研究課題名：英語聴取におけるプロソディ情報依存度の分布傾向

 

氏名：中村智栄

所属：慶應義塾大学大学院

　　　政策・メディア研究科修士課程2年

 

要旨

　本研究では、文を構成している要素を統語的まとまりに分け、その区切れを示すプロソディの「統語的韻律機能」の側面に焦点を当て、日本人英語学習者が英語音声文理解においてプロソディ情報をどの程度利用しているのかを実験により検証した。

　具体的には、文章内の構造を示す文法的要素としての構文情報(syntactic　information)と、発話音声上で表現されるプロソディ情報(prosodic information)それぞれが示す統語機能を故意に矛盾させた文章（統語-プロソディミスマッチ文）を音声刺激として用いた2種類の聴覚実験により、日本人英語学習者の音声文理解における注目情報を明らかにする実験を行った。実験には早稲田大学で英語授業を履修する130名が参加し、英語力の指標として被験者が予め受験していたVersant English Testのスコアを用いた。被験者の英語力レベルは初級から中級である。

　実験1では、被験者はランダムに提示される11文の統語-プロソディミスマッチ文と11文のノーマル文、計22文の音声を聞き、聞いた音声文の主要部名詞をスクリーンに提示される2つの選択肢から選んで回答した。実験2では、実験1で使用した文章にローパスフィルタ処理をした音声を提示し、音声内の構文情報が曖昧になりプロソディ情報が強調された音声を聞いて、文章の主部位置が特定できるかどうかを検証した。

　結果として、実験1においては、タスクへの正解回答率と被験者の英語力に強い相関が見られた。一方、実験2においては、タスクへの正解回答率と被験者の英語力に負の相関が見られた。この結果から、初級英語学習者がプロソディ情報を重要な手がかりとする一方、英語力が上がるに伴い構文情報に基づいた音声文理解を行うようになる可能性が示された。その理由としては、英語力の高い被験者は英語音声文理解においてプロソディ情報よりも、語意や語形態など構文情報に関連した情報に頼った処理を行う一方、初級英語学習者は構文情報に関連した知識が浅いことから音声によるプロソディ情報に基づいた回答を行うことが推測された。本研究により、英語力の相違に伴い英語音声文理解の方略がプロソディ情報に基づいた判断から構文情報に基づいた判断へと変化が生じる可能性が示唆された。

 

1. Introduction

This study focuses on the syntactic function of prosody in English spoken sentence and examines how Japanese EFL learners use prosodic cues in understanding aurally presented English sentences.　 Although language input is processed by a complex interaction of use of syntactic and prosodic information, demonstrating which of the information L2 learners tend to use more is attracting many researchers’ attention. From results of previous studies, it can be assumed that prosodic cues are always the most important information for second language learners [1][2][3][7].　 Our purpose is to determine whether Japanese EFL learners rely heavily on prosodic cues in analyzing sentence structure in English as has been stated in previous studies.

 

2. Method

2.1. Task 1

The methodology for task 1 is similar to that of Read and Schreiber’s study on L1 sentence processing [8] and the study of Harley et al on L2 sentence processing [4].　 This task was to determine which of the information between prosodic and syntactic cues Japanese EFL learners would attend to the sentence structure, under conditions where syntax and prosody were placed in conflict (ambiguous sentences).　 The construction of ambiguous sentences was as follows: Two sentences were recorded separately, at normal speed with appropriate prosody that matched the sentences syntax, for example:

1) 　[[The new teachers]↑[watch baseball on TV.]]

2) 　[[The new teacher’s watch]↑[has stopped.]]

The sound was then edited to make replacement of the italicized and underlined segment of 1) by the italicized and underlined segment of 2), creating a prosodic contour in the stimulus sentence 1) that competed with the sentence syntax (shown as sentence 3)).　

3) 　[[The new teachers]↑[watch has stopped.]]

The result of the sentence splicing was such that the stress falls before the head noun of the subject noun phrase. There were 11 ambiguous sentences edited in the same manner, and another 11 non-edited normal sentences in which subject noun phrase consisted of two words.　 Participants were asked to select the head noun of the subject noun phrase in the sentence from two possible answers after listening to each sentence.

 

2.2. Task 2

Task 2 was performed to determine the sensitivity to the prosodic information in spoken sound.　 22 ambiguous and normal sentences that were used in experiment 1 were low-pass filtered at 50Hz.　 The method of low pass filtering is used to eliminate phonological information of the sound such as fricative sounds, while it still maintains the prosodic information such as fundamental frequency and intonation.　 After low-pass filtering at 50Hz, the output sentences have unclear phonological information and emphasized prosodic information.　 Conducting phonetic analysis on filtered sentences, we can find the position of rest between the segment of the subject noun phrase and the following words (e.g. in sentence 4) shown below, marked “↑”).

4) 　[[Your friend’s book]↑[was found in the box.]]

After listening to the sentence twice, on the third time, participants were asked to press the “1” key at the position of rest between the segment of the subject noun phrase and the following words.　 Since phonological information was unclear by the effect of low-pass filtering, participants had to listen to prosodic information to accomplish the task.　 Response time from the onset of the third play of the sound to pressing the key was measured and compared to correct response time.　

 

2.3. Subjects

130 adult subjects participated in this study.　 All of them were students in an English class in the faculty of law at Waseda University.　 They were asked to take the Versant English Test [6] prior to this research and we used the score as the barometer of their English level.　 Their English level was distributed from primary to intermediate.

 

3. Conclusions

The percentage of correct answers in task 1 was strongly correlated with the participants’ English levels.　 On the contrary, the percentage of correct answers in task 2 was negatively correlated with their English levels.　 From these results, we came to the conclusion that prosodic cues are the most important information for Japanese EFL learners at primary level, but as the learners’ English level becomes higher, they tend to rely more on syntactic cues in sentence comprehension.　 As the main reasons for this result, it can be inferred that as learners’ English improve, they use other information concerning to syntactic cues such as semantic and morphological information instead of prosodic information, while learners at primary level rely heavily on prosodic cues because they do not have other knowledge to rely on.　 This study revealed the possibility that the strategy of Japanese EFL learners transmutes from dependence on prosodic cues to syntactic cues in analyzing aurally presented English sentence.

 

References

[1]       Dussias, P., “Syntactic ambiguity resolution in L2 learners”, SSLA, 25, 529-557, 2003

[2]       Felser, C., Roberts, L, Marinis, T., & Gross, R., “The processing of ambiguous sentences by first and second language learners of English”, Applied Psycholinguistics, 24, 453-489, 2003

[3]       Fodor, J., “Learning to parse?” Journal of Psycholinguistic Research, 27, 285-319, 1997

[4]       Harley, B., Howard, J., & Hart, D., “Second language processing at different ages: do younger learners pay more attention to prosodic cues to sentence structure?”, Language Learning, 45, 43-71, 1995

[5]       Hincks, R., “Using speech recognition to evaluate skills in spoken English” Lund University, Dept. of Linguistics Working Papers 49, 58-61, 2001

[6]       Marinis, T., Roberts, L, Felser, C., & Clahsen, H., 　“Gaps in second language sentence processing”, SSLA, 27, 53-78, 2005

[7]       Papadopoulou, D., & Clahsen, H., ”Parsing strategies in L1 and L2 sentence processing”, SSLA, 25, 501-528, 2003

[8]       Read, C., & Schreiber, P., Why short subjects are harder to find than long ones, “The state of the art”, 78-101, Cambridge University Press, 1982