台灣英語學習者母音感知與產製 - 政大學術集成

全文

(1)國立政治大學語言學研究所碩士論文. National Chengchi University Graduate Institute of Linguistics Master Thesis. 政治大 Advisor: Dr. I-Ping Wan 立. 指導教授：萬依萍博士. sit. y. ‧. ‧ 國. 學. Nat. 台灣英語學習者母音感知與產製. n. al. er. io. Vowel Perception and Production by EFL Students in Taiwan. Ch. engchi. i n U. v. 研究生：陳品宇撰 Student: Chen, Pin-Yu 中華民國一〇七年六月 June, 2018. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(2) Vowel Perception and Production by EFL Students in Taiwan. BY Chen, Pin-Yu. 立. 政治大. ‧. ‧ 國. 學 A Thesis Submitted to the. y. Nat. io. a. er. in Partial Fulfillment of the. sit. Graduate Institute of Linguistics. n. v l Requirements for the Degree n i of Ch. U i e h n c g Master of Arts. June, 2018. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(3) 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. Copyright © 2018 Pin-Yu Chen All Rights Reserved. i. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(4) Acknowledgement 本篇論文首先我要相當感謝萬依萍老師無私給予指導，並且讓我在語音暨心理音韻實驗室擔任助理，讓我能夠一進到語言所就能夠有豐富的研究資源並且實地嘗試搜集語料、做實驗。實驗室的經歷充滿了我這輩子難能可貴的各種經驗，如果沒有加入萬老師的計畫我想我絕對不可能時實地進到小朋友的家，倚在地上跟小朋友一邊錄音一邊玩耍。閔閔翔翔有一次叫我爸爸時我跟 Vicky 都快笑瘋了。而到醫院去收集語料雖然真的非常辛苦但是真的讓我成長許多，我知道研究路上需要注意的各種細節、還有難熬的時刻，要收的語料有後就是花了一天也收不到半分鐘。沒有這些成長，我無法有能力完成這篇論文。完成這篇論文也要感謝兩位口試委員王世平老師及甯俐馨老師給予的意見及指導。我也要相當感謝政大語言所給我如此優良的環境以及學術資源讓我能夠在研究的路上相當順遂、不用為了尋找資源而感到煩惱。我也要感謝研究所的黃瓊之老師、戴智偉老師、蕭宇超老師、何萬順老師、張郇慧老師的惇惇教誨。我絕對無法忘記每次上蕭老師的課前有多緊張，還有黃老師的課每次都是腦筋急轉彎。戴老師則是屢屢充實我課前及課後的時間。大學時期的賴秋月老師、許麗珠老師、羅莉蓓老師也都是我必須要感謝的對象，如果沒有這幾位老師打好我的語言學基礎，我想我是無法走進語言學的研究領域。因為我這篇論文是需要收集語料的，除了歸功在萬老師的實驗室有收集與料的經驗以外，我在碩一下的時候也曾經修了戴老師的田野調查一門課。我永遠無. 立. 政治大. ‧. ‧ 國. 學. n. al. er. io. sit. y. Nat. 法忘懷這門課我跟著三位學長還有老師就這麼走了一學期，每個禮拜都在討論要怎麼搜集最原始的又最關鍵的語料，實驗著各式的引誘句子，揣測可能發生的情況，實地用專業錄音機還有麥克風錄製語料，學會怎麼用各種軟體。最後我們還去到了南投一個禮拜，實地探訪了布農族部落，在那趟旅程中我獲益良多。另外，我也必須要感謝我身邊所有的朋友尤其是那些當了我的受試者的朋友。感謝政大的雯淇 Vicky 在這三年研究所一路跟我當 partner 到最後，從一開始進實驗室的每兩個禮拜的收案到後來時常週末要一起崩潰還有一起合寫 paper 以及一起投稿泰國，雖然我們最後在搶誰先畢業，不過整個研究生的生涯真的幾乎都跟你合作無間。同時，政大語言所的曉蒨學姊、子權學長、孟孟學姊、實驗室的馨芸學姊、旺貞學姊、侃彧學姊、助教學姐等人無論是在學術上或是一些生活經驗上都給予過我很多的指導以及協助。我在剛寫這篇論文時根本不太會用 Praat 也不知道要怎麼寫 Script，甚至還異想天開要做 intonation，幸好遇到曉蒨學姊的協助，教導我使用 Praat 相關的知識。在實驗室時總是要處理各種大量的 tokens 要計算，而孟孟學姊總是在我面對龐大資料無從著手，打開 Excel 卻不知該怎麼辦時伸出援手，輕而易舉地寫出幾行公式給我。馨芸學姊則是教會我行政上的瑣事。碩二下時我曾經想過要到國外唸書，旺貞學姊一聽到馬上提供我各種資訊，. Ch. engchi. i n U. v. 真的讓我倍感溫馨。每次我在學校無聊時則都要找侃彧聊聊，探詢他是否有什麼新的招式可以處理各式統計。在政大的日子裡也要感謝周柏宇、唐寧、林凱源、我最後一年的室友以及阿喵跟阿屁。住在宿舍時，我老是跑去唐寧房間串門子還 ii. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(5) 有找凱源陪我吃宵夜、泡溫泉。阿喵則是陪伴了我無數個夜晚，總是在停車場靜靜地陪我，雖然你可能只是在等我拿出罐罐來，阿屁雖然整天很吵但是我寫不出論文時，但卻都一直在我身旁一起等待靈感乍現。能夠完成這篇論文需要感謝的還有黎明的同學：蘇聖富、吳貫維、楊又頲、范高齊、郭育菽、林柏榮、李昀等人願意為我錄音也在我願意在我研究乏味的期間陪伴我吃飯聊天玩耍，也要相當感謝暨南大學資管系的王茂權，碩一下時的研究邀約成為了我說是論文的初想。同時我也要相當感謝張懷剛在最後的協助以及陪伴，陪伴我各種週末還有幫我找受測者。在研究所期間，我去過泰國兩次發表，都是隻身前往幸好第二次暨大的怡竹也陪伴我一同去玩了數天，那趟旅程裡充滿了各種奇遇、先是你自己佯裝泰國人走水路到後來我們要去 Asoke 按摩時坐錯船兩個人驚恐地被拆散，還有各種酒後的談天一起以為自己終將無法畢業這些記憶都將無法被抹滅。暨大時期，詩雅、雨彤、均均我們四人總是無話不談，上了研究所後倍感空虛，幸好雨彤跟均均後來也來到政大延續我們的情緣尤其是均均時不時我拉出去玩耍，以上總總真的都是我能夠在這乏味的研究所撐下去的支柱。最後，我也要感謝我的父母、家人在我背後支持我一路唸書到 25 歲。如果沒有我家人的體諒以及支持，我想我是無法如此無後顧之憂地完成這本論文。. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 2018/08/25. 陳品宇. iii. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(6) Table of Contents Acknowledgement ........................................................................................................ ii Table of Contents ..........................................................................................................iv Index of Tables ............................................................................................................ vii Index of Figures ............................................................................................................ix Abstract ....................................................................................................................... xii Chapter 1 Introduction ................................................................................................... 1. 政治大. Chapter 2 Literature Review .......................................................................................... 7. 立. 2.1 Acoustic Features of Vowels ............................................................................ 7. ‧ 國. 學. 2.1.1 English Vowels ...................................................................................... 7 2.1.2 Mandarin Vowels ................................................................................ 10. ‧. 2.1.3 Previous studies on English Vowels produced by Taiwan Mandarin. y. Nat. er. io. sit. Speakers ....................................................................................................... 12 2.2 Asymmetry of Production and Perception ..................................................... 14. al. n. v i n C h ......................................................................... 2.2.1 Tense and Lax Vowels 14 engchi U 2.2.2 L1 and L2 Interferences ...................................................................... 15 2.2.3 Vowel Intimacy ................................................................................... 16 2.2.4 Influences on the Vowel Space ........................................................... 17 2.2.5 L1 Language Transfer ......................................................................... 18 2.2.6 General Discussion of L2 Acquisition ................................................ 19. 2.3 Models............................................................................................................ 20 2.3.1 Contrastive Analysis Hypothesis ........................................................ 21 2.3.2 Speech Learning Model ...................................................................... 21 iv. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(7) 2.3.3. Perceptual Assimilation Model .......................................................... 22 2.3.4. Application of the models .................................................................. 23 2.3.5 More on the Developmental Process .................................................. 23 2.3.6 ABX Discrimination Task ................................................................... 24 2.4 Research Questions and Hypothesis .............................................................. 26 2.4.1 Research Questions ............................................................................. 28 2.4.2 Hypothesis........................................................................................... 28 Chapter 3 Methodology ............................................................................................... 31. 治政大 3.1 Data Collection and Elicitation ...................................................................... 31 立 ‧ 國. 學. 3.1.1 Participants.......................................................................................... 31 3.1.2 Materials ............................................................................................. 32. ‧. 3.1.3 Procedures ........................................................................................... 34. sit. y. Nat. 3.1.4 Instruments .......................................................................................... 38. n. al. er. io. 3.1.5 Data Security Management Principles ................................................ 40. i n U. v. 3.1.6 Surveys................................................................................................ 40. Ch. engchi. 3.2 Data Analysis ................................................................................................. 41 3.2.1 Vowel Fetch......................................................................................... 41 3.2.2 Perception Evaluation ......................................................................... 41 3.2.3 Model Construction ............................................................................ 42 Chapter 4 Results and Findings ................................................................................... 43 4.1 Production Performance................................................................................. 43 4.1.1 English Vowels Produced by Taiwan Mandarin EFL Learners .......... 44 4.1.2 Mandarin Vowels Produced by Taiwan Mandarin EFL Learners ....... 55 v. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(8) 4.1.3 English Vowels Produced by American English Native Speakers ...... 67 4.2 English Vowel Perception .............................................................................. 71 4.2.1 Word Identification Task ..................................................................... 72 4.2.2 ABX Discrimination Task ................................................................... 76 4.2.3 General Perceptual Performance......................................................... 88 Chapter 5 Discussion ................................................................................................... 90 5.1 Development of the Production of English Vowels (Experiment I) .............. 91 5.2 Development of Production of Mandarin Vowels (Experiment II)................ 94. 治政 5.3 Development of Perception of English Vowels大 (Experiment III & IV) ......... 95 立 ‧ 國. 學. 5.4 Symmetry and Asymmetry between Learners’ Perception and Production Model ................................................................................................................... 98. ‧. 5.4.1 Symmetry between Vowel Perception and Production ....................... 98. sit. y. Nat. 5.4.2 Asymmetry between Vowel Perception and Production ..................... 99. n. al. er. io. 5.5 Inter-lingual Influences on EFL Learners in Taiwan ................................... 100. i n U. v. 5.5.1 Vowel Perception .............................................................................. 101. Ch. engchi. 5.5.2 Production of English Vowels ........................................................... 102 5.5.3 Production of Mandarin Vowels ....................................................... 104 Chapter 6 Conclusion ................................................................................................. 106 Suggestions and Limitations .............................................................................. 107 References .................................................................................................................. 109. vi. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(9) Index of Tables Table 2- 1 Formant Structure of English Vowels Presented by American English Native Speakers ..................................................................................................... 8 Table 2- 2 Vowels in Mandarin Chinese ...................................................................... 10 Table 2- 3 Formant Structure of Mandarin Vowels. Performed by Taiwan Mandarin. Speakers ............................................................................................................... 11 Table 2-4 Formant Structures of English Vowels Presented by Taiwan Mandarin Speakers ............................................................................................................... 13. 政治大. Table 3- 1 English Vowel Stimuli List ......................................................................... 33. 立. Table 3- 2 Mandarin Vowels Stimuli List .................................................................... 34. ‧ 國. 學. Table 3- 3 Combination of Test Words in Experiment IV ........................................... 38. ‧. Table 4- 1 Formant Structures of English Vowels produced by Taiwan EFL Learners. sit. y. Nat. .............................................................................................................................. 45. io. al. er. Table 4-2 Formant Structures of Mandarin Vowels Produced by Taiwan EFL Speakers. v. n. .............................................................................................................................. 57. Ch. engchi. i n U. Table 4-3 Formant Structures of English Vowels Produced by American Native Speakers ............................................................................................................... 68 Table 4-4 Responses on Target Stimuli at Word Decision Task Performed by All Levels of Taiwan Mandarin Learners and American English Native Speakers ... 72 Table 4-5 Error Responses in Other Stimuli ................................................................ 74 Table 4-6 Performance on ABX Discrimination Task in Random Arrangement Made by all levels of Taiwan Mandarin Learners and American English Native Speakers ............................................................................................................... 77 vii. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(10) Table 4-7 Error Rate of ABX Discrimination Task in Random Arrangement Made by all levels of Taiwan Mandarin Learners and American English Native Speakers .............................................................................................................................. 79 Table 4-8 Error Rate of Combinations in ABX Discrimination Task Made by all levels of Taiwan Mandarin Learners and American English Native Speakers .............. 80 Table 4-9 Error Types of All Stimulus Combinations on ABX ................................... 82 Table 4-10 Mixed Review of Error Type of Stimulus Combination on ABX ............. 86. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. viii. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(11) Index of Figures Figure 2- 1 English Vowel Grid of American English Native Speakers ........................ 9 Figure 2-2 Mandarin Vowel Grids of Taiwan Mandarin Native Speakers................... 12 Figure 2- 3 The perceptual processing when participants undertaking a ABX decision task ....................................................................................................................... 25 Figure 4-1 Vowel Grids of English produced by Taiwan EFL Learners ...................... 46 Figure 4-2 Schematic Spectrograms of Vowels Produced by Elementary Taiwan EFL Learners................................................................................................................ 49. 政治大. Figure 4-3 Schematic Spectrogram of Vowels Produced by Intermediate Taiwan EFL. 立. Learners................................................................................................................ 50. ‧ 國. 學. Figure 4-4 Schematic Spectrogram of Vowels Produced by High Intermediate Taiwan. ‧. EFL Learners........................................................................................................ 52. sit. y. Nat. Figure 4-5 English Vowel Grid of Male Speakers ....................................................... 54. io. al. er. Figure 4-6 English Vowel Grid of Female Speakers.................................................... 55. v. n. Figure 4-7 Vowel Grids of Mandarin Monophthongs Produced by All Taiwan EFL. Ch. engchi. i n U. Learners................................................................................................................ 58 Figure 4-8 Schematic Spectrogram of Mandarin Monophthongs Produced by Elementary Taiwan Mandarin EFL Learners ....................................................... 61 Figure 4-9 Schematic Spectrogram of Mandarin Monophthongs Produced by Intermediate Taiwan Mandarin EFL Learners ..................................................... 62 Figure 4-10 Schematic Spectrogram of Mandarin Monophthongs Produced by High Intermediate Taiwan Mandarin EFL Learners ..................................................... 63 Figure 4-11 Schematic Spectrogram of Mandarin Monophthongs Produced by All ix. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(12) Taiwan Mandarin EFL Learners .......................................................................... 64 Figure 4-12 Vowel Grid Among All Levels of Male Participants................................ 66 Figure 4-13 Vowel Grid Among All Levels of Female Participants ............................ 67 Figure 4-14 Vowel Grids of English Produced by American English Native Speakers .............................................................................................................................. 69 Figure 4-15 Schematic Spectrogram of English Monophthongs Produced by American English Native Speakers ...................................................................................... 71. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. x. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(13) 國. 立. 政. 治. 大. 學. 論. 文. 提. 要. 研究所別：語言學研究所論文名稱：台灣英語學習者母音感知與產製指導教授：萬依萍博士研究生：陳品宇論文內容：（共 24,261 字，分六章）外國口音是外語學習者明顯的特質。不同的口音往往能暗示該 L2 學習者的 L1 為何。這說明了 L1 對 L2 有相當的影響。在影響發音的各種因素中，母音便. 政治大. 是其中一項不容忽略的特質。本篇論文主要研究的是台灣以英語為 L2 而台灣華語為 L1 的學習者(TM)在母音的產製以及認知的表現。同時也針對 L2 在不同的. 立. 學習歷程中母音的產製及認知建構進行研究。為了上述研究目的，本論文一共對. ‧ 國. 學. 台灣華語以英語為 L2 的學習者採取了四種實驗還有以美式英語為 L1 的母語人士(AE)採取三種實驗來進行母音的產製以及認知分析及研究。實驗一是以及實驗二為母音產製實驗，實驗一為英語並且所有母音都包容在. ‧. [hVd]的環境而實驗二為中文。實驗三以及實驗四則是母音認知測驗，皆為英語. y. Nat. 實驗，而在認知實驗中母音則包容在[bVt]的環境中。英語母語人士在本次實驗扮. sit. 演角色為驗證測驗內容信度以及作為控制組以比對 L2 組別之表現。本論文共採. er. io. 納三個傳統分析方式嘗試對於學習者的發展進行建構及探討 L1 及 L2 之間的交. al. n. v i n Ch (Flege, 1995)以及 CPH 對比分析假說(Lado, i U e n g c1957)。 h 實驗 I、III、IV 結果指出 TM 以及 AE 兩組有顯著的差異。綜合實驗 II，我互影響，三種分析方式分別是認知類化模型 PAM (Best, 1995)、言語學習模型 SLM. 們更近一步發現 L1 以及 L2 之間無論是產製或是認知上都有明顯的交互影響。本次研究的大部份結果都與 SLM 還有 PAM 預測，惟有零星錯誤與先前的研究不一致。同時我們也藉由中級學習者在[e]、[æ] 以及 [ɛ]的表現上發現母音學習會有 U 型成長的趨勢。大致上來說，L2 組在本次研究的認知表現以及產製表現都相當一致，僅少數零星的不一致會出現在[ɪ] 還有 [o]、[ɔ] 以及[ʊ]上。本篇論文描述了台灣英語學習者的母音建構過程以及進階解釋其在認知及產製上的相互影響並且同時研究了學習者的 L1、L2 之間的影響遷移。關鍵字：母音習得、第二語言習得、中介語、聲學分析、母音分析. xi. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(14) Abstract This thesis investigates vowels produced and perceived by English learners in Taiwan. Three main questions were inspected: i) learners’ development process of perception and production of English and Mandarin vowels of Taiwan Mandarin English learners ii) the symmetry and asymmetry between the perception and production models iii) whether there is any inter- or intra-interference on L1 and L2 vowels. We have constructed the model of the process of the production and perception development of English (L2) vowels of Taiwan Mandarin (TM) speakers who are. 政治大. studying English as a second language by inspecting their speaking and listening. 立. performances on certain tasks in this thesis. We have conducted four experiments on. ‧ 國. 學. TM L2 learners of English and three experiments on American English native (AE). ‧. speakers.. sit. y. Nat. The first and second experiments were to test the pronunciation English and. io. al. er. Mandarin vowels, where the English vowels were situated in the fixed environment of. v. n. [hVd], and where the Mandarin vowels were combined in a random onset and a fixed. Ch. engchi. i n U. coda [l]. The third and fourth experiments were two English vowel perception tests, consisting of English vowels placed in [bVt]. The AE group performed as the control group showing that the test items in the experiments are recognizable and to provide a contrast to the vowels performed by the L2 learners. Three standard models, the Contrastive Analysis Hypothesis (Lado, 1957), the Perceptual Assimilation Model (Best, 1995) and the Speech Learning Model (Flege, 1995) were employed to explain and construct the development model and the inter-group and intra-group interferences. The results of the performance of the TM and AE groups in Experiments I, III and xii. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(15) IV provide evidence for differences in the perception and production of vowels between the native speakers and the L2 learners. Along with experiment II, the experiments I, II and IV further implies the interand intra- interferences on L1 and L2 vowels in both production and perception. Most of our findings are congruent with the PAM and the SLM, while some of the findings on errors are not consistent with those in previous studies. Possible evidence for Ushaped learning is found in our results in the performance of [e], [æ] and [ɛ] by the intermediate levels. Overall, the perceptual errors were in symmetry with their. 治政大 as [ɪ] and [o], [ɔ] and [ʊ], production performance, while only a few vowels, such 立 ‧ 國. 學. showed more asymmetrical errors in production rather than in perception. This thesis depicts a model of the development of the processing of English vowels by TM. ‧. speakers and further examines how symmetry and asymmetry take place and inspects. n. al. er. io. sit. y. Nat. the degree of L1 transfer.. i n U. v. Keywords: Vowel Acquisition, Vowels, SLA, Second language acquisition, Acoustic. Ch. analysis, Interlanguage, Vowel Analysis. engchi. xiii. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(16) Chapter 1 Introduction The English produced by Taiwan Mandarin speakers is often considered to be strongly accented. The presence of a foreign accent represents the mispronunciation of particular phones in speech (e.g., Flege, 1989a; Flege, 1989b; Munro & Derwing, 1995). Nonnative speakers are often distinguished from native speakers by linguistic features. Linguistic features can be covered by features lie within several linguistic fields, such as syntax, semantics, pragmatics, phonetics and phonology. These linguistic features. 政治大. may reveal the cultural background of the speaker. Accent and pronunciation, which. 立. are encompassed by phonetics and phonology, are two elements that explicitly suggest. ‧ 國. 學. the language background of the speaker. A native speaker can quickly detect the. ‧. linguistic features of an utterance merely by listening to a string of speech. Before we examine the deep structure of the syntax in an utterance and realize the speech act of. y. Nat. er. io. sit. an utterance, we may first be aware of the quality of the sound and the pitch and melody in the speech. We can understand that the sound of the speech can be divided into at. al. n. v i n least two dimensions for furtherC analysis, h e ntheg segmental c h i Uand the suprasegmental parts.. We suppose that the vowel structures, as the formant quality, are one of the essential factors which may make one’s speech carrying a foreign accent, together with the consonants, stress and the other segmental and suprasegmental features that will influence the speakers’ pronunciation. Indeed, the vowels produced by Taiwan Mandarin speakers are different from those produced by English native speakers. Since the formant structures may dominate the recognition and the features of a vowel, they may cause listeners to be aware of the differences in pronunciation between a native English speaker and a non-native English speaker. This study, thus aims at investigating 1. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(17) the vowel production and perception of vowels among Taiwan Mandarin L2 learners of English. Vowels are the nucleus of a syllable. The syllables in most languages are combined with consonants and vowels. However, languages do not share the same phoneme inventory with each other. The phoneme inventories of different languages cannot be identical, and even the same languages spoken in different areas will vary from each other to a significant degree. However, they still share some certain features, and the key aspects of the production of vowels are claimed to be an unobstructed airstream,. 治政大and the rounding of the lips. and the influence of tongue height, tongue advancement, 立 ‧ 國. 學. In addition, the length of a vowel may also be an important feature for speakers; for instance, the production of monophthongs and diphthongs yields different durations. It. ‧. is also possible that a L2 learners may produce the L2 vowels by means of their L1. sit. y. Nat. vowel inventory, also known as language transfer.. n. al. er. io. It has been widely claimed that L2 learners may have difficulties in acquiring the. i n U. v. second language. The ability that L2 learners have to acquire the second language may. Ch. engchi. be affected by the age at which they are exposed to the second language and also by the amount of the use of their L1 (Flege, Frieda, & Nozawa, 1997; Flege, Yeni-Komshian, & Liu, 1999). Different language background factors (e.g., use of L1, dominant language at home, education, L2 background, etc.) may all play a role in the formation of one’s L2 performance. The linguistic features of one’s L1 could cause crucial interferences in the acquisition or production of one’s L2. Most native speakers acquire their segmental and suprasegmental skills in their early childhood, while nonnative speakers start to learn 2. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(18) the target L2 language over a relatively wide range of ages. It is hard to be precise in predicting the possible output of each L2 learner due to the fact that the different factors that might influence the acquisition of the L2 vary greatly among individuals. However, it is widely acknowledged that one’s L2 performance may be most influenced by one’s L1 and the interferences within their interlanguages1 since a group of learners sharing the same L1 often perform systematical and consistent errors. It is a common phenomenon that an L2 English learner in Taiwan will speak English with a Taiwan Mandarin accent. The accent is distinct, for the segmental and. 治政大 speakers. For instance, the suprasegmental features are divergent from those of native 立 ‧ 國. 學. English phonemes and intonation produced by Taiwanese L2 learners are not the same as those produced by English native speakers. Language transfer seems to be inevitable. ‧. as L2 learners learn or acquire a target language, and in this study, we mainly investigate. sit. y. Nat. the issues associated with L2 learning. A learner must to some degree apply his/her. n. al. er. io. prerequisite knowledge of the L1 to either understanding or producing the L2. That is,. i n U. v. one may or must apply some of the rules in the L1 to the L2 (e.g., Odlin, 1989; Yip,. Ch. engchi. 1995). Language transfer can be predicted by contrastive analysis if the amount of data we have obtained is large enough and analyzed by certain features (Odlin, 1989). Thus, Taiwan-accented English can be inferred to be a possible consequence of language transfer. To be more specific, many have asserted that the Taiwanese English accent is a result of the combination of inter- and intra-lingual interferences from Taiwan. 1. Interlanguage refers to a learner’s L2 output, which is a language produced systematically (Gass. & Selinker, 2008). 3. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(19) Mandarin to English L.-M. Chen, Tseng, and Tso (2007). Some other studies have investigated the rates of accuracy in the production of tense and lax vowels. Y. Chen, Robb, Gilbert, and Lerman (2001), Lai (2005) and Liao (2006) have shown that the perception and production of vowels by English as a foreign language (EFL) learners in Taiwan are deviant from those of English native speakers. These studies have constructed Mandarin speakers’ English vowel production models of English vowels produced by Mandarin speakers and some have constructed perception models. Since most studies pay attention merely to the front vowels and. 治政大 levels together, the range place all of the data from Taiwanese EFL learners at different 立 ‧ 國. 學. of the deviation among the speakers may be too broad. Also, some studies collected. speech samples from Beijing Mandarin speakers, which are not appropriate to use to. ‧. describe the speech model in Taiwan. It is known that different dialects will yield. sit. y. Nat. certain properties, and Taiwan Mandarin has already been seen as an independent. n. al. er. io. dialect from that of Beijing Mandarin.. i n U. v. The vowels structures produced and perceived by Taiwanese EFL learners are thus. Ch. engchi. the main areas that we aim to examine and analyze in this study. First, we will analyze the formant quality of the L2 vowels (English) and the L1 vowels (Mandarin) produced by different levels of Taiwanese EFL learners. Then, an L2 vowel quadrant of Taiwanese EFL learners will be built. Also, we will follow the style of the analyses that appeared in Flege (1995), Best (1995), Best and Tyler (2007), Hung (2012) and many others, which examined L2 vowel perception. In so doing, we will see if there is any. 4. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(20) assimilation or dissimilation2 of L2 English vowels and L1 Mandarin vowels. In order to make our stimuli consistent, all the target vowels are placed in the target words comprising the context of /h___d/ and /b___t/, which follows the conventions of most studies. We also adopt Flege’s (1995) and Best’s (1995) point that EFL learners will assimilate their L1 features into their L2 and thus pronounce with a non-native accent. Flege (1995) constructed a speech learning model which elaborated the development of interlanguage speech production. This model hypothesized the possible development. 治政大construct a new category or of language transfer and how L2 learners deviate and 立 ‧ 國. 學. assimilate the same category from the L1. Best (1995), on the other hand, provided insight into perceptual development and formed the concept of the perceptual. ‧. assimilation model. This model simulates the possible stages in the development of an. sit. y. Nat. interlanguage perception acquisition. Hung (2013) adopted both models and. n. al. er. io. investigated the effect of duration on on the perception of English vowels by Taiwanese. i n U. v. EFL learners at the phonetic level. Her study showed that Taiwanese learners. Ch. engchi. assimilated the English tense vowel [ɪ] to the Mandarin sound [i] and thus merged these two sounds into the same category. Together, these two models may give possible reasons for the phenomena of the Taiwanese English accent. In this study, we will introduce the overall process involved in the learning of vowel production and perception of Taiwanese EFL learners. We will show the. 2. The term assimilation does not refer to the specific phonological patterns but to the process. learners classify a L2 vowel in to one of their L1 vowel category. Discrimination, on the other hand, is the situation which learners deviate his/her L1 and L2 vowel inventory, while it does not necessarily mean that learners. 5. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(21) symmetry and asymmetry in the perception and production of the vowels by an L2 learner, and also the stages of the developments in the language transfer involved in both the perceptual and production acquisition from the L1 to the L2. The following second section will review a number of the studies that are relevant to the scope of our research, and the third section will illustrate the methodology of this study. The fourth section will show the preliminary findings and some hypotheses. Section five will provide concluding remarks.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 6. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(22) Chapter 2 Literature Review The present study investigates the symmetry and asymmetry in the process of the assimilation of English vowel production and perception as well as discrimination in Mandarin as performed by Taiwanese Mandarin speakers studying English as a foreign language (EFL) at different levels of proficiency. The acoustic characteristics of the vowels in Taiwan Mandarin and those of the vowels in American English will be investigated in 2.1, along with the American. 政治大. English vowels produced by Taiwan Mandarin speakers. The perception and production. 立. asymmetry in the American English vowels performed by native Taiwan Mandarin. ‧ 國. 學. speakers, and the discrimination of English vowels and their assimilation to Mandarin. ‧. vowels by Taiwan Mandarin speakers will be discussed in 2.2. The inter- and intralingual interferences in the processing model, including the details of the learning. y. Nat. er. io. sit. models that have been widely used and will be adopted in this study will be reviewed in 2.3. Section 2.4 will conclude this chapter and list the research questions of this study.. n. al. Ch. 2.1 Acoustic Features of Vowels. engchi. i n U. v. Vowel qualities are distinctive for their formant structure, especially the first and second resonant formant structures (F1 and F2). F1 and F2 are related to the movement of the tongue of the speaker. F1 is influenced by the height of the tongue, and F2 by the advancement of the tongue. A vowel inventory will be carried out first for both languages, and then the aggregation of F1 and F2 of each vowel will be illustrated. 2.1.1 English Vowels The American English vowels discussed in this study are adopted by a number of studies (e.g., L.-M. Chen et al., 2007; Y. Chen et al., 2001; Hillenbrand, Getty, Clark, 7. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(23) & Wheeler, 1995; Ladefoged & Johnson, 2014; Lai, 2010; Liao, 2006; Peterson & Barney, 1952), and the vowels will thus include ten monophthongs and three diphthongs. The monophthongs are [i, ɪ, ɛ, e, æ, ʌ, ɝ, ɑ, ɔ, ʊ, u], and the diphthongs are [aɪ, aʊ, ɔɪ] in the phonetic representation. Although Hillenbrand et al. (1995), Ladefoged and Johnson (2014), Lai (2010), Liao (2006) and Peterson and Barney (1952) have addressed the relevant issues, there are insufficient vowel grids and no detailed formant structures in their analyses. This study will adopt the acoustic features based on Y. Chen et al. (2001), which correspond best to the vowel systems we utilize in this. 治政大 in correspondence with the study. The vowel grid adopted in Chen et al. (2001) is more 立 ‧ 國. 學. needs of our study. Table 2-1 and Figure 2-1 show the formant structure of American English vowels as analyzed by Y. Chen et al. (2001). They conducted an elaborate. ‧. experiment which recruited twenty male and twenty female native speakers of. sit. y. Nat. American English. Y. Chen et al. (2001) originally constructed this model to carry out. n. al. er. io. a contrastive analysis of the vowel production of Mandarin speakers and of American English native speakers.. Ch. i n U. engchi. v. Table 2-1 Formant Structure of English Vowels Presented by American English Native Speakers Mean (M), standard deviation (SD) and range values in Hertz (Hz) for the F1 and F2 frequency of eleven vowels produced by American male and female speakers Male (n =20) F1 (Hz). Female (n =20) F2 (Hz). F1 (Hz). Vowel. M. SD. Range. M. SD. Range. M. /i/. 283. 30. 215-352. 2322. 218. 1144-2803. 339. /e/. 394. 37. 299-511. 1927. 438. 1077-2471. /u/. 315. 42. 239-478. 1055. 149. /o/. 437. 52. 338-677. 1099. 231. SD. F2 (Hz) Range. M. SD. Range. 42. 276-458. 2861. 289. 1003-3219. 438. 36. 358-511. 2553. 498. 1116-3180. 604-1387. 380. 58. 105-491. 1275. 230. 922-1770. 737-2465. 471. 53. 312-599. 1186. 192. 872-1771. 8. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(24) / ɑ/. 745. 135. 431-1102. 1260. 200. 970-2478. 1005. 107. 617-1196. 1532. 135. 1109-1940. / ɪ/. 432. 44. 325-578. 1864. 254. 1205-2296. 492. 49. 411-637. 2267. 300. 1368-2681. /ɛ/. 578. 73. 282-724. 1793. 136. 1521-2222. 737. 84. 458-944. 2141. 202. 963-2406. /æ/. 757. 83. 583-989. 1713. 148. 1436-2377. 1025. 79. 737-1235. 1993. 182. 1601-2485. /ʌ/. 614. 56. 439-784. 1429. 120. 1189-1829. 769. 89. 544-1022. 1750. 151. 916-2019. /ʊ/. 455. 45. 318-544. 1318. 132. 1063-1632. 514. 76. 318-659. 1614. 247. 962-2093. /ɔ/. 681. 152. 242-1003. 1233. 385. 931-2578. 938. 75. 784-1114. 1433. 312. 1101-3090. Note: Table 2-1 illustrates the mean F1 and F2 frequency of eleven vowels in English. The table contains both male and female vowel formant structures. The eleven vowels are distinct from each other. /i/, / ɪ/, /e/ are relatively high for their low values. 政治大. in the F1 and relatively fronted for the high values in F2.. 立. 1500. 1000. /i/. Nat. /e/. /ɛ/. io. /ɛ/. Ch. /ʊ/ /ʌ/. /æ / /ʌ/. n. al. /u/ /u/ /ʊ/ /o//o/. /e/ / ɪ/. / ɪ/. /ɔ/ / ɑ/. /æ /. engchi NM. ‧. /i/. 500. /ɔ/ / ɑ/. 0 0 200 400. y. 2000. sit. 2500. er. 3000. 學. 3500. ‧ 國. Figure 2-1 English Vowel Grid of American English Native Speakers. i n U. v. 600 800 1000 1200. NF. Note: Figure 2-1 illustrates the relative spaces of the English vowels of American English native speakers of both genders, where NM stands for male, and NF for female. The grids show the relative space of the tongue height and tongue advancement. The vertical axis shows the F2 values, and the horizontal axis shows the F1 values. The dots on the chart represent the place of the tongue of the vowel. The brackets left to the dots help indicate the vowel. 9. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(25) 2.1.2 Mandarin Vowels Turning to Mandarin studies, the surface vowels in Taiwan Mandarin in this study follow some analyses from Jeng (2011), Y.-H. Lin (2007) and Wan and Jaeger (2003). Mandarin Chinese includes ten monophthongs and four diphthongs. The monophthongs are [i, y, e, æ, a, ə, ɚ, ɤ, ɑ, u, o], and the diphthongs are [ai, ei, ɑu, ou] or [aj, ej, au, ou]. The vowels are listed in Table 2-2. In Table 2- 3, the formant structure of the Mandarin vowels performed by the Taiwan Mandarin speakers analyzed by Jeng (2011) are listed as follows. Figure 2-2. 治政大tongue height of each vowel. depicts the movement of the tongue advancement and the 立 ‧ 國. Front. 學. Table 2- 2 Vowels in Mandarin Chinese. Central. ‧. Unrounded Rounded. y. ɛ. ə/ ɚ. al. Low Low. Ch. engchi. sit. Mid. y. er. e. Unrounded Rounded. n. Mid. io. i. Nat. High. Back. i n U. ɤ. u o. v. a. ɑ. Note: Table 2-2 illustrates the vowels in Mandarin by their place of articulation. There are five front vowels with one rounded, two central vowel,s and four back vowels with two rounded. The Low vowels, [a] and [ɑ], pair in Mandarin.. 10. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(26) Table 2- 3 Formant Structure of Mandarin Vowels Performed by Taiwan Mandarin Speakers Male (n =20) F2 (Hz). F1 (Hz). F2 (Hz). Vowel. M. SD. M. SD. Vowel. M. SD. M. SD. /i/. 298. 50. 2243. 148. /i/. 396. 66. 2782. 218. /e/. 520. 26. 1923. 251. /e/. 679. 86. 2508. 174. /u/. 326. 46. 718. 92. /u/. 397. 64. 838. 119. /o/. 445. 60. 814. 87. /o/. 552. 102. 994. 123. / ɑ/. 743. 81. 743. 125. / ɑ/. 1013. 145. 1641. 166. / ɪ/. 365. 30. 2582. 66. 488. 2126. /ə /. 515. 54. 101. 1588. 109. /y/. 284. 33. 1319 立 1973. 2281. 138. */æ/. n/a. */ʌ/. n/a. */ʊ/. n/a. */ɔ/. n/a. ‧ 國. F1 (Hz). Female (n=20). 治 98 396 / ɪ/ 政大 125 676 /ə / 119. /y/. 381. 50. n/a. */æ/. n/a. n/a. 學. n/a. n/a. n/a. n/a. n/a. */ʌ/. n/a. n/a. n/a. n/a. n/a. n/a. n/a. */ʊ/. n/a. n/a. n/a. n/a. n/a. n/a. n/a. */ɔ/. n/a. n/a. n/a. n/a. y. ‧. n/a. Nat. n/a. er. io. sit. Note: Table 2-4 shows the formant structures of Mandarin vowels. The formant values are as measured and normalized by Jeng (2011). There are fewer vowels in. al. n. v i n Mandarin than in English. TheC highlighted h e n garea i U the phonemes that exist in c hindicates English vowels which are absent in Mandarin Chinese.. 11. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(27) Figure 2-2 Mandarin Vowel Grids of Taiwan Mandarin Native Speakers 3000. 2500. 2000. 1500. 1000. 500. 0 0 200. /i/. /i/ /y/. / ɪ/. /y/. /u/ /ə /. /e/ /ə /. /e/. /u/ /o/. / ɪ/. /o/. 400 600. / ɑ/. 800 1000. / ɑ/. 1200. 政治大 Note: Figure 2-2 illustrates 立the relative spaces of the Mandarin vowels of Taiwan TMM. TMF. ‧ 國. 學. Mandarin native speakers of both genders, where TMM stands for male, and TMF for female. The grids show the relative space of the tongue height and tongue advancement.. ‧. The vertical axis shows F2 values, and the horizontal axis with the F1 values. The dots. y. Nat. er. io. phonemes.. sit. on the chart represent the space of the vowel. The brackets on their left indicate the. al. n. v i n We can find that there are asterisk C h marks on someUof the vowels in Table 2- 3. The engchi. asterisk means that the vowels do not exist in Taiwan Mandarin and are that they are the counterpart of English vowels and Mandarin vowels. This lack of vowel inventory plays an important role in this study and the following shows the different related findings. 2.1.3 Previous studies on English Vowels produced by Taiwan Mandarin Speakers Y. Chen et al. (2001) and Liao (2006) both investigated the production of English vowels produced by Mandarin speakers. They described and analyzed the formant values they observed in detailed, and the data is further organized as Table 2-4. As the 12. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(28) elicitation form of Chen et al. (2001) better suits the format and phoneme systems adopted in this study than that used in other studies, we will adopt results in Chen et al. (2001). Table 2-4 Formant Structures of English Vowels Presented by Taiwan Mandarin Speakers Male (n =20) F1 (Hz). Female. F2 (Hz). F1 (Hz). F2 (Hz). M. SD. Range. M. SD. Range. M. SD. Range. M. SD. Range. /i/. 328. 50. 255±498. 2206. 203. 1534±2592. 340. 59. 236±471. 2641. 442. 996±3144. /e/. 422. 107. 282±888. 2064. 343. 1049±2585. 458. 113. 228±821. 2465. 419. 1158±2950. /u/. 368. 42. 276±471. 立 1028. 147. 774±1521. 403. 56. 296±511. 1172. 180. 697±1622. /o/. 444. 82. 316±724. 1111. 271. 781±1885. 496. 125. 309±857. 1157. 261. 793±2289. / ɑ/. 742. 152. 356±1104. 1284. 374. 794±2671. 906. 157. 432±1232. 1429. 240. 967±2000. / ɪ/. 412. 91. 276±717. 2046. 249. 1198±2405. 434. 108. 279±787. 2444. 423. 902±3104. /ɛ/. 606. 111. 311±950. 1823. 183. 1003±2377. 762. 178. 207±1084. 2078. 365. 846±2814. /æ/. 693. 103. 430±979. 146. 1582±2215. 846. 202. 235±1164. 1956. 387. 829±2545. /ʌ/. 728. 123. Nat. 1800. 303±1029. 1368. 164. 936±1696. 807. 243. 224±1144. 1534. 292. 808±2175. /ʊ/. 377. 57. 301±558. 1054. 176. 681±1414. 424. 66. 282±558. 1188. 215. 744±1606. /ɔ/. 633. 127. 309±1070. 161. 774±1643. 752. 189. 1243. 272. 882±2119. 政治大. 學. n. 1094. Ch. engchi U. er. io. al. sit. y. ‧. ‧ 國. Vowel. v ni. 329±1108. Note: Table 2-4 reveals the normalized formant structure of the vowels produced by Taiwan Mandarin speakers. Table 2-4 shows that /i/ and /ɪ/ are two vowels that present quite similar features both in F1 and F2 in the production of Taiwan Mandarin speakers, and such is also the case with /ɛ/-/æ/ and /u/-/ʊ/. Each of these three pairs show a pattern of formant structures that are deviant when compared to those produced by American English native speakers. However, the levels of the English proficiency and the language backgrounds of the Taiwan Mandarin participants in this study are not controlled and 13. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(29) are not equivalent enough to contend for problems we have configured so far. Liao (2006), in addition to the above study, has also provided a model of the English vowels structures produced by Taiwan EFL learners. The problem of both of these studies is that only a general mean value is provided for each vowel, and no values are provided by all of the levels of the learners in the studies. 2.2 Asymmetry of Production and Perception L2 learners show an asymmetrical behavior in the production and perception of vowels (e.g., Best, 1995; Y. Chen et al., 2001; Flege, 1995; Flege, Bohn, & Jang, 1997;. 治政 among many others). A number of scholars have found大 that Taiwan Mandarin speakers 立 ‧ 國. 學. are able to perceive most English vowels but fail to produce them all in a native-like way (e.g., Chang, 2007; L.-M. Chen, Chang, Yang, & Chou, 2006; Lai, 2005; C.-Y. Lin,. ‧. 2013). Although the focus of the research and the test items of these studies differ, they. sit. y. Nat. all contribute to the knowledge of how L2 learners learn vowels. The mastering of. n. al. er. io. tenseness, for example, as well as duration, in English vowels becomes an essential key. i n U. v. for L2 learners since English vowels are mostly paired with a tense and a non-tense. Ch. engchi. counterpart. Mandarin vowels do not differ in tenseness (Lai, 2005, 2010; Y.-H. Lin, 2007). 2.2.1 Tense and Lax Vowels Thus, the differences between the perception and the production of tensed and lax vowels are widely discussed. Learners of English in Taiwan or learners whose L1 does not contain a phonemic difference of tense/lax vowels will tend to differentiate tense vowels from their lax counterparts in English (or vice versa) the duration of the vowels (Hung, 2013). However, there is still a high correlation between the perception of 14. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(30) vowels and the production of vowels (C.-Y. Lin, 2013). This correlation, which may be presented in asymmetrical or symmetrical forms will be part of our focus in studying the performance of the learners in this study. 2.2.2 L1 and L2 Interferences We found evidence that the means of the F1 and F2 of English vowels produced by Taiwanese EFL speakers and those of American native speakers have been shown are different (Lai, 2005, 2010). Also, Taiwanese EFL speakers are more familiar with the front vowels and perform better on them than on other vowels, where those front. 治政大 1989b; Flege, Bohn, et al., vowels existed in their L1 (L.-M. Chen et al., 2006; Flege, 立 ‧ 國. 學. 1997; C.-Y. Lin, 2013). Flege, Bohn, et al. (1997) believed that the degree of the vowel intimacy between the L1 and the L2 would positively influence the learner’s production. ‧. and perception of the vowels in the L2. They made an analysis of Mandarin speakers’. sit. y. Nat. English vowel production, which they evaluated by the F0, F1 and F2 values and also. n. al. er. io. one on perception, which was measured by B0, B1 and B3 (Bark) values. They found. i n U. v. that [ɪ, æ, ɛ] are the three English vowels most likely to fail to produce by Mandarin. Ch. engchi. EFL learners. However, some others (e.g., Hollich, Jusczyk, & Luce, 2001; Locke, 1980) argued that vowel intimacy may also be a barrier to hinder the learner’s ability to create a new phonetic category and that this problem seems to be somewhat influential. Chen (2006) was curious about the correlation between one’s perception and one’s production of vowels and if L1 will interfere L2 at this level. Several students majoring in English were chosen to be participants in her study and eleven vowels were included in an experiment of using production and perception tasks by containing test items that were set in an /h__d/ context, which means that all of the target vowels were put into 15. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(31) that context, and this convention model was first introduced by Fox (1982) and was later adopted by many (e.g., L.-M. Chen et al., 2006; Y. Chen et al., 2001; Flege, 1989b, 1995; Lai, 2005). The results show that it is easier to perceived and to pronounce the front vowels than the mid and back vowels. Also, tense vowels seemed to be less complicated or less unfamiliar for Taiwanese ESL learners than lax vowels, suggesting a language transfer from the L1 to the L2. What is noteworthy here is that for L1 users, back vowels may be steadier; however, some L2 learners, including Taiwanese EFL learners, are less available to perceive or to construct a category for back vowels in L2,. 治政 which may be a barrier caused by the vowel intimacy. 大立 ‧ 國. 學. 2.2.3 Vowel Intimacy. The hindrance caused by the vowel intimacy, thereafter, infers a high-density of. ‧. back vowels, which means that it may be difficult for L1 users to perceive a new. sit. y. Nat. category that is nearby when they have already constructed a robust category for some. n. al. er. io. vowels. This view has long been supported by Locke (1980) in that L2 learners might. i n U. v. use their L1 grid to apply to their L2 grid. American L2 learners of French, for instance,. Ch. engchi. often fail to pronounce the phoneme [y] and use [i] as a replacement. The field of child language acquisition has shed some insight on this. American infants tend to be more familiar with the word patterns ending with ‘-at’ forms, such as cat, hat, mat, rat, vat, etc., while they might be less familiar with those ending with ‘-up’ forms, as cup or pup. Word patterns which end with ‘-at’ are therefore seen as high density, and those which end with ‘-up’ are seen as low density. Infants may be less able to pick up (i.e., acquire) new words occurring in a high-density group and may do so more easily with words in a low-density group, for high-density words are more stable to the perception and a 16. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(32) robust category may have been established for them which it would be difficult to change (Clark, 2016; Hollich et al., 2001). We thus infer that this phenomenon may be carried across to second language acquisition. However, the reason why Taiwan Mandarin speakers tend to perform better on tense vowels may be that they draw a direct analogy from their L1 to L2 but actually fail to construct a real L2 vowel category or have not yet constructed one. Whether this phenomenon is caused by a similarity in the vowel grids or for another reason remains unknown. Chang (2007) sought to study Mandarin speakers’ and Southern-Min speakers’. 治政大 whether their L1 would vowel perception and production of English. She wondered 立 ‧ 國. 學. interfere with or assist them in their acquisition of the L2. Four vowels, /ei/, /e/, /o/ and /ou/ [sic, [e], [ɛ], [o], [ɔ]] were chosen as be the target phonemes, and were claimed by. ‧. Chang to be tense and lax vowel pairs. The reason why she chose these four vowels is. sit. y. Nat. that in Mandarin all of these four vowels can be matched with one phoneme in L1 in. n. al. er. io. Mandarin, while in Southern-Min, there is a gap between [e] and [ɔ]. Chang utilized. i n U. v. contrastive analysis to see whether the differences between L1 and L2 were influential. Ch. engchi. or not. The results showed that there were no significant differences between the Mandarin speakers’ and Southern Min speakers’ perception and production of English vowels. Perception and production do not correlate with each other in the L1 but in the L2. The diphthongs [sic, tense vowels] are more difficult for L2 learners to pronounce than single vowels [sic, lax vowels], which is claimed to be due to the lack of diphthongization. 2.2.4 Influences on the Vowel Space Lin (2013) sought to investigate the perception and performance of the front 17. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(33) vowels made by EFL learners in Taiwan as well as the correlation between the perception and production. She focused on two pairs of front vowels, namely, /i/-/ɪ/ and /ɛ/-/æ/. Previous studies, including Fox (1982) and Baker and Trofimovich (2006), had already revealed that the pronunciation of [i]-[ɪ] and [ɛ]-[æ] may be the most difficult problem for most EFL learners. Though Fox (1982) and some others also indicated that learners’ perception of vowels does correlate to their production behavior, Lin (2013), sought to see if the hypothesis would also fit in EFL learners in Taiwan by conducting both perception and production experiments. The participants were twenty sixth-grade. 治政大the others were male. Both students in Taiwan and half of them were female and 立 ‧ 國. 學. experiments included [i]-[ɪ] pair and [ɛ]-[æ] pair words (mono and simple syllable),. and the results showed that the [ɛ]-[æ] pair was more difficult than the [i]-[ɪ] pair for. ‧. the learners in Taiwan in both the perception and production tasks. At the same time,. sit. y. Nat. the results also revealed that a significant correlation between the learners’ perception. n. al. er. io. and production did exist.. i n U. v. Issues which are commonly discussed as possible factors influencing the L2 are. Ch. engchi. the background of the participants, the learner’s age of arrival, the length of residence and also the amount of use of L1. While all have been claimed to affect the performance of an L2, the age of arrival and the amount of use of the L1 have a greater influence than the other influences (Baker & Trofimovich, 2006; Flege, 1995; Flege, Frieda, et al., 1997; Flege et al., 1999). 2.2.5 L1 Language Transfer L.-M. Chen et al. (2007) investigated the production of English vowels by both high and low achieving Taiwan Mandarin English learners. The high achievers and low 18. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(34) achievers here are respectively college students and elementary school students. These students had all grown up in Taiwan and had received their English education in Taiwan. L.-M. Chen et al. (2007) indicated that more than fifty percent of the Mandarin speakers in their study showed difficulties in distinguishing English vowels. The error rates for the front vowels in their study are higher than those for the back vowels. They also noted that they did not find any evidence that the amount of use of the L1 made an effect on their participants, and that the low achievers in their study did not show a higher degree of improvement than the high achievers. The inference may be a false. 治政大to the L1 of both test groups interpretation of the theory due to the fact that the exposure 立 ‧ 國. 學. is quite similar. Both of the groups, in fact, shared the same language background, and the age of exposure to English should be counted as the one of exposure to an ESL. ‧. environment, for many studies have shown the importance of exposure to native adult. sit. y. Nat. speakers in establishing a native-like accent.. n. al. er. io. Lai (2010) found that Taiwanese EFL learners showed a tendency to assimilate. i n U. v. English vowels such as [i] and [ɪ] in a merge with the Mandarin vowel [i]. Also, she. Ch. engchi. showed that the higher achievement the learner made, the better the discrimination the learner might have. While learners with a lower achievement might assimilate two vowels from the target language to one vowel in the L1 regardless of the tenseness or other features, high achievers might be able to better assimilate the vowels and to discriminate them from those of the L1. 2.2.6 General Discussion of L2 Acquisition As discussed above, we can acknowledge that Taiwanese EFL learners often confuse and mispronounce English tense and lax vowels. Lai (2010) was interested in 19. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(35) the way that learners discriminate English vowels and how they assimilate the L2 vowels to their L1 segments and conducted two perception experiments, including English vowel discrimination and English vowel assimilation experiments. Ninety Taiwanese EFL learners aged from nineteen to twenty-two participated in her study. The participants were divided into two groups according to their level of achievement in English. The first one was a vowel decision task with minimal pairs, and the second asked the participants to transcribe the sound played in Zhuyin Fuhao (aka the Mandarin Phonetic System) if they felt that the sound they had heard was familiar to. 治政大showed that the sensitivity of any sound in Zhuyin. The results of the perception task 立 ‧ 國. 學. both low-achieving EFL (LEFL) and high achieving (HEFL) toward English vowels share a similar pattern. The discrimination tasks show that HEFL may be more sensitive. ‧. to tense/lax vowels, while LEFL may not. In assimilation tasks, misperceptions of. sit. y. Nat. differences were made, and this phenomenon was thus suspected to be the probable. al. n. 2.3 Models. er. io. reason why learners may perform mismatched sounds in tense/lax vowels.. Ch. engchi. i n U. v. A number of studies have tried to investigate this phenomenon through different forms of analysis, techniques and frameworks. The three theories of the Contrastive Analysis Hypothesis, the Speech Learning Model or the Perceptual Assimilation Model are the most famous and most commonly applied analytical frameworks for use in perception and production research, and some have been introduced in the former section of this chapter. Notwithstanding the use of the different models mentioned above, the different stages of progress in language learning and U-Shaped Learning Theory are also widely 20. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(36) applied in related research. We may find a growing trend in utilizing error analysis to discuss how language learners may develop their interlanguage to acquire their idealist target language. In this section, we will also give a brief review on how these theories may be applied in studies. 2.3.1 Contrastive Analysis Hypothesis The Contrastive Analysis Hypothesis compares two languages, which are usually the L1 and L2, and make structural analyses to predict a learner’s potential errors in order to isolate what is important, difficult or easy in second language learning (Gass. 治政大output of language learners. & Selinker, 2008). It is widely used to predict the possible 立 ‧ 國. 學. However, it does not deal with the problem of over-production. Lado (1957) first posited the hypothesis, and it was later applied in many studies. Though the Contrastive. ‧. Analysis Hypothesis has been challenged by many and is notorious for its over-. sit. y. Nat. simplicity and lack of explanation in respect to several aspects of psychological and. n. al. er. io. cognitive issues which need to be taken into consideration, it still plays a crucial role in. i n U. v. the segmental analysis of the performance and production of an L2.. Ch. 2.3.2 Speech Learning Model. engchi. Flege (1995) created Speech Learning Model which postulated the possible stages of production that language learners might go through. The main purpose of the Speech Learning Model is to construct hypotheses to explain the developments in the production of an L2 by learners of the language. It is composed of four postulates and seven hypotheses which illustrate how language learners process their L2 phoneme inventory based on their L1. The core concept of the Speech Learning Model is that language learning should be seen as dynamic and can be separated into stages. Different 21. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(37) stages in the Speech Learning Model composed several hypotheses. The seven hypotheses, labeled one to seven (H1 to H7), normally happen in order, but, in the case of the last few ones, the attainments would depend on the individuals. H1 to H4, for example, proposed that different phones between L1 and L2 may be initially allophonic to L2 learners, and later on, when one can differentiate the phonetic differences and phonemic differences, a new category of vowels would be gradually constructed. For H6 to H7, the representation of the new vowel categories may rely much on the performance in the previous steps as to whether the L2 learners have succeeded in. 治政 making correct judgements on the classification or not.大立 ‧ 國. 學. 2.3.3. Perceptual Assimilation Model. Best (1995), on the other hand, proposed the Perceptual Assimilation Model. ‧. (PAM), which simulates the perceptual process model that a learner may tend to have. sit. y. Nat. to go through when learning a new concept, especially the segmental features in. n. al. er. io. languages. The main purpose of the PAM was to elaborate the speech perception of a. i n U. v. non-native speaker. It elaborates a direct realist model of cross-language speech. Ch. engchi. perception and development. It gives hints for us to realize how L2 learners perceive the L2 segments by the similarities to or discrepancies from the segments of their L1. He raised some assertions such as the following. L2 learners perceive a new sound by assimilating a new sound to one already stored in his/her vowel inventory, dissimilate a new sound from an already stored one and create a new category for it, or do neither for the learners cannot perceive the distinct features of the sound. A number of scholars have then further revised them into different ones at certain level. Below we will review different studies relating to the acquisition or learning of English vowels. 22. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(38) 2.3.4. Application of the models Chang (2007), as mentioned before, used Contrastive Analysis Hypothesis in her study and compared all of the possible pairs which appear in the three languages which she included in her account, i.e., Taiwan Mandarin, Southern-Min and English. After a series of comparison and contrasting analyses, she gave recommendations and suggestions as to how the features were transferred from one to another. Hung (2012) adopted both Flege’s (1995) SLM and Best’s (1995) PAM. She analyzed all of the values retrieved from speakers and cross-examined them with the. 治政大tried to see if the there was similar characteristics performed by her participants. She 立 ‧ 國. 學. any assimilation of each sound and to see if learners dissimilated the different phonemes from each other. Both purposes were done under the SLM and PAM.. ‧. 2.3.5 More on the Developmental Process. sit. y. Nat. As L2 learners construct their language knowledge, they may experience several. n. al. er. io. processing stages (Brown, 2007; Gass & Selinker, 2008; Lennon, 2008; Zobl, 1984).. i n U. v. Zobl (1984) first proposed a learning progress that learners may undergo on their path. Ch. engchi. to picking up a new language. Many have later revised the theory and created their own model of the learning progress (e.g., Brown, 2007; Gass & Selinker, 2008; White, 1998). Though not all of the theories applied their practices in studies related to phonetics and phonology, they may still shed light on our study. Brown (2007) proposed that there may be four different stages of language learning for a language learner: first, a pre-systematic stage, then, an emergent stage, next, a systematic stage, and finally, a post-systematic stage (Brown, 2007; Lennon, 2008; Zobl, 1984). A beginning learner may be unaware of any specific linguistic 23. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(39) features of the L2, and they may make random errors in that language. Then, they may start to establish their own rules in this interlanguage, while there may be some errors on the rules they make. Later, those such rules begin to be self-revised by the learners themselves, and the errors will be more consistent and fewer in number. Finally, they may develop a mature system and only sporadic errors may occur within the interlanguage. Repeated errors may still be seen under this stage for ‘fossilization’ could appear in their process of L2 learning. The U-shape learning theory suggests that L2 learners may seem to produce more. 治政 occasional errors or mistakes in the middle of their way大 towards learning the language 立 ‧ 國. 學. (Bowerman, 1982; Gass & Selinker, 2008; P. M. Lightbown, 1985). That is, the learners may have overgeneralized some rules about the target language and perform as though. ‧. going backwards to a stage that may seem to be the same as that of a previous level.. y. sit. n. al. er. io. 2007).. Nat. This phenomenon may often appear in the emergent stage and systematic stage (Brown,. 2.3.6 ABX Discrimination Task. Ch. engchi. i n U. v. Perceptual examination can be assessed by the use of an ABX discrimination test. The model was first introduced by Fujisaki and Kawashima (1970) to see if a listener can differentiate the difference between item A and item B. Listeners have to listen to stimulus A and stimulus B, and, after a while, stimulus X, and they will have to decide whether stimulus X is more like stimulus A or stimulus B. The participants in an ABX discrimination test relies heavily on their short-term memory, for short-term memory has been claimed to play a crucial role in speech perception (Fujisaki & Kawashima, 1970; Pisoni, 1975). The decision that the listeners make involves a string 24. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(40) of reactions implying that the ABX is a result of retrieving one’s phonetic information. It is also noteworthy that if A and B are the same and listeners will not be informed that the stimuli are identical or not, we can further ask listeners to first identify if A and B are the same phonetic segment. We can see in Figure 2-3 that a listener will have to first identify three stimuli, A, B and X, and then make a comparison among these three. In our study, we will not use such a complicated decision test, but we will inform our participants that A and B will not be identical. The pause between A, B and X may influence the participants’ short-term memory. Therefore, we follow the procedure in. 治政大second, each set is separated Pisoni (1975) in which each stimulus is separated by one 立 ‧ 國. 學. by four seconds, and every set of ten sets is separated by an eight-second pause.. Figure 2-3 is adopted from Pisoni (1975). It represents the process one might go. ‧. through when undertaking an ABX decision task. The participant has to listen to A, B. sit. y. Nat. and then X during the task. If the stimulus is stored in his/her vowel inventory, it will. n. al. er. io. be placed in the phonetic short-term memory, while if it is not, it will be placed in the auditory short-term memory.. Ch. engchi. i n U. v. Figure 2-3 The perceptual processing when participants undertaking a ABX decision task 25. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(41) This figure originally illustrated the use of short term memory, while here we utilize it to explain the operational procedure of our task. 2.4 Research Questions and Hypothesis Though we can see that many studies have been done to investigate vowels produced by Mandarin speakers and some of the vowels perceived and produced by Taiwanese EFL learners, we have not found any which completely analyze the formant value of the full scales of the formant values of Taiwanese EFL learners. Liao (2006) studied the performance of Taiwanese EFL learners, but there was only production. 治政大perception and production of analysis in her study. Those studies which have examined 立 ‧ 國. 學. vowels did not include all of the vowels in their experiments. Also, the backgrounds of the participants in each study vary from each other and the level of the English. ‧. proficiency of the learners ranged widely from research to research. In the present study,. sit. y. Nat. we seek to investigate both the perception and production of English vowels done by. n. al. er. io. Taiwan Mandarin EFL learners. The participants consist of three groups, which are. i n U. v. LEFL, MHFL, and HEFL, for L2 vowel perception improves with one’s experience. Ch. engchi. and ability (Best & Tyler, 2007). A model of the process of the learning of English vowels may be built through this pseudo-longitudinal or cross-sectional experiment. Although the formant structures of the vowels produced by Taiwan Mandarin speakers have been analyzed by Y. Chen et al. (2001), Jian (2007), Lai (2005) and many others, we are still curious about and hold the suspicion that the perception and production of vowels produced by Taiwanese English learners may be systematic and developmental. Such a developmental view would be somewhat consistent under controlled conditions, such as language background, learning environment and use of the L1 and the L2. We 26. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(42) hope to find a correlation of between the learner’s perception and production as well as the symmetry or asymmetry and the model of the developing process of development. In order to be systematic and to preclude any possible variations, all of the hypotheses that we apply and the questions that we ask below are based on the study of performance of Taiwan Mandarin English learners who share mostly the same educational background (at the least, they are all EFL learners and had studied under the system of Taiwan’s standard education system, which includes six years of elementary school, three years of junior high school and three years of high school or five years of junior. 治政 college. This can, therefore, help exclude the variations大 of age of arrival and amount of 立 ‧ 國. 學. L1 use, as these two points have much influence on the attainment of L2 accents. according to the empirical evidence. Following lists the questions that will be explored. ‧. io. sit. y. Nat. n. al. er. in this thesis.. Ch. engchi. i n U. v. 27. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.

(43) 2.4.1 Research Questions Question 1. What is the process of the development of the model of the perception of English vowels in Taiwan Mandarin English learners? Question 2. What is the process of the development of production of the model of the production of English and Mandarin vowels in Taiwan Mandarin English learners? Question 3.. 治政 Is there an asymmetry between the perception大 and the production model? 立 ‧ 國. 學. Question 4.. When developing both the perception and production models for English vowels,. ‧. do the vowel categories of Taiwan Mandarin English learners deviate from those of. sit. y. Nat. their L1?. n. al. er. io. 2.4.2 Hypothesis. i n U. v. We assume that for Research Question One, learners who somewhat share the. Ch. engchi. same or similar education and L1 background will show a similar development in their accent and may go through a similar learning process. Thus, we seek to construct a model to understand the process of the development of vowel perception and production in English that a L2 learner with a background in Taiwan Mandarin may undergo by use of the three experiments adopted in this study. Before we started this study, we conducted a pilot study of the performance of Taiwan Mandarin EFL learner who was qualified to be a participant in our study. Our preliminary study showed that though the participant was quite good at perceiving the 28. DOI:10.6814/THE.NCCU.GIL.003.2018.A07.