Study on Cross-Language Transfer in Phonological ...

10 downloads 0 Views 5MB Size Report
2.2.1 A Case Study of Cross-Language Transfer between two Non-Native .... It is considered that learning another non-native language will ...... bawm /bɔːm/.
Study on Cross-Language Transfer in Phonological Knowledge between Non-Native Languages: A Case of L3 Japanese Effects on L2 English Vowel Length Perception

School of Knowledge Science Japan Advance Institute of Science and Technology Tianjiao Wang March, 2016

Copyright © 2016 by Tianjiao Wang

Master's Thesis

Study on Cross-Language Transfer in Phonological Knowledge between Non-Native Languages: A Case of L3 Japanese Effects on L2 English Vowel Length Perception

1350007

Tianjiao Wang

Examiners: Professor Takashi Hashimoto (Main Examiner) Associate Professor Huynh Van Nam Professor Taketoshi Yoshida Associate Professor Hideaki Kanai

School of Knowledge Science Japan Advanced Institute of Science and Technology February, 2016

Copyright © 2016 by Tianjiao Wang

Contents

Contents

i

List of Figures

vii

List of Tables

x

1. Introduction

1

1.1 Research Background……………………………………………………………….…5 1.2 Research Purpose………………………………………………………………..….…5 1.3 Research Method…………………………………………………………..……….…5 1.4 Structure of This Thesis…………………………………………………………….…5 2. Literature Review

6

2.1 Cross-language Transfer between Two Non-native Languages……………………....5 2.1.1 Second Language (L2) and Third Language (L3)………..…………...……...…5 2.1.2 Cross-Language Transfer (CLT) of L3 Learner…………………………………5 2.1.3 Direction of CLT……………………………………………………………..….5 2.2 Previous Studies of Cross-Language Transfer between two Non-Native languages …5 2.2.1

A Case Study of Cross-Language Transfer between two Non-Native languages in Grammatical Domain………………………………………………………..5

2.2.2

A Case Study of Cross-Language Transfer between two Non-Native languages in Syntactic Domain..…………………………………………………………..6

2.3 The Factors Underlying CLT………………….……………………………………..16 2.3.1

Similarity and Psychotypology..……………………………………………..16

2.3.2 Perceptual Linguistic Salience………………………………………………….19 3. Transfer of Vowel Length Perception from Japanese as L3 to English as L2

29

3.1 Perceptual Assimilation Model (PAM)…………………...………………………….29 3.1.1

What is Perception? …………………...………………………...…………...29

i

3.1.2

Speech Perception and Phonological Knowledge Learning………………….29

3.1.3

Category perception……………………………………….………………….29

3.1.4

Perceptual Assimilation Model……………………………………….………29

3.1.5

Necessity to Clarify Possibility to Extend PAM……………………………...29

3.2 Vowel length in Chinese, English and Japanese……………….…………………….30 3.2.1

Importance of Vowel in Language Learning…………………………………29

3.2.2

An Overview of Chinese English and Japanese Vowel Length………………29

3.2.3

Vowel Length in Chinese……………………………………………………..29

3.2.4

Vowel Length in English……………………………………………………..29

3.2.5

Vowel Length in Japanese…………………………………………..………..29

3.3 Research Hypothesis………………………………………..………………………..32 4. Preliminary Experiment

43

4.1 Experimental Design……………………………………………...………………….43 4.2 Participants………………….………………………………………………………..44 4.3 Experiment Procedure…………………………………………………..................…46 4.4 Results……………………………………………………………………………...62 4.4.1 Result 1: Identification Accuracy across the Two Groups………………...…...62 4.4.2 Result 2: Other Factors May Influence the Identification Accuracy……...…...62 4.5 Discussion……………………………………………………………………………63 5. Main Experiment

65

5.1 Main Experiment……………………………………………………………………..65 5.1.1 Experiment Design……………………………………………………………..65 5.1.2 Participants……………………………………………………………………..65 5.1.3 Experiment Procedure………………………………………………………….65 5.2 Results………………………………………………………….………………….…66 5.2.1

Result 1: Accuracy of Identification Task……………………………………66

5.2.2

Result 2: Relation between Participants’ Age and Accuracy of Identification Task…………………………………………………………………………...66

5.2.3

Result 3: Relations between English Learning Year of Participants, Score of

ii

CET Test, Accuracy of Identification Task…………………………………..46 5.2.4

Result 4: Relation between Word Familiarity and Accuracy of Identification Task

6. Discussion

86

6.1 Reconsideration of Research Questions and Hypothesis………………………….…66 6.2 Limitation of This Research…………….……………………………………………71 6.3 Generalization for Education………………………………………………………...74 7. Conclusion

86

7.1 Summary of the Research…………………………………………………….…66 7.2 Conclusion………………………………………………………………………….66 7.3 Limitation of This Research…………………………………………………………71 Acknowledgements………………………………………………………………………...…56 Reference………………………………………………………………………….……57 Appendices……………………………………………………………………………..…58

iii

List of Figures

Figure 4-1. The interface of the identification task’s application…………….………………..7 Figure 4-2. Process of sound material generation………………………………………….…8 Figure 4-3. Two groups mean identification accuracy for the English and Japanese stimuli...10 Figure 4-4. Time used in identification task across the two groups…………………...……..17 Figure 4-5. Mean English stimuli replay times across the two groups……………….………23 Figure 5-1. Flowchart of the experiment……………………………………………………25 Figure 5-2. User interface of the identification task application…………………………...…32 Figure 5-3. Mean of the accuracy rates………………………………………………….……36 Figure 5-4. Mean of participants’ age …………………………………...…………………38 Figure 5-5. Accuracy rates of word types between below and above the mean age………38 Figure 5-6. Mean of English learning years………………………………………….......…...41 Figure 5-7. Accuracy rates of word types between below and above mean English learning years…………………………………………………………………………………………..48 Figure 5-8. Mean of CET score…………………………………………………………..51 Figure 5-9. Accuracy rates of each group in different value of familiarity…………..53 Figure 6-1 A model for describing cross-language transfer between non-native language…..73

iv

List of Tables

Table 3-1 Comparisons of the corresponding vowel length in Chinese, English and Japanese.………………………………………………………………………………….…..14 Table 4-1 The English and Japanese words used in the preliminary experiment…..11 Table 5-1. Stimuli used in the experiment…………………...…………….…..20 Table 5-2. Pearson’s correlation coefficients between the English learning years and the accuracy rates of four word types………………………………………………………..24 Table 5-3. Results of word familiarity in control group and experimental group...………26 Table 5-4. Result of Word Familiarity of Familiar/Unfamiliar Group in Control and Experimental group….……………..…………………………………………………………34

v

Chapter 1 Introduction In this study, we aim to demonstrate the cross-language transfer between non-native languages. Specifically, we concentrate on the case of the cross-language transfer of learning L3 Japanese influences on the phonological perception of L2 English. In addition, we propose to supplement perceptual assimilation model (PAM) which is a theory to explain how native speakers perceive non-native speech sounds. In this chapter, we introduce the research background, research significance and purpose, as well as the overall structure of this thesis.

1.1 Research Background With the wide spread of English in the world and the increasing mobility of the world's population, learning two or even more than two foreign languages is more and more necessary and universal (Jessner, 2006) Learning a third language (L3) is a complex process since learners of the third language already know two languages and there exists cross-language transfer (CLT) among three languages. It is considered that learning another non-native language will develop some new skills (such as language learning and language maintenance skill), thus would make the improvement of the quality of the leaners’ whole language system (Herdina & Jessner, 2000). The studies on CLT have been focused predominantly on the mother tongue effect on the target leaning language while the other relationship (e.g. transfer between non-native languages) has not got enough attention (Cenoz et al., 2001). For understanding the nature of language knowledge and the essence of cross-linguistic influence, it is important to study the transfer between non-native languages. Chuang (2002), Cheung (2011), and Griessler (2001) indicate the possibility of

1

non-native languages transfer, specifically, from L3 to L2, in the domains of grammar and lexicon. While study on the transfer of phonological knowledge between non-native language is needed for learners to enhance their language intelligibility (i.e. avoiding the communicative misunderstandings, conveying ideas correctly), such study is more challenging because phonological knowledge is likely to be more tacit than other linguistic knowledge (e.g. language users are often not consciously aware of how they produce and perceive specific sounds). The perceptual assimilation model (PAM) is a theory that accounts for how native listeners perceive non-native speech sounds (Best, 1995). According to PAM, the listeners assimilate non-native sounds to the most similar native phonemes, and also they discriminate the non-native contrasts regarding the extent of similarity with the native sound category. Although many studies have provided the evidence for PAM to support the validity in second language learning situations, it is still unclear that whether this model can still be applied to the transfer between non-native languages.

1.2 Research Purpose In the present study, we concentrated on the cross-language transfer between two non-native languages in phonological domain which the previous studied seldom to be touched, and aim to understand how the adults’ speech perception changing as learning another non-native language base on the exiting theory. After reviewing the existing theory and the characteristics of the investigated target, we set the following three specific questions. i.

Whether does learning L3 influence the phonological perception of L2?

ii.

If the influence exists, does this influence give a positive effect to the phonological perception of L2?

iii.

Whether can PAM be applied to the influence of non-native languages? This influence will be able to be observed if the second and third languages share

some common characteristics whereas native language does not have. Therefore, we 2

chose the vowel lengths of Chinese (L1), English (L2), and Japanese (L3) as an investigative target because the contrasts of short and long vowels exist in English and Japanese but not in Chinese.

1.3 Research Method Generally, according to PAM, a category that is in L1 but not in L2 tends to be assimilated into another category in L1, which causes the low identification correctness of L2. We suppose that there are sound categories of vowel length perception in L2 and L3, which are shared between L2 and L3 to some extent and are different from any categories in L1. In addition, we hypothesized that learning L3 will lead the transfer of the phonological knowledge for identifying these internal categories. Thus, our hypothesis is the following: Hypothesis: Learning L3 (Japanese) has a positive effect on the L2 (English) vowel length perception. That our hypothesis was positively proved means that the category of vowel length in L2 is created rather than be assimilated into L1. In order to assess how learning L3 (Japanese) influences the L2 (English) vowel length perception, we conducted the questionnaire, an experiment consisting of identification task, and word familiarity task. At first, we surveyed the tasks used in the previous studies about assessing vowel length perceptions. In speech perception studies, the discrimination task and identification task are often used to assess if a listener could distinguish the differences of stimuli. These two tasks can reflect the different cognitive load in the activity of perceptual process. The discrimination task reflects the ability to perceive the sensory differences of the sounds (McGuire, 2010). The identification task reflects the ability to perceive the sound according to the internal categories (Tsukada, 2011). Because we suppose that the ability to perceive sounds depends on internal categories and we aim to investigate how the internal categories are modified by leaning non-native languages, 3

the identification task is selected in the present study. In order to assess the other influential factors that would affect the identification accuracy, we conduct the questionnaire and word familiarity task between two groups of participants.

1.4 Structure of This Thesis This thesis consists of seven chapters as follows. Chapter 2 contains a survey that reviewed the literatures on Cross-language Transfer, the characteristics of the third language learners to position this research in the studies of cross-language transfer. Furthermore, the several researches related to the influence factors to phonological transfer were reviewed for understanding how learning L3 influences on the perception of L2. Chapter 3 gives an account on the issue formulation of this research. Firstly, we introduce an influential model (Perceptual Assimilation Model) of human speech perception. Secondly, the necessity of to extend this model to two non-native languages transfer is introduced. Thirdly, the reason of selecting those languages as research targets is explained by describing the language characteristics of Chinese, English, and Japanese. At last, we present our research hypothesis. Chapter 4 presents our basic idea of the experiment design including the task of application, sound material, etc. In order to test the validity of our experimental design, we conducted the preliminary experiment with the cooperation of a small number of participants. Then, we discussed the results and proposed some settings which is needed to improve in the main experiment. Chapter 5 presents the main experimental design. Chapter 6 discusses the relationship between the accuracy of identification task and the influence factors of the type of tasks, participant’s age, their familiarity to the English words used in the experiment. In addition, the potentiality of this research for understanding the direction (especially backward transfer) of phonological transfer 4

between non-native language has been elaborated. Moreover, the language teaching strategy and methods for realizing the English vowel length has been proposed for linking the research findings to the educational practice of foreign language in the perspective of Knowledge Science. Chapter 7 explains the conclusion. Then the limitations of the present study are pointed out and the suggestions for the future direction of this research were provided.

5

Chapter 2: Literature Review

Chapter 2 Literature Review In this chapter, firstly, the important technical terms (such as, L2, L3, Cross-language Transfer, etc.) in this research were defined. In addition, the characteristics of the third language learners were surveyed and to provide a foundation for describing the importance of research on the phenomenon of third language leaners. Furthermore, by reviewing of two previous studies about non-native languages transfer in the domain of grammar and syntax, the position of this research in the studies is elaborated. Finally, the underlying factors that are most relevant with this study of cross-linguistic transfer between non-native languages were given.

2.1 Cross-language Transfer between Two Non-native Languages 2.1.1 Second Language (L2) and Third Language (L3) The third language refers to one or more language(s) is (are) currently learned by the languages learners who have already mastered one non-native language. Some of the researchers use the term of second language as an “umbrella” term which instead of third language and even multilingual (Elis, 1994; Larsen-Freemna and Long, 2000), and seldom to distinguish these two concepts. However, Jorda (2005) has pointed out that the "the third language" is a general concept which does not refer to the number of the languages which the language learners have learned, but refer to the one or more languages which the language learners are being learned after their native language and the first non-native language of the learner.

2.1.2 Cross Language Transfer (CLT) of L3 Learner

6

Chapter 2: Literature Review

In psychology field, the term “transfer” refers to the influence of skills, knowledge, and attitude in a situation which skills, knowledge, and attitude are formed. From the 1950s, the term “transfer” starts to be used in linguistic research field, used to refer to the influence of a language in learning another language (Odlin, 1989). Odlin (1989) indicated that transfer is the influence resulting from similarities and differences between the target language and any other languages that has been (and perhaps imperfectly) acquired. Jarvis and Paclenko (2008) have defined the cross-language transfer as “One of the individual’s language knowledge influence on the other language’s usage or knowledge”. According to the definition from Odlin (1989), the language transfer not only refers to the influence of the native language but also the influence of any other language which the learner has learning. Since then, language transfer is no longer restricted to the native language and target language. The transfer within the three or more languages began to be concerned by the researchers. The third language leaning is more complicated than that of the second language learning due to the complexity of cross-language transfer. The second language learners have two potential resources of transfer language systems (L1 and L2). However, there are more than two possible resources of transfer existing in the third language learners’ mind. In the third language acquisition, the other combinations of cross-language transfer are also need to be considered (e.g. L2 to L1, L2 to L3, L3 to L2 etc.). The cross-language transfer is a distinguished feature of third language learning. While the studies of second language acquisition mainly focuses on the transfer from the mother tongue to the target language, which has neglected the other possible language combinations of cross-linguistic transfer. Odlin (1989) believes that when learning the third language or more languages, the individual’s first and second language knowledge and the existing learning experience will inevitably effect on the other languages’ knowledge and the cognitive development. So the third language or multilingual learning with which to have some

7

Chapter 2: Literature Review

characteristics differentiating from the first language (L1) and second language learning, and is worthy of our in-depth study and discussion.

2.1.3 Direction of CLT According to the direction of cross language transfer of which can be divided into, the earlier learned to the afterward learned language that is, forward transfer; afterward learned to the earlier learned language that is, backward or reversed transfer. In resent years, many studies have supported this view: cross language transfer is not only occurs by the forward direction (Cheung, 2011). The unidirectional effects of the third language may also in turn have a reverse impact on their use of the earlier learned language, that is, the reverse transfer or backward transfer. The earlier second and third language leaning studies shows that the earlier acquired language learning experience will have an impact on the generation of the lately acquired language (i.e. forward language transfer), the L1 is relatively fixed, which will easily to trigger transfer. In the resent years, the backward transfer has looked into this issue. Many of the research results have suggested that the afterward learned languages also could be the sources which contribute to the earlier learned language leaning. One research area related with backward transfer is the “Language attrition”, also known as "language loss", means the bilingual or multilingual, reduce or stop to use that language, that of the language ability will gradually subsided with the time pass by. For example, in the native language environment, dialect attrition caused by leave the dialect districts. In the non-native language environment, native language attrition caused by immigration, dialect attrition caused by leave the nations (Van Els, 1986).

2.2 Previous Studies of Cross-Language Transfer between two Non-Native languages 2.2.1 A Case Study of Cross-Language Transfer between two Non-Native 8

Chapter 2: Literature Review

languages in Grammatical Domain Cheung (2011) studied on the backward transfer form L3 German to L2 English among native Chinese (Cantonese) in the perspective of grammar aims to contribute to this research area. There is no past tense in Cantonese language. The Chinese L2 English learners used to make errors in refer to the past situations because the native language influence. In German language there are two ways to indicate the past time situations, one is Präteritum, and the other is Perfekt (the grammar structure is similar with Present Perfect Tense in English). There is an overlap of German Perfeke and English present perfect tense. The German counterpart of the English past simple tense used to indicate past situations in the Präteritum. Through conducted a narrative writing task and an acceptability judgment task. It is found that L3 German participants were significantly more tend to use English present perfect tense in past situations just like the German perfect and apply them in English writing than L2 participants. And the L3 learners were less likely to use the English present simple tense with reference to the past situations, which is a typical error affect of L1 Chinese. It proved that the occurrence of the positive transfer form L3 to L2 in express past situations.

2.2.2 A Case Study of Cross-Language Transfer between two Non-Native languages in Syntactic Domain Contrary to Cheung (2011), another study from Chuang (2002) also studied on the backward transfer form L3 German to L2 English among native Chinese (Cantonese) in the domain of syntactic. This study aims to investigate whether learning third language German could facilities the second language English learning. In the experiment, the participants were splited into two groups of Taiwan students (1) A group of participants with English as L2 German as L3 (2) Another group of participants who learn only English as L2. The experiment focused on the how the two groups of participant use the English relative clauses. Therefore, three tests (grammaticality judgment, translation and

9

Chapter 2: Literature Review

sentence-combination) were employed by 4 grammatical categories of English relative clauses. However, the statistical results showed that as German as third language did not assist learning English as second language, which still be influenced by the first language Chinese.

2.3 The Factors Underlying CLT Several factors (such as the distance of two languages, language proficiency, recency of use, multilingual competence, metalinguistic awareness and the language learning order etc.) have been reported to correlate with cross-language transfer in the phonological domain (Williams & Hammarberg, 1998). In this section, the factors of “Psychotypology” and “Linguistic Salience” are selected as the determining factors to cause the transfer between two non-native languages.

2.3.1 Similarity and Psychotypology To perceive the similarities between the target language and the other language is an essential process in cross language transfer, and the intralinguistic similarities will improve the target language learning (James, 1980). Kellerman (1977) asserting that cross-language similarity is relevance to how the language learner perceived distance of the resource language and received language. The “similarity” is difficult to be defined since that of the criteria is differing from one to another language. The concept of “Psychotypology”(Kellerman, 1983) is separated from the objective language distance (or linguistic typology). Kellerman indicated that “Psychotypology” is the “language that is perceived as typologically closer”. (Cenoz, Hufeisen and Jessner, 2001.) The language knowledge whether be transferred is determined by the perceived distance of the learner between the resource language and received language rather than the actual language distance. For example, “the learners of French and English natives 10

Chapter 2: Literature Review

of a non-IndoEuropean language tend to transfer the vocabulary and structures of language knowledge from the other Indo-European language they know rather than their native languages” (Cenoz et al., 2001). The more similar of the common characteristics of the resource language and the received language shared the more prone to generate transfer. This kind of common composition is both determined by the features of the resource language and the received language. Only when the two languages have the same psychological structure, the original language could be transferred to the target language learning.

2.3.2 Perceptual Linguistic Salience In general, salience refers to the relative importance or prominence of signs. It is considered to help human to give an attention to the importance parts by ranking large amounts of information. The perceptual linguistic salience is the degree to which it pops-out from the languages or parts of languages and captures the attention. These effects of perceptual linguistic salience are regarded as a trigger to the language learners to hear some sounds more easily due to a certain characteristic of language (Schwartz, Jean-Luc, et al., 2005) In this research, the author considers that the vowel length perception is related to one of the perceptual linguistic salience, because for Japanese speakers, vowel length plays an important role to determine the meanings of a word and to the Japanese learner it is also relatively easier to identify comparing with the vowel length in English or other languages. Therefore, considering about this important factor, the language transfer from L3 Japanese to L2 English is expected to occur.

11

Chapter 3

Transfer of Vowel Length Perception from Japanese as L3 to English as L2 The previous chapter surveyed the related terms and literatures on two non-native languages cross-language transfer. In this chapter, in order to give an account on the formulation of this research, the following two parts are explained beforehand. The first part is to describe the important theoretical phonological transfer model (Perceptual Assimilation Model) between native language and non-native language for introducing the necessity to establish a mechanism of phonological transfer by clarifying the possibility to extend the existing model. The second part is to describe the language characteristics of Chinese, English and Japanese to explain the reason of selecting those languages as research targets. Lastly, the research hypothesis was stated.

3.1 Perceptual Assimilation Model (PAM) 3.1.1 What is Perception? Perception refers to the process that the individual draw upon the sensory organs, nervous system and brain detection to interpret the information obtained from the outside world. Human perception can be divided into visual perception, auditory perception and speech perception. Speech perception is the process of extracting the corresponding linguistic elements from the continuous speech sound signal. (Matsumoto, 2009). When the listener heard the sounds and words, they usually compare the language stimuli that be heard in their minds of the language. This criterion is based on the experience of the listener (Kreiman et al, 1993).

3.1.2 Speech Perception and Phonological Knowledge Learning Speech perception refers to the process of speech recognition that the auditory organ transmits the sound waves (Wang, Li 1992). The listeners who receive the stimulus 12

firstly carries on the preliminary analysis and then find out the phonetic phonology characteristics encode on the basis of the relevant knowledge of speech signal in the memory system to complete the speech recognition. Phonology knowledge plays an important role in speech production and perception. It refers to “the knowledge of the sound structure of language”. Unlike the morph-syntax knowledge, the speaker is able to know how to analyze the component parts of the sentence, word order of the phrases and word formation in accordance with some conventional rules. Phonology knowledge is more tacit. We cannot directly know the underpin system and the rules governing the system (e.g. it is difficult for us to say the phoneme /f/ is created by the upper teeth touch lower lip lightly to force the air over the surface of the lower lip and into the upper teeth.) In the phonological knowledge learning, the language learner is the listener as well as the speaker, since it is necessary for them to carry out the speech perception and speech production. When the language learners speak a language, the speech production causes their phonological perception. Through judging their own phonological perception, they further adjust their following speech production. So this cycling process leads the proficiency of foreign language learners continue to be improved. Thus it can be seen that, the speech perception plays an important role in the target language learn. Only if the learners conduct the correct perception, they could possible to achieve the successful speech production. When the learner being as a listener to perceive the speech sound in the communication, she (or he) will make the judgment on the speech, which will also affect the speech production.

3.1.3 Category perception Categorical perception (CP) is one of the important models of speech perception. It refers to the continuum of speech stimuli which are perceived as discrete, a limited number of categories (i.e. Phoneme). Specifically, the perception of the individual is not consistent with the physical boundaries of the speech sound. The listeners are difficult

13

to detect the speech differences within the category, but very sensitive to the differences between the categories (Liberman et al. 1976). It is need to be explained that CP has been considered as a speech-specific phenomenon (Liberman et al., 1961) in the early stage. As the deepening of the research, since the 1970s, it was found that CP is not a unique phenomenon of speech perception, but also existing in a general domain. Miller et al. (1976) found that the subjects are presenting categorical perception in change the buzzing sound of VOT synthesized continuum. Kuhl et al. (1983) found the monkeys to the sound perception are also presenting categorical perception. All of these findings indicate that CP is not unique to speech perception, and is not unique to humans, but may widely existing in the cognitive model. Since the 1970s, as the study of category perception extensively developed in early childhood speech acquisition, the speech scientists found that as the same as the animal to the human speech perception, from birth to one-year-old, the infants could distinguish the majority of the languages’ segments, however, after one-year-old, the infants could only perceive some individual sound segments.

3.1.4 Perceptual Assimilation Model The term assimilation refers to how the human interprets new experiences in terms of their current understanding. The assimilation is conducted to place the new information in terms of the previous schemas. Perceptual assimilation model (PAM) is an influential model to explain how native listeners perceive a non-native sound. It posits that the construction (phonological system) of native language (L1) plays a crucial role in native listeners perceiving non-native sound, and takes the articulatory gesture of native language sound as the fundamental, acclaimed that 1) the non-native sounds are that the articulatory gestures were not overlap with the native sounds, 2) the non-native language segments tend to be perceived according to their similarities with the native language segments. The similarities and differences between different speech sounds should be determined by the language learners to be able to perceive the sound

14

characteristics of the target language. The second language speech sounds and that of the distance with the closest to the native language speech sounds are determined by the extent of the difficulty to distinguish two L2 phonemes. These two sounds are difficult to be discriminated by the second language listeners when the two speech phonemes are similar to one native language phoneme. When the native listeners perceive the non-native sound, basically the following three sorts of situations will be encountered (Best, 1995). I. be same or similar to a certain category of native language II. be similar to the sounds in native speech sounds but cannot be categorized as a native language category III. the nonnative sound cannot be found in the native language category, that is, these kinds of sounds is considered to be the non-speech sound in the native language to the non-native listeners According the above three situations, the non-native sounds could be perceived as the following patterns (Best, 1995). a)

Two-Category Assimilation (TC type). Both of the two non-native sounds are assimilated as a different native sound category, therefore, discrimination is supposed to be excellent.

b)

Single-Category Assimilation (SC type), two non-native speech sounds are assimilated to the one same native sound category, therefore, discrimination is supposed to be poor.

The perceptual distance between the two and the closest to the native speech determines the difficulty of distinguishing the two speech sounds. When two sounds elements are similar to the same native phoneme, the two sounds are difficult to be distinguished. For example, English phoneme /r/ and /l/ are both similar to the native phoneme /r/ in Japanese language, so when Japanese native speakers in learning English pronunciation and it is difficult for them to distinguish English / r / and / l /. When all the phonetic categories of the second language and native language are

15

completely different, the second language phoneme is most likely to be perceived by the native listeners, for example, in Zulu language, the mouth click sound in English does not exist, English native speakers have never had contact with the Zulu language nevertheless can also distinguish the Zulu click sounds contrasts (Best, 1995). Best et al. (1988) summarized that there are mainly two reasons for that “(a) linguistic salience: The Zulu click contrasts are easily to be distinguished since the acoustic characteristic of quite salient. (b) the zulu click contrasts fall outside of the English phonological space and be considered as the non-speech sounds”.

3.1.5 Necessity to Clarify Possibility to Extend PAM Although the theoretical modes of PAM for describing the mechanism of phonological influence between native language and non-native language have been proposed, there is still not any mechanism to show how phonological influence happens among non-native languages in the third language leaners’ mind. The resent researches on the speech perception have indicated that the human perceptual system is adaptable and adjustable as the changes of the sensory input throughout the whole process of life. Therefore, in the Multilanguage learning process, including L1 L2 and L3 language speech perception and phoneme setting may restructure the acoustic perceptual space encompassing each of the language system (Leather&, James 1996). No matter to the speech perception or the output of production, such a realignment of perceptual space may make the monolingual parameter changed. The results lead to the multilingual users no longer possesses the same acoustic perception space with the monolingual users. Therefore, the current model about explaining the cognitive learning process is not adequate to be discussed with the third language leaners’ issues.

3.2 Vowel length in Chinese, English and Japanese We categorize vowel length as either short or long vowel. The phonological systems are various in different languages. For some of the languages like English, Japanese use 16

vowel length contrast systematically to convey different meanings of the word. Especially, vowel length contrasts are fundamental in Japanese language of convey lexical meaning (Ingram and Park, 1996), (large duration difference of long and short vowel).

3.2.1 Importance of Vowel in Language Learning Correctly product the vowels is an important part of phoneme learning. According Prator and Robinett (1985, LinMa, p25), “the essential meaning of the word is carried by vowels. One wants to learn a foreign language or the words to be understood by the others must be able to distinguish the vowel system of the language, and to reflect the differences in the pronunciation of these vowels.”

3.2.2 An Overview of Chinese English and Japanese Vowel Length The characteristics of vowel length contrasts in Chinese, English and Japanese are described in Table 3-1.

Language Vowel Length Example

Chinese

English

Japanese

Non-contrastive

Contrastive

Contrastive

-

Ship [ʃip] vs. Sheep [ʃiːp]

(ryokou) vs. (ryookou)

Table 3-1 Comparisons of the corresponding vowel length in Chinese, English and Japanese.

3.2.3 Vowel Length in Chinese The vowel length belongs to the non-distinctive phonetic features in Chinese Mandarin. Mandarin words are not simply resort on the vowel length to signal the phonemic contrasts, and the differences of the vowel length of the speech are often attached with other phonetic features. For example, the non-light tone Mandarin pronunciation is long,

17

wide range, relatively complete and sound, but the light tone Mandarin pronunciation is exactly the opposite. Chinese Mandarin has five vowel phonemes (Cheng, 1996, Svantesson, 1984), they are /i/ /y/ /e/ /a/ /u/.

3.2.4 Vowel Length in English The vowel system of Chinese and English are differing considerably, both in the vowel inventory, vocal position and the duration cues. The most apparent characteristic of Chinese an English is vowel length. In English, there are differences in the meaning of the vowels, so English is a language that belongs to the duration of distinctive features in words, such as, the word of eat /i:t/ and it /it/, fool /fu:l/ and full /ful/.

3.2.5 Vowel Length in Japanese Vowels in Japanese contrasts in length, e.g. Koto which means a “Japanes harp” Kooto which means “coat”. Vowel length contrasts are fundamental in Japanese of convey lexical meaning (Ingram,Park ,1996). The previous studies have shown that duration is the major acoustic and perceptual correlate of vowel length contrasts in Japanese language (Hirata and Tsukada, 2009).

3.3 Research Hypothesis As mentioned in session 3.1.4, according to PAM, the listeners assimilate non-native sounds to the most similar native phonemes, and also they discriminate the non-native contrasts regarding the extent of similarity with the native sound category. Although many studies have provided the evidence for PAM to support the validity in second language learning situations, it is still unclear that whether this model can still be applied to cross-language transfer between non-native languages. Therefore, it is important to known the fact about phonological backward transfer between non-native languages and also to better understand its mechanism. In this paper, we aim to show how learning L3 influences the phonological perception of L2. This 18

influence will be able to be observed if the second and third languages share some common characteristics whereas native language does not have. Therefore, we chose the vowel lengths of Chinese (L1), English (L2), and Japanese (L3) as an investigative target because the contrasts of short and long vowels exist in English and Japanese but not in Chinese. In order to assess how learning L3 (Japanese) influences the L2 (English) vowel length perception, we conducted an identification task. In the identification task, participants were asked to identify the category of sounds. This task reflects the ability to perceive the sound according to internal phonological categories (Tsukada, 2011). Generally, according to PAM, a category that is in L1 but not in L2 tends to be assimilated into another category in L1, which causes the low identification correctness of L2. We suppose that there are internal categories about vowel length perception shared between L2 and L3, which are different from the categories in L1. In addition, we suppose that learning L3 will lead the cross-language transfer of the phonological knowledge for identifying these internal categories. Thus, our hypothesis is the following: Learning L3 (Japanese) has a positive effect on the L2 (English) vowel length perception. If the hypothesis was positively proved means that the category of vowel length in L2 is created rather than be assimilated into L1.

19

Chapter 4 Preliminary Experiment The previous chapter demonstrated the account of the non-native cross-language, and proposed our hypothesis. In order to test our hypothesis, in this chapter, the basic ideas of experimental design were carried out to investigate how the two groups of listeners perceive the English and Japanese vowel length with the cooperation of a small number of participants. Then, we presented the data analysis and results. Lastly, we identified the possible problems that needed to be considered to improve in the main experiment.

4.1 Experimental Design In order to test of our hypothesis of whether L3 Japanese could influence the vowel perception of L2 English. We conducted our experiment in Japan with the two groups of participants in different Japanese language proficiency (intermediate and advanced level). Therefore, we surveyed the previous study about how to assess the listeners’ speech perception. In speech perception of previous studies, the discrimination task and identification task are often used to assess if a listener could distinguish the differences of the stimuli. However, the discrimination task and identification task can reflect the different cognitive load in the activity of perceptual process. In the discrimination task, the participants are asked to distinguish a sound from that of another. This task reflects the ability to perceive the sensory differences of the sounds (McGuire, 2010). In the identification task, they are asked to identify the category of the sounds. This task reflects the ability to perceive the sound according to the internal categories (Tsukada, 2011). We suppose that the ability to perceive sounds according to internal categories is enhanced due to backward transfer, therefore, in the present study we conduct the identification task. Before the identification task, we administrated the questionnaire to understand the language background of the participants.

20

The contents of the questionnaire including the basic personal information of the participants, the language leaning years and the experience of residence in English and Japanese speaking countries.

4.2 Participants A total of 8 third language learners with English as L2 and Japanese as L3 of Chinese native graduate students from Japan Advanced Institute of Science of Technology (JAIST) were recruited, with 4 participants who have got the JLPT N1 (Japanese Language Proficiency Test for Japanese Advanced Level) as experimental group, and 4 participants who have got the JLPT N2 (Japanese Language Proficiency Test for Intermediate Level) as control group. Amongst them 7 person have reported their age. The average age for the 4 participants in control group was 26.2 (rang from 23 to 28). The average age for the 4 participants in experimental group was 28.6 (rang from 27 to 28). Among these 8 participants, up to 4 are female and 4 are male.

4.3 Experiment Procedure A. Procedure Prior to the genuine task, the first five trails in each language were used to take practice for the participants. The sample words for taking exercise will not be used in the genuine task. The stimuli were blocked by language (English, Japanese). During the identification task, the participants listened the one sample sound each trail. The task of the listener is to label the single sample sound for each trail (E.g. to click the buttons “short” or “long” to select the one which match with sample sound.). We used an application developed by the experimenter (Fig.4-1) The experiment consist two stages, test 1 English two forced-choice identification task and test 2 Japanese two forced-choice identification tasks. Both control and experiment groups participated in all the two stages.

21

Figure 4-1. The interface of the identification task’s application

B. Listening Material The figure 4-1 indicates how the used listening materials were created. The GhostReader application (B) was used for the target words speech material. This application uses the iVox (A) which is a speech database of real voices recorded by real people. Therefore, we consider that the speech material (C) generated by the GhostReader could provide a relatively natural and less-ambiguous environment.

22

(A)

Record voice and turn it into data

Provide sources for speech generation Generate speech material

(B)

(C) Cart.mp3 Figure 4-2. Process of sound material generation

C. Stimuli The speech file got from GhostReader system that CV1C words in English and C1V1C2V2 words in Japanese (where V1 was either short or long in each language). The discrepancy in the word structure is not ideal since the occurrence of the CVC word structure is quite few in Japanese. The stimuli that were used in this study were shown in table 4-1. It is specifically focus on the 4 pairs of English vowel contrasts /ʌ/-/ɑ:/, /ɪ/-/iː/, /ɔ/-/ɔː/, /u/-/uː/, and 5 pairs of Japanese vowel contrasts /ʌ/-/ɑ:/, /e/-/eː/, /ɪ/-/iː/, /ɔ/-/ɔː/, /u/-/uː/. Thus, a total of 64 stimuli were employed in the study: 4 vowels×2counterpart×3 conditions+5 practice trails= 29 trails (in English), 5 vowels×2counterpart×3 conditions+5 practice trails = 35 trails (in Japanese).

23

Table 4-1 The English and Japanese words used in the preliminary experiment

4.4 Results The analysis of the data obtained from the language background questionnaire and identification task were conducted.

4.4.1 Result 1: Identification Accuracy across the Two Groups The mean accuracy on identification task of participants from both control group and experiment group are shown in Figure 4-3. The X-axis represents the groups (control and experimental group) with the English or Japanese identification test while the Y-axis represents the mean identification accuracy. The control group on the English test is 0.6354.

24

The experiment group on the English test is 0.6667. The control group on the Japanese test is 0.8250. The experiment group on the Japanese test is 0.8000. Due to the limitation of the small sample size, the statistical analysis was not conducted. From the virtual representation of data, the differences of the control group and experiment group on the English and Japanese test were not shown. This result did not support the hypothesis.

1

Accuracy Rate of Identification Task

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Control Group (English)

Experiment Group (English)

Control Group (Japanese)

Experiment Group (Japanese)

Figure 4-3. Two groups mean identification accuracy for the English and Japanese stimuli

4.4.2 Result 2: Other Factors May Influence the Identification Accuracy The figure 4-4 shows each participant’s the time used in identification task and the mean time used of each group. The mean time of control group is 795.5 seconds and that of the experiment group is 601.3 seconds. The figure 4-5 shows each participant’s replay times of stimuli and the mean of each group. The mean replay times of control group are 2.41 and the mean replay times of experiment group are 2.17. By comparing the mean time used and mean listening replay times between two groups, it indicated that the participants from control group take more times and listened relatively

25

much more times than the participants from the experimental group. Therefore, we considered that it is important to control these kinds of factors which might influence the accuracy rate of identification task. 1600 1400

Time (Second)

1200 1000

Mean =798.5s

800

Mean =601.3s

600 400 200 0 C1

C2

C3

C4

E1

Control Group

E2

E3

E4

Experiment Group

Figure 4-4. Time used in identification task across the two groups

3.5

Replay Times (Time)

3

Mean =2.41

2.5

Mean =2.17

2 1.5 1 0.5 0

C1

C2

C3

Control Group

C4

E1

E2

E3

E4

Experiment Group

Figure 4-5. Mean English stimuli replay times across the two groups

26

4.5 Discussion From the data analysis of last sub-session, we found some inappropriate settings of the preliminary experiment. We demonstrated these problems in this session, and improved the following four points in the main experiment. P1: Exposition Effects of Japanese in control group Although the participants of preliminary experiment in the control group had the relatively lower Japanese proficiency, due to their studying and living in Japan, these participants had exposed to the Japanese environment. Considering this kind of exposition could be an influence factor to the vowel length identification tasks, we selected to recruit pure Chinese English speakers by conducting a part of the main experiment in a Chinese university in the main experiment. P2: Order Effects to Language Blocks and Stimuli within Each Block In order to counterbalance the order of Japanese block and English block and the order of stimuli within each block, the order of language blocks and stimuli within each block were randomized by the application used in the main experiment. P3: Constant Practice Effect from Different Replay Times across Two Groups From the analyzing the result from the preliminary experiment, there was a distinction to the replay times across the control and experiment group. To avoid the effect of constant practice (improving the accuracy of vowel length identification tasks by keep listening to stimuli), the maximum relay time was limited up to three times by the application used in the main experiment. P4: Influence from Inappropriate Stimuli Although the stimuli used in the preliminary experiment were generated by an application based on the real voices database, there is still some inappropriateness in stimuli which may influence the results of identification tasks. In order to reduce this

27

influence, the stimuli were evaluated by the English and Japanese native speakers and the problematic ones were replaced in the main experiment. P5: Add the Non-Word Stimuli for English and Japanese The criteria of participants recruiting, the modified English word stimuli and added English and Japanese non-word stimuli, and the improvements of the application, would be elaborated in the next chapter.

28

Chapter 5 Main Experiment The previous chapter presented the basic ideas of experimental plan and design for investigating how the two groups of listeners perceive the English and Japanese vowel length, and identified the possible problems that needed to be considered to improve in the main experiment. In this chapter, based on the experiment design, we presented the experimental procedures, participants and stimuli. In the end, the analysis of results were shown.

5.1 Main Experiment 5.1.1 Experiment Design As described in Chapter 3, the research hypothesis is that learning the third language (L3: Japanese) has a positive influence on the vowel length perception of the second language (L2: English) and the characteristic of this influence is different with the assimilation that is explained in the former mechanism of cross-language transfer PAM. To test our hypothesis, the identification task was conducted. In the identification task, the participants were asked to identify the category of sounds. This task reflects the ability to perceive the sound according to internal phonological categories (Tsukada, 2011). It means that a higher accuracy of identification task reflects a higher ability of perceiving phonological categories (such as vowel length). Moreover, several investigation instruments were used for clarifying whether the factor from the experiment participants (such as age, English learning year) and the factor from the material used in experiment have effects on the accuracy of identification task. Furthermore, the learners who are Chinese native and have learned English as L2 are

29

selected as Control Group while the learners who is Chinese native and have learned English as L2, Japanese as L3 are selected as Experimental group. In order to avoid the effect of language exposition, the participants of Control Group are confirmed to not have a living or study experience in Japan.

5.1.2 Participants The control group students were recruited from the 2nd and 4th grade of undergraduate course and the 1st and 2nd grade of graduate course at Western Languages College of Mudanjiang Normal University with Chinese as L1 and English as L2, and without the experience of learning Japanese. The mean age of the control group was 21.6 ranged from 19 to 24 (mean = 21.6 years, SD = 1.60). The experimental group students were recruited from the 1st and 2nd grade of Master’s course at Japanese Advanced Institute of Science and Technology (JAIST) with Chinese as L1 English as L2 and Japanese as L3. The mean age of the experimental group was 24.9 ranged from 22 to 27 (mean = 24.9 years, SD = 1.61). All participants in the both groups speak Mandarin Chinese with normal vision, and have no hearing impairment. In this experiment, the meaning of L2 and L3 represent the order of a non-native language and do not reflect the ability level of language.

5.1.3 Experiment Procedure This experiment consisted of the following three sessions as shown in Figure 5-1. Session A: The pre-questionnaire to explore the background information, language learning experience and proficiency of the participants. Session B: The speech identification task (including English and Japanese speech identification tasks) to investigate the speech perception of short and long vowels in English and Japanese words and non-words. In this task, the

30

participants need to identify and classify the target phonemes in the stimuli (e.g., a word "sit" contains a short vowel, and task of the participants is to classify it into "short" or "long"). Session C: The word familiarity test aims to understand the extent of the familiarity with the word stimuli.

Figure 5-1. Flowchart of the experiment

A. Language Background and Proficiency Questionnaire In order to understand the participants’ background information, language learning experiences, and language proficiency, we administrated the Language Background and Proficiency Questionnaire (Appendix 5-1). The contents related with their basic information (name, age, gender), the non-native languages they are being learned and that of leaning years, and their resident time span of foreign language countries. We measured the English proficiency level of all participants by the College English Test Band 4 (CET4) which is considered as a requirement for graduation of Chinese college students and an authority test of English learners’ proficiency in China. All of the participants in the control group haven’t the experiences of living in the English-speaking country or Japan, and their span of the English learning years is from 7 to 13 (mean = 11.1, SD = 1.640). We measured the Japanese proficiency level of the experimental group participants by Japanese-Language Proficiency Test for Japanese Advanced Level (JLPT N1). All of the

31

experimental group students have taken JLPT N1 and have been living in Japan (range from 2 months to 4.5 years) as they reported. Their span of the English learning years is from 10 to 15 (mean = 12.6, SD = 1,717). They all have at least 3 years of Japanese language learning experience untill the research was conducted. None of them have the experience of living in the English-speaking country.

B. Identification Task b1. Stimuli In order to understand if the participants applied their phonological knowledge or used their memory of pronunciations learned, both the word and the non-word of English and Japanese stimuli are designed in the experiment. The non-word refers a group of letters or speech sounds that looks or sounds like a word but that is not accepted as such by native speakers. A total of 118 stimuli was employed in the identification task: 4 kinds of vowels

2

counterpart × 3 conditions = 30 trails for English word and the same setting for non-word with the C1VC2 structure (C=consonant; V=vowel), 5 vowels× 2counterpart× 3 conditions = 30 trails for Japanese word and the same setting for non-word with the C1V1C2V2 structure. All the stimuli are shown in Table 5-1. The V in the English stimuli was either short or long vowel (/ʌ/-/ɑ:/, /ɪ/-/iː/, /ɔ/-/ɔː/, /u/-/uː/). Two words were paired to have the same C1 and C2 (such as /bɪd/-/biːd/ and /dɔd/-/dɔːd/). The V1 in the Japanese stimuli was also either short or long vowel (/ʌ/-/ɑ:/, /e/-/eː/, /ɪ/-/iː/, /ɔ/-/ɔː/, /u/-/uː/). Two words were paired to have the same C1, V2, and C2 (such as /kado/-/kɑ:do/ and /gise/-/gi:se/).

32

Table 5-1. Stimuli used in the experiment

English word

English non-word

Short

Long

Short

Long

but /bʌt/ cut /kʌt/ hut /hʌt/ bid /bɪd/ fit /fɪt/ pick /pɪk/ gock /gɔk/ dom /dɔm/ pot /pɔt/ full /ful/ bull /bul/ pull /pul/

bart /bɑ:t/ cart /kɑ:t/ heart /hɑ:t/ bead /biːd/ feet /fiːt/ peak /piːk/ gork /gɔːk/ dorm /dɔːm/ port /pɔːt/ fool /fuːl/ bool /buːl/ pool /puːl/

jat /dʒʌt/ luc /lʌk/ nap /nʌp/ shik /ʃɪk/ dik /dɪk/ cid /sɪd/ bom /bɔm/ bok /bɔk/ dod /dɔd/ guul /gul/ jom /dʒum/ phull /ful/

jart /dʒɑ:t/ larc /lɑ:k/ narp /nɑ:p/ sheak /ʃɪːk/ deek /dɪːk/ ceed /sɪːd/ bawm /bɔːm/ bawk /bɔːk/ dord /dɔːd/ goul /guːl/ joom /dʒuːm/ phool /fuːl/

Japanese word Short kabu kado hato kesu seken seki kita biru shiru joshi koto koji suji shukan shuki

Long kaabu kaado haato keesu seeken seeki kiita biiru shiiru jooshi kooto kooji suuji shuukan shuuki

Japanese non-word Short yanu lahi nahi tesu keso nero gise shise chira toni sota kopa muyo muse yuho

Long yaanu laahi naahi teesu keeso neero giise shiise chiira tooni soota koopa muuyo muuse yuuho

33

b2. Material The speech sound of the stimuli were synthesized obtained by using a Text-to-Speech application (GhostReader) and were evaluated by one native English speaker and three Japanese native speakers, respectively, to make sure a clear articulation.

b3. Procedure The participants’ perception of vowel length contrasts was tested by means of a standard forced-choice identification task. The task of the participants is to label the single sample sound each trail (i.e. to click the buttons (short or long) to select the one which match with sample sound). The Figure 5-2 show the user interface of the identification task application. The participants listened to stimuli under Audio-Technica FC707 headphones by clicking the play button [ui3] with a self-pace. They were forced to select the length of vowel, V for English and V1 for Japanese, between the options of long or short by clicking the buttons [ui4]. This task was done under the following conditions, which were considered useful for reducing the other possible effects on the performance of identification task. l

Spelling blindness The spellings of stimuli were not displayed in the questions in [ui1].

l

Non-listening forbiddance It was not allowed to go to the next stimulus without clicking the play button [ui3] at least one time.

l

Stimuli replay limitation The maximum replay time [ui2] was set up to three. The task consisted of two blocks, English and Japanese, which order was randomized.

The order of words and non-words in each block were also randomized. The practice trials (5 words) were provided at the beginning of each block to help the participants to

34

understand how to use the application.

ui2: remaining replay times ui1:question

ui3:play button ui4: answer selection buttons next question button Figure 5-2. User interface of the identification task application

C. Word Familiarity Task The control group’s and experimental group’s familiarities with English word were investigated using a 5-point Likert scale (1: Not at all familiar, 2: Slightly familiar, 3: Uncertain, 4: Moderately familiar, 5: Extremely familiar).

5.2 Results We show the analysis of the experimental result by considering the main factor of identification task accuracy (Result 1) firstly and following by presenting how the other influence factors of participant’s age (Result 2), English Learning Year of Participants,

35

Score of CET Test (Result 2), Word Familiarity (Result 3) influence the main factor.

5.2.1 Result 1: Accuracy of Identification Task Figure 5-3 shows the mean accuracy rates of the identification task for four types of words. The results were analyzed using a 2x4 mixed design ANOVA between group (control, experimental; between subjects) and word type (English word, English non-word, Japanese word, Japanese non-word; within subjects). The ANOVA showed significant main effect of group, F(1, 38) = 43.2, p < .0001, and of word type, F(3, 38) = 14.8, p < .0001. However, these main effects will not be discussed further, as the interaction between group and word type was significant, F(1, 38) = 7.31, p = .0002

1.00 0.95

Control

Experimental

Accuracy Rate

0.90 0.85 0.80 0.75 0.70 0.65 0.60 0.55 0.50 English Word

English Non-word

Japanese Word

Japanese Non-word

Figure 5-3. Mean of the accuracy rates (the error bars are standard errors)

The simple main effects of group within each word type were significant for the all

36

word types; F(1, 38) = 7.69, p = .0062 for English word; F(1, 38) = 10.2, p = .0017 for English non-word; F(1, 38) = 58.1, p < .001 for Japanese word; F(1, 38) = 25.1, p < .001 for Japanese non-word. The simple main effects of word type in each group were also significant for the both groups; F(3, 38) = 3.54, p = .0169 for the control group; F(3, 38) = 18.9, p < .001 for the experimental group. Therefore, the experimental group obtained a higher mean accuracy than the control group in all the four tests (i.e. English word, English non-word, Japanese word, Japanese non-word). And the there were not significant differences between word test and non-word test of both English and Japanese to the control group and experiment group.

5.2.2 Result 2: Relation between Participants’ Age and Accuracy of Identification Task Figure 5-4 compares the mean of participants’ age between the control group and the experimental group. An independent-sample t-test was conducted to compare the mean age of these two groups. There was a significant different in the mean of age for control group (mean = 21.6, SD = 1.61) and experimental group (mean = 24.9, SD = 1.61); t(38) = -6.44, p < .001, d=0.72.

37

Mean of Age 26 25

Age

24 23 22 21 20

Control Group Experimental Group

Figure 5-4. Mean of participants’ age (the error bars are standard errors)

We divided the all participants into two group according to the mean age over all participants (23.2). Figure 5-5 compares the mean accuracy rates for four word types between the below mean age group and the above mean age group. This results were also analyzed using a 2x4 mixed design ANOVA between group (below the mean age, above the mean age; between subjects) and word type (English word, English non-word, Japanese word, Japanese non-word; within subjects). The ANOVA result showed significant main effect of age F(1, 38) = 17.9, p