<%BANNER%>

Effects of Production Training and Perception Training on Lexical Tone Perception

MISSING IMAGE

Material Information

Title:
Effects of Production Training and Perception Training on Lexical Tone Perception Are the Effects Domain General or Domain Specific?
Physical Description:
1 online resource (144 p.)
Language:
english
Creator:
Lu, Shuang
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Linguistics
Committee Chair:
Kaan, Edith
Committee Co-Chair:
Wayland, Ratree
Committee Members:
Cowles, Heidi Wind
Smith, Brenda Jo

Subjects

Subjects / Keywords:
music -- perception -- production -- tone -- training
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:
The relationship between speech perception and production has been debated for a long time. The Motor Theory of speech perception (Liberman et al., 1989) claims that perceiving speech is identifying the intended articulatory gestures rather than perceiving the sound patterns. It seems to suggest that speech production precedes speech perception, i.e., only when people can accurately ‘feel’ their vocal tract gestures, can they successfully perceive speech. One prediction that falls out from the Motor Theory is that training people to produce sounds should be more effective for improving the perceptual ability than training people to only perceive sounds. My dissertation aims to test this prediction by comparing the effects of laboratory perception training and production training as well as the effects of different musical backgrounds (i.e., vocalists and instrumentalists) on the perception of lexical tones. The present dissertation encompasses three experiments. The first experiment compared native English non-musicians, instrumentalists and vocalists on the perception and production of Mandarin tones. The results supported previous findings, demonstrating the advantage of musical training on lexical tone perception. However, no additional benefit was found for musicians with vocal experience in addition to instrumental experience on either lexical tone perception or production. The second experiment recorded both behavioral and electrophysiological data to examine the effectiveness of a perception-only training and a perception-plus-production training on lexical tone perception. The data showed that after training participants in both groups improved on tone discrimination. Moreover, the participants in both groups did not differ in tone processing at the intentional or unintentional level after training. The third experiment, using identification training rather than the discrimination training, demonstrated that the identification-plus-imitation group identified the tones more quickly in the post-training task than in the pre-training task while the identification-only group did not show any improvement in terms of reaction time. These results suggested that the employment of the motor system does not specifically benefit the tone perceptual skills.
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 Shuang Lu.
Thesis:
Thesis (Ph.D.)--University of Florida, 2013.
Local:
Adviser: Kaan, Edith.
Local:
Co-adviser: Wayland, Ratree.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2013
System ID:
UFE0045917:00001

MISSING IMAGE

Material Information

Title:
Effects of Production Training and Perception Training on Lexical Tone Perception Are the Effects Domain General or Domain Specific?
Physical Description:
1 online resource (144 p.)
Language:
english
Creator:
Lu, Shuang
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Linguistics
Committee Chair:
Kaan, Edith
Committee Co-Chair:
Wayland, Ratree
Committee Members:
Cowles, Heidi Wind
Smith, Brenda Jo

Subjects

Subjects / Keywords:
music -- perception -- production -- tone -- training
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:
The relationship between speech perception and production has been debated for a long time. The Motor Theory of speech perception (Liberman et al., 1989) claims that perceiving speech is identifying the intended articulatory gestures rather than perceiving the sound patterns. It seems to suggest that speech production precedes speech perception, i.e., only when people can accurately ‘feel’ their vocal tract gestures, can they successfully perceive speech. One prediction that falls out from the Motor Theory is that training people to produce sounds should be more effective for improving the perceptual ability than training people to only perceive sounds. My dissertation aims to test this prediction by comparing the effects of laboratory perception training and production training as well as the effects of different musical backgrounds (i.e., vocalists and instrumentalists) on the perception of lexical tones. The present dissertation encompasses three experiments. The first experiment compared native English non-musicians, instrumentalists and vocalists on the perception and production of Mandarin tones. The results supported previous findings, demonstrating the advantage of musical training on lexical tone perception. However, no additional benefit was found for musicians with vocal experience in addition to instrumental experience on either lexical tone perception or production. The second experiment recorded both behavioral and electrophysiological data to examine the effectiveness of a perception-only training and a perception-plus-production training on lexical tone perception. The data showed that after training participants in both groups improved on tone discrimination. Moreover, the participants in both groups did not differ in tone processing at the intentional or unintentional level after training. The third experiment, using identification training rather than the discrimination training, demonstrated that the identification-plus-imitation group identified the tones more quickly in the post-training task than in the pre-training task while the identification-only group did not show any improvement in terms of reaction time. These results suggested that the employment of the motor system does not specifically benefit the tone perceptual skills.
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 Shuang Lu.
Thesis:
Thesis (Ph.D.)--University of Florida, 2013.
Local:
Adviser: Kaan, Edith.
Local:
Co-adviser: Wayland, Ratree.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2013
System ID:
UFE0045917:00001


This item has the following downloads:


Full Text

PAGE 1

1 EFFECTS OF PRODUCTION TRAINING AND PERCEPTION TRAINING ON LEXICAL TONE PERCEPTION ARE THE EFFECTS DOMAIN GENERAL OR DOMAIN SPECIFIC? By SHUANG LU A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORI DA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013

PAGE 2

2 2013 Shuang Lu

PAGE 3

3 To my family and friends

PAGE 4

4 ACKNOWLEDGMENTS The journey of pursuing my Ph D in the United States is truly unforgettable. I cannot imagine having accomplished this journey without the great support and help from the people listed below. First of all, I would like to express my deepest gratitude to my advisor Dr. Edith Kaan, who taught me how to do research, how to write, how to apply for grants and everything I know about academia. I am thankful to Dr. Kaan for all her patience intelligence, caring and guidance during the past few years Working with her has been my greatest pl easure Likewise, I am thankful to the co chair of my committee Dr. insights have enriched the analyses of my dissertation and have always encouraged me to think beyond the obvious. Dr. Wind Cowles and Dr. Brenda Smith are the other two members of my committee, and both have provided invaluable support and help for me. Dr. Cowles introduced me into the fascinating world of psycholinguistics. It has been a great pleasure to work as her teaching assistant for the last four years. Many thanks are directed to Dr. Smith for her insightful suggestions on my dissertation and for helping me recruit musicians for my dissertation research. Besides, I would like to thank all of my pro fessors in the Department of Linguistics at the University of Florida for teaching me all the wonderful classes and providing me great help and advice Special thanks go to Dr. Paula Golombek, Dr. Caroline Wiltshire, Dr. Jason Rothman and Dr. Brent Henders on. Appreciation is also given to our office manager Kelli Granade for taking care of the grant money for my dissertation research and for all her helps throughout the past few years

PAGE 5

5 I would also like to thank all my fellow graduate students in the Ling uistic Department and all my wonderful lab mates in the Brain and Language Laboratory for making graduate school an incredible memorable experience for me. I am thankful for all the research assistants who helped me with the preparation and conduction of m y dissertation research, including Joe Kirkham, Eric Holgate, Carlie Overfelt and Kayla Johnson. I am also thankful to all the participants who participated in my experiments. My family and friends have always been the strongest support ers for me. I am ve ry grateful for all my friends in China and in the U S who always bring me true happiness and strength. I would also like to express my greatest appreciation to my parents PP and my other family members. This journey would not have been possible without their love, support and encouragement.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 10 LIST OF ABBREVIATIONS ................................ ................................ ........................... 13 ABSTRACT ................................ ................................ ................................ ................... 15 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ..... 17 Project Motivation ................................ ................................ ................................ ... 17 Background ................................ ................................ ................................ ............. 18 Views of the Relation between Production and Perception .............................. 18 Previous Training Studies ................................ ................................ ................. 20 Perception training ................................ ................................ ..................... 20 Production training ................................ ................................ ..................... 21 Comparison between perception training and production training ............. 23 Tones in Music and Language ................................ ................................ ......... 25 Tone Perception and Production by Non native Speakers ............................... 26 Effect of Music Background on Lexical Tone Perception ................................ 27 Effects of Laboratory Training on Lexical Tone Perception .............................. 29 Perception training ................................ ................................ ..................... 29 Production training ................................ ................................ ..................... 30 Interim Summary ................................ ................................ .............................. 30 Event related Potentials (ERPs) ................................ ................................ ....... 31 ERP Components that are Re levant to Auditory Perception ............................ 32 The Current Study ................................ ................................ ................................ .. 33 Research Questions ................................ ................................ ......................... 33 Structure of the Dissertation ................................ ................................ ............. 33 Intellectual Merit ................................ ................................ ............................... 36 Broader Implications ................................ ................................ ......................... 36 2 EXPERIMENT 1: EFFECTS OF LONG TERM MUSICAL PERCEPTION AND PRODUCTION TRAINING ON LEXICAL TONE PERCEPTION AND PRODUCTION ................................ ................................ ................................ ........ 38 Methods ................................ ................................ ................................ .................. 38 Participants ................................ ................................ ................................ ....... 38 Stimuli ................................ ................................ ................................ ............... 40 Procedure ................................ ................................ ................................ ......... 41

PAGE 7

7 Data Analysis ................................ ................................ ................................ ... 45 Discrimination task ................................ ................................ ..................... 45 Imitation task ................................ ................................ .............................. 46 Predictions ................................ ................................ ................................ ........ 47 Results ................................ ................................ ................................ .................... 47 Discrimination Task ................................ ................................ .......................... 47 Accurac y ................................ ................................ ................................ .... 47 Reaction times ................................ ................................ ........................... 48 Accuracy percentage for each tone pair ................................ .................... 48 Imitation Task ................................ ................................ ................................ ... 50 Summary ................................ ................................ ................................ ................ 51 3 EXPERIMENT 2: EFFECTS OF LABORATORY PERCEPION AND PRODUCTION TRAINING ON TONE DISCRIMINATION A BEHAVIORAL AND ERP STUDY ................................ ................................ ................................ ... 57 Methods ................................ ................................ ................................ .................. 57 Participants ................................ ................................ ................................ ....... 57 Stimu li ................................ ................................ ................................ ............... 58 Behavioral task ................................ ................................ .......................... 58 ERP ................................ ................................ ................................ ........... 60 Procedure ................................ ................................ ................................ ......... 61 Day 1 ................................ ................................ ................................ ......... 62 Day 2 ................................ ................................ ................................ ......... 64 Day 3 ................................ ................................ ................................ ......... 65 Data Analysis ................................ ................................ ................................ ... 65 Behavioral task ................................ ................................ .......................... 65 ERP ................................ ................................ ................................ ........... 66 Predictions ................................ ................................ ................................ ........ 68 Results ................................ ................................ ................................ .................... 69 Behavioral Task ................................ ................................ ................................ 69 Accuracy ................................ ................................ ................................ .... 69 Reaction times ................................ ................................ ........................... 72 Percentage accuracies of different tone pairs ................................ ............ 74 Summary of behavi oral results ................................ ................................ ... 75 ERP ................................ ................................ ................................ .................. 76 100 200ms window ................................ ................................ ................. 77 MMN ................................ ................................ ................................ .......... 78 Late negativity ................................ ................................ ............................ 79 Summary of ERP results ................................ ................................ ............ 80 4 EXPERIMENT 3: COMPARISON OF PERCEPTION TRAINING AND PRODUCTION TRAINING ON TONE IDENTIFICATION ................................ ....... 97 Methods ................................ ................................ ................................ .................. 97 Participants ................................ ................................ ................................ ....... 97 Stimuli ................................ ................................ ................................ ............... 98

PAGE 8

8 Procedure ................................ ................................ ................................ ......... 9 9 Data Analysis ................................ ................................ ................................ 101 Results ................................ ................................ ................................ .................. 103 Accuracy ................................ ................................ ................................ ......... 103 Reaction Times ................................ ................................ .............................. 105 Percen tages of Correct Identification for Each Tone ................................ ...... 107 Summary ................................ ................................ ................................ .............. 109 5 CONCLUSIONS ................................ ................................ ................................ .... 117 Summary of Results ................................ ................................ .............................. 118 Discussion ................................ ................................ ................................ ............ 120 Effects of Vocal Experience and Instrumental Experience on Lexical Tone Perception and Production ................................ ................................ .......... 120 Effects of Laboratory Perception training and Production Training on Tone Perception ................................ ................................ ................................ ... 124 Discrimi nation training paradigm ................................ .............................. 124 Identification training paradigm ................................ ................................ 128 Final remarks ................................ ................................ ........................... 129 APPENDIX A MUSICAL BACKGROUNDS FOR THE ENGLISH MUSICIANS IN EXPERMENT1 ................................ ................................ ................................ ..... 131 B LANGUAGE AND EDUCATION QUESTIONS ................................ ..................... 132 C PERSONAL HISTORY FORM FOR LANGAUGE EXPERIMENTS WITH EVENT RELATED POTENTIALS OR FMRI ................................ ......................... 134 D HANDEDNESS QUESTIONNAIRE ................................ ................................ ...... 136 LIST OF REFERENCES ................................ ................................ ............................. 137 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 144

PAGE 9

9 LIST OF TABLES Table page 2 1. Mean accuracy percentages of the tonal test, the rhythmic test and the composite performance in AMMA for the Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians ............... 53 2 2. by all 3 judges and 2 judges respectively for each tone type and group ............. 53 3 1. Mean accuracy percentages of the tonal test, the rhythmic test and the composite performance in AMMA for the perception only and perception plus production groups ................................ ................................ ....................... 82 3 2. The numb ers and the percentages of trials that were included in data analysis after artifact rejection in the pre and post training tasks for the perception only and perception plus production groups respectively ................................ ... 82 4 1. Mean accuracy percentages of the tonal test, the rhythmic test and the composite performance in AMMA for the identification only and identification plus imitation groups ................................ ................................ ........................ 111 A 1. Musical backgrounds for the American English musicians in Experiment1 ...... 131

PAGE 10

10 LIST OF FIGURES Figure page 2 1. Waveforms and spectro grams with the pitch contours of the Mandarin Tone1, Tone2, Tone3 and Tone4 as sociated with the syllable [ma] ............................... 53 2 2. D Prime Scores with the standard errors for Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians respectively ................................ ................................ ................................ ......... 54 2 3. Mean reaction times with the standard errors of all the correct responses for the Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians respectively ............................. 54 2 4. The mean accuracy percentages with the standard errors for the same pai rs in the Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians respectively ............................. 55 2 5. Mean accuracy percentages with the standard errors fo r each of the different pairs in the Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians ................................ ................. 55 2 6. Mean percentages and the standard error s of the correctly identified productions for each tone in the Chinese non musicians, the English vocalists, instrumentalists and non musicians respectively ................................ 56 2 7. Distribution of the part of T3 ................................ .. 56 3 1. Spectrograms of Tone1, Tone2 and Tone4 for the syllable [t h u] ......................... 82 3 2. Step i llustration of the perception only training and the perception plus production training ................................ ................................ .............................. 83 3 3. Mean d prime scores with the standard errors of the pre training task, training session and t he post training task for the perception only and the perception plus production group respectively ................................ ................................ ..... 83 3 4. Mean d prime scores with the standard errors for the trained and untrained stimul i in the pre and post training tasks for the perception only and the perception plus production group respectively ................................ ................... 84 3 5. Mean d prime scores with the standard errors for the tested and u ntested stimuli in the training session for the perception only and the perception plus production group respectively ................................ ................................ ............. 84

PAGE 11

11 3 6. Mean reaction times with the standard errors of all the correct responses in the pre training task, the training session and the post training task for the perception only and production plus perception groups respectively ................. 85 3 7. Mean reaction times w ith the standard errors of the trained and untrained stimuli in the pre and post training tasks for both groups respectively .............. 85 3 8. Mean reaction times with the standard errors of the tes ted and untested stimuli in the training session for the perception only and the perception plus production groups ................................ ................................ ............................... 86 3 9. Mean percentage accuracies with standard errors of each tone pair in the pre and post training tasks for the perception only and perception plus production groups ................................ ................................ ............................... 86 3 10. Mean percentage accuracies with standard errors of each tone pair in the training se ssion for the perception only and the perception plus production groups ................................ ................................ ................................ ................ 87 3 11. The ERPs of the frontal electrodes (F3, Fz and F4), the central electrodes (C3, Cz and C4) and the parietal el ectrodes (P3, Pz and P4) co llapsed over all six conditions ................................ ................................ ................................ 88 3 12. The ERPs of the Fz electrode in the pre and post training tasks for the perception only and perception plus production gro ups respectively ................. 92 3 13. Deviant minus standard difference waves of the Fz electrode for the six conditions in the pre and post training tasks for the two groups ........................ 95 4 1. Step illustration of the identification only training and the identification plus imitation training ................................ ................................ ............................... 111 4 2. Mean d prime scores with the st andard errors for the identification only and the identification plus imitation groups in the pre training task, the training session and the post training task respectively ................................ ................ 112 4 3. Mean d prime scores with the standard errors for the trained and untrained stimuli in the pre and post training tasks for the identification only and the identification plus imitation group respectively ................................ ................. 112 4 4. Mean d prime scores with the standard errors of the tested and untested stimuli in the training session for the identification only and the identification plus imitation groups respectively ................................ ................................ ..... 113 4 5. Mean reaction times with the standard errors of all the correct responses in the pre training task, the training session and the post training for the identification only and the identification plus imitation group ............................ 113

PAGE 12

12 4 6. Mean reaction times with the standard errors of the trained and untrained stimuli in the pre and post training tasks for both groups respectively ............ 114 4 7. Mean response times with the standard errors of the tested and untested stimuli in the training session for the identification only and identification plus imitation groups respectively ................................ ................................ ............ 114 4 8. Mean percentages of correct identification with the standard errors for each tone type in the pre and post training tasks for the identification only and the identification plus imitation groups ................................ ................................ ... 115 4 9. Mean percentages of correct identification with standard errors for each tone type in the training session for the identification only and the identification plus imitation groups respectively ................................ ................................ ..... 115 4 10. Error distribution for the identification only and the ident ification plus imitation groups ................................ ................................ ................................ .............. 116

PAGE 13

13 LIST OF ABBREVIATIONS EEG Electroencephalogram ERP Event related potentials F0 F undamental frequency MMN Mismatch N egativity P AIR 11 Tone pair Tone1 Tone1 P AIR 22 Tone pair Tone2 Tone2 P AIR 33 Tone pair Tone3 Tone3 P AIR 44 Tone pair Tone4 Tone4 P AIR 12 Tone pair s Tone1 Tone2 &Tone2 Tone1 P AIR 13 Tone pair s Tone1 Tone3 & Tone3 Tone1 P AIR 14 Tone pair s Tone1 Tone4 & Tone4 Tone1 P AIR 23 Tone pair s Tone2 Tone3 & Tone3 Tone2 P AIR 24 Tone pair s Tone2 Tone4 & Tone4 Tone2 P AIR 34 Tone pair s Tone3 Tone4 & Tone4 Tone3 SD Standard deviation T1 Tone1 T2 Tone2 T3 Tone3 T4 Tone4 T12 One condition in the ERP data: deviant T1 in T2 standards T14 One condition in the ERP data: deviant T1 i n T4 standards T21 One condition in the ERP data: deviant T2 in T1 standards T24 One condition in the ERP data: deviant T2 in T4 standards T41 One condition in the ERP data: deviant T4 in T1 standards

PAGE 14

14 T42 One condition in the ERP data: deviant T4 in T2 s tandards

PAGE 15

15 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EFFECTS OF PRODUCTION TRA I NING AND PERCEPTION TRAINING ON L EXICAL TONE PERCEPTION ARE THE EFFECTS DOMAIN GENERAL OR DOMAIN SPECIFIC? By Shuang Lu August 2013 Chair: Edith Kaan Cochair: Ratree P Wayland Major: Linguistics The relationship between speech perception and production has been debated for a long tim e. The Motor Theory of speech perception (Liberman et al., 1989) claims that perceiving speech is identifying the intended articulatory gestures rather than perceiving the sound patterns. It seems to suggest that speech production precedes speech perceptio can they successfully perceive speech. One prediction that falls out from the Motor Theory is that training people to produce sounds should be more effective for improving the perc eptual ability than training people to only perceive sounds. My dissertation aims to test this prediction by comparing the effects of laboratory perception training and production training as well as the effects of different musical backgrounds (i.e., voca l ists and instrumentalists ) on the perception of lexical tones. The present dissertation encompasses three experiments The first experiment compared native English non musicians, instrumentalists and vocalists on the perception and production of Mandarin tones. The results supported previous findings, demonstrating the advantage of musical training on lexical tone perception. However,

PAGE 16

16 no additional benefit was found for musicians with vocal experience in addition to instrumental experience on either lexic al tone perception or production. The second experiment recorded both behavioral and electro physiological data to examine the effectiveness of a perception only training and a perception plus production training on lexical tone perception. The data showed that after training participants in both groups improved on tone discrimination Moreover, the participants in both groups did not differ in tone processing at the intentional or unintentional level after training The third experiment using identificatio n training rather than the discrimination training, demonstrated that the identification plus imitation group identified the tones more quickly in the post training task than in the pre training task while the identification only group did not show any imp rovement in terms of reaction time. These results suggested that the employment of the motor system does not specifically benefit the tone perceptual skills.

PAGE 17

17 CHAPTER 1 INTRODUCTION Project Motivation The relationship between speech perception and pro duction has been debated for a long time. The Motor Theory of speech perception (e.g. Liberman et al., 1967; 1989) asserts that speech perception is accomplished by identifying the intended articulatory gestures rather than perceiving the sound patterns. T he Direct Realist Theory (e.g. Best, 1995; Fowler, 1986) agrees with the Motor Theory in that the perceptual primitives are the vocal tract gestures, but claims that the primitives are the actual vocal tract gestures rather than intended gestures mediated by an innate neural module. Despite the subtle differences between the Motor Theory and the Direct Realist Theory, both of them imply that speech production precedes speech perception, i.e., can they successfully perceive speech. Moreover, several studies have shown that the accurate production of sounds in the second language precedes their accurate perception (e.g. Flege & Eefting, 1987; Sheldon & Strange, 1982; Goto, 1971), indicating that speech perception might rely on speech production. Thus, according to the Motor Theory, the Direct articulatory gestures) is expected to be more effective in improvin g the ability of comprehension than training people to perceive sounds. The current di ssertation aims to test this prediction by comparing the effects of laboratory perception training and production training as well as the effects of different musical bac kgrounds (i.e., vocal training and instrumental training) on the perception of lexical tones. The findings of this research will contribute to the current models of speech perception and will illuminate

PAGE 18

18 the relationship between the auditory system and the articulatory system as well as the relations hip between music and language. Background Views of the Relation between Production and Perception The Motor Theory was first proposed by Liberman and colleagues in the 1950s. Since its first appearance, the theo ry has undergone several revisions. The most recent version of the Motor Theory claims that (1) speech is perceived by recognizing the intended vocal tract gestures; and (2) speech perception and speech articulation are linked through an innate mechanism ( Liberman & Mattingly, 1989). Results from several behavioral and neuropsychological studies have supported these two asse rtions. For example, Adank, Haggort & Bekkering (2010) found that no matter whether people received auditory feedback (of their own voi ce) or not, vocal imitation significantly improved the comprehension of unfamiliar accent. Ojemann (1983) demonstrated that speech perception and speech production share many cortical areas, such as the left inferior frontal cortex and the superior tempora l lobes. The discovery of mirror neurons (Di Pellegrino et al., 1992) was also argued to be in support of the Motor Theory (e.g. Galant ucci Fowler & Tuvey, 2006; but see Lotto, Hickok & Holt 2009). Mirror neurons are neurons that are activated both durin g action execution and during action observation. It suggested that perception may be channeled through the production pathway. On the other hand, the Motor Theory has been criticized on theoretical and empirical grounds (e.g. Lotto et al., 2009; Massaro & Chen, 2008). For instance, some systems that control speech production was not associated with deficiency in speech comprehension (e.g.

PAGE 19

19 Moineau Dronkers & Bates, 2005). However, these studi es only focused on the single word level and most patients in the studies were over 40 years old. Thus, the retained comprehensive ability could probably be attributed to that the kinesthesia of the vocal tract gestures had been stored in the long term mem ory through years of articulatory practice. Patients may have used this stored information to aid in their comprehension, resulting in the gradual deterioration of perceptual ability. Hickok and Poeppel propose a dual stream model of speech processing and claim that the networks between speech perception and production are partially overlapping and part ially distinct (e.g. Hickok, 2001; Hickok & Poeppel, 2004 & 2007). They agree with the Motor Theory that there is a link between speech perception and produ ction, but deny that speech production is m andatory for speech perception. Other researchers even take a position opposite from the Motor Theory, arguing that speech production relies on speech perception (e.g. the Directions Into Velocities of Articulator s Model (DIVA), Guenther Ghosh & Tourville, 2006; and the Speech Learning Model, Flege, 1995 a ). This assertion has been supported by a magnetoencephalography (MEG) study by Levelt et al. (1998), who showed that the initiation of articulatory processes was slightly slower than the auditory cortical activations during speech production. In sum, previous literature demonstrates mixed results and accounts regarding the relationship between speech production and perception. Further research is needed to directl y compare the effect of production learning on speech perception and effect of perceptual learning on speech production, in order to uncover the nature of relationship between speech perception and speech production. In the following sections, I will

PAGE 20

20 brief ly review studies regarding perception training and production training of novel phonological categories. Previous Training Studies novel phonemic contrasts can be improved th rough relatively short term laboratory perception training or production training. Moreover, some studies also demonstrated that participants can transfer the improvement gained through the perception training to the production domain and vice versa. These results suggest that speech perception and production are linked to each other. Perception training Jamieson & Morosan (1986) trained the Canadian francophone adults to identify and discriminate the English voiced and unv /. The results showed that after 90 minutes of identification training with synthesized CV tokens participants identification performance and inter catego ry discrimination improved significantly compared with the con trol group. Moreover, the improvement with the synthetic CV tokens was transferred to the natural CV tokens that were not included in the training. Logan, Lively & Pisoni (1991) improved the training method used in Jamieson & Morosan (1986) by utilizing a large variety of speakers and phonetic contexts (e.g. word initial, word final and intervocalic) to train native speakers of Japane se to distinguish the English /r/ and /l / The participants in this study received fifteen 40 minute training sessions in whi ch they were asked to identify the stimulus from a minimal pair contrasting in /r/ and /l/ and received feedback Logan et al. (1991) reported that after training but the improvements were more prom inent in the word final and intervocalic conditions

PAGE 21

21 than in the word initial condition. In addition, Logan et al. (1991) performed two generalization tests and found that the participants identified /r/ and /l/ more accurately when novel words were produce d by an old talker than when novel words were produced by a new talker that they had never heard in the training sessions. Bradlow, Pisoni, Akahane Yamada & Tohkura (1997 ) followed Logan et al. (1991) but extended their training sessions to a longer peri od of time to examine the The results demonstrated that the Japanese participants improved significantly on the identification of English /r/ and /l/ and performed equally we ll on the two generalization tests no matter whether the novel words were produced by an old talker or a new talker. nificantly in the post test compared with the pre test, even though th e training sessions did not include any overt production compone nt The perception and production improvements were even retained three month after training (Bradlow et al., 1999). Thes e results provide evidence for the link and transfer between speech perception and production. Production training As reviewed in the previous section, several studies have investigated the effectiveness of perception training on the identification/discri mination of novel phonemes. However, only a small body of research has directly examine d the effect of production training on the perception of novel sound contrasts and these studies demonstrated controversial results Hirata (2004) examined the producti on and perception of Japanese pitch and duration contrasts by native English speakers. The participants in this study received a

PAGE 22

22 production training in which they first saw the fundamental frequency (F0) contours of the model utterance produ ced by native J apanese speakers, and t hen produce d the utterance by themselves overlaid the F0 contour of the model utterance participants were asked to reproduce the utteranc e until they thought their F0 contour matched the model. Moreover, they could listen to the model utterance whenever they wanted during their reproduction. The participants received ten of 30 minute training sessions during three and a half weeks. Hirata ( 2004) found that the participants improved significantly on both production and perception of Japanese pitch and duration contrasts after training. However, since the participants in Hirata (2004) could listen to the model utterances as many times as they wanted during the training sessions, it is unclear whether the improvements were due to the production training alone or the production training plus the perceptual exposure. Hattori (2009) obtain ed different result from Hirata (2004) through examining t he perception and production of the English /r/ and /l/ by native Japanese speakers. Participants in this study received 10 sessions of 30 40 minute prod uction training, including one on one explicit instruction and feedback regarding the appropriate artic ulatory gestures and positions of the production of /r/ and /l/ and repeated imitations. Hattori (2009) found that the Japanese speakers improved significantly on the production of English /r/ and /l/, and even showed native like accuracy after the produc tion training. However, their perception did not improve after training. Hattori (2009) took these results as evidence to assume that speech perception and production

PAGE 23

23 might have separated mental represen tations, and learning in one domain might not transfe r to the other domain. Comparison between perception training and production training Some recent studies also compared the effectiveness of perception training and production training on the perception of L2 phonemic contrasts. For example, Herd (2011) trained native English learners of Spanish to perceive and produce the Spanish intervocalic contrasts between the tap / / the trill /r/ and the dental stop /d/ using three training paradigms. In the perception training the participants heard one word that contained one of the sounds and saw a corresponding minimal pair. T hen they were asked to choose the word they heard f rom the minimal pair and received feedback In the production training the participants were first presented with the waveform and spectrogram of a produce the stimulus themselves. N ext th ey compared the waveform and spectrogram of their own production with the mode l and were encouraged to reproduce until they thought the two waveforms and spectrograms matched. The participants who received the production training c ould never hear the nativ In the combination training the participants received three sessions of perception training and three sessions of production training as described above Herd (2011) found that all ception and production of the / r/ contrast in Spanish and the improvements in perception were also generalized to new speakers, new dialects and new words However, the perception of the / d/ contrast did not improve in any of the training group s S pecifically, the participants who received the perception training showed equal perception ability of the / d/ contrast in the pre and post tests, while the participants who received the production training and the

PAGE 24

24 combination training declined significantly in their perception of this contrast after training. Moreover, participants in all the three training groups improved in the production of the trill /r/. Taken together, Herd (2011) concluded that the perception training was more effective than production training to transfer to the other domain. Nevertheless, since the participants who received the perception training could production training were not allowed to hear the model productions it was unclear whether the advantage of the perception training was due to the extra exposure or the perception training alone In addition, the declined perception in the production training and the combination training group was probably due to that the participants he ard their own inaccurate productions, which offset the training effect in perception. As previously mentioned, one concern in this line of research is that it is just impossible to completely isolate speech production from speech perception, i.e. we percei ve what we are producing inevitably. Therefore, some studies have compared perception only training to perception plus production training. For example, Leach & Samuel (2007) demonstrated that perception training plus a production task was more powerful to identification of words in noise (i.e. lexical configuration tasks) than perception training alone. However, they also found that the engagement task s (e.g. phonemic restoration). Baese Berk (2010) compared the effectiveness of perception only training and perception plus production training on the perception and production of novel sound contrast by native English listeners. In the perception only tr aining participants heard

PAGE 25

2 5 one syllable and saw a corresponding picture demonstrating the distribution of tokens on a continuum. The participants were asked to pay close attention to the picture and press a button on a button box to advance to the next tria l when they were ready. The perception plus production training was exactly the same as the perception training except that participants were asked to imitate the stimulus they heard instead of just pressing a button on the button box. The participants in the perception plus production training were only allowed to hear and imitate each stimulus once. Baese Berk (2010) found that the perception perception of novel sound contrasts than the percept ion plus production training. Moreover, even though the participants who received the perception only training also improved in their production, the participants in the perception plus production training showed more significant improvement. Baese Berk (2 010) concluded that the perception only training transferred more effectively than the perception plus production training, and the perceptual learning might be hindered by the additional production task. Taken together, previous studies have examined the effects of perception training and production training on the perception of novel phonemic c ontrasts and demonstrated mixed results. Very few studies have compared the effects of perception training and production training on the perception of suprasegmen tal phonological contrasts. The current study investigates this question by focusing on lexical tones. Tones in Music and Language Music and language share several acoustic characteristics, such as duration, rhythm, and pitch. Among these characteristics pitch has particularly attracted varying pitch to convey

PAGE 26

26 information. In music, melody involves two types of pitch information: contour and interval code. Contour refers to the pitch direction varia tions between successive tones while interval code is the pitch distance between notes. In language, pitch differences are tone languages, which use pitch differences to distinguish lexical meaning and to carry grammatical distinction (Chao, 1948). For example, in Mandarin Chinese, the syllable [ma] produced with four different tones (Tone1 has high level pitch, Tone has a high rising pitch, Tone3 a low dipping pitch, and Tone4 a high falling pitch) will r esult in (Chao, 1948). In i ntonation languages, such as English pitch variation is used to convey prominence and/or pragmatic meaning ra ther than encode lexical meaning. At the word level, English exhibits lexical stress, with some syllables being perceptually more prominent than others. The pitch of the stressed syllable is typically higher than the unstressed syllables. At the sentence l evel, pitch variation is used to distinguish different sentence types (question or statements) and to flag information as new or unpredictable. Tone Perception and Production by Non native Speakers Previous studies have shown that non native listeners of t one languages have difficulties in both comprehending and producing lexical tones (e.g. Gandour 1983; Wang et al. 1999; White, 1981). These difficulties have been attributed to: (1) different perceptual dimensions by non native speakers compared with nativ e speakers, and (2) the interference from suprasegmental features in the native language. For example, Gandour (1983) found that native English listeners tended to pay more attention to the pitch onset, offset and the average pitch, while native speakers o f tone languages focus

PAGE 27

27 more on the pitch contour. Neurophysiological research has also provided support for these differences. Krishnan et al. (2005) reported that the brain stem responses are a more accurate/consistent reflection of pitch contour in tone language listeners than in non tone language listeners. Using Event Related brain Po tentials (ERPs), Chandrasekaran, Krishnan & Gandour (2007) found that a larger mismatch negativity (MMN) was elicited by tone contrasts in tone language listeners than in n on tone language listeners. As for the native language interference, Shen (1989) claimed that English speakers tended to perceive the Mandarin Tone 4 as stressed and Tone3 as unstressed. Moreover, English speakers are likely to consider Mandarin Tone2 as t he rising intonation in question sentences, and Tone4 as the falling intonation in the final position of statement sentences. In terms of production, Chen (1974) claimed that non tone language speakers needed to learn to widen their pitch range in order to successfully produce a tone language. Several studies have found that native English learners of Chinese have difficulty in producing all four tones (e.g. Miracle, 1989 ). Shen (1989) reported that Tone4 was the most difficult tone for non native speakers, which was ascribed to the interference of the English intonation system. English speakers usually produced Tone4 at a lower pitch onset and with a less steep falling slope than the native Chinese speakers, because they associated Tone4 with the falling in tonation in English, which starts at a lower pitch and falls down gradually Effect of Music Background on Lexical Tone Perception Some evidence has suggested that long term experience with pitch differences is domain general, i.e. capacities in language d omain percolate to music domain and vice versa. For example, Pfordresher & Brown (2009) showed that tone language speakers

PAGE 28

28 outperformed non tone language speakers in both vocally imitating and perceptually discriminating musical pitch. On the other hand, m usic training has also been reported to facilitate speech perception (e.g., Anvari, Trainor, Woodside & Levy, 2002; Slevc & Miyake, 2006). Several behavioral studies have demonstrated that musicians were better than non musicians in lexical tone discrimi nation and identification. For example, Lee & Hung (2008) performed a behavioral identification task on English musicians and non musicians. Participants were asked to identify four Mandarin tone s associated with the syllable [sa] by 32 speakers after a br ief tutorial. The identification by musicians was more accur ate than that by non musicians. Schwanh u er (2007) also found that English musicians were better than non musicians in identification and discrimination of Thai tones. At the level of brainstem, Wong et al. (2007) found that pitch tracking was more faithful and robust in musicians than in non musicians. Using ERPs, Chandrasekaran, Krishnan & Gandour (2009) demonstrated that the mean amplitude of the MMN was larger in English musicians than in Engl ish non musicians during Mandarin tone processing W hile there are comprehensive studies on trained instrumental musicians, relatively little is known about formally trained vocal musicians and the differences between the vocalists and instrumentalists. Be sides auditory pitch discrimination ability (perception), vocal pitch control (production) is another essential skill for successful vocal musicians. Using non linguistic pitch stimuli, Zwissler (1971) found that accurate singers performed significantly be tter on the auditory pitch discrimination task than the inaccurate singers, suggesting that there is a positive relationship between auditory

PAGE 29

29 pitch discrimination abilities and accurate singing. However, it is unknown whether this advantage in vo cal musici ans could also carry over to the language domain. Nikjeh (2006) also used musical stimuli and examined the intentional and unintentional pitch discrimination and pitch production by English vocalists and instrumentalists. The results demonstrated that the vocalists and instrumentalists did not differ significantly in pitch perception or pitch production. Nikjeh (2006) claimed that these result w as probably due to that the perception and production tasks were too easy to show different effect of music backgr ound between the vocalists and instrumentalists. Effects of Laboratory Training on Lexical Tone Perception Several behavioral studies have shown that short term perceptual and production training is effective in improving the comprehension and production o f lexical tones by non tonal language speakers. Moreover, the effects of perception training and production training can be inter transferable. Perception training Wang and colleagues conducted a series of tone experiments on the effectiveness of perceptio n training. Wang et al. (1999) trained native American English speakers for two weeks to identify Mandarin Chinese tones in 100 natural Chinese words. After the perception training, the English speakers improved significantly in identifying the tones. The improvement also generalized to untrained stimuli and was retained six months after training. Wang Jongman & Sereno (2003a) further showed that the improvement in identification and discrimination of lexical tones gained through perceptual training transf erred to production. Wang et al. (2003b) conducted a functional magnetic resonance imaging (fMRI) study on the same group of participants as in Wang et al. (1999). They found that after training the improvements observed in

PAGE 30

30 the behavioral study were associ Effects of perception training have also been found for other tone languages. Wayland & Guion (2004) demonstrated that native English listeners revealed significantly improved ability in perceiving Thai lexical tone contrasts after an auditory training. Finally, using ERPs, Kaan et al. (2007) examined the effect of perceptual identification training on the unintentional processing of Thai tones by nave speakers, as measured by the MMN. They found that E nglish speakers showed an increased MMN only for high deviant tone stimuli, while the MMN for low deviant stimuli did not differ before and after training. Production training Leather (1990) was the first to examine the effect of production training on per ception. In this study, Leather trained native Dutch sp eakers to produce the syllable [yu] with four Mandarin tones. The result showed that after the production training the Dutch speakers were able to perceive the differences in tones. Furthermore, Leathe r (1990) also observed an effect of perceptual training on production: after the Dutch listeners were trained to identify the four tones, their productions were more accurate. However, since Leather (1990) only used one syllable in both the training and th e post training tasks, it is unclear whether the improvement generalizes to other stimuli. Interim Summary After reviewing previous studies on the effects of perception training and production training on tone perception in both language and music domains, we can identify several gaps in the current literature. (1 ) Although s ome research has suggested that vocal experience might positively affect music pitch perception, it is unknown whether this advantage in vo cal musicians could also carry over to the

PAGE 31

31 lan guage domain (2) Very few studies have directly compared the effectiveness of production training and perception training on the perception of lexical tones. (3) To our knowledge, no study has investigated which stage of processing can be affected by prod uction training: will the production and perception training have a different effect on lexical tone perception at the behavioral (intentional) level? Or will the two types of training also differentially affect unintentional processing (i.e. when the audi tory stimuli The current study aims to provide more data in terms of these three aspects by using both behavioral and neurophysiological methods to compare the effects of production training and perception train ing (in both music and language domain) on the intentional and unintentional processing of lexical tones. In the next section, I will discuss the neurophysiological method used in the present study to examine the unintentional processing of lexical tones: the event related potentials (ERPs). Event related Potentials (ERPs) Event related potentials (ERPs) are a good method to study the unintentional processing of lexical tones while the auditory stimuli are presented to participants ERPs do not require par provide high temporal resolution The participants can, for instance, read a responses to the auditory stimuli are recorded. In contrast, behavioral tasks provide data regarding partici fully process the stimuli. Therefore, a passive ERPs task, combined with a behavioral task could give a thorough view of how people intentionally and unintentionally process lexical tones, and indicate which level of processing is affected by the type of training (i.e. perception vs. production training) Moreover, previous studies on speech sound training have shown

PAGE 32

32 that changes in neural activity may precede the behavioral i mprovement (e.g. Tremblay Kraus & McGee, 1998). Recording neurophysiological measures may therefore be more informative than only recording behavioral measures, since perceptual changes after training may occur at the unintentional level but not (yet) at the intentional /behavioral level. ERP Comp onents that are Relevant to Auditory Perception Several ERP components have been found to be relevant to auditory perception. The Mismatch Negativity (MMN) is a negative wave elicited by auditory stimuli that are infrequently presented and deviate from a f requently presented standard stimuli. The MMN usually occurs between 100 300ms after the onset of deviant stimuli (e.g. Ntnen & Alho, 1995), and has been claimed to reflect pre attentive processing (although this component may be sensitiv e to intentiona l manipulations, e.g., Woldorff, Hackley & Hillyard, 1991). Many studies have shown that the behavioral improvement in speech perception is usually accompanied by an increased MMN (e.g. Tremblay et al., 1997). Sometimes the increased MMN is observed even b efore the behavioral improvement occurs (e.g. Tremblay et al., 1998). The Late Negativity is another negative wave that occurs around 350 to 600ms after the onset of deviant stimuli. The late negativity has been associated with the reorientation of attenti on (Shestakova et al., 2003). Some studies have shown that the late negativity became smaller after training (e.g. Kaan, et al., 2008), suggesting that after training the deviant stimuli became easier to distinguish and required less attention. Besides MMN and late negativity, the N1 and P2 components have also been reported to be sensitive to the physical features of auditory stimuli. The N1 is a negative wave that occurs between 80 and 120ms; the P2 is the second positive peak, which is normally observed at 180 250 ms after the onset of

PAGE 33

33 the stimulus. Previous research has shown that the amplitude of P2 increas es after training (e.g. Atienza, Cantero & Dominguez Marin, 2002). The P1 and P300 (P3a and P3b) might also be relevant to the current study. The P1 and P300 are the first and third positive peaks that often occur at about 100ms and 300ms after the stimulus onset respectively. The amplitude of P1 is modulated by selective attention (e.g. Van Voorhis & Hillyard, 1977). P3a often follows the MMN and has been argued to be an index of an automatic attention switch toward a deviant stimulus (Escera et al., 1998). If these ERP components are differentially affected by the type of training (perc eption plus production or perception only), it suggests that the additional production component during training changes the unintentional processing of the auditory stimuli after training. The Current Study Research Questions The present study aims to ans wer t hree main questions: Are the effects of production training and perception training domain general or domain specific ? That is, can the instrumental and vocal expertise percolate to lexical tone perception and production ? Is laboratory perception plus production training more effective than percepti on only training in facilitating lexical tone perception at the intentional level ? Does perception plus production training affect the unintentional perception of lexical tones differently from perception on ly training? Structure of the Dissertation Three experiments investigating the relationship between perception and production are presented in the current dissertation. The first experiment examines the production and perception training in the music domai n, while the second and third experiments focus on the language domain. Experiment 1 compares English vocalists

PAGE 34

34 and instrumentalists to see which type of musical training is more advantageous to lexical tone perception and production Participants were tes ted in (1) a same/different discrimination task, in which they decided whether the two syllables they heard had a same or different tone; and (2) a production task, in which they heard a tone and imitated immediately. Experiment 2 collec ted both behaviora l and electrophysiological ( ERP s) data to examine the effectiveness of a perception plus production training and a perception only training on the intentional and unintentional processing of tones by native English speakers. Participants were trained and t ested over the course of three consecutive days. baseline. During recording, participants were watching a silent movie while a stream of syllables was presented. After the ERP recording, p articipants did a behavioral same/different discrimination task. On the second day, participants received either a perception only or a perception plus production training. The trainings for both groups were exactly the same except that the perception plus production training required participants to imitate the stimuli, while the perception only training had participants utter a word unrelated to the stimuli after a same/different discrimination training. On the third day, both groups did the same ERP and behavioral tasks as on the first day. If the employment of the motor system benefits the perceptual skills, the perception plus production group should show more improvement and an increased sensitivity to the tone contrasts as a function of training than the perception only group. However, this experiment is biased toward the perception plus production group because the imitation task may require the participants in this group to pay more attention to the stimuli than the participants in the perception onl y group In other words the advantage of the

PAGE 35

35 perception plus close attention on the stimuli rather than the specific training type. In order to untangle this problem, Experiment 3 was designed to be biased for the perception only group. If the perception plus production group still show more improvement than the perception only group, it is clear that the production training is more effective than the perception training and the emp loyment of the articulatory system benefits the perceptual skills. In Experiment 3, participants were trained and tested over the course of two consecutive days. On the first day, participants first did an identification task, in which they heard one stim ulus and then identified whether this stimulus was a level, rising or falling tone. After the baseline task, the participants received either a perception only or a percep tion plus production training. T he trainings for both groups were exactly the same ex cept that after an identification training the perception plus production training required participants to imitate the stimuli, while the perception only training had participants utter the tone type s of stimuli (i.e. level, rising or falling). On the sec ond day, all participants did the same identification task as on the first day. This experiment is biased toward the perception only group because the pre and post training tests asked participants to identify the tone types of the stimuli ; but the tone t ypes were only reinforced to the participants who received the perception only training during the production process The results of the three experiments will be presented and used to investigate the effectiveness of perception training and production tr aining on lexical tone perception. Lastly, I will discuss the relationship between the results and other rele vant literature on perception training and production training and will suggest directions for future research in this area.

PAGE 36

36 Intellectual Merit Th e current study will contribute to psycho /neuro linguistic literature on speech learning and perception. First, there has been a long debate about whether production relies on perception or vice versa. Results of this study will illuminate the relationshi p between speech perception and production. Second, most behavioral and electrophysiological studies have thus far ignored the effect of production training on the perception of lexical tones, focusing only on the effect of perception training upon product ion. If our predictions are borne out, the perception plus production training should be more effective than the perception only training on lexical tone perception. This result may inspire researchers to re evaluate the importance of production training i n second language learning. Third, the proposed research also introduces methodology that allows linguists to integrate production training into the study of speech perception. The training paradigm developed in the present study can be easily extended to investigate other aspects of speech. Broader Implications This study is primarily fundamental in nature. Howev er results from the present and similar studies will eventually have implications for language education and language software development. With t he economic development of China, there is an increasing demand for learning Chinese. Findings of the current study will help to develop more effective teaching methods for learning lexical tones, which have been claimed as an obstacle in learning Chinese by non native speakers. Newer versions of language learning software can also be programmed to utilize production training for different tone learning. Moreover, if both perception only and perception plus production training lead to changes at the neural level, then further identification of the neural

PAGE 37

37 substrates for pitch perception and vocal pitch control may have implications for treatment strategies for patients who have neurological, language and vocal impairments, such as dyslexia, aphasia, and heari ng/production impairment.

PAGE 38

38 CHAPTER 2 EXPERIMENT 1 : EFFECTS OF LONG TERM MUSICAL PERCEPTION AND PRODUC TION TRAINING ON LEXICAL TONE PERCEPTION AND PRODUCTION The first experiment examines the effects of production training and perception training in the m usic domain on lexical tone perception and production Previous research has demonstrated that musicians were more accurate than non musicians in lexical tone discrimination and identification (e.g. Alexander, Wong & Bradlow, 2005; Wong et al., 2007 a ). Thi s experiment fu r ther compare d American English vocalists and instrumentalists to see which type of musical background was more advantageous to lexical tone perception and production Learning to sing is very different from learning an instrument. Singers r eceive extensive modeling practice (i.e they model their singing on perfect tones and remember the sensations of their productions on the basis of feedback) during vocal training, while instrumental musicians learn to match their hand position and finger placement (depending on different types of instruments) with the notes during instrumental practice. Therefore, one would expect that vocal musicians have more advantage than instrumental musicians in imitating lexical tones. Moreover, if there is indeed a strong connection between production and perception, as suggested by the motor theory ( e.g. Liberman & Mattingly, 1989) the vocalists might also be superior to the instrumentalists in the perception of lexical tones. This experiment tested these predicti ons through a same/different discrimination task and a production imitation task. Methods Participants A total of 60 participants were recruited from the University of Florida student body : 30 native speakers of American English with formal music training (15 vocalists

PAGE 39

39 and 15 instrumentalists), 15 native speakers of American English without music training and 15 native speakers of Mandarin Chinese without music training. The English vocalists ( 1 male and 14 female s ) were between the ages of 19 and 30 (M = 22.53, SD = 2.95) and the English instrumentalists (4 m ale s and 11 female s ) were between the ages of 18 and 23 (M = 20.2, SD = 1.42) All English vocalist s had at least 3 years of continuous formal vocal training (M = 8.33 SD = 4.15 ) starting at or befo re the age of 19 (M = 13.60 SD = 4.05 ) and were still singing at the time of their participation. Thirteen of the vocalists had also received some instrumental training before they had formal vocal training primarily on piano. The English instrumentalist s had diverse instrumental backgrounds. All of them had at least 3 years of continuous formal instrumental training (M = 8.87, SD = 3.09), starting at or before the age of 14 (M = 9.6, SD = 2.16) and were still playing at least one instrument at the time o f their participation. None of the English musicians claimed that they had perfect /absolute pitch, which refers to the ability to identify the pitch of any musical note without a reference pitch. musical backgrounds. The English non musicians (4 males and 11 female s ) were between the ages of 19 and 23 (M = 20.6, SD = 1.24). None of the English participants had previous exposure to any tonal languages The Chinese non musicians (6 males and 9 femal e s ) were between the ages of 22 and 27 (M = 23.73, SD = 1.94). All of them were originally from mainland China, and did not know any tone language other than Mandarin Chinese. All non musicians had less than 2 yea rs of music al tr aining and had not performe d music (played an instrument or s ang) within the past 4 years. An additional 11 participants were recruited, but were excluded from data analysis either because they did not meet

PAGE 40

40 the musical criteria or because of technical problems with the experimental programs. All participants received course credit or financial compensation for their participants. The experiment was approved by the University of Florida Institutional Review Board. Stimuli Stimuli consist ed of six monosyllabic syllables ([t h i], [li], [mi], [t h o], [lo] and [mo]) associated with four Mandarin tones (T one 1: high level; T one 2: high rising; T one 3: low dipping; T one 4: high falling) These particular syllables were selected because American English has these segments as well. Therefore, Ameri cans speakers c ould focus their attention on the lexical tones and were not distracted by unfamiliar syllables. Moreover, all the syllables with the four tones result ed in real words in Mandarin Chinese. Two female native s peakers of Mandarin Chinese were r ecruited to record the stimuli using the Shure SM58 microphone and Marantz PMD660 digital recorder in a sound attenuating booth. Each speaker produce d each stimulus twice, yielding a total of 96 tokens (6 syllables 4 tones 2 speakers 2 tokens / speake r) Only female talkers were recorded for stimuli because previous study had shown that adult female talkers were more intelligible than male adult speakers (Bradlow Torretta & P isoni 1996). All the target words were produced in the phrase Qing shuo TAR GET zhege zi (please say TARGET this word), in order to better mimic natural speech and to reduce the phonetic differences (e.g. duration and creaky voice) between the four tones. This design might be a confound ing factor in the current study, because the tone of the target word could be influenced by the surrounding tones ( Francis et al., 2006 ). Nevertheless our pilot intelligibility task showed that if we recorded the syllables in isolation even English speakers who ha d no exposure to lexical tones cou ld easily notice the difference s between the T one 3 and the other three tones according to the

PAGE 41

41 longer duration and obvious turning point. Thus, we decide d to exploit the frame recording in order to prevent the ceiling effect. The target syllables were segme nted by using PRAAT 5.1.25 (Boersma & Weenink, 2010) and then were normalized for amplitude. Five native Mandarin speaking informants were asked to judge the stimuli for intelligibility. Only words with a mean accuracy rate of 95% or above were included in the experiment. Figure 2 1 shows the waveforms and spectrograms with the pitch contours of the Mandarin Tone1, Tone2, Tone3 and Tone4 associated with the syllable [ma]. Procedure Participants were first asked to read through the informed consent statemen t and complete a questionnaire that includes language, educational background and musical training experience ( Appendix B ). Informed consent was obtained from all of the participants after the purpose and procedures of the experiment were fully explained b y the experimenter Then the participants received a standard hearing test, in which they were wearing headphones and beeps of varying pitch levels (between 250 and 8k Hz at 25 dB) were presented randomly to either their left or right ear. They were asked to raise the corresponding hand to indicate in which ear they heard the sound. This task ensured that all the participants could hear the stimuli well and our results were not confounded by partici working memories were also tested on site through a backward digit span test In the test, participants heard strings of numbers and were asked to immediately repeat them back in the given order or reverse order. Working memory has been shown to affect the perception of pitch differences ( e.g. Payne, 2003 ) We therefore includ e d this factor in data analysis to make sure that the advantages of tone discrimination were not due to

PAGE 42

42 better working memory The auditory working memory test was conducted in the par musicians and English for the English musicians and non musicians. The mean accuracy percentages of the auditory working memory test were 81.19% (SD = 10.50%), 54.76% (SD = 9.12%), 58.81% (SD = 12.80%) and 60.71% (SD = 11.37%) for the Chinese non musician, the English non musician, the English vocalists and the English instrumentalists respectively. The accuracy percentage for each participant was submitted to a one way ANOVA with group as the be tween subject factor. The result showed that there was a main effect of group, F (3, 56) = 17.21, p < .001. Further pairwise comparisons yielded that the Chinese non musicians responded more accurately in the auditory working memory test than the English n on musicians [ t (28) = 7.36, p < .001], the English vocalists [ t (28) = 5.12, p < .001] and the English instrumentalists [ t (28) = 5.24, p < .001]. Importantly, n o difference was found between the three English groups p s > 0.80 The differences between th e Chinese non musicians and the three English groups might be due to that the pronunciations of the Chinese number words are shorter than the pronunciations of the English number words (Stigler, Lee & Stevenson, 1986) In addition, p titude was tested on site by using the Advanced Measures of Music Audiation (AMMA, Gordon, 1989). During the test, pairs of brief musical samples were presented to the participants and they were asked to indicate whether the samples in each pair were the same or different; if different, participants had to indicate whether the discrepancy is rhythmic or tonal in nature. This test provide d a standard ized and effective way of determining the participant abilities irrespective of music training or e xperience M ean accuracy percentages of the

PAGE 43

43 tonal test, the rhythmic test and the composite performance in AMMA for the Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians are presented in Table 2 1 The accuracy percentages were generated automatically by the AMMA program for each participant and were submitted to three one way ANOVAs with group as the between subject factor. The results showed that the effect of group was not significant for the tonal t est [ F (3, 56) = .63, p = .60], the rhythmic test [ F (3, 56) = 50 p > 68 ], or the composite performance [ F (3, 56) = .79, p > .50] indicating that the four groups did not differ in tonal or rhythmic skills. After the AMMA test the participants d id a b ehavioral discrimination task and a production imitation task on lexical tones The order of the two tasks was counterbalanced across participants in order to prevent practice effect s In both tasks, stimuli were presented to participants via headphones (S ennheiser HD 280 PRO) using the E Prime 2.0 software (Psychology Software Tools, Pittsburgh, PA). The discrimination task employed a forced choice AX paradigm, in which the stimuli were presented in pairs. Each trial contain ed two different tokens of the same syllable produced by one speaker with the tokens being separated by inter stimulus interval (ISI) of 500 milliseconds The ISI was set to 500ms because previous studies have shown that participants were sensitive to nonnative sound contrasts at the 5 00ms ISI ( e.g. Cowan & Morse, 1986; Pisoni, 1973; Van Hessen & Schouten, 1992 ) The participants were asked to indicate whether the second stimulus ha d the same or different tone as the preceding one by pressing a button on a button box. In order to pr even t the dominant hand bias half of the participant s did the S ame D ifferent task, in which the leftmost bu tton on the button box indicate d same while the rightmost

PAGE 44

44 denoted different ; the other half of the participants ha d the reverse pattern. A s hort t utorial and 8 practice tr i a ls were presented b efore the experimental tr i a ls. In the tutorial, participants were presented with the syllable [ma] associated with the four Mandarin tones which were produced by a different female native Mandarin speaker Whi le the participants were listening to the four tones they were also presented with the corresponding tone figures which helped them to understand the tone contours. The 8 practice tr ia ls consist ed of different syllables from the experimental stimuli Dur ing practice, participants received feedback after they made their decisions, and t hese data were dropped from analysis. Next, a total of 288 experimental tr ia ls w as presented, with 144 same and 144 different trials T he discrimination of trials involves categorization skills because participants need to form categories for the pitch contours and to ignore other within category acoustic and articulatory features. To the contrary, ipants to pay more attention to the within category properties (e.g. Jamieson & Morosan, 1989). The participants were encouraged to respon d as quickly and accurately as possible. No feedback was given in the experimental trials and automatically triggered the next trial after a 1500 millisecond inter trial interval accuracy and reaction time were logged. A short break with at least 30 seconds was given halfway through the task, in order to prevent fatigue. In th e self paced imitation task, the participants first heard one stimulus via headphones (Sennheiser HD 280 PRO) and were asked to imitate the sound immediately. After their imitation in each trial they had to click the mouse to start the next trial. Before the experimental tr ia ls, the participants received a short tutorial on

PAGE 45

45 Mandarin tone production and were given 8 practice tr ia ls which consisted of different syllables produced by the same female talker as in the tutorial. These practice data were drop ped from analysis. After the practice trials, a total of 96 (6 syllables 4 tones 2 speakers 2 tokens/ speakers) experimental trials were presented in two blocks, with a 30 second break in between. All participants were recorded through a microphone (S hure SM58 ) and a d igital recorder ( Marantz PMD660 ) in a sound attenuating booth. They were encouraged to repeat or correct whenever they feel necessary. Lastly, t he participant s were asked to self e valuate their performance and g a ve suggestions on both di scrimination and imitation tasks The entire experiment took about 1.5 hours per participant. Data Analysis Discrimination task Data obtained in the discrimination task were converted to d prime scores, which were calculated as the hit rates (correct respo nses in the different tr ia ls) minus the false alarm rates (incorrect responses in the same tr ia ls). Hit rate s of 1 were corrected with 1 1/2N, alse alarm rate s of 0 w ere replaced by 1/2N, where N equal ed the total number of the same tr ia ls. A one way ANOVA with group as the between subject factor was performed on the d prime scores to see whether there were differences between the four groups in lexical tone discrimination. Furthermore, reaction tim e (RT) was calculated only for the correct responses. For each participant the RT outliers that w ere beyond the mean RT of all the correct responses plus 2.5 standard deviation s for that participant were replaced by this value. The RT data were then submitted to a one way

PAGE 46

46 ANOVA with group as the between subject factor, in order to examine whether the English musicians discriminated tones more quickly than the English non musicians In addition, t he accuracy percentage s fo r each tone pair w ere also calculated for each participant A repeated measure ANOVA was conducted on the accuracy percentage s with tone pair as the within subject factor and group as the between subject facto r According to the literature (e.g. Roberts & Russo, 1999), all the p values and the F values were adjusted using the Gre enhouse Geisser correction and the post hoc paired t tests were adjusted using the Bonferroni correction for multiple comparisons. The uncorrected degrees of freedom were reported i n both the ANOVA tests and the t tests. The result was considered statistically significant when the p value was less than .05, and was considered as marginally significant when the p value was between .05 and .10. Imitation task The production responses f rom each participant were segmented by using PRAAT 5.1.25 (Boersma and Weenink, 2010 ) Only the best production for each trial was selected for data analysis, resulting in a total of 96 productions per participant. For each trial the pitch contour of the b est production best resembled the contour of the target stimulus, as determined by the experimenter using the PRAAT program. Thirty six native speakers of Mandarin Chinese (20 male, 16 female) who did not participate in the experiment were recruited as ju imitations. All judges were between the ages of 20 and 31 (M=22.28, SD=1.99) and with randomly assigned to 3 judges. The production was conside red accurate when two of the judges had the same identification, which was consistent with the target stimulus.

PAGE 47

47 The accuracy percentage for each tone type by each participant was calculated. A repeated measure ANOVA were conducted on the production accurac y data with t ones as the within participant factor and group as the between par ticipant facto r. Predictions According to previous studies, we should find that the English musicians outperform the English non musicians in both the discrimination task and th e imitation task as a result of musical training. Moreover, the English vocalists should perform better than the English instrumentalists in the imitation task, since the vocalists have been trained to control their vocal pitch and to tune their sensations of production during their vocal training If the vocal expertise benefits the perceptual skills, as suggested by the Motor theory ( e.g. Liberman and Mattingly, 1989) we should also find that the English vocalists were superior to the instrumentalists in the discrimination task. Results Discrimination Task Accuracy Mean d prime scores with the standard errors for the Chinese non musicians, the English vocalists, the English instrumentalists and the English non music ians are presented in Figure 2 2 The mean d prime scores were 4.25 (SD = .72), 3.12 (SD = .40), 2.84 (SD = .63) and 2.41 (SD = .90) for the Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians respectively. A one way ANOVA was performed on t he d prime scores with group as the between subject factor, showing that there was a main effect of group, F (3, 56) = 19.56, p < .00 1. Pairwise comparisons yielded that the Chinese non musicians were more accurate in tone discrimination than the English v ocalists ( p < .001), the English

PAGE 48

48 instrumentalists ( p < .001) and the English non musicians ( p < .001). Moreover, the English vocalists performed significantly better than the English non musicians ( p = .038). However, t he difference between the English ins trumentalists and the English non musicians ( p > .57) and the difference between the English vocalists and the English instrumentalists ( p = 1.00) did not reach significance. To sum up, the English vocalists outperformed the English non musicians in tone discrimination, but the English instrumentalists did not show such advantage. Reaction times Mean reaction times with the standard errors of all the correct responses for the four groups are illustrat ed in Figure 2 3 The mean reaction times were 1023ms ( SD = 256.98), 1198 ms (SD = 299.73), 1034 (SD = 430.17) and 1055 ms (SD = 257.45) for the Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians respectively. A one way ANOVA with group as the between subject factor showed that the effect of group was not significant, F (3, 56) = .97, p > .41. In sum, the four groups did not differ in the reaction times for the tone discrimination task. Accuracy percentage for each tone pair Because there are more different pairs (Tone1 Tone2: pair12; Tone1 Tone3: pair13; Tone1 Tone4: pair14; Tone2 Tone3: pair23; Tone3 Tone4: pair34) than the Tone1: pair11; Tone2 Tone2: pair22; Tone3 Tone3: pair33; Tone4 Tone4: pair44), each same pair had more experimen tal trials (i.e. 36 trials for each pair) than each different pair (i.e.24 trials for each pair). Therefore, we decided to perform analys e s on the s ame pair s and the d ifferent pair s separately The accuracy percentages for each tone pair were calcula ted and were collapsed over the order of

PAGE 49

49 presentation within the trials (e.g. pair12 consisted of Tone1 Tone2 and Tone2 Tone1) for e ach of the four groups. Figure 2 4 illustrates the accuracy percentages with the standard errors for the same pairs of the C hinese non musicians, the English vocalists, the English instrumentalists and the English non musicians. The accuracy percentages for each participant were submitted to a 4 (pair) 4 (group) repeated measures ANOVA with pair as the within subject factor and group as the between subject factor. The result showed that the effect of pair [ F ( 3 168 ) = 20.23 p < 00 1] effect of group [ F ( 3 56 ) = 8.46 p > 00 1] and the interaction of pair group [ F ( 9 168 ) = 3.11 p = 00 4] were significant. Pairwise com parisons revealed that f or the s the Chinese non musicians outper formed the English vocalists ( p = .0 0 5), the instrumentalists ( p < .001), an d the English non musicians ( p = .0 14 ). No difference was found between the three English groups. For th e pair11 t he Chinese non musicians outperf ormed the English vocalists ( p = .0 0 5 ), and the instrumentalists ( p = .019 ); T he English non musicians performed equally well as the Chinese non musicians on the discrimination of pair11 (( p = .44). No other diffe rence was found in other comparisons between the four groups. For the pair22 t he vocalists marginally outperformed the instrumentalists ( p = .067). The Chinese non musicians were only superior to the English non musicians ( p = .032) and the English instru mentalists ( p < .001), but not the English vocalists ( p = .58 ). Moreover, the English instrumentalists did not show an advantage on the discrimination of pair22 over the English non musicians ( p > .97 ). For the pair33 n o difference was found across all th e four groups For the pair44 t he Chinese non musicians outperformed the three English groups ( p s < .01), but the three English groups did not differ from each other ( p s = 1.00 ).

PAGE 50

50 Figure 2 5 presents the accuracy percentages with the standard errors for ea ch of the different pairs in the Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians. A 6 (pair) 4 (group) repeated measures ANOVA with pair as the within subject factor and group as the between subjec t factor yielded a main effect of pair [ F ( 5 280 ) = 14.75 p < 00 1], a main effect of group [ F ( 3 56 ) = 11.97 p < 00 1] and a marginally significant interaction of pair group [ F ( 15 280 ) = 1.84 p = .06]. Pairwise comparisons displayed that for the different pairs t he vocalists and the instrumentalists performed almost as accurately as the Chinese non musicians ( p = 1.00) and both musician groups outperformed the English non musicians ( p s < .001). No difference was found between the English vocalis ts and instrumentalists ( p = 1.00 ). In addition, participants discriminated the pair12 and pair13 more accurately than the pair14 ( p s < .001) pair23 ( p s < .009 ) pair24 ( p s < .04 ) and pair34 ( p s < .001) Imitation Task Table 2 2 presents the mean percenta that were agreed upon by all 3 judges and 2 judges respectively for each tone type and group. Figure 2 6 depicts the mean percentages and the standard errors of the correctly identified productions by the native Chi nese judges for each tone across the four groups The mean percentages of correct productions for all participant were submitted to a 4 (tone) 4 (group) repeated measures ANOVA with tone as the within subject factor and group as the between subject facto r. The result yielded a main effect of tone [ F ( 3 168 ) = 374.41 p < 00 1], a main effect of group [ F ( 3 56 ) = 6.98 p < 00 1], and a significant interaction of tone group [ F ( 9 168 ) = 3.66 p = 00 2]. Pairwise comparisons presented that t he Chinese n on musicians outperf ormed the English

PAGE 51

51 vocalists ( p = .024 ), the English instrumentalists ( p = .004) and the English non musicians ( p = .001). No difference was found between the three English groups. For Tone1 and Tone4, all four groups performed equally w ell ( p s > .32). For T one 2, the Chinese non musicians only outperforme d the English non musicians ( p = .0 07 ). No difference was found between the three English groups. For Tone3 the Chinese non musicians were superior to the English vocalists the English instrumentalists and the English non musicians ( p s < .001). The production accuracy for Tone3 was low even for the native Chinese speakers. This was probably due to the tone assimilation effect during the stimulus manipulation process. As described in the previous section, the experimental stimuli this word). Because t has a falling tone, the rising portion of the Tone3 might be assimilated by the following Tone4 Therefore, w hen the Tone3 was presented in isolation in the imitation task, the participants were more likely to perceive it as Tone4, and in turn produced it as Tone4. Figure 2 7 demonstrates the perc entages of the productions of Tone3 that were identified as Tone4 by the judges Summary This experiment compared native Chinese non musicians, native English non musicians, instrumentalists and vocalists on the perception and production of Mandarin tones. We expected that the English musicians should be better than the English non musicians at both discriminating and imitating lexical tones. Moreover, the vocalists might have more advantage than the instrumentalists in imitating and discriminating lexical tones. Our results demonstrated that the English musicians were only better than

PAGE 52

52 the English non musicians at lexical tone discrimination but not at lexical tone imitation. Specifically, the English vocalists and instrumentalists were better than the Engl ish non musicians on the discrimination but not In addition, o ur results showed some indication that vocalists were better at discriminating lexical tones than instrumentalists (only for pair22 ). In term s of lexical tone production, the English musicians and non musicians produced Tone1 (level tone) and Tone4 (falling tone) as accurately as the native Mandarin speakers, while produced Tone2 (rising tone) and Tone3 (falling rising tone) less accurately tha n the native Mandarin speakers. Moreover, the English musicians did not differ from the English non musician in the production of Tone2 and Tone3.

PAGE 53

53 Table 2 1. Mean accuracy percentages of the tonal test, the rhythmic test and the composite performance in AMMA for the Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians Group Tone Rhythm Composite Chinese non musicians 67% 63% 66% English vocalists 60% 64% 59% English instrumentalists 60% 58% 58% Engli sh non musicians 64% 59% 62% Table 2 2. Mean p ercentage s of correct productions that were agreed upon by all 3 judges and 2 judges respectively for each tone type and group Group Agreed by 3 judges Agreed by 2 judges T1 T2 T3 T4 T1 T2 T 3 T4 Chinese non musicians 78% 74% 33% 89% 17% 20% 12% 10% English vocalists 69% 46% 1% 83% 25% 36% 10% 14% English instrumentalists 61% 48% 3% 68% 30% 31% 10% 24% English non musicians 68% 42% 4% 69% 19% 31% 10% 23% Figure 2 1. Waveforms and spectrograms with the pitch contours of the Mandarin Tone1, Tone2, Tone3 and Tone4 associated with the sy llable [ma] (Note: t he blue lines indicate the pitch contours of the tones )

PAGE 54

54 Figure 2 2. D Prime Scores with the standard errors for Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians respectively Figure 2 3. Mean reaction times with the standard errors of all the correct responses for the Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians respectively

PAGE 55

55 Figure 2 4 The mean accuracy percentages with the standard errors for the same pair s in the Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians respectively Figure 2 5. Mean accuracy p ercentages with the standard errors for each of the different pairs in the Chinese non musicians, the English vocalists, the English instrumentalists and the English non musicians

PAGE 56

56 Figure 2 6. Mean percentages and the standard errors of the correctly identified productions for each tone in the Chinese non musicians, the English vocalists, instrumentalist s and non musicians respectively Figure 2 (Note: t he tone before the dash refers to the target and the tone after the dash indicates the )

PAGE 57

57 CHAPTER 3 EXPERIMENT 2 : EFFECTS OF LABORATORY PERCEPIO N AND PRODUC TION TRAINING ON TONE DISCRIMINATION A BEHAVIORAL AND ERP STUDY Since the long term perception training and production training in the music domain did not show significant difference in improving the perception and production of lexical tone, we further examined the effects of perception and production training in the language domain. The second experiment collec ted both behavioral and electrophysiological (ERPs) data to in vestigate the effectiveness of a laboratory perception only training and a perception plus production training on the intentional and unintentional processing of tones by native English speakers. Methods Participants A total of 22 native speakers of Ameri can English were recruited in the present study. None of the participants had participated in the Experiment1. Participants were randomly assigned to two groups: 11 participants receive d only perception trai ning (perception only group); 11 participants rec eive d a perception plus product ion training (p erception plus production group) All participants (2 m ales and 20 females) were between the ages of 18 and 2 5 years old (M = 19.14 SD = 1.28 ), and were undergraduate students at the University of Florida. Non e of the participants had previous exposure to any tone languages. All participants were right handed according to Edinburgh Handedness inve ntory (Oldfield, 1971), had normal vision and had no history of speech or neurological impairment according to self report hearing and auditory working memory (forward and backward digit span) were tested on site as described in Chapter 2. The mean auditory working memory accuracy percentages for the perception only and the perception plus production grou ps were

PAGE 58

58 63.31% (SD = 11.64%) and 56.49% (SD = 11.50% ) respectively. There was no significant difference between the auditory memory scores for the two groups t ( 20) = 1.38, p > .18 ) Moreover, none of the participants had rece ived more than two years of f ormal music al training and had not been performing music within the past five years by the time of their participation was also tested on site by using the Advanced Measures of Music Audiation (AMMA, Gordon 1989) as describe d in Chapter 2 Table 3 1 presents the mean accuracy percentages of the tonal test, the rhythmic test and the composite performance in AMMA for the perception only and the perception plus production groups. The participants in the two group s did not differ in the musical tone perception [ t (2 0) = 0.62 p > 54], the music rhythm perception [ t (2 0) = 0.35 p > 73] or the composite performance [ t (2 0) = 0.60 p > 55] An additional four participants were recruited, but were excluded from data analysis due to excessive eye or body movement artifacts or technical problems. All participants receive d financial compensation or course credit for their participation. The experimental procedures were approved by the University of Florida Institutional Review B oard. Stimuli Behavioral task Stimuli consist ed of eight monosyllabic syllables ([p h a], [p h i], [k h h o], [t h a], [t h i], [t h h o]) associated with three linear tones that resemble Mandarin Tone1 (high level), Tone2 (high rising), and Tone4 (high f alling). Tone3 (low dipping) was not included because it has been shown to be the most confusable tone for both native Mandarin speakers and non native speakers (e .g. Kirkham et al., 2011). The s yllables [t h a], [t h i], [t h [t h o] [k h o] and [p h a] were use d in t he pre and post training tasks and the s yllables [p h a], [p h i], [k h [k h o] were used in the training ses sion The syllables [t h a],

PAGE 59

59 [t h i], [t h [t h o] were not i ncluded in the training session because they were used to test whether participa nts could generalize the improvements gained through training to untrained stimuli. All of these syllables exist in American Eng lish, thus American speakers could focus their attention on the lexical tones and not be distracted by unfamiliar syllables. Two female native speakers of American English were asked to produce each syllable twice with a high pitch in a sound attenuated booth yielding four tokens for each syllable Only female speakers were recorded for stimuli because adult female speakers have b een rated to be more intelligible than male adult speakers (Bradlow et al., 1996). Three linear pitch contours were superimposed starting from the voiced part of each syllable, using the pitch synchronous overlap and add (PSOLA) function in the Praat softw are (Boersma & Weenink, 2010 ) The procedures of stimulus manipulation were sim ilar to Wong & Perrachione (2007b). The onset value of Tone1 (T1) was the mean fundamental frequency (F0) of all syllables produced by the two speakers. The offset of T1 was ide ntical to its onset. Tone2 (T2) h ad the same ending point as T 1, and its starting point was 26% lower than its ending point. Originally the onset of Tone4 (T4) was set to 10% higher than T1 and dropped by 82%, according to the values of Mandarin T4 reporte d in Shih (1988). P ilot data of seven participants however, displayed a ceiling effect, i.e. there was little room for participants to improve through the training session. Then we decided to set the onset of Tone4 the same as Tone1 and its offset 26 % low er than its onset. Therefore, the onset of T2 was identical to the offset of T4. All stimuli were then normalized for intensity. Except for F0 and intensity, all other acoustic features (e.g. duration and voice quality characteristics) were kept identical As mentioned previously, each

PAGE 60

60 syllable was produced twice by two speakers, resulting in four tokens for each syllable. The three tones were generated for each token which means that (1) within each token the three to nes had the same acoustic features except for F0 ; (2) the same tones for different tokens had the same F0, but different duration and voice quality. All stimuli were judged as perceptually natural by three native Mandarin speakers and three native America n English speakers. A second pilot experiment with 8 participants was carried out to ensure that the discrimination of the tones in untrained native English speakers did not show a ceiling effect. Figure 3 1 shows the waveforms and spectrograms with the pi tch contours of the Tone1, Tone2 and Tone4 of one token of the syllable [t h u]. All the three sounds had the same duration, intensity and other acoustic features. ERP In the ERP experiment, the stimuli were the [t h u] syllable associated with the three tones (Tone1, Tone2 and Tone4) These stimuli were not included in the behavioral tasks or the training sessions. The stimulus recording and manipulation were exactly the same as described in the behavioral task section. The mean duration of all th e tokens was 553 millisecond (SD = 54.66). On average, pitch contour st arted at 108 ms (SD = 14.18) after the stimulus onset. The experiment consist ed of three types of blocks : in the first block T1 was the standard, and both T2 and T4 were the deviants; in the secon d block T2 was the standard, with T1 and T4 being the deviants; in the third block T4 was the standard, and T1 and T2 were the deviants (Ntnen, Pakarinen & Rinne, 2004). W ithin each block each token of the deviant s was presented 25 times resulting in 200 deviant stimuli (25 2 tokens/speaker 2 speakers 2 tones). The se deviant stimuli were 20% of the standard stimuli and were presented in a pseudo random order, with at least two

PAGE 61

61 standard stimul i preceding each deviant. Within each block there were also 100 clean standards (25 2 tokens/speaker 2 speakers), which were not immediately preceded or followed by any deviants. These clean standards were compared with the deviant stimuli in data analysis. The offset to onset inter stimulus interval was varied between 500 and 650ms to prevent Each block was split equally into two sub blocks, with each sub block lasting for about 12 minutes. The order of the six sub blocks was randomized across partic ipants with a different order being presented before and after the training session Before the experimental blocks, all participants also received a practice block, in which each token of the three tones was equally presented. The practice block which w as excluded from data analysis, lasted for about 10 minutes, in order to familiarize the participants with the experiment al setting. Procedure The experiment took place over the course of three consecutive days. All participants were asked to do the pre and post training tasks on the first and third days. On the second day participants either receive d a perception only training or a perception plus production training on tones. Note that testing production without perception is hardly possible; the two t raining types were therefore similar in terms of auditory exposure to the stimuli, but differ ed in the production component: the production training required participants to imitate the stimuli w hile the perception training had participants utter a word un related to the stimuli. In this way, the production training and perception training c ould be directly compared, with participants in both groups receiving the same exposure and attention to the same stimuli.

PAGE 62

62 Day 1 Baseline EEG data were recorded on the fi rst d ay from all participants in both groups. The ERP experiment use d the passive oddball paradig m, in which participants were watching a silent movie in a sound attenuating booth while the stimuli were presented binaurally through inserted ear buds at a c onstant and comfortable hearing level Participants were asked to ignore the sounds and focus their attention on the movie. After each block, they receive d comprehension questions regarding to the movie. In order to control the attention differences across groups, participants were also asked to count the number of times a particular object occurs in the movie. Only participants who answer ed more than 60% of the questions correctly, and were able to produce the correct count (plus or minus 3 ) were included for further data analysis. The movies in the pre and post training sessions were different. The order of the two movies was counterbalanced across participants, with half of the participants watch ing the first movie in the pre training session while the o ther half of the participants watch ing the first movie in the post training session. The ERP experiment was divided into six blocks according to the three conditions. The order of the blocks was randomized among participants. For each participant, the ord er of stimulus presentation in each block was the same before and after training. Participant s were encouraged to take break s after each block to maintain concentration during blocks and to prevent fatigue. After the ERP recording, participants did a behav ioral discrimination task. The discrimination task use d a forced choice AX paradigm, in which the stimuli (the syllables [t h a], [t h i], [t h [t h o] [k h o] and [p h a]) were presented in pairs and were separated by 500 ms inter stimulus interval The ISI was s et to 500ms because previous studies have

PAGE 63

63 shown that participants were sensitive to nonnative sound contrasts at the 500ms ISI ( e.g. Cowan & Morse, 1986; Pisoni, 1973; Van Hessen & Schouten, 1992 ) In the different pairs the two stimuli in each pair had tw o different tones which were generated from the same token of the same syllable produced by one speaker In the same pairs the two stimuli in each pair had the same tones, but were generated from different tokens of the same syllable produced by one speake r. The participants were asked to focus only on the tones and ignore other acoustic features of the stimuli and indicate whether the two stimuli in each pair had the same or different tone s by pressing a button on a button box. For half of the participants the leftmost button on the button box other half of the participants. This design ensured that the results were not biased by nce. A total of 288 trials (6 syllables) were presented, with 144 144 The order of the tria ls was randomized for each participant. The experimental trials were preceded by a short tutorial on the three tones and 6 practice tr ials which were excluded from data analysis The participants were encouraged to respond as quickly and accurately as possible. Their accuracies and reaction time s were logged by the E prime 2.0 software (Psychology Software Tools, Pittsburgh, PA) No feed back was given through this task. A break of at least 30 seconds was given halfway through the task, in order to prevent fatigue The ERP and behavioral discrimination tasks on the first day took about 4.5 hours for each participant.

PAGE 64

64 Day 2 On the second da y, t he participants receive d either a perception only or a perception plus production training, which took about one hour The structure of each trial is illustrated in Figure3 2. Perception only training Participants in the perception only group : (1) fi rst hear d one stimulus; (2) hear d another stimulus after 500ms; ( 3) determine d whether these two stimuli ha d the same or different tones; (4) received feedback; (5) hear d the first stimulus again and s aw the tone and the graphic indication of the tone sim ultaneously; (6) S aid d the second stimulus again and s aw the tone and graphic indication at the same time; and (8) Sa id ed three s production in (8) Perception plus production trai ning Participa nts in the production group receive d exactly the same training as participants in the perception only group, except that they were in steps (6) and (8) They were encouraged to imitate t he tones as accurately as they c ould and were allowed to try only once. In order to match the production training to the perception training the participants in the production group did not receive any feedback about their imitation s All p roductions were recorded by a Marantz PMD660 digital recorder for assessment of the were not used in data analysis. box was iden tical in the pre and post training tasks and the training session. Regardless of the training type a total of 192 trials (the syllables [p h a], [p h i], [k h [k h o] ) The

PAGE 65

65 order of the tria ls was randomized for each participant. The experimental trials were preceded by 6 practice trials which familiarized the participants with the training process and were excluded from data analysis. The participants were encouraged to respond as quickly and accurately as possible. Their accuracies and reaction times in step (3) were logged by the E prime 2.0 software (Psychology Software Tool s, Pittsburgh, PA). A break of at least 30 seconds was given halfway through the training session. After the 30 second break, the participants could take another break as long as they wanted and started the second half of the training session by pressing a ny button on the button box. Originally, we planned to conduct two sessions of training during the course of two consecutive days, with each session took about one hour. However, our pilot data on on abilities improved significantly even after the first day of training. Therefore, we decided to only include one training session This decision may result in some unexpected results, which will be discussed in Chapter 5. Day 3 On the third day, b oth g roups did the same ERP and behavioral tasks as described under Day1. After the participants finished all tasks, they were asked to self evaluate their performance and give suggestions on all the tasks. The entire experiment took about 10 hours for each par ticipant. Data Analysis Behavioral task Data obtained from the pre and post training discrimination task s and the training session were converted to d prime scores, which were calculated using the same formulas as described in Chapter 2. Separate d prime scores for the trained and

PAGE 66

66 untrained stimuli in the pre and post training tasks and the training sessions were also Besides the d prime he pre and post training tasks and the training session were also calculated to see whether the participants responded faster in the post training tasks than in the pre training tasks. The RTs were c ompute d only for the correct trials RTs that greater th an the mean RT of all the correct responses plus 2.5 standard deviatio ns for an individual participant were replaced by this value. After the outlier replacement, the mean RTs of the total stimuli, trained stimuli (the syllables [k h o] and [p h a]) and untrai ned stimuli (the syllables [t h a], [t h i], [t h [t h o] ) for each participant in the pre and post training tasks and the training session were computed. Furthermore, t he percentage of accurate responses for each tone pair type w as also calculated in order to explore which tone pair type showed th e greatest improvement and which type was the most resistant for improvement for each group. These behavioral data were submitted to several repeated measure ANOVAs, which will be described in detail in the result section. ERP Continuous EEG w as recorded from 39 Ag/AgCI scalp electrodes (Fp1/2, F7/8, F5/6, F3/4, Fz, FT7/8, FC5/6, FC3/4, FCz, T7/8, C5/6, C3/4, Cz, TP7/8, CP5/6, CP3/4, FPz, P7/8, P5/6, P3/4, Pz, O1/2) at 512 Hz. Horizontal eye movements were recorded by electrodes on the left and right outer canthi, while the vertical eye movements were recorded by electrodes placed above and below the right eye. Blinks and eye movements were rejected off line from averaging on the basis of the EOG. The EEG was referenced to the right mastoid, off line band pass filtered from 0.16 to 30 Hz, and arithmetically re referenced to the mean of both mastoids after recording. The

PAGE 67

67 continuous EEG was divided into epochs of 900 ms afte r the stimulus onset plus a 200 ms pre stimulus baseline. Table 3 2 presents the numbers and the percentages of trials that were included in data analysis after artifact rejection in the pre and post training tasks for the perception only and perception plus production groups respectively Artifact free epochs were analyzed from 2 00 to 900 ms relative to the onset of the stimulus, using the 2 00 ms window preceding the onset as a baseline. The grand averaged difference waves were generated for each of the six conditions by subtracting the average responses to the clean standard stimuli from av erage responses to the corresponding deviant stimuli. The six conditions were (1) deviant T1 in standards T2 (T12); (2) deviant T1 in standards T4 (T14); (3) deviant T2 in standards T1 (T21); (4) deviant T2 in standards T4 (T24); (5) deviant T4 in standard s T1 (T41); and deviant T4 in standards T2 (T42). The mean amplitude of the time windows spanning the P2 the MMN and the late negativity of the difference waves were computed for each channel, participant and condition in the pre and post training tasks Time windows for these components were defined based on visual inspection of the raw ERPs and the literature. The 100 200ms window and the 200 400 ms window encompass the P2 and the MMN respectively. The mean amplitude of the late negativity was comput ed between 500 and 800 ms after stimulus onset. All these data were submitted to several repeated measures ANOVAs which will be described in detail in the result section. According to the literature (e.g. Roberts & Russo, 1999), all the p values and the F values were adjusted using the Greenhouse Geisser correction and the post hoc paired t tests were adjusted using the Bonferroni correction for multiple comparisons.

PAGE 68

68 The uncorrected degrees of freedom were reported in both the ANOVA tests and the t tests. The result was considered statistically significant when the p value was less than .05, and was considered as marginally significant when the p value was between .05 and .10. Predictions If production indeed facilitates perception, as implied by the Motor Theory (e.g. Liberman & Mattingly, 1989), we should find that the participants who receive d perception plus production training outperform ed the participants in the perception only group in the post training tasks. For the behavioral task, the improvement from pre training to post training should be more pronounced in the perception plus production group than in the perception only group. If the additional production during the perception plus production training had an effect on unintentional auditory proc essing of the tones, the perception plus production group should demonstrate a larger MMN. Both groups may also reveal increased P2 amplitude after training, which has been found in several studies (e.g. Atienza et al., 2002). However, P2 was expected to b e smaller for the perception only group than for the perception plus production group after training. Moreover, both groups might demonstrate a larger late negativity in the post training task, because previous studies have show n that late negativity incr eased with more exposure to the auditory stimuli (e.g. Shest akova et al., 2003) T his late component has been associated with neural processes of auditory rule extraction (e.g. Zachau et al., 2005). To sum up, if participants in the perception plus product ion group outperformed participants in the perception only group after training, it support ed the claim that perception relied on production (e.g. Liberman et al., 1989; Fowler, 1986). If the

PAGE 69

69 advantage of production training only occu rred in the ERPs, this indicate d that: (1) the short term production training affected the unintentional perception, but neither of the trainings was long enough to affect the behavioral measures; or (2) the production training did not additionally improve attentive perception. Follow up studies would be needed to tease these two possibilities apart. However, if the perception only group yielded better results after training than the perception plus production group, the additional production training d id not enhance perception, which was contrary to the Motor Theory. Results Behavioral Task Accuracy Mean d prime scores with the standard errors of the pre and post training tasks and the training session for the perception only and the perception plus production groups are presen ted in Figure 3 3. The mean d prime scores of the pre training task, training session and the post training task were 1.79 (SD = 0.94), 2.55 (SD = 1.49) and 2.52 (SD = 1.25) for the perception only group and 2.11(SD = 1.12), 2.55 (SD = 1.38) and 2.99 (SD = 1.59) for the perception plus production group respectively. prime scores for the trained and untrained stimuli in the pre and post training task s were calculated separately for the perception only and the percepti on plus production groups in order to examine whether the partic ipants could generalize the improvements gained through the trained stimuli to the untrained stimuli The trained stimuli included the syllables [k h o ] and [p h a], which were used in the pre tra ining task, the training session a nd the post training task. T he untrained stimuli were the syllables [t h a], [t h i], [t h h o] which were only tested in

PAGE 70

70 the pre and post training tasks. In addition, the untested syllables [k h h i] were only u sed in the training session. Since the untrained stimuli were not included in the training session, we first focused our analysis on the mean d prime scores for the trained and untrained stimuli in the pre and post training tasks. Figure 3 4 illustrates t he mean d prime scores with the standard errors of the trained and untrained stimuli in the pre and post training tasks for both groups. T he mean d prime scores for the trained stimuli in the pre and post training task s were 1.74 (SD = .91) and 2.53 (SD = 1.35) for the perception only group, and 1.98 (SD = 1.05) and 2.88 (SD = 1.34) for the perception plus production group. For the untrained stimuli the mean d prime scores in the pre and post training task s were1.84 (SD = .98) and 2.54 (SD = 1.22) for t he perception on ly group, and 2.16 (SD = 1.08) and 2.95 (SD = 1.55) for the perception plus production group. The mean d prime scores for each participant were submitted to a 2 (test time: pre and post training) 2 (stimulus type: trained and untrained) 2 (group) repeated measures ANOVA, with test time and stimulus type as the within subject factors and group as the between subject factor. The result yielded a main effect of test time [ F (1, 20) = 53.51, p < .001] Paired samples t tests showed that bo th groups performed more accurately in the post training task than in the pre training task as a result of training [ t (10 ) = 6.04 p < .001 and t (10 ) = 4.60. p = .001] However, the test time group interaction was not significant [ F (1, 20) = .22, p > .64], suggesting that the two groups did not differ from each other in the extent of improvements. There was no difference between the mean d prime scores for the trained and untrained stimuli [ F (1, 20) = 2.94,

PAGE 71

71 p > .10] indicat ing that the participants generalized the improvements gained through the trained stimuli to the untrained stimuli. Moreover, prime scores for the tested ( [k h o ] and [p h a]) and untested ( [k h h i] ) stimuli in the training session were calculated separately for the perception only and the perception plus production groups Figure 3 5 illustrates the mean d prime scores with the standard errors of the tested and un tested stimuli in the tr aining session for both groups. The mean d prime scores of the test ed and untest ed stimuli were 2.44 (SD = 1.55) and 2.60 (SD = 1.27) for the perception only group, and 2.71 (SD = 1.38 ) and 2.37 (SD = 1.30 ) for the perception plus production group. The mea n d prime scores for each participant were submitted to a 2 (stimulus type: t est ed and unt est ed) 2 (group) repeated measures ANOVA, with stimulus ty pe as the within subject factor and group as the between subject factor. The result yielded a significant interaction of stimulus type group, F (1, 20) = 4.46 p = .047. Further independent samples t tests showed that the two groups did not differ in the discrimination of the tested stimuli [ t (20 ) = 0.44 p > .66] or the untested stimuli [ t (20 ) = .42 p > .68] In addition, paired samples t tests revealed that the perception only group discriminated the tested and untested stimuli equally well during the training session, t (10 ) = 0.91 p > .39 However, for the perception plus production group the diff erence between the tested and untested stimuli almost reached significance [ t (10 ) = 2.14 p = .0 58 ], with the discrimination of the tested stimuli being more accurate than that of the untested stimuli. This was probably due to that the generalization abil ity was inaccurate productions in the perception plus production group.

PAGE 72

72 In sum, both the perception only and the perception plus production groups improved on the tone discrimination abilities in the post training task compar ed with the pre training task. Moreover, both groups generalized the improvements gained through the train ed stimuli to the untrained stimuli During the training session, the perception only group discriminated the tested and untested stimuli equally well while the perception plus production group discriminated the tested stimuli more accur ately than the untested stimuli Reaction times stimulus in each trial and were rounded to f ull milliseconds. Mean reaction times with the standard errors of all the correct responses in the pre training task, the training session and the post training task for the perception only and production plus perception grou p are illustrated in Figure 3 6 The mean reaction times in the pre training task, the training session and t he post training task were 1317ms (SD = 321), 1762ms (SD = 620) and 1369 ms (SD = 539 ) for the perception only group, and 1539ms (SD = 357), 1904ms (SD = 395) and 1492 ms (SD = 336 ) for the perception plus production group. The reaction times in the training sessions were longer than in the pre and post traini ng tasks. This was probably because that the participants responded more carefully/slowly in a clear training situation in w hich they received feedback after they made their decisions. In addition, t he training session took about one hour, which was much longer than the pre and post training tasks. Therefore, the p articipants might get tired of maintaining concentration during the training session resulting in slower response times. In order to get rid of the this factor we performed further a nalysis first on the reaction times in the pre and post training tasks to test whether participants

PAGE 73

73 discriminated tones more quickly in the post training task than in the pre training task as a result of training. The mean response times for the trained and untrained stimuli in the pre and post training tasks were calculated separately for the perception only and the pe rception plus prod uction Figure 3 7 illustrates the mean response times with the standard errors of the trained an d untrained stimuli in the pre and post training task s for both groups. For the trained stimuli, the mean reaction times in the pre and post training tasks w ere 1 343 (SD = 359) and 1369 ms (SD = 525 ) for the perception only group, and 1 538 ms (SD = 392 ) and 1 530 ms (SD = 362 ) for the perception p lus production group. For the untrained stimuli, the mean reaction times in the pre and post training tasks were 1 304 (SD = 305) and 1369 ms (SD = 525) for the perception only group, and 1540 ms (SD = 348) and 1472 ms (SD = 331) for the perception plus production group. The reaction times for each participant were submitted to a 2 (test time : pre and post training ) 2 (st imulus type : trained and untrained ) 2 (group) repeated measures ANOVA with test time and stimulus type as the within subject factors and group as the between subject factor. The results showed that there was a significant interaction of test time stim ulus type group, F (1, 20) = 5.26 p = .033. No other significant effect or interaction was found. Further paired samples t tests on the trained and untrained stimuli did not show any significant difference between the reaction times in the pre and post training tasks for either group. Additionally, the mean response times for the tested and untested stimuli in the training session were also calculated separately for the perception only and the perception plus production groups Figure 3 8 illustrates t he mean response times with

PAGE 74

74 the standard errors of the tested and untested stimuli in the training session for both groups. The mean reaction times of the tested and untested stimuli in the training session were 1 755 ms (SD = 577 ) and 1770 ms (SD = 666 ) for the perception only group, and 1 921 ms (SD = 432 ) and 1 883 ms (SD = 3 74 ) for the perception plus production group. The reaction times for each participant were submitted to a 2 (stimulus type: t est ed and unt est ed) 2 (group) repeated measures ANOVA, with st imulus type as the within subject factors and group as the between subject factor. The result did not yield any significant effect or interaction. In sum, neither the perception only training nor the perception plus production training made the participant s discriminate tones more quickly in the post training task than in the pre training task. Moreover, there was no difference between the reaction times of the trained /tested stimuli and the untrained /untested stimuli for the two groups. Percentage accurac ies of different tone pairs Percentage s of accurate discriminations for each tone pair (Tone1 Tone2: pair12; Tone1 Tone4 :pair14; Tone2 Tone4: pair24; Tone1 Tone1: pair11; Tone2 Tone2: pair22; Tone4 Tone4: pair44) were calculated collapsed over the order of presentation within each trial for each participant Analys e s were only performed on the percentage accuracies in the pre and post training tasks, because (1) the training session had fewer tr ia ls than the pre and post training tasks, with 192 tr ia l s in the training session and 288 in the pre and post training tasks; (2) the syllables used in the training session were different from the syllables used in the pre and post training tasks. Therefore, we only focused on the pre and post training tasks in order to make direct comparisons between the perception only and the perception plus production groups.

PAGE 75

75 Figure 3 9 depicts the mean percentage accuracies of each tone pair in the pre and post training tasks for the perception only and perception plus production groups. The percentage accuracy data were submitted to a 2 (test time: pre and post training task) 6 (pair) 2 (group) repeated measures ANOVA with test time and pair as the within subject factors and group as the between subject factor. The result only yielded a main effect of test time, F ( 1 20 ) = 10.03 p = 00 5 indicating that the participants discriminated the tone pairs more accurately in the post training task than in the pre training task. Even though the perception only group shows more improvement in Figure 3 9, the repeated measures ANNOVA did not yield any significant effect or interaction with group. The percentage accuracies for each tone pair in the training session were also computed for the perception only and the perceptio n plus production groups respectively Figure 3 10 presents the mean percentage accuracies with standard errors of each tone pair in the training session for both groups. The percentage accuracies in the training session were submitted to a 6 (pair) 2 (g roup) repeated measures ANOVA with pair as the within subject factor and group as the between subject factor. The result yielded a marginally significant interaction of pair group F ( 5 10 0 ) = 2 47 p = 081. Further independent samples t tests only sho wed a marginally significant difference for the pair44 [ t (20 ) = 1.75 p = .0 95], indicating that the perception plus production group discriminated pair44 more accurately than the perception only group during the training session. Summary of behavioral results In the behavioral tasks, both the perception only and the perception plus production groups improved on the tone discrimination abilities as a result of training

PAGE 76

76 Moreover, the participants in both groups generalized the improvements gained through the trained stimuli to the untrained stimuli after training However, d uring the training session, the perception only group discriminated the tested and untested stimuli equally well while the perception plus production group discriminated the tested sti muli more accurately than the untested stimuli. This was probably due to that during the perception plus production training their generalization skills. In terms of reaction times, neither of the two group s discriminate d tones more quickly in the post training task than in the pre training task. Moreover, there was no difference between the two groups in the reaction times of the trained/tested stimuli and the untrained/untested stimuli. ERP Figure 3 11 pr esent s the ERPs of the frontal electrodes (F3, Fz and F4), the central electrodes (C3, Cz and C4) and the parietal electrodes (P3, Pz and P4) collapsed over all six conditions in the pre and post training tasks for the perception only and perception plus production groups respectively. Descriptively, the ERPs for both groups showed a large P2 component, which was especially visible at the frontal and central electrodes between 100 and 200 m illisecond s. Moreover, a MMN component was elicited for the deviant versus standard stimuli, with a prominent frontal distribution between 200 and 40 0 m illisecond s. A late negativity component was also observed at the frontal and central electrodes between 500 and 800 m illisecond s. F urther analyses will focus on these th ree components. Figure 3 1 2 presents the ERPs of the Fz electrode for the six conditions in the pre and post training tasks for the perception only and perception plus production groups respectively. Figure 3 13 show s

PAGE 77

77 the deviant minus standard difference waves of the Fz electrode for the six conditions in the pre and post training tasks for the two groups. 100 200ms window According to the literature, we expected that the participants in both groups should show an increased P2 as a result of training ( e.g. Atienza et al., 2002) The mean amplitude s of the difference waves between 100 and 200ms were computed for four sites: right frontal (RF) included the electrodes F4, F6, F8, FC4 FC6 and FT8 ; left frontal (LF) included the electrodes F3, F5, F7, FC3, FC5 and FT7; right p osterior (RP) included the electrodes CP4, CP6, TP8, P4, P6 and P8; left p osterior (LP) included the electrodes CP3, CP5, TP7, P3, P5 and P7. The mean amplitude differences were submitted to a 2 (test time) 2 (hemisphere: right and le ft ) 2 (anteriority: anterior and posterior) 6 (condition: T12, T14, T21, T24, T41, T42) 2 (group) repeated measures ANOVA with test time, hemisphere, anteriority and condition as the within subject factors and group as the between subject factor. Th e result yielded significant interaction s of condition group [ F (5, 100 ) = 2.78 p = .048 ] and test time hemisphere anteriority group [ F (1, 20) = 8.53 p = .008] Since the interaction of condition group was significant, further repeated measure s ANOVAs were performed for each condition separately, with test time, hemisphere and anteriority as the within subject factors and group as the between subject factor. We only focused on the significant intercept, main effect of group and significant inte raction of test time group. For the T 12, T24, and T41 conditions n either the intercept nor the group effect was significant. For the T14 condition there was a main effect of group, F (1, 20) = 6.13 p = .022. The percep tion only group showed a larger po sitivity for the deviants versus standards than the perception plus production group For the T21 condition the

PAGE 78

78 intercept was significant [ F (1, 20) = 4.88 p = .041] which suggested that the ERPs for the deviants were significantly more negative than t hose for the standards Moreover, there was a significant interaction of test time group [ F (1, 20) = 5.73 p = .027] For the T42 condition the effect of group was marginally significant, F (1, 20) = 3.40 p = .080. The perception plus production group revealed a larger positivity for the deviants versus standards than the perception only group. However, a s described previously the pitch contours of the stimuli started approximately at 108ms after the stimulus onset ; from Figure 3 13 we saw that the di fference s between the two groups started at or right after the stimulus onset. Therefore, these effect and interaction of group was probably due to some different acoustic features of the stimulus onset rather than the pitch difference. MMN Figure 3 12 dem onstrates that the MMN component was more prominent in the T 21, T24 and T41 conditions while less prominent or absent in the T12, T14 and T42 conditions. Moreover, previous literature has shown that the MMN is largest over the frontal electrodes (e.g. Nt nen & Michie, 1979; Alain, Woods & Knight, 1998). Figure 3 11 also presents that the MMN was larger at the frontal electrodes than at the parietal electrodes. Therefore, our analyses on the MMN will focus on the frontal electrodes. T he mean amplitudes of the difference waves b etween 200 and 4 00ms were computed for two si tes: right frontal (RF) included the electrodes F4, F6, F8, FC4 FC6 and FT8 ; left frontal (LF) i ncluded the electrodes F3, F5, F7, FC3, FC5 and FT7 The mean amplitudes were submitted to a 2 (test time) 2 (hemisphere: right and left) 6 (condition: T12, T14, T21, T24, T41, T42) 2 (group) repeated measures ANOVA with test time, hemisphere and condition as the within subject factors and group as the

PAGE 79

79 between subject factor. The results sh owed that there were main effects of test time [ F (1, 20) = 6.68 p = .018], and condition [ F ( 5 100 ) = 5.13 p = .0 05 ] The MMN was smaller in the post training task than in the pre training task. In addition, the intercept was also significant [ F (1, 20 ) = 13.66 p = .001], which means that the negativity for the deviants was significantly larger than that for the standards Six 2 (test time) 2 (hemisphere) 2 (group) repeated measures ANOVA s were performed for each condition with test time and hemis phere as the within subject factor and group as the between subject factor. No significant main effect or interaction was found for the T12, T14 and T24 conditions. T he intercept s for the T21 [ F (1, 20) = 9.19 p = .007] T41 [ F (1, 20) = 36.63 p < .001] and T42 [ F (1, 20) = 4.33 p = .050] conditions w ere significant, indicating that the negativity for the deviants was significantly larger than that for the standards in these three conditions No other main effect or significant interaction was found in t he T21 and T42 conditions In the T41 condition there was a significant interaction of hemisphere group [ F (1, 20) = 5.80 p = .026] The MMN was larger over left than right frontal sites in the perception only group and vice versa in the perception p lus production group. Late negativity Figure 3 21 illustrated that the late negativity for the perception only group increased in the post training task compared to the pre training tasks for the T12, T14, T21, T24 and T41 conditions. However, the late ne gativity for the perception plus production group decreased or remained the same in the post training task compared to the pre training tasks. The mean amplitudes of the difference waves between 500 and 8 00ms were computed for the four regions : RF LF RP and LP These mean amplitudes were

PAGE 80

80 submitted to a 2 (test time) 2 (hemisphere: right and left) 2 (anteriority: anterior and posterior) 6 (condition: T12, T14, T21, T24, T41, T42) 2 (group) repeated measures ANOVA with test time, hemisphere, anteri ority and condition as the within subject factors and group as the between subject factor. The result yielded a significant intercept [ F (1, 20) = 15.84 p = .001] a main effect of anteriority [ F (1, 20) = 10.64 p = .004] and significant interactions of hemisphere group [ F (1, 20) = 6.37 p = .020], test time anteriority [ F (1, 20) = 9.85 p = .005] and test time hemisphere anteriority condition [ F (2.98, 59.60) = 2.88 p = .044]. Moreover, t he interactions of test time hemisphere group [ F (1, 20) = 3.71 p = .069 ] hemisphere anteriority group [ F (1, 20) = 4.10 p = .056] and hemisphere anteriority condition group [ F (3.59, 71.76) = 2.33 p = .071] were marginally significant. To investigate the source of the interactions, two 2 (test time) 2 (hemisphere) 6 (condition) 2 (group) repeated measure ANOVA s were performed separately on the frontal electrodes (F3/4 F5/6, F7/8, FC3/4, FC5/6 and FT7/8) and p osterior electrodes (P3/4, P5/6, P7/8, CP3/4, CP5/6 and TP7/8) Analysis on the frontal electrodes revealed a significant intercept [ F (1, 20) = 18 62 p = .000 ] and a significant interaction of hemisphere group [ F (1, 20) = 6. 56 p = .0 19 ] The negativity was larger over the left hemisphere in the perception only group, but was larger over the right hemisphere in the perception plus production group. Analysis on the posterior electrodes yielded a significant intercept [ F (1, 20) = 4.88, p = .039] and a main effect of test time [ F (1, 20) = 4.57, p = .045] with an increased negativity after training. Summary of ERP results In the 100 200ms window the group effect was not significant for the T 12, T21, T24, and T41 conditions For the T14 condition t he perception only group showed a

PAGE 81

81 larger positivity for the deviants versus s tandards than the perception plus production group. For the T21 condition the ERPs for the deviants were significantly more negative than those for the standards. For the T42 condition t he perception plus production group revealed a marginally larger po sitivity for the deviants versus standards than the perception only group. However these significant effect and interaction with group w ere probably due to some different acoustic features of the stimulus onset r ather than the pitch difference s, since the different effects started before the onset of the pitch contours. In the 200 400ms window, t he MMN was smaller in the post training task than in the pre training task. Moreover, for the T41 condition the two groups differed in the lateralization of the MMN component, with the perception only group showing a larger MMN at the left frontal electrodes while the perception plus production group displaying a larger MMN at the right frontal electrodes. Interpretations of these MMN results will be discussed in Chapter 5. In the 500 800ms window, both groups showed an increased late negativity at the posterior electrodes in the post training task compared to the pre training task. This increased late negativity might suggest that both the perception only and the perception plus production groups improved at the unintentional level of tone abstraction as a result of training. In sum, analyses on the ERP components did not show any significantly different training effects between the perception only and the perce ption plus production groups.

PAGE 82

82 Table 3 1. Mean accuracy percentages of the tonal test, the rhythmic test and the composite performance in AMMA for the perception only and perception plus production groups Group Tone Rhythm Composite Perception only group 48% 43% 44% Perception plus production group 54% 45% 49% Table 3 2 The numbers and the percentages of trials that were included in data analysis after artifact rejection in the pre and post training tasks for the perception only and perception plus production groups respectively Trials Perception only Perception plus production Pre training Post training Pre training Post training T12_dv 58 (56%) 61 (59%) 56 (54%) 68 (66%) T14_dv 62 (55%) 61 (54%) 64 (57%) 72 (64%) T1_cst 69 (61%) 66 (58%) 62 ( 55%) 74 (65%) T21_dv 71 (61%) 65 (56%) 65 (56%) 71 (62%) T24_dv 62 (54%) 61 (53%) 63 (55%) 73 (64%) T2_cst 65 (56%) 66 (58%) 60 (53%) 72 (63%) T41_dv 63 (57%) 64 (57%) 61 (54%) 68 (60%) T42_dv 68 (55%) 69 (56%) 65 (52%) 79 (64%) T4_cst 61 (54%) 62 (5 4%) 61 (53%) 73 (64%) Figure 3 1. Spectrograms of Tone1, Tone2 and Tone4 for the syllable [t h u]. (Note: t h e mean duration of the stimuli was 553 millisecond s ; t he blue lines indicate the pitch contours of the tones; o n average, the pitch contour s started at 108 millisecond s (SD = 14.18) after the stimulus onset )

PAGE 83

83 Figure 3 2. Step illustration of the perception only training and the perception plus production training Figure 3 3. Mean d prime scores with the standard errors of the pre training task, training session and the post training task for the perception only and the perception plus production group respectively

PAGE 84

84 Figure 3 4. Mean d prime scores with the standar d errors for the trained and untrained stimuli in the pre and post training tasks for the perception only and the perception plus production group respectively Figure 3 5. Mean d prime scores with the standard errors for the tested and untested stimuli in the training session for the perception only and the perception plus production group respectivel y

PAGE 85

85 Figure 3 6. Mean reaction times with the standard errors of all the correct responses in the pre training task, the training session and the post training task for the perception only and production plus perception groups respectively Figure 3 7. Mean reaction times with the standard errors of the trained and untrained stimuli in the pre and post training tasks for both groups respectively

PAGE 86

86 Figure 3 8. Mean reaction times with the standard errors of the tested and untested stimuli in the training session for the perception only and the perce ption plus production groups Figure 3 9. Mean percentage accuracies with standard errors of e ach t one pair in the pre and post training tasks for the perception only and pe rception plus production groups

PAGE 87

87 Figure 3 10. Mean percentage accuracies with standard errors of each tone pair in the training session for the perception only and the perception plus production groups

PAGE 88

88 Figure 3 11. The ERPs of the frontal electrodes (F3, Fz and F4), the central electrodes (C3, Cz and C4) and the parietal electrodes (P3, Pz and P4) co llapsed over all six con ditions. A) P re training task for the perception only group B) Post training task for the p erception only group. C) Pre training task for the perception plus production group. D) Post training task for the percept ion plus production group

PAGE 89

89 Figure 3 11 Continued

PAGE 90

90 Figure 3 11 Continued

PAGE 91

91 Figure 3 11 Continued

PAGE 92

92 Figure 3 12. The ERPs of the Fz electrode in the pre and post training tasks for the perception only and perception plus production groups respectively. A) T12 condition. B) T14 condition. C) T21 condition. D) T24 condition. E) T41 condition. F) T42 condition

PAGE 93

93 Figure 3 12 Continued

PAGE 94

94 Figure 3 12 Continued

PAGE 95

95 Figure 3 13 Deviant minus standard difference waves of the Fz electrode for the six conditions in the pre and post training tasks for the two groups

PAGE 96

96 Figure 3 13 Continued

PAGE 97

97 CHAPTER 4 EXPERIMENT 3: COMP ARISON OF PERCEPTION TRAINING AND PRODUCTION TRAINING ON TONE IDENTIFICATION This experiment is a follow up study of the second experiment. The second experiment was a little biased toward the perception plus production group, because the imitation task in the perception plus production training probably required the participants to pay more attention to the stimuli than the production task in the perception only training. The third e xperiment was designed to be biased toward the perception only group. If t he perception plus production group could show more improvement than the perception only group, it indicated that the perception plus production training was more effective than the perception only training and the employment of the articulatory system ben efits the perceptual skills. In this experiment, a ll participants were trained and tested on tone identification, instead of tone discrimination, over the course of two consecutive days. Participants were also divided into two groups: in the identification only group the participants only received an identification training, while in the identification plus imitation gro up the participants received an identification training plus an imitation of the stimulus The trainings were exactly the same except that the identification plus imitation training required participants to imitate the stimuli, while the identification only training had participants utter the tone type of the stimuli (i.e. level, rising or falling). Methods Participants Twenty native speake rs of American English were recruited in the current experiment, with 10 being assigned to the identification only group and the other 10 in the identification plus imitation group. All participants (3 males and 17 females) were

PAGE 98

98 between the ages of 19 and 23 years old (M = 20.05, SD = 1.12), and were undergraduate students at the University of Florida. None of the participants had previous exposure to any tone l auditory working memory and music abilities were tested on site as described in Chapter 2. The mean accuracy percentages of the auditory working memory test were 58.57% (SD = 14.70% ) and 53.57% (SD = 9.52% ) for the perception group and the imitation group respectively. The two groups did not differ from each other in t he auditory working memory test t (18) = 90 p > .37 Table 4 1 presents the mean accuracy percentages of the tonal test, the rhythmic test and the composite performance in AMMA for the identification only and the identification plus imitation groups. Th e participants in the two groups did not differ in the musical tone perception [ t (18 ) = 0.97 p > 34], the music rhythm perception [ t (18 ) = 0.16 p > 87] or the composite performance [ t (18 ) = 0.75 p > 46]. None of the participants had particip ated in the first or the second experiment Other recruitment criteria were exactly the same as described in Experiment 2 An additional 4 participants were recruited, but were dropped from the analysis either because they had displayed ceiling effect (per centage accuracy > 98%) in the pre training task or because they did not meet the recruitment criteria. All participants received course credit for the participation. The experimental procedures were approved by the University of Florida Institutional Revi ew Board. Stimuli Twelve monosyllabic syllables ( [k h a ], [k h ] [k h i], [k h o ] [p h a], [p h [p h i], [p h o], [t h a], [t h i], [t h h o]) associated with three linear tones (Tone1, Tone2 and Tone4) served as stimuli in the current experiment Four more sylla bles ( [k h a], [k h i], [p h h o], ) were included in the current experiment than in the Experiment 2 to ensure that the

PAGE 99

99 training sessions had enough trials and were long enough to show the training effect. Except for these four syllables, other stimuli w ere also used in Experiment 2. The syllables [k h [k h o ] [p h a], [p h i], [t h a], [t h i], [t h h o] were used in the pre and post training tasks, and the syllables [k h a ], [k h [k h i], [k h o ] [p h a], [p h h i] and [p h o] were used in the training sess ion. The syllables [t h a], [t h i], [t h [t h o] were not included in the training session in order to test whether participants could generalize the improvements gained through the trained stimuli to the untrained stimuli. All of these syllables exist in American Eng lish, thus American speakers could focus their attention on the lexical tones and not be distracted by unfamiliar syllables. The stimulus recording and manipulation were exactly the same as described in Chapter 3. Procedure The experiment too k place over the course of two consecutive days. On the first day, participants first did an identification task, in which they heard one stimulus and then identified whether this stimulus had a level, rising or falling tone by pressing the corresponding b utton on a button box The button box was connected to the computer automatically triggered the next tr ial with a 3 000 ms inter trial interval. The order of the trials was randomized by the E prime 2.0 software (Psychology So ftware Tools, Pittsburgh, PA). For all the participants, the leftmost button Before the real experiment, part icipant received a short tutorial on the three tones through seeing the tone (F0) graphs associated with the three tone types. After the tutorial there were six practice trials with the syllable [p h u], which ensured that the participants were clear about t heir task in th e experiment. These practice tr i a ls were later dropped from the data analysis.

PAGE 100

100 During the tutorial and the practice tr ia ls, the experimenter stayed in the experimental room, so that the participants could ask any question they had. A total of 96 (8 syllables 3 tones 2 speakers 2 tokens/speaker) trials were presented in this identification task. automatically. No feedback was given through this task. The ident ification task took about fifteen minutes for each participant. After the identification task, the participants received either an identification only or an identification plus imitation training which took about an hour The trainings were exactly the s ame except that the identification plus imitation training required the participants to imitate the stimuli; while the identification only training had the participants say the tone types of stimuli (i.e. level rising or falling ). Figure 4 1 illust rates the structure of each trial in the identification only and the identification plus imitation training respectively. Identification only training. The p articipants in the identification only group: (1) first heard one stimulus; (2) identifi ed whether the stimulus had a level, rising or falling tone; (3 ) received feedback and saw the tone graph associat ed with the correct answer; (4 ) heard the stimulus again and saw the tone and the graphic indication of the tone simultaneously; (5) Said the type of the tone (i.e. level rising or falling ) The next trial started three seconds after th ). Identification plus imitation training. The p articipants in the identification plus imitation group received exactly the same training as the participants in the identification only group, except that they were asked to imitate the stimulus instead of saying the tone type in step (5 ) The participants were encouraged to imitate the tones

PAGE 101

101 as accurately as they could, and were onl y allowed to produce once. In order to match the identification plus imitation training to the identification only training, the participants in the identification plus imitation group did not receive any feedback about their production s. All participants for assessment o T hese production data were not analyzed Regardless of the training type, a total of 192 trials was presented in the training session, which w as divided into two blocks. Each block had 96 trials (8 syllables 3 tones 2 speakers 2 tokens/speaker). As mentioned in the previous section, only the syllables [k h a ], [k h [k h i], [k h o ] [p h a], [p h h i] and [p h o] were used in the training session. The trials in both blocks were identical, but the orders of the trials were randomized in each block by the E prime 2.0 software (Psychology Software Tools, Pittsburgh, PA). A break of at least 30 seconds was given after the first block in order to prevent fatigue. On the second day, all participants did the same identification task as on the first day. The participants were encouraged to respond as quickly and accurately as possible in all the tasks. After the participants completed all the tasks, they were asked to fill out a debriefing questionnaire including their self evaluations of their performances in the tasks and their suggestions. The entire experiment took abo ut 2 hours for each participant Data Analysis T he percentages of correct identification for each participant in the pre training task, the training session and the post training task were computed. Then these percentages were converted to d prime scores acco rding to the Table

PAGE 102

102 for m interval forced choice or identification in Macmillan & Creelamn (1991, p. 319). The percentage of correct identification equaled 100% were replaced by 99% which was the maximum percentage that could be found in the Table A5.2 Therefore, the maximum d prime score a participant could obtain was 3.62. This procedure for calculating d prime scores was different from the procedures described in Experiment 1 &2 because there was no existing formula that could be used identification responses in the literature. different for the discrimination and identification paradigm s some researchers claimed that the d prime scores could be compared direc tly (e.g. Keating, 2004). Furthermore, the percentages of correct identification for the trained stimuli (i.e. the syllables [k h [k h o ] [p h a] and [p h i]) and the untrained stimuli (i.e. the syllables [t h a], [t h i], [t h h o] ) in the pre training and the post training tasks were calculated and were converted to d prime scores separately. Besides the d times (RTs) for the correct responses in the pre training task, the training session and the post trainin g tasks were computed T he RT outliers that were larger than the mean RT of all the correct responses plus 2.5 standard deviations for each participant were replaced by this value. total sti muli, the trained stimuli and the untrained stimuli in the pre training task, the training session and the post training task were calculated separately in order to see whether the participants responded faster in the post training task than in the pre tr aining task as a result of training. Moreover, the percentages of correct identification for each tone were calculated separately in order to investigate which tone had the greatest improvement and which tone was the most resistant for improvement.

PAGE 103

103 All th e data were then submitted to several repeated measures ANOVAs and paired t tests (when applicable) which will be described in more detail in the Results section below. T he p values were adjusted using the Greenhouse Geisser correction, and the p ost hoc pa ired t tests were adjusted using the Bonferroni correction for multiple comparisons. The result was considered statistically significant when the p value was less than .05. If the p value was between .05 and .10, the result was described as marginally sign ificant. Results Accuracy Mean d prime scores with the standard errors for the identification only and the identification plus imitation groups in the pre training task, the training session and the post training task are illustrated in Figure 4 2 The me an d prime scores of the pre training task, the training session and the post training task were 1.30 (SD = 0.98), 2.04 (SD = 1.05) and 2.11 (SD = 1.11) for the identification only group and 1.35 (SD = 0.73), 2.33 (SD = 0.77) and 2.34 (SD = 0.92) for the i dentification plus imitation group. Furthermore, e d prime scores for the trained and untrained stimuli in the pre training task, the training session and the post training task were calculated separately in order to investigate whet her the participants generalized the improvements gained through the trained stimuli to the untrained stimuli. The trained stimuli included the syllables [k h [k h o ] [p h a] and [p h i], which were used in the pre training task the training session and the post training task, while the untrained stimuli were the syllables [t h a], [t h i], [t h h o] which were only tested in the pre and post training tasks. Besides, the syllables [k h a], [k h i], [p h h o] were only used in the training session. Since the untrained stimuli were not included in the training session

PAGE 104

104 we first focus ed our analysis on the mean d prime scores for the trained and untrained stim uli in the pre and post training tasks. Figure 4 3 presented the mean d prime scores with the standard error of the trained and untrained stimuli in the pre and post training task s for both groups. The mean d prime scores fo r the trained stimuli in the p re and post training task s were 1.26 (SD = .99 ) and 2.11 (SD = 1.15 ) for the identification only group, and 1.34 (SD = .76 ) and 2.36 (SD = .85 ) for the identification plus imitation group. For the untrained stimuli the mean d prime scores in the pre and post training tasks were 1.3 4 (SD = 1.00 ) and 2.14 (SD = 1.11 ) for the identification only group, and 1.39 (SD = .81 ) and 2.30 (SD = .97 ) for the identification plus imitation group. The mean d prime scores for each participant were submitted to a 2 (test time: pre and post training) 2 (stimulus type: trained and untrained) 2 (group) repeated measures ANOVA, with test time and stimulus type as the within subject factors and group as the between subject factor. The result showed that there was a main e ffect of test time [ F (1, 18) = 93.28, p < .001] indicating that both groups performed more accurately in the post training task than in the pre training task No other significant effect or interaction was found. There was no difference between the mean d prime scores for the trained and untrained stimuli, suggesting that participants generalized the improvements gained through the trained stimuli to the untrained stimuli. Furthermore prime scores for the tested ([k h h o], [p h a] and [p h i] ) and untested ( [k h a], [k h i], [p h h o] ) stimuli in the training session were calculated separately for the identification only and the identification plus imitation groups. Figure 4 4 presents the mean d prime scores with t he standard errors of the

PAGE 105

105 tested and untested stimuli in the training session for both groups. The mean d prime scores of the tested and untest ed stimuli were 2.05 (SD = 1.09) and 2.08 (SD = 1.03 ) for the identification only group, and 2. 32 (SD = .79 ) and 2.34 (SD = .79 ) for the identification plus imitation group. The mean d prime scores for each participant were submitted to a 2 (stimulus type: tested and untested) 2 (group) repeated measures ANOVA, with stimulus type as the within subject factor and gr oup as the between subject factor. No significant effect or interaction was found, indicating that the two groups did not differ in the identification of the tested and untested stimuli in the training session. In sum, participants in both the identificat ion only and the identification plus imitation groups improved on the tone identification abilities through the training session. Even though the experimental design was biased toward the identification only group, the identification only group did not sho w more improvement than the identification plus imitation group after training. Moreover, the participants in both groups generalized t he improvements gained through the trained stimuli to the untrained stimuli. Reaction T imes imes were measured from the o nset of the stimulus and were rounded to full milliseconds. Mean reaction times with the standard errors of all the correct responses in the pre training task, the training session and the post training for the identification only and the identification plus imitation g roup are presented in Figure 4 5 The mean reaction times in the pre training task, the training session and the post training task were 1675 ms (SD = 611.13), 1946 ms (SD = 694.48) and 1587 ms (SD = 804.97) for the identification only group, and 1943 ms (SD = 560.09 ),

PAGE 106

106 1 840 ms (SD = 722.22 ) and 1510 ms (SD = 567.42 ) for the identification plus imitation group. Because the training session had different stimuli and experimental procedure from the pre and post training t asks, we first performed analysis on the reaction times in the pre and post training tasks to test whether the participants identified tones more quickly in the post training task than in the pre training task as a result of training The mean response t imes for the trained and untrained stimuli were calculated separately for the identification only and identification p lus imitation groups. Figure 4 6 illustrates the mean reaction times with the standard errors of the trained and untrained stimuli in the pre and post training tasks for both groups. For the trained stimuli, the mean reaction times in the pre and post train ing tasks were 1670 (SD = 661 ) and 1590 ms (SD = 814 ) for the identi fication only group, and 1894 ms (SD = 524 ) and 1458 ms (SD = 511 ) for the identification plus imitation group. For the untrained stimuli, the mean reaction times in the pre and post training tasks were 1 677 (SD = 576 ) and 1586 ms (SD = 800 ) for the identi fication only group, and 1458 m s (SD = 511 ) and 1568 ms (SD = 636 ) for t he identification plus imitation group. The reaction times in the pre and post training tasks were submitted to a 2 (test time : pre and post training ) 2 (stimulus type : trained and untrained ) 2 (group) repeated measures ANOVA with test time and stim ulus type as the within subject factors and group as the between subject factor. The results showed that there was a main effect of test time [ F (1, 18) = 14.11, p = .001] and a significant interaction of test time group [ F (1, 18) = 6.24, p = .022]. Two independent samples t tests showed that the two groups did not differ in the reaction times for the trained stimuli [ t (18) = 0.84, p = .41] or for the untrained stimuli [ t (18) = 1.16, p = .26] before training. Further paired samples t tests revealed t hat only the

PAGE 107

107 identification plus imitation group identified the tones more quickly in the post training task than in the pre training task [ t (9) = 4.71, p = .001]; the identification only group did not show any improvement with respect to reaction time [ t ( 9 ) = .89 p > 39 ] after training Moreover, the mean response times for the tested and untested stimuli in the training session were also calculated for the identification only and the identification plus imitation groups respectively Figure 4 7 prese nts the mean response times with the standard errors of the tested and untested stimuli in the training session for both groups. The mean reaction times of the tested and untested stimuli in the training session were 1 927 ms (SD = 659 ) and 1 967 ms (SD = 735 ) for the identification only group, and 1 842 ms (SD = 688 ) and 18 36 ms (SD = 760 ) for the identification plus imitation group. The reaction times for each participant were submitted to a 2 (stimulus type: tested and untested ) 2 (group) repeated measures AN OVA, with stimulus type as the within subject factors and group as the between subject factor. The result did not yield any significant effect or interaction. To sum up, although the experiment al design was biased toward the identification only group, the identification plus imitation group gained more improvement in terms of reaction time than the identification only group after training. Moreover, the reaction times for the trained and untrained stimuli did not diffe r. Percentages of Correct Identificati on for Each T one The mean percentages of correct identification for each tone type (Tone1: T1; Tone2: T2; and Tone4: T4) were computed separately for each participant. Because the training session had different procedure and stimuli from the pre and post training tasks, a nalyses were first performed on the percentages of correct identification in the

PAGE 108

108 pre and post training tasks. Moreover, we did not perform analyses for the trained and untrained stimuli separately because the number of the experimental tr ials in the pre and post training tasks was too small to be separated for the trained and untrained stimuli and for each tone type. Figure 4 8 illustrates the mean percentages of correct identification for each tone type in the pre and post training tas ks for the identification only and the identification plus imitation groups. For the identification only group, the mean percentages of correct identification of the T1, T2 and T4 were 60.00% (SD = 33.59%), 74.69% (SD = 24.45%) and 66.88% (SD = 23.49%) in the pre training task, and 85.63% (SD = 21.86 %), 83.75 % (SD = 17.79 %), 76.56 % (SD = 21.51 %) in the post training task For the identification plus imitation group, the mean percentages of correct identification of the T1, T2 and T4 were 71.56% (SD = 26.86% ), 70.63% (SD = 14.82%) and 68.44% (SD = 16.23%) in the pre training task, and 86.56% (SD = 16.86%), 87.81% (SD = 10.97%), 86.56% (SD = 11.88%) in the post training task. The mean percentages of correct identification for each participant were submitted to a 2 (test time: pre and post training task) 3 (tone type: T1, T2 and T4) 2 (group) repeated measures ANOVA with test time and tone type as the within subject factors and group as the between subject factor. The result showed that there was a main effe ct of test time [ F ( 1 18 ) = 64.13 p < 00 1] indicating that participants in both groups identified all the three tones more accurately in the post training task than in the pre training task. No other effect or interaction was significant. Furthermore, the mean percentages of correct identification for each tone type in the training session were also computed for the identification only and the identification

PAGE 109

109 plus imitation groups respectively. Figure 4 9 presents the mean percentages of correct identif ication with standard errors for each tone type in the training session for the two groups. The mean percentages of correct identification for the T1, T2 and T4 were 80.63% (SD = 20.77%), 83.75% (SD = 16.22%) and 81.09% (SD = 17.94%) for the identification only group, and 87.66% (SD = 12.20 %), 89.53 % (SD = 6.83 %) and 87.97 % (SD = 8.90 %) for the identification plus imitation group. The mean percentages of correct identification in the training session were submitted to a 3 ( tone type ) 2 (group) repeated me asures ANOVA with tone type as the within subject factor and group as the between subject factor. The result did not yield any significant effect or interaction suggesting that the two groups did not differ in the identification accuracy for any of the to ne type during the training session. We further calculated the number of errors each participant mad e for each error type in the pre and post training tasks Figure 4 10 depicts the error distribution for the identification only and the identification pl us imitation groups Before training the participants were more likely to misidentify T1 as T2 T2 as T1, T2 as T4, and T4 as T1. After training participants made few er er rors for these error types: T1 as T2: t (19 ) = 4.56 p < .0 01; T2 as T1: t (19 ) = 2.8 8 p = .0 10; T2 as T4: t (19 ) = 2.59 p = .0 18; and T4 as T1: t (19 ) = 5.23 p < .0 01. Summary This experiment compared the effectiveness of a laboratory identification only training and an identification plus imitation training on tone identification. The results demonstrated that the identification accuracy improved after both the identification only and the identification plus imitation training. Moreover, the participants in both groups generalized the improvements gained through the train ed

PAGE 110

110 stimuli to the untrained stimuli. In terms of the reaction time, the identification plus imitation group identified the tones more quickly in the post training task than in the pre training task but the identification only group did not show any improve ment. These results indicated that even though the experimental design was biased toward the perception only group, the identification plus imitation training was more effective to improve the tone identification than the identification only training.

PAGE 111

111 Ta ble 4 1. Mean accuracy percentages of the tonal test, the rhythmic test and the composite performance in AMMA for the identification only and identification plus imitation groups Group Tone Rhythm Composite Identification only group 48% 47% 46% Identific ation plus imitation group 57% 48% 52% Figure 4 1. Step illustration of the identification only t raining and the identification plus imitation training

PAGE 112

112 Figure4 2 Mean d prime scores with the sta ndard errors for the identification only and the identification plus imitation groups in the pre training task, the training session and the post training t ask respectively Figure 4 3. Mean d prime scores with the standard errors for the trained and untrained stimuli in the pre and post training tasks for the identification only and the identification plus imitation group respectively

PAGE 113

113 Figure 4 4. M ean d prime scores with the standard errors of the tested and untested stimuli in the training session for the identification only and the identification plus imitation groups respectively Figure 4 5. Mean reaction times with the standard errors of all the correct responses in the pre training task, the training session and the post training for the identification only and the identification plus imit ation group

PAGE 114

114 Figure 4 6. Mean reaction times with the standard errors of the trained and untrained stimuli in the pre and post training tasks for both groups respectively Figure 4 7. Mean response times with the standard errors of the tested and untested stimuli in the training session for the identification only and identification plus imitation groups respectively

PAGE 115

115 Figure 4 8. Mean percentages of correct identification with the standard errors for each tone type in the pre and post training tasks for the identification only and the identifica tion plus imitation groups Figure 4 9. Mean percentages of correct identification with standard er rors for each tone type in the training session for the identification only and the identification plus imitation groups respectively

PAGE 116

116 Figure 4 10 Error distribution for the identification only and the identification plus imitatio n groups. (Note: t he tone before the d ash refers to the correct answer while the tone after the slash indicates participants )

PAGE 117

117 CHAPTER 5 CONCLUSION S This dissertation sought to compare the effectiveness of perception training and production training on lexical tone perception and to examine whether the effects were domain specific or domain general. T hree main res earch questions were investigated in the present study : Are the effects of production training and perception training domain general or domain specific ? That is, can the instrumental and vocal expertise percolate to lexical tone perception and production ? Is laboratory perception plus production training more effective than percepti on only training in facilitating lexical tone perception at the intentional level ? Does perception plus production training affect the unintentional perception of lexical tones differently from perception only training? If the music expertise carries over to the language domain, w e predicted that vocal musicians should have more advantage than instrumental musicians in imitating lexical tones. I f production indeed facilitates per ception as suggested by the motor theory ( e.g. Liberman & Mattingly, 1989) the vocalists might also be superior to the instrumentalists in the perception of lexical tones. Moreover, the laboratory perception plus production training should be more effect ive than perception only training in facilitating lexical tone perception at both intentional and unintentional levels. However, if the perception only group yielded better results after training than the perception plus production group, the additional pr oduction training did not enhance perception, which was contrary to the Motor Theory In order to test these predictions, we conducted t hree experiments which were pr esented in Chapters 2, 3, and 4. T he first experiment examined the production and percepti on training in the music domain while the second and third experiments focus ed

PAGE 118

118 on the language domain. In the following sections, I will summarize the main findings of the three experiments and discuss the results in terms of the effectiveness of perceptio n training and production training and the relationship between perception and production. Summary of Results Experiment 1 compare d English vocalists and instrumentalists to see which type of musical training is more advantageous to lexical tone perceptio n and production Participants were tested in a same/different discrimination task and a production task. The results demonstrat ed that the English vocalists outperformed the English non musicians in overall tone discrimination while the English instrument alists did not exhibit such advantage. However, the English instrumentalists and vocalists did not show significant difference in lexical ton e discrimination. Moreover the English vocalists and (pa ir 12, pair13, pair14, pair23, pair24, and pair34) almost as accuratel y as the Chinese non musicians, and both musician groups outperformed the English non musicians. Nevertheless t he English musicians did not show any advantage in the discrimination of t (pair11, pair22, pair33 and pair44) compared to the English non musicians. For the pair22, the English vocalists were marginally more accurate than the English instrumentalists. Unexpectedly, we did not find any difference between the English musicians and the English non musicians in lexical tone production according to the evaluation by native Mandarin speakers. Specifically, the English musicians and non musicians produced Tone1 (level tone) and Tone4 (falling tone) as accurately as the native Mandarin speakers. Taken together, these results suggest that the music training improves the pitch perception in the language domain. Moreover, to some extent the vocal training in addition to instrumental training was an advantage for lexical tone

PAGE 119

119 perception. However, lexical tone production was not significantly facilitated by either type of music training. Experiment 2 collected both behavioral and electrophysiological (ERPs) data to examine the effectiveness of a perception plus production training and a perception only training on the intentional and unintentional processing of tones by native English speakers. In the behavioral tasks, both the perception only and the perception plus production groups improved on the tone discrimination ab ilities through the training session Moreover, the participants in both groups generalized the improvements gained through the trained stimuli to the untrained stimuli. In the ERP tasks, the MMN was smaller in the post training task than in the pre traini ng task. Moreover, for the T41 condition the two groups differed in the lateralization of the MMN component, with the perception only group showing a larger MMN at the left frontal electrodes while the perception plus production group displaying a larger M MN at the right frontal electrodes. Besides, both groups showed an increased late negativity at the posterior electrodes in the post training task compared to the pre training task In sum, analyses on the ERP components did not show any signifi cantly diff erent training effects between the perception only and the perception plus production groups. As described in prev ious chapters, Experiment2 was a little biased toward the perception plus production group, because the imitation task probably required the participants to pay more attention to the stimuli than the perception only group. However, the perception plus production group did not gain more improvement through the additional imitation task compared with the perception only group.

PAGE 120

120 Experiment 3 was d esigned to be biased toward the perception only group. If the perception plus production group could show more improvement than the perception only group, it indicated that the perception plus production training was more effective than the perception only training and the employment of the articulatory system benefits the perceptual skills. In Experiment 3, participants were trained and tested in an identification task, instead of a discrimination task. The results demonstrated that the tone identification accuracy improved in both the identification only and the identification plus imitation groups after training. Moreover, the participants in both groups generalized the improvements gained through the trained stimuli to the untrained stimuli. In terms of the reaction time, the identification plus imitation group identified the tones more quickly in the post training task than in the pre training task while the identification only group did not show any improvement. These results indicated that the identifi cation plus imitation training was more effective to improve the tone identification than the identification only training, even though the experimental design was biased toward the perception only group. Discussion Effects of Vocal Experience and Instrum ental Experience on Lexical Tone Perception and Production Consistent with previous studies (e.g. Alexander et al. 2005; Lee & Hung, 2008), the current study demonstrated that the English musicians outperforme d the English non musicians on l exical tone pe rception, suggesting that the processing systems associated with lexical tone perception and musical pitch perception overlap to a great extent. Specifically, the English vocalists and instrumentalists outperformed the English non musicians on the discrimi

PAGE 121

121 discrimi pairs, participants need to form categories for the pitch contours and to ignore other within category acoustic and articulatory features To the contrary, the discrimination of category properties (e.g. Jamieson & Morosan, 1989). Therefore, our results suggested that even thoug h the English musicians had acuter discrimination skills of lexical tone, their pitch categorization skills were not significantly better than those of the non musicians. Our results were highly comparable with the findings in Wayland, Herrera & Kaan (201 0), which demonstrated that although the musicians outperformed the non musicians in the identification of non linguistic pitch contours, their ability to quickly categorize the non linguistic pitch contours across different tokens was not superior to that of non musicians both before and after 3 days of identification training. Taken skills but not the categorization skills of lexical tone. In terms of the effects of dif ferent music experience, we found that the English vocalists and instrumentalists did not differ in the discrimination skills of lexical tone. Using behavioral and electrophysiological methods, Nikjeh (2006) examined the discrimination of non linguistic pi tch by vocalists and instrumentalists and also failed to find a difference between the two types of musicians These results suggested that reg ardless of training type long term music background benefited auditory pitch discrimination in both music and lan guage domain. Interestingly, our result demonstrated that the discrimination of pair22 was marginally better in the English vocalists than in the English instrumentalists. This result seemed to provide some

PAGE 122

122 indication that the vocal training was more effec tive than the instrumental training to But why did we only find the difference for pair22 rather than other tone pairs? One possible explanation for this question is that the English listeners a re more familiar with the linear tones (Tone1 and Tone4) than the contour tones (Tone2 & Tone3), since the level and falling tones are very similar to the level and falling intonation used at sentence final position in English. It might be speculated that the different effect of music background only shows up for unfamiliar pitch contours rather than familiar pitch contours Pair 33 did not show such difference because of the stimulus manipulation which resulting in Tone 3 being more likely to be perceive d as Tone4 As for lexical tone production, the current data showed that the English musicians did not differ significantly from the English non musicians as judged by native Mandarin speakers. Specifically, the English musicians and non musicians produce d Tone1 (level tone) and Tone4 (falling tone) as accurately as the native Mandarin speakers, while produced Tone2 (rising tone) and Tone3 (falling rising tone) less accurately than the native Mandarin speakers. Moreover, the English musicians did not diffe r from the English non musician in the production of Tone2 and Tone3. This result was inconsistent with previous studies (e.g. Amir, Amir, & Kishon Rabin, 2003; Nikjeh, 2006), which demonstrated that regardless of training type music background enhanced pi tch production abilities. The discrepancy between the findings in the current study and previous research might be attributed to the acoustic and articulatory features of lexical tone. Burnham & Mattock (2007) claimed that besides pitch, lexical tone compr ises several other acoustic and articulatory features, including duration, amplitude and

PAGE 123

123 phonation type. The English musicians might only have advantages in the production of pitch, but not other acoustic or articulatory features associated with Mandarin t ones Therefore, their productions were not judged as more accurate than that of the English non musicians by the native Mandarin speakers. Further acoustic analysis on be needed to examine the validity of this explanation In addition, the production of lexical tone did not differ in the English vocalists and instrumentalists. This result was compatible with Nikjeh (2006) which also failed to find significant differences between vocalists and instrume ntalists in music pitch producti on Nikjeh (2006) argued that this result was probably due to that the self paced imitation task was too easy for the musicians to show the different training effect between vocalists and instrumentalists. Another possible explanation for the lack of distinction between the English vocalists and instrumentalists in both the discrimination and imitation tasks is the effects of different types of instrumental training. Spiegel & Watson (1984) reported that musicians whose primary instrument was brass string or woodwind had smaller discrimination thresholds than musicians whose primary instrument was keyboard. Spiegel & Watson claimed that the musicians who needed to tune their own instruments might be more sensitive to small frequency changes than th e musicians who did not tune their own instruments. In the current study, most of the instrumentalists played brass, string or woodwind instruments. Moreover, the vocal musicians also need to tune their articulatory apparatus and their sensation of production Therefore, the English vocalists and instrumentalists in the current study did not differ in either the discrimination or the production of lexical tones. Further

PAGE 124

124 investigations on vocalists and instrumentalists with sam e instrumental expertise should provide more insights on the different effects of vocal training and instrumental training on lexical tone perception and production. Effects of Laboratory Perception training and Production Training on Tone Perception Sinc e the perception training and production training in the music domain did not show significant difference in improving the perception and production of lexical tone, w e further proposed a behavioral and electrophysiological investigation to examine the eff ects of laboratory perception only training and perception plus production training on tone perception. If production indeed facilitated perception, as implied by the Motor Theory (e.g. Liberman & Mattingly, 1989), we anticipated that the perception plus p unintentional discrimination of tone than the perception only training. To the contrary, if production relied on perception, as claimed by the Speech Learning Model (Flege 1995 a ), we would expect that the perception only training might be more effective to than the perception plus production training. In order to investigate these predictions, we co nducted two experiments using two training paradigms. Discrimination training paradigm In the behavioral tasks, both the perception only and the perception plus production groups improved on the tone discrimination abilities as a result of training. Moreov er, the participants in both groups generalized the improvements gained through the trained stimuli to the untrained stimuli. These results suggested that the additional production component in the perception plus production training did not result in more

PAGE 125

125 improvement in the tone discrimination. There were several possible accounts for this perceptual learning. However, this seems to be unlikely since Adank et al. (2010) fo und that no matter whether people heard their own voice or not, vocal imitation significantly improved the comprehension of unfamiliar accent. Moreover, Kraljic & Samuel (2005) wn bad productions. Second, the lack of additional improvement in the perception plus production group may have been due to our training session being too short to show the effect of the production component. In other training studies (e.g. Wang et al., 19 99; Wayland & Li, 2008; Baese Berk, 2010), participants usually received at least two sessions of training with each session took at least one hour. In our study, t he perception plus production training only took one hour, during which the participants wer e only allowed to imitate each stimulus once. Moreover, the participants did not receive any feedback on their productions. Therefore, the perceptual learning through the production process was very limited. Further research should include longer training order to really facilitate perceptual learning. Third, the lack of significant difference between the two groups may also be attributed to the individual differences, with some fast learners demonstrating large improvement while others revealing small or no improvement after training An analysis on individual improvements would be necessary to further investigate the difference between the two training groups. Fourth, if the lack of distinction between the two groups is not due to the short training session or the individual differences we could surmise that the production leaning might be hard

PAGE 126

126 to transfer to the perceptual learning, indicating that speech perception does not rely on speech production. In the ERP task, both groups showed a smaller MMN in the post training task than in the pre training task. This result were comparable with Kaan et al. (2008), which demonstrated that after training the MMN decreased for th e low falling tone in English speakers. Kaan et al. (2008) claimed that the larger MMN in the pre training task was due to the differences in F0 onset rather than the F0 direction. The English speakers were more sensitive to the F0 onset (e.g. Gandour, 198 3) before training and started paying more attention to the F0 direction after training. However, the detection of F0 direction was harder than the detection of F0 onset for the English speakers, therefore, the MMN amplitude decreased in the post training task compared to the pre training task. This explanation could account for the result in the current study as well. In the current study, there were six conditions in the ERP task: T12, T14, T21, T24, T41 and T42 In four ( T12, T21, T24 and T42 ) of the six conditions, the deviants had different F0 onset from the standards, resulting in a larger MMN in the pre training task in both the perception only and perception plus production groups. F rom Figure 3 13 we also observed that the MMN was larger in the T21 T41 and T24 conditions than in the T12 T14 and T42 conditions respectively. Francis & Ciocca (2003) found that Cantonese listeners were more accurate to discriminate the tone pairs when the first token in the pair had a lower F0 (low high order) than whe n the first token had a higher F0 (high low order). Given these findings, it might be that, the MMN was more prominent if the pitch onset of the deviant stimuli was lower than the pitch onset of the standard stimuli; and the MMN was most obvious if the pit ch offset of the deviant stimuli was lower than the

PAGE 127

127 pitch onset of the standard stimuli. In the T21 and T24 condition s the pitch onset of the deviant stimuli (T2) was lower than the pitch onset of the standard stimuli ( T1 &T4) T herefore, the T21 and T24 conditions showed larger MMN than the T12 and T42 conditions. In the T 41 condition, the pitch offset of T4 was lower than the pitch onset of T1 therefore, this condition showed the most prominent MMN. Future analysis on the different stimulus presentatio n order in the behavioral data may yield further insights on this account. In addition both groups showed an increased late negativity at the posterior electrodes in the post training task compared to the pre training task. This finding was inconsistent with Kaan et al. (2008), which presented a decreased late negativity after training. They claimed that the smaller late negativity in the post training task might indicate less intentional reorientation from deviant stimuli to standard stimuli. In other st udies, the late negativity has also been associated with neural processes of auditory rule extraction (e.g. Zachau et al., 2005). Therefore, the increased late negativity in the current study might suggest that both the percepti on only and the perception p lus production groups improved at the unintentional level of tone abstraction as a result of training. Nevertheless, the perception only and the perception plus production groups did not show any significantly different training effects in any of the ERP components. This result suggested that the additional production component in the perception plus production training did not result in more improvement in the tone discrimination at the unintentional level More research is needed to determine the cause o f the lack of distinction between the two types of training. However, these findings provide an

PAGE 128

128 important first step in examining the effectiveness of perception training and production training on tone perception. Identification training paradigm Using d iscrimination training paradigm, the participants who received the perception plus production training did not show different improvements compared to the participants who received the perception only training in the intentional and unintentional discrimin ation of tone. We further examined the effectiveness of these two types of training using an identification training paradigm. Conforming to the discrimination training paradigm, t he results demonstrated that the tone identification accuracy improved in bo th the identification only and the identification plus imitation groups after training. Moreover, the participants in both groups generalized the improvements gained through the trained stimuli to the untrained stimuli. One difference observed between the two training groups was that the identification plus imitation group identified the tones more quickly in the post training task than in the pre training task while the identification only group did not show any improvement in terms of reaction time. It sh ould be remembered that this experiment was a little biased toward the identification only group because the identification only training reinforced the tone types of the stimuli during the production process while the identification plus imitation trainin g directed the participants to pay more attention to the sounds rather than the tone types. Even though the experimental design was biased toward the identification only group, the identification plus imitation group showed more improvement suggesting that the identification plus imitation training was more effective to improve the tone identification than the identification only training.

PAGE 129

129 Final remarks In the current study, two training paradigm yielded different results: using the discrimination training paradigm, the participants who received the perception plus production training did not show more improvements compared to the participants who received the perception only training in the intentional and unintentional discrimination of tone; however, usi ng the identification training paradigm, the participants who received the identification plus imitation training showed a larger decrease in reaction times than the participants who received the identification only training in tone identification. Our res ults were inconsistent with some previous research. For example, Flege (1995 b ) trained native Mandarin speakers to identify the English word final /t/ and /d/ using both identification training and discrimination training. He found that both training parad Moreover, Wayland & Li (2008) trained native English and Chinese speakers to discriminate Thai mid and low tone contrast, using both identification and discrimination training. They al so found that the two training procedures were equally effective in Nevertheless, these two studies only focused on the accuracy data but did not analyze the reaction time data. Therefore, i t was unclear whether reaction time could yield and different effects between the two training paradigms. Additionally, the identification training used in these two studies consisted of a two alternative forced choice identification procedure, which might be too easy for the participants to gain significant improvements. In our study, the identification training employed a three alternative forced choice procedure, which might be more effective to show the different training effects.

PAGE 130

130 Our results suggested that the identification training was more effective than The discrimination training might encourage participants to pay more att ention to within category acoustic differences that are irrelevant to the key identify of the categories (e.g. Jamieson & Morosan, 1989), while the identification training encourages participants to form distinct categories and to focus more on the differences between categories. However, it was still uncle ar why different training paradigm presented different results in the current study. A possible account for this question is that the categorization ability might be crucial for production learning to be able to transfer to the perceptual modality. The dis crimination training paradigm did not yield different results between the perception only and the perception plus production groups is probably due to that the participants did not gain enough categorization skill s through the discrimination training. Ther efore, the transfer from the production learning to the perceptual learning did not take place. On the contrary, the participants who received the identification training gained enough categorization skills, which enable the production learning to take pla ce and in turn to facilitate the percept ual learning In one word, the categorization skill might be the prerequisite for the occurrence of the production learning which in turn may facilitate the perceptual learning. By this account, speech production ma y rely on speech perception since the categorical learning precedes the production learning, which was contrary to the Motor theory. Future research can compare the effect of perceptual learning on speech production with the effect of production learning o n perceptual learning in order to uncover the relationship between speech perception and production.

PAGE 131

131 APPENDIX A MUSICAL BACKGROUNDS FOR THE ENGLISH MUSICIANS IN EXPERMENT1 Table A 1. Musical backgrounds for the American English musicians in Experim ent 1 Mu sician number Years of music training (vocal/instrumental) Age onset of training (vocal/instrumental) Instrument V01 4/10 18/12 Voice, piano V0 2 7/16 14/5 Voice, piano V0 3 10/NA 10/NA Voice V0 4 3/7 19/10 Voice, piano, guitar V0 5 7/5 14/5 Voice, pian o V0 6 3/1 18/6 Voice, piano V0 7 3/6 16/10 Voice, piano V0 8 15/NA 6/NA Voice V0 9 14/3 11/8 Voice, piano V 10 12/1 6/5 Voice, piano, guitar, ukulele, vibraphone V1 1 14/15 16/8 Voice, piano, guitar V1 2 6/10 16/6 Voice, piano, guitar, ukulele V 13 10/3 11/7 Voice, guitar, piano V 14 9/17 14/5 Voice, piano, flute V1 5 8/7 15/11 Voice, saxophone, flute, clarinet I01 NA/10 NA/9 Flute, piccolo I02 NA/13 NA/6 Piano, clarinet I03 NA/10 NA/11 Clarinet I0 4 NA/6 NA/9 Piano, flute, guitar I0 5 NA/4 NA/ 14 Guitar I0 6 NA/8 NA/10 Trumpet I0 7 NA/9 NA/11 Clarinet, piano I0 8 NA/7 NA/11 Trumpet, flute I0 9 NA/3 NA/11 Flute, piccolo I1 0 NA/10 NA/10 Trumpet, sax, euphonium, clarinet, piano I1 1 NA/12 NA/10 Horn, piano I 12 NA/11 NA/8 Piano, flute, piccolo I 13 NA/12 NA/6 Piano, flute, guitar I 14 NA/6 NA/11 Trumpet, bass guitar I 15 NA/12 NA/7 Piano, saxophone, clarinet, bass guitar

PAGE 132

132 APPENDIX B LANGUAGE AND EDUCATION QUESTIONS Note: Your participation is voluntary; you may decline to answer questions. Age:___ Gender: M / F Hispanic? Y / N Race: ______________ Handedness: R / L / both Education ________________ 1b. What is your major (if declared)? ______________________________________ Language 2. Please indicate what languages you speak (including English), whether you learned them at school and at what age/grade you learned them. Next, for e ach language, rate how fluent you are in writing, reading, speaking and listening, using a scale from 1 (=not fluent at all) to 5(= native or near native). Language Learned at school? At home? Age/grade acquired Writ ing Read ing Spea king Liste ning English 3. Which language did you learn first as a child? ___________ 4. Do you consider yourself bi(multi)lingual? ________ Mother _____ __________________________Age acquired:__________ _______________________________Age acquired:__________ _______________________________Age acquired:__________ Father _______________________________Age acquired:__________ _______________________________Ag e acquired:__________ _______________________________Age acquired:__________

PAGE 133

133 6a. If you were not born in the US, at what age did you move to the States? __________ 6b. In which country or countries did you live before you moved to the US? ________________ _______________________________________________________ 7. If you were born in China, what part of China are you from? _________________ What part of China are your parents from?_____________________________________ What dialects of Chinese do you speak?_ ______________________________________ 8. Have you had, or do you have, any known problems with reading, hearing, speech?_______________If yes, please explain: ________________________________ Music experience 9 Do you have perfect (absolute) pitch? Y / N 10a. Have you played or do you play a musical instrument (including voice) or have you ever received musical training? Y / N below for comments Instrument, Including voice Started at age Stopped at age Started with formal training at age Stopped with formal training at age Still playing? Still receiving formal training? Comments

PAGE 134

134 APPENDIX C PERSONAL HISTORY FORM FOR LANGAUGE EXPERIMENTS WITH EVENT RELATED POTENTIALS OR FMRI The following form contains questions about your health. You may request that this form be seen only by the Principal Investigator in charge of the project. If you do no t request this, the form may also be seen by the research assistant. If you request that the form be seen only by the Principal Investigator, check here: ____

PAGE 135

13 5 Health related questions. Note: Your participation is voluntary; you may decline to answer q uestions. Please check the appropriate response. Yes No 1a. Are you taking any medications? ( ) ( ) 2 system (e.g. beta blockers, anti depressants, epilepsy medication)? ( ) ( ) 2. Do you easily get an allergic reaction on your skin? ( ) ( ) 3. Are you allergic to band aids, or stickers on your skin? ( ) ( ) 4. Are you allergic to dust mite? ( ) ( ) 5. Is your corrected vision significantly less than 20/20? ( ) ( ) 6. Is your hearing ok, as far as you know? ( ) ( ) 7a. Have you ever had a head injury or have you been knocked unconscious? ( ) ( ) ( ) ( ) 8. Have you ever had a seizure? ( ) ( ) 9. Have you ever suffered from a neurological or psychological disorder (e.g. epilep sy, depression, schizophrenia)? ( ) ( ) 10. Have you ever been diagnosed with a language or reading related problem (e.g. dyslexia)? ( ) ( ) 11a. Have you participated in other scientific studies? ( ) ( )

PAGE 136

136 APPENDIX D HAN DEDNESS QUESTIONNAIRE Note: Your participation is voluntary; you may decline to answer questions. Which is your dominant hand?_________________________________________ Have you ever had any tendency to change your preferred hand?______________ Do you have any immediate family members that are left handed (parents, siblings?) Please indicate your preferences in the use of hands in the following activities by putting + in the appropriate column. Where the preference is so strong that you would never try t o use the other hand unless absolutely forced to, put ++. If in any case you are really indifferent, put + in both columns. Please try to answer all questions, and leave only a blank if you have no experience at all of the activity or task. L R 1. Writi ng 2. Drawing 3. Throwing 4. Scissors 5. Toothbrush 6. Knife (without fork) 7. Spoon 8. Broom (upper hand) 9. Striking match (match) 10. Opening box (lid) i. Which foot do you prefer to kick with? ii. Which eye do you use when using only one?

PAGE 137

137 LIST OF REFERENCES Adank, P., Haggort P., & Bekkering, H. (2010). Imitation i mproves language comprehension. Psychological Science 21(12), 1903 1909. Alexander, J.A., Wong, P.C. M., & Bradlow, A. R. (2005). Lexical tone perc eption in musicians and non musicians. Proceedings of Interspeech 2005, Lisbon, Portugal. Alain, C., Woods, D.L., & Knight, R. T. (1998). A distributed cortical network for auditory sensory memory in humans. Brain Research 812, 23 37. Amir, O., Amir, N., & Kishon Rabin, L. (2003). The effect of superior auditory skills on vocal accuracy. Journal of the Acoustical Society of America 133 (2), 1102 1108. Anvari, S. H., T rainor, L. J., Woodside, J., & Levy, B. A. (2002). Relations among musical skills, phonolog ical processing, and early reading ability in preschool children. Journal of Experimental Child Psychology 83, 111 130. At ienza, M., Cantero, J. L., & Dominguez Marin, E. (2002). Th e time course of neural changes underlying auditory perceptual learning. L earning and Memory 9, 138 150. Baese Berk, M. M. (2010). An Examination of the relationship between speech perception and production. Doctoral dissertation. Northwestern University. Best, C. T. (1995). A direct realist view of cross language speech percept ion. In Strange, W. (Ed), Speech Perception and Linguistic Experience: Issues in Cross Language Research York, Baltimore, 171 204. Boersma, P., & Weenink, D. (2010). Praat: doing phonetics by computer [Computer program]. Version 5.1.25, retrieved from < http://www.praat.org/ > on March 31, 2010 Bradlow, A., Aka hane Yamada, R., Pisoni, D., & Tohkura, Y. (1999). Training Japanese listeners to identify English /r/ and /l/: Long term retention of learning in perception and production. Perception and Psychophysics 61, 977 985. Bradlow, A., Pisoni, D., Aka hane Yamada, R., & Tohkura, Y. (1997). Training Japanese listeners to identify English /r/ and /l/: some effects of perceptual learning on speech production. Journal of the Acoustical Society of America 101, 2299 2310. B radlow, A.R., Torretta, G.M., & Pisoni, D.B. (1996). Intelligibility of normal speech I: Global and fine grained acoustic phonetic talker characteristics. Speech communication 20, 255 272.

PAGE 138

138 Burnham, D., & M attock, K. (2007). The perception of tones and phones. In O. S. Bohn & M. J. Munro (Eds.), Language experience in second language speech learning (pp. 259 280). Amsterdam: John Benjamins. Chand rasekaran, B., Krishnan, A., & Gandour, J. (2009). Relative infl uence of musical and linguistic experience on early cortical processing of pitch contours. Brain and Language 108, 1 9. Chand rasekaran, B., Krishnan, A., & Gandour, J. (2007). Neuroplasti city in the processing of pitch dimensions: A multidimensional scaling analysis of the mismatch negativity. Restorative Neurology and Neuroscience 25, 195 210 Chao, Y. R. (1948). Mandarin primer Cambridge: Harvard University Press. Chen, G. T. (1974). The pitch range of English and Chinese speakers. Journal of Chines e Linguistics, 2, 159 171. Cowan, N., & Morse, P.A. (1986). The use of auditory and phonetic memory in vowel discrimination Journal of the Acoustical Society of America, 79, 500 507 Di Pellegrino, G., Fadiga, L., Fogassi, L., Gallese, V., & Rizzolatti, G. (1992). Understanding motor events: A neurophysiological study. Experimental Brain Research 91, 176 180. Escera, C., Alho, K., Winkler, I., & Ntnen, R. (1998). Neural mechani sms of involuntary attention to acoustic novelty and change. Journal of Cog nitive Neuroscience 10, 590 604. Flege, J.E. (1995 a ). Second language speech learning: Theory, findings and problems In W. Strange (Ed.), Speech Perception and Linguistic Experience: Issues in Cross language Research (pp. 233 277). Baltimore: York. Flege J.E. (1995 b ). Two procedures for training a novel second language phonetic contrast. Applied Psycholinguistics 16, 425 442. Flege, J.E., & Eefting, W. (1987). Cross language switching i n stop consonant perception and production by Dutch speakers of Engl ish. Speech Communication 6(3), 185 202. Fowler, C. A. (1986). An event approach to the study of speech perception from a direct realist perspective. Journal of Phonetics 14, 3 28. Francis, A.L., & Ciocca, V. (2003). Stimulus presentation order and the pe rception of lexical tones in Cantonese. Journal of Acoustical Society of America 114 (3), 1611 1621.

PAGE 139

139 Francis, A., Ciocca, V., Wong, N.K.Y., Leung, W.H.Y. & Chu, P.C.Y. (2006). Extrinsic context affects perceptual normalization of lexical tone. Journal of Acoustical Society of America 199(3), 1712 1726. Galantucci, B., Fowler, C. A., & Tuvey, M. T. (2006). The motor theory of speech perception reviewed. Psychonomic Bulletin and Review 13 (3), 361 377. Gandour, J. (1983). Tone perception in Far Eastern langu ages. Journal of Phonetics 11, 49 175. Gordon, E. E. (1989). Manual for the advanced measures of music audiation Chicago: GIA. Goto, H. (1971). Auditory perception by normal Japanese ad ults of the sounds "l" and "r". Neuropsychologia, 9, 317 323. Guenther F.H., Ghosh, S. S., & Tourville, J. A. (2006). Neural modeling and imaging of the cortical interactions underlying syllable production. Brain and Language 96, 280 301. Hattori, Kota (2010). Perception and production of English /r/ /l/ by adul t Japanese s peakers. D octoral dissertation. University of College London. Herd, W. (2011). The perceptual and production training of /d, r/ in L2 Spanish: behavioral, psycholinguistic, and neuroli nguistic evidence. D octoral dissertation. University of Kansas. Hickok, G. (2001). Functional anatomy of speech perception and s pee c h production: p sycholinguistic i mplications. Journal of Psych olinguistic Research 30 (3), 225 235. Hickok, G., & Poeppel, D. (2004). Dorsal and ventral streams: a fram ework for understanding aspects of the functional anatomy of language. Cognition 92, 67 99. Hickok, G., & Poeppel, D. (2007). The cortical organizati on of speech processing. Nature Reviews Neuroscience 8, 393 402. Hirata, Y. (2004). Computer assisted pronunciation training for native English speakers learning Japanese pitch and durational contrasts. Computer Assisted Language Learning 17, 357 376 Ja mieson, D. G. & Morosan, D. E. (1986). Training non native speech contrasts in adults: Acquisition of the English /T/ /D/ CONTRAST BY Francophones. Perception and psychophysics 40, 205 215.

PAGE 140

140 Jamie son D.G., & Morosan, D.E. (1989). Training new, non native s peech contrast: A comparison of the prototype and perceptual fading techniques. Canadian Journal of Psychology 43, 88 96. Kaan, E., Barkley, C., Bao, M., & Wayland, R. (2008). Thai lexical ton e perception in native speakers of Thai, English and Mandarin Chinese: an event related potentials training study. BMC Neuroscience 9, 53. Kaan, E., Wayland, R., Bao, M., & Barkley, C. (2007). Effects of native l anguage and training on lexical tone perception: An ERP study. Brain Research 1148, 113 122. Keating, P. (2004). D prime (signal detection) analysis. In University of Californi a at Los Angeles Phonetics Lab Statistics Page. Retrieved June 17, 2013 from http://www.lingui stics.ucla.edu/faciliti/facilities/statistics/dprime.htm Kirkham, J., Lu, S., Wayland, R., & Kaan, E. (2011). Comparison of vo calists and instrumentalists on lexical tone perception and production tasks. Proceedings of the 17th International Congress of Phonetic Sciences 1098 1101. Kraljic, T., & Samuel, A.G. (2005). Perceptual learning for speech: Is there a return to normal? Cognitive Psychology 51, 141 178. Krishnan, A Xu, Y, Gandour, J.T., & Cariani, P. (2005). Encoding of pitch in the human brai nstem is sensitive to language experience. Cognitive Brain Research 25, 161 168. Leach, L., & Samuel, A. G. (2007). Lexical configuration and lexical engagement: when adults learn new words. Cognitive Psychology 55, 306 353. Leather, J. (1990). Perceptual and productive learning of Chinese lexical tone by Dutch and English speakers. In Leather, L. and James, A. (Eds): New Sounds 90: Proceedings of the Amsterdam Symposium on the Acquisition of Second Language Speech Amst erdam: University of Amsterdam, 305 341. Lee, C.Y., & Hung, T. H. (2008). Identification of Mandarin tones by English speaking musicians and nonmusicians. Journal of Acoustical Society of America 124, 3235 3248 Levelt, W.J.M., Praamstra, P., Meyer, A.S., Helenius, P., & Salmelin, R. (1998). An MEG study of picture naming. Journal of Cognitive Neuroscience 10, 553 567. Liberman, A.M. Cooper, F.S., Shankweiler, D. P., & Studdert Kenned y, M. (1967). Perception of the speech code. Psychological Review 74, 431 461. Liberman, A. M., & Mattingly, I G. (1989). A specialization for speech perception. Science 243, 489 494.

PAGE 141

141 Logan, J., Lively, S., & Pisoni, D. (1991). Training Japanese listeners to identify English /r/ and /l/: A first report. Journal of the Acoustical Society of America 89, 874 886. Lo tto, A.J., Hicko k, G. S., & Holt, L. L. (2009). Reflections on mirror neurons and speech perception. Trends in Cognitive Sciences 13, 110 114. Macmillan, N. A., & Creelman, C. D. (1991). Cambridge, England: Cambridge University Press. Massaro, D.W. & Chen, T.H. (2008). The motor theory of speech perception revisited. Psychonomic Bulletin and Review 15, 453 457. Miracle, W. C. (1989). Tone production of American students of Chinese : A pr eliminary acoustic study. Journal of Chinese Language Teachers Association 24, 49 65. Moineau, S., Dronkers, N. F., & Bates, E. (2005). Exploring the processing continuum of sin gle word comprehension in aphasia. Journal of speech, language, and Hearing Res earch 48, 884 896. Ntnen, R., & Michie, P T. (1979 ). Early selective attention effects on the evoked potential: A critical review and reinterpretation Biological Psychology 8, 81 136 Ntnen, R., & Alho, K. (1995). Mismatch negativity a unique m eas ure of sensory processing in audition. International Journal of Neuroscience 80, 317 337. Ntnen, R., Pakarinen, S., & Rinne, T. (2004). The mismatch negati vity (MMN): towards the optimal paradigm. Clinical Neurophysiology 115 (1), 140 144. Nikjeh, D.A (2006). Vocal and instrumental musicians: electrophysiologic and psychoacoustic analysis of pitch discrimination and production. Doctoral dissertation. University of South Florida. Ojemann, G. A. (1983). Brain organization for language from the perspectiv e of electrical stimula tion mapping. Behavioral and Brain Science 6, 189 230. Oldf ield, R. C. (1971). The assessment and analysis of hande dness: The Edinburgh inventory. Neuropsychologia 9, 97 113. Payne, T.W. (2003). Working memory capacity and pi tch dis crimination. Doctoral d issertation, Georgia Institute of Technology. Pfordresher, P. Q., & Brown, S. (2009). Enhanced production and perception of musical pitch in tone language speakers. Attention, Perception, and Psychophysics 71(6), 1385 1398. Pisoni, D .B. (1973). Auditory and phonetic memory codes in the discrimination of consonants and vowels. Perception & Psychophysics 13, 253 260.

PAGE 142

142 Roberts, M. & Russo, R. (1999). London: Routledge. Schwanh u er B. (2007). Lexical tone perception and production: The role of language and musical background Doctoral dissertation. University of Western Sydney. Sheldon, A., & Strange, W. (1982). The acquisition of /r/ and /l/ by Japanese learners of English: evidence that spe ech production can precede speech perception. Applied Psycholinguistics, 3, 243 261. Shen, X. S. (1989). Toward a register approach in teaching Mandarin tones. Journal of Chinese Language Teachers Association 24, 27 47. Shestakova, A., Huotilainen, M., Cep onien, R., & Cheour, M. ( 2003). Event related potentials associated with second language learning in children. Clinical Neurophysiology, 114, 1507 1512. Shih, C. (1988). Tone and intonation in Mandarin. Working Papers of the Cornell Phonetics Laboratory 3, 83 109. Slevc, L. R., & Miyake, A. (2006). Individual differences in second language proficiency: does musical ability matter? Psychological Science 17, 675 681 Speigel, M., & Watson, C. (1984). Performance on frequency discrimination tasks by musician s and non musicians. Journal of the Acoustical Society of America 76, 1690 1696. Stigler, J.W., Lee, S., & Stevenson, H. W. (1986). Digit memory in Chinese and English: Evidence for a temporally limited store. Cognition 23, 1 20. Tremb lay, K., Kraus, N., Carrell, T. D., & McGee, T. (1997). Cent ral auditory system plasticity: Generalization to novel stimuli following listening training. Journal of the Acoustical Society of America 102(2), 3762 3773. Tremblay, K., Kraus, N., & McGee, T. (1998). The time cour se o f auditory perceptual learning: Neurophysiological changes during speech sound training. Neuroreport 9(16), 3557 3560. Van Hessen, A.J., & Schouten, M.E.H. (1992). Modeling phoneme perception: II. A model of stop consonant discrimination. Journal of t he Acoustical Society of America 92, 1856 1868. Van Voorhis, S. & Hillyard, S. A. (1977). Visual evoked potentials and s elective attention to points in space. Perception and Psychophysics 22(1), 54 62. Wang, Y., Jongman, A., & Sereno, J. A. (2003a). Acous tic and percept ual evaluation of Mandarin tone productions before and after perceptual training. Journal of the Acoustical Society of American, 113(2), 1033 1043.

PAGE 143

143 Wang, Y., Sereno, J. A., Jongman, A., & Hirsch, J. (2003b). fMRI evi dence for cortical modific ation during learning of Mandarin lexical tone. Journal of Cognitive Neuroscience, 15(7), 1 9. Wang, Y., Spence, M. M., Jongman, A., & Sereno, J.A. (1999). Training American listeners to perceive Mandarin tone. Journal of the Acoustical Society of America 106, 3649 3658. Wayland, R., & Guion, S. (2004). Training native English and native Ch inese speakers to perceive Thai tones. Language Learning, 54(4), 681 712. Wayland, R., Herrera, E., & Kaan, E. (2010). Effects of musical experience and training on pitch contour perception. Journal of Phonetics 36, 250 267. Wayland, R., & Li, B. (2008). Effects of two training procedures in cross language perception of tones. Journal of Phonetics 36, 250 267. White, C. M. (1981). Tonal perception errors and interference from English intonation. Journal of Chinese Language Teachers Association 16, 27 56. Woldorff, M.G., Hackley, S. A., & Hillyard, S. A. (1991). The effects of channel selective attention on the mismatch negativity wave elicited by deviant tones. Psychophysi ology 28(1), 30 42. Wong, P.C.M., & Perrachione, T.K. (2007b). Learning pitch patterns in l exical identification by native English speaking adults. Applied Psycholinguistics 28, 565 -585. Wong, P.C.M., Sk oe, E., Russo, N. M., Dees, T., & Kraus, N. (20 07a) Musical experience shapes human brainstem encoding of linguistics pitch patterns. Nature Neuroscience 10(4), 420 422. Zachau, S., Rinker, T., Krner, B., Kohls, G., Maas, V., Henninghausen, K., & Schecker, M. (2005). Extracting rules: early and late mis match negativity to tone patterns. Neuroreport 16, 2015 2019. Zwissler, R. (1971). An investigation of pitch discrimination skills of first grade children identified as accurate singers and those identified as inaccurate singers. Doctoral dissertation, Un iversity of California at Los Angeles.

PAGE 144

144 BIOGRAPHICAL SKETCH Shuang Lu was born in Jinan, Shandong Province, China. She received her B.A. in English from Shandong Jianzhu University in 2005. She earned her M.A. in Linguistics from Leiden University in th e Netherlands in 2006 She joined the University of Florida a s a Ph.D. student in the fall of 2008 and received her Ph.D. in Linguistics in the summer of 2013.