Discussion of research designs. Jaspaert, Kroon & van Important considerations on point of reference, first explicit and detailed discussion of longitudinal studies vs. static Hout 1986 group comparisons and their respective advantages and disadvantages.
Andersen 1982 Berko-Gleason 1982 Berko-Gleason 1993 de Bot & Weltens 1985
Article
Table 4. Theoretical articles
Monika S. Schmid
Does exactly what it says on the label: setting the agenda. Discussion of problems involved in studies which focus on communities or the individual, how to establish the research design and point of reference. Overview of studies done on intergenerational language loss in linguistic enclaves, features typically prone to loss. Argues for the use of spontaneous data, practical problems of analysis notwithstanding, proposes a model for the measuring of pauses and hesitation phenomena. Debate of the concept of communicative competence, model for how to define and measure it in spontaneous speech data. Debate of different models of fossilisation and their impact for L2 attrition studies. Discusses issues in designing and conducting research. Early discussion of language loss and psycholinguistics.
Overview of studies and factors that have been investigated in research of L2 influence on L1. An attempt to account for the data used in Håkansson 1995 on the basis of minimalism and the initial hypothesis of syntax. A discussion, criticism and elaboration of Andersen’s 1982 ‘blueprint for research’. Discussion of L1 attrition within generative frameworks of acquisition and markedness, issue of temporary access problem or total erosion. Seliger 1996 Overview article, pointing out problems of establishing points of reference, argues for the need to include acquisitional factors. Sharwood Smith & A debate of attrition within the competence/performance concept – how can it be established whether a specific case of van Buren 1991 ‘loss’ is competence loss or performance loss? Discussion of attrition and parameter setting. Sharwood Smith 1989 Discussion of the Crosslinguistic Influence Hypothesis versus the regression hypothesis, distinguishes central and peripheral deficits, makes predictions on attrition process on the basis of lossinducing properties. Thompson 1982 Implications of language loss phenomena for language policy. Valdman 1982 Implications of language loss phenomena for foreign language teaching. Weilemar 2003 Contrastive discussion of different generative-based approaches and theory in L1 attrition studies Yoshitomi 1992 Overview of research and theories: regression, critical period. Introduction of neurobiology, brain plasticity.
Nakuma 1997b Nakuma 1998 Oxford 1982 Pan & Berko-Gleason 1986 Pavlenko 2000 Platzack 1998 Preston 1982 Seliger & Vago 1991
Maher 1991 Nakuma 1997a
Lambert & Moore 1986 Lambert 1982 Lambert 1989
Outline of Andersen and Preston model, attempt to link this to situation of Ingrian Finnish (based on author’s intuitions, not actual data). General discussion of the nature of language skills and how to test them.
Dynamic model of multilingualism, complex relationships in multilingualism, time-dependent changes in psycholinguistic system. Overview of the situation of the Hebrew language in the US. Overview of the situation of Dutch the language in emigration.
Jessner 2003
Kaufman 2000 Klatter-Folmer & Kroon 1997 Kokko 1998
Summary
Article
Language attrition bibliography
English English English
Czech Danish Dutch
Bolonyai & Dutkova-Cope 2002 Søndergaard 1996 Ammerlaan 1996; de Bot & Clyne 1994; de Vries & de Vries 1997; Hulsen 2000; Jaspaert & Kroon 1992; Raidt 1997; Schatz 1989 French de Bot, Gommans & Rossing 1991 Hebrew Soesman 1997 Indonesian Giesbers 1997 Portuguese Schoenmakers-Klein Gunnewiek 1997 English Sharwood Smith 1983a Dutch Hebrew Olshtain & Barzilay 1991; Seliger 1991 Brown 2001 Italian Portuguese Major 1992 Spanish Porte 1999; Porte 2003 Finnish English Hirvonen 1995; Jarvis 2003, Kravin 1992 Russian/Estonian Kokko 1998; Riionheimo 1998 French English Bullock & Gerfen forthc. Hebrew Ben-Rafael 2001 German Brons-Albert 1994; Jordens et al. 1986 Dutch English Altenberg 1991; Gross 2000; Köpke 1999; Schmid 2002; Schmid forthc.; Waas 1996 French Köpke 1999 Hiller-Foti 1985 Italian Greek English Pelc 2001 Hebrew English Armon-Lotem 2000; Kaufman 1992; Kaufman 1998; Kaufman & Aronoff 1991 Hungarian English Bolonyai 1999; Bolonyai & Dutkova-Cope 2002 Hebrew Vago 1991 Italian Jaspaert & Kroon 1988; Jaspaert & Kroon 1989 Dutch English Bettoni 1991 El Aissati 1997 Mor. Arabic Dutch Polish English Ronowicz 1999 Portuguese French Schoenmakers-Klein Gunnewiek 1998 Schoenmakers-Klein Gunnewiek 1998 Dutch Russian English Andrews 1998; Bode 1996; Isurin 2000a; Pavlenko 2003; Polinsky 1994; Schmitt 2001; Turian & Altenberg 1991
L2
L1
Table 5. Languages investigated in L1 attrition studies
Monika S. Schmid
Spanish
Leisiö 2001 Laufer 2003 Anderson 2001; Brewer Bomar 1982; Hakuta & d’Andrea 1992; Montrul 2002; Silva-Corvalán 1991; Toribio 2000; Toribio 2001; Uribe de Kellett 2002 Grosjean & Py 1991; Py 1986 Håkansson 1995 Håkansson 1996 Huls 1992 Gürel 2002a; Ya˘gmur 1997 Levine 2000
L2
French French Japanese Spanish Portuguese Ghanaian lg. Spanish Hebrew Portuguese English Irish English Japanese English Micronesian languages Spanish English Turkish German
Dutch English
L1
Gonzalez Moncada 1996 Kuhberg 1992
Grendel 1993; Weltens 1988 Dugas 2000; Gardner 1985; Moorcroft & Gardner 1987; Raffaldini 1988 Hansen 1999; Nagasawa 1996; Russell 1999 Bahrick 1984; Hedgcock 1991; Nakuma 1997 Cohen 1989 Nakuma 1998 Cohen 1990 Berman & Olshtain 1983; Olshtain 1989 Murtagh 2003 Fujita 2002; McCormack 2001; Reetz-Kurashige 1999; Tomiyama 1999; Yoshitomi 1999 Hayashi 1999
Table 6. Languages investigated in L2 attrition studies
Yiddish
Turkish
French English French Dutch English English
Finnish Hebrew English
Swedish
L2
L1
Language attrition bibliography
L1 Andrews 1998; Bullock & Gerfen forthc.; El Aissati 1997; Giesbers 1997; Hirvonen 1995; Major 1992; Vago 1991 L2 Gonzalez Moncada 1996; Tomiyama 1999; Yoshitomi 1999 Morphology L1 Altenberg 1991; Anderson 2001; Bettoni 1991; Bode 1996; Bolonyai & Dutkova-Cope 2002; Bolonyai 1999; Brewer Bomar 1982; Brown 2001; de Bot & Clyne 1994; El Aissati 1997; Giesbers 1997; Gross 2000; Håkansson 1995; Hirvonen 1995; Jarvis 2003; Jordens et al. 1986; Kaufman 1992; Levine 2000; Major 1992; Pavlenko 2003; Raidt 1997; Riionheimo 1998; Schmid 2002; Schmid forthc.; Schmitt 2001; Schoenmakers-Klein Gunnewiek 1997; Schoenmakers-Klein Gunnewiek 1998; Seliger 1991; Silva-Corvalán 1991; Vago 1991 L2 Kuhberg 1992; Moorcroft & Gardner 1987; Nagasawa 1996; Nakuma 1997; Reetz-Kurashige 1999; Tomiyama 1999; Yoshitomi 1999 Semantics L1 Brown 2001; Grosjean & Py 1991; Hiller-Foti 1985; Jaspaert & Kroon 1992; Pavlenko 2003; Pelc 2001; Schoenmakers-Klein Gunnewiek 1998 Syntax L1 Armon-Lotem 2000; Brown 2001; de Bot & Clyne 1994; El Aissati 1997; Giesbers 1997; Grosjean & Py 1991; Gürel 2002a; Håkansson 1995; Jarvis 2003; Köpke 1999; Montrul 2002; Schmid 2002; Schmid forthc.; Schoenmakers-Klein Gunnewiek 1998; Sharwood Smith 1983b; Ya˘gmur 1997 L2 Nakuma 1997; Reetz-Kurashige 1999; Tomiyama 1999; Yoshitomi 1999 L2 Yoshitomi 1999 Lexicon L1 Altenberg 1991; Andrews 1998; Ammerlaan 1996; Ben-Rafael 2001; Brewer Bomar 1982; Brons-Albert 1994; de Bot & Clyne 1994; Giesbers 1997; Hulsen 2000; Jarvis 2003; Kaufman 1992; Kaufman & Aronoff 1991; Köpke 1999; Laufer 2003; Pavlenko 2003; Polinsky 1994; Schoenmakers-Klein Gunnewiek 1997; Soesman 1997; Ya˘gmur 1997 L2 Cohen 1989; Grendel 1993; Reetz-Kurashige 1999; Russell 1999; Tomiyama 1999; Yoshitomi 1999 CodeL1 Ben-Rafael 2001; Brown 2001; Schatz 1989; Toribio 2000; Toribio 2001 switching L2 Berman & Olshtain 1983; Kuhberg 1992; Tomiyama 1999 Attitudes L1 Brown 2001; Jaspaert & Kroon 1988; Raidt 1997; Schmid 2002 Use L1 Brown 2001; El Aissati 1997; Huls 1992; Jaspaert & Kroon 1988; Raidt 1997; Schmid 2002; Soesman 1997 L2 Yoshitomi 1999
Phonology
Linguistic level
Table 7. Linguistic levels in attrition studies
Monika S. Schmid
L1 de Bot, Gommans & Rossing 1991; Giesbers 1997; Hakuta & d’Andrea 1992; Jaspaert & Kroon 1988; Jaspaert & Kroon 1989; Kaufman 1998; Kravin 1992; Leisiö 2001; Turian & Altenberg 1991; Uribe de Kellett 2002; Waas 1996 L2 Bahrick 1984; Fujita 2002; Gardner 1985; Gonzalez Moncada 1996; Hedgcock 1991; Murtagh 2003; Raffaldini 1988; Weltens 1988; Yoshitomi 1999 L1 Hiller-Foti 1985; Py 1986; Sharwood Smith 1983a; Søndergaard 1996 L2 Kuhberg 1992; Nakuma 1997; Olshtain 1989; Tomiyama 1999; Yoshitomi 1999
Interlanguage L1 Altenberg 1991; Brons-Albert 1994; Bullock & Gerfen forthc.; Giesbers 1997; Grosjean & Py 1991; Isurin 2000a; Jarvis 2003; Kaufman 1992; Kaufman & Aronoff 1991; Köpke 1999; Laufer 2003; Leisiö 2001; Major 1992; Pavlenko 2003; 1998; Pelc 2001; Schoenmakers-Klein Gunnewiek 1997; Sharwood Smith 1983a L2 Berman & Olshtain 1983; Kuhberg 1992; Reetz-Kurashige 1999 Simplification L1 Bettoni 1991; Bode 1996; Laufer 2003; Sharwood Smith 1983b; Silva-Corvalán 1991 L2 Reetz-Kurashige 1999 Sociolinguistic L1 Andrews 1998; de Bot, Gommans & Rossing 1991; Hakuta & d’Andrea 1992; Hulsen 2000; Jaspaert & Kroon 1988; Jaspaert & Kroon 1989; Olshtain & Barzilay 1991; Raidt 1997; Schatz 1989; Schmid 2002; Schmid forthc.; Waas 1996; Ya˘gmur 1997 L2 Gardner 1985 UG L1 Gürel 2002a; Håkansson 1995; Montrul 2002; Pelc 2001; Sharwood Smith 1983a; Toribio 2001, Weilemar 2003 L2 McCormack 2001 4-M/Abstract L1 Bolonyai & Dutkova-Cope 2002; Bolonyai 1999; Gross 2000; Schmitt 2001 Level model Psycholing. L1 Ammerlaan 1996; Bode 1996; Hulsen 2000; Jessner 2003; Köpke 1999; Riionheimo 1998; Seliger 1991; Soesman 1997 L2 Hansen 1999; Hayashi 1999; Hedgcock 1991; Obler 1993 Regression L1 Brewer Bomar 1982; Jordens et al. 1986 L2 Berman & Olshtain 1983; Fujita 2002; Hansen 1999; Nagasawa 1996; Olshtain 1989; Reetz-Kurashige 1999; Critical L2 Hansen 1999; Hayashi 1999; Hedgcock 1991; Obler 1993 Threshold
Theoretical framework
Table 8. Theoretical frameworks in attrition research
Other
Overall
Linguistic level
Language attrition bibliography
Competence elicitation tasks:
Truth value judgments
FL-style exam
Fill-in/cloze test
Explicit morphological or syntactic task
Grammaticality/ acceptability judgments
McCormack 2001
L2
L1
Hansen 1999; Hayashi 1999 Ammerlaan 1996; Hakuta & d’Andrea 1992; Jaspaert & Kroon 1989; Montrul 2002; Murtagh 2003 Weltens 1988 Bode 1996; Hakuta & d’Andrea 1992; Jaspaert & Kroon 1989 Bahrick 1984; Berman & Olshtain 1983; Fujita 2002; Gonzalez Moncada 1996; Hayashi 1999; McCormack 2001; Moorcroft & Gardner 1987; Murtagh 2003; Olshtain 1989; Tomiyama 1999 Gürel 2002a; Montrul 2002
Anderson 2001; Ben-Rafael 2001; Bolonyai & Dutkova-Cope 2002; Bolonyai 1999; Brons-Albert 1994; Gross 2000; Jarvis 2003; Kaufman & Aronoff 1991; Kravin 1992; Porte 2003; Schmitt 2001; Seliger 1991; Turian & Altenberg 1991; Uribe de Kellett 2002 Tomiyama 1999; Yoshitomi 1999 Andrews 1998; Ben-Rafael 2001; Bettoni 1991; Brown 2001; Brown 2001; Bullock & Gerfen forthc.; de Bot & Clyne 1994; de Bot, Gommans & Rossing 1991; Giesbers 1997; Hiller-Foti 1985; Huls 1992; Kaufman 1998; Leisiö 2001; Levine 2000; Riionheimo 1998; Schatz 1989; Schmid 2002; Schmid forthc.; Sharwood Smith 1983b; Silva-Corvalán 1991; Søndergaard 1996 Kuhberg 1992; Nakuma 1997; Nagasawa 1996; Raffaldini 1988 Armon-Lotem 2000; Ben-Rafael 2001; El Aissati 1997; Jarvis 2003; Kaufman 1992; Kaufman 1998; Köpke 1999; Montrul 2002; Olshtain & Barzilay 1991; Pavlenko 2003; Schoenmakers-Klein Gunnewiek 1998; Sharwood Smith 1983b; Toribio 2000; Toribio 2001 Cohen 1989; Fujita 2002; Hansen 1999; Hayashi 1999; Hedgcock 1991; Kuhberg 1992; Olshtain 1989; Reetz-Kurashige 1999; Russell 1999; Yoshitomi 1999 Altenberg 1991; de Bot, Gommans & Rossing 1991; Grosjean & Py 1991; Gürel 2002a; Hakuta & d’Andrea 1992; Jarvis 2003; Köpke 1999; Laufer 2003; Pelc 2001; Polinsky 1994; Seliger 1991; Sharwood Smith 1983b McCormack 2001 Altenberg 1991; Bode 1996; El Aissati 1997; Kaufman 1998; Köpke 1999; Sharwood Smith 1983b; Vago 1991; Ya˘gmur 1997
L2 L1 L2 L1 L2
L2 L1
L1
L2
L2 L1
Elicited narrative
L1
L2 L1
Spontaneous speech
Free speech
Interview
Elicitation method
Type of data
Table 9. Elicitation methods
Monika S. Schmid
Other
Sociolinguistic questionnaire
Self-report
Translations Can-Do scales
Verbal fluency Picture naming
Lexical tasks
Self-report and – evaluation
Elicitation method
Type of data
L2
L2 L1
L2 L1 L2 L1
Gardner 1985; Gonzalez Moncada 1996; Weltens 1988 Hakuta & d’Andrea 1992; Jarvis 2003; Porte 1999; Waas 1996 Gardner 1985; Gonzalez Moncada 1996; Murtagh 2003; Weltens 1988 Ammerlaan 1996; Brown 2001; Hulsen 2000; Jaspaert & Kroon 1988; Laufer 2003; Raidt 1997; Schmid 2002; Soesman 1997; Ya˘gmur 1997 Hansen 1999; Hayashi 1999; Murtagh 2003; Yoshitomi 1999 Brown 2001; El Aissati 1997; Grosjean & Py 1991; Gürel 2002a; Håkansson 1995; Hirvonen 1995; Hulsen 2000; Jaspaert & Kroon 1989; Jaspaert & Kroon 1992; Jordens et al. 1986; Kravin 1992; Laufer 2003; Major 1992; Montrul 2002; Polinsky 1994; Raidt 1997; Schoenmakers-Klein Gunnewiek 1997; Schoenmakers-Klein Gunnewiek 1998; Seliger 1991; Sharwood Smith 1983a; Soesman 1997; Toribio 2001 Gonzalez Moncada 1996; Hansen 1999; Hayashi 1999; Raffaldini 1988; Weltens 1988; Yoshitomi 1999
Hakuta & d’Andrea 1992; Waas 1996; Ya˘gmur 1997 Ammerlaan 1996; Hulsen 2000; Isurin 2000a; Schoenmakers-Klein Gunnewiek 1998; Soesman 1997 L2 Dugas 2000 L1 Ben-Rafael 2001; Hirvonen 1995; Polinsky 1994; Schoenmakers-Klein Gunnewiek 1998 L1 El Aissati 1997; Waas 1996; Ya˘gmur 1997
L1 L1
Language attrition bibliography
Monika S. Schmid
References Aissati, A. El, & Schaufeli, A. (1998). “Language maintenance and loss: Evidence from language perception and production”. In G. Extra & L. Verhoeven (Eds.), Bilingualism and Migration (pp. 363–377). Berlin: Mouton de Gruyter. Aissati, A. El. (1997). Language Loss Among Native Speakers of Morrocan [sic] Arabic in the Netherlands. Tilburg: Tilburg University Press. Almalech, M. (1997). “First language attrition by speakers of Bulgarian in Israel”. Sapostavitelno Ezikoznanie/Sopostavitel’Noe Jazykoznanie/Contrastive Linguistics (Sofia, Bulgaria), 22 (3), 55–59. Altenberg, E. P. (1991). “Assessing first language vulnerability to attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 189–206). Cambridge: CUP. Ammerlaan, T. (1991). “Testing linguistic variables in first language attrition”. In F. Lovejoy (Ed.), Studies in Language Loss (pp. 20–41). Sydney: Holborn Press. Ammerlaan, T. (1996). “You Get a Bit Wobbly.” Exploring bilingual lexical retrieval processes in the context of first language attrition. Unpublished Doctoral Dissertation, Katholieke Universiteit Nijmegen. Ammerlaan, T. (1997). “ ‘Corrosion’ or ‘loss’ of immigrant Dutch in Australia: An experiment on first language attrition”. In J. Klatter-Folmer & S. Kroon (Eds.), Dutch Overseas: Studies in maintenance and loss of Dutch as an immigrant language (pp. 69– 97). Tilburg: Tilburg University Press. Andersen, R. W. (1982). “Determining the linguistic attributes of language attrition”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 83–118). Rowley, MA: Newbury House. Anderson, R. T. (2001). “Lexical morphology and verb use in child first language loss: A preliminary case study investigation”. International Journal of Bilingualism, 5 (4), 377– 402. Andrews, D. R. (1998). Sociocultural Perspectives on Language Change in Diaspora. Amsterdam and Philadelphia: John Benjamins. Armon Lotem, S. (2000). “Language attrition: why are resumptive pronouns so susceptible?”. In S. C. Howell, S. Fish & K. Thea (Eds.), Proceedings of the 24th Annual Boston University Conference on Language Development (pp. 58–67). Somerville, MA: Cascadilla. Arrieta, K. M. C. (1999). “Nominalisations in Basque: A case in language attrition”. Dissertation Abstracts International, Section A: The Humanities and Social Sciences (DAIA), 59 (8), 2950. Bahrick, H. (1984). “Fifty years of second language attrition: Implications for programmatic research”. Modern Language Journal, 68, 105–118. Ben-Rafael, M. (1994). “Contact de langue et simplification”. Cahier du francais contemporain, I, 77–87. Ben-Rafael, M. (1999). “Emigration et contact de langues: emprunts et configurations sémantiques en Franbreu”. In F. S. Martinez & C. H. Gonzalez (Eds.), Las lengas en la Europa Communitaria III (pp. 423–438). Amsterdam: Rodopi. Ben-Rafael, M. (2001). Contact de Langues: le français parlé des francophones israéliens. Unpublished Doctoral Dissertation, University of Tel Aviv.
Language attrition bibliography
Ben-Rafael, M. (2002). “Le franbreu: les traits essentiels”. In D. Mendelson (Ed.), La culture francophone en Israel (pp. 79–96). Paris: l’Harmattan. Berko-Gleason, J. (1982). “Insights from child language acquisition for second language loss”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 13–23). Rowley, MA: Newbury House. Berko-Gleason, J. (1993). “Neurolinguistic aspects of first language acquisition and loss”. In K. Hyltenstam & A. Viberg (Eds.), Progression and Regression in Language: Sociocultural, neuropsychological and linguistic perspectives (pp. 147–177). Cambridge: CUP. Berman, R. A., & Olshtain, E. (1983). “Features of first language transfer in second language attrition”. Applied Linguistics, 4 (3), 222–234. Bettoni, C. (1986). “Italian language attrition in Sydney: The role of birth order”. In C. Bettoni, T. De Mauro, P. Crino, & J. J. Kinder (Eds.), Italian Abroad: Studies on language contact in English-speaking countries (pp. 61–85). (Sydney). Camilla, MA: Frederick May Fndtn. for It. Studies. Bettoni, C. (1989). “Language variety among Italians: Anglicisation, attrition and attitudes”. I.T.L.: Review of Applied Linguistics, 83–84, 37–50. Bettoni, C. (1991). “Language shift and morphological attrition among second generation Italo Australians”. Rivista di Linguistica (RdLing). Pisa, Italy, 3 (2), 369–387. Bode, S. de (1996). First Language Attrition: Productive morphology disintegration and neurobiological support. A case study. Unpublished MA Thesis, California State Polytechnic University, Pomona. Bolonyai, A. (1999). The Hidden Dimensions of Language Contact: The case of HungarianEnglish bilingual children. Unpublished Doctoral Dissertation, University of South Carolina. Bolonyai, A., & Dutkova-Cope, L. (2001). “L1 attrition of verbal morphology in bilingual children and adults”. In X. Bonch-Bruevich, W. J. Crawford, J. Hellermann, C. Higgins, & H. Nguyen (Eds.), The Past, Present, and Future of Second Language Research: Selected Proceedings of the 2000 Second Language Research Forum (pp. 104–123). Somerville, MA: Cascadilla. Bot, K. de, & Clyne, M. (1989). “Language reversion revisited”. Studies in Second Language Acquisition, 11, 167–177. Bot, K. de, & Clyne, M. (1994). “A 16 year longitudinal study of language attrition in Dutch immigrants in Australia”. Journal of Multilingual and Multicultural Development, 15 (1), 17–28. Bot, K. de, & Fase, W. (Eds.). (1991). Migrant Languages in Europe (=International Journal of the Sociology of Language, 90). Berlin: Mouton de Gruyter. Bot, K. de, & Hulsen, M. (2002). “Tests, self-assessments and perceptions”. In V. Cook (Ed.), Portraits of the Language Learner (pp. 253–274). Clevedon: Multilingual Matters. Bot, K. de, & Lintsen, T. (1989). “Perception of own language proficiency by elderly adults”. I.T.L.: Review of Applied Linguistics, 83–84, 51–61. Bot, K. de, & T. Lintsen. (1986). “Foreign-language proficiency in the elderly”. In B. Weltens, K. de Bot, & Th. van Els (Eds.), Language Attrition in Progress (pp. 131–141). Dordrecht: Foris. Bot, K. de, & Weltens, B. (1985). “Taalverlies: Beschrijven versus verklaren”. Handelingen van het 38e Nederlands Filologengcongres, 51–61.
Monika S. Schmid
Bot, K. de, & Weltens, B. (1991). “Recapitulation, regression, and language loss”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 31–51). Cambridge: CUP. Bot, K. de, & Weltens, B. (1995). “Foreign language attrition”. Annual Review of Applied Linguistics, 15, 151–164. Bot, K. de, Gommans, P., & Rossing, C. (1991). “L1 loss in an L2 environment: Dutch immigrants in France”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 87–98). Cambridge: CUP. Bot, K. de (1991). “Language loss”. In B. Spillner (Ed.), Sprache und Politik (pp. 63–65). Frankfurt: Peter Lang. Bot, K. de (1996). “Language loss”. In H. Goebl, P. H. Nelde, Z. Stary, & W. Wölck (Eds.), Contact Linguistics. An international handbook of contemporary research, Vol. I (pp. 579– 585). Berlin/New York: Walter de Gruyter. Bot, K. de (1998). “The psycholinguistics of language loss”. In G. Extra & L. Verhoeven (Eds.), Bilingualism and Migration (pp. 17–28). Berlin: Mouton de Gruyter. Boyd, S. (1986). “Using the present to predict the future in language contact: The case of immigrant minority languages in Sweden”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 99–115). Dordrecht: Foris. Boyd, S. (1993). “Attrition or expansion: Changes in the lexicon of Finnish and American adult bilinguals in Sweden”. In K. Hyltenstam & A. Viberg (Eds.), Progression and Regression in Language: Sociocultural, neuropsychological, and linguistic perspectives (pp. 386–412). Cambridge: CUP. Brewer Bomar, K. E. (1982). “Second language lexical and syntactical interference on the first language of two four year-old Spanish speakers”. Dissertation Abstracts International (DAI). Ann Arbor, MI, 42 (12), 5105A. Brons-Albert, R. (1992). “Verlust der Muttersprache in fremdsprachiger Umgebung”. Info DaF, 19 (3), 315–325. Brons-Albert, R. (1993). “Interferenzfehler in der Muttersprache von in den Niederlanden lebenden Deutschen”. In B. Spillner (Ed.), Nachbarsprachen in Europa (pp. 96–104). Frankfurt: Peter Lang. Brown, J. C. (2001). L1 Attrition Among Native Speakers of English Residents Abroad: A sociolinguistic case study from Foggia, Italy. Unpublished MA Thesis, University of Surrey. Bullock, B. E., & Gerfen, C. (Forthcoming). “Frenchville French: A case study in phonological attrition”. To appear in: International Journal of Bilingualism. Caramazza, A., & Zurif, E. B. (1978). “Comprehension of complex sentences in children and aphasics: A test of the regression hypothesis”. In A. Caramazza & E. B. Zurif (Eds.), Language Acquisition and Language Breakdown: Parallels and divergencies (pp. 145– 161). Baltimore: John Hopkins University Press. Clark, J. L. D. (1982). “Measurement considerations in language attrition research”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 138–152). Rowley, MA: Newbury House. Clyne, M. (1968). “Decay, preservation and renewal: Notes on some Southern Australian German settlements”. AUMLA, 29, 33–43.
Language attrition bibliography
Clyne, M. (1973). “Thirty years later: Some observations on ‘refugee German’ in Melbourne”. In H. Scholler & H. Reilly (Eds.), Lexicography and Dialect Geography (pp. 96–106). Wiesbaden: Frank Steiner. Clyne, M. (1981). “Second language attrition and first language reversion among elderly bilinguals in Australia”. In W. Meid & K. Heller (Eds.), Sprachkontakt als Ursache von Veränderungen der Sprach- und Bewußtsseinsstruktur: eine Sammlung von Studien zur sprachlichen Interferenz (pp. 25–32). Innsbruck: Institut für Sprachwissenschaft. Clyne, M. (1981). Deutsch als Muttersprache in Australien: Zur Ökologie einer Einwanderersprache. Wiesbaden: Franz Steiner. Clyne, M. (1992). “Linguistic and sociolinguistic aspects of language contact, maintenance and loss: Towards a multifacet theory”. In W. Fase, K. Jaspert, & S. Kroon (Eds.), Maintenance and Loss of Minority Languages (pp. 17–36). Amsterdam and Philadelphia: Benjamins. Clyne, M., & Pauwels, A. (1997). “Use, maintenance, structures, and future of Dutch in Australia”. In J. Klatter-Folmer & S. Kroon (Eds.), Dutch Overseas: Studies in maintenance and loss of Dutch as an immigrant language (pp. 33–49). Tilburg: Tilburg University Press. Cohen, A. (1986). “Forgetting foreign-language vocabulary”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 143–158). Dordrecht: Foris. Cohen, A. D. (1989). “Attrition in the productive lexicon of two Portuguese third language speakers”. Studies in Second Language Acquisition, 11, 135–150. Davies, A. (1986). “Language loss and symbolic gain: The meaning of institutional maintenance”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 117–127). Dordrecht: Foris. Dorian, N. C. (1982). “Language loss and maintenance in language contact situations”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 44–59). Rowley, MA: Newbury House. Dressler, W. U. (1991). “The sociolinguistic and patholinguistic attrition of Breton phonology, morphology, and morphonology”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 99–112). Cambridge: CUP. Dugas, L. G. (2000). “Attrition of pronunciation accuracy among advanced American learners of French”. Dissertation Abstracts International, Section A: The Humanities and Social Sciences (DAIA), 60 (11), 3986. Edwards, J. R. (1992). “Sociopolitical aspects of language maintenance and loss: Towards a typology of minority language situation”. In W. Fase, K. Jaspert & S. Kroon (Eds.), Maintenance and Loss of Minority Languages (pp. 37–54). Amsterdam and Philadelphia: Benjamins. Els, T. van (1986). “An overview of European research on language attrition”. In B. Weltens, K. de Bot & T. van Els (Eds.), Language Attrition in Progress (pp. 3–18). Dordrecht: Foris. Falstein, S., & Waas, M. (1993). “Migrants face erosion of their mother tongue”. Post Migration, 90, 19–20. Fase, W., Jaspaert, K., & Kroon , S. (Eds.). (1992). Maintenance and Loss of Minority Languages. Amsterdam and Philadelphia: John Benjamins.
Monika S. Schmid
Fase, W., Jaspaert, K., & Kroon, S. (Eds.). (1995). The State of Minority Languages. International perspectives on survival and decline (= European Studies on Multilingualism 5). Lisse: Swets & Zeitlinger. Fishman, J. (1992). “Conference summary”. In W. Fase, K. Jaspaert & S. Kroon (Eds.), Maintenance and Loss of Minority Languages (pp. 395–403). Amsterdam and Philadelphia: John Benjamins. Freed, B. F. (1982). “Language loss: Current thoughts and future directions”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 1–5). Rowley, MA: Newbury House. Fromm, E. (1970). “Age regression and unexpected reappearance of a repressed childhood language”. International Journal of Clinical Experimental Hypnosis, 18 (2), 79–88. Fujita, M. (2002). “Second language English attrition of Japanese bilingual children”. Dissertation Abstracts International, Section A: The Humanities and Social Sciences (DAIA), 63 (1), 59–60. Furbee, N. L., & Stanley, L. A. (1996). “Language attrition and language planning in accommodation perspective”. Southwest Journal of Linguistics, 15 (1), 45–62. Gardner, R. C. (1982). “Social factors in language retention”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 24–43). Rowley, MA: Newbury House. Gardner, R. C., Lalonde, R. N., & MacPherson, J. (1985). “Social factors in second language attrition”. Language Learning: A Journal of Applied Linguistics, 35 (4), 519–540. Gardner, R. C., Lalonde, R. N., Moorcroft, R., & Evers, F. T. (1987). “Second language attrition: The role of motivation and use”. Journal of Language and Social Psychology, 6 (1), 29–47. Gardner, R., Lalonde, R., & MacPherson, J. (1983). Social factors in second-language attrition: An initial investigation. London, Ontario: University of Western Ontario (Research Bulletin #584). Giesbers, H. (1997). “Dutch in Indonesia: Language attrition or language contact?”. In J. Klatter-Folmer & S. Kroon (Eds.), Dutch Overseas: Studies in maintenance and loss of Dutch as an immigrant language (pp. 163–180). Tilburg: Tilburg University Press. Ginsberg, R. B. (1986). “Issues in the analysis of language loss: Methodology of the Language Skills Attrition Project”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 19–36). Dordrecht: Foris. Godsall-Myers, J. E. (1982). “The attrition of language skills in German classroom bilinguals: A case study”. Dissertation Abstracts International, Ann Arbor, MI 43, 1. Gonzalez Moncada, A. (1996). “Language attrition and retention among non-native speaker EFL teachers”. Dissertation Abstracts International (DAI), 57 (3), 1118A. Goral, M. (2004). “First-language decline in healthy aging: Implications for attrition in bilingualism”. Journal of Neurolinguistics, 17 (1), 31–52. Grendel, M., Weltens, B., & de Bot, K. (1993). “Language attrition: Rise and fall of a research topic?”. Toegepaste Taalwetenshap in Artikelen, 46/47, 59–68. Grosjean, F., & Py, B. (1991). “La restructuration d’une première langue: l’intégration de variantes de contact dans la compétence de migrants bilingues”. La Linguistique, 27, 35–60. Gross, S. (2000). The Role of Abstract Lexical Structure in First Language Attrition: Germans in America. Unpublished Doctoral Dissertation, University of South Carolina.
Language attrition bibliography
Gürel, A. (2002a). Linguistic Characteristics of Second Language Acquisition and First Language Attrition: Turkish overt versus null pronouns. Unpublished Doctoral Dissertation, Montréal: McGill University. Gürel, A. (2002b). “First language attrition: The effects of second language”. In B. Skarabela, S. Fish & A. H. J. Do (Eds.), Proceedings of the 26th Annual Boston University Conference on Language Development, I, II (pp. 255–266). Somerville, MA: Cascadilla. Gürel, A. (2004). “Selectivity in L2-induced L1 attrition: A psycholinguistic account.” Journal of Neurolinguistics, 17 (1), 53–78. Hagen, A. M., & de Bot, K. (1990). “Structural loss and levelling in minority languages and dialects”. Sociolinguistica, 4, 136–149. Håkansson, G. (1995). “Syntax and morphology in language attrition: A study of five bilingual expatriate Swedes”. International Journal of Applied Linguistics, 5 (2), 153–171. Hakuta, K., & d’Andrea, D. (1992). “Some properties of bilingual maintenance and loss in Mexican background high-school students”. Applied Linguistics, 13 (1), 72–99. Hansen, L. (Ed.). (1999). Second Language Attrition in Japanese Contexts. Oxford: Oxford University Press. Hansen, L. (1999). “Investigating second language attrition: An introduction”. In L. Hansen (Ed.), Second Language Attrition in Japanese Contexts (pp. 3–20). Oxford: Oxford University Press. Hansen, L. (1999). “Not a total loss: The attrition of Japanese negation over three decades”. In L. Hansen (Ed.), Second Language Attrition in Japanese Contexts (pp. 142–153). Oxford: Oxford University Press. Hansen, L. (2001). “Language attrition: The fate of the start”. Annual Review of Applied Linguistics, 21, 60–73. Hayashi, B. (1999). “Testing the regression hypothesis: The remains of the Japanese negation system in Micronesia”. In L. Hansen (Ed.), Second Language Attrition in Japanese Contexts (pp. 154–168). Oxford: Oxford University Press. Hedgcock, J. (1991). “Foreign language retention and attrition: A study of regression models”. Foreign Laguage Annals, 24, 43–55. Hill, J. H. (1989). “Language attrition in progress – Book review”. Language in Society, 18 (4), 594–598. Hiller-Foti, I. (1985). “Erhebung über den Verlust der Muttersprache Deutscher Immigranten in Sizilien”. Rassegna Italiana di Linguistica Applicata, 17 (1), 107–110. Hinskens, F. (1986). “The selection of linguistic variables in empirical research on variation and change in dialects”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 53–74). Dordrecht: Foris. Hirvonen, P. (1998). “The Finnish-American language shift”. In J. Niemi et al. (Eds.), Language Contact, Variation, and Change (pp. 136–150). Joensuu, Finland: Faculty of Humanities. Hofmann, C. (1994). “Language loss and language recovery: The case of the Russlanddeutsche”. In P. M. Mair & W. V. Davies (Eds.), The Changing Voices of Europe: Social and political changes and their linguistic repercussions past, present and future (pp. 311–324). Cardiff: University of Wales Press.
Monika S. Schmid
Hormann, C. (1994). Acts of Identity in Aussiedler-German: Operationaliszing Le Page’s sociolinguistic identity theory. Unpublished Doctoral Dissertation, University of Texas at Austin. Huffines, M. L. (1991). “Pennsylvania German: Convergence and change as strategies of discourse”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 125–137). Cambridge: CUP. Huls, E., & van de Mond, A. (1992). “Some aspects of language attrition in Turkish families in the Netherlands”. In W. Fase, K. Jaspaert, & S. Kroon (Eds.), Maintenance and Loss of Minority Languages (pp. 99–115). Amsterdam and Philadelphia: John Benjamins. Hulsen, M. (2000). Language Loss and Language Processing. Three generations of Dutch migrants in New Zealand. Unpublished Doctoral Dissertation, Katholieke Universiteit Nijmegen. Hulsen, M., de Bot, K., & Weltens, B. (1999). “Language shift, language loss and language processing: An investigation of three generations of Dutch immigrants in New Zealand”. In E. Huls & B. Weltens (Eds.), Artikelen van de Derde Sociolinguistische Conferentie (pp. 221–232). Delft: Eburon. Hulsen, M., de Bot, K., & Weltens, B. (2002). “‘Between two worlds.’ Social networks, language shift, and language processing in three generations of Dutch migrants in New Zealand”. International Journal of the Sociology of Language, 153, 27–52. Hyltenstam, K., & Stroud, C. (1996). “Language maintenance”. In H. Goebl, P. H. Nelde, Z. Stary, & W. Wölck (Eds.), Contact Linguistics. An international handbook of contemporary research, Vol. I (pp. 567–578). Berlin/New York: Walter de Gruyter. Isurin, L. (2000a). “Mechanisms of first language forgetting”. Dissertation Abstracts International, Section A: The Humanities and Social Sciences, (DAIA) 60 (9), 3342. Isurin, L. (2000b). “Deserted island: Or, a child’s first language forgetting”. Bilingualism: Language and Cognition, 3 (2), 151–166. Isurin, L., & McDonald, J. L. (2001). “Retroactive interference from translation equivalents: Implications for first language forgetting”. Memory and Cognition, 29 (2), 312–319. Jackson, A. E., & Mari, K. (1995). “What happens to their English? A study of the English proficiency of Japanese elementary school returnees”. Shoin Literary Review, 107, 1–34. Jamieson, P. (1980). “The pattern of urban language loss”. Australian and New Zealand Journal of Sociology, 16, 102–109. Jarvis, S. (2003). “Probing the effects of the L2 on the L1: A case study”. In V. Cook (Ed.), Effects of the Second Language on the First (pp. 81–102). Clevedon: Multilingual Matters. Jaspaert, K., & Kroon, S. (1987). “The relationship between global language proficiency tests and language and language loss”. In F. Beukema & P. Coopmans (Eds.), Linguistics in the Netherlands (pp. 91–100). Dordrecht: Foris. Jaspaert, K., & Kroon, S. (1989). “Social determinants of language loss”. I.T.L.: Review of Applied Linguistics, 83–84, 75–98. Jaspaert, K., & Kroon, S. (1991). “Social determinants of language shift by Italians in the Netherlands and Flanders”. International Journal of the Sociology of Language, 90, 77–96. Jaspaert, K., & Kroon, S. (1992). “From the typewriter of A.L.: A case study in language loss”. In W. Fase, K. Jaspaert, & S. Kroon (Eds.), Maintenance and Loss of Minority Languages (pp. 137–147). Amsterdam and Philadelphia: John Benjamins.
Language attrition bibliography
Jaspaert, K., Kroon, S., & van Hout, R. (1986). “Points of reference in first-language loss research”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 37–49). Dordrecht: Foris. Jessner, U. (2003). “A dynamic approach to language attrition in multilingual systems”. In V. Cook (Ed.), Effects of the Second Language on the First (pp. 234–246). Clevedon: Multilingual Matters. Jordens, P., de Bot, K., & Trapman, H. (1989). “Linguistic aspects of regression in German case marking”. Studies in Second Language Acquisition, 11 (2), 179–204. Jordens, P., de Bot, K., van Os, C., & Schumans, J. (1986). “Regression in German case marking”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 159–176). Dordrecht: Foris. Kaufman, D. (1998). “Children’s assimilatory patterns and L1 attrition”. In A. Greenhill, M. Hugehes, H. Littlefield, & H. Walsh (Eds.), Proceedings of the 22nd Annual Boston University Conference on Language Development, I, II (pp. 175–188). Somerville, MA: Cascadilla. Kaufman, D. (2000). “Attrition of Hebrew in the United States: Sociolinguistic perspectives”. In E. Olshtain (Ed.), Immigration, Identity, and Language (pp. 409–420). Jerusalem: Magnes Publications Hebrew University. Kaufman, D., & Aronoff, M. (1991). “Morphological disintegration and reconstruction in first language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 175–188). Cambridge: CUP. Kaufman, H. D. (1992). “First language attrition as a creative interplay between two languages”. Dissertation Abstracts International, Ann Arbor, MI 52, 9. Kirschner, C. (1996). “Language attrition and the Spanish English bilingual: A case of syntactic reduction”. Bilingual Review/La Revista Bilingue (BR/RB), 21 (2), 123–130. Klatter-Folmer, J., & Kroon, S. (1997). “Dutch overseas: Introductory remarks”. In J. KlatterFolmer & S. Kroon (Eds.), Dutch Overseas: Studies in maintenance and loss of Dutch as an immigrant language (pp. 1–19). Tilburg: Tilburg University Press. Kokko, O. (1998). “Loss and maintenance of linguistic features in language attrition”. In J. Niemi et al. (Eds.), Language Contact, Variation, and Change (pp. 151–172). Joensuu, Finland: Faculty of Humanities. Köpke, B. (1999). L’attrition de la première langue chez le bilingue tardif: Implications pour l’étude psycholinguistique du bilinguisme. Unpublished Doctoral Dissertation, Université de Toulouse-Le Mirail. Köpke, B. (2000). “Effet du pays d’accueil sur le maintien de la langue. Le cas des immigrés d’origine allemande”. Education et Societes Plurilingues, 9, 59–65. Köpke, B. (2001). “Quels changements linguistiques dans l’attrition de la L1 chez le bilingue tardif? ”. Travaux Neuchâtelois de Linguistique, 34/35, 355–368. Köpke, B. (2002). “Activation thresholds and non-pathological first language attrition”. In F. Fabbro (Ed.), Advances in the Neurolinguistics of Bilingualism: Essays in honor of Michel Paradis (pp. 119–142). Udine: Forum. Köpke, B. (2002). “Besonderheiten in der Verarbeitung spät erworbener Sprachen: Beispiele aus einer Studie zum ‘Sprachverlust’ von deutschen Immigranten in Frankreich und Kanada”. In J. Müller-Lancé & C. Maria Riehl (Eds.), Ein Kopf – viele Sprachen: Koexistenz, Interaktion und Vermittlung (pp. 31–44). Aachen: Shaker.
Monika S. Schmid
Köpke, B. (2004). “Neurolinguistic aspects of attrition” Journal of Neurolinguistics, 17 (1), 3–30. Köpke, B., & Nespolous, J.-L. (2001). “First language attrition in production skills and metalinguistic abilities in German-English and German-French bilinguals”. In T. Ammerlaan, M. Hulsen, H. Strating, & K. Ya˘gmur (Eds.), Sociolinguistic and Psycholinguistic Perspectives on Maintenance and Loss of Minority Languages (pp. 221– 234). Münster: Waxmann. Kouritzin, S. G. (1998). “Cast away cultures and taboo tongues: Face[t]s of first language loss”. Dissertation Abstracts International, Section A: The Humanities and Social Sciences (DAIA), 59 (1), 63. Kravin, H. (1992). “Erosion of a language in bilingual development”. Journal of Multilingual and Multicultural Development, 13, 307–325. Kuhberg, H. (1992). “Longitudinal L2 attrition versus L2 acquisition, in three Turkish children. Empirical findings”. Second Language Research (SLRe). Utrecht, Netherlands, 8 (2), 138–154. Lambert, R., & Moore, S.J. (1984). “Recent research on language skill attrition”. ERIC/CLL News Bulletin, 8, 1–8. Lambert, R. D. (1982). “Setting the agenda”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 6–10). Rowley, MA: Newbury House. Lambert, R. D. (1989). “Language attrition”. I.T.L.: Review of Applied Linguistics, 83–84, 1– 18. Lambert, R. D., & Moore, S. J. (1986). “Problem areas in the study of language attrition”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 177–186). Dordrecht: Foris. Laufer, B. (2003). “The influence of L2 on L1 collocational knowledge and on L1 lexical diversity in free written expression”. In V. Cook (Ed.), Effects of the Second Language on the First (pp. 19–31). Clevedon: Multilingual Matters. Leisiö, L. (2001). Morphosyntactic Convergence and Integration in Finland Russian. Unpublished Doctoral Dissertation, University of Tampere. Levine, G. S. (1996). “Elderly second-generation speakers of Yiddish: Toward a model of L1 loss, incomplete L1 acquisition, competence, and control”. Southwest Journal of Linguistics, 15 (1), 109–120. Levine, G. S. (2000). Incomplete L1 Acquisition in the Immigrant Situation: Yiddish in the United States. Tübingen: Niemeyer. Levy, J. (1982). “Policy implications/complications arising from native language attrition in U.S. ethnolinguistic minority groups”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 191–201). Rowley, MA: Newbury. Lowe, P. Jr. (1982). “The U.S. government’s foreign language attrition and maintenance experience”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 176– 190). Rowley, MA: Newbury House. Maher, J. (1991). “A crosslinguistic study of language contact and language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 67–84). Cambridge: CUP.
Language attrition bibliography
Major, R. C. (1993). “Sociolinguistic factors in loss and acquisition of phonology”. In K. Hyltenstam & A. Viberg (Eds.), Progression and Regression in Language: Sociocultural, neuropsychological and linguistic perspectives (pp. 463–478). Cambridge: CUP. Major, R. C. (1992). “Losing English as a first language”. The Modern Language Journal, 76, 190–209. McCormack, B. (2001). The Maintenance and Loss of Reflexive Anaphors in L2 English. Unpublished Doctoral Dissertation, University of Durham, UK. Montrul, S. (2002). “Incomplete acquisition and attrition of Spanish tense/aspect distinctions in adult bilinguals”. Bilingualism: Language and Cognition, 5 (1), 39–68. Moorcroft, R., & Gardner, R. C. (1987). “Linguistic factors in second language loss”. Language Learning: A Journal of Applied Linguistics, 37 (3), 327–340. Münstermann, H., & Hagen, T. (1986). “Functional and structural aspects of dialect loss: A research plan and some first results”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 75–96). Dordrecht: Foris. Murtagh, L. (2003). Retention and Attrition of Irish as a Second Language. Unpublished Doctoral Dissertation, Rijksuniversiteit Groningen. Nagasawa, S. (1996). “Attrition of speaking skills in advanced speakers of Japanese L2: A multiple case study”. Dissertation Abstracts International, Section A: The Humanities and Social Sciences (DAIA), 56 (11), 4381. Nagasawa, S. (1999). “Learning and losing Japanese as a second language: A multiple case study of American university students”. In L. Hansen (Ed.), Second Language Attrition in Japanese Contexts (pp. 169–199). Oxford: Oxford University Press. Nakuma, C. (1997a). “A method for measuring the attrition of communicative competence: A pilot study with Spanish L3 subjects”. Applied Psycholinguistics, 18 (2), 219–235. Nakuma, C. (1997b). “Loss of communicative competence: Measurability and description of a method”. International Review of Applied Linguistics in Language Teaching, 35 (3), 199–209. Nakuma, C. (1997c). “Cleaning up spontaneous speech for use in L2 attrition research: A proposal”. Journal of Multilingual and Multicultural Development, 18 (2), 135–144. Nakuma, C. K. (1998). “A new theoretical account of ‘fossilisation’: Implications for L2 attrition research”. International Review of Applied Linguistics in Language Teaching, 36 (3), 247–256. Obler, L. K. (1982). “Neurolinguistic aspects of language loss as they pertain to second language attrition”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 60–79). Rowley, MA: Newbury House. Obler, L. K. (1993). “Neurolinguistic aspects of second language development and attrition”. In K. Hyltenstam & A. Viberg (Eds.), Progression and Regression in Language: Sociocultural, neuropsychological and linguistic perspectives (pp. 178–195). Cambridge: CUP. Obler, L. K., & Albert, M. L. (1989). “Language decline in aging”. I.T.L.: Review of Applied Linguistics, 83–84, 63–73. Obler, L. K., & Mahecha, N. R. (1991). “First language loss in bilingual and polyglot aphasics”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 53–65). Cambridge: CUP.
Monika S. Schmid
Olshtain, E. (1986). “The attrition of English as a second language with speakers of Hebrew”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 187– 204). Dordrecht: Foris. Olshtain, E. (1989). “Is second language attrition the reversal of second language acquisition?”. Studies in Second Language Acquisition, 11 (2), 151–165. Olshtain, E., & Barzilay, M. (1991). “Lexical retrieval difficulties in adult language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 139–150). Cambridge: CUP. Oxford, R. L. (1982). “Research on language loss – A review with implications for foreignlanguage teaching”. Modern Language Journal, 66, 160–169. Oxford, R. L. (1982). “Technical issues in designing and conduction research on language skill attrition”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 119–137). Rowley, MA: Newbury House. Pan-Alexander, B., & Berko-Gleason, J. (1986). “The study of language loss: Models and hypotheses for an emerging discipline”. Applied Psycholinguistics, 7, 193–206. Paradis, M. (1993). “Linguistic, psycholinguistic, and neurolinguistic aspects of ‘interference’ in bilingual speakers: The Activation Threshhold Hypothesis”. International Journal of Psycholinguistics, 9 (2), 133–145. Pavlenko, A. (2000). “L2 influence on L1 in late bilingualism”. Issues in Applied Linguistics, 11 (2), 175–205. Pavlenko, A. (2003). “‘I feel clumsy speaking Russian’: L2 influence on L1 in narratives of Russian L2 users of English”. In V. Cook (Ed.), L2 influence on L1 (pp. 32–61). Clevedon: Multilingual Matters. Pelc, L. A. (1998). “Supralexical and lexical attrition of Greek in Greek-English bilinguals”. In A. Greenhill, M. Hugehes, H. Littlefield, & H. Walsh (Eds.), Proceedings of the 22nd Annual Boston University Conference on Language Development (pp. 621–625). Somerville, MA: Cascadilla. Pelc, L. A. (2001). L1 Lexical, Morphological and Morphosyntactic Attrition in Greek-English Bilinguals. Unpublished Doctoral Dissertation, New York: City University of New York. Pfaff, C. W. (1991). “Turkish in contact with German: Language maintenance and loss among immigrant children in Berlin (West)”. International Journal of the Sociology of Language, 90, 97–129. Platzack, C. (1998). “The initial hypothesis of syntax: A minimalist perspective on language acquisition and attrition”. In H. Clahsen (Ed.), Generative Perspectives on Language Acquisition (pp. 369–414). Amsterdam and Philadelphia: John Benjamins. Polinsky, M. (1994). “Structural dimensions of first language loss”. Chicago Linguistic Society, 30, 257–276. Polinsky, M. (1997). “American Russian: Language loss meets language acquisition”. In W. Browne, E. Dornisch, N. Kondrashova, & D. Zec (Eds.), Annual Workshop on Formal Approaches to Slavic Linguistics: The Cornell meeting 1995 (pp. 370–406). Ann Arbor, MI: Michigan Slavic. Pons, C. R. (1989). “Book review of ‘Language attrition in progress’ ”. Studies in Second Language Acquisition, 11 (2), 217–218. Porte, G. (1999). “English as a forgotten language”. ELT Journal, 53 (1), 28–34.
Language attrition bibliography
Porte, G. (2003). “English from a distance: Code-mixing and blending in the L1 output of long-term resident overseas EFL teachers”. In V. Cook (Ed.), Effects of the Second Language on the First (pp. 103–119). Clevedon: Multilingual Matters. Prather, W. J. (1998). “Native language loss in second language learners referred to special education”. Dissertation Abstracts International, Section A: The Humanities and Social Sciences (DAIA), 58 (8), 3087–3088. Preston, D. (1982). “How to lose a language”. Interlanguage Studies Bulletin, 6 (2), 64–87. Py, B. (1986). “Native language attrition amongst migrant workers: Towards an extension of the concept of interlanguage”. In E. Kellerman & M. Sharwood Smith (Eds.), Crosslinguistic Influence in Second Language Acquisition (pp. 163–172). New York, NY: Pergamon. Raffaldini, T. (1988). “Attrition of communicative ability among former year abroad students of French”. Dissertation Abstracts International (DAI). Ann Arbor, MI, 49 (4), 807A. Raidt, E. H. (1997). “Interference, shift and loss of Dutch in South Africa”. In J. KlatterFolmer & S. Kroon (Eds.), Dutch Overseas: Studies in maintenance and loss of Dutch as an immigrant language (pp. 215–236). Tilburg: Tilburg University Press. Reetz-Kurashige, A. (1999). “Japanese returnees’ retention of English-speaking skills: Changes in verb usage over time”. In L. Hansen (Ed.), Second Language Attrition in Japanese Contexts (pp. 21–58). Oxford: Oxford University Press. Rhodes, B. M. (1997). “Research into the attrition of dialect”. Transactions of the Yorkshire Dialect Society, 19 (97), 22–29. Rhydwen, M. (1999). “Language loss, our loss”. In R. S. Wheeler (Ed.), The Workings of Language: From prescriptions to perspectives (pp. 129–138). Westport, CT: Praeger. Riionheimo, H. (1998). “Morphological attrition and interference in language contact: Sketching a framework”. In J. Niemi et al. (Eds.), Language Contact, Variation, and Change (pp. 246–268). Joensuu, Finland: Faculty of Humanities. Ronowicz, E. (1999). “Methodological issues of testing language attrition in children in a natural bilingual environment”. Poznan Studies in Contemporary Linguistics (PSiCL), 35, 105–117. Russell, R. A. (1999). “Lexical maintenance and attrition in Japanese as a second language”. In L. Hansen (Ed.), Second Language Attrition in Japanese Contexts (pp. 114–141). Oxford: Oxford University Press. Schatz, H. F. (1989.“ Code-switching or borrowing? English elements in the Dutch of DutchAmerican immigrants”. I.T.L.: Review of Applied Linguistics, 83–84, 125–162. Schaufeli, A. (1996). “Word order patterns in contact: Turkish in the Netherlands”. Southwest Journal of Linguistics, 15 (1–2), 153–169. Schmid, M. S. (2002). First Language Attrition, Use and Maintenance: The case of German Jews in Anglophone countries. Amsterdam and Philadelphia: John Benjamins. Schmid, M. S. (Forthcoming). “Measuring first language attrition in free speech: What’s in a mistake?”. International Journal of Bilingualism. Schmitt, E. (2000). “The lost word: Language attrition among children”. In S. C. Howell, S. Fish, & K. Thea (Eds.), Proceedings of the 24th Annual Boston University Conference on Language Development (pp. 645–656). Sommerville, MA: Cascadilla.
Monika S. Schmid
Schmitt, E. (2001). Beneath the surface: Signs of language attrition in immigrant children from Russia. Unpublished Doctoral Dissertation, University of South Carolina. Schoenmakers-Klein Gunnewiek, M. (1989). “Structural aspects of the loss of Portuguese among migrants: A research outline”. I.T.L.: Review of Applied Linguistics, 83–84, 99– 123. Schoenmakers-Klein Gunnewiek, M. (1997). “Dutch language loss in Brazil and the conceptual hypothesis”. In J. Klatter-Folmer & S. Kroon (Eds.), Dutch Overseas: Studies in maintenance and loss of Dutch as an immigrant language (pp. 99–119). Tilburg: Tilburg University Press. Schoenmakers-Klein Gunnewiek, M. (1998). Taalverlies door Taalcontact? Een onderzoek bij Portugese migranten. Unpublished Doctoral Dissertation, Katholieke Universiteit Brabant. Seliger, H. W. (1985). Primary Language Attrition in the Context of Other Language Loss and Mixing. MS, Queens College. Seliger, H. W. (1989). “Deterioration and creativity in childhood bilingualism”. In K. Hyltenstam & L. K. Obler (Eds.), Bilingualism Across the Lifespan (pp. 173–184). Cambridge: CUP. Seliger, H. W. (1991). “Language attrition, reduced redundancy, and creativity”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 227–240). Cambridge: CUP. Seliger, H. W., & Vago, R. M. (1991). “The study of first language attrition: An overview”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 3–15). Cambridge: CUP. Seliger, H. W., & Vago, R. M. (1996). “Primary language attrition in the context of bilingualism”. In W. C. Ritchie & T. K. Bathia (Eds.), Handbook of Second Language Acquisition (pp. 605–626). San Diego, CA: Academic Press. Sharwood Smith, M. (1983a). “On explaining language loss”. In S. Felix & H. Wode (Eds.), Language Development at the crossroads (pp. 49–69). Tübingen: Narr. Sharwood Smith, M. (1983b). “On first language loss in the second language acquirer”. In S. Gass & L. Selinker (Eds.), Language Transfer in Language Learning (pp. 222–231). Rowley, MA: Newbury House. Sharwood Smith, M. A. (1989). “Crosslinguistic influence in language loss”. In K. Hyltenstam & L. K. Obler (Eds.), Bilingualism Across the Lifespan (pp. 185–201). Cambridge: CUP. Sharwood Smith, M., & Van Buren, P. (1991). “First language attrition and the parameter setting model”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 17– 30). Cambridge: CUP. Silva-Corvalán, C. (1991). “Spanish language attrition in a contact situation with English”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 189–206). Cambridge: CUP. Soesman, A. (1997). “An experimental study on native language attrition in Dutch adult immigrants in Israel”. In J. Klatter-Folmer & S. Kroon (Eds.), Dutch Overseas: Studies in maintenance and loss of Dutch as an immigrant language (pp. 181–194). Tilburg: Tilburg University Press. Søndergaard, B. (1981). “Decline and fall of an individual’s bilingualism”. Journal of Multilingual and Multicultural Development, 2 (4), 297–302.
Language attrition bibliography
Thompson, R. T. (1982). “Implications of language attrition research for national language policy”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 202–206). Rowley, MA: Newbury House. Tomiyama, M. (1999). “The first stage of second language attrition: A case study of a Japanese returnee”. In L. Hansen (Ed.), Second Language Attrition in Japanese Contexts (pp. 59–79). Oxford: Oxford University Press. Toribio, A. J. (2000). “Code switching and minority language attrition: Papers from the 1999 conference on the L1 & L2 acquisition of Spanish and Portuguese”. In R. P. Leow & C. Sanz (Eds.), Spanish Applied Linguistics at the Turn of the Millennium (pp. 174–193). Somerville, MA: Cascadilla. Toribio, A. J. (2001). “On Spanish language decline”. In A. H. J. Do, L. Dominguez, & A. Johansen (Eds.), Proceedings of the 25th Annual Boston University Conference on Language Development (pp. 768–779). Somerville, MA: Cascadilla. Turian, D., & Altenberg, E. P. (1991). “Compensatory strategies of child first language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 207–226). Cambridge: CUP. Uribe de Kellett, A. (2002). “The recovery of a first language: A case study of an English/Spanish bilingual child”. International Journal of Bilingual Education and Bilingualism (IJBEB), 5 (3), 162–181. Vago, R. M. (1991). “Paradigmatic regularity in first language attrition”. In H. W. Seliger & R. M. Vago (Eds.), First Language Attrition (pp. 241–251). Cambridge: CUP. Valdman, A. (1982). “Language attrition and the administration of secondary school and college foreign language instruction”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 155–175). Rowley, MA: Newbury House. Ventureyra, V. A. G., Pallier, C., & H.-Y. Yoo (2004). “The loss of first language phonetic perception in adopted Koreans”. Journal of Neurolinguistics, 17 (1),79–91 Vries, J. de & Vries, G. de (1997). “Demolinguistic aspects of language maintenance and shift among Dutch immigrants in Canada”. In J. Klatter-Folmer & S. Kroon (Eds.), Dutch Overseas: Studies in maintenance and loss of Dutch as an immigrant language (pp. 121– 137). Tilburg: Tilburg University Press. Waas, M. (1991). “Aspects of non-pathological language loss: Selected bibliography”. In F. Lovejoy (Ed.), Studies in Language Loss (pp. 55–72). Sydney: Holborn Press. Waas, M. (1991). “On language attrition”. In F. Lovejoy (Ed.), Studies in Language Loss (pp. 5–21). Sydney: Holborn Press. Waas, M. (1993). “Loss of first language skills in the community: Intermediate stage”. Language Problems and Language Planning, 17 (3), 225–237. Waas, M. (1994). “‘Deutschies’ und ‘Australianer’. Sprachverlust unter Deutschsprachigen in Australien”. Muttersprache, 4, 312–319. Waas, M. (1996). Language Attrition Downunder. Frankfurt: Peter Lang. Waas, M. (1997). “First language loss: Reflex responses, repartee and sound symbolism”. Language Problems and Language Planning, 21 (1), 119–133. Waas, M., & Ryan, A. (1993). “Onomatopoeia in language attrition”. Linguistische Berichte, 148, 477–482. Weilemar, L. (2003 Generative Theory in Language Attrition Research: A critical evaluation. Unpublished MA Thesis, Vrije Universiteit Amsterdam.
Monika S. Schmid
Weltens, B. (1988). The Attrition of French as a Foreign Language. Dordrecht: Foris. Weltens, B., & Cohen, A. D. (1989). “Language Attrition”. Studies in Second Language Acquisition, 11 (2), 127–133. Weltens, B., & van Els, T. (1986). “The attrition of French as a foreign language: Interim results”. In B. Weltens, K. de Bot, & T. van Els (Eds.), Language Attrition in Progress (pp. 205–221). Dordrecht: Foris. Weltens, B., van Els, T. J. M., & Schils, E. (1989). “The long-term retention of French by Dutch students”. Studies in Second Language Acquisition, 11, 205–216. Ya˘gmur, K. (1997). First Language Attrition among Turkish Speakers in Sydney. Tilburg: Tilburg University Press. Ya˘gmur, K., Bot, K. de, & Korzilius, H. (1999). “Language attrition, language shift and ethnolinguistic vitality of Turkish in Australia”. Journal of Multilingual and Multicultural Development, 20 (1), 51–69. Yoshitomi, A. (1992). “Towards a model of language attrition: Neurobiological and psychological contributions”. Issues in Applied Linguistics, 3 (2), 293–318. Yoshitomi, A. (1994). “The attrition of English as a second language of Japanese returnee children”. Dissertation Abstracts International (DAI), 55 (3), 553A. Yoshitomi, A. (1999). “On the loss of English as a second language by Japanese returnee children”. In L. Hansen (Ed.), Second Language Attrition in Japanese Contexts (pp. 80– 111). Oxford: Oxford University Press.
A new blueprint for language attrition research Monika S. Schmid Vrije Universiteit Amsterdam
Introduction I hesitate to call the present chapter a ‘conclusion’ to either this volume or the conference that begat it. It seems to me that the questions, approaches and findings that were presented do not lend themselves to closure – on the contrary, they open up perspectives even as they reveal shortcomings. I will, therefore, make an attempt here to sum these up and try to look forward as well as back. What has been established beyond doubt here is that great progress has been made over the past decades in the area of theoretical interpretation and evaluation of the phenomena witnessed in language attrition. The frameworks and models which have been applied to research on attrition have become more sophisticated, reflecting the relatively greater wealth of researcher’s experiences. This fine-tuning of how research questions are asked has, in many instances, led to a more and more microlinguistic approach – which has made it possible to answer (albeit tentatively) some very specific questions. The more general ones, however, remain unanswered. On the basis of the studies published in this volume it is, for example, possible to say that language attrition can to some degree affect binding properties of some anaphoric pronouns in both L1 and L2 (see Gürel, this volume; McCormack, this volume), or that different types of morphemes may be affected differently by the attritional process (see Gross, this volume; Schmitt, this volume). It is not possible, however, to re-investigate the findings from the latter studies in the light of the theoretical considerations that informed the first two, or vice versa. Nor is it possible to relate the findings from any of these studies to an overall evaluation of the individual proficiency of the speakers that were
Monika S. Schmid
tested. In other words, while interesting facts and discoveries are being made about specific and local phenomena of language attrition, the global questions remain unanswered and possibly unanswerable. This problem is compounded by the fact that the majority of attrition studies appear to fall into one of two categories. The first category comprises those investigations which inductively proceed from a clearly defined theoretical framework, such as GB, minimalism, the Abstract Level model etc., setting out to investigate one or more predictions made by this theory with respect to language attrition. These studies usually have a local, microlinguistic focus, investigating one or two linguistic features, usually on no more than one linguistic level (e.g. morphology). Sometimes, but by no means always, they also investigate a small number of informants. They typically elicit their data with one or two well-considered tests, such as a grammaticality or truth value judgment. In the second category are those studies that have chosen a deductive approach. Such studies typically use a broader range of elicitation techniques, since an approach that does not depart from a clearly specified and theoretically informed hypothesis but from an observation of data obviously has to cast its net much wider. In these cases, the researcher attempts to approach the data with an open mind, investigating every potentially interesting phenomenon and drawing conclusions from this. What is, as yet, lacking are studies that attempt to amalgamate both approaches, combining data that have been elicited with a wide range of methods and investigated from all possible angles with a theoretically rigid framework.1 Ideally, such an approach would furthermore cover a wide range of languages and a large number of informants. Obviously, this is not a task that can be achieved by one researcher. But it is the necessary next step if real progress is to be made in language attrition research – that is, progress on more fundamental and global issues.
.
Diagnosis
In the light of these considerations, it seems crucial that we now turn ourselves to the task of building a broader base on which language attrition research can be conducted. This fundament will have to rest on four pillars: 1. strong international contacts within the research community 2. a broad and interdisciplinary theoretical perspective
A new blueprint for language attrition research
3. methodological consistency (a common research framework) 4. the relevance of language attrition from the perspective of society The first three of these were explicitly or implicitly addressed by the conference, the fourth was only addressed incidentally as it concerned various specific studies (e.g. Jääskeläinen 2002; Kipp 2002; Schmid 2002). . International contacts It was pointed out in the introduction to this volume that the loss of the close links, both personal and scientific, between researchers on language attrition from the late 1980s onwards brought with it the loss of target-oriented impetus in research that the field had previously enjoyed. Since research in the 90’s was characterised by the solitary existence of the PhD student, many researchers were not aware of each other and each other’s work. One example of this is the case of the PhD theses of Gürel (2002, McGill, Canada) and McCormack (2002, Durham, UK), who investigate similar linguistic features within the same theoretical framework (GB) and with a comparable elicitation technique in different attrition settings (L1 vs. L2) and on different languages (Turkish/English vs. English/Japanese). Due to the lack of a clearly defined attrition research community, both researchers were – until the Amsterdam conference – unaware that related work was being conducted. This problem is compounded by the fact that of the postgraduates who completed a PhD on any area of L1 attrition so far, not one has gone on to conduct a further study on this topic. Senior expertise, combined with a thorough overview of the field, is thus desperately hard to come by – which typically makes the onset of any PhD project comparatively slow and onerous: most graduates feel that it is up to them to re-invent the wheel each time. The organisers of this conference thus felt the necessity to (re-)create a strong international network which would facilitate co-operation, exchange of ideas and research designs. It was also hoped that this network might in time become a hub of information about different projects and a source of support and advice for young researchers and PhD students. The most important step taken in that direction, on the suggestion of Kees de Bot and under supervision of some of the senior colleagues present at the conference, was the formation of a graduate research group. This group of some 14 young and 4 senior researchers from 10 countries has so far met twice in Amsterdam, and further meetings are envisaged on a roughly yearly basis.
Monika S. Schmid
Within this network, and hopefully within a larger, more informal context, it is anticipated that research questions, hypotheses, methods, data and results will inform one another from the research design stage onwards. The first step towards this was the development of a common and crosslinguistically applicable research design (described in more detail below). This design will be relatively easy to adapt to other research conditions, and will hopefully make it possible to arrive at the point where individual studies can compare their findings and thus cease to exist in a vacuum. At least as important as these methodological aspects, however, is an open and constructive debate between various theoretical models. . Theoretical perspectives Language attrition has enjoyed attention from various disciplines (see Köpke & Schmid, this volume). It is easy to see why the topic should be so attractive to linguists of every denomination: while full-fledged, ‘normal’, adult language behaviour is the basis for most theoretical models, testing the hypotheses put forward by various theories on data obtained from such speakers is tricky. This is why speakers who do not conform to one or more of these criteria – acquirers (normal or impaired), L2 speakers, or speakers with pathologically conditioned language impairment – have so often been the focus of research. Not only is attrition a different ‘condition’ from the aforementioned situations, allowing different conclusions to be drawn, but the actual testing procedure may also be less complicated, since many tests are difficult to adapt for children or impaired speakers. And, of course, attriters may also, for various reasons, cooperate more easily in tests than children or aphasics. The studies presented in this volume demonstrate that on the theoretical level, great strides have been made in the past decade. This is doubtlessly and demonstrably the outcome of the wealth of PhD studies – and more than half the articles contained here are indeed reporting on the findings from such projects. A comparison of the present collection with earlier volumes shows that the range and depth of theoretical models informing the study of language attrition has indeed grown more sophisticated. An example of this development is the case of attrition studies within the generative framework. While in the 80s and early 90s these studies wrestled with general and rather broad and vague notions such as competence and performance, the frameworks currently adapted to attrition research allow for the formulation of more precise and testable hypotheses, and therefore more rigorously suggestive findings
A new blueprint for language attrition research
(for a discussion of the development of generative-based attrition studies see Weilemar 2003). However, reviewing the literature on language attrition over the past years leaves one with the feeling that a peak may have been reached in the progress that has been achieved on the theoretical level. One reason for this may be that many of the questions and hypotheses put forward on this front are inherently comparative and crosslinguistic in nature. Consider the following three research questions that were the point of departure for some of the studies presented here: 1. Does the definition of binding domain in the L1 change under attrition? (Gürel) 2. Are content morphemes more vulnerable to attrition than early system morphemes, and are these in turn more vulnerable to attrition than late system morphemes? (Gross) 3. Is there a relationship between ethnolinguistic vitality perceptions and first language attrition? (Ya˘gmur) All of these questions concern phenomena that are highly dependent on characteristics of individual languages, whether typological or sociolinguistic, and it thus seems safe to assume that crosslinguistic investigations may be necessary in order to find generalisable answers. Furthermore, the explanatory value of various linguistic theories has often been put to the test on the basis of language attrition data. What is lacking, however, is an interdisciplinary approach that would assess the power of different theoretical models to account for the phenomena observed in any one study (Schmid 2002 made an attempt towards such an approach, but had to concede that the limited nature of the data under investigation precluded anything but the most tentative conclusions). In my opinion, it is as much the monodisciplinary approach to theoretical explanation that is responsible for the lack of any real progress within attrition research over the past ten years as the fact that no study has taken a crosslinguistic, comparative approach. Both are necessary, and both require co-operation, exchange and openness between researchers. . Methodological aspects: Developing a common research design The third important impediment to progress in language attrition research has been the variety of research designs and data analysis, compounded by the fact that it is rare for any one study to employ more than one method of
Monika S. Schmid
data elicitation. This inconsistency of test methods is clearly evident across the literature on language attrition, including the present collection. In the introduction to this volume, three types of data were identified: data elicited through self-assessments, data elicited by formal tests targeting a particular linguistic structure, and data elicited by inviting spontaneous speech production. While it seems clear that an ideal study of attrition should combine all three kinds of data, since each of them allows assessment of different factors which the others cannot capture, the reality of attrition research appears very far from achieving this goal: Of the studies included in the present collection, only two (Montrul and Ya˘gmur) include data of more than one type, and none include data from all three. The idea of a ‘blueprint for research’, which was first invoked by Andersen (1982), seems even more compelling 20 years down the line. However, on the basis of the work conducted in the meantime, we are in the position of putting forward a suggestion for such a blueprint that is somewhat more detailed and concrete, containing a number of specific tests and data elicitation methods (see below, Section 2). These ideas were developed within the graduate network and are currently being adapted to and pilot-tested in a variety of languages. Having conducted the analyses, it will then be possible to provide detailed instructions pertaining to the adaptation and administering of the tests as well as to their coding and analysis. Scientists will then be invited to contribute the data and analyses obtained by these methods to a common ‘pool’ of language attrition data, which will be shared across the research network. In this way, it will be possible to obtain results and answers that are far more detailed and reliable than has previously been the case. . Societal relevance .. Multilingualism The year 2001 was the European Year of Languages, and the EU Commission’s stated objective in announcing this event was the attainment of “a European Union where everybody speaks several languages. All those leaving compulsory education should be able to communicate in at least two European languages in addition to their mother tongue and then be able to build on that knowledge for the rest of their lives.” (http://www.europa.eu.int/comm/education/policies/ lang/languages/actions/year2001_en.html, my emphasis) An examination of the actions and outcomes of the work undertaken reveals an impressive body of work – which, however, exclusively focuses on the languages spoken in ad-
A new blueprint for language attrition research
dition to the mother tongue. In other words, all efforts have been going to the teaching of second and foreign languages, while the preservation of the first language and its transmission to the next generations in the case of migrants seems to be deemed a non-issue. The point to be made here is that there is a danger of neglecting – and thus losing – a great wealth of language proficiency, the preservation of which would cost far less effort than the teaching of an unknown language from scratch. It thus seems vital to investigate the processes and factors underlying the preservation, attrition, and early acquisition of heritage languages at this point, particularly since these questions will be of growing interest with the increasing inter-European mobility expected over the next decades. However, the prevailing opinion in many European countries seems to be that knowledge of the heritage language is something that is unnecessary at best and detrimental to integration into the dominant society at worst. Language is one of the most important markers of a ‘deviant’ identity of migrants to those who subscribe to xenophobic ideas, and linguistic assimilation the most often and loudly demanded act of integration on the part of an immigrant. The prevalent idea seems to be that linguistic assimilation works like an organ transplant, where the defective and useless L1 is surgically removed and the more useful dominant language is put in its place. In this vein, the Dutch cabinet has recently announced plans for the cancellation of all funding for teaching in heritage languages in public schools, since the focus is to be on the teaching of Dutch and all across the EU, the opinion that immigrants should make the switch to the dominant language even in their homes is being heard. While proficiency in the dominant language is clearly an important asset for anyone who is a part of a certain society, the idea underlying such demands – that the sacrifice of the first language is a necessary step within this integration process – is unsupported by linguistic research. It seems essential to make a step towards re-interpreting multilingualism not as an indicator of deficiency, but as an asset to any community. This is already the case for the so-called elite bilingualism. The current situation is thus that, for example, English- or German-speaking parents in the Netherlands consider their language a valuable asset for their children, and make efforts to help them achieve fully native-like proficiency – efforts which are encouraged by the host community. Turkish parents, on the other hand, will at every step be confronted with the attitude that it is detrimental for their children to learn Turkish, that this will impair their chances for assimilation and career opportunities, and that education in the monolingual Dutch mode is really preferable.
Monika S. Schmid
One consequence of this long-standing attitude towards non-prestige immigrant languages is that subsequent generations typically have a dramatically reduced proficiency in their heritage languages. This, however, does not seem to lead to the desired effect of full-scale integration into the host society. In recent years, it has been observed that especially young speakers from the most important migrant groups (e.g. Turkish and Moroccan Arabic in the Netherlands, Turkish and Polish in Germany, Arabic in France etc.), even though clearly being dominant in the host language, are developing a ‘contact-like’ variety of this language as a marker of their identity – e.g. ‘Türkendeutsch’ (Androutsopoulos 2001). These varieties, sometimes stigmatised as ‘bad Dutch’ or ‘bad German’, are far more detrimental to the integration of these speakers into and acceptance by the host society than full-fledged bilingualism might be. Research into the preservation and loss of migrant languages has usually focused on the second generation. However, the key to whether or not a language is passed on, and to what degree – i.e. the key to large-scale bilingual education which is in effect cost-neutral to the community – lies in the first generation (see Py & Grosjean 2002:22). In view of the goal of a multilingual European Union, where everyone is able to communicate in at least two languages in addition to their native one, it thus seems vital to a. assess the factors that have a favourable effect on the maintenance of first language proficiency in immigrants as opposed to the societal conditions that will be conductive to first language loss/attrition in order to enable governments and educators to help maintain proficiency in the first language in addition to that on the foreign/second languages b. assess the factors that have a favourable effect on the transmission of first language proficiency to subsequent generations of immigrants in order to enable governments and educators to help maintain proficiency in the first language across immigrant generations. .. Endangered languages In a keynote presentation to the 25th annual meeting of the Deutsche Gesellschaft für Sprachwissenschaft in 2002, Dieter Wunderlich identified as the most pressing and important concern for linguistics in the 21st century the quest to save or at least document endangered languages. This surely is a view that many would share, and it is reflected in the fact that many funding bodies,
A new blueprint for language attrition research
both in the governmental and in the business sector, have special programs for research on endangered languages. However, one of the most difficult methodological problems research on endangered languages faces is the fact that if a speech community is dramatically reduced in numbers, it is very difficult to assess to what degree the original linguistic system is still intact, and to what degree it has eroded or changed due to language contact. Since there typically is no written tradition and thus no documented history of endangered languages, it may be all but impossible to reconstruct the language in its pre-critical state. If we assume the processes underlying language attrition and language death to be similar (and all findings thus far indicate that they might well be), language attrition can provide very valuable findings for language death studies: The comparison of the linguistic system of the attriter with that of the non-attrited monolingual speaker in the speech community of origin may allow us to discover the rules that govern the attritional process – and these, in turn, may be invaluable for the reconstruction of dead or dying languages. . Conclusion In the introduction to this volume, we summed up the purposes that the project which led to this volume was intended to achieve, namely: 1. To arrive at a clearer, and more explicit, picture of the theoretical models that have been used to account for the findings in language attrition studies than is usually presented 2. To achieve a better communication between researchers engaged in language attrition studies, in order to make it unnecessary for everyone to ‘re-invent the wheel’ at the outset of their study 3. To achieve a level of consciousness about the benefits and disadvantages of particular research designs and data elicitation methods 4. To begin forming an international network of attrition researchers 5. To achieve communication between established scientists and young researchers We are now in the position to formulate a number of action points to continue progress in these directions: 1. Set up a research project designed to use data which it is possible to evaluate from different theoretical angles; share multilingual and comparable data across projects and approaches. This will hopefully make it possible to test
Monika S. Schmid
the explanatory power of various theoretical frameworks, to show where they can account for the phenomena witnessed in attrition across different languages – and where they cannot. 2. Achieve international visibility of the language attrition network through continued regular (yearly) graduate workshops, a follow-up conference, a language attrition discussion list, as well as through the more informal contacts each of us have. 3. Develop a unified language attrition research design, including tests that will elicit the three different kinds of data. Develop clear guidelines for the setting up, administering, analysing and interpreting tasks involved with each of the tests. The first two of these speak for themselves (and are being tackled even as I write this). On the third, some progress has been made towards such a design, which will be outlined below.
. Prescription: A common research design The preliminary considerations involved in deciding on a variety of tasks to elicit data to use in language attrition research suggest that the test battery should consist of tests that a) tap into various aspects of language attitudes and linguistic proficiency, b) are recognised for their validity (and, ideally, have been standardised across the non-attrited population) and c) are economic in their usage, that is to say, allow for data collection and analysis within a reasonable amount of time and are short enough for the testing procedure to take no more than two to three hours (in order to avoid informant fatigue). On the basis of these deliberations, it has been considered reasonable to include the following set of elicitation methods: I.
Introspection and self-assessments a. Sociolinguistic questionnaire: This instrument is an indispensable part of any study involving multilinguals, considering the large variation encountered in this type of subjects (e.g. Grosjean 1998). The questionnaire that we propose to use is an amalgamation of earlier instruments of this type used in language attrition research (e.g. Hulsen 2000). It contains questions on the subject’s background, language behaviour and linguistic environment.
A new blueprint for language attrition research
b. Can-Do Scales: This instrument provides an assessment of selfperceived language proficiency in both L1 and L2. Self-assessments have often been criticised, especially when used as the only measuring instrument for any one study (Hulsen 2000; Schmid 2002, see also Schmid & Köpke, this volume), since they often do not correlate with actual proficiency revealed by formal tests. However, in other cases, self-assessments did apparently correspond to actual language skills, and one of the purposes of including them in the research design is to investigate the validity of self-assessments in more detail in a broader context. For the present research design, the ALTE can-do statements of the Council of Europe’s Common European Framework of Reference for Language Learning and Teaching (www.alte.org/can_do) are available as a standardised tool. These statements are subdivided into different levels of everyday activity within the four domains of listening, speaking, reading and writing. Since the statements were primarily devised with a view to L2 acquisition, they start at a relatively low level of proficiency. Since we do not anticipate L1 attrition among adults to ever reach such dramatic proportions (deterioration to beginner level never having been documented in such cases), the basic levels of the statements (A1-A2) have been removed from the scales. c. Matched guise experiment: This technique focuses on language as a social identity marker. The participant is asked to listen to a passage of text read by several (fully native-like bilingual) speakers, and to score these with respect to fifteen character traits on a 7-point Likert scale (see Lambert et al. 1960). The participant is unaware that the speaker(s) recording the text in the different languages are, in fact, the same person. This task is commonly used to detect differences in attitudes towards the various languages or varieties of a multilingual or multilectal speaker. Since attitudinal factors have often been evoked, but seldom rigidly demonstrated, in language attrition studies, this test seemed an appropriate measure. II. Formal elicitation tasks a. C-test: The first test in this battery has been designed as a broad measure of overall language proficiency. It is based on earlier versions of the cloze test2 and consists of 5 or 6 small texts (around 70 words) in which the second half of every second word has been removed, usually starting with the second sentence (for a detailed description of the C-test
Monika S. Schmid
cf. Grotjahn 1992). The informant’s task is to complete the missing parts. This requires the speaker to make full use of the natural redundancy of a text, which makes it possible to measure not only relatively low level skills (command of vocabulary, grammar, idioms) but also higher order skills such as awareness of intersentential relationships, global reading, etc. Further advantages of the C-test are that it is based on authentic material and that it is relatively easy to use and to score. It has been extensively studied and found to be a highly valid and reliable predictor of overall proficiency.3 b. Verbal fluency task: This task is aimed at tapping specifically into lexical retrieval, which has been recognised to be one of the most important and earliest consequences of first language loss (e.g. Köpke 1999). The informant is required to produce as many words of a certain type as possible within a period of 60 seconds. The task may focus on semantic aspects of lexical retrieval by eliciting words of a specific semantic category (animals, vegetables, etc.) or on phonological aspects focussing on words starting with the same letter/sound. The fluency task is known to be highly reliable even in case of light retrieval difficulties and has the advantage of requiring less material and time than other tests detecting lexical retrieval difficulties, such as e.g. picture naming tasks. c. Grammaticality judgments: Despite some criticism and drawbacks, this test has been recognised as one of the best means to investigate specific morpho-syntactic features (see Altenberg & Vago, this volume). III. Spontaneous speech a. The Chaplin film retelling task: This instrument to elicit spontaneous speech has been developed within the ESF (European Science Foundation) network on learner languages (for a detailed description see Perdue 1993). It is recognised to produce relatively free speech data which are still easier to compare across informants than data elicited by informal conversation. Despite the considerable amount of work necessary to transcribe and analyse spontaneous speech data, this nevertheless remains the only means of getting an overall impression of the informant’s real language proficiency. Spontaneous speech data allow not only to look for problematic areas that have not yet been detected, they also provide the researcher with proficiency data, showing what the immigrant is still able to do in his language via
A new blueprint for language attrition research
type/token, stylistic analysis, overall morphosyntactic complexity etc. (see Schmid, forthc.). b. Semi-structured interview: At the end of the testing procedure, the researcher will engage the informant in an (apparently) informal conversation, inviting opinions on the testing procedures, but also on language skills etc. This conversation will be guided by a catalogue of questions, which are posed in a casual way to avoid the impression that this interview is still part of the testing procedure. We hope that the research design outlined here will make it possible to give a more detailed and more suggestive picture of language attrition in process. For each immigrant community, it will be possible to determine rates of language loss on various levels of language and communication in comparison with each of the other languages investigated within the network. That the results are comparable is ensured by the setup, which will evaluate findings not directly across languages but in relation to the native-speaker control group scores. To illustrate this last point: one of the research instruments, the filmretelling task, will attempt to evaluate the extent of loss in the lexicon on the basis of type-token frequencies (among other measurements). Type-token frequencies assess how many different lexical items appear in a given stretch of linguistic data, i.e. how often each word is used on average. A larger number of different words will indicate a larger mental lexicon and manifest itself in a lower type-token-frequency score, since more words are present and consequently, each different word is used less often. However, the average scores of proficient native speakers will vary greatly across languages, since a highly inflecting language, like German (or Arabic), will have more bound morphemes (inflectional endings, infixes etc.) and thus a lower average type-token frequency than a language like English which relies on grammatical words like prepositions etc. to make the distinctions that the former might make through, for example, case marking. Since grammatical words are typically very high frequency, this will result in different overall type-token scores for each language. In a crosslinguistic investigation of language loss, it is therefore necessary to calculate these scores not in absolute figures, but in percentages of the average native speaker score. This research design is commonly being developed, adapted, tested and applied for a number of languages (including Dutch, English, German, Russian and Spanish). We will be happy to make it available to anyone who wishes to apply it, assist with the adaptation to other languages where such help is needed and provide support for the data analysis.
Monika S. Schmid
. Conclusion A word-count on the manuscript for this volume, which at the time of writing is as good as ready to go to the publishers, reveals that it is at present just under 125,000 words. I should say that that was enough, and will close with a quotation from Goethe’s Faust: Der Worte sind genug gewechselt, Lasst mich auch endlich Taten sehn! Indes Ihr Komplimente drechselt, Kann etwas Nützliches geschehn. (Enough words have been exchanged, now let us proceed to action! While you fiddle around constructing compliments useful things could be done.)
Notes I would like to thank Barbara Köpke and Chris McCully for their helpful comments on earlier versions of this article. . There are some notable exceptions to this point within sociolinguistic and psycholinguistic frameworks, e.g. Hulsen 2000, Köpke 1999 and Ya˘gmur 1997. . A comparison between the C-test and the cloze-test shows that the C-test is more reliable with advanced speakers (which immigrant L1 speakers most likely are) whereas the clozetest is more reliable with lower level speakers. . It has to be noted that the C-test, as well as some other tests included in this category, may be a somewhat problematic instrument in a test battery that aims to achieve crosslinguistic validity: Tests that rely extensively and exclusively on written material cannot be adapted to languages that do not have a writing system. If and when the network is expanded to include languages for which this is the case, solutions to this problem will have to be found.
References Andersen, R. W. (1982). “Determining the linguistic attributes of language attrition”. In R. D. Lambert & B. F. Freed (Eds.), The Loss of Language Skills (pp. 83–118). Rowley, MA: Newbury House. Androutsopoulos, J. (2001). “Ultra korregd Alder! Zur medialen Stilisierung und Aneignung von ‘Türkendeutsch’.” Deutsche-Sprache, 29 (4), 321–339.
A new blueprint for language attrition research
Grosjean, F. (1998). “Studying bilinguals: Methodological and conceptual issues”. Bilingualism: Language and Cognition, 1, 131–149. Grotjahn, R. (Ed.). (1992). Der C-Test. Theoretische Grundlagen und praktische Anwendungen. Band 1. Bochum: Universitätsverlag Dr. N. Brockmeyer. Gürel, A. (2002). Linguistic Characteristics of Second Language Acquisition and First Language Attrition: Turkish overt versus null pronouns. Unpublished Doctoral Dissertation, McGill University Montréal. Hulsen, M. (2000). Language Loss and Language Processing. Three generations of Dutch migrants in New Zealand. Unpublished Doctoral Dissertation, Katholike Universiteit Nijmegen. Jääskeläinen, K. (2002). “Language attrition and language prejudices”. Paper presented at the International Conference on First Language Attrition, Amsterdam Aug. 2002. Kipp, S. (2002). “Social networks, individual biography and individual difference – their impact on language use 30 years on”. Paper presented at the International Conference on First Language Attrition, Amsterdam Aug. 2002. Köpke, B. (1999). L’attrition de la première langue chez le bilingue tardif: Implications pour l’étude psycholinguistique du bilinguisme. Unpublished Doctoral Dissertation, Université de Toulouse-Le Mirail. Lambert, W. E., Hodgson, R. C., Gardner, R. C., & Fillenbaum, S. (1960). “Evaluational reactions to spoken language”. Journal of Abnormal and Social Psychology, 60, 44–51. McCormack, B. (2001). The Maintenance and Loss of Reflexive Anaphors in L2 English. Unpublished Doctoral Dissertation, University of Durham, UK. Perdue, C. (1993). Adult Language Acquisition: Cross-linguistic perspectives. Cambridge: CUP. Py, B., & Grosjean, F. (2002). “Compétence bilingue: approche expérimentale”. La notion de compétence en langue. Notions en questions, 6, 19–27. Schmid, M. S. (2002). First Language Attrition, Use and Maintenance: The case of German Jews in Anglophone countries. Amsterdam and Philadelphia: John Benjamins. Schmid, M. S. (Forthcoming). “Measuring first language attrition in free speech: What’s in a mistake?”. International Journal of Bilingualism. Weilemar, L. (2003). Generative Theory in Language Attrition Research: A critical evaluation. Unpublished MA Thesis, Vrije Universiteit Amsterdam. Ya˘gmur, K. (1997). First language Attrition among Turkish Speakers in Sydney. Tilburg: Tilburg University Press.
Author index
A Ager 180 Ahmed 62 Aksu-Koç 148 Alanen 118 Allen 109 Altenberg 10, 17, 18, 25, 26, 31, 32, 105, 106, 121, 200, 283, 319, 323, 328, 330–332, 360 Ammerlaan 7, 10, 12, 20–23, 26, 34, 35, 137, 139, 140, 147, 319, 328, 330–333 Andersen 3, 22, 135–138, 143, 149–151, 161, 193, 195, 275, 282, 285, 326, 327, 354 Anderson 319, 329, 330, 332 Andersson 50 Andrews 48, 319, 328, 330–332 Androutsopoulos 356 Appel 67, 78, 178 Argenti 208 Armon Lotem 319, 328, 330, 332 Aronoff 9, 17, 283, 321, 328, 330–332 Arthur 106 As 34 Au 208 Auer 168 Avermaet 7 B Bader 108, 120 Bard 116, 121 Barzilay 21, 26, 148, 185, 284, 321, 328, 331, 332 Bassano 120
Basser 208 Belcher 81, 84 Ben-Rafael 165, 171, 181, 182, 185, 319, 328, 330, 332, 333 Beniak 178, 302 Berko-Gleason 61, 189, 326, 327 Berman 10, 148, 324, 329–332 Berry 109, 111 Besemeres 81, 101 Bettoni 4, 17, 147, 319, 328, 330–332 Bever 108, 109 Bialystok 109 Birdsong 106, 110, 111, 115–117, 119, 121–123 Blackwell 114, 120, 123 Blanc 168 Blanche-Benveniste 180 Bley-Vroman 121 Bock 287 Bode 9, 13, 26, 319, 328, 330–332 Bolonyai 7, 17, 285, 302, 319, 328, 330–332 Bonch-Bruevich 7 Bond 81 Bot 4, 5, 7–9, 12, 14, 16, 17, 19, 22, 25, 26, 31, 33, 50, 61, 82, 92, 98, 101, 105, 112, 114, 118, 119, 122, 134, 136, 138, 147, 189, 191, 192, 194, 198, 203, 231, 237, 240, 254, 281, 285, 319, 321, 323, 326, 328, 330–332, 351 Bouba 12, 18, 259, 266 Bourhis 136, 149 Boyd 50 Bradac 122
Author index
Brewer Bomar 16, 319, 329–331 Brons-Albert 195, 319, 328, 330–332 Broselow 255 Brown 12, 319, 328, 330, 332, 333 Bruhn-Garavito 233 Buren 17, 18, 21, 61, 193, 244, 245, 276, 327 C Cairns 106, 111, 116, 117, 121, 124 Canagarajah 55 Caplan 25 Caramazza 20 Chan 266, 274 Chantefort 178 Chaudron 106, 107, 110, 115, 120 Chen 244 Chernitsky 53 Choi 218 Chomsky 107, 226, 227, 243, 244, 246, 259, 265 Clahsen 18, 26 Clark 3, 25, 137, 138, 326 Clyne 12, 50, 147, 168, 189, 191, 198, 203, 326, 328, 330, 332 Cohen 4, 76, 148, 324, 326, 329, 330, 332 Cole 65 Comrie 310 Connor 55, 81, 84 Cook 47, 48, 87, 260 Coughlan 68 Coveney 181 Cowan 113 Cowart 106, 109, 113, 117–119, 122–124 Crain 268 Curtiss 208 Cuza 18, 27 D Dörnyei 85, 88, 91, 102 Dal Negro 27, 141, 142 Damasio 289
De Guerrero 62, 78 De Kadt 27, 141, 142 Dehaene 208 Dekydtspotter 233 Dewaele 12, 31, 81, 86, 88, 89 Digman 90 Donato 62 Dorian 26, 168, 178, 293 Dufour 89 Dugas 26, 324, 329, 333 Dupoux 208, 217 Durkin 110 Dussias 21, 31 E Eckman 108, 237, 240 Edwards 136 Eefting 52 Eichen 208 Ellis 113, 114, 117, 120–122 Elordui 27, 141, 142 Els 3, 4, 8, 9, 33, 138, 326 Enç 239 Engeström 65 Ervin-Tripp 52 Eysenck 90 F Faerch 70 Fase 7, 136, 189 Ferguson 136 Fernández 118, 124 Ferrand 218 Field 168 Filiaci 259 Finer 255 Flege 52 Flynn 115 Frawley 62, 67, 71, 73, 74, 78 Freed 2–4, 33, 135, 326 Frei 180 Frenck-Mestre 11 Fromm 21 Fujita 324, 329, 331, 332 Fuller 285
Author index
G Gürel 12, 17, 18, 23, 25, 26, 31–33, 225, 235, 239, 245, 246, 259, 260, 320, 329–333, 349, 351, 353 Gadet 180, 181 Garcia 49, 51 Gardner 12, 324, 326, 329–333 Gass 107 George 229, 250 Giesbers 191, 193, 320, 328, 330–332 Giles 13, 136 Ginsberg 3, 138, 326 Giorgi 265 Gommans 191, 192, 203, 231, 328, 331, 332 Gonzalez-Reigosa 81 Gonzo 135, 136, 139, 161 Goodglass 26 Gopnik 218 Goral 8, 26, 34 Goss 113 Grabois 9, 13, 20, 53 Grainger 20 Green 22, 23, 87, 240 Greenberg 106 Greidanus 114 Grosjean 17, 21, 25, 27, 48, 105, 320, 329–333, 356, 358 Gross 17, 31, 32, 281, 289, 293, 301, 320, 328, 330–332, 349, 353 Grotjahn 360 H Haarmann 136 Hagège 166, 178 Hagen 25, 285, 326 Hakuta 20, 25, 320, 329, 331–333 Hall 26 Hamers 168 Han 107, 118, 120, 121 Hansen 16, 61, 244, 324, 326, 329, 331–333 Harley 19, 20 Harres 147
Harris 83, 84, 101, 102 Hatasa 113 Haugen 136, 193, 195, 198, 293 Hawkins 266, 274 Hedgcock 16, 61, 324, 329, 331, 332 Heycock 259 Hickerson 52 Hiller-Foti 4, 165, 320, 328, 330–332 Hiramatsu 111 Hirose 217 Hirvonen 27, 320, 328, 330, 333 Hlavac 285, 293 Hout 30, 31, 87, 139, 189, 326 Hughes 194 Hulk 275 Huls 320, 329, 330, 332 Hulsen 7, 13, 14, 21–23, 25, 26, 61, 82, 88, 92, 98, 137, 139, 320, 328, 330, 331, 333, 358, 359, 362 Hupet 282 Håkansson 19, 50, 61, 300, 320, 327, 329–331, 333
I Isurin 9, 13, 17, 26, 320, 328, 331, 333
J Jääskeläinen 30, 31, 351 Jacobson 168 Jake 17, 32, 286, 300–302 Jakobson 34, 326 Jarvis 48, 50, 52, 53, 320, 328, 330–333 Jaspaert 7, 10, 12, 14, 20, 26, 87, 115, 121, 136, 139, 147, 189–192, 194, 196, 199, 321, 322, 326, 328, 330–333 Javier 81 Jordan 26 Jordens 16, 25, 30, 31, 116, 231, 321, 328, 330, 331, 333 Jun 208
Author index
K Köpke 1, 11–14, 17, 18, 20–23, 25, 26, 34, 105, 106, 120, 122, 134–137, 141, 165, 166, 168, 225, 231, 259, 260, 321, 326, 328, 330–332, 352, 359, 360, 362 Kail 120 Kakehi 217 Kanno 229, 232 Kaplan 26 Kasper 70, 71 Kaufman 7, 9, 17, 19, 141, 142, 283, 300, 321, 327, 328, 330–332 Kawasaki 233 Keijzer 16, 31, 33 Kellerman 71, 106, 116 Kemper 282 King 178 Kinginger 101 Kipp 14, 30, 351 Klatter-Folmer 7, 327 Klein 115, 121, 322, 323 Knightly 208 Knowlson 82 Knowlton 289 Kornfilt 229, 239 Krashen 110 Kravin 276, 321, 328, 331–333 Kroll 89 Kroon 4, 7, 9, 10, 12, 14, 20, 26, 33, 34, 87, 136, 139, 189–192, 194, 196, 199, 321, 322, 326–328, 330–333 Kuhberg 10, 16, 324, 329–332 Kunst-Wilson 218 Kurbel 295 Kynette 282 L Lüdi 168 Lachman 11 Lai 81 Lakshmanan 117 Lambert 2–4, 12, 33, 135, 189, 190, 327, 359
Lantolf 66, 67, 71, 73, 74, 78 Lasnik 227, 246 Latomaa 54 Laufer 51, 321, 329–333 LeBihan 208 Leeman Bouix 178 Leisiö 17, 321, 329, 331, 332 Lenneberg 20, 208 Leont’ev 62, 64, 65 Levelt 286, 287, 295, 301 Levin 301 Linden 114 Linebarger 114 Long 19 M Mägiste 11 Macevichius 27, 141, 142 Maher 17, 327 Major 17, 22, 52, 259, 321, 328, 330, 331, 333 Manzini 226, 230, 243 Marcos 81 Martin 98 Martineau 178 Martohardjono 115, 121, 124 Matos 218 Mayberry 208 Mayer 148, 167 McCafferty 62, 78 McCormack 18, 25, 32, 226, 243, 249, 324, 329, 331, 332, 349, 351 McCreary 68 McDaniel 111, 116, 117 McLaughlin 259 Mehler 208, 217 Meng 108, 120 Merino 300 Miceli 20 Michael 89 Milroy 14 Minkowski 12 Mistler-Lachman 11 Mond 320 Montalbetti 229
Author index
Montrul 9, 19, 26, 30, 32, 259–261, 265, 267, 274–277, 321, 329–333, 354 Moore 3, 327 Moreau 181 Mori 54, 55 Morton 20 Mougeon 178, 302 Müller 275 Munnich 112, 114 Münstermann 25 Murphy 122 Musen 289 Muysken 18, 30, 31, 178, 196, 323 Myers-Scotton 168, 283, 286–288, 293, 300–302, 312
N Nagata 122 Nespoulous 34, 120, 122 New 13, 105, 184, 218 Newport 208 Nicoladis 9, 13, 20 Novakovich 82–84, 101
O Obler 61, 331 Odlin 196, 198, 201 Oh 208 Olshtain 10, 16, 20, 21, 26, 61, 148, 185, 284, 321, 325, 328–332 Otheguy 49, 51 Oxford 3, 137, 138, 327 Özsoy 239 Öztürk 239
P Pallier 10, 14, 20, 32, 207, 208, 217, 218 Pan 189, 327 Paolillo 121 Paradis 23, 34, 109–112 Pater 233
Pavlenko 12, 17, 31, 33, 47, 48, 51, 53, 54, 56, 81, 88, 100–102, 165, 168, 226, 260, 284, 322, 327, 328, 330–332 Pelc 10, 11, 17, 18, 20, 25, 31, 34, 322, 328, 330–332 Penke 25, 26 Perdue 360 Pergnier 168 Pfaff 300 Pianesi 265 Pinker 107 Platzack 19, 61, 327 Polinsky 25, 27, 105, 106, 259–261, 300, 322, 328, 330, 332, 333 Poplack 168 Posner 181 Poulisse 71 Prévost 233 Pritchet 117 Progovac 255 Py 4, 17, 21, 25, 50, 105, 320, 322, 329–333, 356 R Raidt 17, 322, 328, 330, 331, 333 Rappaport 301 Reber 109 Rice 113 Richards 135 Rochon 25 Roediger 115 Romaine 193, 302 Rossing 191, 192, 203, 231, 319, 328, 331, 332 S Söter 81, 82, 100 Salmons 283 Saltarelli 135, 136, 139, 161 Sanders 181 Schütze 106, 115–117, 119, 122, 123 Schachter 105, 107, 117, 118, 121 Schelstraete 282 Schlyter 275, 276
Author index
Schmid 1, 12–18, 23, 27, 28, 31, 34, 50, 61, 82, 86, 105, 106, 124, 134–137, 141, 165, 166, 168, 178, 192, 193, 195, 198–202, 204, 225, 231, 259, 281, 283, 285, 293, 317, 322, 326, 328, 330–333, 349, 351–353, 359, 361 Schmitt 10, 16, 17, 31, 32, 299, 300, 302, 322, 328, 330–332, 349 Schoenmakers-Klein Gunnewiek 14, 26, 27 Scholes 116 Seidenberg 109 Seliger 5, 10, 15, 17, 18, 21, 25, 34, 50, 105, 106, 121, 165, 226, 276, 283, 284, 323, 327, 328, 330–333 Shapard 82–84, 101 Sharwood Smith 4, 17, 18, 21, 22, 61, 105, 106, 111, 193, 244, 245, 276, 284, 323, 327, 328, 330–333 Shimron 53 Silva-Corvalán 259, 323, 329–332 Slabakova 261, 265, 267, 274–277 Slobin 148–150, 159, 161, 303 Smith 4, 17, 18, 21, 22, 61, 105, 106, 111, 124, 134, 193, 244, 245, 276, 284, 323, 327, 328, 330–333 Smits 27, 141, 142 Snyder 118 Soesman 12, 189, 191, 323, 328, 330, 331, 333 Sorace 18, 19, 61, 112, 117, 118, 121, 122, 259, 266, 275 Squire 289 Stanovich 110 Stavans 82–84, 102 Strating 7, 137, 139 Stromswold 289 Supalla 208 T Tao 53 Tarone 71, 123 Taylor 136 Tees 207, 208
Teranishi 117 Thomas 226, 254, 256 Thompson 53, 327 Thornton 268 Timberlake 310 Tokowicz 89 Tomioka 114 Toribio 61, 323, 329–333 Trapman 231 Tsimpli 260 Turian 10, 323, 328, 331, 332 U Ullman 289 Ushakova 62 V Vago 5, 10, 17, 31, 33, 34, 50, 105, 283, 323, 327, 328, 330, 332, 360 Valdés 260 Valois 302 Vendler 262 Vergnaud 311 Verkaik 76 Vihman 259 Vygotsky 62, 63, 65–67 W Waas 11, 13, 21, 26, 50, 61, 88, 140, 147, 165, 166, 191, 201, 323, 328, 331, 333 Walter 178, 180 Walters 134 Wang 19, 20 Waters 25 Wei 293 Weilemar 31, 327, 331, 353 Weinreich 195 Weltens 4, 5, 8, 16, 33, 138, 194, 254, 325, 326, 329, 331–333 Werker 207, 208 Wexler 226, 230, 243 White 226, 233, 237, 240, 266 Wijst 76
Author index
Williams 53 Willis 116 Wilson 134 Winter 109 Wulfeck 114, 121
Yamamoto 98 Yip 107, 117 Yoshitomi 61, 244, 325, 327, 329–333
Y Yaˇgmur 27
Z Zajonc 218
Subject index
4-M model(s) 32 A Access 11, 21, 67, 69, 78, 88, 105, 107, 115, 120, 144, 196, 219, 245, 302, 303, 319, 327 Activation 21–23, 81, 87, 209, 211 Adoptees 32, 208–214, 216, 217, 219 Agrammatism 25 Anger 85, 86, 89–91, 93–95, 97, 99–101 Aphasia 5, 16, 25, 34, 121, 134, 317, 326 Arabic 3, 320, 328, 356, 361 Aspect 261, 262, 265, 267, 275, 276 Attitude 12–14, 107, 319, 355, 356 B Binding 32, 226–230, 233, 237, 239, 243–245, 247, 248, 250, 254–256, 320, 324, 349, 353 Borrowing 32, 47–50, 61, 190, 193, 195, 196, 198, 200, 293, 322 Brain imaging 209 Brain plasticity 207, 219, 327 C Case 283, 285, 287, 292, 302, 308, 309–315 Case-marking 200, 229, 283, 285, 301, 313, 361 case study/case studies 5, 140, 193, 261, 320 Children 9, 10, 13, 19, 20, 26, 32, 34, 62, 66, 68, 69, 78, 83, 111–114,
116, 135, 141, 150, 195, 208, 212, 218, 259, 275, 299, 300, 303–307, 324, 326, 352, 355 Chinese 3, 212 CI 47, 48, 52, 53 CLI 17, 18 Cloze test 114, 231, 332 Codeswitching 29, 32, 168, 184, 288, 293, 294, 299, 300, 304–307, 313–315 Competence 9, 11, 16, 21–23, 48, 54, 55, 61, 87, 89, 106–109, 112, 116, 118, 120, 124, 136, 142, 161, 166, 167, 189–194, 197, 204, 225, 226, 230, 239, 245, 260, 261, 266, 276, 277, 281, 323, 325, 327, 332, 352 Conceptual attrition 56 Contact 12, 14, 17, 19, 20, 22, 23, 30, 32, 134–136, 139, 148–150, 155, 159, 161, 165, 166, 168, 179–185, 191–193, 195, 197, 198, 203, 207, 212, 231, 237, 239, 240, 266, 276, 281, 283, 287, 289, 290, 293, 294, 299, 313, 319–321, 323, 326, 357 Convergence 47, 52–54, 114, 115, 181, 266, 285, 287, 294, 299, 300, 302, 305–307, 309, 313–315, 320 Critical period 19, 20, 207, 208, 327 Crosslinguistic influence 17, 47, 190, 201, 204, 283, 327 D Dementia 25 Dutch 4, 13, 17, 35, 85, 89, 142, 147, 191, 192, 293, 319–329, 355, 356, 361
Subject index
E Editing tests 26 Education level 10, 11, 23, 34, 86 Emotion 21, 51, 81, 84, 86, 88, 93, 100, 176 English 12, 18, 22, 32, 48–56, 68, 75, 78, 81–85, 89, 97–99, 112–114, 117, 141, 142, 144–147, 150, 151, 156, 159, 193, 195–202, 217, 225–227, 229–234, 237–240, 243, 246–248, 250, 251, 254–256, 262–266, 273, 283–285, 287–289, 291, 293, 300, 302–307, 309–315, 319–325, 328, 329, 351, 361 Ethnolinguistic vitality (EVT) 13, 134, 146–149, 154, 323, 353 Explicit knowledge 23, 109–114, 116, 120, 124 Explicit memory 218
F Feelings 49, 55, 56, 85, 86, 90, 92–94, 100, 101, 176, 183, 184 Film retelling 360 Finnish 50, 52, 54, 55, 85, 212, 320–322, 327–329 fMRI 208, 210, 211, 219 free speech 69, 332, 360 French 12, 32, 82, 85, 89, 97–99, 165–175, 177–184, 208–219, 319–325, 328, 329
G Garden-path sentences 107 German 13, 32, 81, 85, 97, 98, 140, 142, 147, 192, 193, 198–202, 282–285, 287–291, 295, 319–324, 328, 329, 356, 361 Grammatical norms 109 Grammaticality 11, 12, 22, 28, 31, 52, 105–124, 229, 246, 248, 251, 261, 281, 283, 295, 332, 350, 360 Grammaticality judgment 11, 12, 105–115, 117–124, 248, 251, 283
Greek
11, 18, 85, 98, 212, 322, 328
H Hebrew 51, 53, 166, 168–172, 174, 175, 177–184, 212, 319, 321, 323–325, 327–329 Heritage speakers 32, 260, 261, 266–269, 271–277 I Immigration length 12 Implicit knowledge 109–113, 120, 124, 210 Implicit memory 34, 109, 110, 115, 211, 218 Incomplete learning 32 Inhibition 22, 23 Inner speech 66, 85, 86, 90, 95, 99–101 Input 10, 17, 18, 61, 114, 226, 239, 244, 247, 255, 260, 262, 275, 290, 322 Instructions 117, 121–123, 211, 354 Inter-subject variability 117, 118 Interlanguage 15, 17, 74, 123, 166, 183, 201, 256, 266, 294, 326, 331 Interview task 112, 114 Intra-subject variability 118 Italian 10, 18, 85, 97, 99, 147, 319–321, 328 J Japanese 3, 18, 32, 54, 68, 69, 78, 85, 209, 211, 212, 217, 243, 246–248, 250, 252, 254–256, 324–326, 329, 351 K Korean
32, 208–219
L L2 influence 18, 19, 31, 47, 48, 53–55, 168, 201, 275, 322, 327
Subject index
Language change 17, 68, 109, 134, 139, 144, 199, 231, 259, 283, 285, 313, 322 Language death 259, 293, 317, 357 Language identification 209 Language shift 3–5, 9, 17, 134–136, 160, 161, 284, 293, 317 Language Skills Attrition Project 3, 326 Lexical borrowing 47–50, 190, 193, 195, 200 Lexical retrieval 21–23, 26, 34, 49, 161, 282, 284, 325, 360 Linguistic framing 53 Linguistic knowledge 109, 116, 118 Literacy 11, 20, 55, 275, 324 Loan translation 48 Longitudinal research design 190 LSAP 3, 326 M Marathi 2 Memory 2, 16, 21, 34, 62, 101, 109–111, 115, 120, 122, 155–157, 171, 173, 211, 218, 289, 320 Mental calculation 85, 86, 90, 91, 95, 97, 100 Metalinguistic 11, 49, 52–54, 73, 88, 110–112, 116, 119, 123 Metalinguistic task 110, 111 Migrants 4, 8, 148, 152, 191, 322, 355 Minimalism 19, 327, 350 Minorities 4, 135 Minority language(s) 5, 7, 25, 33, 136, 260, 317, 323, 326 Morphological interference 200, 201 Multicompetence 48 N Number recognition O Object-regulated
211
66, 67
Object-regulation 68, 69 On-line judgment task 114 On-line processing 108, 109, 113, 120, 235, 261 Other-regulated 67, 69 Other-regulation 67, 68, 73 P Parsing 116, 117 Performance 10, 11, 13, 21, 22, 25–27, 32, 34, 61, 89, 101, 106–111, 115, 120, 123, 159, 165, 204, 209, 210, 226, 237, 245, 252, 254, 261, 271–274, 327, 352 Phonetic perception 211 Phonology 20, 52–54, 87, 106, 118, 119, 166, 207, 208, 211, 217, 295, 320, 325, 330 Phonotactics 211, 217 Picture naming 11, 14, 21, 333, 360 Plasticity 207, 208, 219, 327 Plural 146, 199–201, 285, 287, 293 Point(s) of reference 8, 87, 133, 136, 147, 183, 189, 191, 192, 194, 326, 327 Portuguese 68, 85, 321–324, 328, 329 Preferentiality 117 Prescriptive 112, 113, 116, 122, 263, 282 Preterit 262, 277, 278 Pro-drop 226, 227, 229, 239, 240, 303, 313 Production task 116, 148, 154–158, 161 Proficiency 1–3, 5, 7, 13, 29, 84, 86, 89, 91, 92, 99, 101, 102, 113, 123, 136, 138, 139, 141, 142, 148, 151, 159, 189, 190, 197, 204, 225, 231, 238, 247, 260, 261, 267, 276, 277, 319–322, 324, 325, 349, 355, 356, 358–360 Pronominals 226, 227, 230, 237 Psychophysics 119 Puberty 9, 15, 20, 34, 208, 260, 276
Subject index
R Rating scales 122 Reduction 17, 34, 94, 150, 151, 156, 177, 184, 199, 283, 284, 299, 300, 306–309, 312–315, 325 Regression 15, 16, 19, 23, 30, 121, 244, 245, 254–256, 319, 321, 324, 326, 327, 331 Relative clauses 147, 148, 150, 155, 156, 159–161 Repetition 73, 114, 157 Repetition task 114 Response bias 121 Restructuring 21, 47, 50–52, 54, 74, 225, 226, 239, 284, 322, 323 Retraining 211, 216 Russian 32, 48–54, 56, 85, 247, 255, 299–315, 319–323, 328, 361 S Sanskrit 2 Self-assessments 25, 325, 354, 358, 359 Self-regulation 67–71, 73–75, 77 Self-regulatory strategies 69, 71–73 Self-report data 13, 14 Self-reported proficiency 92, 99, 101 Semantic extension 51 Semantic narrowing 51, 52 Semantic transfer 193, 198–200 Sensitive period 20, 23 Sentence generation 11, 22 Shift 3–5, 9, 17, 26, 47, 53, 69, 98, 133–136, 160, 161, 165, 263, 266, 284, 293, 317, 322, 326 Simplification 15–18, 24, 180, 184, 199, 285, 299, 300, 306–309, 312–314, 323, 331 Social networks 30 Sociocultural theory 31, 61, 62, 66, 67, 77 Spanish 2, 18, 32, 49, 53, 74, 78, 81–83, 85, 97, 99, 102, 178, 254, 259–267, 273, 275, 277, 300, 319–325, 328, 329, 361
Speech fragment detection 209 Spontaneous speech 22, 27, 148, 149, 160, 327, 332, 354, 360 Strategic loanwords 195, 197 Subject characteristics 122 Swearing 85, 86, 94, 101 Swedish 85, 209, 320, 329 Syntactic interferences 193, 195, 201, 202 T Tamil 55 Task characteristics 122 Tense 167, 179, 180, 247, 261–263, 265, 266, 268, 270, 271, 275, 276, 300, 323, 325 Timed grammaticality judgment task 108, 120 Transfer 17, 18, 32, 55, 63, 161, 190, 193, 198–200, 226, 237, 239, 244, 247, 250, 255, 256, 260, 266, 324 Translation equivalents 50, 51, 54 Translation task 114, 141, 142 Trauma 13, 134 Truth-value judgments 26, 332 Turkish 13, 18, 32, 133, 139, 142–153, 155, 156, 158–161, 212, 225–227, 229–234, 237, 239, 240, 320, 323, 324, 329, 351, 355, 356 U UG 18, 32, 226, 245, 248, 250, 256, 331 Ungrammaticality 108, 109, 116, 122 Universal grammar 15, 18, 22, 30, 32, 61, 226, 245, 260, 265 Untimed grammaticality judgment tasks 120 Urdu 2 V Van Els taxonomy 9 Verbal fluency 11, 26, 147, 333, 360
Subject index
W Web questionnaire 84, 88
Word finding difficulties 21, 34 Word recognition 209, 217, 219
In the series Studies in Bilingualism (SiBil) the following titles have been published thus far or are scheduled for publication: 1 2 3 4 5
6 7
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
FASE, Willem, Koen JASPAERT and Sjaak KROON (eds.): Maintenance and Loss of Minority Languages. 1992. xii, 403 pp. BOT, Kees de, Ralph B. GINSBERG and Claire KRAMSCH (eds.): Foreign Language Research in Cross-Cultural Perspective. 1991. xii, 275 pp. DÖPKE, Susanne: One Parent – One Language. An interactional approach. 1992. xviii, 213 pp. PAULSTON, Christina Bratt: Linguistic Minorities in Multilingual Settings. Implications for language policies. 1994. xi, 136 pp. KLEIN, Wolfgang and Clive PERDUE: Utterance Structure. Developing grammars again. In cooperation with Mary Carroll, Josée Coenen, José Deulofeu, Thom Huebner, Anne Trévise. 1992. xvi, 354 pp. SCHREUDER, Robert and Bert WELTENS (eds.): The Bilingual Lexicon. 1993. viii, 307 pp. DIETRICH, Rainer, Wolfgang KLEIN and Colette NOYAU: The Acquisition of Temporality in a Second Language. In cooperation with Josée Coenen, Beatriz Dorriots, Korrie van Helvert, Henriette Hendriks, Et-Tayeb Houdaïfa, Clive Perdue, Sören Sjöström, Marie-Thérèse Vasseur, Kaarlo Voionmaa. 1995. xii, 288 pp. DAVIS, Kathryn A.: Language Planning in Multilingual Contexts. Policies, communities, and schools in Luxembourg. 1994. xix, 220 pp. FREED, Barbara F. (ed.): Second Language Acquisition in a Study Abroad Context. 1995. xiv, 345 pp. BAYLEY, Robert and Dennis R. PRESTON (eds.): Second Language Acquisition and Linguistic Variation. 1996. xix, 317 pp. BECKER, Angelika and Mary CARROLL: The Acquisition of Spatial Relations in a Second Language. In co-operation with Jorge Giacobbe, Clive Perdue & Rémi Porquier. 1997. xii, 212 pp. HALMARI, Helena: Government and Codeswitching. Explaining American Finnish. 1997. xvi, 276 pp. HOLLOWAY, Charles E.: Dialect Death. The case of Brule Spanish. 1997. x, 220 pp. YOUNG, Richard and Agnes Weiyun HE (eds.): Talking and Testing. Discourse approaches to the assessment of oral proficiency. 1998. x, 395 pp. PIENEMANN, Manfred: Language Processing and Second Language Development. Processability theory. 1998. xviii, 367 pp. HUEBNER, Thom and Kathryn A. DAVIS (eds.): Sociopolitical Perspectives on Language Policy and Planning in the USA. With the assistance of Joseph Lo Bianco. 1999. xvi, 365 pp. ELLIS, Rod: Learning a Second Language through Interaction. 1999. x, 285 pp. PARADIS, Michel: A Neurolinguistic Theory of Bilingualism. 2004. viii, 293 pp. (+ index). AMARA, Muhammad Hasan: Politics and Sociolinguistic Reflexes. Palestinian border villages. 1999. xx, 261 pp. POULISSE, Nanda: Slips of the Tongue. Speech errors in first and second language production. 1999. xvi, 257 pp. DÖPKE, Susanne (ed.): Cross-Linguistic Structures in Simultaneous Bilingualism. 2001. x, 258 pp. SALABERRY, M. Rafael: The Development of Past Tense Morphology in L2 Spanish. 2001. xii, 211 pp. VERHOEVEN, Ludo and Sven STRÖMQVIST (eds.): Narrative Development in a Multilingual Context. 2001. viii, 431 pp. SCHMID, Monika S.: First Language Attrition, Use and Maintenance. The case of German Jews in anglophone countries. 2002. xiv, 259 pp. + CD-rom. PILLER, Ingrid: Bilingual Couples Talk. The discursive construction of hybridity. 2002. xii, 315 pp. DIMROTH, Christine and Marianne STARREN (eds.): Information Structure and the Dynamics of Language Acquisition. 2003. vi, 361 pp. CALLAHAN, Laura: Spanish/English Codeswitching in a Written Corpus. 2004. vi, 170 pp. + index. SCHMID, Monika S., Barbara KÖPKE, Merel KEIJZER and Lina WEILEMAR (eds.): First Language Attrition. Interdisciplinary perspectives on methodological issues. 2004. viii, 364 pp. + index.