Second Language Attrition in Japanese Contexts
This page intentionally left blank
Second Language Attrition in Japanese Conte ts
EDITED BY LYNNE HANSEN
New York Oxford Oxford University Press 1999
Oxford University Press Oxford New York Athens Auckland Bangkok Bogota Buenos Aires Calcutta Cape Town Chennai Dar es Salaam Delhi Florence Hong Kong Istanbul Karachi Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi Paris Sao Paulo Singapore Taipei Tokyo Toronto Warsaw and associated companies in Berlin Ibadan
Copyright © 1999 by Oxford University Press, Inc. Published by Oxford University Press, Inc. 198 Madison Avenue, New York, New York 10016 Oxford is a registered trademark of Oxford University Press. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of Oxford University Press. Library of Congress Cataloging-in-Publication Data Second language attrition in Japanese contexts / edited by Lynne Hansen p. cm. Includes bibliographical references and index. ISBN 0-19-512304-2 1. Language attrition—Japan. 2. Linguistic change. I. Hansen, Lynne, 1940- . P40.5.L28S43 1999 99-22791 495.6'8—dc21
1 3 5 7 9 8 6 4 2 Printed in the United States of America on acid-free paper
For Jadyn, Gabe, Rachael, Tiana, and Tora-chan with boundless affection
This page intentionally left blank
Foreword
Research topics seem to have their own life-cycle: It takes quite some effort (or the support of high-status colleagues from neighboring disciplines) to get a topic on the research agenda. Then the people working on the topic have to show that what they do is relevant, theoretically interesting, and sufficiently embedded in established paradigms to get their grant proposals through. It is in those early stages that the theorizing is generally rather wild and ambitious, since few data are available to restrict the hypotheses' space. In the next phase results from empirical research begin to have their impact, and early truths are replaced by new doubts, redefining what the major issues are. If a topic gets through this phase, it is likely to become part of the larger field of research. Language attrition research as a subdiscipline of the field of applied linguis tics has already gone through the early stages, and there is consensus about what the real issues are at the moment. Three issues and their interaction now define the field: the what, the how, and the why of language attrition. Although the questions are simple, answering them is far from easy, given the specific problems encountered in attrition research and the fact that almost every study explores unknown territory with all the methodological uncertainty that entails. At the same time this "newness" makes the field of language attrition a fascinating field, in which a shared interest in learning is more prevalent than paradigmatic clashes. Indeed, in contrast to other subfields of the language sciences, such as language acquisition, there still is space for everyone who wants to work on it, and there is a group of nonorthodox devotees willing to share their expertise with colleagues from other parts of the world. A strong part of this group is represented in this volume. The articles in this volume present a unique set in the sense that they all have one language in com-
viii
Foreword
mon. Japanese is studied in various situations of attrition and relearning in different circumstances of language contacts, and it is studied with learners and losers of different ages. The linguistic focus of most of the contributions leads to rich descriptions of language use and detailed analyses of specific constructions, which allow for comparisons with other processes of language change, such as first language acquisition and language pathology. One of the main conclusions is that language attrition is not an endstage of a process, but a normal part of changes in proficiency over time, both in the first language and in the second language. The studies also show that language attrition is gradually becoming a part of the field of second language acquisition (SLA). There are remarkable parallels with respect to the topics studies: the roles of cross-linguistic influence, age, individual differences, language setting, social-psychological factors such as attitudes and motivation in language attrition. There is also a fairly recent interest in the psycholinguistics of language attrition, which links the topic with the long-established one of memory research in psychology. One may wonder why there is so much interest in language attrition in Japan. Is language so much a part of the Japanese soul that losing it is regarded as something extremely dramatic? Or is it the contrast between Japanese and languages like English that attracts researchers to study it? Maybe the reasons are more mundane, like more generous allowances for sabbaticals at Japanese universities. For the present volume the explanation is probably a very simple one: Lynne Hansen has been working on language attrition with Japanese for a long time, and she has no doubt been a source of inspiration to other researchers in this field. The development of language attrition as a serious research topic so far has been marked by a number of edited volumes. There is little doubt that this volume marks another step in its evolution, and, it should be added, a significant step. Kees de Bot University of Nijmegen
Contents
Contributors
ix
1. Investigating Second Language Attrition: An Introduction 3 Lynne Hansen and Anita Reetz-Kurashige Part I: The Loss of English as a Second Language by Japanese Children 2. Japanese Returnees' Retention of English-Speaking Skills: Changes in Verb Usage over Time 21 Anita Reetz-Kurashige 3. The First Stage of Second Language Attrition: A Case Study of a Japanese Returnee 59 Machiko Tomiyama 4. On the Loss of English as a Second Language by Japanese Returnee Children 80 Asako Yoshitomi Part II: The Loss of Japanese as a Second Language by Adult Foreigners 5. Lexical Maintenance and Attrition in Japanese as a Second Language 114 Robert A. Russell
x
Contents 6. Not a Total Loss: The Attrition of Japanese Negation over Three Decades 142 Lynne Hansen 7. Testing the Regression Hypothesis: The Remains of the Japanese Negation System in Micronesia 154 Brenda Hayashi 8. Learning and Losing Japanese as a Second Language: A Multiple Case Study of American University Students 169 Sumiko Nagasawa
Referenc Index
Contributors
Lynne Brigha Brenda Hayashi Miyagi Gakuin Women's College Sumiko Nagasawa Kurume University Anita Reetz-Kurashige UCLA Extension, American Language Center Robert A. Russell Brigham Young University Machiko Tomiyama International Christian University Asako Yoshitomi Tokyo University of Foreign Studies
XI
This page intentionally left blank
Second Language Attrition in Japanese Contexts
This page intentionally left blank
1 Investigating Second Language Attrition: An Introduction Lynne Hansen and Anita Reetz-Kurashige
Language loss affects all of us. It occurs in every corner of the world, taking its toll among young and old alike. Only recently, however, has the forgetting of language attracted much scholarly attention. During the past three decades a new area of study examining language regression has grown out of the converging efforts of researchers from several fields. The facts they deal with are relevant to the various disciplines because they relate not only to what is being lost (the focus of linguists) but also to how it happens (the focus of psycholinguists and neurolinguists) and why it happens (the focus of sociolinguists, sociologists, and anthropologists). In these several approaches to the study of language loss, a number of terms are used to refer to the same or related themes: language attrition (Seliger & Vago 1991; Waas 1996; de Bot & Weltens 1995) and language regression (Hyltenstam & Viberg 1993); language loss (Fase et al. 1992; Pauwels 1986), and language shift (Fishman 1964; Gal 1979); codeswitching and code mixing (Heller 1988; Eastman 1992; Myers-Scotton 1993, 1997); and language death (Dorian 1981; Schmidt 1985; Grenoble & Whaley 1996) and language obsolescence (Dorian 1989). These terms have proved useful in referring to the gradual waning or disappearance of language in individuals or groups, and each represents a different emphasis. The primary concern of this volume is second language (L2) attrition, the disintegration or loss of the structure of a language learned after the mother tongue (L1). A person who experiences such loss, a language attriter, is by definition bilingual.1 The disappearing second language does not simply vanish, leaving a vacuum; rather, the communicative functions it previously filled are taken over by a different language, most often the attriter's L1. As a language is forgotten, it is replaced by another. In each of the studies that follow, Japanese is either the attriting language or the replacing language. Part I reports studies of Japanese 3
4
Second Language Attrition in Japanese Contexts
returnee children who had learned English while living abroad and then returned to their homeland. Part II looks at several groups of adults who had learned Japanese as a second language in a natural milieu in Japan or at school. Language loss is a natural outcome in language contact situations, given the dynamism and variability of linguistic phenomena. Any living language is constantly changing, either progressing or regressing. Our linguistic knowledge is not static. As a language is learned or lost, it is merely adapting to the ever changing constraints of the speaker's mind and to the conditions of the social milieu. Whether in individuals or in speech communities, whether in acquisition or in attrition, language change involves "the same functional and formal parameters of linguistic structure, and is embedded in the same matrix of sociocultural and neuropsychological determining factors" (Hyltenstam & Viberg 1993:25). This outlook is elucidated most fully in Hyltenstam & Viberg 1993; also in Andersen 1982, 1983; Jakobson 1941; Rickford 1991; Slobin 1977. The study of language attrition, then, brings to light one aspect of this pervasive linguistic dynamism and variability. Others are found in such areas of research as dialectology, sociolinguistics, historical linguistics, pidgin and Creole languages, indigenized World Englishes, languages in contact in bilingual communities, bilingualism in individuals, and language acquisition. Language attrition constitutes part of the larger picture of linguistic dynamism as it adds essential information to our understanding of how the various manifestations of language change relate to each other in their sociocultural as well as their neuropsychological dimensions. Within a sociocultural perspective, language attrition research contributes to a comprehensive account of language progression and regression in different societal situations. Investigation of the environmental conditions that foster language loss in the individual is needed to complement a growing number of societal studies that explore the cultural determinants of language shift (Dorian 1981; Gal 1979; Grenoble & Whaley 1998; Hill & Hill 1986; Moore 1988; Schmidt 1985; Kulick 1992; Woolard 1989). Within a neuropsychological perspective, the constraints of the human brain in processing linguistic dynamism are seen in the universal developmental sequences found in both L1 and L2 acquisition. In language loss, the regression hypothesis, the idea that linguistic distinctions will be lost in an inverse order to which they are learned, provides a theoretical framework for examining these neuropsychological constraints as revealed in language attrition (for a discussion of regression theory in language attrition research, see "Processes in Attrition" in this chapter). Among the areas of study that examine linguistic dynamism, that of L2 attrition is the most recent to develop. The earliest groundwork for the new field was laid by a conference on language loss held at the University of Pennsylvania (Lambert & Freed 1982). A second conference in 1986 in the Netherlands presented state of the art inquiries on British, Dutch, Israeli, Swedish, and North American research (Weltens, de Bot & van Els 1986). In the intervening period, investigations of L2 loss have steadily accumulated. The L2 attrition research published in English has reported on a wide spectrum
Investigating Second Language Attrition
5
of languages, including English loss by Hebrew L1 speakers (Berman & Olshtain 1983; Olshtain 1986; 1989); French loss by Dutch native speakers (Weltens, van Els & Shils 1989; Weltens 1989) and by English-speaking Canadians (Gardner, Lalonde, Moorcroft & Evers 1987); German loss by Turkish native speakers (Kuhberg 1992), and by English native speakers (Godsall-Meyers 1981); L2 Hindi-Urdu by English speakers (Hansen 1980; 1983); L2 Portuguese by English/Hebrew bilinguals (Cohen 1986), and L2 Spanish by English speakers (Bahrick 1984a, b). These studies tested theoretical constructs to further our understanding of the universal dynamics of language loss. This volume continues in the same vein of hypothesis testing while documenting features of loss that have significance for linguistic comparison. We have assembled research on the loss of L2 English by bilingual Japanese children and the loss of L2 Japanese by English and Ponapean speaking adults in order to test fundamental hypotheses and to identify the particular influences of the Japanese language in the dynamic circumstances of language shift in individuals. While the Japanese language is not a main independent variable in these studies, our research documents the processes of language change in subjects whose L1 or L2 is Japanese. Thus we offer a body of evidence on language loss that has theoretical and comparative value. Through these studies, we see better the patterns of linguistic interface between Japanese and English in the backand-forth flow between acquisition and loss. We document how three sets of variables influence second language maintenance: personal characteristics, language acquisition history, and attrition circumstances. Applications of research findings in these three areas will be considered at the conclusion of this chapter. First, however, in the following section, we wish to highlight some of the initial findings and fundamental issues in the second language attrition field, with a view toward setting a background to the various topics addressed in the volume. Overviews of the papers then follow. Finally, the reader is invited to contemplate potentials of the study of second language loss as we outline future applications.
Findings and Issues in Second Language Attrition Variables Affecting Attrition In this section we will introduce three variables that have been found to relate to second language attrition: age, proficiency level, and attitude/motivation. The Age Advantage A sharp distinction between child and adult language attrition is noted in numerous anecdotal accounts in the literature. For example, in the dissertation study of one of the authors, L2 Hindi-Urdu data were collected longitudinally from two pre-school American children during three sojourns of the family on the Indian subcontinent (Hansen 1980). In India and Pakistan the children came frequently to be judged as native speakers of their second language; their mother was far
6
Second Language Attrition in Japanese Contexts
less proficient. Back in the United States, between L2 exposure periods, both children each time suffered what seemed to be a complete loss of their HindiUrdu within a few months; the mother noticed no decline in her second language skills. Today, over twenty years after the family's final departure from South Asia, the children, now in their mid-twenties, comprehend not a word from recordings of their own animated conversations in Hindi-Urdu; the mother still understands much of them. The importance of age as a variable in child language attrition has been substantiated in several longitudinal case studies (Cohen 1989; Hansen 1980; Hansen-Strain 1990; Kaufman & Aronoff 1991; Koike 1990; Yukawa 1996), leading to the conclusion that the younger a child, the more rapid the pace of language attrition. Cohen (1989), Hansen (1980), Hansen-Strain (1990), and Yukawa (1996), looking at language loss by siblings together in the same linguistic environment, losing L2 Portuguese, L2 Hindi-Urdu, L2 Japanese, and L1 Japanese, respectively, found that the younger siblings lost their language skills more quickly than the older. A single exception to this in the case study literature is Kuhberg's (1992) report of the L2 German attrition of two Turkish children, ages nine and seven. The more rapid attrition of the nine-year-old is explained by the researcher as due to the "stronger pressure for the older child to give absolute priority to Turkish" (Kuhberg 1992:145). Group studies of child language attrition are also beginning to accumulate. In a study of Israeli children who had learned English abroad, Olshtain (1986, 1989) found that her younger subjects, ages five to seven, suffered more loss of their second language than did the older children, ages eight to fourteen. In another cross-sectional study of L2 English loss, Yoshida & Arai (1990) measured the listening comprehension of Japanese returnee children. In tasks that included repetition of increasingly complex sentences, the researchers found that the younger groups made significantly more errors. Although rapid attrition of unused language skills is the rule for young children, retention is a common outcome for adults. For those who attain relatively high levels of proficiency, an initial plateau of several years is reported, during which skills are relatively unaffected (Edwards 1977; Hansen 1996; Schumann, van Os & Weltens 1985; Snow, Padilla & Campbell 1984). Such is the case for Dutch students who learned French at school, according to a series of research reports from the Netherlands (Grendel 1993; Weltens 1989; Weltens & Grendel 1993; Weltens, van Els & Schils 1989). Weltens (1989), for example, reports that general receptive L2 proficiency was not susceptible to loss after four years of disuse by former students of French. When students attain lower levels of proficiency, however, a different outcome has been reported. In an American context, for example, Bahrick's (1984a) large-scale study of 773 former high school and college students of Spanish found significant L2 attrition during the first five or six years following training. The loss leveled off then, and knowledge tended to remain stable over several decades. Explanations for differences between groups of adult attriters have been sought in a number of variables, a key one being the L2 proficiency level before the onset of attrition.
Investigating Second Language Attrition
1
Proficiency Level: Inverse Hypothesis and Critical Threshold Hypothesis In a review of the L2 attrition literature, Vechter, Lapkin, & Argue (1990) state that "the higher the degree of attainment, the lower the degree of attrition." This inverse hypothesis postulates that there is an inverse relationship between proficiency level prior to the onset of attrition and the rate and/or amount of loss. High levels of proficiency predict better survival of L2 German skills over time according to Godsall-Myers (1981). Especially receptive skills like vocabulary recognition and reading comprehension show high correlation between advanced L2 learning and L2 maintenance (Bahrick 1984a; Olshtain 1989; Weltens 1989). For mature second language speakers, as for mother tongue speakers, there seems to be an initial period when little attrition is observed despite language disuse. In fact, in cases of both children and adults, "residual learning" or apparent improvement of second language skills during periods of diminished use has been documented (Cohen 1975; Hansen-Strain 1992; Scherer 1957; Weltens et al. 1989). In his study of L2 Spanish attrition, Bahrick (1984a) concludes that the amount of content to be lost during the first five years following training "becomes a progressively smaller portion of total knowledge with higher levels of training" (116). He further reports that the bulk of the loss of Spanish knowledge (reading, vocabulary, idioms, grammar, and word order) occurs within a few years after training, but "the remainder is immune to further losses for at least a quarter of a century" (111). Bahrick coined the term permastore to refer to L2 content that is accessible and recognizable for the long term (up to 50 years!). In response to Bahrick's (1984a) findings, Neisser (1984) offers a different interpretation from a cognitive psychological perspective. He explains that Bahrick's subjects, instead of acquiring a set of isolated items, have discovered a structured system of relationships for Spanish, also called a schema, a structured mental representation of complex information. For example, the recall of a Spanish word is not a simple reproduction of a previously memorized item, but, rather, it is like a problem-solving response that involves deeper processes of memory retrieval. Thus, information tied into this extensive and redundant cognitive structure is resistant to forgetting, while less well connected pieces of information are vulnerable to loss. Neisser postulates that "some response strengths reach a critical threshold during learning; beyond that threshold they become immune to interference or decay" (33). His argument is based on Ausubel's subsumption theory, a process of relating new material to an existing cognitive field, thus making learning more meaningful and more resistant to loss (Brown 1987:65-69). Language attrition research supports Neisser's idea of a critical threshold, a point at which the learner has a stable enough mental representation of some linguistic elements and/or structures that they, at least temporarily, are resilient to loss. For example, as mentioned above, several studies of the L2 French learned in Dutch schools found little L2 attrition of recognition skills among the subjects, all very competent in the L2 (Grendel 1993; Weltens 1989; Weltens &
8
Second Language Attrition in Japanese Contexts
Grendel 1993; Weltens, van Els & Schils 1989). Clark and Jorden (1984:58) recommend large scale studies to determine more precisely the "critical point in overall language proficiency below which attrition is rapid and extensive, but above which a large proportion of the initially acquired material is retained." Level 3 on the Foreign Service Institute (FSI) scale, according to FSI language instructors' informal observation, may be that point, the researchers add. Attitudes and Motivation At the first conference on language loss in 1980, Gardner proposed that social factors be included as predictor variables in attrition research. Citing Edwards's (1976, 1977) work on second language retention among Canadian government workers in Ottawa, the researcher argued that attitudinal/motivational factors are related to maintenance, in that they influence the degree to which individuals seek out opportunities to use a language (Gardner 1982). In Canadian bilingualism research, social factors, particularly attitudinal/ motivational variables, have been integral parts of the study of language progression and regression. Motivation was claimed to have two aspects: integrative and instrumental. A language learner with integrative motivation learns another language because of "a sincere and personal interest in the people and culture represented by the other language group," whereas a learner with instrumental motivation is more interested in "the practical value and advantage of learning a new language" (Lambert 1974:98). Growing out of this work (e. g., Gardner & Lambert 1959), numerous studies conducted in many parts of the world indicate varying degrees of relationship between L2 learning and attitudes and motivation. The effects of attitudes, motivation, and language use on L2 attrition have been studied by Gardner, Lalonde & MacPherson (1985) in Canadian students who had taken intensive French courses in Quebec. Less favorable attitudes and lower language use among the subjects correlated with the attrition of speaking and understanding as reported in "can-do" scales, but not to reading. The researchers suggest that attitudinal/motivational factors may exert their greatest influence on attrition primarily during the acquisition period by influencing the level of competence acquired. A subsequent study (Gardner, Lalonde, Moorcroft & Evers 1987) examined the role of motivation and use in L2 attrition in high school students of French over the summer vacation. While the best predictor of retention was found to be the level of proficiency attained, the researchers conclude that attitude and motivation are important indirect influences. 2 Processes in Attrition We now turn to accounts of several processes that have been examined in second language attrition research: the regression hypothesis, influence from the replacing language, and retrieval failure.
Investigating Second Language Attrition
9
The Regression Hypothesis The most widely used theory for describing the nature of the attrition process is the regression hypothesis. Dating back to Jakobson (1941), the hypothesis describes the path of language loss as the mirror opposite of acquisition. This proposition holds true at the interlinguistic skills level: receptive skills precede productive skills in acquisition and the reverse is true in attrition. The progression of both receptive skills (from listening comprehension to reading comprehension) and productive skills (from speaking to writing) is largely mirrored in attrition. Bahrick (1984a), Weltens (1987), and Yoshida et al. (1989), among others, have found support for the regression hypothesis at the level of interlinguistic skills. At the intralinguistic level (within morphology, syntax, the lexicon), documenting that the stages of development are reversed in attrition is more difficult. Tracking both acquisition and attrition is time consuming, and a universal or predictable development ladder has been established for only a limited number of linguistic structures. However, Berman & Olshtain (1983), Cohen (1975), and Olshtain (1986) have demonstrated through testing that the regression hypothesis holds. In a longitudinal study of the acquisition and attrition of negation in Hindi-Urdu by two American children, Hansen (1980) concludes that "the forgetting data from both children could be interpreted as a recapitulation in reverse of the acquisitional sequence." Kuhberg's (1992) longitudinal L2 acquisition and L2 attrition study of three Turkish children's German found that "attrition was largely a mirror image of acquisition": First-learned, basic syntactic patterns were retained longest. Yoshitomi (1992:295) concludes that "the generalizability of reverse order [the regression hypothesis] at the intra-skills level is limited, however, because the hypothesis has been tested only on a limited number of specific syntactic structures." Influence from the Replacing Language In both L2 acquisition and attrition, L1 transfer can be viewed as a resource actively used to restructure a linguistic system in transition. The extensive literature on L1 influence in L2 learning spans several decades (Gass & Selinker 1992; Odlin 1989), and suggests a number of constraints on the process: linguistic markedness (Eckman 1977); psycholinguistic markedness (Kellerman 1983); frequency (Anderson 1982); transfer more likely to occur if L1 and L2 are perceived as being close (like Japanese and Korean) than if they are perceived as being distant (like Japanese and English) (Kellerman 1983; Ringbom 1987); and similarity between L1 and L2 structures necessary for transfer to occur (Cowan 1986; Wode 1978; Zobl 1980). In the L2 attrition field the interference theory of memory loss holds that forgetting is actually interference between the replacing language and the attriting language (Weltens & Grendel 1993:137), but the transfer that occurs in the process has just begun to engage the attention of researchers. In a pioneering study of features of first language (Hebrew) transfer in second language (English)
10
Second Language Attrition in Japanese Contexts
attrition, Berman and Olshtain (1983) found that structural properties in which the attriting language differs quite markedly from the dominant language showed the greatest attrition. Additional work on the attrition of English as a second language by Hebrew speakers led Olshtain (1989) to suggest that typological universals affect interference from the dominant language in the use of the more restricted one. Further L2 attrition studies can be expected to shed new light on the constraints on linguistic dynamism which und lie the processes of language learning and loss. Retrieval Failure Theory The retrieval failure theory of language loss follows an information pro cessing approach whereby forgotten information is not gone, but has become inaccessible. Forgetting is said to be "much like being unable to find something that we have misplaced somewhere. Forgetting occurs because the information we seek is temporarily inaccessible; if only we had the right retrieval cue, the information we seek could be successfully retrieved" (Loftus & Loftus 1976). A decrease in accessibility to one's knowledge of a language is evidenced by strategies such as "progressive retrieval," a strategy in which a learner, in search of a certain target item, starts with an inappropriate form but eventually arrives at the correct form (Sharwood-Smith 1983; Cohen 1986, 1989; Olshtain 1989), circumlocution (Olshtain & Barzilay 1991; Turian & Altenberg 1991) and other hesitation and repair behaviors (Russell 1996; Hansen, Gardner & Pollard 1998, Kenny 1996). In the retrieval failure view, a critical element in measuring language attrition is, of course, processing time. An early symptom of attrition is an increase in the time needed to retrieve linguistic items (Olshtain 1994). Thus, while spontaneous speech is probably the most relevant data for research on language acquisition, in language attrition research it needs to be supplemented with well-controlled, timed experiments that force subjects to apply rules they are uncertain about (de Bot & Weltens 1995).
Overview of This Volume Part I of this collection examines language attrition in children, Japanese nationals who returned home following residence abroad. As an introduction to this section, Reetz-Kurashige defines the population and elaborates on the conditions of drastic reduction of English input back in Japan. She then reports on her own study which tracks changes in the L2 verb usage of 18 returnees over a year's time. Tomiyama's contribution next summarizes her 19-month case study of one returnee, with comparative data from two additional children. Rather than a focus on a single linguistic skill, the broader scope of this investigation allows for the examination of the differential attrition of lexicon, morphology, and syntax, and of productive skills versus receptive skills. Finally, Yoshitomi reports on her longitudinal case study of four girls over a year's time. Again, a broad
Investigating Second Language Attrition
11
focus in the research questions leads to comparative information on the attrition of several linguistic subskills, and, more important in her view, on the ability to coordinate them. The subjects of the three studies were all elementary school children who had spent from one to seven years in an English-speaking country and had been back in Japan for as long as a year and a half at study onset. Although the English proficiency levels of the subjects in the three studies cannot be compared using the same standardized measure, each of the researchers relates her attrition findings to subjects' age, length of residence abroad, L2 proficiency at study onset, and period of L2 disuse. Together the investigations report the accumulated effects of attrition of such subskills as fluency, lexical retrieval in speaking, accuracy in phonology, morphology, and syntax, in addition to assessing global attrition in speaking skills. With regard to the age variable, the quantitative and qualitative evidence from the three studies confirms the proposition that, for children, language retention increases with age. There is also agreement that the amount of loss is somewhat proportional to the subjects' level of L2 proficiency. The more these children knew, the more they were likely to retain. A further shared finding is that as length of disuse increases, the linguistic characteristics of loss change. Tomiyama specifies in her chapter that initial reduction in fluency and lexical retrieval precede morphosyntactic loss. In her observations of a highly proficient eight-year-old, she was unable to find any notable change in performance in phonology, morphology, receptive lexicon, and syntax over a 19-month period. The significance of her report lies not only in the loss that was observed but also in the loss that was not observed. Subsequent reports of the child's attrition over an additional two-and-a-half-year period (written and reported while this manuscript was being reviewed) show that the subsequent period of retention is followed by a stage of stability in productive lexicon and attrition in syntax. Evidence of morphological attrition starts to appear at this stage showing fluctuation in performance rather than a steady decline (Tomiyama 1995). Finally, in the third stage, we see the deterioration in his overall command of English, as seen in discourse skills. Phonology and receptive lexicon remain stable throughout the latter stages (Tomiyama 1996). By including 18 subjects in her study, Reetz-Kurashige was able to document a range of individual differences in patterns of retention/attrition of English verbs. Age and length of residence abroad emerged as interdependent factors. Children older than eight years who lived at least two years abroad retained 80 percent or better accuracy of verb forms. The most striking finding in the study was evidence of a threshold in attrition, a level of proficiency in speaking that mitigated against morphosyntactic loss. The study found that accuracy and clarity (as measured by targetlike use) in storytelling at the beginning of the study were strong predictive factors for retaining target like use of verb forms. On a comparative linguistic level Reetz-Kurashige's work supports Olshtain's (1986, 1989; Berman & Olshtain 1983) claim that typological and pragmatic similarities between the L1 and L2 lead to positive reinforcement for L2 retention,
12
Second Language Attrition in Japanese Contexts
whereas L1 and L2 dissimilarities lead to interference and attrition. A distinctive difference between Israeli and Japanese children is reported: the widespread use by the Asians of the English verb stem rather than analyzed forms. This appears to reflect a typological feature of Japanese grammar and may be attributable to the frequent use of the base form, shuushikei, in Japanese. After documentation of the accumulative effects of the English attrition of four subjects, Yoshitomi draws an original point from her analyses: "Overall, the measure of general accuracy (frequency of error-free clauses) exhibited attrition much more clearly than any other syntactic or lexical measurement." Thus she concludes that language attrition first appears in a reduction in the overall ability to coordinate linguistic subskills spontaneously and communicatively in real time. Erosion in specific linguistic subskills becomes apparent only subsequent to this loss in coordinative competence. Thus, Yoshitomi argues that maintenance programs for returnees should include opportunities for spontaneous interaction in order to utilize this competence. Part II of this volume examines the second language attrition of Japanese in adults. In the first two contributions the focus is on the L2 loss of Americans in the United States who had learned Japanese while working in Japan (Russell, Hansen). The concluding chapters look at the loss of L2 Japanese that had been learned in two very different educational settings: one in primary schools in Micronesia (Hayashi); the other in various classroom settings in the United States, as well as in Japan in combination with L2 exposure outside of class (Nagasawa). Russell's study examines the Japanese lexicons of L1 English speakers who had learned Japanese as missionaries in Japan. Production data, collected three times over a two-year period, is examined separately for two subgroups differentiated by the amount of formal L2 instruction they had received after leaving Japan. Significant differences are not found between the instructed and uninstructed groups. Using 26 dependent variables, the researcher finds only subtle evidence for loss over the first five years of reduced L2 Japanese use, lending support to claims for an initial plateau in the L2s of highly proficient attriters. The results are also consistent with the retrieval failure view: lexical skills are not so much lost as they are simply more difficult to access in real time. Russell's findings confirm those of previous studies that have found reports of decreasing vocabulary size under conditions of disuse but which, at the same time, have failed to find evidence for significant changes in lexical variability or density. Hansen examines the loss of negation in the same population of American adults as Russell, after 25-35 years of Japanese disuse. Acquisition data collected from learners in Japan had established the acquisition sequence for negation and had demonstrated that the negated structures are learned by this population during their sojourn in Japan. The attrition data support the regression hypothesis. The results are explained in a linguistic framework, including a discussion of structural salience and a consideration of the naturally occurring high frequency of verbal negation in Japanese. Significantly greater retention for
Investigating Second Language Attrition
13
attriters who had spent more time in Japan suggests that a critical threshold for second language maintenance may be crossed by learners who spend three years in intensive L2 use. Together, these two chapters provide a partial attrition profile of the returned missionary population. Little loss is found during the first few years of L2 disuse. There is considerable loss after several decades, however, with sizable individual differences in retention, particularly among those who had experienced shorter L2 exposure periods (the women). In the next chapter, Hayashi, documents the remains of L2 Japanese in elderly Micronesians who learned the language as school children during the Japanese occupation of their islands. With the analytical focus again on negation, the interview and elicitation data that are reported suggest the same constraints in the acquisition/attrition of Japanese negated structures as have been reported in previous studies of the progression (Kanagy 1991) and regression (Hansen & Iwata 1992; see also Chapter 6, this volume) of Japanese L2 negation. Hayashi also comments on the robustness of formulaic expressions and pragmatic markers of negation (e.g., dame, iya, chigau) in the L2 Japanese of her subjects. Nagasawa's subjects were seven American graduate students who had spent the summer in Japan. She followed them during the subsequent year back at the university in circumstances of greatly decreased L2 input, relating their retention/attrition of Japanese to proficiency level, attitudes and motivation, and language learning background. The study differs from the previous ones in that, rather than a discontinuation of second language input, all of these students continued regular L2 exposure (albeit very much reduced) over the observation period, including two 80-minute Japanese courses each week. Thus, in Nagasawa's careful longitudinal analysis, more than in those of the previous researchers, we see the ebb and flow of linguistic dynamism. Oral proficiency was tracked by Nagasawa through Oral Proficiency Interview (OPI) data over a seven-month period. In addition to ratings using the American Council on the Teaching of Foreign Languages (ACTFL) scale, the interviews yielded data for the analysis of Japanese particle use, sentence structure, and fluency. The major findings were (1) an advanced-high rating (that is, 2+) or higher on the ACTFL scale appeared to be a threshold level for resisting L2 attrition; (2) the four subjects who maintained their high motivation for learning Japanese also maintained their proficiency to a greater extent than the three who did not, with retentive power in either instrumental or integrative motivation; (3) the four subjects in the maintenance group had instruction prior to (or simultaneously with) their language contact, whereas the three attriters had experienced extensive contact in Japan (19-36 months) before their formal instruction in Japanese. Although Nagasawa takes a cautious approach in interpreting and generalizing her findings, the results of her study do point to a hypothesis for future studies: in the initial stages of second language learning, classroom instruction better leads to subsequent retention of the second language than does immersion without instruction.
14
Second Language Attrition in Japanese Contexts
The Language Attrition Puzzle: Research and Applications Amassing and Assembling the Pieces Some evidence from empirical research on language attrition may appear contradictory to findings from other research. For example, the relationship of proficiency level to attrition reported by Reetz-Kurashige (see Chapter 2, this volume), seems to run counter to Weltens's (1989) findings, which indicated that attrition is independent of training level (de Bot & Weltens 1995:153), and to Bahrick's conclusion that "the total amount of content to be forgotten during the five years following training is relatively constant for individuals at different levels of training" (Bahrick 1984a:116). Why such apparent contradictions? The answer lies in the study designs and instruments. Measurement of attrition in productive skills consistently shows more attrition than measurement of recognition skills. Subjects' age and level of L2 competency at study onset also influence attrition, so that, in the above example, the younger and less competent subjects of Reetz-Kurashige's study retained less than either Welten's highly competent or Bahrick's older subjects. Repeated measures of the same subjects permits tracking individuals' loss patterns, which comparisons between groups does not allow. Tomiyama's (1996; see also Chapter 3, this volume) evidence of decreased fluency and increased code switching in her subjects revealed aspects of loss that comparisons between groups cannot show. Similarly, tracking productive use of language can provide evidence like Reetz-Kurashige (Chapter 2, this volume) and Yoshitomi's (Chapter 3, this volume) data on incremental downward shifts toward use of simpler syntax; or it can provide evidence like Nagasawa (Chapter 8, this volume) and Tomiyama's (1996) data on the fluctuating nature of language change in highly proficient L2 attriters. Thus, when we look closely at task, age, levels of language competency, amount of time over which attrition is measured, and the design of time series assessments (repeated measure or comparisons between groups), we find that the separate study results are like pieces of a puzzle that we are still putting together. We hope that the chapters in this volume will be viewed as a modest contribution toward sorting and assembling some of the pieces. As we fit findings from individual studies into the larger picture of language attrition, a consideration of task—for example, the language abilities assessed —is particularly important for distinguishing common, complementary, and contrary results. The initial important distinction in task is between the testing of productive versus recognition/receptive skills. The common finding reported by Weltens's literature review (1987:23) and confirmed in this volume (see Chapters 3 and 8) is that productive skills are vulnerable to long-term loss, whereas recognition skills are much less so. Within the separate areas of recognition and productive skills, we must further specify the type of task as the following examples show. Single-word recall is easier when subjects are prompted in L2 to recall L1 than when prompted in L1 to recall L2 vocabulary items (Bahrick 1984a; Kroll et al. 1993); story retelling after hearing a tape compared with storytelling
Investigating Second Language Attrition
15
from pictures demonstrates that the latter is more prone to error (Chapter 2, this volume); in assessing conversational skills, subjects' recall of common idioms is relatively immune to attrition, whereas recalling grammatical structures typologically different from their L1 is candidate for loss (Olshtain 1989). In order to bring unambiguous evidence to bear on hypotheses in the language attrition field, we need reliable measurements of linguistic ability. The more precisely we define and measure attrition, the more congruent our findings appear and the better the puzzle pieces fit together. Inasmuch as no tried-andtrue test battery for measuring language proficiency is yet available, a variety of instruments is needed to investigate declining L2 proficiency from many different perspectives. In addition to syntactic and lexical skills which have most often been studied, future research should develop reliable instruments for the measurement of phonological, pragmatic, and discourse skills. This broadened focus of attrition research, coupled with the extension to larger samples of the approaches exemplified in this volume, can give us a more complete picture of the dynamism that underlies language loss. Applications of Research on Second Language Attrition
In considering the applications of second language attrition research, we again point out the three sets of variables involved in L2 loss: characteristics of the acquisition process, the attrition period, and the attriter. As our understanding of them and of their interaction in the language attrition process is expanded through research, applications can be made in the areas of L2 instruction, language maintenance programs, and strategy training for language retention. L2 Instruction In relating language teaching to subsequent maintenance of the language, one of the most fundamental questions is whether particular teaching methods are superior to others in terms of the long-term retention of the acquired skills. So far, a very tentative conclusion is that explicit teaching engenders more durable skills than immersion or direct methods. One source of evidence for this, cited by Weitens & Cohen (1989:128) with respect to young children is their "rapid and apparently total loss of L2 competence acquired under natural circumstances in the L2 environment once the learners return to the L1 environment again" (Burling 1978; Hansen 1980). Since these children were pre-school age, however, maturity is recognized as a confounding variable. It remains to be seen the extent to which the reported age differences in attrition accrue from the fact that the older subjects learned explicit grammar, whereas the children did not. Nagasawa's (Chapter 8, this volume) interesting finding is that adults who initially had classroom instruction experienced less attrition than adults who had no explicit instruction in the initial stages of exposure to Japanese. In order to make claims about the superiority of one exposure type or teaching method over others, however, we need to test empirically the impact of various learning histories on the language maintenance of larger groups of attriters.
16
Second Language Attrition in Japanese Contexts
The emphasis given to particular language skills in the L2 classroom is another variable to include in studies of the effects of learning conditions on subsequent language attrition. The important dichotomy here is between productive and receptive skills. Since the latter appear more durable, we should look at the effects on retention of an emphasis in instruction on listening and reading as opposed to speaking and writing. Given the differences found so far in the development and maintenance of comprehension and production abilities, we may find differences in retention between students whose instruction focused mainly on oral productive skills and those whose instruction stressed primarily reading and listening. The extent to which literacy anchors linguistic knowledge is an area just beginning to be addressed (Hansen & Newbold 1998; Hansen & Chantrill 1998). In a recent study of the attrition of Japanese as a second language by American adults, Hansen & Newbold (1998) found that the level of Japanese writing that had been attained by the subjects was a strong predictor of their subsequent retention of the language. While several earlier reports of child language attrition also show a positive relationship between language retention and the attainment of reading and writing skills (Hansen-Strain 1990; Ohlstain 1989; see also Chapter 2, this volume), the designs of these studies result in a confounding of the literacy variable with both age and the original level of L2 proficiency. The effect on language retention of the intensity of the training program is still another area remaining to be examined in attrition research. The literature on language acquisition, however, indicates that students enrolled in more intensive programs make more progress than those in non-intensive programs with the same amount of time in the classroom (Edwards 1976; Stern 1976). Thus, it may be that language acquisition in intensive and nonintensive programs do involve different processes which result in differences in subsequent loss of the language. It would be useful for educators to know whether the language learned in a particular number of hours in an intensive course would have the same staying power for students (over a vacation period or for the long-term) as would the language learned in the same number of hours spread over a longer time. Language Maintenance Programs Although t so ubiquitous around the world as second language teaching programs, pioneering programs for second language maintenance do exist and should be the object of applied classroom research. In Japan, for example, contributors to the present volume are familiar with programs designed to help returnees maintain their English skills. In the Saturday classes available to the children, little chance is given them to use their second language in communicative situations. As of now, only activities involving local skills or individual subskills seem to be included, such as spelling tests, memorization games, and simple listening or reading comprehension exercises. Few activities include interactive production exercises that involve communication gaps and the use of global language skills (Yoshitomi 1992). Thus, since the returnees' ability to use subskills erodes more slowly than
Investigating Second Language Attrition
17
their global ability (see Chapter 4, this volume), their successful performance in these activities is likely to give them, their teachers, and their parents a false impression that the children are retaining English quite well. This underestimation of the children's L2 attrition may have a negative effect on language maintenance as it masks the need to use English communicatively in order to retain the ability to coordinate linguistic subskills. Allotting some time in class for students to talk with a competent speaker one-to-one, even infrequently, is considered by Yoshitomi (Chapter 4, this volume) to be more beneficial than making them work on "shallow" activities that do not necessitate the coordination of their language skills. What applied classroom research needs to investigate, then, is whether or not the increase in maintenance programs of class activities that involve real-time spontaneous interaction aids language retention more than activities involving individual subskills. In addition to face -to-face interaction, technology offers exciting and innovative options for supporting communication-based language use (Fotos 1996). Worldwide correspondence networks which include e-mail, chatrooms and computer lined classrooms now provide opportunities for spontaneous communicative practice. This technology has handed returnees around the world an immediate and inexpensive means to communicate with native speakers of languages learned abroad. Interactive Computer Assisted Learning (CAL) programs can be designed to utilize the global language skills of students for use in maintenance programs as well (Otto 1988). Applied research should explore the effects on retention of such communicative on-line uses. Training in Language Retention Strategies During this decade many language teachers have recognized one of the most valuable gifts they can give their students: help in learning how to learn. Materials for teaching strategies for successful language learning have proliferated (Brown 1989, 1991; Ellis & Sinclair 1989; Rubin & Thompson 1994), based on a substantial L2 research literature that examines the characteristics and strategies of the "good language learner" (For reviews see Dickenson 1992; McDonough 1995; O'Malley & Chamot 1990; Oxford 1990). Similarly, as we enter a new century, the work of language attrition researchers can provide a foundation for teaching learners to maintain their language effectively (Berko Gleason 1988). While much less is known about language attrition than about language acquisition, one feature they clearly share is the enormous difference between individuals in their success in learning or keeping a second language. Some are able to maintain language skills, while others with similar learning histories and attrition circumstances lose them. Just as language acquisition researchers during the 1980s attempted to understand the characteristics and strategies of "the good language learner," so now language attrition researchers in the new millennium and beyond have the challenge of finding the characteristics and strategies of "the good language keeper."
18
Second Language Attrition in Japanese Contexts
Notes 1. The definition assumed here is the one current in the literature on bilingualism as detailed by the following entry from The Cambridge Encyclopedia of Language (Crystal 1993:362). What is a bilingual? The obvious answer is: someone who speaks two languages. But this answer will not suffice. It does not allow for those who make irregular use of one or other language, or those who have not used the language at all for many years (so-called "dormant" bilinguals). Nor does it allow for the many people who have developed a considerable skill in comprehending a foreign language, but who do not speak it; or those who have learned to read in another language, but who cannot speak or write it. ... And above all, this definition says nothing about the level of proficiency that has to be attained before speakers can legitimately claim to be bilingual. The notion of proficiency raises some very complex issues. Again, the "obvious" answer is to say that people are bilingual when they achieve native-like fluency in each language. But this criterion is far too strong. People who have "perfect" fluency in two languages do exist, but they are the exception, not the rule. The vast majority of bilinguals do not have an equal command of their two languages: one language is more fluent than the other, interferes with the other, imposes its accent on the other, or simply is the preferred language in certain situations.. . . Scholars now tend to think of bilingual ability as a continuum: bilingual people will find themselves at different points on this continuum, with a minority approaching the theoretical ideal of perfect, balanced control of both languages, but most being some way from it, and some having very limited ability indeed. 2. Although Gardner stood by his statistics that support his socioeducational model, in the face of academic criticism (Au 1988), he admitted that "association" is the more accurate description of the relationships between attitudes, motivation, and language proficiency, rather than "causality" (Gardner 1988).
PART I
The Loss of English as a Second Language by Japanese Children
This page intentionally left blank
2 Japanese Returnees' Japanese Returnees' Retention of English-Speaking Skills: Changes in Verb Usage over Time Anita Reetz-Kurashige
This study tracked the English-speaking ability of 18 school-age Japanese returnees for a period of 12-18 months. The data included interviews, free talking, and storytelling. We measured the changes in the subjects' retention of ability to tell or to retell stories and specifically examined children's changes in verb forms over their first three years back in Japan. The purpose of the research was to provide pedagogically useful information on attrition, highlighting verb forms.
The Problem and Its Setting Returnees Returnees, called kikoku shijo in Japanese, who are the main subject of this study, need to be defined. Tetsuya Kobayashi of Kyoto University (1983:i) has described kikoku shijo as those who have had the experience of being born in Japan and brought up in the mainstream of Japanese culture. When they reached a certain age, they went overseas with their parents. While they were being brought up overseas, they received some influence from the local culture. After a few years, they returned home and were brought up in the mainstream Japanese culture once again. Over 10,000 kikoku shijo have returned to Japan each year since 1988 (Shoowa:63). As Figure 2.1 shows, the 1992 volume of returnees reached 13,219. The number of returnees was steady through the early 1990s, totaling 12,000 to 13,000 between 1992 and 1996. The annual number of returnees is expected to remain above 10,000 through this century.
21
22
The Loss of English as a Second Language
Figure 2.1 Annual Numbers of Returnees by School Level Consistently, the school-age returnees tend to be young. Through the 1980s and early 1990s, slightly over 60% of the returnees were in elementary school, just under 25% in junior high school, and a relatively small 15% in high school. While the tendency has been for children to attend high school and university in Japan, two new trends in education have helped push up the proportion of high school students going abroad with their parents. First, the Japanese Ministry of Education (Mombushoo) has advised Japanese high schools to recognize academic credits earned in high schools abroad. Secondly, an increasing number of private colleges and universities have introduced a separate kikoku shijo entrance examination or special admissions considerations for high school graduates who have spent three years of high school overseas with their parents. 1 Nevertheless, the preponderance of younger returnees is not expected to change, given the relatively young age of the adults sent abroad. English has emerged as the preferred language of instruction for Japanese children studying abroad. In 1992, over 96% of Japanese children on a prolonged stay in North America attended local schools, as opposed to Nihonjin gakkoo (Japanese schools) administered abroad by the Japanese Ministry of Education. In Europe, 65% of resident Japanese children attended local schools (Mombushoo 1993:5). The importance of acquiring English as a second language is not lost on the parents of returnees. After Mandarin Chinese, English is the most widely spoken language in the world, with an estimated 440 million native speakers. 2 English is spoken as a first or second language in at least 45 countries, or one-third of the world. By focusing this study of returnees' L2 on English, we address the most widely acquired L2.
Japanese Returnees' Retention of English-Speaking Skills
23
Problem of Returnees L2 Maintenance Returnees face the problem of trying to keep their English once they return and are surrounded by and actively engaged in daily communication in Japanese. Understandably, drastic reduction in L2 input and interaction creates circumstances for L2 loss." Input frequency, an environmental factor, modifies linguistic knowledge," according to Yoshitomi (1992:304) because items that are frequent in input are neurologically processed and stored differently from items that are processed infrequently. The problem is that the returnees function in a stunningly different language environment in Japan, where English-language speakers are scarce and the opportunity for spontaneous and natural interaction in English is even scarcer. Parents of bilingual children frequently experience frustration in trying to recreate a social setting that would bring learner and native speakers together. Although English as a Second Language (ESL) scholars like Kyoko Yashiro and others are working to introduce foreign language study at the upper elementary school (koogakuen) level (see Yashiro 1991), the Ministry of Education has resisted expansion of the curriculum because it runs counter to the current objective of eliminating Saturday classes to give a full two-day weekend. In 1993, however, a few target elementary schools throughout Japan were selected to begin teaching foreign languages, including English, in the fifth and sixth grades. Because this is a pilot effort, we expect families to remain still largely on their own in the task of encouraging their children's L2 abilities. Accordingly, the desire to develop a curriculum for returnees' L2 maintenance has stimulated the need to explore the ESL attrition process of these children. Issues in Returnee Attrition Research No complete model of attrition has been accepted to date. The regression hypothesis, which postulates L2 loss occurring in the reverse order of L2 acquisition has received qualified support in studies by Cohen (1975), Berman & Olshtain (1983), Jordens, de Bot & Trapman (1989), Hansen (1980), HansenStrain (1990), Hansen & Shen (1999), Kuhberg (1992), and Olshtain (1986, 1989). Olshtain's (1989) attempt to trace changes in linguistic features of Israeli returnees' ESL attrition found no clear regression patterns. However, she documented linguistic features that are prime candidates for attrition and found extensive evidence of change in verb forms. The critical threshold notion, that attainment of some level of L2 competence mitigates against loss, has also been partially supported in studies by de Bot & Clyne (1989), Weltens (1987), Weltens et al. (1989). Bahrick's (1984a) survey, Fifty Years of Second Language Attrition, which plotted plateaus of recall and recognition remains the benchmark study. Specification of threshold levels of ESL proficiency in listening, speaking, reading, writing, and various syntactic components has never been completed.
24
The Loss of English as a Second Language
In the past ten years, case studies of returnees' L2 maintenance by Kuhberg (1992), Olshtain (1989), Tomiyama (see Chapter 3), and Yoshitomi (see Chapter 4), have documented the sequence of loss of at the intraskills level, including code switching, lexical retrieval, fluency, phonology, morphology, and syntax. Although based on a limited number of subjects, these studies provide a base of evidence for generalizing on patterns of returnees' L2 retention. A major two-volume study of Japanese returnees' ESL retention was published in 1989 and 1990: Kikoku shijo no gaikokugo hoji ni kansuru choosa kenkyu hookokusho (A survey on the foreign language retention of returnees) by the Kaigai Shijo Kyooiku Shinkoo Zaidan (Japan Overseas Educational Services, JOES). The survey reported on differences in returnees' retention and related these to personal characteristics of age and time elapsed since return. In one study, Yoshida & Arai (1990) measured listening comprehension in tasks that included repetition of increasingly complex sentences. The responses were evaluated for correctness of morphemes and grammatical structure. In the grammatical analysis, they found that the younger groups had higher errors, most frequently omissions and substitution. The youngest children who had been back the longest (over two years) also had the highest rate of substitution of words and phrases. Older children who had been back the longest showed the highest accuracy in repetitions. Yoshida & Arai identified linguistic weaknesses in the returnees' speech through error analysis and found that inaccurate verb substitution ranked the highest. The above studies point to a number of unsolved issues. The regression hypothesis has proved difficult to assess for lack of documentation on acquisition patterns (Olshtain 1989). The critical threshold notion has not been quantified for ESL. Documentation on Japanese returnees' loss of ESL is still fragmentary. The bulk of attrition research to date has been building from the bottom up, documenting retention of specific skills and charting changes over time. This study continues that effort. Attrition is admittedly an incremental process, and there is no clear distinction between acquisition and attrition. This presents a hidden problem in study design. Studies that group subjects by the time elapsed since the end of L2 training and the onset of testing, and then test for proficiency and draw conclusions on retention, make the implicit assumption that all subjects learn about the same amount over the same acquisition period. However, studies of L2 acquisition (Wong Fillmore 1976, 1991) have shown that children of similar age and length of time in the L2 environment vary considerably in their L2 competence. There is a pedagogic need to distinguish between failure to acquire and attrition because relearning a second language may take less time than initial learning (Hansen 1980). Furthermore, there is a need to develop longitudinal tracking evidence and compare it with evidence from cross-sectional or between groups study designs. Accordingly, in this study I investigated attrition patterns by comparing each subject with his or her own previous performance. In this way, this study can distinguish between language loss and failure to ever acquire some forms, and, more important, it can document actual shifts in language use.
Japanese Returnees' Retention of English-Speaking Skills
25
Purpose and Scope of Study This study takes a close look at returnees' verb usage, which Olshtain (1986, 1989) and Yoshida & Arai (1990) identified as highly susceptible to change and error, in order to shed light on several unanswered points. First, we have not yet answered what changes in language occur over time in Japanese returnees' speech. Specifically, we want to ascertain if there are any patterns of change in verb usage. Then, we can examine whether any personal characteristics of the children relate to their retention patterns, and, if so, what factors predict L2 retention. The following research questions have been framed to assess returnees' competence over time in using verbs in storytelling. (1) What common changes and shifts over time occur in returnees' use of verb forms in ESL narrations? Can any patterns of change over time be identified? (2) Do the personal characteristics of returnees' age, length of time abroad, length of time back, and proficiency level at study onset relate to their retention/attrition in using verbs? If so, what factors relate favorably to (are predictive of) L2 retention? Method The investigation through storytelling included two tasks and three subject groups. Storytelling included a retelling after a tape and an original story told from a picture book. The three subject groups were 18 returnees, 10 Japanese children of elementary school age living in Honolulu and attending local English speaking schools, and 14 native English speakers in the same range. These three groups were compared in lexical and morphosyntactic assessments. Then the study assessed the returnees individually at two or three times during approximately 12-19 months. Narrations were analyzed quantitatively and qualitatively. Measurements included type-token counts and ratios to determine vocabulary diversity and targetlike usage (TLU) assessments of accuracy in morphology. Qualitative analysis focused on comparisons between subjects' initial and final descriptions of the same story pictures. Subjects Three groups of elementary school age children comprise the subjects of this study: returnees whose L2 was measured for change in Japan, a baseline group of Japanese children whose L2 narration was recorded in Honolulu, and a native English-speaking group recorded in Honolulu. Table 2.1 (see Appendix) contains the personal data on age, length of residence, and length of schooling. The average age of the children was nine years old. Returnees had lived abroad for an average of 2.4 years, compared with the baselines' 3.5 yrs.3 The native English speakers had lived in the United States continuously and attended school there. The following section contains additional information of each group.
26
The Loss of English as a Second Language
Returnees Tables 2.1 and 2.2 (in the Appendix) sh ow the personal and test characteristics of the returnees. At study onset, the eighteen returnees had lived from 1-5 years abroad and had attended school in English from 1-4 years, 2.2 years being the average (Table 2.1). In Table 2.2, returnees are grouped by time elapsed between return and study onset (incubation period). As shown, twelve children were back less than three months at study onset, and six children had been back from 8-22 months. I deliberately included subjects who had been back over a year in order to gather evidence on plateaus or thresholds of retention. Table 2.2 shows that all subjects had an initial and a final assessment, and eleven had an interim one also. Initial and final assessments varied from 7 months for Subjects 1 and 3, to 18 months for Subject 17. Eleven of the returnees were interviewed three times and the other seven were interviewed twice. The returnees took a qualifying literacy test; Table 2.2 also shows those scores. Subject eligibility required scoring 50th percentile or better on the language section of the Stanford Achievement Test, Abbreviated, Primary 1, Form J, (Psychological Corp., Harcourt, Brace, and Jovanovich, 1988), and a primary -level (grade 1) reading test. The purpose of a literacy requirement was to avoid mixing in preliterates whose attrition behavior is known to be different; Berman & Olshtain (1983) and Hansen-Strain (1990) clearly associated literacy with stability of knowledge. Initial arrangement of data in Table 2.2 by incubation period is consistent with cross-sectional analysis and allows us to chart changes for subjects within the framework of time lapsed since their return. After the verb analysis, subjects are ranked by retention levels, and it is in that format that we later examine the relationship between subjects' retention and their personal characteristics. Baseline Subjects The purpose of the baseline group was to obtain examples of levels of competence reached by children relative to their length of residence and schooling abroad. We examined their proficiency and L2 learner errors. Through these baseline data we can identify language variations in vocabulary, morphology, and syntax that may also occur in the returnees' speech. Their language variations inform us that some structures have not yet been acquired. The ten Japanese children, ages eight to eleven, in this group were attending the Rainbow School, a Saturday Nihongo gakkoo in Honolulu. Native Speakers of English We evaluated the returnees against same-age native English speakers rather than against a standard norm of (adult) English speakers because research has shown that linguistic and discourse features change significantly with age. Children's vocabulary and syntax develop during their first decade as they learn to handle more detail and narrative complexity. Children's use of verb tense and aspect (Ellis 1987), pronouns (Clancy 1982), and lexical and syntactic complexity
Japanese Returnees' Retention of English-Speaking Skills
27
(Slobin 1990) develop incrementally. The immature storyteller often switches apparently randomly from past to present tense and back, whereas the older child shows a more definite time sense, usually selecting the past tense for narration (Ellis 1987; Slobin 1990). It would be unfair to assert that a young returnee lacks stability in verb tense if native English-speaking youngsters lack the same skill. This subject group consisted of 12 students, ages six to twelve, who were attending elementary school in Honolulu. Storytelling Storytelling was the task selected to measure changes in verb use primarily because it offers repeatability and comparability. In addition, the particular stories used in this research have been used in other studies of bilingual children's L2 attrition. Frog, Where are You? by Mercer Mayer (1969) and A Boy, A Dog, and A Frog by the same author (1967) are picture books with detailed drawings and no text. They are companion books that have the same three characters (a boy, a dog, and a frog) and many development sequences that invite variety in vocabulary and syntax. Because there is no text, a child must rely on his or her own recall abilities. The frog stories were used in attrition studies by Berman & Olshtain (1983), & Olshtain (1986, 1989), and Yoshitomi (1994). The Book is a set of 19 drawings with no text; but an accompanying audiotape in English provides approximately one explanatory sentence per picture. The tape includes many phrasal verbs, a negative verb, and a passive voice. The Book was used in Hansen-Strain's (1990) study of English-speaking children remembering and forgetting Japanese. Testing Procedures For the first task, The Book was given each child, who looked at the pictures while listening to the English tape. Those pictures were put aside and the child was then given Frog, Where Are You? and told to look at all the pictures and afterward tell the story in English. Then storytelling and taping began. When the child finished, he was asked to look again at the pictures for The Book and retell that story. The nonsequential procedure is meant to diminish short-term recall of the exact words on the tape. Before the second taping, I reviewed the returnees' first stories and identified vocabulary weakness and failure to complete constructions or to describe images. Some subjects had skipped over details in the more complex pictures. I wanted to hear if they could formulate a proper sentence based upon a prompt. Through informal trials, I found that some subjects could integrate a prompt into the narrative, indicating that they knew the semantic and syntactic use of the prompt. Other subjects simply repeated the prompt and did not integrate it into the narrative, suggesting that the lexical item and/or its use were beyond their knowledge. Subjects who ignored the prompt totally provided no clear information. To explore further, an interventionary approach was adopted for subsequent storytelling as follows:
28
The Loss of English as a Second Language
(1) Warm up period to review the two stories with the subject before storytelling begins. The interviewer invited the subject to ask the names of anything before taping began. (2) During storytelling, if the subject asked for a word, paused more than 10 seconds, or repeated herself several times for want of a word, the interviewer supplied it.4 (3) If the subject skipped a picture or an explanation, which other children usually explained, the interviewer probed further, with something like, "What's happening here?" Samples of the prompting are (a) The subject below picked up on a prompt: Returnee: he was going down a steep hill Interviewer: cliff Returnee: cliff. And the deer stopped/and the dog fell down the cliff. (b) Conversely, another subject ignored a prompt: Returnee: he has worried look. Interviewer: Yes, he looks worried. Returnee (later referring to the same frog): * is again worried look. After the initial storytelling, we faced the problem that subsequent telling of the same story becomes familiar, so the task may involve less language creation and more recall of one's previous performance. As a check on this practice effect, four subjects (7,11,16, and 14) who had been back in Japan for different lengths of time, from less than a month to 19 months, were asked to tell the companion story, A Boy, A Dog, and A Frog in the second session (Time 2). There was no discernible effect that this story change had on these subjects' subsequent (Time 3) narration of Frog, Where Are You? Data Analysis All narrations collected through testing were transcribed according to a modified version of Slobin's (1989) Frog Story Procedures. (See the notes for transcription notation). The following measures were obtained from the data. Measure of Vocabulary Diversity Based on Type-Token Counts "In mathematical linguistics the total number of words in a text may be referred to as the number of text tokens and the number of different words as the number of text types. The ratio of different words in a text to total words in the text is known as the lexical density or Type-Token Ratio" (Richards, Platt & Weber, Dictionary of Applied Linguistics 1988:300). Overall lexical density is calculated by: [different words (types) total words (tokens)]. The analysis is carried one step further in this study by classifying all types and tokens by part of speech (verbs, nouns, articles and so forth). This breakdown provides an additional
Japanese Returnees' Retention of English-Speaking Skills
29
qualitative measure of speech production that relates to the grammatical composition of each narration. Verb Analysis: Targetlike Usage Measured Over Time The methodology in the verb analysis followed these steps: (1) Count verb phrases as one unit and classify the unit by type. Actual types included present, past regular and past irregular tenses; present and past progressive aspect; negatives; passive voice; infinitives; and past participles. tried to touch was broken got stuck did not come hear singing was able to get up
tried - past regular, to touch = other (infinitive) passive voice, was = past irregular, broken = other (past participle) passive voice, got = past irregular, stuck - other (past participle) negative, did = past irregular, come (verb stem) hear = present tense, singing = other (present participle) was = past irregular, able = adjective, to get up = other (infinitive)
(2) Apply the TLU formula as set forth in Pica's (1983) methods of morpheme quantification : No. Correctly Supplied Morphemes/Forms in Obligatory Context = SOC No. Obligatory Contexts + No. Supplied in Nonobligatory Context = OC + SNOC The numerator is a count of the number of verbs correctly supplied by a subject in his/her story. It is labeled SOC for supplied in obligatory contexts and it indicates accuracy. The denominator counts the subject's total verbal production, correct and incorrect, per story. It combines OC, or obligatory contexts, and SNOC, supplied in non-obligatory contents. OC indicates the number of verbs that the subject supplied for correct verb usage given the context of the narration. SNOC indicates the number of verbs the subject supplied in addition to the number of verbs required for the given context. Consider the sample calculation of one subject's sentences: The bee chasing the dog. The owl chased the boy, and the boy fall of from the tree. (Correct forms: The bee was chasing the dog. The owl chased the boy, and the boy fell off the tree.) Employing the TLU formula for past regular (chased) - 1/1 or 100%; for past irregular (fall instead of fell) = 0/1 or 0%; for past progressive (chasing) = 0/1 or 0% because the auxiliary is missing. The overall TLU = 1/3 or 33%. In counting correctness (SOC) the general rule is, if the sentence is grammatically correct and the verb forms are correct on the sentence level, the verb is counted as correct. This rule allows for shifting of verb tense in the narrative and is consistent with tense shifting among the native speakers of elementary school age. There are some instances, however, in which, a correct form at the sentence
30
The Loss of English as a Second Language
level has not been accepted as correct: for example, when the narrator has established consistency in using the past tense and then abruptly shifts to the present tense for one or two verbs. the boy and dog caught the frog .. . they watched the frog . .. while boy and dog sleep the frog slip out next morning boy and dog saw the frog was not in the jar In this example, slip is an error because it is incorrect as either the present or past form. Sleep is arguably correct as a present tense form at the clause level but incorrect at the narrative level because TLU would normally call for the past tense form, slept. I counted sleep as incorrect here because past tense had been established as the narrative form. When this type of shift occurred it was also footnoted. The following section summarizes the study results and elaborates on the findings. Results Overview In the initial comparison of the three groups, type-token counts indicated that the recent returnees resembled the baseline group in vocabulary diversity. This implied that discernible loss in productive vocabulary had not set in for youngsters who had returned within three months. From the quantitative TLU analysis and from qualitative comparison of each child's stories over a period of approximately 12-18 months, we found that the 18 subjects demonstrated some common changes and shifts in their language patterns and specifically their use of verbs. These changes could be generalized best by ranking subjects in terms of TLU accuracy at final assessment. Some changes and errors appeared to occur sequentially and were characteristic of particular levels of ESL competence: 1. Shifts in tense and aspect occurred from past to progressive and from past progressive to present progressive. These changes were made by lowretention speakers whose stability of language was eroding. 2. Common errors in morphology included dropping the bound -ed morpheme in the past tense, the -s for the third person singular, omitting the auxiliary in the progressive, or failing to have the auxiliary agree in number with the subject. Moderate-to low-retention speakers lost bound and free morphemes. 3. Use of the verb stem was a particularly prevalent feature of attrition for low-retention speakers. These interlanguage speakers showed instability
Japanese Returnees' Retention of English-Speaking Skills
4.
5. 6.
7.
31
in tense and morphemes from study onset and increased their reliance on unanalyzed forms over time. Returnees' use of negatives over time showed few word order errors, some shifts to simpler forms, and progressively increasing errors in bound and free morphemes. This suggests an attrition pattern of morpheme loss occurring before reduction in complexity and reductions in complexity (which constitute a type of compensatory strategy) occurring before syntactic errors. Simplification often appeared to be an attempt to avert syntactic errors. Most subjects who scored high in TLU accuracy and retention still showed declines in vocabulary diversity and/or increased redundancy. Lexical loss appeared be a precursor to morphosyntactic loss. Qualitative comparison of each subject's stories over time documented shifts in usage that occurred without error: reduced vocabulary, increased redundancy, "downshift" in verb tense, and selection of simpler forms, particularly in forming or avoiding negation. Initial proficiency scores were the best predictors of retention. Age and length of residence emerged as interdependent in that children over eight years old and with at least two years of school abroad retained over 80%TLU. Children who were either under eight years old or had less than two years of schooling abroad showed 30%-79% retention of TLU. Both conditions of being over eight years old and having over two years study abroad needed to be met for over 80%TLU to continue for more than twelve months.
Elaboration of Results Vocabulary Diversity of Subject Groups The three subject groups were compared for vocabulary diversity using the first storytelling of The Book. Table 2.3 in the Appendix shows average type-token counts overall and by parts of speech for all subjects. As we can see, the tape of The Book had a type-token ratio of .43 compared with a similar .44 for native speakers, baselines, and recent returnees. Returnees back the longest had the lowest, .39. The tape, however, contained the highest absolute vocabulary diversity, with a total of 213 words tokens and 92 different word types. In contrast, native speakers averaged only 185 tokens and 86 types. Recent returnees and the baseline group were nearly identical, with a typetoken ratio of .44 (62/142) and .44 (64/145), respectively. The type-token counts by parts of speech (in bold print in the table) as well are nearly identical for the baselines and recent returnees. Groups 2 and 3, returnees who had been back longer, showed lower vocabulary diversity of .41 (62/152) and .39 (52/132), respectively. We recognize the strong similarity of the ratios for each subject group, that is, the strong overall similarity of the groups in lexical density. Noteworthy is the almost identical type counts of the baselines and recent returnees.
32
The Loss of English as a Second Language
This aggregate analysis of The Book is the clearest demonstration we have that the narrative speech of the recent returnees and baselines who are still living abroad are virtually indistinguishable on the point of the lexical density by parts of speech. This evidence suggests that for some recent returnees, productive vocabulary did not decline immediately upon leaving the English-speaking environment. Because their vocabulary corresponds so closely with bilinguals still in the L2 environment, these limitations indicate failure to acquire rather than lexical loss following return. In the following subsections we examine changes over time in targetlike usage of verbs forms, including negative forms and relate the results again to the subjects' personal characteristics. Tracking Targetlike Usage of Verbs In both tables, the subjects are listed from high to low retention based on their overall score (SOC no. verbs) at final testing. The overall fraction score is equivalent to accuracy productivity and, hence, is more meaningful than the %TLU. Change over time (Change/time) in the tables, compares each subject's final %TLU with his/her initial %TLU. Interim scores were ignored in this calculation but presented in the tables and included in the discussion. These tables break down accurate use by tense and aspect (e.g., present, past regular, irregular; present and past progressive), giving the SOC and %TLU for each category. They do not show errors explicitly. There are two reasons for this: the nature of the errors themselves and the nature of the TLU calculation. It was impossible to determine if an error like he jumping should be classified as past or present progressive, or an error like he fall, under simple present or simple past. Additionally, unless the subject supplied at least one (SOC) verb correctly to generate a numerator, the identifiable mistakes counted in the denominator (OC SNOC) could not be reflected in %TLU (e.g., 0 5 = no. TLU score). Thus, Tables 2.4 and 2.5 (see Appendix) focus on changes in accuracy only since no credit was given for erroneous but still communicatively competent language. The tables also include a final column of notes, which itemizes specific characteristics of each subject's narrations. These notes, made during preparation of TLU tables, were not quantified. I tried, but could not confidently order the occurrence of errors in a sequential (or a regression) pattern because there were only two or three assessments per subject. The interviews were exhaustive rather than frequent. So, notes 1-10 below summarize evidence of the subjects' strengths and weaknesses, the latter in the form of information on types of errors. Tables 2.4 and 2.5 (see Appendix) relate the enumerated notes to the subjects (i.e., notes are listed for each returnee whose stories showed evidence of the characteristic). 1. Used verb stem in place of analyzed verb form 2. Missed third person singular -s
Japanese Returnees' Retention of English-Speaking Skills
3. 4. 5. 6. 7. 8. 9. 10.
33
Missed auxiliary in progressive aspect Failed to have progressive auxiliary and subject agree in number Used inappropriate lexical choice; verb overgeneralized in meaning Relied on present tense (50% of total verbs) Shifted from past to present tense Shifted to progressive aspect Relied on past tense (50% of total verbs) Used stable TLU of verbs (80% + accuracy)
Tables 2.4 and 2.5 rank subjects by their overall score (SOC total no. verb) at final assessment and place subjects into three categories: high retention, moderate retention and low retention. The highest scorers all had 80-100%TLU, the moderate group measured approximately 50-79%TLU at final assessment, and the low retention group all had under 50%TLU at final testing. The characteristics of subjects in the three groups will be discussed after we examine the linguistic information in the two tables. The following three examples from Table 2.4 (Subject 13 in high retention, 2 in moderate retention, and 16 in low retention) illustrate interpretation of the data. The top scorer, Subject 13, had an overall score of 28/28 initially and 29/29 finally, so no change over time in accuracy. From first to final assessment, Subject 13 never used the present tense. The subject consistently and accurately relied the past tense and showed decline in SOC for regular past tense but increase in SOC for irregular past. The notes indicate that Subject 13 relied on the past tense (note 9), and showed stable use of verb forms (note 10). If we look at Subject 2, in the moderate group, we see output increased but accuracy decreased over time with 21/21 in Time 1 and 17/25 in Time 3, with a very high 32% loss in TLU over 12 months. Present tense use did not expand over time. The subject continued to rely on the past tense, but with less accuracy: past regular score dropped from 100%-40% accuracy and past irregular from 100%-81%. The notes indicate Subject 2 relied on the past tense (note 9) and only initially showed stable use of verb forms (note 10). Next, with Subject 16 in the low retention group, we see a precipitous drop in overall accuracy from 20/28- 9/26, a 36% decline over 12 months. Except for 66%TLU of past tense initially, the subject scored 50% or less accuracy for all tenses used. Without positive SOC scores we lack information on errors, however, the notes on Subject 16 provide details. Subject 16 used verb stems in place of present and past forms (note 1), missed third person singular -5 (note 2), missed auxiliaries in the progressive (note 3), and finally, failed to have progressive auxiliary and subject agree in number (note 4). Thus, we can conclude that differences became progressively marked over time in these three subjects' abilities to access accurately verb forms to describe the action in the stories. In the following section we will take a detailed look at common changes in verb usage and associate them where relevant, with subjects' rank order in proficiency (i.e., with the three levels of retention).
34
The Loss of English as a Second Language
Patterns of Change in Verb Usage In response to the first research question (Can any patterns of change over time be identified?) the study found that certain changes in verb usage were common in the attrition process. The following paragraphs describe these findings. (1) Shifts in tense and aspect occurred from simple past to past progressive and to present progressive. Tables 2.4 and 2.5 indicate that Subjects 18, 11, 1, and 6, from the high-and moderate-retention groups, made this shift (note 8). Subject 18 was a competent speaker who doubled her use of the present progressive aspect. Subjects 6, 1, and 11, ranked in descending order, were moderate-level speakers whose stability in regular and irregular past verb forms eroded. Increased errors in past verb forms and shift to progressive forms, not always accurately, occurred in the their Time 2 and Time 3 narrations. The examples below compare the subjects' descriptions of the same segments of the stories. They show that the more competent speakers, Subjects 18 and 6, made a tense shift without error, but the less capable, Subjects 11 and 1, dropped the auxiliary in the process. Subject 18 (age 13;9, abroad 1.0 year, high retention, auxiliary correct)7 back < 1 month
back 12 months
the boy ran from the owl and climbed up on a rock and he grabbed something
the boy is trying to get away the boy climbed up a rock and he is holding to a tree
Subject 6 (age 8;7, abroad 2.3 years, moderate-high retention, auxiliary correct) back > 1 month
back 5 months
when the boy went to sleep the frog got out of the bottle they were going to find the frog in the woods
while the boy and the dog is sleeping the frog got out of the jar the boy and the dog is looking the frog
Subject 11 (age 10;5, abroad 1.6 years, moderate-low retention, no auxiliary) back 19 months
back 35 months
boy saw (= looked) at the tree hole
and he looking in the hole
Subject 1 (age 6;6, abroad 1.6 years, moderate retention, auxiliary difficult to distinguish) back < 1 month he's going to climb up the rock
back 7 months the boy still calling the frog
(2) Common errors in morphology included omitting the auxiliary in progressive and failing to have the subject and auxiliary agree in number. As shown in Tables 2.4 and 2.5, errors of omission or agreement of the auxiliary in the progressive
Japanese Returnees' Retention of English-Speaking Skills
35
aspect (notes 3 and 4) were made by subjects in the moderate and low retention groups: in descending order of ranking, Subjects 1, 15, 9, 7, 16, and 5. There were no discernible differences among these two levels of subjects who made these two different errors. All these subjects initially showed evidence of omitting the auxiliary in the progressive (evidenced by SOC of 0 for progressive aspect). Their use of the auxiliary was unstable at study onset, which was within a month of return for Subjects 16 and 1, and within three months for Subjects 15 and 9. The following comparisons of subjects' initial and subsequent descriptions illustrate increasing error over time. Subject 15 (age 11 ;6, abroad 2.10 years, moderate retention, past irregular and progressive auxiliary errors) back 3 months
back 11 months
the boy fell from the tree
the boy surprised and fall from the branch the bee chasing the dog
and bees chasing the dog
Subject 16 (age 11;6, abroad 1.6 years, low retention, auxiliary missing) back < I month
back 19 months
and the b e e s . . . [go] went down bees chasing with dog
bees house fell down in ground and bees chasing dog
Subject 1 (age 6;6, abroad 1.6 years, moderate retention, shift in tense, subject and auxiliary don't agree) back < 1 month
back 7 months
the boy and the dog was looking the frog
the boy and the dog is looking the the frog
(3) Two bound morpheme errors that occurred over time were ropping -ed for the past regular and -s for the present third person singular. Errors in the past tense are reflected only in %TLU scores. Subjects with moderate to low retention (in descending rank order: 6, 2, 10, 11, and 16) showed reduced accuracy in the regular past. Dropping the -s morpheme for the third person singular was also a common error for subjects of moderate retention (in descending rank order with 4, 10, 1, 11, 7, 9, and but we are less sure that this morpheme was solidly in place at study onset. The examples below show that the -ed was initially used correctly, whereas the -s morpheme was not always in place initially. Subject 10 (age 9; 11, abroad 1.0 year, moderate retention, -ed for past used erroneously) back 2 months
back 14 months
he wanted to get the book he tried to pay for this book
he buyed the book but he didn't have money . . .
36
The Loss of English as a Second Language
Subject 11 (age 10;5, abroad 1.6 years, moderate retention, -ed dropped) back 19 months
back 35 months
he dropped the book again and the man picked up the book and handed it to Mr. S
in the bus he drop his book and the nice man pick up his book he take it for him
Subject 9 (age 9; 10, abroad 2.10 years, moderate retention, shift from past to present and -s omitted) back 3 months
back 11 months
and he drops the book again and he dropped the book and the nice man picked his book up and the nice man pick it up and handed it to Mr. Sakamoto and hand [him] it to him and he [sitting on a chair] sat on a and he noticed that a page of the chair book was XXX look the book [the book] XXX Subject 4 (age 7;10, abroad 2 years, moderate retention, dropped -s morpheme) back 9 months
back 14 months
he drops the book . he took the book and he got a bus
he drops that book he pick up the book and he go on the bus
(4) Increasing use of the unanalyzed verb stem wa a prevalent feature of attrition for low-retention speakers. Data in this study point to a single form, the verb stem, as a basic "chunk" that the least proficient speakers recalled and relied on. Pienemann et al. (1988) informed us that learners acquire mastery of the third person -s morpheme rather late in syntactic development. When the -s is not supplied, the "unanalyzed" simple present (verb stem) is used. As the following examples show, it is impossible to say which analyzed verb form, present or past, the subject failed to provide. Subject 15 (Time 3) the boy surprised and fall from the tree branch Subject 9 (Time 3) 'owl stop the chasing/boy climb up the rock We want to relate these increased errors to personal variables and indicate where relationships are strongest. Initially, I grouped the subjects by time back in Japan and arrayed the erroneous use of verb stem in that format. Then I arranged the subjects by TLU scores and showed verb stem data in that format. By comparing
Japanese Returnees' Retention of English-Speaking Skills
37
these two formats, it was clear that errors were not strongly related to time elapsed since return, but they were strongly related to subjects' TLU score. Data in Table 2.6 (see Appendix) show that the number of verb stems that subjects used erroneously usually increased over time and increased disproportionately for the less proficient speakers. The implication of this to the threshold notion is discussed below. The following examples illustrate limitations in vocabulary and syntax that constrained the subjects who made verb stem errors. Subject 15 (age 11;6, abroad 2.10 years, moderate retention, verb stem errors) back 3 months
back 11 months
and the boy fell from the tree
the boy surprised and fall from the tree branch the boy get up the rocks deer stand up
so he climbed on the wall and the deer pick him up
Subject 7 (age 9;1, abroad 2 years, moderate-low retention, verb stem errors) back 8 months
back 23 months
frog went ou boy looked at (= looked for) the frog
the frog go out boy wake and look at the bin (= jar)
Subject 16 (age 11 ;6, abroad 1.6 years, low retention, verb stem errors) back < 1 month
back 12 months
he found his book that he w a n t . . . but he got card, money card
he find book that he want but he have a credit card
Reverting to the verb stem in place of simple present or past forms may reflect a typological feature of Japanese grammar. Japanese verbs have no morphemes to distinguish person or number. The plain form shuushikei 8 is used conversationally by Japanese to convey both present and future tense, specifically to convey intent, action, or states that will occur from now. So the wide use of shuushikei combined with infrequent use of subject and object pronouns makes it easy or natural for Japanese interlanguage speakers to rely on the verb stem in English. (5) Negatives showed few word order errors, some shifts to simpler forms and progressively increasing morpheme errors. Word order errors did not increase noticeably over time. For negative constructions as well as positive, there was some "downshifting" from past to present tense by Subjects 17, 10, 15, 7,and 5 (in descending order of competence in Tables 2.4 and 2.5), which indicates that simplification was an adjustment made by all levels of all speakers. (See also Tables 2.4 and 2.5, note 7.) Shifts from doesn't + verb to no + noun, called nofronting, were made by Subjects 8 and 4, in the high- and moderate-retention groups, respectively. The most obvious changes were increased errors in bound
38
The Loss of English as a Second Language
morphemes (i.e., -ed on past tense), free morphemes (i.e., agreement of subject and in auxiliary number), and in lexical choices. (See also Tables 2.4 and 2.5, notes 3, 4, and 5.) At the point of one year after their return, the returnees' syntax of negative sentences was not noticeably affected. At two years after return, only one proficient speaker (Subject 14) showed signs of syntactic decline with word-order errors. Of the three subjects tested at three years after return, one, the oldest returnee (Subject 17), once shifted from negative to positive phrasing in mid sentence as a corrective measure, and the youngest (Subject 5) had an increase in errors and made the most errors overall. The following examples illustrate changes in phrasing and increases and decreases in accuracy over time. Subject 14 (age 11;1, abroad 5.5 years, high retention, word order errors) back 11 months
back 26 months
but he couldn't find it/the frog it wasn't there
the frog was not in the jar ... nowhere was the frog nowhere the frog came the frog wasn't anywhere
Subject 17 (age 12;3, abroad 3.10 years, high retention, negation abandoned, switched to positive sentence) back 17 months
back 35 months
he realized that he forgotten the wallet at his house
but [he didn't have a] he forgot his wallet at home
Subject 2 (age 7;4, abroad 2.10 years, moderate retention, lexical and article errors) back < 1 month
back 14 months
he noticed he didn't have any money he thought he didn't put a money to the pocket Subject 4 (age 7;10, abroad 2 years, moderate retention, shift to no-fronting) back 9 months
back 21 months
I don't have a money
he has no money
Subject 7 (age 9;1, abroad 2 years, moderate retention, shift to present, lexical + morpheme errors) back 8 months
back 23 months
but he ... didn't have the money he remember
but he don't (= doesn't) have the money he forgot the pillow (= purse) at house
Japanese Returnees' Retention of English-Speaking Skills
39
Subject 5 (age 8;7, abroad 1.10 years, increased free morpheme errors) back 17 months
back 33 months
frog was not there frog is not there he has not (= no) money
dog's head don't (= didn't) get out there aren't a frog there is not (= no) money
Subjects continued throughout their testing to attempt to produce relatively difficult, complex clauses with negatives and showed surprisingly little loss of syntax. The interview process itself was stimulating and encouraged subjects to try their hardest. For the subjects of this study, over the first year to a year and a half after return, their syntax was sturdier than their lexicon and morphology. A lot of infrequently used words dropped out of the returnees' vocabulary, and their morpheme errors increased over time, but grammatical structures that were properly in place at the time of their return were retained. The apparent robustness of syntax is further considered in the following discussion. Personal Fa ors and Their Relation to Loss We need to show the range of retention in verb usage for the 18 subjects before we can relate the independent variables of age, time spent abroad, time back, and TLU scores at study onset to the dependent variable of retention. By ranking the subjects according to their TLU scores in Tables 2.4 and 2.5, we establish the basis for evaluating the personal variables later. The results of Tables 2.4 and 2.5 are summarized in the more concise Tables 2.7 to 2.9 (see Appendix), which present the overall TLU scores and percentages for both stories (refer to Tables 2.4 and 2.5 for base data). The subjects separate into three groups: (1) > 80% TLU over the assessment period indicates high retention and near native speakers; (2) 50%-79% TLU at final assessment describes moderate retention and interlanguage speakers; (3) > 50% TLU at final assessment indicates low-retention and interlanguage speakers with unstable syntax. Tables 2.7 to 2.9 show the data for groups 1-3, respectively. The division between the subjects in Tables 2.7 and 2.8 is somewhat arbitrary because subjects scored differently on the two stories. Scores tended to be higher on The Book, which had an accompanying tape, than on Frog, Where Are You ? which required original storytelling from pictures. For example, the performance of Subject 6 is borderline, scoring high retention on the retelling of The Book but only moderate retention on Frog, Where Are You? (1) < 50% TLU indicates high retention by near native speakers. As shown in Table 2.7, these children maintained TLU scores at over 80%, indicating a high level of communicative ability. Their speech continued to be characterized by idiomatic, nativelike phrasing, although five out of six experienced either decline in vocabulary or increase in redundancy. The same five out of six increased their requests for vocabulary to be supplied; however, it is important to recognize that they continued to incorporate quickly and accurately any prompts given. Five subjects with residence abroad of 2.6 years or longer (14,
40
The Loss of English as a Second Language
12, 13, 17, and 8) maintained stable use of the past tense in narrative, whereas Subject 18, with residence abroad of under two years, shifted over time to the present tense. This group generally had error-free syntax, accurate use of two-word verbs, and prepositions. Only Subject 14, who was back over two years, made errors in word order. (2) 50%-79% TLU indicates moderate retention by interlanguage speakers. Generally, the following features characterize this group. They maintained communicative ability and use of idiomatic and formulaic expressions. There seems to be a threshold level even at this intermediate level of L2 acquisition at which the subjects' English has an internal cohesiveness. Erosion in vocabulary, however, led to overuse of common verbs (especially get in phrasal form, e.g., get on, get up, get to). Bound and free morpheme errors already occurred in the speech of these returnees at study onset. Bound morpheme errors included omission of -s morpheme for third person singular, omission of -ed in past regular; free morpheme errors included omission of the auxiliary in the progressive, failure to produce a correct irregular past tense, and failure to have subject and progressive auxiliary to agree in number. The evidence is clear that children of comparable age, time abroad, and time back varied widely in proficiency at both the beginning and the end of the study. Table 2.8 shows that the moderate retention group had the widest variation. Initial TLU scores ranged from 48% to 100%, and final scores from 47% to 83% at study onset. Retention had the widest variation in this group: from 28% gain (Subject 10) to 32% loss (Subject 2). We found through parent and subject interviews that the quality of educational experience these children had abroad also varied from excellent elementary schools and close rapport with teachers to never getting out of an ESL class where English input was primarily from nonnative speakers. The differences in educational experiences coupled with interlanguage speakers' characteristically unstable command of some morphosyntactic structures probably explain much of the variation. (3) > 50% TLU indicates low retention and unstable interlanguage speakers. The test results for this group are summarized in Table 2.9. The striking characteristic of returnees with low retention scores of > 50% TLU was the instability of their syntax at study onset, which was within a month of one subject's (16) return to Japan. These children (9, 16, 3, and 5) had not acquired sufficient knowledge to maintain communicative ability outside the L2 environment. It appears that they had failed to reach a critical threshold of speaking competency. Their attrition was an accelerated version of loss experienced by returnees with moderate retention. This group had a very limited vocabulary, was uncertain of grammar, and put full attention to producing a relatively fluent story with no apparent concern for accuracy. For these subjects, reversion to the verb stem in place of a correct present or past form was most extreme. Like a spinning top losing speed, their English showed ever increasing variation and instability. We should not. however, dismiss children in this group because of low scores. These youngsters struggled to communicate in English. They were inventive in their use of English, and with determination and composure they carried off storytelling that was clearly difficult for them.
Japanese Returnees' Retention of English-Speaking Skills
41
Thus, in response to the second part of the research questions, Do personal characteristics of returnees' age, length of time abroad , length of time back, and proficiency level at study onset relate to their accuracy in using verbs? If so, what factors relate favorably to (are predictive of) L2 retention? We can see from Tables 2.7-2.9 that high competence at study onset was the most predictive factor of L2 retention. Short time elapsed since return was not a common characteristic of subjects with high retention. Also, contrary to Yoshida & Arai (1990) findings, being older did not guarantee high retention. I concluded that it is not one factor but a combination of factors that predict strong retention. Near native-speaking English ability was virtually a requirement for minimal attrition. Children over eight years old, with literacy skills in English built up over two years in school abroad, were the returnees with highest L2 retention. Children under eight years and children with less than two years schooling abroad experienced attrition of 30% to 79% in accuracy of verb forms. Thus, the factors of age and length of stay are interdependent. The significance of these findings to the critical threshold notion is considered in the following discussion.
Discussion This study of 18 returnees found common changes and shifts in use of the English verb in storytelling. Specific changes could be related to particular levels of proficiency as measured by TLU of verb forms. Evidence of these changes has both theoretical and pedagogical implications. On the theoretical side, the study findings support the inverse hypothesis that the higher the subject's proficiency, the lower the degree of attrition. It also substantiates the notion that attaining a critical threshold of competency mitigates against language loss. The study further found that more than age or length of time abroad, proficiency in speaking was the best predictor for retaining accuracy in speaking. This finding that links retention patterns to proficiency contradicts Weltens's 1989 conclusion that "attrition is independent of training level" (de Bot & Weltens 1995:153). Weltens's results agreed with Bahrick's massive study in which the amount forgotten over a five-year period following training was "relatively constant for individuals at different levels of training" (Bahrick 1984a:116). How can we explain this apparent contradiction? The answer, as described in the Introduction to this volume, lies in the study designs. First, this study measured speaking skills, whereas most of Weltens's and much of Bahrick's measured reading and listening. Productive skills, as research has shown, are more susceptible to attrition than recognition skills. Second, this study focused on children whose L2 competency was much below Weltens's college students' or Bahrick's adults'. Third, this study reported individual loss patterns through tracking the same subjects over time, whereas Weltens and Bahrick both measured attrition by comparing groups. Thus, a closer look at task, subject characteristics, and measurement design resolves the apparently contradictory findings of this study and Weltens's and Bahrick's. Although these results also support the notion of a critical threshold level that secures retention, given the limited time frame of this study, a question
42
The Loss of English as a Second Language
arises about the relative robustness of syntax. Does it reflect structural stability of the returnees' L2 or simply that a much longer period must elapse before syntactic attrition sets in? Considering research by Weltens (1989) and Bahrick (1984a), we believe that long-term syntactic stability probably characterizes returnees who have achieved a high level of speaking competence. A longer tracking period, however, would be the definitive way to answer this question. Comparing this study with related research from the Kikoku shijo no gaikokugo hoji, this study extends the findings of Yoshida & Arai (1990), which reported characteristics of retention but not patterns of change. This study sheds light on what happens to verb usage over time; what is altered or lost and how it is altered or lost. Many differences that Yoshida & Arai found in speaking ability between younger and older children were substantiated in this study. Both studies agree on the influence of age on retention, the wide gap between recognition and recall abilities, and the susceptibility of vocabulary and morphology to loss versus the relative robustness of syntax that has stabilized at some threshold level. On a comparative linguistic level, Japanese children's English attrition followed patterns similar to Israeli children's English attrition as reported by Berman & Olshtain (1983) and Olshtain (1986, 1989). There were, however, distinctive differences. This study supports Olshtain's claim that typological and pragmatic similarities between the L1 and L2 lead to positive reinforcement for L2 retention, whereas L1 and L2 dissimilarities lead to interference and attrition. Japanese children's erroneous use of the English verb stem in place of analyzed forms may be attributable to the wide use of the base form shuushikei in Japanese. Turning to the pedagogical implications of the findings, our comments on assessing competencies, defining instructional goals, providing appropriate-level L2 input, and designing classroom tasks may be useful to ESL teachers of returnees. In assessing competencies, teachers need to be aware that current measures go beyond the four skills to assess fluency: accuracy in vocabulary, expressions, and grammar; idiomatic usage; discourse responsiveness; and motivational factors. Defining instructional goals for returnees requires considering the wide discrepancies between their receptive and productive abilities. Yoshida & Arai (1990:25) made the point that listening ability is easy to keep and should be used as a key to retain and build up other skills. Reading is also an important anchor for the literate child to retain the L2. Building up productive or speaking skills requires intensive practice, but recognition skills require virtually no effort for retention. Providing L2 input at the appropriate level is essential. In teaching, we know that vocabulary, idioms, grammatical complexity, and speed of discourse need to be appropriate for each child's level so that the L2 input fits, yet reasonably challenges, each child's abilities. The experience of prompting subjects during this research made it clear that children will pick up only cues that fit their language level. Prompting a returnee who hesitates in speaking can jog his or her memory and ability to say nearly forgotten words and phrases. Speed and complexity of input must also match the learner's level.
Japanese Returnees' Retention of English-Speaking Skills
43
In designing classroom tasks, naturally, the instructional needs of near-native speakers are different from the needs of less competent speakers. Instruction of high-level students should maximize contextualized speaking practice, and "real" discourse in the classroom should have some time pressure. Less proficient speakers with unstable syntax need more explicit instruction, including grammatical explanations. In many areas, the interlanguage speaker needs to relearn rather than recall. Time constraints are not appropriate for hesitant speakers. Both high-and low-proficiency students benefit from conversational practice that resembles "on line" discourse. Yet, reliance on short responses will fail to build grammatical, syntactic complexity. Exercises that require explanations, comparisons, and complicated descriptions will challenge returnees to produce clauses rather than phrases and complex rather than simple sentences. Although this study used TLU as an analytical tool that measures only accuracy, it should be emphasized that some returnees who used inaccurate or erroneous language are still successful storytellers. They are able to convey their ideas even with very limited vocabulary and rudimentary grammar. Teachers working to improve returnees' ESL skills need to encourage communicative ability more than accuracy. It is hoped that understanding some of the dynamic processes of language retention and attrition will increase teachers' and parents' awareness and ability to make their own personal observations and to help keep returnees' L2 abilities. I want to point out that dropping efforts to maintain a second language is also an option for returnee families who put their priority in other areas. There is a choice. There is no right or wrong. Our time is limited and we want to be effective.
Appendix Table 2.1. Subjects' Personal Data Age1
Subjects
(Years)
School3 (Grades)
LOR2 (Years)
Months since Return
Returnees
18 17
16 15 14 13 12 11 10 9 8 7
6
5 4 3 2 1 Avg.
13.7 12.2 11.5 11.5 11.1 10.8 10.8 10.4
9.9 9.8 9.2 9.1 8.6 8.6 7.8 7.5 7.3 6.5 9.8
7 1-5 (1/2) 4-6(1/2) 2-4 K-3 1-4 1-4 1-2 3 1-3 1-3 1-3(1/2) K-2(l/2) K-l K-l 1-2(1/2) K-l (1/2) K-l (1/2) 2.2 yrs
1.0 1.8 1.5 2.8 5.4 4.7 4.7 1.5 1.0 2.8 3.6 2.0 2.2 1.8 2.0 1.5 2.2 1.5 2.4
Baseline
10 9 8 7 6 5 4 3 2 1 Avg.
11.8 10.5 10.0
9.5 9.5 9.3 8.9 8.8 8.7 8.3 9.5
1-6
5.5 3.0 3.0 4.5 1.5 2.6 2.3
2-5 2-5 1-4 2-3 2-3 1-3 1-3 1-3 1-2 3.5 yrs
6.6 3.6 2.2 3.5
Native Speakers
14
13 12 11 10 9 8 7
6 5
11.8 11.5 11.5 11.3 10.3
9.8 9.6 8.9 8.9 7.8
11.8 11.5 11.5 11.3 10.3 9.8 9.6 8.9 8.9 7.8
K-6 K-5 K-5 K-5 K-4
K-4 K-4 K-3 K-3 K-2
44
<1
17 <1
3 11 <2 <2 19 3 3 <2 8 <1
17 9 <1 <1 <1
Table 2.1. (continued) Subjects
4 3 2 1 Avg.
Age1
LOR2
School3
Months since
(Years)
(Years)
(Grades)
Return
7.8 7.0 6.8 6.8 9.3
7.8
K-2
7.0 6.8 6.8 9.3
K-2 K-l K-l 2.9
Notes: 1Age and time periods are in decimals; 5.5 - 5 years, 6 months. -LOR = Length of residence in L2/L1 country. Given for L2 country for returnees and baseline. 3 Years in L2 school for returnees and baseline; for L1 school for native speakers.
45
Table 2.2. Returnees' Interview Times and Literacy Scores Subject
(ID no.)
Age
LOR-L2
(Years)
(Years)
Literacy Test1 Time 1
Time 2
Time 3
(percentile score)
Group 1 (Recent Returnees Back 0-3 Months)
18
16 15 13 12 10 9
8 6 3 2 1 Avg.
13.7 11.5 11.5 10.8 10.8
9.9
9.8 9.2 8.6 7.5 7.3 6.5 9.8
1.0 1.5 2.8 4.7 4.7 1.0 2.8
3.6 2.2 1.5 2.2 1.5
2.5
<1 1 3 <2 <2 2 3 <2 <1 <1 < 1 <1 <2
nd2 6 nd 5 5 5 nd 5 5 7 5 7 5.6
12 12 11 10 10 14 11 10 14 nd 14 nd
82 99 99 99 99 79 98 99 68 88 66
11.8
89.6
26 23 21
95 99 79
23.3
91.0
99
Group 2 (Back 8-11 Months)
14 7 4 Avg.
11.1
5.4
9.1 7.8 9.3
2.0 2.0 3.1
11 8 9 9.3
17 14 14 15.0
Group 3 (Back 17-19 Months)
17 11 5
12.2 10.4
Avg.
10.4
8.6
3.8 1.5 1.8 2.4
17 19 17 17.7
nd 25 nd 25.0
35 35 33
99
34.3
88.7
88 79
Nates: The grouping of subjects in Table 2.2 is not intended to indicate especially logical groupings of subjects according to anticipated retention patterns, The study findings on subjects' relative ranking on a retention-attrition continuum, presented later in Tables 2.4—2.9, does group the subjects by L2 accuracy in verb forms and permits further analysis of the personal variables of age, length of time abroad, and accuracy of verb forms at study onset. 1 Literacy test was the Stanford Achievement Test, Abbreviated, Primary 1, Form J (Psychological Corp., Harcourt, Brace & Jovanovich 1988). 2 nd means no test data for test time shown.
46
Table 2.3. Vocabulary Diversity in Storytelling: (The Book) Comparison of Returnees, Baselines, and Native Speakers
Tape token ratio = .43
Overall
Verb
Noun
Articles
Pronoun
Preposition
Conjunc.
Modifiers
Others
Type Token
Type Token
Type Token
Type Token
Type Token
Type Token
Type Token
Type Token
Type Token
92
213
29
45
19
46
3
21
8
39
13
26
8
17
7
11
5
11
17 25 18
2
3
6
4
5
1 3 2
2
6
4
9 23 15
4
10 6
4 10 7
0 2 1
0 3 2
13 10 13
4 1 3
5 1 4
3 0 1
4 0 1
9 27 16
2 5 4
2 4
1 4 4
2 7 3
12 33 20
2 15
2 16 7
1 11 4
2 18
Returnees' Range: low high Avg. ( n = l l )
46 70 62
113 165 142
13 19 18
20 32 29
13 17 16
23 27 29
Range: low high Avg. (n = 3)
59 64 62
125 178 152
16 22 19
28 33 30
13 17 16
28 47 4
Range: low high Avg. (n = 3)
38 62 52
124 139 132
18 15 17
26 26 26
18 13 15
35 27 32
Range: low high Avg. (n = 10)
48 82 64
110 199 145
12 24 19
21 39 29
15 24 18
24 41 31
Group 1 : Back 0-3 Months (Type-token Ratio =.,44) 3 39 3 2 10 3 22 4 16 7 33 7 13 2 14 5 28 6 13 5 2 Group 2: Back 8-11 Months (Type-token Ratio = .41) 5 5 19 3 22 58 3 4 20 4 24 6 15 2 4 12 19 3 23 7 12 Group 3: Back 17-22 Months (Type-token Ratio == .39) 2 17 7 25 6 13 3 2 21 3 26 2 10 2 20 5 23 6 12 3 2 Baselines (Type-token Ratio = .44) 10 3 10 6 12 2 2 7 4 25 8 33 15 27 4 16 5 27 8 18 3
6
6
Native Speakers (Type-token Ratio = .44) Range: low high Avg. (n = 14) 1
55 116 116 254 86 185
Note: Returnee data are from first testing
18 32 25
21 51 36
15 26 18
24 51 38
11 23 17
11 23 15
3 9 6
25 44 35
5 13 9
11 27 19
8 4
6
5
6
Table 2.4. Returnees' Targetlike Usage" of Verb Morphemes and Tense over Time for The Book Subjects (High to low retention)
Overall Change/ LOR/L2 SOC Time Present Age (Yrs) No. Verbs %TLU T1-T3 SOC %TLU
Past Reg. SOC %TLU
Past Past Other Present Irreg. Prog. Prog. (Infin. SOC %TLU SOC %TLU SOC %TLU Participles)
Notes
High Retention 13.timel:<2 mo. time 2: 5 time 3: 10
10.8
14. timel: 11 mo. time 2: 17
11.1
12. time i: < 2 mo. time 2: 5 time 3: 10
10.8
17. time 1: 17 mo. time 2: 35
12.2
6. time 1 : < 1 mo. time 2: 5 time 3: 14
8.6
8. time l:< 2 mo. time 2: 5 time 3: 10
9.2
4.7
5.4
4.7
3.8
2.2
3.6
— — —
13 10 8
100 100 100
12 15 16
100 100 100
0
0%
0 0 0
0%
0 0
— —
7 8
100 100
15 18
100 100 100
0%
0 2 0
— 100 —
10 10 6
100 100 100
16/21 22/26
76 85
9%
0 0
— —
6 9
30/30 21/23 21/25
100 91 84
0
—
26/27 21/21 20/24
96 100 83
28/28 28/28 29/29
100 100 100
24/24 28/28
100 100
24/24 24/24 24/24
1
3
9,10
— —
0 0
—
2 2
9,10
— —
3 2 1
100 100 100
— —
0 0
—
1
2 2
100
100 100
0 0
11 10 17
100 100 100
0 0
55 82
10 13
77 87
0
0 0
1 1
100 100 100
— —
0
0 0
0
9,10
0 0
9,10
5,9,10
100 100 66
16 10 12
100 100 92
0 0 0
— 1
1
100 100
-16%
66 —
9 8 2
2
2 0
—
2
100
3 0 5
— — —
10 7 2
100 100 50
15 14 14
95 100 78
0
— — —
0 0 3
—
0
-13%
0 0 0
0 0
100
1
1
9,10
Moderate Retention timel1:
7.3
4. time 1:9 mo. time 2: 14
7.8
10. time 1:2 mo. time 2: 5 time 3: 14
9.9
l. time 1: 1 mo. time 2: 7
6.5
2.2
2.0
1.0
1.5
11. time 1: 19mo. 10.4 time 2: 25 (BDFb) time 3: 35
1.5
15. time 1: 3 mo. time 2: 11
11.5
2.8
7. time 1:8 mo. time 2: 14 time 3: 23
9.1
2.0
—
-32%
0 0 0
-30%
56
18/30 22/43
60 51
21/28 11/24 12/25
75 46 48
13/22 12/24
60
13/20 19/27 12/21
65 70 57
21/21 13/16 17/25
100 81 68
31/35 17/29
89 59
16/26 16/25 14/25
61
100
16
—
4 3 2
100 71 81
0 0 0
— —
40
10 13
11 11
100 50
4 1
100 100
12 5
83 100
0 0
— —
0 3
-5%
2 2 2
50 50 100
3 3 3
100 40 38
9 9 9
100 50 60
0 0 0
— — —
1 8
50 44
2 3
22
-9%
15 7
83 70
0 1
4
5 14 1
83 3 20
9
50
75 7
6
6
60
50
—
60
-27%
3
44 1 19
1 5
50 50
6 2
75
-10%
— —
4 6 6
50
-8%
0 0 0
60 66 75
1 0 2
9,10
— 100
3 0
1,2 7,8
2 2 0
100 100 —
0 0 0
2,9 2,3,9
— 20
0 2
— 66
0 1
2,3
60
0 64 0
— 0 —
0 — 0
— 0 —
3 — 2
0 1,2,5
4
60 100
0 0
— —
0 0
— —
0 1
6 6 5
75 60 63
0 0 0
— — —
0 1 -2
— 100 0
1 6 1
0 0
— —
0
9
1,3,7
2,5,7,8 1,2,3,4
(continued)
Table 2.4. (continued) Overall Subjects (High
LOR/L2
to low retention)
Age
(Yrs)
9. time 1: 3 mo. time 2: 1 1
9.75
2.8
Change/
SOC
Past
Time Present
No. Verbs %TLU
T1-T3
Reg. SOC %TLU
Past
Present
Past
Other
Irreg.
Prog.
Prog.
(Infin.
SOC %TLU
SOC
%TLU
SOC
%TLU
Participles)
— —
0 0
SOC
%TLU
1
100 80
3 2
50 30
5 2
45 30
0 0
— —
0 0
4 3 3
66
12 9 4
66
0
0
0
33 —
50 33
0 0
0
0
0
0
Notes
Low Retention
16. time 1:< ! mo. time 2: 6 time 3: 12
11.5
5. time 1:<1 mo.
7.5
1.5
time 2: 33 a
50
20/28
71 50
13/26 9/26 1.5
time 2: 7 3. time 1: 17 mo.
9/18 9/26
8.6
1.8
6/20 4/21 14/24 15/22
35
35
-15%
0 1 -36%
50
1,3,5
2 0 2
2 1,2.3
1
0 2
—
—
18
5 1
25 18
0 0
— —
0 0
— —
1 0
1,2,3 1,2,3
1 9
33 74
2 0
40 —
10 2
83 100
0
10%
— —
1 2
100 66
0 2
4,5,6,7
0
58 68
0
37
1,
-11%
30 19
5
Notes: TLU = Targetlike usage. Formula = SOC (OC + SNOC). See explanation on p. 29. b BDF means subject told the companion story, The Boy, the Dog, and the Frog. 1. Relied on verb stem in place of present and past forms. 2. Missing third person singular s morpheme. 3. Missing auxiliary in progressive. 4. Failure of progressive auxiliary to agree in number with subject. 5. Inappropriate lexical choice/overgeneralization of verb meaning. 6. Relied on present tense. 7. Shift to present from past tense. 8. Shift to progressive aspect. 9. Relied on past tense. 10. Stable TLU of verb forms (high or low).
0
Table 2.5. Returnees' Targetlike Usagea of Verb Tense and Aspect over time for Frog, Where Are You? Subjects (High to low retention)
Age
Overall LOR/L2 SOC (Yrs) No. Verbs
Change/ Time Present %TLU T1-T3 SOC %TLU
Past Present Irreg. Prog. SOC %TLU SOC %TLU
Past Reg. SOC
%TLU
100 76 100
28
5
Past Prog. SOC
%TLU
Notes
High Retention
14. time 1: 11 mo. time 2: 17 (BDF) time 3: 26
11.1
12. time l: < 2 mo. time 2: 5 time 3: 10
10.8
18. time 1: < 1 mo. time 2: 12
13.7
13. time 1: < 2 mo. time 2: 5 time 3: 10
10.8
17. time 1: 17 mo. time 2: 35
12.2
8. time 1 :< 2 mo. time 2: 5 time 3: 10
9.2
5.4
4.7
1.0
4.7
3.8
3.6
43/43 44/57 65/68
100 77
40/41 48/50 43/45
99 96
50/60 63/68 35/37 40/42 36/39
96
96
0 -4%
39/42 46/51
93
36/41 26/27 33/37
88 95
90
89
1 2
100 100 100
100 55 100
3 2
75 100
95
11 1 10
100
10 16 13
3 1
20 7
100 95 100
13 24 21
100 90
14 27
100 78
96
5
10
-3%
4
10%
7 23
70 96
5 7
83 100
26 16
93 100
8 15
98 79
2
100
5 11 6
100 100 100
13 17 15
100 100 100
10 9 10
90 100 83
1 0 0
100
-3%
3 3 1 1 9
100
-3%
100
14 13
93 93
22 21
96 87
1 1
100 100
1 3
100 100
5.7.10
— —
7 11 5
64 100 83
26 13 22
92 100 92
0
1%
0 0 0
— —
3 2 3
100 100 75
9,10
95 95 92
100 94
100 100 100 80
83 93
0 8 19
5 4
100 100
0 0
0 1
9,10
33 100
10
9,10
(continued)
Table 2.5. (continued) Subjects (High to low retention)
Age
Overall LOR/L2 SOC (Yrs) No. Verbs
Change/ Time Present %TLU T1-T3 SOC %TLU
Past Reg. SOC
%TLU
Past Present Irreg. Prog. SOC %TLU SOC %TLU
Past Prog. SOC
%TLU
Moderate Retention 8.6
2.2
6. time 2: 5 time 3: 14 10. time 1:2 mo time 2: 5 time 3: 14
9.9
4. time 1:9 mo time 2: 14 time 3: 21
7.8
11. time 1:19 mo time 2: 25 (BDF) time 3: 35
10.4
2. time 1: < 1 mo time 2: 5 time 3: 14
7.3
1. time 1: 1 mo time 2: 11
6.5
1.0
2.0
1.5
2.2
1.5
38/38 31/33 38/50
100 94 75
23/61 38/54 41/54
48 70 76
26/46 25/30 25/36
57 83 69
26/39 29/35 15/25
66 83 60
20/24 27/32 22/38
83 84 58
19/32 19/38
50 50
0
0
— — —
11 11 7
100 100 87
25 18 22
100 100 88
25 78
60
13 21 14
52 68 77
0
—
0
33 —
2 1 7
100 100 87
5,9
1
50 0 100
5 3 1
45 75 100
7,9,
1
-25%
0
50
28%
3 0 8
80
1 14 10
60 100 56
1 4 8
33 80 73
13 8 11
72
12%
10 9 5
80 80
0 5 1
0 83 33
1 0 0
14 — —
1,2,3,6
16 48 14
3 5 5
75 83 83
12 9 4
86 90 50
8 1 5
66 100
-6%
1 11 1
70
0 0 0
— —
7 8
-25%
0 2 0
—
6 6 0
100 85 nd
10 13 19
77 93 83
0 5 0
nd 100 —
1 5 3
50 100 37
4,5,9
50
2 3
22 100
15 5
83 70
1 2
50 17
1 1
0 100
1,2,3,6
-5%
0 8
100
1 0
Low Retention 9. time 1: 3 mo time 2: 11
9.8
7. time 1: 8 mos time 2: 14 (BDF) time 3: 23
9.1
16. time 1: < 1 mo time 2: 6 (BDF) time 3: 19
2.8
2.0
9/19 17/36
47 47
19/25
76 57 47
20/35
15/42 11.5
1.5
22/34
16/29 8/22
3. time 1: 1 mo time 2: 7
7.5
5. time 1: 17 mo time 2: 33
8.6
1.5
27/36
17/35 1.8
22/37
6/20
0%
0 2
— 100
2 1
50 17
6 9
50 60
0 0
— —
1 5
33 38
25 29 0
6 4 4
86 80 44
8 5
-29%
1 2 0
10
80 42 50
1 3 1
100 50 33
2 0 0
100 — —
1,2,3,5 1,2,3,9
4
66 32 100
3 2 4
75
7 5 -14
78 62 0
4 1 1
66 17 100
2 0 2
50 —
5
100
1,2,3,4
75
7 7
100 54
5 7
71 80
0
—
16 1
80 25
1 1.00
100 33
0
65 55 26
-39%
6 1
74 50
-24%
12 1
100
3 I
60 30
-30%
1 3
100 30
4 0
Nates: "TLU = Targetlike usage. Formula: = SOC 4 (OC + SNOC). See explanation on p. 29. "BDF means subject told the companion story, The Boy, the Dog, and the Frog. 1. Relied on verb stem in place of present and past forms. 2. Missing third person singular s morpheme. 3. Missing auxiliary in progressive. 4. Failure of progressive auxiliary to agree in number with subject. 5. Inappropriate lexical choice/overgeneralization of verb meaning. 6. Relied on present tense. 7. Shift to present from past tense. 8. Shift to progressive aspect. 9. Relied on past tense. 10. Stable TLU of verb forms (high or low).
66 80 50 50
30
0
i
—
2,3,5
1,2,5
0 33
1,2,3,5
Table 2.6. Verb Stem Errors in Storytelling Number of Verb Stems Used Erroneously 1
Subjects
Initial Stories
Final Stories
High Retention
14 12 18 13 17 8
0 1 2
0 0 2 0 3 3
0 5 0
Moderate Retention
6 10 4 2 11 15 1 7
1
0 19 7 2 10 11 22 7
21 10 7 17 14 20 19 Low Retention
9 16 3 5
15 12 28 14
13 2 21* 5*
Notes: Number of months subjects had been back in Japan varied; number of months between subjects' initial and final stories varied. *indicates low output 1 Subjects ranked by targetlike usage scores.
54
Table 2.7. High Retention: < 80% Targetlike Usage" Returnees' Verb Accuracy in Storytelling over Time Frog, Where Are You?
The Book
Subjects
14. back: 11 mo.
back: 17 (BDF1) back: 26
12. back: < 2 mo. back: 5 back 10
SOC Total Overall Change Verbs %TLUa %TLU
Total Verbs
Overall %TLUa
100 100 nd2
43/43 44/57 65/68
100
100 100 100
40/41 48/50 43/45
99 96
-3
nd nd
50/60 63/68
83 93
10
28/28 28/28 29/29
100 100 100
35/37 40/42 36/39
95 92
-3
16/21 22/26
76 9
39/42 46/51
93 90
-3
26/27 21/21 20/24
96 100
95
-13
36/41 26/27 33/37
24/24 28/28
24/24 24/24 24/24
18. back: < 1 mo. back: 12 mo.
13. back: < 1 mo. back: 5 back: 10
17. back: 17 mo. back: 35
8. back: < 2 mo. back: 5 back: 10
soc
85
83
Change %TLU
11.1
5.4
10.8
4.7
13.7
1.0*
10.8
4.7
12.2
3.8
9.2
3.6
77 96
-4
96
95
88 89
Notes: *See Tables 2.4 and 2.5 for complete data. TLU - Targetlike usage. Formula: = SOC - (OC + SNOC). See explanation on p. 29. 'BDF means subject told The Boy, the Dog, and the Frog in Time 2. -nd means no test data for test time shown
55
Age LOR-L2 (Yrs.) (Yrs.)
1
Table 2.8. Moderate Retention: 50-79% Targetlike Usagea Returnees' Verb Accuracy in Storytelling over Time The Book
Frog, Where Are You?
SOCSubjects
6. back: < 1 mo.
Total Verbs
back: 5 back: 14
30/30 21/23 21/25
100 91
10. back: 2 mo. back: 5 back: 14
16/26 16/25 14/25
61 60
4. back: 9 mo.
31/35 17/29
89
back: 14 (BDF1) back: 21
2. back: < 1 mo. back: 5 back: 14
nd2
84
56
59
21/21 13/16 17/25
100 81
21/28 11/24 12/25
75 46
13/22 12/24
60
1. back: < 1 mo back: 7
18/30 22/43
60
7. back: 8 mo.
13/20 19/27 12/21
65 70
11. back: 19 mo. back: 25 (BDF) back: 35
15. back: 3 mo. back: 11
back: 14 (BDF) back: 23
SOCTotal Verbs
Overall %TLUa
-26
38/38 31/33 38/50
100 94 75
-5
23/61 38/54 41/54
48 70 76
26/46 26/30 25/36
57 87 69
20/24 27/32 22/38
83 84 58
26/39 29/35
66 83
Overall Change %TLUa %TLU
68
48
50
51
57
-30
-32
Change %TLU
-5
19/32
21/42
60 50
-10
19/25 20/35 15/42
76 57 47
-8
Notes: *See Tables 2.4 and 2.5 for complete data. TLU = Targetlike usage. Formula: = SOC - (OC + SNOC). See explanation on p.29. 'BDF means subject told The Boy, the Dog, and the Frog in Time 2 3 nd means no test data for test time shown.
56
1.0
7.8
2.0
7.3
2.2
10.4
1.5
11.5
2.8
6.5
1.5
9.1
2.0
-25
55 50
-9
9.9
12
21/38 19/38
-10
2.2
28
-6
15/26
8.6
-25
60
-27
Age LOR-L2 (Yrs.) (Yrs.)
-29
Table 2.9. Low Retention: > 50% Targetlike Usagea Returnees' Verb Accuracy in Storytelling over Time The Book
Subjects
SOCTotal Overall Change %TLUa %TLU Verbs
SOC Total %TLUa
9/19 17/36
47 47
22/34 16/29
65 55 26
-39
9. back: 3 mo. back: 11
9/18 9/26
50
16. back: < 1 mo.
20/28 13/26 9/26
71 50
3. back: < 1 mo. back: 7
6/20
30
5. back: 17 mo.
14/24 15/22
back: 6 (BDP)
back: 12
back: 33
Frog, Where Are You?
35
35
4/21
58
68
-15
8/22
-36
27/36
Overall Change %TLU (Yrs.)
LOR-L2
Verbs
9.8
2.8
11.5
1.5
7.5
1.5
8.6
1.8
0
-7
17/35
74 50
-24
10
22/37 6/20
60 30
-30
Notes: *See Tables 2.4 and 2.5 for complete data. TLU = Targetlike usage. Formula: = SOC 4- (OC + SNOC). See explanation on p. 29. 'BDF means subject told The Boy, the Dog, and the Frog in Time 2.
57
Age (Yrs.)
58
The Loss of English as a Second Language
Notes Notation Used in Transcription Stories were transcribed clause by clause following a modified version of D. I. Slobin's (1989) Frog Story Procedures, (unpublished procedures manual). Each line is a clause, providing the subject uttered a subject and verb. /
End-of-clause line where the text format was limited and run-on line was used [word] Subject's self-correction; the word given is the subject's intended word (I:...) Interviewer's comment XXX Indecipherable or garbled word The sentence has a grammatical or lexical error, e.g., finding" (looking for)—the * indicates a lexical error, the word in ( ) was not used by the subject, but is a possible and correct lexical choice). Underlining is added to draw attention to a particular word.
1. While schools differ in their admissions procedures, kikoku shijo eligibility is typically restricted to applicants whose families went abroad because of a parent's job. 2. Campion's Interactive Encyclopedia [CD ROM] 1994. In English. 3. The one-year longer average length of stay in the L2 environment of the baseline group compared with the returnees deserves comment. The baseline group provided a frame of reference for L2 lexical and grammatical development against which the returnees could be compared. The more advanced lexicon and grammatical competency in English of the baseline group indicates their steady linguistic development in English. However, their inaccuracies also dispelled any notion that three years in the L2 school system produces near-native competency in speaking English. This baseline group still showed evidence of L2 learner errors. The baselines' stories replete with learner errors demonstrates the near futility of trying to distinguish between failure to acquire language and language loss. The baseline data, in fact, make the best case for using longitudinal tracking to assess language loss. 4. Subjects' repetitions of prompts were not counted in vocabulary or output tallies. 5. T-tests on these (% change in TLU) results showed there was > 2.75 difference in tvalue between the means of the two tests, indicating that both instruments (The Book and Frog, Where Are You?) measured the same thing. The difference in measurement results of the two instruments was not significant at 99%. 6. No credit was given for erroneous but still communicatively competent language. 7. Age 13;8 means 13 years, 8 months; abroad 1.0 year means child resided in L2-speaking country one year. 8. Shuushikei is commonly used for the following types of exp -essions: a Boku ga yobu yoo: (male speaker) I'll call you/her/him/etc. b. Tsukau (rising intonation): Will you/she/he/etc., use it? c. Issho ni iku (rising intonation): Will (someone) go/comewith/you/me/her/ etc.? [or] Shall we go together? d. Gakko nu iku: I/you/she/etc., will go to
3 The First Stage of Second Language Attrition: A Case Study of a Japanese Returnee Machiko Tomiyama
Despite a short history of language attrition as a field of linguistic research, we have witnessed new insights being accumulated, hypotheses being generated, and evidence and counter evidence being provided. However, as young a field as it is and the methodological paths to the inquiry being multifold, many areas still remain basically unexplored. The present study is devoted to one such area, namely, a longitudinal study—in a genuine sense, of natural second language (L2 English) attrition within an individual in the first language (L1 Japanese) environment. Previous empirical research that has contributed significantly to the literature of L2 attrition has been based mostly on cross-sectional data of foreign language skills loss in adults (e.g., Godsall-Meyers 1981; Bahrick 1984; Myers 1984a,b; Moorecroft & Gardner 1987; de Bot & Clyne 1989; Weltens et al. 1989). These employ tests and self-reports or both, as measures to assess attrition. Longitudinal studies that deal with the attrition of naturally acquired L2 of younger children have been limited to investigations of specific grammatical structures (e.g., Olshtain 1989; Reetz-Kurashige this volume) or a subset of skills (e.g., Cohen 1989). While Yoshitomi's pioneering work (1994) on the attrition of Japanese returnees provides a wealth of information on various linguistic features as well as on subskills of English, the data are drawn from comparisons of four necessarily different individuals to represent the variables selected for investigation. Kuhberg's longitudinal research (1992) probing into the question of regression hypothesis is a study methodologically and typologically similar to the present study but deals with L1 Turkish, L2 German. Other longitudinal studies, such as the one by Kaufman & Aronoff (1991) and another by Turian & Altenberg (1991) deal with L1 attrition observed in younger subjects (two to three years of 59
60
The Loss of English as a Second Language
age at the onset). The former focuses on the morphosyntactic attrition, whereas the latter focuses on the attrition of compensatory strategies. This continuing project was launched in 1992 with the purpose of tracking the same individual's overall change in naturally acquired L2 when this individual was placed in an environment where L2 input and opportunities for use were deprived. The present contribution focuses on the first stage of the overall developmental attrition process of one child, which covers the period 2-19 months after being removed from the dominant language environment. At present, the tracking has extended over four years. (For the subsequent stages, see Tomiyama 1995, 1996, 1998). A longitudinal observation of the same individual in a natural setting (as opposed to cross-sectional data based on tests and self-reports, or both) is valuable in the following ways: 1. It is possible to establish the fact of acquisition and subsequent attrition; in attrition research, caution must be exercised in distinguishing between non- acquisition and loss; 2. It is possible to avoid the assumption taken by cross-sectional studies that the length of instruction or residence can be equated with the level of acquisition; with this assumption, the cross-sectional approach examines the degree of attrition by comparing the subject's levels of proficiency and elapsed time after being away from the L2-rich environment; 3. It is possible to integrate isolated linguistic behavior from a macroscopicpoint of view. Since this contribution is based on a case study classified as qualitative research, it should be understood that the main thrust of the project is to describe as fully as possible what was observed or not observed and to provide insights and directions for future large-scale experimental studies. At this stage of development in the study of language attrition, when no complete theory has been advanced nor relevant factors have been discerned from the irrelevant, it is crucial to document any notable phenomena with an open mind. Based on previous research, however, the following questions were highlighted for the study: 1. Are phonology, lexicon, morphology, and syntax differentially affected? 2. Are productive skills more susceptible to attrition than receptive skills? 3. How is the speed of attrition? 4. How is fluency affected? 5. How does the subject compensate for the skills he is losing? 6. Can individual differences be observed?
Method Subject's Profile The subject was a Japanese boy who will be called Ken. Ken comes from an upper-middle-class family. His parents are well educated and proficient in
The First Stage of Second Language Attrition
61
English. He left Japan at the age of 1;3 when his father was transferred to San Jose, California, and returned home when he was 8;0. During his seven years in the San Jose area, he went through a local preschool, kindergarten, and elementary school up to the second grade. As a rule set by his parents, only Japanese was to be spoken at home, but according to his mother, English became his dominant language by the final year. His report card for the second grade showed "outstanding" achievements in reading, spelling, and language arts. His playmates were English-speaking children in the neighborhood. Upon returning to Japan, he entered a local public school in Tokyo as a second grader. Naturally, L2 input became very limited. Table 3.1 lists chronologically the time of data-collecting sessions, classroom-type input, and miscellaneous L2 input events. Watching videos or other L2 activities, which could not
Table 3.1. Summary of Input
Stage
I
Total Hours
Month
Months after Returning
May/92 June
2 3
1
July
4
2
Aug.
5
Sept.
6
3
Oct.
7
4
Nov.
8
5
Dec.
9
Jan./93
10
6
Feb. March April May June July Aug. Sept..
11 12 13 14 15 16 17 18
7
Oct.
19
Data Collecting Session
Structured Input
Misc. Input
1 Computer lesson (1 hr.)/week 1 Computer lesson (1 hr.)/week 1 Computer lesson
Reading
(1 hr.)/week
1-2 Computer lessons (1 hr.)/month
1-2 Computer lessons
(1 hr.)/month 1-2 Computer lessons (1 hr.)/month 1-2 Computer lessons (1 hr.)/month 1-2 Computer lessons
(lhr)/month
8 9 10
11 12
13
16.5
3 English lessons (80 min.)/month 3 English lessons (80 min.)/month 24
Visitors Visitors Visitors
62
The Loss of English as a Second Language
be documented, were excluded from the list. The information was based on the report from the subject's mother. Starting from three months after his return, Ken went to a computer lesson once a week, one hour per session, for two months. This was a private lesson taught by an English-speaking teacher in English. There was a break for two months, and from six to ten months after returning to Japan, he went to the lesson sporadically once or twice per month. Other than the data-collecting sessions, there was a blank period for about seven months. At the point of 18 months after his return, he enrolled in a group English lesson for returnees. The lessons were 80 minutes per session, which he attended three times a month on the average. Miscellaneous events pertaining to L2 input were also limited. There were a couple of occasions when English-speaking visitors came to Ken's home and conversed with him in English. Reading in English stopped at about two months after returning to Japan. Approximately, there was a total of less than 16.5 hours of data-collecting sessions and 24 hours of English instruction over the period of a year and a half. The figures show that the total exposure hours of L2 drastically decreased from what he had experienced in the United States, even considering the undocumented input hours. According to his mother, he has adjusted well to Japanese life and is doing well in school, both academically and socially. He is a very mature, bright, and cooperative child. Data Collection Data collection began two months after Ken returned to Japan. The data reported here are based on those up to 19 months, hereafter labeled Stage l. 1 The data from the subject were gathered monthly, with a few exceptions, by the researcher visiting the subject's home (see Table 3.1 for the data-collecting schedule). Each session lasted for about one hour per subject, which was videotaped and audiotaped. The taped materials were then transcribed. In a typical session, a game was played in the beginning, mainly as a warm-up and to establish rapport in English. This was usually followed by free conversation to record spontaneous speech. Elicited speech was also collected at various points. As formal measures, the Peabody Picture Vocabulary Test (PPVT) and the Bilingual Syntax Measure (BSM) were used in the beginning to formally assess his English proficiency, and to determine the baseline for receptive lexicon and morphology, respectively. The PPVT and BSM were used repeatedly at later sessions to examine the progression of attrition. A picture book, A Boy, A Dog, A Frog, and A Friend (Mayer 1971) was employed as another elicitation device. This frog series by Mayer is now commonly used by attrition researchers. The book has a series of detailed pictures but no written text, and the subjects are free to produce a story along the pictured story line in any way they wish. Wacky Wednesday by LeSieg (1974) was used to systematically elicit his productive
The First Stage of Second Language Attrition
63
vocabulary and to examine his morpheme use as well as sentence structur in a controlled environment; that is, to force the subject always to refer to certain objects or notions whenever speech is elicited. The task was to find out and describe all the wacky things in the pictures. In general, in order not to sacrifice naturalness for the convenience of data gathering, every effort was made to secure an environment that made the subject feel at ease and enjoy himself. Baseline Data In the first session (i.e., two months after the return), Ken showed no signs of difficulty in expressing himself and sustaining a conversation with the researcher for a long period of time. He would respond to the researcher's English in English right from the beginning, and he would continue to do so throughout the session. His mannerisms and pronunciation were nativelike. Formally, his raw score on the PPVT was 101, an age equivalent of 9;3, when his chronological age wa's 8;2. He measured Level 4 (intermediate English) on the BSM. The reason he did not measure Level 5 (proficient) was that he had not acquired conditionals.
Findings A Precursor Up until five months after returning, Ken would react to the researcher in English right from the beginning of the session. However, at the sixth-month session, he did not respond in English even when the recording had started. The whole discourse of the first twelve minutes was the researcher speaking in English and Ken responding in Japanese or talking to himself in Japanese. At seven months Ken rebounded. He responded in English immediately from the beginning without codemixing or codeswitching. There was hardly any noticeable change from the sessions prior to the one at six months. At that point the sixth-month session appeared to be an exception. However, at eight months, instances of codeswitching started to appear. The sixth-month session, where there was a chunk of Japanese discourse, was indeed a precursor to a more regular attrition pattern. Codeswitching Codeswitching is an oft-observed phenomenon in the process of L2 attrition (e.g., Olshtain 1989; Kuhberg 1992) as well as L1 attrition (e.g., Kaufman & Aronoff 1991; Turian & Altenberg 1991). Kuhberg (1992) claims that for his subjects, codeswitching is "the developmentally most relevant feature in the
64
The Loss of English as a Second Language
attrition process" (p. 148). He therefore divides the attrition stages from the perspective of codeswitching as follows: Stage 1: No codeswitching Stage 2: Predominantly free morpheme codeswitching and some bound morpheme codeswitching Stage 3: Increased codeswitching; some free morpheme codeswitching and predominantly bound morpheme codeswitching Ken's codeswitching observed at this stage was all free-morpheme switching. No instance of bound-morpheme codeswitching was found. For Ken, however, a more notable pattern emerged when the type of switched utterances was focused for analysis: The switched utterances were concentrated on interjections, conversational fillers, and emotional utterances such as expressing frustration and excitement. The following is an example taken from the transcript2 of the eighth month session. (1)
1 R: Woo. Tough, tough, tough. 2 K: ee wakannai yo [= What? I can't figure it out!] 3 K: ee wakannai <even more frustrated than before> [= What? I can't figure it out!] 4 R: Where's the answer? 5 K: aa atte-ru [= Wow! It's correct!] 6 R: Is that correct? 7 K: <mumbles something to himself> 8 R: Huh? 9 K: Uh, huh. Let's do this one. [K05]
Another example is onaka hetta na [= I'm hungry!] right in the middle of a storytelling task later in the same session. Conversational fillers were also susceptible to codeswitching as in (2). (Japanese fillers are underlined.) (2) 1 2 3 4 5
K: R: K: R: K:
And went to the, uh, uhm, to the king. Mm. Palace. Uh, huh unto ne (2 sec) de (5 sec) eeto ne [= uh, well, uhm]
The First Stage of Second Language Attrition
65
6 R: Yeah. 7 K: And he gave it to the king. [K05] (See also [3], [5], [11], and [12] for other instances of fillers indicating hesitation.) As shown above, interjections, fillers, and emotion-laden utterances were very likely to be switched to L1 at this stage. Lexical Attrition Lexical Retrieval
Difficulty
Codeswitching is one observable surface phenomenon with possibly multiple motivations for its use. One apparent motivation for Ken's codeswitching was to compensate for his lexical retrieval difficulty. From the eighth month on, there were many examples of his switching to Japanese to request a particular English word. He even had trouble with a seemingly familiar word like English. Below are some examples: (3)
1 K: There was uhm (4 sec) to eigo-tte nan te iu n dakke [= Uhm, how do we say English?! 2 R: English. 3 K: English. [K05 = 8 months after return] (4) 1 K: sangatsu-tte May to March dotchi dakke [- Which was March. May, or March?] 2 R: March. 3 K: March? 4 R: Yeah. You're gonna move in March? [K07 =11 months after return] (5) 1 K: <whispers> eeto (1 sec) nawatobi-tte nan te iu n da [= U h m , . . . how do you say jump rope?] 2 R: Jump rope. 3 K: Yeah, jump rope. [K07] Previously, when he had encountered difficulty in locating the exact word, he would prompt himself in English or request the researcher in English as in (6) and (7). (6)
1 K: Uhm (2 sec) the gold fish is in the what. Lotion. Lotion bottle. [K03 = 6 months after return] (7) 1 K: Huh? The table (2 sec) what, what do you call it, legs? 2 R: Yeah. One of the legs. [K03]
Lexical retrieval difficulty also manifested as long pauses as in (8), which is also taken from Wacky Wednesday data:
66
The Loss of English as a Second Language
(8)
1 2 3 4 5 6 7
K: R: K: R: K: R: K:
And there- (5 sec) there's (6 sec) Yeah, there's (3 sec) You wanna talk about this one? Yeah. Do you know the name for it? denshinbashira [= utility pole] [K08]
This example clearly shows that he was looking for the word pole, which was accessible to him six months before as shown in (9), and that he was waiting for the word to come to him. (9)
There's not the pole here. [K03]
Even though Ken started to experience trouble with immediate retrieval with some words, they were' not completely lost. Examples (10) and (11) show that with chanting and prompting, he was able to recover what was not readily available to him. (10) 1 R: But, uh, she'll be coming soon, you said, to make guacamole. 2 K-: Uhm (2 sec) in (2 sec) <whispers> gogatsu-tte nan da eeto [= What is May? Uhm] 3 R: January 4 K: February, March, April, May. 5 R: Right. 6 K: She's coming in May. [K09] ( 1 1 ) 1 R: Do you know how to say this? 2 K: eeto [= uhm] 3 R: It starts with an F 4 K: The faucet is upside down. [K08] Tracking of Lexical Items As shown in (8) and (9), the development of attrition of lexical items can be observed clearly by looking at the data taken from the controlled environment of Wacky Wednesday. Unlike spontaneous data, this type of data ensures that a certain vocabulary item indeed existed initially (i.e., acquisition confirmed through his spontaneous use) and enables the tracking of each item over time. Quantitatively, there were 83 items, including objects and concepts that were possible to keep track of in his use. In Stage 1,13 items (16%) were seemingly lost.3 Qualitative change can be observed in terms of how a particular situation or an item was referred to. Table 3.2 provides examples of the change over three sessions. It lists items that were used spontaneously in the third session (at six months) and their subsequent change in the expression to refer to the same object or phenomenon. It also shows the strategies Ken used to compensate for the difficulty
The First Stage of Second Language Attrition
67
Table 3.2. Examples of Lexical Attrition No. 3 (6 Months)
pole driver's seat
leg sideways piled
No. 8 (12 Months)
denshinbashira [switching] [avoidance]
feet, foot sideway going up [paraphrase]
No. 12 (17 Months)
stand [approximation] untenseki ga aru tte (= How do you say there is a driver's seat?) [direct request in L1 ] foot wrong [paraphrase] this way [diexis]
Notes: Compensatory strategies are given in [ ]. Japanese utterances are given in italics.
in lexical retrieval: codeswitching, approximation, avoidance, direct request in L1, and paraphrase. All of these strategies are very much in common with those cited in L2 acquisition (e.g., Faerch & Kasper 1983), L2 attrition (e.g., Olshtain 1989), and L1 attrition (e.g., Turian & Altenberg 1991) studies. In addition, Ken frequently used deixis when the word was not available to him, as in (12). To express three houses being piled up, he said: (12)
The three houses. Uh, has to be (7 sec) doesn't belong this way. <points upward> [K12]
In short, Ken's attrition in productive lexicon was evident in Stage 1. His lexical retrieval difficulty manifested in various aspects, including long pauses, codeswitching, and employment of other compensatory strategies.
Fluency In parallel with codeswitching and lexical attrition, there was a change in Ken's fluency. Indications of his nonfluency were abundant in spontaneous speech starting at the eighth month. However, to obtain more tangible evidence of attrition in fluency, quantitative data from the storytelling sessions are provided below. Pauses, repetitions, and self-repairs were selected as features manifesting his nonfluency. Table 3.3 lists the number of occurrences of pauses, repetitions, and self-repairs as well as the total length of pauses in seconds. The figures are based on the data taken at nine months, thirteen months and seventeen months from the storytelling task using A Boy, A Dog, A Frog and A Friend. Since the total number of words for each storytelling was of course not the same, the number of occurrences was converted to a percentage figure showing how many times or how many seconds he paused or made repetitions per 100 words.
T
T h e L o E s o s s f n h a i g a s l e c o n d L a n g u a g e
Table 3.3. Deterioration of Fluency Session No. 6 (9 Months)
No. 9 (13 Months)
No. 12 (17 Months)
Pauses: Sec. No.
23.0% 6.6%
24.0% 8.6%
40.0% 10.5%
No. of repetitions
1.6%
2.9%
3.0%
Self-repairs: Specific no. No. of wrong words
4
0 0
2
Storytelling Features
0
6
1
As the table shows, except for the difference between nine months and thirteen months for the total length of pauses, Ken progressively paused more and paused longer each time. He also made more repetitions as time progressed. For self-repairs, not only the number of instances but also the nature of each repair was examined. At nine months, all of his repairs were instances of making the statement more specific or accurate; for instance,
(13)
the turtle was (2 sec) the boy thought the turtle was dead. [K06]
On the other hand, at 13 months and 17 months his repairs were mainly for the instances where he made a wrong choice of word as in (14), or where he made a change in the sentence structure (false start) as in (15). (14) (15)
his right hand, right, right foot [K09] the boy (1 sec) was trying, when the boy (1 sec) was trying to (2 sec) put the fishing pole [K09]
Also, his repairs were sometimes unsuccessful, as in (16). (16)
and the dog (2 sec) were, and the tur- and s- the thing that (he) the thing (3 sec) the thing (2 sec) that pulled (he) the dog's tail and the dog (2 sec) got in-to the water, [K09]
In sum, Ken's fluency as defined by pauses, repetitions, and self-repairs deteriorated progressively.
Morphological Attrition Compared to the attrition of productive lexicon and fluency, attrition of morphology was less evident. The grammatical morphemes examined were regular plural,
69
The First Stage of Second Language Attrition
third-person singular, progressive, auxiliary (both singular and plural), copula (both singular and plural), possessive, regular past, and irregular past. The BSM data and spontaneous speech data elicited at the initial data gathering session (two months) served as the baseline of Ken's performance (i.e., acquisition). "Acquisition" was operationally defined as supplying the morphemes 90% of the time or more in obligatory contexts. At the first session at two months—Ken had attained the criteria for all of the above morphemes (as shown on Tables 3.4 and 3.5). The BSM data elicited subsequently at 9 months and 16 months showed that all of these morphemes were supplied 100% of the time for each obligatory context except for one time when he missed the plural for nose at 16 months (see Table 3.4). The spontaneous speech data, on the other hand, revealed that the plural and the irregular past dropped to less than 90% at 13 months and 17 months (see Table 3.5). Even so, these morphemes were supplied over 80%, which is the criterion some researchers use in L2 acquisition (Ellis 1994; Huebner 1983).The storytelling data supplement the BSM and spontaneous speech data where the number of occasions was small or particular morphemes were not elicited. These data show a similar pattern (see Table 3.6). However, a drop from 100% level at 13 months to 69% at 19 months for the irregular past may call for a more detailed analysis. There were three irregular verbs that were correctly used at 13 months but incorrectly used at 19 months: make, take, and bite. In other words, for these verbs, attrition, not nonacquisition, could be confirmed. At 19 months, make occurred twice. For both times the zero morph was used. Take occurred six times altogether. Of those six times, the irregular past was supplied four times, while the zero morph was used for the other two occasions. Bite occurred twice. For one occasion, the -ed form was used and for the other, the correct irregular form was used. It appears that the supply of the correct form had become rather unstable around the end of Stage 1. Qualitatively, in the spontaneous speech data, Ken's self-repair of irregular verbs could be observed. Example (17) observed at 13 months, and (18), at 19 months, show these repairs.
Table 3.4. Accuracy of Morphemes BSM-—Data Session
Plural Third person Progressive Auxiliary Copula Possessive Regular past Irregular past
No. 1
No. 6
(2 Months)
(9 Months)
100(11/11) 100 (2/2) 100(11/11) 100(10/10) 100(11/11) 100 (2/2) 100(1/1) 100(3/3)
100 (7/7) 100(1/1) 100(6/6) 100 (6/6) 100 (9/9) 100(1/1)
100(4/4)
No. 11 (16 Months) 86 (6/7) 100(1/1) 100 (5/5) 100(6/6) 100 (6/6) 100(3/3) 100(2/2) 100(3/3)
The Loss of English as a Second Language
70
Table 3.5. Accuracy of Morphemes—Spontaneous Data Session
Plural Third person Progressive Auxiliary Copula Possessive Regular past Irregular past
No.l (2 Months)
No. 9 (13 Months)
100(16/16) 100 (6/6)
83(15/18) 100(1/1) 100(9/9) 100(10/10) 100(18/18) 100(6/6)
100 (47/47) 100 (4/4) 100(5/5) 100 (27/27)
89(8//9)
No. 1 (17 Months) 80(16/20)
100 (4/4) 100 (4/4) 100 (42/42) 100 (2/2) 86(19/22)
Note: A dash shows a structure was not elicited.
(17) Get, got uhm,... [K09]
(18)
There's two boys come, came from America or something. [Kl3]
In brief, although morpheme attrition was not as evident as lexical attrition, irregular past and plural had started to show possible signs of instability by the end of Stage 1. Syntactic Attrition In the area of syntax, there was even less evidence to claim that attrition took place. Word order and other frequently studied structures in acquisition research such as negatives and interrogatives were all unaffected. However, a glimpse of indication was found in the relative clause structure. The baseline data of spontaneous speech at two months showed that he had acquired the relative pronouns: He spontaneously used the relative clause structure six times, and the relative pronouns were supplied 100% of the time for all those occasions. Examples of his spontaneous use are given in (19) and (20). Table 3.6. Accuracy of Morphemes—Storytelling Data Session
No. 6 (9 Months) Plural Third person Progressive Auxiliary Copula Possessive Regular Past Irregular past
— — 100 (3/3) 100 (3/3) 100(7/7) — 100 (3/3) 100* (17/17)
No. 9 (13 Months)
No. 13 (1 Months)
— — 100 (8/8) 100 (8/8) 100 (8/8) —
— — 100 (3/3) 100 (3/3)
100 (5/5) 100(16/16)
Note: A dash shows a structure was not elicited.
100 (7/7) — 100 (6/6) 69(18/26)
71
The First Stage of Second Language Attrition
(19) (20)
Roy got a cap that says Roy. [ KO1 ] Or whatever flavor that she liked. [KO1]
Starting at around 13 months, however, a few instances of missing relative pronouns could be observed: (21) (22)
There's two boys come, came from America or something? [K13, spontaneous speech] There was a boy, had a dog and a frog pet. [K09, storytelling]
These sparing instances may prove to be a prelude to a more regular pattern of attrition. Receptive Lexicon A very clear case of nonattrition was the receptive lexicon. Ken's receptive skill was virtually unaffected within Stage 1. As shown in Table 3.7, Ken's raw score on the PPVT was 101 at the start (i.e., two months after his return). At 12 months he scored 101 again, and at 17 months he scored 105. In light of the definite lexical retrieval difficulty Ken was undergoing, the differential attrition of productive versus receptive skills in the lexicon was evident. Phonological Attrition Another unaffected area was Ken's phonological skills. There were very few instances of phonological problems, the entire list short enough to be given in Table 3.8. The first item is a performance type of mistake, in as much as all other instances of / / were pronounced correctly. Interestingly, the other items are all loan words in Japanese, and therefore the mistakes possibly reflect cases of phonological transfer. It should be noted, however, that since there was no baseline information for these three items, they may simply be cases of nonacquisition.
Table 3.7. Receptive Vocabulary—PPVT Scores Session
Score
No.l (2 Months)
No. 8 (12 Months)
No. 12 (17 Months)
101
101
10
72
The Loss of English as a Second Language Table 3.8. Phonological Problems Session
Item
No. 9 (13 Months)
then model
/zen/ /m del/
No. 12 (17 Months)
hotel cottage
/houtel/ /kouted/
Pronunciation
Individual Differences In order to examine whether Ken's pattern of attrition reflect individual differences, two other returnees with backgrounds and experiences similar to Ken's will be introduced briefly here for comparison. The returnees, Eugene and Lily, were also involved in the author's investigation on L2 attrition (see Tomiyama 1994). They are siblings and moved to Connecticut at the age of 5;8 and 2;8 and returned home when they were 10;0 and 7;0, respectively. Eugene is an active, outgoing, and sociable child, whereas Lily is a bit shy and cautious but extremely mature for her age. Their overall attrition process showed much the same pattern. Productive lexicon, morphology, syntax, and receptive lexicon were differentially affected. Lexical retrieval difficulty and its consequences for fluency were similar for both. Eugene's codeswitching pattern and timing were similar to Ken's, although perhaps Eugene's progression was a bit swifter and more condensed. Eugene's attrition of morphology and syntax were even less evident than Ken's. The biggest individual difference could be found in Lily's data with respect to codeswitching. Ken and Eugene used the strategy of codeswitching to compensate for their attrited productive vocabulary. On the other hand, Lily did not revert to L1; instead, she used a strategy of waiting. In other words, very long pauses could be found in her data around the same time when instances of codeswitching appeared in the boys' data. Example (23) is taken from the spontaneous speech data showing Lily's long pauses at six months. (23) 1 2 3 4
R: L: R: L:
What do you do? If I, you and I are going to play this game. You just (11 sec) Yeah. (2 sec) Like (2 sec) I was it and I tag you (3 sec) then somebody uhm (2 sec) tags you then you're free. [A04]
Other data from Wacky Wednesday also showed a number of instances where she became silent due to lexical difficulty. On the other hand, in the data from as late as 16 months no instances of codeswitching could be found. She never reverted to using L1 unless the researcher directly asked her to switch to Japanese. Moreover, there was a difference in the manifestations of nonfluency. Naturally, the manifestation of Lily's nonfluency was most evident in an increase
The First Stage of Second Language Attrition
73
in pauses. On the other hand, Eugene's nonfluency became conspicuous in terms of repetitions. The following is a typical example of Eugene's repetition from the spontaneous speech data. (24) 1 E: You can't go to the uhm, you can't go to the uh, you can't go to, the, let's say there's like uhm, a problem, but you can't go to the problem first, because you have to, there's like a story and you have to read a long story, 2 R: Mmhm. 3 E: and then you have to, uhm, answer 4 R: Yeah, the questions? [A04] Differences in the use of compensatory strategies may have come from their differences in personal style. As mentioned above, Lily is a very cautious child. She aims at perfect utterance every time, no matter how long it takes, and there is both hesitation on her part to ask for assistance and a strong inhibition against using L1. In speech performance, this translates to her being a "pauser" and a "noncodeswitcher." On the other hand, Eugene will utter whatever comes to his mind first, however imperfect the sentences may be. This makes him a "repeater," and since there is no strong inhibition on his part against using L1, he codes witches. Lily's sense of inhibition against using L1 combined with her desire to produce perfect utterances and to provide accurate information also manifested itself as frequent employment of the avoidance strategy. Compared with the boys' utterances, Lily's response was abundant with "I don't know" and "I forgot" routines. She often used these expressions to avoid facing the task of verbalizing in L2. The following examples illustrating this are all from the same session dealing with the topic, her favorite things. (25) 1 2 3 4 5
R: L: R: L: R:
What was your favorite cartoon in America? I don't know. You don't know? You didn't watch cartoons at all? Yeah, I did, but I liked them all. You liked them all? What kinds of cartoons did y u watch. Do you remember? 6 L: (6 sec) Bugs Bunny. 7 R: What about Disney ones. Disney cartoons? Donald Duck? Mickey Mouse? 8 L: I forgot? 9 I 10 R: What's your favorite okazu? (= dish). 11 L: (9 sec) I don't know. [A04]
In sum, the overall progression of attrition for the three children was very similar. However, the surface realizations of attrition, their choices of compensatory strategies in particular, were sometimes different, reflecting their personal styles.
74
The Loss of English as a Second Language
Discussion Differential Attrition of Subskills It was observed that different linguistic skills (i.e., phonology, lexicon, morphology, and syntax) were not affected in the same way by the attrition process. For Ken, 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 throughout the course of the observation. Kaufman & Aronoff (1991) report a similar pattern. They studied longitudinally the process of a young child's L1 (Hebrew) attrition in the L2 (English) environment. The subject's productive lexicon was affected initially, but syntax was unaffected. This stage was followed by a disintegration of morphosyntax. On the other hand, Kuhberg (1992) reveals a different pattern in his longitudinal L2 attrition research. From two Turkish children's German attrition data, he found that the morphemic level was affected first. It was only from the second stage onward that the lexicon was negatively affected. The fact that syntax (in terms of word order rules) was intact is the same as in the present study. Studies concerning the attrition of a school-learned foreign language by adults, such as those by Moorcroft & Gardner (1987) (L1 English, L2 French) and Weltens et al. (1989) (L1 Dutch, L2 French) report the same pattern as Kuhberg's study. In her literature review, Yoshitomi (1992) relates the differential order among the subskills to the subjects' initial proficiency level. She summarizes that more proficient students lose more vocabulary than grammar, while the pattern is reversed in less proficient students. This is because advanced students have a relatively stable knowledge of grammar and a larger amount of lexical knowledge. She also adds that in L1 attrition, lexicon is affected more than grammar. This attests to the assumption that the more stable the knowledge, the better it is retained, because native speakers presumably have acquired the grammar of their L1 thoroughly. The initial proficiency level may explain the present results and those from the above-mentioned studies. The proficiency level of Ken was very high; he was on par with his native-speaker peers academically and socially. It can safely be assumed that his achievement in L2 acquisition was far more advanced than that of students who participated in the above-mentioned studies of foreign language attrition. Kaufman & Aronoff's subject, although very young, was "a fluent speaker of Hebrew (L1), whose level paralleled that of her Hebrew speaking peers" (1991:176). On the other hand, Kuhberg's results cannot be sufficiently explained by the difference in proficiency alone. His subjects' proficiency levels seem to be as high as those of the subjects of the present study. According to Kuhberg, their fluency, communicative competence, and syntactic patterns were almost nativelike; yet deviant morphological patterns appeared before difficulty in lexicon. However, equating the proficiency levels of Kuhberg's subjects to the present study's might be a risky assumption. Due to the lack of an adequate L2 proficiency measure for children in each language, none of the subjects involved in
The First Stage of Second Language Attrition
75
these studies was assessed objectively by standardized tests across all skills. Assessments such as "nativelike" and "parallel to native-speaker peers" are subjective and may indeed reflect different levels of proficiency. Other possible factors for explaining the differing results are the influence of the language that is being lost, the influence of the dominant language used in the environment, and the interactive effect between the two languages. Andersen (1983), along with Weltens et al. (1989), have noted that these factors play an important role in the attrition process. The languages involved in Kuhberg's study were Turkish (L1) and German (L2), whereas there was a common language, English, involved in Kaufman & Aronoff's (1991) and the present study. On the other hand, Kaufman & Aronoff's focus was L1 attrition, whereas the present study was focused on L2 attrition. As more cross-linguistic data become available, the influence of these factors on the attrition process will be more precisely discerned. Productive versus Receptive Skills The present result supports the previous findings regarding the lag between productive and receptive skills (Weltens 1989; Yoshitomi 1992). The subjects' productive skill in the lexicon was first to be affected in the process as evidenced by their lexical retrieval difficulty. On the other hand, their receptive skill was virtually unaffected throughout the observation. This is in accordance with studies which report long-term retention of receptive vocabulary, such as those of Weltens et al. (1989), Bahrick (1984), and Cohen (1989). The Speed of Attrition Studies based on longitudinal data are in agreement with respect to the onset of attrition. Kuhberg (1992) reports no substantial change in any areas for five months. Kaufman & Aronoff's study (1991) shows the first sign of attrition at seven months. The present subject also showed little change for seven months. The onset of attrition, therefore, is assumed to be quite similar in L2 attrition and in L1 attrition, occurring at around six months after removal from the dominant language environment. On the other hand, the speed of attrition was more rapid for the subjects in Kuhberg's study than for the present subject. By the 15th month, one of Kuhberg's subjects became ashamed of her L2, which led him to terminate the study. The attrition of Kaufman and Aronoff's subject was also swifter. They report that by the 12th month, the subjects' L1 verbal system started to disintegrate with increasing unwillingness to speak in L1. At 19 months, however, Ken was still speaking spontaneously and willingly with morphology being relatively unaffected despite lexical retrieval difficulty and decreased fluency. It is possible that Ken was at an "initial plateau," that is, "a period of a few years during which skills are relatively unaffected, before attrition actually sets in" (Weltens & Cohen 1989:130). However, this notion is based on data mostly
76
The Loss of English as a Second Language
from adults experiencing attrition of a foreign language, and whether it is applicable to children losing their naturally acquired L2 is subject to further investigation. Differences in the absolute speed of attrition are likely to be attributed to a number of interacting factors such as chronological age, initial proficiency level, amount of use and input of attriting language, motivation, and pressure to conform to the hosting environment, to name a few. Fluency Deterioration of fluency marked by slower speech, more frequent hesitations, pauses, repetitions, and self-repairs is commonly noted as the first sign of attrition in longitudinal studies, including the present study and Kuhberg's (1992). In their study of child L1 (Russian) attrition, Turian & Altenberg (1991), also report that "reduced ability to be quick and easy and expressive" was one of the salient features when initial change was noted. Whether as a result of lexical retrieval difficulty or morphological difficulty, loss of control in any subskill is very likely to affect fluency, which can be achieved only through successful coordination of well-functioning subskills. Compensatory Strategies Strategies for compensating for the affected skills employed by the subject were very similar to the ones reported in L2 attrition research such as Olshtain (1989) and Cohen (1989). In the area of L1 attrition, Turian & Altenberg (1991), also cite similar strategies that are common to the ones observed in L2 acquisition (e.g., Faerch & Kasper 1983). A variety of strategies, including codeswitching (L1 based strategy4, paraphrase, and approximation (L2-based strategy), direct appeal (cooperative strategy) as well as avoidance (functional reduction strategy) appeared in the corpus of the present data. The result of this study, therefore, is to add another piece of evidence to support Turian & Altenberg's conclusion that "the cause of incomplete linguistic knowledge (process of acquisition versus attrition) is not a factor in selecting the strategies that are used to cope with the deficiency" (1991:216). Individual Differences Based on the data from Ken and two other returnees referred to (Eugene and Lily), it can be assumed that the pattern of attrition is similar across individuals. Although literature that alludes to individual differences is scarce, Kuhberg (1992) also emphasizes the similarity of the attrition pattern that his two subjects underwent. A notable difference observed in this study was the choice of compensatory strategies. To accommodate for their lexical retrieval difficulty, the boys resorted
The First Stage of Second Language Attrition
77
mainly to codeswitching, while Lily opted for avoidance. Their choice seems to reflect their personal style and their attitudes toward facing a task. Yoshitomi (see Chapter 4, this volume) cites personality and proficiency level as two factors influencing the choice. An individual difference that Kuhberg (1992) observed in his study was the speed of attrition. The attrition of one of the girls was faster, which Kuhberg attributes to the amount of pressure to give priority to L1 and her willingness to give in to the pressure. In either case, personal variables seemed to play an important role in the entire process of attrition. The results of Yoshitomi (see Chapter 4, this volume) as well as the present results concur with that observation. Affective Factors Previous literature suggests that "very frequent and/or pragmatic-laden items such as closed class vocabulary, idioms/fixed expressions, and interjections and fillers (e.g., urn)" (Yoshitomi, 1992:297) are immune to loss despite lack of use. However, it was noted in this study that switching to L1 began with emotionladen utterances, interjections, and conversational fillers. In her own research, Yoshitomi (see Chapter 4, this volume), also points out that contrary to what previous research claims, these items were indeed susceptible to loss. One possible interpretation of the present results is that these items are not lost or even forgotten but that the subject no longer feels natural expressing those emotions in L2. In other words, as the attrition progresses, speaking in L2 becomes much like a performance on stage, which makes him feel unnatural and takes much energy. Speaking in L1, on the other hand, is an off-stage natural everyday behavior where he can feel relaxed and express his genuine emotions. He takes a break by expressing his emotions and filling pauses in L1. Therefore, it is precisely these emotion-laden utterances that are most appropriate and natural to be expressed in L1. Some evidence to support these on-stage/off-stage acts can be found in the data. Ken's conversational fillers in L1 were often uttered in a whisper or low volume (See example [5]). When he switched to L1 for a direct appeal for a wanted word, he also tended to whisper (see example [8]). Data from the 17th month session revealed that 46% of conversational fillers and 55% of requests for a word were whispered. Ken's altering of speech volume could be interpreted as a reflection of how he perceived the act of speaking in L2 and in L1 at that stage of attrition. Speaking in L2 is an on-stage performance, whereas speaking in L1 is a natural behavior.
Concluding Remarks This study provided a brief account of the overall process of young children's L2 attrition over a period of 19 months. The first stage of attrition could be characterized by loss of fluency and codeswitching caused primarily by difficulty in
78
The Loss of English as a Second Language
lexical retrieval. Some signs of attrition in morphology and syntax were present, but they did not yet show definite attrition patterns. Phonology and receptive lexicon remained stable throughout the stage. The present study also suggested directions for further investigation. Interactive effects among variables on the speed, degree, and manner of attrition should be more clearly discerned. Among the more notable variables are: languages involved, initial proficiency level, amount of use, and personal factors such as motivation, affect, and styles. Although age factor was not specifically addressed in this study, chronological age and literacy, which usually parallel age, are also significant. Classroom-oriented research, namely, application of L2 attrition findings to language maintenance programs and its effect, needs to be launched. Although limitations of case studies should be recognized, longitudinal research is valuable in documenting the details over time as well as providing previously unacknowledged phenomena. It is also advantageous particularly for attrition research in which nonacquisition and attrition must be clearly distinguished. Further attrition research would benefit from both qualitative and quantitative types of data. From a larger perspective, attrition studies, in line with L1 and L2 acquisition, should continue to be recognized as a field that will help us better understand how the human mind deals with language.
The First Stage of Second Language Attrition
79
Notes 1. The stages were divided, in retrospect, according to Ken's characteristic attrition pattern of lexicon, morphology, and syntax. 2. The following symbols are used in the transcription: ? I (xsec) (xxx) < > [=] Underlining
downward intonation upward intonation cut off conversational turns omitted pause in seconds pause less than one second transcriber unsure of words uttered explanation of situation English translation is used for emphasis in Japanese text.
3. These items were "seemingly" lost since some were recovered later at Stage 2. See Tomiyama (1995) for the details of Stage 2. 4. The categorization of strategies given in parentheses are from Faerch& Kasper (1983).
4 On the Loss of English as a Second Language by Japanese Returnee Children Asako Yoshitomi
In the last two decades, the number of Japanese families going abroad to English-speaking countries on business has increased dramatically as a result of the growth in the country's economy. Consequently, the number of children in such families who spend several years during their youth exposed to a second language (L2 English), acquiring the L2 fairly naturally, and later suffering L2 attrition after their return to Japan, has increased considerably. There is thus a rising need in Japan to develop language maintenance programs for both the first language (L1 Japanese) and second language (L2 English) of such children. Knowledge about the language attrition process itself, upon which such programs should be based, is insufficient at present. Although previous research in the field has provided tentative answers to some key issues in L2 attrition (see Chapter 1, this volume), it is apparent that qualitative case studies that examine the language attrition process from a more holistic perspective are lacking; that is, studies that examine not only the linguistic aspects of language loss but also the factors that affect the process, such as differences in personality and aptitude as well as various developmental, psychological, cognitive, sociological, and sociopsychological variables. This chapter reports some of the key findings of a case study that was conducted as research for a dissertation.'The case study was a preliminary attempt to take such a holistic perspective by carefully examining the linguistic and sociopsychological attributes of the process of L2 (English) attrition of four Japanese returnee children from English-speaking countries. Because no complete theory or model is currently available to guide researchers to "test" specific questions, a hypothesis-generating type of approach was taken, in which the nature of the data served to guide analysis.2 The purpose of this chapter is twofold: (1) to verify whether some of the main findings in the field can be supported by the data obtained in the case study 80
Loss of English by Japanese Returnee Children
81
on Japanese returnees; and (2) to discuss the pedagogical implications the results may have on improving language maintenance programs in Japan.
Method Subjects Speech data from four female Japanese returnee children were examined.The girls differ in age (i.e., older versus younger) and incubation period (shorter versus longer) at the time of initial data collection. A description of the subjects' age at initial data collection, length of stay in the United States, length of incubation period, and siblings are given in Table 4.1 in the Appendix. Henceforth these subjects are referred to as JYS (Japanese girl, younger group, shorter incubation), JYL (Japanese girl, younger group, longer incubation), JOS (Japanese girl, older group, shorter incubation), and JOL (Japanese girl, older group, longer incubation), respectively. The Japanese returnee children come from families of middle to upper-middle social class. They all lived in large metropolitan areas in America. Their fathers, businessmen in a relatively large corporation, are well educated. All mothers, except for JOL's, were graduated from a university. JOL's mother finished high school. Family size is relatively small, varying from three to five. The Japanese children and their parents were all aware that they would return to Japan after a stay of two to five years in America. Hence, all children went to local American schools on weekdays and attended supplementary Japanese schools on Saturdays. After returning to Japan, they started to go to Englishmaintenance classes on Saturdays, as do the majority of returnee children in Japan. In addition, JYS and JOS attended English classes for returnees at their regular Japanese elementary schools once a week for one hour. The children might choose to use English in the family as well, although infrequently. Despite the children's willingness to maintain their English skills, their opportunities to use English are fairly limited. They will not start receiving formal English education until they enter junior high school. Obviously, the availability of subjects severely restrict the possible number of individual cases that could be included in a single study.3 This study limits itself to the analysis of four returnees for a span of one year, who were chosen because they met the following criteria regarding sex, age, length of stay, and length of incubation period. Sex. To avoid possible differences attributable to sex, all subjects were female. Age. The selection of children under age twelve was made to avoid the inclusion of an additional variable that is introduced by the start of formal English education in Japanese junior high schools. Length of stay. The subjects were children whose lengths of stay in the United States were three to four years. Nomoto (1973; cited in Minoura 1981) found that for most children, English becomes the dominant language after two to four years of stay.
82
The Loss of English as a Second Language
Incubation period. The subjects' incubation periods were either zero months or longer than 12 months. Previous studies imply an initial "plateau" of language loss. In other words, L2 acquirers, especially highly proficient acquirers, resist attrition for a while even after they discontinue their use of English. Because the study was to be conducted for only one year, linguistic samples beyond one year of incubation could be collected only in this manner. Ma
ials and Procedures for Data Collection
Data collection consisted of several different observation procedures in order to collect data on different aspects of language use and to decrease the bias that might result from the application of a single task. The five tasks that were adopted are described below: Free interaction. The core of the data are the free speech samples collected through a free conversation between the subject and a native-English-speaking examiner. The examiner initiated the conversation by laying out semiplanned topics that were believed to be relevant and interesting to the child (e.g., school and friends; family; English and experience with English; the future or imaginary situations); the child was free to expand or to move to other topics of her interest. Story description. Following the free interaction, each subject was asked to describe stories by looking at picture books. All the picture books used include pictures with a clear story line but no written text. The subjects described the same two picture books (Sunshine and Frog, Where Are You?) every session and a different third book in session two (A Boy, A Dog, A Frog, and A Friend), session three (Frog Goes to Dinner), and session four (One Frog Too Many). This method was adopted so that the subjects' linguistic performance on the same two stories could be compared across sessions and so that possible practice effects resulting from describing the same stories repeatedly could be complemented by examining the subjects' performance on a new story line each session. The child looked through the whole book once, then went back to the first page of the book and described the story from the beginning at her own pace. When she was finished with one book, she went on to the second book and followed the same procedures. Planned speech. After the story description, the subject was asked to think about a topic that was discussed during the free interaction, expand on that topic, and plan a short speech about it. She was given as much time as she wanted to prepare the speech and was allowed to make notes and rehearse the speech if she wished. Listening comprehension. As the final English task in each session, a controlled listening comprehension test was given, using Comprehensive Language Evaluation (CYCLE) developed by Curtiss & Yamada, specially made to measure children's formal linguistic knowledge. The children's task is to select a picture out of a set that corresponds to the sentence they hear. The sentences differ syntactically in that the only difference between two sentences, for example, may be a single bound morpheme such as the plural noun marker -5.
Loss of English by Japanese Returnee Children
83
Interview and questionnaire. Finally, interviews were conducted in Japanese between the researcher and the child, and then between the researcher and the parents, respectively. The aim was to obtain supplementary information on socioeconomic, educational, sociopsychological (i.e., motivational and attitudinal), and personality factors both from the returnees themselves and from their parents as second-party observers. Parental interviews were based on the replies they gave on a questionnaire they filled out while the data were collected. Questions included those that asked the parents to rate their children's English abilities as well as their own, to describe their child's personality, and to report on how frequently their child uses English and what kind of attitude they and their child have toward the use and maintenance of the English language. The children were also interviewed in Japanese and asked to self-evaluate their English language skills and to answer questions concerning their L2 use in general, their attitudes toward maintaining and using English, and their motivational intensity as to how much they are willing to retain their English skills. All tasks except for the interviews were conducted by a female nativeEnglish speaker in English. Each session was approximately 150-180 minutes long and was videotaped. The interviews with the children and their parents were administered by the researcher in Japanese at the end of each data-collecting session. Data were collected every three months for one year, resulting in a total of four data sets for each subject. The first data sets obtained from the two returnees with shorter incubation periods (i.e., JYS and JOS) were collected within one month after their return to Japan. Hence these sets of data were considered baseline data, where the attrition process started. The other sets of data obtained from the two returnees with shorter incubation periods and all data obtained from the two returnees with longer incubation periods (i.e., JYL and JOL) were thus compared with the baseline data.4
Data Analysis All speech samples collected through production tasks in English (i.e., free interaction, story description, and planned speech) were transcribed and analyzed by the use of the Codes for the Human Analysis of Transcript (CHAT) transcription and coding format and Computerized Language Analysis package of analysis programs developed by Brian MacWhinney and his colleagues (MacWhinney 1991). The returnees' English speech data were analyzed in terms of accuracy and fluency. Accuracy was assessed in terms of measurements (A) through (H), fluency was assessed in terms of measurements listed in (I), and supplementary data for interpreting the speech data were collected through measurement (J), respectively listed below: A. Scores of CYCLE and comprehension errors or difficulties (defined here either as the difficulty in comprehension, resulting in a query or a misunderstanding of what the examiner said) during free interaction.
84
The Loss of English as a Second Language
B. Production errors in phonology during free interaction and story description. C. Use of verb morphology during free interaction and story description. (The verb morphology examined included copula be, auxiliary be, simple verbs in their base form, the third person singular -5 marker, past regular and irregular forms of simple verbs, and modal auxiliaries. The percentage of contexts in which verb morphology was supplied in obligatory contexts (%SOC) and the percentage of verb morphology used in a targetlike manner (%TLU) were calculated across the four data sets. %SOC was calculated as the percentage of correct verb morphology supplied in linguistic contexts in which the verb morphology is required. %TLU was calculated as the percentage of correct forms used among the total verb morphology). D. Use of definite and indefinite articles during story description and planned speech. (The %SOC and %TLU were calculated for the definite and indefinite articles, respectively.) E. Type-token ratio of total vocabulary and open-versus closed-class vocabulary items used during free interaction and planned speech. F. Types and frequency of communication strategies used during free interaction and story description. (The terminology used to discuss the use of communication strategies was based on Tarone's [1980] classification.) G. Use of complex syntactic structures during free interaction and story description. (The frequency and correctness of complex structures, the percentage of attempted complex structures among total clauses uttered, and the percentage of successful end products as a result of the attempts were calculated. A complex structure is defined as two or more clauses in which one or more of the clauses is embedded in the main clause. Examples of complex structures are clauses embedded through the use of relative pronouns and subordinate conjunctions.The passive construction is also regarded as a complex structure. Structures including subordinate clauses introduced by "I think," "He said," and the like were excluded because such embedding involves a mere juxtaposition of simple clauses. Complex structures were considered deviant whenever an error was included in any one of the clauses constituting the complex structure. Only the end products were examined.) H. Frequency of error-free clauses during story description and planned speech. (Errors included all types; that is, grammatical, semantic, and lexical errors. If a clause included any type of error, it was regarded as an error clause. Only the end products were examined.) I. Amount of speech per time unit, calculated as the average number of words uttered per minute, and the frequency of false starts and repairs, calculated as the average number of clauses in which one repair, false start, or filler occurred (i.e., a repair or false start or filler occurred once in every X number of clauses) during the story description task.5 (It was considered that the decrease in the amount of speech per time unit and/or the increase of repairs, false starts, or fillers used was an indication of less fluency.6)
Loss of English by Japanese Returnee Children
85
J. Replies obtained from the interview and questionnaire. (This set of data
was used as supplementary information to guide interpretations of the speech data and to examine aspects of the attrition process that were difficult to tap through an analysis of speech data from English tasks alone. The children's and their parents' replies from the interview and questionnaire were tallied. These results were taken into consideration whenever they seemed to provide useful clues to an interpretation of the children's linguistic performance. The results of the analyzes were compared across subjects and across tasks wherever possible).
Summary of Results In this section, the results obtained from measurements (A) through (J) above are briefly summarized. Analysis of Accuracy (A) Scores of the listening comprehension task did not render evidence of attrition (see Appendix, Table 4.2). Although JOS's scores on the CYCLE test are slightly better than those of the other three girls, the difference is not considerable.7 If there is any change over time at all, it is the improvement of performance observed in JOL, JOS, and JYL. At the same time, there were indications that CYCLE was not discriminative enough to measure the returnees' loss in listening comprehension skills. Such an interpretation finds support in the fact that difficulties and mistakes in comprehension were more apparent in the subjects' performance in the other linguistic tasks, especially in the free interaction. The number of such comprehension errors during free interaction were, nonetheless, rare. Even JOL, who seemed to have lost English relatively severely compared with the other returnees, did not show so many difficulties in understanding English. When she did have difficulties, they appeared to be due more to local errors (i.e., lack or loss of a particular vocabulary item) than to global errors (i.e., the inability to comprehend the sentence structure as a whole). This result implies that English listening skills are relatively immune to attrition at least for the first year of incubation, which is in accordance with the claims made in the literature that comprehension skills are generally maintained better than production skills, and that there is an initial plateau of regression. (B) In general, production errors in phonology occurred infrequently in all four subjects' speech (see Appendix, Table 4.3). There may be a very slight decline in productive phonological skill within the first year of incubation, which is seen in JYS's case, but the level of skill stabilizes after the initial regression and stays relatively intact. At least 8 among the total of 16 errors observed in the subjects' speech appeared to be clear cases of phonological
86
The Loss of English as a Second Language
errors. All of these errors seem to be influenced by the returnees' L1. Six involved difficulty in distinguishing a number of minimal pairs that are typically considered difficult for Japanese learners of English, including [s] versus [ ], [z] versus [ ], [1] versus [r], and [b] versus [v]. The remaining two involved words which, in Japanese, are loan words from English. These words were pronounced by JOL according to the Japanese accent and mora system. (C) Only a slight increase in errors in the returnees' use of verb morphology was observed over the sessions during both the free interaction and the story description (see Tables 4.4 to 4.7 in the Appendix). The result may be due to the fact that the data for each morpheme were somewhat limited. Nonetheless, possible indications of attrition were found in all four returnees' spontaneous speech samples, which involved an increase in the number of verbs used in their base forms without obligatory morphological markers. The increase in errors was generally more noticeable with subjects with longer incubation periods (i.e., JOL and JYL) than with those with shorter incubation periods (i.e., JOS and JYS), suggesting that the attrition rate seems faster after an incubation period of one or more years. Although the result suggests that syntactic skills may be more vulnerable to loss than phonological skills, indications of English attrition in verb morphology were generally not considerable. (D) Although slightly more errors in the use of definite and indefinite articles in story description and planned speech were observed toward the later sessions, all four returnees seem to have acquired the English article system well and retained it fairly well (see Appendix, Table 4.8). As in the case of verb morphology, it was found that attrition was more clearly evidenced by the difference between returnees with shorter and longer incubation periods than by the difference observed over time. In other words, if there were any decline in the returnees' ability to use articles at all, it occurred before the data collection started while the returnees (i.e., JYL and JOL) virtually had no opportunities to interact with a native speaker on an individual basis. (E) Type-token ratio of open versus closed class vocabulary used during free interaction, and planned speech did not indicate clear evidence of attrition (see Appendix, Table 4.9). Although the ratio of open-class vocabulary items in the speech of the returnees with longer incubation periods tended to decrease slightly, the overall use of the vocabulary items did not provide any clear evidence of attrition. Differences were observed more between the returnees with shorter versus longer incubation periods than over time, suggesting again that the returnees were able to maintain vocabulary more or less over the data collection sessions during which they had the chance to interact with a native speaker of English. Lexical errors in terms of word choice were occasionally observed in the returnees' speech. Most frequent were lexical errors where short, simple verbs were overgeneralized to cover a wide variety of verb meanings, and where incorrect word choice was made either in terms of verb transitivity or parts of speech. Incorrect selection of anaphora was also observed. More Japanese expressions (mainly nouns and interjections) were found to be mixed in the free interaction toward the latter data-collecting sessions.
Loss of English by Japanese Returnee Children
87
(F) There were some noticeable differences in the use of communication strategies due to levels of English proficiency (see Appendix, Tables 4.10 and 4.11). JOS, the returnee who seemed to retain the highest English proficiency, had little need to rely on communication strategies. The other returnees increasingly made use of communication strategies over time, reflecting their decreasing accessibility to lexical items. In general, the less proficient, the more the use of communication strategies was observed. The returnees increasingly had to use paraphrase, namely approximation, word coinage, and circumlocution or exhibit progressive retrieval in order to convey their messages. They began to mix more Japanese in spontaneous speech over time as well. As attrition became severe, as in JOL's case, different types of communication strategies, such as the use of the L1, nonverbal appeals for assistance, explicit requests for help, and avoidance became dominant. Apparently, with further attrition, the ability to compensate for lost vocabulary items starts to erode as well. (G) The percentage of attempts to use a complex structure among total clauses increased with JOS and JYL, decreased with JYS, and stayed more or less the same with JOL (see Appendix, Table 4.12). While JOS and JYL attempted more frequently to use complex structures over time, their success rate tended to decrease. A clear decline in the success rate was observed with JYS as well. In contrast, the success rate of JOL's attempts to use complex structures increased over time, but the frequency of her attempts was low compared with that of the other returnees. The result suggests that in terms of the ability to use complex structures in spontaneous speech, all the returnees except for JOL regressed over time. JOL perhaps had already declined in her ability to use complex structures before the data collection started, and she actually benefited from the data collection by having the opportunity to use knowledge that was becoming difficult to access. This measure revealed a clearer case of regression in the returnees' syntactic skill than did the assessment of their ability to use verb morphology or articles. It should be noted that the measurement of embedded clauses differs from the other two in that the assessment of the returnees' ability to use verb morphology and articles involve only one subsystem of their syntactic skill. In the assessment of their ability to use complex structures, however, the returnees' ability to use various subsystems, including the use of subordinate conjunctions, relative pronouns, passive constructions, and other grammatical structures is combined. (H) Frequency of error-free clauses during story description and planned speech revealed more evidence of loss. All the subjects declined to some degree in their accuracy in using English in that the percentage of error-free clauses among the total number of clauses uttered generally decreased over time (see Appendix, Table 4.13). The decline was larger with returnees with longer incubation periods. Especially with JOL, almost every clause included some kind of grammatical error in the story description task. As already observed in the other measurements of accuracy, such as the use of verb morphology and articles, the difference between the returnees with shorter incubation periods and those with longer incubation periods was larger than the difference across the four data collecting sessions.
88
The Loss of English as a Second Language Analysis of Fluency
(I) The amount of speech per time unit decreased only in JYS's speech. In the other three returnees' speech, the amount of speech per time unit either stayed approximately constant or even increased, as in JOL's and JYL's cases (see Appendix, Table 4.14). In terms of the amount of speech per time unit, therefore, no attrition was observed. When the amount of speech per time unit was compared among the returnees, however, the returnees with shorter incubation periods again spoke more than the returnees with longer incubation periods. The average number of clauses per filler exhibited no attrition, but rather showed improvement over time. In JOS's and JYL's cases in particular, the average number of clauses per filler increased considerably. This, together with the finding that the amount of speech per time unit increased in JYL's speech, implies that not only did JYL speak faster, but she also used fillers much less frequently in her speech over time. The only measurements that indicated possible signs of attrition were the average number of clauses per repair and the average number of clauses per false start. The average number of clauses per repair decreased for JOS and JYS, whereas the average number of clauses per false start decreased for JOL and JYS. In other words, all the returnees except for JYL exhibited a small degree of regression in fluency, in that the frequency of repairs or false starts in their speech increased over time. While the returnees produced approximately the same amount of speech over the sessions, the content of their speech came to include an increasing number of repairs or false starts. As for JYL, all four measures of fluency indicate improvement of performance, contrary to the prediction that all returnees would exhibit some indication of decreased fluency in English. As will be mentioned in result (J) below, JYL's personality and attitude may have had positive effects on her maintenance (or even improvement) of her fluency in English. Replies to the Interview and Questionnaire
(J) Replies to the interview and questionnaire showed that all returnees wish to maintain their English skills more or less but self-evaluate loss especially in vocabulary. Despite some loss in vocabulary, the returnees reported using paraphrase or circumlocution to compensate for such lack of ability. The use of such strategy may camouflage lexical loss to a certain extent. There were differences in the returnees' use of vocabulary due to the returnees' personalities, in that the more talkative returnees produced more types and tokens of vocabulary items and seemed to benefit more from planning. Personality seems to be involved in the use of communication strategies as well. The more talkative returnees used paraphrase and progressive retrieval much more frequently than transfer or avoidance, whereas the less talkative returnees resorted more to transfer and avoidance than paraphrase and progressive retrieval.
Loss of English by Japanese Returnee Children
89
JYL is a risk taker who used the largest number and the widest variety of communication strategies most frequently among the four subjects. She seems to have high communicative competence, being able to convey her intended messages effectively despite considerable regression in accuracy over time. She is talkative and fluent and shows the strongest motivation to retain her L2. Her attitude toward the data-collecting sessions was highly positive. JOL, on the other hand, seems to be a non-risk taker and aborted messages even when she had little need to resort to strategies to express her ideas. She is social, but her command of L2 has declined to the extent that her ability to communicate effectively has become severely limited. Her attitude toward data collection was positive but not particularly active. JYS is talkative and fluent. JYS, together with JYL, offered a large amount of speech data in most tasks. JYS in general seems more proficient than JYL, possibly reflecting the difference in the length of incubation periods. JYS's personality seems to resemble JOS's in terms of extroversion, strong motivation, and favorable attitude toward the data-collecting experience. JOS is the fluent but quiet girl who seems to take few risks if any. In general, JOS's speech data were usually smaller in quantity, making it difficult to interpret her language ability. The questionnaire reveals that she is less outgoing than the other three returnees. She was also the only subject who was unwilling to experience the same kind of data-collecting sessions she experienced in this study in the future. This, together with the fact that the limited speech she offered was generally more normative than the other returnees, suggests that the quantity of JOS's speech may be limited, not because of lack of ability or attrition, but because of her quiet personality and rather negative attitude toward the data-collecting experience. It seems there is not one type of personality in particular that is strongly linked to better retention of English proficiency. It does seem that extroverts with positive attitude toward the situation of language use are better in communication and are willing to make the most out of the opportunities to use English. Such willingness, in turn, appears to have positive effects on language retention. According to the replies to the questionnaire, the returnees parents attribute the biggest cause of their children's loss in English skills to the disuse or rare use of English in everyday life. All returnees reported that they scarcely use any English in their daily lives other than at language maintenance schools or during supplementary English activities at school, adding up to a maximum of three hours of contact with English per week, among which less than thirty minutes is spent on one-to-one interaction in English. Parents support their children's wish to maintain English, and were glad that the data collection sessions offered their children an opportunity to use English. Despite such positive attitude towards English use and retention, they mostly do not tell their children to make any specific kind of effort. In general, the returnees express their willingness to retain their English abilities, but seem to make little "real" effort other than reading English books once in a while.
90
The Loss of English as a Second Language
Discussion Since limit in space does not allow for a detailed discussion of all the results summarized above, I will restrict myself to discussing what I consider the most important findings of the case study. Put very simply, the results can be further reduced to the following three points: (A) The returnees exhibited little language attrition over the data collection sessions. Although high retention in phonological skills was predicted, the retention observed in the returnees' ability to use other language subskills in production, including verb morphology, articles, and lexicon was better than expected. (B) Indications of language attrition were more evident in the regression in the returnees' ability to combine the language subskills. The returnees made more errors in their use of complex structures and produced less error-free clauses over the sessions. More subtle indications of loss were exhibited in terms of their decrease in fluency and the change in their use of communication strategies. (C) 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. The returnees with longer incubation periods seemed to have lost more during the period before data collection than over the data collection sessions despite the overall- increase in incubation period. Similarly, the returnees with shorter incubation periods seemed to have lost less over the data-collecting sessions than the returnees with longer incubation periods did during their first year of incubation. Assuming that JOL and JYL had roughly the same level of English proficiency as JOS and JYS when they left the United States, we can say that JOS and JYS did not lose as much of their English proficiency as JOL and JYL did during the first year of their incubation periods. There are at least three possible reasons for this finding. One is that the assumption itself is wrong; that is, JOL and JYL were never as proficient as JOS and JYS: They started out at a lower level. The second possible reason is that it would take a few more months for JOS's and JYS's English to erode to a level equivalent to JOL's and JYL's English at session one. Data collection started when the incubation periods of JOL and JYL were 15 months and 13 months, respectively. In the cases of JOS and JYS, less than one full year had passed since their return to Japan at the time of session four. Hence attrition might have been more evident if data were collected from the returnees for a longer period of time. Perhaps JOS's and JYS's English would have eventually eroded to the level equivalent to JOL's or JYL's. The third possible reason is that the very act of collecting English speech data from these returnees gave them the opportunity to use English, and therefore helped them to maintain their English skills to a certain degree. Although all four returnees attend English maintenance schools on Saturdays, they rarely have a chance to speak with a native speaker on a one-to-one basis in their daily lives. Hence the chance provided by the data collection, although only once every three months, may have affected the rate or amount of language attrition. This may account for the fact
Loss of English by Japanese Returnee Children
91
that JOS and JYS did not lose as much of their English proficiency as JOL and JYL within the one-year incubation period. The third reason is the most probable. This I base on my experience of having dealt with more than 100 returnee children in the JOSEF project (See Nakazawa [1989 a,b]; Nakazawa & Yoshitomi [1990]; Seya [1990]; Yoshida & Arai [1989]; and Yoshida et al. [1990] for detail), knowing more than 60 friends who are returnees, having taught nearly 300 university students who are returnees, and being a returnee myself. Three of four returnees reported that they became more motivated to maintain their English after data collection started. Especially for the two returnees with longer incubation periods, the initial data collection made them realize how much English proficiency they had lost. Although they thought they had a basic idea of how much English had already been lost during the incubation periods, by talking to a native speaker they discovered that they had lost more than they had initially thought. Such a "shock" may have triggered them to think more seriously about making efforts to maintain their English. This reaction appeared to apply especially in the case of JYL, who started reading more books in English and joined an English pen pal club after the data collection started. In fact, JYL's loss is relatively small, given the fact that her incubation period was more than 13 months at session one. Indications of positive consequences of the data collection process were also observed in the slight improvement of productive phonology by JOL, the increase in JOL's success rate in attempting to use complex structures, the increase of the frequency of attempts to use complex structures observed with JOS and JYL, and some improvement in fluency measures seen in the form of increase in the amount of speech per time in JOL and JYL's cases, and the decrease of frequency in using fillers in JOS and JYL's cases. Again, my interpretation is that the difference between the returnees with shorter incubation periods and those with longer incubation periods represents the "real" attrition that possibly occurs within approximately one year of incubation, and it appears that the relatively small amount of regression or improvement unexpectedly exhibited by the returnees was a consequence of the practice offered by the datacollection process itself. What is lost, then? What should be noted is that when the returnees' lexical skills or particular subsystems of syntactic skills, such as their use of verb morphology and articles, are examined, attrition is exhibited only slightly; but when error rates are calculated all together, attrition becomes much more evident. This finding suggests that although the regression in the individual subskills of English is not considerable, the small degree of regression in various parts of their linguistic skills has a cumulative effect on the returnees' overall linguistic performance. As a result, their increased difficulty in producing in English only becomes evident when they face a situation in which they must coordinate all their subskills simultaneously and spontaneously. Such difficulty can be "measured" only when their overall linguistic performance is examined rather than their ability to use individual skills or subsystems. Perhaps what is being lost initially is their ability to coordinate all the linguistic skills rather than their ability to use the linguistic elements within a particular linguistic subskill.
92
The Loss of English as a Second Language
Such finding suggests a potential danger in measuring the returnees' subskills as a means of assessing their overall linguistic proficiency; by doing so, the general regression is overlooked, and thus the returnees are judged to have lost very little of their English proficiency, when actually their linguistic knowledge is gradually but surely eroding. The fact that most returnees seem optimistic in their estimation of their English loss may also be due to the same reason. When asked to work on a listening comprehension task, they perform well. When their grammatical performance is assessed, they show little regression. When their vocabulary skill is investigated, only a limited degree of decreased accessibility can be detected. Both the examiner and the returnees conclude from the returnees' performance that attrition is not severe. But by the time their attrition can be observed clearly in terms of measurements on individual sub-skills, the returnees have lost a considerable amount of English proficiency. It should also be noticed that much of what may be affected first in the process of attrition is not easily assessable in terms of common linguistic measures. For example, the returnees may not be able to remember a specific lexical item in a given situation. They might resort to a communication strategy such as paraphrase or they may avoid using the item, in which cases the returnees' strategy is not necessarily detectable, especially when the degree of attrition is slight and the returnees are able to use the strategies effectively. The returnees may pause more than they used to, but their interlocutor may judge that they are merely thinking of what to say and not how to say something. The returnees may also decide to say only what they are able to say and not risk attempting to say other things. As a result, their speech production will render few errors and few pauses, and they are judged as being quiet or shy, rather than having lost the ability or confidence to use English. Their decreased confidence may show up in terms of sentence stress and intonation, but not in terms of the actual linguistic forms used in their speech. In all the examples given above, the returnees are the only ones who are aware, at least subconsciously, of their attrition. That such a phenomenon commonly exists is evident by the fact that many returnees, especially adults, report a sense of insecurity or lack of confidence in using English when native English speakers who know them say their English is "perfect." The returnees' self-evaluation is often considered a reflection of their modesty, when in fact what they are reporting is their attrition. Lack of fluency in turn discourages the returnees from forcing themselves to practice for fear of feeling embarrassed by making mistakes or pausing in search of an expression that they know they used to have in their linguistic knowledge but which has become inaccessible or difficult to access. Especially in a linguistic environment where the returnees almost never have the need to use English, the effect of such fear could virtually deprive the returnees of what few opportunities remain in their daily lives to use English. Eventually, decreased accessibility would result in complete loss. It is only at that point or at the point at which accessibility has become noticeably low that the regression shows up in terms of a linguistic measurement of English subskills.
Loss of English by Japanese Returnee Children
93
Conclusion In conclusion, I would like to return to the research topics raised at the beginning of this chapter, namely: (1) Can the main findings in the field be supported by the data obtained in the case study on Japanese returnees; and (2) What are the pedagogical implications of the case study on improving language maintenance programs in Japan? In order to provide the answer for the first question, I need to clearly state my beliefs as to how language attrition takes place. As stated in Yoshitomi (1992), I support the concept that language knowledge is regarded as consisting of the connections and associations of information (Schumann 1990; Jacobs & Schumann 1992). According to this view, a regression in one part of the connection or association would definitely affect other parts. In this sense, attrition is least likely to take place in an isolated portion of knowledge. Particular groups or subsystems of connections certainly may be affected more, or earlier, than others, but language is a multimodal system in which no one subsystem can be used in total disconnection with others. This is why it is virtually impossible to single out a particular subskill of language and examine it, much less regard the result of a measurement assessing particular subskills to reflect the actual language proficiency of the language learners. I consider the result of the present case study to be insightful in that it precisely illuminates the point that language attrition initially occurs as accumulative effects of erosion in various subsystems of the returnees' linguistic knowledge. The study found that it is only through an examination of the speakers' ability to utilize and to coordinate their linguistic knowledge fully and to produce the most effective and expressive linguistic forms and structures in spontaneous speech that their regression is likely to be accurately measured, especially in the early stages of attrition. These initial stages, in which erosion reveals itself only in the form of accumulative defects, are commonly overlooked in language attrition studies. I believe, however, that it is precisely the coordinative language ability that should be measured if one wishes to accurately assess the language proficiency of language learners and attriters. Unfortunately, a reliable measure is yet to be developed. In light of this view toward language attrition and based on the result obtained in the present case study, I maintain that previous studies offering some answers to the questions raised in the field (see the introductory chapter of this volume) call for reinterpretation. Here I would like to focus on four claims made in the field; namely, the loss of one type of skill over another, the regression hypothesis, the existence of an initial plateau, and the occurrence of residual learning. First, previous findings in the field suggest that loss in production skills occurs faster and to a greater extent than in receptive skills. What the literature shows, however, may be that when the returnees' English proficiency does erode to the extent that it is measurable, it is in the assessments of production skills that the regression first becomes more evident. Most probably, productive skills are more vulnerable than receptive skills. What must not be overlooked, however,
94
The Loss of English as a Second Language
is that the receptive skills, although not necessarily measurable in terms of individual assessments, erode simultaneously. Perhaps defects in productive skills are merely easier to detect than defects in receptive skills. Obviously, production calls for much more planning and coordination on the speaker's side than does comprehension, so that breakdowns in linguistic abilities are more likely to become apparent in the measurement of productive skills than in that of receptive skills. Linguistic skills are interrelated, and particular skills will unlikely erode in total isolation. As attrition progresses, regression in receptive skills will most likely accelerate the erosion of productive skills. Similarly, reports have been made that lower-level skills seem to be more vulnerable than global skills. Global skills, however, result from the cumulative conduct of lower-level skills and thus cannot be treated as totally isolated from lower-level skills. Naturally, when there is only slight regression in a particular lower-level skill, the defect is unlikely to be discovered in terms of global skills. But the fact that the regression is hard to measure does not mean that attrition is not taking place. Claims have also been made that vocabulary tends to be lost more than grammar in the cases of highly proficient, naturalistic L2 acquirers. Such a claim is problematic because losing one grammatical rule and one vocabulary item certainly cannot be regarded as the same "amount" of loss, so a simple comparison is meaningless. Furthermore, the process of attrition is gradual, from decreased accessibility to total loss, so that it is difficult, if not impossible, to determine at any given time that a certain grammatical rule or vocabulary has totally eroded from the learner's linguistic repertoire. Speakers always have the choice to abort, using an uncertain rule or word and instead use different expressions to convey their messages. As long as they use "safe" rules and words, they will be regarded as having maintained the language. Second, intuitively appealing ideas such as the regression hypothesis should be reinterpreted. To the extent that many aspects of language knowledge become affected by the regression of certain portions of knowledge, attrition can be regarded as taking place simultaneously and uniformly. To the extent that a certain regression triggers another, attrition can be regarded as having some kind of order. One would be tempted to say that whenever learners fail to use a linguistic form that is considered difficult from the researcher's point of view, that which is learned later is lost first. The linguistic knowledge represented in the learner's brain, however, may simply not be in the form of common grammar rules or syntactic descriptions thought up by linguists. In other words, regression may very well take place in reverse order of acquisition, but the order may not be in the form of digital stages, consisting of prescribed linguistic rules, but in the form of analog stages, consisting of the weakening and eventually disappearing neuronal connections that each carry simple and overlapping linguistic knowledge. Third, the notion of initial plateau must be regarded with care as well. Although returnees do seem to lose less in the initial stages of attrition than later on, it is highly unlikely that they would exhibit total resistance to loss, even during the very first period of L2 disuse. Most possibly, their attrition involves the aspects
Loss of English by Japanese Returnee Children
95
of language use that are difficult to assess through common linguistic measurements. As a result, the returnees may seem superficially immune to loss at the beginning. My belief, however, is that there is no such thing as a real initial plateau: Attrition sets in as soon as language use is discontinued, but initially only in the form of erosion in the ability to coordinate various linguistic subsystems simultaneously and spontaneously. Finally, residual learning, by the same token, may be a mere consequence of a particular linguistic element being overlooked in initial measurements but being successfully assessed in later assessments. The fact that attrition starts in the form of decreased accessibility conflates the problem, because it is always possible that a certain task or measurement would give the learners an opportunity to reactivate their eroding association, thus providing a chance to perform as if they have learned something residually. Regarding the second research question raised in this chapter, I would like to stress that the use of English in an interactive situation, in which the returnees are forced to speak spontaneously and to coordinate their linguistic subskills, is crucial in maintaining English proficiency. Unfortunately, such opportunity seems to be seldom provided in language maintenance programs offered in Japan at present. It is usually the returnees themselves who first realize how serious the degree of their language attrition has become, at the point where a common linguistic measurement of particular linguistic subskills would not overtly detect much regression. What happens then is that the returnees increasingly lose confidence in their use of English, while their parents or teachers notice little loss. Less confidence presumably results in fewer attempts to use or to interact in English. This creates the vicious cycle of further decreasing the returnees' opportunities to use English. With less practice, the returnees gradually lose fluency, making more false starts and repairs. Hence, it is important to regard the slightest decrease in confidence and fluency on the part of the returnees as indications of language loss. It is at this point that returnees should definitely be given an opportunity to interact intensively with a native speaker on a one-to-one basis. I believe that even a couple hours of such opportunity per month would be much more effective in maintaining, or slowing the attrition of, the returnees' L2 than merely making them participate in nondemanding classroom activities in a group for two hours every week. At present, maintenance classes seem to offer activities involving only local skills or individual subskills. Few activities seem to consist of interactive production exercises which involve the use of global language skills. Some activities are communicative in nature, yet they are not made to force the returnees to utilize their language skills to the fullest, much less encourage them to extend or to stretch their abilities. In essence, the maintenance programs lack activities involving comprehensible input i + 1 as well as comprehensible output i + 1, (Swain 1985, 1995, 1998). It is important for the teachers and the parents to become aware that the returnees' L2 starts to regress immediately after they return to Japan. Underestimation of their degree of L2 attrition has a negative effect on their language
96
The Loss of English as a Second Language
maintenance, since unawareness of attrition may deprive the returnees of their incentive to make the most out of their limited opportunities to use English despite their positive attitude and motivation to maintain the L2. Especially in the case of children, it would be essential for the teachers and parents to acknowledge their language loss, and provide them with more effective exercises. With the development of technology, it may also be feasible to set up a correspondence network between the Japanese returnees in Japan and the nativespeaking children in the United States so that the returnees will be offered the chance to use English for real communicative purposes with native children of their age. The potential effect of such a network can be explored in future applied research.
Appendix Table 4.1. Subject Description Younger children (4th grade) [Y]
Older children (5th grade) [O]
Japanese, shorter incubation from [JS]
9:7 yrs. old Lived in U.S. for 3:8 yrs, 5:10 to 9:6 Returned 3 weeks before initial data collection Only child
11:3 yrs. old Lived in U.S. for 4: 11 yrs, from 6:3 to 11:2 Returned 4 weeks before initial data collection Has one elder brother
Japanese, longer incubation [JL]
9:6 yrs. old Lived in U.S. for 5:5 yrs, from 3:0 to 8:5 Returned 13 months before initial data collection
11:3 yrs. old Lived in U.S. for 3:3 yrs, from 6:9 to 10:0 Returned 15 months before initial data collection
Table 4.2. Results of the Listening Comprehension Task Level Session
6
7
— P P
P P P P
8
9
65 55 P 70
65 55 75 —
65 — — 35
60 — — —
65 60 55 50
65
— — —
50 65 70 70
— — — —
JOS
1
2 3 4
—
JOL
1 2 3 4
—
78 P P
P 76 73 P JYS
1 2 3 4
— P — —
P P P P
1 2 3
P P — —
78 P P P
JYL
4
Note: P(ass) = 80 points or more among 100
97
Table 4.3. Number of Phonological Errors JOS
JOL
JYS
JYL
1 2 3 4
0
0 0 0
0 2 1 2
2 1 0 3
2 2 1
Total
0
5
6
5
Session
0
Table 4.4. Summary of Verb Forms Used by JOS During Free Interaction Session %SOC %T-L
1 100
92
Copula (is) 2 3 4 100 100 (100) 100 100 (100)
Session %SOC %T-L
1 100 100
Copula (was) 2 3 4 (100) 100 100 (100) 100 100
Session %SOC %T-L
1 89 93
Session %SOC %T-L
1 100 100
Verb Stem 2 3 100 88 100 97
4 100 100
Past (regular) 2 3 4 (100) 100 88 (100) 100 100
Copula (are) 2 3 4 100 (100) 100 100 (100) 100
1 57
100
1 (100) (100)
Copula (were) 2 3 4 (100) — (100) (100) (100) —
3rd Person Sing, (-s) 2 3 4 95 (100) 83 (100) 79 (67) 63 (100)
1
Past (irregular) 2 3 4 100 95 100 100 100 100
1
96 100
During Story Description Session %SOC %T-L
Copula (is) 2 3 100 (100) (100) 80 (100) (100)
1
4 (0) (0)
100 100
Copula (was) 2 3 4 100 100 100 100 91 70
1 100 100
Auxiliary (is) 2 3 100 (100) 91 (100)
1 — —
Session %SOC %T-L
(100)
85
Copula (were) 2 3 4 (50) 25 33 (100) (100) (100)
Session %SOC %T-L
1 100 100
Auxiliary (was) 2 3 4 90 100 95 100 78 100
1
98
1
4 (0) (0)
Auxiliary (were) 2 3 4 (100) 33 (100) (50) (100) (100)
Table 4.4. (continued) During Free Interaction 3rd Person Sing, (-s) Session
%SOC %T-L
1 71 100
2 86 100
3 57 100
Verb Stem
4
1
(100) (100)
(100) (100)
4 97 97
1
Past (regular) Session
%SOC %T-L
1 100 86
2 100 100
3 100 93
96 100
2 0 (100)
4 3 100 (100) 33 (100)
Past (irregular)
2 100 100
3 96 100
4 98 100
Notes: Verb morphology with little or no data across all sessions was excluded from the table. Dashes (—) indicate no data. Percentages in parentheses indicate that there were less than three occurrences of the form in the data.
99
Table 4.5. Summary of Verb Forms Used by JOL During Free Interaction Session %SOC %T-L
1 99 96
Copula (is) 2 2 3 4 100 93 100 91 82 75
Session %SOC %T-L
1 77 80
Verb Stem 2 2 3 4 85 100 0 65 68 0
Session %SOC %T-L
PAST (regular) 2 2 3 4 1 50 (100) 0 0 (100) (0) 0 (0)
1 100 75
Copula (are)
2
— —
3
4
— —
(0) (0)
Copula (was) 1 14 (100)
2
3
4
50 33 (33) (100) (100) (100)
3rd Person Sing, (-s) 1 2 3 4 (0) 25 (0) — (0) (100) (0) —
1 70 59
PAST (irregular) 2 2 3 4 63 62 46 92 93 100
During Story Description Session %SOC %T-L
1 100 83
Copula (is) 2 3 4 2 80 100 100 62 69 50
Session %SOC %T-L
1 (100) (100)
Auxiliary (is) 2 3 4 2 64 88 82 100 100 90
Session %SOC %T-L
1 (100) 5
Verb Stem 2 3 4 2 73 92 100 19 24 33
Session %SOC %T-L
1 (0) (0)
Past (regular) 2 3 4 2 — 18 — — (100) —
Copula (are) 2 1 2 3 4 (50) (50) (0) (33) (50) (100) (0) (100)
1 — —
Auxiliary (are) 2 3 4 2 (100) 25 (0) (100)(100) (0)
3rd Person Sing, (-s) 32 3 4 2 1 0 7 14 0 0 50 86 0
1 44 100
Past (irregular) 2 3 4 2 34 11 15 100 (100) (100)
Notes: Verb morphology with little or no data across all sessions was excluded from the Cable. Dashes (—) indicate no data. Percentages in parentheses indicate that there were less than three occurrences of the form in the data.
100
Table 4.6. Summary of Verb Forms Used by JYS During Free Interaction Session %SOC %T-L
1
100 100
1
Copula (am) 4 2 3 100 (100) (100) 100 (100) (100)
97 90
Copula (is) 2 3 100 97 96 91
Copula (are) 2 3 67 57 100 100
4 67 100
1 94 97
Copula (was) 4 2 3 100 100 100 94 100 100
\
2 100 100
Auxiliary (am) 3 4 100 (100) 100 — (100)
Session %SOC %T-L
58 100
Session 1 %SOC %T-L
Copula (were) 3 4 100 (100) 86 100 (100) 100
2
(0) (0)
Session %SOC %T-L
Auxiliary (is) 4 2 3 — (100) (100) (100) — (100) (100) (100)
Session %SOC %T-L
Auxiliary (was) 4 2 3 100 (100) — (100) 100 (100) — (100)
Session %SOC %T-L
Modal (could) 2 3 100 (100) 100 (67) 100 100
1
1
1
Session %SOC %T-L
100 93
Verb Stem 2 3 98 97 99 97
Session %SOC %T-L
1 95 97
Past (regular) 2 3 100 83 100 83
1
1
Auxiliary (are) 4 1 2 3 (50) — (100) (100) — (100) (100) (100)
1
100 100
4
1
(0) (0)
(100)
4 100 96 4 100
88
4 100 88
(100)
Modal (can) 4 2 3 92 100 100 80 100 100 Modal (will) 4 2 3 100 — (100) 75 100 —
3rd Person Sing. (-s) 4 2 3 85 100 96 100 100 89 92 100
1
Past (irregular) 2 3 4 94 82 98 91 100 100 100 100
1
During Story Description Session %SOC %T-L Session %SOC %T-L
1 95 95
Copula (is) 2 3 100 100 92 100
4 100
83
Copula (was) 1 2 2 33 4 (100) 100 91 100 (100) 86 100 63
1
101
Copula (are) 1 2 3 (50) (67) 100 (100) (100) 100
4 (0) 0
Copula (were) 1 22 3 3 4 4 (100) (25) (100) (50) (100) (100) (33) (100) (continued)
Table 4.6. (continued) During Free Interaction Auxiliary (is) Auxiliary (are) Session 2 1 2 3 4 1 2 3 4 %SOC (100) 91 100 100 (100) (33) 71 (100) %T-L (100) 95 91 100 (100) (100) 100 (100) Auxiliary (was) Session 2 1 2 3 %SOC — 100 100 (100) %T-L — 88 100 (100)
session % SOC %T-L
Third Person Sing, (-s) 4 1 2 3 94 100 95 75 100 96 100 100
Session %SOC %T-L
1 100 100
4
Verb Stem 1 2 3 4 100 (33) (100) (100) 75 (100) (100) 40 1 (100) (100)
Past (regular) 2 3 4 100 (100) 100 100 (100) 88
Past (irregular) 3 2 4 83 100 96 100 96 100
Notes: Verb morphology with little or no data across all sessions was excluded from the table. Dashes (—} indicate no data. Percentages in parentheses indicate that there were less than three occurrences of the form in the data.
102
Table 4.7. Summary of Verb Forms Used by JYL During Free Interaction Session %SOC %T-L
Copula (am) 2 3 1 4 (0) (100) — 78 — (100) — 100
1 98 98
Copula (is) 2 3 97 97 98 89
4 99 93
Copula (are) 3 2 100 67 94 (100)
89 89
Session %SOC %T-L
1 67 (100)
4 69 92
1 100 100
Copula (was) 3 2 88 89 94 93
Session %SOC %T-L
Copula (were) 3 2 1 4 (0) (50) (50) (100) (100) (100) — —
1 — 100
Auxiliary 2 100 —
4
(am) 3 4 — — — —
Auxiliary (are) 4 3 2 (0) — 67 (0) — 100
Session %SOC %T-L
1 — —
Auxiliary (is) 2 3 4 100 — (100) 80 — (50)
1 (100) (100)
Session %SOC %T-L
1 — —
Auxiliary (was) 4 2 3 100 — (100) 100 — (100)
1 100 67
Modal (can) 3 2 100 90 100 100
4 97 100
Session %SOC %T-L
1 33 (0)
Modal (could) 4 3 2 — 100 100 — 93 78
1 — —
Modal (will) 3 2 — 100 — 100
4 90 90
Session %SOC %T-L
1 100 77
Verb Stem 2 3 96 87 74 78
4 96 87
Session %SOC %T-L
1 33 80
Past (regular) 2 3 67 88 73 75
4 63 56
3rd Person Sing, (-s) 4 3 1 2 — 59 80 8 — 83 100 (100) 1 89 100
Past (irregular) 3 2 4 82 79 73 100 97 87
During Story Description Session %SOC %T-L
1 (100) (100)
Session %SOC %T-L
1 100 100
Copula (is) 3 4 2 — 83 (100) — 100 (100) Copula (were)
2
(0) (0)
3
— —
4 0 (0)
1 100 100
Copula (was) 33 22 44 100 95 100 83 91 69
Auxiliary (was) 22 3 34 4 82 95 100 (100) 100 83 100 (100)
11
(continued)
103
Table 4.7. (continued) During Free Interaction Verb Stem
Session
%SOC %T-L
Session %SOC %T-L
11 22 (0) (100) 0 10
1 100 71
33 44 (50) (100) 8 8
Past (regular)
3rd Person Sing, (-s)
111 0 (0)
2
3
4
11
100 65
94 80
98 84
85 100
222 0 (0)
3 33 4 44 31 0 100 (0)
Past (irregular)
22 84 100
33 100 100
4
4
85 100
Notes: Verb morphology with little or no data across all sections was excluded from the table. Dashes (—) indicate no data. Percentages in parentheses indicate that there were less than three occurrences of the form in the data.
104
Table 4.8. Articles Used in Obligatory Context JOS a/an
the
a/an
the
a/an
the
a/an
Used Required %SOC
74 74 100
0 0
14 22
1 4
137
59 61
1 8
(50)
59 10 100
96.7
87.5
Used Required %SOC
110 112
3 3
Used Required %SOC
119 120
Used Required %SOC
110 111
Data
1
3
4
JYL
JYS
the
Session
2
JOL
98.2
99.2
99.1
140
63.6
(25)
57 82
3 11
69.5
27.3
2
55
2
75
5 10 50
163 165
2 6
102 104
(100)
(100)
2 2 (100)
73.3
59 75 78.7
33.3
97.9
201 203 99.0
98.8
98.1
16 20
92 121
80.0
76.0
25 26
62 80
96.2
77.5
12 13 92.3
98 112 87.5
Note: Percentages in parentheses indicate that there were less than three occurrences of the form in the data.
105
16 18 88.9
1 3 (33.3)
4 9 44.4
Table 4.9. Types and Tokens of Vocabulary Used during Spontaneous versus Planned Speech Session
Data
JOS
JOL
JYS
JYL
Free Interaction
1
2
3
4
Type Token Ratio
483
349
785
484
2,085
2,159
5,267
2,582
Type Token Ratio
248
258
511
459
2,875
1,623
3,063
2,717
Type Token Ratio
1,227
Type Token Ratio
.28
291
.19
.17
.16
.17
256
486
448
1,594
2,713
2,535
.24 255 895
.15
.16
.23
.18
.16
.18
271
500
448
1,899
2,624
3,158
9
14
.29
.14
Planned Speech
1
2
3
4
Type Token Ratio Type Token Ratio
63 119
.53
85
101
249
234
79
38 59
Type Token Ratio
48 82
.65
.30
86 190
95
299 .55
59
66 .30
106
84 210
44
71
.47
.45
.32 838
.44 126 266
37 57
47
159 .46
Type Token Ratio
.43
.34
172
71 161
.67
.40 67 118 .57
Table 4.10. Types and Frequency of Paraphrase Used by the Returnees Session
Paraphrase Types
JOS
JOL
JYS
JYL
1
Total Clauses Approximations Word Coinages Circumlocutions
346 5(1.4) 0(0) 0(0)
306 11(3.6) 0(0) 1 (0.3)
556 10(1.8) 0(0) 1 (0.2)
832 45 (5.4) 1 (0.1) 10(1.2)
2
Total Clauses Approximatio Word Coinages Circumlocutions
224 5(2.2)
254 17 (6.7) 0(0) 2 (0.8)
656 17(2.6) 2 (0.3) 1 (0.2)
573
Total Clauses Approximations Word Coinages Circumlocutions
278 3(1.1)
239 23 (9.6) 0(0)
534 44(8.2)
2 (0.8)
570 13(2.3) 0(0) 8(1.4)
Total Clauses Approximations Word Coinages Circumlocutions
224 4(1.8) 0(0) 1 (0.4)
262 21 (8.0) 0(0) 3(1.1)
563 11 (2.0) 4 (0.7) 0(0)
492 69(14) 2 (0.4) 4(5.8)
3
4
0(0) 0(0)
0(0) 0(0)
48 (8.4) 1 (0.2) 2 (0.3)
3 (0.6) 7(1.3)
Note: Numbers in parentheses indicate the percentage of respective paraphrase type that occurred among the total number of clauses.
Table 4.11. Frequency of Japanese Used JYS
JYL
Session
Japanese (J) Usage
JOS
JOL
1
J Words J Word as Error/Strategy Total Words % J Word as Error/Strategy
0 0 557 0
18 16 379 4.2
3 0 995 0
3 0 583 0
2
J Words J Word as Error/Strategy Total Words % J Word as Error/Strategy
0 0 745 0
20 17 813 2.1
4 1 1525 0
1 0 815 0
J Words J Word as Error/Strategy Total Words % J Word as Error/Strategy
0 0 778 0
30 26 746 3.5
9 7 1319
.005
0 0 809 0
J Words J Word as Error/Strategy Total Words % J Word as Error/Strategy
0 0 761 0
23 20 923 2.3
2 2 879 .002
0 0 920 0
3
4
107
Table 4.12. Percentage of Attempted and Successful Complex Clauses Session
Data
JOS
JOL
JYS
JYL
1
Attempted/Total Successful/Attempted
5.8 100
1.0 33.3
7.6 88.1
4.8 37.5
2
Attempted/Total Successful/ Attempted
5.8 76.9
1.2 0.0
5.6 78.3
9.9 28.1
3
Attempted/Total Successful/ Attempted
12.9 77.8
2.1 20.0
6.5 73.0
7.5 35.0
4
Attempted/Total Successful/ Attempted
8.9 85.0
1.1 66.7
2.7 66.7
12.2 21.7
Table 4.13. Percentage of Error-Free Clauses among Total Clauses Uttered JOS
Session
Data
1
Error-Free Clauses Total Clauses % Error-Free Clauses
312 346
Error-Free Clauses Total Clauses % Error-Free Clauses
190 224
Error-Free Clauses Total Clauses % Error-Free Clauses
236 278
Error-Free Clauses Total Clauses % Error-Free Clauses
197 224
2
3
4
90.2
84.8
84.9
87.9
108
JOL 157 306 51.3
99 254 39.0
53 239 22.2
83 262 31.7
JYS 478 556 86.0
565 656 86.1
470 570 82.5
460 563 81.7
JYL 617 832 74.2
377 573 65.8
306 534 57.3
201 492 40.9
Table 4. 14. Measurement of Fluency Session
Speech per Time Unit
1
JOS
JOL
JYS
JYL
Total Words Total Seconds Words/Second
553 335 1.7
268 455 0.6
972 640 1.5
574 930 0.6
2
Total Words Total Seconds Words/Second
736 480 1.5
601 660 0.9
1485 1130 1.3
807 940 0.9
3
Total Words Total Seconds Words/Second
776 410 1.9
594 650 0.9
1277 1090 1.2
808 915 0.9
4
Total Words Total Seconds Words/Second
759 465 1.6
759 660 1.2
864 680 1.3
915 845 1.1
Average Number of Clauses per Repair 1 2 3 4
27.7 12.0 24.8 9.5
2.5 2.2 3.1 2.9
5.9 2.3 3.0 2.0
2.0 3.9 2.9 2.9
Average Number of Clauses per False Start 1 2 3 4
16.6 20.0 15.5 30.8
5.9 2.3 3.0 2.0
3.6 1.5 1.8 1.7
2.8 3.0 5.8 3.3
Average Number of Clauses per Filler 1 2 3 4
20.8 13.3 62.0 61.5
109
2.0 3.9 2.9 2.9
8.5 5.3 4.5 8.6
10.1 17.9 15.2 31.2
110
The Loss of English as a Second Language
Not 1. This chapter is a condensed version of Yoshitomi (1994), a Ph.D. dissertation submitted to the University of California, Los Angeles. 2. Hypothesis-generating type of research, based on real-life language use represented by samples of free speech in various situations, would be useful in extracting features, skills, strategies, or whatever relevant types of knowledge that could be later tested more specifically in diagnostic type of measurements. Especially in studying children at different stages of cognitive developments, measurements of high validity will be hard to develop. As Andersen puts it, "To some extent the researcher should let the nature of the data itself guide him in choosing which features he should study" (1982:87). This approach, according to Andersen, is a common one taken in language variation research in L1 acquisition, as well as in sociolinguistic variation research in the use of a native language, and in pidgin and Creole studies. The approach seems relevant to language attrition research as well, in as much as people suffering from language loss are likely to exhibit different variations in their language use as compared with native speakers, and this difference in variation is precisely one of the major concerns in studying attrition (Olshtain 1989). 3. Ideally, the sample subjects should represent combinations of at least the following characteristics, excluding, sociopsychological variables: different age at the time of arrival (early versus later or late) in the United States, different age at the time of departure (early versus later or late) from the United States, different lengths of stay (relatively long versus relatively short) in the United States, different lengths of incubation (relatively long versus relatively short) since returning to Japan, and different sex (male versus female). This would require a minimum of 32 subjects, each filling the respective "cells" of possible combinations of characteristics. Moreover, it would be best to have a number of cases for each cell. 4. The contextual information of the data together with the date from the acquisition phase provided the means for evaluating the individual subjects' proficiency. Therefore, no proficiency test was given, as in the structure-centered normative tests employed in many of the previous cross-sectional studies. As reviewed in Yoshitomi (1994), there are some methodological problems in previous L2 attrition studies. In essence, no single reliable and valid general proficiency measurement is currently available, especially for the purpose of assessing the individuals' levels of language proficiency in "real-life" communicative situations. This lack of measurement is especially apparent for young children, who are the subjects of the present study. According to Bachman & Clark, most currently available measures are inadequate because, "they are based on a model of language proficiency that does not include the full range of abilities required for communicative language use, and they are based on norm-referenced principles of test development that only permit interpretation of ability levels relative to the performance of specific groups of language users" (1987:21). Knowledge of grammar or language structure rather than of communication skills has been the main focus of previous studies. Although such focus in itself may reveal interesting characteristics about language attrition, future studies should also look into other types of skills or competence if we are to truly grasp what is going on in the process of language loss. I would also like to add that based on my experience of having observed and talked with many returnees, their communicative abilities in L2 are roughly the same
Loss of English by Japanese Returnee Children
111
at the time of return to Japan if their length of stay in an English-speaking country was three to five years, if their age of arrival in the country was between five to eight years, and if they attended a regular English-speaking school while they were abroad. 5. This kind of measurement should be used cautiously, because even native speakers may exhibit many repairs, false starts, and fillers. The attempt here was to measure any increase in the occurrences of repairs, false starts, and fillers in the individual returnee's speech across time; it was not a direct comparison of the frequency of repairs, false starts, and fillers among the returnees. 6. Fluency can be defined in many ways. Most definitions rely on the rater's intuition in judging the fluency of a language learner. For example, the Foreign Service Institute interview scale defines fluency as the "overall smoothness, continuity, and naturalness of speech." Here I adopt Riggenbach's (1989) approach and attempt to measure fluency in terms of "fluency-related markers." Riggenbach lists hesitations, repairs, and rate of speech as possible features ascribed to fluency. Hesitations include filled and unfilled pauses; repairs refer to either repetitive repairs or restarts, and the rate of speech is the numbers of words uttered per minute. Riggenbach uses a microanalytic approach, which involves the measurement of unfilled pauses in microseconds. Such microscopic measurement was impossible in this study; hence, only hesitations in terms of pauses filled by fillers (e.g., uh, um) were counted. What Riggenbach calls "repetitive or retraced repairs" are what I call "repairs." They refer to reformulations in speech in which part of the original utterances is repeated. What Riggenbach calls "(unretraced) restarts" are what I call''false starts," defined as reformulations in speech in which the original utterance is rejected and rephrased in terms of different expressions (p. 43). 7. All statistics reported in this study are descriptive statistics. The amount of data and the nature of the data did not allow for inferential statistics in many cases. Furthermore, the nature of the study being more qualitative than quantitative, and more of a hypothesis-generating type than of a hypothesis-testing type, it was assumed that descriptive statistics could render sufficient information for the purpose of the study. In preliminary studies such as that reported in this paper, merely pursuing "statistical significance" may result in ignoring essential characteristics of language use in the attrition process (Roger Andersen, personal communication). I do admit, however, that the use of inferential statistics would be compulsory in future studies, where the research design is experimental and the purpose of study more fine-tuned to "test" specific hypotheses.
This page intentionally left blank
PART II
The Loss of Japanese as a Second Language by Adult Foreigners
5 Lexical Maintenance and Attrition in Japanese as a Second Language Robert A. Russell
In recent years, researchers have shown increasing interest in the phenomenon of second language (L2) loss or attrition. A few of the issues that have been addressed in the literature include identification of the particular skills and subsystems of L2 that are lost, the order in which they are lost, the rates at which different skills are lost, the processes and variables affecting L2 loss, and the amount and timing of attrition.1 The present study deals with the maintenance or loss of productive L2 lexical skills. Studies that have dealt with questions of lexical attrition include those, for example, by Cohen (1986, 1989), Graham (1990), Brown & Graham (1991), Olshtain & Barzilay (1991, on first language [L1] lexical attrition), and Grendel (1993). One such question deals with whether the lexicon is lost more or less rapidly than other L2 subsystems, such as grammar and morphology. Although intuition might suggest, for example, that vocabulary is lost more rapidly than grammar, Weltens, van Els & Schils (1989) report, to the contrary, that after zero, two, and four years of disuse of French as a foreign language, learners displayed greater attrition in grammar than in phonology, the lexicon, or other areas measured in their study. Using a time-restricted lexical decision experiment, Grendel's (1993) study of Dutch learners of French as a foreign language found that, while speed of lexical accessibility might be slowed as a result of disuse, there was no significant evidence of actual loss over a period of several years, at least with respect to certain aspects of orthography and of lexical associations. Still other studies claim to have observed lexical attrition in varying amounts over varying periods of disuse: McMahon (1946) reported a 7.4% loss of vocabulary in Latin as a foreign language over a three-month period and, for French as a foreign language, Messelink & Verkuylen (1984) report no loss in the first year but 15% loss in the second year, as cited in Weltens (1987). With the notable exceptions of Cohen (1986, 1989) and Olshtain (1986), 114
Lexical Maintenance and Attrition
115
most studies of lexical attrition to date have focused on receptive lexical knowledge or on limited aspects of such knowledge elicited under experimental conditions. According to an earlier hypothesis by Anderson (1982), however, "active" vocabulary is more likely to be lost than "passive" vocabulary, a hypothesis that has received some confirmation from a study by Gonzo & Saltarelli (1983), in which the effects of attrition were more pronounced in a lexical retrieval task than in one involving lexical recognition (see de Bot & Weltens, 1995). In a study dealing with changes in certain aspects of both receptive and productive lexical knowledge, Graham (1990) reports no clear, linear pattern of vocabulary loss in adult speakers of Spanish as a second language, either over the two-year period of disuse during which the experiment was conducted or as a function of the amount of time elapsed since the onset of disuse (ranging from zero to four years prior to the first test). Graham's study was limited, however, primarily to an analysis of the relative frequency of occurrence in the subjects' data of the "most frequent words" as determined by independent frequency lists. Further questions have been raised with respect to the effects on lexical attrition of such variables as the learner's age or other personal factors, the proficiency level achieved before the onset of disuse, the type of vocabulary lost (such as function versus content), the methods by which the vocabulary was acquired, and the setting in which it was acquired. The results of studies that have addressed such questions have been varied and in many cases inconclusive.2 It should also be noted that most of the attrition-related studies reported in the literature thus far have dealt with Spanish, French, German, English, or other Indo-European languages as second or foreign languages. The picture of lexical attrition that has begun to emerge, however incompletely, is thus heavily colored by semantic and structural characteristics that may be peculiar, to some degree, to the particular subset of target languages studied. We clearly need studies that deal with second languages that have lexical semantic properties and inflectional, compounding, and other morphological processes that are different from those typically found in the Indo-European languages. Only then can we begin to form a more accurate view of the processes of lexical acquisition and attrition that hold across all languages.
Purp se of the Study The overall objective of the present study is to examine patterns of lexical maintenance and attrition specifically among native English-speaking learners' of Japanese as a second language (JSL). The data on which the research is based came from transcriptions of oral monologue responses to a number of questions that were administered to a group of (initially) eighty subjects on three different occasions over a period of two years. The subjects, all of whom were university students in their early twenties who had recently returned from approximately two years of missionary service in Japan, varied as to the year of their return from Japan and in the amount of formal instruction they had received in Japanese following their return. Of the four major categories of language attrition identified
116
The Loss of Japanese as a Second Language
by van Els, de Bot, & Weltens (1983), this study falls most nearly into the third category—that is, the loss of L2 or foreign language (FL) in an L1 environment. In the present case, however, we are looking for the possibility of attrition not in "foreign languages learned at school" but in a second language learned largely in a host culture setting with a subsequent return to the L1 environment. Specific questions addressed in this chapter relate to the nature and extent of attrition in productive lexical ability as reflected in total word counts, lexical variability (based on type-token ratios), lexical density (based on ratios of content, or open-class, words to total words), lexical accuracy (based on lexical error rates), lexical frequency orders, and relative frequencies of filler words, loan words, and English words in the JSL data. The study also addresses the question of whether extended formal instruction in the language upon returning from Japan correlates significantly with changes in any of the above-mentioned variables related to productive lexical competence. Each question will be discussed in detail below. Although it is difficult to measure precisely, it can safely be assumed that, both in general and, usually, in specific contexts of elicitation, vocabulary size will be larger for native speakers of a particular language than for nonnative speakers of that language (Chapelle 1994; Aitchison 1987). It is also natural to assume that the size of L2 vocabulary should increase as the learner's interlanguage develops (Blum-Kulka & Levinson 1983; Laufer & Nation 1995). Conversely, under conditions of attrition, as vocabulary is forgotten or becomes inaccessible for production, it follows that vocabulary size (especially of content words) will decrease and that the decrease should be reflected in the total number of words (both tokens and types) produced in a given context. In Cohen's (1989) longitudinal study of third language (L3) Portuguese oral data elicited from two bilingual English-Hebrew children, for instance, a decrease in vocabulary size was observed following a period of disuse. For the present study as well, it was predicted that a decrease in vocabulary size would also be observed among JSL-speaking subjects over a two-year period of observation. With regard to lexical variability, as measured by the ratio of types to tokens, one might expect, intuitively, that native speakers of a given language would exhibit higher degrees of variability than nonnative speakers because of the generally larger number of open-class (or content) lexical items available to native speakers (as opposed to nonnative speakers) in comparison with the small number of closed-class (or function) words, where a much smaller difference is likely to be found between native and nonnative speakers. In the acquisition of a second language, it might also be expected, therefore, that as the size of learners' content vocabulary increases with increased levels of general proficiency, lexical variability should also increase, at least for longer samples of production. Likewise, one might also expect lexical density (the ratio of content tokens to total tokens) to correlate positively with levels of lexical and general proficiency—that is to say, native speakers and more advanced learners of a given language, with their larger content vocabularies, might naturally be expected to exhibit higher ratios of content to total tokens than those exhibited by learners of lower proficiency.
Lexical Maintenance and Attrition
\\7
In the few studies that directly address these issues, however, there seems to be no clear evidence of a correlation between either lexical variability or density on the one hand and L2 proficiency level on the other. Laufer (1991), for example, observed no significant increase (in an acquisition setting) in lexical variability or lexical density in written compositions by advanced English as a Second Language (ESL) learners in her study over a two-semester period of instruction (see also Laufer & Nation 1995). That observation is weakened, however, by the absence of independent evidence of any change in general proficiency among the subjects over the two-semester interval. In an unpublished paper dealing with data in an oral production mode, Brown & Graham (1991) also found no significant evidence of a decline in lexical variability (measured by type-token ratios) over a two-year period of disuse by L2 speakers of Spanish and Japanese.3 In the latter study, however, there was no control over the variable of subjects' further exposure to the L2 during the period of relative disuse. The failure of existing studies to find a correlation between lexical variability and density on the one hand and changes in proficiency on the other hand may be due, in other words, at least in part to their failure to control critical variables. The present study does attempt, to some degree, to control the variable of further exposure, but due to the post hoc nature of its design, it was not possible to control for the variable of subjects' initial proficiency. This failure is partially mitigated by the fact that all subjects in the study had undergone the same predeparture, intensive two-month language training course and had spent an identical period of 22 months in the host country, acquiring the L2 informally under highly similar conditions. Based, then, on the foregoing considerations, another question in this study is whether lexical variability (type-token ratios) and lexical density (content-to-total-word ratios) for relatively advanced JSL learners performing in an oral mode will be found to have decreased over a two-year period of relative disuse of the L2, and whether the presence or absence of formal instruction following a two-year period of rather intensive informal acquisition in the host culture will be found to have had an effect on any changes in lexical variability or density. With regard to lexical accuracy or quality, it was predicted that, even if broad measures of lexical variability and density turn out not to exhibit significant change, learners in the study might nevertheless show a decreasing ability to use words accurately, as reflected in (1) an increased ratio of various types of lexical errors per T-unit of discourse, (2) a decreased ratio of the number of words per (lexical) error-free T-unit, and (3) a decrease in the number of error-free T-units as a percentage of the total number of T-units.4 It was also expected, however, that any decrease in lexical accuracy might be more pronounced in the group of learners who had little or no subsequent formal instruction than in the group who received formal instruction. The further formal instruction to which subjects were exposed emphasized reading but also included a considerable amount of conversational practice as well. It was expected, intuitively at least, that the additional exposure would help reinforce vocabulary skills already acquired by those subjects. It remains an empirical question, of course, whether and to what
118
The Loss of Japanese as a Second language
extent instruction in reading and writing skills will transfer to, or reinforce, oral skills—a question, however, that lies beyond the scope of this study. The weighted frequencies of the lexical items most commonly used by the JSL subjects in this study were also calculated and compared, both quantitatively and qualitatively, with those of a group of Japanese native speakers (JNS). It was predicted that the JSL groups' frequency orders would correlate more closely with each other than with those of the JNS group, and that there might also be a decrease in the degree of correlation between JSL and JNS frequency orders over time. It was further expected that there might be differences between JSL and JNS subjects with regard to the frequency of different lexical strata. The lexical strata referred to here are those into which the Japanese lexicon has traditionally been divided (Shibatani 1990:142-153): native Japanese words (wago or Yamato-kotoba), Sino-Japanese words (kango), and loan words from languages (primarily Indo-European) other than Chinese (gairaigo). In addition to the three traditional lexical strata, however, a fourth category must be added when analyzing the interlanguage speech of JSL learners—words borrowed from the learner's L1 (in this case English), which may be substituted when the corresponding expressions cannot be retrieved from L2 stores. While a detailed analysis of words by syntactic category and of lexical strata is in progress and will be reported at a later date, the present report focuses only on the relative frequency of loan words and of English words in the JSL data. Specifically, it was predicted that JSL learners would use relatively higher frequencies of both loan words and English words than did the JNS subjects; that the relative frequencies of both strata would increase over time, indicating increased difficulty in the retrieval of corresponding native Japanese and Sino-Japanese forms; and that the learners with further formal instruction would show a lower rate of increase in loan word and English word usage than those without formal instruction.
Method Subjects For the present study, a group of (initially) 80 subjects was randomly selected from among hundreds of students at Brigham Young University who had spent approximately two years as missionaries in Japan prior to the beginning of the study. Of the 80 subjects who participated in the administration of the first test, only 65 were able to participate in all three elicitations. The 65 subjects who participated in all three tests were then divided into two groups, one group consisting of learners with a significant amount of additional formal instruction in Japanese (eight or more semester hours) over the two-year period of the study, and the other group consisting of learners without such instruction (zero to four semester hours). Ten subjects were then selected at random from each of the two groups. As indicated, the purpose of the study was to observe the nature and
Lexical Maintenance and Attrition
119
extent of attrition or other changes in the subjects' productive JSL lexical competence over an extended period of time, and to measure the effects, if any, of further formal instruction on JSL lexical maintenance. The study deals with productive (versus receptive) lexical skills, in an oral (versus written) mode, with free monologue (versus dialogue or more controlled elicitation) tasks, with young adult JSL learners (all of whom were male,5 were native speakers of English, had completed approximately two years of residence in Japan as missionaries, and had undergone an identical period of largely informal acquisition of Japanese under highly similar conditions). The subjects varied minimally with respect to the amount of exposure they had had to Japanese prior to leaving for Japan (the majority had had virtually no previous exposure to the language). They also varied as to the amount of experience in learning other languages, the year of their return from Japan, and the amount of formal instruction and further informal exposure to Japanese following their return. The subjects also varied, certainly, with respect to actual level of proficiency in the language, general intelligence, language aptitude, and other personal factors that could not be controlled in the present study. Of the factors mentioned above, the only variation that could systematically be controlled was that relating to the amount of formal instruction subjects received upon returning from Japan. For normative purposes, data were also elicited from five native speakers of Japanese; the same set of questions and test conditions were used as for the JSL subjects.6 Data Collection A mixed cross-sectional and longitudinal design was used for the study. As described above, oral monologue data were elicited from two groups, each consisting of ten subjects, at three different points in time over a period of approximately two years between late 1986 and early 1989. The subjects were given the same elicitation task each time. With approximately one year intervening between test administrations, the likelihood of test-learning effects was minimal. The subjects were asked to record their oral responses, in Japanese, to each of six questions. Four minutes were allowed for each response. Two of the questions were based on religious topics familiar to the subjects, but these were omitted from the present study because they would not permit comparison with a larger population of JSL learners. A third question involving a picture-guided story retelling was also omitted because data were not available from the JNS group for comparative purposes. The three remaining questions, on which the present study is based, are included for reference in Appendix A. The subjects' responses were recorded on audiocassette tapes and were later transcribed and edited for accuracy. A further editing process was required to ensure that words were segmented consistently in preparation for later treatment with text analysis software.7
120
The Loss of Japanese as a Second Language
Data Analysis The transcribed text files were initially processed using WordCruncher® software (a text indexing, retrieval, and analysis program developed by Electronic Text Corporation, 1988) to generate lists of words and their frequencies. Typetoken ratios were then calculated using these lists. As discussed above, typetoken ratios are interpreted as measures of lexical variation, with higher ratios reflecting greater variety in the productive lexicon. In the present study, a decrease in type-token ratios over time would be taken to indicate decreased lexical variability, an indication of lexical attrition. Contrary, however, to Laufer's observation (1991:447, n.9) that lexical variation "is sensitive to the length of the composition" (leading her to limit the calculation of type-token ratios to the first 250 words of each essay), there was no correlation in our data between text length and type-token ratios. Type-token ratios in the present study were therefore based on the entire text of each subject, regardless of length. The type-token lists abstracted from the text files in the present study were also sorted further by frequency, so that word frequency orders could easily be generated for each subject. A software program was also developed to facilitate the tagging and frequency analysis of words in each file based on grammatical category and lexical stratum. The relative frequencies of content words (nouns, verbs, adjectives, and adverbs) versus function words (postpositions, conjunctions, complementizers, fillers, etc.,) were also analyzed and compared with those of the JNS group. The text files were segmented into T-units and analyzed for lexical errors by three different native speakers of Japanese, all of whom were graduate students in applied linguistics and had received training in procedures required specifically for the present analysis.8 It was difficult in a number of cases to assign errors to unique categories but also, at times, even to distinguish between lexical and syntactic or other errors. This difficulty may have resulted, in part, from inadequacies in the categories chosen for analysis but also, it surely reflects the underlying difficulty of clearly distinguishing, in theoretical terms, the boundaries between different components of grammatical competence. For example, it might be argued that an error such as Nihon ni *kita mae wa ("Before I came to Japan") is really a grammatical error involving inappropriate tense marking (past tense kita "came," instead of the required nonpast form kuru "come"), and not a lexical error. It can also be argued, however, that the incorrect tense marking itself results, in part at least, from an incomplete understanding of the lexical semantics of the word mae, "before." At any rate, error counts used in the present analysis were limited to total numbers of lexical errors. A detailed analysis of individual error types is in progress and will be reported at a later date. Brief mention should also be made of the procedure used for analyzing the frequency orders of different lexical items used by both JNS and JSL subjects. The 20 most frequent lexical items (including ties) were identified for each subject. Since there was considerable variation in individual item frequencies and some overlap in item inventories, simply calculating the average rank without
Lexical Maintenance and Attrition
121
taking item frequencies into account would have led to distortion in the rankings. Weighted average frequencies were therefore calculated using Rosengren's formula (1972), discussion of which is also found in Francis & Kucera (1982). Based on the resulting weighted averages, the 25 most frequent lexical items occurring in common between the JSL and JNS groups were then ranked and compared using the Spearman rank-order correlation. Mean values for other variables of interest (such as type-token ratios, lexical errors, etc.) were compared using analysis of variance (ANOVA); the two factors used in comparing learners' data were group and time, including two levels within group (ENS1, the JSL learner group with little or no further formal instruction; and ENS2, the JSL learners with further formal instruction) and three levels within time (representing the administration of the test at three different points in time, although only two levels were available for the analysis of lexical errors). A repeated measures design was used, as the subjects within each group were the same at each time level. For comparisons between JSL learners and the Japanese native speaker group, the general linear model of ANOVA was used because of the different numbers of subjects in the three groups. The Scheffe post hoc comparison was used in all cases to locate the differences among means. Because of the large number of dependent variables in this study, F-values, significance levels, and other detailed statistical data for all but the Spearman procedure will usually not be displayed in the body of the text, but they have been summarized in Appendix B for reference (see Tables 5.7 and 5.8).
Results and Discussion The data in Table 5.1 (see Appendix B) summarize the average total numbers of tokens and types, and the type-token ratios for each of the three groups: ENS 1 (JSL learners without further formal instruction), ENS2 (JSL learners with further formal instruction), and JNS (the reference group of native speakers of Japanese). The means for both groups of JSL learners are given for each of three elicitations; the JNS group was tested only once. As expected, the JNS group means for both total tokens and types were somewhat higher than those of both JSL groups, although not to a level of significance.9 The JNS type-token ratio was also slightly higher (32.6% to approximately 28% for both JSL groups at Time 1), although, once again, not significantly so. As predicted, there was a significant decrease in the total numbers of both tokens (