ACQUISITION OF JAPANESE VOCABULARY BY CHINESE BACKGROUND LEARNERS
The Roles of Transfer in the Productive and Receptive Acquisition of Cognates and Polysemy
T. KATO
A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy
School of Modern Language Studies University of New South Wales October, 2006 ACQUISITION OF JAPANESE VOCABULARY BY CHINESE BACKGROUND LEARNERS -The Roles of Transfer in the Productive and Receptive Acquisition of Cognates and Polysemy-
Toshihito Kato
ABSTRACT
As is widely known, Japanese and Chinese not only share the common logographic orthography called “kanji”or“hanzi” respectively, but also share a number of kanji compounds as cognates, many of which share the same or similar meaning. The major objective of this dissertation is to investigate the roles of transfer and the difficulty in
Chinese background learners’ (CBLs’) use and acquisition of Japanese kanji compounds and kanji words. In particular, under what condition and how CBLs transfer
Chinese words into Japanese counterparts is investigated.
The results of a lexicality judgement test, an oral production test, and a translation test showed that acquisition of partially deceptive cognates, which share the same orthography with partly the same and partly different meanings, was often prolonged. It was also found that the difficulty of acquisition of partially deceptive cognates varied according to their cross-linguistic semantic condition and task type. In the oral production test, CBLs frequently used L1 words by adapting them into L2 phonology both successfully and unsuccessfully when they had no prior knowledge of the L2 counterparts. In addition, negative transfer was detected even when CBLs had a correct knowledge of the L2 word. The results of the translation test revealed that CBLs are liable to misinterpret the meaning of partially deceptive cognates when one of their meanings happens to make sense within the context. Additionally, it is suggested that
ii CBLs might create different types of interlanguage depending upon the cross-linguistic semantic condition and relative frequency of the L2 input for each meaning of the partially deceptive cognates.
The transferability of polysemy was found to be constrained by prototype condition, learners’ existing L2 knowledge, and task type. While transferability correlated well with the perceived prototypicality of the L1 items in CBLs’ oral production, transfer was also at work for the less prototypical items in their comprehension task. The findings indicate that the transferability of Chinese words into their Japanese counterparts is constrained by multiple factors. Further, both positive and negative transfer influence CBLs’ production, comprehension, and interlanguage construction of
Japanese vocabulary in a complex manner.
iii ACKNOWLEDGEMENTS
I would like to take this opportunity to express my gratitude to those whose support and help made this dissertation possible. First and foremost, I would like to thank Associate
Professor Chihiro Kinoshita Thomson, my academic supervisor. She has given me insightful suggestions and encouragement throughout the years that I have been at the
University of New South Wales. Without her guidance and patience, this dissertation would not have been possible. I am also extremely grateful to my co-supervisor Dr.
Yew-Jin Fang for her patience and encouragement, especially in the later stages of thesis writing.
Many thanks are also due to Ms. Kazue Okamoto, Yukika Ojima, Kikuko Nakamura and other lectures at the University of New South Wales who helped me to collect the valuable data for this study. My gratitude also goes to Ms. Jessica Wu, whose help with
Chinese language and data collection has been indispensable to me. I am deeply indebted to Associate Professor Tatsuhiko Matsushita of Obirin University for his help and stimulating discussion. This motivated me to pursue the topic in this study.
I also thank Dr. Kayoko Evon for her advice on statistical analysis of the data. I thank
Dr. Paula McAndrew, who read the draft and polished the English of the thesis. I would like to express my gratitude to all my family members in Japan who have given me their support throughout the period of my stay in Sydney.
iv TABLE OF CONTENTS Page INTRODUCTION……………………………………………………………………. 1 0.1 Introduction ………………………………………………………………… 1 0.2 Objectives of the Present Research……...…………………………………... 10 0.3 Organisation of Chapters…………………………………………………….. 10
PART I: REVIEW OF RELATED RESEARCH…………………………………. 12 1. RESEARCH ON LANGUAGE TRANSFER……………………………………… 13 1.1 A Brief History of Transfer Research and the Contrastive Analysis Hypothesis………………………………………………………………….…13 1.2 The Constraints of Transfer or Transferability……………………………… 22 1.3 Transfer as a Production Procedure…………………………………………. 34 1.4 Transfer as a Reception Procedure………………………………………….. 45 1.5 Transfer as a Learning Procedure……………………………………………. 55 1.6 Avoidance……………………………………………………………………..59
2. RESEARCH ON L2 VOCABULARY ACQUISITION AND L2 LEXICO- SEMANTIC DEVELOPMENT…………………………………………………… 67 2.1 A Brief Overview of L2 Vocabulary Research…………………………….. 67 2.2 Receptive versus Productive Vocabulary…………………………………… 70 2.3 Learnability of L2 Vocabulary………………………………………………. 77 2.4 L2 Lexico-semantic Development………………………………………….. 85 2.5 Polysemy and Prototype…………………………………………………….. 99 2.6 Cognates……………………………………………………………………. 110 2.7 Bilingual Mental Lexicon…………………………………..………………. 121 2.8 Japanese Kanji Compounds………………………………………………… 130 2.9 Summary of Findings and Problems in the Existing Literature….………… 140
PART II: CHINESE BACKGROUND LEARNERS’ (CBLs’) USE AND ACQUISITION OF JAPANESE VOCABULARY…………………....143 3. THE ROLES OF TRANSFER IN CBLs’ USE AND ACQUISITION OF FOUR TYPES OF JAPANESE KANJI COMPOUNDS (STUDY A)……………. 145 3.1 Introduction………………………………………………………………… 145
v 3.2 Research Question and Hypotheses………………………………………… 147 3.3 Methodology………………………………………………………………... 148 3.3.1 Participants…………………………………………………………… 148 3.3.2 Instrument……………………………………………………………. 149 3.3.3 Procedures……………………………………………………………. 157 3.4 Results and Discussion…………………………………………………….. 157 3.4.1 Use and Acquisition of Good Cognates……………………………... 158 3.4.2 Use and Acquisition of Non-cognate Kanji Compounds….…………. 160 3.4.3 Use and Acquisition of Totally Deceptive Cognates………………….163 3.4.4 Use and Acquisition of Partially Deceptive Cognates……………….. 168 3.5 Summary of Results and Conclusion………………………………………. 185
4. THE ROLES OF TRANSFER IN CBLs’ USE AND ACQUISITION OF PARTIALLY DECEPTIVE COGNATES (STUDY B)…………………………. 190 4.1 Introduction………………………………………………………………… 190 4.2 Research Question and Hypotheses……………………………………….. 192 4.3 Methodology……………………………………………………………….. 193 4.3.1 Participants……………………………………………………………195 4.3.2 Instruments…………………………………………………………… 196 4.3.2.1 Oral Production Test………………………………………..... 196 4.3.2.2 Translation Test……………………………………...………. 197 4.4 Results and Discussion……………………………………………………... 199 4.4.1 CBLs’ Use and Acquisition of Partially Deceptive Cognates in the Oral Production Test…………………………………………. 201 4.4.1.1 Production of (J
vi 4.4.2.4 Comprehension of (J>C) in Unshared Meaning…...………… 246 4.5 Summary of Results and Conclusion…………………………………..…... 252
5. TRANSFERABILITY OF CHINESE POLYSEMOUS VERBS “(KAI)” & “ (KAN)” AND ACQUISITION OF JAPANESE CORRESPONDING WORDS BY CBLs (STUDY C)………………………………………………………………… 258 5.1 Introduction……………………………………………………………….... 258 5.2 Research Question and Hypotheses………………………………………... 263 5.3 Survery on the Prototypicality of Chinese Polysemy (Investigation 1)……. 264 5.3.1 Instrument and Procedures…………………………………………… 264 5.3.2 Participants…………………………………………………………… 266 5.3.3 Results………………………………………………………………... 267 5.4 Acceptability Judgement Test on Transferability of Chinese Polysemy (Investigation 2) ……………………………………………………..…….. 269 5.4.1 Instrument and Procedures…………………………………………… 269 5.4.2 Participants…………………………………………………………… 271 5.4.3 Results………………………………………………………………... 271 5.5 Oral Production Test on Transferability of Chinese Polysemy (Investigation 3)…………………………………………………………….. 276 5.5.1 Instrument and Procedures…………………………………………… 276 5.5.2 Participants…………………………………………………………… 277 5.5.3 Results………………………………………………………………... 277 5.6 Discussion…………………………………………………………………... 282
5.6.1 Transferability of Chinese Polysemous Verb “(kai)” and Acquisition of its Corresponding Words in Japanese…….....……….. 284
5.6.2 Transferability of Chinese Polysemous Verb “(kan)” and Acquisition of its Corresponding Words in Japanese…………...…… 291 5.7 Summary of Results and Conclusion………………………………………. 294
6. CONCLUSION…………………………………………………………………… 299 6.1 Summary of Findings…………………………………………………….... 299 6.2 Contributions to the Field…………………………………………………... 304 6.3 Limitations and Suggestions for Future Studies……………………………. 306 6.4 Pedagogical Implications………………………………………………….... 308
vii REFERENCES………………………………………………………...…………….. 313 APPENDIX………………………………………………………………………….. 334
LIST OF TABLES Table 1.1 Simplified Version of Stockwell, et al.’s (1965) Hierarchy of Difficulty.… 18
Table 2.1 Degree of Cross-linguistic Semantic Correspondence for Each Type of Japanese Kanji Compounds……………………………………………… 133
Table 3.1 Demographic Information of Participants……………..…………..……... 149
Table 3.2 Target Items in the Acceptability Judgement Test……………..………… 155
Table 3.3 Means and Standard Deviations for Good Cognate (S)………...………… 159
Table 3.4 ANOVA for Good Cognate (S)………………………………....………… 159
Table 3.5 Means and Standard Deviations for (NN) or Non-Cognate Kanji
Compounds whose Meanings are Not Comprehensible for CBLs…….… 160
Table 3.6 ANOVA for Non-Cognate (NN)……………………………....………..… 161
Table 3.7 Means and Standard Deviations for (NC) or Non-Cognate Kanji Compounds whose Meanings are Comprehensible for CBLs…………… 161
Table 3.8 ANOVA for Non-Cognate (NC)……………………………....………..… 162
Table 3.9 Means and Standard Deviations for (DER) or Totally Deceptive Cognates which are Erroneously Used in the Chinese Meaning …………………..… 163
Table 3.10 ANOVA for Totally Deceptive Cognates (DER)…………....………..… 163
Table 3.11 Means and Standard Deviations for (DCR) or Totally Deceptive Cognates which are Correctly Used in the Japanese Meaning ………………….… 165
Table 3.12 ANOVA for Totally Deceptive Cognates (DCR)…………....………..… 166
Table 3.13 Means and Standard Deviations for (OCR) or Partially Deceptive Cognates which are Correctly Used in the Japanese Meaning …………………..… 169
Table 3.14 ANOVA for Partially Deceptive Cognates (OCR)…………....………… 169
Table 3.15 Means and Standard Deviations for (OER) or Partially Deceptive Cognates which are Erroneously Used in the Chinese Meaning …………………..… 179
viii Table 3.16 ANOVA for Partially Deceptive Cognates (OER)…………....……….… 179
Table 4.1 Demographic Information of Participants………………………………... 196
Table 4.2 Target Items for Each Hypothesis and Test……………………………… 198
Table 4.3 Number and Percentage of the Participants who have Orally Produced Each Target Item (CBLs, EBLs and JNSs)……….………..………….…..203
Table 4.4 Number and Percentage of the Participants who have Orally Produced Each Target Item (CIMs and EIMs)…………………….…………………204
Table 4.5 Number and Percentage of the Participants who have Orally Produced Each Target Item (CADs and EADs)…………………………………..… 204
Table 4.6 Number and Percentage of the Participants who have Orally Produced Each Target Item (CBLs, EBLs and JNSs)……….………..………….…..210
Table 4.7 Number and Percentage of the Participants who have Orally Produced Each Target Item (CIMs and EIMs)…………………….…………………210
Table 4.8 Number and Percentage of the Participants who have Orally Produced Each Target Item (CADs and EADs)…………………………………..… 211
Table 4.9 Number and Percentage of the Participants who have Orally Produced Each Target Item (CBLs, EBLs and JNSs)……….………..………….…..219
Table 4.10 Number and Percentage of the Participants who have Orally Produced Each Target Item (CIMs and EIMs)…………………….…………...……219
Table 4.11 Number and Percentage of the Participants who have Orally Produced Each Target Item (CADs and EADs)……………………...…………..… 220
Table 4.12 Number and Percentage of the Participants who have Orally Produced Each Target Item (CBLs, EBLs and JNSs)……….………..……...….…. 224
Table 4.13 Number and Percentage of the Participants who have Orally Produced Each Target Item (CIMs and EIMs)…………………….………...…...… 224
Table 4.14 Number and Percentage of the Participants who have Orally Produced Each Target Item (CADs and EADs)………...………………………..… 225
Table 4.15 Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CBLs, EBLs and JNSs)…..…… 233
Table 4.16 Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CIMs and EIMs)………….…… 233
ix Table 4.17 Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CADs and EADs)……..….….... 234
Table 4.18 Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CBLs, EBLs and JNSs)…..…… 238
Table 4.19 Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CIMs and EIMs)………….…… 239
Table 4.20 Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CADs and EADs)…………...… 239
Table 4.21 Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CBLs, EBLs and JNSs)…..…… 242
Table 4.22 Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CIMs and EIMs)………….…… 243
Table 4.23 Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CADs and EADs)…...…….….. 243
Table 4.24 Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CBLs, EBLs and JNSs)….….… 247
Table 4.25 Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CIMs and EIMs)…………..….. 248
Table 4.26 Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CADs and EADs)………...…… 249
Table 5.1 The Mean Value of Nine-point Scale Prototypicality Judgement on Chinese Polysemy “(kai)”………………………..…………………….. 268
Table 5.2 The Mean Value of Nine-point Scale Prototypicality Judgement on Chinese Polysemy “(kan)”…………………….…………………….…. 268
Table 5.3 Results of the Lexicality Judgement Test (Transferability of “(kai)”)… 272
Table 5.4 Results of the Lexicality Judgement Test (Transferability of “(kan)”)... 273
Table 5.5 Frequency of Target Items Appeared in Eight Volumes of Japanese Textbooks (Japanese Words that Correspond to Each Usage of “(kai)”) 275
Table 5.6 Frequency of Target Items Appeared in Eight Volumes of Japanese Textbooks (Japanese Words that Correspond to Each Usage of “(kan)”)275
Table 5.7 Results of the Oral Production Test (Transferability of “(kai)”)………. 278
x Table 5.8 Results of the Oral Production Test (Transferability of “(kan)”)……… 279
Table 5.9 Participants’ Use of Words in Oral Production Test (Japanese Words that Correspond to Each Usage of the Chinese “(kai)”)……….. 281
Table 5.10 Participants’ Use of Words in Oral Production Test (Japanese Words that Correspond to Each Usage of the Chinese “(kan)”)……..... 282
LIST OF FIGURES
Figure 1.1 Færch & Kasper’s (1986a, 1987, 1989) Categorisation of Transfer in Production…………………………………………………. 35
Figure 1.2 Mechanism of Transfer in Learning and Transfer in Communication…… 58
Figure 2.1 Semantic Divergence between Hebrew Word noah and English Corresponding Words……………………………………………. 81
Figure 2.2 Japanese ESL Learners’ Understanding of the English Word “WEAR”…. 84
Figure 2.3 Receptive Use of Same Translation Pairs………………………………... 97
Figure 2.4 Productive Use of Same Translation Pairs………………………………... 97
Figure 2.5 Totally Deceptive Cognates……………………………………………... 113
Figure 2.6 Partially Deceptive Cognates (Inclusion)……………………………….. 113
Figure 2.7 Partially Deceptive Cognates (Overlapping)……………………………. 113
Figure 2.8 Partially Deceptive Cognates Between Japanese and Chinese………….. 133
Figure 2.9 Differing Patterns of CLI on (J>C) According to Task Type .………….. 139
Figure 2.10 Differing Patterns of CLI on (J Figure 3.1 The Interaction between L1 and Level for Good Cognates (S)………...... 159 Figure 3.2 The Interaction between L1 and Level for (NN) ………………..………. 161 Figure 3.3 The Interaction between L1 and Level for (NC) ………………………….162 Figure 3.4 The Interaction between L1 and Level for (DER) ………...…………….. 164 Figure 3.5 The Interaction between L1 and Level for (DCR) …………...………….. 166 xi Figure 3.6 Semantic Ranges of Four Target Items for (J>C)……………………….. 169 Figure 3.7 The Interaction between L1 and Level for (OCR) …………...………….. 170 Figure 3.8 Mean Scores of Correct Judgement for (OCR) or Partially Deceptive Cognate Used in Japanese Meaning Correctly………………………….. 171 Figure 3.9 Mean Scores of Correct Judgement for “差別(sa-betsu)” ……………….. 172 Figure 3.10 Mean Scores of Correct Judgement for “程度(tei-do)” ……….……….. 173 Figure 3.11 Mean Scores of Correct Judgement for “得意(toku-i)” ……….……….. 174 Figure 3.12 Mean Scores of Correct Judgement for “反対(han-tai)” .………………. 176 Figure 3.13 Semantic Ranges of Four Target Items for (J Figure 3.14 The Interaction between L1 and Level for (OER) …………...…..…….. 179 Figure 3.15 Mean Scores of Correct Judgement for (OER) or Partially Deceptive Cognates Used in Chinese Meaning Erroneously……………………….. 180 Figure 3.16 Mean Scores of Correct Judgement for “了解(ryou-kai )”.……………...180 Figure 3.17 Mean Scores of Correct Judgement for “問題(mon-dai)”.…………..…. 182 Figure 3.18 Mean Scores of Correct Judgement for “地方(chi-hou)”..…………..…. 183 Figure 3.19 Mean Scores of Correct Judgement for “関心(kan-shin)”..……………. 184 Figure 4.1 Hypotheses where Semantic Range of the Partially Deceptive Cognate is Wider in L1 (J Figure 4.2 Hypotheses where Semantic Range of the Partially Deceptive Cognate is Wider in L2 (J>C)………………………………………………………. 193 Figure 4.3 Semantic Patterns of the Target Items for Hypothesis 1………….……... 203 Figure 4.4 Semantic Patterns of the Target Items for Hypothesis 3….……………... 209 Figure 4.5 Semantic Patterns of the Target Items for Hypothesis 5………………… 218 Figure 4.6 CBLs’ Production Using L1-based IL in the Case of (J Figure 4.7 CBLs’ Production Using L1-based IL in the Case of (J>C)…………….. 223 Figure 4.8 Semantic Patterns of the Target Items for Hypothesis 7…………….….. 224 Figure 4.9 L1-based Lexical Creativity when CBLs Lack the Knowledge xii of L2-specific Meaning of (J>C) ……………………………………….. 228 Figure 4.10 Semantic Patterns of the Target Items for Hypothesis 2………………... 233 Figure 4.11 Semantic Patterns of the Target Items for Hypothesis 4………………... 238 Figure 4.12 Semantic Patterns of the Target Items for Hypothesis 6………………... 242 Figure 4.13 Semantic Patterns of the Target Items for Hypothesis 8………………... 247 Figure 5.1 Four Groups of Target Items Categorised in Terms of L1 Prototypicality and Potential of Positive Transfer………………...…………………….. 285 LIST OF ABBREVIATIONS SLA Second Language Acquisition CAH Contrastive Analysis Hypothesis CLI Cross-linguistic Influence L1 First Language L2 Second (or Nth) Language IL Interlanguage TL Target Language NL Native Language CBLs Chinese Background Learners CBGs Chinese Beginning Learners CIMs Chinese Intermediate Learners CADs Chinese Advanced Learners EBLs English Background Learners EBGs English Beginning Learners EIMs English Intermediate Learners EADs English Advanced learners JNSs Japanese Native Speakers xiii INTRODUCTION 0.1 Introduction There is a popular myth among learners of Japanese that Chinese background learners (hereafter CBLs)1 have an uncontested advantage over other language background learners in learning Japanese. The myth derives from the fact that with only a few exceptions, most kanji or logograph used in the Japanese language originated in the Chinese written language, and CBLs have pre-existing familiarity with the characters. CBLs starting to learn Japanese already possess the ability to recognise kanji as characters, and they potentially know the meaning of most of the Japanese kanji compounds and kanji words2 which comprise of meaning units, namely kanji characters. This argument won popularity and played a role in giving a false sense of security to CBLs and at the same time discouraging other learners who study Japanese along side CBLs in the same classrooms. Contrary to this “fact” and learners’ beliefs, however, many teachers of Japanese tend to be intolerant of CBLs’ use of their first language (hereafter L1) in learning Japanese as a second language (hereafter L2). This seems to be caused by teachers’ (either conscious or unconscious) belief that the learners’ use of L1 should be restrained due to its negative effects: L1 interference or negative transfer. In addition to teachers’ beliefs, previous studies suggest that CBLs’ use of their L1 in learning Japanese is generally a 1 Some researchers include the speakers of Mandarin as a second language or the speakers of some Chinese dialects as CBLs. In the present research, however, CBLs are referred to as the learners who speak Mandarin as a first language, and are educated in China or Taiwan at least till the end of their primary school. 2 In the present research, “kanji compounds” are defined as any Japanese words which are written in two or three kanji characters regardless of their origin or cognancy with Chinese counterparts (e.g. 政治, 都合), while “kanji words” refer to any non-Chinese origin words which are made up of both kanji and hiragana syllabics (e.g. 住む, 来る). 1 negative effect in their development3 of total Japanese proficiency (Hatasa1992; Matsunaga 1999). Based on the results of a kanji test, a reading test, and a grammar test, Hatasa (1992) argues that CBLs’ use of L1 knowledge only works at the initial stage of their learning Japanese vocabulary and reading comprehension, and the use of L2 knowledge produces better outcomes in the long run. Matsunaga (1999) also claims that CBLs’ transfer of kanji knowledge from their L1 is advantageous for comprehending the Japanese texts that contain many kanji words, while its effects are negative as far as oral and aural skills are concerned. Thus, teachers or researchers tend to view CBLs’ use of L1 as negative effects despite some positive effects in earlier stages of their learning vocabulary and reading comprehension. Seemingly, the results of these studies on CBLs’ use of L1 knowledge in learning Japanese contradict the myth or belief held by learners of Japanese, but the above studies are insufficient in understanding the nature of CBLs’ use of L1. That is, it is crucial to discuss CBLs’ L1 transfer separately according to each aspect of language due to the fact that some linguistic features are more susceptible to L1 transfer than others (Shirai 1992; Ellis 1994; Kellerman 1995; Jarvis 1997; Jarvis & Odlin 2000; Nakahama 2003): L2 phonology or lexicon are more likely to be affected by learners’ L1 than L2 syntax or morphology. 3 The term “development” is often used as a synonym of “acquisition” and “learning” (Sharwood Smith 1994), although some researchers use “acquisition” as informal modes of learning (e.g. Krashen 1976). However, despite their synonymous relationships, emphasis may be placed on different aspects: “development” focuses on the process itself, while the other two terms call attention to the locus of development (Sharwood Smith 1994). The present research adopts this distinction. 2 It has been claimed that L1 involvement cannot be avoided in learners’ L2 lexico- semantic development, because L2 word form will be automatically mapped to the L1 semantic equivalents (e.g. Jiang 2000, 2002, 2004a, 2004b). If the L1 involvement is an unavoidable process, it is critical for teachers of Japanese to consider how CBLs can effectively and positively utilise their L1 knowledge, rather than how to minimise their use of L1 at least in their lexico-semantic development of Japanese. In order to do so, it is necessary to understand the nature and condition of transfer involving CBLs’ use and acquisition of Japanese vocabulary more clearly. Although Japanese and Chinese share a number of kanji compounds as cognates, their meanings in each language are not always the same (Bunkachou 1978). Some of the shared compounds have slightly or completely different meanings. Moreover, there are also some kanji compounds that exist only in Japanese (and vice versa). In fact, numerous descriptive studies have reported the examples of errors made by CBLs or predicted the types of errors that they tend to make in their use of Japanese kanji compounds (e.g. Moriya 1979; Shi & Wang 1983; Miura 1984; Miyajima 1994; Ookouchi 1997; see Chen 2003a for review) or kanji words (e.g. Hishinuma 1980; Suzuki 1988). These contrastive studies on the lexical issues of Japanese and Chinese primarily argue that the semantic discrepancy between two languages is the cause of negative transfer errors (i.e. use of L2 words in the meaning of L1 counterparts or over-extension of the semantic range from L1 words to L2 equivalents). For example, it may not be uncommon for those who have taught Japanese to CBLs to hear such an erroneous utterance as “*問題がありますか？(Are there any *problems?)” after their oral 3 presentation or speech, while they meant to ask “Are there any questions?”. The source of the error has been considered to be learners’ ignorance of semantic discrepancy of the cognate pair between Japanese “問題(mon-dai) and its Chinese counterpart “ (wen-ti)4”. Even though the former is not appropriate in the above context, the latter is appropriate in the corresponding expression in Chinese “ (Are there any questions?)”. Even when the lexical items do not share the identical or similar forms between Japanese and Chinese, CBLs still produce errors which are considered to be caused by their L1. For example, Hishinuma (1980) reports some CBLs’ errors in their speaking or writing of Japanese such as “*テレビを開く(lit. *open TV)” and “*薬を食べる(lit. *eat medicine)” where they meant “turn on TV” and “take medicine” respectively. Hishinuma (1980) argues that despite their non-cognate status, the corresponding expressions in Chinese (i.e. “(lit. open TV)” or “ (lit. eat medicine)”) are the trigger of such errors. At a first glance, such arguments appear to be reasonable due to the fact that CBLs’ erroneous utterances in these examples are L1-like or subsume some semantic features of their L1. Yet, the nature of CBLs’ L1 transfer is still uncertain since a number of non-structural effects of transfer, which is induced or inhibited by factors other than the similarities or differences between learners’ L1 and L2 (e.g. task type or lexical prototypicality), have 4 Whereas simplified Chinese characters are used in mainland China, traditional forms are used in Taiwan and Hong Kong. The present research uses traditional forms of Chinese characters to describe Chinese words because of the relative similarity to Japanese kanji characters. The description of Mandarin Chinese in the Roman alphabet also varies in Mainland China (Chinese Pin Yin) and in Taiwan (Wade-Giles). In the current research, “Chinese Pin Yin” is used to describe the pronunciation of Chinese words due to its relative consistency with English pronunciation. 4 been identified (Odlin 1989; Shirai 1992; Ellis 1994; Kellerman 1995; Jarvis 1997, 2000). Thus, it is possible that what contrastive studies have predicted is simply one of the factors that trigger their L1 transfer. In addition, Kellerman (1983) claims that interference or negative transfer is only the tip of the iceberg as evidence for the L1 influence on L2. It is essential to consider not only negative transfer but also positive transfer and other types of cross-linguistic influence (hereafter CLI)5 in order to fully understand the effects of L1 on CBLs’ use and acquisition of Japanese vocabulary. Despite the abundance of descriptive studies that focus on negative aspects of CBLs’ lexical transfer, studies arguing CBLs’ positive transfer in learning Japanese vocabulary have been rare (e.g. Bunkachou 1978; Takebe 1979; Matsushita 2001, 2002), and it appears that no empirical research has explored this area. This is not surprising given the fact that positive transfer has been largely neglected in the entire field of second language acquisition (hereafter SLA) (Ringbom 1987, 1991, 1992; Ellis 1994; Jarvis & Odlin 2000; Odlin 2003). Ringbom (1991) asserts the importance of studies on positive transfer in relation to transfer in comprehension because transfer is primarily a facilitating, not inhibiting, effect on learning. Whereas two empirical studies have attempted to examine the L1 influence on the acquisition of Japanese-Chinese cognates both by CBLs in their learning Japanese (Chen 2003b) and by Japanese native speakers (hereafter JNSs) learning Chinese (Miura 1997), both of these are found to have quite a few methodological problems. The largest problem of these studies is that they studied learners of only one L1 5 In the present research, I will follow Odlin (2003), and use the term “cross-linguistic influence (CLI)” synonymously and interchangeably with “transfer”. Definition of terms will be given in Chapter 1. 5 background, and thus did not compare the test results with learners of other L1s. For this reason, the claimed L1 influence is not convincing because other factors relevant to ease or difficulty of acquisition (e.g. intralingual and developmental factors) may also have been at play in their results of multiple-choice test. In order to identify L1 influence on learners’ use and acquisition of L2 more rigorously, Jarvis (1997, 2000) proposed three potential effects of L1 influence that must be considered. They are: (1) intra-L1 group homogeneity in learners’ IL performance, (2) inter-L1 group heterogeneity in learners’ IL performance, and (3) intra-L1 group congruity between learners’ L1 and IL performance (IL=interlanguage). (Jarvis: 2000: 253) In detecting L1 influence or language transfer, it is not enough to find similarities between the learners’ L1 performance and that of interlanguage (hereafter IL). Identification of transfer requires finding similarities in the IL performance of learners who share the same L1, and differences in the IL performance of learners of different L1s. On the basis of Odlin (1989), Nakahama (2003) also suggests three important conditions for testing CLI. First, it would be ideal to examine the IL of at least two groups of learners of typologically different L1s: one distant from and one close to their L2. Second, a comparison of learners from different L1 backgrounds at different proficiency levels is crucial in distinguishing developmental errors from L1-induced 6 errors. That is, it allows us to understand the differing patterns of language development according to learners’ L1 by expecting more target-like performance for those learners whose L1 shares the same linguistic features as the L2. If learners from different L1 backgrounds commonly show difficulty with a particular L2 item, it is possible that developmental factors are affecting their acquisition of that item more strongly than transfer factors are. Lastly, it is essential to analyse learners’ L1 (i.e. close comparison of the similarities and differences between their L1 and L2), and performance of native speakers (of their L2) in order to understand learners’ both L1-like and target-like utterance of the particular linguistic items. Based on these methodological frameworks regarding the identification of language transfer or CLI (Odlin 1989; Jarvis 1997, 2000; Nakahama 2003), the present research uses a group of English background learners (hereafter EBLs) of Japanese of different proficiency levels as well as JNSs in order to compare their performance with that of the target group, namely, CBLs. Although it may be indisputable that Japanese and English are typologically distant (e.g. Nakahama 2003), Japanese and Chinese have also been claimed to be typologically unrelated in spite of numerous etymologically related lexical items (e.g. Matsumi 2002). Since the present research specifically deals with Japanese kanji compounds and kanji words in which Japanese and Chinese share not only logographic orthography but also a large number of cognates, the target items to be investigated are surely comparable between two different L1s of learners. That is, one (Chinese) bears a close resemblance to the L2 (Japanese), and the other (English) is totally dissimilar to the L2 (Japanese). 7 Another issue that needs to be noted with regard to learners’ language background is that all the CBLs who participated in the present research are students or graduates of Australian universities, and thus, strictly speaking, Japanese is their third (or fourth) language; English being their L2 (or their third language). However, according to Sharwood Smith (1994), “second language (L2)” is normally used as a cover term for any language other than the L1 that is currently being learnt by particular learners irrespective of the number of other non-native languages possessed by them. Therefore, the present research will follow this, and use “L2” synonymously and interchangeably with “target language (hereafter TL)”. Although there is no guarantee that CBLs do not transfer their knowledge of English in their use and acquisition of Japanese vocabulary, previous studies such as those quoted below suggest that such cases should be rare if there are any. Kellerman (1983) argues that learners’ perception of language distance or psychotypology constrains the occurrence of transfer: if learners perceive certain L2 as close to their L1, it will promote transfer, whereas it will inhibit transfer if they perceive the L2 as distant. Ringbom (1978, 1987) empirically shows this in his investigation of both Finnish-speaking Finns with Swedish as L2 and Swedish-speaking Finns with Finnish as L2 in their acquisition of English. The results indicate that both groups transferred from Swedish to English but not from Finnish regardless of whether Finnish was their L1 or L2. Ringbom (1978, 1987) maintains that this is due to the typological proximity of each language to English. That is, Swedish is one of the Germanic languages, and typologically similar to English, while Finnish is typologically unrelated to English. Thus, it is more likely that the close similarity of lexical items between 8 Japanese and Chinese facilitate CBLs’ lexical transfer from Chinese to Japanese, but not from English to Japanese. The reason that English-speaking CBLs of Japanese, instead of CBLs with no knowledge of additional languages other than Japanese, were chosen as participants of the present research, despite its shortcomings (i.e. knowledge of English as an uncontrollable variable), is because of its methodological advantages. Since the current research examines CBLs’ use and acquisition of Japanese-Chinese cognates focusing on their semantic aspects, use of another language than their L1 and TL is extremely beneficial in giving cues or clarifying the meaning of each target item. More specifically, the target items of the present research include deceptive cognates6, thus it is predicted that the meaning of the target items in a comprehension task may be equivocal with the context given in the TL alone (i.e. both Japanese meaning and Chinese meaning of the deceptive cognate may make sense in the context). It may also be that the context given solely by TL may be too difficult for less proficient learners. In order to cope with such problems, English translations can be given to clarify the meaning of each sentence and target item. Moreover, if the cue is provided in their L1, it may work as stimuli for their automatic response in cognates. Use of English as a medium language also enables CBLs to conduct the same tests as EBLs, which is a prerequisite in the comparison of the results between learners of different L1s. This issue will be discussed in more details in the methodology sections (see Chapter 3, Chapter 4 and Chapter 5). 6 “Deceptive cognates” or “false cognates” are referred to as the cognates which share the same or similar orthography (or sound) between two languages but have divergent meanings such as English “fabric”and French “fabrique (factory)”. Detailed classification and discussion of cognates will be given in Chapter 2. 9 0.2 Objectives of the Present Research Assuming that there are differences in the rate of transfer or the level of difficulty for each lexical item, what kinds of items are more likely to be transferred than the others? What kinds of Japanese kanji compounds and kanji words are easier or more difficult to learn for CBLs? Does their transferability or difficulty for CBLs vary according to the task type or proficiency level of learners? These questions were the motivation for this project. The major objective of this research is to examine the roles of transfer and the difficulty in CBLs’ use and acquisition of Japanese kanji compounds and kanji words. In particular, under what conditions and in what ways do CBLs transfer Chinese words into their Japanese counterparts are investigated. Another important goal of the current research is to attempt to ascertain what variables may override or obscure the occurrence of transfer. The variables that will be examined are (1) cross-linguistic semantic conditions of lexical items (i.e. match or mismatch), (2) task type, (3) learners’ L2 proficiency level, (4) frequency of the L2 input, (5) context the word is embedded in and (6) prototypicality of L1 lexical items. In order to achieve the above objectives, I will conduct three types of tests which include both receptive and productive measures: an oral production test, a lexicality judgement test and a translation test. The results of these three tests will be compared among CBLs, EBLs and JNSs in order to distinguish the influence of transfer from other factors. 0.3 Organisation of Chapters This thesis is organised into two main parts. Part I is a review of related research, and consists of Chapter 1 and Chapter 2. In Chapter 1, I will briefly review the history of 10 transfer research mainly in relation to the Contrastive Analysis Hypothesis, followed by research on the constraints of transfer or transferability. Subsequently, I review research on transfer in three dimensions, i.e., production, reception and learning. I also review “avoidance” as an important phenomenon of CLI. Chapter 2 is a review of the relevant literature on L2 vocabulary acquisition and L2 lexico-semantic development. It includes a brief overview of L2 vocabulary acquisition studies as well as a more detailed review of each area of research that is relevant to the present research. In Part II, I will present three independent but closely related studies on CBLs’ use and acquisition of Japanese vocabulary. Though a cohesive whole, each study benefits from distinct research questions, methodology, results and discussion. Chapter 3 (Study A) explores the roles of transfer in CBLs’ use and acquisition of four types of Japanese kanji compounds: three of them are Japanese-Chinese cognates with differing degrees of semantic overlap between the two languages. On the basis of the results of Study A, Chapter 4 (Study B) focuses on one type of Japanese-Chinese cognates, i.e., partially deceptive cognates, which were found to be the most difficult and problematic for CBLs. In Chapter 5 (Study C), I will attempt to investigate CBLs’ acquisition of Japanese words that correspond to each usage of the Chinese polysemous verbs “(kai)” and “(kan)” in relation to their transferability. Following these three studies, in Chapter 6, I summarise the results of each study to display the general findings and implications of the thesis. I also describe limitations of the entire thesis and conclude by discussing its pedagogical implications. 11 PART I: REVIEW OF RELATED RESEARCH In Part I, I will review the previous literature relevant to the present research. The two major areas of study which are most relevant to the current research are investigations into language transfer and L2 lexico-semantic development. In addition, as the main objective of the present research is to examine CBLs’ use and acquisition of Japanese kanji compounds and kanji words, I will also review related contrastive studies between Japanese and Chinese, particularly on cognates. First, a brief historical background of transfer research in relation to the Contrastive Analysis Hypothesis will be provided, followed by a more detailed review of research on language transfer or CLI. After reviewing research on the constraints of transfer or transferability, I will discuss three dimensions of transfer: transfer as a production procedure, transfer as a reception procedure, and transfer as a learning procedure. Additionally, I will also review another type of CLI, namely avoidance, due to its relevance to the current research. The remainder of Part I focuses on studies of L2 vocabulary acquisition and L2 lexico- semantic development. In reviewing the L2 vocabulary acquisition research, I will first explain its past and current status in the field of SLA studies, and areas of the research that includes. Subsequently, I will discuss more specific areas of L2 vocabulary research which are relevant to the present research. They include research on (1) receptive versus productive vocabulary, (2) learnability of L2 vocabulary, (3) L2 lexico- semantic development, (4) polysemy and prototype, (5) cognates, (6) bilingual mental lexicon, and (7) Japanese kanji compounds. Finally, I will summarise the findings and problems in the existing literature. 12 CHAPTER ONE RESEARCH ON LANGUAGE TRANSFER 1.1 A Brief History of Transfer Research and the Contrastive Analysis Hypothesis This chapter will begin with a brief historical review of transfer research in relation to the Contrastive Analysis Hypothesis (hereafter CAH). This will be followed by a more detailed discussion of transfer research in general7 and explicitly how it relates to the present research. In reviewing the history of transfer research as well as SLA research, there are two noteworthy stages prior to the advent of IL study: contrastive analysis and error analysis. Fries (1945) provided the pioneer concept that later influenced Lado’s (1957) work on CAH, popular until the early1970s. Fries indicated that the most efficient materials for learners were those that were based upon a scientific description of the language to be learnt, carefully compared with a parallel description of the native language (hereafter NL)8 of the learner. This view suggests that learners’ errors in their TL arise from differences between their L1 and L2. This contention is more clearly stated in Lado (1957: 2): 7 There are a number of studies that refer to language transfer from the perspective of Universal Grammar. However, they specifically deal with “syntax” or grammar, and their main interests are not in the transfer itself. Thus, they are not within the scope of this review. 8 As the term “native language (NL)” and “first language (L1)” is often used as an identical concept, and interchangeably among many authors of SLA, the current research will follow this. As I mentioned earlier, the term “target language (TL)” and “second language (L2)” will also be used interchangeably to indicate the language that learner is currently learning, regardless of the number of the languages the learner already speaks. 13 We assume that the students who come into contact with a foreign language will find some features of it quite easy and others extremely difficult. Those elements that are similar to his native language will be simple for him, and those elements that are different will be difficult (cited by Odlin 1989). CAH assumed that close comparison between the TL and the NL of the learners would elucidate their learning problems; hence contrastive analysis was largely carried out in order to predict the errors that learners tend to make and to decide the content of teaching materials. Although contrastive analysis played an important role in the history of transfer research (and that of SLA research at the same time), the claims made by Fries and Lado have since been challenged. By the 1970s, several empirical studies showed that learning difficulties do not always arise from the differences between learners’ L1 and L2, and the difficulties are not always predicted by the CAH (e.g. Briere 1968; Whitman & Jackson 1972, cited by Shirai 1992). By this time, many SLA researchers had also begun to consider the nature of L1 as one of the many possible causes of learning problems. In the 1970s, the same kinds of errors were found in learners of different L1s (e.g. Dulay & Burt 1973, 1974), and researchers found out that L1 transfer cannot account for every error that learners make. As a result of their investigation into Spanish speaking children learning English, Dulay & Burt (1973) concluded that a learners’ L1 has almost no influence on their L2 acquisition. Their claim shed light on the existence of systematic or developmental errors that every learner tends to make regardless of his/her language background. Richards (1971) called these errors intralingual and developmental errors in opposition to the interlingual errors that were caused by the interference of the learners’ L1. As a 14 consequence, contrastive analysis was seen as less valid and error analysis gained importance among applied linguists. One of the most serious problems of the CAH was that it was based on behaviourist theories, even though not all the advocates of the CAH had behaviourist views (Odlin 1989, 2003). According to Gass & Selinker (1994), behaviourists viewed language learning as a habit formation, and thus many proponents of the CAH assumed that learning a new language largely consisted of acquiring new “habits”. As a result, L1 transfer was regarded as the use of old habits which prevent the formation of TL habits, although these behaviourist views have been discredited now (Ellis 1994). Another serious problem with the CAH was that it failed to acknowledge sources of difficulty other than the learners’ L1, hence it was too simplistic and restrictive (Ellis 1994; Gass & Selinker 1994). In other words, CAH overpredicted the occurrence of negative transfer by claiming that where there were differences in the corresponding items or structures between learners’ L1 and L2, interference would inevitably occur. However, some researchers, so-called “minimalists”9 such as Dulay & Burt (1973, 1974), have seriously underestimated the roles of the L1 in the acquisition of L2. Ellis (1994) argues that one of the reasons for their underestimation is likely to be due to their own theoretical biases. As it was not easy to determine whether certain errors were caused by L1 transfer or the universal nature of language development10, minimalists tended to estimate transfer errors minimally or too conservatively. However, now there is clear evidence that the L1 plays a major role in L2 acquisition (Ellis 1994). Jarvis 9 The term “minimalists” is Ellis’ (1994) term, and has nothing to do with “minimalist program” used among Chomskyan Linguists. 10 Ellis (1994) states that CLI is often difficult to ascribe due to the fact that errors deriving from L1 frequently correspond with learners’ developmental errors. Odlin (1989) also shares a similar view. 15 (1997) also claims that the primary problem with the CAH does not concern contrastive analysis itself, but its formation and theoretical underpinnings. In fact, a few modified versions of the CAH have been proposed to complement the original or “strong” version of the CAH put forwards by Lado (1957). The strong version of the CAH is often compared with the later “weak” version (e.g Wardhaugh 1970) as in “a priori versus a posteriori” or “predictive versus explanatory” (Gass & Selinker 1994; Ellis 1994). That is, the strong version maintains that one could make predictions about learning difficulty based on a comparison between two languages, while the weak version starts with an analysis of learners’ recurring errors, and attempts to account for those errors on the basis of NL-TL difference (Gass & Selinker 1994; Larsen-Freeman & Long 1991). Although the latter might not have predictive power, it was still claimed to have a posteriori explanatory power (Gass & Selinker 1994; Larsen-Freeman & Long 1991). Thus, it came to be part of error analysis, which is more useful in detecting the source of error in a broader approach by analysing errors of other sources as well (Larsen- Freeman & Long 1991). However, some researchers held a different view, i.e., contrastive analysis is only worthwhile if it is predictive (Schachter 1974; James 1980). From such a perspective, the strong version was theoretically untenable, and the weak version was impractical and inadequate (Ellis 1994). A “moderate” version of the CAH is put forward by Oller & Ziahosseiny (1970). This assumes that learners’ perceived similarities and differences are the basis for learning, and thus wherever patterns are minimally distinct in form or meaning in one or more 16 systems, confusion will arise. That is, learning (in phonology, syntax and semantics) will be the most difficult where the most subtle distinctions are required either interlingually and intralingually. Based on their study, Oller & Ziahosseiny found that the participants whose L1 used the Roman alphabet made more spelling errors, which was caused by interference from their L1 spelling, than the participants whose L1 did not use the Roman alphabet. Hence, they concluded that knowledge of the Roman alphabet made it more difficult to learn another Roman writing system. However, Ringbom (1987) contends that Oller & Ziahosseiny’s (1970) moderate version of the CAH is based on a study of spelling errors, and therefore their claim cannot be generalised to all learning situations. Ringbom (1987) points out that spelling is a highly specific task in which there is usually only one correct answer, and something similar is regarded as an error. Comprehension does not always require full knowledge of the L2 form or meaning because a fair amount of understanding can be reached by successful inferencing (Ringbom 1987). Thus, understanding L2 language items where subtle distinctions are required are not always difficult in comprehension tasks due to its approximate nature. However, the moderate version of the CAH may be predictive and useful as far as learners’ L2 production is concerned particularly when a high level of accuracy is required. Although there seems to be no dispute with regard to the above criticism of each version of the CAH, some supporting evidence has also been shown for the CAH. Larsen-Freeman & Long (1991) point out that not all of the contrastive analyses were merely lists of binary predictions of the form (i.e. similarity / difference = ease / difficulty), and Stockwell, et al.’s (1965) “hierarchy of difficulty” is considerably more 17 sophisticated than the strongest version of the CAH (Lado 1957). Table 1.1 shows a simplified version of Stockwell, et al.’s (1965) hierarchy of difficulty which is adapted to the case of acquisition of Japanese by English speakers. According to Stockwell, et al. (1965), the greatest difficulty for learners would be predicted in “split” where a single form in the L1 is manifest as two or more in the L2, followed by “new” where a form exists in the L2 but not in the L1. A slightly easier category is “absent” where a form exists in the L1 but is absent in the L2, and this is followed by “coalesced” which is the opposite of “split”, i.e., where several forms in the L1 are represented by one form in the L2. The least difficult category is “correspondence” where there is a complete match in certain linguistic items between two languages. As Larsen-Freeman & Long (1991) claim, the most significant difference between Stockwell, et al. (1965) and Lado (1957) is that the former do not predict the greatest difficulty in “new” and “absent” where the differences between the two languages may be largest. Table 1.1: Simplified Version of Stockwell, et al.’s (1965) Hierarchy of Difficulty (Based on Information Given in Shibuya (2002)) Type of Difficulty L1: English L2: Japanese English Japanese 1.Split x to exist aru (inanimate) x iru (animate) y 2.New ø ———— x—wa (topic marker) 3.Absent x ————ø plurality / article — 4.Coalesced x present perfect ~teiru x “have-en” y present progressive “be -ing” 5.Correspondence x ————x simple past “-ed” ~ta 18 Unlike many of the studies arguing against the CAH, Shirai (1992) indicates that there are some empirical studies that support Stockwell, et al.’s (1965) claim in the area of lexico-semantics using a modified version of the hierarchy of difficulty (e.g. Tanaka 1983; Takahashi 1984). Shirai also referred to other studies in which corresponding L1 systems in an article (Master 1987) or in plural morphemes (Young 1990) facilitated learning of the L2. Accordingly, Shirai (1992: 111) states: We thus cannot deny the possibility that the CA hypothesis may still have some validity in a priori prediction of difficulty at the level of certain linguistic items, if other variables can be controlled. In sum, despite researchers’ various attempts to predict learners’ difficulty or to explain their errors on the basis of careful comparison between their L1 and L2, no version of the CAH appears to be successful in predicting learners’ difficulty or explaining their errors in all learning situations. This may be largely owing to the fact that L2 acquisition is an extremely complex process affected by numerous variables such as linguistic, social, and psychological factors (Gass & Selinker 1994; Jarvis 1997). Contrastive analysis of learners’ L1 and L2 alone can neither predict the difficulty that learners’ face nor explain the complex process of their learning. However, whereas the central notions of the CAH (Lado 1957), namely the “difference” and “difficulty” equation, have been widely rejected, the solution lies not in its abandonment, but in careful revision and extension (Odlin 1989; Ellis 1994) or in utilising it as one of the methodological options (Larsen-Freeman & Long 1991). Therefore, although the present research does not assume that structural discrepancy 19 across languages or L1 influence is the only factor contributing to learners’ difficulty, or affecting L2 acquisition, it still does acknowledge the importance of comparing linguistic features between learners’ L1 and L2 as one of the methodological options in detecting CLI. As discussed above, in earlier attempts to predict learners’ difficulty, it was erroneously equated with the differences between their L1 and L2. However, as Ellis (1994) puts it, while it is incorrect to say that L1/L2 differences will lead to difficulty, it is also clear that they might do. Indeed, whereas L1 transfer has been one of the most central issues in the field of SLA research, there have been large discrepancies in the previous findings of the field. Jarvis (2000: 246) states: Perhaps no area of second language research has received as much attention and remained as elusive as the influence of the first language (L1). Despite the myriad studies that have been conducted in this area over the past four decades, there still remains a surprising level of confusion in the field concerning when, where, in what form, and to what extent L1 influence will manifest itself in learners’ use or knowledge of a second language (L2). For example, Dulay & Burt (1973) investigated Spanish speaking children in their acquisition of English, and reported that only 3% of total errors were due to language transfer. On the contrary, Tran-Thi-Chau (1975) reported that 51% of total errors were due to language transfer after investigating adult Chinese speakers in their acquisition of English. As Gass & Selinker (1994) state, the question of the L1 transfer had been posed dichotomously: is language transfer of major importance in forming IL or is it 20 not? By the late 1970s, however, researchers’ interest in the role of L1 had been shifted from dichotomous perspective to more sophisticated one such as “when” and “under what condition” does transfer take place (Gass & Selinker 1994). The reasons for the discrepancy in the results of previous research on transfer, apart from researchers’ theoretical biases, may include the differences in experimental design between different studies (e.g. language level, task type, learners’ ages, L1-L2 typological proximity), and the lack of well-defined and broadly-accepted criteria for the language transfer (Ellis 1994). Ellis (1994) claims that transfer is not simply a matter of interference or falling back on the NL as is closely associated with behaviourist theories, nor just a question of the influence of the learners’ NL, since other previously acquired languages can also have an effect. For this reason, Kellerman & Sharwood Smith (1986) coined a superordinate term “crosslinguistic influence (CLI)” which was theory-neutral and included “transfer”, “interference”, “avoidance”, “borrowing” and other phenomena. Nevertheless, the term “transfer” has persisted, because usage does not always conform to reason, as is often the case in terminological disputes (Ellis 1994). Dechert & Raupach (1989) indicate that the term “transfer” has been used with 17 different definitions, but Selinker (1992) stresses that “language transfer” can be considered as a cover term for a whole class of behaviours, process, and constraints. One of the most widely acknowledged definition of transfer may be Odlin’s (1989: 27): Transfer is the influence resulting from similarities and differences between the target language and any other language that has been previously (and perhaps imperfectly) acquired. 21 Although this definition is somewhat vague, it is broad enough and provides an adequate basis for many aspects of CLI (Ellis 1994). For this reason, the term “transfer” in the present research will be used as a cover term for many aspects of CLI as in Odlin (1989: 27). At the same time, like recent studies on transfer (e.g. Odlin 2003; Nakahama 2003), the term “transfer” will be used synonymously and interchangeably with “cross- linguistic influence (CLI)”. 1.2 The Constraints of Transfer or Transferability Among the SLA researchers’ strong interest in the study of language transfer, one of the most frequently discussed issues has been the constraints of transfer or transferability (Jarvis & Odlin 2000). Studies on transferability were pioneered by Kellerman in his work on the transfer strategy of idiomatic expressions (1977) and on polysemy “BREAK” (1978, 1979). Kellerman (1978, 1979) considered that the transferability of a certain language item was determined by how learners perceived the item. Thus, Kellerman (1983) states that, unlike the CAH, predictions of transfer need to be based not only on the structure of L1 and L2, but also on the learner’s perception of how the structures of L1 are perceived by the learner as being either transferable or non- transferable, because language transfer is not an “all or nothing” phenomenon. A series of Kellerman’s studies on “BREAK” (1978, 1979) reveals that there are three constraints on the learners’ transfer of their L1 knowledge. 22 The first constraint concerns the learner’s perceptions of “coreness” or markedness of an L1 item. That is, if a certain L1 item is perceived as core or prototypical11 by the learner, it is likely to trigger transfer. On the other hand, if a certain L1 item is perceived as non-core or peripheral, it is not a good candidate for transfer. Kellerman (1978) asked the Dutch learners of English to decide whether the 17 examples of Dutch polysemy “BREKEN” could be translated by English equivalent “BREAK”. As a result, Dutch learners of English were willing to translate a sentence like “Hij brak zijn been. (He broke his leg.)”, which was considered to be prototypical usage of the word, directly into English using “broke” for “brak”. However, they were not prepared to give direct translation to a sentence like “Het ondergrondse verset werd gebroken.(The underground resistance was broken.)”, which was considered to be less prototypical usage, although this translation was possible. I will discuss this issue more extensively in the next chapter. The second constraint concerns the learners’ perception of L1-L2 distance or psychotypology. This indicates that if learners perceive a particular L2 as close to their L1, it will facilitate transfer, while discourage transfer if they feel the L2 is distant from their L1. A good example of this is Ringbom’s (1978, 1987) studies which examined the acquisition of English by Finnish-speaking Finns with Swedish as L2, and Swedish- speaking Finns with Finnish as L2. He found that both groups transferred from Swedish to English but not from Finnish, no matter whether Finnish was their L1 or L2. This result can be easily explained by the notion of psychotypology because Swedish is one of the Germanic languages, and shares plenty of linguistic aspects with English, 11 Kellerman (1986) says that the term “markedness” used in his earlier studies is equivalent to “prototypicality”, i.e., unmarked items are prototypical, and marked items are less prototypical. 23 whereas Finnish is not typologically related to English. Kellerman (1979) also argues that learners’ psychotypology can be revised as they obtain more information about the TL: learners may assume that the TL is very similar to their L1 in earlier stages, but later they may come to perceive it very different or vice versa. The third constraint is what Kellerman (1979) refers to as the learner’s actual knowledge of the L2. Learners could benefit from positive evidence of the transferability for the particular items if they were exposed to L2 input (Kellerman 1995), hence intuitions about transferability are prone to developmental change (Kellerman 1984). That is, when correct knowledge of a certain L2 item is available for learners, they can make a decision regarding the choice of the L2 items based on their L2 knowledge, and thus learners do not need to rely on their L1 knowledge in compensation for the lack of their L2 knowledge. Despite their considerable contributions to the studies of transfer or SLA, limitations to Kellerman’s studies on transferability have also been suggested (Færch & Kasper 1987; Kellerman 1984, 1986; Meara 1984; Ellis 1994; Shirai 1995). Færch & Kasper (1987) claim that these studies explore learners’ transferability competence or what they believe or do not believe to be transferable rather than their decision making in specific communicative situations. Kellerman (1984, 1986) himself also notes that the main problem with these studies lies in how they can be related directly to learners’ spontaneous performance, since the model proposed in his studies will not necessarily predict second language performance. 24 Meara (1984) states that only a single set of idioms based on one verb (i.e.“BREAK”) is discussed, and only three languages (i.e. English, Dutch and German) are involved in Kellerman’s studies. Thus, it is unknown whether the similar principles would apply to the languages which are typologically distant such as English and Chinese. Meara (1984) further claims that it is not clear to what extent this technique could be extended to other areas of lexicon study due to the fact that idioms make up only a limited part of our language use. It has also been argued that Kellerman’s studies on transferability are mostly based on positive transfer items (Ellis 1994; Shirai 1995). Thus, it is unclear whether or not learners still show the similar transferability on the negative transfer items or potentially non-transferable items. Finally, Ellis (1994) also points out that the study needs to be careful to equate “translatability” with “transferability” given that the participants of Kellerman’s (1978) transferability experiment were only asked whether or not each usage of Dutch “BREKEN” could be translated into English “BREAK”. In addition to Kellerman’s works on transferability, there also exist some studies which view the constraints of transfer in a fundamentally different way. Kellerman (1979) emphasises that the transferability of a certain item (based on prototypicality) is determined solely by the L1, and is thus independent of the nature of the L2, although it may interact with another constraint: psychotypology. In contrast to Kellerman’s approach which places L1 as a transfer-determining factor, Andersen (1983) views transferability more as related to the formal properties of the L2. On the basis of previously suggested language acquisition principles (e.g. Slobin 1973; Zobl 1980), Andersen (1983) claims that L1 transfer interacts with the process of generalising structure from the L2 input in his “Transfer to somewhere principle”. 25 There are two general rules in this principle. First, any linguistic element transferred from the L1 to the L2 must not violate the “natural acquisition principle” (Andersen 1983). In other words, no L1 feature will be transferred to the L2 until learners have arrived at a certain stage of development where they are ready to learn the particular L2 items. Second, L2 input must provide learners with an impetus for generalising that certain L2 features work in the same way as L1 corresponding elements (Andersen 1983). That is, learners must perceive that certain linguistic elements are congruent or similar between their L1 and L2, and therefore they can be used in the same way. To put it another way, when learners are confronted with evidence that certain L2 structures share some features with their L1 in the form of L2 input, they might assume that transfer is possible in that particular structures. Thus, Andersen (1983) stresses the important role of frequency of the L2 input as a requirement for transfer: the more frequently a form occurs in the L2 input, the more likely it is to be transferred and incorporated into the IL. Whereas Andersen (1983) argues the role of frequency of L2 input in the occurrence of transfer, other studies discuss the role of L2 input frequency in the acquisition of a particular item (e.g. Ellis 2002; Gass & Mackey 2002). These studies suggest that input frequency generally plays an important role in the L2 acquisition of certain linguistic items, but input frequency alone is not sufficient to account for the acquisition of every language item. For instance, definite articles are always frequent in English, but they are not necessarily learnt easily or completely in early stages by ESL learners (Ellis 2002). Gass & Mackey (2002) argue the complexity of the task of deciphering the role and importance of frequency by giving examples of areas where acquisition proceeds at its 26 own pace regardless of the frequency of the input, and where input is infrequent but acquisition is unimpeded. In their conclusion, Gass & Mackey (2002) assert that we must investigate how frequency effects interact with other aspects of the SLA process particularly between frequency of L2 input and the role of the L1. Considering these claims, it appears that frequency of the L2 input may affect L2 acquisition in mainly two different ways. One is more direct influence, i.e., the more frequently certain items appear, the earlier learners acquire them, despite its interaction with other factors including CLI. The other is more indirect influence by promoting transfer, given that transfer also affects L2 acquisition by interacting with other factors. As Kellerman (1995) argues, a number of studies indicate that typological similarities rather than differences between learners’ L1 and L2 stimulate transfer (e.g. Kellerman 1978, 1979, 1983, 1986; Ringbom 1978, 1987; Zobl 1980; Andersen 1983; Shirai 1992). This is the opposite of what the strong version of the CAH (Lado 1957) has predicted, however. Whereas numerous non-structural factors (e.g. personality, age or task type) have been identified, certain degrees of structural congruence are still considered as a necessary condition for the occurrence of transfer (Selinker 1992; Gass & Selinker 1994). This is because the main effect of transfer is that learners can find the TL for items and structures that are L1-like (Selinker 1992). Odlin (2003) also indicates that what is considered CLI often depends on interlingual identifications, i.e., learners’ judgements that something in their L1 and something in the L2 are similar. Therefore, a constraint of transfer could be anything that prevents learners either from noticing a similarity itself or from deciding that the similarity is useful (Odlin 2003). 27 Yet, despite the strong support of the notion of similarity as the main driving force behind CLI, exceptions have also been suggested (Sridhar & Sridhar 1986; Odlin 1992). Accordingly, Kellerman (1995) proposed the “Transfer to nowhere principle” which suggests transfer may be triggered not only by similarities but also by differences between L1 and L2. Whereas Andersen’s (1983) “Transfer to somewhere principle” focuses on the conditions that will induce learners to make interlingual identifications, this principle considers how CLI can occur even when there is no basis for interlingual identification (Odlin 2003). However, Odlin further claims that although Kellerman (1995) intends the “Transfer to nowhere principle” to complement the “Transfer to somewhere principle”, the examples that Kellerman offers as instances where the “Transfer to somewhere principle” cannot account for are actually able to be explained by it. For this reason, Odlin (2003) concludes that the “Transfer to nowhere principle” is superfluous. However, we need more research to judge whether the “Transfer to nowhere principle” is valid or not since other studies do support it (e.g. Jarvis 1997, 2000). In addition to the structural similarity or difference between learners’ L1 and L2, sociolinguistic factors also have been claimed to influence the conditions of transfer. Odlin (1989) maintains that negative transfer is less likely to occur in classroom settings than in natural settings in that in the classroom learners are more aware of the correct and standard form of the TL, and they may receive error corrections in the classroom. However, Tarone (1982) suggests that L1 transfer is more evident in the learners’ “careful style” than in their “vernacular style” because learners tend to utilise their L1 as one of the potential resources when they are paying attention to how they speak. 28 Given that classroom settings are usually “careful style” environments, the findings of these two studies contradict each other. What is more, Shirai (1992) argues that L1 transfer is strong when learners learn another language in input poor environment such as foreign language situation on the grounds that learners tend to use a “grammar translation approach” which necessitates them connecting L1 to L2.12 Again, more studies are necessary to suggest a convincing and well-grounded condition of transfer in this area. There is a surprising claim that virtually no constraints on transfer exist (Thomason & Kaufman 1988: cited by Jarvis & Odlin 2000). However, a number of factors that promote or inhibit transfer have been identified, although different researchers have attempted to find the constraints or conditions of transfer from different perspectives (e.g. linguistic, psycholinguistic or sociolinguistic) and in different linguistic levels (e.g. lexicon, syntax or phonology). Jarvis (1997) claims that CLI carries a different significance in different areas of L2 acquisition: L1 grammatical influence is relatively small and short-lived, while phonology is the area of L2 where L1 influence is strongest or most conspicuous. Tanaka & Abe (1989) also claim that transfer is weak in the area of morphology or syntax because it is difficult for learners to find interlingual equivalents. In contrast, transfer is strong in the area of lexico-semantics, phonology, and discourse, since it is easier and more straightforward to find interlingual equivalents. 12 Shirai (1992) refers to this factor in “learning environment” under the subcategory of “learner characteristics” instead of “sociolinguistic context” in which he includes such factors as “hearer-role” and “subject matter”. Despite different ways of categorisation in each study, these issues may be comparable due to the fact that they are all arguing “learning context or environment”. 29 Dulay, Burt, & Krashen (1982) contended that transfer involving morphology and syntax was minimal, and Rutherford (1983) and Zobl (1986) were sceptical about the occurrence of basic word order transfer. Yet, Shirai (1992) emphasises that transfer does occur even in the areas where transfer is considered to be weak, if L1-L2 mapping is simple. Indeed, evidence has shown that transfer does operate in certain language contexts even in the domains or language levels that had been considered to be non- transferable or less transferable, such as basic word-order patterns (Odlin 1990), semantically opaque idioms (Sridhar & Sridhar 1986), and morphology (Jarvis & Odlin 2000). Nakahama (2003) also provided convincing evidence for the differing patterns in the acquisition of Japanese postpositional markers and morphological changes between Korean and English L1 speakers. At the same time, Odlin (1990) admits that learners are generally less likely to transfer the basic word order of their L1 to the L2, given that their metalingual awareness of the TL may be much stronger in this domain than in others such as lexico-semantics or discourse. In order to synthesise previous findings, Ellis (1994) reviewed an enormous number of previous studies on the constraints of transfer, and categorised them into six major factors. They include: (1) language level (e.g. phonology or lexis), (2) social factors (e.g. learning contexts or the relationship of the interlocutors), (3) linguistic markedness, (4) perceived prototypicality, (5) language distance or psychotypology, and (6) developmental factors (e.g. Transfer to somewhere principle). Additionally, Ellis (1994) referred to some non-structural factors such as (1) personality, (2) age, and (3) the nature of the tasks that a learner is performing. 30 Jarvis (2000) made some modifications to the list of factors proposed by Ellis (1994). Jarvis’s list includes: (1) age, (2) personality, motivation, and language aptitude, (3) social, educational, and cultural background, (4) language background (all previous L1s and L2s), (5) type and amount of TL exposure, (6) TL proficiency, (7) language distance between the L1 and TL, (8) task type and area of language use, and (9) prototypicality and markedness of the linguistic feature13. The newly added factors are (3) social, educational and cultural background, (4) language background, and (6) TL proficiency. Factor (3) and (4) seem to be particularly significant variables for Jarvis (2000) because he constructed the list for the purpose of his study conducted in Finland, dealing with both Finnish speaking Finns and Swedish speaking Finns of virtually the same cultural and educational background. As for (6) TL proficiency, Ellis (1994) referred to this factor in (6) developmental factors in his list, although he assumes “developmental factors” in a broader sense including the interplay between natural principles of language learning and transfer. Nakahama (2003) asserts that certain aspects of L1 may affect the learners’ IL at different stages of development. That is, while some studies claim that beginners tend to rely more on their L1 knowledge (e.g. Taylor 1975; Corder 1983; Major 1986; Ringbom 1987), other studies indicate that the possibilities of transfer may increase as learners’ L2 knowledge increases (Kellerman 1983; Klein 1986). Kellerman (1983) explains that transfer errors involving relatively difficult structures (e.g. pronominal 13 Although Ellis (1994) differentiated (3) linguistic markedness and (4) perceived prototypicality in his categorisation, Jarvis (2000) put these concepts into one category. The largest differences between these two concepts are that the former is determined by researchers’ linguistic analysis of learners’ TL, while the latter is based on learners’ perception of particular linguistic items in their L1. Detailed discussion about the prototypicality will be given in the following chapter. 31 copies in relative clauses) may only appear when the learners are at a sufficiently advanced stage of development to be able to produce the structures themselves. For this reason, Ellis (1994) states that although some aspects of L2 development may reflect the gradual restructuring of L1 features to those of the TL, other aspects do not. In some cases, transfer is only visible in the later stages, while in others early transfer is never eradicated. In other words, L1 and developmental factors function together in forming learners’ IL. Thus, transfer may both delay and accelerate L2 learners’ natural developmental process (Ellis 1994). Although the “Transfer to somewhere principle” (Andersen 1983) has attempted to elucidate the necessary conditions for such interplay (i.e. between transfer and the natural development process), still not all of the conditions of transfer seem to be fully predicted. As Nakahama (2003) says, the issue of CLI and proficiency level is far from simple. Apart from the factors that promote or inhibit transfer proposed by Ellis (1994) and Jarvis (2000), Shirai (1992) additionally suggests “cognitive load” as one of the important conditions of transfer: when cognitive load is high, L1 transfer tends to occur. On the basis of an information-processing framework in the field of psycholinguistics (e.g. McLaughlin 1978, 1987; Færch & Kasper 1986a; Bialystok 1988), Shirai (1992) argues that when L2 learners confront tasks with a high cognitive load (e.g. discussion of complex matters), their attention is diverted and their accuracy decreases. When L2 learners cannot pay attention to their performance, they are likely to be dependent on automatised knowledge, which does not require any use of working memory. The automatised knowledge possibly available for L2 learners is their IL knowledge and L1 knowledge, but the automatised IL knowledge is not always stable and sufficient for 32 communication needs. Thus, L2 learners tend to fall back on L1 knowledge, which is fully automatic and readily available (Shirai 1992). As Ellis (1994) states, while there is clear evidence that these constraints influence the occurrence of transfer, little is known about how these factors interact. Therefore, future research needs to explore the interaction among various constraints of transfer. Among them, the present research will specifically deal with (1) cross-linguistic semantic conditions of lexical items, (2) task type (i.e. receptive versus productive), (3) learners’ L2 proficiency, (4) frequency of the L2 input, (5) context in which the word is embedded, and (6) perceived prototypicality of L1 lexical items. Whereas “cognitive load” is relevant to the present research to a large extent, and will be discussed extensively, it will not be actively controlled as a variable due to its methodological difficulty. That is, it is hard to determine exactly when learners’ cognitive load is high and when it is not, even though it can be generally inferred that their cognitive load is higher when they are required to give prompt utterances than when they are not. It has been claimed that transfer procedure can be used in production, reception and as a learning process (Færch & Kasper 1987, 1989; Ringbom 1991, 1992). According to Ringbom (1992), production and reception (or comprehension in his term) are communication processes relating to performance, but learning involves a change in competence or a representation of an L2 form in memory. In other words, the roles of the L1 in L2 use are not necessarily the same as the ones in L2 development. Ellis (1994) also stresses the importance of the distinction between transfer in communication and learning. He says that this distinction is not always made in the discussion of transfer, and most researchers ignore this problem. As a result, the 33 evidence of transfer in performance is often erroneously identified with that of transfer in learning (Ellis 1994). In the following sections, transfer of three different dimensions (i.e. production, reception and learning) will be considered both theoretically and empirically in relation to the conditions of transfer according to “task type”. However, their scope will be focused on the lexicon due to its relevance to the present research. 1.3 Transfer as a Production Procedure Færch & Kasper (1986a, 1987, 1989) believe that L2 acquisition involves a complex interaction of skills at different levels of cognitive control, and consider transfer as a psycholinguistic procedure. Thus, transfer can both be conceived of as the creative activation of L1 knowledge at different levels of consciousness and the activation of highly automatised L1 knowledge in the absence of conscious control. Based on learners’ actual production data from two corpus they constructed14, Færch & Kasper (1986a, 1987, 1989) classified transfer in production into three categories of “strategic transfer”, “automatic transfer” and “subsidiary transfer”. Their classification of transfer in production is valuable in that it gave insight into the different roles of learners’ consciousness in different types of transfer. Despite its significant contribution to transfer research, Færch & Kasper’s (1986a, 1987, 1989) classification of productive transfer is not without problems. Thus, this section will review research on transfer in production by placing their classification as a starting point. The first type of productive transfer, which is used as a communication strategy or as a means of solving problems, is known as “strategic transfer”. Strategic transfer has three 14 PIF corpus of learner language consists of 120 video-taped, face-to-face conversations about everyday topics between Danish learners at various educational levels and native speakers of English. BO data consists of 48 audio-taped face-to face conversations between German first-year students and English native speakers. The conversations were based on the role-plays involving everyday situations. 34 subcategories of “borrowing”, “foreignizing” and “literal translation” (see Figure 1.1). “Borrowing”15 refers to the strategy which learners use an L1 lexical item with L1 form or L1 pronunciation. For instance, German word kopf (head) is used without tailoring it to L2 English in the following sentence: I shoot him with gun in *kopf (James 1998: 150). “Foreignizing”16 refers to the strategy which learners select an L1 lexical item and adjust its elements phonologically or morphologically to L2. For instance, Portuguese word nocivo (harmful) is tailored to the English-like morphology in: Smoking can be very *nocive to health (James 1998: 150). If the L2 word is created on the basis of direct translation of L1 word, it will be referred to as “literal translation”17. Examples of this include sleep suit for pyjamas based on a German word “Schlafanzug”(James 1998: 150). Borrowing (‘Substitution’; Odlin 1989) Strategic transfer Foreignizing (‘Coinage’; Jamas 1998) Transfer in Literal translation (‘Calques’; Odlin 1989; James 1998) production Automatic transfer (L1-based slips of the tongue; Poulisse 1999) Subsidiary transfer Figure 1.1: Færch & Kasper’s (1986a, 1987, 1989) Categorisation of Transfer in Production James (1998) refers to three types of “interlingual misformation errors” which are similar to Færch & Kasper’s three subcategories of strategic transfer. The main objective of James’s study is to describe and diagnose errors, and thus his intension is only on negative transfer. On the other hand, Færch & Kasper (1986a, 1987, 1989) view strategic transfer as a means by which learners compensate their lack of competence or knowledge regardless of its positive or negative effect. For instance, Færch & Kasper (1987) found, based on retrospective interviews conducted after the collection of oral 15 Odlin (1989) called the similar phenomenon as “substitution”. 16 James (1998) called the similar phenomenon as “coinage”. 17 Odlin (1989) and James (1998) called the similar phenomenon as “calques”. 35 data, that although Danish learners did not know certain German words (e.g. Schulpflicht = “compulsory education”), they were able to produce them because the elements of the compound words were known to them, and equivalent compounds were existing in Danish (e.g. skolepligt, lit. “school duty”). Færch & Kasper (1986a) claim that the activated L1 lexical item is an unanalysed chunk in the case of borrowing, while foreignizing and literal translation involve a combination of L1 and IL knowledge18. Thus, it is a prerequisite for learners to decompose the L1 item into smaller elements before these combinatory transfers can be used. Transparency of the L1 item may also be an important condition for these combinatory transfers to occur (Færch & Kasper 1986a). Poulisse (1990) examined the strategies used by Dutch learners of English to solve lexical problems in the L2 while performing three types of tasks: a picture description task, an oral interview and a story retell task. Poulisse (1990) reported that L1-based solutions to lexical problems using these three types of “strategic transfer” particularly “foreignizing” were frequent in her study. However, they were hardly observed in Chen (1990) which examined similar strategies of Chinese learners on their L2 English. Kellerman (1995) supposed that the contradictory results of these studies were due to the role of psychotypology or learners’ perceived language distance between L1 and L2, given that these strategies reflect knowledge of regular morphological (probably phonological as well) relations between the two languages and the probabilistic usefulness of L1-based lexical creativity. 18 Færch & Kasper (1986a) refer to learners’ knowledge of the TL as “IL knowledge” while other researchers call the same concept as “L2 knowledge” or “TL knowledge”. Strictly speaking, all of the knowledge regarding the TL learnt by learners is IL, but these three terms are used interchangeably in the present research. 36 Automatic transfer, the second type of productive transfer, may occur in a situation where there is competition between an L1 plan and an IL plan19, and where the L1 plan wins because it is more highly automatised. Alternatively, automatic transfer may occur in situations where there is no experienced need for developing a plan in IL, but where an automitised L1 plan is triggered off as a response to a certain situational cue (Færch & Kasper 1986a, 1987, 1989). For instance, the Danish word “men” (but) is used without any adaptation to L2 phonology in the following example: “I really don’t know *men maybe I better like to live there” (Færch & Kasper 1987: 128). According to Færch & Kasper, although their participants certainly knew the equivalent English expressions, this L1 word was triggered due to their paying less attention to the L2 word form. Swan (1997) maintains that the learners produce a simplified form, one closer to or identical with the L1 pattern because they have more fully-automated control over it, and can assemble it more quickly and easily than the correct TL equivalent. In cognitive terms, learners cannot access the TL form and retrieve it from storage quickly enough to use it for communication, and are driven back to more easily accessible material (Swan 1997). Considerations of this kind of transfer are explained by the multiple-competence models (e.g. Tarone 1983; Ellis 1985a) or the distinction between linguistic knowledge and the learner’s control of that knowledge (e.g. Bialystok & Sharwood Smith 1985; Bialystok 1994) which suggest that certain errors may coexist with accurate knowledge of the L2 items. In other words, these errors (or mistakes) are reflections of L2 performance weakness rather than the lack of L2 knowledge. This is also parallel with 19 A plan is referred to as means of realising communicative goals (Færch & Kasper 1986a). Plans are hierarchically organised involving the higher-order levels (e.g. pragmatics/discourse), the intermediate levels (e.g. syntax/lexis), and the lower levels (e.g. morphology/phonology) (Clark & Clark 1977). 37 James’s (1998: 86) claim that “learners know more than we credit them with knowing. We seize upon every deviance, assuming it to be an error, whereas it is more likely to be a mistake”. Corder (1967) refers to “errors” as failures in competence, and “mistakes” as failures in performance.20 Whereas Ellis (1985b) states that the distinction between errors and mistakes is not practically observable, James (1998) defines “error” as being an instance of language that is unintentionally deviant and is not self-corrigible, and “mistake” as either intentionally or unintentionally deviant and self-corrigible. Thus, James argues that mistakes can be differentiated from errors if learners are able to repair their own utterances. As a similar phenomenon, Poulisse (1999) defines “a slip of the tongue” as an unintended and nonhabitual deviation from a speech plan. In other words, the slips are the outcomes of speech plans which differ from the originally intended one, and the errors would not have been made if the speaker had paid more attention (Poulisse 1999). Considering the fact that the source of the erroneous utterances in automatic transfer is limited to L1 (or any previously learnt languages other than TL) but that this is not necessarily the case for the slips of the tongue, it seems reasonable to equate automatic transfer with “L1-based slips” which are slips of the tongue caused by learners’ L1. Thus, studies on the slips may promote our understanding of automatic transfer. Poulisse (1999) also indicates that those speakers who produce these slips should be able to repair them when they perceive them or they are pointed out. As the notion of 20 In the present research, the term “error” is used as a cover term for both concepts of ‘failure in competence’ and ‘failure in performance’, while the term “mistake” is used only in the sense of the latter. 38 reparability is crucial in the studies of slips, it may be possible to detect automatic transfer or L1-based slips if the speaker repairs his or her own utterances. The concept of subsidiary transfer, the third type of productive transfer, is rather fuzzy in that it forms a continuum of attention and automatisation placing automatic transfer at the one end and strategic transfer at the other (Færch & Kasper 1987). According to Færch & Kasper (1986a, 1987, 1989), subsidiary transfer tends to occur in a situation where there is focal attention neither on the relevant IL planning procedure nor on the transferred L1 knowledge. This is in contrast to strategic transfer where focal attention is on the IL production problem and its solution by means of L1. The largest difference between subsidiary and automatic transfer is that the latter involves the activation of a compiled sub-plan21 in the L1, but this is not the case for subsidiary transfer. Psychotypology or performance constraints are usually the trigger of this type of transfer (Færch & Kasper 1987, 1989). Færch & Kasper (1987) admit that the three types of transfer in production they have proposed are not discrete categories. Moreover, they show only one example of subsidiary transfer at the lexical level and it is also claimed that this type of transfer occurs when learners start executing a syntactic plan according to how lexicalisation may take place in their L1: “I have *four-one three sisters and one brother (Danish: ‘jeg har fire søskende’, søskende = siblings)” (Færch & Kasper 1987: 130). From its definition and example suggested by Færch & Kasper, it is difficult to grasp the concept of subsidiary transfer. Subsidiary transfer seems elusive, and thus problematic as a 21 Compiled sub-plan is prefabricated ready-made subroutines or an unanalysed chunk stored in the language user’s declarative knowledge (Færch & Kasper 1987). 39 heuristic for the research of transfer in production, although it may be feasible theoretically. Færch & Kasper (1986a, 1987, 1989) presume that strategic transfer is a conscious procedure and a creative activation of L1 knowledge which differs from the definition of automatic transfer or slips of the tongue. In addition, Færch & Kasper (1986a) argue that content words are semantically the most important elements of an utterance, and tend to be consciously chosen rather than automatically activated. In other words, strategic transfer is more likely to be used for content words, because learners are usually aware of lexical gaps and attempt to solve the problem (Færch & Kasper 1986a). However, some conflicting results with their claims have been shown. Poulisse (1999) reports a number of examples of L1-based slips on the basis of detailed classification of slips of the tongue. The results of her study show that almost 30% of the total number of the slips collected were caused by L1 influence, and 118 (out of 596) examples of L1-based slips were content words. In addition, 388 out of 596 L1-based slips (65%) were categorised as “L1 lexical substitutions” in which an L1 word is used with L1 pronunciation. This is also referred to as “non-adapted language switches” in Poulisse & Bongaerts (1994). Whereas this type of slip appears to be similar to the “borrowing” which is one of the subcategories of strategic transfer in Færch & Kasper (1986a, 1987, 1989), it contradicts the previous assumption: slips of the tongue are similar to automatic transfer, but not to strategic transfer. 40 All of the slips, including L1 lexical substitutions, detected by Poulisse (1999) were either actually followed by self-repair or considered to be reparable due to the frequent use of the same word correctly. What is more, Poulisse (1999) reports 23 examples of “lexical foreignizings” and 64 of “L1 / L2 lexical substitutions” as subcategories of L1- based slips. Lexical foreignizings are referred to as L1 substitutions which have been adapted to the L2 either morphologically, phonologically or in both ways. The example of this type can be seen in: “I must uh 1 uh put uh, the flashes in uh, of the, the bottles in erm, in uh 1 in cupboards” (Dutch “flessen”= English “bottles”) (Poulisse 1999: 149). L1 / L2 lexical substitutions can be characterised as false cognates. This type of L1- based slips comprises L2 words used with L1 meaning as in the following example: “the most important part of a clock, a bell sorry” (Dutch “klok” = English “bell” and “clock”) (Poulisse 1999: 150). Poulisse (1999: 149-150) assumes the mechanism for this type of slips is that: the concept for ‘bell’ activates the English lemma ‘bell’ and Dutch lemma ‘klok’. The Dutch lemma ‘klok’ spreads its activation to the Dutch word form ‘klok’, which spreads its activation to the English word form ‘clock’. In addition, some activation may spread from the concept ‘bell’ to the associated concept ‘clock’ and from there to the lemmas ‘clock’ and ‘klok’, which also activate the word forms ‘clock’ and ‘klok’. Since the English word form ‘clock’ receives activation from two sources, viz. from the Dutch word form ‘klok’ and from the English lemma ‘clock’, it is encoded before the intended word form ‘bell’. Lexical activation can spread from one concept to another associated concept, and from one word form to a similar word form from another language (Poulisse 1999). It also 41 should be noted that 59 out of 64 examples of L1 / L2 lexical substitutions were repaired by the speakers (the rests were also considered to be reparable), which indicates that such utterances were not resulting from their lack of L2 knowledge. Again, these results are inconsistent with such a claim as “One marked aspect of automatic transfer is that it is never combinatory” (Færch & Kasper 1986a: 60). Poulisse (1993) claims that the learners’ slips of the tongue that are blends of L1 and L2 words can be partly explained by the Bilingual Speech-Production Model that she proposed on the basis of Levelt’s (1989) L1 Speech-Production Model. The model assumes that L1 substitution errors where L1 lexical items are unintentionally used in substitution for the L2 equivalent words are not unlike intralingual substitution errors involving two semantically related lexical items (e.g. low/high, fingers/toes). Poulisse (1993) argues that translation equivalents generally share all semantic features and differ in the language feature only. For instance, when Dutch-English bilinguals activate an English word “boy”, other related words such as “person”, “girl”, “man” and “Jongen (Dutch= “boy”)” are also activated. As a result, “boy” is normally selected because it meets the largest number of criteria, i.e., [+male][+child] [+English] and has the closest meaning to the preverbal message. Occasionally, however, “Jongen”or “girl” may be unintentionally selected when they fail to activate [+English] or [+male] accidentally. Alternatively, blends of two related words (either interlingually or intralingually) that are activated together may be produced due to the speaker’s lesser attention to the form or under higher cognitive loads (Poulisse 1993; Poulisse & Bongaerts 1994). 42 Moreover, Poulisse (1997) reports that whereas many of the words involved in L2 learners’ unintentional language switch or slips of the tongue were function words (N=316) or editing terms (N=302), content words were also involved (N=131). Interestingly, however, 53.4% of the accidentally selected L1 content words were corrected by the learners, as opposed to 30.7% of the L1 function words. Poulisse (1997) believes that this is due to the learners’ stronger attention to the selection of the content words (and thus fewer mistakes), since they are more important in the communication of messages. This view is basically congruent with Færch & Kasper’s (1986a) claim except that they deny the occurrence of automatic transfer to content words entirely. Færch & Kasper’s (1986a) claims were largely based on theoretical grounds. Whereas their arguments are on the basis of learners’ oral production data in the previously mentioned two corpus, they did not find adapted automatic transfer or automatic transfer of content words. Færch & Kasper (1987, 1989) show how different types of transfer can be distinguished by paying attention to “strategy markers” or “signs of uncertainty” such as hesitations or appeals during participants’ oral tasks. If these markers or signals (i.e. pausing, sighing or laughing) appeared before or after the L1- like utterances, they are more likely to be strategic transfer. On the other hand, the detection of the slips in Poulisse (1999) is only based on the learners’ actual self-repairs or the correct use of the same word in other contexts by the same individuals. It cannot be denied that Færch & Kasper’s method is subjective in a way, and it seems that there is room for researchers’ theoretical biases as to whether these utterances are strategic (conscious and creative) or non-strategic (unconscious and automatic). 43 Thus, one could argue that some automatic transfers or slips can be results of using the knowledge of both L1 and L2: usually L1 items are phonologically or morphologically adapted to L2.22 Furthermore, as Poulisse’s (1997, 1999) examples show, L1 content words can also be seen to be automatically, unintentionally transferred to L2 production, particularly when cognates are involved. In contrast to Poulisse’s (1997, 1999) abundance of evidence for adapted-automatic transfer and automatic transfer of content words23, Færch & Kasper suggest no evidence to support their claims that such types of automatic transfer do not occur. As far as the above two particular points are concerned, we must conclude that Poulisse’s (1997, 1999) claims are more plausible. In sum, Færch & Kasper’s (1986a, 1987, 1989) classification of productive transfer did offer a valuable theoretical point of departure for the present research. That is, strategic transfer occurs when learners use L1 knowledge as a means of solving problems, so it is a conscious and creative procedure. When an L1 lexical item is used as it is without phonological or morphological adaptation, it is referred to as “borrowing”. When an L1 word is adjusted into L2 pronunciation or morphology, it is referred to as “foreignizing”. In contrast, automatic transfer occurs when learners pay less attention to the language form probably due to their high cognitive load, since L1 knowledge is more highly automatised than IL knowledge. In other words, like slips of the tongue, automatic transfer is an unconscious and unintentional procedure. 22 Poulisse (1993) argues that learners consistently apply L1 morphology when speaking the L1, and L2 morphology when speaking L2 even when they misselect a lexical item from the unintended language. 23 Poulisse’s examples indicate that automatically transferred content words can also be phonologically adapted to L2. 44 Nevertheless, some modifications are necessary for the purpose of the present research of Færch & Kasper’s (1986a, 1987, 1989) classification. Although Poulisse does not refer to the inconsistency of her results with Færch & Kasper’s claims regarding the slips of content words or phonologically adapted slips, Poulisse’s (1997, 1999) studies were useful in addressing the shortcomings of Færch & Kasper’s classification of transfer in production. Thus, the present research assumes that automatic transfer may involve content words as well as phonological adaptation from L1 to L2. The research will not consider subsidiary transfer due to its ambiguous definition. In the following section, I will discuss transfer of another dimension: reception (or comprehension). 1.4 Transfer as a Reception Procedure Whereas a number of studies have explored the issues of transfer, it should be noted that the majority of previous research focuses on transfer in production particularly of negative transfer, so transfer in reception and positive transfer have been largely neglected (Ringbom 1987, 1991, 1992; Ellis 1994; Jarvis & Odlin 2000; Odlin 2003). This might be mainly due to the fact that the detection of transfer phenomenon is far more difficult in the area of reception than in production (Ellis 1994), and in the area of positive transfer than negative transfer. For instance, the detection of receptive transfer is a difficult process given that it does not produce language output, and that a learner’s mental organisation is not directly observable. In addition, it is extremely difficult to judge whether or not a correct utterance by a learner is because of the influence of transfer. Even when the produced language form is L1-like, it is also possible that the learner’s production is pure reflection of L2 input in the case of positive transfer. Ringbom (1991: 180) stresses the importance of studies on receptive transfer and positive transfer: 45 If CLI on comprehension is studied, its primary facilitating, not inhibiting, effect on learning will become clear. And if we accept that comprehension and production constantly interact in the learning process, the importance of ‘positive transfer’ will also be recognised for production. Ringbom (1987) asserts that comprehension is the learners’ ability to process incoming data relating to previous knowledge structures, whereas production is the ability to activate knowledge structures without a direct linguistic stimulus from outside. Færch & Kasper (1986b) view comprehension as the integration of three types of information: input (verbal and non-verbal), the listener’s existing knowledge (linguistic and world), and contextual information. Comprehension is a matching process which is both input- driven (bottom-up) and knowledge-driven (top-down), and the proportion of each is dependent on the nature of the comprehension task. When gaps exist between the input and the learner’s knowledge, inferencing will be used in an attempt to bridge them, and hence input is further processed (Færch & Kasper 1986b; Ringbom 1987). Ringbom (1987) argues that when L2 learners encounter an unfamiliar word, they can make use of several different cues to infer its meaning: intra-linguistic, cross-linguistic, and/or extra-linguistic. Moreover, existing knowledge is more easily activated by the linguistic cues of incoming data if similarities (cross-linguistic and intra-linguisitc) are perceived by the learners. In other words, the learners’ L1 constitutes “potential knowledge” that can be drawn on more easily in decoding (i.e. comprehension) than in encoding (i.e. production) of their L2 particularly when learners’ L1 and L2 share many cognates (Ringbom 1992). 46 Odlin (1989) states that L1 structures can influence the interpretation of L2 messages, and sometimes that influence leads to learners inferring something very different from what native speakers of that language would infer. Odlin intends only negative aspects of L1 influence here because he refers to this as “misinterpretation”. According to Færch & Kasper (1987, 1989), receptive L1 transfer constitutes a subcategory of inferencing, and implies that the learner attempts to interpret incoming L2 utterances on the basis of his or her L1 knowledge regardless of its result (positive or negative transfer). Some evidence has been shown with regard to the learners’ use of receptive transfer particularly facilitating. Ard & Homburg (1983) found that there were significant differences between Spanish and Arabic ESL students in their results of a vocabulary section in the Michigan Test of English Language Proficiency. They suggested that Spanish learners did especially well with the words that had identical or similar spelling to Spanish forms. The vocabulary section of the test used for Ard & Homburg’s analysis consisted of synonym-type questions where examinees select the correct meaning of the italicised word in a short sentence out of four choices, and fill-in-the- blank questions where they choose one word from among the four possible words that could be inserted into the given sentence. In this sense, each part of the test measures how test takers infer the meaning of the target words using contextual or other clues (e.g. L1 knowledge), unless they had learnt and knew the meaning. However, not only the correct answer but also the other three words tended to make sense in the context because usually they were the same part of speech (Ard & 47 Homburg 1983). For this reason, contextual clues did not work very well to find the right answer. Accordingly, Ard & Homburg (1983) concluded that the results reflected the Spanish learners’ extensive use of receptive L1 transfer given that 60% of the target words resembled Spanish words in form and meaning, but only 1% of them resembled Arabic words24. Ringbom (1992) reports the results of English section in the Finnish National Matriculation Examination. This was a standardised test comprising (a) reading comprehension, (b) listening comprehension, and (c) an essay. The results show that Swedish speaking Finns performed consistently better than Finnish speaking Finns in all of these sections. Based on these results, Ringbom concluded that Swedish speakers have had much more help from their L1 than did the Finnish speakers whose L1 is unrelated to English. Although the reading and listening comprehension tests were multiple-choice with four alternatives, the results were based on the examinations of ten-year period (1980-1989), and the examinees were 255,000 Finnish and 16,000 Swedish speaking students. Thus, Ringbom (1992) believes that such generalisations can be justified. In addition to the studies that attempted to show the L1 effects on L2 text comprehension in typologically close European languages, there are also some studies that involve Japanese and Chinese languages, which are typologically unrelated but share numerous cognates. Thus, there may be some unique features of L1 influence in such conditions. Hatasa (1992) conducted a kanji test, a reading test and a grammar test 24 This is not based on psychotypology or learners’ perceived language distance, but Ard & Homburg’s typological comparison of these languages. 48 on both Chinese background learners (CBLs) and English background learners (EBLs) of three different proficiency levels of Japanese, and found that CBLs performed significantly better in kanji test and reading test but not in grammar test. However, the overall improvement of CBLs in the reading test was small compared to the drastic rise of EBLs’ score as their proficiency level increased. Consequently, Hatasa (1992) concluded that although CBLs could attain a higher proficiency level in a short period of time, their reliance on knowledge of Chinese characters might be effective only at the initial stage of learning to read Japanese. Matsunaga (1999) demonstrated a series of reading and oral tests on CBLs and non- CBLs to compare their performance between each task. The results of Matsunaga’s study indicate that there was no positive correlation between the CBLs’ text comprehension and oral skills of Japanese. Namely, at the beginning level, CBLs’ transfer of kanji knowledge from the L1 is advantageous for comprehending the Japanese texts that contain many kanji words, but not for oral tasks. For this reason, Matsunaga (1999) stresses the importance of the learning of accurate pronunciation of Japanese kanji words as well as the development of solid oral proficiency in order for CBLs to fully develop their reading skills. Previous empirical studies of CBLs’ use of L1 knowledge in learning Japanese commonly indicate that it positively affects their vocabulary acquisition and reading comprehension especially in their earlier stages of development but not in other skills or tasks. Whereas Swedish speakers performed significantly better than Finnish speakers not only in reading comprehension but also in listening comprehension and writing in Ringbom (1992), CBLs performed better only in vocabulary (kanji) and reading comprehension in Hatasa (1992) and Matsunaga (1999). 49 These results are not surprising if we consider the fact that although Japanese and Chinese languages share a tremendous number of cognates which are written identically or similarly, their pronunciations are not always similar (Kato 2005). Moreover, these two languages are typologically unrelated and share few similar grammatical features. Thus, it is natural that CBLs did not perform better than the learners of other language backgrounds in oral and grammar tests. As referred to earlier, CBLs’ semantic transfer from Chinese to Japanese particularly involving kanji characters needs to be argued separately from that of other language levels (e.g. syntax or morphology) because the potentiality of their transfer is considerably different in the area of lexico-semantics. While studies conducted by Ard & Homburg (1983) and Ringbom (1987, 1992) have shed light on learners’ use of positive transfer especially when their L1 and L2 share many cognates, there are also some limitations of their studies. Firstly, they only give a macro view of the facilitative effects of learners’ L1 in comprehension tasks, thus a more micro views of receptive transfer, i.e., how an individual L2 word is inferred or learnt on the basis of learners’ L1 knowledge, is by no means clear. Secondly, these macro views of CLI may offset the effects of negative transfer (if there are any) against positive transfer, so it appears as if there are no effects of negative transfer at all. It is necessary to observe each lexical item individually in order to detect negative transfer where positive transfer may be overwhelming. Likewise, studies dealing with CBLs’ L1 influence on reading comprehension of Japanese texts, which are more relevant to the present research, also have the same shortcomings. Although Hatasa (1992) and Matsunaga (1999) showed that the positive effects of CBLs’ use of L1 knowledge in reading Japanese texts were only limited to 50 earlier stages, these studies also lack micro views, and thus questions arise: (1) Is only positive transfer at work for beginning CBL’s comprehension of Japanese texts? (2) Is transfer not in operation for intermediate or advanced CBLs in their reading Japanese? (3) When and under what condition does either positive or negative transfer occur for CBLs in reading Japanese? There have been a few studies that specifically explore both the facilitating and inhibiting effects of learners’ L1 knowledge on L2 reading comprehension (Holmes 1986; Holmes & Ramos 1993). Holmes (1986) attempted to analyse the recognition of cognates in English texts by Portuguese students with very little knowledge of English. Think-aloud protocols were used to collect data as participants read silently through an English text and orally summarised their understanding of it in Portuguese. As a result of matching cognates between the oral summaries and the original texts, he found that a lack of grammar knowledge resulted in frequent misrecognition of cognates: nouns were recognised as verbs, adjectives as nouns. In addition, whereas participants were very dependant on the previous knowledge of the topic to predict the text content, their lack of a check-up or monitoring of initial guesses tended to result in misrecognition of cognates. Holmes assumes that the main reason of participants’ “reckless guessing” or the inconsistency of their summary with the original text is due to a cognitive overload: they could not use contextual clues or knowledge of the world to guess the meaning of words properly while conducting an oral task. Unlike Holmes (1986), Holmes & Ramos (1993) asked Brazilian learners of English of a similar proficiency level to write a summary of an English article in Portuguese within a group of four or five students. The main objective of this research was to confirm the 51 hypothesis that participants would check their first guesses in a group discussion by examining grammatical, textual and semantic contextual clues, hence a topic familiar to the participants was chosen. The results of the study reveal that the cognates were almost all recognised successfully in context, due to the participants’ monitoring of their initial hypothesis. Moreover, it is interesting to note that while they detected some examples of misrecognition of false cognates in the written summary of participants, such misrecognitions happened only when the context made sense accidentally, and no examples of misrecognition of false cognates were found when they were not coherent with contextual clues. As a result, Holmes & Ramos (1993: 101) concluded that “false cognates rarely led students astray”. Their findings are consistent with Ringbom’s (1992) claim which argues that learning to understand cognates (at least approximately) does not involve much effort for learners. According to Ringbom (1992), in comparison to the production of cognates, the comprehension of cognates causes much less problems for learners because the context frequently rules out a misunderstanding of deceptive cognates. Even when deceptive cognates are understood in a sense slightly different from their actual meaning, usually it does not hinder the understanding of the general meaning of the text due to the approximate nature of comprehension. Besides, many of the cognates are not really learnt at all, but are part of the learners’ “potential vocabulary”, namely, what they already know without ever having come across it before (Ringbom 1992). However, another study suggested that not all “potential knowledge” was exploited in the reading comprehension of closely related L2. Nagy, et al. (1993) investigated the relationship between Spanish speaking children’s ability to identify Spanish-English 52 cognates and their comprehension of English texts. Nagy, et al. (1993) found that the children’s performance in a multiple-choice comprehension test was largely dependant on the students’ ability to recognise cognate relationships. Although the results seem to be inconsistent with the claims made by Holmes & Ramos (1993) or Ringbom (1992), it is also possible that the participants’ age (grade four to six) and their limited knowledge of either Spanish or English (they were attending an elementary school in the United States) played a role in the difficulty of recognising Spanish-English cognates. Empirical studies conducted by Holmes (1986) and Holmes & Ramos (1993) are of great value in that they investigated the effects of L1 knowledge on the basis of controlled learners’ data in a real reading activity using an authentic L2 text. Yet, the conclusion of Holmes & Ramos (1993) cannot be easily generalised to the receptive transfer or acquisition of cognates as a whole, because they intentionally provided the reading material whose topic was well-known to the participants, and the reading task was held through discussion in a group of four or five students. It is unknown what happens when learners need to read a text of an unfamiliar subject or they need to read the text individually without any help from others. Indeed, there have been claims that context for L2 learners plays a relatively less important role than for native speakers particularly when their L2 proficiency is limited (Nagy 1997) or that contextual clues do not always work (Schmitt 1997; Huckin & Coady 1999). It has also been claimed that learners use multiple sources of information in determining the meaning of novel words (Paribakht & Wesche 1999; Mori & Nagy 1999). Among them, contextual clues and word morphology (word roots, affixes or inflectional suffixes) are the two major sources of information that learners use to infer the meaning 53 of unknown words in studies of non-character based languages (De Bot, Paribakht & Wesche 1997; Paribakht & Wesche 1999). Mori (2002) investigated individual differences in the ability to integrate information from word parts (or radicals) and context in interpreting novel kanji compounds using EBLs of Japanese as participants. Mori reported that guesses based on a single source are generally not as good as those from multiple clues. Moreover, those who rely too much on the kanji clues alone would have difficulty in expanding their vocabulary knowledge because multiple exposures are less likely to help their vocabulary learning. She also found that their ability to guess from kanji clues would not necessarily improve with proficiency (Mori 2002). Whereas Mori’s (2002) study revealed the mechanism of EBLs’ guessing the meaning of novel Japanese kanji compounds where their L1 lexical information is less likely to be useful, CBLs may infer the meaning of Japanese kanji compounds in a different manner because of their greater potentiality of transfer. As we have seen, most research arguing CLI on receptive skill focuses on reading comprehension, but there is also a study on vocabulary acquisition through audio and video (Duquette & Painchaud 1996) which refers to L1 influence. Duquette & Painchaud (1996) conducted three sets of vocabulary tests on English speaking university students of French as a second language, before, immediately after, and ten days after either watching a video about how to drive a car or listening to the same dialogue without visual support. They found that students’ inferencing of the word meaning was more facilitated when only audio was available. They believe that this result is owing to students’ more attention to the purely linguistic cues when only audio is available particularly due to the existence of French-English cognates. This study shows that L1 lexical knowledge especially of cognates facilitates L2 learners’ listening 54 comprehension. However, Duquette & Painchaud’s main objectives were to investigate whether students can discover the meaning and learn the unknown words through watching a video or listening to an audio-tape. Therefore, there is no further information relevant to the current study with regard to the mechanism of L1 transfer on listening comprehension. In sum, previous studies regarding L1 influence on L2 comprehension show that it generally has positive effects, but they do not give us much information about the conditions of transfer for individual words. It was claimed that learners do not always exploit their potential knowledge, but when and to what extent they use L1 knowledge is still a mystery. Holmes & Ramos (1993) suggest that learners have few problems even for the comprehension of deceptive cognates due to their extensive use of contextual and topical clues. However, inference based on context does not always work effectively for L2 learners (Nagy 1997; Schmitt 1997; Huckin & Coady 1999). Thus, it would be worthwhile to investigate an interaction of receptive transfer (i.e. use of L1 lexical clues) and the use of contextual clues in CBLs’ guessing of the meaning of novel Japanese kanji compounds. 1.5 Transfer as a Learning Procedure It is generally accepted that transfer in communication (production or reception) is common, but some researchers deny the role of L1 in shaping the learners’ IL system (Ellis 1994). For instance, Corder (1983) assumes that transfer is primarily a communication strategy (“borrowing” in his term), and it is a performance phenomenon rather than a learning process. However, Schachter (1983) suggests that learners construct a hypothesis about the nature of the L2, and restructure it using both inductive 55 and deductive inferencing. Then, one source of the hypothesis which learners draw on is their L1, although it can contribute to the construction of both correct and incorrect IL. It seems that Schachter’s view has gained much more support than that of Corder’s as to whether or not learners transfer directly from their L1 into their IL (e.g. Selinker 1992; Kellerman 1995; Ellis 1994; Ringbom 1991, 1992; Færch & Kapser 1987, 1989). Selinker (1992) believes that the creation of IL reflects developmental processes dealing with the L2 input, as well as a process of selectively using L1 knowledge and that of other ILs known to the learner. Kellerman (1995) also claims that transfer is the result of learners selectively exploiting their knowledge of L1 while coping with the complexities of the L2 input. Thus, the more widely acknowledged view is that transfer is not only used as a means of solving communication problems or guessing incoming L2 items on the basis of L1 knowledge, but also as a learning procedure in the learners’ attempt to establish hypotheses about L2 items. The hypotheses may be based on their L1 knowledge alone or as the result of an interaction between L1 knowledge and the L2 input (Færch & Kasper 1987, 1989). While a few researchers do not accept the idea that L1 knowledge is directly taken into learners’ IL (e.g. Corder 1983), there seems to be no disagreement as to whether transfer results in learning indirectly: through the use of transfer either in production or comprehension. The main difference between L2 comprehension and L2 production (of lexical items) concerns their relationship to learning (Ringbom 1992). Although production of L2 words usually occurs after learners have learnt them (except for the case of communication strategy), all words need not have been learnt before the comprehension activity takes place. Therefore, transfer in comprehension is much more 56 likely to result in learning or change of representation in memory than transfer in production (Ringbom 1992), even though the input that works for comprehension may not always work for acquisition (Ellis 1994). Færch & Kasper (1986b) and Ringbom (1991, 1992) stress the interdependence of comprehension and learning, because they regard “inferencing” as the first stage of vocabulary learning. Færch & Kasper (1986b) argue that it is not comprehension itself but lack of comprehension that facilitates learning. That is to say, inferencing is indispensable in order for the acquisition of new words to occur by using content schema or world knowledge to discover their meaning and contextual uses (Færch & Kasper 1986b). For this reason, Ringbom (1992) asserts that observing the transfer in comprehension is much more important than the transfer in production, which is a consequence of previously occurred transfer, in order to better understand the relationship between transfer and learning. While Ellis (1994) acknowledges the legitimacy of Ringbom’s claim, he maintains that transfer in production can also contribute significantly to IL development. Kellerman (1995) also states that L1-based (communication) strategies are essential for learners to develop their IL system, referring to Poulisse’s (1990) study on compensatory strategies in solving lexical problems. Corder (1983) assumes that transfer in production in the sense of communication strategy (or “borrowing”) results in learning, even though his claim is based not on lexical items but on grammar. Corder (1983) said that successfully borrowed forms are eventually incorporated into IL grammar, both as the correct and the incorrect forms. Other studies have also suggested similar views (Færch & Kasper 57 1980; Nation 1990). Figure 1.2 shows my understanding of the mechanism of transfer in learning and transfer in communication. L2 Input Creation of IL based on Directly L1 hypothesis Knowledge about the TL (Learning) Receptive Use Indirectly Purely as a Productive Use communication strategy (Non-learning) Figure 1.2: Mechanism of Transfer in Learning and Transfer in Communication Yet, it is not always possible to determine whether an utterance is the result of a “learning” or a “communication strategy” (Corder 1978, 1983). For instance, an utterance, which is traceable to learners’ L1, may be the process of testing the hypothesis that their L1 can be used in the particular context due to the similarity between their L1 and L2. In this case, the utterance is more likely to result in learning. However, it is also possible that the same utterance is borrowed from the L1 for communicative purposes without the intention of incorporating it into the learners’ IL system. In that case, the utterance is simply the result of communication strategy (Corder 1978, 1983). As Ellis (1994) points out, it is possible to make a clear distinction between the transfer in communication and learning theoretically, but it is difficult to do so empirically due 58 to the fact that the main data for the study of transfer come from language performance. For this reason, the present research sees transfer in learners’ “use” of a particular L2 word and its “acquisition” as existing on a continuum, rather than as two separate procedures. 1.6 Avoidance Ellis (1994) claims that it is insufficient to focus exclusively on production errors (i.e. negative transfer) or facilitation (i.e. positive transfer) in a discussion of CLI, because many of the subtle manifestations of transfer will be missed. Kellerman (1983) also asserts the elusiveness of L1 effects on L2 by pointing out that interference or negative transfer, which is a linguistic product, is only the tip of the iceberg as far as evidence for the influence of L1 on L2 is concerned. This is because avoidance becomes visible only by virtue of the relevant structure being under-produced by one group of learners with a particular L1 in terms of another group with a different L1 (Kellerman 1983). While there had been some studies referring to similar concepts (e.g. Lado 1961), the first comprehensive argument about avoidance was made by Schachter (1974). Schachter (1974) investigated the use of relative clauses of four different L1 groups (Japanese, Chinese, Arabic and Persian) by collecting 50 English compositions from each group, and using English native speakers as a control group. The results of her study show that Japanese and Chinese speakers produced significantly fewer relative clauses compared with the Arabic and Persian speakers. In addition, the error rate of Japanese and Chinese groups on relative clauses was also significantly lower than that of the other two groups. Hence, Schachter (1974) concluded that Japanese and Chinese speakers avoided the use of relative clauses, and produced fewer errors due to the fact 59 that placement of relative clauses in Japanese and Chinese is completely different from English, while in Arabic and Persian it is similar to English. Based on this result, Schachter criticised error analysis in that it is totally dependant on error in the absence of other information, and it cannot account for the phenomenon of avoidance. Moreover, she emphasised that contrastive analysis should not be totally abandoned due to its predictability of learners’ difficulties, which may be overlooked by analysis of error rate or the raw number of errors alone. Nevertheless, the following studies of avoidance have pointed out many problems in Schachter’s study. Kleinmann (1977) questioned the validity of Schachter’s study in that her Chinese and Japanese participants might have lacked the knowledge of relative clauses, which possibly is the cause of underproduction. Hence, Kleinmann claimed that the learners’ knowledge of the particular structure is a prerequisite for detecting avoidance, and true avoidance occurs only when learners have the choice to use or not to use that structure. Kleinmann (1977) investigated Arabic speakers and Spanish/Portuguese speakers in their use of passives, present progressives, infinitive complements, and direct pronouns in English. These four grammatical structures were chosen based on contrastive analysis, predicting differential difficulty for each of the two groups: two of them are more difficult for one group, and two are more difficult for the other group. Despite his criticism of the methodology of Schachter’s (1974) study, the findings of Kleinmann’s study supports her study in a way: the second language learners do resort to an avoidance strategy. The significance of Kleinmann’s study lies in his attempt to indicate the presence of the learners’ knowledge of the four grammatical structures 60 using comprehension tests, so his results cannot be attributed to a lack of knowledge of the avoided structure. As Kamimoto, et al. (1992) say, the absence of requisite knowledge cannot lead to avoidance because there is nothing to avoid, i.e., the learner is merely ignorant. Dagut & Laufer (1985) examined the Hebrew speaking learners’ use of English phrasal verbs (e.g. turn up, let down) because there were no phrasal verbs in their L1. Dagut & Laufer found that Hebrew speaking learners significantly avoided using phrasal verbs, preferring a one-word verb instead. Thus, they concluded that the avoidance was caused by systematic incongruence between learners’ L1 and L2. They also argued that the learners’ non-use of phrasal verbs were the result of genuine avoidance rather than of ignorance, because these expressions were not unknown to them. In contrast to Dagut & Laufer (1985), Hulstijn & Marchena (1989) indicated that Dutch learners of English also tended to prefer one-word verbs even though equivalent phrasal verbs did exist in Dutch. Hulstijn & Marchena reported that Dutch learners of English did not avoid phrasal verbs categorically, but they were likely to avoid idiomatic phrasal verbs. Laufer & Eliasson (1993) compared the use of phrasal verbs between Swedish and Hebrew learners of English, and attempted to verify which factor could predict the phenomenon of avoidance accurately. As a consequence, they found that L1-L2 differences still played the largest role in the avoidance of phrasal verbs. Seliger (1989) examined Israeli learners’ use of the English passive form by comparing it with English native speakers’ use in a task describing the processes or procedures of specific actions. In addition, Seliger (1989) confirmed the existence of the Israeli learners’ knowledge about how to construct the passive form using an active-to-passive 61 transformation task. The results of Seliger’s study show that the Israeli learners’ use of the passive form differs significantly from that of English native speakers: while Israeli learners’ used few passive forms in certain contexts, English native speakers used few passive forms in other contexts. For this reason, Seliger concluded that the Israeli learners’ underuse of the passive form in certain contexts was caused by avoidance. Seliger (1989: 30) also claimed that “the cause of avoidance would not be ignorance of the structure but rather interference from distribution rules in L1”. Kamimoto, et al. (1992) point out that despite the Israeli learners’ ability to transform from active to passive form, this is not avoidance at all but they are simply transferring the distribution of the passive form from Hebrew to English. In addition, Kamimoto, et al. contend that to know a construction is also to know its distribution. Thus, they believe that whereas Israeli learners knew what a passive form was, they simply did not know when or in what context passive forms should be used. In other words, Israeli learners did not use the passive form as frequently as English native speakers simply because it was rare in their L1. Bley-Vroman & Houng (1988) claim that one structure may be very frequent in one language, but the corresponding structure in the other language may not be so frequent. Hence, they supposed that Chinese and Japanese speakers in Schachter’s (1974) study were not necessarily avoiding the use of relative clauses, but they were simply transferring the distribution of relative clauses from their L1. To test their hypothesis, Bley-Vroman & Houng (1988) compared the frequency of relative clauses between an original English version of the American novel “The great Gatsby” and its published Chinese translation. Their investigation revealed that whereas the original English 62 version used 93 relative clauses in the first five chapters, the Chinese version used only 32 relative clauses in the equivalent chapters. Therefore, Bley-Vroman & Houng argue that underproduction of certain L2 structure should not be so easily attributed to avoidance, and it is also important to compare the relative frequencies of the structures used by native speakers in both languages. Bley-Vroman & Houng (1988) indicated that the frequencies of relative clauses were much lower in Chinese texts than in the equivalent English texts; however, they did not show such evidence in Japanese texts. Accordingly, Kamimoto, et al. (1992) compared the frequencies of relative clauses in the original “The great Gatsby” with its three different published Japanese translations using a similar method of counting25.Asa consequence, they found that only 44 to 50 relative clauses were used in the Japanese versions in contrast to 100 relative clauses in the English version. However, whereas Kamimoto, et al. acknowledged some advantages of comparing the frequencies of a certain structure across languages using a source text and its translations, they questioned Bley-Vroman & Houng’s method of counting due to its Anglo-centric aspect. Instead, Kamimoto, et al. stressed the importance of looking at the relative clauses which did not appear as relative clauses in the original text, and they did count all the relative clauses in one of the translations regardless of whether they corresponded to relative clauses in the original version or not. The result showed that 396 relative clauses were used in the Japanese texts, which were almost four times as many than in the original English text. 25 Although Bley-Vroman & Houng did not count those relative clauses that did not have relative pronouns or adverbs, Kamimoto, et al. did count them. Thus, Bley-Vroman & Houng found 93 relative clauses, while Kamimoto, et al. found 100 in the same original English texts. 63 Furthermore, Kamimoto, et al. attempted a similar comparison between the well-known Japanese book “Kotoba to bunka” and its English translation “Words in Context” as well as newspaper articles in “Asahi Shinbun” and the “New York Times” referring to the same incident and using the same number of clauses. Their findings indicate that relative clauses are used in Japanese texts at least as frequently as in English texts, although it does not mean every English relative clause functions in the same way in Japanese. Accordingly, Kamimoto, et al. (1992) conclude that study of avoidance should not take a grammatical structure purely as a syntactic unit without considering its function and frequency across languages. Finally, they also suggest that the definition of avoidance and the methodology of detecting it should be reconsidered, otherwise avoidance merely becomes a synonym for underproduction. In summary, detection of avoidance is extremely difficult particularly when grammatical items where interlingual mapping is not straightforward (e.g. relative clauses or passive forms) are involved. This is because the distribution or frequency of certain linguistic forms for certain functions may differ according to languages. For instance, “Someone took my wallet” is usually expressed in Japanese in passive form, but English speaking learners of Japanese may not use the passive form in describing this concept due to the different distribution in their L1. In other words, detection of avoidance must be based on the cross-linguistic comparison not only of the corresponding linguistic forms, but also of the equivalent function or meaning of certain grammatical items. In this respect, detecting the phenomenon of avoidance in lexical items seems to have fewer problems given that the form-meaning relationship is relatively simple in lexical levels. 64 In this chapter we have seen that CLI is an elusive phenomenon and it is difficult to grasp its nature due to the fact that a number of factors affect the occurrence of transfer. The present research attempts to investigate how some of these factors interact with each other. Among these factors, “task type” is possibly one of the most influential factors, since the literature review has revealed that the mechanism of transfer and its constraints are fundamentally different in production and reception. Psycholinguistic research on speech production suggested that transfer could be either a conscious or unconscious procedure, but the definition and the identification of these types of transfer have not been without problems. Hence, the current research will clearly define each type of transfer and detect them in a more reliable manner. Research on reading comprehension found that positive transfer overwhelms negative transfer due to the approximate nature of comprehension tasks and the assistance of contextual or topical clues. However, the nature of the receptive transfer is far from clear due to the fact that no previous research on transfer has given carefully chosen target items to the learners for the purpose of investigating receptive transfer. Therefore, the present research will examine CBLs’ receptive transfer in relatively controlled tasks. Although learners’ L1 knowledge may be transferred and incorporated directly into their IL as a source of their own hypotheses, their use of L1 knowledge either in production or comprehension can also result in learning. Thus, the present research will not always attempt to make a clear distinction between transfer in communication and learning empirically, and consider these two processes as a continuum, even though they are regarded as separate processes theoretically. 65 In the following chapter, I will review related research on L2 vocabulary acquisition and L2 lexico-semantic development. 66 CHAPTER TWO RESEARCH ON L2 VOCABULARY ACQUISITION AND L2 LEXICO-SEMANTIC DEVELOPMENT In Chapter 1, we looked at the question of when and under what conditions transfer would take place primarily by considering the task type. In this chapter, we turn our attention to the issue of the relationship between learners’ interlingual semantic transfer and their L2 lexico-semantic development. The chapter begins with a brief overview of the research on L2 vocabulary acquisition. This is followed by a review of the research on “receptive versus productive vocabulary” which is crucial in arguing for task type as the major factor in transfer. It then discusses several areas of L2 vocabulary and psycholinguistic research that are most relevant to the above mentioned issue. They are: learnability of L2 vocabulary, L2 lexico-semantic development, polysemy and prototype, cognates, bilingual mental lexicon, and Japanese kanji compounds. 2.1 A Brief Overview of L2 Vocabulary Research For many years, vocabulary has been undervalued or somewhat peripheral in the fields of second language teaching and research (Richards 1976; Levenston 1979; Meara 1984; Carter 1987; Carter & McCarthy1988; Laufer 1990a, 1997; Coady 1997; Zimmerman 1997; Schmitt 2000). This neglect is partly due to a heavy concentration on syntax and phonology in linguistic research. In particular, vocabulary was felt to be a less important element in learning a second or foreign language at a time when structuralism and contrastive analysis were dominating the language acquisition research (Carter 1987; Carter & McCarthy1988). 67 By the late 1990s, however, there had been a growing concern with vocabulary teaching and learning to redress this neglect. Laufer (1997: 140) claimed that: Vocabulary is no longer a victim of discrimination in second language learning research, nor in language teaching. After decades of neglect, lexis is now recognised as central to any language acquisition process, native, or non-native. After seven years of this claim, Read (2004) refers to the current status of L2 vocabulary studies as a “boom”. As a consequence of this increased attention to lexis, an extensive body of literature has been produced in the field of L2 vocabulary studies in recent years. These comprise of a wide range of topics or approaches and the way of classifying them varies from researcher to researcher. Barcroft (2004) presented ten areas of research related to L2 vocabulary acquisition. They are: (1) incidental vocabulary learning, (2) lexical requirements for comprehension, (3) input enhancement and text-based factors, (4) vocabulary learning strategies, (5) combined indirect and direct vocabulary instruction, (6) methods of direct instruction, (7) word-based determinants of learnability, (8) bilingual mental lexicon, (9) receptive versus productive vocabulary knowledge, and (10) lexical input processing. Barcroft (2004) uses the term “L2 vocabulary acquisition research” as a cover term for the concept of L2 vocabulary research in general including vocabulary teaching and instruction. In his definition, teaching or instruction of vocabulary can be regarded as “intentional learning” from the learners’ point of view as opposed to “incidental learning” where learners acquire new words from context without intending to do so. Barcroft (2004) also indicates that the first six of these research areas constitute a 68 continuum between highly incidental and highly intentional acquisition, given that attention is not a dichotomous entity (Gass 1999; Wesche & Paribakht 1999), while the last four research areas involve other issues that cannot be classified on this continuum. On the other hand, Bogaards & Laufer (2004) also list areas of L2 vocabulary acquisition in their review of the research over the past two decades. They include 12 themes in their work. These are: (1) the construct of vocabulary knowledge (e.g. the distinction between receptive and productive knowledge, and between knowledge and use), (2) the relationship between vocabulary knowledge and language proficiency, (3) the role of word frequency in vocabulary learning, (4) task effect on learning, (5) the use of dictionaries, paper and electronic, (6) interactive tasks, (7) explicit versus implicit learning, (8) incidental versus intentional learning, (9) learning new words versus learning new meanings of already known words, (10) patterns of vocabulary development over time, (11) strategies used by learners to comprehend and learn new words, and (12) testing vocabulary knowledge. As we have seen, L2 vocabulary acquisition research is currently one of the most popular areas of SLA research, and its topics and approaches are diverse. As it is beyond the scope of this chapter to review each area or theme of L2 vocabulary research, the most relevant topics to the present research will be reviewed here. Prior to discussing the main issues of this chapter, it is meaningful to review one area of L2 vocabulary research that is directly related to the issues of task type: receptive versus productive vocabulary. 69 2.2 Receptive versus Productive Vocabulary Nation (2001: 25) defines receptive and productive vocabulary as the following: receptive vocabulary use involves perceiving the form of a word while listening or reading and retrieving its meaning. Productive vocabulary use involves wanting to express a meaning through speaking or writing and retrieving and producing the appropriate spoken or written word form. The general assumption is that a word is known receptively first, and later becomes available for production (e.g. Nation 1990, 2001; Meara 1996; Melka 1997; Read 2000). Melka (1997) believes that receptive knowledge and productive knowledge are on a continuum rather than two separate areas of knowledge, although it may be a problem where to locate the threshold at which the word passes from receptive to productive status. Even though this distinction has been argued for in a number of studies, the approach or intention of such work frequently differs (Webb 2005). Webb (2005) indicates that some studies compare the effectiveness of receptive learning and productive learning (Griffin & Harley 1996; Waring 1997a; Schneider, et al. 2002; Mondria & Wiersma 2004; Webb 2005), while other studies focus on receptive and productive vocabulary size (Laufer 1998; Laufer & Paribakht 1998; Laufer & Nation 1999; Waring 1997b) or argue whether receptive knowledge is gained before productive knowledge (Aitchison 1994; Channell 1988; Melka 1997). Furthermore, although Webb (2005) did not comment on this,, there have also been some studies that distinguish receptive and productive acquisition of vocabulary within the interactionalist framework (Ellis, et al. 1994; Ellis & He 1999; de la Fuente 2002). 70 This next section will consider each of these approaches to the research. As for the first type of research, it has been argued that receptive learning (e.g. matching words with their L1 translations or definitions in L2) is less likely to go further than receptive knowledge, while productive learning (e.g. cloze exercise or writing tasks) is more likely to lead to productive knowledge (Griffin & Harley 1996; Waring 1997a). However, these results are based on learning word pairs, i.e., participants in the study were simply required to give the L1 or L2 translation equivalent in reception or production, respectively. In other words, vocabulary gains were only measured with tests of meaning and form in these studies. Given that knowledge of a word involves much more than simply knowing its meaning and form (Nation 2001), Webb (2005) attempted to test each target word on the knowledge of orthography, syntax, grammatical functions, association, and meaning and form both receptively and productively. In order to determine the relative efficacy of receptive learning and productive learning, Webb (2005) conducted ten different English vocabulary tests on Japanese college students. These tests covered the five aspects outlined above both receptively and productively. In Webb’s study there were two experimental groups: one group learnt the target words from reading three glossed sentences and the other group learnt them in a sentence writing task. The results showed that when the same amount of time was spent on both tasks, participants learnt better in the reading task. However, when participants spent as much time as they required to complete each task (the writing task required more time), the writing task was more effective. Webb (2005) strongly argues that writing a sentence is more effective than reading three sentences given that his 71 study showed productive learning to be superior to receptive learning not only in developing productive knowledge but also in gaining receptive knowledge. That is, learners can gain both productive and receptive knowledge through productive learning, while they can only attain receptive knowledge through receptive learning. The second type of research indicated by Webb attempts to estimate the respective sizes of receptive and productive vocabulary (Laufer 1998; Laufer & Paribakht 1998; Laufer & Nation 1999; Waring1997b). Laufer (1998) investigated the gains in three types of vocabulary knowledge, namely, “passive”, “controlled active” and “free active”26 for two groups of learners during one year of school instruction. Laufer (1998) found that passive vocabulary size, which was measured by Vocabulary Levels Test27, progressed very well, and controlled active vocabulary, which was measured by the productive version of the Levels Test28, showed modest progress. In contrast, free active vocabulary, measured by Lexical Frequency Profile29, did not progress at all. In addition, whereas the size of passive vocabulary was larger than that of controlled active vocabulary in both groups of participants, the gap between each type of vocabulary size enlarged in the more advanced group. Moreover, although passive and 26 Laufer uses two different terms for each concept interchangeably in her articles, i.e., “active” and “productive”, “passive” and “receptive”, respectively. 27 Vocabulary Levels Test (Nation 1983, 1990), which has been proved to be a useful tool for diagnostic vocabulary testing, consists of five parts, representing five levels of word frequency in English: the first 2000 words, 3000 words, 5000 words, the university word level (beyond 5000 words) and 10000 words. The test involves word-definition matching which was designed to minimise the chance of successful guessing by including as little reading as possible. 28 The active or productive version of the Vocabulary Levels Test (Laufer & Nation 1995, 1999) is the same in overall structure and the target words as the original version, but the test-takers are given a set of sentences with a blank. They are required to write the missing target word in each blank in which a variable number of initial letters are provided for each blank to ensure that no words other than the target word fit in. 29 Lexical Frequency Profile (Laufer & Nation 1995) is based on the relative frequency of words in the learners’ composition where the percentage of word families that belong to each frequency band (e.g. first 1000 most frequent words, second 1000 most frequent words, words in the University Word List, and other words) is calculated. 72 controlled active vocabulary size was well correlated with each other, the size of free active vocabulary did not show significant correlation with either of them. The third type of studies on receptive and productive vocabulary deals with the effects of input and interaction in vocabulary acquisition (Ellis, et al. 1994; Ellis & He 1999; de la Fuente 2002). The reason that these studies have not been mentioned in any literature reviews of receptive and productive vocabulary research (e.g. Read 2000; Webb 2005; Mondria & Wiersma 2004) may be because they are more likely to be associated with studies on “input and interaction” rather than “vocabulary”. Indeed, the main objective of such studies is to provide evidence for the role of negotiation in facilitating the acquisition of L2 vocabulary, i.e., whether or not learning is more effective when learners negotiate the meaning of a word with interlocutors than when they simply get premodified input. However, they are worth mentioning here to compare the use of the terms and interpretation of the results with other studies on receptive and productive vocabulary. Based on similar previous studies (e.g. Ellis, et al. 1994; Ellis & He 1999), de la Fuente (2002) investigated the effects of different oral interactions on the comprehension, receptive acquisition, and productive acquisition of new L2 words. In her (2002) study de la Fuente defined “comprehension” as the ability to correctly carry out an instruction: to identify target items on a map and place them on a worksheet. In order to measure both receptive and productive acquisition of new L2 words, de la Fuente used two (receptive and productive) vocabulary knowledge scales (taken from Wesche & 73 Paribakth 1996) where participants were asked which of the four points of the scale30 best indicated their vocabulary knowledge of each word. Thus, “receptive acquisition” was measured by the number of target words that each participant was able to understand whilst listening to these target words read aloud by the researcher. Likewise, productive acquisition was measured as the number of target words that each participant was able to produce where the researcher gave images (i.e. pictures) of each target word to the participants. The results showed that learners’ comprehension of instructions and the target words contained in these instructions was better when they had opportunity to negotiate. Both receptive and productive acquisition of the target words were more facilitated when they had opportunity to negotiate the target words than when they were exposed to premodified input. Furthermore, although production of target words during negotiation did not have an effect on learners’ receptive acquisition, it did have an effect on their productive acquisition. As we have seen, it is apparent that each of these three types of research that compare the receptive and productive vocabulary uses the very terms “receptive and productive vocabulary” in a different manner. Read (2000) claims that the terms “receptive (passive)” and “productive (active)” are too broad, and many researchers use two different definitions for each concept interchangeably, i.e., “recognition versus recall”, and “comprehension versus use”. 30 The four points in “Receptive Vocabulary Knowledge Scale” include: (1) I do not remember having heard that word before, (2) I have heard that word before, but I do not know what it means, (3) I have heard that word before, and I think it means ______(give L1 translation), (4) I know that word. It means ______(give L1 translation). The four points in “Productive Vocabulary Knowledge Scale” include: (1) I have never produced the word in (L2) for that image. I do not know it, (2) I have produced that word before, but I can’t remember, (3) I think that is a ______in (L2), (4) I know that is a ______in (L2). 74 Recognition here means that the test-takers are presented with the target word and are asked to show that they understand its meaning, whereas in the case of recall they are provided with some stimulus designed to elicit the target word from their memory (Read 2000: 155, emphasis original). For instance, participants give the L1 translation of the target word or choose the most appropriate definition of the target word from several alternatives in recognition tests, while they give the target word in response to its L1 translation or fill in a blank with the target word in a sentence in recall test (Read 2000). On the other hand, Comprehension here means that learners can understand a word when they encounter it in context while listening or reading, whereas use means that the word occurs in their own speech or writing (Read 2000: 156, emphasis original). In testing comprehension of a word, participants need to listen to speeches or read texts containing a number of target words in order to examine how well they understand the words in context. Similarly, in testing use of a word, target words are usually elicited by tasks such as story-retelling, translation or picture description (Read 2000). In other words, assessment of these aspects of vocabulary must be based on comprehensive and context dependant tasks. As Read (2000) asserts, it is a problem when the terms “reception” and “production” are used indiscriminately for these two separate distinctions. This is because we should not easily assume that learners who successfully complete recall tests can also use the target word correctly and appropriately in their speaking and writing. In particular, Read (2000) criticised Laufer & Nation’s (1999) interpretation of the blank-filling tests (i.e. Productive version of Vocabulary Levels 75 Test), which was aimed to measure controlled productive vocabulary knowledge. Their conclusion that “the approximate number of words at a particular frequency level which were ‘available for productive use’ ” (Read 2000: 41) conflicts with Laufer’s (1998) finding that controlled active vocabulary size did not correlate with free active vocabulary size. Yet, Read (2000) does not necessarily claim that “comprehension” and “use” are the only valid measures of receptive and productive vocabulary. Testing target words in a whole written or spoken text may be inefficient in terms of testing time and participants’ energy. The range of vocabulary that can be included in the test may also be restricted depending on the topic of the text. Moreover, participants may be able to infer the meaning of the target words in context (where the knowledge of the words is to be measured) or they may not use the target words in their writing or speech. Thus, in fact, researchers tend to prefer controlled and selective tasks in which target words are given individually or in short sentences (Read 2000). In summary then, it seems that researchers need to distinguish clearly between two separate definitions of the concepts of receptive and productive vocabulary and state clearly exactly what the test that they are using is measuring (i.e. recognition versus recall,andcomprehension versus use) regardless of the type of the tests they adopt. It is also crucial for researchers to realise that the ability to recall a target word by simply responding to a L1 translation equivalent or by filling a blank cannot be equated with the ability to use it correctly and appropriately in a meaningful context. Since the present research gives or elicits specific target words in a short sentence, I will choose relatively controlled and selective tasks. As Read (2000) claims, in free and 76 comprehensive tasks, such as story telling or interview, participants may not use the target words. In addition, many target words need to be examined by a number of participants, hence free and comprehensive tasks are not suitable for the purpose of the current research. For these reasons, the present research uses the terms “receptive and productive vocabulary” in a similar sense of “recognition and recall” (Read 2000), respectively. Yet, whereas the term “productive vocabulary” in the present research can be equated with “recall”, “receptive vocabulary” is not exactly the same as Read’s (2000) definition of “recognition” . Two types of receptive tests in the present research (i.e. acceptability judgement test and translation test) include some contextualized short sentences, hence participants may be able to guess the meaning of the target words. As the current research aims not only to measure the participants’ knowledge of each target word but also to examine how they infer its meaning, “receptive vocabulary” here is somewhat a combination of the concepts of both “recognition” and “comprehension”31. Accordingly, the present research does not assume that the participants are always able to use the target items correctly and appropriately in various contexts even when they could successfully recall the target items in the limited context of the present research. 2.3 Learnability of L2 Vocabulary One of the main issues in this chapter includes “learnability” or “ease or difficulty” of L2 words. This concept has not been extensively explored in the field of L2 vocabulary research, with the exception of a series of Laufer’s studies (Dagut 1977; Levenstone 31 The present study uses the term “reception” as a synonym of “comprehension” whose concept is a mixture of “recognition” and “comprehension” in Read’s (2000) sense. 77 1979; Takahashi 1984; Laufer 1988, 1989, 1990a, 1990b, 1990c, 1991,1992, 1997; Swan 1997). These studies were conducted in order to answer such questions as “why some foreign words are more difficult to learn than others” (Laufer 1990a: 573). This question motivates not only SLA researchers but also language teachers in that the answer may make it possible to decide on a different treatment for different words in terms of amount of explanation, practice, and selection for assessment items (Laufer 1990a). According to Nation (2001), different words have different learning burdens for learners with different language backgrounds. Nation (1990, 2001) shows the general principle of learning burden: the more an L2 word represents patterns and knowledge that learners are already familiar with, the less learning burden. These patterns and this knowledge can be from either their L1, L2 or any other languages the learner knows (Nation 2001). Laufer (1990a, 1997) also argues that both “intralexical factors” (occurring within the TL) and “interlexical factors” (caused by the influence of languages other than learners’ TL) affect the difficulty of L2 vocabulary acquisition. On the basis of her previous studies (Laufer 1988, 1989, 1990b) particularly on “synforms” (similar lexical forms), Laufer (1997) outlined ten facilitating and 12 difficulty-inducing intralexical factors32 which affect vocabulary learning. 32 The ten facilitating factors include: (1) familiar phonemes, (2) phonotactic regularity, (3) fixed stress, (4) consistency of sound-script relationship, (5) inflectional regularity, (6) derivational regularity, (7) morphological transparency, (8) generality, (9) register neutrality, (10) one form for one meaning. Twelve difficulty-inducing factors include: (1) presence of foreign phonemes, (2) phonotactic irregularity, (3) variable stress and vowel change, (4) incongruency in sound-script relationship, (5) inflectional complexity, (6) derivational complexity, (7) deceptive morphological transparency, (8) synformy, (9) specificity, (10) register restriction, (11) idiomaticity, (12) one form with several meanings. 78 Laufer (1988) classified synforms into ten categories, each representing a different type of similarity between the target word and the error produced, based on empirical research which tested more than 500 learners of English on all possible synformic confusions. The results of these studies showed that the most problematic synforms were those which differed according to suffixes (e.g. industrial / industrious; comprehensive / comprehensible) and synforms identical in consonants but different in vowels (e.g. adopt / adapt; proceed / precede). However, Laufer (1992) also found that the difficulty order of each type of synformy varied according to learners’ L1 background (i.e. Semitic, Germanic, and Romance). This may suggest that the learners of different L1 have different routes of progression towards the mastery of synformic confusions (Laufer 1992). Although Laufer (1997) explored the formal aspects (i.e. orthographic or phonetic) of intralexical difficulty-inducing factors extensively, she did not include semantic aspects in her lists with one exception: one form with several meanings. Yet, it has been suggested that semantic confusion may be more persistent than formal confusion within TL words in learners’ L2 development (Hulstijn & Tangelder 1991). Based on their empirical study, Hulstijn & Tangelder (1991) found that meaning similarity within the TL words caused more interference than form similarity for both native speakers and advanced learners but not for intermediate learners. As a result, they claimed that native speakers and advanced L2 learners continued to be susceptible to semantic interference, but are less likely to have formal confusion among TL words. In other words, acquisition of semantic aspects of L2 words takes much longer than that of formal 79 aspects as far as intralexical factors are concerned. Ard & Gass (1987) present similar views. Laufer (1990a) suggested three facilitating and four difficulty-inducing interlexical factors with regard to lexical correspondences between learners’ L1 and L2. This next section will outline these factors. The first factor is referred to as “similarity of form to L1 words” which can be seen to include both sides of the coin. That is, where L2 words are similar both in form and in meaning to words in the learners’ L1 (i.e. cognates), it facilitates the learning of the L2 words. On the other hand, where L2 words are similar to L1 words in form but different in meaning(s) (i.e. false or deceptive cognates), it will induce difficulty (Laufer 1990a). Laufer (1990a) also points out that whereas the facilitative effects caused by interlexical similarity may be considerable in comprehension, learners may use the L1 words in production without any modification due to their erroneous assumption: they are the “same” as L2 words, not “similar”. Moreover, even when cognates share similar meanings between L1 and L2, the gap in frequencies and register restrictions may give rise to their overuse, underuse, or inappropriate utterances due to the influence of the use of their L1 counterparts (Laufer 1990a). The second factor that affects the difficulty of L2 vocabulary is “meaning relations between words in L1 and L2”. On the basis of Dagut’s (1977) study, Laufer (1990a) argues that difficulty of L2 words may be caused by two types of one-to-many correspondence: “divergence” where one word in L1 is represented by several words in L2, and “convergence” where several words in L1 are equivalent to one word in L2. Such distinction of patterns of cross-linguistic correspondences recalls Stockwell, et 80 al.’s (1965) “hierarchy of difficulty” in which “split (=divergence)” is considered to be more difficult than “coalesced (=convergence)”. Similarly, Laufer (1990a) maintains that “divergence” is the difficulty-inducing factor, and gives no discussion about the ease or difficulty of “convergence”. One of the examples of error triggered by “divergence” is “It isn’t comfortable to live so far away (Hebrew noah means ‘comfortable’ and ‘convenient’)” (Dagut 1977: 225). According to Laufer (1990a), these errors arise from the learners’ application of one of the L2 lexical items to the whole (or a part of the) semantic area of the L1 lexical items that is not covered by the L2 word chosen by them (see Figure 2.1). L2: English L1: Hebrew Comfortable noah Convenient Figure 2.1: Semantic Divergence between Hebrew Word noah and English Corresponding Words It is important to note that Laufer (1990a) did not refer to Takahashi’s (1984) work which explored the Stockwell, et al.’s (1965) “hierarchy of difficulty” using a modified version of the hierarchy. Takahashi (1984) investigated the relative degree of difficulty for L2 learners dealing with four interlingual semantic patterns: congruence, convergence, divergence and semantic gap. In her categorisation, “congruence” is the pattern in which a pair of L1 and L2 items are mutually associated and congruent with each other. “Convergence” refers to the pattern in which two or more L1 items become coalesced into one item in the L2. On the other hand, “divergence” refers to the pattern in which one L1 item corresponds to two or more L2 items. “Semantic gap” is a condition which presents some discrepancy between L1 and L2 congruent items due to 81 certain collocational restriction (Takahashi 1984). For instance, Japanese adjective semai can be characterised by the features of (1) width (e.g. semai michi = narrow road) or (2) space (e.g. semai heya = *narrow (small) room), while English corresponding word “narrow” are characterised only by width. Thus, the equation of “semai” = “narrow” causes no problem in the sense of width (congruence) but it may trigger transfer error when used in the sense of space (semantic gap). Takahashi (1984) conducted a blank-filling test on 300 Japanese ESL learners of three different proficiency levels on each of above four categories, i.e., congruence, convergence, divergence, and semantic gap, using both adjectives and verbs. The results show that the difficulty order of each semantic pattern is generally consistent with Stockwell, et al.’s (1965) “hierarchy of difficulty”. Takahashi (1984) argues that her participants relied on one-to-one correspondence between L1 and L2 lexical items, and thus semantically congruent items tended to result in positive transfer, while semantically incongruent items were more likely to result in negative transfer. Takahashi (1984) gives possible reasons why the difficulty order of adjective condition was (congruence < convergence < divergence < semantic gap), while that of verb condition was (congruence < divergence < convergence < semantic gap). The inconsistent results of adjective condition and verb condition may be caused by interaction with the “specific exemplar”33 factor. That is, acquisition of verbs is more likely to be influenced by specific exemplar which is the first example or usage of the item acquired by the learners. Takahashi (1984) also believes that the reason for the 33 “Specific exemplar” is referred to as an exemplar that is actually chosen by the learner. The distinction between exemplar and specific exemplar can be identified with distinction between input and intake (Tanaka 1983, 1987). 82 consistent difficulty of “semantic gap” for the participants is due to its deceptive nature: the words in this category can be translated into primary equivalents (e.g. narrow = semai) in certain contexts, but not in other contexts (e.g. narrow semai). Takahashi’s (1984) results generally show that when the semantic ranges between a learners’ L1 and L2 are congruent, positive transfer tends to occur, while negative transfer may induce errors when L1-L2 semantic ranges are incongruent, even though there exist different levels of semantic discrepancy. In addition, specific exemplar factor is likely to interact with each semantic condition, so difficulty of each item may be subject to change. As a consequence, Takahashi (1984) indicates that some of the “coalesced” or “convergence” items can be more difficult than “split” or “divergence” items when learners rely on one-to-one correspondence. In other words, when learners create an association between an L2 lexical item and one of its L1 translation equivalents as a specific exemplar, which is usually the most basic or frequent usage of the item, they are likely to face difficulty in using other meanings or usages than the specific exemplar in L2. A further study that indicates the potential difficulty of “coalesced” or “convergence” is Imai (1993). This work investigated Japanese ESL learners’ meaning representation of an English verb, “WEAR”. It used a four-point scale acceptability judgement test on 17 correct and 13 erroneous usages of “WEAR” in a total of 30 short sentences. The results showed that Japanese ESL learners’ acceptability did not differ from an American control group for such items as “wear a uniform” or “wear a dress”, while Japanese ESL learners’ acceptability was significantly lower for such items as “wear a cap”, “wear shoes” or “wear a ring”. Imai (1993) argues that the results are caused by the semantic 83 discrepancy between the L2 word “WEAR” and its L1 translation equivalents: the former usages can be translated by the Japanese corresponding word “kiru” but the latter usages cannot. As a result, Imai (1993) concludes that Japanese ESL learners are dependent on one-to-one correspondence (i.e. wear = kiru) in their understanding of “WEAR”(seeFigure2.2). L2: English Object L1: Japanese (well uniform/dress kiru understood) Strong Association wear shoes haku (not properly cap kaburu understood) ring tsukeru Figure 2.2: Japanese ESL Learners’ Understanding of the English Word “WEAR” This example clearly shows that learners tend to have difficulty in the acquisition of an L2 word where multiple L1 items become coalesced into one item in the L2 due to their strong association between an L2 word and its primary equivalent or specific exemplar. However, Imai’s (1993) results are based on learners’ acceptability judgement. Thus, it is unknown how they will use this type of the L2 word in more authentic tasks such as reading texts or oral production. It may be worthwhile to further examine learners’ learnability of L2 words by reconsidering both types of semantic discrepancy (i.e. divergence and convergence) using various types of tasks. Other types of interlexical difficulty-inducing factors include “cultural differences” such as different connotative meanings across languages or lack of a certain concept (lexical voids) in learners’ L1 or L2, and “different collocations” in which L2 words co- occur with different types of words in L1. Along with her comprehensive studies on intralexical difficulty-inducing factors (Laufer 1988, 1989, 1990b, 1997), Laufer 84 (1990a) is of great value in that whereas there were some studies which argued each interlexical difficulty-inducing factor in isolation (e.g. Lado 1972; Dagut 1977; Levenstone 1979; Ringbom 1983; Takahashi 1984; Imai 1993), no studies attempted to combine all the possible factors and discuss them comprehensively. However, due to a relative lack of research in this area, some of the arguments in this article are based on non-empirical literature or her own teaching experience. Therefore, more empirical research is needed in this area. These interlexical difficulty-inducing factors will be further discussed in the following section in relation to L2 lexico-semantic development. 2.4 L2 Lexico-semantic Development Unlike child first language acquisition, L2 lexico-semantic development involves the mapping of lexical and conceptual systems between learners’ L1 and L2 (e.g. Strick 1980; Ijaz 1986; Sonaiya 1991; Zughoul 1991; Swan1997; Jarvis 1997, 2000; Mori 2003). Tanaka (1983) proposed a “Semantic Transfer Hypothesis” which assumes that language transfer is both pervasive and persistent in the domain of L2 lexico-semantics. This is largely due to the fact that: The two languages may differ in the number and nature of distinctions made within a common, shared concept or in the linguistic distinctions made between semantic categories. In many instances, a lexical item in the L2 cannot be directly mapped onto a concept existing in the L1, and the L2 learner has to restructure existing L1 concepts or develop a new concept that corresponds to a lexical item in the L2 (Ijaz 1986: 405). 85 Although relatively many errors which can be traced to failure in interlingual semantic mapping have been reported (Dagut 1977; Levenston 1979; Ringbom 1983; Laufer 1990a, Sonaiya 1991; Zughoul 1991; Lennon 1991, 1996; Singleton 1999), only a limited number of studies have attempted to comprehensively reveal the nature of L2 learners’ semantic development in relation to L1-L2 semantic mapping (e.g. Strick 1980; Ijaz 1986; Jiang 2000, 2002, 2004a). Despite its relatively neglected status, empirical studies on interlingual semantic transfer or L2 lexico-semantic development did begin to appear by the early 1980’s (Strick 1980; Tanaka 1983; Takahashi 1984; Ijaz 1986; Tanaka & Abe 1985; Tanaka, et al. 1990; Shirai 1995; Jarvis 1997, 2000) and some of these will be outlined below. Ijaz (1986) compared the meaning ascribed to English spatial prepositions by adult English speakers with those of advanced ESL learners. The results of a semantic- relatedness test and a cloze-type test indicated that the ESL learners differed substantially from the native speakers in the semantic boundaries ascribed to the words. For instance, many L2 learners avoided the use of those L2 words in which the L2 made finer semantic distinctions in a semantic range than did the L1. In addition, for those words in which the L1 made a distinction that was not made in the L2, the L2 learners also sought to make the same distinction linguistically salient in the L2. Finally, Ijaz (1986) concluded that word usage in a second language was shown to be strongly influenced by a “Semantic Equivalence Hypothesis” which suggests that conceptual patterns and linguistic coding practices in the learners’ L1 provide the essential criteria for those in the L2. 86 Jarvis (1997) also showed that where the conceptual structures underlying the L1 lexicon differ from those underlying native speakers’ use of the L2, the learner might face substantial difficulties. Jarvis (1997, 2000) investigated the English lexical reference of Finnish-speaking Finns and Swedish-speaking Finns using three elicitation tasks: a written narrative, and two types of lexical listing tasks. In the written narrative task, participants were asked to retell the story of an eight-minute silent film. In the first lexical listing task, participants were shown clippings of 27 preselected objects and events from the film and were asked to list all of the content words that they considered to be appropriate to refer to each of the objects or events. In the second lexical listing task, participants were provided with preselected lists of English vocabulary items to choose from as well as the same film clippings of the objects and events. The results showed that learners’ L1 background and L1-based concepts play an important role in determining the referential lexical choices in the L2. That is to say, L1-based concepts tend both to motivate and limit the lexical options that a learner refers to a certain object or event. Accordingly, Jarvis (1997, 2000) asserts that learners need to acquire not only the phonological and orthographical features of the L2 lexicon, but also new concepts in order to attain native-like competence in their L2 lexical reference or choice of appropriate lexical items in L2. Whereas most studies on interlingual semantic transfer mainly argue the incomplete or erroneous features of the IL (e.g. Tanaka 1983; Takahashi 1984; Tanaka & Abe 1985; Ijaz 1986; Sonaiya 1991; Zughoul 1991; Jarvis 1997, 2000), Swan (1997: 167) states: 87 …some kind of equivalence hypothesis is probably indispensable in second language learning, especially during the early stages. Mother tongue influence is responsible not only for errors, but also for much of what is correct in an interlanguage”. In other words, the Semantic Equivalence Hypothesis enables L2 learners to learn new languages without learning to categorise things in the world again or relabelling the concepts that they have already learnt in their L1 (Swan 1997). Singleton (1999) also claims that many of the meanings and meaning hierarchies that are lexicalised in learners’ L1 will be recyclable with only minimal adjustment in the L2. Viberg (1998) demonstrated this with an instruction-giving test in which participants were shown the actions on the video and assigned the task of instructing a fellow student in Sweedish using three equivalents of the English verb “PUT”: ställa, lägga and sätta. Results showed that Polish participants used appropriate Swedish verbs significantly more than the Finnish and Spanish participants. Viberg (1998) concluded that the Polish participants obtained positive influence from their L1, which had similar distinctions of the verbs, with respect to the semantic representation of the words. As Jarvis (1997) claims, although most L2 words have similar concepts or words in learners’ L1, their semantic content, originally transferred from L1, needs to be restructured or a new lexical entry needs to be created in order to obtain native-like proficiency in the L2. Thus, Jiang (2000, 2002, 2004a, 2004b) sees the acquisition of the lexico-semantics, like other linguistic features, as a developmental process. Jiang (2004b) observes that whereas relatively many researchers acknowledge that new L2 words are initially mapped to pre-existing L1 concepts (e.g. Strick 1980; Ringbom 1983; Graham & Belnap 1986; Giacobbe 1992; Shirai 1995; Jarvis 1997, 2000; Hall 88 2002; Jiang 2000, 2002, 2004a, 2004b), there is little consensus as to the extent to which L2 learners restructure their L1 semantic structures or create new concepts. While some researchers are optimistic about the L2 learners’ semantic restructuring, and believe that plentiful exposure to contextualised L2 input does work in forming new concepts or reorganising L1 semantic structure (Ringbom 1983; Giacobbe 1992), others argue that the restructuring process is slow and usually incomplete (Strick 1980; Ijaz 1986; Sonaiya 1991; Jarvis 1997, 2000; Jiang 2000, 2002, 2004a, 2004b). According to Jiang (2000, 2002, 2004a, 2004b), L2 learners’ incomplete restructuring in the development of their semantic structures may be caused by limited exposure to the L2 or their successful use of L2 words on the basis of L1 concepts. Jiang (2000, 2002) points out that although many of the translation equivalents between L1 and L2 do not share the identical concepts or semantic boundaries, the core meanings are often the same. Hence, L2 learners tend to use the L2 words without making errors on the basis of the L1 semantic structures. Even when learners produce an error, it is not always noticed by them. For these reasons, Jiang (2000, 2002) claims that increased exposure to the L2 input is less likely to result in restructuring. In other words, restructuring occurs only when the context of the input is powerful enough to make the learners aware of the gaps (subtle in many cases) in the semantic boundaries of the L2 words and their L1 translation equivalents. What is more, Jiang (2000, 2002, 2004a, 2004b) further claims that increased exposure will strengthen the connections between the L2 word form and transferred L1 concept due to repeated co-activation, presupposing that the L1 concept functions as “mediator” of the L2 word in the learners’ mental lexicon. This claim can be supported by the 89 connectionist model of second language representation and processing (e.g. Gasser 1990; Shirai 1992; Sokolik & Smith 1992; Ellis & Schmidt 1998) in which human behaviour is explained in terms of neuronal activation. Connectionists assume that linguistic knowledge is represented over many nodes34 and in the strength of connection / association between nodes. That is, the more often particular nodes are activated, the stronger the connections become. As a consequence, stronger connections become more easily activated, and which are considered to be learning (Shirai 1992). According to Harrington (2002), the connectionist models provide an explicit characterisation of the developmental process because they posit a single, unified account for learning and processing. Shirai (1992) argues that L2 learners tend to search L1-L2 equivalents in their use of L2, particularly of lexical items, due to their easier interlingual mapping and frequent use of such a strategy makes the connection / association of these pairs stronger. To be more precise, the connection between L2 lexical items and their L1 equivalents is actually a connection between L2 word form and L1 semantic content rather than L1 word form and L2 semantic content. Thus, such associative links between L2 words and L1 equivalents (i.e. L2 word forms mapped on L1 semantic contents) can be regarded as the L2 learners’ erroneous IL formed by interlingual semantic transfer unless the semantic range of the L2 words is identical with that of the L1 equivalents. To show evidence of incomplete semantic restructuring or L2 words mapped onto L1 concepts even in later stages of L2 development, Jiang (2002) demonstrated an online 34 “Node” is referred to as a place where knowledge is represented in the brain. For instance, the words “bread” and “butter” are represented on different nodes, but may have a strong connection due to their strong linguistic and conceptual association (Shirai 1992). “Node” is also called as “unit”. 90 semantic judgement task using 40 English word pairs that share the same Chinese translation, such as “problem” and “question” (both translated into wenti), and 40 related pairs that do not share the same translation in Chinese as well as 80 totally unrelated pairs. In the task, 25 advanced ESL learners of Chinese and 27 English native speakers were asked to decide whether the two words presented simultaneously on the screen were related in meaning. Prior to the task, Jiang (2002) hypothesised that the degree of semantic relatedness was negatively correlated with response time. That is, the more related the meanings of word pairs were, the more quickly participants’ would respond. Jiang maintains that this is because identification of a shared meaning is easier when two words are more closely related than when they are not. The results showed that while native English speakers did not differ in their reaction time for both same-translation and different-translation pairs, Chinese advanced ESL learners responded to the same-translation pairs more than100 milliseconds faster than to the different-translation pairs. Jiang (2004a) also conducted the same task for Korean advanced ESL learners using the Korean same-translation and different-translation pairs, and found similar results. As a result of these two reaction time studies, Jiang argues that those L2 word pairs which share the same translation in learners’ L1 share the same semantic content, and thus they are responded to significantly faster than those word pairs which do not share the same translation. Accordingly, Jiang (2002, 2004a) concluded that L1 semantic structures were still mediating L2 word use even for advanced L2 learners. Jiang (2002: 633) further claimed that: 91 …. a decrease in transfer errors may not necessarily imply a substantial change in the semantic content of L2 words but rather be a result of learning and using explicit lexical knowledge. In other words, even when learners have explicit knowledge regarding the correct semantic boundaries of the L2 words, it does not necessarily indicate that new semantic content has been integrated into lexical entries. Jiang (2000, 2002) explains this by giving an example of two English words, “hat” and “cap”, which share the same translation in Chinese (i.e. maozi). Chinese ESL learners often use these two words interchangeably because they are mapped onto the same L1 concept or lexical entries. Under those circumstances, both “cap” and “hat” are activated for the preverbal message (e.g. the concept of “cap”) before the speaker drawing on explicit knowledge and chooses the correct word to produce. As a result, learners sometimes produce “hat” where the context requires “cap” or vice versa. They can also occasionally correct their own mistakes or wrongly chosen counterpart of the same-translation pairs (Jiang 2000, 2002). The reason for the learners’ inconsistent use of these words and self-correction can be accounted for by the automaticity of the lexical retrieval or production (Jiang 2002). According to Jiang (2002), what is represented within the lexical entry is retrieved automatically and spontaneously. In contrast, what is represented outside a lexical entry or stored in a general memory system can also be retrieved for speech production, but it is not automatic and consistent. The accuracy of the speech production for the latter varies according to the learners’ focus of the attention: meaning or form (e.g. Han 2000). Learners can correct their own mistakes when they have the explicit knowledge 92 of the L2 words and when attentional resources are available (e.g. Kormos 2000). Hence, Jiang (2002) believes that when lexical choice is often accompanied by self- correction, L1 lemma mediation or L2 words mapping onto L1 lexical entries is still likely to be sustained. Jiang’s view that both of the same-translation pair are activated for the preverbal message seems to be congruent with the view of Poulisse (1993) and Poulisse & Bongaerts (1994). These studies suggest that semantically related words (both interlingually and intralingually) as well as the intended word are activated, and slips of the tongue or unintentional substitutions take place, when learners have less attention to the form. Jiang (2004b) also argues that her online reaction time studies (2002, 2004a) are extremely beneficial in revealing learners’ invisible mental representation and process. This is not only because variables are easily controlled (i.e. same-translation and different-translation pairs isolate L2 words’ relationship with L1 translation equivalents), but also because emphasis on speed makes participants’ performance unaffected by deliberate thinking and conscious application of explicit knowledge (Jiang 2004b). Harrington (2005) also argues that reaction time measures serve as stable index of L2 development. As a result of a lexical decision task carried out by ESL learners of two different proficiency levels and by English native speakers, Harrington found that reaction time differences systematically discriminated between the groups: advanced ESL learners responded faster than intermediate learners, and English native speakers responded faster than the other two groups. In addition to the fact that more frequently occurring words were responded to faster than less frequent words in all groups, accuracy scores were negatively correlated with reaction time. This indicates 93 that there is no trade-off in speed and accuracy, and increasing response stability reflects learners’ L2 development (Harrington 2005). Limitations of such online judgement tasks are acknowledged by Jiang (2004b). Like other studies on word recognition or mental representations, they are dependant on some kinds of inferences or assumptions, given that such tasks examine the target items independently from contexts, and the language used is far from authentic language. For these reasons, Jiang (2004b) conducted a more direct paper-based task which aims to investigate how learners use words in context: a revised version of the sentence completion task in Ijaz (1986). Jiang (2004b) points out that although Ijaz (1986) regarded all the correct responses in her sentence completion task (cloze-type test) as a reflection of learners’ semantic development, the test could not distinguish the actual semantic development and positive semantic transfer. Besides, Jiang also mentions that Ijaz’s (1986) participants were free to use explicit lexical knowledge in the test. In an attempt to overcome the shortcomings of Ijaz’s test, Jiang (2004b) used six pairs of English words which share the same Chinese translation (criterion / standard, complicated / complex, accurate / precise, safe / secure, insist / persist, doubt / suspect) as target items. Ten advanced Chinese ESL learners and ten English native speakers were asked to choose one answer out of four choices to fill in the blank in a sentence. The purpose of the test was to examine participants’ ability to use these word pairs distinctively in each context, so four possible answers were prepared as in the following: (1)= Only “criterion” is appropriate, (2)=Only “standard” is appropriate, (3)=Both words are appropriate, and (4)=not sure. The results showed that whereas the overall average accuracy rate of English native speakers was 94%, that of Chinese ESL 94 learners was 65%. As a consequence, Jiang concluded that Chinese ESL learners’ accuracy level reflected the level of semantic development that had occurred on the grounds that the target items were not distinguished in their L1, and for which minimum explicit knowledge was available. Therefore, their performance cannot be attributed to positive transfer or the involvement of explicit knowledge (Jiang 2004b). Nevertheless, the method of investigation has made it difficult to generalise the results of Jiang’s study. Firstly, the participants were regarded as advanced Chinese ESL learners, but it is hard to determine what makes them advanced learners. Although Jiang (2004b) indicated that their years of formal instruction were from eight to 20 years, years of residence in the USA were from one to five years, and their self-rating of their English proficiency in ten point scales ranges from four to nine (1=minimum, 10=near native), such data does not guarantee that they are in an advanced level. Furthermore, the proficiency levels of the participants may differ to a large extent. This can be seen in the results of her sentence completion task that the accuracy rate largely varied from participant to participant (from 35% to 90%). Secondly, while Jiang (2004b) stresses the importance of controlling learners’ use of explicit knowledge, it is questionable whether sufficiently control was achieved. Jiang (2004b) says that the differences between these six pairs of English words are rarely explained in instruction based on her long experience of learning and teaching English in China. However, even though her participants had not been explicitly taught such differences in the classroom, they might be able to learn them through other sources (e.g. looking up dictionaries or asking English speaking friends). Moreover, a fundamental question arises: why did Jiang (2004b) attempt to control explicit 95 knowledge in the written multiple-choice sentence completion task in the first place? While her online judgement tasks (Jiang 2002, 2004a) were suitable for eliminating participants’ explicit knowledge in that they were instructed to respond as quickly and accurately as possible, it was not necessarily the case for the offline sentence completion task. Participants were given as much time as they required in this paper- based task where explicit knowledge was likely to be used if it was available. Finally, the most serious problem of Jiang’s (2004b) paper-based task might be the choice of target items and lack of comparison in the results with learners of other L1. Jiang (2004b) specifically selected the above mentioned six pairs of the same- translation words whose semantic differences were subtle and hard to describe. While Jiang (2004b) adopted only the example sentences in which nine out of ten English native speakers chose to use the same word, she reported that there was much disagreement among the native speakers in their word choice between these pairs. This inconsistency even among English native speakers indicates that differentiating the pairs is an extremely hard, if not impossible, task for non-native speakers regardless of their L1 background. Indeed, the difficulty caused by semantically similar lexical items within the L2 has been found to trigger more errors than any other kind of lexical errors (Martin 1984; Hulstijn & Tangelder 1991; Takahashi & Tanaka 1992; Hiki 1995). Thus, the results of this task cannot be easily ascribed to Chinese participants’ same- translation effect or L1 lemma mediation unless learners of other L1 backgrounds significantly outperformed them. Jiang (2004b) analysed the difficulty in L2 learners’ semantic restructuring (e.g. successful use of L2 words based on transferred L1 meaning) by giving an example of 96 the same-translation pair (standard / criterion = biaozhun). According to Jiang (2004b), Chinese ESL learners are unlikely to face misinterpretation or comprehension difficulty in their receptive use of these L2 words, because each word will activate all of its semantic elements contained in the broader L1 concept (see Figure 2.3). On the other hand, the learners may use these words incorrectly or create awkward sentences in their production since the concept of both “standard” and “criterion” are activated for the preverbal message, and one of them may be produced unintentionally (see Figure 2.4). L2: English L1: Chinese L2: English L1: Chinese standard standard biaozhun biaozhun criterion criterion Figure 2.3: Receptive Use of Same Figure 2.4: Productive Use of Same Translation Pairs Translation Pairs Despite Jiang’s (2004b) comprehensive analysis both in receptive and productive use of such L2 words, it seems that the word pair “standard” and “criterion” is not a good example to explain this. As Jiang herself notes, the distinction between “standard” and “criterion” is subtle and even English native speakers’ use of these words are not consistent in many cases. This, again, indicates that not only the same translation effect but also the semantic similarity of the word pair itself may be the cause of the difficulty in Chinese ESL learners’ productive use of these pairs. As reviewed earlier, the moderate version of the CAH (Oller & Ziahosseiny 1970) predicts that learning will be the most difficult where the most subtle distinctions are required, although this is not necessarily the case in comprehension. 97 In summary, the series of studies conducted by Jiang (2000, 2002, 2004a) has significantly contributed to our understanding of L2 learners’ semantic development both theoretically and empirically. However, the validity of the paper-based sentence completion task in Jiang (2004b), which was intended to complement her findings in uncontextualised online judgement tasks, can be questioned. The difficulty of the six same-translation pairs used in this study might also be caused by their subtle semantic distinction. For this reason, Jiang’s (2004b) results cannot be directly attributed to Chinese ESL learners’ failure in semantic restructuring. Hence, research on L2 learners’ semantic development in more naturalistic situation is necessary to take into account the following points: (1) compare the test results not only with the native speakers but also with the learners of other L1 background(s) in order to distinguish CLI from other factors (i.e. intralingual and developmental factors), (2) explicit knowledge cannot be eliminated due to the fact that learners use both implicit and explicit knowledge in their actual use of the L2. One further limitation of Jiang’s work (2000, 2002, 2004a, 2004b) may be that Jiang only targeted same-translation pairs in which one L1 word corresponds to two L2 words (i.e. “divergence”), and did not investigate the learners’ difficulty where two L1 words correspond to one L2 word (i.e. “convergence”). This is because empirical studies (e.g. Takahashi 1984; Imai 1993) showed that not only “divergence” but also “convergence” might induce learners’ difficulty due to their one-to-one correspondence of L1 and L2 lexical items. Moreover, Jiang assumes L2 learners’ semantic development basically as a one to one (or at most a one to two) mapping of L1 concept and L2 form(s). Thus, Jiang’s model cannot account for the L2 learners’ semantic development involving more complex semantic mapping, namely polysemy. 98 2.5 Polysemy and Prototype Most basic verbs are polysemous and tend to have many extended meanings or usages as well as a core meaning (Tanaka 1990; Shirai 1990). Nagy (1997: 68) describes the nature of polysemy: When words in two languages overlap only partially in meaning, it is likely that some of the non-overlap is due to irregular or conventionalised polysemy. The pervasive lack of one-to-one mapping between words in two languages, seen even in the most abridged bilingual dictionaries, suggests that a substantial amount of the polysemy in any given language is language-specific, and hence irregular. For the purpose of understanding L2 learners’ semantic development of such words, it is important to consider not only interlingual lexical mapping, but also the notion of “prototype” (Shirai 1995). Prototype theory has been developed in the field of psychology with regard to human categorisation. The basic concept of the theory is that most natural categories are composed of a core member which is the clearest or best examples of the category and of other category members of decreasing similarity to the core member (Rosch 1973; Taylor 1989). Berlin & Kay (1969) investigated colour terms in many different languages, and found that although the boundaries of different colour terms were arbitrary in each language, the best example for the colour chosen by the speakers of different languages was very similar. For instance, the colour “red” includes many different shades: some are close to orange, pink or purple. However, the focal point of the red is crimson or bright red, and this colour is likely to be chosen almost universally as the best example of “red” despite 99 the different boundaries from other colours in varied languages (Berlin & Kay 1969). Based on a series of psychological experiments, Rosch (1973) claimed that “robin” or “sparrow” was a typical member of “birds”, while “penguin” or “ostrich” was the least typical member. These studies suggest that even the constituents of the same category may range from typical members (i.e. prototype) to less typical (i.e. peripheral) members. The concept of prototype has been widely used in such fields as psychology (e.g. Berlin & Kay 1969; Rosch 1973), linguistics, particularly within cognitive semantics (e.g. Fillmore 1975; Lakoff 1987), first language acquisition (e.g. Bybee & Slobin 1982; Shirai & Andersen 1995) as well as SLA (e.g. Tanaka 1983; Ijaz 1986; Andersen & Shirai 1994; Hiki 1995; DeKeyser 1995; Hu 2002). Due to such wide usage, the methodology and interpretation of prototype-related research often differ (Tanaka 1990; Sugaya 2004). In addition, even in the field of SLA, the definition of prototype, or the way it is applied to the research, frequently varies from researcher to researcher (Hiki 1995; Shirai 1995; Sugaya 2004). This discrepancy within the field of SLA may give rise to confusion (Hiki 1995; Shirai 1995; Sugaya 2004).35 Shirai (1995) gave an example of conflicting claims made in previous prototype-related studies in SLA: Whereas Kellerman (1978, 1979) claims that prototypical items are more transferable to L2, Tanaka & Abe (1985) assert that negative transfer tends not to 35 Although different studies have used different terms in representing the concept of “prototype” such as “core” “unmarked” “best example” or “central example” each of them can also be regarded as prototype- related studies (Hiki 1995; Shirai 1995). For instance, Kellerman (1986) mentioned that the term “markedness” used in his earlier works was equivalent to “prototypicality”. Whereas prototype-related studies in SLA are not restricted to lexicon or polysemy in a narrower sense, this review mainly focuses on lexical items or polysemy because the acquisition of grammatical items such as tense-aspect morphology (e.g. Shirai & Andersen 1995; Shirai 2002; Sugaya 2002) or others (Yamaoka 1988; DeKeyser 1995; Hu 2002) is beyond the scope of this review. 100 occur in the case of prototypical items. However, in a closer analysis, Shirai (1995) found that Kellerman’s concern was only about the prototypicality of learners’ L1, while Tanaka & Abe were mainly arguing about that of learners’ L2. Moreover, other studies use native speakers rather than learners of the language as participants to determine prototypicality (e.g. Shirai 1990; Tanaka 1990). As a consequence, Shirai (1995: 63) identified three notions of prototype: (1) Second language prototype (L2 prototype), which is subcategorized into the following two: (1a) Native speaker prototype (NS prototype), which is the prototype that native speakers of the target language have. (1b) Interlanguage prototype (IL prototype), which is the prototype that L2 learners have about the target language. (2) First language prototype (L1 prototype), which is the prototype that second language learners have about their mother tongue. Others have classified these three notions differently from Shirai (1995). Hiki (1995) categorised each notion of the prototype independently, and referred to the first category (1a) as “target language (TL) prototype”. On the other hand, Sugaya (2004) divided these three notions into two approaches. The first includes the TL prototype and L1 prototype equivalent to (1a) and (2), respectively. The second approach is an IL prototype which corresponds to (1b). Despite the use of different terms or different ways of subcategorisation, these three studies basically argue the same concepts of prototypicality. The reason for the different classification adopted by Sugaya (2004) may reflect her view of prototype: both NS (TL) and L1 prototype items are more easily 101 learnt or transferred because they are universally easier to be recognised, while IL prototype involves with learners’ prototype formation (so the latter is fundamentally different from the former). In many SLA studies, however, NS (TL) prototype is often utilised as the criteria to compare with IL prototype, i.e., whether or not the latter approximate to the former (e.g. Ijaz 1986; Tanaka, et al. 1990; Shirai 1995; Hiki 1995). In addition, Shirai (1995) claims that a series of Kellerman’s studies (1977, 1978, 1979, 1983, 1986), which specifically deals with L1 prototype, focuses not on factors that determine acquisition itself, but on factors that determine transferability. On the other hand, Tanaka (1983) and Tanaka & Abe (1985), mainly argue for L2 prototype, and are concerned more with acquisition or learnability (i.e. which items are more easily learnt than others). That is to say, the L1 prototype correlates with transferability, while the L2 prototype correlates with learnability (Shirai 1995). The former concept was originated in SLA (i.e. Kellerman 1978, 1979), and has been used only in SLA, while the latter was based on Rosch’s (1973) claim that prototypical items in a category were acquired more easily than peripheral items. For these reasons, Shirai’s (1995) definition and terminology of prototype are more convenient when arguments are mainly concerned with L1 prototype. Therefore, the current research, whose focus is transferability and L1 prototype, follows Shirai’s (1995) terminology. In this way, it is easier to discuss L1 prototype as opposed to L2 (NS and IL) prototype than putting NS and L1 prototype in the same category as opposed to IL prototype (i.e. Sugaya 2004). Another important distinction that needs to be made when comparing the results of prototype-related studies in SLA is how prototypicality is measured. Tanaka (1990) 102 claimed that two different types of prototypicality had not been clearly distinguished and had confounded researchers. “Theoretical prototype” is based on “concreteness” because it is involved with the physical phenomenon, while “psychological prototype” is based on the awareness of a certain group of people. In other words, theoretical prototype is measured and determined by researchers deductively in their linguistic or typological analysis, so it can be equated with the concept of “markedness” in linguistic typology (e.g. Eckman 1977, 1985). On the other hand, psychological prototype deals with the target population’s awareness of a particular item: which items within the same category are psychologically more salient than others. Despite the different criteria depending upon such variables as culture, age or gender of the target group, it is possible to obtain a stable mean value for a certain group of people (Tanaka 1990). To determine the prototypicality of lexical items in SLA studies, various methods have been adopted. Corpus data has been used to identify the prototypicality due to its capability of calculating the accurate frequency of the target items in large body of texts (Tanaka1990; Shirai 1990). Tanaka (1990) used an English corpus of 7.3 million words (Birmingham University) to analyse the frequency of an English polysemy “HOLD”. Shirai (1990) used corpora of both spoken and written English of 120,000-word each (UCLA Oral Corpus and Brown Corpus, respectively) to analyse the frequency of another polysemy, “PUT”. However, one shortcoming of such analyses is that corpus data tends to be influenced or biased according to its topic or genre. Thus, it is difficult to find a corpus that truly reflect and represent the target language in general (Shirai 1990, 1998). 103 Free elicitation tests, in which participants write the most typical sentence(s) using the target item, have also been used to determine both NS prototype (Ijaz 1986; Shirai 1990, 1995; Tanaka 1990; Tanaka, et al. 1990; Matsuda 2000) and IL prototype (Ijaz 1986; Tanaka, et al. 1990; Shirai 1995; Matsuda 2000). As in the case of corpus data, the identification of prototype in this method is based on the frequency of the target item. Hence, these methods are usually followed by researchers’ linguistic analysis (Ijaz 1986; Shirai 1990, 1995; Tanaka 1990; Tanaka, et al. 1990; Matsuda 2000) since the characteristics or the features of the prototype need to be extracted from the most frequently appearing usages of the polysemy in corpus data or participants’ production. Other methods that have been used to determine prototypicality in SLA studies include prototypicality judgement test, similarity / dissimilarity judgement test, and card sorting task. In his case study, Tanaka (1990) asked 20 English native speakers to rate the prototypicality of 36 different usages of the English polysemy “HOLD” in a nine-point scale. Hiki (1995) conducted a prototypicality judgement test to 113 ESL learners and 33 English native speakers in which they were asked to make a judgement on English verbs of perception, “SEE, LOOK AT, HEAR and LISTEN TO”, using a seven-point scale (1=very good example, 7=very bad example). Hu (2002) also conducted a prototypicality judgement test on Chinese ESL learners on English grammatical rules such as article or tense, but participants had only two choices: which usage of the target item is more typical (e.g. A dog is man’s friend. / A dog is chasing a cat.). Kellerman (1978) examined the prototypicality (“coreness” or “markedness” in his words) of the Dutch polysemy “BREKEN” using a dissimilarity judgement task and a card sorting task before performing the transferability experiments (mentioned in the 104 previous chapter). In the dissimilarity judgement task, 35 participants were given a pair of sentences using “BREKEN” and were asked to judge the degree of dissimilarity of each “BREKEN” on a five-point scale. In the card sorting task, 50 participants sorted 17 different sentences using “BREKEN” on separate cards into groups of similar meanings. Kellerman analysed the data from each task multidimensionally, and obtained the rank order of coreness for each usage of “BREKEN”. Based on the results of these tasks, Kellerman (1978, 1979) found that the rank order of the coreness was strongly correlated with that of the transferability of the Dutch “BREKEN” into the English corresponding word “BREAK”. Tanaka (1990) believes that the prototypical items identified by theoretical prototype and psychological prototype are not always the same. For instance, “break the vase” maybeamoreprototypicalusagethan“break the tradition” or “break one’s heart” from the perspective of theoretical prototype, because it involves concrete action. Yet, Tanaka (1990) supposes that English native speakers may perceive “break the tradition” or “break one’s heart” as also being prototypical usage because of such factors as frequency, social / cultural importance, cognitive saliency or ease of memorisation. Indeed, although Tanaka (1990) found that the prototypical usage of “HOLD” determined by a corpus data and free elicitation task (i.e. subject [+human] hold [continue to have] object [+solid]) generally corresponds to the result of his prototypicality judgement task, there were also some items that English native speakers rated higher where objects were not necessarily solid or concrete as in “hold a meeting” or “my luck continues to hold”. 105 Apart from this, Shirai (1990) reports on the discrepancy between the prototypicality identified by a free elicitation task and corpus data. Shirai (1990) asked 42 English native speakers to write what they considered to be the most typical sentence using an English polysemy “PUT”, and found that the most frequently produced sentences were similar to “X put(s) a/the book on the table/desk”. As a result, Shirai determined that the prototypical usage of “PUT” was the following: subject [+human] put [physical transfer] object [+solid] locative [+horizontal surface, +small, +alienable]. However, this prototypical usage of “PUT” (i.e. physical transfer) did not appear as frequently as that of non-physical transfer or metaphoric extensions in the oral or the written corpus. This result implies that the actual frequency of use in native speaker discourse does not always match with the prototype determined by free elicitation task. In his conclusion, Shirai (1990) speculates that the use of psycholinguistic procedure, in which participants rate the prototypicality or idiomaticity of the target expressions, may also bring about different results. One possible limitation of such psycholinguistic rating is that it must be based on continuous rather than dichotomous measures, so deciding on the cut-off point may be difficult (Shirai 1990). Shirai (1998) also points out that identification of the prototypicality may be much more difficult if the judgement task is on grammatical items, since there may be a number of other factors that affect people’s judgement other than the nature of the target structure itself (e.g. combination of words). Moreover, free elicitation tasks are only available to those participants who are familiar with the target grammatical terms such as “relative clause” or “aspect” (Shirai 1998). However, the identification of prototypicality for lexical items seems to have less problems. 106 As shown above, previous prototype-related studies in SLA suggest the importance of clarifying the definition of prototype, i.e., whose prototype is discussed and how it is measured, in order to satisfactorily investigate the learnability or transferability of lexical items. In fact, most studies in this paradigm have focused on L2 prototype to examine the learnability of lexical items (Tanaka 1983; Tanaka & Abe 1985; Tanaka, et al. 1990; Ijaz 1986; Shirai 1995; Hiki 1995), and had similar findings: prototypical items are easier to be learnt. In Tanaka (1983), the prototypical or basic meaning of the English locative “IN” was determined as “containment” or “enclosure” on the basis of semantic analysis from previous descriptive studies (i.e., theoretical prototype). Thus, “The bird is in the box” was regarded as prototypical, while “The bird is in the tree” was not. As a result of multiple-choice cloze-type tests, where 320 Japanese ESL learners were asked to choose the most appropriate prepositions36 according to the pictures given, Tanaka (1983) found that prototypical usages were generally regarded as easy regardless of positive transfer items or negative transfer items.37 However, the results also showed that extended or less prototypical usages were more difficult in negative transfer items than in positive transfer items. Ijaz (1986) asked 17 native speakers of English to write sentences using the English prepositions “ON, ON TOP OF, ONTO, OVER, and ABOVE” in order to determine the 36 Prepositions are often categorised as “function words” as opposed to “content words”, and thus regarded as less typical member of lexicon. However, in this review, they are considered to be an analogous item to polysemy of content words due to their multi-functional or polysemous nature. 37 Positive transfer (or transferable) items are referred to as the L2 words that can be translated into the primary equivalent or most easily associated counterpart in L1 as in “John is standing on thebed”(on= ueni). Negative transfer (or non-transferable) items, on the other hand, are the L2 words that cannot be translated using the primary equivalent as in “There are many apples on the tree” (on ueni). 107 prototypical meaning features for each preposition. For instance, the most typical usage of “ON” was found to be [+Contact +Verticality -Movement] as in “I sat on the lawn”. Ijaz also conducted a sentence completion (multiple-choice cloze type) test to compare preposition choice between ESL learners of two different L1 (German and Urdu) and native speakers. The results showed that whereas there was no significant difference among these groups in the case of prototypical meanings, ESL learners’ choice of preposition was significantly deviated from that of native speakers in the case of non- prototypical meanings. In other words, learners’ knowledge of prepositions tends to approximate native speakers more closely in the prototypical meanings, but it tends to be more affected by their L1 in non-prototypical ones. Tanaka, et al. (1990) elicited the most typical usage of the English polysemy “MAKE” from 300 Japanese ESL learners of three different proficiency levels using free elicitation tasks. They found that the prototype of “MAKE” for Japanese ESL learners takes a noun phrase having the semantic features [+concrete object, + small size] as in “make a cake” or “make a doll”. The results of an acceptability judgement test on non- prototypical instances of “MAKE” (e.g. make an atmosphere, make a turn) show that transferable (positive transfer) items were more highly accepted than non-transferable (negative transfer) items by the least proficient group but not the other groups. However, prototypical items were judged to be more acceptable than non-prototypical items regardless of participants’ proficiency level. Hiki (1995) conducted a prototypicality judgement test and a multiple-choice cloze type test on two sets of perception verbs which have similar meanings, i.e., “HEAR / LISTEN TO” and “SEE / LOOK AT”. Hiki found that learners’ lexical choice of these 108 verbs was very different from that of the native speakers. In addition, IL prototype was found to be associated with learners’ lexical choice, and NS (TL) prototype did not reflect their lexical choice. However, one limitation of her study is that it disregards CLI due to the diversity of participants’ L1. Therefore, it cannot elucidate the relationship or interplay between prototype factors and transfer factors, both of which are likely to influence the learnability of polysemy. Shirai (1995) asked three different levels of Japanese ESL learners (N=212) and 41 English native speakers to write one sentence that contains the most typical use of the English polysemy “PUT”, and two more sentences with its typical instances. The prototypical usage of “PUT” was determined based on the native speakers’ response (i.e. physical transfer), but it was also produced most frequently by all of the three groups of ESL learners. As a result of a seven-point scale acceptability judgement test on 20 different instances of “PUT”, including both positive transfer items and negative transfer items, Shirai (1995) found that Japanese ESL learners’ development regarding the knowledge of “PUT” was constrained by both prototype and transfer factors: less prototypical items within the L2 are more difficult particularly where L1 positive transfer is not available (i.e. negative transfer items). The results of these L2 prototype-based studies on polysemy commonly show that both intralexical semantic factors (i.e. prototypicality) and interlexical semantic factors (i.e. transfer) affect the acquisition of polysemy in L2, and they interact with each other to determine its ease or difficulty. More specifically, prototypical items are generally easier to learn regardless of potential transferability (i.e. positive transfer items or negative transfer items), while less prototypical items are more difficult, particularly in 109 the case of negative transfer items. Hence, L2 acquisition of polysemy cannot be argued without discussing these two factors: prototype and transfer. However, since Kellerman’s pioneer studies (1977, 1978, 1979, 1983, 1986) and despite the relative abundance of studies focusing on L2 prototype, no research has further explored the relationship between L1 prototype and transferability. As reviewed earlier, although Kellerman found a strong correlation between the prototypicality and transferability of L1 lexical items, many limitations have also been found in Kellerman’s framework. Obviously, more work is needed to investigate the relationship between L1 prototype and transferability in order to complement Kellerman’s studies particularly by eliciting L2 learners’ transfer in their L2 performance. 2.6 Cognates A cognate is defined as “a word in one language that has the same origin as a word in another language” (Longman Dictionary of Contemporary English 1995). This suggests that cognates refer only to words etymologically related such as the German word hund and the English word hound. However, in the field of SLA and psycholinguistics, there are different views. Odlin (1989) indicates that the term “cognate” in SLA research is broader than the use of the term by historical linguists: any lexical similarities may be considered as cognates. Carroll (1992: 104) also argues that “cognates are any pair of words which are treated by learners as belonging to distinct linguistic systems but are also treated as ‘the same thing’ within those systems”. Thus, both Odlin (1989) and Carroll (1992) suggest the existence of etymologically unrelated cognates. 110 It has been reported that some 3,000 English words have cognates in Spanish, and a few simple rules allow these words to be expanded into 10,000 Spanish items (Lobo 1966, cited by Meara 1993). Moreover, it has also been claimed that French and English share some 6,500 homographic cognates, which share the identical orthography, and some 17,000 parographs which are cognates but not identically written (Séguin & Tréville 1992, cited by Meara 1993). If we assume that these cognates share the similar orthography (or pronunciation) as well as the meaning, L2 vocabulary acquisition should be greatly facilitated (Ringbom 1978, 1987, 1991, 1992; Palmberg 1985, 1987; Odlin 1989; Nation 1990, 2001; Laufer 1990a; Hall 2002). However, as Carroll (1992) states, cognates present a paradox for learning theory: while they appear to facilitate the learning of a foreign language by allowing unknown words to be easily recognised, cognate-paring also appears to hinder long term learning. This is because the so-called “false cognates” or “false friends” may cause erroneous production or misinterpretation. For example, whereas the word fabrique in contemporary French means “factory”, the word fabric in contemporary English means “cloth” or “material” (Granger 1996). Yet, the number of false cognate or deceptive cognate38 is limited. According to Hammer & Monod (1976), the ratio of “good cognates”, which share the same meanings between learners’ L1 and L2, to “deceptive cognates” between French and English is 11:1. Moreover, not all deceptive cognates are totally unrelated in their meanings (James 1998). There are also some cognates which have partial semantic overlap between two languages: 38 Although the three terms “false friends”, “false cognates” and “deceptive cognates” are used synonymously and interchangeably in the present study, “deceptive cognates” will be used mainly except where authors of each literature use other two terms. 111 One of the most common problems in second language acquisition is when there is only a partial semantic identity of cognates. Thus, the translation of English succeed into Spanish as suceder will be acceptable in some contexts but not in others: For example, while Truman sucedió a Roosevelt (“Truman succeeded Roosevelt”) is acceptable Spanish, Sucedió en su trabajo (“He succeeded in his work ”) is not (Anthony 1952-3, cited by Odlin 1989: 79). The reason for such semantic discrepancy or partial overlap of cognates across languages is that words begin to live their own life in the new language, while the words in the donor language continue to evolve independently (Granger 1996). Ringbom (1987) collected a very large corpus of lexical errors of Swedish and Finnish learners of English, and classified “false friends” into three types of correspondences: (1) Cognate form but completely different semantics for the forms, (2) Cognate form with very limited similarity in semantics, and (3) Cognate form with partial semantic equivalence. However, the distinction between (2) and (3) seems to be unclear in this classification. Granger (1996) distinguished “totally deceptive cognates” from “partially deceptive cognates”. The former refers to historically related words in L1 and L2 that were synonymous but have totally divergent meanings now (see Figure 2.5). The latter refers to L1 words that correspond to two or more words in L2 or L2 words that correspond to two or more words in L1. Van Roey (1990) explored the cognate relationship between English and French, and subdivided partially deceptive cognates into two main categories: “inclusion” and “overlapping”. Inclusion is where the semantic space of one cognate is included in that of other, either all the meanings of the English word are 112 included in those of the French counterpart (F>E) or vice versa (F Overlapping, on the other hand, is where there is a common semantic space but there are also specific extensions of meaning for both cognates (see Figure 2.7). For instance, the English word fatal and the French word fatal share the meaning of “having a very bad effect” but fatal in English also can mean “resulting in someone’s death” and the one in French also mean “inevitable” (Granger 1996). Fabric Fabrique (English) (French) Figure 2.5: Totally Deceptive Cognates (F>E) (F dossier gang (English) (French) dossier gang (French) (English) Figure 2.6: Partially Deceptive Cognates (Inclusion) fatal fatal (English) (French) Figure 2.7: Partially Deceptive Cognates (Overlapping) 113 According to Granger (1996), partially deceptive cognates induce further difficulties for learners. For instance, the meaning of the English word dossier is much more restricted than the corresponding French meaning. Whereas the French word dossier refers to any collection of documents, its English counterpart is used only for confidential information. Such nuances of meaning typically boost the difficulty of learning partially deceptive cognates because even the so-called “common meanings” are rarely identical (Granger 1996). Another factor that hinders accurate learning of cognates is the different patterns of cognancy between languages, i.e., the frequency or register of cognates is not necessarily the same across languages (Laufer 1990a; Meara 1993). Romance vocabulary in English is generally less frequent and more formal than equivalent Germanic vocabulary (Meara 1993; Hancin-Bhatt & Nagy 1994). For instance, while the word embargo in English has a specific technical sense and is rarely used, embargo in Spanish is used in everyday life. Accordingly, Meara (1993) suggests that teaching Spanish speakers to rely heavily on their English cognates may be misleading, because they may have a poor understanding of the everyday words. Moreover, they may speak in an inappropriate or (at worst) offensive register (Meara 1993). Meara (1993) also claims that there are four main patterns of cognancy. First, where L1 and L2 share very few cognates such as English and Arabic. Second, where L1 and L2 are closely related, and the bulk of the vocabularies between the two languages are mapped by some simple phonological rules. Italian and Spanish, or Russian and Polish are examples of this pattern. While the first and second patterns are relatively simple, the third and fourth are more complicated. The third pattern is where a large proportion 114 of the basic L2 words are cognate with the L1, but the less frequent and more difficult items in the L2 are non-cognate. German speakers learning English are an example of this pattern. The fourth pattern is where the cognates do exist between two languages, but their use is restricted to certain domains or registers. This can be seen in the case of Romance speakers learning English, because Romance words in English tend to be less frequent and more formal. CBLs’ learning of Japanese may correspond to the fourth pattern given that Japanese and Chinese share tremendous number of cognates in the terminology of natural science or social science, while they share relatively less vocabulary in daily conversation (Matsushita 2002). Some studies have concentrated on the issue of cognates in the field of applied linguistics, and this can be seen to reflect the importance of cognates in teaching or learning foreign languages, particularly when the students in the classroom share the same L1. However, most of these studies are pedagogically oriented (e.g. Anthony 1952-1953; Sheen 1979; Banta 1981; Meara 1993) or pure descriptive studies (e.g. Hammer & Monod 1976; Van Roey 1990). Hence, only a limited number of empirical studies have been conducted on the issues of cognates from the perspective of SLA (Ard & Homburg 1983; Lightbown & Libben 1984; Palmberg 1985; Ringbom 1978, 1987; Holmes 1986; Holmes & Ramos 1993; Nagy, et al. 1993; Granger 1996). Although some different types of cognate relationships have been explored, for example, how meanings overlap between two languages (i.e. good cognates, partially / totally deceptive cognates), the question of the relative ease or difficulty of learning each type of cognate still remains unanswered. Carroll (1992: 108) claims that “to the extent that lexical entries of the cognate-pair match, learning will be facilitated….and to 115 the extent that they differ, learning will be hindered”. Carroll’s (1992) claim recalls the strong or original version of the CAH (Lado 1957), i.e., the greatest difficulty would be predicted where the greatest difference exists between learners’ L1 and L2. In contrast, Carroll’s claim contradicts with the moderate version of the CAH (Oller & Ziahosseiny 1970) which suggests that the learning will be the most difficult where the most subtle distinctions are required. In brief, if the strong version of the CAH predicts the difficulty of cognates correctly, totally deceptive cognates will be the most difficult category for learners. On the other hand, if the moderate version of the CAH is correct, partially deceptive cognates will be the most difficult. Yet, as numerous studies on language transfer or CLI suggest, structural similarities or differences between learners’ L1 and L2 are only one of the many factors which may facilitate transfer or make certain linguistic items difficult to acquire. At the same time, however, structural conditions still play certain roles in the transfer or difficulty inducing factors (Odlin 1989, 2003; Larsen-Freeman & Long 1991; Selinker 1992; Gass & Selinker 1994; Ellis 1994). In addition, the difficulty of the same word may vary depending on the nature of the task required (e.g. Ringbom 1987; Ellis 1994). As discussed in the previous chapter, cognates are generally associated with learners’ positive transfer in comprehension tasks (Ringbom 1987, 1991, 1992; Ard & Homburg 1983; Holmes 1986; Holmes & Ramos 1993). Holmes & Ramos (1993) further indicate that L2 learners rarely misinterpret the meaning of deceptive cognates in reading tasks due to the extensive use of contextual and topical clues in their inferencing. This may be a valid claim, however it is possible that deceptive cognates may cause errors in production, since some of them require subtle semantic distinctions between learners’ 116 L1 and L2. Granger (1996) also asserts that most of the evidence supporting the facilitative effect of cognates is derived from reception data mainly of reading comprehension; hence it is by no means certain that this positive effect is found in the same proportions in production data. Lightbown & Libben (1984) investigated both recognition and production of French- English cognates by French ESL students. They conducted a composition test, a cloze test and a word acceptability judgement test based on the story of a short film. Ten concepts were targeted, all of which were cognates of concrete nouns. The results of the study showed that there is no clear indication of overuse or underuse of cognates in general, but the proficiency level of participants interacted with the appropriateness of specific cognates in specific contexts. On the basis of these results, Lightbown & Libben claim that even when the languages are as closely related as French and English, learners tend to be suspicious of the cognate relationship, and do not always use, or even recognise, cognates as a means of solving problems or compensating their lack of lexical knowledge. For this reason, Lightbown & Libben (1984: 407-408)stressthe importance of input on the part of L2 counterpart for both positive and negative transfer of cognates to occur: .…learners will not accept or attempt to use a particular cognate unless they have actually encountered the specific word in a target language context. Having encountered it, they may tend for a period to overuse the word, overextending its meaning to include related concepts – whether or not these correspond to the usage of the word in their L1.... In order for the process to begin, however, the learner must encounter the cognate. 117 Lightbown & Libben’s (1984) claim is congruent with Andersen’s (1983) “Transfer to somewhere principle” which proposes that learners generalise the transferability of certain structures on the basis of the L2 input they encounter: if they perceive certain L2 features are similar to or work in the same way as L1 corresponding elements, transfer is facilitated. Although Andersen (1983) assumes only transfer of grammatical structures, the same principle may also apply to the transferability of cognates because they are the lexical items in which learners are most likely to perceive cross-linguistic similarities. Apart from this, Lightbown & Libben’s (1984) claim seems to be similar to that in Kellerman’s (1978, 1979) studies on the English-Dutch cognates “BREAK/BREKEN” in which he proposed that learners’ actual L2 knowledge is one of the three constraints of transfer. Whereas both Kellerman and Lightbown & Libben found L2 knowledge or L2 input to be an important factor in transferability, they see the L2 input from different perspectives. Although the former assumes the actual L2 knowledge only as positive evidence whereby learners are unnecessary to be dependent on their L1 knowledge, the latter supposes that once learners are aware of the existence of cognates in the form of L2 input, they begin to use them actively. In other words, Kellerman (1979, 1984) argues that the actual (and correct) L2 knowledge inhibit negative transfer, while Lightbown & Libben (1984) view L2 input as a transfer (of both positive and negative) inducing factor. These different views might be derived from the nature of the tasks or target items each party used. Kellerman (1978, 1979) questioned his participants about the possibility of direct translation from L1 words to their L2 counterparts focusing only on one 118 polysemous verb, while Lightbown & Libben (1984) conducted two types of production tests as well as one comprehension test targeting ten concrete nouns. That is, Kellerman’s claim was based on the transferability (which Ellis (1994) calls “translatability”) of certain usages within a polysemous verb, and thus the participants had known some (at least the most prototypical and frequent) usages could be translated into the English verb “BREAK”. On the other hand, the participants of Lightbown & Libben’s study were not aware that the target items were cognates. Moreover, they did not know what the target items were in the first place. For these reasons, it is apparent that these two claims are in different dimensions, and therefore they are not inconsistent or exclusive. In her study, Granger (1996) attempted to empirically confirm the hypotheses proposed by previous descriptive studies. The first hypothesis was that French learners of English tend to overuse the French cognates when they are faced with Germanic-Romance doublets such as “help / aid” or “get / obtain” (Van Roey 1990). However, the results of Granger’s study did not support this hypothesis. Granger (1996) compared the frequency of Germanic and Romance words used by French learners of English and native speakers of English in two 10,000 word corpora of English academic essays on the same topic: one written by advanced French learners of English and one by native speakers of English. The analysis of this corpus data revealed that the French learners used a significantly higher number of Germanic words, instead of Romance words, than the native speakers of English. Granger (1996) views this result as a universal feature of advanced IL because French learners’ overuse of Germanic words is largely due to a significant overuse of the most frequent Germanic verbs. This is also supported by Hasselgren’s (1994) claim that core words in L2, which tend to be learnt early, widely 119 usable and less susceptible to errors, are hugely overused even among advanced learners. The second hypothesis Granger (1996) attempted to confirm was that incorrect use of deceptive cognates is the least enduring type of interference (Hammer & Monod 1976). Based on the same corpus data outlined above, Granger found that approximately one third (260 out of 750) of the total lexical errors in the French learners’ corpus was caused by the misuse of deceptive cognates. Due to the fact that even advanced learners frequently used deceptive cognates erroneously, Hammer & Monod’s (1976) hypothesis was also questioned. In addition, although some instances of totally deceptive cognates were found, the majority of errors were partially deceptive cognates. Granger (1996) argues that the difficulty with partially deceptive cognates is their seemingly equivalent nature. If French learners are relying on the equivalence of important in the meaning “of great value”, they are liable to infer equivalence in the other meaning “large”. Accordingly, French learners produce such an error as: “Their population is as *important as the rest of Europe” (Granger 1996: 117). Studies on cognates in European languages suggest that difficulty presented by cognates for L2 learners may vary depending upon the degree of semantic overlap across languages and nature of the task type. That is, cognates generally have facilitating effects on comprehension, while interfering effects are not uncommon in production particularly where partially deceptive cognates are involved. As Granger (1996) notes, however, we still cannot largely generalise the previous findings because the studies examined only limited aspects of L2 learners’ IL regarding cognates. In other words, an investigation of other language modes (e.g. spoken) using different types of tasks, 120 different proficiency levels or different target languages (e.g. different patterns of cognancy) may bring about different results. 2.7 Bilingual Mental Lexicon The research on bilingual mental lexicon was originally based on that of L1 mental lexicon (e.g. Aitchison 1994). Whereas many similarities have been found between L1 and L2 situations in terms of the mechanism of mental lexicon, some significant differences have also been revealed. Among them, the largest difference is the existence of CLI (Singleton 1999). Bilingual mental lexicon research aims to elucidate the structure or organisation of the mental lexicon system of bilinguals, and their analyses are largely based on the participants’ reaction time of each word, or pair of words, which appears on a computer screen, using such methods as lexical decision, picture naming or word translation task. However, bilingual mental lexicon research in the field of psychology and vocabulary research in applied linguistics or SLA have previously existed as two largely separate traditions with little interaction. Meara (1997: 109) claims that: ….one of the main reasons why SLA has failed to take advantage of research on bilingual lexicons, and why psychologists remain largely ignorant of work done by applied linguists is that the two bodies of research belong to two very different research traditions: linguists and psychologists tend to think about language in fundamentally different ways. 121 According to Meara (1997), whereas there are a few signs that the barriers between two traditions (i.e. bilingual mental lexicon studies and L2 vocabulary acquisition studies) are beginning to break down, there is still little collaboration among the two traditions. It has almost been a decade since Meara’s claim, and recently, there seems to be a growing interest in L2 vocabulary acquisition within laboratory based psycholinguists (e.g. Dijkstra, et al. 1998; Talamas, et al. 1999; Kroll & Tokowicz 2001; Kroll, et al. 2002; Jiang 2000, 2002, 2004a; Hall 2002). Kroll & Tokowicz (2001) reviewed laboratory based bilingual mental lexicon studies in relation to adult L2 learners’ development of conceptual representations, and concluded that the achievement of L2 proficiency requires not only adequate representation of L2 words but also the acquisition of control processes that allow the relative activation of the two languages (i.e. L1 and L2) to be modulated. That is, both L1 words and L2 words are always available for bilinguals to some extent, and competition exists between the two languages. Kroll, et al.’s (2002: 137) online study explored questions such as “How do adult second language (L2) learners acquire lexical representations for L2 words and then connect them to existing representations within the cognitive network for words in the first language (L1) and their meanings?” They considered two models that had been proposed regarding the development of bilingual lexicon (e.g. Potter, et al. 1984). The “word association model” assumes that L2 words are mediated via direct connection to their translation equivalents in L1, while the “concept mediation model” predicts that L2 words are connected directly to their meanings without L1 mediation. On the basis of previous studies, Kroll, et al. (2002) argue that adult L2 learners’ lexical representations are susceptible to developmental change. That is, in the early stages of 122 L2 acquisition, learners are likely to perform in accordance with the “word association model” (e.g. Chen & Leung 1989), while they are more likely to perform in accordance with the “concept mediation model” as their L2 proficiency increases (e.g. Talamas, et al. 1999). For instance, Talamas, et al. (1999) conducted a translation recognition task in which participants were asked to decide whether the second of two words was the correct translation of the first using English-Spanish word pairs. These pairs were related either in form (e.g. man-hambre (hunger) which is orthographically similar to hombre (man)) or meaning (e.g. man-mujer (woman) which is semantically similar to hombre (man)). As a result, Talamas, et al. found that less fluent bilinguals were prone to form-related interference, while more fluent bilinguals were more susceptible to semantic interference. Their results support the offline studies conducted by Hulstijn & Tangelder (1991) which also found that acquisition of semantic aspects of L2 words takes much longer than that of formal aspects. As we have seen in the earlier section, recent laboratory based psycholinguistic studies have made significant contributions to our understanding of L2 learners’ lexico- semantic development (e.g. Jiang 2002, 2004a). In addition to these online studies which focus on the same-translation pairs, one study attempts to reveal L2 learners’ lexical development specifically targeting cognates. Hall (2002) explored the early stages of vocabulary development of L2 learners by claiming that such learners automatically exploit existing lexical material in order to establish an initial memory representation. Hall (2002) views early L2 vocabulary development as a matter of pattern-matching and assimilation with current lexical knowledge, and proposes a 123 “Parasitic hypothesis”. This hypothesis posits that early L2 vocabulary acquisition is concerned with connecting novel L2 words to the frame representation of a translation equivalent in L1. In this way, learners can maximise the use of an already established language structure (i.e. L1 lexical entries) by exploiting the L1 lexicon in parasitic fashion. Whereas Hall (2002) notes that there is much more to learn new words than registering their forms and connecting them with L1 representations, he strongly argues that all learners must establish some memory representation of a form to which the other characteristics of wordship may be anchored. Thus, he assumes that the language faculty will automatically try to detect and exploit L1 word forms particularly when cognates are involved, because they can easily afford storage and access cues for L2 forms. Hall also maintains that these strategies are consonant with general learning principles such as the “Least effort principle” (e.g. Slobin 1977) or the “Principle of economy” (e.g. Haiman 1983). In order to confirm the Parasitic hypothesis, Hall (2002) conducted a word familiarity report task on 95 Spanish speaking learners of English. In this task, participants were instructed to record whether they had seen a total of 90 target items before, and provide a translation equivalent in Spanish. Thirty non-existent words were created as experimental stimuli and randomly shown on a video monitor with 60 real English words, half of which were English-Spanish cognates. For the purpose of detecting the roles of form overlap with participants’ L1, half of the stimuli were designed as pseudo- cognates: non-existent words in English, but which have similar form to Spanish words, as in *companary which shares the similar form with the Spanish word companario 124 (bell tower). In addition, Hall (2002) controlled such variables as frequency of real words (non-existent words have zero frequency), word lengths, morphological complexity and cognate status (pairs must share at least two thirds of their form in regard to the number of letters) for all of the 90 target items. The results showed that participants found pseudo-cognates to be more familiar than (non-existent) non-cognate words, and a much narrower range of translations was provided for the former than the latter. Hall (2002) believes that these results support the Parasitic hypothesis in that participants used L1 words to confirm or create hypotheses about word meaning when they encountered novel words, despite the fact that the data was based on artificial words out of their contexts of use. Moreover, Hall claims that participants are not only sensitive to form overlap of cognates, but also assume their semantic overlap (i.e. shared form indicates shared meaning) because of the higher rates of translation conformity for pseudo-cognates. Although Hall’s (2002) attempt to explain the early L2 lexical development in the form of Parasitic hypothesis gave us an insight into the learners’ automatic exploitation of cognates, limitations and questions must be raised with regards to this study. As Hall (2002) himself notes, he limits the scope of the study to L2 learners’ initial exposure to cognates, and hence the Parasitic hypothesis cannot explain the later process of acquisition: when and how the learners’ first hypothesis (i.e. all cognates share the same meaning) will be restructured. Several questions can be raised here. Firstly, if we assume that all learners have such a presupposition (i.e. shared form indicates shared meaning), do deceptive cognates 125 always trigger misrecognition or misinterpretation in the first encounter? Hall’s (2002) results apparently conflict with Holmes & Ramos’s (1993) claim that deceptive cognates hardly induce misinterpretation in reading comprehension because of the contextual and topical cues. However, these contradictory results can be easily ascribed to the different methodologies that each study adopted. Whereas the participants in Hall (2002) were shown pseudo-cognates or artificial words individually independent of contexts, those of Holmes & Ramos (1993) read an authentic article in more naturalistic conditions. Secondly, the Parasitic hypothesis also contradicts Lightbown & Libben’s (1984) claim that learners are initially suspicious about the transferability of cognates, and that they may begin to use the cognates after they have encountered the cognates in the form of L2 input. Again, the discrepancy in the results of these two different studies may be due to their varied experimental conditions. Although Lightbown & Libben (1984) conducted two types of production (composition and cloze) tests as well as a comprehension (acceptability judgement) test, Hall’s (2002) results are based on a familiarity report task in which participants were requested to state whether or not they have seen the target items (including non-existent words) before. The language skills or knowledge required in a familiarity judgement of pseudo-cognates are not only different from those required in production tasks of unknown cognates but also differ significantly from those necessary for comprehension tasks in contexts, where many factors may constrain learners’ transfer. Hall (2002) argues that there is no reason for participants to believe that these non- existent words are anything other than existing words in English on the grounds that 126 they had limited knowledge of English, and that they were required only to judge familiarity with the target items. However, participants’ familiarity with the words cannot be directly equated with their “acquisition” of the words because they were simply asked whether they had seen the words before, and to give a possible L1 translation equivalent. On the other hand, Lightbown & Libben’s (1984) tasks were more realistic since they used real words in context. For these reasons, Hall’s (2002) results are difficult to generalise as “early stages of vocabulary development in second language learners”(Hall 2002: 69), or even as the acquisition of cognates in the first encounter. In sum, Hall’s (2002) Parasitic hypothesis seems to be plausible only in a strictly controlled experimental environment, and acquisition of cognates in reality is largely influenced by the variables occurring in actual language use and learning situations. Although it is valuable, as a pure psycholinguistic study, to understand the learners’ mental process in the situation where variables affecting their learning are highly controlled, acquisition study must attempt to predict the whole process of learners’ IL system by considering many possible variables occurring in their real learning environment. The discrepancy of the results between laboratory based experiments and the studies with more naturalistic data may suggest that the variables affecting the acquisition of novel cognates, such as context, have a huge impact on the learners’ initial hypothesis: all cognates share the same meanings. Thus, it is critical to further investigate the picture of interaction between such variables and learners’ such initial presuppositions in their encounter to the novel cognates within naturalistic conditions. It also will be worthwhile to examine under what condition and how learners restructure their initial hypothesis. 127 The increased interest in laboratory based psycholinguists in L2 acquisition of European languages has led to some bilingual mental lexicon studies focusing on Japanese and Chinese (e.g. Kayamoto, 1996, 2000; Tamaoka & Matsushita, 1999; Tamaoka, et al. 2003; Chiu, 2003). Chiu (2003) compared the CBLs’ reaction time for three different types of Japanese kanji compounds, which were read aloud and were known to the participants in the study: cognates, non-cognates with on-yomi (Chinese origin pronunciation), and non-cognates with kun-yomi (Japanese native pronunciation). The results showed that the reaction time of cognates was significantly longer than that of non-cognates with on-yomi, but shorter than that of non-cognates with kun-yomi. On the basis of these results, Chiu believes that CBLs tend to access the meaning of the cognates in their mental lexicon via phonological information of the Chinese counterpart, while they access the meaning of non-cognates via its Japanese pronunciation. Thus, she explains that the longer reaction time of cognates than non- cognates with on-yomi is due to a weaker associative link between the form and Japanese pronunciation in the former. To be more precise, processing of Japanese-Chinese cognates for CBLs involves the activation of the Chinese sound, whose associative link to the form is much stronger, so the learners’ access to the Japanese sound is susceptible to interference by the Chinese sound. On the other hand, Chiu (2003) assumes that the processing of non-cognates for CBLs does not involve activation of the Chinese sound, hence there is nothing to interfere with the Japanese pronunciation. Chiu speculates that the reason for the longer reaction time of non-cognates with kun-yomi than for the others is due to their relative difficulty of learning, which is unrelated to the Chinese sound. Finally, she concludes 128 that the interlexical relationship between Japanese and Chinese impedes the CBLs’ phonological processing of cognates in Japanese, although the speed of accessing their Japanese sounds will improve as their proficiency increases. Tamaoka, et al. (2003) conducted a series of online judgement tasks on advanced CBLs using kanji compounds which (1) share the same orthography and the same meaning between Japanese and Chinese (i.e. good cognates), (2) exist only in Japanese, (3) exist only in Chinese, and (4) exist neither in Japanese nor in Chinese (i.e. non-existent words). They found that advanced CBLs’ reaction time was shorter where the kanji compounds exist in both languages, while it was much longer where they exist in Chinese but not in Japanese. According to Tamaoka, et al. (2003), even in judging whether or not the target words exist in Japanese, semantic representations are activated by both Japanese and Chinese where the kanji compounds exist in both languages (e.g. 青春(youth)), so CBLs can judge them correctly and quickly. However, advanced CBLs’ judgements are relatively slower where the kanji compounds exist only in Japanese (e.g. 食事(meal)) because there are no facilitation effects from Chinese. On the other hand, CBLs need to suppress the activation from Chinese in judging the kanji compounds that exist only in Chinese in order to deny their existence as Japanese words. As a result of inhibitory effects from Chinese, their judgements tend to be slower and susceptible to misjudgement. This is in contrast to the results for those kanji compounds that exist neither in Japanese nor Chinese in which the learners rejected the target words correctly and quickly. 129 The difficulty of denying the kanji compounds that exist only in Chinese as Japanese words was apparent due to the fact that nearly half of the words were judged erroneously by advanced CBLs. Accordingly, Tamaoka, et al. (2003) conclude that conceptual interdependence of kanji or kanji compounds between Japanese and Chinese is very strong, and this supports the “concept mediation model” (e.g. Kroll & Tokowicz 2001; Kroll, et al. 2002) which is considered to reflect later stages of L2 acquisition. In sum, existing bilingual mental lexicon studies on CBLs of Japanese generally suggest that conceptual or semantic representations of cognates are shared even for advanced learners who are likely to access L2 words directly to their meanings without mediation of L1 words. In the following section, I will review previous descriptive studies and a few empirical studies that have attempted to reveal the difficulty of Japanese kanji compounds or Japanese-Chinese cognates for CBLs. 2.8 Japanese Kanji Compounds According to Miyajima (1993), 21,656 out of 40,393 words (53.6%) in Reikai Japanese Dictionary (1956) are kanji compounds, although not all of them have cognate relationship with Chinse words. A number of descriptive studies have discussed the issues of Japanese kanji compounds or Japanese-Chinese cognates with respect to CBLs’ learning of Japanese (Bunkachou 1978; Takebe 1979; Arakawa 1979; Moriya 1979; Miura 1984; Hida & Lu 1986; Chou 1986; Ookouchi 1997; Ueno & Lu 1995; Lin 2002; see Chen 2003a for review), many of which are based on the classification of cognates similar to that in European languages. Among them, one of the most significant works is Bunkachou’s (1978) classification of Japanese kanji compounds. Bunkachou (1978) divided Japanese kanji compounds into four categories of S (same), 130 D (different), O (overlap) and N (none) according to its orthographic and semantic similarity to Chinese. Within this framework, (S) indicates the kanji compounds which share the same orthography with Chinese counterparts and have the same meanings. For instance, the Japanese word “学校” (gak-kou) shares the same orthography with the Chinese word “39”(xue-xiao) and has the same meaning (i.e. school). Only the pronunciations are different. (D) indicates kanji compounds which share the same orthography with Chinese counterparts but have completely different meanings. For instance, the Japanese word “汽車”(ki-sha) shares the same orthography with the Chinese word “ ”(qi-che) but they have different meanings. The former means “a (steam) train” and the latter, “a car” or “an automobile”. (O) indicates kanji compounds which share the same orthography with Chinese counterparts with partly the same or partly different meanings according to the context. For instance, both Japanese word “所有”(sho-yuu) and Chinese word “ ”(suo-you) mean “possession” or “to own” but only the Chinese one can also mean “all” in a different context. (N) indicates kanji compounds whose forms do not exist as a Chinese word. For instance, the Japanese word “宿題”(shuku-dai = homework) does not exist and makes no sense as Chinese; “ ”(zuo-ye) is used for “homework” in Chinese. 39 Some Japanese kanji characters (e.g. “学”) and their Chinese counterparts (e.g. “”) do not always share exactly the same script due to the simplification of characters. However, Arakawa (1979) asserts that the differences in character form (such as simplified or traditional) should be disregarded in contrastive studies on Japanese-Chinese cognates as long as they have the same origin character. In the present study, thus such pairs are also regarded as identical characters. This seems reasonable since cognate pairs do not always share exactly the same script in European languages as well. 131 Kanji compounds (S), (D) and (O) are considered as Japanese-Chinese cognates, but (N) is not. (S) can be referred to as “good cognates”, (D) as “totally deceptive cognates”, and (O) as “partially deceptive cognates” in Van Roey’s (1990) and Granger’s (1996) categorisation. According to Bunkachou (1978), among 2,000 kanji compounds which they collected from beginning and intermediate Japanese textbooks, (S) takes up approximately two thirds, (N) does one fourth, and (D) and (O) do only one tenth of the total number of kanji compounds. This indicates that the good cognates by far outnumber the deceptive cognates. However, given that the ratio of good cognates to deceptive cognates between French and English is 11:1 (Hammer & Monod 1976), there exist more deceptive cognates between Japanese and Chinese (7:1) than between English and French. Kanji compounds (O) or partially deceptive cognates can be classified into three subcategories according to the way in which their meanings are overlapped (Ueno & Lu, 1995; Lin 2002). They are: (1) there are extra meanings in Japanese or the semantic space of (O) is wider in Japanese (J>C), (2) there are extra meanings in Chinese or the semantic space of (O) is wider in Chinese (J (J> 132 Inclusion (J>C) Inclusion (J 所有 (Chinese) (Japanese) 単位 (Japanese) (Chinese) 得意 (Japanese) (Chinese) Figure 2.8: Partially Deceptive Cognates Between Japanese and Chinese Table 2.1: Degree of Cross-linguistic Semantic Correspondence for Each Type of Japanese Kanji Compounds Type of kanji compound L2 L1 L2: Japanese L1: Chinese (S) = Same x ————x 学校 (school) (school) (N) = None x ————ø 宿題 (homework) e.g. (homework) (D) = Different x ————y 汽車(steam train) (car) x 得意 (proudly / skilful) (proudly) (J>C) x e.g. (skilful) y x 所有(possession) (possession / all) (O) = Overlap (J Recently, there have been a few empirical studies which attempt to investigate the difficulty of learning Japanese-Chinese cognates either for JNSs or CBLs among four categories of (S), (D), (O) and (N) and with their subcategories (Miura 1997; Chen 2003b). Miura (1997) conducted a multiple-choice translation test on kanji compounds (S), (D) and (O) (i.e. Japanese-Chinese cognates) on JNSs who were learning Chinese at a university in Japan. In the test, participants were asked to choose the most appropriate translation of each kanji compound in short sentences both from Japanese to Chinese and Chinese to Japanese. The results indicated that the kanji compounds (S) or “good cognates” marked the highest percentage of correct answers, and kanji compounds (D) 133 or “totally deceptive cognates” were likely to be answered correctly if the words had been taught in their course. Kanji compounds (O) or “partially deceptive cognates” were divided into two subcategories of (J>C) and (J (1997) also suggested that the difficulty of (O) might vary depending on the way in which its meanings overlap across languages, (i.e. (J>C) or (J Since Miura’s (1997) participants are JNSs, Chen (2003b) is the first empirical study which examined the difficulty of four types of Japanese kanji compounds for CBLs. Chen compared the difficulty of each of the four categories of Japanese kanji compounds (S), (D), (O), and (N) by conducting a multiple-choice test for CBLs of three different proficiency levels at a university in Taiwan. The results showed that the 134 percentage of correct answer was highest in (S), and this was followed by (N), (O) and (D). Seemingly, this result conflicts with Miura’s (1997) claim that (D) is not difficult once it has been taught. Apart from the fact that the participants’ L1 and L2 are reversed in these two studies, the contradictory results might be largely due to the differences in the selection of target items. The Japanese-Chinese cognates selected as target items in Miura’s test could be said as beginning or intermediate level, and at least a part of these items were formally taught in the participants’ classes. In contrast, all the target items in Chen’s study were selected from a vocabulary list of Japanese proficiency test level 2 or above, i.e., vocabulary for advanced learners, and therefore would rarely be taught in beginning or intermediate classes, particularly in a non-Japanese speaking environment. In addition, Chen excluded two items from analysis whose rate of correct answer was extremely high because they were regarded as items that had been taught. For this reason, it is not surprising that Chen reports (D), or totally deceptive cognate, as the most difficult, while Miura (1997) states that (D) is not difficult as long as it has been taught. Chen (2003b) attempted to control the participants’ prior knowledge of each target item, and used only advanced Japanese kanji compounds assuming that no participants had this prior knowledge. This implies that Chen’s study is not necessarily examining the difficulty of learning Japanese kanji compounds, but the difficulty of guessing the meaning of novel kanji compounds, although she does not clearly state so. Besides, Chen (2003b) did not actually compare the difficulty of (O) with other types of compounds. As the meaning of each kanji compound in the multiple-choice test was 135 asked individually without any contexts, it was not clear which meaning of (O) was being questioned. Under such circumstances, success or failure of guessing the meaning of unknown partially deceptive cognates without context seems to be simply a matter of whether the meaning in question corresponds to that of the L1 counterpart. This is because learners are likely to treat deceptive cognates as if they were good cognates when first exposed to them unless other clues are available as in the Parasitic hypothesis (Hall 2002). Chen’s (2003b) study has similarities with Hall’s (2002), namely, that both attempted to control the participants’ prior knowledge of the target items, and asked participants the meaning of novel cognates without any context. As a result, these two studies had similar results: learners tended to suppose that all cognates share the same meaning as their L1 counterparts in spite of the semantic discrepancy or even non-existence of the cognate pair. According to Chen, (N) was not as difficult as (D) and (O) for CBLs. She explains the reason for this is due to the nature of kanji or logograph. Even when the kanji compound itself does not exist and make sense in their L1, the elements of each compound or individual characters are often understandable for CBLs. Thus, they could guess the meaning of many of the Japanese kanji compounds using their L1 knowledge. However, as Chen (2003b) herself notes, not all non-cognate Japanese kanji compounds are comprehensible to CBLs. For this reason, the difficulty of (N) for CBLs may vary according to the comprehensibility of its meaning using their L1 knowledge. 136 One of the shortcomings of Chen’s study was that she chose only advanced kanji compounds and assumed that all target items were unknown to participants. This naturally produced no significant difference among participants of different proficiency levels in their choice of the answers and difficulty (error rate) of each type of kanji compounds for them. In other words, participants of all proficiency levels simply fell into a trap, and erroneously guessed the meaning of deceptive cognates (i.e. (D) and (O)) due to lack of available clues other than their L1 knowledge. Thus, in order to compare the difficulty of learning the meaning of each type of the kanji compounds (i.e. lexico-semantic development), instead of the difficulty of guessing their meaning without any context, it is necessary to include kanji compounds of varied level or frequencies as target items for each category. Whereas Miura (1997) and Chen (2003b) are significant in that they pioneered the empirical studies of Japanese-Chinese cognates or Japanese kanji compounds on the basis of learner data, there are quite a few methodological problems with these studies. The most significant problem is that, although the main concern was L1 influence on the participants’ choice of words, the studies did not compare the test results with other controlled groups, i.e., learners of other L1 backgrounds. Hence, the proposed difficulty of deceptive cognates could also be caused by intralingual or developmental factors which commonly affect learners of other language backgrounds. Like Granger (1996), which, based on corpus data of learners’ composition, maintains that partially deceptive cognates are more difficult than the other types of cognates, there have been some descriptive studies which claim that kanji compounds (O) are the hardest for CBLs to learn due to their semantic ambiguity (Bunkachou 1978; Lin 2002; 137 Matsushita 2001). Among these, Matsushita (2001) extensively explored cross- linguistic semantic congruence and incongruity of kanji compounds (O). He postulates that the problems that CBLs face may differ depending upon the task type (receptive or productive) as well as the cross-linguistic semantic condition. For instance, where (O) has wider semantic ranges in Japanese such as “注意(chuu-i)” (see Figure 2.9), CBLs may not have problems in producing40 (O) in the sense shared between Japanese and Chinese (i.e. to pay attention). However, it may be problematic in receptive activity to understand the Japanese word when the cognate is not used in the same meaning as Chinese (i.e. to warn). Likewise, where (O) has smaller semantic ranges in Japanese, such as “招待(shou-tai)” (see Figure 2.10), CBLs may not have problems in listening or reading (O) as a Japanese word (i.e. to invite). However, Matsushita (2001) predicts that CBLs may make errors in their attempt to produce the semantic range of the non-cognate L2 word where the cognate is used in the L1 (i.e. to serve) due to their semantic over-extension from the L1 cognate counterpart. Matsushita’s (2001) hypotheses on Japanese-Chinese partially deceptive cognates seem to be congruent with the previous findings particularly of laboratory based psycholinguistic studies on L2 lexico-semantic development (e.g. Jiang 2002, 2004a) or acquisition of cognates in the initial encounter (e.g. Hall 2002), given that they primarily consider the L1 influence on learners’ acquisition of L2 words by taking no 40 Matsushita (2001) did not explicitly make a distinction between speaking and writing in language production, and listening and reading in reception, respectively. Yet, the problems for CBLs might vary among them. Unlike the cognates between the languages using Roman alphabet or phonogram, CBLs must know how to pronounce the cognates in Japanese in order to use them in speaking, but this is not the case in writing. Similarly, CBLs need to know the pronunciation of the cognates to understand their meaning in listening task, while the phonetic information is not prerequisite in reading task. However, these principles may not always hold true when learners’ guessing or communication strategies are involved. 138 account of other possible factors. Thus, it is worthwhile to examine Matsushita’s (2001) hypotheses to see whether or not factors other than cross-linguistic semantic conditions are at work to facilitate or inhibit CBLs’ transfer using more naturalistic tasks. This could possibly elucidate the more realistic picture of CBLs’ use and acquisition of Japanese-Chinese partially deceptive cognates, since they may be affected not only by cross-linguistic semantic conditions but also by other factors such as task type, L2 proficiency, frequency of L2 input or the context in which they are embedded. :Production based on L1 knowledge :Reception based on L1 Knowledge L2: Japanese L1: Chinese zhu-yi chuu-i Positive transfer 注意 (to pay attention) (to pay attention / to warn) jing-gao Negative transfer (to warn) Figure 2.9: Differing Patterns of CLI on (J>C) According to Task Type (Based on Matsushita 2001) L2: Japanese L1: Chinese shou-tai PositivePositive transfer?transfer? 招待 zhao-dai (to invite) motenasu (to invite / to serve) もてなす Negative transfer? (to serve) Figure 2.10: Differing Patterns of CLI on (J 139 2.9 Summary of Findings and Problems in the Existing Literature In this chapter we have considered the roles of L1 in learners’ L2 lexico-semantic development from various perspectives. As we have seen, an extensive body of research in different areas commonly suggests that L2 learners tend to map L2 word forms onto pre-existing L1 concepts in their L2 lexico-semantic development. Even though researchers’ opinions vary as to the extent to which learners restructure the semantic content of the L2 words, the more recent and prevalent views seem to suggest that the creation of new lexical entries or new concepts for L2 is less likely to occur. When novel L2 words share similar word forms with their L1 counterparts as in cognates, learners are liable to assume that they share the same meanings as well. Even when the L2 words and their L1 translation equivalents do not share similar word forms, learners are still likely to use and learn L2 words on the basis of the semantic contents of their L1 counterparts. Such over-extension of the L1 meanings into L2 translation equivalents may result in production error or odd utterance where one word in L1 is represented by multiple words in L2 (i.e. divergence). In addition, it has also been claimed that learners may rely on interlingual one-to-one mapping by connecting one of the usages of the L2 word (i.e. specific exemplar) and an L1 item, and ignore other usages or meanings where one word in L2 is equivalent to several words in L1 (i.e. convergence). There appears to be sufficient evidence to claim that interlingual semantic transfer is both pervasive and persistent, but some of the evidence and claims are based on laboratory experiments (e.g. Hall 2002; Jiang 2002, 2004a) in which target words are examined in unnatural online tasks and independently of language contexts. Indeed, 140 some studies based on more naturalistic situations (e.g. Holmes & Ramos 1993; Lightbown & Libben 1984) showed contradictory results with such laboratory experiments regarding learners’ use of novel cognates. The discrepancy of the results between the strictly controlled laboratory data and more naturalistic data may indicate that learners’ semantic transfer from L1 words into the L2 equivalents is constrained by the variables occurring in more naturalistic and authentic situations where learners actually use and acquire L2 words. As I noted earlier, those laboratory based psycholinguistic experiments, many of which measure participants’ reaction time for each stimuli or target item, are valuable and useful in that they may be able to reveal learners’ invisible mental process or organisation of mental lexicon by controlling many variables that may affect their response. Besides, there has also been some evidence that such reaction time data predicts learners’ lexical development effectively: advanced learners responded faster and more accurately than intermediate learners (Harrington 2005). Nevertheless, as is reviewed in Chapter 1, L1 transfer and L2 acquisition involve multiple factors, and these factors interact with each other. Therefore, it is essential to further examine L2 learners’ lexico-semantic development by considering the interaction between transfer factors and the variables (e.g. task type, proficiency, L2 input or context) affecting their use and acquisition of L2 words or cognates in particular. In a similar vein, previous studies on polysemy have indicated that acquisition of such words in L2 involves not only interlingual semantic transfer, but also its prototypicality both in learners’ L1 and L2. It has been relatively clear that prototypical items in L2 are generally easy to be learnt irrespective of potential transferability (i.e. positive or 141 negative transfer items), while less prototypical items in L2 are more difficult, particularly when negative transfer is involved. However, much less is known with regard to the interaction between the prototypicality of learners’ L1 items and their transferability. That is, Kellerman’s (1978, 1979) findings on the transferability of Dutch “BREKEN” are only based on learners’ judgement as to whether or not each of its usages can be translated into English counterparts. Therefore, the results of his studies cannot be directly identified with learners’ transferability of polysemy in their performance or actual use of the L2, i.e., reception or production of the L2 words in context. With the results of the previous relevant studies in mind, the current research attempts to verify the existing findings and to give new insights into the acquisition of cognates and polysemy by using three types of tests which are different from those employed in the studies reviewed in Chapter 1 and Chapter 2. It is apparent through the review of the related literature that some of the critical factors regarding learners’ use and acquisition of cognates or polysemy have not been adequately explored (e.g. comparison of the results between the target group and the learners of different L1). In order to fill this void, three independent but closely related studies were undertaken. In Part II, they are presented separately with each of the research questions, methodologies, results and discussions. 142 PART II: CHINESE BACKGROUND LEARNERS’ (CBLs’) USE AND ACQUISITION OF JAPANESE VOCABULARY Part I has reviewed the existing related research. In Chapter 1, we have seen that numerous factors promote or inhibit the occurrence of language transfer, and they tend to interact with each other. We have also noted that L2 learners make use of their L1 knowledge in the process of producing and understanding messages in the L2, and this would (although not always) be incorporated into their IL system. Accordingly, the present study considers learners’ transfer in their “use” of a particular L2 word and its “acquisition” as a continuum, even though it acknowledges that they are two separate processes theoretically. In Chapter 2, we obtained some insight into the L2 lexico-semantic development in conjunction with interlingual semantic mapping of cognates and polysemy. The literature review has identified several problems which are yet to be solved in the studies of lexico- semantic transfer and development. For example, little is known regarding the interplay between transfer factors and the variables affecting learners’ use and acquisition (i.e. creation and restructuring of their IL) of L2 words. The following chapters will address some issues in the acquisition of Japanese vocabulary, focusing on cognates or polysemy using Chinese background learners (CBLs) as a target population. In Chapter 3 (Study A), I will investigate the roles of transfer in CBLs’ use and acquisition of four types of Japanese kanji compounds. It also attempts to consider their possible differences in the process and relative difficulty of learning, using a lexicality 143 judgement test. Chapter 4 (Study B) further explores CBLs’ use and acquisition of Japanese kanji compounds focusing on partially deceptive cognates whose acquisition was found to be more difficult than others. An oral production test and a translation test will be conducted in order to analyse the possible differences in the process and difficulty of learning according to cross-linguistic semantic condition and task type. Chapter 5 (Study C) will examine CBLs’ use and acquisition of Japanese words that correspond to each usage of two Chinese polysemous verbs “ (kai)” and “ (kan)” by considering their transferability. A prototypicality judgement survey, a lexicality judgement test, and an oral production test will be conducted to reveal the transferability of the two polysemous verbs in CBLs’ receptive and productive use of the L2. Finally, Chapter 6 will synthesise the results of all the studies, and discuss the overall and more general findings of the thesis. It will also include discussion of the limitations of the study and implications for future research as well as pedagogical implications. 144 CHAPTER THREE THE ROLES OF TRANSFER IN CBLs’ USE AND ACQUISITION OF FOUR TYPES OF JAPANESE KANJI COMPOUNDS (STUDY A)41 3.1 Introduction The literature review on receptive transfer offered earlier has revealed that L2 comprehension is greatly facilitated when the text includes many cognates. It has also been indicated that there are different types of cognate relationship according to the level of cross-linguistic semantic overlap: “good cognates”, “totally deceptive cognates”, and “partially deceptive cognates”. Whereas there seems to be no dispute with regard to the relative ease of acquisition involving good cognates, varied results have been shown regarding the difficulty of acquisition of deceptive cognates. For instance, research on reading comprehension of cognates found that even beginning learners rarely misinterpret the meaning of (English-Portuguese) deceptive cognates due to their active use of contextual cues (Holmes & Ramos 1993). On the other hand, analysis of advanced learners’ academic essays revealed that one third of their total lexical errors were caused by the misuse of (English-French) deceptive cognates, the majority of which were partially deceptive cognates (Granger 1996). Studies on cognates in European languages suggest that the difficulty of learning of cognates differs depending upon the task type and the type of semantic overlap arcoss 41 A significant portion of this study has been published in Japanese as “中国語母語話者による日本語の 漢語習得 —他言語話者との習得過程の違い— (Acquisition of Japanese Kanji Compounds by Chinese Native Speakers: Differences in the Acquisition Process from Speakers of Other Languages)” in NIHONGO KYOIKU (Journal of Japanese Language Teaching) No.125, Apr. 2005. 145 languages. That is, generally learners have less problems in their receptive use of deceptive cognates, but they are more likely to produce errors particularly when partially deceptive cognates are involved. However, we cannot largely generalise the previous findings because they investigated only limited aspects of learners’ use of cognates. For instance, they targeted learners of limited proficiency levels, hence learners’ developmental patterns (i.e. creation and restructuring of the IL with respect to cognates) are by no means clear. It is possible that investigation into the learners of different proficiency levels and of different language backgrounds (e.g. non-European languages) using different tasks may give rise to different results. Some descriptive studies on Japanese-Chinese cognates or Japanese kanji compounds used similar classification of cognates to that of European languages. Bunkachou (1978) categorised Japanese kanji compounds into four types according to their different degree of orthographic and semantic correspondence between Japanese and Chinese: (S) or “good cognates”, (D) or “totally deceptive cognates”, (O) or “partially deceptive cognates” and (N) or “non-cognates”. An empirical study has attempted to investigate the relative ease or difficulty of each type of Japanese kanji compounds for CBLs, and found that (D) or totally deceptive cognates were the most difficult category (Chen 2003b). However, its methodological problems made it difficult to generalise the results. The present study investigates the possible differences in CBLs’ use and acquisition process of four types of Japanese kanji compounds by taking account of learners of three different proficiency levels. It also examines the relative ease or difficulty of acquisition of 146 each type of kanji compounds in relation to transfer factors. In order to investigate these, the results of an acceptability judgement test will be compared between CBLs and EBLs across different proficiency levels. 3.2 Research Question and Hypotheses The research question for Study A and hypotheses are as follow: Research question: Are there any differences in CBLs’ use and developmental patterns of four types of Japanese kanji compounds (S), (N), (D) and (O), and their relative difficulty in comparison to EBLs ? Hypothesis 1: For good cognate (S), CBLs will outperform EBLs due to their positive transfer. Hypothesis 2a: For non-cognate (N), CBLs will not outperform EBLs when CBLs’s L1 knowledge is not useful. Hypothesis 2b: CBLs will outperform EBLs when non- cognate (N) is comprehensible for CBLs using their L1 knowledge. Hypothesis 3: For totally deceptive cognate (D), CBLs will underperform EBLs due to their negative transfer. Hypothesis 4: The acquisition of partially deceptive cognate (O) for CBLs will be prolonged due to its cross-linguistic semantic overlap. 147 3.3 Methodology 3.3.1 Participants The participants in this study consisted of 57 CBLs (16 beginning (CBGs), 25 intermediate (CIMs) and 16 advanced (CADs)), 45 English background learners (EBLs) (15 beginning (EBGs), 18 intermediate (EIMs) and 12 advanced (EADs)) and 21 Japanese native speakers (JNSs). Whereas the target participants of the present study were CBLs, EBLs participated in this study to allow me to compare whether or not the test results are due to CBLs’ L1 influence, given that EBLs have different L1-L2 pair associations. Likewise, JNSs took part in the study in order to examine how knowledge of the learners in each proficiency level of each kanji compound approximates to or differs from that of native speakers. The learners’ proficiency levels were determined according to the results of a widely recognised Japanese placement test called SPOT (Simple Performance- Oriented Test)42. Those participants who scored 50 points or above out of 65 points in the version A of the SPOT were regarded as advanced learners; those with 35 to 49 points as intermediate; those with 20 to 34 points as beginning, and below 20 points were not included in this study. The test results of those who scored below 20 points in the SPOT were eliminated from the analysis because they are likely to have serious difficulty not only with the target items but also with the sentences in which they are embedded. This decision was made as a result of small-scale pilot test prior to the main study. The levels roughly correspond to divisions 42 SPOT is a cloze type test in which test takers fill in a blank in each of the 65 (Version A) or 60 (Version B) short sentences with a hiragana letter while listening to an audio tape for the same sentences on their answer sheet. It can measure learners’ proficiency levels of Japanese easily and briefly. Whereas the SPOT uses an audio tape, it does not necessarily measure learners’ listening comprehension, but integrated proficiency of Japanese. This has been proven by the correlations with other Japanese tests. For details of the SPOT, see Ford-Niwa, et al. (1995) or Kobayashi, et al. (1996). 148 used at the University of New South Wales where the SPOT is utilised as a placement test instrument. All participants were students or graduates of Australian Universities at the time of the test, and they participated in this study voluntarily. Participants were recruited immediately after the placement tests for Japanese courses held in the University of New South Wales at the beginning of the semester one and after Japanese classes during the semester in 2004. Additionally, some graduates of Australian Universities through my personal network were asked via email whether they could participate in the study. Participants’ demographic information is shown in Table 3.1. Table3.1: Demographic Information of Participants N Gender Age* Stay in Australia Japanese Stay in Japan (M/F) (yrs)* Study (yrs)* (mts)* CBGs 16 4/12 20.2 4.8 2.4 0.8 CIMs 25 3/22 20.8 6.2 4.0 1.4 CADs 16 4/12 24.7 7.5 6.3 9.7 EBGs 15 9/6 21.5 N/A 3.0 0.2 EIMs 18 4/14 23.4 N/A 5.8 5.8 EADs 12 8/4 25.6 N/A 7.7 13.4 JNSs 21 5/16 28.9 1.4 N/A N/A *mean values 3.3.2 Instrument In order to answer the research question and test the hypotheses formulated in this study, I have constructed an acceptability judgement test (see Appendix A). The acceptability judgement test consisted of 100 short sentences, each of which included an underlined word. A total of 28 target items and 72 kanji words which were aimed for other studies were randomly organised, thus it was very unlikely that participants were aware of the purpose 149 of the study. The task was to judge whether the usage of the underlined word in each sentence was correct or not, and to give a correct answer if it was judged to be erroneous. This allows us to understand the reason for the participants’ judgement on the target items by seeing how they corrected the words. Since this study treats kanji compounds as independent words and does not require participants to know kanji characters, phonetic clues for each kanji character including the target items were given in the form of hiragana syllabics (i.e. rubi). This enables participants to judge the word correctly without knowledge of the kanji characters. In other words, even when EBLs cannot read or recognise the target items in the kanji characters, they are still able to judge correctly if they are familiar with the meaning and the pronunciation of the word in hiragana syllabics. In the test, English cue was given to each short sentence, and the corresponding words to the target items were also underlined in order to clarify the meaning of the words in question. In an acceptability judgement test, particularly when the target item is used erroneously in the sentence, the meaning of the word that the test writer intended is often unclear. For instance, if we ask CBLs to judge whether the deceptive cognate “汽車(ki- sha)” (i.e. “train” in Japanese and “car” in Chinese) is used correctly or not in the following sentence without English cue, it is unknown whether the participants have understood the target item correctly (i.e. “train”) or not (i.e. “car”) particularly when they judged that it was correct, as shown below. “私は毎日汽車で大学に来ます。 (I come to university by train everyday) 150 This problem is avoided when we give an English cue to the sentence such as “I come to university by car everyday”. Those CBLs who did not have prior knowledge about the meaning of Japanese word “汽車(ki-sha)” may judge it as “correct” by guessing its meaning on the basis of their L1 knowledge, while those CBLs who had existing knowledge of the meaning of Japanese word “汽車(ki-sha)” would judge it as “wrong”, and correct it to “車 (kuruma)” or “car” in Japanese as in “私は毎日車で大学に来ます。(I come to university by car everyday) This acceptability judgement test is basically a Japanese-English matching task of the underlined words, thus EBLs are considered to perform as well as CBLs if they can understand the meaning of the target word in Japanese. Given the nature of the judgement test, even when EBLs can guess the meaning of the target items through contexts with English clue, it does not help their judgements. Two possible shortcomings to this method were overcome as follows. First, problem of giving an English cue for each sentence is that participants may be intolerant of the English cue for certain words when they have other L1-L2 translation pairs (i.e. one-to-one mapping) in their mind. In other words, some participants may judge the target items in terms of their preference to the English cue rather than whether they are used correctly in the sentences or not. In order to avoid (or at least to decrease) this possibility, the Japanese sentences and their English cues were modified or rewritten on the basis of the learners’ responses in the pilot study. 151 Second, this method could be weakened if the English cue does not help participants clarify the context and the meaning of the target word, that is if their language proficiency in that language (English here) is not sufficient. However, all the participants in the present study were university students or graduates of universities in Australia. Thus, their English was considered to be proficient enough to understand the cues in English at least to clarify the meanings of the target items and sentences in Japanese. Overall, I believe that there are more advantages than disadvantages in giving English cues to each sentence in this particular study. That is, English cues are effective in clarifying the intended meaning of the target items particularly when both Japanese and Chinese meanings of the deceptive cognates make sense in the sentence. For this reason, English cue was given for all of the short sentences. The 28 target items consisted of four kanji compounds for each of the four categories of kanji compounds, i.e., (S), (D), (O), and (N), and their subcategories. It has been claimed that the difficulty of (O) differs according to the way meanings overlap between Japanese and Chinese (Miura 1997; Matsushita 2001). That is, some of them have wider semantic ranges in Japanese, while others have wider semantic ranges in Chinese (Ueno & Lu 1995; Lin 2002). Therefore, in the present study, the former is referred to as (J>C), and the latter is referred to as (J 152 Although (N) is kanji compounds whose forms exist only in Japanese, some of them are relatively easier than others for CBLs to guess the meanings using their L1 knowledge (Chou 1986; Chen 2003b). This is largely due to the nature of kanji character as a logograph; every character has a basic meaning. Thus, the meaning of kanji compounds may be easier to guess if the meaning of each kanji character is transparent. For example, it is quite possible to guess the meaning of “腕時計(ude-do-kei)”(=watch), because the basic meanings of the characters are “腕(arm)”, “時(time)”, and “計(to measure)”. On the other hand, guessing of their meaning is extremely hard when the compounds have meanings which no longer correspond to the basic meaning of each character. For example, it is almost impossible to guess the meaning of “大切(tai-setsu)” (=important) from its each of the constituents “大(big)” and “切(to cut)”. It is possible that the acquisition process or developmental patterns of these words vary depending upon the comprehensibility of their meanings. Accordingly, for the purpose of the present study, (N) will be divided into two subcategories: (NN) and (NC). (NN) refers to those kanji compounds which exist only in Japanese, and the meanings are not comprehensible for CBLs using the knowledge of Chinese. On the other hand, (NC) refers to such kanji compounds whose meanings are comprehensible using the learner’s knowledge of Chinese. The selection of the target items for these categories was based on Chou’s (1986) survey. Chou (1986) asked 20 Chinese native speakers of no prior knowledge of Japanese whether they could understand the meanings of the kanji compounds which exist only in Japanese. On the basis of the Chou’s (1986) results, (NN) was selected from those words which only less than three out of 20 153 participants could understand, while (NC) was selected from those words which more than 13 participants understood the meaning. Unlike Chen’s (2003b) study, which measured the difficulty of inferring the meaning of novel kanji compounds for CBLs by using advanced level words, the present study aims to investigate whether participants of each proficiency level have knowledge of the meaning for each kanji compound, and their difficulty of inferring its meaning when they did not have prior knowledge. For this reason, all of the 28 target items in the present study were chosen from beginning or intermediate level in the two lists of kanji compounds which categorises them based on the semantic patterns between Japanese and Chinese (Bunkachou 1978; Ueno & Lu 1995). In constructing an acceptability judgement test, two types of sentences are necessary: those sentences in which target items are used correctly (CR), and those in which they are used erroneously (ER). Since deceptive cognates, such as (D) and (O), have different meanings in Japanese and Chinese, they were tested both in the Japanese meaning (CR) and in the Chinese meaning (ER). As for (D), four sentences were prepared using its Japanese meaning which was used correctly (DCR), and another four sentences were prepared using its Chinese meaning which was used erroneously (DER) in the sentence. As an example of (DCR), a Japanese word “階段(kai-dan)”, which means “stairs” (“phase” or “stage” in Chinese), is used correctly in the following sentence: そのビルはエレベーターがないので、 階段で上がらなければなりません。(The building does not have lifts, so we need to go up by the stairs.). In contrast, a Japanese word “新聞(shin-bun)”, which means “newspapers” is 154 erroneously used in its Chinese meaning (i.e. “news”) in: 昨日、テレビで新聞を見ました。(I watched news on TV yesterday.). This is an example of (DER). Likewise, four sentences were prepared for (J>C) using the Japanese meaning (OCR), and four sentences were prepared for (J Japanese. For these reasons, (S), (NN) and (NC) were tested only as correct Japanese sentences. The target items and the sentences in which they are embedded are shown in Table 3.2. Table 3.2: Target Items in the Acceptability Judgement Test Category CR Target Sentences in which the target items are embedded /ER Items (S) CR 雑誌 毎月ファッション雑誌を買っています。 “Good (zas-shi) (I buy fashion magazines every month.) Cognates” 数学 私は学生の時、数学がきらいでした。 (suu-gaku) (When I was a student, I hated mathematics.) 鉛筆 最近、鉛筆を使う人は少なくなってきました。 (en-pitsu) (These days, less people are using pencils.) 政府 日本政府はアメリカ政府の要求を断れません。 (The Japanese government cannot decline the U.S. government’s (sei-fu) request.) CR 階段 そのビルはエレベーターがないので、階段で上がらなければなりませ (D) (kai-dan) ん。 “Totally 野菜 (The building does not have lifts, so we need to go up by the stairs.) Deceptive (ya-sai) 野菜をたくさん食べるのは健康にいいです。 Cognates” 迷惑 (Eating a lot of vegetables is good for our health.) (mei-waku) 毎晩となりの犬がうるさくて、とても迷惑です。 (The neighbor dog is noisy every night, so it is very annoying.) 約束 (yaku- 日曜日は友達と約束があるので、カラオケに行けません。 (On Sunday, I have an appointment with my friend, so I cannot go to soku) karaoke.) 155 ER 新聞 昨日、テレビで新聞を見ました。 (shin-bun) (I watched news on TV yesterday.) 汽車 私は毎日、汽車で大学に来ます。 (ki-sha) (I come to university by car everyday.) 貧乏 彼は仕事の経験が貧乏なのでまだ課長にはなれません。 (bin-bou) (His working experience is scarce, so he cannot be a manager yet.) 出世 私は20年前の今日、11時に出世しました。 (I was born today 20 years ago at 11pm.) (shus-se) (J>C) CR 反対 日本とオーストラリアは季節が反対です。 (han-tai) (The seasons in Japan and in Australia are opposite.) 得意 彼は料理を作るのが得意です。 O atal eetv Cognates Deceptive Partially (O) (toku-i) (He is good at cooking.) 差別 人種差別はまだ世界中にあります。 (sa-betsu) (Racial discrimination still exists all over the world.) 程度 DVDプレーヤーがほしいんですが、安いのはありますか？はい＄300程 (I’d like to buy a DVD player, but have you got a (tei-do) 度でありますよ。 cheep one? Yes, there are some around $300.) (J 3.3.3 Procedures 156 Before participants took part in the study, they were asked to fill in a subject consent form as well as their demographic information including ages, place of birth, first and second languages (third or fourth language if applicable), duration of living in Australia, duration of learning Japanese, and duration of living or staying in Japan. This was followed by the SPOT which was conducted specifically for the identification of participants’ proficiency level of Japanese. After participants had finished the SPOT, they were asked to take the acceptability judgement test. In the test, participants were asked to choose “”ifthe underlined word in each sentence was used correctly, and choose “” if it was used erroneously. Also they were asked to give a correct answer if they chose “”. There was no time limit for the test, so participants were able to spend as much time as they require. 3.4 Results and Discussion This section presents the results of the analyses of all the seven categories of Japanese kanji compounds, i.e. (S)(DCR)(DER)(OCR)(OER)(NN)(NC), for each L1 and proficiency level of participants. In order to analyse the results of the acceptability judgement test, the mean scores of each category were obtained for seven groups of participants according to their L1 and proficiency level (i.e. CBGs, CIMs, CADs, EBGs, EIMs, EADs and JNSs). One point was given for a participant’s correct judgement of individual target item, and no points were given for their wrong judgement or no response. Even when participants did not give a correct answer as a result of choosing “”, one point was given in the case of a correct judgement because they denied or doubted the usage of 157 the target item. The exception to this was when they corrected the pronunciation of the word in hiragana (although it was correct), and did not give an alternative word. This is because participants regarded the meaning of the target word as appropriate when they corrected its pronunciation only. Since each category included four target items, the maximum score for each category was “4”, and minimum score was “0”. In order to demonstrate differences between CBLs and EBLs of three different proficiency levels in their judgements of Japanese kanji compounds, a series of separate two-way ANOVAs with L1 (Chinese / English) and Level (beginning=BG / intermediate=IM / advanced=AD) as independent variables were performed on each of the seven categories of Japanese kanji compounds. The results of each of the seven categories of kanji compounds will be discussed in relation to the hypotheses formulated in an earlier section. 3.4.1 Use and Acquisition of Good Cognates (S) Hypothesis 1, which predicted that CBLs would outperform EBLs in their judgements of good cognate (S) due to their positive transfer, was confirmed by the results of the acceptability judgement test. Mean scores and standard deviations for the acceptability judgement test for each group on good cognates (S) are provided in Table 3.3. The results of the two-way ANOVA showed that both independent variables were significant (L1, F=10.948, p< .001; Level, F=23.423, p< .000). Also, the interaction between L1 and Level was found to be significant (F=13.484, p< .000). The results of the ANOVA are reported in Table 3.4, and the interaction between L1 and Level is depicted in Figure 3.1. The comparison of mean scores revealed that CBLs outperformed EBLs. A post hoc 158 comparison (Tukey HSD) among the participants of the three different proficiency levels indicated that intermediate and advanced learners judged significantly better than beginning learners (p< .000). However, there was no significant difference between the judgement of intermediate and advanced learners. Table 3.3: Means and Standard Deviations for Good Cognate (S) L1 LEVEL MSDn Chinese BG 3.87 .342 16 IM 3.96 .200 25 AD 4.00 .000 16 English BG 3.27 .594 15 IM 4.00 .000 18 AD 4.00 .000 12 Table 3.4: ANOVA for Good Cognate (S) Source SS df MS F Sig. L1 .872 1 .872 10.948 .001 LEVEL 3.730 2 1.865 23.423 .000 L1 * LEVEL 2.147 2 1.074 13.484 .000 4.2 4.0 3.8 3.6 3.4 L1 ns Chine se Mea 3.2 English BG IM AD LEVEL Figure 3.1: The Interaction between L1 and Level for Good Cognates (S) These results can be explained by CBLs’ positive transfer in the sense that novel L2 words can be inferred and interpreted successfully on the basis of their L1 knowledge (i.e. receptive transfer). Assuming that successful use of L1 knowledge or guessing based on L1 knowledge is likely to be incorporated into learners’ IL system, it could be argued that CBLs have much more chance to learn good cognate (S) as well as to guess its meaning 159 than EBLs do. From the perspectives of language development, CBLs are more likely to learn this type of cognate in relatively earlier stages of their L2 acquisition than EBLs do with the assistance of their L1. 3.4.2 Use and Acquisition of Non-Cognate Kanji Compounds (N) Hypothesis 2a, which predicted that CBLs would not outperform EBLs in judging (NN) or non-cognate kanji compounds whose meanings are not comprehensible for CBLs using their L1 knowledge, was also confirmed by the results of the acceptability judgement test. Mean scores and standard deviations for the acceptability judgement test for each group on (NN) are provided in Table 3.5. The results of the two-way ANOVA revealed significant effects for Level (F=17.548, p=.000). However, no significant effect for L1 was found (F=0.106, p=.745), nor any significant interaction between L1 and Level (F=1.900, p=.155). The results of the ANOVA are reported in Table 3.6, and the interaction between L1 and Level is depicted in Figure 3.2. These results seem to indicate that CBLs’ transfer is not at work or the influence of their L1 is minimal in their interpretation of unknown kanji compounds of this type. Table 3.5: Means and Standard Deviations for (NN) or Non-Cognate Kanji Compounds whose Meanings are Not Comprehensible for CBLs L1 LEVEL MSDn Chinese BG 3.13 .957 16 IM 3.56 .583 25 AD 3.87 .342 16 English BG 2.73 1.033 15 IM 3.78 .428 18 AD 3.92 .289 12 Table 3.6: ANOVA for Non-Cognate (NN) Source SS df MS F Sig. L1 .047 1 .047 .106 .745 LEVEL 15.582 2 7.791 17.548 .000 L1 * LEVEL 1.687 2 .844 1.900 .155 160 4.0 3.8 3.6 3.4 3.2 3.0 L1 2.8 nN1 Chine se Mea 2.6 English BG IM AD LEVEL Figure 3.2: The Interaction between L1 and Level for (NN) Hypothesis 2b, which predicted that CBLs would outperform EBLs in judging (NC) or non-cognate kanji compounds whose meanings are comprehensible for CBLs using their L1 knowledge, was confirmed. Mean scores and standard deviations for the acceptability judgement test for each group on (NC) are provided in Table 3.7. The results of the two- way ANOVA indicated significant effects for L1 (F=8.825, p=.004), and Level (F=14.626, p=.000), and a significant interaction between L1 and Level (F=4.562, p=.013). The results of the ANOVA are reported in Table 3.8, and the interaction between L1 and Level is depicted in Figure 3.3. Table 3.7: Means and Standard Deviations for (NC) or Non-Cognate Kanji Compounds whose Meanings are Comprehensible L1 LEVEL MSDn Chinese BG 3.56 .629 16 IM 3.72 .458 25 AD 3.94 .250 16 English BG 2.73 .961 15 IM 3.56 .511 18 AD 3.92 .289 12 Table 3.8: ANOVA for Non-Cognate (NC) Source SS df MS F Sig. L1 2.777 1 2.777 8.825 .004 LEVEL 9.205 2 4.602 14.626 .000 L1 * LEVEL 2.871 2 1.436 4.562 .013 161 4.2 4.0 3.8 3.6 3.4 3.2 3.0 L1 2.8 ns Chine se Mea 2.6 English BG IM AD LEVEL Figure 3.3: The Interaction between L1 and Level for (NC) The comparison of mean scores revealed that CBLs outperformed EBLs. A post hoc comparison (Tukey HSD) among the participants of the three different proficiency levels also indicated that intermediate and advanced learners judged significantly better than beginning learners (p< .001). However, there was no significant difference between the judgement of intermediate and advanced learners. These results are considered to be a reflection of CBLs’ positive transfer by which they can guess the meaning of a novel word using their L1 knowledge where many of the beginning learners did not have prior knowledge of the word. As is hypothesised, CBLs are likely to use their L1 knowledge to guess the meaning of (NC) where necessary, while they do not or cannot use their L1 knowledge fully in guessing the meaning of (NN). Therefore, it could be argued that the acquisition of (NC) is more likely to be facilitated than the case of (NN) for CBLs given that successfully transferred items tend to be integrated into their IL. 3.4.3 Use and Acquisition of Totally Deceptive Cognates (D) 162 Hypothesis 3, which predicted that CBLs would underperform EBLs in judging (D) or totally deceptive cognates due to their negative transfer, was partly confirmed. Mean scores and standard deviations for the acceptability judgement test for each group on (DER) or totally deceptive cognates which are erroneously used in the Chinese meaning within the Japanese sentence are provided in Table 3.9. The results of the two-way ANOVA revealed significant effects for L1 (F=38.442, p=.000), and Level (F=20.781, p=.000), and a significant interaction between L1 and Level (F=6.739, p=.002). The results of the ANOVA are reported in Table 3.10, and the interaction between L1 and Level is depicted in Figure 3.4. Table 3.9: Means and Standard Deviations for (DER) or Totally Deceptive Cognates which are Erroneously Used in the Chinese Meaning L1 LEVEL MSDn Chinese BG 1.69 1.195 16 IM 2.40 .913 25 AD 3.69 .479 16 English BG 3.27 .594 15 IM 3.56 .511 18 AD 3.83 .389 12 Table 3.10: ANOVA for Totally Deceptive Cognates (DER) Source SS df MS F Sig. L1 22.392 1 22.392 38.442 .000 LEVEL 24.210 2 12.105 20.781 .000 L1 * LEVEL 7.851 2 3.925 6.739 .002 4.0 3.5 3.0 2.5 2.0 L1 ns Chine se Mea 1.5 English BG IM AD LEVEL 163 Figure 3.4: The Interaction between L1 and Level for (DER) The comparison of mean scores revealed that CBLs underperformed EBLs. A post hoc comparison (Tukey HSD) among the participants of the three different proficiency levels revealed that there were significant differences among all of the three levels (p< .05). These results can be regarded as proof of CBLs’ negative transfer due to the nature of deceptive cognates, and are consistent with the claims made by a number of descriptive studies that CBLs are susceptible to L1 interference for this type of words. However, the consideration of CBLs’ developmental patterns reveals that these words are not necessarily difficult, because the mean scores increase with the development of their proficiency level. In addition, considering the fact that EBGs’ mean score is relatively high (3.27 out of 4.00), one could argue that the difficulty of these words is not necessarily high unless negative transfer is at work for CBLs. One possible reason for such a developmental pattern of CBLs is that it is relatively easy for them to notice the cross-linguistic semantic discrepancy of this type of deceptive cognates, and thus restructuring of their IL is also smooth. If this is the case, it is parallel with Miura’s (1997) claim that learners tend to produce errors of totally deceptive cognates when they have not learnt these words yet, while they are less likely to produce errors if they had existing knowledge about the different meaning in L2. A plausible explanation for this is that CBLs’ negative transfer either in comprehension or production for this type of cognates, which are content words and hence the most important information in the 164 message, tends to result in communication breakdown or promote negative feedback (in the case of production) for them. Table 3.11 shows mean scores and standard deviations for the acceptability judgement test for each group on (DCR) or totally deceptive cognates which are correctly used in the Japanese meaning within the Japanese sentence. The results of the two-way ANOVA showed significant effects for Level (F=19.856, p=.000), and a significant interaction between Level and L1 (F=3.418, p=.037). However, in contrast to the results of (DER), no significant effect for L1 was observed (F=0.733, p=.394). The results of the ANOVA are shown in Table 3.12, and the interaction between L1 and Level is depicted in Figure 3.5. A post hoc comparison (Tukey HSD) among the participants of the three different proficiency levels revealed that there were significant differences among all of the three levels (p< .05). Table 3.11: Means and Standard Deviations for (DCR) or Totally Deceptive Cognates which are Correctly used in the Japanese Meaning L1 LEVEL MSD n Chinese BG 3.13 .500 16 IM 3.36 .569 25 AD 3.81 .403 16 English BG 2.60 .910 15 IM 3.56 .616 18 AD 3.83 .389 12 Table 3.12: ANOVA for Totally Deceptive Cognates (DCR) Source SS df MS F Sig. L1 .257 1 .257 .733 .394 LEVEL 13.924 2 6.962 19.856 .000 L1 * LEVEL 2.397 2 1.198 3.418 .037 165 4.0 3.8 3.6 3.4 3.2 3.0 2.8 L1 2.6 ns Chine se Mea 2.4 English BG IM AD LEVEL Figure 3.5: The Interaction between L1 and Level for (DCR) These results suggest that despite the nature of deceptive cognates which are known to be susceptible to negative transfer, CBLs’ transfer is constrained where judgement of the correct usage in a certain context is involved. This implies that CBLs do not always interpret the meaning of (D) or totally deceptive cognates as the same as that of their Chinese counterparts. If CBLs had correct existing knowledge about the meaning of (D), as most advanced learners presumably did, negative transfer did not occur. Furthermore, even when the correct meaning of (D) was unknown for CBLs, they seem to interpret it using information from various sources including the knowledge of both L1 and TL that was available as well as contextual clues. In the acceptability judgement test, whereas CBLs tended to misinterpret the meaning of (D) where it was used in Chinese meaning (but made sense in the context), they were more likely to infer its Japanese meaning successfully where it was used correctly in the sentence. Yet, in a more realistic situation of language use such as reading a newspaper, there are no such erroneously used deceptive cognates. In reality, it is expected that CBLs are more likely to infer the meaning of (D) successfully rather than erroneously. This is consistent with Holmes & Ramos’s 166 (1993: 101) claim that “false cognates rarely led students astray” based on their group reading task whose topic was familiar to the participants. Questions still remain, however, because inferencing based on context does not always work (Nagy 1997; Schmitt 1997; Huckin & Coady 1999). This may be particularly true of beginning learners whose knowledge of the TL is minimal. Unlike the acceptability judgement test in this study, more realistic reading comprehension does not provide cues for clarifying the context. In other words, success or failure in guessing the meaning of (D) for beginning learners seems to be largely dependent on the quality of the contextual clues or how much contextual information they can obtain. As previous studies indicated (e.g. Hall 2002; Chen 2003b), learners tend to suppose that all cognates share the same meanings as their L1 counterparts in the first exposure if they are independent of contexts. Hence, it seems to be reasonable to assume that CBLs are less likely to misinterpret the meaning of (D) where sufficient contextual clues are available, while they are liable to fall into a trap, and infer its meaning erroneously if such clues are lacking or short. In sum, hypothesis 3, which posited that CBLs would underperform EBLs in judging totally deceptive cognates (D) due to their negative transfer, has over-predicted the occurrence of negative transfer. A more realistic picture would be that learners tend to infer and interpret the meaning of novel deceptive cognates successfully and correctly as long as sufficient contextual information is available. Besides, the mean scores of this type of cognates increase with the development of learners’ proficiency levels. Accordingly, although this type of cognates is susceptible to negative transfer if contextual clues are 167 deficient in earlier stages of CBLs’ L2 lexico-semantic development, it could be argued that their restructuring process is relatively smooth. 3.4.4 Use and Acquisition of Partially Deceptive Cognates (O) Hypothesis 4, which predicted that the acquisition of partially deceptive cognates (O) for CBLs would be prolonged due to its cross-linguistic semantic overlap, was generally confirmed. One distinctive feature of (O) that diverges from the kanji compounds of other categories is its polysemous nature at least either in Japanese or in Chinese. Figure 3.6 shows the semantic ranges of four target items for (J>C) between Japanese and Chinese. As Figure 3.6 indicates, each of the target items is a case of “convergence” since they are equivalent to two separate words in learners’ L1. As Jiang (2000, 2002, 2004a, 2004b) claims, most of the L2 words are mapped onto learners’ pre-existing L1 concepts, and the creation of new lexical entry is less likely to occur in their L2 lexico-semantic development. Hence, when the semantic range of the L2 words is not congruent with that of the L1 counterparts, it may be easier to observe learners’ misuse of the L2 words on the basis of their L1 semantic contents. Table 3.13 shows mean scores and standard deviations for the acceptability judgement test for each group on (OCR) or partially deceptive cognates which are correctly used in the Japanese meaning within the Japanese sentence. The results of the two-way ANOVA revealed significant effects for Level (F=17.047, p=.000), and a significant interaction between Level and L1 (F=3.708, p=.028). However, no significant effect for L1 was found 168 (F=3.836, p=.053). The results of the ANOVA are reported in Table 3.14, and the interaction between L1 and Level is depicted in Figure 3.7. L2: Japanese L1: Chinese L2: Japanese L1: Chinese de-yi fan-dui toku-i han-tai 得意 (proudly) 反対 (to disagree) (proudly / skillful) na-shou (to disagree / xiang-fan opposite) (skillful) (opposite) L2: Japanese L1: Chinese L2: Japanese L1: Chinese cheng-du cha-bie tei-do sa-betsu 程度 (extent) 差別 (distinction) (extent / about) zuo-you (distinction / qi-shi discrimination) (about) (discrimination) Figure 3.6: Semantic Ranges of Four Target Items for (J>C) (Shaded parts indicate the meaning in question) Table 3.13: Means and Standard Deviations for (OCR) or Partially Deceptive Cognates which are Correctly Used in the Japanese Meaning L1 LEVEL MSDn Chinese BG 1.88 1.025 16 IM 2.40 .764 25 AD 2.56 .892 16 English BG 1.53 .990 15 IM 3.11 .900 18 AD 3.25 .754 12 Table 3.14: ANOVA for Partially Deceptive Cognates (OCR) Source SS df MS F Sig. L1 3.015 1 3.015 3.836 .053 LEVEL 26.796 2 13.398 17.047 .000 L1 * LEVEL 5.828 2 2.914 3.708 .028 169 3.5 3.0 2.5 2.0 1.5 L1 ns Chine se Mea 1.0 English BG IM AD LEVEL Figure 3.7: The Interaction between L1 and Level for (OCR) These results seem to suggest that either CBLs’ L1 transfer is not at work or EBLs’ semantic transfer is also operative given that one L2 (Japanese) word is represented by two L1 (English) words for this category of words. However, in addition to these overall results of (OCR) or partially deceptive cognates which are correctly used in the Japanese meaning, I will discuss the results for individual target items for this type of cognates, since the mean scores and developmental patterns of each target item within this category largely varied. Figure 3.8 shows each group’s mean scores of (OCR) or partially deceptive cognates which are correctly used in the Japanese meaning within the sentence. As can be seen, learners’ developmental patterns of the target items in this category differ from those of other categories. Unlike (S), (N) and (D), where mean scores of advanced learners usually closely approximates to JNSs, there are some words whose mean scores are below 0.5 out of 1.0 even in the case of advanced learners. This indicates that the gap from JNSs regarding the knowledge of the words’ meaning is larger for some of the words in this category. Moreover, not only CADs but also EADs have a word whose mean score is below 0.5. This 170 implies that the difficulty of these words cannot be explained solely by CBLs’ negative transfer. One contributing factor to the relative difficulty of the words in this category seems to be their cross-linguistic semantic convergence for both CBLs and EBLs, but the mean scores vary largely from word to word. The possible reasons for this will be discussed for each target item. 1.2 1.0 .8 .6 .4 T OKUI .2 HANTAI TEIDO ean M 0.0 SABETS U CBG CIM CAD EBG EIM EAD JNS L1 Figure 3.8: Mean Scores of Correct Judgement for (OCR) or Partially Deceptive Cognate Used in Japanese Meaning Correctly Figure 3.9 shows each group’s mean scores of “差別(sa-betsu)”. As can be seen, the mean scores of “差別(sa-betsu)” are generally higher (except EBGs), and become higher as participants’ proficiency level increases. A plausible explanation for this is that between the two meanings of “差別(sa-betsu)”, “discrimination” is more frequently used than “distinction” in Japanese, and learners are more likely to encounter the former as an input of this word. In order to confirm this, the first 100 examples of Japanese word “差別(sa- betsu)” in an Internet search43 were collected, and the frequency of each usage was counted. The result showed that “差別(sa-betsu)” was used 83 times in the sense of “discrimination”, while it was used only 17 times in the sense of “distinction”. 43 It is realistic to expect students today to have access to the Internet, and that they are getting significant language input from there. Hence, I performed an Internet search to get a sense of the usage for these words. 171 1.2 1.0 .8 .6 .4 L1 nSABETSU Chine se Mea .2 English BG IM AD LEVEL Figure 3.9: Mean Scores of Correct Judgement for “差別(sa-betsu)” Thus, although the meaning in question “discrimination” is not shared with the cognate counterpart in Chinese “(cha-bie = distinction)”, CBLs tend to associate the Japanese word “差別(sa-betsu)” with another Chinese word “ (qi-shi = discrimination)” as a translation equivalent. If this is the case, “差別(sa-betsu)” is regarded as (D) or a totally deceptive cognate by CBLs, and most CADs have knowledge of this meaning of the word. However, it is not known whether they have knowledge of the other meaning, “distinction”, even though it shares the same meaning with the Chinese cognate counterpart “(cha- bie)”. Figure 3.10 shows each group’s mean scores of “程度(tei-do)”. As Figure 3.10 indicates, the mean scores of “程度(tei-do)” are very low even for advanced learners of both language backgrounds. For this reason, acquisition of this particular meaning or usage of the word seems to be extremely difficult for L2 learners regardless of their language background. The data suggest that CBLs’ association such as “程度(tei-do)” = “ (zuo-you)” (=about) or EBLs’ association such as “程度(tei-do)” =“about” was very weak if there was any as many of them denied (misjudged) the correct usage of “程度(tei-do)”. I would argue that 172 the main reason for their weaker association of such a pair is due to the lower frequency in the usage of “about” compared to “extent” within the same word “程度(tei-do)”. .5 .4 .3 .2 .1 L1 nTEIDO Chine se Mea 0.0 English BG IM AD LEVEL Figure 3.10: Mean Scores of Correct Judgement for “程度(tei-do)” In order to confirm this, the first 100 examples of Japanese word “程度(tei-do)” in an Internet search were collected, and the frequency of each usage was counted. It was found that “程度(tei-do)” was used 72 times in the sense of “extent”, while it was used only 28 times in the sense of “about”. In other words, “程度(tei-do)”isusedmoreofteninthesense of “extent” than that of “about”, and learners tend to learn the former usage first as a specific exemplar. As a result, they are liable to create such an associative link as “程度(tei- do)” = “(cheng-du)” extent). If this is the case, CBLs are likely to regard “程度(tei- do)” as (S) or good cognate. Another possible reason for the weak association of “程度(tei-do)” = “ (zuo-you)” (=about) for CBLs in particular is that once they have created a pair-association such as “程 度(tei-do)” = “(cheng-du)” extent)itishardtodiscardsuchanassociativelink because the Japanese word “程度(tei-do)” and the Chinese counterpart “(cheng-du)” are 173 the cognates, and seemingly share exactly the same semantic contents as well. According to Chiu (2003), CBLs tend to process Japanese-Chinese cognates as Chinese words while reading Japanese texts silently. Chiu (2003) claims that such a processing pattern prevents reinforcing the connection between meanings of the word, which comes with orthography, and their Japanese pronunciation. Although Chiu’s arguments are based on learning Japanese pronunciation of good cognates by CBLs, they shed some light on the strong associative link between cognates. Thus, it is possible that the pair-association formed by the strong connection between the Japanese word “程度(tei-do)” and the Chinese counterpart “(cheng-du)” hinders creating new connections such as “程度(tei-do)” = “ (zuo-you)” (=about). Figure 3.11 shows each group’s mean scores of “得意(toku-i)”. As Figure 3.11 shows, mean scores of “得意(toku-i)” grow gradually with the development of participants’ proficiency level for both CBLs and EBLs. Therefore, it could be argued that the Japanese word “得意(toku-i)” is more likely to be associated with the Chinese word “ (na-shou)” (=skillful) rather than its cognate counterpart “(de-yi)” (= proudly) by CBLs. 1.1 1.0 .9 .8 .7 .6 .5 KUI L1 .4 nTO Chine se Mea .3 English BG IM AD LEVEL Figure 3.11: Mean Scores of Correct Judgement for “得意(toku-i)” 174 In order to confirm such an assumption that CBLs are more likely to encounter “得意(toku- i)” in the sense of “skilful” rather than “proudly”, the relative frequency of each meaning was counted in the first 100 examples of “得意(toku-i)” in an Internet search. The result showed that “得意(toku-i)” was used 81 times in the sense of “skilful”, only twice in the sense of “proudly”, and 17 times in another meaning (i.e. customer). In other words, as for this sample, only 2% of the learners’ input of “得意(toku-i)” is in the sense of “proudly”, and more than 80% of their input is in the sense of “skilful”. This indicates that CBLs are likely to regard “得意(toku-i)” as (D) or totally deceptive cognate. In their formation of such an associative link, the frequency of each usage within this word is likely to play a role. Figure 3.12 indicates each group’s mean scores of “反対(han-tai)”. As can be seen, even though the overall mean scores of “反対(han-tai)” for each group are not necessarily low, they fall in advanced learners for both L1 groups. Considering the fact that the mean scores of beginning and intermediate learners of each L1 group are near 0.7, it can be assumed that many of the beginning and intermediate learners had prior knowledge of this usage. This also indicates that CBLs tend to associate Japanese word “反対(han-tai)” with the Chinese word “(xiang-fan)”(=opposite) probably due to its higher frequency than the other meaning “to disagree”. 175 1.0 .9 .8 .7 .6 AI L1 .5 nHANT Chine se Mea .4 English BG IM AD LEVEL Figure 3.12: Mean Scores of Correct Judgement for “反対(han-tai)” One possible reason that the mean scores of advanced learners are lower than those of intermediate learners in each L1 group is advanced learners’ confusion of the two different meanings within this word. As far as advanced learners are concerned, it is likely that they have obtained a certain amount of input of “反対(han-tai)” in both meanings, i.e., “opposite” and “to disagree”. Therefore, when advanced learners have noticed the polysemous nature of “反対(han-tai)”, they can learn both its meanings. However, if they fail to do so or persist with one-to-one mapping, they simply replace their existing associative link, i.e., “反対(han-tai)” = “(xiang-fan)”(=opposite) with the new link, i.e., “反対(han-tai)” = “ (fan-dui)”(=to disagree). This assumption was supported by the alternative word given by participants in their lexicality judgement test. Three CBGs and three CIMs judged the correct usage of the Japanese word “反対(han-tai)” as erroneous, and wrote “” as an alternative word by inferring it might exist in Japanese. On the other hand, five CADs and two EADs corrected it into “逆(gyaku)” which also means “opposite” in Japanese, and appropriate in the context of the sentence. Thus, the reason that the mean scores of advanced learners are lower than 176 those of intermediate learners could be ascribed to their knowledge of a new word which has the same meaning as the target item. That is, some of the advanced learners might have restructured their associative link from (1) “反対(han-tai)” = “(xiang-fan)”(=opposite) to (2) “反対(han-tai)” = “ (fan-dui)”(=to disagree) erroneously as a result of gaining the knowledge of “逆(gyaku)” by which they might have created a new associative link “逆 (gyaku)” = “(xiang-fan)”(=opposite). The results of (OCR) or partially deceptive cognates used in Japanese meaning correctly showed that many advanced learners of both L1 backgrounds denied the correct usage of some items and resulted in a wrong judgement (e.g. “程度(tei-do)”). This is in contrast to the results of the target items in other categories where most advanced learners accepted the correct usages in Japanese sentences. It seems that this is a reflection of their strong associative link between one of the two meanings of the target items and their L1 translation equivalents (i.e. specific exemplar). To put it another way, however, many CBLs accepted the correct usage of some items even though every item in this category elicited negative transfer (e.g.“差別(sa-betsu)”or “得意 (toku-i)”). In other words, if CBLs interpreted the meaning of each target item as the same as their L1 cognate counterpart, they would have rejected the usage of each item, and resulted in a wrong judgement. Hence, like the results of (D), CBLs do not always interpret (J>C) as a Chinese word, but they also use their existing L2 knowledge or contextual clues in judging the meaning of this type of words. It also seems that L2 input plays an important 177 part in CBLs’ creation of IL or forging an associative link between one of the meanings of the L2 word and an L1 word in the case of semantic convergence. Figure 3.13 shows the semantic ranges of four target items for (J Chinese. Table 3.15 indicates mean scores and standard deviations for the acceptability judgement test for each group on (OER) or partially deceptive cognates which are erroneously used in the Chinese meaning within the Japanese sentence. The results of the two-way ANOVA showed that both independent variables were significant (L1, F=3.935, p< .050; Level, F=14.038, p< .000). However, no significant interaction between Level and L1 (F=1.347, p=.265) was found (see Table 3.16 and Figure 3.14). L2: Japanese L1: Chinese L2: Japanese L1: Chinese mon-dai ryou-kai 問題 wen-ti 了解 liao-jie (problem / (to understand45) question44) (problem / question) (to understand) shitsu-mon ri-kai 質問 理解 (question) (to understand) L2: Japanese L1: Chinese L2: Japanese L1: Chinese kan-shin chi-hou 関心 guan-xin 地方 di-fang (interest) (region)) kiotsukau (interest / to care) ba-sho/tokoro (region / place) 気をつかう 場所 / 所 (to care) (place) Figure 3.13: Semantic Ranges of Four Target Items for (J 44 Although the Japanese word “問題(mon-dai)” also has both meanings of “question” and “problem”, the situation that the former can be used in is much more restricted than in Chinese. That is, “問題(mon-dai)” is acceptable in the sense of “questions in a test”, while another Japanese word “質問(shitsu-mon)” must be used in the sense of “asking a question”. 45 Whereas both Japanese word “了解(ryou-kai)” and its Chinese counterpart “ (liao-jie)” can mean “to understand”, the latter is used in wider contexts. For instance, the sentence “You don’t understand me” can be translated into Chinese using “ (liao-jie)”, while the Japanese translation cannot use “了解(ryou-kai)”. 178 Table 3.15: Means and Standard Deviations for (OER) or Partially Deceptive Cognates which are Erroneously Used in the Chinese Meaning L1 LEVEL MSD n Chinese BG 1.31 1.302 16 IM 1.92 1.077 25 AD 2.44 1.094 16 English BG 1.33 .724 15 IM 2.72 .752 18 AD 2.83 .937 12 Table 3.16: ANOVA for Partially Deceptive Cognates (OER) Source SS df MS F Sig. L1 4.009 1 4.009 3.935 .050 LEVEL 28.610 2 14.305 14.038 .000 L1 * LEVEL 2.744 2 1.372 1.347 .265 Figure 3.15 shows the mean scores of (OER) or partially deceptive cognates which are erroneously used in Chinese meanings within the Japanese sentence. Like the case of (OCR), overall mean scores are much lower than the target items in other categories, and the mean scores of advanced learners for some words are far from JNSs regardless of their L1. The factors that contribute to the varied levels of mean scores and developmental trends will be discussed for each target item. 3.0 2.5 2.0 1.5 L1 Chine se Means 1.0 English BG IM AD LEVEL Figure 3.14: The Interaction between L1 and Level for (OER) 179 1.2 1.0 .8 .6 .4 MONDAI .2 RYOUKAI KANSHI N ean M 0.0 CHIHOU CBG CIM CAD EBG EIM EAD JNS L1 Figure 3.15: Mean Scores of Correct Judgement for (OER) or Partially Deceptive Cognates Used in Chinese Meaning Erroneously Figure 3.16 indicates each group’s mean scores of “了解(ryou-kai )”. As can be seen, the mean scores of CBLs were much lower than those of EBLs for “了解(ryou-kai =*to understand)”, hence it is likely that negative transfer was at work in CBLs’ interpretation of this word. The main reason for this seems to be CBLs’ over-extension of the semantic ranges of their L1 word into its L2 counterpart. Since Chinese meaning of the target item makes sense in the sentence, CBLs are likely to interpret the meaning of the cognate on the basis of their L1 counterpart unless they have explicit knowledge of the word through positive evidence of the L2. .9 .8 .7 .6 .5 .4 .3 OUKAI L1 .2 nRY Chine se Mea .1 English BG IM AD LEVEL Figure 3.16: Mean Scores of Correct Judgement for “了解(ryou-kai )” 180 Moreover, as Jiang (2000, 2002, 2004a, 2004b) mentioned, the distinction between the two L2 words which share the same L1 translation equivalent tends to be hard even when learners have explicit knowledge of both words. As far as this target item is concerned, it seems to be congruent with Jiang’s claim because the same translation pair “了解(ryou-kai)” and “理解(ri-kai)” shares a similar meaning (i.e. to understand). Therefore, they might have been tolerant of the erroneous usage of “了解(ryou-kai)” and could not deny it because both concepts are expressed in the same word in their L1, despite their knowledge of these two L2 words. Figure 3.17 shows each group’s mean scores of “問題(mon-dai)”. It could be argued that acquisition of this word is very difficult due to the fact that advanced learners’ mean scores are lower than those of intermediate learners. This implies that there is no development in terms of the knowledge of the meaning of this word over three levels particularly for CBLs. A plausible explanation for this, again, would be learners’ confusion between the two meanings within this word. It is easy to suppose that even beginning learners tend to have input of “質問(shitsu-mon)” (= question) because “質問がありますか？(Are there any questions?)” is often asked of students by teachers. This can be seen in the results in that all correct judgements were successfully corrected to “質問(shitsu-mon)” or “しつもん(shitsu- mon)”. That is, relatively many beginning and intermediate learners had prior knowledge about the L2 word “質問(shitsu-mon)”, and in this case CBLs’ actual L2 knowledge inhibited their negative transfer. 181 .9 .8 .7 .6 .5 NDA I L1 .4 nMO Chine se Mea .3 English BG IM AD LEVEL Figure 3.17: Mean Scores of Correct Judgement for “問題(mon-dai)” However, the Japanese word “問題(mon-dai)” can be used to mean “question” in limited contexts (e.g. questions in a test), and many advanced learners are also likely to have the chance of obtaining the input of this word in this sense of “question” in addition to “problem”. As a result, some advanced learners appear to restructure their IL regarding the meaning of “問題(mon-dai)” erroneously by creating a new associative link such as “問題 (mon-dai)” = “(wen-ti)” (=question). Alternatively, some advanced learners may believe that both “問題(mon-dai)” and “質問(shitsu-mon)” are acceptable in the given context (i.e. “Teacher, I have a question”). Figure 3.18 indicates each group’s mean scores of “地方(chi-hou)”. As Figure 3.18 shows, the overall mean scores of “地方(chi-hou =*place)” are relatively high for each group, and they increase as learners’ proficiency level develops. In addition, all of the correct judgements were accompanied by a correct alternative word, i.e., “所(tokoro)” or “場所(ba- sho)”, which means “place” in Japanese. This suggests that many CBLs tend to have such an associative link as “所(tokoro)” or “場所(ba-sho)”=“(di-fang)” based on their L2 182 input, so it was relatively easy for them to deny the erroneous usage of “地方(chi-hou =*place)” in Japanese. 1.1 1.0 .9 .8 .7 .6 L1 .5 n CHIHOU Chine se Mea .4 English BG IM AD LEVEL Figure 3.18: Mean Scores of Correct Judgement for “地方(chi-hou )” Similarly, many EBLs are likely to have such an associative link as “所(tokoro)” or “場所 (ba-sho)” = “place”. Therefore, CBLs and EBLs show similar developmental patterns as far as the acquisition of Japanese word “地方(chi-hou)” is concerned. In other words, it can be assumed that both CBLs and EBLs tend to acquire the meaning of “所(tokoro)” or “場所(ba- sho)” in relatively early stages, and most advanced learners have such knowledge. Yet, whereas the result shows that the Chinese word “(di-fang)” is less likely to be transferred to the Japanese cognate counterpart “地方(chi-hou)” in the sense of “place” (negative transfer), it is unknown whether it can be transferred in the sense of “region” (positive transfer). Figure 3.19 shows each group’s mean scores of “関心(kan-shin)”. Although the mean scores of “関心(kan-shin =*to care)” go up with the development of learners’ proficiency level for both L1 groups, overall mean scores are low. As can be seen in Figure 3.19, even the mean scores of advanced learners are below 0.5 for this target item. The difficulty of 183 this item could be caused by the differences in the part of speech between the target item and the context. That is to say, even though the Japanese word “関心(interest)” is a noun, the underlined Chinese meaning of the word in the sentence was a verb (to care). Whereas some participants did judge this correctly and gave an alternative word “気をつかう (kiotsukau)”or“心配する(shinpai-suru)” successfully, most participants judged it wrongly or did not answer this question. Thus, the differences in the part of speech might have contributed to the difficulty of the judgement for this item. .5 .4 .3 .2 .1 L1 nKANSHIN Chine se Mea 0.0 English BG IM AD LEVEL Figure 3.19: Mean Scores of Correct Judgement for “関心(kan-shin)” The results of (OER) showed similar trends to those of (OCR): the mean scores largely varied from item to item even in the case of advanced learners. This implies that some of the meanings of partially deceptive cognates (O) are more difficult than others. Although cognancy may play a certain role in the formation of associative links or reinforcement of their connection in the sense that cognate pairs are more easily and strongly linked each other, other factors seem to play larger roles in CBLs’ use and acquisition of (O). I would argue that a more significant factor is the relative frequency of each meaning within a word. 184 In other words, even when CBLs are familiar with one of the meanings of the L2 word as a specific exemplar, they do not necessarily have the knowledge of its other meaning. 3.5 Summary of Results and Conclusion This study examined the possible differences in CBLs’ acquisition processes or developmental patterns of four types of Japanese kanji compounds by comparing their test results with EBLs. The acquisition process of Japanese kanji compounds for CBLs was found to differ according to the varied degree of orthographic and semantic correspondence between the Japanese and the Chinese. As predicted in Hypothesis 1, CBLs’ use and acquisition of good cognate (S) was facilitated owing to their positive transfer. CBLs can infer the meaning of (S) easily and correctly, and therefore, successfully inferred or transferred items are likely to be incorporated into their IL. As posited in Hypothesis 2, acquisition of non-cognate kanji compounds (N) was affected by positive transfer when they are comprehensible for CBLs using their knowledge of Chinese, while their acquisition is less likely to be influenced by transfer factors when they are incomprehensible for them. As Hypothesis 3 predicted, those CBLs who did not have prior knowledge of the correct meaning of totally deceptive cognate (D) were liable to fall into a trap, and inferred its meaning erroneously where it was used in Chinese meaning (and made sense in the context). Nevertheless, when (D) was used correctly in Japanese meaning within the context, CBLs inferred the meaning of (D) at least as well as EBLs did. This implies that contextual clues play a significant part in facilitating the guessing of the meaning of (D) 185 where they are available. Moreover, most advanced learners had correct knowledge of the meaning for this type of kanji compounds. Therefore, these results suggest that negative transfer involving (D) is not as pervasive and persistent as it has been considered in descriptive studies when a comprehension task is concerned. As hypothesis 4 predicted, many advanced learners failed to acquire some of the meanings of partially deceptive cognate (O). Since acquisition of this type of kanji compounds was found to be difficult not only for CBLs but also for EBLs, cross-linguistic semantic overlap (i.e. convergence or divergence) of these words appears to be an important difficulty- inducing factor. It is also assumed that learners tend to create one-to-one associative links between a particular usage of the L2 word and its L1 translation equivalent, and the creation of such IL is more strongly influenced by the frequency of the L2 input that they obtain rather than whether the counterpart is cognate or not. Therefore, where one of the meanings within an L2 word is more frequently used, and learners get more input of that particular meaning, acquisition of the less frequently used meaning tends to be prolonged regardless of the cognancy status. Moreover, learners may be confused about the distinction of multiple meanings or usages within an L2 word when they have encountered each of them as L2 input. The results of this study elucidated the roles of transfer in CBLs’ use and the acquisition of each type of kanji compound, and gave insight into the mechanism of their ease or difficulty. The most significant contribution of the current study in relation to the previous research is its far deeper understanding of the structure of CBLs’ bilingual lexicon or cross- 186 linguistic lexical mapping between Japanese and Chinese. Although Bunkachou (1978) has been extremely influential in the area of research on Japanese-Chinese kanji compounds or cognates, it must be pointed out that their view was too simplistic and naive. The current study has revealed the complex nature of CBLs’ mental lexicon or the IL that they tend to create by attempting to separate it from their inferences where CBLs did not have prior input of the word. However, the current study also has its limitations. Firstly, the number of target items for each category was limited, hence it would be premature to claim definitive conclusions based solely on this study. Secondly, another limitation of this study may be its inability to distinguish participants’ “knowledge” of the word which they have acquired already and “use” of the word by which they are inferring its meaning from various clues. Whereas this study assumed that learners’ L1 knowledge would be integrated into their IL through transfer as a means of communication strategy (i.e. by guessing the meaning of novel L2 words on the basis of L1 knowledge), there is no guarantee that this always happens (Corder 1978). As Ellis (1994) said, it is possible to make a clear distinction between the transfer in communication and learning theoretically, but it is difficult to do so empirically. In my future research, I would consider conducting a pre-test as one of the methodological options to check participants’ prior knowledge, although they are likely to learn the target items through the pre-test when lexical items particularly of cognates are dealt with. In addition, the comprehensibility of Japanese kanji compounds (N) for CBLs cannot always be discussed dichotomously because there may be something between the two end 187 points, and it may be subject to individual variation. Although the current study examined only the possible positive transfer items of (N), there might be possible negative transfer items or deceptive (N) as well. That is, an item may be seemingly comprehensible using their knowledge of Chinese, but it results in wrong interpretation if CBLs rely on their L1. Therefore, the nature of (N) would be much more complex than what was discussed in this study. Although this study has revealed the difficulty and the nature of CBLs’ acquisition of (O) or partially deceptive cognates to some degree, their use and acquisition of partially deceptive cognates also seem to be more complex than this study has found. Generally, CBLs’ L1 influence was not observed very clearly among eight partially deceptive cognates examined in the acceptability judgement test. I would argue that this is caused by the existing IL or associative link that learners have created based on their exposure to L2 input. That is, many of the CBLs and EBLs appear to have constructed IL of each target item (either correctly or erroneously) as a result of the L2 input, so two different L1 groups showed similar developmental patterns. Additional limitations of the study concern the task that was given to the participants. The acceptability judgement test investigated both the participants’ knowledge of each kanji compound and their ability to infer its meaning in certain contexts, but the task itself was not as natural and realistic as is the case in learners’ authentic language use such as reading a newspaper. The nature of the task, in which there were only two options for participants to choose from, might also have minimised the gap of the results between the two L1 188 groups. In other words, CBLs’ L1 influence may be more conspicuous if we use more comprehensible tasks and less frequent words as target items for which learners are less likely to have existing IL based on L2 input. Finally, whereas the acceptability judgement test in this study elicited negative transfer of (O) for both (J>C) and (J 189 CHAPTER FOUR THE ROLES OF TRANSFER IN CBLs’ USE AND ACQUISITION OF PARTIALLY DECEPTIVE COGNATES (STUDY B) 4.1 Introduction The results of Study A showed that, among the four types of Japanese Kanji compounds (S), (N), (D) and (O), (O) or partially deceptive cognates are the most difficult to learn not only for CBLs but also for EBLs. This is due to interlingual semantic convergence or divergence. The results of the acceptability judgement test also revealed that many advanced learners had difficulty in learning some of the meanings of (O) or partially deceptive cognates. Although CBLs’ judgement was not significantly different from that of EBLs in many items among (O), there were also a few items in which CBLs significantly underperformed compared to EBLs. It was also discovered that CBLs do not always interpret the meaning of Japanese-Chinese deceptive cognates as the same as that of their Chinese counterparts even when they appeared to lack prior knowledge of their meaning in L2. This suggests that CBLs use contextual clues as well as their L1 knowledge in guessing the meaning of novel cognates in L2 reading comprehension. However, it is still obscure how CBLs produce and understand cognates in L2 because the acceptability judgement test required them only to judge whether each usage of the cognates was correct or not. In addition, whereas the acceptability judgement test examined both types of semantic overlap of partially deceptive cognates (i.e. (J>C) and (J 190 compounds (O), but the roles of transfer in CBLs’ use and acquisition of Japanese- Chinese partially deceptive cognates are by no means clear. Therefore, Study B further explores CBLs’ possible positive transfer and negative transfer of partially deceptive cognates using both productive and receptive measures: an oral production test and a translation test. CBLs’ transfer of partially deceptive cognates in their oral production will be analysed on the basis of Færch & Kasper’s (1986a, 1987, 1989) classification of productive transfer. Among the three types of transfer in production, this study will consider strategic transfer and automatic transfer. CBLs’ receptive transfer in the present study will be assumed as a subcategory of inferencing (Færch & Kasper 1987, 1989), and hence it attempts to analyse how CBLs infer the meaning of novel cognates in L2: under what condition they use or do not use their L1 knowledge in their comprehension of Japanese-Chinese cognates. As the target items of the current study are partially deceptive cognates, each type of transfer above potentially involves both positive transfer and negative transfer according to the way in which semantic ranges overlap between Japanese and Chinese. As we have seen, however, cross-linguistic semantic congruence does not always result in positive transfer, and semantic discrepancy does not always trigger negative transfer. Thus, the factors that influence the occurrence of transfer other than the cross-linguistic semantic condition will also be considered if there are any. On the basis of Matsushita’s (2001) prediction of CBLs’ difficulty in their use of Japanese-Chinese partially deceptive cognates according to cross-linguistic semantic condition and task type, more detailed hypotheses are raised. 191 4.2 Research Question and Hypotheses The research question and hypotheses are the following: Research question: To what extent would the match and mismatch of semantic ranges of partially deceptive cognates across languages be able to predict the occurrence of positive transfer and negative transfer, respectively? :Production based on L1 knowledge :Reception based on L1 Knowledge L2: Japanese L1: Chinese hon-yaku H1H5H5 翻訳 fan-yi (translation) H2 AR tsuu-yaku H3 (translation / translator / 通訳 interpretation / interpreter) (interpretation / interpreter) H4 Figure 4.1: Hypotheses where Semantic Range of the Partially Deceptive Cognate is Wider in L1 (J (a) Where semantic range of the partially deceptive cognate is wider in L1, i.e., (J Hypothesis 1: its production in shared meaning will result in positive transfer. Hypothesis 2: its comprehension in shared meaning will result in positive transfer. Hypothesis 3: negative transfer will be triggered in an attempt to produce the semantic range that is not covered by the cognate in L2. Hypothesis 4: comprehension of the non-cognate L2 word whose meaning is covered by a cognate in L1 will not trigger transfer. 192 L2: Japanese L1: Chinese H5 fan-dui han-tai 反対 H6 (disagree) H7 xiang-fan (disagree / opposite) & H8 (opposite) Figure 4.2: Hypotheses where Semantic Range of the Partially Deceptive Cognate is Wider in L2 (J>C) (b)Where semantic range of the partially deceptive cognate is wider in L2, i.e., (J>C): Hypothesis 5: its production in shared meaning will result in positive transfer. Hypothesis 6: its comprehension in shared meaning will result in positive transfer. Hypothesis 7: its production in unshared meaning will be avoided. Hypothesis 8: its comprehension in unshared meaning will result in negative transfer. 4.3 Methodology In order to examine the role of transfer in both CBLs’ productive and CBLs’ receptive use of cognates in Japanese L2, an oral production test and a translation test were constructed. According to the hypotheses raised in the previous section, three Japanese- Chinese partially deceptive cognates each were selected to test the eight hypotheses. As a result, a total of 24 partially deceptive cognates became target items: 12 items were used in the oral production test and 12 in the translation test. These target items were selected from the lists that categorise Japanese-Chinese deceptive cognates based on their cross-linguistic semantic patterns (Bunkachou 1978; Ueno & Lu 1995). Whereas each Japanese kanji compound is usually a noun, it can be used in various grammatical functions in conjunction with “na”“no” (adjective), “ni”(adverb)or“suru” (verb). Given that the number of partially deceptive cognate for each hypothesis was limited, 193 the grammatical function of target items in both the tests was not considered in their selection. Similar to Study A, participants’ prior knowledge of each target item was not controlled, so it was not clear whether a particular target item was known or unknown to each participant. However, if I conducted pre-tests for the participants to check their prior knowledge of each target item, it was expected that such tests would affect their performance significantly in the main study particularly because the target items in this study are cognates. In addition, it is impossible to assure that all target items are completely novel to all participants unless we use non-existent words or artificial words. Hulstijn (2003) states that an experiment with artificial words ranks relatively high on reliability in terms of controlling participants’ prior knowledge, but low on ecological validity. In contrast, use of real words may result in high validity but low reliability. As the present study aims to examine CBLs’ use of cognates in realistic and naturalistic situations, existent L2 words rather than artificial words were selected as target items. Although this is not without problems (i.e. the inability to control participants’ prior knowledge), such problems would be minimised by considering CBLs’ “use” of the target items and their “acquisition” as a continuum, and by comparing their performance with that of EBLs. 194 4.3.1 Participants The participants for both the oral production test and the translation test were 40 CBLs (27 CIMs and 13 CADs). As control groups, 33 EBLs (23 EIMs and 10 EADs) and 21 JNSs also sat for these tests. All participants were students or graduates of Australian universities, and they participated in this study voluntarily. As in Study A, participants were recruited immediately after the placement tests for Japanese courses held in the University of New South Wales at the beginning of semester 1 and after Japanese classes during the semester in 2004. Some graduates of Australian Universities were also asked via email whether they could participate in the study. The proficiency level of participants was measured by the SPOT (Simple Performance- Oriented Test) (Ford-Niwa, et al. 1995; Kobayashi, et al. 1996). Those participants who scored 50 points or above out of 65 points in the version A of the SPOT were regarded as advanced learners; those with 35~49 points as intermediate, and those with below 34 points were not included in this study. This decision was made as a result of findings in a small-scale pilot test prior to the main study. The levels roughly correspond to divisions used at the University of New South Wales where the SPOT is utilised as a placement test instrument. Many of the participants also took the acceptability judgement test in Study A. As a few target items of the current study (in the oral production test) were also examined in Study A, participants were asked to sit for the oral production test first, followed by the translation test and then the acceptability judgement test if they took all tests. This aimed to minimise the task effects in their test results. Participants’ demographic information is shown in Table 4.1. 195 Table 4.1: Demographic Information of Participants N Gender Age* Stay in Australia Japanese Stay in Japan (M/F) (yrs)* Study (yrs)* (mths)* CIMs 27 5/22 21.5 5.4 4.2 1.6 CADs 13 4/9 24.9 7.8 6.5 9.8 EIMs 23 7/16 23.8 N/A 5.9 5.5 EADs 10 7/3 25.0 N/A 7.9 13.0 JNSs 21 5/16 28.9 1.4 N/A N/A *Mean values 4.3.2 Instruments 4.3.2.1 Oral Production Test The oral production test, in the form of 34 flashcards each of which contains a picture and a short sentence describing it, was prepared to elicit 36 words. Among them, 12 flashcards were aimed to elicit the 12 target items in the current study (see Appendix B). The rest of the 24 flashcards were aimed to elicit other Japanese words for other studies (see Appendix C and Appendix F). The sentences each had a blank where the target item was to be provided by the participants. In order to confirm whether or not these flashcards effectively elicit the concepts in question, pilot studies were carried out on seven JNSs and five learners of Japanese who did not participate in the main study. As a result, some of the pictures and the ranges of the blank in the sentences were modified. As 36 flashcards were shown randomly, it was very unlikely that participants had noticed the purpose of the study. In one of the classrooms at the University of New South Wales, participants were individually asked to fill in the blank with the most appropriate word or phrase they can think of with reference to the picture while reading out the short sentence. The researcher administered all of the oral production tests. Participants were required to fill in the blank as quickly as possible, but they were not allowed to ask questions. If they 196 could not produce any words to fill in the blank in ten seconds, they were to move on to the next picture. Most participants completed the test within five minutes. The test procedure was audio-taped for subsequent analysis. 4.3.2.2 Translation Test As a translation test, I had written 12 short Japanese sentences using each of the 12 target items (see Appendix D). Additionally, English translation was given to those words (other than the target items) which were considered to be difficult to understand as a result of the pilot tests (e.g. “income”, “bankrupt” or “abortion”). The purpose of this was to promote participants’ translation of the target items by providing them with as many contextual clues as possible. The task was to translate the whole sentence including the target item into English. Participants were not informed of either what the target items were, or what the aims of the study were. The task of translating the whole sentence instead of translating only the target item was selected in order to avoid participants’ excessive attention to the target item, and their translation without considering the meaning of the whole sentence. In the same classroom as the oral production test, participants were asked to translate as much as they could even when they encountered unknown words or grammatical structures. There was no time limit for the test, so participants could spend as much time as they required. As is mentioned earlier, all of the participants were students or graduates of Australian universities, so their English was considered to be proficient enough to participate in the translation test in which only the choice of the English word in translating each target item was analysed. The target items used in both tests are shown in Table 4.2. 197 Table 4.2: Target Items for Each Hypothesis and Test Hypo- Semantic Test Target Sentences in which target items are embedded. thesis condition type items (Within the brackets are blank in the oral production test.) 1 production Oral ( oiietransfer Positive 輸入 日本はオーストラリアからたくさんの牛肉を 輸入)しています。 (yu-nyuu) (Japan imports a lot of beef from Australia.) 緊張 田中さんはとても(緊緊張し) ています。 (kin-chou) (Mr. Tanaka is very nervous.) 失敗 ロケットの打ち上げは(失失敗)しました。 Match (ship-pai) (The launch of the rocket failed.) 2 (income) oiietransfer Positive 一定 一定の収入 のある人じゃないと結婚したくありません。 Translation (it-tei) (I don’t want to marry anyone without a regular income.) 招待 今日のパーティーにアメリカ人の先生を招招待しました。 (shou-tai) (I invited an American teacher to the today’s party.) 所有 会社が倒産(bankrupt)したので所所有の土地と建物を売らなければなり (sho-yuu) ません。(Since the company has bankrupted, we must sell the lands J Mismatch B:どうして知ってるんですか？ (How do you know?) A:テレビの(**広告)で見たんです。 *コマーシャル, CM or 宣伝(sen (I saw it on the TV commercial.) -den) is the correct utterance. 4 釈明 証拠(proof)があるので釈釈明してもだめです。(解釈) Non-transfer Translation (shaku-mei) (There is a proof, so there is no point explaining it.) 開催 2000年にシドニーでオリンピックが開開催されました。(挙行) (kai-sai) (In the year 2000, the Olympic games were held in Sydney.) 上達 日本で二ヶ月間ホームステイをして、日本語はどのくらい上上達しま (jou-tatsu) したか。 (進歩) (After two months of home stay in Japan, how much has your Japanese improved?) Words in brackets are cognates involved. 5 production Oral (formally) oiietransfer Positive 訪問 オーストラリアのハワード首相が日本を公式 に(訪訪問)しまし (hou-mon) た。(Australia’s prime minister Howard formally visited Japan.) 無理 明日までにレポートを50ページ書くのは(無無理な) 要求です。 (mu-ri) (It is an unreasonable request to write a 50-page report by 注意 tomorrow.) (chuu-i) 天井が低いので頭に(注注意し) て下さい。 Match (The ceiling is low, so please pay attention to your head.) 6 適当 大学の卒業式にTシャツとジーンズは適適当ではありません。 (T-shirts and jeans are not appropriate in the graduation ceremony oiietransfer Positive (teki-tou) Translation 是非 at university.) (ze-hi) 今日のクラスでは人工中絶(abortion)の是是非についてディスカッショ 姿勢 ンをしました。(In today’s class, we had discussions about the pros (shi-sei) and cons of abortion.) コンピューターを使う時、座る姿姿勢が悪いと肩(shoulder)が痛くなり やすいです。(When using a computer, you are likely to get sore shoulders if your sitting posture is bad.) J>C 7 production Oral 反対 日本とオーストラリアの季節は反対です。 Mismatch Avoidance (han-tai) (The seasons in Japan and in Australia are opposite.) 得意 私は高校生の時、数学が一番得意)でした。 (toku-i) (When I was a high school student, I was good at maths best.) 必要 日本の大学で勉強するにはたくさんのお金が(必要)です。 (hitsu-you) (You need a lot of money to study at a Japanese university. ) 198 8 依頼 父はお客さんに急ぎ(urgent)の仕事を依依頼されたので日曜日も会社に eaietransfer Negative 行きました。(My father was requested to do urgent work by the Translation (i-rai) 呼吸 customer, so he went to the office on Sunday.) (ko-kyuu) テニスのダブルスの試合で、マイクさんとティムさんのペアはあま 夢中 り呼呼吸があっていませんでした。(In the doubles match of tennis, the (mu-chuu) combination of Mike and Tim’s pair was not good.) マイクさんは今、キャシーさんに夢夢中です。 (Mike is crazy about Kathy now.) 4.4 Results and Discussion The hypotheses of the present study were formulated to address the evidence for both positive and negative transfer triggered by the semantic condition of cognates between Japanese and Chinese. They postulate that the roles of transfer in CBLs’ use and acquisition of partially deceptive cognates vary according to the way in which their semantic ranges overlap across languages and task type. At the same time, however, the current study also assumes that some other factors (e.g. proficiency level, context or L2 input) may constrain the occurrence of transfer. Thus, such factors which possibly affect CBLs’ use and acquisition of Japanese-Chinese partially deceptive cognates will also be discussed in relation to each hypothesis. In order to compare the results of the both oral production test and translation test between CBLs and EBLs, a series of chi-square tests were carried out. L1 (Chinese and English) and Level (intermediate and advanced) were the independent variables, and semantic conditions were the dependent variables. The results of the chi-square tests revealed that there were significant differences in the rate of producing the target items between CBLs and EBLs for six out of 12 items in the oral production test (please refer to the following sections for the results of statistical analysis). Among the six items, CBLs produced correct utterances in three cases, and possible L1-induced errors in three cases. 199 In order to control the proficiency level of the participants, the results of the oral production test were also compared between CIMs and EIMs, and CADs and EADs, respectively using the chi-square test. Significant differences were found between CIMs and EIMs in their rate of producing the target items for seven out of 12 items. Among the seven items, CIMs produced correct utterances in four cases, and possible L1- induced errors in three cases. However, there was no significant difference between CADs and EADs in their rate of production for any of the 12 target items. The results of the translation tests for CBLs and EBLs were compared in terms of the percentage of the participants who had used particular English words as translation for each target item. As a consequence, it was found that CBLs used certain English words in their translation more frequently than EBLs for nine out of the 12 items, and the chi- square tests show that these differences are significant. Among the nine items, CBLs used certain English words correctly in their translation for four cases, and erroneously with expected L1-like English words for five cases. Likewise, the results of the translation test were also compared between each proficiency level using a chi-square test. The results show that CIMs used certain English words in their translation more frequently than EIMs for nine out of the 12 items, and the chi-square tests reveal that such differences are significant. Among the nine items, CIMs used certain English words correctly in their translation for four cases, and erroneously with expected L1-like English words for five cases. However, significant differences were found only in two out of 12 items between CADs and 200 EADs. For both of the two cases, CADs used certain English words erroneously with expected L1-like English words. These results seem to suggest that there are distinct patterns of preferences between learners of two different L1 backgrounds in their production and comprehension of the target items among intermediate levels, but not among advanced levels. These results may also imply that transfer for the purpose of communication strategy is at work more actively in CIMs’ production and comprehension of the cognates, while CADs are more likely to use their existing L2 knowledge or IL in the tests. In addition, there also existed some possible L1-induced utterances other than the expected utterances, which may indicate that various kinds of CLI are operative. Thus, it would be over simplistic to conclude that the occurrence of transfer is determined solely by the cross-linguistic semantic condition of the cognates. In this section, the results of the oral production test and the translation test will be analysed on the basis of each hypothesis. The results of the oral production test (hypotheses 1, 3, 5 and 7) will be discussed first, followed by the results of the translation test (hypotheses 2, 4, 6 and 8). 4.4.1 CBLs’ Use and Acquisition of Partially Deceptive Cognates in the Oral Production Test CBLs’ transfer in production will be analysed on the basis of Færch & Kasper’s (1986a, 1987, 1989) classification into three types of productive transfer which differ according to the level of consciousness and automaticity: strategic transfer, automatic transfer and subsidiary transfer. However, subsidiary transfer will not be considered in the present study due to its ambiguous definition and difficulty of detection. 201 Subcategories of strategic transfer include “borrowing”, “foreignizing” and “literal translation”. Borrowing is the strategy in which learners use L1 lexical items with L1 pronunciation, while foreignizing refers to the strategy in which they choose L1 lexical items and adjust them phonologically or morphologically into L2. In the present study, literal translation in which learners translate each element of the L1 word directly into L2 will not be considered because the target items of the current study are cognates. Although Færch & Kasper (1986a) stated that automatic transfer would involve neither content words nor phonological adaptation of L1 lexical items into L2, much counter- evidence was shown by Poulisse (1999) in her study of slips of the tongue. Thus, the present study assumes that automatic transfer may involve both content words and phonological adjustment. 4.4.1.1 Production of (J Hypothesis 1, which predicted that production of the cognate in shared meaning, where semantic range is wider in Chinese, would result in positive transfer, was confirmed by the results of the oral production test. This hypothesis was tested using the three items, “緊張(kin-chou)”, “失敗(ship-pai)” and “輸入(yu-nyuu)”. Their cross-linguistic semantic patterns between the Japanese and the Chinese are shown in Figure 4.3. Table 4.3, Table 4.4 and Table 4.5 show comparisons of the number and percentage of the participants who have orally produced each target item among CBLs, EBLs and JNSs, between CIMs and EIMs, and between CADs and EADs, respectively. Three chi-square tests were performed for each of the three items between (1) CBLs and EBLs, (2) CIMs and EIMs, and (3) CADs and EADs in order to examine the differences between two L1 groups in their choice of the word by controlling proficiency levels. 202 Table 4.4 indicates that more CIMs produced the two target items “緊張(kin-chou)” and “失敗(ship-pai)” than EIMs did, and the chi-square tests show that these differences are significant. These results can be explained by the positive transfer triggered by semantic congruence of the cognate pair. L2:Japanese L1:Chinese L2:Japanese L1:Chinese L2:Japanese L1:Chinese kin-chou ship-pai yu-nyuu jin-kou 46 緊張 jin-zhang 失敗 shi-bai 輸入 (nervous) - (to fail) 5 (import) (import ) (nervous / (to fail / be shu-ru busy) defeated) G4 isogashii makeru nyuu-ryoku (import / 忙しい 負ける 入力 (busy) (be defeated) (input) input) Figure 4.3: Semantic Patterns of the Target Items for Hypothesis 1 (Shaded parts indicate the meaning in question) Table 4.3: Number and Percentage of the Participants who have Orally Produced Each Target Item (CBLs, EBLs and JNSs) 47 CBLs % EBLs %JNSs% kin-chou kin-chou 30 75 19 58 21 100 緊張 agaru 4 10 2 600 sin-pai 4 10 2 600 No answer 2 5 10 30 0 0 Total 40 100 33 100 21 100 χ(1) = 2.49, p > .05 CBLs % EBLs %JNSs% ship-pai ship-pai 31 78 13 39 21 100 失敗 ochiru 5 13 1 300 No answer 4 10 19 58 0 0 Total 40 100 33 100 21 100 χ(1) = 10.97, p < .001* CBLs % EBLs %JNSs% yu-nyuu yu-nyuu 14 35 11 33 21 100 輸入 shu-nyuu 13 33 0 000 shin-kou 3 8 0 000 No answer 10 25 22 67 0 0 Total 40 100 33 100 21 100 Shaded parts indicate the expected utterance (*statistically significant) χ(1) = .02, p > .05 46 Although the word “輸入(yu-nyuu)” was included as (J 203 Table 4.4: Number and Percentage of the Participants who have Orally Produced Each Target Item (CIMs and EIMs) CIMs % EIMs % kin-chou kin-chou 20 74 9 39 緊張 agaru 3 11 2 9 sin-pai 3 11 2 9 No answer 1 4 10 43 Total 27 100 23 100 χ(1) = 6.23, p < .05* CIMs % EIMs % ship-pai ship-pai 21 78 4 17 失敗 ochiru 3 11 0 0 No answer 3 11 19 83 Total 27 100 23 100 χ(1) = 18.12, p < .001* CIMs % EIMs % yu-nyuu yu-nyuu 4 15 2 9 輸入 shu-nyuu 12 44 0 0 shin-kou 3 11 0 0 No answer 8 30 21 91 Total 27 100 23 100 Shaded parts indicate the expected utterance (*statistically significant) χ(1) = .04, p > .05 Table 4.5: Number and Percentage of the Participants who have Orally Produced Each Target Item (CADs and EADs) CADs % EADs % kin-chou kin-chou 10 77 10 100 緊張 agaru 1 8 0 0 sin-pai 1 8 0 0 No answer 1 8 0 0 Total 13 100 10 100 χ(1) = 2.65, p > .05 CADs % EADs % ship-pai ship-pai 10 77 9 90 失敗 ochiru 2 15 1 10 No answer 1 8 0 0 Total 13 100 10 100 χ(1) = .67, p > .05 CADs % EADs % yu-nyuu yu-nyuu 10 77 9 90 輸入 shu-nyuu 1 8 0 0 shin-kou 0 0 0 0 No answer 2 15 1 10 Total 13 100 10 100 Shaded parts indicate the expected utterance χ(1) = .67, p > .05 204 Unlike these two items, there was no significant difference in the rate of producing “輸 入(yu-nyuu)” between CIMs and EIMs. However, it does not necessarily mean that transfer was not operative in CIMs’ attempts to produce the concept of “import” because there were other possible L1-induced utterances which were not observed in the other two items. As Table 4.3 shows, only CBLs produced “*shu-nyuu” or “*shin-kou” each of which can be traced to a Chinese word. I would argue that these utterances are the results of unsuccessful “foreignizing” which is a subcategory of strategic transfer (Færch & Kasper 1986a, 1987, 1989). In Chinese, both “G4(shu-ru)” and “ (jin- kou)” are used for the concept of “import”, and they might have triggered “*shu-nyuu” and “*shin-kou” respectively. Whereas “*shin-kou” for “ (jin-kou)” is perfect in terms of phonological adjustment, the word “*進口(shin-kou)” does not exist in Japanese.48 The cognate pair “輸入(yu-nyuu)” and “G4(shu-ru)” shares the common meaning “import”. It could be argued that some CIMs have seen the word somewhere or had visual input of “輸入(yu-nyuu)” as an L2 word, and thus they might have been aware of the existence of the cognate. As Lightbown & Libben (1984) stress, it is important for learners to receive input of the L2 counterpart for both positive and negative transfer of cognates to occur. Even though some CIMs might have been aware of the existence of the cognate, they did not seem to have phonological information of the L2 counterpart. This can be assumed by the fact that 44% of the CIMs produced “*shu-nyuu” while only 11% of the CIMs produced “*shin-kou”. No EBLs produced such utterances. 48 It can be assumed that some CIMs produced “shin-kou” because of such semantic conditions as (J>C) where one L2 word (i.e. 輸入(yu-nyuu)) is represented by two L1 words (i.e. G4(shu-ru)”and (jin- kou)). Thus, this utterance does not necessarily reflect the semantic condition of (J 205 Yet, the data suggest that the pronunciation of “輸入(yu-nyuu)” was difficult to guess. Although 44% of the CIMs appropriately adjusted phonologically from “ru” to “nyuu”, they failed to adjust “shu” to “yu”. As Matsushita (2001) indicated, it would not be difficult to make phonological adjustment for those kanji characters that are in CBLs’ existing knowledge of Japanese. It can be inferred that 44% of the CIMs successfully adjusted the Chinese sound “ru” into the Japanese “nyuu” simply because they were familiar with kanji compounds such as “入学(nyuu-gaku)” (entrance into school) or “入 国(nyuu-koku)” (entry into a country) as positive evidence. On the other hand, they could not adjust the Chinese sound “shu” into the Japanese “yu” due to their lack of knowledge regarding the Japanese words that are made up of “輸(yu)” such as “輸送(yu- sou)” (transport) or “密輸(mitsu-yu)”(smuggling). The data also indicate that CBLs tend to rely on the L1 sound of the kanji character when they cannot guess the sound in L2, presumably due to their perception that many kanji compounds or individual characters have similar pronunciations in Japanese and Chinese. An important question here is whether we should regard those utterances induced by foreignizing as negative transfer error or not. It might be almost impossible for JNSs to think of the word “輸入(yu-nyuu)” or its meaning “import” by hearing “*shin-kou” unless they understand Chinese. However, some JNSs may be able to guess the word that CBLs meant by hearing “*shu-nyuu” in the same context. 49 As Table 4.4 shows, only 9% of the EIMs produced the correct utterance “yu-nyuu” and 91% of them could not produce any words to express the concept of “import”, while 15% of the CIMs 49 The Japanese sound “shu” can be written as “syu” in another style of description. In this way, it might be easier to understand that CBLs’ L1-based phonological guessing in this particular case (i.e. G (shu) for (syu)) is close to the sound of its Japanese counterpart (yu) given that it differs only in the first sound “s”. 206 produced the correct word and 44% of them produced “*shu-nyuu” at least. Hence, it could be argued that CBLs’ use of L1 knowledge in L2 production using foreignizing is not always negative transfer. Sometimes it can also be considered as positive transfer even when the phonological adaptation is not complete. Now it needs to be explained why CIMs did not produce errors induced by foreignizing such as “*shu-nyuu” for other items in this category (i.e. “失敗” and “緊張”). One possible explanation of this is that CIMs did use foreignizing, and guessed their pronunciation successfully. The Japanese pronunciation of “失敗(ship-pai)” is fairly close to that in Mandarin Chinese “5(shi-bai)”.50 Due to the phonological proximity, CBLs’ foreignizing or even “borrowing” by which learners produce L1 words without phonological adaptation might have been successful in communication. In addition to the possibility that CBLs used foreignizing or borrowing successfully for “失敗(ship-pai)”, it is also possible that some CBLs had prior knowledge of the word in L2 or IL which consisted of L2 phonology and L1 semantic content. Such IL was supposedly constructed as a result of an aural input51 of the word in L2 (i.e. “ship-pai”) which enabled CBLs to be aware of the existence of the cognate and its pronunciation. If this is true, it is consistent with Færch & Kasper’s (1987, 1989) claims: transfer is 50 Although it is beyond the scope of the present study, some CBLs might have guessed the Japanese pronunciation of “失敗(ship-pai)” on the basis of some dialects of Chinese. Based on a strategy interview with CBLs after their writing test of Japanese kanji words, Kato (2005) reports that some of the participants associated the Japanese sound of cognates with that of Taiwanese. According to the participants in Kato (2005), pronunciation of Japanese-Chinese cognates in such a dialect tends to be more similar to that of Japanese than in Mandarin. 51 Unlike the cognate pairs in European languages whose pronunciation as well as orthography are usually similar, Japanese-Chinese cognates do not always share similar pronunciation even when their orthography is identical. In addition, kanji or logographic characters carry far less phonetic information, if not none, than alphabetic letters do. For these reasons, we cannot expect that CBLs will automatically have the phonetic information of the cognate in L2 simply via “visual input” such as in reading comprehension. 207 used as a learning procedure in the learners’ attempt to establish a hypothesis about the L2 words, and such a hypothesis may be based on either their L1 knowledge alone or as the result of an interaction between their L1 knowledge and the L2 input. Although CBLs produced “失敗(ship-pai)” correctly, they did not necessarily have correct knowledge about the semantic gap of the cognate pair. The reason that their utterances resulted in positive transfer might be simply due to the semantic match with regard to the meaning in question (i.e. “to fail”). Similar explanation can also be made for “緊張(kin-chou)” except that pronunciation of the pair between “緊張(kin-chou)” and “-(jin-zhang)” may not be as similar as that of “失敗(ship-pai)” and “5 (shi-bai)”.52 Therefore, foreignizing for “緊張(kin-chou)” might not have been successful unless CBLs had prior knowledge of the pronunciation for each of the constituents in advance. It is difficult to make a clear distinction as to whether the learner had existing IL (i.e. L2 phonology and L1 semantic content) or used strategic transfer successfully by simply hearing their utterances unless they produce clear errors. Yet, considering the fact that more EIMs, who cannot use strategic transfer to compensate their lack of L2 knowledge, produced “緊張” (39%) than the other two words (i.e. “輸入” (9%) and “失敗” (17%))53, it can be inferred that more CIMs had existing IL of “緊張(kin-chou)” than that of “輸入(yu-nyuu)” or “失敗(ship-pai)”. 52 Although it is learners’ perception whether the pronunciation of the cognate in Japanese and Chinese sounds similar or not, generally some cognate pairs share more similar sounds than others. The main reason for different levels of similarity in their pronunciation is that kanji characters was not introduced to Japan at once in one time period but were introduced at different times and from different parts of China. The latest large-scale introduction of kanji was in the era from Song Dynasty to Ming Dynasty (1127-1662 A.D.) when the basis of contemporary Mandarin was formed. Thus, those words introduced before this period sound less similar to contemporary Mandarin (The Japan Foundation 1988). 53 Such results may reflect the frequency of input or use of each item for the learners. Given that most of the participants are university students, it is natural that “nervous” or “failure” are more relevant and frequently used than “import”. 208 4.4.1.2 Production of the Semantic Range that is not Covered by a Cognate in L2 Hypothesis 3, which predicted that negative transfer would be triggered in an attempt to produce the semantic range that was not covered by a cognate in L2, was confirmed in all of the three target items as a result of the oral production test. Hypothesis 3 was tested by eliciting negative transfer such as “問題(mon-dai)”, “翻訳(hon-yaku)” and “広 告(kou-koku)”. Their semantic patterns are shown in Figure 4.4. Table 4.6, Table 4.7 and Table 4.8 show comparisons of the number and percentage of participants who have orally produced the target items among CBLs, EBLs and JNSs, between CIMs and EIMs, and between CADs and EADs, respectively. Three chi-square tests were carried out for each of the three items between (1) CBLs and EBLs, (2) CIMs and EIMs, and (3) CADs and EADs. Table 4.7 indicates that more CIMs produced all of the three target items than EIMs did, and the chi-square tests revealed that these differences are significant. As can be seen in Table 4.6, the predicted utterance “問題(mon-dai)” was produced only by CBLs. This can be explained by the negative transfer triggered by the semantic mismatch (i.e. divergence) of the cognate between the Japanese and the Chinese. L2:Japanese L1:Chinese L2:Japanese L1:Chinese L2:Japanese L1:Chinese mon-dai hon-yaku kou-koku 問題 wen-ti 翻訳 fan-yi 広告 guang-gao (problem/ F) (translation) AR (advertisement) M question) (problem/ (translation / (advertisement shitsu-mon question) tsuu-yaku translator / komaasharu / commercial) 質問 通訳 interpreter / コマーシャル (question) (interpreter / interpretation) (commercial) interpretation) Figure 4.4: Semantic Patterns of the Target Items for Hypothesis 3 (Shaded parts indicate the meaning in question) 209 Table 4.6: Number and Percentage of the Participants who have Orally Produced Each Target Item (CBLs, EBLs and JNSs) 54 ( ) indicates self-repairs CBLs % EBLs %JNSs% mon-dai mon-dai 10 (3) 25 0 0 0 0 問題 shitsu-mon 30 75 33 100 21 100 Total 40 100 33 100 21 100 χ(1) = 9.56, p < .01* CBLs % EBLs %JNSs% hon-yaku hon-yaku 10 (1) 25 1 3 1 5 翻訳 tsuu-yaku 19 48 13 39 20 95 han-I 4 10 0 00 No answer 7 18 19 58 0 0 Total 40 100 33 100 21 100 χ(1) = 6.82, p < .01* CBLs % EBLs %JNSs% kou-koku 28 70 7 21 0 0 kou-koku komaasharu 3 8 4 12 11 52 広告 CM 3 8 3 9943 sen-den 2 5 1 315 No answer 4 10 18 55 0 0 Total 40 100 33 100 21 100 Shaded parts indicate the expected utterance (*statistically significant) χ(1) = 17.25, p < .001* Table 4.7: Number and Percentage of the Participants who have Orally Produced Each Target Item (CIMs and EIMs) ( ) indicates self-repairs CIMs % EIMs % mon-dai mon-dai 8(2) 30 0 0 問題 shitsu-mon 19 70 23 100 Total 27 100 23 100 χ(1) = 8.11, p < .01* CIMs % EIMs % hon-yaku hon-yaku 7 26 0 0 翻訳 tsuu-yaku 9 33 5 22 han-i 4 15 0 0 No answer 7 26 18 78 Total 27 100 23 100 χ(1) = 6.93, p < .01* CIMs % EIMs % kou-koku 24 89 5 22 kou-koku komaasharu 0 0 0 0 CM 0 0 0 0 広告 sen-den 0 0 0 0 No answer 3 11 18 78 Total 27 100 23 100 Shaded parts indicate the expected utterance (*statistically significant) χ(1) = 22.99, p < .001* 54 The chi-square test was conducted only between CBLs and EBLs. 210 Table 4.8: Number and Percentage of the Participants who have Orally Produced Each Target Item (CADs and EADs) ( ) indicates self-repairs CADs % EADs % mon-dai mon-dai 2(1) 15 0 0 問題 shitsu-mon 11 85 10 100 Total 13 100 10 100 χ(1) = 1.69, p >.05 CADs % EADs % hon-yaku hon-yaku 3(1) 23 1 10 翻訳 tsuu-yaku 10 77 8 80 han-i 0 0 0 0 No answer 0 0 1 10 Total 13 100 10 100 χ(1) = .67, p >.05 CADs % EADs % kou-koku 4 31 2 20 kou-koku komaasharu 3 23 4 40 広告 CM 3 23 3 30 sen-den 2 15 1 10 No answer 1 8 0 0 Total 13 100 10 100 Shaded parts indicate the expected utterance χ(1) = .34, p >.05 However, not necessarily all CBLs who produced “問題(mon-dai)” lacked the correct knowledge regarding the cross-linguistic semantic boundary of the cognate pair. Three CBLs (two CIMs and one CAD) repaired their own utterances from “問題(mon-dai)” to “質問(shitsu-mon)”55 immediately after their first utterance. This indicates that these CBLs produced negative transfer errors in spite of their knowledge regarding the correct utterance in L2. Thus, these examples can be considered as CBLs’ activation of highly automatised L1 knowledge in the absence of conscious control, namely, automatic transfer (Færch & Kasper 1986a, 1987, 1989) or L1-based slips of the tongue (Poulisse & Bongaerts 1994; Poulisse 1999). As Table 4.6 shows, “翻訳(hon-yaku)” was also produced only by CBLs with the exception of one EBL. In addition, one CAD repaired the negative transfer error “翻訳 55 When participants produced the target item and gave an alternative word, it was counted as the use of the target item even when it was followed by correct utterance. Thus, the number of participants who produced “mon-dai” and “hon-yaku” in Table 4.6, Table 4.7 and Table 4.8 includes those who repaired and gave an alternative word later. 211 (hon-yaku)” into the correct L2 utterance “通訳(tsuu-yaku)” immediately after her first utterance. This example can also be considered as the automatic transfer or L1-based slips. The reason that one EAD produced this error cannot be directly traced back to his L1 because “interpreter” is distinguished from “translator” in English as well. Interestingly, however, one JNS produced “翻訳者(hon-yaku-sha)” which means “translator” for the concept of “interpreter” erroneously. This may indicate that the distinction of the concept between “通訳(interpreter)” and “翻訳者(translator)” is somewhat ambiguous, and even Japanese native speakers may confuse or misuse them. If this is the case, the errors produced by an EAD and a JNS may be induced by the difficulty of semantically similar lexical items within the TL (Martin 1984; Hulstijn & Tangelder 1991; Takahashi & Tanaka 1992; Hiki 1995) rather than the influence of other languages. It is not clear to what extent this intralingual factor contributed to the utterances of CBLs, but it could still be argued that negative transfer was operative in their production of “翻訳(hon-yaku)”. This is because 26% of the CIMs produced this erroneous utterance, while no EIM did so, and the difference was statistically significant (see Table 4.7). Besides, 15% of the CIMs produced “*han-i” which can be inferred as an unsuccessful foreignizing from the Chinese word “AR(fan-yi)” to Japanese counterpart “翻訳(hon-yaku)”. This example suggests that those who produced “hon- yaku” are likely to have had an aural input (either in or outside of the classroom) of the word previously because it might be difficult to guess its pronunciation due to its Japanese and Chinese phonological dissimilarity and the relatively low frequency of each kanji character appearing in other kanji compounds. 212 As a result of the aural input, CBLs may construct an IL by overextending the semantic range of their L1 word “AR(fan-yi)” (i.e. translation / translator / interpretation / interpreter) into that of the L2 counterpart “翻訳(hon-yaku)” (i.e. translation) due to the Semantic Equivalence Hypothesis (Ijaz 1986) which states that conceptual patterns in the learners’ L1 provide the essential criteria for those in the L2. Therefore, it is entirely possible that those who produced “hon-yaku” have had previous aural input of this L2 word, and established L1-based IL as a result of the interaction between L2 input and L1 knowledge. That is, these CBLs’ IL for “翻訳” is composed of L2 phonology (i.e. hon-yaku) and L1 semantic content (i.e. translation / translator / interpretation / interpreter). This semantic over-extension also seems to apply to those who have not had aural input of “hon-yaku” previously because their attempted phonological adjustment (i.e. *han-i) is also based on the semantic range of their L1 word “AR(fan- yi)”. As Table 4.6 shows, “広告(kou-koku)” was produced by 70% of the CBLs, but it was also produced by 21% of the EBLs. No JNSs produced such utterance. Such EBLs’ utterances may be caused by their semantic transfer, since EBLs also use their L1 knowledge and create L1-based IL. Another possible reason for such utterances from EBLs is the ambiguity of the semantic distinction between (1) “広告(kou-koku)” and (2) “CM(shii-emu)”, “コマーシャル(komaasharu)”or “宣伝(sen-den)”. That is, even though the semantic range of the former is generally restricted to the written form as in a magazine advertisement, and that of the latter is generally restricted to TV or radio commercial, such a distinction in English is not as strict as it is in Japanese.56 The word 56 Like the English word “advertisement”, the Japanese word “広告(kou-koku)” is also a general term for various types of “advertisement”. However, the acceptability of “広告(kou-koku)” in the sense of TV 213 “advertisement” in English is defined as “a picture, set of words, a film etc that is used to advertise a product or service” (Longman Dictionary of Contemporary English: italic my own), although “commercial” may be more frequently used when TV or radio is involved. As can be seen in Table 4.7, no CIMs and EIMs produced correct utterances such as “CM(shii-emu)”, “コマーシャル(komaasharu)”or “宣伝(sen-den)”. I would argue that this is reflected by the frequency of learners’ input for each word in L2. In other words, “広告(kou-koku)” seems to be more frequently used than “CM(shii-emu)”, “コマーシ ャル(komaasharu)”or “宣伝(sen-den) especially in the classroom, so participants are more likely to have an aural input of “広告(kou-koku)” than others. The results show that no CIMs and EIMs produced the correct utterance and this indicates that these intermediate learners had constructed L1-based IL regarding the concept of “commercial message” by overextending the semantic range of the L1 word “M (guang-gao)” or “advertisement” (i.e. commercial / written advertisement). However, given that 61% of the CADs and 80% of the EADs produced the correct utterances (see Table 4.8), such IL is also likely to be restructured as their proficiency level increases supposedly by gaining such input as “CM(shii-emu)”, “コマーシャル(komaasharu)”or “宣伝(sen-den) which may make them notice the erroneous nature of their IL. Automatic transfer occurs when learners have correct explicit knowledge of the L2 word but the L1 counterpart is more automatised. Thus, it is difficult to judge whether learners simply did not have correct knowledge of the L2 words or did have explicit commercial seems to be much lower than the case of the English “advertisement”. As a native speaker of Japanese and Japanese language teacher, I would not use and accept such usages at all. This is also consistent with the results of the 21 JNSs who participated in this study. 214 knowledge but L1 forms were more automatised unless they repaired their own utterances. For this reason, there might have been more examples of automatic transfer such as “問題(mon-dai)” for “質問(shitsu-mon)”, “翻訳(hon-yaku)” for “通訳(tsuu-yaku)” or “広告(kou-koku)” for “CM(shii-emu)”, etc. in the results of the oral production test. Yet, I do not have clues if participants did not repair their utterances even when they realised their slips or they did not even notice their slips in spite of their correct knowledge. The only thing I am certain of is that automatic transfer or L1-based slip is in operation when learners attempt to produce the semantic range of the cognate which is covered in L1 but not by the L2 counterpart. These results support Jiang’s (2002) claim that the same translation pairs such as two L2 words “problem” and “question” both of which are translated as “wenti”inL1may trigger negative transfer for Chinese ESL learners. That is, Chinese ESL learners tend to use these L2 words interchangeably since they are mapped to the same L1 concept. According to Jiang (2002), even advanced ESL learners of Chinese background may use these same translation words erroneously in spite of their correct explicit knowledge with regard to the semantic content of each L2 word. Poulisse (1999) also reports similar examples of L1-based slips in the Dutch-English partially deceptive cognate “klok (clock / bell)” in which Dutch ESL learners produce “clock” as a slip where the context requires “bell”. Given that semantic overlap of the target items in Jiang (2002) and Poulisse (1999) were (E 215 I would argue that the reason for the different rate of negative transfer errors that CIMs produced among the three items in this category is owing to the number of participants who had existing L2 knowledge or L1-based IL for each item. We can assume that relatively many CIMs had prior knowledge of the correct semantic boundaries between “問題(mon-dai)” and “F)(wen-ti)”, because 70% of the CIMs produced correct utterance “質問(shitsu-mon)” and a further 3 (7%) CIMs repaired their slips with correct utterances (see Table 4.7). This is even clearer in the case of CAD, where 85% of them correctly produced “質問(shitsu-mon)”, and one (7%) CAD repaired her negative transfer error with a correct utterance, which indicates this participant had correct explicit knowledge. On the other hand, few CIMs (if there were any) seem to have had prior knowledge regarding the semantic discrepancy between “広告(kou-koku)” and “M(guang-gao)” since no CIMs produced correct utterances such as “CM(shii-emu)”, etc. Under those circumstances, CIMs’ negative transfer is more likely to be caused by their L1-based IL. Although 61% of the CADs produced correct utterances, which suggests that they had correct explicit knowledge about it, 31% of the CADs produced negative transfer errors. It is not clear how many of these negative transfer errors are caused by automatic transfer or simply lack of correct L2 knowledge (L1-based IL), but both possibilities are plausible. As for “翻訳(hon-yaku)”, 33% of the CIMs produced the correct utterance “通訳(tsuu- yaku)”, so it suggests that at least these participants had existing knowledge regarding the semantic gap of “翻訳(hon-yaku)” and “AR(fan-yi)”. Therefore, one could argue 216 that both types of negative transfer (i.e. automatic transfer and production on the basis of L1-based IL) co-exist for these three items in varied proportions. Moreover, considering the fact that 77% of the CAD produced the correct utterance “通訳(tsuu- yaku)” as opposed to 33% of the CIM, CBLs’ knowledge on the semantic boundaries of Japanese word “翻訳(hon-yaku)” and “通訳(tsuu-yaku)” seems to develop as their proficiency level increases. As we have seen, CBLs’ correct utterances can also come from different levels of their knowledge about the target items and use of their L1 knowledge. That is, CBLs could have (1) had complete knowledge of the L2 word, (2) had L1-based IL (i.e. L2 phonology and L1 semantic content), (3) used foreignizing successfully while being aware of the existence of the cognate (i.e. visual input only), or (4) used foreignizing successfully without being aware of the existence of the cognate (i.e. no prior input). It is unclear from the results of the oral production test which utterances were caused by which level of CBLs’ L2 (IL) knowledge and L1 use, but it is clear that CADs tend to have more correct L2 knowledge than CIMs do. This is natural in a way, because CADs would have had more chances to encounter L2 input than CIMs. Therefore, if we consider that these different levels of CBLs’ L2 (IL) knowledge and L1 use are a continuum along which most of them pass through, it elucidates the positive nature of their transfer: constructing and restructuring their IL about the particular L2 word by actively utilising both L2 input and their L1 knowledge. 4.4.1.3 Production of (J>C) in Shared Meaning Hypothesis 5 predicted that production of the cognate in shared meaning, where the semantic range is wider in Japanese, would result in positive transfer. The hypothesis 217 was tested using “無理(mu-ri)”, “訪問(hou-mon)”, and “注意(chuu-i)” in the oral production test, and it was confirmed. Semantic patterns of the target items can be seen in Figure 4.5. Table 4.9, Table 4.10 and Table 4.11 show the number and percentage of the participants who have orally produced the target items according to L1 and proficiency level. Three chi-square tests were run for each of the three items between (1) CBLs and EBLs, (2) CIMs and EIMs, and (3) CADs and EADs. As Table 4.10 shows, two items, “無理 (mu-ri)” and “訪問(hou-mon)”, were produced by more CIMs than EIMs, and the chi- square test indicates that these differences are significant. Whereas there was no significant difference between CIMs and EIMs in their rate of producing “注意(chuu-i)”, there was a significant difference between CBLs and EBLs. These results can be explained by positive transfer triggered by semantic congruence of the cognate pair. L2:Japanese L1:Chinese L2:Japanese L1:Chinese L2:Japanese L1:Chinese wu-li fang-wen zhu-yi mu-ri DJ hou-mon =F chuu-i , 無理 (unreasonable) 訪問 (official visit) 注意 (to pay attention) (unreasonable/ qiang-po (visit) bai-fang (to pay attention jing-gao compulsorily) /towarn) 7 O= (compulsorily) (casual visit) (to warn) Figure 4.5: Semantic Patterns of the Target Items for Hypothesis 5 (Shaded parts indicate the meaning in question) Table 4.9: Number and Percentage of the Participants who have Orally Produced Each Target Item (CBLs, EBLs and JNSs) 57 CBLs % EBLs %JNSs% mu-ri mu-ri (na) 28 70 13 39 19 90 無理 fukanou (na) 3 8 0 000 dekinai 2 5 1 300 57 The chi-square test was conducted only between CBLs and EBLs. 218 No answer 7 18 19 58 2 10 Total 40 100 33 100 21 100 χ(1) = 6.88, p < .01* CBLs % EBLs %JNSs% hou-mon hou-mon 13 33 5 15 21 100 訪問 tazuneru 4 10 5 15 0 0 iku 5 13 6 18 0 0 No answer 18 45 17 52 0 0 Total 40 100 33 100 21 100 χ(1) = 2.93, p > .05 CBLs % EBLs %JNSs% chuu-i chuu-i 12 30 3 9 4 19 注意 kiotsukete 22 55 25 76 17 81 No answer 6 15 5 15 0 0 Total 40 100 33 100 21 100 Shaded parts indicate the expected utterance (*statistically significant) χ(1) = 4.84, p < .05* Table 4.10: Number and Percentage of the Participants who have Orally Produced Each Target Item (CIMs and EIMs) CIMs % EIMs % mu-ri mu-ri (na) 17 63 5 22 無理 fukanou (na) 3 11 0 0 dekinai 2 7 0 0 No answer 5 19 18 78 Total 27 100 23 100 χ(1) = 8.57, p < .01* CIMs % EIMs % hou-mon hou-mon 7 26 1 4 訪問 tazuneru 3 11 3 13 iku 4 15 6 26 No answer 13 48 13 57 Total 27 100 23 100 χ(1) = 4.30, p < .05* CIMs % EIMs % chuu-i chuu-i 9 33 3 13 注意 kiotsukete 12 44 15 65 No answer 6 22 5 22 Total 27 100 23 100 Shaded parts indicate the expected utterance (*statistically significant) χ(1) = 2.80, p > .05 Table 4.11: Number and Percentage of the Participants who have Orally Produced Each Target Item (CADs and EADs) CADs % EADs % mu-ri mu-ri (na) 11 85 8 80 無理 fukanou (na) 0 0 0 0 dekinai 0 0 1 10 No answer 2 15 1 10 Total 13 100 10 100 χ(1) = .08, p > .05 219 CADs % EADs % hou-mon hou-mon 6 46 4 40 訪問 tazuneru 1 8 2 20 iku 1 8 0 0 No answer 5 38 4 40 Total 13 100 10 100 χ(1) = .09, p > .05 CADs % EADs % chuu-i chuu-i 3 23 0 0 注意 kiotsukete 10 77 10 100 No answer 0 0 0 0 Total 13 100 10 100 Shaded parts indicate the expected utterance χ(1) = 2.65, p > .05 As Table 4.10 shows, 63% of the CIMs produced “無理(mu-ri)”, whereas only 22% of the EIMs did so. One could argue that this is caused by positive transfer both through learning procedure (i.e. construction of L1-based IL) and as a strategic transfer such as foreignizing or borrowing. Due to the phonological proximity of the cognate pair “無理 (mu-ri)” and “DJ(wu-li)”, and the relative prevalence of each character in other Japanese kanji compounds, strategic transfer might have been successful even when CBLs did not have phonological information of the cognate in L2. Table 4.10 also shows that 26% of the CIMs produced “訪問(hou-mon)”, but only 4% of the EIMs did so. Use of “訪問(hou-mon)” in oral production seems to be generally difficult for participants because only 46% of the CADs and 40% of the EADs produced the word (see Table 4.11), which may suggest that relatively fewer participants had L2 input of this word previously. Despite learners’ relative lack of L2 input, strategic transfer such as foreignizing seems to be effective for CIMs but not for EIMs. Whereas 33% of the CIMs and 23% of the CADs produced “注意(chuu-i)”, only 13% of the EIMs did so, and no EADs produced this utterance. Although this target item was not the only best utterance (“気をつけて(kiotsukete)” was also appropriate) in the 220 context of the flashcard, more CBLs used this word than EBLs did, and the chi-square tests revealed that this difference was significant (see Table 4.9). Thus, one could argue that CBLs have a relative preference for “注意(chuu-i)”, more so than EBLs do. That is, although more CBLs (44% of the CIMs and 77% of the CADs) produced “気をつけて (kiotsukete)” rather than “注意(chuu-i)”, which may reflect the frequency of learners’ input for each word58, 30% of the CBLs still chose “注意(chuu-i)”, in preference to “気を つけて(kiotsukete)”. A plausible explanation for such CBLs’ relative preference for “注 意(chuu-i)”, is the influence of their L1 counterpart “,(zhu-yi)” which can be used in the same situation. Seemingly, hypothesis 1 and hypothesis 5 are similar in that each of them assumed that semantic congruence of the cognate pair would facilitate positive transfer, and both of them were confirmed in the results of the oral production test. Because of the semantic match, CBLs’ strategic transfer tended to be successful as long as the phonological adaptation was appropriately made in each hypothesis. Nevertheless, I would argue that the semantic ranges underlying the IL which CBLs create differentiates the nature of the two semantic conditions predicted by each hypothesis, i.e., (J CBLs tend to create IL of partially deceptive cognates on the basis of the semantic ranges of their L1 counterparts unless they notice the correct semantic boundaries of these cognate pairs by encountering the positive evidence in L2. Negative transfer would be triggered when CBLs attempt to produce (J CBLs assume for the cognate is not always covered in L2. For instance, CBLs may use 58 Considering the fact that 81% of the JNSs used “気をつけて(kiotsukete)” and 19% of them used “注意 (chuu-i)” in the same oral production test, it can be inferred that JNSs generally prefer to use “気をつけて (kiotsukete)” in the same situation. 221 “緊張(kin-chou)” in the sense of “busy” in their L2 utterance if the semantic content of their IL is copied from its L1 counterpart (see Figure 4.6). Even when they have explicit knowledge regarding the correct semantic boundaries of the cognate pair, automatic transfer may be triggered. Although this particular example was not examined, the same principle was examined and confirmed by hypothesis 3. On the other hand, CBLs’ utterance will not result in negative transfer even when their IL is based on the semantic content of its L1 counterpart in the case of (J>C), because all the semantic contents that their IL contains are included in the L2 counterpart of the cognate. For instance, whereas the Japanese word “注意(chuu-i)” means “to pay attention” as well as “to warn” in different contexts, CBLs’ IL includes only the former if it is based on its L1 counterpart (see Figure 4.7). L2: Japanese CBLs’ L1-based IL kin-chou jin-zhang 緊張 positive transfer (nervous) - isogashii (nervous / busy) 忙しい negative transfer (busy) Figure 4.6: CBLs’ Production Using L1-based IL in the Case of (J L2: Japanese CBLs’ L1-based IL zhu-yi chuu-i positive transfer , 注意 (to pay attention) jing-gao (to pay attention / to warn) (to warn) Figure 4.7: CBLs’ Production Using L1-based IL in the Case of (J>C) 222 4.4.1.4 Production of (J>C) in Unshared Meaning Hypothesis 7 predicted that production of the cognates in unshared meaning, where the semantic range is wider in Japanese, would be avoided due to lack of the concept in the Chinese counterpart. Semantic patterns of the target items are shown in Figure 4.8. Table 4.12, Table 4.13 and Table 4.14 show the number and percentage of the participants who have orally produced the target items according to L1 and proficiency level. Three chi-square tests were performed for each of the three items between (1) CBLs and EBLs, (2) CIMs and EIMs, and (3) CADs and EADs. As a result, the hypothesis was denied in all of the three target items, “反対(han-tai)”, “得意(toku-i)” and “必要(hitsu-you)”, because there were no significant differences between CIMs and EIMs in their rate of producing them. These results seem to be derived from the nature of the IL that CBLs tend to establish in this type of semantic condition. L2:Japanese L1:Chinese L2:Japanese L1:Chinese L2:Japanese L1:Chinese fan-dui de-yi bi-yao han-tai toku-i 2 hitsu-you :I59 反対 (disagree) 得意 (proudly) 必要 (need) 59 The semantic distinction between the Chinese word “:I(bi-yao)” and “ I(xu-yao)” may seem unclear because each word can be translated as “need” in English. However, the context that each word can be used in is different. For instance, these two Chinese words are not interchangeable in the following two sentences. “?H$!L0:I (There is no need to say anything.) ”/ “3@6> I %Q3;” (You need at least 50 million yen to buy a house in Japan.) 223 (disagree / xiang-fan (proudly / na-shou (need / xu-yao opposite) skillful) N require) I (opposite) (skillful) (require) Figure 4.8: Semantic Patterns of the Target Items for Hypothesis 7 (Shaded parts indicate the meaning in question) Table 4.12: Number and Percentage of the Participants who have Orally Produced Each Target Item (CBLs, EBLs and JNSs) 60 CBLs % EBLs %JNSs% han-tai 25 63 19 58 12 57 han-tai gyaku 5 13 6 18 9 43 sou-han / 513 0 000 反対 shou-han No answer 5 13 8 24 0 0 Total 40 100 33 100 21 100 χ(1) = .18, p > .05 CBLs % EBLs %JNSs% toku-i toku-i 16 40 13 39 21 100 得意 jou-zu 18 45 12 36 0 0 yokatta 2 5 8 24 0 0 No answer 4 10 0 000 Total 40 100 33 100 21 100 χ(1) = .00, p > .05 CBLs % EBLs %JNSs% hitsu-you hitsu-you 17 43 16 48 19 90 必要 kakaru 11 28 9 27 2 10 iru 3 8 2 600 No answer 9 23 6 18 0 0 Total 40 100 33 100 21 100 Shaded parts indicate the expected utterance χ(1) = .26, p > .05 Table 4.13: Number and Percentage of the Participants who have Orally Produced Each Target Item (CIMs and EIMs) CIMs % EIMs % han-tai 16 59 15 65 han-tai gyaku 1 4 0 0 sou-han / 519 0 0 反対 shou-han No answer 5 19 8 35 Total 27 100 23 100 χ(1) = .19, p > .05 CIMs % EIMs % toku-i toku-i 9 33 8 35 得意 jou-zu 14 52 7 30 yokatta 2 7 8 35 No answer 2 7 0 0 Total 27 100 23 100 χ(1) = .01, p > .05 60 The chi-square test was conducted only between CBLs and EBLs. 224 CIMs % EIMs % hitsu-you hitsu-you 10 37 10 43 必要 kakaru 7 26 5 22 iru 1 4 2 9 No answer 9 33 6 26 Total 27 100 23 100 Shaded parts indicate the expected utterance χ(1) = .22, p > .05 Table 4.14: Number and Percentage of the Participants who have Orally Produced Each Target Item (CADs and EADs) CADs % EADs % han-tai 9 69 4 40 han-tai gyaku 4 31 6 60 sou-han / 000 0 反対 shou-han No answer 0 0 0 0 Total 13 100 10 100 χ(1) = 1.97, p > .05 CADs % EADs % toku-i toku-i 7 54 5 50 得意 jou-zu 4 31 5 50 yokatta 0 0 0 0 No answer 2 15 0 0 Total 13 100 10 100 χ(1) = .03, p > .05 CADs % EADs % hitsu-you hitsu-you 7 54 6 60 必要 kakaru 4 31 4 40 iru 2 15 0 0 No answer 0 0 0 0 Total 13 100 10 100 Shaded parts indicate the expected utterance χ(1) = .09, p > .05 Unlike the cases of hypotheses1, 3, and 5, where CBLs may produce either positive or negative transfer based on their IL which consists of L2 phonology and L1 semantic content, the IL that CBLs construct for the cognate in the case of hypothesis 7 does not necessarily include the semantic range of its L1 counterpart. This is because the meaning in question or the target semantic range of the cognate in L2 is not covered by the L1 counterpart. For instance, one of the meanings in “得意(toku-i)” (i.e. skillful) is not covered by the Chinese counterpart “2(de-yi)”. As a result, the IL that CBLs create for the cognate may contain L2 semantic content alone if they have encountered this L2-specific meaning before gaining the input of Japanese-Chinese common 225 meaning in L2. In other words, if CBLs come into contact with positive evidence that the cognate in L2 is used as the meaning which is not covered in their L1 counterpart, they may create IL whose semantic content is totally different from its L1 counterpart. As is mentioned in Study A (Chapter 3), CBLs are more likely to encounter the L2- specific meaning of “得意(toku-i)” (i.e. skillful) than L1-L2 common meaning (i.e. proudly). This was confirmed by counting the relative frequency of each meaning from the first 100 examples of “得意(toku-i)” in an Internet search. The result showed that “得 意(toku-i)” was used 81 times in the sense of “skilful”, only twice in the sense of “proudly”, and 17 times in another meaning (i.e. customer). That is, as far as this sample is concerned, only 2% of the learners’ input is the L1-L2 common meaning of “得意(toku-i)” (i.e. proudly), and more than 80% of their input is the L2-specific meaning of “得意(toku-i)” (i.e. skillful). Accordingly, it could be argued that the IL regarding the semantic content of the cognate created by CBLs was a reflection of their L2 input rather than its L1 counterpart, so they produced these items as frequently as EBLs in the oral production test. If this is the case, L2 input of the particular meaning of the cognate is crucial for CBLs to construct such L2-based IL. As can be seen in Table 4.13 and Table 4.14, the rate of producing “反対(han-tai)” slightly increases from CIMs (59%) to CADs (69%), while it decreases from EIMs (65%) to EADs (40%). These results seem to be caused by the interaction with the use of another word “逆(gyaku)” which also means “opposite”, given that 43% of the JNSs produced “逆(gyaku)” and 57% of them produced “反対(han-tai)” in the same test (see Table 4.12). Although no EIMs produced “逆(gyaku)”, 60% of the EADs did so in their utterances. At the same time, only 4% of 226 the CIMs produced “逆(gyaku)” and 31% of the CADs produced it for the same elicitation task. These results are consistent with those of the acceptability judgement test in Study A which examined the same word. The results of the acceptability judgement test on “反対 (han-tai)” (i.e.日本とオーストラリアは 季節が反対です。The seasons in Japan and in Australia are opposite. ) showed that advanced learners had the worst mean scores for both L1 groups. The reason for this was explained by the IL or associative link in which learners simply connect one of the meanings of the L2 word with one L1 translation equivalent. It was also shown in the acceptability judgement test that many intermediate learners or even beginners have knowledge of “反対(han-tai)” in the sense of “opposite” presumably due to its relatively high frequency of input for learners. As their proficiency level develops, however, learners are likely to encounter another word “逆 (gyaku)”, which also means “opposite”. Since the Japanese word “反対(han-tai)” also means “to disagree” and this meaning is shared in Chinese, it was argued that some advanced learners might have restructured their IL regarding the meanings of the word erroneously, and replaced the first IL (1) “反対(han-tai)” = “opposite” with new ones (2a) “逆(gyaku)” = “opposite” and (2b) “反対(han-tai)” = “to disagree”. In the process of such (erroneous) restructuring, L2 input of each word, i.e., “反対(to disagree)”, “逆 (opposite)”, must play a crucial role. This assumption in the previous chapter was confirmed in the results of the oral production test because 59% of the CIMs and 65% of the EIMs produced “反対(han-tai)” in the sense of “opposite”, and 31% of the CADs produced “逆(gyaku)”. 227 Interestingly, however, 19% of the CIMs produced “*sou-han” or “*shou-han” whose sound can be traced to the Chinese word “&8(xiang-fan)” which means “opposite”. This can be explained as an unsuccessful foreignizing by which CBLs attempt to adjust an L1 item into L2 phonology (see Figure 4.9). I would argue that the reason why these CIMs produced such utterances using strategic transfer is the lack of their existing L2 knowledge or L2-based IL. That is, these CIMs did not have prior knowledge of “反対 (han-tai)” in the sense of “opposite”, so they relied on the Chinese word “&8 (xiang- fan)” to express the concept of “opposite”. These examples indicate that the production of (J>C) in unshared meaning will be largely dependent on whether or not CBLs have an IL based on positive evidence of the L2 input. L2: Japanese L1: Chinese fan-dui han-tai 反対 (to disagree) xiang-fan Creation of non- (to disagree / opposite) existent L2 word & (opposite) Figure 4.9: L1-based Lexical Creativity when CBLs Lack the Knowledge of L2- specific Meaning of (J>C) Kamimoto, et al. (1992) claim that the absence of requisite L2 knowledge does not lead to avoidance since the learner is merely ignorant. The CBLs in the present study seem to have produced “反対(han-tai)” in the sense of “opposite” where they had such L2- based IL, while they did not use it when such IL was lacking. If we follow the definition by Kamimoto, et al. (1992), no avoidance is observed here. Hence, one could argue that the reason why CIMs’ overall use of “反対(han-tai)” did not differ from EIMs is because CIMs’ potential to produce the cognates in this semantic condition is, like the case of 228 EIMs, determined by their existing L2-based IL which is usually established on the basis of the L2 input they obtained. However, although CBLs do not avoid using the L2-specific meaning of cognates when they have its L2-based IL, they may produce negative transfer when such IL is yet to be created. That is, they may produce non-existent words in L2 such as “*sou-han (*相反)” or “*shin-kou(*進口)” by phonetically adapting L1 words which do not have cognate relationship with any L2 words. Like other examples of unsuccessful foreignizing such as “*shu-nyuu(輸入)” or “*han-i(翻訳)”, advanced learners rarely produced these kind of errors. This seems to be due to their rich knowledge of the L2 vocabulary which reduces their need to take a risk and reliance on transfer as a means of solving communication problems (Ellis 1994; Corder 1983). The results of the oral production test indicate that the use of “得意(toku-i)” is not necessarily easy for learners. As Table 4.13 shows, 52% of the CIMs and 30% of the EIMs produced “上手(jou-zu)” which has a similar meaning but is not appropriate in the context of the oral production test (i.e. 数学が得意でした。I was good at maths.). “上手 (jou-zu)” (=skillful) is normally used for physical dexterity, while this is not the case for “得意(toku-i)” (=skillful). It suggests that learning to use synonyms distinctively in an appropriate context is not a simple task. This can also be supported by the fact that 31% of the CADs and 50% of the EADs produced “上手(skillful)” erroneously (see Table 4.14). The acceptability judgement test which measured CBLs’ receptive transfer in Study A also tested “得意(toku-i)” in the sense of “skillful” (i.e. 彼は料理が得意です。He is good at cooking) which is not shared with Chinese. In contrast to the case of “反対 (han-tai)”, however, the results showed that the mean scores rose as learners’ 229 proficiency levels increased. The results of the oral production test and the acceptability judgement test on “得意(toku-i)” suggest that the same usage of the same word can be relatively easy in comprehension, but not in production due to the competition with synonyms. In Study A, the relative ease of “得意(toku-i)” in the sense of “skillful” was explained by the frequency effect of the L2 input for learners. Whereas it is used relatively frequently in daily conversation, the other meaning of “得意(toku-i)” (i.e. proudly), which is shared with the Chinese, is not used as frequently as “skillful”. Thus, it was argued that both CBLs and EBLs tend to come into contact with the word “得意(toku-i)” in the sense of “skillful” first, and create an L2-based IL or associative link such as “得 意(toku-i)” = “skillful”. Such L2-based IL contains the L2-specific meaning of the word, but it is still not complete as L2 knowledge: lacking another semantic range “proudly”. As we have seen in Study A, where some of the correct usages of partially deceptive cognates were strongly denied even by advanced learners, such CBLs’ L2-based IL on (J>C) occasionally results in under-extension. That is, even when the L1 semantic range of the cognate (e.g. “proudly”) can be incorporated into their IL, not all CBLs can acquire its complete L2 knowledge supposedly due to the strong associative link they have created or/and lack of positive evidence in the form of L2 input. Finally, the reason that the rate of producing “必要(hitsu-you)” did not differ even between CIMs and EIMs can also be accounted for in a similar way to the cases of the other two target items in this category. Due to lack of the concept in the L1 counterpart, production of the correct utterance in unshared meaning of (J>C) is largely dependent on CBLs’ L2-based IL which is usually constructed on the basis of their L2 input. For 230 this reason, it could be argued that the rate of producing “必要(hitsu-you)” both by CBLs and EBLs reflects the frequency of the L2 input they have obtained. When CBLs do not have such L2-based IL, they may use foreignizing and create new words which do not exist in L2. However, no such utterances were detected for “得意(toku-i)” and “必 要(hitsu-you)”. These results may reflect learners’ intuition about the cognancy: some L1 words are more likely to have cognate relationship with L2 words than others. In other words, L1-based lexical creativity may vary word-to-word, and possibly learner- to-learner. 4.4.2 CBLs’ Use and Acquisition of Partially Deceptive Cognates in the Translation Test Færch & Kasper (1987, 1989) claim that receptive L1 transfer constitutes a subcategory of inferencing. Thus, in the present study, CBLs’ use of L1 lexical information in guessing the meaning of cognates in L2 contexts will be examined. When the target items are novel as L2 words for participants, they need to infer the meaning of the cognates. As we have seen in the previous section, however, even when partially deceptive cognates are not novel for CBLs, they do not always have complete L2 knowledge regarding the semantic contents of these cognates. CBLs tend to overextend the semantic ranges of the cognates in L2 on the basis of their L1-based IL particularly in the case of (J As the present study assumes CBLs’ transfer in their “use” of a particular L2 word and its “acquisition” as a continuum, their existing IL will be considered as a result of either the success or the false success of inferencing in their previous encounter(s) to the 231 cognate unless they created such IL directly from their L1. For this reason, discussion for the results of the translation test will focus on the way in which participants infer the meaning of novel cognates. 4.4.2.1 Comprehension of (J Hypothesis 2 predicted that comprehension of the cognate in shared meaning between Japanese and Chinese, where semantic range is wider in Chinese, would result in positive transfer. This hypothesis was tested using the three items, “一定(it-tei)”, “所有 (sho-yuu)” and “招待(shou-tai)”. Their semantic patterns are shown in Figure 4.10. Table 4.15, Table 4.16 and Table 4.17 show the number and percentage of the participant who have translated the target items in different translations according to L1 and proficiency level. Three chi-square tests were run for each of the three items between (1) CBLs and EBLs, (2) CIMs and EIMs, and (3) CADs and EADs, and the results of the translation test confirmed the hypothesis only in one out of three items. As Table 4.16 shows, 93% of the CIMs translated “一定(it-tei)” correctly as “certain amount of”, “regular”, or similarly, but only 22% of the EIMs did so, and this difference was found to be significant in the results of the chi-square test. Therefore, one could argue that positive transfer facilitated CBLs’ comprehension of “一定(it-tei)”. L2:Japanese L1:Chinese L2:Japanese L1:Chinese L2:Japanese L1:Chinese it-tei sho-yuu shou-tai 一定 yi-ding 所有 suo-you 招待 zhao-dai (regular) / (to own) H (to invite) ' kanarazu (regular / subete (to own / all) motenasu (to invite / 必ず without fail) 全て もてなす to serve) (without fail) (all) (to serve) 232 Figure 4.10: Semantic Patterns of the Target Items for Hypothesis 2 (Shaded parts indicate the meaning in question) Table 4.15: Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CBLs, EBLs and JNSs)61 CBLs % EBLs %JNSs% it-tei certain amount of 10 25 5 15 10 48 一定 regular / stable / steady 26 65 5 15 10 48 No translation given for 410 23 70 1 5 the target item Total 40 100 33 100 21 100 χ(1) = 27.65, p < .001* CBLs % EBLs %JNSs% sho-yuu all/every 35 88 0 0 0 0 所有 own / hold 5 12 4 12 20 95 No translation given for 00 29 88 1 5 the target item Total 40 100 33 100 21 100 χ(1) = 55.47, p < .001* CBLs % EBLs %JNSs% take care of / look after 14 35 0 0 0 0 shou-tai serve / treat / entertain 4 10 0 0 0 0 invite 16 40 20 61 21 100 招待 introduce 0 0 3 900 No translation given for 615 10 30 0 0 the target item Total 40 100 33 100 21 100 Shaded parts are possible positive or negative transfer (*statistically significant) χ(1) = 19.71, p < .001* Table 4.16: Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CIMs and EIMs) CIMs % EIMs % it-tei certain amount of 5 19 0 0 一定 regular / stable / steady 20 74 5 22 No translation given for 2718 78 the target item Total 27 100 23 100 χ(1) = 25.98, p < .001* CIMs % EIMs % sho-yuu all/every 25 93 0 0 所有 own / hold 2 7 0 0 No translation given for 0023 100 the target item Total 27 100 23 100 χ(1) = 42.59, p < .001* CIMs % EIMs % take care of / look after 11 41 0 0 shou-tai serve / treat / entertain 3 11 0 0 invite 7 26 10 43 招待 introduce 0 0 3 14 No translation given for 62210 43 the target item Total 27 100 23 100 61 The chi-square test was conducted only between CBLs and EBLs. 233 Shaded parts are possible positive or negative transfer (*statistically significant) χ(1) = 16.56, p < .001* Table 4.17: Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CADs and EADs) CADs % EADs % it-tei certain amount of 5 38 0 0 一定 regular / stable / steady 6 46 5 50 No translation given for 215 5 50 the target item Total 13 100 10 100 χ(1) = 3.20, p > .05 CADs % EADs % sho-yuu all/every 10 77 0 0 所有 own / hold 3 23 4 40 No translation given for 006 60 the target item Total 13 100 10 100 χ(1) = 13.61, p < .001* CADs % EADs % take care of / look after 3 23 0 0 shou-tai serve / treat / entertain 1 8 0 0 invite 9 69 10 100 招待 introduce 0 0 0 0 No translation given for 000 0 the target item Total 13 100 10 100 Shaded parts are possible positive or negative transfer (*statistically significant) χ(1) = 3.73, p > .05 On the other hand, negative transfer seems to be triggered for “所有(sho-yuu)” and “招待 (shou-tai)” because only CBLs translated them incorrectly using particular English words which are considered to be L1-induced (i.e. “all” or “every” for “所有(sho-yuu)” and “take care of ”, “serve” etc. for “招待(shou-tai)”). In addition, the difference in the rate of choosing such CBLs’ possible L1-induced translation between CIMs and EIMs was found to be significant in the results of the chi-square test. The reason for the conflicting results with the hypothesis seems to be that the hypothesis overlooked the semantic range of the cognate which CBLs assume. For instance, the semantic contents that CBLs expect for the Japanese word “所有(sho-yuu)” (i.e. to own) tend to be that of the Chinese counterpart “ H(suo-you)” (i.e. to own / all). 234 Likewise, CBLs are likely to assume that the Japanese word “招待(shou-tai)” means “to serve / to invite” by overextending the semantic ranges of its Chinese counterpart “ ' (zhao-dai)”. Therefore, the data show that CBLs interpret the meaning of partially deceptive cognates whose semantic range is wider in Chinese not only as shared meaning (e.g. to own; to invite) but also as unshared meaning (e.g. all; to serve). In other words, potentially both positive transfer and negative transfer are operative in CBLs’ comprehension of (J The discrepancy between the results of the present study and Jiang’s (2004b) claim can be attributed to the target items that each study dealt with. The target items in Jiang (2004b) were the six pairs of English words which share the same Chinese translation (i.e. criterion / standard, complicated / complex, accurate / precise, safe / secure, insist / persist, doubt / suspect). Jiang (2004b) noted that the distinction between these pairs was very subtle, and there was much disagreement among ten English native speakers in their choice of the words between these pairs. On the other hand, the target items in the present study are Japanese-Chinese cognates whose multiple meanings are not always similar (e.g. to own / all; regular / without fail). In other words, the Chinese participants in Jiang (2004b) would not largely misinterpret the meaning of each L2 word simply because each of the same translation pairs had considerably similar 235 meanings. However, use of the same translation pairs which share dissimilar meanings as target items in the current study has revealed that CBLs still misinterpret the meaning of the same translation pairs where one lexical item in L1 corresponds to two or more wordsinL2asin(J The results of the translation test indicate that only “一定(it-tei)” triggered positive transfer while the other two items triggered a considerable amount of negative transfer within this same semantic condition. This can be explained by CBLs’ use of contextual cues in guessing the meaning of the target items. CBLs need to consider the whole meaning of the sentence in order to write a meaningful sentence in the translation task. In other words, if the meaning that CBLs inferred fits into the sentence, they may be satisfied and unsuspicious with their guessing. For example, the sentence “所有の土地を 売らなければなりません (we must sell the lands that we own)” was erroneously interpreted and translated as “we must sell *all the lands” by overextending the semantic range of the Chinese counterpart. Similarly, the sentence “アメリカ人の先生 を招待しました (I invited an American teacher)” was translated as “I served / entertained / took care of an American teacher”. Such misinterpretations were induced largely because the translated sentences were meaningful and made sense. In contrast, if their guessed meaning did not fit into the context, they need to think of an alternative interpretation. For example, when CBLs guessed the meaning of “一定(it- tei)” as “without fail” in the context of “一定の収入(income)のある人(the person who has regular income)”, the sentence does not make sense. Thus, I would argue that the reason why no such translation was given in the translation test is that “without fail” is normally used adverbially, while the context required an adjective or an adjectival 236 phrase. This might have lead CBLs to abandon their first hypothesis. This type of learners’ use of linguistic knowledge as one of the sources of contextual inferencing (e.g. by restricting word category or part of speech) is described in Nagy (1997). If CBLs’ inferred meaning clearly disagrees with the context, it is easy to see the misfit of the deceptive cognates. Yet, they are more likely to misinterpret the meaning of partially deceptive cognates when their inferred meaning or their hypothesis happen to make sense in the context. Therefore, it could be argued that the target meaning of (J (1993) in that although they also detected some examples of misinterpretation of false cognates, such examples were found only when they made sense in the context accidentally. As successful use of receptive transfer is likely to be incorporated into learners’ IL system (Færch & Kasper 1987, 1989; Ringbom 1991, 1992), CBLs may construct erroneous IL as a result of such false success. 4.4.2.2 Comprehension of the Non-Cognate L2 Word whose Meaning is Covered by a Cognate in L1 Hypothesis 4 predicted that comprehension of the non-cognate L2 word whose meaning is covered by a cognate in L1 would not trigger transfer. This hypothesis was tested using the three items, “釈明(shaku-mei)”, “開催(kai-sai)” and “上達(jou-tatsu)”, each of which shares some semantic ranges with cognate pairs such as “解釈(kai-shaku) / S (jie-shi)”, “挙行(kyo-kou) / (ju-xing)” and “進歩(shin-po) / <(jin-bu)”, 237 respectively (see Figure 4.11). Table 4.18, Table 4.19 and Table 4.20 show the number and percentage of the participants who have translated the target items using different English words according to L1 and proficiency level. L2:Japanese L1:Chinese L2:Japanese L1:Chinese L2:Japanese L1:Chinese kai-shaku kyo-kou shin-po 解釈 jie-shi 挙行 ju-xing 進歩 jin-bu (to interpret) S [hold (formal (progress in sci- < ceremonies etc.)] ence or thoughts) [ to hold shaku-nei (to interpret kai-sai jou-tatsu (progress) /to (ceremonies / 釈明 開催 上達 explain) meetings)] (to explain) [open (meetings / (progress in exhibitions etc.)] person’s skills) Figure 4.11: Semantic Patterns of the Target Items for Hypothesis 4 (Shaded parts indicate the meaning in question) Table 4.18: Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CBLs, EBLs and JNSs)62 CBLs % EBLs %JNSs% explain 22 55 4 12 5 24 shaku- excuse 3 8 0 01571 mei argue 2 5 2 600 clarify 2 5 0 000 釈明 No translation given for 11 28 27 82 1 5 the target item Total 40 100 33 100 21 100 χ(1) = 14.50, p < .001* CBLs % EBLs %JNSs% held 24 60 20 61 20 95 kai-sai opened 5 12 2 600 took place 0 0 6 18 0 0 開催 started 3 8 0 000 hosted 4 10 4 12 0 0 No translation given for 410 1 315 the target item Total 40 100 33 100 21 100 χ(1) = .00, p > .05 CBLs % EBLs %JNSs% jou-tatsu improve/make progress 33 83 27 82 21 100 上達 No translation given for 717 6 18 0 0 the target item Total 40 100 33 100 21 100 Shaded parts indicate the most frequently used word (*statistically significant) χ(1) = .01, p > .05 62 The chi-square test was conducted only between CBLs and EBLs. 238 Table 4.19: Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CIMs and EIMs) CIMs % EIMs % explain 15 56 3 13 shaku- excuse 1 4 0 0 mei argue 2 6 2 9 clarify 1 4 0 0 釈明 No translation given for 83018 78 the target item Total 27 100 23 100 χ(1) = 9.74, p < .01* CIMs % EIMs % held 13 48 12 52 kai-sai opened 4 15 1 4 took place 0 0 5 22 開催 started 2 7 0 0 hosted 4 15 4 18 No translation given for 415 1 4 the target item Total 27 100 23 100 χ(1) = .08, p > .05 CIMs % EIMs % jou-tatsu Improve/make progress 21 78 18 78 上達 No translation given for 622 5 22 the target item Total 27 100 23 100 Shaded parts indicate the most frequently used word (*statistically significant) χ(1) = .00, p > .05 Table 4.20: Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CADs and EADs) CADs % EADs % explain 7 54 1 10 Shaku- excuse 2 15 0 0 mei argue 0 0 0 0 clarify 1 8 0 0 釈明 No translation given for 323 9 90 the target item Total 13 100 10 100 χ(1) = 4.79, p < .05* CADs % EADs % held 11 84 8 80 kai-sai opened 1 8 1 10 took place 0 0 1 10 開催 started 1 8 0 0 hosted 0 0 0 0 No translation given for 000 0 the target item Total 13 100 10 100 χ(1) = .08, p > .05 CADs % EADs % improve 12 92 9 90 239 jou-tatsu improve 12 92 9 90 上達 Totaljou-tatsu 13 100 10 100 Shaded上達 parts indicate the most frequently used word (*statistically significant) χ(1) = .04, p > .05 As it was posited that no transfer would occur, there were no expected English words other than the correct translations. Hence, three chi-square tests were performed between (1) CBLs and EBLs, (2) CIMs and EIMs, and (3) CADs and EADs using the most frequently produced utterances. As a result, this hypothesis was confirmed in two out of three items. That is, there were no significant differences between CIMs and EIMs in their choice of the word as a translation for “開催(kai-sai)” and “上達(jou- tatsu)”. However, 56% of the CIMs used the word “explain” to translate “釈明(shaku- mei)” while 13% of the EIMs did so. Moreover, the chi-square test shows that this difference is significant. The results of the formers (i.e. “開催(kai-sai)” and “上達(jou- tatsu)”) can be explained by non-transfer for the non-cognate L2 words. Also, 54% of the CADs used the word “explain” to translate the sentence “証拠(proof)が あるので 釈明してもだめです。 (There is proof, so there is no point explaining it.)”, but only 10% of the EADs did so. This difference was found to be significant in the results of the chi-square test. Hence, the data suggest that some kinds of CLI are at work. I would argue that CBLs associated “釈明(shaku-mei)” with the Chinese corresponding word “ S(jie-shi)” which means “to explain” in the same context.63 Such CBLs’ guessing could also be caused by the fact that these words share the same character “釈 (shaku)” /“S(shi)” despite their simplification of the characters. Indeed, the individual character “S” by itself can be “explain” in Chinese. If this is the case, the significant 63 Its Japanese cognate counterpart “解釈(kai-shaku) can mean “to explain” in a limited context, but it is not appropriate in the context given in the translation test. 240 difference not only between CIMs and EIMs but also between CADs and EADs in their choice of the word “explain” in the translation task can be explained by CBLs’ positive transfer. As the results of Study A show, CBLs use transfer for non-cognate L2 words or kanji compounds (N) depending upon their comprehensibility using L1 knowledge. Thus, a plausible explanation might be that this semantic condition itself would not trigger receptive transfer of non-cognate L2 words for CBLs, but some kinds of CLI are operative even in the level of individual kanji character according to its comprehensibility for them. 4.4.2.3 Comprehension of (J>C) in Shared Meaning Hypothesis 6 predicted that comprehension of the cognate in shared meaning between Japanese and Chinese, where the semantic range is wider in Japanese, will result in positive transfer. The hypothesis was tested using “適当(teki-tou)”, “是非(ze-hi)”and “姿 勢(shi-sei)” in the translation test (see Figure 4.12 for their semantic patterns). Table 4.21, Table 4.22, and Table 4.23 show the number and percentage of the participants who have translated the target items in different English words. Three chi-square tests were carried out for each of the three items between (1) CBLs and EBLs, (2) CIMs and EIMs, and (3) CADs and EADs. This hypothesis was confirmed in two out of three items by the results of the translation test, because significant differences were found between CIMs and EIMs in their choice of English words in translating “適当(teki-tou)” and “是非(ze-hi)”. One could argue that these results are due to CIMs’ positive transfer triggered by semantic congruence of the cognate pair. 241 L2:Japanese L1:Chinese L2:Japanese L1:Chinese L2:Japanese L1:Chinese shi-dang shi-fei zi-sh teki-tou 1P ze-hi "9 shi-sei # 適当 (appropriate) 是非 (pros & cons) 姿勢 (posture) (appropriate / ma-hu (pros & cons / wu-bi (posture / zi-tai irresponsible) at any cost) attitude) K B: ( (irresponsible) (at any cost) (attitude) Figure 4.12: Semantic Patterns of the Target Items for Hypothesis 6 (Shaded parts indicate the meaning in question) Table 4.21: Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CBLs, EBLs and JNSs)64 CBLs % EBLs %JNSs% inappropriate / 23 58 13 39 13 62 teki-tou (not) appropriate 65 (not) suitable 15 38 1 3 7 33 適当 (not) acceptable 1 2 0 015 (cannot) wear / 12 14 43 0 0 (not) to be worn No translation given for the 00 5 15 0 0 target item Total 40 100 33 100 21 100 χ(1) = 24.39, p < .001* CBLs % EBLs %JNSs% ze-hi pros and cons / right or 23 58 6 18 18 86 wrong 是非 issue 3 7 1 300 No translation given for the 14 35 26 79 3 14 target item Total 40 100 33 100 21 100 χ(1) = 11.67, p < .001* CBLs % EBLs %JNSs% shi-sei posture / pose 29 73 18 55 15 71 姿勢 position 6 15 7 21 4 19 No translation given for the 512 8 24 2 10 target item Total 40 100 33 100 21 100 Shaded parts indicate possible positive transfer (*statistically significant) χ(1) = 1.70, p > .05 Table 4.22: Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CIMs and EIMs) CIMs % EIMs % inappropriate / 15 56 4 17 teki-tou (not) appropriate (not) suitable 11 41 0 0 適当 (not) acceptable 0 0 0 0 (cannot) wear / 1414 61 (not) to be worn No translation given for the 005 22 target item Total 27 100 23 100 64 The chi-square test was conducted only between CBLs and EBLs. 65 “適当(teki-tou)” (= appropriate) was used in negative meaning in the translation test, hence the translations given for this word include negative suffixes or “not”. 242 χ(1) = 32.22, p < .001* CIMs % EIMs % ze-hi pros and cons / right or 15 56 2 9 wrong 是非 issue 1 4 0 0 No translation given for the 11 40 21 91 target item Total 27 100 23 100 χ(1) = 12.15, p < .001* CIMs % EIMs % shi-sei posture / pose 19 70 13 57 姿勢 position 4 15 3 13 No translation given for the 415 7 30 target item Total 27 100 23 100 Shaded parts indicate possible positive transfer (*statistically significant) χ(1) = 1.77, p > .05 Table 4.23: Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CADs and EADs) CADs % EADs % inappropriate / 8 62 9 90 teki-tou (not) appropriate (not) suitable 4 31 1 10 適当 (not) acceptable 1 7 0 0 (cannot) wear / 000 0 (not) to be worn No translation given for the 000 0 target item Total 13 100 10 100 χ(1) = .80, p > .05 CADs % EADs % ze-hi pros and cons / right or 8 62 4 40 wrong 是非 issue 2 15 1 10 No translation given for the 323 5 50 target item Total 13 100 10 100 χ(1) = 1.05, p > .05 CADs % EADs % shi-sei posture / pose 10 77 5 50 姿勢 position 2 15 4 40 No translation given for the 181 10 target item Total 13 100 10 100 Shaded parts indicate possible positive transfer χ(1) = .04, p > .05 As Table 4.22 shows, 61% of the EIMs translated “ジーンズは適当ではありません (Jeans are not appropriate / not suitable)” as “cannot wear jeans” or “Jeans are not to be worn”, while only 4% of the CIMs did so. Although they are acceptable and correct as translations for this word, it seems that most of those who translated them as such lacked the knowledge of the meaning of “適当(teki-tou)” as an individual word. This 243 can be inferred from the fact that 100% of the EADs translated it using “appropriate” or “suitable” (see Table 4.23). This suggests that most of the EADs had prior knowledge regarding the meaning of “適当” (i.e. appropriate), and hence they translated it as such. For these reasons, the differences in the translation of “適当(teki-tou)” between CIMs and EIMs can be generally ascribed to CIMs’ use of L1 knowledge and EIMs’ use of contextual clues in guessing its meaning unless they had existing L2 knowledge. It can also be said that guessing the meaning of “適当(teki-tou)” in this context was relatively easy because there was no example of odd translations. Even though the participants’ translation for “是非(ze-hi)” was similar in trend to “適当 (teki-tou)”, guessing from the contextual clues seems to be more difficult. As Table 4.22 shows, whereas 56% of the CIMs translated the sentence “人工中絶(abortion) の是 非” successfully as “right or wrong of abortion” or “pros and cons of abortion”, only 9% of the EIMs did so. The rest of the EIMs (91%) could not write any translation for this word. Considering the fact that only 62% of the CADs and 40% of the EADs translated the word properly (see Table 4.23), it can be inferred that not many learners acquire this particular meaning of “是非(pros and cons)” even when they reach the advanced level. I would argue that the biggest factor contributing to “適当(appropriate)” being more successfully translated than “是非(pros and cons)”, despite their same semantic condition, is due to the respective frequency of the L2 input for these particular meanings of each word: learners are likely to encounter the former more often than the latter in L2.66 66 For instance, this particular meaning of “適当(teki-tou)” is often seen in the instructions of tests as in “ カッコに最も適当な言葉を入れなさい (Fill in the brackets with the most appropriate word) ”. In addition to the frequency of each item, their differences in register may also play a role: the former is relatively casual and the latter is more formal. 244 Unlike the other two target items in this category, there was no significant difference in the choice of English translation for “姿勢(shi-sei)” even between CIMs and EIMs. This result seems to be due to the relative ease of guessing the meaning of the target item in the given context: “コンピューターを使う時、座る姿勢が悪いと肩(shoulder) が痛くなりやすいです。(When using a computer, you are likely to get sore shoulder if your sitting posture is bad.)”. Hence, the ceiling effect might have been at work given that 70% of the EIMs guessed its meaning correctly (see Table 4.22). Whereas it is also possible that some participants had prior knowledge of the word, the data suggest that the ceiling effect has concealed the CIMs’ transfer either way. Although both hypothesis 2 and hypothesis 6 posited that positive transfer would occur where semantic ranges of the cognate pair are congruent, the roles of transfer are not necessarily the same in each hypothesis. As we have seen, where the semantic range of the cognate is wider in Chinese (hypothesis2), CBLs need to determine which meaning of the L1 counterpart can be used to interpret the cognate in L2 even when the meaning in question is shared across languages. Thus, CBLs are liable to misinterpret the meaning of the cognate in L2 when their erroneously inferred meanings happen to make sense in the context. To make matters worse, they may create erroneous IL as a result of their false success. On the other hand, where the semantic range of the cognate is wider in Japanese and its shared meaning is concerned (hypothesis 6), CBLs’ receptive transfer usually results in positive transfer since the whole semantic range of its L1 counterpart is included within the cognate in L2. However, the acquisition of multiple meanings within a partially 245 deceptive cognate is not necessarily easy even in the case of hypothesis 6. This is because the IL that CBLs create on the basis of their success in guessing the meaning of the cognates in L2 contexts do not always cover other semantic ranges that they actually contain such as “適当(irresponsible)” “是非(at any cost)”, “姿勢(attitude)”. That is, CBLs tend to create L1-based IL of (J>C) which only include the semantic ranges of its L1 counterpart, and lack L2-specific meanings if they guessed its meaning successfully in the context. The possibility for their restructuring of such L1-based IL into complete L2 knowledge would be dependent upon whether or not they encounter and notice the L2- specific meaning of the cognate. 4.4.2.4 Comprehension of (J>C) in Unshared Meaning Hypothesis 8 predicted that comprehension of the cognate with an unshared meaning, where the semantic range is wider in Japanese, would result in negative transfer or misinterpretation. This was confirmed in all of the three items by the results of the translation test. The hypothesis was tested using “呼吸(ko-kyuu)”, “依頼(i-rai)”and “夢中 (mu-chuu)”. Their semantic patterns can be seen in Figure 4.13. Table 4.24, Table 4.25, and Table 4.26 show the number and percentage of the participants who have translated the target items in different ways. Three chi-square tests were performed for each of the three items between (1) CBLs and EBLs, (2) CIMs and EIMs, and (3) CADs and EADs. As can be seen in Table 4.25, more CIMs chose certain English words more than EIMs in translating these three items, and chi-square tests show that these differences are significant. These results can be accounted for as a negative transfer triggered by cross-linguistic semantic discrepancy 246 of the cognate pair. In these examples, the different strategies used by the two different L1 groups in guessing the meaning of the target items can be clearly observed. L2:Japanese L1:Chinese L2:Japanese L1:Chinese L2:Japanese L1:Chinese hu-xi yi-lai mon-zhong ko-kyuu i-rai mu-chuu C+ 呼吸 (breath) 依頼 (to rely on) 夢中 (in a dream) (breath / bu-tiao (torelyon/ wei-tuo (in a dream / ru-mi harmony) request) be crazy about) <. * 4E (harmony) (request) (be crazy about) Figure 4.13: Semantic Patterns of the Target Items for Hypothesis 8 (Shaded parts indicate the meaning in question) Table 4.24: Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CBLs, EBLs and JNSs)67 CBLs % EBLs %JNSs% breath 20 50 0 0 0 0 ko- work together / cooperate 718 12 36 3 14 kyuu match / combination 5 12 4 13 14 67 呼吸 hard / tough 4 10 3 915 communicate 4 10 11 33 0 0 No translation given for the 00 3 9314 target item Total 40 100 33 100 21 100 χ(1) = 22.73, p < .001* CBLs % EBLs %JNSs% relied on / depended on 8 20 0 0 0 0 i-rai requested / asked 18 45 9 27 17 80 依頼 has / got 6 15 18 55 2 10 assigned / given 3 8 2 6210 called 2 4 1 300 No translation given for the 38 3 900 target item Total 40 100 33 100 21 100 χ(1) = 7.41, p < .01* CBLs % EBLs %JNSs% dreaming 22 55 9 27 0 0 mu- crazy 8 20 0 0943 chuu obsessed / infatuated 2 5 3 915 夢中 fall in love with 4 10 0 0838 No translation given for the 410 21 64 3 14 target item Total 40 100 33 100 21 100 67 The chi-square test was conducted only between CBLs and EBLs. 247 Shaded parts indicate possible negative transfer (*statistically significant) χ(1) = 5.69, p < .05* Table 4.25: Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CIMs and EIMs) CIMs % EIMs % breath 16 59 0 0 ko-kyuu work together / cooperate 415 8 35 呼吸 match / combination 1 4 3 13 hard / tough 2 7 3 13 communicate 4 15 9 39 No translation given for the 000 0 target item Total 27 100 23 100 χ(1) = 20.04, p < .001* CIMs % EIMs % relied on / depended on 7 26 0 0 i-rai requested / asked 10 37 2 9 依頼 has / got 4 15 16 70 assigned / given 2 7 2 9 called 1 4 1 3 No translation given for the 311 2 9 target item Total 27 100 23 100 χ(1) = 6.93, p < .01* CIMs % EIMs % dreaming 22 81 7 30 mu- crazy 1 5 0 0 chuu obsessed / infatuated 0 0 0 0 夢中 fall in love with 2 7 0 0 No translation given for the 2716 70 target item Total 27 100 23 100 Shaded parts indicate possible negative transfer (*statistically significant) χ(1) = 13.29, p < .001* Table 4.26: Number and Percentage of the Participants who have Used Different English Words in Translating the Target Items (CADs and EADs) CADs % EADs % breath 4 31 0 0 ko- work together / cooperate 323 4 40 kyuu match / combination 4 31 1 10 呼吸 hard / tough 2 15 0 0 communicate 0 0 2 20 No translation given for the 003 30 target item Total 13 100 10 100 χ(1) = 3.73, p > .05 CADs % EADs % relied on / depended on 1 8 0 0 i-rai requested / asked 8 61 7 70 依頼 has / got 2 15 2 20 assigned / given 1 8 0 0 called 1 8 0 0 248 No translation given for the 001 10 target item Total 13 100 10 100 χ(1) = .80, p > .05 CADs % EADs % dreaming 0 0 2 20 mu- crazy 7 55 0 0 chuu obsessed / infatuated 2 15 3 30 夢中 fall in love with 2 15 0 0 No translation given for the 215 5 50 target item Total 13 100 10 100 Shaded parts indicate possible negative transfer χ(1) = 2.85, p > .05 As Table 4.25 and Table 4.26 show, 59% of the CIMs and 31% of the CADs translated the phrase “呼吸が合う (work in harmony)” as something to do with “breath”, while no EBLs did so. It appears that EBLs mostly used contextual clues, while CBLs preferred using kanji or L1 lexical clues. I would argue that the reason why the English word “breath” was extensively used in CBLs’ translation is due to the relative difficulty of guessing the meaning of the phrase using context alone. Given that only 48% of the EIMs and 50% of the EADs gave appropriate translations such as “cooperate” or “combination”, it seems that many CBLs relied on their L1 knowledge to infer the meaning of the phrase due to a lack of sufficient information from the context. Another reason for CBLs’ extensive use of “breath” in their translation might be the false match of the meaning within the context. For example, many CBLs translated the phrase as “(In the doubles match of tennis) Mike and Tim were out of breath / couldn’t breathe / hardly breathe”, each of which make sense in the sentence or are grammatically correct at least. As can be seen in Table 25 and Table 26, 26% of the CIMs and 8% of the CADs translated “依頼(i-rai)” as “relied on” or “depended on”, whereas no EIMs and EADs 249 did so. Although this meaning is shared between Japanese and Chinese, it is not appropriate in the given context. Examples of such misinterpretation of the sentence “父はお客さんに急ぎ(urgent)の仕事を依頼された (My father was requested an urgent work by the client)” by CBLs include “the guest had an urgent work that’s been dependent on my father / a client is relying on my father for an urgent work”. In this example, the rate of erroneous translation induced by CBLs’ negative transfer has largely decreased with the development of their proficiency level (i.e. 26%→8%). One could argue that this is caused by their positive evidence of the L2 input: actual L2 knowledge constrained their negative transfer. As Table 4.25 shows, 81% of the CIMs translated “夢中(mu-chuu)” as “dreaming”, while only 30% of the EIMs did so. This difference was found to be significant through the results of the chi-square test; hence it could be argued that transfer is at work in CIMs’ attempts to guess the meaning. However, unlike the other two items where no EBLs gave CBLs’ possible L1-induced translations such as “breath” for “呼吸(ko- kyuu)” or “relied / dependent on” for “依頼(i-rai)”, 30% of the EIMs and 20% of the EADs also translated “夢中(mu-chuu)” as “dreaming” in the following context: “マイ クさんはキャシーさんに夢中です(Mike is crazy about Kathy.)”. A plausible explanation for this is that not only CBLs but also EBLs used kanji or lexical clues in guessing its meaning. As most of the target items in the translation task seem to be semantically less transparent68, I would argue that EBLs used contextual clues more often than kanji clues. However, “夢中(mu-chuu)” appears to be semantically more transparent and easier to be guessed if they had knowledge about the meaning of each constituent of the compound, 68 Semantically transparent kanji compounds are the words whose meaning can be easily guessed because the meaning of each element or character is clear. 250 i.e., “夢(dream)” and “中(inside)”. As a result, some of the EBLs also could have used kanji clues to guess the meaning of the compound. These results indicate that CBLs are prone to inferring the meaning of the cognate erroneously where the target meaning of (J>C) is not shared between Japanese and Chinese unless the context gives very powerful clues. This could be due to the nature of the semantic condition: CBLs would misinterpret the cognate if they relied on the meaning of its L1 counterpart. In addition, whereas guessing from the context does not always work (Nagy 1997; Schmitt 1997; Huckin & Coady 1999), L1 lexical clues are always available for CBLs as long as they find the cognates. That is, CBLs are more likely to utilise bottom-up approaches to interpret the cognate in Japanese than EBLs since they have much more potential to use lexical clues in comprehending such kanji compounds. Therefore, it seems to be difficult for CBLs to guess the meaning of the cognate in this semantic condition correctly and to discover its L2-specific meaning unless the context strongly denies their L1-based inferences. In other words, successful guessing and creation of L2-based IL would be largely dependent on the quality of the context in which the cognate is embedded and its input frequency: the more frequently CBLs encounter the L2-specific meaning of the cognate in different contexts, the more likely that they would be able to guess its meaning correctly and create L2-based IL of (J>C). 4.5 Summary of Results and Conclusion This study attempted to investigate the roles of transfer in CBLs’ use and acquisition of partially deceptive cognates using both productive and receptive measures. Overall 251 analyses of the data showed that there were distinct patterns of both positive and negative transfer depending upon the way in which the semantic ranges of the cognates overlap between Japanese and Chinese. As a consequence, the current study strongly supported the Semantic Transfer Hypothesis (Tanaka 1983; Takahashi 1985), which claims that language transfer is both pervasive and persistent in the domain of L2 lexico-semantics. In addition, the overall findings of the current study generally supported Stockwell et al.’s (1965) hierarchy of difficulty. However, the findings also suggested that convergence (coalesced) is not always easier than divergence (split), because CBLs tend to construct their either L1-based or L2-based IL on (J>C) which lacks another meaning of the partially deceptive cognates. The present study also supported the claims made by previous studies of various areas. First, it strengthened the theoretical and empirical research on transfer in production (Færch & Kasper 1986a, 1987, 1989; Poulisse 1999), which proposed that learners transfer their L1 lexical items both strategically (intentionally) and automatically (unconsciously). Second, it also supported the studies on transfer in reception (Færch & Kasper 1986b; Ringbom 1991, 1992), which claimed that receptive L1 transfer constitutes a subcategory of inferencing, and learners interpret incoming L2 data on the basis of L1 knowledge. Third, the current study further strengthened the claims made by Jiang’s (2000, 2002, 2004a) studies on L2 lexico-semantic development that even advanced L2 learners might use two L2 words that share the same translation in L1 interchangeably, despite their correct explicit knowledge of the two L2 words using more naturalistic tasks. Fourth, the present study also supported and complemented the descriptive study on Japanese-Chinese partially deceptive cognates (i.e. Matsushita 252 2001), which posited that the difficulty or problems for CBLs vary according to cross- linguistic semantic condition and task type. In addition to the findings with regard to the abovementioned studies, the current study also gave new insights into the variables that affect CBLs’ use and acquisition of Japanese-Chinese cognates by taking account of their interplay with transfer factors. Oral production of cognates was found to be influenced by multiple factors. When CBLs lacked prior knowledge of the L2 word to describe a certain concept, and they supposed that the cognate should exist for the concept, strategic transfer may be used to solve the communication problem. When CBLs had visual input of the cognate in an L2 context without phonetic information, they are more likely to infer its pronunciation by adapting the L1 sound into L2 (i.e. foreignizing). In the process of foreignizing, some cognates are easier than others to phonetically adjust due to different levels of phonological proximity or learners’ existing knowledge of the pronunciation for individual kanji characters. Even when CBLs have correct explicit knowledge of an L2 word regarding the cross-linguistic semantic boundaries with its L1 counterparts, automatic transfer may become apparent where one word in L1 is represented by multiple words in L2. The results of the translation test revealed that CBLs use both L1 lexical clues and contextual clues in guessing the meaning of novel cognates in L2. However, they relied more strongly on L1 lexical clues and misinterpreted the meaning of some cognates where the check up function of the context did not work very well. On the other hand, EBLs used contextual clues more extensively due to their inaccessibility to kanji or lexical clues. Therefore, CBLs’ success or failure of inferencing was largely dependent 253 on the semantic condition of the cognates (i.e. match or mismatch) and the quality of contextual clues. That is, CBLs can easily abandon their first hypothesis if the inferred meaning using L1 knowledge clearly disagrees with the context, but they are liable to misinterpret the meaning of partially deceptive cognates when their guessed meaning happens to make sense or agree with the context. The results of the current study generally suggest that CBLs tend to create IL regarding the meaning of L2 cognates on the basis of their L1 counterparts and L2 input. However, I proposed that the nature of the IL that CBLs create is fundamentally different according to the type of semantic condition and the relative frequency of the L2 input for particular meanings of the partially deceptive cognates. CBLs tend to create L1- based IL and use (J On the other hand, CBLs create either L1-based or L2-based IL for (J>C) depending upon the type of the L2 input they obtained more frequently. That is, if CBLs encountered the L1-L2 common meaning of the cognate more frequently, they are more likely to create L1-based IL. In this case, all the semantic contents that CBLs assume for their IL of (J>C) is included within the cognate in L2, and hence it will trigger positive transfer both in production and comprehension. When CBLs came across the L2-specific meaning of (J>C) more often, they are more likely to create its L2-based IL. In other words, the semantic range of (J>C) which is not covered by its L1 counterpart is less susceptible to transfer as long as they have encountered its L2-specific meaning, 254 and created L2-based IL. Yet, it would be considerably difficult for CBLs to produce such a semantic range of the cognate in L2 if they did not have prior knowledge of its L2-specific meaning. The same principle applies to the comprehension of the L2-specific meaning of (J>C). That is, CBLs should not have problems in understanding the L2-specific meaning of the cognate once they have gained its knowledge, while they may misinterpret it or have comprehension difficulty if they have not come across its L2-specific meaning previously. I proposed that the relative frequency of the input for a particular meaning within (J>C) plays a significant role in the process of constructing either L1-based or L2-based IL for CBLs. That is to say, the more frequently CBLs encounter the particular meaning of (J>C) in L2, the more likely that they create IL or a specific exemplar with regard to its particular meaning regardless of whether or not it is shared between Japanese and Chinese. Consequently, acquisition of the other meanings of (J>C) would be prolonged. The present study is valuable in that it attempted to examine CBLs’ use and acquisition of partially deceptive cognates in both productive and receptive measure using more realistic tasks based on the results of Study A. At the same time, however, the nature of the tasks might have brought about the limitations of the current study. That is, it was difficult to control many variables which might affect the results of each test. It cannot be denied that such factors as frequency of particular meanings in L2, existence of similar meanings in L2, register or genre of the target items were not controlled very well. These factors might have affected whether or not the hypotheses were supported to some extent, so prediction of transferability or ease of learning target items was not 255 solely based on the L1-L2 semantic patterns. As a result, the arguments might have included certain degree of conjecture. Therefore, it is crucial for future research to select or construct target items which control for various intervening factors. In addition, as the number of participant and target item examined for each hypothesis was not large, the findings of the current study cannot be easily generalised. Hence, research of a larger scale may be necessary in order to have more decisive conclusions. Another limitation of the study, like the case of Study A, is its inability to distinguish CBLs’ “existing knowledge” of the cognate in L2 and their “successful use” of the cognate in the form of strategic transfer or receptive transfer. However, I suggested that there are different levels of CBLs’ knowledge about partially deceptive cognates even when CBLs used them successfully in the L2 contexts. That is, CBLs may or may not have been aware of the existence of the cognate in L2. If CBLs had noticed the existence of the cognate, they may or may not have been familiar with its L2 pronunciation. Even when CBLs are familiar with the L2 phonology of the cognate, the semantic content that they assume it carries may only be based on its L1 counterpart or specific exemplar (i.e. one of the meanings of L2 word). Therefore, whereas the present study was unable to distinguish each level of CBLs’ knowledge about the target items, it did shed some light on the types of strategy that CBLs tend to use and the types of the IL that they are likely to construct. The following chapter will explore CBLs’ use and acquisition of Japanese words that correspond to each usage of two Chinese polysemous verbs “ (kai)” and “ (kan)” by taking account of their transferability based on CBLs’ perceived prototypicality. 256 257 CHAPTER FIVE TRANSFERABILITY OF CHINESE POLYSEMOUS VERBS “ (KAI)” & “ (KAN)” AND ACQUISITION OF JAPANESE CORRESPONDING WORDS BY CBLs (STUDY C)69 5.1 Introduction This study investigates the transferability of Chinese polysemous verbs “ (kai = to open)” and “ (kan = to see)” into Japanese corresponding words “開ける(akeru) / 開く (hiraku)” or “見る(miru)” both in CBLs’ receptive and productive use of Japanese. CBLs’ acquisition of the Japanese words that correspond to each usage of Chinese polysemy “ (kai)” and “ (kan)” will then be considered in relation to their transferability. These two verbs were selected as target items because I, as a Japanese language teacher, often encounter CBLs’ errors which are possibly caused by the negative transfer of these two Chinese verbs. For instance, CBLs’ such errors as “*本を見る (lit. *see a book)”, “*医者を見る(lit. *see a doctor)”, and “*電気を開ける (lit. *open the light)” can be explained by cross-linguistic contrastive studies because the translation equivalents in their L1 are consistent with these errors: “ (read a book; lit. see a book)”, “ B (see a doctor)”, and “ &(turn on the light; lit. open the light)”. At the same time, however, I have never encountered errors such as “*車を開く(lit. *open a car)” or “*冗談 を開く(lit. *open a joke)” despite the fact that their translation equivalents in Chinese 69 A significant portion of this study has been published in Japanese as “中国語母語話者による 日本語の語彙習得-プロトタイプ理論，言語転移理論の観点から- (Acquisition of Japanese Vocabulary by Chinese Native Speakers: Based on the Prototype and Transferability Theory)” in Acquisition of Japanese as a Second Language, No.8, Dec. 2005. 258 use the same verb “ (kai)”, i.e., “ (drive a car; lit. open a car)”, “ (tell a joke; lit. open a joke)”. If the semantic over-extension of the L1 items into L2 corresponding expressions is the only source of negative transfer, such errors should also appear as often as the former type of errors. Kellerman (1983) argues that prediction of transfer needs to be based not only on the structure of L1 and L2, but also on the learners’ perception of how the structures of L1 are perceived by the learner as being transferable or non-transferable. That is, if a certain L1 item is perceived as prototypical by the learner, it tends to facilitate transfer. On the other hand, if a certain L1 item is perceived as less prototypical, it is likely to inhibit transfer. Therefore, the present study explores the transferability of two Chinese polysemous verbs, “ (kai)” and “ (kan)”, into Japanese corresponding words on the basis of Kellerman’s (1978, 1979) findings on the transferability of polysemy. Subsequently, I will consider the relationship between the transferability of these two verbs and acquisition of their Japanese corresponding words by CBLs: how they interact with each other. In this next section, I will briefly review relevant aspects of prototype theory and SLA studies on polysemy again. As Shirai (1995) claims, it is important to consider the notion of “prototype” as well as interlingual lexical mapping in order to examine the L2 learners’ semantic development of polysemy. The basic concept of the prototype theory is that most natural categories are composed of a core member which is the clearest or best examples of the category and of other category members of decreasing similarity to the core (Rosch 1973). However, the way prototype theory is applied to research largely 259 varies among researchers even in the field of SLA (Hiki 1995; Shirai 1995; Sugaya 2004). In order to understand and compare the findings of previous prototype-related studies in SLA, it is crucial to clarify whose prototype is being discussed. Shirai (1995) suggested important notions of prototype: “L1 prototype” is the prototype that L2 learners have about their mother tongue, and “L2 prototype” is the prototype that either native speakers or L2 learners have about the TL. According to Shirai (1995), the former correlates with transferability of L1 items into L2 items, while the latter correlates with the learnability of L2 items. The current study focuses on the relationship between the L1 prototype and the transferability of polysemy. Another important distinction in comparing the results of prototype-related studies in SLA is how prototypicality is measured. “Theoretical prototype” is based on concreteness because it involves physical phenomenon, while “psychological prototype” is based on the awareness of a certain group of people (Tanaka 1990). The former is measured and determined by researchers deductively in their linguistic or typological analysis, but the latter usually measures target population’s awareness through judgement task or elicitation task, i.e., which items within the same category are psychologically more salient than others (Tanaka 1990). As L1 prototype involves how learners perceive their L1 items, prototypicality of each target item will be determined on the basis of psychological prototype in the present study. One significant aspect of the current study is that the target items include both potentially transferable (positive transfer) items and potentially non-transferable (negative transfer) items. Most of the target items in Kellerman’s (1978) transferability 260 experiment on “BREAK” were potentially transferable items and included few potentially non-transferable items. It is possible that the transferability of polysemy from learners’ L1 to L2 and acquisition of its corresponding words in L2 may differ between potentially transferable items and potentially non-transferable items, so use of both types of words may enable us to examine other factors that possibly constrain the occurrence of transfer. In order to explain the concept of potentially transferable items and potentially non-transferable items, let us take another Chinese polysemous verb “ (chi = to eat)” as an example. The Chinese polysemous verb “(chi = to eat)” is considered to have both prototypical and extended meanings or usages, but it is hard to find potentially transferable items for CBLs of Japanese other than the most prototypical usage as in the following: (1) )<17(おなかいっぱい食べた。/I have eaten and I am full.) (2).627?(薬を飲むのを忘れないように。/ Don’t forget to take the medicine.) (3),.3 。(この種の紙は墨汁を吸わない。/ This type of paper does not absorb ink.) (4) 05<A(僕の彼女はよくやきもちをやく。/ My girlfriend often shows her jealousy.) (5) -:7<"((米国で大変苦労した。/ Ihada lot of difficulty in the U.S.) The usage of “(chi = to eat)” in (1) is assumed to be the most basic and prototypical instance because it involves such an action of putting food into a mouth. In the case of (2), whereas the action involves putting something into a mouth, the object is not food. In the case of (3), the subject is not animate and involves neither food nor the action of putting something into a mouth. Further, the instances of (4) and (5) are rather idiomatic, and thus it is hard to imagine the Japanese primary equivalent “食べる(to eat)” from them. As can be seen, the Chinese verb “(chi = to eat)” can be replaced with the 261 Japanese primary equivalent “食べる(to eat)” only in the case of (1). In other words, (1) is the only case of potentially transferable item, and the others are potentially non- transferable items. As for the case of the Chinese polysemous verb “ (kai = to open)”, such instances as “ =(open the window)” and “# 35/(open /turntopage 35)” are likely to result in positive transfer if CBLs utilise an associative link between the Chinese “ (kai)” and its primary equivalent in Japanese “開ける(akeru) / 開く(hiraku)”70. This is because these concepts are expressed in Japanese using “開ける(akeru)” or “開く(hiraku)”(i.e. 窓 を開ける35ページを開く). On the other hand, if CBLs rely on such associative links, “ (drive a car; lit. open a car)” and “ (tell a joke; lit. open a joke)” will result in negative transfer since these concepts are not expressed in Japanese using “開ける (akeru)” or “開く(hiraku)”(i.e. 車を運転する 冗談を言う). The verb “ (kan = to see)” in contemporary Mandarin Chinese corresponds to “見る (miru = to see)” in Japanese. Ookouchi (1990) indicates that although idiomatic expressions often do not match across languages, Japanese and Chinese have many expressions which correspond with each other. Particularly, he uses the examples of the Chinese “ (kan)” and Japanese “見る(miru)” such as “ ;(look after a child; lit. see a child)” = “子供を見る(look after a child; lit. see a child)” and “99 (make an attempt; lit. do and see)” = やってみる (make an attempt; lit. do and see)”. However, there also 70 In the present study, both “開ける(akeru)” and “開く(hiraku)” are regarded as primary equivalents of Chinese polysemy “ (kai)” given that these two verbs have basically the same meaning, and are used interchangeably in many cases. 262 exist some potentially non-transferable items such as “ !(read a novel; lit. see a novel)” = “小説を読む (read a novel)” or “ B (see a doctor)” = “医者にいく (see a doctor; lit. go to a doctor)”. 5.2 Research Question and Hypotheses Based on Kellerman’s (1978, 1979) studies on transferability of Dutch polysemy into English or German counterparts, and the suggested limitations for them (Kellerman 1984, 1986; Meara 1984; Færch & Kasper 1987; Ellis 1994; Shirai 1995), the present study aims not only to reconfirm Kellerman’s results in different languages (i.e. Chinese and Japanese), but also to explore the unanswered question for two decades or more: how Kellerman’s findings on the transferability of polysemy can be applied to learners’ spontaneous L2 performance? The research question and hypotheses are: Research question: Would the learners’ perceptions of prototypicality of L1 items be able to predict not only what they believe to be transferable but also the transferability of L1 items in their L2 performance? Hypothesis 1: Learners’ perceptions regarding the prototypicality of polysemy within their L1 correlate well with transferability in their L2 reading comprehension . Hypothesis 2: Learners’ perceptions regarding the prototypicality of polysemy within their L1 correlate well with transferability in their L2 oral production. 263 In order to answer the research question and test the hypotheses put forward in this study, a survey on the prototypicality of Chinese polysemy and two types of Japanese vocabulary tests (i.e. an acceptability judgement test and an oral production test) were conducted. The instruments, procedures, participants, and results of each investigation will be described separately before a comprehensive discussion and conclusion is presented. 5.3 Survey on the Prototypicality of Chinese Polysemy (Investigation 1) Investigation 1 aims to reveal the prototypicality of ten usages of the Chinese polysemous verbs “ (kai)” and “ (kan)” that Chinese native speakers71 intuitively perceive (i.e. psychological prototype). 5.3.1 Instrument and Procedures In order to determine Chinese native speakers’ perceived prototypicality or psychological prototype, which deals with target populations’ awareness of particular linguistic items, a prototypicality judgement task was constructed based on Tanaka (1990). Although different methods have been used to determine psychological prototype such as free elicitation tasks, prototypicality judgement tasks or similarity / dissimilarity judgement tasks (Sugaya 2004), a prototypicality judgement task was chosen in the present study. This is because it allows us to investigate the relative distance between the most prototypical usage of the polysemy and each of its extended 71 The term “Chinese native speakers” will be used as opposed to “Chinese background learners (CBLs)” in the present study. Whereas the latter is assumed to be learners of Japanese, the former includes those who have no knowledge of Japanese. 264 usages gradationally. A prototypicality judgement task is also suitable for investigating the prototypicality of a number of usages within a polysemy. The task was devised to rate each of the ten usages of Chinese polysemy “ (kai)” and “ (kan)” using a nine-point scale ranging from “1” (least prototypical instance of the word) to “9” (most prototypical instance of the word). In order to familiarise participants with the concept of prototype, five instances of a Chinese polysemous verb “ (chi = to eat)” were included with the ratings as examples. The rating for each usage of “” was the mean value rated by five Chinese native speakers who had received detailed instruction on the concept of prototype by the author. The following is the English translation of the instruction and the examples of the rating (see Appendix D for the original Chinese versions of the survey). Instruction Please rate the prototypicality of the underlined word using “1” to “9”. If you think the underlined word in a sentence is the most prototypical usage of the word, please choose the highest number i.e., “9”. If you think the underlined word is a less prototypical usage of the word, please choose a lower number (1-3 for less typical instances, 7-9 for more typical instances). “Prototypical usage” indicates the most primitive or basic meaning of the word. For example, if you were asked for the most prototypical meaning or usage of a Chinese verb “(chi = eat)”, you are more likely to think of such an instance as “ )<17 (I have eaten and I am full.)” than “ -:7<"((Ihada lot of difficulty in 265 the U.S.)”. The former is more prototypical than the latter because the most prototypical or most basic meaning of the verb “(chi = eat)” is to put food into your mouth. Similarly, other words also have such prototypical usages. Please use your intuition to judge the prototypicality of the underlined words below (You can use the same number for the rating of different instances). Example )<17(I have eaten and I am full.) →9 .627?(Don’t forget to take the medicine.) →6 ,.3 。(This type of paper does not absorb ink.) →3 05<A(My girlfriend often shows her jealousy.) →1 -:7<"((Ihada lot of difficulty in the U.S.) →1 In order to collect a large amount of data from Chinese native speakers, a survey was constructed in the form of an email which included a brief objective of the study, a request for participation, the prototypicality judgement task, and questions regarding participants’ demographic information. The survey was written in Chinese, and two different versions were prepared using either traditional or simplified Chinese characters (Appendix E). No languages other than Chinese was used in the survey so as to exclude participants of non-Chinese native speakers or of Chinese native speakers who cannot read Chinese characters properly due to their lack of formal education in Chinese. 5.3.2 Participants The participants consisted of 72 Chinese native speakers. The survey was sent to approximately 170 Chinese native speakers by email through my acquaintances, and a total of 97 responses were received. Among the 97 responses, the study excluded 17 participants who had not completed high school in either Taiwan or China. Additionally, 266 eight of them were excluded due to lack of reliability, such as leaving many questions blank or rating “9” for all instances. Among the 72 Chinese native speakers, 43 were from Taiwan (18 living in Taiwan, 9 living in Japan, and 16 living in English speaking countries), and 29 were from People’s Republic of China (13 living in China, 6 living in Japan, and 10 living in English speaking countries). Given that all of the participants had completed at least high school either in Taiwan or China, and their place of residence at the time of the survey was diverse (six different countries), the samples were considered to be sufficiently random and reflect Chinese native speakers’ intuition about the prototypicality of the chosen verbs. 5.3.3 Results Table 5.1 and Table 5.2 show the mean value of prototypicality for each usage of the Chinese polysemous verbs “ (kai)” and “ (kan)” as rated by the 72 Chinese native speakers. In the present study, the prototypicality of each usage is determined by the mean value of the rating: the higher the mean value, the more prototypical the usage. Therefore, the results indicate that among the ten instances of “ (kai)”, “ =(open the window)” is the most prototypical, and “ (tell a joke; lit. open a joke)” is the least prototypical usage. Likewise, among the ten instances of “ (kan)”, “ %(watch TV)” is the most prototypical, and “ (try to taste; lit. eat and see)” is the least prototypical usage. 267 Table 5.1: The Mean Value of Nine-point Scale Prototypicality Judgement on Chinese Polysemy“ (kai)” Sentences given in the judgement task Mean Translations in Japanese 1 M 3.96 来週もまた会議を開開くので，必ず出席して下さ (We will hold a meeting next week again, so please い。 attend it.) 3.94 72 6 J I,3dZ5& W W< 張さんの将来の夢は自分の会社をつつくる こと (Mr. Zhang’s future dream is to establish his own です。 company.) 7 QFU*k6RW Wk 3.85 夏休みが終わって，来週から大学が始始まりま (After the summer holidays, university is starting す。 next week.) 8 .B(PtW W+ 3.39 呉さんは車を運運転するのが大好きです。 (Miss Wu loves driving acar.) 9 WW: :O-H 2.72 お湯が沸沸きました。コーヒーを飲みませんか？ (The water has boiled, let’s drink coffee.) 10 j :W W2D 2.36 冗談を言言わないで下さい。 (Please don’t tell a joke!) Table 5.2: The Mean Value of Nine-point Scale Prototypicality Judgement on Chinese Polysemy“ (kan)” Sentences given in the judgement task Mean Translations in Japanese 1 N "/X8 8cV 8.88 郭さんは部屋でテレビを見見ています。 (Mr. Guo is watching TV in the room.) 2 .3[#58 8 8.22 呉さんの趣味は小説を読読むことです。 (Miss Wu’s hobby is reading novels.) 3 % ,88' 6.92 今度また会会いに来ます。 (I will be back to see you next time. ) 4 $1mS3ao`8 88q b# 6.90 頭が痛いなら，医者に行行った方がいいですよ。 (If you have a headache, you should see a doctor.) 5 8 8+ 6.33 車に気気をつけて！ (Watch out for the car!) 6 )CRX)YYn)84 4.71 私が働いている間，母が子供を見見てくれます。 (When I’m working, my mother looks after the kid.) 7 Ar:8 803@# # 3.81 明日ビーチに行くかどうかは明日の天気を見見て (Let’s see tomorrow’s weather to decide whether 考えましょう。 we will go to the beach or not.) 8 )g788 E 2.75 彼は山田さんを見見下していると思います。 (I think he is looking down on Mr. Yamada.) 9 )88 ], 2.31 彼は来ないと見見ましたが。 (I guess he won’t come.) 10 , *4$+' 2.00 これが日本の納豆です。食べてみみて下さい。 (This is Japanese natto,pleasetryand taste it.) 72 Whereas “店を開く (open a shop)” is an acceptable and correct expression in Japanese, “*会社を開く (lit. *open a company)” is not. “*会社をつくる / 設立する(establish a company)” would be used instead. 268 5.4 Acceptability Judgement Test on Transferability of Chinese Polysemy (Investigation 2) Investigation 2 aims to examine the transferability of the Chinese polysemous verbs “ (kai)” and “ (kan)” into Japanese corresponding words in a comprehension task. The transferability of “ (kai)” and “ (kan)” was measured as the percentage of participants who accepted each Japanese sentence using “開ける(akeru) / 開く(hiraku)” or “見る (miru)”. 5.4.1 Instrument and Procedures An acceptability judgement test was conducted in order to examine whether there is a correlation between (1) the prototypicality of each usage of the Chinese polysemy “ (kai)” and “ (kan)” obtained in Investigation 1, and (2) CBLs’ acceptability of their corresponding Japanese expressions using “開ける(akeru) / 開く(hiraku)” or “見る (miru)”. The acceptability judgement test was used in Study A as well as in Study C. The task was to judge whether the usage of the underlined word in each sentence was correct or not, and to give a correct answer if it was judged to be erroneous. In constructing the acceptability judgement test, Kellerman (1978) and Shirai (1995) were consulted. In Kellerman (1978), participants were given 17 sentences in their L1 (Dutch) using the Dutch polysemy “BREKEN”, and were asked to judge whether they could be translated into English using the English “BREAK”. Therefore, whereas the judgement task examined participants’ knowledge or their intuition about whether Dutch “BREKEN” could be translated into English “BREAK”, it was not clear whether or not transfer would actually occur in learners’ L2 use. 269 On the other hand, Shirai (1995) asked Japanese ESL learners to judge the acceptability of 20 instances of the English polysemy “PUT” on a seven-point scale by giving them L2 (English) sentences which included “PUT”. The main purpose of Shirai (1995) was to investigate the interaction of L1 prototype and L2 prototype, and the formation of learners’ IL prototype rather than the transferability of “PUT” itself. Therefore, he used a seven-point scale to judge the acceptability of “PUT”, but the acceptability judgement test in the present study aims to examine whether or not receptive transfer will occur in L2 reading comprehension. For this reason, participants were asked to judge whether each instance of “開ける(akeru) / 開く(hiraku)” or “見る(miru)” was used correctly in L2 (Japanese) sentences. The sentences were written by the author using either “開ける(akeru) / 開く(hiraku)” or “見る(miru)” for both potentially transferable items (correct usage) and potentially non- transferable items (erroneous usage). As a result, six correct sentences and seven erroneous sentences were inserted into a total of 100 sentences with 28 sentences in Study A as well as with another 59 sentences (see Appendix A) for the purpose of other studies. As was mentioned in Study A, the English translation was given for each short sentence, and the words corresponding to the target items were also underlined in order to clarify the meaning of the words in question. The pronunciation of each kanji character was given in the form of hiragana syllabics in case some participants did not recognise or did not know how to sound out the kanji characters. The test was held in one of the classrooms at the University of New South Wales, and the researcher was in the room for the duration of the test. There was no time limit for the test. Please refer to Study A for the more detail. 270 5.4.2 Participants The participants of the acceptability judgement test consisted of 57 CBLs (16 CBGs, 25 CIMs, and 16 CADs), 44 EBLs (15 EBGs, 17 EIMs, and 12 EADs) and 21 JNSs. They were the same participants as for Study A in Chapter 3.73 The EBLs participated in this study as a control group in order to clearly detect the CBLs’ L1 influence. The JNSs also took part in the study in order for me to compare the performance of the learners with that of native speakers. 5.4.3 Results Table 5.3 and Table 5.4 indicate the results of the acceptability judgement test. The percentages in the tables show the transferability of the Chinese “ (kai)” and “ (kan)” in the task: the percentage of participants who accepted each usage of the Japanese “開 ける(akeru) / 開く(hiraku)” or “見る(miru)” regardless of whether they are used correctly or erroneously in the sentence. When participants denied the usage of “開ける (akeru) / 開く(hiraku)” or “見る(miru)”, the alternative word given by the participants was also examined to exclude the same verb, i.e., “開ける(akeru) / 開く(hiraku)” or “見 る(miru)”, in different tense-aspect or inflection. Only when participants gave a different verb from these three, was it counted as denial of these verbs. A chi-square test was conducted between CBLs and EBLs in order to compare the rate of acceptability of each sentence using “開ける(akeru) / 開く(hiraku)” or “見る(miru)”. As a result, significant differences were found in five items, i.e., *テレビを開ける 73 Although 18 EIMs participated in Study A, the data of one EIM was excluded in Study C due to the considerable number of blanks regarding the target items. 271 (turn on TV; lit.*open TV), パーティーを開く(give a party; lit. open aparty),*会社を開 く(establish a company; lit.*open a company), *冗談を開く(tell a joke; lit.*open a joke), and *医者を見る(*see a doctor). Thus, it is likely that CBLs’ receptive transfer is operative for at least these five items. As can be seen, these five items include both a potentially transferable item and potentially non-transferable items. Besides, some of them are relatively prototypical items, while others are less prototypical. Table 5.3: Results of the Acceptability Judgement Test (Transferability of “ (kai)”) Sentences given in the acceptability judgement test L1 Proto- The percentage that “開ける/ 開く” was typicality accepted regardless of correct or erroneous use (%) (Shaded are erroneous sentences, and others are ( ) indicates the number of participants correct sentences) CBLs EBLs Chi- JNSs N=57 N=44 square N=21 1 この部屋は暑いので窓を開けて下さい。 8.96 100 100 n.s. 100 (This room is hot, so please open the window.) (57) (44) (21) 2 *つまらないので，テレビを開けて下さい。 7.17 49 11 16.09** 0 (I am bored, so please turn on the TV.) (28) (5) (0) 3 教科書の35ページを開けて下さい。 7.00 95 95 n.s. 100 (Please open the textbook to page35.) (54) (42) (21) 4 来週の日曜日にパーティーを開くので，来て下さい。 4.25 68 36 10.29** 100 (I will give a party next Sunday, so please come.) (39) (16) (21) 5 *去年，弟とソフトウェアの会社を開きました。 (Last 3.94 75 55 4.85* 10 year, I established a software company with my brother.) (43) (24) (2) 6 *長い夏休みの後，３月には学校が開きます。 3.85 16 11 n.s. 0 (After the long summer holidays, school will start in March.) (9) (5) (0) 7 *私は毎日，車を開いて大学に来ます。 3.39 5 2 n.s. 0 (I drive to university everyday.) (3) (1) (0) 8 *冗談を開かないで下さい。 2.36 35 2 16.24** 0 (Please don’t tell a joke.) (20) (1) (0) (**p<.01, *p<.05) A Spearman’s rho was calculated between (1) the transferability of Chinese “ (kai)” and “ (kan)” into Japanese “開ける(akeru) / 開く(hiraku)” or “見る(miru)” by CBLs based on the results of the acceptability judgement test, and (2) the prototypicality of “ (kai)” and “ (kan)” obtained from the prototypicality judgement survey in Investigation 1. As a consequence, no significant correlation was found between the 272 results of the acceptability judgement test and the prototypicality of “ (kai)” and “ (kan)” (r=.542(N=13), p>.05). The reason for the inconsistency of the results with those of Kellerman’s (1978, 1979) studies may be due to the differences in the experimental designs or the skills that were elicited in the tests. As mentioned above, whereas Kellerman’s experiments did not test learners’ L1 transfer in their L2 use, the acceptability judgement test in the current study examined whether or not learners interpret the L2 words on the basis of their L1 knowledge. Table 5.4: Results of the Acceptability Judgement Test (Transferability of “ (kan)”) Sentences given in the acceptability judgement test L1 Proto- The percentage that “” was accepted typicality regardless of correct or erroneous use (%) (Shaded are erroneous sentences, and others are ( ) indicates the number of participants correct sentences) CBLs EBLs Chi- JNSs N=57 N=44 square N=21 1 田中さんはいつもテレビを見ています。 8.88 100 100 n.s. 100 (Mr. Tanaka is always watching TV.) (57) (44) (21) 2 *山下さんは小説を見るのが大好きです。 8.22 11 5 n.s. 5 (Mr. Yamashita loves reading novels.) (6) (2) (1) 3 *頭が痛いのなら，お医者さんを見た方がいいですよ。 6.90 81 39 18.72** 0 (If you have a headache, you should see a doctor.) (46) (0) 75 (17) 4 私が働いている間，母が子供を見てくれるんです。 4.71 37 30 n.s. 100 (While I am working, my mother looks after my child.) (21) (13) (21) 5 明日魚釣りに行くかどうかは天気をみて考えましょう。 3.81 79 91 n.s. 100 (Let’s see tomorrow’s weather to decide whether we will go (45) (40) (21) fishing or not.) (**p<.01) Apart from the L1 prototypicality of each usage of “ (kai)” and “ (kan)”, there might be other factors that constrained the occurrence of transfer. One possible factor is the CBLs’ existing knowledge of the L2 words that correspond to particular usages of “ 75 Although 15 EBLs (34%) corrected the underlined word into “*医者さんに会った方がいい”, such answers were regarded as correct judgement because they denied the use of “*医者さんを見た方がいい”. It seems that some EBLs tend to have such an associative link as “see = 会う(au)” rather than “see = 見る (miru)” even though “see a doctor” is the corresponding expression in English. 273 (kai)” and “ (kan)”. Assuming that generally learners are more likely to acquire the lexical items that are encountered frequently than infrequently (e.g. Ellis 2002; Gass & Mackay 2002), their existing knowledge of the L2 words will be argued in relation to the frequency of the L2 input for each target item. In order to gain some information about the frequency of the L2 input for each target item, four Japanese textbook series were analysed. The frequency of certain words appearing in textbooks and frequency of input learners receive is not necessarily the same. However, the learners’ TL input is more likely to be dependent on the textbooks that they use in a foreign language situation where TL input outside of the classroom is minimal than in a second language situation especially in earlier stages of language acquisition. Table 5.5 and Table 5.6 indicate the frequency of Japanese expressions that correspond to each usage of the Chinese “ (kai)” and “ (kan)” appearing in the four series of Japanese textbooks. As can be seen, one item appears more than 100 times, i.e., テレビ / 映画を見る(watch TV / movie), while some items do not appear at all in the eight volumes of the Japanese textbooks. The results of the acceptability judgement test and the investigation of the frequency for each target item in Japanese textbooks will be furtherdiscussedinsection5.6. 274 Table 5.5: Frequency of Target Items Appeared in Eight Volumes of Japanese Textbooks76 (Japanese Words that Correspond to Each Usage of “ (kai)”) Nakama1 Nakama2 Kimono1 Kimono2 Kimono3 Wakatta Textbooks Mirai5 Mirai6 Total Target items （窓 /ドアを）開ける / 開く 20001101216 Open (the window / door) （電化製品を）つける 0 0 0 0 0 1 0 10 11 Turn on (the electronic appliances) （本 / 雑誌を）開ける/開く 00000000 0 Open (books / magazines) パーティーを開く 00000000 0 Give /havea party (lit. open aparty) 会議を開く 00000003 3 Hold /havea meeting (lit. open a meeting) 会社をつくる/設立する 0 0 0 0 0 0 0 0 0 Establish a company ～が始まる 0 0 0 0 0 5 2 3 10 (Something) begins （乗り物を）運転する 0 0 0 0 4 2 0 10 16 Drive vehicles 冗談を言う 0 0 0 0 0 1 0 0 1 Tell a joke Shaded are potentially non-transferable items, and others are transferable items Table 5.6: Frequency of Target Items Appeared in Eight Volumes of Japanese Textbooks (Japanese Words that Correspond to Each Usage of “ (kan)”) Nakama1 Nakama2 Kimono1 Kimono2 Kimono3 Textbooks Wakatta Mirai5 Mirai6 Total Target items （テレビ / 映画を）見る 116 13130 155411 123 Watch (TV / movie) （本 / 雑誌を）読む 4 4 4 3 4 26 24 24 86 Read (books / magazines) 医者にいく 0 0 0 0 0 0 4 0 4 See a doctor (lit. go to a doctor) 医者にみてもらう 0 0 0 0 0 0 0 0 0 See a doctor (lit. seen by a doctor) 子供を見る 00000000 0 Look after achild(lit.seeachild) 天気を見る 00000000 0 See weather Shaded are potentially non-transferable items, and others are transferable items 76 These textbooks are used in secondary schools and universities in Australia. Generally, “Kimono1-3” is for year 8-10, “Mirai5-6” is for year 11-12, “Wakatta” is for preparation of the university entrance exam, and “Nakama1-2” is for university beginners. 275 5.5 Oral Production Test on Transferability of Chinese Polysemy (Investigation 3) Investigation 3 aims to examine the transferability of the Chinese polysemous verbs “ (kai)” and “ (kan)” into their Japanese corresponding words using a picture elicitation method. The transferability of “ (kai)” and “ (kan)” was measured as the percentage of participants who used the Japanese primary equivalents “開ける(akeru) / 開く (hiraku)” or “見る(miru)” in their utterances. 5.5.1 Instrument and Procedures Fourteen flashcards were constructed in order to examine whether there is a correlation between (1) the prototypicality of each usage of the Chinese polysemy “ (kai)” and “ (kan)” obtained in Investigation 1, and (2) the percentage of CBLs who used Japanese “開ける(akeru) / 開く(hiraku)” or “見る(miru)” in their utterances. Each of the 14 flashcards included a picture and a descriptive sentence with a blank to be filled in by the participants (see Appendix F). The 14 sentences were chosen from the 20 Chinese sentences in the prototypicality survey, and translated into Japanese with minor changes.77 The pictures were drawn to elicit the Japanese words that correspond to each usage of “ (kai)” and “ (kan)”. No English translation was given to the Japanese sentence on the flashcards. The task required participants to look at the picture and read aloud sentences on the flashcards, and to choose appropriate words or phrases to fill in the blanks. The 14 flashcards were shown randomly with another 22 flashcards which aimed to examine 77 Names of people were changed. Chinese names were used in the prototypicality judgement survey, while Japanese names were used in the oral production test because EBLs may be unfamiliar with the Chinese names. 276 other words for other studies including the 12 flashcards for Study B in Chapter 4. In order to narrow the gap between the test situation and that of the authentic communication, participants were asked to provide the answer as promptly as possible, and the next flashcard was shown when they could not respond within ten seconds. Many of the participants took both the oral production test and the acceptability judgement test which were discussed in Investigation 2. Since the two tests examined the same words, the participants were asked to sit for the oral production test first to minimise the task effects between two types of the tests when the participants took both tests. The test procedure was audio-taped for later analysis. Please refer to Study B for the more information on the oral production test. 5.5.2 Participants The participants of the oral production test consisted of 40 CBLs (27 CIMs and 13 CADs), 33 EBLs (23 EIMs and 10 EADs), and 21 JNSs. The same participants also participated in Study B (see Chapter 4). Many also took the acceptability judgement test mentioned in the previous section. EBLs participated in this study as a control group to clearly see the L1 influence on CBLs’ utterances. JNSs also took part in the study in order to compare the performance of learners with that of native speakers. 5.5.3 Results Table 5.7 and Table 5.8 indicate the results of the oral production test. The percentages in the tables show how frequently either “開ける(akeru) / 開く(hiraku)” or “見る(miru)” were chosen by participants. Even when participants produced errors in their utterances due to misuse of tense-aspect or inflection of the verb, or of postpositional particle, they were regarded as participants’ use of the target items as long as it was clear that they 277 had attempted to use one of the three verbs above. The results of the CBLs indicate that not all of the usages of the polysemy in L1 trigger transfer. No participants used the Japanese word “開ける(akeru) / 開く(hiraku)” in their utterances for three of the corresponding expressions of Chinese “ (kai)”, i.e., “ (school starts; lit. school opens)”, “ (drive a car; lit. open a car)”, and “ (tell a joke; lit. open a joke)”. Table 5.7: Results of the Oral Production Test (Transferability of “ (kai)”) Sentences given in the oral production test with pictures L1 Proto- The percentage that “開ける/ 開く”was (Inside of the brackets was blank in the test) typicality used (%) (Shaded are potentially non-transferable items, and ( ) indicates the number of participants CBLs others are potentially transferable items) EBLs Chi- JNSs N=40 N=33 square N=21 1 この部屋は暑いので窓を（開けて）下さい。 8.96 98 100 n.s. 100 (This room is hot, so please open the window.) (39) (33) (21) 2 暗い所で勉強すると目が悪くなるので，（電気をつけた 7.17 38 0 15.58** 0 ）ほうがいいですよ。(It is not good for your eyes if you (15) (0) (0) study in a dark place, so you should turn the light on.) 3 テキストの35ページを（開けて / 開いて）下さい。 7.00 100 100 n.s. 100 (Please open the textbook to page 35.) (40) (33) (21) 4 来週，家でパーティーを（開く）ので来て下さい。 4.25 8 9 n.s. 29 (We will give a party next week, so please come.) (3) (3) (6) 5 来週もまた会議を（開く）ので，必ず出席して下さい。 3.96 13 15 n.s. 71 (We will hold a meeting next week again, so please attend it.) (5) (5) (15) 6 中山さんの将来の夢は，自分の会社を（つくる）ことで 3.94 8 6 n.s. 5 す。(Mr. Nakayama’s future dream is to establish his own (3) (2) (1) company.) 7 夏休みが終わって，来週から大学が（始まります）。 3.85 0 0 n.s. 0 (After the summer holidays, university is starting next week.) (0) (0) (0) 8 上田さんは車を（運転する）のが大好きです。 3.39 0 0 n.s. 0 (Mr. Ueda loves driving a car.) (0) (0) (0) 9 冗談を（言わないで下さい）。 2.36 0 0 n.s. 0 (Please don’t tell a joke!) (0) (0) (0) (**p<.01) A chi-square test was carried out between the results of CBLs and EBLs in order to compare the rate of using “開ける(akeru) / 開く(hiraku)” or “見る(miru)” in oral production. As can be seen in Table 5.7 and Table 5.8, the CBLs produced two items, “*電気を開ける(turn on the light; lit.*open the light)” and “*医者を見る(see a doctor)”, more frequently than the EBLs did, and the chi-square tests indicate that these differences are significant. It should be noted that these two items were accepted by 278 more CBLs than EBLs in the acceptability judgement test, and such differences were also significant in the results of the chi-square test. Further, both of them are relatively high in their perceived prototypicality. Table 5.8: Results of the Oral Production Test (Transferability of “ (kan)”) Sentences given in the oral production test with pictures L1 Proto- The percentage that “見る” was used(%) (Inside of the brackets was blank in the test) typicality ( ) indicates the number of participants (Shaded are potentially non-transferable items, and CBLs EBLs Chi- JNSs N=40 others are potentially transferable items) N=33 square N=21 1 田中さんは部屋で（テレビを見ています）。 8.88 100 100 n.s. 100 (Mr. Tanaka is watching TV in the room.) (40) (33) (21) 2 山下さんの趣味は小説を（読む）ことです。 8.22 3 0 n.s. 0 (Mr. Yamashita’s hobby is reading novels.) (1) (0) (0) 3 カゼをひいたんですか？（医者にいった) ほうがいいで 6.90 38 0 12.57** 29 78 すよ。(Have you caught a cold? You should see adoctor!) (15) (0) (6) 4 私が会社で働いている間，母が子供を（みて）くれます 4.71 8 6 n.s. 71 。 (3) (2) (15) (While I’m working, my mother looks after the kid.) 5 明日ビーチに行くかどうかは，明日の天気（をみて）考 3.81 10 18 n.s. 76 (4) えましょう。(Let’s see tomorrow’s weather to decide (6) (16) whether we will go to the beach or not.) (**p<.01) A Spearman’s rho was calculated between (1) the transferability of “ (kai)” and “ (kan)” into “開ける(akeru) / 開く(hiraku)” or “見る(miru)” by the CBLs based on the results of the oral production test, and (2) the prototypicality of “ (kai)” and “ (kan)” obtained from the prototypicality judgement survey in Investigation 1. As a result, a strong correlation was found between the CBLs’ use of “開ける(akeru) / 開く(hiraku)” or “見る(miru)” and the prototypicality of “ (kai)” and “ (kan)” (r=.719(N=14), 78 Although the Japanese expressions that correspond to Chinese “ B (see a doctor)” include not only “医者にいく(lit. go to a doctor)” but also “医者にみてもらう(lit. seen by a doctor)”, 25 out of 40 CBLs produced the former and only one CBL produced the latter (see Table 5.10). As a result, 14 CBLs produced the erroneous utterance “*医者を見る(*see a doctor)”. The correct utterance “医者にみてもらう(lit. seen by a doctor)”, which was produced by 29% of the JNSs, also use “見る(see)” in a way. However, this expression was also regarded as a potentially non-transferable item because it needed to be accompanied by an additional grammatical function, i.e., receiving favourable actions. 279 p<.01). Accordingly, the results of the oral production test supported Kellerman’s (1978, 1979) results: transferability of polysemy correlates well with learners’ perceptions of prototypicality in their L1. In order to consider participants’ use of words other than the target items, and to ensure that potentially transferable items are consistently produced by JNSs in the oral production test, the CBLs’ utterances were compared with those of the JNSs and the EBLs. Table 5.9 and Table 5.10 show the participants’ use of words in the oral production test according to each L1 group and proficiency level. As can be seen in these two tables, not every JNS produced the same utterance which indicates that there are also variations or preferences for the use of words even among JNSs. For instance, only 29% of the JNSs produced “パーティーを開く(lit. open a party)”, and 67% and 5% of them produced “パーティーをする(lit. do a party)” and “パーティーをやる(lit. do a party)”, respectively. All of these three expressions mean “to give a party”, and are correct in Japanese. Thus, it seems that “パーティーを開く(lit. open a party)” is not the single best utterance in the given context although there were also some JNSs who produced this expression in preference to the other two expressions. 280 Table 5.9: Participants’ Use of Words in Oral Production Test (Japanese Words that Correspond to Each Usage of the Chinese “ (kai)”)80 Shaded rows indicate CBLs’ transferability: the percentage of CBLs who produced “開ける(akeru) / 開く (hiraku)” regardless of correct or erroneous utterances (positive transfer or negative transfer) 80 CBLs’ results on “*電気を開ける(turn on the light; lit.*open the light)” include three self-repairs. That is, two CIMs and one CAD produced correct utterance “電気をつける(turn on the light)” immediately after this erroneous utterance. Hence, their first (erroneous) utterances were recorded in the Table. 281 Table 5.10: Participants’ Use of Words in Oral Production Test (Japanese Words that Correspond to Each Usage of the Chinese “ (kan)”) Shaded rows indicate CBLs’ transferability: the percentage of CBLs who produced “見る(miru)” regardless of correct or erroneous utterance (positive transfer or negative transfer) 5.6 Discussion In the present study, transferability of the Chinese polysemous verbs “ (kai)” and “ (kan)” into the Japanese corresponding words was examined both in an acceptability judgement test and an oral production test. In addition, correlations between the L1 prototypicality of each usage of the polysemy judged by 72 Chinese native speakers and the percentage of CBLs who transferred the two Chinese polysemous verbs into their Japanese counterparts were calculated for both tests. Hypothesis 1, which predicted that learners’ perceptions regarding the prototypicality of polysemy within their L1 correlate well with transferability in their L2 reading comprehension, was not confirmed by the results of the acceptability judgement test. No significant correlation was found between 282 the prototypicality and the transferability of the Chinese “ (kai)” and “ (kan)” in the acceptability judgement test which investigated the occurrence of receptive transfer. In contrast, hypothesis 2, which predicted that learners’ perceptions regarding the prototypicality of polysemy within their L1 correlate well with transferability in their L2 oral production, was confirmed by the results of the oral production test. The results of this test showed a strong correlation between these two, and hence supported Kellerman’s (1978, 1979) findings: the higher the learners’ perceived prototypicality of L1 item is, the more transfer tends to occur. Accordingly, Kellerman’s framework on the transferability of polysemy, which has not explored its transferability in learners’ L2 use, seems to have predictive power for the occurrence of transfer in production. As far as receptive transfer is concerned, however, it appears that such prediction cannot be made. Although the transferability of “ (kai)” and “ (kan)” in the oral production test showed a strong correlation with their L1 prototypicality as a whole, there were also a few individual items whose percentage of transfer could not be explained solely by the prototypicality of the L1 items. In this section, the results of the acceptability judgement test and the oral production test will be compared in order to discuss the possible factors that constrained the occurrence of transfer in both tests in relation to (1) L1 prototypicality, (2) potential of positive transfer (i.e. transferable or non-transferable items), and (3) learners’ existing L2 knowledge. CBLs’ acquisition of Japanese words that correspond to each usage of Chinese “ (kai)” will be discussed first, and followed by those of “ (kan)”. 283 5.6.1 Transferability of Chinese Polysemous Verb “ (kai)” and Acquisition of its Corresponding Words in Japanese We will examine cases in four groups. Figure 5.1 indicates the four groups of L2 words categorised in terms of L1 prototypicality and potential of positive transfer. The first group consisted of the two items, “窓を開ける(open the window)” and “35ページを開ける (open / turn to page 35)”, taken from the Japanese words that correspond to the Chinese polysemy “ (kai)”. In this group, there was nearly 100% of transferability both in the acceptability judgement test and the oral production test. These two expressions are the corresponding words of two usages of the Chinese polysemy “ (kai)” which 72 Chinese native speakers rated as highly prototypical, and, at the same time, are potentially transferable from L1. However, it is not possible to conclude to what extent the CBLs’ transfer is involved in the results of two tests, because the EBLs also showed similar acceptability and successful oral production of these usages. Since almost all of the participants judged and produced these items correctly, some participants of each L1 group might have had prior knowledge of word usage. Frequency of the L2 input is one possible reason that these participants had such existing knowledge of these items. According to Table 5.5, “(窓 / ドアを) 開ける(open the window / door)” appears 16 times, and “(本 / 雑誌を)開ける(open books / magazines)” does not appear in the eight volumes of the Japanese textbooks. Given that these expressions do not necessarily appear very frequently in the Japanese textbooks 284 analysed, frequency of learners’ L2 input alone cannot satisfactorily account for such results.81 Prototypical in L1 Equivalent Words Group 2 Group 1 e.g. e.g. “テレビをつける “窓を開ける (turn on TV)” “医者に (open the window)” “ いく テレビを見る (see a doctor ; (watch TV)” Potentially lit. go to a doctor)” Potentially Non-transferable Transferable from L1 from L1 Group 4 Group 3 e.g. e.g. “冗談を言う (tell a “パーティーを開く joke)” (give aparty; lit. open aparty)”“子 供を見る (look after achild)” Non-prototypical in L1 Equivalent Words Figure 5.1: Four Groups of Target Items Categorised in Terms of L1 Prototypicality and Potential of Positive Transfer I would also argue that another possible reason why almost all of the participants correctly produced these items is their intrinsic ease of acquisition. That is, highly prototypical items in one language are usually highly prototypical in other languages (Sugaya 2004), so they are often universally easy to learn. In order to confirm this, I have informally asked 20 JNSs to rate the prototypicality of eight usages of a Japanese polysemous verb “(開ける(to open)” in a nine-point scale. As predicted, “窓を開ける (open the window)” and “35ページを開ける(open / turn to page 35)” were rated to be the 81 However, it is also possible that these expressions are frequently used by teachers as classroom instruction. 285 most prototypical usages out of eight usages of the verb. Therefore, intrinsic ease of acquisition for these items can be explained by L2 prototype factors. The second group involves potentially non-transferable L2 words whose L1 equivalents are perceived as prototypical. These items are found to be susceptible to negative transfer or over-extension both in receptive and productive tasks. This can be seen clearly by the results of the acceptability judgement test where 49% of the CBLs accepted the erroneous expression “*テレビを開ける(*open the TV)” in the sense of “turn on the TV”, and resulted in wrong judgement. The results of the chi-square test showed that significantly more CBLs accepted this erroneous use of “開ける(open)” than EBLs did (χ(1) = 16.09, p < .01). Moreover, 38% of the CBLs, a group which only included intermediate and advanced learners, produced the erroneous utterance “* 電気を開ける (*open the light)”in the sense of “turn on the light”, which no EBLs produced (χ(1) = 15.58, p < .01). It is interesting to note that three CBLs produced the correct expression, “電気をつける (turn on the light)” immediately after their erroneous utterance “*電気を開ける(*open the light)”. Such self-repair indicates that some CBLs produce negative transfer errors in spite of their correct knowledge of the L2 words. Therefore, one can argue that for those L2 words whose L1 equivalents are perceived as prototypical but are non- transferable from L1, negative transfer is at work not only when they lack the knowledge of the L2 word, but also when they have correct knowledge of the L2 expression in the form of automatic transfer. 286 The third group includes potentially transferable items whose L1 equivalents are perceived as non-prototypical. In this group, the results of the two tests seem to diverge. Since 68% of the CBLs correctly judged the correct expression “パーティーを開く (give a party; lit. open a party)” in the acceptability judgement test which was significantly higher than EBLs, the CBLs’ comprehension of this expression is likely to be facilitated by their positive transfer. In contrast, only 8% of the CBLs produced “パー ティーを開く (give a party; lit. open a party)” in the oral production test despite the fact that Chinese corresponding expression “ >8(give a party; lit. open a party)”could be positively transferred. “会議を開く(hold a meeting)” was included as a target item in the oral production test but not in the acceptability judgement test. Thus, the comparison of the transferability between the two tasks cannot be directly made. However, only 13% of the CBLs produced “会議を開く(hold a meeting; lit. open a meeting)” in the oral production test. As both “パーティーを開く(give a party; lit. open a party)” and “会議を開く(hold a meeting; lit. open a meeting)” are not the single best utterance in each context in the oral production test, i.e., “パーティーをする / やる(lit. do a party)” and “会議をする(lit. do a meeting)” are also appropriate, the CBLs’ rate of production for each target item was compared with that of JNSs. Given that 71% of the JNSs produced “会議を開く (hold a meeting; lit. open a meeting)” in the same oral production test, this expression seems to be underproduced by the CBLs and advanced learners (see Table 5.9). As for “パーティーを開く(give a party; lit. open a party)”, as only 29% of the JNSs produced this expression (see Table 5.9), it is inconclusive as to whether this item was 287 underproduced by CBLs or the result simply reflects frequency of input assuming that the JNSs’ use of these words may reflect the frequency of learners’ input to a certain extent. Yet, Table 5.9 shows that, like the case of “会議を開く(hold a meeting; lit. open a meeting)”, only advanced learners of both CBLs and EBLs produced “パーティーを開く (give a party; lit. open a party)”. This may imply that these CADs’ utterances are more likely to be produced by their existing L2 knowledge rather than the transfer based on their intuition about transferability triggered by the L1 prototypicality. This is because learners tend to rely on their L1 knowledge when L2 knowledge is not sufficient (Corder 1983). In other words, less proficient learners are more likely to transfer these items unless transfer is suppressed by their perception that these items are not good candidates for transfer. Therefore, the L2 words categorised in this group seem to be susceptible to under-extension (i.e. even though the item can be transferred from L1 equivalent, learners do not exploit the benefit of transfer) in production tasks if learners have not encountered the positive evidence in L2. On the other hand, considering the fact that, in the acceptability judgement test, significantly more CBLs accepted the correct expression “パーティーを開く(give aparty; lit. open a party)” than EBLs, it is likely that CBLs used L1 knowledge in their comprehension of this item. Therefore, it seems that transfer is inhibited in the CBLs’ production of these items probably due to their perception that such usages are L1- specific. However, receptive transfer is still operative, and hence these items tend to result in positive transfer even where their L1 corresponding words are perceived as less prototypical. 288 The final group is potentially non-transferable items whose L1 equivalents are perceived as non-prototypical. In this group, no CBLs used “開ける開く(open)” in their utterances except 8% of the CBLs produced “*会社を開く(*open a company)” in the sense of “establish a company”. The results of the acceptability judgement test showed that two erroneous expressions, “*車を開く(drive a car; lit.*open a car)” and “*学校が開 く(school starts; lit.*school opens)”, were accepted (judged wrongly) only by 5% and 16% of the CBLs, respectively. However, 75% and 35% of the CBLs accepted (judged wrongly) the other two erroneous expressions in this group “*会社を開く(establish a company; lit.*open a company)” and “*冗談を開く(tell a joke; lit.*open a joke)”, respectively. Further, significantly more CBLs accepted these two erroneous expressions than EBLs did. Thus, like the case of Group 3, transfer is likely to be suppressed in the CBLs’ productive use of these items, while receptive transfer seems to be at work for some of the corresponding expressions of Chinese “ (kai)” even where they are perceived as less prototypical in L1. A question still remains, however. Among the four items which are similarly non- transferable from L1 and are perceived as non-prototypical in their L1 equivalents, most CBLs denied (judged correctly) two erroneous expressions “*車を開く(drive a car; lit.*open a car)” and “*学校が開く(school starts; lit.*school opens)”. On the other hand, many CBLs accepted (judged wrongly) another two erroneous expressions “*会社を開く (establish a company; lit.*open a company)” and “*冗談を開く(tell a joke; lit.*open a 289 joke)”. I would argue that the reason for the different rates of transfer for these items is due to CBLs’ varied degree of existing knowledge of each expression in L2. As can be seen in Table 5.5, “運転する(drive)” and “始まる(begin)” were used 16 times and 10 times in the eight volumes of Japanese textbooks respectively, while “会社をつくる/ 設 立する (establish a company)” did not appear at all, and “冗談を言う(tell a joke)” appeared only once. It is entirely possible that many CBLs had prior knowledge of the L2 words regarding the concepts of “to drive” and “to begin”, and hence they denied such erroneous expressions as“*車を開く(*open a car)” and “*学校が開く(*school opens)” relatively easily. Indeed, all of the correct judgements were accompanied by such words as “運転 する(drive)” and “始まる(begin) / 始める(begin)82” in the acceptability judgement test. In contrast, not many CBLs had prior knowledge of the L2 words for the concepts of “to establish a company” and “to tell a joke”, so they tended to interpret “*会社を開く (*open a company)” and “*冗談を開く(*open a joke)” on the basis of their L1 knowledge, and accepted (judged wrongly) these erroneous expressions in the acceptability judgement test. The data would suggest that learners’ existing knowledge of the L2 words plays certain roles in constraining the occurrence of transfer. 5.6.2 Transferability of Chinese Polysemous Verb “ (kan)” and Acquisition of its Corresponding Words in Japanese 82 The Japanese verb “始まる(begin)” is an intransitive verb, while “始める(begin)” is a transitive verb, and the latter was not appropriate in the context of the acceptability judgement test. Whereas some participants gave the word “始める(begin)” erroneously as an alternative to “開く(open)”, it was regarded as correct judgement because they denied the usage of “開く(open)” in the judgement test. 290 The CBLs’ acquisition of Japanese words that correspond to each usage of Chinese “ (kan)” can be explained in a similar way to those of “ (kai)”. The first group includes a potentially transferable item whose L1 corresponding word is perceived as prototypical, i.e., “テレビを見る(watch TV)”. This item was transferred and correctly answered by 100% of the CBLs in both the acceptability judgement and the oral production tests. This result suggests that acquisition of this type of item, like the case of “窓を開ける (open the window)” and “35ページを開ける (open / turn to page 35)”, is relatively easy. As Table 5.6 shows, “(テレビ / 映画を)見る (watch TV / movie)” appears 123 times in the eight volumes of Japanese textbooks, so the relative ease of acquisition for this item can be explained by both frequency factors and L2 prototype factors. The second group consisted of potentially non-transferable items whose L1 equivalents are perceived as prototypical. The results of the acceptability judgement test showed that 81% of the CBLs accepted (judged wrongly) the erroneous expression “*医者を見る (*see a doctor)”, while only 11% of the CBLs did so for “*小説を見る (read a novel; lit.*see a novel)”. Similarly, in the oral production test, 35% of the CBLs produced the erroneous utterance “*医者を見る(*see a doctor)”, while only 3% of the CBLs produced an error such as “*小説を見る(lit.*see a novel)” in the sense of “read a novel”.83 As can be seen in Table 5.4 and Table 5.8, more CBLs accepted the erroneous expression “*医 者を見る(*see a doctor)” in the acceptability judgement test, and produced the same expression in the oral production test, than EBLs did. These differences are also found 83 Chinese words that correspond to Japanese “読む(read)” include not only “ (kan)” but also “@(du)”. While the former only means “read silently”, the latter means both “read silently” and “read aloud”. However, “ (kan)” is more commonly used in referring to “silent reading” which was the intended concept in the oral production test. Thus, it is unlikely that CBLs produced “小説を読む(read novels)” successfully as a result of positive transfer from “@!(read novels)”. 291 to be significant in the results of the chi-square tests (i.e. χ(1) = 18.72, p < .01; χ(1) = 12.57, p < .01). However, no significant difference was found between these two groups in their judgement and production of “*小説を見る(read a novel; lit.*see a novel)” in the two tests. These results would suggest that, whereas CBLs’ transfer is operative both in their comprehension and production of the concept about “see a doctor”, their use of transfer is minimal when the concept of “read a novel” is involved. Such results are in conflict with Kellerman’s (1978, 1979) findings, i.e., the higher the learners’ perceived L1 prototypicality, the more transfer tends to occur. That is to say, although the L1 prototypicality of the former (i.e. “ B (see a doctor)”= 6.90) is lower than that of the latter (i.e.“ !(read a novel; lit. see a novel)”= 8.22), more transfer occurred in the former. It is possible that the CBLs’ existing knowledge about the L2 word constrained the occurrence of their transfer. As can be seen in Table 5.6, “医者にい く(see a doctor; lit. go to a doctor)” was used four times in the eight volumes of Japanese textbooks, while “(本 / 雑誌を)読む(read books / magazines)” was used as many as 86 times. Thus, one would assume that most CBLs had existing knowledge of theL2word“読む(read)” due to abundant input. Accordingly, it can be inferred that even though both of these L2 expressions are susceptible to negative transfer or over-extension due to the higher prototypicality of their L1 equivalents, the CBLs’ errors or erroneous IL, such as “*小説を見る(*see a novel)” in the sense of “read a novel”, are restructured in relatively earlier stages along with their acquisition of the correct L2 word, i.e.,“読む(read)”. This is parallel with Kellerman’s (1979, 1984, 1995) claim that learners could benefit from positive 292 evidence of the transferability for the particular items if they were exposed to L2 input. Again, it points to the strong possibility that the CBLs’ actual L2 knowledge constrained the occurrence of transfer. Whereas no self-repair was detected for these items, one could argue that, like in the case of “*電気を開ける(*open the light)”, some of the CBLs’ erroneous utterances were the results of automatic transfer triggered by their relatively high L1 prototypicality in spite of their correct L2 knowledge. The third group consisted of potentially transferable L2 words whose L1 corresponding words are perceived as non-prototypical. The results of the oral production test indicate that only 8% and 10% of the CBLs produced “子供を見る(look after a child; lit. see a child)” and “天気を見る(see the weather)”, respectively. Since 71% and 76% of the JNSs produced these expressions in the same test, like in the case of “会議を開く(hold a meeting)”, the data would suggest that the CBLs’ transfer in production was inhibited for these items, presumably because they were perceived as L1-specific. On the other hand, “子供を見る(look after a child; lit. see a child)” and “天気を見る(see the weather)” were accepted (judged correctly) by 37% and 79% of the CBLs in the acceptability judgement test, respectively. Unlikethecaseof“パーティーを開く(give a party)”, however, the CBLs’ judgements of these two items did not significantly differ from those of EBLs. The main reason for this may be that “パーティーを開く(give a party; lit. open a party)” is a potentially non- transferable item for EBLs, while “子供を見る(look after a child)” and “天気を見る(see the weather)” are transferable items for EBLs due to the fact that their English translation equivalents include “see” or “look”. In other words, although CBLs’ transfer was at work in the judgement of “子供を見る(look after a child)” and “天気を見る(see the 293 weather)”, the EBLs’ positive transfer might have obscured the differences in the judgement of these two items between CBLs and EBLs. If this is the case, then again, the L2 words categorised in this group are susceptible to under-extension as far as oral production is concerned, but transfer is still at work, and thus tends to result in positive transfer in reading comprehension of these items. Among the Japanese words that correspond to each usage of the Chinese polysemy “ (kan)”, no item was examined for potentially non-transferable L2 words whose L1 equivalents were perceived as non-prototypical. Therefore, the relationship between the transferability of “ (kan)” and acquisition of its Japanese corresponding words by the CBLs in this group is unknown. However, it can be inferred that the results would be similar to those of “ (kai)” if the above discussions are valid. 5.7 Summary of Results and Conclusion This study has examined the transferability of two Chinese polysemous verbs “ (kai)” and “ (kan)” both productively and receptively in order to understand its relationship with CBLs’ acquisition of the corresponding words in Japanese. The single most important finding of the current study is the impact of task type with regard to the transferability of polysemy: prototypicality correlated well with transferability in oral production, while this was not the case in judgement task. The overall results of the present study suggest that whereas those items which are perceived as prototypical in the learners’ L1 tend to facilitate transfer regardless of their potential transferability and task type, potentially non-transferable items are, more often 294 than not, restructured as a result of learners’ gaining correct L2 knowledge probably through the L2 input. Consequently, negative transfer tends to decrease as the CBLs’ proficiency level increases although negative transfer may persist in the form of automatic transfer even after the learners have grasped the correct knowledge about the L2 word. As for those items which are perceived as non-prototypical in the learners’ L1, however, transfer tends to be inhibited in production, while transfer is still operative in reception even for the same usage of the same word. It is also assumed that CBLs’ under-extension of potentially transferable items in their production and over-extension of non-transferable items in reception will be restructured as they are exposed to positive evidence of the L2 input. The main reason for the considerable gap in the rate of transfer between the acceptability judgement test and the oral production test, despite the fact that the same usages of the same words were compared, could be that completely different types of transfer were elicited in each task. That is, each test in the present study examined potentially different skills. As no previous study has explored the transferability of polysemy based on learners’ L2 oral production or L2 reading comprehension, it would be natural even if we had different results from previous studies. Yet, the results of the oral production test supported Kellerman’s (1978, 1979) results: perceived prototypicality of a particular L1 item correlates well with its transferability. On the contrary, CBLs tended to rely on their L1 knowledge to interpret novel L2 items even when their perceived L1 prototypicality was low, and thus no such correlation was found in the results of the acceptability judgement test. 295 It seems that Kellerman’s (1978, 1979) findings are valid as a general and overarching principle in predicting the transferability of polysemy. However, it may also be necessary to consider the subordinate conditions of this general principle according to task type. They are: (1) where perceived prototypicality of a particular L1 item is relatively high, learners tend to be willing to transfer it to L2 both in production and comprehension tasks, and (2) where perceived prototypicality of a particular L1 item is relatively low, learners tend to be reluctant to transfer it to L2 in production tasks, while they are still willing to use L1 knowledge in comprehension tasks. The current study is worthwhile not only because it attempts to verify the universality of Kellerman’s (1978, 1979) transferability principle using non-European languages, but also because it investigates the transferability of polysemy in learners’ L2 oral production and L2 reading comprehension, neither of which has been previously explored. The present study is also valuable in that it sheds some light on the different roles of L1 prototypicality in the transferability of polysemy depending upon the task type, and on the important roles of learners’ existing L2 knowledge in the occurrence of transfer. Although the present study has attempted to complement previous studies and overcome some of their limitations, it is not without problems. One possible criticism of the current study is the method of eliciting prototypicality. That is, although the original studies on transferability of polysemy used a similarity sorting task where the term “prototypicality” was not mentioned to participants (Kellerman 1979), the current study adopted more direct rating. It is not known whether or not these two different methods 296 are measuring the same intuitions of learners, but direct rating is one of the popular methods of eliciting prototypicality in other areas of study (e.g. cognitive-semantics). Whereas the direct rating may disregard the distinction between the concepts of abstract and concrete, similarity sorting also has shortcomings. Determining the rank order using similarity sorting task may not be very practical with long lists of sentences, and it would be prone to order effect. Given the nature of the task, participants may feel frustrated or fatigued, and could produce unreliable data. Balancing these pros and cons, the current study has adopted the direct rating as a way of eliciting prototypicality with regard to Kellerman’s (1979) transferability framework. Another limitation of the current study is that the boundary between prototypical items and non-prototypical items could be seen to be somewhat arbitrary. However, this is largely due to the nature of the psycholinguistic rating used in the prototypicality judgement task. As Shirai (1990) mentioned, the rating must be based on continuous rather than dichotomous measures, and therefore deciding on the cut-off point is often difficult. I would argue that the boundary drawn in the current study is reasonable in that the prototypical items ranged from 8.96 to 6.90 in their mean value of the rated scores, while non-prototypical items ranged from 4.71 to 2.36. However, there was no definite reason that the border must be drawn there. Finally, like Kellerman’s (1979) original transferability study, the current study was unable to differentiate transferability and learners’ knowledge derived from L2 input, particularly when positive transfer was involved. In order to account for the decrease of negative transfer errors, learners’ existing knowledge of the L2 words was argued on 297 the basis of how frequently they appeared in eight volumes of Japanese textbooks. Although the L2 lexical knowledge for beginning or early intermediate learners, particularly in a foreign language environment, is likely to be dependent on the textbooks that they have used, it cannot be equated with the frequency with which the L2 words appeared in the textbooks because other factors may also affect acquisition of these words. Future studies need to explore the method that can distinguish transferability and learners’ knowledge derived from L2 input by considering the learners of different proficiency levels including true beginners, and their development of sensitivity on transferability. 298 CHAPTER SIX CONCLUSION 6.1 Summary of Findings Motivated by the reported pervasiveness and persistency of interlingual semantic transfer, this thesis explored CBLs’ lexico-semantic development focusing on their acquisition of Japanese-Chinese cognates and Japanese kanji words that correspond to each usage of two Chinese polysemy “¤(kai)” and “¯(kan)”. The major objective of this research was to investigate the roles of transfer and the difficulty in CBLs’ use and acquisition of Japanese kanji compounds and kanji words (i.e. under what condition and in what ways do CBLs transfer Chinese words into Japanese counterparts). In addition to this objective, I attempted to examine the interaction between transfer factors and the variables of task type, L2 proficiency, frequency of the L2 input, context the word is embedded, and prototypicality which affect CBLs’ use and acquisition of these Japanese words. The thesis made many meaningful findings that advance our understanding of cross- linguistic lexical mapping. Lexico-semantic transfer and its associated difficulties arise from (1) interlingual semantic conditions, (2) type of cognate, (3) proficiency levels, and (4) lexical prototypicality. The thesis addressed varied developmental patterns of cognates across different L1 groups and proficiency levels in Study A, differing potentials of both positive and negative transfer according to the way meanings of cognates overlap across languages in Study B, and different rates of transfer depending upon the prototypicality of the word and task type in Study C. 299 The results of Study A showed that CBLs’ developmental patterns regarding the knowledge of meaning(s) of Japanese kanji compounds vary according to the extent to which their orthography and semantic ranges correspond between Japanese and Chinese. Among the four types of kanji compounds, CBLs’ use and acquisition of (S) or “good cognate” seem to be facilitated due to their consistent positive transfer. Developmental patterns of (N) or “non-cognate kanji compounds” for CBLs may differ depending upon the comprehensibility of their meaning using their knowledge of Chinese. That is, transfer is less likely to occur if it is perceived as non-comprehensible for CBLs, while either positive or negative transfer may influence their use and acquisition of non-cognate kanji compound if it is perceived as comprehensible on the basis of their L1 knowledge. CBLs’ use and acquisition of (D) or “totally deceptive cognate” were found to be susceptible to negative transfer in earlier stages of their development particularly where contextual clues are not strong enough to let CBLs abandon their L1-based hypothesis about its meaning. Yet, the results of the acceptability judgement test showed that most advanced learners had the correct L2 knowledge regarding the meaning of totally deceptive cognates. Unlike other types of Japanese kanji compounds, acquisition of (O) or “partially deceptive cognate” was found to be far more difficult for CBLs in that many of the advanced learners lack the knowledge about some of the meanings of this type of cognate. Based on the results of Study A, CBLs’ use and acquisition of partially deceptive cognates (O) were further investigated using both productive and receptive measures. The results of Study B indicated that CBLs transfer their L1 knowledge both consciously (i.e. strategic transfer) and unconsciously (i.e. automatic transfer) in their oral production. Although all of the examples of detected automatic transfer were 300 erroneous utterance due to its nature, CBLs’ strategic transfer particularly of foreignizing resulted in both positive transfer and negative transfer. In addition to CBLs’ transfer as a communication strategy or means of solving problems, it was also assumed that some CBLs had existing L2 knowledge or IL regarding the meaning(s) of partially deceptive cognates. For this reason, I suggested a hierarchy of different levels of CBLs’ L2 or IL knowledge and use of strategic transfer for their correct utterances of partially deceptive cognates, which may be different from the cases of cognates in European languages. When CBLs produced the cognate successfully in the L2 context, they may have used strategic transfer with or without prior knowledge about the existence of the cognate. If they were aware of the existence of the cognate, they may or may not have been familiar with its L2 pronunciation. Even when CBLs were familiar with the L2 phonology of the cognate, the semantic content they assume it carries is not necessarily the same as JNSs’ norm, i.e., there may be over- extension or under-extension of the actual semantic ranges in the TL. CBLs also used their L1 knowledge both successfully and unsuccessfully in their comprehension of novel L2 cognates. The overall results of the translation test suggest that although CBLs tended to rely on a bottom up approach or L1 lexical clues to infer the meaning of the L2 cognates, they were still able to deny and abandon their first hypothesis if their L1-based inferencing clearly disagreed with the context. Also where the check up function of the context did not work well, i.e., when their guessed meaning made sense or agreed with the context accidentally, CBLs were liable to misinterpret the meaning of the cognate. 301 As a result of such misinterpretation or even correct guessing of the meaning of partially deceptive cognates in particular L2 contexts, CBLs may create erroneous IL based on their (false) success: successfully transferred L1 knowledge will be incorporated into their IL system (Corder 1983). In connection with this, I proposed two distinct types of the IL that CBLs tend to create regarding the semantic contents of partially deceptive cognates, namely, L1-based IL and L2-based IL. Where the semantic range of the cognate pair is wider in L1, i.e., (J L2. Therefore, these type of cognates are susceptible to over-extension or negative transfer both in production and reception. However, where the semantic range of the cognate pair is wider in L2, i.e., (J>C), CBLs may create either L1-based IL or L2-based IL according to the type of the L2 input they obtain more frequently: L1-L2 common meaning or L2-specific meaning. That is, if CBLs come across the L1-L2 common meaning of the cognate more frequently, they are more likely to construct L1-based IL. CBLs’ IL for this type of cognate is expected to promote positive transfer both in production and comprehension when this L1-L2 common meaning is involved because all the semantic content that they assume for it is included within the cognate in L2. Conversely, if CBLs encounter the L2-specific meaning of this type of cognate more frequently, they are more likely to create its L2- based IL. Once this kind of IL has been constructed, CBLs’ production or comprehension of this type of cognate is less likely to be affected by transfer. At the same time, however, CBLs may have difficulty in their production or comprehension of 302 the L2-specific meaning of (J>C) if they have not encountered it as an L2 input previously. The reason that such erroneous IL sometimes persists and is difficult to be restructured might be that CBLs are less likely to doubt their IL. In the case of CBLs’ over- extension of (J However, CBLs may also hypothesise that both their L1-based IL such as “翻訳 (translation / *translator / *interpretation / *interpreter)” and the correct L2 utterance “ 通訳interpretation / interpreter)” can be used interchangeably for the concept of “interpreter”. In the case of CBLs’ under-extension of (J>C), restructuring of their IL may be difficult because it is likely to contain more frequently used meaning of the partially deceptive cognates in L2. Therefore, CBLs’ difficulty in the semantic restructuring of this type of partially deceptive cognates seems to be largely dependent on the frequency of the L2 input of the lesser used usage and the strength of the context it tends to appear in. In other words, it would be extremely hard for CBLs to notice their under-extension if their unknown usage or meaning of the partially deceptive cognates rarely occurs and the context they encountered it in (if they did) is not powerful enough to deny their existing IL. 303 The results of Study C showed that transferability of Chinese polysemy into Japanese corresponding words exerts a very strong influence on CBLs’ use and acquisition of these Japanese words. If CBLs perceive a certain usage of the Chinese polysemy as prototypical, transfer tends to be facilitated regardless of the potential transferability of its Japanese corresponding words and task type. Yet, CBLs’ negative transfer would decrease with the development of their L2 proficiency because of encountering positive evidence in the L2, although it may persist as an automatic transfer in spite of their correct L2 knowledge of the word. On the contrary, if CBLs perceive a certain usage of the Chinese polysemy as non-prototypical, transfer is more likely to be inhibited in their production of its corresponding word in Japanese. However, transfer is still at work in their comprehension even for the same usage of the same word. Accordingly, I suggested an overall picture of CBLs’ developmental pattern with regard to Japanese words that correspond to each usage of Chinese polysemy: CBLs’ under-extension of potentially transferable items in their production and over-extension of non-transferable items both in their production and reception will be restructured as they are exposed to positive evidence of the L2 input. 6.2 Contributions to the Field In general, the most significant contribution of this thesis is the advancement in understanding the structures of interlingual semantic mapping of learners whose L1 and L2 share numerous cognates but are typologically unrelated. In addition, this thesis has elucidated the complex nature of the IL on partially deceptive cognates that learners tend to construct according to the type of semantic overlap between their L1 and L2. Furthermore, this thesis has shed light on the structures of L1-L2 mapping of 304 polysemous words between Japanese and Chinese using a modified version of Kellerman’s (1979) framework. This thesis has also made some original contributions in the areas of research on L1 transfer and on L2 lexico-semantic development particularly of cognates and polysemy. First, it provides substantial evidence for various types (e.g. strategic or automatic) and levels (i.e. in production, reception and learning) of L1 transfer particularly by eliciting certain target items in multiple tasks. This research is also valuable in that although these different types of transfer have been found previously and the data collected is mostly in European languages (e.g. English, Danish, Dutch, or Swedish), the thesis has shown that similar phenomena do occur among non-European languages especially those in which logographic characters are used. Furthermore, the thesis attempted to investigate the interaction between learners’ transfer and other variables in their acquisition of certain lexical items, and suggested overall picture of such interplay. Second, this research has supported the findings of empirical studies on the cognates of European languages that learners are less likely to misinterpret the meaning of novel deceptive cognates when contextual information is sufficient and its check up function works (e.g. Holmes & Ramos 1993), and partially deceptive cognates are more susceptible to errors than other types of cognates (Granger 1996). Furthermore, it has revealed the mechanism of L2 learners’ difficulty in their productive and receptive acquisition of partially deceptive cognates by taking account of their conscious or unconscious use of L1 knowledge in production, roles of contexts in comprehension, and the type of IL they tend to create. 305 Third, this research has found the variables that affect learners’ transfer and acquisition of lexical items, and proposed subordinate conditions of Kellerman’s (1978, 1979) general principle on transferability of polysemy according to task type. Namely, (1) where perceived prototypicality of a particular L1 item is relatively high, learners tend to be willing to transfer it to L2 both in production and comprehension tasks, and (2) where perceived prototypicality of a particular L1 item is relatively low, learners tend to be reluctant to transfer it to L2 in a production task, while they are still willing to use L1 knowledge in a comprehension task. Fourth, this research has not only supported previous findings about continued L1 semantic involvement in later stages of adult L2 learners’ lexico-semantic development, but also given insight into a different type of L2 learners’ difficulty: under-extension in the case of “convergence” or where multiple L1 words are represented by one L2 word. That is, most claims made in this line of research have been based on the investigation of same translation pairs (i.e. two L2 words that share the same L1 translation equivalent) or the cases of “divergence”, but “convergence” has been rarely examined (except Takahashi 1984, and Imai 1993). Although this type of difficulty for learners has been largely ignored or overlooked in recent studies of this area, the results of the current research showed that learners might have comprehension difficulty or be prone to create erroneous IL (or specific exemplar) in the case of “convergence”. Moreover, it shed light on the roles of L2-based IL in learners’ L2 lexico-semantic development. 6.3 Limitations and Suggestions for Future Studies Now, I would like to turn to limitations of the present research and suggestions for future studies. The limitations of the current research include methodological problems. 306 The first methodological problem is its inability to distinguish CBLs’ “existing knowledge” of the target items and their “successful use of transfer”, although many different levels of CBLs’ knowledge and their use of transfer (as a communication strategy) were proposed. Second, the present research was not able to control such variables as frequency, register, or genre of the target items well. Future research needs to take these variables into consideration and control them accordingly. Apart from the limitations mentioned in each study, a whole series of studies in this thesis was conducted in an English speaking country, and thus all the CBLs who participated in these studies had a certain amount of English language knowledge as well as Japanese language knowledge. Although the methodological benefits owing to the learners’ multilingual nature were put forward, it is not clear to what extent the learners’ knowledge of English has affected the results of each test (if any). There has been growing concern about the roles of both learners’ L1 and L2 in their acquisition of a third language (e.g. Cenoz et. al 2001; Herdina & Jessner 2002). It would be interesting to investigate the respective roles of both Chinese and English in CBLs’ acquisition of Japanese in English speaking countries. Another limitation of the overall research is that as each study was carried out in a foreign language environment where the TL input for learners outside of the classroom is minimal, it is unknown whether we would obtain the same results if it was conducted in a second language environment. As is shown in the results of each study, L2 input may play a crucial role in learners’ transfer and acquisition of lexical items. Thus, future studies should explore whether or not similar results can be found from the CBLs who are learning Japanese in a second language environment such as in Japan. 307 An additional limitation is that since the present research investigated CBLs’ knowledge and use of each target word only with regard to its form and meaning(s) in minimal contexts, the results of each study cannot be equated with their complete knowledge of the word or ability to use the word in any contexts. Nation (2001) showed nine aspects which are involved in “knowing a word” (i.e. spoken form, written form, form of word parts, form and meaning, concept and referents, associations, grammatical functions, collocations, and constraints on use). In other words, participants of the current research are not necessarily familiar with the grammatical functions of the word or capable of using the target items in different contexts even when they had correct knowledge about the form and meaning of the word. Hence, future studies would benefit by testing other aspects of learners’ lexical knowledge in order to see how and to what extent CBLs’ knowledge or use of these aspects of words are influenced by their L1. 6.4 Pedagogical Implications As a Japanese language teacher, I would like to suggest the pedagogical implications of the present research. Generally, our primary concern in teaching Japanese kanji compounds and kanji words to CBLs is how to promote their positive transfer and minimise their negative transfer (as a product). The results of Study A, which gave insight into CBLs’ differing developmental patterns for each type of kanji compounds, indicate that most CBLs are able to acquire the TL meaning of (S), (N), and (D) by the time they reach an advanced level in their general proficiency. Hence, all we need to do is to provide CBLs with as much input of these kanji compounds as possible within certain contexts to promote their IL construction and restructuring (if necessary). 308 In this way, we can facilitate CBLs’ positive transfer of good cognates (S) and of some of the non-cognate kanji compounds (N). At the same time, CBLs’ negative transfer of totally deceptive cognates (D) and of some of the (N) will be reduced as they are exposed to the positive evidence of the L2 input in specific contexts. Similarly, CBLs’ acquisition of Japanese words that correspond to each usage of Chinese polysemy can also be promoted by the positive evidence of the L2 input, since CBLs’ over-extension and under-extension seem to be simply due to their lack of positive evidence in the L2. Matsushita (2002) proposed “the preceding module of vocabulary learning for Chinese background learners of Japanese” in which he assumes that it would be most effective to introduce intermediate or advanced level kanji compounds / kanji words to post- beginners when CBLs are the only group in the classroom. Such increased input of kanji compounds and kanji words for CBLs may accelerate their vocabulary learning of Japanese by promoting their IL construction and restructuring process. Matsushita (2002) also argues that this would facilitate not only the vocabulary learning of CBLs, but also their learning of grammar, as teachers would be able to provide a wider variety of and more familiar examples in introducing new grammatical items. In addition to Matsushita’s (2002) proposal, I would like to further suggest that CBLs’ acquisition of Japanese kanji compounds and kanji words will be facilitated through increased input even where CBLs are not the only members of the classroom. Let us take an example of beginner or post-beginner Japanese classes in an Australian university in which the course syllabus targets EBLs. Given that the number and type of kanji script to be taught is largely restricted in these classes because of its difficulty for 309 EBLs, the classes are not necessarily the most favourable learning environments for CBLs, particularly in terms of vocabulary learning. CBLs may miss opportunities for learning new Japanese kanji compounds or kanji words if they are not given the chance to associate them with corresponding L1 words. Hence, teachers can facilitate CBLs’ acquisition of these Japanese words by simply showing them the kanji script in addition to the pronunciation of the word in hiragana syllables even when the kanji script is not to be taught in the course syllabus. The kanji script for these words can be introduced as “optional” or just shown as a “reference” so that non-CBLs will not find them too difficult and lose their motivation to study Japanese. Yet, CBLs do not always acquire native-like semantic content of partially deceptive cognates (O) by simply obtaining their contextualised input. As is shown in Study B, CBLs tend to create either L1-based IL or L2-based IL as a result of gaining the L2 input of this type of cognates. In order to encourage CBLs to notice the gap between their IL and the correct L2 knowledge, explicit instruction may be necessary and effective since natural contexts are not always powerful enough to reveal such differences (Jiang 2004a). Thus, teachers can explicitly point out the semantic gap of partially deceptive cognates between Japanese and Chinese where possible. It might also be beneficial to provide CBLs with either negative evidence or positive evidence of the over-extended usage or under-extended usage of the partially deceptive cognates, respectively. CBLs often over-extend their L1-based IL onto the smaller L2 semantic content in the case of “divergence” or (J 310 least to reduce their negative transfer, it is also important to deal with CBLs’ automatic transfer of this type of partially deceptive cognates. However, as they have correct knowledge regarding the sematic content of the cognate in L2 already, the only way to suppress their unconscious mental process seems to be “frequent use” of the correct word by which they can automatise it in their L2 use. On the other hand, CBLs’ both of their L1-based IL and L2-based IL are often under- extended in the case of “convergence” or (J>C). Therefore, providing CBLs with positive evidence of the under-extended usage may help their restructuring process. In providing CBLs with examples of both negative and positive evidence of over-extended and under-extended usages of partially deceptive cognates, teachers should give sentences whose contexts can highlight the semantic differences between their IL and the actual L2 semantic content, and should avoid equivocal contexts where both meanings of the partially deceptive cognates make sense. As Bunkachou (1978) has observed, good cognates (S) make up two-thirds, non- cognates (N) one-fourth, and deceptive cognates (D) and (O) less than one-tenth of two thousand kanji compounds that are taught in beginner and intermediate Japanese courses in Japan. The number of partially deceptive cognates (O) is limited, and therefore, it may be possible to pay special attention to this type of cognates in constructing course syllabus, developing materials, and teaching in the classroom, if CBLs make up a large proportion of the students. Whereas there have been some studies that categorise Japanese-Chinese cognates according to the extent to which their meanings overlap between two languages (e.g. Bunkachou 1978; Moriya 1979), the 311 most comprehensive and detailed description of Japanese-Chinese deceptive cognates can be found in Ueno & Lu (1995). Finally, it may also be effective to provide training for teachers of Japanese dealing with the importance of facilitating CBLs’ positive transfer and on the basic correspondence between Chinese words and their Japanese equivalents. This is important because CBLs are the largest L1 group among learners of Japanese in the world, and they are also learning Japanese overseas in vast numbers. It is hoped that increasing efforts will be made to link research with practice in order to enhance CBLs’ learning of Japanese. 312 REFERENCES Aitchison, J. (1994). Words in the Mind? An Introduction to the Mental Lexicon (2nd ed). Oxford: Blackwell. Andersen, R. (1983). Transfer to somewhere, in Gass, S. and Selinker, L., eds. Language Transfer in Language Learning. Rowley, MA: Newbury House, pp177-201. Andersen, R. and Shirai, Y. (1994). Discourse motivations for some cognitive acquisiton principles. Studies in Second Language Acquisition, Vol.16, No.2, pp133- 156. Anthony, E. M. (1952-3). The teaching of cognates. Language Learning, Vol.4, No.1, 79-82. Arakawa, K. (1979). Chuugokugo to kango: bunkachou “Chuugokugo to taiousuru kango” no hyou o kanete (Chinese language and Japanese kanji compounds: Assessing Bunkachou’s “Correspondences between Japanese and Chinese Character Compounds”). Aichi Daigaku Bungaku Ronsou (Collections of Articles on Literature, Aichi University), Vol.61, pp1-28. Ard, J. and Gass, S. (1987). Lexical constraints on syntactic acquisition. Studies in Second Language Acquisition, Vol.9, No.2, pp233-251. Ard, J. and Homburg, T. (1983). Verification of language transfer, in Gass, S. and Selinker, L., eds. Language Transfer in Language Learning. Rowley, MA: Newbury House, pp147-176. Banta, F. G. (1981). Teaching German vocabulary: The use of English cognates and common loan words. Modern Language Journal, Vol.65, No.2, pp129-136. Barcroft, J. (2004). Second language vocabulary acquisition: A lexical input processing approach. Foreign Language Annals, Vol.37, No.2, pp200-208. Berlin, B. and Kay, P. (1969). Basic Color Terms: Their Universality and Evolution. Berkeley: University of California Press. Bialystok, E. (1988). Psycholinguistic dimensions of second language proficiency, in Rutherford, W. and Sharwood Smith, M., eds. Grammar and Second Language Teaching. New York: Newbury House, pp31-50. Bialystok, E. (1994). Analysis and control in the development of second language proficiency. Studies in Second Language Acquisition, Vol.16, No.2, pp157-168. Bialystok, E. and Sharwood Smith, M. (1985). Interlanguage is not a state of mind: An evaluation of the construct for second-language acquisition. Applied Linguistics, Vol.6, No.2, pp101-117. 313 Bley-Vroman, R. and Houng, C. (1988).Why do Chinese use few relative clauses in English? University of Hawaii. Working Papers in ESL, Vol.7, pp93-98. Bogaards, P. and Laufer, B. (2004). Introduction, in Bogaards, P. and Laufer, B., eds. Vocabulary in a Second Language. Amsterdam: John Benjamins, ppVII-XIV. Briere, E. (1968). A Psycholinguistic Study of Phonological Interference. Mouton: The Hague. Bunkachou (1978). Chuugokugo to Taiousuru Kango (Correspondences between Japanese and Chinese Character Compounds). Tokyo: Ookurashou Press. Bybee, J. L. and Slobin, D. I. (1982). Rules and schemas in the development and use of the English past tense. Language, Vol.58, No.2, pp265-289. Carroll, S. (1992). On cognates. Second Language Research, Vol.8, No.2, pp93-119. Carter, R. (1987). Vocabulary: Applied Linguistic Perspective. London: Allen and Unwin. Carter, R. and McCarthy, M. (1988). Development in the teaching of vocabulary: 1945 to the present day, in Carter, R. and McCarthy, M., eds. Vocabulary and Language Teaching. London: Longman, pp39-59. Cenoz, J., Hufeisen, B. and Jessner, U. (2001). Cross-linguistic Influence in Third Language Acquisition: Psycholinguistic Perspectives. Clevedon, UK: Multilingual Matters. Channell, J. (1988). Psycholinguistic considerations in the study of L2 vocabulary acquisition, in Carter, R. and McCarthy, M., eds. Vocabulary and Language Teaching. London: Longman, pp83-96. Chen, H. C. and Leung, Y. S. (1989). Patterns of lexical processing in a nonnative language. Journal of Experimental Psychology: Learning, Memory and Cognition, Vol.15, No.2, pp316-325. Chen, S. Q. (1990). A study of communication strategies in interlanguage production by Chinese EFL learners. Language Learning, Vol.40, No.2, pp155-187. Chen, Y. M. (2003a). Chuugokugo o bogo to suru nihongo gakushuusha ni okeru kango shuutoku kenkyuu no gaikan: Imi to youhou o chuushin ni (Learning the meaning and usage of kanji compound words: A review of Japanese kanji compound words as acquired by native Chinese speakers). Dainigengo Shuutoku, Kyoiku no Kenkyuu Saizensen 2003, Nihongo Gengo Bunka Kenkyuukai (The State of the Art in Second Language Acquisition and Instruction Research-2003 November Version), pp96-113. Chen, Y.M. (2003b). Chuugokugo o bogo to suru nihongo gakushuusha no kango shuutoku ni tsuite: Dougigo, ruigigo, igigo, datsurakugo no 4 taipu kara no kentou (Acquisition of Japanese kanji compounds by Chinese background learners of Japanese: 314 Consideration into four types of good cognates, partially deceptive cognates, totally deceptive cognates, and non-cognate kanji compounds). Nihongo Kyoiku Gakkai Shuuki Taikai Yokoushuu (Proceedings of the Autumn Conference of the Society for Teaching Japanese as a Foreign Language), pp174-179. Chiu, H. C. (2003). Taiwanjin nihongo gakushuusha no nihongo kanji jukugo no on’in shori ni tsuite: Tango taipu, tango no shuutoku nenrei, shuujukudo no kanten kara no kentou (The phonological processing of Japanese kanji words by Taiwanese learners of Japanese: An analysis of word type, age of acquisition, and fluency level). Nihongo Kyoiku (Journal of Japanese Language Teaching), Vol.116, pp89-98. Chou, J. (1986). Nicchuu kango taiou no mondai: Bunkachou “Chuugokugo to taiou suru kango” ni tsuite. Sukejin Daigaku Gaigo Gakuin Nihongo Bungakukei (Department of Japanese Literature, College of Foreign Language Studies, Furen University), Vol.12, pp69-89. Clark, H. and Clark E. (1977). Psychology and Language. New York: Harcourt Brace Jovanovich. Corder, S. P. (1967). The significance of learner’s errors. International Review of Applied Linguistics in Language Teaching, Vol.5, No.4, pp161-170. Corder, S. P. (1978). Language distance and the magnitude of the learning task. Studies in Second Language Acquisition, Vol.2, No.1, pp27-36. Corder, S. P. (1983). A role for the mother tongue, in Gass, S. and Selinker, L., eds. Language Transfer in Language Learning. Rowley, MA: Newbury House, pp85-97. Dagut, M. (1977). Incongruities in lexical gridding – An application of contrastive semantic analysis to language teaching. International Review of Applied Linguistics in Language Teaching, Vo.15, No.3, pp221-229. Dagut, M. and Laufer, B. (1985). Avoidance of phrasal verbs – A case for contrastive analysis. Studies in Second Lanugage Acquisition, Vol.7, No.1, pp73-79. de Bot, K., Paribakht, T. S., and Wesche, M. B. (1997). Towards a lexical processing model for the study of second language vocabulary acquisition. Studies in Second Language Acquisition, Vol.19, No.3, pp309-329. Dechert, H. and Raupach, M. (1989). Introduction, in Dechert, H. and Raupach, M., eds. Transfer in Language Production. Norwood, NJ: Ablex, ppix-xvii. DeKeyser, R. M. (1995). Learning second language grammar rules: An experiment with a miniature linguistic system. Studies in Second Language Acquisition, Vol.17, No.3, pp379-410. de la Fuente, M. J. (2002). Negotiation and oral acquisition of L2 vocabulary: The roles of input and output in the receptive and productive acquisition of words. Studies in Second Language Acquisition, Vol.24, No.1, pp81-112. 315 Dijkstra, T., Jaarsveld, H. V. and Brinke, S. T. (1998). Interlingual homograph recognition: Effects of task demands and language intermixing. Bilingualism, Language and Cognition, Vol.1, No.1, pp51-66. Dulay, H. and Burt, M. (1973). Should we teach children syntax? Language Learning, Vol.23, No.2, pp245-258. Dulay, H. and Burt, M. (1974). Natural sequences in child second language acquisition. Language Learning, Vol.24, No.1, pp37-53. Dulay, H., Burt, M. and Krashen, S. (1982). Language Two. New York: Oxford Press. Duquette, L. and Painchaud, G. (1996). A comparison of vocabulary acquisition in audio and video contexts. The Canadian Modern Language Review, Vol.53, No.1, pp143-172. Eckman, F. (1977). Markedness and the contrastive analysis hypothesis. Language Learning, Vol.27, No.2, pp315-330. Eckman, F. (1985). Some theoretical and pedagogical implications of the markedness differential hypothesis. Studies in Second Language Acquisition, Vol.7, No.3, pp289- 307. Ellis, N. (2002). Frequency effects in language processing: A review with implications for theories of implicit and explicit language acquisition. Studies in Second Language Acquisition, Vol.24, No.2, pp143-188. Ellis, N. and Schmidt, R. (1998). Rules or associations in the acquisition of morphology? The frequency by regularity interaction in human and PDP learning of morphosyntax. Language and Cognitive Processes, Vol.13, No. 2-3, pp307-336. Ellis, R. (1985a). Sources of variability in interlanguage. Applied Linguistics, Vol.6, No.2, pp118-131. Ellis, R. (1985b). Understanding Second Language Acquisition. Oxford: Oxford University Press. Ellis, R. (1994). The Study of Second Language Acquisition. Oxford: Oxford University Press. Ellis, R. and He, X. (1999). The roles of modified input and output in the incidental acquisition of word meanings. Studies in Second Language Acquisition, Vol.21, No.2, pp285-301. Ellis, R., Tanaka, Y. and Yamazaki, A. (1994). Classroom interaction, comprehension and the acquisition of L2 word meanings. Language Learning, Vol.44, No.3, pp449- 491. 316 Færch, C. and Kasper, G. (1980). Processes and strategies in foreign language learning and communication. Interlanguage Studies Bulletin, Vol.5, No.1, pp47-118. Færch, C. and Kasper, G. (1986a). Cognitive dimensions of language transfer, in Kellerman, E., and Sharwood Smith, M., eds. Crosslinguistic Influence in Second Language Acquisition. Oxford: Pergamon, pp49-65. Færch, C. and Kasper, G. (1986b). The role of comprehension in second language learning. Applied Linguistics, Vol.7, No.3, pp257-274. Færch, C. and Kasper, G. (1987). Perspectives on language transfer. Applied Linguistices, Vol.8, No.2, pp111-136. Færch, C. and Kasper, G. (1989). Transfer in production: Some implications for the interlanguage hypothesis, in Dechert, H., and Raupach, M., eds. Transfer in Production. Norwood, NJ: Ablex, pp173-193. Fillmore, C. (1975). An alternative to checklist theories of meaning, Papers from the First Meeting, Berkeley Linguistics Society. Berkeley: Berkeley Linguistics Society, University of California, pp123-131. Ford-Niwa, J., Kobayashi, N. and Yamamoto, H. (1995). Nihongo nouryoku kan’i shiken (SPOT) wa nani o sokutei shiteiruka: Onsei teepu youin no kaiseki (What does a “Simple Performance-Oriented Test” measure? A study of the tape listening factor). Nihongo Kyoiku (Journal of Japanese Language Teaching), Vol.86, pp93-102. Fries, C. (1945). Teaching and Learning English as a Foreign Language. Ann Arbor: University of Michigan Press. Gass, S. (1999). Incidental vocabulary learning. Studies in Second Language Acquisition, Vol.21, No.2, pp319-333. Gass, S. and Mackey, A. (2002). Frequency effects and second language acquisition: A complex picture? Studies in Second Language Acquisition, Vol.24, No.2, pp249-260. Gass, S. and Selinker, L. (1994). Second Language Acquisition: An Introductory Course. Mahwah, NJ: Lawrence Erlbaum. Gasser, M. E. (1990). Connectionism and universals of second language acquisition. Studies in Second Language Acquisition, Vol.12, No.2, pp179-199. Giacobbe, J. (1992). A cognitive view of the role of L1 in the L2 acquisition process. Second Language Research, Vol.8, No.3, pp232-250. Graham, C. R. and Belnap, R. K. (1986). The acquisition of lexical boundaries in English by native speakers of Spanish. International Review of Applied Linguistics in Language Teaching, Vol.24, No.2, pp275-286. 317 Granger, S. (1996). Romance words in English: From history to pedagogy, in Svarvik, J., ed. Words: KVHAA conference 36. Stockholm: Almquist and Wiksell, pp105-121. Griffin, G. F. and Harley, T. A. (1996). List learning of second language vocabulary. Applied Psycholinguistics, Vol.17, No.4. pp443-460. Haiman, J. (1983). Iconic and economic motivation. Language, Vol.59, No.4, pp781- 819. Hall, C. J. (2002). The automatic cognate form assumption: Evidence for the parasitic model of vocabulary development. International Review of Applied Linguistics in Language Teaching, Vol.40, No.2, pp69-87. Hammer, P. and Monod, M. (1976). English-French Cognate Dictionary. Edmonton: The University of Alberta. Han, Z. (2000). Persistence of the implicit influence of NL: The case of the pseudo- passive. Applied Linguistics, Vol.21, No.1, pp78-105. Hancin-Bhatt, B. and Nagy, W. (1994). Lexical transfer and second language morphological development. Applied Psycholinguistics, Vol.15, No.3, pp289-310. Harrington, M. (2002). Cognitive perspectives on second language acquisition, in Kaplan,R.B.,ed.The Oxford Handbook of Applied Linguistics. New York: Oxford University Press, pp124-140. Harrington, M. (2005). Response time as a developmental index of L2 proficiency, in Horiba, Y., ed. Text Understanding and Learning: The Effects of Text’s Linguistic Features on Comprehension and Memory. Tokyo: Graduate School of Language Sciences, Kanda University of International Studies, Centre for Language Sciences, pp105-121. Hasselgren, A. (1994). Lexical teddy bears and advanced learners: A study into the ways Norwegian students cope with English vocabulary. International Journal of Applied Linguistics, Vol.4, No.2, pp237-260. Hatasa, Y. A. (1992). Transfer of the Knowledge of Chinese Characters to Japanese. Unpublished doctoral dissertation, University of Illinois at Urabana-Champaign. Herdina, P. and Jessner, U. (2002). A Dynamic Model of Multilingualism: Perspectives of Change in Psycholinguistics. Clevedon, UK: Multilingual Matters. Hida, Y. and Lu, Y. X. (1986). “Chuugokugo to taiousuru kango” o shindansuru (Diagnosing “Correspondences between Japanese and Chinese Character Compounds”). Nihongogaku (Japanese Linguistics), Vol.5, No.6, pp72-85. Hiki, K. (1995). An Exploratory Study into Second Language Learner Knowledge of Semantically Similar Lexical Items: The Case of Verbs of Perception, Unpublished doctoral dissertation, Indiana University. 318 Hishinuma, T. (1980). Chuugokugo to nihongo no gengo kanshou: Chuugokujin gakushuusha no goyourei (The linguistic interference between Chinese & Japanese— The explanation of mistakes made by Chinese learners of Japanese). Nihongo Kyoiku (Journal of Japanese Language Teaching), Vol.42, pp58-72. Holmes, J. (1986). Snarks, quarks and cognates: An elusive fundamental particle in reading comprehension. The ESPecialist, Vol.15, pp13-40. Holmes, J. and Ramos, R. G. (1993). False friends and reckless guessers: Observing cognate recognition strategies, in Huckin, T., Haynes, M., and Coady, J., eds. Second Language Reading and Vocabulary Learning. Norwood, NJ: Ablex, pp86-108. Hu, G. (2002). Psychological constraints on the utility of metalinguistic knowledge in second language production. Studies in Second Language Acquisition, Vol.24, No.3, pp347-386. Huckin, T. and Coady, J. (1999). Incidental vocabulary acquisition in a second language: A review. Studies in Second Language Acquisition, Vol.21, No.2, pp181-193. Hulstijn, J. (2003). Incidental and intentional learning, in Doughty, C., and Long, M., eds. Handbook of Second Language Acquisition. Oxford: Blackwell, pp349-381. Hulstijn, J. and Marchena, E. (1989). Avoidance: Grammatical or semantic causes. Studies in Second Language Acquisition, Vol.11, No.3, pp241-255. Hulstijn, J. and Tangelder, P. (1991). Intralingual Interference: Lexical Errors as a Result of Formal and Semantic Similarities among English Word-pairs. Paper presented at the First Annual Meeting of the European Second Language Association (EUROSLA): Salzburg. Ijaz, I. H. (1986). Linguistic and cognitive determinants of lexical acquisition in a second language. Language Learning, Vol.36, No.4, pp401-451. Imai, M. (1993). Gaikokugo gakushuusha no goi gakushuu ni okeru mondaiten: Imi hyoushou no kenchi kara (What is missing in L2 word meaning representation? Problems in second language vocabulary learning). Kyoiku Shinrigaku Kenkyuu (Japanese Journal of Educational Psychology), Vol.41, pp243-253. James, C. (1980). Contrastive Analysis. London: Longman. James, C. (1998). Errors in Language Learning and Use. London: Longman. The Japan Foundation (1988). Hyouki (Description): Japanese Teachers’ Handbook. Tokyo: Bonjin Press. Jarvis, S. (1997). The Role of L1-based Concepts in L2 Lexical Reference. Unpublished doctoral dissertation, Indiana University. 319 Jarvis, S. (2000). Methodological rigor in the study of transfer: Identifying L1 influence in the interlanguage lexicon. Language Learning, Vol.50, No.2, pp245-309. Jarvis, S. and Odlin, T. (2000). Morphological type, spatial reference, and language transfer. Studies in Second Language Acquisition, Vol.22, No.4, pp535-556. Jiang, N. (2000). Lexical representation and development in a second language. Applied Linguistics, Vol.21, No.1, pp47-77. Jiang, N. (2002). Form-meaning mapping in vocabulary acquisition in a second language. Studies in Second Language Acquisition, Vol.24, No.4, pp617-637. Jiang, N. (2004a). Semantic transfer and its implications for vocabulary teaching in a second language. The Modern Language Journal, Vol.88, No.3, pp416-432. Jiang, N. (2004b). Semantic transfer and development in adult L2 vocabulary acquisition, in Bogaards, P. and Laufer, B., eds. Vocabulary in a Second Language. Amsterdam: John Benjamins, pp101-126. Kamimoto, T., Shimura, A. and Kellerman, E. (1992). A second language classic reconsidered – The case of Schachter’s avoidance. Second Language Research, Vol.8, No.3, pp251-277. Kato, T. (2005). Learning strategies employed by Chinese-background learners in learning Japanese vocabulary. Japanese Studies, Vol.25, No.3, pp271-286. Kayamoto, Y. (1996). Nihongo kanji to chuugokugo kanji no keitaiteki on’inteki sai ga chuugokugo bogowasha ni yoru nihongo kanji no yomi ni oyobosu eikyou (Orthographical and phonological effects on reading Japanese kanji by Chinese native speakers). Hiroshima Daigaku Kyoikugakubu Kiyou (The Bulletin of Faculty of Education, Hiroshima University), Vol.45, pp345-352. Kayamoto, Y. (2000). Nihongo o gakushuusuru chuugokugo bogowasha no kanji no ninchi: Joukyuushachoujoukyuusha no shinnaijisho ni okeru on’in jouhoushori (Processing phonological information: Recognition of Japanese characters by advanced- and superior-level native speakers of Chinese). Kyoiku Shinrigaku Kenkyuu (Japanese Journal of Educational Psychology), Vol.48, pp315-322. Kellerman, E. (1977). Towards a characterisation of the strategy of transfer in second language learning. Interlanguage Studies Bulletin, Vol.2, No.1, pp58-146. Kellerman, E. (1978). Giving learners a break: Native language intuitions as a source of predictions about transferability. Working Papers on Bilingualism, Vol.15, pp59-92. Kellerman, E. (1979). Transfer and non-transfer: Where we are now. Studies in Second Language Acquisition, Vol.2, No.1, pp37-57. Kellerman, E. (1983). Now you see it, now you don’t, in Gass, S., and Selinker, L., eds. Language Transfer in Language Learning. Rowley, MA: Newbury House, pp112-134. 320 Kellerman, E. (1984). The empirical evidence for the influence of L1 on interlanguage, in Davies, A., Criper, C., and Howatt, A., eds. Interlanguage. Edinburgh: Edinburgh University Press, pp98-122. Kellerman, E. (1986). An eye for an eye: Crosslinguistic constraints on the development of the L2 lexicon, in Kellerman, E. and Sharwood Smith, M., eds. Crosslinguistic Influence in Second Language Acquisition. New York: Pergamon Press, pp35-48. Kellerman, E. (1995) Crosslinguistic influence: Transfer to nowhere? Annual Review of Applied Linguistics, Vol.15, pp125-150. Kellerman, E. and Sharwood Smith, M. (1986). Crosslinguistic Influence in Second Language Acquisition. New York: Pergamon Press. Klein, W. (1986). Second Language Acquisition. Cambridge: Cambridge University Press. Kleinmann, H. (1977). Avoidance behavior in adult second language acquisition. Language Learning, Vol.27, No.1, pp93-108. Kobayashi, N., Ford-Niwa, J. and Yamamoto, H. (1996). Nihongo nouryoku no atarashii sokuteihou (SPOT) (SPOT: New method of testing Japanese language proficiency). Sekai no Nihongo Kyoiku (Japanese-Language Education around the Globe), Vol.6, pp201-218. Krashen, S. (1976). Formal and informal linguistic environments in language acquisition and language learning. TESOL Quarterly, Vol.10, No.2, pp157-168. Kroll, J. F. and Tokowicz, N. (2001). The development of conceptual representation for words in a second language, in Nicol, J. L., ed. One Mind, Two Languages: Bilingual Language Processing. Cambridge, MA: Blackwell, pp49-71. Kroll, J. F., Michael, E., Tokowicz, N. and Dufour, R. (2002). The development of lexical fluency in a second language. Second Language Research, Vol.18, No.2, pp137- 171. Kormos, J. (2000). The role of attention in monitoring second language speech production. Language Learning, Vol.50, No.2, pp343-384. Lado, R. (1957). Linguistics Across Cultures. Ann Arbor: University of Michigan Press. Lado, R. (1961). Language Testing. New York: McGraw-Hill. Lado, R. (1972). Patterns of difficulty in vocabulary, in Allen, H. B. and Campbell, A., eds. Teaching English as a Second Language. New York: McGraw-Hill, pp275-288. Lakoff, G. (1987). Women, Fire, and Dangerous Things: What categories reveal about the mind. Chicago: The University of Chicago Press. 321 Larsen-Freeman, D. and Long, M. H. (1991). An Introduction to Second Language Acquisition Research. New York: Longman. Laufer, B. (1988). The concept of “synforms” (similar lexical forms) in vocabulary acquisition. Language and Education, Vol.2, No.2, pp113-132. Laufer, B. (1989). A factor of difficulty in vocabulary learning: Deceptive transparency. AILA Review, Vol.6, pp10-20. Laufer, B. (1990a). Words you know: How they affect the words you learn, in Fisiak, J., ed. Further Insights into Contrastive Linguistics. Holland: Benjamins, pp573-593. Laufer, B. (1990b). Why are some words more difficult than others? Some intralexical factors that affect the learning of words. International Review of Applied Linguistics in Language Teaching, Vol.28, No.4, pp293-307. Laufer, B. (1990c). Ease and difficulty in vocabulary learning: Some teaching implications. Foreign Language Annals, Vol.23, No.2, pp147-155. Laufer, B. (1991). Some properties of the foreign language learner’s lexicon as evidenced by lexical confusions. International Review of Applied Linguistics in Language Teaching, Vol.29, No.4, pp317-330. Laufer, B. (1992). Native language effect on confusion of similar lexical forms, in Mair, C. and Marcus, M., eds. New Directions in Contrastive Linguistics: Innsbrucker Beitrage zur Kulturwissenschaft, Vol.2. Innsbruck: University of Innsbruck, pp199-209. Laufer, B. (1997). What's in a word that makes it hard or easy: Some intralexical factors that affect the learning of words, in Schmitt, N. and McCarthy, M., eds. Vocabulary: Description, Acquisition and Pedagogy. Cambridge: Cambridge University Press, pp140-155. Laufer, B. (1998). The development of passive and active vocabulary: Same or different? Applied Linguistics, Vol.19, No.2, pp255-271. Laufer, B. and Eliasson, S. (1993). What causes avoidance in L2 learning: L1-L2 difference, L1-L2 similarity, or L2 complexity? Studies in Second Language Acquisition, Vol.15, No.1, pp35-48. Laufer, B. and Nation, I. S. P. (1995). Vocabulary size and use: Lexical richness in L2 written production. Applied Linguistics, Vol.16, No.3, pp307-322. Laufer, B. and Nation, I. S. P. (1999). A vocabulary size test of controlled productive ability. Language Testing, Vol.16, No.1. pp33-51. Laufer, B. and Paribakht, T. S. (1998). The relationship between passive and active vocabularies: Effects of language learning context. Language Learning, Vol.48, No.3, pp365-391. 322 Lennon, P. (1991). Error: Some problems of definition, identification and distinction. Applied Linguistics, Vol.12, No.2, pp180-196. Lennon, P. (1996). Getting ‘easy’ words wrong at the advanced level. International Review of Applied Linguistics in Language Teaching, Vol.34, No.1, pp23-36. Levelt, W. (1989). Speaking: From Intention to Articulation. Cambridge: Cambridge University Press. Levenston, E. A. (1979). Second language acquisition: Issues and problems. Interlanguage Studies Bulletin, Vol.4, No.2, pp147-160. Lightbown, P. M. and Libben, G. (1984). The recognition and use of cognates by L2 learners, in Andersen, R. W., ed. Second Languages: A cross-linguistic perspective. Rowley, MA: Newbury House, pp393-417. Lin, Y. H. (2002). Nikka, nikkan jiten kara mita nicchuu doukeigo kijutsu no mondaiten: Doukeiruigigo o chuushin ni (Misrepresentation of homographs found in Japanese-Chinese dictionaries). Sekai no Nihongo Kyoiku (Japanese-Language Education around the Globe), Vol.12, pp107-121. Lobo, F. (1966). A 10,000 Word Spanish Vocabulary Expanded from 3,000 English Cognates. PhD thesis. Washington: Georgetown University. Longman Dictionary of Contemporary English (3rd ed.). (1995). London: Longman. Major, R. (1986). The ontogeny model: Evidence from L2 acquisition of Spanish r. Language Learning, Vol.36, No.4, pp453-503. Martin, M. (1984). Advanced vocabulary teaching: The problem of synonyms. Modern Language Journal, Vol.68, No.2, pp130-137. Master, P. (1987). A Cross-linguistic Interlanguage Analysis of the Acquisition of the English Article System. Unpublished doctoral dissertation, University of California. Matsuda, F. (2000). Nihongo gakushuusha ni yoru goi shuutoku: Saika, ippanka, tenkeika no kanten kara (Lexical acquisition by learners of Japanese: From the perspectives of differentiation, generalization and typicalization). Sekai no Nihongo Kyoiku (Japanese-Language Education around the Globe), Vol.10, pp73-89. Matsumi, N. (2002). Dainigengo no goi o shuutokusuru (Acquiring second language vocabulary), in Kaiho, H. and Kashiwazaki, H., eds. Nihongo Kyoiku no tameno Shinrigaku (Psychology for Japanese Language Education). Tokyo: Shinyousha, pp97- 110. Matsunaga, S. (1999). The role of kanji knowledge transfer in acquisition of Japanese as a foreign language. Sekaino Nihongo Kyoiku (Japanese-Language Education around the Globe), Vol.9, pp87-100. 323 Matsushita, T. (2001). Shokyuu Nihongo Bunpou Gakushuu ni tsukaeru Chuu-joukyuu Goi no Kentou: Chuugokugokei Nihongo Gakushuusha no tameno Goigakushuu Senkou Mojuuru Kaihatsu ni mukete (A Consideration into Useful Intermediate- Advanced Level Vocabulary for Grammar Learning at the Beginning Level: Development of Preceding Module of Vocabulary Learning for Chinese background Learners of Japanese). Paper presented at Japanese Culture Forum in Dalian University of Foreign Studies, 2001. Matsushita, T. (2002). Chuugokugo o bogo to suru nihongo gakushuusha no tameno goi gakushuu senkou mojuuru no teian: Daini gengo shuutoku riron, gengo ninchi, taishou bunseki, goiron no seika o fumaete (A proposal of the preceding module of vocabulary learning for Chinese background learners of Japanese: Based on the theories of language acquisition, cognitive linguistics, contrastive analysis, and lexicology). Riyu Xuexi yu Yanjiu (Learning and Study of Japanese), Vol.108, pp50-54. McLaughlin, B. (1978). The monitor model: Some methodological considerations. Language Learning, Vol.28, No.2, pp309-332. McLaughlin, B. (1987). Theories of Second Language Learning. London: Edward Arnold. Meara, P. (1984). The study of lexis in interlanguage, in Davies, A., Criper, C., and Howatt, A., eds. Interlanguage. Edinburgh: Edingurgh University Press, pp225-235. Meara, P. (1993). The bilingual lexicon and the teaching of vocabulary, in Schreuder, R. and Weltens, B., eds. The Bilingual Lexicon. Amsterdam: John Benjamins, pp279- 297. Meara, P. (1996). The dimensions of lexical competence, in Brown, G., Malmkjaer, K. and Williams, J., eds. Performance and Competence in Second Language Acquisition. Cambridge: Cambridge University Press, pp35-53. Meara, P. (1997). Towards a new approach to modelling vocabulary acquisition, in Schmitt, N. and McCarthy, M., eds. Vocabulary: Description, Acquisition and Pedagogy. Cambridge: Cambridge University Press, pp109-121. Melka, F. (1997). Receptive vs productive aspect of vocabulary, in Schmitt, N. and McCarthy, M., eds. Vocabulary: Description, Acquisition and Pedagogy. Cambridge: Cambridge University Press, pp84-102. Miura, A. (1984). Nihongo kara chuugokugo ni haitta kango no imi to youhou (The meanings and usages of some kanji compounds borrowed from Japanese into Chinese). Nihongo Kyoiku (Journal of Japanese Language Teaching), Vol.53. pp102-112. Miura, K. (1997). Nicchuu doukeigo ga gakushuusha ni ataeru eikyou: Nihonjin no chuugokugo gakushuusha o taishounishite (Influence of Japanese-Chinese cognates on learners: Targeting Japanese native speakers learning Chinese). Gengo Bunka 324 (Language and Culture): Aichi Shukutoku Daigaku Gengo Bunka Gakkai (Language and Culture Centre, Aichi Shukutoku University), Vol.6, pp89-96. Miyajima, T. (1993). Nicchuu doukeigo no buntaisa (Stylistic differences of Japanese- Chinese cognates). Handai Nihongo Kenkyuu (Japanese Studies, Osaka Univeristy), Vol.5, pp1-18. Miyajima, T. (1994). Goiron Kenkyuu (The Study of Lexicology). Tokyo: Mugi Shobou. Mondria, J-A. and Wiersma, B. (2004). Receptive, productive, and receptive + productive L2 vocabulary learning: What difference does it make? in Bogaards, P. and Laufer, B., eds. Vocabulary in a Second Language. Amsterdam: John Benjamins, pp79- 100. Mori, Y. (2002). Individual differences in the integration of information from context and word parts in interpreting unknown kanji words. Applied Psycholinguisics,Vol.23, No.3, pp375-397. Mori, Y. (2003). Vocabulary Acquisition in Japanese, in Hatasa, Y. A., ed. An Invitation to Second Language Acquisition Research in Japanese: In Honor of Seiichi Makino. Tokyo: Kurosio Publishers, pp171-186. Mori, Y. and Nagy, W. (1999). Integration of information from context and word elements in interpreting novel kanji compounds. Reading Research Quarterly, Vol.34, No.1, pp80-101. Moriya, H. (1979). Shiryou: Nicchuu Doukeigo – Sono Imiyouhou no Sai (Teaching Material: Differences in meaning and usage between Japanese and Chinese cognates). Nihongo Gakkou Ronshuu (Collection of Articles from Japanese Language Schools), Vol.6, pp159-168. Nagy, W. E. (1997). On the role of context in first-and second-language, in Schmitt, N. and McCarthy, M., eds. Vocabulary: Description, Acquisition and Pedagogy. Cambridge: Cambridge University Press, pp64-83. Nagy, W. E., García, G. E., Durgunoglu, A.Y. and Hancin-Bhatt, B. (1993). Spanish- English bilingual students’ use of cognates in English reading. Journal of Reading Behavior, Vol.25, No.3, pp241-259. Nakahama, Y. (2003). Cross-linguistic Influence on the Development of Referential Topic Management in L2 Japanese Oral Narratives. Unpublished Ph.D. dissertation, Georgetown University. Nation, I. S. P. (1983). Testing and teaching vocabulary. Guidelines, Vol.5, pp12-25. Nation, I. S. P. (1990). Teaching and Learning Vocabulary. New York: Newbury House. 325 Nation, I. S. P. (2001). Learning Vocabulary in Another Language. Cambridge: Cambridge University Press. Odlin, T. (1989). Language Transfer: Cross-linguistic Influence in Language Learning. Cambridge: Cambridge University Press. Odlin, T. (1990). Word-order transfer, metalinguistic awareness and constraints on foreign language learning, in Van Patten, B. and Lee, J., eds. Second Language Acquisition/Foreign Language Learning. Clevedon: Multilingual Matters, pp95-117. Odlin, T. (1992). Transferability and linguistic substrates. Second Language Research, Vol.8, No.3, pp171-202. Odlin, T. (2003). Cross-linguistic influence, in Doughty, C., and Long, M., eds. Handbook of Second Language Acquisition. Oxford: Blackwell, pp436-486. Oller, J. and Ziahosseiny, S. (1970). The contrastive analysis hypothesis and spelling errors. Language Learning, Vol.20, No.2, pp183-189. Ookouchi, Y. (1990). Nihongo to chuugokugo no goi no taishou (A lexical comparison between Japanese and Chinese), in Tamamura, F., ed. Kouza nihongo to nihongo kyoiku (A Course on Japanese Language and Japanese Language Teaching), Vol.7 Nihongo no goi, imi (Vocabulary and meanings of Japanese). Tokyo: Meiji Shoin, pp54-79. Ookouchi, Y. (1997). Nihongo to chuugokugo no doukeigo (Cognates shared between Japanese and Chinese), in Ookouchi, Y., ed. Nihongo to Chuugokugo no Taishou Kenkyuu Ronbunshuu (Collection of Articles on Contrastive Studies between Japanese and Chinese). Tokyo: Kurosio Publishers, pp411-447. Palmberg, R. (1985). How much English vocabulary do Swedish-speaking primary- school pupils know before starting to learn English at school? in Ringbom, H., ed. Foreign Language Learning and Bilingualism. Åbo: Åbo Akademi, pp89-97. Palmberg, R. (1987). Patterns of vocabulary development in foreign language learners. Studies in Second Language Acquisition, Vol.9, No.2, pp201-220. Paribakht, T. and Wesche, M. (1999). Reading and “incidental” L2 vocabulary acquisition. Studies in Second Language Acquisition, Vol.21, No.2, pp195-224. Potter, M. C., So, K. F., Von Eckardt, B. and Feldman, L. B. (1984). Lexical and conceptual representation in beginning and more proficient bilinguals. Journal of Verbal Learning and Verbal Behavior, Vol.23, pp23-38. Poulisse, N. (1990). The Use of Compensatory Strategies by Dutch Learners of English. Dordrecht, The Netherlands: Foris. Poulisse, N. (1993). A theoretical account of lexical communication strategies, in Schreuder, R. and Weltens, B., eds. The Bilingual Lexicon. Amsterdam: John Benjamins, pp157-189. 326 Poulisse, N. (1997). Compensatory strategies and the principles of clarity and economy, in Kasper, G. and Kellerman, E., eds. Communication Strategies: Psycholinguistic and Sociolinguistic Perspectives. London: Longman, pp49-64. Poulisse, N. (1999). Slips of the Tongue: Speech Errors in First and Second Language Production. Amsterdam: John Benjamins. Poulisse, N. and Bongaerts, T. (1994). First language use in second language production. Applied Linguistics, Vol.15, No.1, pp36-57. Read, J. (2000). Assessing Vocabulary. Cambridge: Cambridge University Press. Read, J. (2004). Plumbing the depths: How should the construct of vocabulary knowledge be defined? in Bogaards, P. and Laufer, B., eds. Vocabulary in a Second Language. Amsterdam: John Benjamins, pp209-228. Reikai Japanese Dictionary (1st ed.). (1956). Tokyo: Chuukyou Publishers. Richards, J. (1971). Error analysis and second language strategies. Language Sciences, Vol.17, pp12-22. Richards, J. (1976). The role of vocabulary teaching. TESOL Quarterly, Vol.10, No.1, pp77-90. Ringbom, H. (1978). The influence of the mother tongue on the translation of lexical items. Interlanguage Studies Bulletin, Vol.3, No.2, pp80-101. Ringbom, H. (1983). Borrowing and lexical transfer. Applied Linguistics, Vol.4, No.3, pp207-212. Ringbom, H. (1987). The Role of the First Language in Foreign Language Learning. Philadelphia: Multilingual Matters. Ringbom, H. (1991). Crosslinguistic lexical influence and foreign language learning, in Phillipson, R., Kellerman, E., Selinker, L., Sharwood Smith, M. and Swain, M., eds. Foreign/Second Language Pedagogy Research: A Commemorative Volume for Claus Færch. Clevedon, UK: Multilingual Matters, pp172-181. Ringbom, H. (1992). On L1 transfer in L2 comprehension and L2 production. Language Learning, Vol.42, No.1, pp85-112. Rosch, E. (1973). On the internal structure of perceptual and semantic categories, in Moore, T., ed. Cognitive Development and the Acquisition of Language.NewYork: Academic Press, pp111-144. Rutherford, W. (1983). Language typology and language transfer, in Gass, S. and Selinker, L., eds. Language Transfer in Language Learning. Rowley, MA: Newbury House, pp358-370. 327 Schachter, J. (1974). An error in error analysis. Language Learning, Vol.24, No.2, pp205-214. Schachter, J. (1983). A new account of language transfer, in Gass, S. and Selinker, L., eds. Language Transfer in Language Learning. Rowley, MA: Newbury House, pp98- 111. Schmitt, N. (1997). Vocabulary learning strategies, in Schmitt, N and McCarthy, M., eds. Vocabulary: Description, Acquisition and Pedagogy. Cambridge: Cambridge University Press, pp199-227. Schmitt, N. (2000). Vocabulary in Language Teaching. Cambridge: Cambridge University Press. Schneider, V. I., Healy, A. F. and Bourne, L. E. Jr. (2002). What is learned under difficult conditions is hard to forget: Contextual interference effects in foreign vocabulary acquisition, retention and transfer. Journal of Memory and Language, Vol.46, No.2, pp419-440. Séguin, H. and Tréville, M. C. (1992). Les congénères interlinguaux: Un atout pour accélérer i’acquisition du vocabulaire et faciliter la compréhension des textes? in Courchêne, R., Glidden, J. I., St. John, J. and Thérien, C., eds. Comprehension Based Second Language Teaching. Ottawa: University of Ottawa Press. Seliger, H. (1989). Semantic transfer constraints in foreign language speakers’ reactions to acceptability, in Dechert, H., and Raupach, M., eds. Transfer in Language Production. Norwood, NJ: Ablex, pp21-33. Selinker, L. (1992). Rediscovering Interlanguage. New York: Longman. Sharwood Smith, M. (1994). Second Language Learning: Theoretical Foundations. London: Longman. Sheen, R. (1979). “Faux amis” for francophones learning EFL. SPEAQ Journal, Vol.3, No.1-2, pp95-103. Shi, J. and Wang, J. (1983). Nicchuu doukeigo ni okeru bunpouteki zure (Differences in grammatical functions of cognates between Japanese and Chinese). Nihongo, Chuugokugo Taiou Hyougen Youreishuu (Collection of Corresponding Expressions between Japanese and Chinese), Vol.5, pp56-82. Shibuya, K. (2002). Gakushuusha no bogo no eikyou: Gakushuusha no bogo ga eikyousuru baai to shinai baai (Learners’ L1 influence: When learners’ L1 have an influence on L2 learning and when it does not), in Noda, H., Sakoda, K., Shibuya, K. and Kobayashi, N., eds. Nihongo Gakushuusha no Bunpou Shuutoku (Acquisition of Grammar by Learners of Japanese). Tokyo: Taishuukan, pp.83-99. 328 Shirai, Y. (1990). Putting PUT to use: Prototype and metaphorical extension. Issues in Applied Linguistics, Vol.1, No.1, pp78-97. Shirai, Y. (1992). Conditions on transfer: A connectionist approach. Issues in Applied Linguistics, Vol.3, No.1, pp91-120. Shirai, Y. (1995). The acquisition of the basic verb PUT by Japanese EFL learners: Prototype and transfer. Gogaku Kyoiku Kenkyuu Ronsou (Collection of Articles on Language Education and Research): Daito Bunka University, Vol.12, pp61-92. Shirai, Y. (1998). Gengo gakushuu to purototaipu riron (Language learning and prototype theory), in Okuda, S., ed. Boodaaresu Jidai no Gaikokugo Kyoiku (Foreign Language Education in Borderless Era). Tokyo: Miraisha, pp70-108. Shirai, Y. (2002). The aspect hypothesis in SLA and the acquisition of Japanese. Acquisition of Japanese as a Second Language, Vol.5, pp42-61. Shirai, Y. and Andersen, R. (1995). The acquisition of tense-aspect morphology: A prototype account. Language, Vol.71, No.4, pp743-762. Singleton, D. (1999). Exploring the Second Language Mental Lexicon. Cambridge: Cambridge University Press. Slobin, D. (1973). Cognitive prerequisites for the development of grammar, in Ferguson, C., and Slobin, D., eds. Studies of Child Language Development. New York: Holt, Rinehart and Winston, pp175-208. Slobin, D. (1977). Language change in childhood and history, in MacNamara, I., ed. Language Learning and Thought. New York: Academic Press, pp185-214. Sokolik, M. E. and Smith, M. E. (1992). Assignment of gender to French nouns in primary and secondary language: A connectionist model. Second Language Research, Vol.8, No.1, pp39-58. Sonaiya, R. (1991).Vocabulary acquisition as a process of continuous lexical disambiguation. International Review of Applied Linguistics in Language Teaching, Vol.29, No.4, pp273-284. Sridhar, K. and Sridhar, S. N. (1986). Bridging the paradigm gap: Second language acquisition theory and indigenized varieties of English. World Englishes, Vol.5, pp3-14. Stockwell, R., Bowen, J., and Martin, J. (1965). The Grammatical Structures of English and Spanish. Chicago: Chicago University Press. Strick, G. J. (1980). A hypothesis for semantic development in a second language. Language Learning, Vol.30, No.1, pp155-176. Sugaya, N. (2002). Daini gengo to shiteno nihongo no asupekuto shuutoku kenkyuu gaikan “Dousa no jizoku” to “kekka no joutai” no teiru o chuushinni (A survey of the 329 acquisition of aspect morphology in L2 Japanese: Focusing on the progressive and the resultative state use of -teiru). Gengo Bunka to Nihongo Kyoiku (The State of the Art in Second Language Acquisition and Instruction Research-2002 May version), pp70-86. Sugaya, N. (2004). Purototaipu riron to daini gengo to shiteno nihongo no shuutoku kenkyuu (Prototype theory and acquisition studies on Japanese as a second language). Daini Gengo to shiteno Nihongo no Shuutoku Kenkyuu (Acquisition of Japanese as a Second Language), Vol.7, pp121-140. Suzuki, Y. (1988). Kanji kyoiku no mondai ten: Chuu-joukyuu kanjikei gakusei no baai (Problems in teaching kanji: The case of intermediate and advanced kanji background learners). Kouza Nihongo Kyoiku (A Lecture on Japanese Language Education), Vol.23, pp76-87. Swan, M. (1997). The influence of the mother tongue on second language vocabulary acquisition and use, in Schmitt, N. and McCarthy, M., eds. Vocabulary: Description, Acquisition and Pedagogy. Cambridge: Cambridge University Press, pp156-180. Takahashi, T. (1984). A Study on Lexico-semantic Transfer. Unpublished doctoral dissertation, Columbia University. Takahashi, T. and Tanaka, S. (1992). Mental representation of basic verbs in a second language: TAKE, HOLD, and KEEP. SFE Journal of Language and Communication, Vol.1, pp105-121. Takebe, Y. (1979). Kanji kokumin ni taisuru chuukyuu kanji kyoiku (Teaching of kanji for Chinese people at intermediate level). Nihongo Kyoiku (Journal of Japanese Language Teaching), Vol.37, pp13-23. Talamas, A., Kroll, J. and Dufour, R. (1999). From form to meaning: Stages in the acquisition of second-language vocabulary. Bilingualism: Language and Cognition, Vol.2, No.1, pp45-58. Tamaoka, K. and Matsushita, T. (1999). Chuugokugokei Nihongo Gakushuusha niyoru Nihongo Kanji Niji-jukugo no Ninchi Shori niokeru Bogo no Eikyou (L1 Influence on the Recognition of Two-character Kanji Compounds by Chinese Background Learners of Japanese), Paper presented at the fourth international symposium of Japanese language education and Japanese studies: Japanese language education and Japanese studies in Asia pacific regions – status quo and prospect. Tamaoka, K., Miyaoka, Y. and Matsushita, T. (2003). Nihongo gakushuusha no shinteki jisho no kouzou: Chuugokugo o bogo tosuru choujoukyuu nihongo gakushuusha no kanji jukugo no shori o rei ni (The Structure of Japanese Learners’ Mental Lexicon: Based on an Example of Processing Kanji Compounds by Superior Level Chinese Background Learners of Japanese), Paper presented at Symposium of Cognitive Science and Japanese Language Education. 330 Tanaka, S. (1983). Language Transfer as a Constraint on Lexico-semantic Development in Adults Learning a Second Language in Acquisition-poor Environments. Unpublished doctoral dissertation, Teachers’ College, Columbia University. Tanaka, S. (1987). The selective use of specific exemplars in second language performance: The case of the dative alternation. Language Learning, Vol.37, No.1, pp63-88. Tanaka, S. (1990). Ninchi Imiron (Cognitive Semantics): Eigo Doushi no Tagi Kouzou (Polysemous Structures of English Verbs). Tokyo: Sanyuusha. Tanaka, S. and Abe, H. (1985). Conditions on interlingual semantic transfer, in Larson, P., Judd, E. L. and Messerschmitt, D. S., eds. On TESOL’ 84: A Brave New World for TESOL, Washington D.C.: TESOL, pp101-120. Tanaka, S. and Abe, H. (1989). Gaikokugo gakushuu ni okeru gengo ten’i no mondai (The problem of language transfer in foreign language learning). Eigo Kyoiku (The English Teacher’s Magazine), Vol.3, January, pp78-80. Tanaka, S., Takahashi, T. and Abe, H. (1990). Acquisition of the lexeme MAKE by Japanese learners of English. Eigo Eibungaku Shinchou (A Magazine of English and English Literature), pp406-422. Tarone, E. (1982). Systematicity and attention in interlanguage. Language Learning, Vol.32, No.1, pp69-84. Tarone, E. (1983). On the variability of interlanguage systems. Applied Linguistics, Vol.4, No.2, pp142-164. Taylor, B. (1975). The use of overgeneralization and transfer learning strategies by elementary and intermediate students of ESL. Language Learning, Vol.25, No.1, pp73- 107. Taylor, J. (1989). Linguistic Categorization: Prototypes in Linguistic Theory. Oxford: Oxford Press. Thomason, S. and Kaufman, T. (1988). Language Contact, Creolization, and Genetic Linguistics. Berkeley: University of California Press. Tran-Thi-Chau. (1975). Error analysis, contrastive analysis and students’ perceptions: A study of difficulty in second language learning. International Review of Applied Linguistics in Language Teaching, Vol.13, No.2, pp119-143. Ueno, K. and Lu, X. (1995). Oboeteokitai Nicchuu Doukeiigigo 300 (Japanese-Chinese Deceptive Cognates that should be Remembered). Tokyo: Kouseikan. Van Roey, J. (1990). French-English Contrastive Lexicology – An Introduction. Louvainla- Neuve: Peeters. 331 Viberg, Å. (1998). Crosslinguistic perspectives on lexical acquisition: The case of language-specific semantic differentiation, in Haastrup, K. and Viberg, Å., eds. Perspectives on Lexical Acquisition in a Second Language. Sweden: Lund University Press, pp175-208. Wardhaugh, R. (1970). The contrastive analysis hypothesis. TESOL Quarterly, Vol.4, pp123-130. Waring, R. (1997a). The negative effects of learning words in semantic sets: A replication. System, Vol.25, No.2, pp261-274. Waring, R. (1997b). A comparison of the receptive and productive vocabulary size of some second language learners. Immaculata: Notre Dame Seishin University Okayama, Vol.1, pp53-68. Webb, S. (2005). Receptive and productive vocabulary learning: The effects of reading and writing on word knowledge. Studies in Second Language Acquisition, Vol.27, No.1, pp33-52. Wesche, M. and Paribakht, T. S. (1996). Assessing second language vocabulary knowledge: Depth versus breadth. Canadian Modern Language Review, Vol.53, No.1, pp13-40. Wesche, M. and Paribakht, T. S. (1999). Introduction. Studies in Second Language Acquisition, Vol.21, No.2, pp175-180. Whitman, R. L. and Jackson, K. L. (1972). The unpredictability of contrastive analysis. Language Learning, Vol.22, No.1, pp29-41. Yamaoka, T. (1988). A semantic and prototype discussion of the ‘be easy to V’ structure: A possible explanation of its acquisition process. Applied Linguistics, Vol.9, No.4, pp385-401. Young, R. (1990). Functional interpretations of variation in interlanguage morphology. Paper presented at the 10th Second Language Research Forum. University of Oregon, Eugene. Zimmerman, C. B. (1997). Historical trends in second language vocabulary instruction, in Coady, J. and Huckin, T., eds. Second Language Vocabulary Acquisition. Cambridge: Cambridge University Press, pp5-19. Zobl, H. (1980). The formal and developmental selectivity of L1 influence on L2 acquisition. Language Learning, Vol.30, No.1, pp43-57. Zobl, H. (1986). Word order typology, lexical government, and the prediction of multiple, graded effects in L2 word order. Language Learning, Vol.36, No.2, pp159- 183. 332 Zughoul, M. R. (1991). Lexical choice: Towards writing problematic word lists. International Review of Applied Linguistics in Language Teaching, Vol.29, No.1, pp45- 60. 333 APPENDIX A Name:______Student no: ______ VOCABULARY JUDGEMENT TEST (V1) Please read the following sentences, and (1) choose if the underlined words are used appropriately based on the English translation given, or if they are used inappropriately. (2) If you answer , please correct the underlined word or provide an alternative word that fits into the context. There are NO incorrect vocabulary or grammar usages other than underlined parts. (You do NOT need to change the word written in hiragana into kanji, if the word is used correctly.) U Ʈ(Example) W.' 'NH û :Ċƒ;Y X #54) / (My hobby is listening to music.) 8EY 'Y Y+Y 7 ;M ŨƐ:Ġ¬IJ;53KŒ4) / ļ (Japan’s Shinkansen (bullet train) is very fast.) W.' .7 ' ' û ;ŚŐ%YXŋSG)/ ŋ13G) (I know Mr. Tanaka.) CM 2 G6 / 1. #:Ɓ;đ:4ĺX¤!3% / (This room is hot, so please open the window.) G2 &1' 2. ƑÌxZnhñþXů13G) / (I buy fashion magazines every month.) ># ( Y;Y 3. hqr Ry j{G4ŹÝ²4Ă°Ų SG) / (It takes 2hours and half by aeroplane from Sydney to Brisbane.) 5 P 'N Y OJ ) 4. ޾4 ċ°ƞƕ7}o8ČHG'./ (I stayed at a famous hotel for a week in Tokyo.) P9Y 55 'L >R 5. »Ŭ Ŗ 5lxp^_[: ąX¤ G'./ (Last year, I established a software company with my brother.) 80Q? 5K/0 M - 6. ŨơŨ;Ɲń5ƜĻ T:4bad8Ý!G+Y / (On Sunday, I have an appointment with my friend, so I cannot go to karaoke.) W.' + 5 ) 7. û ;¨ī:Ăĥ¨ R4'. / (When I was a student, I hated mathematics.) 5 P ( Y- G0 8. ޾;Ă° ƍ '§4)/ (Tokyo is a busy city 24 hours a day.) 334 +Y+ KY/ 9. İīƘŃ TY4) Ƅ3K4) / (Sensei, I have a question to ask, may I ask? ) +Y+ 8EYSN -/Y 10. İīŨƐƩ¨:#54ĹŊ'.Y4) / (Sensei, I would like to consult with you about going to Japan for study. ) .7 >5 11. ŚŐ%Y;53KK'Vģ4) / (Mr. Tanaka is a very funny person .) R9Y-2 P '$5 % 12. ƥŬ ĽÃ4)9KöÿXŅ'G'. / (So are you graduating next year? Have you found a job yet?) ' 5 3Y % >R 13. fnb :ýà4) ţř:ä ł ¤ G'. / (Although it is a soccer game, the score gap has become large.) .7 H 14. ŚŐ%Y;2KovXÑ3G) / (Mr. Tanaka is always watching TV.) U SPS 2 5 15. ŷ;ƬƧXîT: ţ4) / (He is good at cooking.) /Y 16. -:v;`{ m 7:4¥ň4ĚR7!U<7SG+Y (The building does not have lifts, so we need to go by the stairs.) / .G . 'L H E 17. Š œ:7RĆ%YXÑ.Ɗ 4)Q / (If you have a headache, you should see a doctor.) CM ; G KY / 18. Ɓ8ũTij8ƙXsnc'3% / (Please knock on the door before entering the room.) : 'YBY H 19. ïŨov4ĠƄXÑG'. / (I watched news on TV yesterday.) 8EY +2 ;Y. 20. ŨƐ5a jp [;µĮ ųŀ4) / (The seasons in Japan and in Australia are opposite.) W.' 9YG 21. û ;Ŭij8 Y# Ra jp [8ƥG'. / (I came from Korea to Australia 3 years ago.) 'PR (# F 'L 22. Ėƥ;ăÕ:Ə ąX2 S.4) / (I would like to establish my own trading company in the future.) W.' G80 'L / 23. û ;ƑŨ´ć4ł¨8ƥG) / (I come to uni by car everyday.) : 335 >5?5 'Y# Y'P :# 29. -:;. %Y:ģ 8ġá7 ęXõ'G'. / (The film has left a deep impression on many people.) % Y Y@2 2 >5 ) 30. æÄŻX÷ģ;Ę7 713 G'./ (These days, less people are using pencils.) %A W) 9 / 31. ìžXƌU.:4ÅXŁ'3% / (I forgot to bring my wallet, so please lend me some money.) 57S CM # 32. ƭ :Ɓ Rw[s:Ĭ #G) / (I can hear the sound of piano from next room. ) +A +A Q N #5W 33. ŨƐĨſ;[ bĨſ: ƣ¹XŇUG+Y/ (The Japanese government cannot decline the U.S. government’s request.) % "Y 34. ~]c%Y:<%Y;ç7:853KÓ³4) / (Mike’s grandmother is 80 years old, but she is full of vigour.) 5 P/ 8EY 0 336 8EY 9Y ('Y 48. ŨƐ4;Ŭ850¢ RăĞ SG) / (In Japan, there are around 50 earthquakes in a year.) CM -( .) 49. Ɓ:ĶĕXē!3"G'P / (Shall I help you clean your room?) MG'. 'P+2 H /) 50. ó%Y; ėįXÑT: łÙ 4) / (Mr. Yamashita loves reading novels.) / 51. 2GR7:4ovX¤!3% / (I am bored, so please turn on the TV.) #Y 337 !Y0 P/ >Q 68. v:ÎŎ8;¼ł7ŸƢ SG)/ (It costs huge amount of money to construct a building.) 78 . ;M / 69. jp4 XĝDT ķ G13%/ ((In a restaurant) Please quickly decide what you will eat.) W.' 8EY $ 5K/0 70. û ;ŨƐ×:Ɲń . %YG) / (I have many Japanese friends.) P 'P / 71. ÁĒ:35| iX¤!3% / (Please open the textbook to page35.) W.' MG/ +' #% 72. û ;óŚ%Y5ĪĄ8Øê'3G) / (I am formally going out with Mr. Yamada.) # "2(P 5 >2Q 73. ¦â8ÌĜÝ Ă;vg źƣ4)/ (If you go abroad for more than 3 months, you need a visa.) 9 92U2 74. ø;pc k:ūƯ7xZ4) / [My sister is an ardent (passionate) fan of Tom Cruise.] 5 + . U H( 75. ~]c%Y:ƀ%Y;Ů ß:8ŷ; ņ 4) / (Although Mike’s father is tall, Mike is short.) "Y& +Y- 76. Ôë]c5[ b;ıĸ'3G)/ (Currently, Iraq and U.S. are at war.) W.' ;.R 0E M! 77. û : š 3TŌƊ;v:30¥7:4ƚÇ U4) / (The place where I am working is on the 30th floor of a building, so the view at night is beautiful.) 0P NSP 78. Œ871.:4 ºƬ Ě SG'./ (As I have been promoted to a manager, my salary has gone up.) 4Y 79. /UK7:8ś³ 2!3G)/ (Although no one is there, the light is on.) 0P%!1 I(NY 80. [ b5ŨƐ:őåË;Ɣď'3G)/ (The results of the investigations between Japan and the U.S. contradict each other.) 8EY #1 Q%Y < / Y 81. ŨƐ:âƠô;űł7Å©4)/ (Japan’s national budget is an enormous amount.) 8EY $ I* 82. ŨƐ×:BY ;53K Ŧ '4) / (Japanese grammar is very difficult.) 8EY 715 . / 83. #U ŨƐ:ŭş4)0P15ĝD3H3%/ (This is Japanese Natto, please try and eat some.) (Y'N%D2 + (N 84. ģĉäƆ;G/Ħ£Ő8SG)/ (Racial discrimination still exists all over the world. ) U SP'Y 85. ŷ; ƫĢ 7:453K W-4) / (It is a pity that he doesn’t have parents.) (P/Y >R / 86. ěŊX¤ 74% / (Don’t tell a joke.) EY %Y# 87. :ƐKòÜ'.E 4)Q/ (You should refer to that book as well.) 338 !%2 Y #2+S 88. Êð« ØŔĩƧX'3G)/ (The policeman is regulating the traffic at the intersection.) 9YG 8EY $ DY P 89. Ŭij8ŨƐ×: ƇÀX;(GSG'. / (I began to learn Japanese 3 years ago.) 0 9# M'7 90. d %Y;8ŪX. %YƤ13G) / (Kelly is keeping a lot of cats at home.) U W 6 S2 91. ŷ;G/Ĉ:8Ťƨ4)/ (Although he is young, he is independent. ) W.' 0N$ (Y 8EY $ Y( 92. û ; Őâģ7:4ŨƐ×:®Ā;M)4)/ (Since I am a Chinese, Japanese kanji is easy.) U '$5 !!Y ?YF 0P 93. ŷ;öÿ:ÈÒ ŽƋ7:4G/Œ8;7UG+Y/ (His working experience is scarce, so he cannot be a manager yet.) W.' G80 TG >R / 94. û ;ƑŨ ć X¤3ł¨8ƥG)/ (I drive to uni everyday.) 2$ 95. 7.;K15ľ:ģ:ŜàKÜ.Ɗ 5úG)/ (I think you had better think more about other people’s convenience, too.) '. K: 96. ƖŨ'P8ůƂ8Ý G'P / (Let’s go shopping together tomorrow.) G 339 APPENDIX B 340 341 APPENDIX C 342 APPENDIX D TRANSLATION TASK Please translate the following Japanese sentences into English, and write in the bracket given. If you cannot translate the whole sentence, please try to translate where you can. . 4 "3( . 1.%f:"h (urgent)$y} 9`¥ 7$ ¤ -b "r * () 2 2.d$j"TCWJ D]F% %5*: () 2 : : 2 3.t$OK<";V[? $ 9 * () "): : "): 2 4. ^le SUEK=9 p %$ 4 * () 0"0 : 5.a$ (income)$6 .! ku 5*: () . : 21 -$ 6. b x (bankrupt)$ £ $ m 94!7&!5*: () #: 7. 2000"CMN >[]PIA cw 7* () * +0 8.T=A:%t @WC:"¡ () 2 :0 ' 9. t$AZE %q(abortion)$ " L 343 APPENDIX E ů¿tįûDåĀ Ĥā±đŲó9ĤšśĠĤtŔk«ŁĬt\ĀÓĢŰŊ´t{a9ĤĪĞĘ{aŨĢkŗ tyĽľÎÑ»ŨĢR¾ŁûŁĬõŗk«tÍáŢi9ãŢčŌªSŀĠŠlłÀt_±şāŐĠ ãtÝŗāŨĢ9ãĒtčŔĻĩ¶PeäÛ9ãtĿÛAzūœµwŤ¶ªmãt©ĻŨùð9 öãŧř_±ŠlłÀĵÞAî©oŌĮčŔģē[·±đįû3/2)*)*3/+$3/)/3-$*,%/- 9 öã©Šl{aAÂĉÚãęō_±ŠlłÀĵÞ9öã]ř_±ŠlłÀA îùhŠœµ9ĹĹ9 A. yĽľ`ă îŒErIÎéÄŨ¥ıtktyĽľ9öãÅsÄŨ¥ıkYİÑŠktŮyĽľAãÉ ŌŀŊéćö:;9öãÅsÄŨYİÑŠkt¹vyĽľAãÉŌŀŊ¹vté ć>EBFvyĽľAHBIyĽľ?9 @@yĽľ>ťktŮŘÿŮ®Ttōċ?9ÐöAöãSņïĉg:d;ŭtyĽōŎŒ Aã¢ÉâĴr:Ĥ ds¢QÑ;]ªðĴr“ĤśØÒŁtü£dÑ¢Ìě”9“Ĥds¢QÑ”U“ĤdÑ¢Ì ě”ÃyĽľAŐĠ:d;Š}ktŮyĽ®TtōŎāMýħ÷ËŨ9ęńAëĎtkŇ ŔŖŢIJőtyĽŒ 9îŒãtţÅçŌĮÄŨt¥ıktyĽľ>ÉŌŪþŒIJęté ć?9 ÐÄĤds¢QÑ9< Ħ^ޤ]ņğÑdŅ9< ŠũŦ]dÜĈ9< Ĥtåèŕ¢Kdn9 344 C“È” 2.1. öěĚt¦AőðÈÈŋûU¹9< 2.2. įûś³ÈzĂ9< 2.3. ĤúNê³AĤÔÔOĤÈū9< 2.4. ŠāõTtà~AîbbÈ9< 2.5. ĤôĠĎÈ]ìøėįû9< 2.6. ĤÈĎ]ªÎÑ9< 2.7. ð]ðXņÈÚĕtĕí]9< 2.8. ĥĶ¼tKāÈĶĉ9< 2.9. ĮlŚÎÈã9< 2.10.ĶĺAÈc=< ãtR¾ îś×ģē¤ĖĸšñtćÕÃĔĉÚ9 gûwx Ũ ďĜ ij ëĎ>îÃĔĉÚ?< Ýŗ ŗ |¦ ëĎ>îÃĔĉÚ?< ãĶŁVʼntwx Ũ ďĜ ij ëĎ>îÃĔĉÚ?< ãfŨVʼntwx Ũ ďĜ ij ëĎ>îÃĔĉÚ?< ãŨVʼntwx Ũ ďĜ ij ëĎ>îÃĔĉÚ?< ãqŁVʼntwx Ũ ďĜ ij ëĎ>îÃĔĉÚ?< ã¬sĊĀŁġtwx Ũ ďĜ ij ëĎ>îÃĔĉÚ?< ãŜŁõŗÖJ āÁśßÏJ< ĹĹãt¡ű= !/2)*)*3/ $3/>±đŲó? $24$, '%341'1 )$.&*&$3' '0$13-'.3/($0$.'2' /1'$. 34&*'2 $%4,38/(132 /%*$, %*'.%' ".*5'12*38/( '6 /43)#$,'2 #4231$,*$ !', $7 345 ÔòA: c.ħ÷c8JcPĪæîėÉÑĔ%· ¢Ĥàēc`ÅÍÿē %æć rxo ú %Øė:ėÉ'ćæî^ī ÊW {*ôËęÌ¢i».G W9ć %W×m{P¨ÀºDWwDĀç?I_ÀôÃWH [/ MWbĬ».ÌQiê=ĒHâ*{P½ĦO>+.ħA:([email protected])M WHÖÌQēÕĉ}WCï».ÌQiê=MW Ĭ».ÌQiĒ[/Ì¢ç? ĠĠ . rxÝü Ē;1s9oåU4 ÞđærxMWĭÂ4 ÞđæÆ Ì¢æеrxW0* ę ¢µåU ČM9 MWĭÂ4 Æ Ì¢æþXrxW0*ę ¢þXåU Č1-3Xrx7-9µrx rxæФvyп7ïnMMWËhĉ-F Nrïøy; W0³ðscFÂā R ô/ðscJ¾¯ė©£F KĚ cF ā (cF KĚ eprxGF Ì¢¹æÐry¿7ïød | Cčq,æPij Ğr;Ē;W ĭZģ*4 Þđærx0 *Öl;CåU Č n4 cFÂā 9 !ě a FĨ 6 Ìą² FĖ) 3 c~"ñFĕ cJ¾¯ė©£F KĚ 1. è 1.11. Ì¢zéĎĒèá 1.12. Ē è° 1.13. J@đóI&ݪăÈö LġèĘ(þL 1.14. Û'c¦ùè¢ĆôĒWo 1.15. ÁA:ÀoĂðSè¢ 2 1.16. Ù¶ãf ėÛèė 1.17. ]uYÓįèk 1.18. Ē6èĪ35 1.19. ÛÀBQ1èôĮĒ¦¸5-§ 1.20. )è Òt¼ 2 2.11.MĚßýĞû/ĥ:(þL 2.12.ÎA:JzéĀä 2.13.c®Ûécííĝc 346 2.14.Ì'7±jĒĜĜ 2.15.cĈ, ´ B. WØ ĒJg¢½ĦìĊ8ďČĐypıĉ} 1. -:IĢ :(1)3IJ (2) ¡Ü (3) ¾Ï(4) q, Ēpıĉ} 2. 9ć: (1) ¾ć (2) ċý (3) q, Ēpıĉ} 3. WėÇõIĢ: (1)3IJ (2) ¡Ü (3) ¾Ï(4) q, Ēpıĉ} 4. W ÇõIĢ: (1)3IJ (2) ¡Ü (3) ¾Ï(4) q, Ēpıĉ} 5. Wµ ÇõIĢ: (1)3IJ (2) ¡Ü (3) ¾Ï(4) q, Ēpıĉ} 6. WėÇõIĢ: (1)3IJ (2) ¡Ü (3) ¾Ï(4) q, Ēpıĉ} 7. WğÂĄėTIĢ: (1)3IJ (2) ¡Ü (3) ¾Ï(4) q, Ēpıĉ} 8. WÚėÿ'ćë: (1)\ (2) J¥ùK ĠĠWEV ******************************************** Toshihito Kato .ħ÷ Casual Lecturer & Ph.D. Candidate Department of Japanese & Korean Studies Faculty of Arts & Social Science University of New South Wales, 2052 NSW Australia Tel: 61(2) 9385-3763 Fax:61(2)9385-3731 ******************************************** 347 APPENDIX F 348 349