<%BANNER%>

Perception and Production of Isolated and Coarticulated Mandarin Tones by American Learners

Permanent Link: http://ufdc.ufl.edu/UFE0041638/00001

Material Information

Title: Perception and Production of Isolated and Coarticulated Mandarin Tones by American Learners
Physical Description: 1 online resource (143 p.)
Language: english
Creator: He, Yunjuan
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: american, coarticulated, isolated, learners, mandarin, perception, production, tones
Linguistics -- Dissertations, Academic -- UF
Genre: Linguistics thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This study tested the effects of language learning experience, tonal condition, tonal sequence, tonal context and syllable position on American English speakers? ability to perceive and produce Mandarin Chinese coarticulated tones in disyllabic words. It showed that experienced learners with more language learning experience outperformed inexperienced ones on perception and production of coarticulated tones. Tonal condition, that is, whether a tone in isolation or in coarticulation, affected the perception and production of all tones. Tones in isolation were identified and produced with higher accuracy rate than tones in coarticulation. Tonal sequence, that is, whether the two syllables of the word had the same tone or two different tones, was found to affect the perception and production of Tone 1. Tone 1 in identical tonal sequence was identified and produced with higher accuracy rate than it was in nonidentical tonal sequence. Tonal context, that is, whether the pitch of the two tones at their intersection was compatible or conflicting, affected the perception and production of Tone 1. Tone 1 in compatible context was perceived and produced better than it was in conflicting tonal context. In terms of the effect of syllable position, it was found that syllable position affected Tone 2 and Tone 4 perception and production accuracy and syllable position affected Tone 3 perception accuracy but not production accuracy. The perception of these tones was better in final syllable position. The production of Tone 2 was better in final syllable position, whereas the production of Tone 4 was better in initial syllable position. For the perception of coarticulated tones, tonal direction errors decreased with increased learning experience, but tonal height errors remained. However, for the production of coarticulated tones, the ability to produce tones in terms of both tonal direction and tonal height quickly improved with increased learning experience. The relationship of perception and production of isolated tones for American learners was still not clear. Perception and production were not correlated for either group of learners. Perception of coarticulated tones predicted production for inexperienced but not experienced learners. This finding was inconsistent with the hypothesis that accurate perception precedes accurate production.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Yunjuan He.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Wayland, Ratree.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-10-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041638:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041638/00001

Material Information

Title: Perception and Production of Isolated and Coarticulated Mandarin Tones by American Learners
Physical Description: 1 online resource (143 p.)
Language: english
Creator: He, Yunjuan
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: american, coarticulated, isolated, learners, mandarin, perception, production, tones
Linguistics -- Dissertations, Academic -- UF
Genre: Linguistics thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This study tested the effects of language learning experience, tonal condition, tonal sequence, tonal context and syllable position on American English speakers? ability to perceive and produce Mandarin Chinese coarticulated tones in disyllabic words. It showed that experienced learners with more language learning experience outperformed inexperienced ones on perception and production of coarticulated tones. Tonal condition, that is, whether a tone in isolation or in coarticulation, affected the perception and production of all tones. Tones in isolation were identified and produced with higher accuracy rate than tones in coarticulation. Tonal sequence, that is, whether the two syllables of the word had the same tone or two different tones, was found to affect the perception and production of Tone 1. Tone 1 in identical tonal sequence was identified and produced with higher accuracy rate than it was in nonidentical tonal sequence. Tonal context, that is, whether the pitch of the two tones at their intersection was compatible or conflicting, affected the perception and production of Tone 1. Tone 1 in compatible context was perceived and produced better than it was in conflicting tonal context. In terms of the effect of syllable position, it was found that syllable position affected Tone 2 and Tone 4 perception and production accuracy and syllable position affected Tone 3 perception accuracy but not production accuracy. The perception of these tones was better in final syllable position. The production of Tone 2 was better in final syllable position, whereas the production of Tone 4 was better in initial syllable position. For the perception of coarticulated tones, tonal direction errors decreased with increased learning experience, but tonal height errors remained. However, for the production of coarticulated tones, the ability to produce tones in terms of both tonal direction and tonal height quickly improved with increased learning experience. The relationship of perception and production of isolated tones for American learners was still not clear. Perception and production were not correlated for either group of learners. Perception of coarticulated tones predicted production for inexperienced but not experienced learners. This finding was inconsistent with the hypothesis that accurate perception precedes accurate production.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Yunjuan He.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Wayland, Ratree.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-10-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041638:00001


This item has the following downloads:


Full Text

PAGE 1

PERCEP TION AND PRODUCTION OF ISOLATED AND COARTICULATED MANDARIN TONES BY AMERICAN LEARNERS By YUNJUAN HE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORID A IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010 1

PAGE 2

2010 Yunjuan He 2

PAGE 3

To my grandma 3

PAGE 4

ACK NOWLEDGMENTS First of all, I would like to thank the c hair of my committee, Dr. Ratree Wayland, who is very knowledgeable at doing research on tones. It was her to attract me to apply for PhD program at the Universi ty of Florida. During my PhD study, she provided lots of valuable advice. Second, I would like to t hank Dr. Caroline Wiltsh ire. From taking a course on research method in linguistics wit h her, I had the opportunity to start thinking about the topic of my dissertation and finished the first draft of the research proposal. Third, I would like to thank Dr. Andrea Pham and Dr. Zhihui Fang, for accepting to be part of my committee and provid ing me with their insightful comments. Fourth, I would like to thank Dr. Cynthia Shen, Ms. Eli nore Fresh and Ms. I-Chun Peir from the department of department of languages literatures & Cultures, who helped me to recruit American students to participate in my study. I would like to thank my classmates Ye Han, Weihua Zhu, Lili Gai for their academic or personal help when I was in Gainesville and Tampa. Last but the least, I would like to thank my grandma Rongj ie Liu, my father Yedong He, my mother Weiping Yu and my husband Feng Li for being supportive and always believe in me. 4

PAGE 5

TABL E OF CONTENTS page ACKNOWLEDGMENTS ..................................................................................................4 LIST OF TABLES ............................................................................................................9 LIST OF FIGURES ........................................................................................................11 ABSTRACT ...................................................................................................................13 CHA PTER 1 INTRODUC TION ....................................................................................................15 Research Questions ...............................................................................................16 Research Design ....................................................................................................17 Significance of the Study ........................................................................................18 2 LITERATURE REVIEW ..........................................................................................19 Suprasegmental Features in Mandarin and English ...............................................19 Mandarin ..........................................................................................................19 Tone in isolation .........................................................................................19 Tone in coarticulation .................................................................................21 English ..............................................................................................................23 Stress .........................................................................................................23 Intonation ...................................................................................................24 Effect of Learning Experience on the Acquisition of L2 Sounds ..............................25 The Effect of Learning Experienc e on the Perceptio n of L2 Sound Features ...25 The Effect of Learning Experienc e on the Production of L2 Sound Features ...27 Conclusion ........................................................................................................30 Perception and Production of Man darin Tones by L2 speakers ..............................30 Perception of Isolated Tones by L2 Listeners ...................................................30 Perception of Coarticula ted Tones by L2 Listeners ..........................................37 Production of Mandarin Isol ated Tones by L2 Speakers ..................................37 Production of Mandarin Coarticu lated Tones by L2 Speakers ..........................38 Conclusion ........................................................................................................41 The Relationship Between Percep tion and Production of L2 Features ...................42 The Acquisition Order of Percept ion and Production of L2 Features ................42 Production precedes perception ................................................................42 Perception precedes production ................................................................43 Good perception can make good production .............................................44 Good production makes good perception ..................................................48 Good production cannot promise good perception ....................................48 Speech Learning Model (SLM) ..................................................................48 5

PAGE 6

Relationship between Perception by L1 Listeners and Production by L2 Speakers .......................................................................................................50 Conclusion ........................................................................................................52 3 PERCEPTION AND PRODUCTION OF ISOLATED MANDAR IN TONES .............54 Research Questions and Hypothesis ......................................................................54 Perception Experiment ............................................................................................55 Method .............................................................................................................55 Participants ................................................................................................55 Stimuli ........................................................................................................56 Results and Analysis ........................................................................................57 Accuracy rate .............................................................................................57 Error type ...................................................................................................59 Discussion ........................................................................................................60 Production Experiment ............................................................................................60 Method .............................................................................................................60 Stimuli ........................................................................................................60 Procedure ..................................................................................................61 Judgment ...................................................................................................61 Results and Analysis ........................................................................................61 Error type ...................................................................................................63 Duration of the tone ...................................................................................64 Discussion ........................................................................................................65 Relationship between Perception and Produc tion of Isolated Tones ......................66 Results .............................................................................................................67 Discussion ........................................................................................................69 Conclusion ..............................................................................................................69 4 PERCEPTION OF COARTICULATED MANDARIN TONES ..................................71 Research Questions and Hypotheses.....................................................................71 Methods ..................................................................................................................74 Participants .......................................................................................................74 Stimuli ...............................................................................................................74 Procedure .........................................................................................................75 Results and Analysis ...............................................................................................75 Accuracy Rate ..................................................................................................75 Effects of language learning experience ....................................................76 Effects of tonal condition ............................................................................77 Effects of tonal sequence ...........................................................................79 Effects of tonal context ...............................................................................81 Effects of syllable position ..........................................................................84 Error type ..........................................................................................................86 Discussion ..............................................................................................................92 Effects of Learning Experience .........................................................................92 Effects of Tonal Condition ................................................................................94 6

PAGE 7

Effects of Tonal Sequence ...............................................................................95 Effects of Tonal Context ...................................................................................95 Effects of Syllable Position ...............................................................................96 Development of Perception of Coarticulated Tones .........................................97 Conclusion ..............................................................................................................99 5 PRODUCTION OF COARTICULATED MANDARIN TONES ...............................100 Research Questions and Hypothesis ....................................................................100 Method ..................................................................................................................102 Participants .....................................................................................................102 Stimuli .............................................................................................................102 Procedure .......................................................................................................103 Judgment ........................................................................................................103 Results and Analysis .............................................................................................104 Accuracy Rate ................................................................................................104 Effects of language learning experience ..................................................104 Effects of tonal condition ..........................................................................106 Effects of tonal sequence .........................................................................108 Effects of tonal context .............................................................................109 Effects of syllable position ........................................................................111 Tone 3 in tone sandhi ..............................................................................113 Error Type ......................................................................................................114 Relationship between Perception and Produc tion of Coarticulated Tones .....120 Discussion ............................................................................................................122 Effects of Learning Experience .......................................................................122 Effects of Tonal Condition ..............................................................................123 Effects of Tonal Seque nce and Tonal Context ...............................................123 Effects of Syllable Position .............................................................................124 Development of Production of Coarticulated Tones .......................................125 Relationship between Perception and Produc tion of Coarticulated Tones .....125 Conclusion ............................................................................................................126 6 GENERAL DISCUSSION AND CONCLUS ION ....................................................127 Summary of Results ..............................................................................................127 Experiments 1 and 2 ......................................................................................128 Experiment 3 ..................................................................................................128 Experiment 4 ..................................................................................................129 General Discussion ...............................................................................................130 Limitations and Future Research ..........................................................................133 Pedagogical Implications ......................................................................................135 APPENDIX LIST OF REFERENCES .............................................................................................137 7

PAGE 8

BIOGRAPHICAL SKETCH ..........................................................................................143 8

PAGE 9

LIST OF T ABLES Table page 2-1 Lexical tones in Mandarin ...................................................................................20 2-2 Tone errors .........................................................................................................39 2-3 Most common errors of tones in disyllabic words by American students (Wang, 1995) ......................................................................................................41 3-1 Percentages of misperception made by inexperienced American learners of Mandarin ............................................................................................................59 3-2 Percentages of misperception made by experienced American lear ners of Mandarin ............................................................................................................59 3-3 Percentages of production confus ion made by inexperienced American learners of Mandarin ...........................................................................................63 3-4 Percentages of production confus ion made by inexperienced American learners of Mandarin ...........................................................................................63 3-5 The mean duration of tones produced by learners who confused Tone 2 as Tone 3, in ms. .....................................................................................................64 3-6 The mean duration of tones produced by learners who confused Tone 3 as Tone 2, in ms. .....................................................................................................64 3-7 The mean duration of tones produced by learners who confused Tone 4 as Tone 1, in ms. .....................................................................................................65 3-8 Individual Inexperienc ed American learners perception and production accuracy rate. .....................................................................................................68 3-9 Individual experienc ed American learners perception and production accuracy rate.. ....................................................................................................68 4-1 Mandarin tonal combinations in t onal env ironments (positive factors are unshaded; negative factors are shaded.) ...........................................................74 4-2 Confusion matrix for tones in init ial position identified by inexperienced learners ..............................................................................................................87 4-3 Confusion matrix for tones in fi nal position identifi ed by inexperienc ed learners ..............................................................................................................87 4-4 Confusion matrix for tones in final position id entified by experienced learners ...87 9

PAGE 10

4-5 Confusion matrix for tones in final position identified by experienced learners, ..88 4-6 Description of Mandarin Chinese tonal categories .............................................89 4-7 Tonal confusions categorized into error types ....................................................90 5-1 Confusion matrix for tones in init ial position produced by inexperienced learners ............................................................................................................114 5-2 Confusion matrix for tones in initial position produced by experienc ed learners ............................................................................................................114 5-3 Confusion matrix for tones in final position pr oduced by inexperienced learners ............................................................................................................115 5-4 Confusion matrix for tones in fina l position pr oduced by experienced learners 115 5-5 Tonal confusions categorized into error types. .................................................118 5-6 Individual inexper ienc ed American learners perception and production accuracy rate.. ..................................................................................................121 5-7 Individual experienc ed American learners perception and production accuracy rate ....................................................................................................121 10

PAGE 11

LIST OF FIGURES Figure page 2-1 Mean F contours of four Mandarin tones in the monosyll able /ma/ produced in isolation.0.........................................................................................................20 2-2 Waveforms and F0 tracks of citation pronunciations of permit (Noun) and per mit (Verb), cited from Gussenhoven, 2004. ...................................................23 3-1 Mean percent correct identificatio n of the four M andarin tones for inexperienc ed and experienced learners of Mandarin on isolated tones.. ..........57 3-2 Inexperienced and experienced learners mean percent correct identification for each tone. ......................................................................................................58 3-3 Mean percent correct production of the four M andarin tones for inexperienc ed and experienced learners of Mandarin on isolated tones.. ..........62 3-4 Mean percent correct production of the four M andarin tones for inexperienc ed and experienced lear ners of Mandarin on each tones.. ...............62 3-5 Mean percentage of accuracy of per ception and produc tion of Mandarin isolated tones by two groups. .............................................................................67 4-1 Mean percent correct ident ification of coarticulated Mandarin tones across all stimuli for inexperienced and experienced American lear ners of Mandarin.. ......76 4-2 Mean percent correct identificatio n of coarticulated tones across both syllables by inexperienced and experienced American learner s of Mandarin. ...77 4-3 Mean percent correct perception fo r Inexperienced and Experienc ed groups for tones in isolation vs. tones in coarticulation. ..................................................78 4-4 Mean percent correct perception fo r Inexperienced and Experienc ed groups for tones in identical tone sequen ce vs nonidentical tone sequence.. ................80 4-5 Mean percent correct perception fo r Inexperienced and Experienc ed groups for tones in compatible vs. conflicting tonal context. ...........................................82 4-6 Mean percent correct ident ification of coarticulated final syllable Tone 1 in different tonal contexts by tw o groups of learners of Mandarin. ..........................84 4-7 Mean percent correct perception fo r Inexperienced and experienc ed groups for tones in initial vs. final syllable positi on. ....................................................... 85 4-8 Frequency of tonal directional misidentification by two groups of American learners ..............................................................................................................91 11

PAGE 12

4-9 Frequency of tonal height misidenti fication by two groups of American learners ..............................................................................................................91 5-1 Mean percent correct production of coarticulated Mandarin tones by inexperienc ed and experienc ed learners of Mandarin ......................................105 5-2 Mean percent correct pr oduction of coarticulated t ones by inexperienced and experienced American learners of Mandarin ....................................................105 5-3 Mean percent correct pr oduction of coarticulated M andarin tones in different tonal condit ions by inexperienced learners and experienced learners .............106 5-4 Mean percent correct pr oduction of coarticulated M andarin tones in different tonal sequences by inexperienced learners and experienced lear ners. ...........108 5-5 Mean percent correct pr oduction of coarticulated M andarin tones in different tonal contexts by inexperienc ed learners and experienced learner s ................110 5-6 Mean percent correct pr oduction of coarticulated M andarin tones in different syllable positions by inexperienc ed learners and experienced learners.. .........112 5-7 Frequency of tonal direction confusion by two groups of American learners ....119 5-8 Frequency of tonal height confusion by two groups of American learners ........119 5-9 Frequency of tonal direction plus height confusion by two groups of American learners ............................................................................................................120 5-10 Mean percentage of accuracy of perception and prod uc tion of coarticulated Mandarin tones by two groups ..........................................................................120 12

PAGE 13

Abstract of Dissertation Pr esented to the Graduate School of the University of Fl orida in Partial Fulf illment of the Requirements for t he Degree of Doctor of Philosophy PERCEPTION AND PRODUCTION OF ISOLATED AND COARTICULATED MANDARIN TONES BY AMERICAN LEARNERS By Yunjuan He May 2010 Chair: Ratree Wayland Major: Linguistics This study tested the effects of language learning experience, tonal condition, tonal sequence, tonal context and syllabl e position on American English speakers ability to perceive and produce Mandarin Chi nese coarticulated tones in disyllabic words. It showed that experienced lear ners with more language learning experience outperformed inexperienced ones on perception and production of coarticulated tones. Tonal condition, that is, whether a tone in is olation or in coarticulation, affected the perception and production of all tones. Tones in isolation were identified and produced with higher accuracy rate than tones in coarti culation. Tonal sequence, that is, whether the two syllables of the word had the same tone or two different tones, was found to affect the perception and production of Tone 1. Tone 1 in identical tonal sequence was identified and produced with higher accuracy rate than it was in nonidentical tonal sequence. Tonal context, that is, whether the pi tch of the two tones at their intersection was compatible or conflicting, affected the perception and production of Tone 1. Tone 1 in compatible context was perceived and produced better than it was in conflicting tonal context. In terms of the effect of syllable position, it wa s found that syllable position 13

PAGE 14

affected Tone 2 and Tone 4 perception and production accuracy and syllable position affected Tone 3 perception accuracy but not production accuracy. The perception of these tones was better in final syllable posit ion. The production of Tone 2 was better in final syllable position, whereas the production of Tone 4 was better in initial syllable position. For the perception of coarticulated tones, tonal direction errors decreased with increased learning experience, but tonal hei ght errors remained. However, for the production of coarticulated t ones, the ability to produce tones in terms of both tonal direction and tonal height quickly improved with increas ed learning experience. The relationship of perception and pr oduction of isolated tones for American learners was still not clear. Perception and pr oduction were not correlated for either group of learners. Perception of coarticu lated tones predicted production for inexperienced but not experienced learners. This finding was inconsistent with the hypothesis that accurate percepti on precedes accurate production. 14

PAGE 15

15 CHAPTER 1 INTRODUCTION In the United States, Mandar in Chinese is very commonly taught as a foreign language at the college level. Because Mandar in is a tonal language, perceiving and producing tones accurately is very important for American speakers in their acquisition of Mandarin Chinese. When learning Chinese as a foreign languag e, students get input of the target language mainly from their t eachers in the classroom. In the current Chinese language classroom, teachers focus on introducing the tonal direct ion of Mandarin tones in isolation and provide practice on perceiv ing and producing tones in isolation on monosyllabic words (Xing, 2006). However, a ccording to one statistical analysis (Wang et al, 1986) among the 31,159 Mandarin words surveyed, 22,941 (74%) were disyllabic words, and only 12% were monosyllabic word s. The remaining 14% of words contain more than two syllables. Therefore, per ceiving and producing tones on monosyllabic words alone is not sufficient for L2 Mandarin learners. Practicing tone use on disyllabic words is necessary for those who actually wa nt to communicate successfully with native speakers. The tonal direction and height of the four Mandarin lexical tones is quite stable when they occur in isolation, whereas these acoustic features may be changed in coarticulated tonal environments (in the current study, the coarticulated tonal environment is the disyllabic word) (Xu 1994, 1997). Therefore, the gap between the tonal practice in classroom and the use of to nes in the real world poses a great problem for learners. American learner s may not be able to perceive and produce coarticulated

PAGE 16

tones on disyllabic words accurately even after they have acquired Mandarin tones in isolation. The effect of language learning expe rience has been examined on adults perception and produc tion of L2 sounds. The ma in research question asked in these studies is whether the performance of experi enced speakers of the target language is better than inexperienced speakers on the acquisition of L2 sounds. Most previous research on segmental feat ures has suggested that the ability to perceive and produce L2 sounds may impr ove with increased learning experience (Mackain et al.1981, Flege et al. 1995; Flege et al.,1997; He, 2004; Levy and Strange, 2008). In terms of suprasegmental features experienced learner s also outperform inexperienced learners on the acquisition of stress in English and tones in Thai and Mandarin (He et. al., 2008; Wayland and Guion, 2003; Wang et al., 1999, 2003). In Wang et al., (1999, 2003) s studies perception and production of isolated Mandarin tones by American speakers, ex perienced learners who received perception training in a language lab identified and produced tones mo re accurately than the inexperienced learners who did not receive th e training. However, no studies have been undertaken to systematically describe how Am erican speakers with different amounts of classroom learning experience perceive and produce Mandarin tones in coarticulation. Research Questions This study is aimed at finding ho w different amount of language learning experience affects L1 American English sp eakers perception and production of L2 Mandarin tones. This study is guided by several research questions: Research Question 1. Do experienced American E nglish speakers perceive and produce Mandarin tones more accurately than inexperienced learners? 16

PAGE 17

Research Question 2. What are the error types of perception and production of tones typical to American speakers wit h different amount of language learning experiences? Research Question 3. What are the linguistic factors that prevent American speakers from percei ving and producing Mandarin tones accurately? Research Question 4. What is the relationship between perception and production of Mandarin tones by American speakers? Research Design To answer the four research questions, four experiments were designed: two perception and two production. First, one pe rception and one production experiment of isolated Mandarin tones were completed by experienced American learners and inexperienced American learners in order to explore whether language learning experience affects the accuracy of identif ying and pronouncing isol ated Mandarin tones. Furthermore, the production study was designed to investigate the acoustic cues that American English speakers of L2 Mandarin tend to use to modify Mandarin isolated tones. In these first two experiments, 22 American students who had studied Chinese for three months participated as inexperi enced learners, and 16 American students who had studied Chinese for fifteen months participated as experienced learners. More importantly, this study extended t he testing stimuli to longer utterances, disyllabic words, which more frequently appear in the word level of real conversation. One perception and one production experiment were conducted to assess the effects of language learning experience, syllable positi on, and tonal environment on American speakers accuracy of perc eption and production of coarti culated Mandarin tones. 17

PAGE 18

Production and perception data from 14 i nexperienced learners and 9 experienced learners who were able to accurately perceiv e tones in isolation were included in the analyses. Significance of the S tudy This study is significant for several reasons First, it is the first study to examine the effect of language learning experi ence on the perception and production of coarticulated Mandarin tones by American learners; earlier studies investigated only learners mastery of isolated, si ngle-syllable tones. Second, it is the first study to test the effect of tonal environment on the acquisi tion of Mandarin coarticulated tones on disyllabic words by L2 learners. Th ird, it is the first study to investigate the nature of the relationship between perception and production of Mandarin coarticulated tones by American learners. Furthering our knowledge of the acquisition of Mandarin tone by American learners will lead to a better underst anding of their difficulties in perceiving and producing Mandarin coarticulated tones and help us to improve Chinese teaching pedagogy. 18

PAGE 19

CHA PTER 2 LITERATURE REVIEW In this chapter, I first introduce the feat ures of tone in Mandarin, which are basic knowledge for understanding the present st udy and the methodology presented in Chapter 3, Chapter 4 and Chapt er 5. I then describe the features of stress and intonation in English, two features which may be transferred to L2 Mandarin tones by American learners. Then, I present a review and critique of currently available literature relevant to my main subjec t: perception and production of t ones by L2 speakers. Finally, I review some of the existing literatur e on the relationship between perception and production of L2 features. Suprasegmental Features in Mandarin and English Mandarin Tone in isolation Mandarin Chinese is a tonal language. It has four lexical tones, the main cue to which is fundamental frequency. They are high level tone (Tone 1), a tone [that] starts with a high F0 value and stays around that level throughout the syllable; rising tone (Tone 2), a tone [that] starts with a low F0, then falls slightly bef ore rising throughout the remainder of the syllable; low rising tone (Tone 3), a tone [that] starts with an F0 value slightly lower than the onset of Tone 2, falls to the lowest F0 then rises sharply to the end of the syllable; [and] falling tone (Tone 4), a tone [that] starts with the highest F0 value of the four tones, continues to rise before reaching the ma ximum about one fifth of the way into the vowel then falls sharply to the end of the syllable (Xu, 1997 ; p.67). Traditionally, Mandarin tones are described by the Scale of Five Pitch Levels (Chao, 1948). This measurement is widel y used in phonetic and phonological studies on Mandarin tones. Table 2-1 summarizes the de scription of the four tones. (5 indicates the highest possible pitch.) 19

PAGE 20

Figure 2-1. Mean F0 contours of four Mandarin tones in the monosyll able /ma/ produced in isolation. The time is normalized, with all tones plotted with their average duration proportional to the average durat ion of Tone 3. (X u, 1997; p. 67) Table 2-1. Lexical tones in Mandarin Tone Descriptive name Pitch value Diacritic Pitch targets by Xu (1993) Ma mother Tone 1 High level 55 high-high Ma hemp Tone 2 Rising 35 low-high Ma horse Tone 3 Low rising 214 low-low Ma scold Tone 4 Falling 52 high-low The fundamental frequency is the primary acoustic cue for Mandarin speakers to identify Mandarin tones ((Liu, 1924; Howie, 1976; Wu, 1986). However, the amplitude of tones also signals the category of Mandarin tones, except, Tone 1, the high flat level tone (Whalen and Xu, 1992). Tone 2, Tone 3 and T one 4 can be identified from just the amplitude contour. Duration is also a cue to tone identity. Tone 3 has the longest duration, Tone 4 has the shortest durati on, and Tone 1 and Tone 2 are in between. In addition, duration affects the perception of Tone 2 and Tone 3 (Blicher et al,1990). Specifically, it was found t hat Tone 2 and Tone 3 were identified more often as Tone 3 when the tones were lengthened. Besides the overall duration of the tone, Moore and Jongman (1997) found that the turning point (the point where falling pitch changes into 20

PAGE 21

rising pitch) and F0 (the change in pitch from onset of the tone to the turning point) are also acoustic cues to differentiate Tone 2 and Tone 3. Tone in coarticulation When produced in isolation, the pitch contour or the F0 value of the tone is quite stable. However, in connected speech, the basic shapes of Mandarin tones often undergo context-induced modifications (Shih, 1986; Shen, 1990; Xu, 1993, 1994). In disyllabic words, it was found by Shen (1990) and Xu, (1994 a) that the (tone value) F0 value/contour of the preceding syllabl e has a carry-over influence on the F0 value/contour of the following tone Furthermore, the F0 va lue of the following syllable also has an anticipatory influence on the F0 contour of the preceding tone Shen (1990) claimed that the bidirectional effects are sy mmetrical: the influences of adjacent tones on each other are similar. Also, she claimed that tonal coarticulation changes the range but not direction of F0. In other words, the pitch value of tones will be changed within their own tone category. However, by conducting a more carefully controlled experiment, Xu (1997) found that the bidirectional effects are asymmetrical. In terms of magni tude, the carry-over effect is larger than the anticipatory effect. The carry-over effect causes the onset of the following tone to assimilate to the offset of the preceding tone: a high offset of the preceding tone raises the onset of the following tone, and a lo w offset lowers the onset of the following tone. On the other hand, anticipatory effect causes the offset of the preceding tone to dissimilate to the onset of the following tone: a high onset of the following tone lowers the offset of the prec eding tone, and a low onset raises the offset of the preceding tone. 21

PAGE 22

Besides the influence of different tonal contexts, Xu (1993) identified two additional types of contextual effects that may also affect pitch shape of a tone. One is a compatible context: the value of the pitch of the offset of the preceding tone and the value of the pitch of the onset of the following tone are similar. For example, in the phrase dn bi zh, (protein) the value of the offs et o f the preceding tone is similar to the value of the onset of the following tone. The other is a conflicting context: the value of the offset of the preceding tone and the value of the on set of the following tone are substantially different. For example, in the phrase c ng yu b n (a kind of pie) the value of the offset of the preceding tone is different from the value of the onset of the following tone. Xu (1993) investigated the percepti on and production of the second syllable associated either with a rising tone or a fa lling tone in trisyllabic words to see how different tonal contexts affect the tonal shape of the tar get tones. In terms of production, it was found that the slope of rising tone keeps rising in the compatible context, while the slope falls in the conflicting context. Therefore, context can actually cause the slope of a tone to change to the opposite of what is heard in isolation. The slope of a falling tone is steeper in the compatible context t han in the conflicting context. Meanwhile, the perception accuracy of the cat egory of the tone on the second syllable is higher in the compatible context than in the conflicting context. Finally, Cheng (1986) introduces the third -tone sandhi rule: If two consecutive syllables both have Tone 3, the first Tone 3 is changed to Tone 2. 22

PAGE 23

English Stress A stressed syllable in American English is characterized by four phonetic properties: higher pitc h, l onger duration, higher int ensity and full vowel quality (Gussenhoven, 2004). Shown in Figure 2-2 (Gussenhoven, 2004), the pitch characteristics of stress in disyllabic Englis h words produced in citation form are 1) If the first syllable is stressed, the pitch level st arts high and falls on the second syllable. The fall is somewhat truncated. 2) If the second syllable is stressed, the pitch level starts low, rises on the approach to the second syl lable, and falls on the second syllable. Figure 2-2. Waveforms and F0 tracks of citation pronunciations of permit (Noun) and permit (Verb), cited from Gussenhoven, 2004. The rules of assigning stre ss (Cruttenden, 2000; p16) are: (i) Verbs and adjectives (a) Stress on the penultimate syllable when the final syllable has a short vowel in an open syllable or is followed by no more than one consonant (b) Otherwise, stress is on the final syllable (subject to rule (iii) below) (ii) Nouns 23

PAGE 24

(a) if the final syllable has a short vowel, disregard it and apply rules under (i) above (b) if the final syllable has a long vowel, it is stressed (subject to (iii) below) (iii) In words of more than tw o syllables with a long final vowel (a) Stress may optionally occur on the antepenultimate rather than the final syllable. Intonation Though English is not a tonal language, it does use intonation to express certain expressions. Unlike Mandarin tones, English intonation can be expressed over several words. In one sentence, a nuclear tone involves the major part of the meaning contributed by the pitch pattern of an intonation-group (Cru ttenden, 2000; p.50). According to the description of Cruttenden (2000), English has seven types of nuclear tones, which Cruttendon divides into two gro ups: falling tones and rising tones. Falling tones are more common in sentence final pos ition than in sentence non-final position. They are 1) High falling: to show more inte rest, more excitement, or more involvement. 2) Low falling: to show less interest, less ex citement and less emotion. 3) rising-falling: to mark the completeness of a declaration. Within certain environ ments, falling nuclear tones can express the feeling of being impr essed, the intention to gossip, or the intention to challenge. Rising tones often occur in sentence non-fina l position. They are: low rising, high rising, fall-rising and mid level. There ar e two subgroups of rising tones. One group contains dependent rising tones, whose me aning depends on the relationship with another tone. The other group contains i ndependent rising tones, which have their own meaning regardless of surrounding tones. Dependent rising tones are 1) Low rising: the most oratorical, typical of a formal reading st yle. 2) High rising: more casual. 3) Fall24

PAGE 25

rising: to emphasize the contrastive nature of the subject. 4) Mid level: no other meaning except of non-finality. Independent rising tones are 1) Low rising: to express the feeling of uncertainty. 2) High rising: to soothe, reassure or patronize. 3) Fall-rising: to express self-justifi cation or to appeal. The common combinations of the nuclear t ones in a sentence are 1) Rising tone + falling tone. 2) Falling tone + low rising tone, which is used in tag questions. The listener will get the sense of disagreem ent. 3) Falling tone + falling tone, which is also used in tag questions. The listener will get the sens e of being required to have agreement. Effect of Learning Experience on the Acquisition of L2 Sounds As indicated in the literature review ed above, little research has been conducted on the effects of learning experience on the perception and/or production of suprasegmental features, such as stress, to ne and intonation. Therefore, I broaden the scope of my literature review to include the effects of learning experience on the acquisition of L2 sound features, includi ng both segmental and suprasegmental features. In the following par agraphs, I review studies of the effect of learning experience on the acquisition L2 sound features first, beginning with perception and then continuing with production. The Effect of Learning Experience on the Percepti on of L2 Sound Features Flege (1997) studied the effect s of experience on the perc eption of English vowels (/i/, / /, /e/ / /) by non-native speakers with vari ous language backgrounds (German, Spanish, Mandarin, and Korean) It was found that the exper ienced speakers, with an average of 7.3 years length of residence (LOR) in the U.S., made significantly more use of spectral cues which changes the qualit y of vowels than did the inexperienced 25

PAGE 26

speakers, who had an average of 0.7 years LOR, when identifying /e/-/ /. Experienc ed Mandarin and Korean speakers, like native English speakers, made significantly more use of spectral cues and fewer temporal cues (such as duration) than did the inexperienced Mandarin and Korean speakers when identifying /i/-/ /. Levy and Strange (2008) conducted a study on the perception of French vowels by L1 American English adults with and without French language experience. The results showed that the experienced gr oup of American English adults discriminated /u-/, /i-y/ and /y-/ better than the inexper ienced group did. However, there was no significant difference in the two groups ability to discriminate /u/ and /y/. Consonantal context affected the inexperienced groups ability to discriminate these vowels, but not the experienced group. The inexperienced group performed better in th e bilabial context than in the alveolar context, whereas the experienced groups performance in these two consonant contexts was comparable. In Fleges (1984) study, experienced L1 Arabic L2 English speakers who had lived in the U.S. for an aver age of 5.8 years and inexper ienced L1 Arabic L2 English speakers who had lived in the U.S. for an aver age of two months were asked to identify /s/ or /z/ as "piece" or "peas": The dur ation of the vowels and fricatives was manipulated, and it was found that the experienced L2 English speakers performance more closely matched native English speakers in all three conditions: only vowel duration increased, only fric ative duration decreased and vowel duration increased + fricative duration decreased. The inex perienced L2 English speakers behavior resembled that of the native English speakers in only two conditions. In the condition of 26

PAGE 27

only fricative duration decreased, their perform ance is dramatically different from the English native speakers. Only one study on discriminating supras egmental features has been conducted: Wayland and Guion in 2003 tested nave and exper ienced learners of Thais ability to discriminate middle and low tones in Thai. This pair of Thai tones is quite confusing to American speakers. However, experienced American learners in the study performed better than American speakers without any learning experienc e on discriminating middle tone and low tone. However, not all studies have found an advantage for experienced learners. In Swedish, a long vowel must combine with a short consonant in a stressed syllable, and a short vowel must be with a long consonant. McAllister et al. (2002) asked experienced and inexperienced groups of L1 English and L1 Spanish speaking adults to identify whether perceived words followed this phonological rule. Experienced and inexperienced groups of listeners did not differ significantly in thei r perception accuracy on this task. The Effect of Learning Experience on the Production of L2 Sound Features In addition to investigating the effects of language experience on perception of English vowels (/i/, / /, /e/ / /) by non-native speakers with various language backgrounds (German, Spani sh, Mandarin, and Korean), Flege (1997) also studied the effects of language experience on the production of these vowels. Based on an intelligibility test, the production of / / by experienced non-native subjects (average 7.3 years residency in the U.S.) was more nativ e like than that of t he inexperienced nonnative subjects (average 0.7 years residency in the U.S.). Acoustic analysis revealed 27

PAGE 28

that experienced L1 German speakers and experienc ed L1 Mandarin speakers produced a larger spectral difference between /i/-/ / and /e/-/ / than inexperienced speakers of the two language groups, and experienced L1 German and L1 Spanish speakers produced significantly diffe rent durations for /e/ and / /, which inexperienced speakers did not make. Flege (1995) found that experienced L1 Japanese speakers of L2 English (average 21 years of residenc e in the U.S.) perform ed better than inexperienced speakers (average 2 years of residence in t he U.S.) when producing English /r/ and /l/. The production of the Englis h liquids by experienced Ja panese speakers was near the native norm based on the judgment of nat ive speakers of English, whereas inexperienced Japanese speakers frequently produced /r/ as /l/ and sometimes produced /l/ as /r/. He (2004) tested two groups of L1 Mandarin speakers production of L2 English /l/ in syllable-final position. One group included inexperie nced L2 English speakers who were college students in China, and the ot her included experienced speakers of English who had studied in Canada for at least one year. It was found that experienced speakers produced more native-like syllable-final /l/s than inexperi enced speakers of English. Trofimovich and Baker (2006) investigated the production of five suprasegmental features (stress timing, peak alignmen t, speech rate, pause frequency, and pause duration) of six English decla rative sentences by L1 Korean learners of L2 English. 30 adults with varying levels of English experie nce (3 months, 3 years, and 10 years of United States residence, respectively) parti cipated in the study. The Korean learners 28

PAGE 29

with extens ive English l anguage experience performed better on producing stress timing than learners with medi um level of Eng lish language experience, and learners with medium level of Englis h language experience performed better than learners with short language experience. However, the effect of language experience was not detected in the production of t he other four suprasegmental features. Instead, the adult learners' age at the time of first extensive exposure to t he L2 (indexed as age of arrival in the United States) influenced the produc tion of speech rate, pause frequency, and pause duration. In contrast to the studies discussed above, McCallister et al. (2002) found no effect of experience on production of lengt h contrasts. Experi enced (average 10 years of residence in Sweden) and inexperienced (a verage 3.6 years of residence in Sweden) groups of L1 English and L1 Spanish speaking adults completed a production experiment of Swedish consonant and vowel length distinctions, and the results showed that the ratios of durations of vowels and consonants were not significantly different between experienced and inexperienced groups. He et al. (2008) investigated how native speakers of Mandarin, a tonal language, employ phonetic cues to differentiate st ressed and unstressed syllables in producing English disyllabic words. Two groups of L1 Mandarin speakers with di fferent L2 English learning experiences participated in the st udy: 8 inexperienced learners of English who were college students in China, and 8 experie nced learners of English who had studied at a university in the U.S. for at least two years. It was found that, compared to inexperienced learners, experienced learner s produced more native-like acoustic characteristics (larger ratios of amplit ude and duration in stressed vs. unstressed 29

PAGE 30

syllab les) to stress English words. However, the ratios were still smaller than English native norms. Both groups performed similarly to the native norm on the ratio of pitch between stressed and unstressed syllables. Conclusion Generally speaking, previous studies s how that experienced learners perform better than inexperienced learners on perceivi ng and producing L2 features, not only in terms of overall accuracy rate, but also in terms of which spectral cues are used in perceiving and/ or producing these L2 f eatures. Experienced learners generally perceive more accurately than inexperienced learners. They also produce the target sounds in a more native-like fashion than inex perienced learners. Ho wever, with certain features, such as vowel and consonant lengt h distinction in Swedish produced by English speakers and English lexical st ress produced by Mandarin speakers, more learning experience does not seem to l ead to improved performance. Since no study has been done to investigate the effect of learning experience on perception and production of Mandarin tones by L2 learners, this study was aimed to fill the research gap. Perception and Production of Ma ndarin Tones by L2 speakers Perception of Isolated Tones by L2 Listeners To understand how L2 listeners identify and discr iminate tones in isolation, several factors that might affect the perception of tones by L2 listeners have been studied. Linguistic experience. It has been found that native listeners can perceive tones better than non native listeners (Gottfried et al, 1997; Lee et al, 1996; Wang, 2004, Wayland & Guion, 2003; Wayland & Guion, 2004). Cantonese listeners performed 30

PAGE 31

better at differentiating Cantonese tones than Mandarin listeners and Englis h listeners ( Lee et al, 1996); Mandarin listeners performed bette r at identifying Mandarin tones than Cantonese listeners and English listeners (Gottfried et al., 1997, Lee et al., 1996); and Thai listeners did better at differentiating the middle and low tone cont rasts in Thai than Chinese and Englis h listener s (Wayland & Guion, 2003; Wayland & Guion, 2004). The advantage of the native lis teners in perceiving the tones in their language accurately is quite straightforward. Since they are acquiri ng the tones from the time they are born, they are very familiar with the acoustic cues of each tone in the tone inventory of the language. This advantage allows native listeners to identify t ones in a shorter time than non native speakers ( Lee et al, 1996), and it also helps nativ e listeners identify tones more accurately than non native speakers when the tones are pr esented with an entire syllable, center-only syllable, initial-only sy llable and silent-center syllable (Gottfried et al, 1997). Not only do non native listeners have problems with identification and differentiation of tones compared to native listeners, but they al so lack categorical perception of tones. In Halle et al.s (2003) study, Mandarin listeners perceived tones categorically, whereas French listeners perceived tones psychophysically and their results showed more variation. Furthermo re, Mandarin listeners performed better at identifying and discriminating Mandarin t ones than French listeners in terms of accuracy. Although the Frenc h listeners didnt show a ca tegorical perception of Mandarin tones, they did show sensitivit y to changes of pitch when the acoustic difference was salient enough. It was also found that native speakers of tonal languages have a right-ear (left-hemis phere) advantage for percepti on (Wang et al, 2001; Wang et 31

PAGE 32

al, 2004). This results has been tested on s peakers of Thai, Mandarin and Norwegian, bilingual speakers whose one of the languages is the te sted language also showed left hemispher e dominance. However, the nonnativ e speakers without L1 tonal background, such as Norwegian and English speakers, do not show the right-ear advantage (Wang et al, 2004). Therefore, it was conc luded that nonnative speakers do not have hemispheric lateralization to native like left-hemisphere dominance. (Wang, 2004, p. 465). It is not clear whether a listener with a tonal background will always outperform a listener without any tonal language experien ce when perceiving foreign tones. Some studies have supported this idea: Cantonese listeners discriminat ed Mandarin tones better than English listeners (Lee et al, 1996), and Wayland & Guion (2004) found that Chinese listeners were better at discriminating and identifyi ng low and mid tones in Thai than American listeners without a background in tone languages. Wayland and Guion attributed the result to the Chinese listener s perceptual transfer from their L1 tonal background to perceive L2 tones. The ability to track pitch direction and contour, which Chinese listeners use to perceiv e Chinese tones, is transfe rred to the discrimination and identification of the pair of Thai tones. It s eems that tonal listeners use similar strategies to identify and discriminate tones which non tonal listeners do not use. Gandour (1983) claimed that listeners of tonal languages (in this case, Cantonese, Mandarin, Taiwanese, and Thai) paid more attention to t he direction of tones, whereas listeners of a nontonal language (English) attached more importance to t he height of the tone. Here, the ability to track the contour of the tone, which English lis teners lack, seems to play a 32

PAGE 33

more important role in identifyin g and discr iminating tones than the ability to hear the tones pitch level. However, other studies have provided cont radictory evidence: Mandarin listeners did not obtain higher tonal discrimination scores than English listeners on the perception of Cantonese tones (Lee et al, 1996). The authors suggest that Cantonese tones are more difficult than Mandarin tones for childr en to acquire. Differences in first language development between the two languages indica te that Cantonese tones are harder to perceive than Mandarin tones. Therefore, Cantonese listeners can discriminate Mandarin tones more easily than Mandarin listeners can discriminate Cantonese tones. Francis et al. (2008) suggested that language background (tonal vs. nontonal) may not affect the error rate of identifying tones, but only the types of errors made when identifying tones. In their study, Mandarin speakers (tonal language speakers) and English speakers (non-tonal speakers) were trained to identify Cantonese lexical tones. Both groups performed better on recognizing Cantonese lexical tones after their training. However, there was no significant difference in the degree of improvement between the two groups. In term s of error type, Mandarin s peakers more often confuse tones with similar tonal directions, whereas English speakers more often confuse tones with similar average F0. For example, Mandarin speakers tended to misidentify low rising (23) tone as high rising (25) tone, whereas English speaker s tended to mistake it for mid level (33) tone. Language learning experience. Experienced L2 learners whose L1 is not a tonal language may have fewer perceptual difficulties in identifying tones than inexperienced learners, since they have been exposed to t he target tones for a while and have 33

PAGE 34

become familiar with the phonetic properties that tones carry. Furthermore, those experienced learners could poss ibly reach nat ive-like perceptual ability. One study has supported this claim: Wayland and Gui on (2003) found that experienced American learners of Thai outperformed inexperienced American lear ners when discriminating the middle and low tones in Thai and performed almo st as well as native speakers of Thai. Perceptual training. Intensive perceptual training with certain tones will improve certain L2 listeners ability to discriminate and identify tones (Francis et al. 2008; Wang et al, 1999; Wang & Kuhl, 2003, Wayland & Guion, 2004). American listeners who had perceptual training identified Mandarin tones more accurately afte r training, whereas American listeners who did not participate in training m ade smaller progress on tone identification (Wang et al, 1999). Surprising ly, though, younger Amer ican listeners (from 6 to 14 years old) did not make greater progress on tone identification than older American listeners (19 years old) (Wang & Kuhl, 2003). Different age groups ability to perceive L2 tones can be improved to the sa me degree, as long as the group members receive the same amount of target info rmation. In another study (Wayland & Guion, 2004), Mandarin listeners also benefitted from perceptual training in discriminating mid and low tone in Thai, whereas American listeners did not show significant improvement from the pretest to the posttest. The authors attributed t he differential effect of perceptual training between the two groups of listeners to t he degree of intensiveness of training. Compared to Chin ese listeners with tonal language experience, American listeners without prior exposure to tones need more practice to discern acoustic properties of the pair of tones In Francis et al. (2007), both L1 Mandarin speakers and L1 American English speakers made improv ement on recognizing Cantonese lexical 34

PAGE 35

tones after receiving perceptual training. In fact, there wa s no significant difference in the two language groups improvement from pretest to posttest. Tonal context. Bent (2006) tested monosyllabic and trisyllabic meaningless words (all /ra/ syllables) attached with Mandarin tones There was an effect of tonal context on the accuracy of discrimination of tone on the second syllable within a tri-syllabic utterance. The fallin g and the rising tones were more easily differentiated when they are preceded by a level tone and followed by a fa lling tone (i.e., level + Tone + falling pattern), whereas the two tones were confus ed in the tonal pattern of falling + Tone + rising. The author explained th is phenomenon using the Perceptual Assimilation Model (PAM). The tonal environments of level + fa lling + falling correspond to the L+H* L+H* intonation pattern in English, whereas the le vel + rising + falling pattern corresponds to the L* + H English intonation pattern. Thes e two patterns show a two-category (TC) assimilation pattern, in which discrimination is expected to be good. On the other hand, the falling + falling+ rising and falling + rising + rising patterns are both mapped onto the same English intonation pattern, L+H* L*, and show a single-category (SC) assimilation pattern, in which discrimination is expected be poor. Syllable structure. Native English speakers have been found to experience more difficulty in discriminating Thai middle and low tone contrasts when these tones are presented in open syllables than in closed syllables (Wayland & Guion, 2003). Differences in acoustic properties of the mi ddle and the low tones are more salient in open syllables than closed syllables. Talker variability. In Leather (1990), L1 Dutch speakers were tested on four Mandarin tones produced within the monosyllable /y/. Part icipants misidentified tones 35

PAGE 36

more often when the stimuli were pronounced by multiple speakers than when all stimuli were produced by the same speaker. Leather concludes that too much inter-speaker variability at too early a st age may prevent the learner fr om discovering with sufficient accuracy the prototypical forms that the exemplars presented expound (Leather, 1990; p98) Speech sounds. Bent et al. (2006) investigated how tonal listeners and non tonal listeners perceive pitch in speech (Mandarin tones) and nonspeech sounds (both pure tones and saw-tooth wave pulse trains). Ma ndarin listeners identif ied Mandarin tones with a higher rate of accuracy than Amer ican listeners. Also, Mandarin listeners exhibited a small range of performance ( 96% to 100% correct) because of a ceiling effect, whereas the English listeners sco res varied widely and did not overlap with any of the Mandarin listeners scores (26% to 88% correct). However, there was no significant difference between Mandarin an d American listeners performance on the nonspeech pitch discrimination. Music experience. Musicians have been found to have an advantage for identifying and discriminating Mandarin tone s than non-musicians (Alexander et al., 2005). Musicians identified and discriminated Mandarin tones with a higher rate of accuracy (89% and 87%, respectively) than non -musicians (69% and 71%). This result contradicts the claim that processing mu sic is unassociated with processing lexical tones (Kimura, 1961). Musical experience c an help listeners to perceive lexical tones more accurately (Alexander et al., 2005). 36

PAGE 37

Perception of Coarticulated Tones b y L2 Listeners To the best of my knowledge, only one study has been conducted to examine how coarticulated tones are perceived by L2 listene rs. Broselow et al. (1987) found that L1 English speakers found the falling tone in M andarin harder to perceive accurately when it was on the first syllable of the disyl labic words than when it was on the second syllable. They attributed the phenomenon to L1 intonation transfer, since falling pitch rarely occurs in word-initial position in English. In terms of error type, they found that the falling tone was frequently identified as a high flat tone by American listeners when it occurred word-finally. They argued that Amer ican listeners took the falling part of the falling tone as a word-ending marker and did not associate it as a part of the tone. Therefore, they only perceiv ed the high pitch onset at the beginning of the tone and identified the tone as a high flat tone. Production of Mandarin Isolated Tones by L2 Speakers Chen (1974) found that native English s peakers did not produce the full pitch range of Chinese tones. Their pitch range was narrower than the native pattern. Wang et al. (2003) tested production of isolated M andarin tones by American speakers. Their participants had taken one or two semesters of Mandarin Chi nese. Between these tests (pretest and post-test), half of the participants received perceptual training. Their preand post-training productions were judged by multiple native speakers of Mandarin Chinese and analyzed by WAVES + ESPS softwar e. The results indicated a specific order of ease of production: Tone 1, T one 4, Tone 2 and then Tone 3, (ranked from easiest to hardest) which is the same order as first language acquisition of Mandarin tones by Mandarin children (Li and Thompson, 1977) In terms of types of errors, it has 37

PAGE 38

been found that Tone 1 was misproduced as Tone 2 or Tone 4, Tone 2 was mainly misproduced as Tone 3, Tone 3 was often misproduced as Tone 2, and Tone 4 was misproduced as Tone 1 or Tone 3. Besides the tonal contour errors, the tonal height (which is synonymous with the term tonal regi ster in other studies on Mandarin tones) errors were also addressed. It was found that the pitch height of Tone 1 was slightly lower than native production, the onset of Tone 2 was hi gher and the valley of Tone 2 was lower than native norms, the valley of Tone 3 was not as low as that of the native one, and both onset and offset of Tone 4 we re lower than ones of the native norms. Compared to the tonal contour errors, the tonal height erro rs are more difficult to overcome. In other words, pitch height and pitch contour are not acquired in parallel. For American students, the shape of pitch contour is relatively easier to produce in a more native-like manner than the value of pitch height. In conclusion, the production of Mandar in tones on monosyllabic words in the post-test was more native-like in terms of the dimensions of tonal height and tonal contour. Therefore, the audito ry training of American adults to identity Mandarin tones not only improves their percept ion of Mandarin tones (Wang 1999), but is also effective in improving their production of the tones without any explicit training or feedback in speech production (Wang 2003). Production of Mandarin Coarticulated Tones by L2 Speakers At least four papers have studied the production of Mandarin coarticulated tone by American Speakers. Shen (1989) examined t he tonal production in a piece of Mandarin text by American students. The participant s had studied Mandarin for one semester. A hierarchy of degrees of difficulty of producin g tones is proposed: Tone 4 is the hardest, 38

PAGE 39

then Tone 1 is the next hardest, and Tone3 a nd Tone2 are the easiest. It was assumed that Tone 4 and Tone 1 are less marked than the others in perception becaus e they both start with high pitch and ma y be perceived as indicative of a stressed syllable. Therefore, Tone 4 and Tone 1 are more subjected to negatively transfer by American students. However, the author did not explain why stressed syllables are less marked than unstressed syllables and harder to produce correctly. Miracle (1989) also examined the tonal productions of American students who had studied Mandarin for at least one academic y ear. The target words were monosyllabic or disyllabic words in different positions in a sentence. In contra st to Shens (1989) finding, this study showed that the error ra te of Tone 2 is the highest, which Miracle attributes to interference from the Englis h intonation system. However, he does not thoroughly explain how the English intonat ion system was employed by American speakers to produce Mandarin tones. Two types of errors were categorized: tonal register (tone range) errors and tonal contour (tone direction) errors. Unfortunately, the author only described what errors were made, but did not explain why Mandarin tones were inaccurately modified. Table 2-2 summarizes Miracles description of tone errors. Table 2-2. Tone errors Tone Tonal register errors Tonal contour errors Tone 1 Too low Falling tone Tone 2 Beginning of the tone too high Falling tone or level tone Tone 3 Too high Rising tone Tone 4 Beginning of the tone is mid N.A Chen (1993) analyzed the tonal errors of American adult learners of Chinese by examining tones occurring in sentences.. Six students in a second-year university conversational Chinese class participated in his production experiment. Two sequential tonal patterns used by American speakers in a four-syllable sentence were detected: one is 55---33---22---53; the other is high---l ow---high---low--(high), and all syllables 39

PAGE 40

carried lev el tones. The reasons for making t he two patterns were attributed to negative transfer from English prosodic features to Mandarin tones. The transfer of English intonation modifies tones into the first pat tern and the transfer of English rhythm modifies tones into the second one. However, he did not explain which kind of English intonation or rhythm was negatively transferr ed into American speakers production of Mandarin sequential tones. He notic ed that contour tones are ei ther distorted flatly or produced in an opposite direction; the level t one either changes the range of the tone or became a contour tone. A hierarchy of difficulty for second-l anguage Mandarin tonal acquisition suggested in his study agrees with the first-language acquisition order reported in Li and Thompsons study (1993): T one 1 is the easiest one to be acquired, Tone 4 and Tone 2 are next, and Tone 3 is the mo st difficult one. This study provides a picture of Chinese tonal acquisition among nat ive English speakers; however, the model for labeling the tone value is not convin cing. Though the author used the numbers to describe tones, he did not provide acoustic measurements to support his description, but instead relies on his own perception. Theref ore, his results were very subjective. Wang (1995) analyzed tonal errors of disy llabic words made by American students who had been studying Mandarin in China for longer than half a year. As in other studies mentioned, Tone 1 and Tone 4 were produced correctly more often than Tone 2 and Tone3. Both tonal register errors and tonal contour errors were found in the production of Tone 2, Tone 3 and Tone 4. Only tonal contour errors were found in the production of Tone 1. The most common errors for each tone are shown in Table 2-3. 40

PAGE 41

Table 2-3. Most common errors of tones in disyllabic words by American students (Wang, 1995) Tone Tone 1 Tone 2 Tone 3 Tone 4 Main contour errors falling rising Low flat (1st syllable) High flat or falling (2nd syllable) rising high flat rising The most interesting part of her study is that she found that Tone 2 on the initial syllable tends to be pronounced as a low tone when the pitch of the onset of the following tone is high. She attributed the erro r to the transfer of English stress patterns (in a disyllabic word, one syllable is stressed and another one is unstre ssed). In English, the stress pattern is associat ed with the height of pitch. T he high pitch represents the stressed syllable and the low pitch represent s the unstressed syllable. Wang assumed that American speakers treated the syllable starting with high pitch in this case the second syllable -as a stressed syllable. Co nsequently, the first syllable is treated as unstressed and thus pronounced wit h a low tone. However, th is combination of high tone and low tone was not found in the production of other tone combinations. Conclusion American students have difficulty producing tones on disyllabic words. However, no previous studies have empirically investigated which tones in what kind of tonal context are harder to produce accurately. Though previous studies listed the main error types for the coarticulated tones, they hav e not revealed what environments triggered tone value changes. From previous studies two linguistic factors which prevent American students from producing Mandarin coarticulated tones accurately have been found: 1) English intonation combinations an d 2) English stress combinations. However, these two reasons have not been fully explained in previous articles. All the tested words in Shen (1989) and Chen (1993) were al ready familiar to their participants, and 41

PAGE 42

the participants in Wang (1995) were familiar with most of the test ed words. Therefore, the effect of experienc e might play a role in tone production. In other words, because of the effect of practice, a word with a difficu lt-to-produce tone pattern might have a higher possibility of being pronounced accurately than a word with an easy-to-produce tone pattern. Therefore, it is bette r to use words which are new to the subjects to test their ability to produce tones. The Relationship Between Percepti on and Production of L2 Features The relationship of perception and production of SLA features will be discussed with more details in the following sections. Since very few studies have been conducted specifically on the relationship between perception and production of acquisition of tones by L2 speakers, in this section, I will broaden my review to includ e previous research on the acquisition of segmental features by L2 speakers. Two su bsections will be incl uded: the order of perception and production of L2 sounds and re lationship between their perception and production. The Acquisition Order of Percepti on and Production of L2 Features Production precedes perception Studies by Goto (1971) and Sheldon et al. (1982) showed that some Japanese subjects were able to produce identifiable /r/ and /l/ tokens even though they were unable to reliably identify nat ive English /r/ and /l/ toke ns. This finding led these researchers to conclude that production can precede perception in the acquisition of a non-native contrast. Flege & Eefting (1987) al so found that Dutch speakers could not identify stops in a /da /-/ ta / synthetic continuum as well as they could produce this contrast. Smith (2001) reanalyzed the results of previous studies on the perception and 42

PAGE 43

production of English /l/ and /r/ by J apanese and Korean speakers. Because good production and mediocre perception performance on the contrast /l/ and /r/ can co-occur in the learners, she concluded that production development can be independent from perception development in second language acquisition. Perception precedes production Borden et al. (1983) examined the relati ons hip between perception and production of English /l/ and /r/ by Korean learners of En glish. They found that their participants perception developed earlier t han their production and accurate perception might be a prerequisite for accurate production. The authors note that the ability to make phonemic perceptual judgments in a /r/ /l/ continuum that are similar to those of English speakers also seems to improve before production (p. 516). Sheldon (1985) statistically reanalyzed the Borden et al.s resu lts. This reanalysis confirmed the findings of Sheldon and Strange (1982) and failed to s upport the argument m ade by Borden et al. (1983) that accurate perception is acqui red before accurate production. One of the important conclusions of Sheldons reanalysis was that the relationship between production and perception depended on the amount of time spent in the Englishspeaking environment by the Kor ean learners, so that as the learners time in the U.S. increases, the probability of environment of perception exceeding production decreases (p.111). This claim is supported by the fact that the speakers studied by Sheldon and Strange (1982) we re advanced learners. Therefore, we can say t hat there is no conclusive connection between speech perception and production on the contrast of /l/ and /r/ by Korean and Japanese speakers. To account for the relationshi p between perception and production of English /l/ and /r/ by Japanese and Korean speakers, the duration of language experience 43

PAGE 44

should be c onsidered. If the speaker is a begi nning learner, they will perceive more accurately than they produce, but advanced learners may produce more accurately than they perceive. For the vowels, Bohn and Flege (1990) found that inexperienced L1 German speakers can perceive L2 English / / and / / accurately. However, they cannot produce the contrast in a native like manner. This re sult confirmed the conclusion of Borden et al. (1983) that inexperienced speakers will perceive L1 segments before they successfully produce them. There are very few studies on the or der of development of perception and production of suprasegmental features. Only Wang (1999) and Wang at el. (2003) tested the relationship of perception and pr oduction of Mandarin isolated tones by American students and found that perception of Mandarin t ones by American speakers preceded production of Mandarin tones, since they can perceive Mandarin tones with a high rate of accuracy but their production of tonal contours do not match the native norm. Good perception can make good production If the statement that good perce ption leads to good production is true, then in order to improve their ability to produce L2 sounds, L2 learners must accurately perceive the L2 sounds. Several studies show that per ception training will help learners to differentiate and/or identify L2 segments. Fu rthermore, the improvement of perception leads to the improvement of production. Ro chet (1995) found that L1 Mandarin Chinese speakers perception of L2 French voice onset time (VOT) ca tegories was more accurate after receiving auditory training. Moreover, the participants production on their 44

PAGE 45

imitation task was also more nati ve like. Log an et al. (1991) and Bradlow et al. (1997) showed that the performance of Japanese speak ers on perceiving English /r/-/l/ was improved by giving them perceptual training. Bradlow et al. (1997, 1999) showed that the transfer of perceptual learning to speech production can also occur. In terms of supr asegmental features, several studies have been conducted to test whether perceptual training will affect the performance of perceiving tones by L2 speakers. It has been shown that for Amer ican English speakers, intensive perceptual training with certain tones will improve their ability to discriminate and identify the tones (Wang et al, 1999; Wang & Kuhl, 2003, Wayland & Guion, 2004). Later, Wang et al. (2003) reported that t he production of the same participants in her 1999 study was also improved by giving them auditory training; identification of trainees post-test tone productions improv ed by 18% relative to their pretest productions. They conducted two production ex periments of isolated Mandarin tones by American speakers. Compared to the tonal contour errors, t he tonal height errors were more difficult to overcome. In other wo rds, pitch height and pi tch contour are not acquired in parallel. For Am erican students, the shape of pitch contour is relatively easier to produce in a native-like manner than pitch height. In conclusion, the production of Mandarin tones on monosyllabic words afte r training is more native like regard with the dimensions of tonal height and tonal contour. Therefore, auditorily training American English speaking adults to i dentity Mandarin tones not only improves their perception on Mandarin tones (Wang 1999), but is also effect ive in improving their tone production without any explicit training or feedba ck in speech product ion (Wang 2003). Leather (1997) found that Dutch speakers with good t one perception also produced Mandarin 45

PAGE 46

tones correctly. Specifically, the tones wh ich were perceived correctly were also produced c orrectly, and the tones which were perceived wrongly were also produced inaccurately. No studies have been conducted to test whether good perception of stress and intonation will result in good production of them when learner s do not receive production training. Good perception cannot guarantee good production Contrary to the evidence reviewed in the previous section, two studies have shown that good perception can not guarantee accurate production. In other words, perceptual accuracy gained from training is not necessarily transferable to production. According to Wang (2004), perception traini ng helps Chinese discriminat e and identify English vowel contrasts /i/-/ / and / /-/ / better. However, the effects of the training were not found on their accuracy of production of these vowels Perez (2005) also showed no significant effect on production when L1 Spanish speakers received training on discrimination and identification of Englis h vowel contrasts /i/-// and /a/-/ /, though there was a great improvement in perception of Eng lish vowels after the training. Furthermore, many studies have shown that accurate perception is a necessary but not sufficient condition for accurate production. It has been found that L2 English voiced stops in coda position were frequently followed by an inserted schwa, completely deleted, or simply devoiced by L1 Mandarin (Flege et al, 1992; Hansen, 2001; Wang, 1995), though there have been no st udies of how Mandarin speakers perceive syllable final stops. According to my own experienc e as a native Mandarin speaker, we are able to perceive voiced stops in a native like manner. Mandarin speakers presumably modify 46

PAGE 47

voiced stops by using deletio n or feature changing strat egies because of L1 language transfer and universal constraints, ye t they do not exper ience corresponding perceptional problems. Since Mandarin does not allow any obstruents to occur in coda position, Mandarin speakers tend to use the stra tegies of insertion and deletion to form a CV structure that the language allows. The choice between using insertion and deletion is based on the disyl labic preference which exis ts in Mandarin phonology (Hansen, 2001; Wang, 1995), There is another possible reason why L2 speakers cannot produce L2 segments accurately. According to my observati on, Mandarin speakers have no difficulty perceiving the trill /r/ in Russian. However, some speakers cannot produce the trilled /r/ accurately. Apparently, the misproduction is not motivated by misperception of the sound. Here, I attribute this kind of misp roduction to an articulatory disability. During the acquisition of tones, it was f ound that L2 speakers cannot produce tones accurately even if they can perceive tone without any difficulty. For American English speakers, the stress pattern and intonation pattern in English have an effect on their production of Mandarin tones (Review of Chen, 1993; Wang, 1995 Section 2.2). Wayland et al. (2006) tested the production and perception of English stress by Thai speakers. Thai speakers tested on Eng lish non-words stressed the first syllable more often than the second syllable when the non-word was categorized as a noun, whereas they stressed the second syllable more often when the non-word was a verb. Wayland et al. also found that Thai speaker s tended to place stress on a syllable with a long vowel more often than those with a shor t vowel or coda consonant(s). However, 47

PAGE 48

neither lexical class nor syllabic structure influenced their pr eferred patterns of stress assignment during their perception experiment. Pickering (2001) examined the differences in English intonation between American teaching assistants and Chinese teaching assistants. It was found that Chinese teaching assistants used more level intonat ion and less rising intonation than American teaching assistants. The author attributed the difference in intonation patterns to the Chinese teaching assistants L1 language tr ansfer rather than perceptual problems. Good production makes good perception To the best of my knowledge, only one st udy has tested the effect of production training on perception of Mandarin tones. Leather (1997) gave Dutch speakers production training on Mandarin tones and then tested their perception. It was found that good production does help Dutch speaker s perceive Mandarin tones correctly. Tones which were produced correctly were perceived correctly, and the tones which were produced wrongly were also perceived wrongly. Good production cannot promise good perception In the previous section, I reviewed studies on production and perception of English /l/-/r/. It was found that advanced Japanese s peakers may produce the English /l/ and /r/ accurately despite being unable to perceive t hese sounds accurately. In other words, good production cannot prom ise good perception Concerning suprasegmental features, none of the previous studies have been conducted under the argument that good pr oduction can promise good perception. Speech Learning Model (SLM) The relationship of perception and produc tion of L2 segments is addressed in the Speech Learning Model (SLM) (Flege, 1995). In the model, it was proposed that good 48

PAGE 49

production is preceded by good per ception of L2 segments. This prediction is tested by giving perceptual training to L2 learners and co mparing their production before and after the training. Most studies showed signi fic ant improvement on the accuracy of production. Auditory perceptual training had a positive effect on the production of French voice onset time categories by Mandarin Chinese speakers (Rochet (1995). Auditory perceptual training improved the production performance of English /l/ and /r/ by Japanese speakers (Bradlow et al., 1997, 1999; Hazan et al. 2005). Audiovisual perceptual Training improved Japanese speakers production of labial segments in English. (Hazan et al. 2005). However tw o studies showed that good perception does not predict accurate production. Wang (2004) revealed that perception training helps Chinese to discriminate and identify better on English vowels contrasts /i/-/I/ and / /-/ /. However, the effects of the training were not found on their production. Perez (2005) also showed no significant effect on production by perception training on discrimination and identification of English vowel contracts /i/-/ / and /a/-/ / by Spanish speakers, though there is a great improvement on perceiving English vo wels after the training. Furthermore, adequately produced L2 sounds may not have been well perceived. Smith (2001) reanalyzed the results of previous st udies on the perception and production of English /l/ and /r/ by Japanese and Korean speakers. Based on the fact that good production and fair perception performance on the contrast /l/ and /r/ can co-occur on the learners, she drew a conclusion t hat production development can be independent from perception development in the second language acquisition. 49

PAGE 50

Relationship betw een Perception by L1 Li steners and Production by L2 Speakers Foreign accents are identified by errors in segments (vowels and consonants) and suprasegmental features (s tress, tone and int onation). Misproduction of sounds may cause native speakers to misunderstand. As we know, misproduced vowels and consonants may change the lexical meaning of words, for example, the vowel / / as in the word pit is substituted with the vowel /i:/ as in the word Pete, which is a typical Chinese accent of English (Chen, 1996); similarly, the consonants /r/ and /l/ are confused in production by Japanese speakers as in the words rice and lice (Goto, 1971; Sheldon et al., 1982). Inaccurate production of suprasegmental features such as tones can also cause lexical meaning changes. Fo r example, American speakers frequently mispronounce ma horse as ma hemp (Wang, 2003). Differences in stress placement by L2 learne rs or by speakers of different dialects can lead to misunderstandings as well. For example, Hungarian and Polish L2 learners tend to place stress on the initial syllable of an English word (Archibald, 1993). Indian English speakers do not raise the pitch or lengthen the duration of stressed vowels the way American English speakers do, and produc e smaller phonetic differences between stressed and unstressed syllables than Am erican English speakers do (Wiltshire & Moon, 2003). All above production errors are serious because they can cause misunderstanding among native listeners.. Considering the need for adequate communication, language teachers should focus on correcting learners pronunication, which will improve their ability to communicate. 50

PAGE 51

Certain no n native pronunciations will not change the meaning of the target word. The following statements were observed by the author. In terms of vowels, American speakers often produce Mandarin / / as English / /. However, this kind of misproduction will not cause any problems wit h lexical meaning change, si nce Mandarin does not have / / as a phoneme in its vowel inventory. The vowel [ ] is an allopne of / /. In terms of consonants, Mandarin speakers often produce English voiced stop /d/ as Mandarin voiceless unaspirated stop /t/ and produce E nglish voiceless aspirated stop /t/ as Mandarin voiceless aspirated stop /t /. So far, no studies have found these production errors to cause a delay of understanding by native speakers of English. To examine misproduction of English stress, He (2006)1 conducted a study on the production of English lexical stress by Mandarin speaker s. In the study, Mandarin speakers used smaller duration ratio and amplitude differenc e to stress English disyllabic words. However, the pitch contours they used to indica te English stress were similar to those of native speakers. Since pitch level is one of the primary acoustic correlates of stress ((Wiltshire and Moon, 2003). it was assum ed that using smaller duration ratio and amplitude difference to stress English words by Mandarin speakers would not affect English native speakers judgment of the location of stress Misproduction of intonation does not change any lexical meanings. However, using different intonation from the native pattern ma y cause pragmatic meaning changes, which not only marks the learner as a non-native speaker but also can cause native listeners to find the learner rude. Pickering (2001) pointed out that Chinese teaching assistants frequently use level t one and falling tone where rising tone is used 1 The state I put here is from my individual study with Dr. Wilshire in fall 2006. 51

PAGE 52

by American teaching assist ants. For example, falling tone is frequently used by Chinese teaching ass istants in response to incorrect answers from American students, where a rising tone was expected. The misuse of the tone confuses native listeners, since falling tone is used to affirm t he response in English. To avoid the misunderstanding caused by the pragmat ic change with foreign accent, language teachers need take the issue seriously and dr aw their students attention to problems with intonation. Regarding tonal errors, L1 American Engl ish speakers may produce native like L2 Mandarin tonal directions, but the contour of the tone is slightly different from native norms. The pitch height of Tone 1 is often slightly lower than native production, the onset of Tone 2 is higher and the offset of Tone 2 is lower than native norms, the valley of Tone 3 is not as low as that of the native one, and both onset and offset of Tone 4 are lower than ones of the nat ive norm (Wang, 2003). It rema ins an empirical question whether the tonal contour mistake affects Mandarin speakers perception. Conclusion We cannot draw a general conclusi on regarding the relationship between perception and production of L2 sounds. Many lingu istic factors affect the relationship. In terms of the effect of language tr aining, good perception may lead to good production. However, L1 transfer and articula tory difficulty may prevent L2 speakers from producing L2 sounds accurately. G ood production may lead to good perception; however, L1 transfer, such as categorical perception, may be barriers that keep L2 listeners from perceiving L2 sounds accurately To determine whether misproduction of 52

PAGE 53

L2 features causes a delay of understanding by L1 listeners, we need to consider the nature of the error and the langua ge background of the L1 listeners 53

PAGE 54

CHA PTER 3 PERCEPTION AND PRODUCTION OF ISOLATED MANDARIN TONES This chapter reports on a study on the perception and production of isolated Mandarin tones by American le arners of Mandarin with diffe rent learning experience. The first section presents the perception expe riment, including the methodology, results, analysis, and discussion. The second section presents the production experiment in the same manner. The third section discusses the relationship between perception and production of isolated Mandarin t ones by American speakers. Research Questions and Hypothesis The present experiments te sted perception and production of Mandarin isolated tones in monosyllabic words by experienced and inexperienced American learners. The experiments were aimed at ans wering the following questions: Research question 1. Does American learners production of isolated Mandarin tones improve with Mandarin learning experience in the classroom? Wang et al.s (1999, 2003) studies show ed that American speakers performed better on perceiving and producing isolated Mandarin tones after they received perception training to identif y these isolated Mandarin tones. In the present study, American learners with more learning experi ence in the classroom were expected to perform with greater accuracy than inexperie nced American learners who had relatively less learning experience in identifyi ng and pronouncing isolated Mandarin tones. Research question 2. What are the typical tonal perce ption and production errors for American learners with different leve ls of language learning experience? Previous studies revealed that American speakers who receive perceptual training tend to misperceive Tone 1 and Tone 4 as each other and Tone 2 and Tone 3 as each 54

PAGE 55

other (Wang, 1999). They confuse tones in their produc tion in the same way (Wang 2003). American learners who learned tones in the classroom in the present study were expected to perform as American speakers in the lab. Research question 3. What is the relationship between perception and production of Mandarin isolated tones at different stages of the learning experience Flege (1995) suggested that accurate perception is the predictor of accurate production. In the Speech Learning Model (SL M), the speech production accuracy of second language (L2) learners is limited by their perceptual accuracy. Thus, in the present study, the ability to perceive isol ated Mandarin tones was expected to develop faster than the ability to produce these tones. Perception Experiment Method Participants All participants were American students at the University of Florida. They all studied Mandarin Chinese from Monday to Friday, one hour each day in the classroom. The perception experiment was conducted at the end of November 2007. At the time, the twenty-two participants in the beginn ing level had studied Mandarin for three months; they were label ed inexperienced learners for this experiment. The sixteen participants in the intermediate level had studi ed Mandarin for twelve months and were labeled experienced learners for this st udy. They had had no previous experience learning a tone language prior to learning Mandarin at the Un iversity of Florida. They had no speech or hearing problems. Partici pants earned extra credit for completing the study. 55

PAGE 56

Stimuli A set of 12 monosyllabic words containing the four lexic al Mandarin tones (three words per tone) was produced by a native female speaker of Mandarin. A single speaker was used rather than multiple speakers in order to avoid the possibility that participants might misperceive the tones because they were unable to generalize pitch changes to different talkers. A female speak er was selected because of Wang et al.s (1999) finding that American st udents found female talkers more intelligible than male ones. All stimulus words were the CV sylla ble /na/. The syllable structure and the segments chosen in the study all exist in Eng lish. This design avoids the possibility that American listeners could misperceive the t ones because they were distracted by the novel segments or syllable structure rat her than focusing on the suprasegmental features. The contour of F0, the primary acoustic cue to i dentify Mandarin tones, and the secondary cues of amplitude envelope and syllable duration (Fu et al., 1998; Liu & Samuel, 2004; Whalen & Xu, 1992) were kept as the speaker naturally produced them. They were not manipulated. Tone 3 did not have creaky voice, and Tone 4 did not have glottalization (Liu & Samuel, 2004), which ar e reported as acoustic cues to identify those two tones additionally. Procedure All stimuli in the perception experiment were presented in random order. Each type of tone was presented three times. For each trial, the participant viewed a syllable na on the screen of the computer. The partici pant clicked on the syllable to hear the corresponding sound and then wrote down the t one value of the syllable of each word on a provided answer sheet. Participants could adj ust the volume to their comfort level. 56

PAGE 57

The participants could hear the sound as many times as they needed. Before the real test, there were three warming up trials. The results from these trials were not analyzed No feedback about their performance was provided during the test. Results and Analysis In this section, both tone identification accuracy rate and tone error type from the perception will be reported. Accuracy rate Figure 3-1. Mean percent co rrect identification of t he four Mandarin tones for inexperienced and experienced learners of Mandarin on isolated tones. Error bars indicate standard error. The overall results in Figure 3-1 show that experienced l earners of Mandarin performed better on identifying isolated Man darin tones than inexperienced learners (94.28% vs. 82.99% correct). A one-way ANOVA showed this difference to be significant at the .01 level [F (1, 150) = 9.213, p=0.003]. 57

PAGE 58

Figure 3-2. Inexperienced and ex perienced learners mean percent correct identification for each tone. Error bars indicate standard error. The accuracy rate of perception for each isolated tone by groups of American learners is illustrated in Fi gure 3-2. A two-way repeated ANOVA, with Tone as within subject factor and Group as between subject fa ctor, showed a main effect for Group [F (1, 36) = 5.86, p<.03], indi cating that experienced lear ners performed better than inexperienced learners on identifyi ng tones overall, and a main effect of Tone [F (3, 34) = 4.415, p<.02] suggesting that their perfo rmance varied across tones. A Turkey-HSD test (Bonferroni adjusted p<. 05) showed that their identific ation of Tone 2 (79%) was significantly worse than their identification of Tone 1 (94.75%), but not significantly worse than Tone 4 (90.34%) or Tone 3 ( 86.89%). The interaction between Group and Tone was nonsignificant [F (3, 34) = .155, p= .926] showing that experienced learners consistently outperformed inexperienced lear ners across all four tones. These results are consistent with those of Wang et al. (1999). 58

PAGE 59

Error ty pe Comparing Table 3-1 to Table 3-2, it is apparent that there is no difference of error type between the two language proficien cy groups. Tone 1 and Tone 4 showed bidirectional confusion patterns, since t here was no significant difference between frequencies of misproducetion of Tone 1 as Tone 4 and misproduction of Tone 4 as Tone 1 by inexperienced group ( 2=.895, p=.344). However, Tone 1 and Tone 4 were identified with very high accuracy rates by experienced group. Tone 2 and Tone 3 showed bidirectional confusion patterns, since there was no significant difference between frequencies of misproducetion of Tone 2 as Tone 3 and misproduction of Tone 3 as Tone 2 by either inexperienced group ( 2=2.533, p=.213) or experienced group ( 2=1.103, p=.294). Table 3-1. Percentages of misperception m ade by inexperienced American learners of Mandarin, with the number of frequency in parentheses Tone Identified as Tone 1 Tone 2 Tone 3 Tone 4 Tone 1 90.9 (60) 3.0 (2) 0 6.1 (4) Tone 2 0 72.7 (48) 27.3 (18) 0 (0) Tone 3 0 (0) 18.2 (12) 81.8 (54) 0 (0) Tone 4 10.6 (7) 1.5 (1) 1.5 (1) 86.4 (57) Table 3-2 Percentages of misperception made by experienced American learners of Mandarin, with the number of frequency in parentheses Tone Identified as Tone 1 Tone 2 Tone 3 Tone 4 Tone 1 100 (48) 0 (0) 0 (0) 0 (0) Tone 2 0 (0) 87.5 (42) 12.5 (6) 0 (0) Tone 3 0 (0) 6.3 (3) 93.8 (45) 0 (0) Tone 4 0 (0) 4.2 (2) 0 (0) 95.8 (46) 59

PAGE 60

Discussio n The results of the accuracy rate of per ception on isolated Mandarin tones suggest that language learning experienc e affects American learners perception of Mandarin tones. The longer they study, the more accu rately they perceive tones in general. In terms of error types, it was found that the most confusing Tones for American learners are Tone 4, which is often conf used with Tone 1, and Tone 2, which is frequently confused with Tone 3. These resu lts suggest that American speakers paid more attention to the pitch value at the onset of tones than at the offset of the tones. This would explain why Tone 4 is confused with Tone 1, since both have a high pitch onset, and the low pitch offset of Tone 4 might be interpreted as simply the end of the utterance. Tone 2 and Tone 3 both have a mid-range onset pitch. Furthermore, both tones show falling + rising contours in thei r phonetic descriptions, though the falling part of Tone 2 is not perceived by native speakers of Mandarin. Therefore, some American speakers did not form a categor ical perception on Tone 2. Production Experiment Method Participants The same participants who completed the perception experiment completed this production experiment as well. Stimuli The tested words in the study were non words. The participants were instructed to produce the syllable /na/ with the indicated tone. A sonorant consonant was selected because a non-sonorant consonant would perturb the F0 of the following vowel. The vowel / a/ was selected to minimize vowel intrinsic pitch effect on pitch production. 60

PAGE 61

Procedure All stimuli in the production experiment were presented in random order. Each tone was presented three times. For each ti me, the participants viewed the word /na/ (with tone indicated) in Pinyin on the scr een of the computer. Parti cipants pronounced each word twice. The first pronunciation was not analyzed. The second try was considered indicative of what the participant really intended and would be taken into analysis. Before the real tes t, there were three warming up trials. The production from these warm-up trials were not analyzed. No feedback about their performance was provided during the test. A total of 456 stimu li (38 subjects x 3 tokens x 4 tones) were included in subsequent analyses. Judgment Two native speakers of Mandarin with a Beijing acc ent transcribed the tones produced by participants based on their percept ion. Answer sheets were provided. The produced stimuli were printed with no tonal di acritics. Judges needed to provide a tonal diacritic corresponding to the tone they heard (Wang 2003). Results and Analysis In this section, the results will be present ed for three measures accuracy rate, error type, and duration of the vowel. Accuracy rate From Figure 3-3, we can see that inexperienced learners and experienced learners showed comparable performance in producing isolated Mandarin tones (84.88% vs. 85.94%). A oneway ANOVA showed no evidence of a difference in accuracy rate for the two groups [F (1, 150) = 0.053, p=0.817]. 61

PAGE 62

Figure 3-3. Mean percent correct production of the four Mandarin tones for inexperienced and experienced learners of Mandarin on isolated tones. Error bars indicate standard error. Figure 3-4. Mean percent correct production of the four Mandarin tones for inexperienced and experienced learners of Mandarin on each tones. Error bars indicate standard error. The accuracy rate of production for each isolated tone by the two groups of American learners is illustra ted in Figure 3-4. A two-wa y repeated ANOVA with Tone as within subject factor and Gr oup as between subject factor showed no main effect for Group [F (1, 36) = .067, p=.797], giving us no evidence of a difference between 62

PAGE 63

experienced learners and inexperienced learners in their tone production accuracy. Howev er, there was a main effect for Tone [F (3, 34) = 9.043, p< .001]. A Tukey-HSD test (Bonferroni adjusted p<.05) showed t hat only Tone 3 (68.42%) was significantly worse than Tone 1 (97.39%), Tone 2 (85. 13%) and Tone 4 (89.47%). The interaction between Group and Tone did not reach significanc e [F (3, 34) = .047, p= .986], showing that experienced learners cons istently outperformed inexperienced learners across all four tones. Error type As shown in Tables 3-3 and 3-4, Tone 1 was produced with a very high accuracy rate by both groups of learners, and Tone 4 was sometimes produced as Tone 1 by both groups of learners. Tone 3 was more often misproduced as Tone 2 than the opposite by both inexperienced learners ( 2=5.927, p<.02) and experienced learners ( 2=3.920, p<.05). Table 3-3. Percentages of production conf usion made by inexperienced American learners of Mandarin, wit h the number of fre quency in parentheses Tone Identified as Tone 1 Tone 2 Tone 3 Tone 4 None Tone 1 97 (64) 3.0 (2) 0 (0) 0 (0) 0 (0) Tone 2 1.5 (1) 86.4 (57) 9.1 (6) 1.5 (1) 1.5 (1) Tone 3 0 (0) 24.2 (16) 68.2 (45) 1.5 (1) 6 (4) Tone 4 9.1 (7) 1.5 (1) 0 (0) 89.4 (59) 0 (0) Table 3-4. Percentages of production confusion made by inexperienced American learners of Mandarin, wit h the number of fre quency in parentheses Tone Identified as Tone 1 Tone 2 Tone 3 Tone 4 None Tone 1 97.7 (47) 2.1 (1) 0 (0) 0 (0) 0 (0) Tone 2 6.3 (3) 85.4 (41) 8.3 (4) 0 (0) 0 (0) Tone 3 2.1 (1) 22.9 (11) 70.8 (34) 0 (0) 4.2 (2) Tone 4 6.3 (3) 2.1 (1) 0 (0) 91.7 (44) 0 (0) 63

PAGE 64

Duration of the tone According to previous studies on duration of Mandarin lexical tones (Fu & Zeng, 2000; Howie, 1976; Tseng, 1990), Tone 3 is the longest tone, Tone 2 and Tone 1 are shorter, and Tone 4 is the shortest tone. In the following paragraphs, the durations of Tone 2, Tone 3 and Tone 4 produced on the sylla ble /na/ were measured. Using both waveforms and wide band spectrograms gener ated by Praat (Boersma and Weenink 2007), the duration of the tone was measured from t he beginning of the nasal consonant to the end of the vowel /a/. The data were analyzed (by using t-tests Table 3-5. The mean duration of tones pr oduced by learners who confused Tone 2 as Tone 3, in ms. Tones T2 produced as T2 T2 produced as T3 T3 produced as T3 Duration 617 539 703 s.d. 165 144 130 The most frequent mistake for inexper ienced and experienced learners when attempting to produce Tone 2 is to produce it as Tone 3. The duration of Tone 3 produced for intended Tone 2 (557 ms) is signif icantly shorter than Tone 3 produced for intended Tone 3 (0.604s) ( p =0.013). However, the durat ion of Tone 3 produced for intended Tone 2 is not significantly longer than Tone 2 produced for intended Tone 2 (0.471s) (p =0.355). Table 3-6. The mean duration of tones pr oduced by learners who confused Tone 3 as Tone 2, in ms. Tones T3 produced as T3 T3 produced as T2 T2 produced as T2 Duration 604 557 471 s.d. 92 97 116 The most frequent error for inexperienced an d experienced learners attempting to produce Tone 3 is to produce it as Tone 2. The duration of Tone 2 produced for intended Tone 3 (557 ms) is significantly longer than that of Tone 2 produced for intended Tone 2 (471 ms) ( p =0.034). However, the duration of Tone 2 produced for intended Tone 3 is not significantly shorter than Tone 3 (604ms) ( p =0.462). 64

PAGE 65

Table 3-7. The mean duration of tones pr oduced by learners who confused Tone 4 as Tone 1, in ms. Tones T4 produced as T4 T4 produced as T1 T1 produced as T1 Duration 263 267 559 s.d. 105 58 195 The most frequent error for both groups of Americ an learners attempting to produce Tone 4 is to produce it as Tone 1. The duration of Tone 1 produced for intended Tone 4 (267 ms) is significantly shorter than that of Tone 1 produced for intended Tone 1 (559 ms) ( p =0.005). However, the duration of Tone 1 produced for Tone 4 is not significantly longer than that of Tone 4 produced for intended Tone 4 (263 ms) ( p =1.000). Discussion Experienced learners were no more accura te in producing isolated tones than were relatively inexperienced learners. Th is result suggested that perhaps learning experience of twelve months is not sufficient to make a difference in production. Tone 4 produced as Tone 1 might be due to a pragmatic concern since an isolated sharp falling intonation is associated with a re jection intonation, but it is not due to a perception problem since none of participants who misperceived the isolated Tone 4 were participants who misproduced the tone. The reasons to choose Tone 2 as a substitution are analyzed from a perc eption consideration and a production consideration. Misperception of Tone 2 as Tone 3 by participants may lead them to misproduce Tone 2 as Tone 3. However, the perception problem is a minor reason for this phenomenon, since only five out of eigh teen participants who misperceived Tone 2 as Tone 3 misproduced Tone 2 as Tone 3 as well. 65

PAGE 66

American speakers could misproduce Tone 3 as Tone 2 because of L1 transfer. Tone 3 is a novel tone for American speaker s because it does not exist in the English intonation inventory. Theref ore, they produced Tone 3 with more difficultly. It was also found that alt hough the pitch contour of the tone category may be produced incorrectly, the durati on of the target tone cat egory is produced correctly. Therefore, we can say that the duration of a tone category is acquired faster than the pitch contour of that tone category. Furthermore, unlik e native speakers of Mandarin, American speakers of Mandarin paid more attention to duration, which is not primary acoustic character to native speakers of Mandarin, than to pitch contour when producing different tones. This is not a uni que phenomenon in the field of acquisition of L2 sounds that L2 speakers paid attention to ac oustic cues different from those used by native speakers to perceive or produce L2 features. For example, L1 Mandarin speakers producing L2 English only focus on the pitch difference between unstressed and stressed syllables and ignore the durati on and amplitude differences (Wang 2008). In this study, the preference of using duration ra ther than pitch to differentiate tones might be due to their English language backg round. The change of pitch does not lead to the change of vowel quality in English. Howe ver, the change of duration of vowels will change the vowels quality, such as /i/-/ / (71 vs. 31 ms) (Flege et al., 1997). Therefore, American speakers are more sensitive to the difference of duration of tones. Relationship between Perception and Production of Isolated Tones In this section, the relationship betw een perception and produc tion of Mandarin isolated tones is discussed in terms of both the two groups and individual learners within the groups. 66

PAGE 67

Results Figure 3-5 shows that both groups were highly accuracy at both perceiving and producing Mandarin isolated t ones (Inexp: 82.99% & 84.50%; Exp: 94.28% & 85.94%). Figure 3-5. Mean percentage of accuracy of perception and production of Mandarin isolated tones by two groups. Erro r bars indicate standard error. The accuracy rate of experienced learners perceiving Mandarin isolated tones is significantly better than thei r accuracy rate of producing the tones [t (126)=1.974, p=0.051], whereas there is no significant difference between the accuracy rates of perception and production of inexperienced learners [t (174)= -.371, p=0.711]. Table 3-8 displays the accuracy rate of perception and produc tion of isolated Mandarin lexical tones by inexperienced Am erican learners. According to a Pearson Correlation test, the perception and pr oduction scores are not correlated (r=0.04, p=0.372). Table 3-9 shows that accuracy rates of perception and production of isolated Mandarin tones by experienced American learners are also not significantly correlated (r=0.013, p=0.671). 67

PAGE 68

Table 3-8. Individual Inexperienced Am erican learners perception and production accuracy rate. These data are across all isolated lexical tones. Participant Perception Production Difference 1 100 100 0 2 100 100 0 3 100 91.75 8.25 4 75 91.75 -16.75 5 100 83.25 16.75 6 67 91.75 -24.75 7 83.5 75 8.5 8 100 91.75 8.25 9 83.5 50 33.5 10 83.25 66.75 16.5 11 83.25 66.75 16.5 12 41.75 91.75 -50 13 75 91.75 -16.75 14 75 100 -25 15 91.75 75 16.75 16 83.5 100 -16.5 17 100 91.75 8.25 18 100 100 0 19 49.75 66.5 -16.75 20 58.25 83.25 -25 21 91.75 91.75 0 22 83.5 58.5 25 Total 82.99 84.5 -1.51 Table 3-9. Individual experienced American learners perception and production accuracy rate. These data are across all isolated lexical tones. Participant Perception Production Difference 1 83.25 91.75 -8.5 2 100 75 25 3 75 91.75 -16.75 4 100 100 0 5 91.75 100 -8.25 6 91.75 100 -8.25 7 100 83.25 16.75 8 100 100 0 9 100 100 0 10 75 75 0 11 100 66.5 33.5 12 100 100 0 13 100 91.75 8.25 14 91.75 25 66.75 15 100 83.25 16.75 16 100 91.75 8.25 Total 94.25 85.84 8.34 68

PAGE 69

Discussio n Inexperienced learners perception developed faster than production of isolated Mandarin tones, whereas experienced learner s showed comparable development of both perception ability and production ability. Overall, there is no correlation between the accuracy of perc eption of isolated Mandarin tones and the accuracy of production of isolated Mandarin tones by either inexperienced or experienced l earners. However, the majori ty of individual students developed the ability to identify isolated Mandarin lexical tones faster than their ability to pronounce these tones. Conclusion Inexperienced learners have achieved accuracy rates of perception and production over 80%, except perception of Tone 2 and production of Tone 3. Experienced American learners of Mandarin achieved an accuracy rate over 80% across all four tones, and t hey performed better on perceiving but not producing isolated Mandarin tones than inexperienced American learner s. These results suggest that learning experience of three months which inexperienced learners had is sufficient for American speakers to be fam iliar with Mandarin isolated lexi cal tones. (not sure what you mean by concept of Mandarin tones. The experienced groups ability to perceptually categorize tones is stronger, as indicated by the fact that they identified tone with a higher accuracy rate. This study also confirmed the results from Wang et al ( 1999, 2003) that to American learners, Tone 2 is the hardest tone to be perceived accurately and Tone 3 is the most difficult tone to be produced accurately. 69

PAGE 70

Different from the previous study by Wang et al in 2003, the present study detected from errors made by American learners that they tend to use duration of tones rather F0 of tones to produce the di fference among Tone 2, Tone 3 and Tone 4. Furthermore, we analy zed the relationship between the perception and production of isolated tones by American learners in te rms of their accuracy rates on identifying and producing tones. As a group, we find that the experienced group s perception ability was better than their production ability,. whereas inexperienced learners perceive and produce isolated tones at a comparable le vel. However, .there was no correlation between the perception and t he production scores among either experienced or inexperienced learners. 70

PAGE 71

CHA PTER 4 PERCEPTION OF COARTICULATED MANDARIN TONES This chapter reports on a study on the per ception of coarticulated Mandarin tones by American learners of Mandarin with different learning ex perience. The first section introduces the research questions and hypot heses. The second section presents the methodology, including participants, stim uli, procedure, and data analysis. The third section reports on the data and st atistical analysis, and the fourth section discusses the results. Research Questions and Hypotheses In the present experiment, perception of M andarin coarticulated tones in disyllabic words by inexperienced and experienced Am erican learners was investigated. The experiment was aimed at answe ring the following questions: Research question 1. Does American learners id entification of Mandarin coarticulated tones improve wit h Mandarin learning experience? Because they had more learning experienc e, experienced learners were expected to perceive Mandarin coarticulated tones be tter than inexperienced learners. In other words, experienced learners were expected to identify these tones with a higher accuracy rate. Research question 2. What are the typical tonal per ception errors for American learners with different amount of language learning experience? There are two dimensions to Mandarin tones. One is tonal direction and the other is tonal height. Therefore, it was assumed that there would be two error types, tonal direction misperception and tonal height misperception. Since tonal direction is more emphasized in the classroom teaching, the frequency of tonal direction (Xing, 2006) 71

PAGE 72

misperception was expected to decrease wit h increasing language learning experience in classroom. Research question 3. What are the linguistic fa ctors that affect A merican learners ability to perceive coarti culated Mandarin tones accurately? Two linguistic factors were expected to influence American learners ability to identify coarticulated Mandarin tones. One is syllable position (initi al syllable vs. final syllable). Previous studies have not given an indication of which tones might be easier to perceive on which syllables, so no hypothes is regarding ease of perception according to syllable position was made. The other factor is tonal environment, which consists of three subordinate factors: tonal condition (isolated tones vs. coarticulated tones), tonal sequence (identical tones vs. nonidentical tones ) and tonal context (compatible context vs. conflicting context). Tones in monosyllabic word were labeled as isolated tones and tones in disyllabic words were labeled as coarticulated tones. Since pitch contours and pitch level of tones in coarticulation are altered due to coarticu latory effects of the neighboring tones; coarticulated tones were expected to receive a lower accuracy rate than those in isolation. If two identical tones occurred next to each other in a disyllabic word, such as Tone 1 + Tone 1, the tonal environment was l abeled an identical tone sequence. If two different tones occurred in a disyllabic wo rd, such as Tone 1 + Tone 2, the tonal environment was called a nonidentical tonal sequence. Considering the effect of repetition of tones, the acousti c features of tones in an id entical tonal sequence should be enhanced due to the repetition. Therefore, tones in an identical tone environment 72

PAGE 73

were expec ted to be identified with a higher accuracy rate than those in a nonidentical tone environment. According to Xu (1994), a compatible tonal context occurs when the value of the pitch of the offset of the preceding tone and the pitch value of the onset of the following tone are similar, such as Tone 2 + Tone 4, which have a high pitch offset and onset respectively, and a conflicting tonal context occurs when the pitch value of the offset of the preceding tone and that of the onset of the following tone are substantially different, such as Tone 1 + Tone 2, which have a high offset but a low onset respectively. The tonal direction in a compatible tonal contex t stays as it is, whereas the degree of pitch change in a conflicting tonal context is not as substantial as in a compatible tonal context. In other words, the pitch range of the tone becomes smaller in a conflicting tonal context than in a compatible tonal c ontext. Therefore, the acoustic features of tones are weakened or changed by the conflicting tonal cont ext. Tones in a compatible tonal context were therefore expected to be identified with a higher accuracy rate than those in a conflicting tonal context. In some circumstances, the pitch dire ction of the tone is even changed. When Tone 1 occurs after Tone 3 or Tone 4, the di rection of Tone 1 changes into rising from flat due to the conflicting tonal environment (Xu, 1997). In an other case, Tone 3 loses its rise and becomes a falling tone when Tone 1, Tone 2 or Tone 4 occurs after it. It was expected that the accuracy rate of identification of these t ones in which tonal direction had changed would be lower than the accuracy rate for those whose tonal direction was unchanged. 73

PAGE 74

Table 4-1. Mandarin tonal combinations in tonal env ironments (positive factors are unshaded; negative factors are shaded.) Tonal context environment Tonal combination Identical tone sequence Nonidentical tone sequence Compatible tonal context Conflicting tonal context T1+T1 X X T1+T2 X X T1+T3 X X T1+T4 X X T2+T1 X X T2+T2 X X T2+T3 X X T2+T4 X X T3+T1 X X T3+T2 X X T3+T4 X X T4+T1 X X T4+T2 X X T4+T3 X X T4+T4 X X Methods Participants The participants in the study were all from the participant pool in the study presented in Chapter 3. The fourteen participants in the beginning level had been studying Mandarin Chinese for three mont hs, and the fourteen participants in the intermediate level had been studyi ng Mandarin Chinese for twelve months. All of them were able to perceive each isolated lexical Mandarin tone with at least a 67% accuracy rate (at least two correctly identified tokens out of every three stimuli). Therefore, the misperceived coarticulated tones from the current study should not be due to participants inability to perceive isolated tones. Stimuli The perception study tested 45 non word stim uli, three stimuli for each of the fifteen possible disyllabic tone combinations. This design was to prevent participants from using their tonal knowledge of words with which they might already learn to identify 74

PAGE 75

tones. The 45 words were produced by the same native female speaker of Mandarin mentioned in Section 3.1 and ex hibited the four lexical M andarin tones. The words were all of the structure /na.na/. As with the isolated tone st imuli, the contour of F0, the amplitude envelope, and syllable duration were kept as the speaker naturally produced them. There is no creaky voice for Tone 3 and no glottalized voice quality for Tone 4. Procedure All stimuli in the perception experim ent were presented in random order and divided into three 15-stimulus blocks. Be tween each block, the participant was required to rest for one minute. Each tone combi nation was presented three times. For each stimulus, the participant viewed the s pelling nana on the computer screen. The participant clicked on the word to hear the corresponding sound and then wrote down the tone values of both syllables of each word on the provided answer sheet. Participants could adjust the volume to thei r comfort level and hear the sound as many times as they needed. Before the real test there were three warm-up trials, and the results from the three warm-up trials were not analyzed. Results and Analysis In this section, the results will be presented from two perspectives: accuracy rate and error type. Accuracy Rate Three factors were considered in the study: language learning experience (inexperienced vs. experienced), syllable position (initial position vs. final position), and tonal environment (identical tonal sequence vs. nonidentical tonal sequence, compatible 75

PAGE 76

tonal context vs. conflicting tonal contex t, changed tonal direction environment vs. unchanged tonal direct ion environment). Effects of language learning experience Figure 4-1. Mean percent correct identification of coarticulated Mandarin tones across all stimuli for inexperienced and experi enced American learners of Mandarin. Error bars indicate standard error. The results in Figure 4-1 show that the experienced le arners of Mandarin outperformed inexperienced learners at ident ifying coarticulated Mandarin tones (69.21% vs. 56.75%). This difference is highly statistically significant [ t (1, 838) = -4.82, p <.001]. There is a significant difference betw een inexperienced and experienced American learners of Mandarin on identifying indivi dual tones in a coarti culated environment. A two-way repeated ANOVA, with Tone as wit hin subject factor and Group as between subject factor, showed a main effect for Group [F (1,166 ) = 21.972, p<.001], indicating significant improvement of accuracy on iden tifying tones from inexperienced learners (50.964%) to experienced learner s (70.149%). A significant main effect of Tone was 76

PAGE 77

also found [F (3,166) = 17.21, p<.001]. Post hoc analys es (Bonferroni adjusted p<.05) showed that Tone 1 perception was significantly more accurate than Tone 4. In addition, both Tone 1 and Tone 4 were significantly more accurate than Tone 2 and Tone 3. Perception accuracy of Tone 2 and Tone 3 we re, however, comparable. There was no significant interaction between Group and T one [F (1,166) = 1.340, p=.263], which means the better performance of identifying coarticulated t ones by experienced learners was consistent across all four tones. Figure 4-2. Mean percent corre ct identification of coar ticulated tones across both syllables by inexperienced and experienced American learner s of Mandarin. Error bars indicate standard error. Effects of tonal condition Figures 4-3 A and B show how inexpe rienced and experienced groups of American learners identify Mandarin tones under different tonal conditions. Overall, the condition of tones in isolation allowed greater accuracy of i dentification (94.089% accurate responses) for listeners than did the condition of tones in coarticulation (57.741% accurate responses). A three way repeated measures ANOVA, with Tonal 77

PAGE 78

A) B) Figure 4-3. Mean percent co rrect perception for (A) I nexperienced and (B) Experienced groups for tones in isolation vs. tones in coarticulation. Error bars indicate standard error. Condition (isolation and coar ticulation) and Tone (Tone 1, To ne 2, Tone 3, Tone 4) as the within-subject factor and Group (inex perienced, experienced) as the between78

PAGE 79

subject factor, yielded a main effect of Tonal Condition [ F (1, 26) = 91.558, p < .001], Group [ F (1, 26) = 7.739, p =.01], Tone [ F (3, 24) = 20.314, p < .001]. It was also found that the interaction of Tonal C ondition X Tone is significant [ F (3, 24) = 14.33 p < .001]. Post hoc analyses (Bonferroni adjusted p< .05) was shown that, averaged across both groups, the accuracy rate of Tone 1, Tone 2, Tone 3 and Tone 4 were not significant different from each other, when they were in isolation. However, when they were in coarticulation, the accuracy rates of Tone 1 and Tone 4 were significantly higher than the accuracy rates of Tone 2 and Tone 3. No other significant interactions were found. Effects of tonal sequence Since a Tone 3 + Tone 3 series always cha nges to Tone 2 + Tone 3, perception of Tone 3 was always in a nonidentical tonal s equence. Therefore, only Tone 1, Tone 2 and Tone 4 will be discussed here for comparison of results under the two tonal sequences. Figures 4-4 A and B show how inexpe rienced and experienced groups of American learners identify Mandarin tones under different tonal sequences. Overall, the identical tonal sequence allo wed greater accuracy of ident ification (70.631% accurate responses) for listeners than did the nonidentical tonal sequence (64.285% accurate responses). However, the three-wa y repeated measures ANOVA with Tonal Environment (identical, nonidentical) and Tone (Tone 1, Tone 2, Tone 3, Tone 4) as the within-subject factors and Gr oup (inexperienced, experienced) as the between-subject factor did not revealed the main effect of Tonal Environment [ F (1, 54) = 2.739, p = .104], but it yielded the ma in effect of Group [ F (1, 54) = 4.965, p < .05] confirming that, averaged across both tonal sequences, experienced learners out performed inexperienced learners. A marginally signif icant interaction of Tonal sequences X Tone 79

PAGE 80

X Group [ F (2, 53) = 2.988, p = .059] was also found. Further analys es were conducted to examine this interaction. A) B) Figure 4-4. Mean percent corre ct perception for (A) Inexper ienced and (B) Experienced groups for tones in identical tone sequence vs nonidentical tone sequence. Error bars indicate standard error. The analyses showed that the inexperienc ed group (Figure 4-3 A) only made more accurate identifications under identical tonal sequences than under nonidentical tonal sequences for Tone 1 (85.79% & 59.01%) [ t (1, 109) = 3.342, p =.001]. However, inexperienced listeners made more errors un der identical tonal sequence for Tone 2 80

PAGE 81

(45.14% & 54.88%) and made almost equally accurate judgments in both tonal sequences for Tone 4 (64.32% & 64.26%). H owever, the difference of accuracy rates due to the tonal sequences for Tone 2 [t (1, 108) = -1.174, p =.243] and Tone 4 [ t (1, 108) = -.008, p =.993] were not significant. Although the experienced group (Figure 4-3 B) showed a tendency to perceive tones more accurately in identical tonal sequence than nonidentical tonal sequence (Tone 1, 86.89% vs. 76.72%; Tone 2, 63.11% vs. 54.76%; Tone 4, 78.54% vs. 71.83%), none of the tones difference in accuracy rate for the two tonal sequences was significant: Tone 1 [ t (1, 108) =1.454, p =.149], Tone 2 [ t (1, 108) =1.030, p =.305], or Tone 4 [ t (1, 108) =.906, p =.367. Effects of tonal context Figures 4-5 A and B show how inexpe rienced and experienced groups of American learners identify Mandarin tones under different tonal contexts. Overall, the accuracy rate of identifying tones in the com patible tonal context (62.5%) is higher than the accuracy rate for the conflicting tonal context (53.87%). The three -way repeated measures ANOVA with Tonal Context (compat ible, conflicting) and Tone (Tone 1, Tone 2, Tone 3, Tone 4) as the within-subject factor s and Group (inexperienced, experienced) as the between-subject factor showed that there were significant main effects of Tonal Context [F (1, 54) =13.655 p=.001], Gro up [F(1, 54) =9.308 p=.004], and Tone [F(3, 52) =16.379, p< .001]. No significant interaction was found. Even though no significant interaction among the factors was found, further analyses were conducted to investigate the effe ct of learning experience. It was found that inexperienced groups of listeners tend to perceive T one 1 (72.07% vs. 58.88%), Tone 2 (58.34% vs 45.76%) and Tone 4 (66.09% vs. 61.36%) more accurately in the 81

PAGE 82

A) B) Figure 4-5. Mean percent corre ct perception for (A) Inexper ienced and (B) Experienced groups for tones in compatible vs. conflicting tonal context. Error bars indicate standard error. compatible tonal context t han in the conflicting tonal c ontext and Tone 3 (38.07% vs. 43.46%) less accurately in the compatible t onal context. Furthermore t-tests indicated that the differences between the accurate perception rates of Tone 1 [ t (1, 110) =1.844, p =.068] and Tone 2 [ t (1, 110) =1.766, p =.080] which occurred in different tonal contexts were only marginally significant and the di fferences between the accurate perception 82

PAGE 83

rates of Tone 3 [ t (1, 82) =-.584, p =.561]; and Tone 4 [ t (1, 110) =705, p =.482] which occurred in different tonal cont exts were not significant. The same results were found for the expe rienced listeners. Although experienced groups of listeners tend to perceive Tone 1 (84.52% vs. 73.84%), Tone 2 (61.93% vs 51.77%) and Tone 4 (73.80% vs. 73.21%) more accurately in the compatible tonal context than in the conflicting tonal cont ext and Tone 3 (64.25% vs. 81.59%) less accurately in the compatible tonal context, t-tests indicate d that the differences between the accurate perception rate of tones whic h occurred in different tonal contexts was marginally signific ant on Tone 1, [ t (1, 109) =.1.781, p =.078] but not significant on Tone 2 [ t (1, 110) =1.455, p =.149]; Tone 3 [ t (1, 82) =-.403, p <.688]; or Tone 4 [ t (1, 110) =.092, p =.927]). In terms of change of pitch direction, in the literature review, I explained that both carry-over and anticipatory influences change t he pitch ranges but not the direction of tones, except Tone 1 after Tones 3 and 4 Tone 1 ri ses instead of staying flat (Xu, 1997). However, this kind of modification dr iven by tonal context does not change the perceptual tonal category to native speakers of Mandarin. In Figure 4-6, the accuracy rate of i dentifying final syllable Tone1 occurring after Tone 1 and Tone 2 was compared with the accu racy rate of identifying final syllable Tone1 occurring after Tone 3 and Tone 4. In the former tonal envir onment, the direction of Tone 1 was maintained, whereas in the la tter tonal environment, the direction of Tone 1 was changed from flat to rising. It was shown that Tone 1 with unchanged direction was identified better than Tone 1 with chang ed direction (75.607% vs. 61.296%). A twoway repeated ANOVA, with Context (uncha nged pitch direction and changed pitch 83

PAGE 84

direction) as within group factor and Group (i nexperienced group and experienced group) as between subject fact or, revealed a significant main effect of Context [ F (1, 54) =8.715, p<.01]. However, there was no significant main effect of Group [F (1, 54) =2.52, p=.118]. There is no inte raction between Context and Group [ F (1, 54) =1.548, p=.219]. These results suggested that both gr oups were more accurate at identifying Tone 1 under the unchanged pitch direction conditi on (i.e., after anot her Tone 1 or after Tone 2) than under the changed pitch direction condition. Figure 4-6. Mean percent correct identification of coarticulated final syllable Tone 1 in different tonal contexts by two groups of learners of Mandarin. Error bars indicate standard error. Effects of syllable position Figures 4-7 A and B show how inexpe rienced and experienced groups of American learners identify Mandarin tones in different syllable positions. Overall, accuracy rate for the initial syllable posit ion (59.90% accurate responses) was lower than the accuracy rate for the final syllable position (69.05% accurate responses). The threeway repeated ANOVA yiel ded significant main effects of Syllable position [F (1, 82) = 33.159, p< .001], Group [F (1, 82) = 22.865, p< .001],and Tone [F (3, 80) = 84

PAGE 85

A) B) Figure 4-7. Mean percent co rrect perception for (A) I nexperienced and (B) Experienced groups for tones in initial vs. final syllable position. Error bars indicate standard error. 16.277, p< .001]. There was also a significant interaction of Tone X Syllable Position [F (3, 80) = 18.613, p< .000], a marginally sign ificant interaction of Tone X Group [F (3, 80) = 2.611, p= .057], and a significant Syllable position X Group X Tone interaction [F (3, 80) = 2.984, p< .05]. 85

PAGE 86

Inexperienced learner s identified Tone 2 ( 41.61% initial vs. 62.52% final), Tone 3 (20.25% initial vs. 63.52% final), and Tone 4 (59.57% initial vs. 67.88% final) better in final syllable position, whereas they identif ied Tone 1 (73.23% initial vs. 57.71% final) better in initial syllable position. However, re sults of T-test analyse se showed that the difference between accuracy percentages as a function of syllable position was significant for Tone 1 [t (1,110)= 2.182, p< .04], Tone 2 [t (1, 110)=3.019, p<.004] and Tone 3 [t (1, 82)=-5.836, p<. 001], but not significant fo r Tone 4 [t (1, 110)=-1.243, p=.217]. Experienced learners identified Tone 2 (54. 77% initial vs. 58.93% final), Tone 3 (46.76% initial vs. 86.55% final), and Tone 4 (67.88% initial vs. 79.14% final) better in final syllable position, whereas they identified Tone 1 at al most the same rate (79.18% & 79.18%)2 in both syllable positio ns. T-testresults, howev er, showed that the difference between accuracy percentages as a function of syllable position was highly significant for Tone 3 [ t (1, 82) =-5.524, p< .001] and marginally for Tone 4 [ t (1, 110) =1.777, p=.078], but not significant for T one 1 [t (110) =.000, p=1] or Tone 2 [ t (1, 110) =.591, p=0.556]. Error type Confusion matrices on identifying coarticu lated tones in differ ent syllable positions are shown in tables 4-2 to 4-5.The data was based on 168 responses for Tone 1, Tone 2, Tone 4 (12 stimuli 14 participants), and 126 responses for Tone 3 (9 stimuli 14 2 Though the average mean of identification accuracy ra te of initial Tone 1 is the same as the average mean of identification accuracy rate of final Tone 1, the standard deviations of these two accuracy rates are different (30.903 & 33.403). 86

PAGE 87

participants ). The results of Tone 3 in the t onal combination of Tone 3 +Tone 3 were not included. As shown in the below tables, three kinds of error types were found: tonal direction misperception and tonal height misperception. The definition of each error type is as following: Tonal direction misperception : the direction of the target c oarticulated tone is different from the identified tonal category. Tonal height misperception : the direction of the target coarticulated tone is the same as the identified tonal category, bu t the tonal height is different. Table 4-2. Confusion matrix for tones in initial position identif ied by inexperienced learners, with the per centages in parentheses Tone Identified as Tone 1 Tone 2 Tone 3 Tone 4 Tone 1 133 (79.17) 14(8.33) 0 21(12.5) Tone 2 24 (14.3) 92 (56.7) 21(12.5) 31 (18.5) Tone 3 12 (8.8) 22 (16.2) 59 (43.4) 43 (31.6) Tone 4 20 (11.9) 14 (8.3) 20 (11.9) 114 (67.9) Table 4-3. Confusion matrix for tones in final position identified by inexperienced learners, with the per centages in parentheses Tone Identified as Tone 1 Tone 2 Tone 3 Tone 4 Tone 1 134 (79.8) 23 (13.7) 3 (1.8) 8 (4.8) Tone 2 9 (5.4) 99 (58.9) 55 (32.7) 5 (3) Tone 3 0 17(13.5) 109 (86.5) 0 Tone 4 9 (5.4) 14 (8.3) 12 (7.1) 133 (79.2) Table 4-4. Confusion matrix fo r tones in final position identif ied by experienced learners, with the percentages in parentheses Token Identified as Tone 1 Tone 2 Tone 3 Tone 4 Tone 1 97 (57.7) 56 (33.3) 3 (1.8) 12 (7.2) Tone 2 21 (12.5) 103 (61.3) 37(22.0) 7 (4.2) Tone 3 8 (6.3) 29 (23.0) 80 (63.5) 9 (7.2) Tone 4 12 (7.2) 30 (17.9) 13 (7.7) 113 (79.2) 87

PAGE 88

Table 4-5. Confusion matrix fo r tones in final position identif ied by experienced learners, with the percentages in parentheses Token Identified as Tone 1 Tone 2 Tone 3 Tone 4 Tone 1 123 (73.2) 17 (10.1) 0 28 (16.7) Tone 2 50 (29.8) 70 (41.7) 15 (8.9) 33 (19.6) Tone 3 29 (23) 31(24.6) 26 (20.6) 40 (31.7) Tone 4 33 (19.6) 16 (9.5) 18 (14.3) 101 (60.1) The tonal category is defined in t he Table 4-6 and how error types were categorized are shown in Table 4-7. The following paragraphs explained how the errors were categorized in details. 1) Syllable initial T1 is a flat tone, whereas the category of T2 is a rising tone and T4 is a falling tone. When syllable initial T1 is perceived as T1, T2 or T3, the error was labeled as a direction misperception. 2) Syllable initial Tone 2 is a rising tone, whereas T1 is a fl at tone, and T4 is a falling tone. When syllable initial T2 is perceived as T1 or T4, (I l abeled) the error as a direction misperception. Syllable initial T3 phonetically moves in the same direction as T2. However, T3 starts lower and ends lower than T2, and turns later than T2. Therefore, T3 misperceived as T2 was labeled as a height misperception. 3) Syllable initial Tone 3 is a falling tone, wher eas T1 is a flat tone, T2 is a rising tone, and T4 is a falling tone. When syllable initial T3 was perceived as T1 or T2, (I labeled) the error as a direction misperception. Syllable initial T3 moves in the same direction as T4. However, T3 is a low falling tone, whereas T4 is a high falling tone. Wh en T3 was misperceived as T4, (I labeled) the error as a height misperception. 88

PAGE 89

4) When syllable initial T4 was perceived as T1, T2 or T3, the error was a direction misperception. 5) Syllable final T1 shows a rising direction when it occurs after T3 or T4, but the degree of rising is not as deep as Tone 2. Therefore, when T1 was identified as T2, (I labeled) the error as a height misperception. When T1 was perceived as T3 or T4 by American learners, the misperception was categorized as a direction misperception. 6) Syllable final T2 presents a rise, whereas T1 has no direction change and T4 falls. Therefore, T2 misperceived as T1 or T4 was labeled as a direction error. Similar to syllable final T2, syllable final T3 rises. However, syllable final T3 does not rise as high as T2. Therefore, syllable fi nal T2 perceived as T3 is due to a confusion of tonal height. 7). Syllable final T3 perceived as T1 or T4 was labeled a direction error. Syllable final T3 perceived as T2 was labeled a tonal height error. 8) Syllable final T4 is a falling tone, unlike T1, T2, or T3. Therefor e, syllable final T4 perceived as T1, T2 and T3 was labeled as a tonal directional error. Table 4-6. Description of Mandar in Chinese tonal categories Tonal category Phonological description Phonetic description T1 Flat High flat T2 Rising Falling-rising (low onsite and high offsite) with a early turning point T3 Falling-rising Falling-rising (low onsite and low offsite) with a late turning point T4 Falling High falling 89

PAGE 90

Table 4-7. Tonal conf usions categorized into error types Target coarticulated tone Tonal direction error Tonal height error 1) Initial high flat T1 T2 T3 T4 2) Initial rising T2 (acoustically fallingrising with a early turning point) T1 T4 T3 3) Initial low falling T3 ( effect of tonal context) T1 T2 T4 4) Initial high falling T4 T1 T2 T3 5) Final rising T1 (acoustically rising after T3 or T4, effect of tonal context) T3 T4 T2 6) Final rising T2 (acoustically fallingrising with a early turning point) T1 T4 T3 7) Final falling-rising T3 T1 T4 T2 8) Final falling T4 T1 T2, T3 Figure 4-8 shows that inexperienced lear ners tended to incur more directional misperception errors than experienced learne rs. This difference was non significant for initial syllable Tone 1 [ 2=2.404, p=.121] and init ial syllable Tone 4 [ 2=2.183, p=.140]. However, it was marginally signi ficant for final syllable Tone 1 [ 2=3.630, p=.057] and significant for initial syllable Tone 2 [ 2=4.325, p=.038], final syllable Tone 2 [ 2=13.105, p=.000], initial syllable Tone 3 [ 2=11.470, p=.001], and final syllable Tone 4 [ 2=9.002, p=.003] Experienced learners did not make a single directional error when identifying final syllable Tone 3, whereas inexperienced learners misperceived the rising direction of final syllable Tone 3 17 times. Figure 4-9 shows that experienced learners made fewer tonal height misperception errors than in experienced learners when they identified final syllable Tone 1 [ 2=12.229, p=.000] and fi nal syllable Tone 3 [ 2=3.829, p=.05] However, on 90

PAGE 91

Figure 4-8. Frequency of tonal directional misidentification by two groups of American learners Figure 4-9. Frequency of tonal height misi dentification by two groups of American learners certain tones the experienced learners made more height perception errors than the inexperienced group, such as initial syllable Tone 2 [ 2=1.12, p=.29], initial syllable Tone 91

PAGE 92

3 [ 2=.162, p=.688] and final syllable Tone2 [2=4.850, p=.028]. However, the differences between the frequencies of misidentifyi ng tonal height of initial syllable Tone 1 and Initial syllable Tone 3 made by two groups of learners are not significant according to chi-square tests. Discussion This study was designed to investigate how coarticulated Mandarin Chinese tones in disyllabic words are identif ied by American learners with different levels of language experience (inexperienced learners vs. ex perienced learners). The purpose of the perception experiment was to test whet her American learners identification of coarticulated Mandarin Chi nese tones improves with language learning experience, whether the syllable position affects identification accuracy, and whether tonal context affects identification types. Effects of Learning Experience The results show that experienced Am erican learners of Mandarin performed better than inexperienced learner s on identifying coarticulated tones in disyllabic words, and this advantage was consistent across all four Mandarin tones. The results also show that experienced learners perform ance was less influenced by the linguistic factors, such as tonal environment and syllable position. In terms of tonal environment, inexperienced learners performance on ident ifying coarticulated tones was more affected by the tonal environment than exper ienced learners. Inex perienced learners identified Tone1 better when it occurred next to an identical tone and in the unchanged tonal direction environment than next to a nonidentical t one or in the changed tonal direction environment, whereas experienced learners did not perform statistically differently in these various environments. Regarding the effect of syllable position, 92

PAGE 93

inexperienc ed learners showed a stronger tendency than experienced learners to perceive final syllable tone more accurately than initial syllable tone in more tonal categories. Therefore, we can say that learning experience makes American speakers more sensitive to pitch change across tonal categori es in coarticulated tonal environments but less sensitive to pitch change within tonal ca tegories triggered by t onal environments. In other words, more exposure to the Chinese language along with learning experience helps American speakers to improve their abi lity to perceive tonal features from phonetic cues to phonological categories. For example, inexperienced American learners misperceive Tone 1 (as Tone 2) more frequently when it is slightly rising due to the tonal environment, but this small degree of rising does not cause Tone 1 to change into Tone 2 in native Mandarin speakers perception. The fact that experienced American learners are more accurate than inex perienced learners at identifying Tone 1 when its pitch direction was changed due to tonal context suggested that with a longer experience with Mandarin, American learners are better at ignoring within-category differences (induced by phonetic contexts ) and focus more on across-category differences. In other words, in comparison to inexperienced learners, they were able to ignore the rise in pitch at the onset of T one 1 when it occurs after Tone 3 and Tone 4, and focus instead on the rest of the pitch cont our to arrive at a constant percept on Tone 1. In addition, learners with more learning ex perience may also be less influenced by their L1 background than learners with less lear ning experience. Although English is not a tonal language, it does use intonation to si gnal pragmatic meanings, such as rising 93

PAGE 94

intonation for interrogative. The negative L1 influence is more often transferred by inexperienc ed American lear ners, and it makes inexper ienced American learners perceive final syllable Tone 2 better than init ial syllable Tone 2. The advantages that experienced learners have can be attributed to the increased L2 input that comes along with their longer learning experience. This idea is supported by Wayland et al. (2003), which reported that in discrim inating the Thai middle and low tone contrast, experienced learners of Thai outperformed inexperienced learners who did not have any learning experience in Thai. Previous studies showed t hat not only in the classroom but also in the laboratory, more L2 percept ion input seems to lead to greater ability to perceive L2 tonal features, such as Mandarin and Amer ican listeners discriminating the mid tone and low tone contrast in Thai (Wayland & Guion, 2004), Americ an listeners identifying Mandarin tones (Wang et al., 1999), and M andarin listeners and American listeners identifying Cantonese tones (Francis et al ., 2008). Extended to L2 input in general, listeners with more L2 perception experi ence have more advantages to perceive L2 tonal features correctly com pared with inexperienced listeners. Effects of Tonal Condition In the present study, three types of tonal environments were tested. Two types of tonal condition were tested. The first type of tonal condition is called tones in isolation and the second type of tonal condition is calle d tones in coarticulation. It found that tones in isolation is identified better than tone s in coarticulation and this kind of superior cross over four tones. The fa ct can be possibly attributed to the effect of attention distribution. It is assumed t hat the attention on a tone in c oarticulation was less than it was in isolation. 94

PAGE 95

Effects of Tonal Sequence The type of tonal environment is defined by considering the tonal sequence. T here are two types of tonal s equence: identical tonal s equence vs. nonidentical tonal sequence. Identical tonal sequence was assum ed to be a positive tonal environment for American learners to identify tones, wher eas nonidentical tonal sequence was assumed to be a negative tonal environment. It was assu med that identical tonal sequence helps learners to identify tones by showing tonal repetition. This was not confirmed by the results. There was no main effect of tonal sequenc e. However, if we look at the effect of the factor on a specific tone, only Tone 1 was identified with a significantly higher accuracy rate in identical tonal sequence than in nonidentical tonal sequence. The difference between the accuracy rates of Tone 2 and Tone 4 in two tonal sequences does not reach the significant level. The t onal combination of Tone 1 + Tone 1 is the only tonal combination in two positive tonal environments, which is identical tonal sequence and compatible tonal context, wher eas Tone 2 and Tone 4 in any tonal combination are with one posit ive and one negative environment. Therefore, Tone 1 in the identical tonal sequence is identified better, but Tone 2 and Tone 4 in identical tonal sequence are not identified better due to the negative interaction with the tonal context that they occur in. Effects of Tonal Context The type of tonal environment is categorized by the relation between the offset of preceding tone and the onset of following tone (compatible and conflicting tonal context). Since a conflicting tonal context reduces the phonetic pitch range of the tones, identification might be more diffi cult in this type of contex t. Therefore, the compatible 95

PAGE 96

tonal context was predicted to improve American learner s t one identification, while the conflicting context was expected to inhibit it. This study showed American learners was affected by the linguistic fa ctor of tonal context. For final syllable Tone 1, t he tonal context changes not only the pitch but also the direction of the tone. Final syllable Tone 1 tu rns into a rising tone when it occurs in a conflicting tonal context. As expected, final syllable Tone 1 in conflicting context was identified less accurately the same tone in a compatible context. Effects of Syllable Position Syllable position affected American learner s perception of coarticulated tones. Tones in final syllable position were ident ified better than tones in initial syllable position. More specifically, coarticulated Tone 2 and Tone 4 were identified better in final than initial position. He re, it might be due to L1 perc eption transfer. There is no lexical tone in English. Howe ver, English does use intonati on to differentiate statements and questions. A falling intonation makes a st atement, and a rising intonation indicates a yes-no question. To perceive the pragmatic function, English listeners pay more attention to the pitch direction of sentence final intonation. Therefore, the American learners in the present study mi ght transfer this kind of att ention to perception of tones in disyllabic words. In other words, the Am erican learners pay more attention to the tonal direction of the word final tone than t he word initial tone. Therefore, the accuracy rate of identifying Tone 2 and Tone 4 is higher in word final position than in word initial position. The effect of syllable positi on on identifying Tone 3 is signi ficant for both groups of learners. Both of groups of learners indentified Tone 3 better in final syllable position. 96

PAGE 97

Tone 3 cha nges from a falling-rising tone to a low falling tone when it occurs on initial syllable position, whereas the tonal direction of Tone 3 on final syllable position remains as the same as it is in isolation. The changed tonal direction of initial Tone 3 might cause the difficulty for both groups of learners to identify the tone. The above results may be also due to a recency effect; final tones are remembered better than initial tones. Therefore, final syll able tones are misperceived less frequently than initial syllable tones. We also noticed that the durat ion of the tone on the final syllable is longer than the one on the initial syllable. The shape of the to ne on the final syllable might be more fully presented, and the fully present ed tone might be recognized more easily. Therefore, the accuracy rate for the final tone is higher than the one for the initial tone. The following argument is speculation and n eeds to be tested in future studies. In certain contexts, Tone 3 changes from a fa lling-rising tone to a falling tone, whereas Tone 1 changes from a high flat tone to a rising tone. Tone 3 chan ges from a complex tone to a simple tone, but Tone 1 changes fr om one simple tone to another. Therefore, the change of Tone 3 is more dramatic in term s of tonal types and it is more difficult to generalize to the same tonal category. Ther efore, even with more learning experience, American learners still had a hard time identifying coarticula ted Tone 3 in initial syllable position. Development of Perception of Coarticulated Tones Two types of errors were found in both groups: tonal direction misperception and tonal height misperception. Experienced learners committed fewer tonal directional errors than inexperienced learners. Howe ver, experienced learners still made a 97

PAGE 98

considerable amount of tonal height errors. Therefore, the ability to i dentify tonal direction may improve faster than the abilit y to identify tonal height among English speakers. Two reasons could explain this phenom enon. The first is the influence of the language teaching method. In Chinese language classrooms, instructors focus on introducing the differences in tonal direction of Mandarin tones in isolation but not on the differences of their tonal height. They only drill their students in the classroom on identifying tonal direction. T herefore, the more time Amer ican learners spend studying in the classroom, the more practice they get in identifying tonal direction. The practice experience improves experienced learners ability to identif y tonal direction. Meanwhile, the lack of practice in identif ying tonal height keeps learners from improving this skill. The second potential reason for this unparallel development is the tonal environment. Within a disyllabic word, the tonal direction of one tone can be implied by the tonal direction of its nei ghboring tone. For example, the tonal directions of Tone 4 and Tone 2 in the tonal combination T one 4 +Tone 2 are implied by the opposite directions of tones in the tonal combinati on. Listeners can notice that the pitch changes of the two contour tones are different from each other. However, the height of one tone is different compared with its neighboring tone. For exam ple, Tone 3 in the tonal combination Tone 3 + Tone 2 is a low falling t one. Its falling direction is more noticeable than its low range in the tonal direction of falling and rising. Theref ore, tonal height is harder to identify in a disyllabic tonal env ironment even for ex perienced American learners. 98

PAGE 99

Conclusion Experienc ed learners identified coarticula ted tones with higher accuracy rate than inexperienced learners. Moreov er, experienced learners tonal identification ability was more influenced by tonal coarticulation and tonal repetition and less influenced by tonal context and syllable position than inexperienc ed learners ability wa s. In other words, with increasing learning experie nce, American learners ability to categorically perceive coarticulated Mandarin tones improves. The results indicate that tonal environments and syllable position play import ant roles in identifying coarticulated tones. This implies that in future models of tone perception, the surrounding tonal environment needs to be considered to account for identification accuracy of lexical tones by L2 learners. Furthermore, language learning experience im proves American l earners ability to identify Mandarin tone direction more than their ability to identify tone height. 99

PAGE 100

CHAPT ER 5 PRODUCTION OF COARTICULATED MANDARIN TONES This chapter reports on a study of the pr oduction of coarticulated Mandarin tones by American learners of Mandarin with different learning ex perience. The first section presents the research questi ons and hypotheses. The se cond section presents the methodology, including participants, stim uli, procedure and data analysis. The third section reports the data and statistical analys is, and the fourth section discusses the results. Research Questions and Hypothesis In the present experiment, production of M andarin coarticulated tones in disyllabic words by inexperienced and experienced Am erican learners was investigated. The experiment was aimed at answe ring the following questions: Research question 1. Does American learners production of Mandarin coarticulated tones improve with Mandarin learning experience? Experienced learners were expected to produce Mandarin coarticulated tones better than inexperienced learners. In other words, experienc ed learners were expected to pronounce these tones with a higher accuracy rate. Research question 2. What are the typical tonal production errors for American learners with different amounts of language learning experience? Recall that Wang et al. ( 2003) categorized two types of errors American learners make when producing isolated Mandarin tones: tonal dire ction confusion and tonal height confusion. They also f ound that the tonal height conf usion was more resistant to improvement. In the present st udy, these two types of errors were expected to occur in 100

PAGE 101

American learners production of coarticulated Mandarin tones in disyllabic words, and the tonal height confusion was expected to be harder for learners to overcome. Research question 3. What are the linguist ic factors that affect American learners ability to produce coarticulated Mandarin tones accurately? As with perception, two linguistic factor s were expected to influence American learners production of coarticulated Mandarin tones. One is syllable position (initial vs. final syllable position). There was no predict ion regarding which tones on which syllable would be produced better. The other factor is tonal environment, including tonal condition (isolation vs. coarticulation), t onal sequence (identical vs. nonidentical tone sequence) and tonal context (compatible vs. conflicting tonal context). Tones in monosyllabic word were labeled as isolated tones and tones in disyllabic words were labeled as coarticulated tones. Considering the effect of coarticulated environment, tones in coarticulation were expected to be produced with a lower accuracy rate than those in isolation. Since all the participants had already lear ned to produce isolat ed Mandarin tones, they were expected to produce these four tones more easily in the identical tonal sequence by simply repeating the tones in s equence. Therefore, tones in identical tone environment were expected to be produced wit h a higher accuracy rate than those in nonidentical tone environment. The concept of compatible and conflicting tonal contexts was introduced in the previous chapter. The tones in compatib le tonal context are connected smoothly, whereas there is a pitch gap between the tones in conflict. American learners were expected to feel more comfor table producing coarticulated tones in a tonal environment 101

PAGE 102

without a tonal gap. Ther efore, tones in the compatible tonal context were expec ted to be produced with a higher accuracy rate than those in the conflicting tonal context. Table 4-1 categorized the possi ble Mandarin tonal combinatio ns on disyllabic words into four tonal environments. Method Participants The participants were selected from the pool of participants in t he study of isolated tone production. Nine inexperi enced American learners and ni ne experienced American learners participated in the st udy. These 18 participant (8 female and 10 male) were all part of the group of 28 who co mpleted the coarticulation per ception task. All of them attained at least a 67% accuracy rate of both perceiving and producing isolated tones in the first experiment. This step was designed to reduce the possibility that the participants inaccurate production of coarti culated tones was due to their inability to perceive or produce Mandarin tones in isolation. Stimuli The study tested 48 non words. This desi gn was intended to prevent participants from using their tonal knowledge of words with lexical meanings to produce tones. As with the perception stimuli, t he target words all had the st ructure /na.na/. The syllable structure was chosen to avoid the possibility of misproducing tones by placing English stress on one syllable, since English speak ers tend to stress CVN over CV syllable structure. There are three reasons to choose nasal /n/ as the onset consonant in the stimuli: 1) the pitch value of the vowe l is not changed due to the consonant. Nasal consonant /n/ does not change the F0 the following vowel. 2) There is no broken pitch 102

PAGE 103

track within one syllable. Nasal /n/ keeps t he pitch track consistent. 3) Due to the character of nasals in the spectrogram, t here is a very cl ear boundary between the two syllables. Procedure All stimuli in the production experiment were presented in random order, and each possible tone combination was presented three times. The stimuli were divided into three blocks. Between each block, the partici pant was required to rest for one minute. This design ensured that participants effort on producing each token was approximately equal. For each stimulus, the participant vi ewed one word in Pinyin on the computer screen. The same word was presented again after the participant finish ed their first try. The second try on the pronunciation was count ed. This step was intended to ensure that the participants tonal errors were not due to their unfamiliarity with the stimuli. In order to ensure that the participant produced coarticulated tones on disyllabic words, no pause was allowed in their production between the two syllables of each word. The 48 stimuli (The production stimuli include T3 + T3, even though the perception stimuli didn't) were preceded by five practice st imuli that were not analyzed. The productions were collected using a solid state reco rder (PMD660/U3B) with a professional microphone (Audio-Techinca AT4041 Car). Befo re the real test, there was three warmup stimuli, and the results from the th ree warm-up stimuli were not analyzed. Judgment Two native speakers of Mandarin with a Beijing acc ent transcribed the tones produced by participants based on their percept ion. Answer sheets were provided. The produced stimuli were printed with no tonal di acritics. Judges needed to provide a tonal 103

PAGE 104

diacritic corresponding to the tone they heard (Wang 2003). The program PRAAT was also used to generate pitch contours to avoid the misperception of the speakers by their categorical perception. It was observed dur ing the rating that the Mandarin speakers tended to perceive all the American learners low tones in initial syllable position, whether low flat tone or lo w falling tone, as Tone 3. Results and Analysis In this section, the results will be presented based on two measures of learner knowledge: accuracy rate and error type. Accuracy Rate Three factors were considered in the study: language learning experience (inexperienced vs. experienced), syllable position (initial position vs. final position), and tonal environment (isolated tonal condition vs. coarticulated tonal cond ition, identical vs. nonidentical tonal sequence, compatible vs. conflicting tonal context). Effects of language learning experience The overall results in Figure 5-1 show that the experienced lear ners of Mandarin performed better than inexperienced learner s on producing coarticulated Mandarin tones (54.81% inexperienced vs. 80.38% ex perienced). This difference is highly statistically significant [ t (1, 538) = -7.876, p <.001]. As shown in Figure 5-2, overall, tonal production by ex perienced American learners of Mandarin were judged to be more a ccurate by native listeners than that of inexperienced learners (79.17% vs. 52.92%), ma in effect of Group [F (1,106) = 54.93, p<.001]. In addition, production accuracy varied significantly across the four tones, main effect of Tone [F (3,104) = 44.999, p<. 001], Post hoc analyses (Bonferroni adjusted p<.05) showed that Tone 1 production was signi ficantly more accurate than all other 104

PAGE 105

tones. In addition, Tone 2 and Tone 4 were si gnificantly more accurate than Tone 3. Production accuracy of Tone 2 and Tone 4 were, however, comparable. A significant interaction between Group and Tone [F (1,104) = 3.702, p=.014] was also obtained. This was due mainly to the fact that exper ienced learners were more accurate than inexperienced learners in t heir production of Tone 1 (p<.001), Tone 3 (p<.001) and Tone 4 (p<.001) but not Tone 2 (p=.210). Figure 5-1. Mean percent correct producti on of coarticulated Mandarin tones by inexperienced and experienc ed learners of Mandarin Figure 5-2. Mean percent corre ct production of coarticulat ed tones by inexperienced and experienced American learners of M andarin. Error bars indicate standard error. 105

PAGE 106

Effects of tonal condition A) B) Figure 5-3. Mean percent correct production of coarticulated Mandarin tones in different tonal conditions by (A) inexperienced learners and (B) experienced learners. Error bars indicate standard error. Figures 5-3 A and B show how inexpe rienced and experienced groups of American learners identify Mandarin tones under tonal condition. Overall, the isolated tonal condition allowed greater accuracy of identification (96.333% accurate responses) 106

PAGE 107

for listeners than did the coarticulated tonal condition (65.722% accurate responses). A three way repeated measures ANOVA with Tonal condition (isolation and coarticulation) and Tone (Tone 1, Tone 2, Tone 3, Tone 4) as the within-subject factor and Group (inexperienced, experienced) as the between-subject factor, yielded a main effect of Tonal condition [F (1, 16) = 51.185, p< .001], Tone [F (3, 14) = 58.605, p< .001], but not main effect of Grou p [F (1, 16) = 2.899, p= .108]. Furthermore, it was found that the intera ction of Tonal Condition X Tone X Group was marginally different [F (3 ,14) = 3.05, p=0.64] For inex perienced groups of listeners, there was a consistent trend to perceive tones more accurately in the compatible tonal context than in the conflicting tonal cont ext (Tone 1, 96.33% vs. 74.10%; Tone 2, 96.33% vs 67.01%; Tone 3, 89.00% vs 24.69%; Tone 4, 100.00% vs. 45.83%). All ttests indicated that the di fferences between the accurate perception rates of tones which occurred in different tonal c ontexts was significant (Tone 1, [ t (1, 79) =1.906, p =.060]; Tone 2, [ t (1, 79) =2.120, p =.037]; Tone 3, [ t (1, 61) =5.888, p =.000); Tone 4, [ t (1, 79) =3.841, p =.000). Although the experienced group (Figure 4-3 B) showed a tendency to perceive tones more accurately in isolated tonal cond ition than coaticulated tonal cond ition (Tone 1, 100.00% vs. 98.63%; Tone 2, 92. 67% vs 73.61%; Tone 3, 95.33% vs 60.50%; Tone 4, 100.00% vs. 83.81%), only Tone 3s (3s) difference in accuracy rate for the two tonal conditions was significant: Tone 3, [ t (1, 61) =2.763, p =.008) whereas the difference in accuracy rate of Tone 2, [ t (1, 79) =1.652, p =.102]; Tone 1, [ t (1, 79) =.618, p =.539]; Tone 4, [ t (1, 79) =1.481, p =.143) did not reach the significant level. 107

PAGE 108

Effects of tonal sequence Since Tone 3 only can be found in the noni dentical tone sequence, the data of Tone 3 was not analyzed in this section. A) B) Figure 5-4. Mean percent corre ct production of coarticulated Mandarin tones in different tonal sequences by (A) inexperienced learners and (B) experienced learners. Error bars indicate standard error. 108

PAGE 109

Figure 5-4 shows, overall, American speaker s tend to produce tones better in identical tonal sequence (76.53%) than in nonidentical tonal sequence (68.52%). A three way repeated measures ANOVA, with T onal sequence (identical and nonidentical) and Tone (Tone 1, Tone 2, Tone 3, Tone 4) as the within-subj ect factor and Group (inexperienced, experienced) as the between-s ubject factor, yielded significant main effects of Tonal Sequence [F(1, 34) = 4.842, p<.05], Group [F(1, 34 ) = 11.692, p=.002] and Tone [F(2, 33) = 20.181, p<.001]. Post hoc analyses (Bonferroni adjusted p <.05) showed that Tone 1 was produced more accurately than Tone 2 and Tone 4, but that production accuracy of Tone 2 and Tone 4 was not significantly different. Additionally, a three-way interaction between Group, Tonal Sequence and Tone was significant [F(2, 33) = 3.848, p< .05]. Inexperienced l earners production of Tone 1 was found to be significantly more accurate in the identical environment than in the nonidentical sequence [t(1, 70) = 2.737, p<.01]. In contrast there was no significant difference in the production accuracy betwe en the two tonal sequences for Tone 1 among experienced learners [t (1, 70) =1.271, p=.208]. There was no significant difference on accuracy rate of Tone 2 or Tone 4 production between the two tonal sequences for ei ther the inexperienced learners [Tone 2: t(1, 70) = -.943, p=.349; Tone 4: t(1, 70) = 0.270, p=. 788] or the exper ienced learners [Tone 2: t(1, 70) =.610, p=.544; Tone 4: [t(1, 70) =-1.497, p=.139]. Effects of tonal context Figure 5-5 shows that, overall, Amer ican speakers production of Mandarin coarticulated in both compatible tonal environment (66.67%) and conflict tonal environment (62.03%) was equally accurate. A three way repeated measures ANOVA, with Tonal context (compatible context and conflict context) and Tone (Tone 1, Tone 2, 109

PAGE 110

A) B) Figure 5-5. Mean percent correct production of coarticulated Mandarin tones in different tonal contexts by (A) inexperienced learners and (B) experienced learners. Error bars indicate standard error. Tone 3, Tone 4) as the within-subject fact or and Group (inexperienced, experienced) as the between-subject factor, yielded a significant main effect of Tone [F(3, 32) = 50.928, p <.001] and Group [F(1, 32)= 26.263, p <.001]. However, The effect of Tonal Context 110

PAGE 111

[F(1, 34) = 2.635, p =.114] did not reach signif icance. Tone was found to interact marginally significantly with Tonal Context [F(3, 34) =2.686, p =.063]. Follow-up tests showed that the production of Tone 1 was nearly significantly more accurate in compatible context than in conflicting context [t(1, 142) = 1.719, p=.088]. On the contrary, no significant difference was observed for the production of Tone 2 [t(1, 142) = 0.295, p=.768], Tone 3 [t(1, 106) = 0.866, p=.389] and Tone 4 [t(1, 142) = 0.000, p=1]. Further investigation into the effect s of tonal context revealed that, for inexperienced learners, Tone 1 was produced with a higher accuracy rate in the compatible context than in the conflicting context, and this difference is marginally statistically significant [t(1, 70) = 1.852, p=.068]. For experienced learners, there was virtually no difference in accuracy rate in the two tonal contexts [t(1, 70) = .583, p=.562]. Effects of syllable position Figure 5-6 shows that, overall, American speakers produce tones more accurately in final syllable pos ition (68.53%) than in in itial syllable position (62.95%). A three way repeated measures ANOVA, with Syllable position (initial syllable position and final syllable position) and Tone (Tone 1, Tone 2, Tone 3, Tone 4) as the within-subject factor and Group (inexperienced, experienced) as the betweensubject factor, yielded a significant main effect of Syllable position [F(1, 52) = 4.435, p< .04], Group [F(1, 52) = 42.715, p<.001] and Tone [F(3, 50) = 45.033 p<.001]. Tone was found to interact significantly with Syllable position [F(3, 50) = 45.033, p<.001] and with Group [F(3, 50) = 3.636, p<.02]. 111

PAGE 112

A) B) Figure 5-6. Mean percent correct production of coarticulated Mandarin tones in different syllable positions by (A) inexperienced learners and (B) experienced learners. Error bars indicate standard error. Follow-up tests on the significant inte raction between Tone and Syllable Position showed that the production of Tone 2 was significantly more accurate in final syllable position than in initial syllable position [t(1, 142) = -3.870, p<.001 ]. On the contrary, 112

PAGE 113

Tone 4 production was significantly more accurate in initial syllable position than in final syllable position [t(1, 142) = 2.001, p<.05]. Howev er, Tone 1 and Tone 3 productions were equally accurate in both syllable positio ns, [Tone 1, t(1, 142) = .300, p=.765; Tone 3, t(1, 104) = -1.239, p=.218]. Further investigation into the effects of syllable position showed that Tone 2 was produced significantly more accurately in fi nal syllable position by both inexperienced learners [t (1, 70) = -2.49, p=.015] and experienced learners [t(1, 70) = -3.05, p=.001]. Inexperienced learners produced Tone 4 signific antly better in initial syllable position [t(1, 70) = 2.21, p=.030], but syllable position did not af fect experienced learners production accuracy of this t one [t(1, 70) = .84, p=.41]. Tone 3 in tone sandhi When Tone 3 occurs before another Tone 3, it is changed to a rising tone which perceptually sounds like Tone 2. The format of this phonological change is: Tone 3 Tone 2] / __ Tone 3 In the present study, the pr oduction of Tone 3 + Tone 3 by two groups of learners was tested. It was found that no learners in the inexperienced group produced the tonal combination as Tone 2 + Tone 3 and only o ne participant once produced the tonal combination as Tone 3 + Tone 3, and six pa rticipants produced the tonal combination as Tone 2 + Tone 2 with different frequencies. In the group of experienced learners, only one participant once produced the tonal combination Tone 3 + Tone 3 as Tone 2 + Tone 3, and the same participant produced t he tonal combination twice as Tone 3 + Tone 3. The other eight participants produced t he tonal combination as Tone 2 + Tone 2 with various frequencies. 113

PAGE 114

Obviously, initial syllable Tone 3 and final syllable Tone 3 tonal combination Tone 3 + Tone 3 were not acquired by either inex perienced learners or experienced learners. Only one participant in each group produced Tone 3 accurately in the tonal combination. Most participants exaggerated the rising part of Tone 3 on both initial syllable Tone 3 and final syllable Tone 3. T herefore the tone combination Tone 2 + Tone 2 is the most common substituti on tonal sequenc e for the intended tone combination Tone 3 +Tone 3. Error Type Confusion matrices for production of coar ticulated tones in different syllable positions are shown in Tables 5-2 through 5-5. The data was based on 108 responses (12 stimuli 9 participants) for Tone 1, Tone 2 and Tone 4, and 81 responses (9 stimuli 9 participants) for Tone 3. The results of Tone 3 in the tonal combination of Tone 3 +Tone 3 were not included. Table 5-1 Confusion matrix for tones in initial position pr oduced by inexperienced learners, with the per centages in parentheses Tone Produced as Tone 1 Tone 2 Tone 3 Tone 4 Low flat Low falling Tone 1 82 (76.6) 1(9) 0 (0) 14 (13.1) 10 (9.3) 0 Tone 2 14 (13) 60 (55.6) 0 (0) 12 (11.1) 20 (18.5) 2 (1.9) Tone 3 3 (3.7) 49 (60.5) 17 (8.6) 12 (14.8) 0 0 Tone 4 36 (33.3) 2 (1.9) 0 (0) 61 (56.5) 0 9 (8.3) Table 5-2 Confusion matrix for tones in initial position pr oduced by experienced learners, with the per centages in parentheses Tone Produced as Tone 1 Tone 2 Tone 3 Tone 4 Low flat Low falling Tone 1 107 (99.1) 0 (0) 0 (0) 0 (0) 1 (0.9) 0 Tone 2 9 (8.3) 67 (62.1) 0 (0) 3 (2.8) 24 (22.2) 5 (4.6) Tone 3 2 (2.5) 29 (35.8) 45 (55.5) 5 (6.2) 0 (0) 0 (0) Tone 4 7 (6.5) 0 (0) 0 (0) 94 (87) 0 (0) 7 (6.5) 114

PAGE 115

Table 5-3 Confusion matrix for tones in final position pr oduced by inexperienced learners, with the per centages in parentheses Tone Produced as Tone 1 Tone 2 Tone 3 Tone 4 Toneless Tone 1 79 (73.1) 18 (16.7) 5 (4.6) 3 (2.8) 3 (2.8) Tone 2 12 (11.1) 84 (77.8) 9 (8.3) 3 (2.8) 0 (0) Tone 3 2 (2.5) 55 (67.9) 23 (28.4) 1 (1.2) 0 Tone 4 15 (13.9) 21 (19.4) 3 (2.8) 38 (35.2) 31 (28.7) Table 5-4 Confusion matrix for tones in fi nal position produced by experienced learners, with the percentages in parentheses Tone Produced as Tone 1 Tone 2 Tone 3 Tone 4 Toneless Tone 1 106 (98.1) 2 (0.9) 0 (0) 0 (0) 0 (0) Tone 2 5 (4.6) 92 (85.2) 10 (9.3) 0 (0) 0 (0) Tone 3 0 (0) 27 (33.3) 53 (65.4) 1 (1.2) 0 (0) Tone 4 0 (0) 2 (1.9) 0 (0) 86 (79.6) 20 (18.5) As shown in the above tables, three kinds of error types we re found: tonal direction confusion, tonal height confusion, and simult aneous direction and height confusion. The definition of eac h error type is as following: Tonal direction confusion : the direction of tonal cat egory of the target tone is different from the produced tone. Tonal height confusion : the height of tonal category of the target coarticulated tone is the same as the identified t one, but the tonal height is different. Tonal direction plus height confusion : the direction and height of tonal category of the target tone are differ ent from the produced tone. The tonal confusions are divided according to error type in Table 5-5. In the following paragraph, the explanati on of Table 5-5 is provided: 1) Initial syllable T1 is a flat tone, whereas T2 is a rising tone, T3 is a falling-rising tone and T4 is a falling tone. When Initial syllable T1 was produced as T2, T3 or T4, the error was labeled as direction confusion. When initial syllable T1 was produced as a low flat tone, the error was treated as height confusion. 115

PAGE 116

2) Initial syllable Tone 2 is a rising tone, whereas T1 is a fl at tone, and T4 is a falling tone. When initial syllable T2 is produced as T1 or T4, the e rror was labeled as direction confusion. Initial syllable T3 phonetically shows the sa me direction as T2. However, T3 starts lower and ends lower than T2, and turns later t han T2. Therefore, T2 produced as T3 was labeled height confusion. When initia l syllable T2 was produced as a lo w flat tone or a low falling tone, tonal direction was not rising and tonal height was not high, so the e rror was labeled as direction plus height confusion. 3) Initial syllable Tone 3 is a falling-rising t one, whereas T1 is a flat tone, and T4 is a falling tone. When Initial syllable T3 was produc ed as T1, T2 or T4, the error was labled as the error direction confusion. Initial syllable T2 is a rising tone, which has the same direction as initial syllable T3. Therefore, when initial syllable T3 was produced as T2, the error was labled as the error height confusion. 4) Initial syllable Tone 4 is a fa lling tone, whereas T1 is a flat tone, T2 is a rising tone and T3 is a falling-rising tone. When initial syllable T4 was produced as T1, T2 or T3, the error was direction confusion. When initial syllable T4 was produced as a low falling tone, the error was treated as height confusion. 5) When Final syllableT1 was produced as T2 T3 or T4 by American learners, the confusion was categorized as tonal direction confusion. 116

PAGE 117

When final syllable T1 was produced as a tone with lower pitch level, truncated duration of pitch and smaller amplitude of pi tch, the confusion was categorized into tonal direction confusion. 6) Final syllable T2 is a rising tone, whereas T1 is flat and T4 is a falling tone. Therefore, T2 misproduced as T1 or T4 was labeled a direction error. Final syllable T2 is a rising tone, as is final syllable T3. However, T3 rises lower than T2. Therefore, final sylla ble T2 produced as T3 was l abeled a tonal height error. 7) Final syllableT3 rises, whereas T1 and T4 are not rising tones. Therefore, T3 produced as T1 or T4 was labe led a tonal direction error. Final syllable T3 shows a rising direction as T2 does, but final syllable T3 falls first and rises lower than T2. Therefore, T3 produced as T2 was labeled a tonal height error. 8) Final syllable T4 is a falling tone, unlike T1, T2 and T3. Therefore, T4 produced as T1, T2 or T3 was labeled a tonal directional error. When T4 was produced as a tone with lower pitch level, truncated duration of pitch and smaller amplitude of pitch, the confus ion was categorized as tonal direction confusion Figure 5-7 shows that inexperienced l earners tended to produce more direction errors than experienced learner s for final syllable Tone 1 [2=23.635, p=. 000], initial syllable Tone 2 [ 2=7.855, p=.005], syllable final Tone 2 [2=5.510, p=.019], initial syllable Tone 4 [ 2=26.975, p=.000], and final syllable Tone 4 [2=41.213, p=.000]. When producing initial syllable Tone 1, ex perienced learners did not make a single mistake on the tonal direction, whereas in experienced learners still made 15 errors on the flat direction of initial syllable Tone 1. However, the di fference in numbers of errors 117

PAGE 118

made by inexperienced learners and exper ienced lear ners was only marginally significant for initial syllable Tone 3 [2=3.366, p=.066] and was non significant for final syllable Tone 3 [ 2=1.035, p=.311]. Figure 5-8 shows that inexperienced l earners tended to produce more height errors than experienced learner s for initial syllable Tone 1 [ 2=7.759, p=.005], initial syllable Tone 3 [ 2=9.890, p=.002], and final syllable Tone 3 [2=19.361, p=.000]. When producing final syllable Tone 1, experienced learners did not even make a single mistake on the tonal height, whereas inexperienced learners still made 3 errors on final syllable Tone 1. However, the difference between inexperienced learners and experienced learners was only marginally significant for final syllable Tone 4 [ 2=3.747, p=.053] and non significant for syllable final Tone 2 [ 2=.058, p=.810] and initial syllable Tone 4 [ 2=.270, p=.603]. Table 5-5 Tonal confusions categorized into error types. Target tone Tonal direction confusion Tonal height confusion Tonal height and direction confusion 1) Initial high flat T1 T2 T3 T4 Low flat tone 2) Initial rising T2 T1 T4 T3 Low flat tone Low falling tone 3) Initial low falling T3 T1 T4 T2 4) Initial high falling T4 T1 T2 Low falling tone 5) Final flat T1 T2 T3 T4 Toneless 6) Final rising T2 T1 T4 T3 7) Final falling-rising T3 T1 T4 T2 8) Final falling T4 T1 T2, T3 Toneless 118

PAGE 119

Figure 5-7. Frequency of tonal direction confusion by two groups of American learners Figure 5-8. Frequency of tonal height confusi on by two groups of American learners 119

PAGE 120

Figure 5-9. Frequency of tonal direction pl us height confusion by two groups of American learners There was no significant difference in the number of simultaneous height and direction errors by inexperienced and exper ienced learners for initial syllable Tone 2 [ 2=.961, p=.327]. Relationship between Perception and Production of Coarticulated Tones In this section, the relationship between perception and production of coarticulated Mandarin tones is analyzed based on the data from the two whole gr oups and individual learners in those two groups. Figure 5-10. Mean percentage of accuracy of perception and production of coarticulated Mandarin tones by two groups. Error bars indicate standard error. 120

PAGE 121

As shown in Figure 5-10, the accuracy rate of experienced learner s producing Mandarin coarticulated tones is better than the accuracy rate of thei r perception of the tones [t (538)=1.974, p =-2.154], whereas inexperienced learners performed slightly worse at producing the coarticulated tones than at perceiving them [ t (538)= 1.056, p =0.292]. However, neither groups differ ence in accuracy rates of perception and production reached significance. Table 5-6 Individual inexperienced Amer ican learners percept ion and production accuracy rate. These data are across all coarticulated lexical tones in different syllable positions. Participant Perception Production Difference 1 80 70 -10 2 38 38 0 3 42 31 -11 4 53 60 7 5 80 89 9 6 46 56 10 7 67 50 -17 8 53 48 -5 9 68 52 -16 Table 5-7 Individual experi enced American learners pe rception and production accuracy rate. These data are across all coarticulated lexical tones in different syllable positions Participant Perception Production Difference 1 80 83 3 2 84 79 -5 3 71 88 17 4 53 83 30 5 68 60 -8 6 87 88 1 7 80 80 0 8 81 92 11 9 63 73 10 Table 5-6 displays the accuracy rate of perception and production of coarticulated Mandarin lexical tones by inexperienced Am erican learners. According to a Pearson Correlation test, perception and production are not correlated (r=.802, p<.01). 121

PAGE 122

Table 3-6 shows that accuracy rates of perception and production of coarticulated Mandarin tones by experienced American learners are also not significantly correlated (r=.352, p=.352). Discussio n This study was designed to investigate how coarticulated Mandarin Chinese tones in disyllabic words are produced by Amer ican learners with di fferent amounts of language experience (inexperi enced learners vs. experienced learners). The purpose of the production experiment was to test whether American learners pronunciation of Mandarin Chinese coarticula ted tones improves with l anguage learning experience, whether the syllable position and tonal contex t affect identification accuracy, and what strategies were used by American students to modify Mandarin coarticulated tones in disyllabic words. Effects of Learning Experience The results showed that experienced learner s produced coarticulated tones with a higher accuracy rate than inexperienced le arners. Experienced learners performed better not only overall, but also on each indivi dual tone. Therefore, we can conclude that learning experience helps American learners produce Mandarin tones more accurately when the tones occur in a coarticulation environment. This result confirmed the findings from previous studies that experienced learners produced L2 features more accurately than inexperienced learners. Language learning experience also made l earners production of coarticulated tones less subject to contextual effects. In the present study, syllable position affected inexperienced learners but not experienced learners production of coarticulated Tone 1. In terms of effect of linguistic factors, both tonal s equence and tonal context influence 122

PAGE 123

the accuracy of production of Tone 1 by i nexperienced learners, but not experienced learners. Since experienced lear ners have had mo re time to study the target language, their tone production is more native like, wh ich means the experienced learners are less influenced by linguistic factors that do not affect the production of tones by native speakers of Mandarin. Regarding the substitution of tones in produc tion, there is more variation in the substitution of errors made by the in experienced group than by the experienced learners. This phenomenon is concluded fr om the comparison between final syllable tonal errors made by inexperienced learner s and experienced learners. The distribution of tonal errors of the ta rget tone was across other th ree tones by inexperienced learners, whereas only one tone was chosen to substitute the target tone by experienced learners. Effects of Tonal Condition Tonal condition affected how accurately Tone 3 was produced by both experienced and inexperienced American learners. However, the effect of tonal condition on Tone 1, Tone 2 and Tone 4 is only found on inexperienced learners. Effects of Tonal Sequence and Tonal Context There was no effect of tonal sequence (i dentical vs. nonidentical sequence) or tonal context (compatible vs. conflicting c ontext) on Tone 2, Tone 3, or Tone 4 production accuracy by either group of learners. However, there was an effect of tonal sequence and tonal context on the accuracy rate of Tone 1 production by inexperienced learners, but not by experi enced learners. Inexperienced learners produced Tone 1 with a higher accuracy rate in the identical t onal sequence than in the non-identical tonal 123

PAGE 124

sequence. In addition, their production of Tone 1 was more accurate in compatible tonal context than in conflicting tonal context. Two modification strategies have been found to be used by American learners to produce Mandarin coarticulated tones. One is no pitch gap in disyllable and another is no double stressed disyllable. The most frequent combination of modifying tones in conflicting tonal context is Tone 4 +Tone 2. The falling tone and the rising tone are connected by the low F0 offset of the falling tone and the low F0 onset of the rising tone. Therefore, there is no pitch gap within a disyllabic word. This kind of avoidance of pitch break was also found in a previous st udy on production of Vietnamese tones by American speakers (Nguyen & Macken, 2008). This modification strategy might be driven by the American learners L1 background. It is rare to have a dramatic pitch shift from one nuclear tone to another nuclear tone in one sentence (Cruttenden, 1997). Meanwhile, the other two most frequent combinations of modifying tones are low tone +high tone for the tonal combinations Tone 2 + Tone 1 and high tone+ toneless for in the tonal combinations Tone 4 + Tone 4. As with the first type of strategy, the second type of strategy can be also attributed to L1 transfer. Disyllabic English words do not allow both syllables to be stressed. A stress ed syllable is associated with high pitch, whereas an unstressed syllable is associat ed with low pitch. American students transfer this pitch constriction from their L1 to their Mandarin learning and modify a tone with high pitch to a low pitch. This strategy was employed on modifying initial Tone 2 and final Tone 4. Effects of Syllable Position Syllable position affected how accurately Tone 2 was produced by both experienced learners and inexperienced Amer ican learners. Both groups showed a 124

PAGE 125

higher production accuracy rate for Tone 2 in t he final syllable position. This result may be attributed to the rising in tonation used to form questi ons in English. On the other hand, the factor of syllable position only a ffected inexperienced lear ners production of Tone 4. For inexperienced learners, the accura cy rate of Tone 4 production on the initial syllable was higher than that on the final syllabl e. No clear explanation can be offered to explain why Tone 4 was produced more accurate ly in initial syllable position. Syllable position, however, did not affect production accuracy of Tone 1 or Tone 3 production. Development of Production of Coarticulated Tones Three types of errors were found in bot h groups: tonal direction confusion, tonal height confusion, and tonal direction plus height confusion. There was no difference between inexperienced and experienced learners frequency of directional confusion, height confusion, and tonal dire ction plus height confusion e rrors. It was found that the ability to produce tones in terms of both tonal direction and tonal height quickly improved. Relationship between Perception and Production of Coarticulated Tones If we take the groups of learners as a whole, we can say that the ability to perceive coarticulated tones developed in parallel wit h the ability to produce coarticulated tones for both inexperienced and exper ienced American learners. Considering individual performance on perc eption and production of coarticulated tones, there is a strong positive correlation between perception performance and production performance by inexperienced Am erican learners, but not experienced American learners. In ot her words, the proficiency of i nexperienced learners perception is the predictor of the prof iciency of their production of Mandarin coarticulated tones. 125

PAGE 126

Accurate perception leads to accurate production for A merican learners with less learning experience. Conclusion This study revealed that American lear ners with more learning experience were more accurate in producing coarticulated Mandarin tones in disyllabic words. Furthermore, it was found that with increas ed experience, production of certain coarticulated tones becomes more resistant to such phonological and phonetic factors as syllable position and tonal environment. It was also found that tonal sequence and tonal context only affected tone 1 producti on; and syllable position affected tone 2 and tone 4 production accuracy. A strong positiv e correlation exists between perception performance and production performance for inex perienced American learners, but not experienced American learners. 126

PAGE 127

CHA PTER 6 GENERAL DISCUSSION AND CONCLUSION In this chapter, the findings from the st udy are first summarized. Second, general discussion based on the findings is provided. Th ird, I propose future research on the acquisition of Mandarin tones by L2 speakers based on the results and limitations of the present study. Finally, I make some pedagogical suggestions regarding teaching Chinese tones. Summary of Results The experiments in this dissertation set out to test whether amount of learning experience affects adult American learners perception and production of Mandarin tones. One of the goals of the study was to investigat e whether syllable position and tonal environment affect English speakers perception and production of Mandarin tones, and another goal was to study t he relationship between perception and production of Mandarin tones in second language acquisition. The first two experiments, presented in Chapter three, investigated the perception and produc tion of isolated Mandarin tones. The third experiment, pr esented in Chapter four, examined the perception of coarticulated M andarin tones in disyllabic words. The fourth experiment, presented in Chapter five, studied the produc tion of coarticulated Mandarin tones in disyllabic words. The main hypothesis tested across all experiments in this study was that L2 learners with more learning ex perience outperform those with less learning experience when perceiving and producing M andarin tones. Results from the four experiments supported this hypothesis. 127

PAGE 128

Experimen ts 1 and 2 Results from the first perc eption experiment suggest th at experienced American learners perceive isolated Mandarin tones more accurately than inexperienced American learners. However, the producti on experiment showed no significant difference between experienced and inexper ienced learners on producing isolated Mandarin tones. The error types in percept ion and production experiments are the same for both groups: Tone 1 and Tone 4 are frequent ly confused with each other, as are Tone 2 and Tone 3. However, although Amer ican learners ignore the pitch contour differences between these pairs of tones in production, they do tend to produce duration differences to differentiate Tone 1 from Tone 4 and Tone 2 from Tone 3. Generally speaking, inexperienced learners percept ion and production abilit ies developed equally. As their language learning experience increased, experienced learners perception ability became better than their production abil ity. (However, there is no correlation between perception and production of isolat ed Mandarin tones for either group of learners.) Experiment 3 The results from experiment 3, which test ed perception of coarticulated Mandarin tones, showed that experienced learner s outperformed inexperienced ones on perception of coarticulated tones; this holds for the pooled results across all four tones and for each tone individually. Therefore, learning experience does aid American learners ability to perceive coarticulated Mandarin tones. Tonal condition, that is, whether a tone in isolation or in coarticulati on, affected the perception of tones by both groups of learners. Tones in isolation were identified with higher accuracy rate than 128

PAGE 129

tones in coarticulation. Tonal sequence, that is, whether the two syllables of the word had the same tone or two different tones, was found to affect the perception of Tone 1 for inexperienced learners. Tone 1 in identical tonal sequence was identified with higher accuracy rate than it was in nonidentical to nal sequence. Tonal context, that is, whether the pitch of the two tones at their intersecti on was compatible or c onflicting, affected the perception of Tone 1 by both groups of learners. Tone 1 in compatible context was perceived better than it was in conflicting tonal context. In terms of t he effect of syllable position, it was found that Tone 2, Tone 3 and Tone 4 of the final syllable were identified more accurately than those of the first syllable. Two types of perceptual errors were found: tonal direction misperception and tonal height misperception. As l earning experience increased, t he frequency of tonal direction errors decreased, whereas the frequency tonal height errors did not change. Experiment 4 The results from experiment 4, which te sted production of coarticulated Mandarin tones, showed that experienced learners outperformed inexperienced ones, not only overall but also for each of the four tones individually. Tonal condition, affected the perception of tones by both gr oups of learners. Tones in isolation were produced with higher accuracy rate than tones in coarti culation. Tonal sequence and tonal context only affected tone 1 production. Tone 1 was pr oduced better in identical tonal sequence and compatible tonal context than it was in nonidentical tonal s equence and conflicting tonal context. Syllable position affected Tone 2 and Tone 4 production accuracy. The accuracy rate of Tone 2 production on the final syllable was higher than that on the initial syllable, whereas, the accuracy rate of Tone 4 production on the initial syllable 129

PAGE 130

was higher than that on the final syllable. With increased experience, production of coarticulated tone becomes less affected by such phonological and phonetic factors as syllable position and tonal sequence and tonal contexts. Three types of errors were found: tonal di rection confusion, tonal height confusion, and tonal direction plus height confusion. It was found that the ability to produce tones in terms of both tonal direction and tonal height quickly improv ed. A strong positive correlation exists between perception per formance and production performance for inexperienced American l earners, but not experi enced American learners. General Discussion The findings confirmed the hypothesis that longer language learning experience makes American learners perceive Mandarin tones more accurately and partially confirmed the hypothesis that longer language learning experience helps American learners produce Mandarin tones more accu rately. Experienced le arners ability to identify tones is significantly better than inexperienced learners for both isolated and coarticulated tones, whereas experienced le arners ability to produce tones is only significantly better than inex perienced learners for coarticulated tones. In order to identify tones accurately, American speakers could acquire an ability to perceive tones categorically, that is, to exaggerate the phonological differences between four lexical Mandarin tones and ignore the phonet ic variability among tokens of the same tone. This ability to categorically perceive tones keeps developing throughout their language learning experience. On the other hand, American speakers might still detect the phonetic change of tones which is driven by t onal context. However, with more learning experience, experienced learners become more fa miliar with the allophonic variations of tones. 130

PAGE 131

The fact that there was no significant di fference in the production of isolated tones between two groups, despite a difference of nine months in learning experienc e suggested that, in comparison to perception abi lity, production ability may improve at a slower pace. This might be attributed to the absence of L1 negat ive transfer. The English stress pattern prohibits two high to nes from occurring in one disyllabic word. However, this constraint does not keep nat ive English speakers from producing tones on one-syllable words correctly. Therefor e, both inexperienced learners and experienced learners produced tones in isolat ion well. When the tones are coarticulated (in two-syllable words), experienced lear ners show an advantage in production and produced tones in coarticulation with signific antly higher accuracy rate. Here, it is assumed that this L1 negative transfer a ffects inexperienced learners more than experienced learners. The results from the four experiments s uggest that coarticulated tones are harder to identify and produce than isolated ones. In ot her words, even if learners are able to perceive and produce an isolated tone accurate ly, they might not be to perceive or produce it with the same accuracy when the tone is coarticulated with another (in the current study, tones in disyllabic words). Tonal sequence and tonal context affect American learners perception and production of Tone 1. In terms of perception, the narrower pi tch range triggered by the conflicting tonal context prevents accurate identification of coarticulated Tone 1. Moreover, a changed pitch direction triggered by the conflicting tonal context makes Tone 1 harder to identify. In terms of pr oduction, there is no pitch gap in the tonal compatible context, such as Tone 1 + T one 1 and Tone 1 + Tone 4. American learners 131

PAGE 132

perform well when producing Tone 1 in these Co mbinat ions. In contrast, the conflicting tonal combination, such as Tone 4 + Tone 1, leads American students to make more errors on production. The results suggest that American learners prefer producing Tone 1 in a tonal environment without a pitch shift between two tones. Syllable position affects American learner s perception and production of Tone 2 and Tone 4. It was found that the accuracy ra tes of perception and production of Tone 2 were higher on final syllable position than init ial syllable position. This result may be attributed to the rising intonation used to form questions in English. However, the accuracy rate of perception of Tone 4 was hi gher on final syllable position, whereas the accuracy rates of production of Tone 4 was hi gher on initial syllable position. No clear explanation can be offered to explain wh y Tone 4 was produced more accurately but was perceived less accurately in initial syllable position. One interesting finding regarding the produc tion of isolated tones is that American learners use duration rather than pitch to differentiate Tones 1 and 4 and Tones 2 and 3. English speakers in the present study acquired the duration difference among tones earlier than pitch contour. For the perception of coarticulated tones, it seems that tonal he ight is harder to acquire than tonal direction, given that tonal direction errors decreased with increased learning experience, but tonal height errors remained. However, for the production of coarticulated tones, it was found that the ability to produce tones in terms of both tonal direction and tonal hei ght quickly improved. The relationship of perception and pr oduction of isolated tones for American learners is still not clear. Perception and produc tion are not correlated for either group of 132

PAGE 133

learners. Some learners perceive isolated tones better than they produce them, while some do the opposite. Perception of coar ticulated tones predicts production for inexperienc ed but not experienced learners. This finding partially contradicts the Speech Learning Models (SLM) (Flege, 1995) proposal that accurate perception is a must for accurate production. Limitations and Future Research This present study is the first to s ystematically exami ne the perception and production of coarticulated Mandarin tones by American adult learners with different amount of learning experience. Because this is the first such study, acquisition of coarticulated Mandarin tones by L2 speaker s needs to be studied further in the future. This study investigated the effect of language learning experience on the perception and production of Mandarin tones. Language learning experience was defined as the amount of language learning ti me in the classroom. The inexperienced learners had learned Mandarin fo r three months, the experienced learners for twelve months. Therefore, the differ ence in learning duration is not as great as in previous studies (Jun and Cowie, 1994; Bohon and Flege, 1992; Flege et al, 1997) on the effect of language experience on perc eption and production of L2 s ounds. Flege et al. (1997) categorized inexperienced and experienced group based on their length of residence in the US. The former group had lived there for only 0.7 years, the latter group for 27.3 years. In future research, we may need to recruit experienced learners from an advanced level who have longer learning ex perience and may perform more differently than beginning learners. 133

PAGE 134

Many previous studies have determined other factors than language learning experience that may also affect L2 lear ners perception and/or production of target segments and suprasegmental features. An in teresting topic would be to investigate how L2 learners linguistic background affe cts their percepti on and production of Mandarin tones. Due to practical limitat ions, the native language background of all participants is American English, which is a non tonal, stress-based language. In the future, studies should involve participant s with various language backgrounds, such as Japanese speakers whose L1 is a pitch-accent language, Thai speakers whose L1 is a tonal language, and bilingual English-Cantonese speakers who can speak a non tonal stress-based language and a tonal rhythm-based language. Another limitation of the study is that the stimuli are not as natural as real language in conversation. It is difficult to balance the elimination of possible research bias caused by uncontrolled stimuli in t he experiments against stimulation of natural behavior in perception and production of Mandar in tones in a language lab. In the future, we need to find a better way to test the perception and production of coarticulated tones by L2 learners. Here I suggest extending the testing units from word level to sentence level. Participants need to be tested on the ent ire sentence, since sentences are more frequently used in r eal conversation than isolated words. Also, the stimuli used in the production ex periments are alphabet ically presented, na for isolated tones and nana for coarticula ted tones. Because the alphabetic form looks phonologically similar to English wo rds, American learners may be influenced to transfer the stress or intonation pattern from English to their production of these words. 134

PAGE 135

The order to eliminate the alphabetic confusion, we need to change the method in the future and use Chin ese characters or pictures to illustrate the tested words. Pedagogical Implications From the results of the pres ent study, we can tell that, for American learners, the ability to perceive and produce Mandarin to nes in isolation does not guarantee the ability to perceive and produce them in coarticulation. In the current classroom, Chinese instructors focus on introduc ing the phonetic and phonologica l features of Mandarin tones in isolation. More specifically, the pitch direction of tones in isolation is emphasized. The biggest problem with this tendency is t hat it misleads learners into thinking Tone 3 is always a falling-rising tone and putting more emphasis on the rising part. Therefore, initial syllable T one 3 is frequently produced as Tone 2. Actually, Tone 3 is changed into a low falling tone when it occurs in initial syllable position. I suggest that Chinese instructors more em phasize the low pitch of Tone 3 to American students and less emphasize the rising part of Tone 3 when they perceive and produce it. I have three arguments to support this idea. First, T one 3 is the only low tone in the Mandarin tonal inventory. Therefore, low pitch is the defining feature of Tone 3. Second, the contour of falling rising does not appear in English at the syllable level and, American students will expect it to be difficult to master the pitch track. However, low pitch does exist in English. Therefore, simplifying the pitch contour will help American students deal with the exotic tone. Third, due to the categorical perception of Mandarin tone, I observed that native speakers of mandarin categor ize all the low tones as Tone 3 when the low tone occurs in the initial syllable of disyllabic words. Therefore, American 135

PAGE 136

learners will not be misunderstood by native speakers if they simply produce initial syllable Tone 3 as a low tone. The second suggestion for teaching coar ticulated Mandarin tones to American students is to emphasize the pitch height as much as pitch direction. This is essential to call Americ an students attention to the im portance of producing two high tones in disyllabic words, although this tonal pattern conflicts with the pitch pattern in English disyllabic words. The third suggestion is to point out ther e is a pitch gap between the two tones in words with the conflicting tonal context, alt hough such a gap does not exist in the pitch contour of English disyllabic words. In the present study, it was found that American students tend to change the tonal direction or tonal height in order to fill the pitch gap. Chinese instructors need to realiz e that the errors of tonal direction and tonal height do not stem from their ability to produce the tone in isolation. The errors are triggered by the tonal context. Therefore, Chinese instruct ors must force their students to notice the pitch gap by providing more practice on the tonal combinations with pitch gap. 136

PAGE 137

LIST OF REFERENCES Best, C. T. (1995). A direct realist view of cross-language speech. In W. Strange (Ed.), Speech per ception and linguistic experienc e: issues in cross-language research (pp. 171). Timonium, MD: York Press. Best, C. T., and Tyler, M. D. (2007). Nonnative and second-language speech perception: Commonalitie s and complementarities In M. J. Munro & O.-S. Bohn (Eds.), Second language speech learning: The role of language ex perience in speech perception and production (pp. 13-34). Amster dam: John Benjamins. Blicher, D.L., Diehl, R.L., and Cohen, L.B. (1990). Effects of syllable duration on the perception of the Mandar in tone2/tone3 distinction: Evidence of auditory enhancement. Journal of Phonetics. 18, 37-49. Bluhme, H. and Burr, R. (1971). An audio-visual display of pitch for teaching Chinese tone. Studies in Linguistics, 22, 51-57. Borden, G., Gerber, A & Milsark, G. (1983). Production and Perception of the /r/-/l/ Contrast in Korean Adults Learning English. Language Learning, 33(3), 499-526. Bohn, O.-S., and Flege, J. E. (1997). Pe rception and production of a new vowel category by adult second language learners. In J. Leather, & A. James (Eds.), Second language speech: structure and pr ocess (pp. 53). Berlin: Mouton de Gruyter. Bradlow, A. R., Pisoni, D. B ., Akahane-Yamada, R ., and Tohkura, Y (1997) Training Japanese Listeners to Identify Englis h /r/ and /l/: Some Effects of Perceptual Learning on Speech Production The Journal of the Acous tical Society of America, 101 2299-2310 Broselow, E., Hurtig, R.R., and Ringen, C. (1987) The perception of second language prosody. In G. Ioup and S.H. Weinberger (eds.), Inter-language Phonology, The Acquisition of Second Language Sound System. Cambridge: Newbury House Publishers. pp. 350-361. Chao, Y. (1948). Mandarin Primer. Cambridge, MA: Harvard University Press. Chen, G.T. (1974). The pitch range of English and Chinese speakers. Journal of Chinese Linguistics, 2, 159-171. Cruttenden A (1997). Intonation. Cambridge University Press. Flege, J. E. and Eefting, W.(1987) "Cro ss-language switching in stop consonant perception and production by Du tch speakers of English", Speech Communication. 6, 185-202. 137

PAGE 138

Flege, J. E. (1993) "Production and percept ion of a novel, second-language phonetic contrast", Journal of the Acoustic a l Society of America 93 1589-1608. Flege, J. (1984). The effect of ex perience on Arabs' per ception of the English/s/vs./z/contrast, Folia Linguistica 18, 117-138. Flege, J. E. (1995). Second-Language Sp eech Learning: Theory, Findings and Problems. in Speech Perception and Linguis tic Experience: Issues in CrossLanguage Research edited by W. Strange ( Timonium, MD). pp. 233. Flege, J.E. (1997). The role of phonetic category formati on in second-language speech learning. In J. Leather and A. James (eds.): New Sounds 97. Proceedings of the Third International Symposium on t he Acquisition of Second-Language Speech Amsterdam: University of Amsterdam, pp. 79-88. Flege, J. E. Bohn, O, and Jang, S (1997) Effects of Experience on Non-Native Speakers' Production and Perception of English Vowels. Journal of Phonetics 25, 437-470 Flege, J., MacKay, I. and Meado r, D. (1999) Native Italian speaker s production and perception of English vowels. The Journal of the Acoustical Society of America 106, 2973-987. Francis, A. L., V. Ciocca, L. M, and Fenn, K.. 2008. Perceptual learning of Cantonese lexical tones by tone and non-tone language speakers. Journal of Phonetics 36, 268-294. Fu, Q. J., Zeng, F. G., Shann on, R. V., and Soli, S. D. (1998). Importance of tonal envelope cues in Chinese speech recognition. The Journal of the Acoustical Society of America 104, 505. Fu, Q., and Zeng, F. (2000). I dentification of temporal envel ope cues in Chinese tone recognition. Asia Pacific Journal of Speech, Language and Hearing 5 45-57. Gandour, J.T. (1983). Tone perception in Far Eastern languages. Journal of Phonetics, 11, 149-175. Gandour, J.T. (1984). Tone dissimilarity judgments by Chinese listeners. Journal of Chinese Linguistics, 12, 235-261. Garca Prez, G. M. (2005). Perception of English vowels by native speakers of Spanish in a regular classroom setting. Revista Virtual de Estudos da Linguagem, 3, 5 Grding, E., Kratochvil, P., Svantesson, J.O., and Zhang, J. (1986). Tone 4 and Tone 3 discrimination in mode rn standard Chinese. Language and Speech, 29, 281-293. 138

PAGE 139

Goto, H. (1971) Auditory perception by normal Japanes e adults of the sounds "l" and "r", Neuropsychologia, 9, 317-323. Gottfried, T. L., and Suiter T. L. (1997). Effect of linguistic experience on the identification of Mandarin Chinese vowels and tones. Journal of Phonetics, 25, 207231. Gussenhoven, C. (2004), The phonology of tone and intonation. Cambridge, UK; New York : Cambridge University Press. Halle P. A., Chang, Y. C., and Best, C. T. ( 2004). Identification and discrimination of Mandarin Chinese tones by Mandarin Chinese vs.French listeners. Journal of Phonetics, 32, 395. He. Y (2004). The production of English /L/ by Mandarin speakers. MA Thesis: University of Victoria. He, Y., Wang, Q, and Wiltshire, C. (2008). Production of English lexical stress by inexperienced and experienced learners of English. Canadian Acoustics 36, 128129. Hewings, M. (1995). Tone choice in the E nglish intonation of non-native speakers. IRAL 36, 251-265. Ingram, J. and Park, S.-G. (1997). Cross-language vowel per ception and production by Japanese and Korean learners of English. Journal of Phonetics 25, 343 370. Jun, S. A.;and Cowie, I. (1994). Interference for new and similar vowels in Korean speakers of English. Ohio State Univ. Working Papers 43: 117. Li, C. N. and Thompson, S. A. (1977). T he acquisition of tone in Mandarin speaking children. Journal of Child Language, 4: 185-199. Liu, S.; Samuel, A. (2004). Perception of Mandarin Lexical Tones When F0 Information Is Neutralized. Language and Speech 47, 2, 109-138. Leather, J. (1990). Perceptual an d productive learning of Chi nese lexical tone by Dutch and English speakers. In J. Leather and A. James (eds.): New Sounds 90: Proceedings of the Am sterdam Symposium on the Ac quisition of Second Language Speech Amsterdam: University of Amsterdam, pp.305-341. Lee, Y., Vakoch, D. and Wurm, L. (1996). Tone perception in Cantonese and Mandarin: A cross-linguistic comparison. Journal of Psycholinguistic Research 25(5), 527544. 139

PAGE 140

Liu, F. (1924). Szu Sheng Shih Yen Lu Chun Yi Shanghai. Liu, S., and Samuel, A. G. (2004). Percept ion of Mandari n lexical tones when F0 information is neutralized. Language Speech 47, 109. Logan, J. S., Lively, S. E., and Pisoni, D. B. (1991). Training Japanese listeners to identify English /r/ and /l/: A first report. Journal of theAcoustical Society of America, 89, 874-886. Howie, J.M. (1976). Acoustical Studies of Mandarin Vowels and Tones Cambridge University Press, Cambridge. MacKain, K. (1981) Categorical perception of English /r/ and /l/ by Japanese bilinguals. Applied Psycholinguistics 2, 369-390. Massaro, D. W., Cohen, M.M., and Tseng, C. (1985). The ev aluation and integration of pitch height and pitch contour in lexica l tone perception in Mandarin Chinese. Journal of Chinese Linguistics, 13, 267-290. McAllister, R., Flege, J. E. and Piske, T. (2002) The influence of L1 on the acquisition of Swedish quantity by native speakers of Spanish, English and Estonian. Journal of Phonetics, 30, 229-258. Miracle, W. C. (1989). Tone production of American students of Chinese: A preliminary acoustic study. Journal of Chinese Language Teachers Association 24: 49-65. Moore, C. B., and Jongman, A. (1997). Speaker normalizat ion in the perception of Mandarin Chinese tones. Journal of the Acoustica l Society of America, 102, 18641877. Rochet, B. (1995) Perception and producti on of second-language speech sounds by adults. In Speech perception and linguistic experience (W. Strange, editor). Timonium, MD: York Press. Schmidt, A., & Meyer, K. (1995). Traditiona l and phonological treatm ent for teaching English fricatives and affricates to Koreans. Journal of Speech and Hearing Research, 38,828-838. Sheldon, A., and Strange, W. (1982). The acquisition of 11-1 and 111 by Japanese learners of English: ev idence that speech production can precede speech perception. Applied Psycholinguistics, 3,243-261. Shen, S. X-N. (1989). Towa rd a register approach in teaching Mandarin tones. Journal of Chinese Language T eachers Association, 24(3): 27-47. 140

PAGE 141

Shen, S. X-N. (1990). Tonal coarticulation in Mandarin. Journal of Phonetics, 18: 281295. Shen, X.S. and Lin, M.C. (1991). A perceptual study of Mandarin tones 2 and 3. language and Speech, 34, 145-156. Shih C (1986). The phonetics of the Chinese tonal system.. Technical memorandum. AT & T Bell Laboratories. Trofimovich, P., and Baker, W. (2006). Learning second-language suprasegmentals: Effect of L2 experience on prosody and fluency characteristics of L2 speech. Studies in Second Language Acquisition, 28, 1-30. Xing, J (2006). Teaching and Learning Chinese as a Foreign Language. A Pedagogical Grammar. Hong Kong: Hong Kong University Press. Wang, H., Chang, B., Li, Y., Lin, L., Liu, J., Sun, Y., Wang. Z., Yu, Y., Zhang, J., and Li, D. (1986). Xiandai Hanyu Cipin Cidian [Frequency Dicti onary of Modern Chinese] Beijing, China: Beiji ng Language Institute Press. Wang, Y., Spence, M.M., Jongman, A., and Sereno, J.A. (1999). Training American listeners to perceive Mandarin tones. Journal of the Acoustical Society of America 106, 3649-3658. Wang, Y. (1995). Error Analysis of Tone production by American Speakers and Pedagogical strategies in Teaching Tones. Phonetics research and Teaching Chinese as a Second Language, 379-394. Wang, Y., Sereno, J., Jongman, A., and Hirsch, J. (2003). fMRI evidence for cortical modification during learning of Mandarin lexical tone. Journal of Cognitive Neuroscience, 15, 7, 1019-1027. Wang, Y., Jongman, A., and Sereno, J.A. (2003). Acoustic and perceptual evaluatio n of Mandarin tone productions befor e and after perceptual training. Journal of the Acoustical Society of America, 113, 1033-1044 Wang, Y. and Kuhl, P.K. (2003) Evaluating the critical per iod hypothesis: Perceptual learning of Mandarin tones in American adults and American children at 6, 10, and 14 years of age. Proceedings of the 15th Internat ional Congress of Phonetic Sciences, 1537-1540. Wang, Y., Behne, D., Jongman, A., and Sereno, J.A. (2004). The role of linguistic experience in the hemispheric processing of lexical tone. Journal of Applied Psycholinguistics 25, 449-466. 141

PAGE 142

Wayland, R. Guion, S. (2003). Percept ual Discriminat ion of Thai Tones by Naive and Experienced Learners of Thai. Applied Psycholinguistics, 24, 1, 113-129. Wayland, R. Guion, S. (2004). Traini ng English and Chinese Listeners to Perceive Thai Tones: A Preliminary Report. Language Learning, 54, 4, 681-712. Whalen, D. H. and Xu, Y., (1992). Information for Mandarin tones in the amplitude contour and in brief segments. Phonetica, 49, 25-47. White, C.M. (1981). Tonal per ception errors and interferenc e from English intonation. Journal of Chinese Language Teachers Association, 16, 27-56. Wiltshire, C and Moon, R. (2003). Phonetic St ress in Indian English vs. American English. World Englishes 22, 291-303. Wu, Z.J. (1986). The Spectrographic Album of Mono-syllables of Standard Chinese Beijing: Social Science Press. Xu, Y. (2004). Understanding t one from the perspective of production and perception. Language and Linguistics, 5, 757-797. Xu, Y. (1997). Contextual t onal variations in Mandarin. Journal of Phonetics 25, 61-83. Xu, Y. (1994). Production and percept ion of coarticulated tones, Journal of the acoustical Society of America, 95, 2240 2253. Xu, Y. (1993). Contextual tonal variation in Mandarin Chinese. Ph.D. dissertation. The University of Connecticut. 142

PAGE 143

143 BIOGRAPHICAL SKETCH Yunjuan He was born in Changchun, China. She received her B.A. in Teaching Chinese as Second Language from Beijing Language and Culture University in 2001. She earned her M.A. in Linguistics from Univ ersity of Victoria in 2004. She joined the University of Florida as a Ph.D. student in fall 2005