Variability in Distributional Structure Influences Learning of Mandarin Accented Speech Restricted; Files Only

Abstract

Listeners adapt to variable speech signals by learning linguistic features and talker characteristics. In a study by Clarke & Garrett (2004) (later CG04), native English listeners adapted to a Mandarin-accented talker with exposure of less than a minute. One mechanism that explains how we learn highly variable speech is distributional learning. Specifically, listeners adjust their phonetic category boundaries when learning unfamiliar speech like foreign accents. Structurally, listeners are sensitive to the distribution of acoustic-phonetic cues such as consonant voice onset times when shifting their phonetic categories. However, less studies focused on whether the structure of distribution would affect the learning of vowels in accents. This study uses the cross-modal word-matching task from CG04 to test two different structures of vowel variation in mandarin-accented english. Participants were exposed to sentences with final word that either have consistent ambiguous shift of vowel /i/ or /ɪ/ or inconsistent inputs that contains both ambiguous and canonical forms. Response times and error rates of 246 native English listeners are analyzed. Listeners showed a robust learning towards accented speech across all conditions, indicated by a significant decrease of response time and error rate during exposure. At the exposure phase, participants showed greater difficulty with inconsistent vowel input, as reflected by significantly slower response times compared to the consistent condition. This slowing effect was significantly stronger in the EE bias group than in the IH bias group, suggesting that Mandarin speakers’ EE bias may be more difficult to adapt due to its proximity to IH. However, error rates at test indicate that participants ultimately formed a stronger pattern when exposed to inconsistent input throughout the experiment, resulting in significantly lower error rates than the consistent condition. Overall, these results suggest that listeners may prefer variable distribution in perceptual learning. 

Table of Contents

Introduction   3

Talker-specific encoding and talker-independent learning               4

Fast adaptation of Foreign Accented Speech           6

Perceptual Learning and Lexical Adaptation              7

Systematicity and Distributional Learning 9

Mandarin-Accented English Vowels                12

Hypotheses    14

Method               15

Participants    15

Design 15

Materials           16

Sentence recordings.                 16

Acoustic manipulation.           17

Norming study.              17

Procedure         17

Analysis pipeline.        20

Results                22

Exposure Phase            22

Error rate.          22

Response time.             22

Test Phase        24

Error rate.          24

Response time.             25

Discussion       26

Rapid adaptation to accented speech          27

Consistent versus variable exposure: evidence for the Varying Hypothesis         28

Initial processing costs of variability                30

Directionality of vowel shifts and phonological constraints            31

Distributional learning in accented speech perception     32

Implications for speech perception                 34

Conclusion      35

Figures                 37

Figure 1               37

Figure 2               38

Figure 3               39

Figure 4               40

Tables 41

Table 1                 41

Table 2                 42

Table 3                 43

Table 4                 44

Appendices    45

Appendix A. Stimuli List with Visual Probe Words 45

Appendix B. Data analysis of RT and error rate with vowel bias type          48

Appendix C. Figures of RT and error rate with vowel bias type       52

References      56

About this Honors Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Emory College
Department	Linguistics
Degree	B.A.
Submission	Honors Thesis
Language	English
Research Field	Language, Linguistics Psychology, Cognitive
Keyword	Statistical learning Speech perception Foreign accent Phonetics
Committee Chair / Thesis Advisor	Lynne Nygaard, Emory University
Committee Members	Yun Kim, Emory University Hiram Maxim, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	File download under embargo until 28 May 2027	2026-04-07 20:09:28 -0400	File download under embargo until 28 May 2027

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions