台灣華語送氣與非送氣子音之對比：以語音實驗為例證 - 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. 指導教授：萬依萍博士 Advisor: Dr. I-Ping Wan. 立. 政治大. ‧ 國. 學 ‧. 台灣華語送氣與非送氣子音之對比：以語音實驗為例證. n. al. er. io. sit. y. Nat. The status of aspirated and unaspirated consonants in Mandarin: Evidence from phonological experiments. Ch. engchi. i n U. 研究生：李冠霆撰 Student: Kuan-Ting Lee 中華民國一 0 四年六月 June, 2015. v.

(2) The status of aspirated and unaspirated consonants in Mandarin: Evidence from phonological experiments. 立. 政 BY 治大 Kuan-Ting Lee. ‧. ‧ 國. 學 er. io. sit. y. Nat. n. a lA Thesis Submitted to the i v Graduate n C hInstitute of Linguistics U eFulfillment in Partial n g c h iof the Requirements for the Degree of Master of Arts. June 2015.

(3) 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v.

(4) 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. Copyright © 2015 Kuan-Ting Lee All Rights Reserved iii. v.

(5) 致謝詞在寫論文的過程中，不論在哪個階段，總是會想致謝詞要寫什麼，但是等到真的開始寫了，卻又不知該如何將滿滿的思緒化成文字。因為能夠將這本論文完成，實在受到太多人的幫助，心中的感激難以言喻。首先，我要感謝我的指導教授萬依萍老師。在把論文催生出來的這段期間，從題目到方向，從語句到分析討論，如果沒有老師耐心地指導，這本論文根本不可能完成；在對自己極度沒有信心以及焦慮的時候，也謝謝老師的鼓勵及加油打氣，教我要有自信。在研究所的三年期間，不論是身為學生也好，或是身為助理也好，都從老師身上學到許多東西，不論是知識或是待人處事的道理，亦或是做事的高效率，感謝老師的教導，也感謝老師的包容。此外，也非常感謝我的三位口試委員，張郇慧老師、林祐瑜老師、謝佳玲老師。謝謝老師們即使百般忙碌，仍非常乾脆地答應來當口試委員，也給予許多內容、統計、架構等等的建議，讓這本論文更為齊全。在語言所的三年期間，也非常謝謝老師們的教導，謝謝蕭宇超老師、何萬順老師、張郇慧老師、黃瓊之老師、賴惠玲老師、莫建清老師、心理系的蔡介立老師、師大的曾文鐽老師，感謝老師們在學術上以及待人處事上都讓我有很多的收穫及成長。也要特別感謝惠鈴助教，總是耐心地回答問題也熱心的協助許多大小事，還有鼓勵及加油。謝謝所上的學長姐、學弟妹們，也謝謝 101 級的所有同學們，韶君、彥棻、宇涵、翊倫、子貽、雯婷、明哲，因為有你們的幽默、有趣及瘋狂，讓研究所生活增添了許多快樂及鮮豔的色彩，其中特別感謝韶君跟彥棻，辛苦你們常常聽我焦慮及給予協助和建議。除了上課以外的時間，最常待的地方就是語音實驗室了。在這實驗室遇到太多好的工作夥伴：感謝涵絜學姊，在我還是小碩一的時候教我許多事情；感謝心怡學姐，好懷念和妳一起在實驗室或是錄音路上八卦、聊天的時光；感謝好夥伴彥棻，和妳一起分工合作完成許多實驗室的大小事，妳的圓融及有趣讓實驗室生活更加開心；謝謝欣瑩、庭瑄、馨云、佳琪和已經畢業的佳琳，謝謝你們的高效率及超棒的工作能力，讓我在分派工作及交接時輕鬆、放心很多。另外，這本論文能夠完成也要謝謝我的受試者們，不但沒有人放鴿子，甚至還會很熱心的問需不需要幫忙多找一點人；還要特別感謝統計所的胡逢升，不只建議我該跑什麼統計，還常常幫我趕跑統計，如果沒有你的話，這本論文的統計根本生不出來，謝謝你！感謝我大學的老師們，特別是李思穎老師和左偉芳老師，在我大學畢業後依舊給我許多鼓勵跟精神上的支持，也在我茫然無措時溫暖的拉我一把；也謝謝系上的助教和學弟妹們，之前為了期末報告找受試者時，有你們的幫助才能讓我順利寫完報告。還要謝謝我大學好朋友們，謝謝你們的加油、關心和幫忙。更要感謝我的摯友們，泱和果，謝謝你們一直都在，不但幫我許多，也陪我度過低潮和分享心事。. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. iv. i n U. v.

(6) 最後，感謝我的家人，尤其是爸爸、媽媽和姐姐，謝謝你們的栽培和後盾，讓我可以專注在課業上，不用擔心太多；也謝謝你們的關心、期許和包容，總是在聽完我的焦慮跟抱怨後給予我安慰及建議。謝謝大阿姨一家人，你們的關懷讓我每次回家時都可以充飽滿滿能量。謝謝我的外婆，您的愛和關心陪著我度過每個時刻。回首研究所的這三年，有太多挫折、焦慮、緊張等情緒跟關卡，謝謝身邊的所有人，也謝謝自己沒有輕易放棄。不論是現實中或是網路上，謝謝所有跟我說過加油還有口試通過後留言祝福的你們！. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. v. i n U. v.

(7) 國. 立. 政. 治. 大. 學. 研. 究. 所. 碩. 士. 論. 文. 提. 要. 研究所別：語言學研究所論文名稱：台灣華語送氣與非送氣子音之對比：以語音實驗為例證指導教授：萬依萍研究生：李冠霆論文提要內容： (共一冊，15,708 字，分五章 ). 政治大. 送氣與非送氣的對比是台灣華語塞音及塞擦音之特性。為討論送氣的議題，本篇論文透過語音實驗，從兩方面來探討：其一為嗓音起始時間(voice onset time)，另一則為語誤的方向性(directionality)及標記 (markedness)。本研究有兩個實驗，皆以四字非詞為實驗材料。實驗一為語誤實驗，此實驗. 立. ‧ 國. 學. ‧. 結合繞口令以及立即回想的方法來誘發語誤；實驗二則為嗓音起始時間實驗，只讓受試者回想，並未結合繞口令。這兩個實驗的設計由於受到中文本身音韻限制及詞彙空缺的限制，韻母使用了單母音(V)及韻母(VG 或 GV)兩種形式，因此實驗二亦將母音對子音數值的影響之比較結果納入討論。嗓音起始時間實驗的結果顯示，使用四字非詞並結合回想作為實驗內容所測量出的數值與前人使用實詞所測量出的數值並無太大不同，並支持送氣子音的數值較非送氣子音的大。此外，塞音的研究結果支持 Cho & Ladefoged (1999)所提出的原則，也就是塞音除阻的位置若越後面，則其數值會越大；若將塞音的數值套用 Cho & Ladefoged (1999)的分類，則中文的非送氣塞音確為非送氣子音，而送氣塞音的數值則介於稍微送氣及送氣子音之間。塞擦音的研究結果若如 Lai (2013)般套用 Cho & Ladefoged (1999)之原則，則會發現送氣塞擦音符合此原則，但非送氣塞擦音卻非如此。另外，子音後接的母音或韻母亦會影響子音之嗓音起始時間。研究結果顯示當大部份子音後接單母音時，子音之嗓音起始時間較長；研究結果亦指出當子音後接的為單母音[a]或以[a]開頭的韻母時，子音的嗓音起始時間會較接其他母音如[u], [i], [ow]時短。語誤實驗的結果顯示，除了[k], [kh]有以有標送氣塞音取代無標非送氣塞音的情形外，其他組塞音及所有塞擦音皆無以有標送氣子音取代無標非送氣子音 (反之亦然)的情形。就方向性而言，研究結果顯示語誤來源出現在語誤之前 (perseveration)的方向性顯著最多。. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 關鍵詞：送氣/非送氣子音、嗓音起始時間、語誤、方向性、標記 vi.

(8) Abstract The contrast of aspiration is one of the characteristics of stops and affricates in Taiwan Mandarin. Through the conduction of two phonological experiments, we discuss the issue of aspiration from two aspects: one is from voice onset time (VOT), and the other one from directionality and markedness of speech errors. Non-word quadruple sets were used in both two experiments. The first experiment was speech errors, which combined tongue twister and immediate recall to induce speech errors; the second experiment was voice onset time, which only had the subjects to recall. Due to the limitation of phonological ill-form and lexical gaps, the rhymes of the two experiments had two forms, V and VG/GV; as a result, how vowels affected the VOT of stops and affricates were also discussed. The findings of the second experiment (voice onset time) showed that the VOT measured by using non-word quadruple set was similar to the VOT measured by using real words as stimuli, and the findings also indicated that the VOT of the aspirated consonants was longer than that of the unaspirated ones. In terms of the results of the stops, they supported the principle proposed by Cho & Ladefoged (1999) that the duration of VOT has a further back relationship with the closure; furthermore, the categories proposed by Cho & Ladefoged (1999) were applied onto the results, it showed that Mandarin unaspirated stops were indeed unaspirated consonants, and the aspirated stops fell between slightly aspirated and aspirated consonants. The findings of the aspirated stops, if applied the principle proposed by Cho & Ladefoged (1999) like Lai (2013), revealed that aspirated affricates supported the principle, yet the unaspirated ones did not. Regarding the influence of vowels on the VOT, it was found that when the succeeding vowel was a single vowel [a] or rhymes beginning with [a], the VOT of the stops and the affricates would be shorter. The findings of the first experiment (speech errors) indicated that with regard to the stops, the aspirated velar stop tended to replace the unaspirated counterpart, while for the other stops, the probability for marked aspirated stops to replace unmarked unaspirated ones or vice versa was of the same. With regard to the directionality of speech errors, the findings showed that perseverations significantly outnumbered anticipations and exchange.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. Keywords: Aspirated/ unaspirated consonants, voice onset time, speech errors, directionality, markedness vii.

(9) Table of Contents Chapter 1 Introduction ................................................................................................... 1 1.1 Voice onset time ............................................................................................... 1 1.2 Speech errors and markedness effect ............................................................... 2 1.3 The framework of the thesis ............................................................................ 4 Chapter 2 Literature Review .......................................................................................... 6 2.1 Voice onset time ............................................................................................... 6 2.1.1 Definition of VOT................................................................................. 7 2.1.2 Stops...................................................................................................... 8 2.1.3 Affricates ............................................................................................. 15 2.1.4 The effect of other factors on VOT ..................................................... 18 2.1.5 Summary ............................................................................................. 19 2.2 Experimental speech errors ............................................................................ 21 2.2.1 SLIP .................................................................................................... 21 2.2.2 Tongue twister paradigm and immediate serial recall ........................ 22 2.2.3 The directionality of speech errors ..................................................... 24 2.2.4 The markedness effect ........................................................................ 25 2.2.5 Summary ............................................................................................. 28 2.3 Research questions ......................................................................................... 28 Chapter 3 Methodology ............................................................................................... 31 3.1 Subjects .......................................................................................................... 31 3.2 Equipment ...................................................................................................... 31 3.3 Materials ........................................................................................................ 32 3.3.1 Experiment 1 (Speech errors) ............................................................. 32 3.3.2 Experiment 2 (Voice Onset Time) ....................................................... 34 3.4 Procedure ....................................................................................................... 35 3.5 Data analysis .................................................................................................. 38 3.5.1 Experiment 1 (Speech Errors) ............................................................ 38 3.5.2 Experiment 2 (Voice Onset Time) ....................................................... 39 3.5.3 Statistical analysis ............................................................................... 42 Chapter 4 Results and Analysis.................................................................................... 43 4.1 The VOT of Mandarin oral stops and affricates ............................................ 43 4.1.1 The VOT of Mandarin oral stops ........................................................ 44 4.1.2 VOT of stops under different vowel weights and vowel context........ 47 4.1.3 The VOT of Mandarin affricates ......................................................... 54 4.1.4 VOT of affricates under different vowel weights and vowel context . 57 4.2 Speech errors .................................................................................................. 63. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. viii. i n U. v.

(10) Chapter 5 Discussion ................................................................................................... 68 5.1 Voice onset time ............................................................................................. 68 5.2 Speech errors .................................................................................................. 73 5.3 Concluding remarks ....................................................................................... 75 Reference ..................................................................................................................... 78 Appendix A .................................................................................................................. 82 Appendix B .................................................................................................................. 84 Appendix C .................................................................................................................. 86 Appendix D .................................................................................................................. 87 Appendix E .................................................................................................................. 88. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. ix. i n U. v.

(11) List of Tables Table 2.1 The consonant chart of Mandarin Chinese (Lin, 2007) .............................. 11 Table 2.2 The mean VOT values (ms) of Taiwan Mandarin Oral Stops .................... 11 Table 2.3 The mean noise duration (ms) & VOT (ms) of Mandarin Affricates ......... 17 Table 2.4 The research focus of VOT in Taiwan Mandarin ....................................... 20 Table 2.5 The eight markedness relations (Fellbaum, 1983: 294-295) ...................... 25 Table 4.1The mean VOT (ms) and standard deviation (S.D.) of the oral stops ......... 44 Table 4.2 The mean VOT (ms) and standard deviation (S.D.) of affricates ............... 54 Table 4.3 The confusion matrix of the stops and affricates ........................................ 64 Table 5.1 The VOT of stops from this study and previous studies............................. 70 Table 5.2 The VOT of affricates in this study and previous studies ........................... 72. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. x. i n U. v.

(12) List of Figures Figure 2.1 The spectrogram of [tha55] ......................................................................... 7 Figure 2.2 The spectrogram of [tʂha55] ...................................................................... 17 Figure 3.1 The flow chart of a trial in the experiments .............................................. 36 Figure 3.2 The waveforms and spectrograms of dental stops .................................... 40 Figure 3.3 The waveforms and spectrograms of dental affricates .............................. 41 Figure 4.1The mean VOT values of the stops ............................................................ 45 Figure 4.2 The VOT of stops of different vowel weight ............................................ 48 Figure 4.3 The comparison of unaspirated stops of single vowels ............................. 49 Figure 4.4 The comparison of aspirated stops of single vowels ................................. 50 Figure 4.5 The comparison of unaspirated stops of rhymes ....................................... 52 Figure 4.6 The comparison of aspirated stops of rhymes ........................................... 53 Figure 4.7 The mean VOT (ms) of the affricates ....................................................... 55 Figure 4.8 The VOT of affricates of different vowel weight...................................... 57 Figure 4.9 The comparison of unaspirated affricates of single vowels ...................... 59 Figure 4.10 The comparison of aspirated affricates of single vowels ........................ 60 Figure 4.11 The comparison of unaspirated affricates of rhymes .............................. 61 Figure 4.12 The comparison of aspirated affricates of rhymes .................................. 62 Figure 4.13 The tokens on each word position........................................................... 66 Figure 4.14 The directionality of speech errors .......................................................... 67. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. xi. i n U. v.

(13) Chapter 1. Introduction This thesis aims to investigate the aspiration in Taiwan Mandarin. The issue of aspiration will be explored in two aspects: voice onset time and speech errors in terms of directionality and markedness effect. In section 1.1, voice onset time will be briefly. 政治大. introduced; a general presentation of speech errors and markedness effect will be. 立. addressed in section 1.2. The organization of the thesis will be stated in section 1.3.. ‧ 國. 學. 1.1 Voice onset time. ‧ sit. y. Nat. Voice onset time has been often used to measure the duration of stops across. n. al. er. io. languages (Chen, Chao, & Peng, 2007; Chen, Tsay, & Hong, 1998; Cho & Ladefoged,. Ch. i n U. v. 1999; Keating, 1984; Kent & Read, 2002; Lisker & Abramson, 1964; to name just a. engchi. few). The values of voice onset time are capable to efficiently distinguish the acoustic quality, such as voicing and aspiration, of the stops. These values reveal the acoustic characteristics of the consonants of a language, and therefore can be utilized as a tool to set a reference in clinical research (K. Chen et al., 1998; Liu, Ng, Wan, Wang, & Chang, 2007) and to compare with the acoustic performance of the consonants across languages.. 1.

(14) A number of relevant studies in Taiwan Mandarin have probed the voice onset time from various perspectives. Some did a comprehensive investigation of the duration of initial consonants (K. Chen et al., 1998), some examined the duration of consonants and vowels under different situations (e.g., speech rate, utterance unit) (Jeng, 2005), while some others have compared the voice onset time of stops in Taiwan Mandarin vs. in Hakka (H. L. Wu, 2009; Peng et al., 2009).. 政治大. When it comes to affricates, there are more affricates in Taiwan Mandarin than in. 立. English. However, affricates have not been much investigated in Taiwan Mandarin.. ‧ 國. 學. Earlier studies examined affricates from their noise duration (Jeng, 2005; K. Chen et. ‧. al., 1998; Liu et al., 1999; Tse, 1988), while later studies have discussed affricates. Nat. io. sit. y. from the voice onset time perspective in that affricates comprise a stop and a. er. homorganic fricative and thus have the quality of both consonants (H. L. Wu, 2009;. al. n. v i n C h 2011; Lai, 2013).UHence, voice onset time is able Jeng, 2011; Ladefoged and Johnson, engchi to measure the duration of affricates as well.. To sum up, voice onset time can be used to discuss the acoustic quality of stops and affricates. The values of the voice onset time are capable to distinguish aspiration in Taiwan Mandarin.. 1.2 Speech errors and markedness effect. The naturalistic approach and the experimental approach are complementary. 2.

(15) Researchers based on the naturalistic corpus data of speech errors, or conducted experiments controlling certain variables to provide empirical evidence (Stemberger, 1992). To elicit speech errors, researchers have applied different techniques and paradigms, including the SLIP (Spoonerisms of Laboratory Induced Predisposition) technique, a classic tool to elicit spoonerisms, proposed by Baars, Motley, and MacKay (1975) and Motley and Baars (1976), and the tongue twister paradigm,. 政治大. which can efficiently elicit consonant errors (Shattuck-Hufnagel, 1992; Wilshire,. 立. 1999). Furthermore, the tongue twister paradigm has been combined with different. ‧ 國. 學. means such as paced reading and acoustic analysis (Goldrick & Blumstein, 2006),. ‧. Nat. io. sit. Cumming, & Norris, 2007; Wilshire, 1999).. y. immediate serial recall (Acheson & MacDonald, 2009; Ellis, 1980; Page, Madge,. er. The elicitation of speech errors has not only been proven to be identical to the. al. n. v i n C h & MacDonald,U2009; Ellis, 1980; Page et al., spontaneous speech errors (Acheson engchi 2007) but also been employed to provide evidence for the markedness effect. Concerning the markedness of the consonants, those consonants that are less frequent. in the language or more difficult to articulate are considered more marked (Hume, 2011); concerning the markedness of voicing and aspiration of the consonants, unaspirated stops are less marked than aspirated stops, and voiceless stops are less marked than voiced stops (Fellbaum, 1983). The relevant speech error studies on the. 3.

(16) effects of markedness in Taiwan Mandarin investigated the initial consonants regarding place of articulation (Wan, 2002) and nasal codas (Hsu, 2011). Wan (2002) found that consonants at different places of articulation tend to substitute for unmarked coronals, while Hsu (2011) found that in the coda position, it is the marked velar nasal that tends to be replaced. The purpose of this thesis is to investigate the voice onset time of the stops and. 政治大. the affricates in Taiwan Mandarin. In addition, the present study aims to elicit speech. 立. errors and to figure out the substitution pattern of markedness effect regarding. ‧ 國. 學. aspiration in Taiwan Mandarin. Therefore, two experiments would be conducted by. ‧. using non-word materials as stimuli.. y. Nat. n. er. io. al. sit. 1.3 The framework of the thesis. i n U. v. In Chapter 2, the relevant studies will be reviewed; in section 2.1, the definition. Ch. engchi. and studies related to voice onset time of stops and affricates will be discussed, and in section 2.2, the experimental studies of speech errors and the markedness effect will be addressed. Finally, in section 2.3, the research questions of the present study will be stated. In Chapter 3, section 3.1 will briefly introduce the background of the subjects for the present study, and in section 3.2, the equipment applied in the two experiments will be described. The materials of the experiments will be elaborated in section 3.3, and the procedure of the experiments will be addressed in section 3.4. In 4.

(17) section 3.5, the data analysis regarding the measurement of the voice onset time and the transcription of speech errors will be discussed. The results and analysis will be in Chapter 4. Section 4.1 will display the values of the voice onset time for the stops and the affricates, and in section 4.2, the speech error analysis will be provided. The discussion and conclusion of this study will be presented in Chapter 5.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 5. i n U. v.

(18) Chapter 2 Literature Review The present study investigates the voice onset time (henceforth, VOT) and the markedness effect of Mandarin unaspirated and aspirated stops and affricates in terms of experimental elicited data. The VOT of either focusing on one language or cross-linguistically has been studied following the pioneer of Lisker and Abramson. 治政 (1964), while the examination of speech errors via 大experimental induction has 立 ‧ 國. 學. flourished since Baars, Motley and MacKay (1975) and Motley and Baars (1976) proposed the classic SLIP technique. Therefore, in this chapter, previous studies. ‧. concerning the VOT, experimental studies of speech errors, and markedness effect on. sit. y. Nat. io. n. al. er. the consonants will be discussed. In section 2.1, the basic idea and definition. i n U. v. regarding VOT and the research of VOT on stops and affricates will be reviewed. In. Ch. engchi. section 2.2, previous studies on experimental induction of speech errors will be addressed; and section 2.3 will display the relations of markedness and previous studies related to markedness in Mandarin will be reviewed. The research questions for this thesis will be addressed in section 2.4. 2.1 Voice onset time In this section, the issues of VOT on stops and affricates will be introduced. First, the definition and basic ideas of VOT will be introduced in 2.1.1. Discussions on 6.

(19) stops cross-linguistically and of Mandarin will be in 2.1.2, while those on affricates will be illustrated in 2.1.3. The effects of other factors on VOT will be addressed in 2.1.4, and the summary of this section will be in 2.1.5. 2.1.1 Definition of VOT Voice onset time (VOT), referred to as the time duration between the burst release of the initial stop and the voicing onset or vocal fold vibration of the following. 政治大. vowel (Ladefoged & Johnson, 2011; Lisker & Abramson, 1964), has been regarded as. 立. a convincing tool to measure and discuss the acoustic properties of consonants. The. ‧ 國. 學. VOT values can be clearly measured through the waveform and the spectrogram (see. ‧. Figure 2.1).. n. er. io. sit. y. Nat. al. Ch h. engchi. i n U. v. Figure 2.1 The spectrogram of [t a55] 1. The spike on the spectrogram or the short vibration on the waveform indicates where the burst of the stop starts, and it is followed by the stop gap, the noise (aspiration), and then the formants of the vowel (Kent & Read, 2002). This figure shows the spectrogram of [tha55] in Mandarin. As it is clearly shown in the. 1. The spectrogram is analyzed using KayPANTAX, provided by Phonetics and Psycholinguistics Lab at National Chengchi University. 7.

(20) spectrogram, there is a spike indicating the release of the stop. The VOT is measured starting from this spike to the onset of the voicing, as represented in the red block of the figure. Kent and Read (2002), in addition, defined the perceptual components of VOT as four acoustic events: transient, frication interval, aspiration, and onset of voicing. As an important phonetic cue, the values of VOT distinguish the voicing and the. 政治大. aspiration of syllable-initial consonants, and they may vary depending on various. 立. factors such as the place of articulation or speech rate.. ‧. ‧ 國. 學. 2.1.2 Stops. In relation to the stop consonants, they have been described in different ways. On. Nat. io. sit. y. one hand, there are three acoustic or phonetic phases of syllable-initial stops: closure,. er. release (which can be further divided into aspirated/unaspirated), and formant. al. n. v i n transition (Kent & Read, 2002).COn hand, stops can be further investigated h the e nother gchi U in terms of voicing and aspiration, and they contrast variously in different languages. Lisker and Abramson (1964) did a classic cross-linguistic study and classified the eleven languages into three groups based on the number of the contrast of the stops; they were “two-category languages” 2 (e.g., German, contrast in voicing, having voiced and voiceless unaspirated stops), “three-category languages” (e.g., Korean),. 2. Two-category languages refer to those languages that the stops contrast in either voicing or aspiration; i.e. voiced stops, voiceless aspirated stops, voiceless unaspirated stops 8.

(21) and “four-category languages” (e.g., Hindi). In their study, they introduced the notion of “voicing lead” and “voicing lag” of VOT. Voicing lead is represented by negative values, for the voicing starts before the stop release, while voicing lag is represented by positive values, for the voicing starts after the stop release. Voicing lag can be further classified as “short lag” and “long lag” depending on how late the voicing quality is; short lag ranges around 0-25 ms, and long lag ranges higher than 35 ms.. 政治大. (Auzou et al., 2000; Keating, 1984; Kent & Read, 2002). As a result, VOT is able to. 立. differentiate unaspirated stops and aspirated stops.. ‧ 國. 學. In their study, Lisker and Abramson (1964) measured the VOT of the initial stops. ‧. in 11 languages and pointed out that it possesses the function of distinguishing. Nat. io. sit. y. voicing, aspiration and the place of articulation effectively. Their result showed that. n. al. er. the VOT ranges for the stops in two-category languages are -125~-75ms (voiced),. Ch. engchi. 0-25ms (voiceless unaspirated), and 60-100ms. iv n (voiceless U. aspirated). Moreover,. regarding aspiration, they found that the VOT values of aspirated stops are longer than unaspirated stops; when it comes to the place of articulation, the values of the velar stops are the longest among labials, apical (dental and alveolar), and velar. Their classic study inspired many researchers to study VOT in different languages. Cho and Ladefoged (1999) integrated previous research findings and proposed three major principles of VOT regarding place of articulation. First, the duration of VOT has a. 9.

(22) further back relationship with the closure. Second, the duration of VOT changes with the degree of extension of the contact area. Third, the duration of VOT changes with the speed of the shift of the articulator. Therefore, the VOT values of the world languages are supposed to generally follow the principle that the values of the alveolar stops are shorter than those of the velar stops, but are higher than those of the bilabial stops (Cho & Ladefoged, 1999; Kent & Read, 2002). While the universal. 政治大. VOT values seem to follow the principle, in the study of Cho and Ladefoged (1999),. 立. the VOT of word-initial voiceless aspirated and voiceless unaspirated stops across 18. ‧ 國. 學. languages was measured. It was reported that the VOT values of velar stops are the. ‧. longest among bilabial and coronal stops; however, in some languages, the difference. Nat. io. sit. y. of the values of bilabial stops and coronal stops do not reach significance, and VOT. er. values are believed to be language-specific. Besides, their study provided the. al. n. v i n C hunaspirated and aspirated boundaries of the VOT values for e n g c h i U stops and classified them into four categories as well: 30 milliseconds for unaspirated stops, 50 milliseconds for. slightly aspirated stops, 90 milliseconds for aspirated stops, and over 90 milliseconds for highly aspirated stops. Mandarin oral stops are voiceless, and they are two-way contrasted in aspiration: voiceless aspirated stops and voiceless unaspirtaed stops. Table 2.1 displays the consonant inventory of Mandarin from Lin (2007).. 10.

(23) Table 2.1 The consonant chart of Mandarin Chinese (Lin, 2007) bilabial stop. labiodental. dental. ph. p. f. affricate m. (central) approximant. w ɥ. palatal. velar k. ʂ. s ts. nasal. alveolopalatal. th. t. fricative. postalveolar. ts. h. tʂ. ʐ tʂ. h. ɕ. kh. x h. tɕ. tɕ. ŋ. n ɻ. (lateral) approximant. j ɥ. w. l. 政治大 As Table 2.1 shows, the six oral stops in Mandarin are as follows: /p/, /p /, /t/, /t /, 立 h. h. ‧ 國. 學. /k/, /kh/, and they could be further classified as bilabial stops, dental stops, velar stops, based on their place of articulation. In addition, the oral stops can only occur. ‧ sit. y. Nat. syllable-initially.. n. al. er. io. The VOT of the stops has been investigated a lot, either in Beijing Mandarin or. i n U. v. in Taiwan Mandarin. Z. Wu (1987) and H. Liu et al. (2007) discussed the VOT of the. Ch. engchi. stops in Beijing Mandarin, and the VOT of the consonants of Taiwan Mandarin has been probed into since Chen, Tsay, and Hong’s (1998) comprehensive study. The VOT values of the stops in Taiwan Mandarin have then been investigated by several studies (H. L. Wu, 2009; H. M. Liu et al., 1999; Jeng, 2005; Lai, 2013; L. M. Chen et al., 2007; Peng, 2009); the figures of the studies are shown in Table 2.2.. Table 2.2 The mean VOT values (ms) of Taiwan Mandarin Oral Stops [p]. [ph]. [t] 11. [th]. [k]. [kh].

(24) Chen, Tsay, & Hong (1998). 13.9. 74.3. 14.7. 81.2. 24.2. 88.5. H. M. Liu et al. (1999). 9. 72. 14. 74. 24. 83. Jeng (2005). 11. 80. 19. 68. 23. 87. Chen, Chao, & Peng (2007). 13.9. 77.8. 15.3. 75.5. 27.4. 85.7. H. L. Wu (2009). 13.60. 62.08. 14.97. 62.21. 34.51. 79.11. Peng (2009). 14.68. 89.4. 14.73. 86.52. 31.05. 109.21. Lai (2013). 47.69. 91.99. 62.04. 171.40. 62.63. 168.40. Z. Wu (1987) explored the issue of Beijing Mandarin’s prevocalic stops and. 政治大. affricates’ duration from two aspects; one was from biological point of view by. 立. measuring the pressure and airstream in the vocal tract, and the other was from. ‧ 國. 學. acoustic point of view through the analysis of intensity and spectrograms. His results. ‧. showed that the duration time of unaspirated stops are 8 times shorter than that of. y. Nat. er. io. sit. aspirated stops. However, his study only provided the mean duration of each stop and affricate under different vowel context instead of an averaged mean figure of each. n. al. Ch. engchi. i n U. v. stop and affricate. H. Liu et al. (2007) analyzed and compared the VOT of the Beijing Mandarin stops by esophageal speakers and normal speakers in terms of place of articulation and aspiration. The figures of each individual consonant were not provided; however, it was reported in their study that regarding normal speakers, the VOT of the velar stops is significantly the longest among initial oral stops, and the values of the unaspirated stops is claimed to be shorter compared with that of the aspirated ones; the findings of which correspond to the general rules and principles.. 12.

(25) Concerning the VOT of Taiwan Mandarin, K. Chen et al. (1998) did a comprehensive survey over the duration time on the initial consonants. The findings suggested that the duration varies due to some factors. Aspiration is one of the two main factors that the duration of the unaspirated sounds (e.g., [p]) is shorter than that of the aspirated counterparts (e.g., [ph]). Besides, the duration is influenced by the place of articulation as well. Take unaspirated stops for example; the duration of labial stop [p] is shorter. 政治大. than that of dental stop [t], which is shorter than that of velar stop [k]. The finding of. 立. which is in line with the principle proposed by Cho and Ladefoged (1999).. ‧ 國. 學. Other research studied the VOT in different aspects; some of them only explored. ‧. the issue in Taiwan Mandarin (H. M. Liu et al., 1999), some of them compared the. Nat. io. sit. y. VOT of stops in Taiwan Mandarin vs. the VOT of English stops produced by. er. Taiwanese native speakers (L. M. Chen et al., 2007), or VOT of Taiwan Mandarin vs.. al. n. v i n C2009; Hakka (H. L. Wu, 2009; Peng, et al., 2009), and the others of them h e nPeng gchi U. compared the VOT of Taiwan Mandarin stops and affricates produced by native Taiwan Mandarin speakers and produced by international learners learning Taiwan Mandarin as their second language (Lai, 2013). The methods of these studies were alike that disyllabic word lists were created and the participants were asked to read them. As a result, most of the results of the Mandarin VOT in previous studies were similar, as in Table 2.2. From their results, the VOT can be further interpreted. 13.

(26) regarding aspiration and place of articulation. With respect to aspiration, the VOT of aspirated stops are longer than those of unaspirated ones; the findings of which confirm the cross-linguistic studies found in the world languages (H. Liu, 2007; H. L. Wu, 2009; H. M. Liu et al., 1999; Jeng, 2005, 2011; K. Chen et al., 1998; L. M. Chen et al., 2007; Peng, 2009; Z. Wu, 1987). All VOT of the aspirated stops from previous studies were higher than 35 milliseconds, proving that the aspirated stops in Taiwan. 政治大. Mandarin fall in the ‘long lag’ category. With respect to the place of articulation, the. 立. VOT values of previous studies are similar. As Table 2.2 shows, except that the. ‧ 國. 學. figures provided by Lai (2013) are apparently higher, which might be due to the. ‧. design of the test materials and the location of the target words, the VOT values of. Nat. io. sit. y. each consonants from the rest of the studies generally fall within a range: [p] between. er. 9 and 14.68 milliseconds, [ph] between 62.08 and 89.4 milliseconds, [t] between 14. al. n. v i n and 19 milliseconds, [t ] fromC62.21 86.52 milliseconds, [k] from 23 to 34.51 h e ton g chi U h. milliseconds, and [kh] from 79.11 to 109.21 milliseconds. The figures from Table 2.2 show that in previous studies, the VOT values of the velar stops are significantly the longest and those of the bilabial stops being the shortest, the findings of which conform to the principle. However, the findings of Jeng (2005) and L. M. Chen et al. (2007) revealed that the VOT values of the voiceless aspirated stops do not follow the principle; instead, voiceless aspirated dental stops have shorter VOT values than. 14.

(27) voiceless aspirated bilabial stops do. 2.1.3 Affricates Affricates, such as [dʒ] or [tʃ] in English, are phonemes that consist two segments and are a composition of stop and homorganic fricative (Kent & Read, 2002; Ladefoged & Johnson, 2011). Kent and Read (2002) stated that affricates have the acoustic properties of stops and fricatives—affricates are equivalent to stops that there. 政治大. is “a period of complete obstruction in the vocal tract,” and are equivalent to fricatives. 立. that there is “a period of frication” but the length is shorter than that of the fricatives.. ‧ 國. 學. Since affricates have the acoustic quality of stops, VOT is believed to serve as an. ‧. efficient tool to do the measurement. Yiu (2008) measured the closure period, fricative. Nat. io. sit. y. period, and VOT of Cantonese stops ([t], [th]), fricatives ([s]), and affricates ([ts],. er. [tsh]). Her results were in accordance with the general rule that the VOT values of the. al. n. v i n Clonger aspirated stops and affricates are their counterpart: the VOT values of [t] h e nthan gchi U are 52.91 milliseconds, [th] being 69.11 milliseconds, [ts] being 79.80 milliseconds, and [tsh] being 106.01 milliseconds.. Mandarin, similar to Cantonese, has more affricates than English does. Unlike English affricates, which contrast in voicing, Chinese affricates contrast in aspiration. The affricates in Chinese involve dental affricates, [ts], [tsh], post-alveolar affricates, [tʂ], [tʂh], and alveolo-palatal affricates, [tɕ], [tɕh], as shown in Table 2.1.. 15.

(28) Contrary to the numerous amount of research in stops, affricates in Chinese have not been much discussed. Z. Wu (1987) pointed out that, in Beijing Mandarin, the duration of voiceless unaspirated affricates is three times shorter than that of voiceless aspirated affricates, inasmuch as the time for aspiration is condensed by the presence of the fricative part. Regarding Taiwan Mandarin, past research has probed the issue of affricates by measuring the noise duration. Tse (1988) measured the noise duration. 政治大. on Mandarin fricatives and affricates and some of the findings were as follows: (1) the. 立. unaspirated affricates are produced to have shorter mean duration than its counterpart,. ‧ 國. 學. (2) voiceless unaspirated affricates and voiceless aspirated affricates have shorter. ‧. mean duration than that of voiceless fricatives. Later studies measuring the noise. Nat. io. sit. y. duration demonstrated corresponding results (H. M. Liu et al., 1999; Jeng, 2005; K.. er. Chen et al., 1998). While past research studied the affricates via the measurement of. al. n. v i n C hown the acoustic property the noise duration, since affricates e n g c h i U of stops, recent research has started to study the affricates by measuring the VOT (H. L. Wu, 2009; Lai, 2013). An example of Mandarin word [tʂha55], with a word-initial affricate, is presented in Figure 2.2.. 16.

(29) Figure 2.2 The spectrogram of [tʂha55]. Mandarin affricates show that an affricate is composed of a stop ([t]) together with a fricative ([s], [ʂ], [ɕ]). That is, there exists a spike representing the burst of stop,. 政治大. followed by the fricative noise and aspiration in its spectrogram, as shown in the. 立. block of Figure 2.3.. ‧ 國. 學. Presented in Table 2.3, the numbers show that by measuring the VOT the results. ‧. Nat. y. are not much different from those done by measuring the noise duration.. [tsh]. [tʂ]. [tʂh]. [tɕ]. [tɕh]. al. 166.3. 87. iv n114.6. 78.6. 145.8. K. Chen et al. (1998). 72.0. 123.5. 84.3. n. Tse (1988). 58.9. C h117.0 55.6 U engchi. H. M. Liu et al. (1999) VOT. 131. er. [ts]. io. Duration. sit. Table 2.3 The mean noise duration (ms) & VOT (ms) of Mandarin Affricates 148. 120. 141. Jeng (2005). 90. 150. 83. 153. 84. 140. H. L. Wu (2009). 66.98. 101.68. 62.46. 118.87. 79.86. 129.18. Lai (2013). 113.63. 183.57. 113.21. 157.57. 114.63. 170.69. The figures in Table 2.3 show that the variation of the affricates is large, for the range of the duration and of the VOT are not small. Except the apparently higher figures of Lai (2013), the duration of unaspirated dental affricate [ts] ranges from 58.9 17.

(30) milliseconds to 90 milliseconds, aspirated dental affricate [tsh] from 101.68 milliseconds to 166.3 milliseconds, unaspirated post-alveolar [tʂ] from 55.6 milliseconds to 87 milliseconds, aspirated post-alveolar [tʂh] from 114.6 milliseconds to 153 milliseconds, unaspirated alveolo-palatal [tɕ] from 72 milliseconds to 84 milliseconds, and aspirated alveolo-palatal [tɕh] from 123.5 milliseconds to 145.8 milliseconds.. 政治大. The figures reveal that there exists a big variation across studies; however, there. 立. are two points in common: aspiration and place of articulation. Regarding aspiration,. ‧ 國. 學. the results are consistent with the general rule that the aspirated affricates have longer. ‧. noise duration/ VOT than unaspirated ones; regarding place of articulation, the results. Nat. io. sit. y. correspond to what Tse (1988) and Lai (2013) found –the variation between different. er. places of articulation of voiceless unaspirated affricates do not reach the level of. al. n. v i n significance, so does it betweenC those aspirated affricates. h eofnvoiceless gchi U 2.1.4 The effect of other factors on VOT. The results of previous studies indicated that some factors would influence the duration time/ VOT values of either stops or affricates, such as speech rate, the position of the target, or vowel context. It has been reported that when the stops or affricates are produced under the context of continuous speech, either in sentences or paragraphs, the VOT values would be shorter than those when they are in isolated. 18.

(31) words or isolated syllables (Howell & Rosen, 1983; Lisker & Abramson, 1964). Concerning the speaking rate, Jeng (2005, 2011) investigated the duration of Taiwan Mandarin consonants under different speech rate. Her results showed that for the affricates, the noise duration would be shorter if the speech rate turns faster; for the aspirated stops, the VOT would be shorter if the speech rate turns faster, and the circumstance of which, however, does not occur to unaspirated stops.. 政治大. Apart from these two factors, vowel context affects the VOT values as well.. 立. However, it has not been discussed much. K. Chen et al. (1998) proposed that the. ‧ 國. 學. vowel context has influence on the duration of the consonants. Previous studies in. ‧. Taiwan Mandarin focused on the influence on stops brought by single vowels [i], [u],. Nat. io. sit. y. and [a] (L. M. Chen et al., 2007; Peng, 2009). Findings from the two studies both. er. suggested that the VOT values of stops are affected by the vowel height (i.e. highness. al. n. v i n C hhave shorter VOTUwhen under the context of low or lowness); in other words, stops engchi vowel [a], and they have longer VOT when under the context of high vowels [i] and [u]. 2.1.5 Summary VOT is an efficient tool to measure the stops and affricates. Table 2.4 lists the previous studies of Taiwan Mandarin reviewed above displaying on stops or affricates or both their research focus are.. 19.

(32) Table 2.4 The research focus of VOT in Taiwan Mandarin unaspirated stops. aspirated stops. Tse (1988) K. Chen et al. (1998). v. v. H. M. Liu et al. (1999). v. v. Jeng (2005). v. v. L. M. Chen (2007). v. v. Peng (2009). v. H. L. Wu (2009). v. Lai (2013). v. 立. unaspirated affricates. aspirated affricates. v. v. v. v v. v. v. v 政 v 治大v. v. v. v. v. ‧ 國. 學. From the data of previous studies on the stops of Taiwan Mandarin, regarding the. ‧. sit. y. Nat. unaspirated stops, the results of previous studies showed that the velar stops are. n. al. er. io. apparently longer the dental ones, and that the dental ones are longer than the bilabial. i n U. v. ones. Unlike unaspirated stops, the VOT values of aspirated stops conform to the. Ch. engchi. principle that the VOT values of the velar ones are the longest, but there shows no tendency for that of dental ones to be significantly longer than that of bilabial ones. Moreover, the aspirated stops [ph], [th], [kh], all longer than 35 milliseconds, are proved to fall at the long-lag category. With respect to Mandarin affricates, the data of each affricate vary across different studies, but they conform to the rule that the VOT values of aspirated affricates are longer than those of the unaspirated ones. Moreover, the VOT of stops and affricates would be influenced by factors such as speech rate 20.

(33) and vowel context. 2.2 Experimental speech errors In this section, the techniques of experimental speech errors will be introduced. The classic SLIP technique will be introduced in 2.2.1, the tongue twister paradigm and the combination of tongue twister and immediate serial recall will be addressed in 2.2.2. The directionality of speech errors will be discussed in 2.2.3, the markedness. 政治大. effect will be introduced in 2.2.4, and the summary section will be in 2.2.5.. 立. 2.2.1 SLIP. ‧ 國. 學. Pioneering in inducing speech errors, Baars et al. (1975) and Motley and Baars. ‧. (1976) devised the classic SLIP technique as a means of error elicitation. The. Nat. io. sit. y. technique aimed to induce spoonerisms from the subjects. The subjects were shown. er. word pairs one by one; the word pairs first primed the subjects with a certain order of. al. n. v i n initial consonants, and the orderCwas U pair, then right after the pair h ereversed n g caththei last the subjects had to respond to the last word pair. For example, the subjects were shown a list give book, go back, get boot, bad goof, RESPOND. It turned out that the. subjects would answer gad boof, reversing the order of the initial consonants. Innovated by the SLIP technique and prompted by the advantages of eliciting speech errors, to investigate speech errors in an experimental way has sprung up. By eliciting speech errors can the researchers control certain variables such as the types. 21.

(34) of the error and their frequency, and the researchers can also avoid slips of the ear (Stemberger, 1992). The results of the experiments not only are proved to be similar to those of the spontaneous speech errors but also could test the hypothesis and provide empirical evidence (Baars, 1980, 1992; Baars & MacKay, 1978; Levitt & Healy, 1985; Motley & Baars, 1976; Stemberger, 1992; to name just a few). 2.2.2 Tongue twister paradigm and immediate serial recall. 政治大. Besides carrying out SLIP to elicit spoonerisms, the tongue twister paradigm has. 立. also been used as another technique to elicit speech errors. The results of the. ‧ 國. 學. read-aloud and recall of the tongue twister experiments of Shattuck-Hufnagel (1992). ‧. showed that when two segments are both at the initial position of the word, the errors. Nat. io. sit. y. would be more likely to occur than when the two segments are not at the same word. er. positions, and the errors would more likely to be induced as well when the two. al. n. v i n segments of the word pair shareCstress. Wilshire (1999) designed tongue twisters in a hen gchi U form of four- syllable list; the initial consonants were organized in the form of ABBA, and the coda consonants in the form of ABAB. The subjects repeated the tongue twisters on the computer screen. The results showed that the errors are likely to occur if the two competing consonants are at the same word position; that is, the word-initial consonant error has its source from another word-initial position, so does the word-final consonant error; the error rates are higher in the word-initial position. 22.

(35) than in the word-final position. Moreover, the phoneme similarity contributes to the error rates as well. The error rates of word-initial consonants are higher when the consonants share as many distinctive features as possible. The author also suggested that the induced slips are phonological errors based on the results above and on that errors are mostly of consonants (instead of vowels) and of single linguistic unit. To study how speech errors are involved in speech production and to present. 政治大. more powerful evidence, tongue twister paradigms have been combined with other. 立. testing methods such as immediate recall. Speech errors induced from immediate. ‧ 國. 學. recall of tongue twisters have been proven to be phonological errors. The findings of. ‧. Ellis’s (1980) three experiments not only confirmed this argument but also showed. Nat. io. sit. y. some similar patterns to the results presented above. First of all, the segments,. er. particularly consonants, are prone to exchange than vowels are; among the exchanged. al. n. v i n consonants, the error rates are C higher that share more distinctive features h e for n gthose chi U. (“feature similarity effect”). Moreover, the results showed “syllable position effect” and “contextual similarity effect” as well: the replaced segment could find its source at the same syllable position, and the frequency is higher for an exchanged consonant to find its source at another word with the same vowels or rhyme (cf. Acheson & MacDonald, 2009; Page, Madge, Cumming, & Norris, 2007). Page et al. (2007) changed the procedure of the experiments from Ellis (1980) and proved the argument. 23.

(36) that the elicited speech errors and phonological errors are related in an opposite way. Acheson and MacDonald (2009) proved the equivalence of spontaneous phonological errors and errors from immediate recall, and the syllable position effect and contextual similarity by further examination from experiments like immediate typed-recall and immediate serial recognition as well. 2.2.3 The directionality of speech errors. 政治大. When it comes to the contextual substitution of speech errors, the directionality. 立. of the speech errors might be considered. The directionality of speech error includes. ‧ 國. 學. anticipation, perseverance, and exchange (Fromkin, 1971, 1973; Jaeger, 2004).. ‧. Anticipation errors are those errors that their source occurs after them, while. Nat. io. sit. y. perseverance errors are the errors that their source comes before them, and exchange. er. errors are the transposition of the two linguistic units. Relevant studies concerning the. al. n. v i n errorsCshowed results. h e ncontradictory gchi U. directionality of speech. It was reported that. anticipations are more frequent than perseverations by Berg (1987), Fromkin (1971, 1973), Nooteboom (1973) and Jaeger (2004) from their corpus studies. However, the results from experimental studies revealed that perseverations exceed anticipations (Acheson & MacDonald, 2009). In addition, studies of tone languages such as Mandarin (Wan, 1999, 2007) and Thai (Gandour, 1977) revealed that the numbers of perseverations of tones and segments exceeded those of anticipations.. 24.

(37) 2.2.4 The markedness effect Markedness theory has been studied in these few years, and the terms ‘unmarked’ and ‘marked’ has been discussed a lot including in speech error studies (For aphasic speech and apraxia speech: Blumstein, 1973; Marquardt, Reinhart, & Peterson, 1979; Wolk, 1986; for normal speech: Hsu, 2011; Shattuck-Hufnagel & Klatt, 1979; Wan, 2002). When it comes to ‘unmarked,’ it is believed to be articulatorily simple, more. 政治大. frequent, and acquired earlier, while ‘marked’ is believed to be the opposite; that is, it. 立. is difficult to articulate, less frequent, complex, and is acquired later (Hume, 2011).. ‧ 國. 學. Concerning the issue of markedness in phonetics and phonology, Fellbaum (1983:. ‧. 294-295) summarized eight types of implicational universals based on previous. Nat. sit. n. al. er. io. markedness relations.. y. research and further proposed eight markedness relations. Table 2.5 lists the eight. Ch. engchi. i n U. v. Table 2.5 The eight markedness relations (Fellbaum, 1983: 294-295) (1) (2) (3) (4) (5) (6) (7) (8). more marked. less marked. voiced stops voiced fricatives fricatives aspirated stops palatalized consonants word-final voice constrast word-medial voice contrast syllable-initial aspirated stops in word-medial position word-final voiced stops. voiceless stops voiceless fricatives stops unaspirated stops unpalatalized consonants word-medial voice contrast word-initial voice contrast syllable-initial aspirated stops in word-initial position word-final voiceless stops. 25.

(38) Among the eight markedness relations in Table 2.5, the first five relations are claimed to be context-free cross-linguistically, while the remaining three are not. Whether the relations are context-free or not, they provide evidence suggesting those that take less effort to articulate are unmarked, and those that take more effort to articulate are marked. A number of research have looked into the markedness effect in terms of speech. 政治大. errors. Shattuck-Hufnagel and Klatt (1979) explored the distinctive features and. 立. markedness of consonants by examining the speech errors via MIT and UCLA. ‧ 國. 學. corpora. They found the most frequent phonological units that are more likely to. ‧. substitute for one another are segments instead of features. Moreover, they found that. Nat. io. sit. y. the replacement between the four segments [s], [ʃ], [t], [tʃ] shows asymmetry. [s]. n. al. er. tends to be replaced by [ʃ] and [tʃ], while [t] tends to be replaced by [tʃ], the finding of. Ch. engchi. which is claimed to be due to the influence of. iv n “palatalization” U. and reveals that. marked segments are not to be substituted for unmarked segments. Previous studies concerning the markedness effect on Mandarin consonants put emphasis on place of articulation (Wan, 2002; Hsu, 2011). Both studies discussed the consonants from the unmarked/marked point of view by supporting evidence from speech errors. From her corpus of naturalistic spontaneous speech error, Wan (2002) examined the substitution pattern of Mandarin consonants at onset position. The. 26.

(39) results showed that dentals (i.e. coronals, the unmarked place of articulation) have the highest rates to be replaced by not only other dental consonants but also by other consonants at different place of articulation. Contrary to Wan’s (2002) speech error investigation on the initial consonants, Hsu (2011) examined the Mandarin nasal codas, [n] and [ŋ]. To investigate whether it is the unmarked labial-dental (i.e. coronal) [n] to be replaced by marked velar [ŋ] or. 政治大. vice versa, she collected natural speech error data and the elicited speech error data. 立. from the experiments. The results of data from her corpus and elicited data both. ‧ 國. 學. suggested that the marked velar [ŋ] has the inclination to be substituted by the. ‧. unmarked labial-dental [n]. The markedness effect on consonants shows different. Nat. io. sit. y. substitution pattern on onset and coda position that at word-onset position, the. er. consonants of unmarked place of articulation are prone to be replaced by consonants. al. n. v i n from other place of articulation;C ath word-coda position, e n g c h i Uthe unmarked nasal consonant tends to be substituted by the marked nasal consonant.. To sum up, if the five context-free relations mentioned in Table 2.5 are applied onto the voiceless oral stops and affricates in Mandarin, it could be inferred that with respect to aspiration, aspirated stops and aspirated affricates are more marked, and unaspirated stops and unaspirated affricates are more unmarked. There has been a debate over the substitution pattern of marked and unmarked, and contradictory. 27.

(40) results were reported in the speech error studies of Taiwan Mandarin. The consonant of unmarked place of articulation tends to be replaced by another consonant from other place of articulation at the onset position, while the unmarked nasal labial-dental tends to be substituted for the marked nasal velar consonant at coda position. The substitution pattern regarding the markedness of aspiration has not been explored yet. 2.2.5 Summary. 政治大. The study of phonological speech errors by experimental induction started from. 立. the pioneering research by Baars et al. (1975) and Motley and Baars (1976). The. ‧ 國. 學. acquired data from experiments have been considered a further evidence for the data. ‧. from naturalistic corpora, and the experimentally elicited data are able to compensate. Nat. io. sit. y. the drawbacks of naturalistic data. Therefore, speech errors have been induced,. er. studied, and used to test the hypothesis related to speech production or short-term. al. n. v i n memory. The speech error dataChave analyzed by the directionality and h eoften n g been chi U markedness effect of the substitution pattern. Previous studies in Taiwan Mandarin have revealed that perseveration exceeds anticipation regarding the directionality of substitution pattern, while the substitution pattern regarding markedness effect on place of articulation showed difference at the onset and coda position. 2.3 Research questions Previous studies on measuring the VOT of Mandarin oral stops and affricates. 28.

(41) gave their subject a list of real words and have them to read, and previous research on speech errors of consonants in Taiwan Mandarin put emphasis on the place of articulation. However, few studies adopted non-word stimuli as a means of measuring the VOT values of the stops and affricates, and few speech error studies in Taiwan Mandarin elicited speech errors through non-word tongue twister paradigm. Therefore, the aim of the thesis is to answer the following three questions by having subjects to. 政治大. do the immediate serial recall task of tongue twisters:. 立. In terms of the VOT values, what are the values of Mandarin unaspirated and. 學. ‧ 國. 1.. aspirated stops and affricates? Would the values be similar to those of the. ‧. previous studies?. Nat. io. sit. y. According to Wan’s (2002) corpus study on initial consonants, the unmarked coronal consonants would be easily replaced by those of different place of. er. 2.. al. n. v i n C h under study, would articulation. As initial consonants e n g c h i U the substitution pattern be similar to that of Wan’s (2002) study? That is, would the unaspirated consonants be replaced by aspirated consonants? This question could be further divided into two sub-categories: (a) Regarding speech errors in terms of place of articulation, would the unaspirated consonants be replaced by aspirated consonants? In other words, with respect to oral stops, will this pattern vary among bilabial, dental, and. 29.

(42) velar stops; with regard to affricates, will the pattern vary among dental, post-alveolar, and alveolo-palatal affricates? (b) Regarding speech errors in terms of aspiration, would the unaspirated consonants be replaced by aspirated consonants? Will the unmarked unaspirated consonants be preferred over the marked aspirated counterparts?. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 30. i n U. v.

(43) Chapter 3 Methodology This research aims to collect 1920 trials by setting up two experiments (48 trials *2 experiments *20 subjects). In section 3.1, the subjects’ background will be introduced, and the equipment used under the study will be described in section 3.2. The design of the materials will be illustrated in section 3.3, and the procedure of the. 治政 experiments will be displayed in section 3.4. Data analysis 大 including the transcription 立 ‧ 國. 學. of speech errors, the measurement of the VOT, and statistical analysis will be addressed in section 3.5.. ‧. 3.1 Subjects. sit. y. Nat. io. n. al. er. Twenty subjects were recruited for the present study. There were nine male and. i n U. v. eleven females, and they were all undergraduate/graduate students of National. Ch. engchi. Chengchi University. Their age ranged between 18 and 24 years old (M = 20.6, S. D. = 1.57) by the time they participated in the experiment. All of the 20 subjects were linguistically naïve. 3.2 Equipment The experiments were conducted in a quiet room. The stimuli were presented by E-prime 2.0 on a ASUS N61V series computer in both experiments. The participants were recorded during the procedure; a uni-directional cardioid dynamic vocal 31.

(44) microphone (Audio-technica ATR1200) and a digital voice recorder, Sony ICD-UX513F, were used as recording equipment. 3.3 Materials 3.3.1 Experiment 1 (Speech errors) The aim of experiment 1 was to elicit speech errors in the syllable-onset consonants (stops and affricates); therefore, the stimuli were designed as the tongue. 政治大. twister paradigm. The evidence provided by Wilshire (1999) suggested it a productive. 立. method to induce speech errors in the tongue twister paradigm that the initial. ‧ 國. 學. consonants were arranged in ABBA order and the word final consonants arranged in. ‧. ABAB order. Furthermore, as Acheson & MacDonald (2009) claimed that by. Nat. io. sit. y. arranging the initial consonants in ABBA order and the rhymes in ABAB order can. er. the stimuli be more difficultly recalled and can the researchers compare the errors. al. n. v i n C hof the present study based on the context, the stimuli e n g c h i U followed their patterns: the initial consonants were set in ABBA order, and the vowels (V) and the rhymes (VG/GV) in ABAB order. The tongue twister stimuli were composed of a sequence of four syllables. The initial consonants were the twelve oral stops ([p], [ph], [t], [th], [k], [kh]) and affricates ([ts], [tsh], [tʂ], [tʂh], [tɕ], [tɕh]). Every consonant was paired with its counterpart in. 32.

(45) each sequence. Due to the limitation of lexical gaps 3 and phonological ill-form 4, the vowels and rhymes that combined with the stops and affricates were different. Ten single vowels and eight rhymes in total were used in this study. For the stops, the single vowels were [u], [a], [i], [ʌ], and the rhymes were VG rhymes, [aw], [aj], [ow]. For the affricates, the single vowels were [i], [y], [u], [a], [ʌ], [ɨ], and the rhymes were GV rhymes, [ja], [jɛ], and VG rhymes, [aw], [aj], [ow]. Among the single vowels, [u]. 政治大. and [a] occurred mostly across the consonants, for they could be collocated with. 立. consonants more than other vowels could. Furthermore, each consonant pair was. ‧ 國. 學. combined with two sets of single V and VG/GV individually. Except that the palatal. ‧. affricates, [tɕ], [tɕh], were combined with GV due to the limitation of phonological. Nat. io. sit. y. ill-form, the remaining consonants were collocated with VG. Moreover, in each. n. al. single vowel, V2 was then. er. sequence, the vowel pair corresponded with each other; that is, if the vowel V1 was a. v i n aCsingle as well. To sum h e nvowel gchi U. up, the consonant. counterpart pair was ordered in ABBA order, and the rhyme pair was in ABAB order. In other words, if the rhyme pair was a single vowel, the sequence order would be [C1V1, C2V2, C2V1, C1V2]. Take one sequence [pu4 pha4 phu4 pa4] for example, in this sequence, the consonants are [p] and [ph], with C1 being [p] and C2 being [ph]; the. 3. Lexical gaps, also known as accidental gaps, refer to those well-formed words that do not violate phonological rules of the language but do not exist yet, e.g., in Taiwan Mandarin, [pjaŋ51] exists but has no words (Fromkin, Rodman & Hyams, 2007). 4 Phonological ill-form refers to the structures that violate the phonological rules of the language and do not exist. *[tɕhaw51] is one example in Taiwan Mandarin. 33.

(46) single vowel pair is [u] and [a], with V1 being [u] and V2 being [a]. There were forty-eight trials in total (12 consonants *2 single vowel pairs *2 VG/GV pairs =48 sequences). Of all the forty-eight trials, only tone 1 and tone 4 were used for two reasons. One reason was that they could produce the most possibilities, and the other one was to avoid tone sandhi. The details of the forty-eight trials were listed in Appendix A.. 政治大. To avoid lexical effect, which may benefit the subjects from semantic association,. 立. the following shows how the test material is done. First, all the sequences were. ‧ 國. 學. non-words. For example, any potential word like qi4 ju4 [tɕhi4 tɕy4] ‘equipment’ was. ‧. avoided. Then, the list was checked and revised by more than three people to exclude. Nat. io. sit. y. any possible semantic association. Third, the stimuli were presented visually instead. n. al. er. of auditorily. The other effort was that the task was to recall the list instead of reading the list.. Ch. engchi. i n U. v. 3.3.2 Experiment 2 (Voice onset time) The aim of the second experiment was to measure the VOT values of the syllable-onset consonants (stops and affricates). The design of the stimuli was basically the same as that of the stimuli of the first experiment, except that the stimuli of this experiment were designed as non-tongue twisters. The order of the initial consonants was AAAA, while the order of the rhymes remained unchanged, ABAB.. 34.

(47) That is, in every sequence, the four initial consonants were the same one. In other words, if the rhyme pair was a single vowel, the sequence order would be [C1V1, C1V2, C1V1, C1V2]. Take the sequence [pu4 pa4 pu4 pa4] for example. The initial consonant, C1, is [p]; V1 is [u], and V2 is [a]. There were forty-eight trials as well, and the whole list was displayed in Appendix B. Furthermore, similar to that of the first experiment, in order to avoid the presence of lexical effect, the sequences included all non-words,. 政治大. presented visually, and recalled by subjects.. 立. 3.4 Procedure. ‧ 國. 學. The experiments were conducted in a quiet room, and the subjects participated in. ‧. the experiments one by one. Before the experiments began, the subject was required. Nat. io. sit. y. to fill out a sheet on his personal backgrounds to double check whether s/he was. er. linguistically naïve or not (Appendix C). After filling out the sheet, the participant. al. n. v i n was introduced briefly how theC experiment go for the two experiments. They h e n gwould chi U were informed that it did not matter if they made any mistakes; instead, they had to make themselves louder so that their sound could be clearly recorded for later acoustic analysis by the microphone placed in front of them. The whole procedure was composed of the following sections. (1) Pre-learning section: The participant was distributed a word list (Appendix D). The listed words included (i) rarely seen in daily life; (ii) polyphones with a. 35.

(48) dominant pronunciation; (iii) rarely shown alone. The word list aimed to train the subjects to produce the intended/correct pronunciation since one might not be able to pronounce the word correctly because of the unfamiliarity with the Chinese characters. After the participant was familiarized with the word pronunciations, the task would be explained in detail. A brief flow chart of how one trial went in the experiments is shown in Figure 3.1.. 學. ＋ (1 second). ‧ 國. 立. 政治大 trial (900 ms). response page. ‧. Figure 3.1 The flow chart of a trial in the experiments. sit. y. Nat. io. n. al. er. When the trial started, after the fixation “+” showed up for 1 second, the stimulus. i n U. v. sequence would appear on the screen. The participant had to memorize the four-word. Ch. engchi. sequence, which would last for only 900 milliseconds. 900 millisecond was used for the reason that based on a pilot study, it has been found that 900 millisecond is the optimal period for the subjects not to anticipate the purpose of the study very soon. Five hundred milliseconds after the disappearance of the sequence, they would hear a beep sound indicating the start of the recall section. Following the beep sound, four. 36.

(49) metronome-like clicks at a rate of 2.5 per second would be played 5. As the clicks were being played, the participant had to recite simultaneously, trying their best to keep up with the speed of the clicks. (2) Practice section: The participant had five chances to practice. During this section, they were allowed to ask any questions. The initial consonants of the stimuli for practice were all different from those used in the experiments, and the rhymes. 政治大. were not totally the same as those used in the experiments.. 立. (3) Experiment 1: The forty-eight trials were randomly presented to the participant,. ‧ 國. 學. and the order of the trials was recorded by E-prime. The experiment started by. ‧. the participant pressing the spacebar. Furthermore, when each trial ended, the. Nat. io. sit. y. participant pressed the spacebar to begin the next trial. The participant was. er. allowed to take a rest after they finish the first experiment.. al. n. v i n Experiment 2: After takingCahfive-minute break,Uthe participant proceeded with engchi. (4). the second experiment. Before the second experiment started, the participant was informed that there were forty-eight trials as well, and the procedure was the same as it was in the first experiment. Again, the participant was informed that it did not matter if they made any mistake; it was their clear and loud reply that mattered.. 5. This method was adapted from Goldrick and Blumstein (2006) and Gormley (2014). The click sounds were used for the purpose of controlling the speech rate of the participant. 37.

(50) 3.5 Data analysis 3.5.1 Experiment 1 (Speech Errors) The audio files were transcribed into texts, and only those consonants that were to be collected were transcribed into IPA. As the consonant error regarding aspiration is the focus of the present study, after the transcription was done, the following principles of error collection were followed: (i) The order of the rhymes uttered by. 政治大. participants should correspond with that of the intended sequence. For example, the. 立. intended rhymes of the sequence was aw aj aw aj, and the utterance of the rhymes. ‧ 國. 學. should be aw aj aw aj as well. (ii) The initial consonant pair of the error tokens that. ‧. matched with that of the intended sequence; for instance, the intended consonant pair. Nat. io. sit. y. was [ts] and [tsh], and the error outcome should be [ts] and [tsh] as well.. er. Non-contextual consonants such as [tɕh] would be excluded. (iii) The number of the. al. n. v i n C hthe same as the oneUof the intended sequence. The syllables of the word sequence was engchi sequences composed of more or less than 4 would be excluded. (iv) The tone was the same as the test sequence. Many trials under the experiment might not be able to be considered. The following examples showed some cases, explaining why certain trials have to be excluded. (1) [thu4 ta4 tu4 tha4] → [thu4 ta4 tu4 ?]. incomplete results. (2) [tʂhaw1 tʂaj1 tʂaw1 tʂhaj1] →[tʂhaw1 tʂaj1 tʂhaw2 tʂhaj1]. 38. tone not matching.

(51) 3.5.2 Experiment 2 (Voice Onset Time) The audio files were transcribed into texts. Those sequences that were produced as errors were excluded, and the principles adopted were the same as they were in the transcription of the first experiment. In other words, only the VOT of the initial consonants for the correct sequences was measured. Many trials under the experiment might not be able to be considered. The following examples showed some cases,. 政治大. explaining why certain trials have to be excluded.. 立. (2) [tɕja1 tɕjɛ1 tɕja1 tɕjɛ1] →[tɕja1 tɕjɛ1 tɕja4 tɕjɛ1]. consonant not matching. 學. ‧ 國. (1) [khu1 khʌ1 khu1 khʌ1] → [khu1 kʌ1 khu1 khʌ1]. tone not matching. ‧. The VOT of the oral stops and the affricates was measured by using PRAAT.. Nat. io. sit. y. When doing the segmentation, the TextGrid of PRAAT helped do the segmentation.. er. With the use of the TextGrid, the author could not only label the IPA of the certain. al. n. v i n Conh the segments soUthat PRAAT could measure the consonant but also put boundary engchi duration between the two boundaries. In terms of the measurement of the VOT values of the stops and the affricates, the methods are described below. (i) Stops After the transcription of the audio files, to measure the VOT of the stops, the present study followed the definition and methods from previous studies. The VOT. 39.