The Intelligibility of Chinese-Accented English

Home , Be language, Gambian English, General American English, Hong Kong English, Oi (interjection)

to International and American Students at a U.S. University

Dissertation

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Jocelyn Brooks Hardman, M.A.

Graduate Program in Education

The Ohio State University

2010

Dissertation Committee:

Keiko Samimy, Co-Advisor

Mary Beckman, Co-Advisor

Shari Speer

Abstract

This study investigated the intelligibility of Chinese graduate students to their

Indian, Chinese, Korean, and American peers. Specifically, the researcher sought to determine the teaching priorities for English for Academic Purposes in the US, where listeners have a wide variety of native languages.

Research on Second Language Acquisition (SLA), International Teaching

Assistants, and English as a Lingua Franca (ELF) has not provided sufficient empirical data on the factors that affect intelligible English communication among academic professionals with many native languages (L1). SLA has focused on the processes and factors affecting the acquisition of second language (L2) phonology; and ITA research has focused on the communication needs of international graduate students teaching

American undergraduates. Both perspectives examine the intelligibility of foreign- accented speech to native English-speaking listeners. World Englishes (WE) and ELF argue for more research from the perspective of L2 listeners, which thus far has largely been limited to linguistic descriptions and case studies.

A psycholinguistic word-recognition-in-noise study was designed to examine to what extent a talker’s L1 and segmental pronunciation accuracy affected intelligibility, and how this varied by a listener’s L1 and word familiarity. Participants included 6 male graduate students (Chinese & American) as talkers and 72 graduate students (Indian,

ii Chinese, Korean, & American) as listeners. The oral English proficiency level of the

international participants was held constant at “graduate TA certification” and all

American listeners were natives of Ohio. Talkers were recorded reading 60 sentences

from the Bamford-Kowal-Bench Standard Sentence Lists, revised for American English.

The stimuli were mixed with white noise at a +5 dB signal-to-noise ratio and presented in

a counterbalanced design to listeners, who transcribed the sentences they heard.

Intelligibility was calculated using a dichotomous measure of key word transcription accuracy.

A series of logistic regression mixed-effects models revealed that talker L1, listener L1, and word familiarity were significant predictors of intelligibility. Talker segmental accuracy did not significantly predict intelligibility. American English L1 talkers were more intelligible than Mandarin L1 talkers to all listeners, except to

Mandarin L1 listeners, who found them equally intelligible. This is an important finding for English Language Teaching, since most teachers are either native speakers of English or share the L1 of their students. Special effort must be made to include more diversity of accents in the audio and video materials for international students preparing to study or conduct research in English L1 countries.

English for Academic Purposes curricula cannot assume that one size fits all of the diverse listeners encountered in the academic context. As the findings indicated, intelligibility improved as listener word familiarity increased, and was also affected differently according to the listener’s L1. Consequently, in the academic setting more attention should be paid to increasing listeners’ discipline-specific vocabulary, which

iii would benefit both international and American listeners. More attention should also be paid to providing both international and American university students with the linguistic training necessary to accommodate the range of accent diversity that has become a reality in higher education today.

Dedication

Dedicated to my grandparents,

Justice H. Hale McCown

1914 – 2005

and

Helen L. McCown

1913 – 2006

Acknowledgments

During the long years and hard work involved in my quest to know better how

people understand one another, George Bernard Shaw’s words have helped me maintain

a sense of humor and perspective: “The biggest problem with communication is the

illusion that it has taken place.” Despite agreeing with this observation, I have still

persevered so that I could contribute a few more drops of knowledge to the ocean of

human speech communication.

Many people have helped me along the way by providing resources and expertise,

both theoretical and practical. In the discipline of education, we frequently separate researchers from teachers who are identified as “the practitioners.” But my experience as both a researcher and teacher has taught me that research is also a practice, and mentors are just as crucial. In this vein, I would like to acknowledge the support of my dissertation committee members, Dr. Keiko Samimy, Dr. Shari Speer, and Dr. Mary

Beckman. As my academic advisor, Dr. Samimy raised fruitful questions that led to my dissertation topic. She also gave me tremendous freedom to explore what really fascinated me and recommended contacts for me in other departments where I could pursue my research questions.

The Spoken English Program was my professional home during my doctoral studies and many of my colleagues have been a source of encouragement and support.

Directors Susan Sarwark and Kathi Cennamo have opened doors in TESOL for me,

vi Administrator Andi Sondrini was prompt and responsive to every request for newly certified ITAs, Jane Smirniotopoulos donated labor and expertise in transcribing Chinese- accented English, and Laurie Maynell shared references and commiserated as we both led dual lives in English for Academic Purposes and Linguistics.

The Linguistics Department has been an invaluable resource at every stage of my research. Dr. Shari Speer was available for consultation on psycholinguistic research design and made available her laboratory equipment and facilities. Other graduate students in the department offered tutorials on E-Prime and donated their time as materials translators and pilot study subjects. Undergraduates Nicole Holliday and

Danielle Brown also served as very dependable research assistants. And of course, no data collection is possible without subjects, so I am especially grateful to the graduate students who took time out of their very busy lives to participate in my study.

Dr. Mary Beckman has guided my research throughout, and has unstintingly devoted the time, attention, and nurturing involved in a true mentor-apprentice relationship; even offering overnight stays when data collection went long into the night.

She has taught me everything I know about phonological theory, Praat, and R; any remaining lacunae are my responsibility. I will never hope to attain her level of knowledge and her work ethic, but she embodies the ideal that I will emulate for the rest of my career.

Researchers at other institutions have also been very generous with their advice and resources. Dr. Ann Bradlow provided copies of the written materials as well as prepublication articles for my reference. Dr. James Flege kindly gave me a tutorial on designing word familiarity tests though he had retired comfortably to Italy. Dr. Stefan

vii Frisch provided words with low to moderate familiarity ratings which I used as filler

words and Dr. Ben Munson and graduate students at the University of Minnesota have

already begun using the Buckeye GTA Corpus and have added much needed text grids

and annotations.

As noted in the dedication, my grandparents Hale and Helen McCown have

supported my education beginning with my undergraduate degree. Without their

financial support, I would not have been able to finish my doctoral degree. My uncle

Bob McCown and his wife Ranna, opened their home to me when I had broken my arm and couldn’t commute. My stepfather Saul Young shared insights and scholarship with

me in the field of business, which has similarities to that of education in terms of the

separation between research and practice. These generosities and camaraderie are all the

more missed since they cannot be with me to celebrate the culmination of my studies.

I wish to acknowledge the many kindnesses and day-to-day support my family

has given me from my mother and my mother in law taking care of my two young boys

when I was in school, to my husband resigning himself to full-time grocery shopper and cook so that I could dedicate the time and intensive study that research requires. And most importantly, I want to thank my husband Jim and our now much older boys, Zeke and Jacob, my mother Lynn McCown and sister Hadley Messner, my best friends

Violette-Anne Onfroy-Curley and Karen Crist, for their unfailing love, companionship, and approval, with or without a Ph.D. after my name. As much as I love traveling, both to other lands and undiscovered countries of the mind, there is still nothing better than coming home to you.

viii

VITA

1987 B.A. Modern Foreign Languages and

Literature (Spanish and Italian),

Kenyon College, Gambier, Ohio

1995 M.A. Hispanic Literature,

Miami University, Oxford, Ohio

1993 – 1995 Graduate Teaching Assistant, Spanish

Miami University, Oxford, Ohio

1995 – 1996 Instructor, Spanish

Wittenberg University, Springfield, Ohio

1996 – 1997 Instructor, Spanish

Miami University, Oxford, Ohio

1997 – 1998 Graduate Teaching Associate, Spanish

The Ohio State University, Columbus, Ohio

1999 – 2000 Graduate Administrative Associate,

Spoken English Program,

The Ohio State University, Columbus, Ohio

ix 2000 – 2006 Graduate Teaching Associate,

Spoken English and ESL Composition,

The Ohio State University, Columbus, Ohio

2006 – 2008 Visiting Assistant Professor,

Spanish and Second Language Acquisition,

Director, Language Program

Antioch College, Yellow Springs, Ohio

2009 – present Lecturer of Spanish,

Department of Languages,

University of Dayton, Dayton, Ohio

Publications

Hardman, J.B. (2003). The advantages of accent diversity in ESL classes. International

Teaching Assistants: A TESOL Interest Section Newsletter, 7(3), 2-7.

Fields of Study

Major Field: Education

Minor Fields: Spanish Linguistics

ESL Literacy

Table of Contents

Abstract ii

Dedication v

Acknowledgments vi

Vita ix

List of Tables xvii

List of Figures xix

Chapter 1: Introduction 1

1.1 Introduction 1

1.2 Statement of the Problem and Research Gap 3

1.3 Objectives of the Study and Research Questions 5

1.4 Significance of the Study 7

1.5 English Language Teaching Contexts at the University Level 8

1.5.1 The US Context: ITA Program Development 9

1.5.2 The Chinese Context: College ELT 11

1.6 Literature Review 13

1.6.1 L2 Phonological Acquisition Theories 13

1.6.2 Research on International Teaching Assistants 16

xi 1.6.3 World Englishes and English as a Lingua Franca 18

1.7 Methods 19

1.8 Definition of Key Terms 20

1.9 Basic Assumptions of the Study 25

1.10 Limitations of the Study 25

1.11 Summary 28

Chapter 2: Literature Review 29

2.1 Introduction 29

2.2 L2 Phonological Acquisition Research 31

2.2.1 Terminology in L2 Research 32

2.2.2 Theories of L2 Phonological Acquisition 34

2.2.2.1 Cross-Language Transfer 34

2.2.2.2 Universals, Transfer, and their Interaction 42

2.3 Factors Affecting Listening Comprehension and Intelligibility 47

2.3.1 Talker Factors that Affect L2 Speech Production 48

2.3.2 Listener Factors that Affect L2 Speech Perception 51

2.3.3 Text Factors: Word Frequency or Word Familiarity 54

2.3.4 Task Factors 56

2.3.4.1 Talker Tasks 56

2.3.4.2 Listener Tasks 57

2.3.5 Phonetic-Acoustic Factors 58

2.3.5.1 Segmental and Prosodic Production 59

xii 2.3.5.2 Speech Rate 63

2.4 Research on Accommodations of the Native Speaker 65

2.4.1 Speaking Accommodations 65

2.4.1.1 Speech Rate 65

2.4.1.2 Pause Phenomena 69

2.4.1.3 Clear Speech 72

2.4.2 Listening Accommodations 74

2.5 English for Specific Purposes and EAP 79

2.5.1 International Teaching Assistants 80

2.5.1.1 Undergraduate Attitudes toward ITAs 81

2.5.1.2 EAP and ITA Research 85

2.5.1.3 Discourse Analysis and ITA Intelligibility 88

2.5.2 International Graduate Students 90

2.6 World Englishes (WE) and English as a Lingua Franca (ELF) 95

2.6.1 Terminology in WE Theory 96

2.6.2 Terminology in ELF Theory 99

2.6.3 Linguistic Description Research in WE 101

2.6.4 Linguistic Description Research in ELF 104

2.6.5 Intelligibility Studies in WE and ELF 107

2.6.6 Listener Factors that Affect ELF Intelligibility 109

2.6.7 Phonetic-Acoustic Factors that Affect Intelligibility 113

2.6.8 Research and Teaching Implications of WE and ELF 118

xiii 2.7 The Literature Gap 122

2.8 Summary 125

Chapter 3: Methods 128

3.1 Introduction 128

3.2 Research Design 129

3.3 Location of the Study 131

3.4 Stimuli 138

3.4.1 Materials 138

3.4.2 Talkers 141

3.4.2.1 Talker Variables 133

3.4.2.2 Talker Speech Rates 147

3.4.2.3 Talker Segmental Accuracy 148

3.4.3 Talker Recording Procedure 151

3.5 Experiment Paradigm 152

3.5.1 Pilot Study 153

3.5.2 Listener Subject Selection 155

3.5.3 Main Experiment Procedure 157

3.6 Data Collected 161

3.6.1 Listener Variables 161

3.6.2 Intelligibility Measure 162

3.6.3 Word Familiarity Ratings 162

3.7 Data Analysis 163

xiv 3.8 Limitations 164

3.9 Summary 168

Chapter 4: Results 171

4.1 Introduction 171

4.2 Effects of Talker and Listener L1 171

4.2.1 Effects of Listener L1 vs. Individual Listener 176

4.2.2 Effects of Talker L1 vs. Individual Talker 177

4.2.3 Individual Key Word Random Effects 179

4.3 Effects of Talker Segmental Accuracy 181

4.4 Effects of Listener Word Familiarity 181

4.4.1 Word Familiarity Ratings by Listener L1 184

4.4.2 Key Word Effects of Model 0 versus Model 3 186

4.5 Key Word Analyses 190

4.5.1 Segmental Error Patterns of Mandarin L1 Talkers 190

4.5.2 Intelligibility Results by Listener L1 200

4.6 Summary 208

Chapter 5: Discussion and Conclusions 210

5.1 Introduction 210

5.2 Effects of Talker and Listener L1 211

5.3 Effects of Talker Segmental Accuracy 212

5.4 Effects of Listener Word Familiarity 213

5.5 Key Word Analyses 214

xv 5.6 Recommendations for Further Research 216

5.7 Conclusions 217

5.7.1 Contributions to Second Language Speech Research 217

5.7.2 Contributions to English Language Teaching 219

References 223

Appendix A: Talker Background Questionnaire 255

Appendix B: Listener Background Questionnaire 260

Appendix C: Recruitment Letter to Mandarin Talkers 265

Appendix D: Recruitment Letter to American English Talkers 266

Appendix E: Recruitment Letter to Mandarin Listeners 267

Appendix F: Recruitment Letter to Hindi Listeners 268

Appendix G: Recruitment Letter to Korean Listeners 269

Appendix H: Recruitment Letter to American English Listeners 270

Appendix I: Recruitment Response Form 271

Appendix J: Informed Consent Form – Talkers 272

Appendix K: Informed Consent Form – Listeners 276

Appendix L: Materials List 280

Appendix M: IPA – CMU Phoneme Correspondence Key 285

xvi

List of Tables

3.1 Sentence item by listening subject rotation 130

3.2 Top 15 US institutions hosting international students, 2008–2009 132

3.3 OSU graduate student enrollment by department, 2000–2009 134

3.4 Fields of study of international students in the US, 2008–2009 135

3.5 Countries of origin of international students in the US, 2008–2009 136

3.6 Countries of origin of international students at OSU, 2008–2009 137

3.7 Mean percent correct scores for each BKB-ST list 139

3.8 Chinese talker variables 145

3.9 American talker variables 146

3.10 Mean percent correct talker key word accuracy ratings 150

4.1 Statistical significance of the fixed effects of the mixed-effects base model 175

4.2 Variance and standard deviation of random effects of the base model 175

4.3 Statistical significance of the fixed effects of Model 3 182

4.4 Variance and standard deviation of random effects of Model 3 183

4.5 Key words of ME1 mistranscribed by 75% or more listeners 192

4.6 Key words of ME2 mistranscribed by 75% or more listeners 193

4.7 Key words of ME3 mistranscribed by 75% or more listeners 194

4.8 Vowel errors in mistranscribed words spoken by Chinese talkers 196

4.9 Consonant errors in mistranscribed words spoken by Chinese talkers 198

4.10 American L1 listener percent mistranscriptions of key words 200

4.11 Hindi L1 listener percent mistranscriptions of key words 202

xvii

4.12 Mandarin L1 listener percent mistranscriptions of key words 204

4.13 Korean L1 listener percent mistranscriptions of key words 206

xviii

List of Figures

3.1 Talker speech rate analyses measured in words per second (WPS) 147

4.1 Word intelligibility means by talker L1 and listener L1 172

4.2 Listener L1 group effects vs. within-group variation, Model 0 176

4.3 Individual talker effects vs. talker L1 group effect, Model 0 178

4.4 Individual key word random effects, Model 0 180

4.5 Word familiarity rating categories by listener L1 185

4.6 Listener L1 effects and key word effects, Model 0 vs. Model 3 186

4.7 Individual key word effects, Model 3 188

4.8 Correlation of key word effects, Model 0 vs. Model 3 189

xix

Chapter 1: Introduction

1.1 Introduction

Today’s global context includes second language (L2) speakers of English from so many different first language (L1) backgrounds and L2 proficiency levels that mutual intelligibility has become a greater challenge. Second language acquisition (SLA) research and International Teaching Assistant (ITA) programs in American universities are interested in the degree to which L2 English learners approximate the proficiency of monolingual L1 speakers. Therefore, the point of reference is necessarily an L1 variety of English, and the standard for intelligible English pronunciation is set by L1 English listeners. This has set the stage for researchers of SLA, World Englishes (WE), and

English as a Lingua Franca (ELF) to make a case for which standards should be used to determine the intelligibility of English for international communication.

While an L1 English reference fits the research and teaching model for L1

English contexts of usage, scholars of WE and ELF contend that many L2 speakers of

English today will not be interacting primarily with L1 English speakers. They assert that many L2 English users already speak their desired target English variety, such as

Indian English, even if it may not be an L1 variety of English. From the WE and ELF perspectives, L1 and L2 speakers of English are equally responsible for effective

international communication, and research and teaching in SLA has been called upon to better reflect this new, more diverse global reality. This dissertation therefore evaluates the intelligibility of L2 speakers of English to both L1 and L2 listeners to help determine the priorities for pronunciation teaching of English in the international context of academia.

Since studies on listening comprehension, pronunciation, and intelligibility have been conducted from a variety of theoretical perspectives and within many disciplines, including linguistics, psychology, SLA, WE, and ELF, several terms have been used that have been defined in various ways. Listening comprehension studies generally focus on investigating the oral comprehension abilities of L2 listeners. The findings inform models of L2 perception at the theoretical level, and the teaching of listening skills for L2 learners at the pedagogical level.

Although the terms “pronunciation” and “intelligibility” are often used interchangeably, they refer to related but distinct concepts. “Pronunciation” refers to the phonetic-acoustic features of an L2 learner’s speech production, including both segmental and prosodic, or suprasegmental, features. “Intelligibility” refers to the extent to which an L2 learner’s speech or pronunciation is accurately received by a listener – that is, whether the listener recognizes the intended word or sentence. Using listener feedback, the current intelligibility study makes recommendations about what should be prioritized in pronunciation teaching for the academic speech community.

1.2 Statement of the Problem and Research Gap

SLA research on L2 phonological acquisition provides insight into the way learners perceive and produce the sounds of a new language. Intelligibility studies identify the aspects of a foreign accent that impact negatively on intelligibility in a specific context, which helps prioritize the sounds that must be produced as accurately as possible for communication to succeed. In SLA research, native speakers of the target language are generally used as baseline models of L2 speech production. Similarly, native speakers are used as judges to determine L2 speech intelligibility. This research has traditionally overlooked the way that L2 English is perceived by other L2 English listeners.

In World Englishes (WE) theory, Kachru (1985) argued for three concentric circles of English usage: Inner, Outer, and Expanding. The Inner and Outer Circles primarily use English intranationally, while the Expanding Circle uses English mainly for international communication. As a result, Kachru asserted that native norms are appropriate for the Inner Circle, but they do not respect the pragmatic sociolinguistic usage of English in the Outer Circle. Instead, he recommended that local norms and standards of English were more appropriate, since they were more desirable and accessible in the local context. In fact, Kachru and Nelson (1996) recommended rejecting the concept of interlanguage, popularized by Selinker (1972, 1992), for the

Outer Circle, since the local variety of English, rather than a native variety, was the target language.

ELF theory borrows the sociolinguistic theoretical framework that WE employs to legitimize nativized varieties of English and applies it to the Expanding Circle.

According to ELF, since L2 speakers of English now outnumber L1 speakers worldwide,

L2 listeners should be used to determine the standards of intelligibility for L2 English speech, rather than an L1 English model. Unlike WE scholars, however, ELF scholars view Expanding Circle interlocutors as second language learners, and therefore make pronunciation teaching recommendations for improving intelligibility in ELF contexts.

As a result of spanning both sociolinguistic and SLA theory, ELF has not yet resolved the role of native speakers of English; nor has sufficient research been conducted on the differing levels of intelligibility required for different contexts of use.

Within the graduate school context in the United States, there is a greater need to understand the standards of L2 English intelligibility, due to the consistently increasing numbers of international students pursuing higher education degrees. International graduate students have been researched in some depth, from the SLA and ITA program perspectives. SLA research has focused on the intelligibility of international graduate students who have just arrived at US campuses. The data from these studies are important to SLA, in that they shed light on the impact of immersion in English. From a teaching standpoint, however, it is already clear that new international graduate students often need to improve their pronunciation. In fact, SLA research from the 1970s and

1980s led to the creation of ITA programs in most universities. These programs now assess graduate students’ oral proficiency level upon their arrival if they are being considered for an ITA position. Furthermore, they are charged with addressing any

problems with intelligible speech production that these students may have, as determined by L1 English listeners. Research has not yet documented in significant phonetic- acoustic detail, however, the L2 English pronunciation of ITAs with a variety of L1s.

In addition, outside the classes they teach, international graduate students struggle with expectations for them to participate orally in seminars and classrooms, contribute to research projects conducted with other international graduate students and professors, and present their research to international academic audiences. However, very little research has examined the oral English needs of international graduate students in their roles as students and researchers. Many departments at US universities enroll a majority of international graduate students and some departments employ many international professors and instructors. The few research studies that have examined the oral English needs of international graduate students have not included the perspective of their international peers. Instead, they have been largely conducted from the perspective of L1

English interlocutors. The current study was designed to address this gap in the research by investigating the intelligibility of international graduate students as determined by their international peers.

1.3 Objectives of the Study and Research Questions

This study had two main objectives. First, the researcher sought to compile a speech corpus of ITAs and American graduate students, the Buckeye GTA Corpus, which she consulted to describe the segmental phonological features of Chinese-accented

English at the oral proficiency level of “graduate TA certification.” Second, the researcher investigated the intelligibility of Chinese graduate students as determined by

their Indian, Chinese, Korean, and American peers. Specifically, she sought to determine to what extent the talkers’ segmental pronunciation accuracy and native language affected intelligibility, and to what extent intelligibility varied according to the listeners’

L1s and word familiarity. These speech recordings and analyses of the talkers’ pronunciation, combined with the intelligibility data from international and American graduate students can inform English pronunciation teaching priorities at both the college level in China and the graduate ITA level in the US.

This study sought to answer the following research questions:

Research Question #1:

To what extent does the talker’s L1 (native speaker: General American

English L1; and foreign-accented: northern Mandarin L1)

affect intelligibility, as measured by listeners’ written transcriptions?

Research Question #2:

To what extent does the listener’s L1 (Hindi, northern Mandarin, Korean,

and General American English) affect intelligibility?

– Is there any interaction between talker L1 and listener L1 that

would evidence an “interlanguage match” benefit, when Mandarin

L1 listeners are transcribing Mandarin L1 talkers’ speech?

Research Question #3:

To what extent does talker segmental accuracy, as measured by key word

accuracy ratings based on whether the talker’s phonemic pronunciation

matched a General American English phonemic transcription of that word,

affect intelligibility?

Research Question #4:

To what extent does listener word familiarity, as measured by a 5-point

Likert scale, affect intelligibility?

1.4 Significance of the Study

The data collected in the current study can contribute to English language teaching as well as research. The recordings of the talkers and listeners have been combined into a speech corpus the researcher termed the “Buckeye GTA Corpus.” This corpus can be used for teaching both English pronunciation and listening comprehension.

For pronunciation teaching, the recordings can serve as models of both native and TA- certified English and may help other aspiring international graduate students qualify as

TAs in a US university. For listening comprehension teaching, students in their home countries who may not have been exposed to many international accents can become more familiar with multiple talkers of the accents in English they are most likely to encounter in a US graduate school.

For research purposes, the quality of the “Buckeye GTA Corpus” will make the phonetic-acoustic features of ITA speech accessible not only for the segmental investigation of the current study, but also for prosodic investigations of future

researchers. This corpus can be used to linguistically describe both the native and nonnative speech of graduate TAs at a US university.

The listener intelligibility data collected by the current study also has significance for SLA research and teaching applications. The graduate students who participated in this study had all passed the minimum proficiency level necessary for teaching contact with American undergraduate students, so the findings will help prepare international graduate students for the next highest proficiency level in English, that is, being intelligible to their peers within the international academic community. Specifically, the findings reveal which sounds a talker may vary without impeding intelligibility and which sounds should constitute the highest priorities for English pronunciation teaching for the diversity of listeners in the academic context.

1.5 English Language Teaching Contexts at the University Level

This study is important for English language teaching (ELT) both in the United

States and China. In the US context, Chinese students are admitted to graduate school based on their scores on the Test of English as a Foreign Language (TOEFL), which has included an oral proficiency section since 2007. In addition, many Chinese graduate students seek funding support in the form of teaching assistantships, which are granted on the basis of demonstrated oral English proficiency. Therefore, the importance of oral

English proficiency, and not just reading comprehension or grammatical ability, increases for students seeking graduate study, teaching assistantships, research opportunities and professional employment in English-speaking countries.

In the Chinese context, it would be helpful to know how English language teachers could better prepare Chinese students for graduate study in American universities. Due to the diversity of the context of US graduate schools, attention should also be paid to how English pronunciation teaching can better focus on especially those sounds which are found to facilitate intelligible communication with both international and American interlocutors.

1.5.1 The US Context: ITA Program Development

The “foreign TA problem” or dilemma refers to the combination of the limited oral English proficiency of ITAs and increasing university dependence on them due to decreasing numbers of Americans pursuing advanced graduate degrees, especially in math, science, and engineering. US university ITA programs are by necessity focused, intensive oral English programs. They are classified as English for Specific Purposes

(ESP), or English for Academic Purposes (EAP), with the specific purpose of training international graduate students to teach monolingual English-speaking undergraduates: in the classroom, in the lab, and during office hours.

Complaints about the “foreign TA problem” from university undergraduates and their parents led many states to enact legislation in the mid-1980s requiring oral English proficiency of TAs. For example, the law for the state of Ohio that went into effect for the 1986-87 academic year reads as follows:

The board of trustees of each state university, college of medicine, technical

college, state community college, community college, and the board of trustees or

managing authority of each university branch shall establish a program to assess

the oral English language proficiency of all teaching assistants providing

classroom instruction to students and shall ensure that teaching assistants who are

not orally proficient in the English language attain such proficiency prior to

providing classroom instruction to students. (Ohio Revised Code 3345.281,

1986)

By 1992, twelve states (California, Florida, Illinois, Iowa, Minnesota, Missouri, North

Dakota, Ohio, Pennsylvania, South Carolina, Tennessee, and Texas) adopted oral English proficiency laws for ITAs. Six more states (Arizona, Georgia, Kansas, Michigan,

Oregon, and Wyoming) instituted university regulations for oral English testing of ITAs

(Byrd, 1992).

The law for the state where this study was conducted determines the mission of the university’s ITA program as follows: (1) to assess prospective ITAs’ oral English proficiency; (2) to provide coursework for those who do not meet the minimum proficiency level, and (3) after coursework, to administer teaching performance testing for TA certification. The university under study has defined “oral English proficiency” for TA certification in the following ways: (1) as a score of at least 230 out of 300 on the

SPEAK test, the institutional version of the Educational Testing Service’s Test of Spoken

English; or (2) the successful completion of the Mock Teaching Test, a teaching performance test developed in house. Although the successful completion of one of these tests is required for TA certification at the university under study, international graduate students may hold graduate assistantships that do not require “direct student contact” – e.g., as graders – until they become certified to teach. Departments will generally support

their international graduate students for at least one year in this way to allow time to achieve teaching certification.

ITA programs will benefit most from the intelligibility data from the American listeners in this study, and the intelligibility data from all listeners will be relevant for the

Chinese context. A description of the TA-certified level of oral proficiency for Chinese- accented English will be helpful for the context of English as a Foreign Language (EFL) in China, where students with the same native language background will be together in the classroom. In the ESL or ITA context in the US, where students with many L1s interact in a classroom, these data will help improve their intelligibility to one another in the language classroom as well as in their own departments.

1.5.2 The Chinese Context: College English Language Teaching

Although the earliest mention of foreign language teaching in China dates back to

1289 for trade purposes, followed by Western missionaries in the 16th and 17th centuries,

the teaching of English in China in the modern period begins in 1949 with Chairman

Mao’s rule (Bolton, 2003). According to Lam (2002), foreign language teaching in

China underwent six phases before, during, and after the Cultural Revolution (1966-

1976) and from the early 1970s, ELT began to enjoy both support and expansion. By

1999, over 50 million Chinese secondary students were studying English, and there were

nearly 500,000 full-time secondary English teachers (Adamson, 2002).

At the undergraduate level in China, English is a required subject and the College

English Test is given nationally to all college-bound high school students (Yang, 2001,

cited in Adamson, 2002). At the university, there are six levels, or “Bands,” for English

proficiency. As of 2000, all college students in China have been required to pass the

Band 4 level prior to graduation, while eight leading universities are authorized to set higher English proficiency standards (Cheng, 2002). According to Wang (1999), there are two strands of college English in China: one for English majors and the other for non-

English majors. The majority of students studying English at the college level are non-

English majors. The non-English major strand consists of two years or four semesters of study, with classes meeting no more than four hours a week. The primary goal for non-

English majors is to “develop a relatively high level of reading competence so as to be able to extract the necessary information from the readings in his or her own field of specialization” (Wang, 1999, p. 46).

The majority of Chinese students who come to the United States for graduate work are non-English majors. While their main focus in college in China was reading proficiency and TOEFL preparation, many seek funding support in the form of a TAship from a US university. In addition, some US university departments require TA certification for a graduate degree, regardless of whether the student teaches or not.

However, many Chinese students are unable to secure TA certification – and the funding and professional advancement it provides – largely because their oral skills are not sufficiently developed.

The current study not only provides detailed information about the range of pronunciation acceptable for TA certification, but also indicates the intelligibility of this proficiency level to the L1 groups most commonly found on US university campuses.

1.6 Literature Review

Below is a summary of Chapter 2, which reviews the three bodies of literature that contributed to the theoretical framework that informs this research. These include research in Second Language Acquisition, specifically the acquisition of L2 phonology,

English as a Lingua Franca, and International Teaching Assistants.

1.6.1 L2 Phonological Acquisition Theories

L2 phonological acquisition research has found that two major influences affect second language speech perception and production: the native language and the universal principles of language. Contrastive Analysis (CA) first predicted that any sounds in the second language that did not exist in the first language would cause learner errors (Lado,

1957). Since subsequent research found that some learners had no trouble acquiring new sounds in the second language, a weak version of CA was used to explain learner errors post hoc in terms of transfer or interference from the first language (Wardhaugh, 1970).

However, since this version of CA could not predict which errors a learner would make, a moderate version was posited, which incorporated degrees of similarity between the L1 and the L2 (Oller & Ziahosseiny, 1970).

The Perceptual Magnet Theory suggested that L1 sounds function as magnets, pulling new sounds toward them and inhibiting the learner’s ability to perceive L2 sounds

(Kuhl, 1991, 1993). Unfortunately, this theory could not account for the successful acquisition of sounds in the L2 which did not exist in the L1, nor for attrition or loss of the L1. The Perceptual Assimilation Model (PAM) was thus proposed, which hypothesized a common phonological space for all languages learned (Best, 1994). The

notion of similarity between an L1 and an L2 sound was used to predict whether an L2 sound would be assimilated to an L1 phonetic category or whether it would be perceived as dissimilar enough to either prompt the creation of a new category or be classified as a non-speech sound.

In the area of second language speech production, the Critical Age Hypothesis

(CAH) from first language acquisition (Lenneberg, 1967) was extended to include second language acquisition (Scovel, 1969; Long, 1990). Based on brain lateralization at puberty, which was assumed to trigger an end to neuronal plasticity, and evidenced by the accentedness of long-term bilingual speech, the CAH posited a critical age or period after which second language speech production could not be native-like. However, since research on bilinguals revealed both early L2 learners who failed to achieve native-like production and late L2 learners who did, the Speech Learning Model (SLM) was proposed, which accounted for these data (Flege, 1995). The SLM also hypothesized a shared phonological space for speech sounds and used ‘equivalence classification’ to explain why a learner might not create a new phonetic category for an L2 sound perceived as similar to an L1 sound. In addition, findings in neuroscience, demonstrating neuronal plasticity after brain lateralization, undermined the original biological justification for a strong CAH. Therefore, since in general the earlier one learned a language the better the ultimate attainment level, then the concept of a ‘sensitive’ or

‘optimal’ period was favored over a ‘critical’ one.

Besides cross-linguistic transfer, universal principles of language, such as

Universal Grammar (Chomsky & Halle, 1968), stylistic features (Bayley & Preston,

1996; Tarone, 1988), and markedness (Greenberg, 1966, 1978; Jakobson, 1941;

Trubetskoy, 1939) were also found to play a role in second language phonological acquisition. The Markedness Differential Hypothesis (Eckman, 1977) combined the

CAH with typological markedness, which refers to the distributions of linguistic representations in binary opposition across languages (those which are more frequent or commonly found are unmarked and those which are infrequent are marked) (Jakobson,

1941; Trubetskoy, 1939). However, the L1-L2 contrastive requirement of the CAH did not account for the learning of sounds which were not contrastive, nor for how a learner might simplify or alter a sound which was marked. The subsequently proposed Structural

Conformity Hypothesis (Eckman, 1991) thus eliminated the CAH, or considerations of contrast between the L1 and L2, from the prediction that learners would have difficulty with marked features. This hypothesis was in keeping with first language acquisition and affirmed that interlanguages functioned like natural languages, “interlanguage” referring to a learner’s evolving cognitive construct of the second language (Corder, 1967;

Selinker, 1972).

Natural Phonology (NP) also hypothesized that interlanguages function like natural languages and therefore predicted that second language learners would apply generative linguistic rules to their interlanguage (Donegan & Stampe, 1979; Stampe,

1979). Unfortunately, NP did not predict which rule a learner would apply; nor could it account for features of language that were not based on surface structures, such as markedness which is based on distribution across languages (Kerschhofer-Puhalo, 1997).

Optimality Theory (Prince & Smolensky, 1993, 1997, 2004) is much like NP, but accounts for markedness by applying ranked constraints rather than rules.

The Ontogeny Phylogeny Model (OPM) (Major, 2001), a revised version of the

Ontogeny Model (Major, 1987), incorporates cross-linguistic transfer as well as universals throughout a second language learner’s development. The OPM predicts that in the early stages of L2 phonological development, the first language will have the strongest influence. At intermediate developmental levels, universal principles of language are predicted to overtake the influence of the L1; and at advanced developmental stages, second language phonological settings will finally be attained.

In L2 phonological acquisition processes, both cross-linguistic transfer and universals are acknowledged as playing significant roles. Language transfer involves interactions between the L1 and L2 depending on perceived similarity between sounds, while universals of language play an increasing role according to the learner’s developmental level. Therefore, the current study takes into consideration the phonological inventories of the L1 and L2, as well as the learner’s developmental level as measured by his/her L2 oral proficiency and segmental production accuracy.

1.6.2 Research on International Teaching Assistants

Research on International Teaching Assistants (ITA) has focused on improving international graduate students’ L2 English production and perception for teaching purposes. The increasing numbers of international graduate students at American universities in the 1960s and 1970s led to the development of oral English proficiency standards and ITA training programs in the 1980s. Whereas pronunciation was initially

thought to be the sole cause, and therefore sole solution, to the ‘ITA dilemma,’ subsequent research reviewed by Nelson (1990) identified a more complex interaction among pronunciation, discourse, culture (both in terms of educational systems and of the larger home/host communities), and the pragmatics of teaching.

At the same time, theoretical and pedagogical perspectives shifted from behavioral to cognitive psychology, from structural to generative linguistics, and from individual L2 learner skills and products to more social, communicative processes.

Consequently, while pronunciation continues to occupy a (more minor) role in teaching and learning, research in the ITA field has moved away from segmental pronunciation studies to place more emphasis on the discourse level of communication, cross-cultural communication, and the social context of college teaching in the US. The researcher argues, however, that since pronunciation continues to be an important part of L2 communicative competence, especially for professional educators, then neither research nor pedagogy is served by neglecting to contribute ongoing empirical studies to the research base which informs L2 pronunciation, intelligibility, and teaching methodologies.

The Chinese participants of the current study have achieved the threshold level of

English required for teaching purposes at a US university. According to World Englishes

(WE) and English as a Lingua Franca (ELF), these Chinese students are also classified as belonging to the Expanding Circle, and the current study describes their segmental pronunciation. The study also investigates the intelligibility of this accent in English for

Inner (Americans), Outer (Indians), and Expanding Circle (Chinese and Koreans)

listeners within a specific context of usage, that of the functional domain of academic study, teaching, and research. Finally, it examines the effects of talker segmental pronunciation accuracy and the listener’s word familiarity on intelligibility, which has implications for the intelligible use of English within the international academic community.

1.6.3 World Englishes and English as a Lingua Franca

World Englishes (WE) theory proposed categorizing the spread of English in the world by three concentric circles consisting of (1) The Inner Circle, where English is primarily a first language used by monolingual speakers; (2) The Outer Circle, where

English is a secondary official or national language used by bi/multilingual speakers; and

(3) The Expanding Circle, or everywhere else English is used (Kachru, 1985). This categorization allowed WE scholars to make a distinction between English use in the

Outer Circle, where an intranational or local norm of English applies, and the Inner and

Expanding Circles, where L1 norms and SLA theories apply (Kachru & Nelson, 1996).

English as a Lingua Franca (ELF) theory borrowed WE theory regarding the

Outer Circle and extended it to the Expanding Circle (Jenkins, 2000, 2007). ELF theory differs from WE, in that while L2 varieties of English are acceptable as norms, L1

English norms may also be used, but only insofar as they contribute to intelligibility. The

‘Lingua Franca Core’ (LFC) was proposed as a minimum core set of English sounds, or phones, and prosodic features which should follow an L1 English norm to ensure mutual intelligibility for international interlocutors (Jenkins, 2000). Sounds outside the LFC

were defined as not impacting negatively on intelligibility, and therefore free to follow a local English norm.

There still remains much to learn about the intelligibility of English as a Lingua

Franca. For example, to what extent does L1 and segmental variation impact intelligibility, and can one core set of English sounds be intelligible to listeners from all

L1 backgrounds, oral proficiency levels, and vocabulary sizes? Finally, since standards of intelligibility depend on the expectations of specific contexts and communicative purposes, the participants in the current study have been selected in order to begin to explore some of these concerns empirically and in greater depth.

1.7 Methods

This dissertation study investigated the impact of talker L1 and pronunciation accuracy on the intelligibility of Chinese-accented English as judged by Hindi L1,

Mandarin Chinese L1, Korean L1, and American English L1 listeners in the US academic context. The goal of this experiment was twofold: first, to identify the effect of a talker’s

L1 and accurate pronunciation of vowels and consonants on sentence intelligibility for northern Mandarin-accented English; and second, to determine the extent to which this effect varied according to the listener’s L1 and word familiarity.

The study was conducted on the main campus of a large public Midwestern university, where its international student body is among the top fifteen highest enrollments in the country and reflects the same top three nationalities most common on

US campuses today. Chinese-accented English was selected because Chinese-speaking students continue to comprise the highest numbers of international graduate students in

the US (combining students from Hong Kong, mainland China, and Taiwan). The four listener L1s (Hindi, Mandarin Chinese, Korean, and American English) were selected because they constitute the top four L1s of students in US graduate schools (Institute of

International Education, 2009).

The psycholinguistic protocol of word recognition in noise was used to elicit the intelligibility response measure. Listeners were presented with Chinese-accented and

American English-accented sentences which they were asked to transcribe. Their intelligibility scores were calculated based on the key words accurately transcribed in each sentence. Using a mixed-effects model with logistic regression, intelligibility was analyzed to evaluate the effects of talker L1, talker segmental accuracy, listener L1, and listener word familiarity. Random-effects variables were also incorporated into the model and analyzed to assess the extent of inter-individual variation across talkers, listeners, and key words.

1.8 Definition of Key Terms

For the sake of consistency throughout this document, a list of key terms and their definitions for the purposes of this study are included below.

ELF – English as a Lingua Franca. ELF refers to English used for

communication by second language users in the Expanding Circle,

regardless of the country in which English was taught (Jenkins, 2005,

2007).

English for Academic Purposes (EAP) – English for Academic Purposes involves

training students to use English appropriately in higher education. It is

included within the wider field of English Language Teaching (ELT), and

is a branch of English for Specific Purposes (ESP); other branches include

Business English or Healthcare English, among others.

Expanding Circle – Kachru (1985, 1992) proposed a model of three concentric

circles to describe English usage and expansion worldwide, including the

Inner, Outer, and Expanding Circles. The Expanding Circle refers to

English usage in countries where English is neither a first language nor a

secondary official or national language. In this circle, which includes

Japan, Korea, China, and many countries in South and Central America

and Africa, English is used primarily for international communication

purposes.

“Graduate TA certification” – (GTA certification). On the campus under study,

this oral English proficiency level is defined as a score of 230 or higher

out of 300 on the SPEAK test, or an unconditional pass on the Mock

Teaching Test (MTT). Because the Educational Testing Service found

that a score of 27 on the iBT TOEFL speaking section corresponded to a

“clear pass” for ITAs at several large research institutions in the US (Xi,

2007), the researcher also included international participants who scored at

this level or higher.

Inner Circle – Kachru (1985, 1992) proposed a model of three concentric circles

to describe English usage and expansion worldwide, including the Inner,

Outer, and Expanding Circles. The Inner Circle of English usage includes

countries where English is spoken as a first language, e.g., the United

Kingdom, United States, Australia, and New Zealand, primarily for

intranational usage.

Intelligibility – Intelligibility refers to “[the] extent to which a speaker’s message

is actually understood by a listener” (Munro & Derwing, 1995a, p. 76). In

the current study, intelligibility is measured by key word recognition, or

the accuracy of the listener’s written transcription of the key words that

they heard as compared to the written materials that the talkers read.

Interlanguage – Interlanguage is a second language learner’s evolving cognitive

construct of the second language, which is characterized by transferred

features of the first language, second language learning and

communication strategies, and overgeneralization of the patterns of the

second language (Corder, 1967; Selinker, 1972).

L1 – A person’s first or native language.

L2 – A language learned or spoken in addition to one’s L1. An L2 could be the

second, third, or fourth, etc. language learned.

Outer Circle – Kachru (1985, 1992) proposed a model of three concentric circles

to describe English usage and expansion worldwide, including the Inner,

Outer, and Expanding Circles. The Outer Circle of English usage includes

countries such as India, Nigeria, and Hong Kong, where English is spoken

as a secondary official or national language, primarily for intranational

usage.

Phonetics – The area of linguistics concerned with describing the speech sounds

that occur in the world’s languages, the patterns they follow, and the

changes they undergo when preceded or followed by other sounds

(Ladefoged, 1993).

Phonology – The area of linguistics concerned with the description of the systems

and patterns of sounds which convey a difference in meaning in a

language (Ladefoged, 1993).

Prosody – Prosody is frequently confused with the term

“suprasegmental.” However, these terms reflect two different

theoretical positions in phonology. Within the Structuralist

framework, the segment is represented by a bundle of features to which

another level of suprasegmental specifications can also be added.

Within the Autosegmental-Metrical framework, prosody is the “meter,” or

organizational structure of speech, which includes a hierarchy of

constructs realized in the alignment of segmental and tonal properties

at many levels above and below the (prosodic) syllable. Prosody is

language specific and in English includes structures such as the

intonational phrase, the intermediate phrase, and accented syllable

(Beckman, 1996).

Segment – In the Autosegmental-Metrical framework, the segment is

the smallest sequential prosodic constituent, representing the

bundling together of distinctive properties into a single speech

sound, such as the English vowel phonemes /i:/ versus /I/ or the

English consonant phonemes /k/ versus /g/. Work in psycholinguistics

suggests that the human brain relies at least as much on larger word

parts (such as syllables and stress feet) to process speech. Work

in first language acquisition suggests that robust segmentation at the

level of the individual consonant or vowel segment is acquired fairly late.

However, because segmentation at this level is derived from natural

boundaries imposed by the nonlinearities in the mapping from

articulation to acoustics (e.g., Stevens, 1989), this level of

prosodic structure is usefully posited for all spoken languages. It

is also the level of analysis that is encapsulated in alphabetic

writing systems such as the Greek alphabet and Korean hangul, and thus

provides a fairly intuitive "basic theory" for linguists and applied

linguists.

World Englishes – Kachru (1997) favors the term “World Englishes” to refer to

the world’s multiple varieties of English. In this way, he preserves the

sense of pluralism he believes is not reflected in the terms “English as an

International Language” (Jenkins, 2000), “English as a Lingua Franca”

(Jenkins, 2005, 2007), or “world language English.”

1.9 Basic Assumptions of the Study

This study relied on “graduate TA certification” for its oral English proficiency measure, which is based on the SPEAK or the Mock Teaching Test. Although these tests measure many more aspects of a student’s speaking proficiency than just his/her pronunciation, the researcher assumed that when a student achieves GTA certification, his or her pronunciation has also passed a certain threshold of acceptability within the academic context for the purposes of undergraduate teaching.

The researcher assumed that all of the adult graduate student participants possessed sufficient English language skills to complete the language background questionnaire (LBQ) written in English and to perform all tasks (reading, talking, listening, and transcribing in English) associated with the experiment procedures.

Furthermore, she assumed that the participants would fill out the LBQ as truthfully as possible and to the best of their knowledge.

1.10 Limitations of the Study

The findings of this study are limited by the location, talkers, methods, and listeners involved. The findings of this research cannot be generalized beyond the US graduate school under study. The sample of participants was drawn from graduate students at a specific university and therefore the findings apply only to the total population of graduate students at that university. In addition, because the current study was conducted at a large research institution with a student body of more than 50,000, and an international student body of more than 4,500, participant familiarity with foreign-

accented English may be significantly different from familiarity levels at other universities of different sizes and L1 distributions.

The talker factors that limit the findings of this research include L1 and regional dialect, L2 English oral proficiency level, age, educational level, and gender. The results of the study reflect only the intelligibility of northern Mandarin-accented English and

General American English and cannot be generalized beyond these L1s and their regional dialects. The three Chinese talkers were natives of the northern region of mainland China and the three American talkers were natives of Ohio. Therefore, the intelligibility findings are limited to the dialects of these talkers. The current study sought to add data to an under-investigated oral English proficiency level, that of “graduate TA certification,” since most previous research has focused on the level of “graduate school admission.” As a result, this study’s findings are limited to L2 speakers of English at that proficiency level. In addition, all talkers were graduate students, so they were similar in age, they had all completed an undergraduate degree, and had been accepted to graduate school for advanced studies. Therefore, the findings cannot be generalized to talkers with significantly different ages and educational backgrounds. A review of the literature revealed mixed results in terms of the effect of gender on intelligibility and listening comprehension. Therefore, the current study held this factor constant by using only male talkers. Consequently, the findings of the study are limited to male speakers.

The findings of the study are also limited by its methods, including stimuli of read, scripted sentences and a listener task of written transcription. In order to avoid the ceiling effect common with stimuli presented in quiet laboratory settings, the current

study degraded the sentence stimuli with white noise at a +5 dB signal-to-noise ratio, which limited the findings to this ratio. Since carrier sentences were used as stimuli, this increased ‘face’ validity over word intelligibility, but limited the research findings to sentence intelligibility. Passage intelligibility was not selected in order to avoid discourse-level effects and topic familiarity, which were beyond the scope of the study.

In response to the oral stimuli, listeners performed the task of written transcription. Since listeners have been found to perform differently on different tasks, the intelligibility findings of this research are limited to this task.

The setting for this research was in a laboratory, in order to focus on the impact of segmental, i.e., vowel and consonant, pronunciation on intelligibility while excluding other variables found in naturalistic speech production. Outside the laboratory, many other factors – both linguistic and extra-linguistic – can affect communication. The linguistic factors include syntax, morphology, and lexicon; the extra-linguistic factors include gestures, eye contact, facial expressions, or lip movement, and background knowledge of the topic being communicated. However, the multitude of variables at play in a more natural setting make conclusions about the role of a single variable impossible.

Therefore, since the researcher sought to establish a baseline understanding of the impact of the phonological factors of foreign-accented speech on intelligibility, the lab setting was chosen as most appropriate for answering the research questions, and the findings are therefore limited to this setting.

Listener factors held constant in the study included L1, regional dialect, oral

English proficiency level, and educational level. L1s and regional dialects represented by

the listeners who participated in this study were Hindi from northern India, Mandarin from northern mainland China, Korean, and American English from Ohio. To the greatest extent possible on the campus under study, the listeners represented as little dialect variation as possible in their L1s. However, this attempt to hold L1 and regional dialect constant also limited generalizability to those selected. All international listeners were graduate TAs, so they were similar in oral English proficiency level, had all completed an undergraduate degree, and had been accepted to graduate school for advanced studies. Therefore, the findings cannot be generalized to listeners with significantly different oral proficiency levels and educational backgrounds.

1.11 Summary

This chapter introduced the theoretical frameworks of Second Language

Acquisition, World Englishes, and English as a Lingua Franca which inform the current study. Previous research on second language phonological acquisition and international teaching assistants was consulted to design a psycholinguistic experiment to explore the intelligibility of Chinese-accented English to graduate students with the four native languages most commonly found in US universities. The chapter identified the research questions as how talker L1 and segmental pronunciation accuracy affect intelligibility for listeners and to what extent that intelligibility varies by listener L1 and word familiarity.

It determined that the findings therefore revealed which sounds a talker may vary without impeding intelligibility and which sounds should constitute the highest priorities for

English pronunciation teaching for the diversity of listeners in the US academic context.

Chapter 2: Literature Review

2.1 Introduction

Theories of second language (L2) phonological acquisition have been informed by research on speech production and perception. To determine how a second language is learned, this research depends on the judgements and speech of native speakers.

Therefore, in the field of Second Language Acquisition (SLA) in general, the norms of the target or second language and its native speakers are used as reference points and models. In keeping with this theoretical framework, intelligibility studies depend on the judgements and comprehension of native listeners. In terms of teaching applications, intelligibility data have been used to determine the teaching priorities of English pronunciation curricula by establishing error gravity hierarchies. Due to the unprecedented spread of English use worldwide, however, some scholars have recommended reevaluating the basic assumptions of SLA research in order to better reflect the fact that the majority of the world’s users of English today are nonnative, rather than native speakers (Graddol, 1997; Crystal, 2003).

Standards of oral proficiency and intelligibility differ according to the needs and requirements of specific listeners and contexts. Accordingly, significant research has been devoted to determining the English oral proficiency level necessary for specific

purposes, such as for ITAs teaching American undergraduates at US universities.

Outside of the teaching context, however, we know very little about the oral proficiency or intelligibility necessary for conducting research or learning in academic settings and communicating with peers, where the burden of successful communication is shared equally by all interlocutors.

Research on intelligibility and listening comprehension from the perspective of nonnative speakers has been conducted by researchers of World Englishes (WE) and

English as a Lingua Franca (ELF). The focus of WE has been mainly on English in the

Outer Circle, while ELF has focused on the Expanding Circle. Since Expanding Circle speakers use English mainly internationally rather than intranationally, ELF research has attempted to identify the core phonology of English necessary for mutual intelligibility among L2 English speakers with a variety of L1s. The current study sought to fill a significant gap in the research among the interrelated areas of Second Language

Acquisition, English for Academic Purposes, and English as a Lingua Franca.

Intelligible, successful communication is influenced by many factors, among them, the talkers and listeners, their L1s and communicative purposes, contexts, and relative social status. Therefore, this chapter first reviews research on L2 phonological acquisition, which has been largely conducted from the perspective of native English speakers. This section culminates in a review of research on speaking and listening accommodations that native speakers (can) make when communicating with nonnative speakers. The chapter also reviews research on two aspects of the academic context, since these have been found to impact intelligibility. For international graduate students,

the purposes of communication and the contexts in which they occur have been researched from two perspectives: International Teaching Assistants (ITAs) and English for Academic Purposes (EAP). Furthermore, even in US universities, English communication increasingly includes more international interlocutors. Therefore, this chapter concludes with reviews of research in World Englishes and English as a Lingua

Franca.

2.2 L2 Phonological Acquisition Research

Second language (L2) phonological acquisition has been investigated in terms of speech perception and speech production. Interest in L2 speech production began earlier, while L2 speech perception research grew more recently. Early researchers were interested in L2 learner production errors, which they investigated in order to develop theories of L2 speech acquisition in general. This focus on errors led to a greater emphasis on the factors that negatively affect L2 speech production, but by the 1970s, attention shifted to include the factors that also positively affected L2 speech acquisition

(Corder, 1967; Selinker, 1972).

Research on L2 speech perception and production has been conducted for the purposes of developing and refining theories of L2 phonological acquisition, which have significant implications for teaching and learning. The factors that facilitate L2 speech production are as important to understand as those that interfere with it, so that teaching methods and learning strategies can address areas of both strength and weakness for L2 learners. The factors that affect L2 perception and listening comprehension also yield important clues as to what will facilitate and impede listeners’ participation as

interlocutors. Intelligibility studies can shed light on how L1 and L2 English speech is perceived by specific types of listeners for particular contexts and purposes. Finally, educational research investigates the efficacy of teaching methods and materials so that teachers can evaluate their options based on the results of empirical research, rather than being limited to the frame of their own experiences.

The research reviewed below, therefore, first identifies factors which affect L2 speech production, or an L2 learner’s pronunciation. These include linguistic factors, such as L2 speech perception, as well as extra-linguistic factors, such as age of L2 learning, degree of L1 and L2 usage, and total experience with the L2. Second, research which has identified factors that affect listening comprehension is covered, including text and task variables, as well as listener variables. Finally, studies investigating L2 speech intelligibility from the perspective of native English-speaking listeners are reviewed.

2.2.1 Terminology in L2 Research

Speech research has been conducted from a variety of theoretical perspectives and within many disciplines, including linguistics, psychology, and second language acquisition, among others, which have used similar terms but defined them differently.

For instance, although the terms “pronunciation” and “intelligibility” have been used interchangeably, they refer to distinct but related concepts. Furthermore, studies of

“intelligibility” and “listening comprehension” frequently involve similar variables, but they are generally conducted for different purposes.

“Pronunciation” describes the phonetic-acoustic features (segmental and prosodic) of an L2 learner’s speech production. For the purposes of this research,

“segmental” features refer to the vowels and consonants of English. “Prosody” refers to the “meter,” or organizational structure of speech with levels above and below that of the

(prosodic) syllable. Finally, the aspects of an L2 English speaker’s pronunciation which differ from an L1 English speaker’s pronunciation have been referred to as

“accentedness” or “degree of foreign accent,” although in the broadest sense, variation is fundamental to speech production and everyone speaks with an accent of some type.

“Intelligibility” refers to how this pronunciation, or L2 learner’s speech, has been perceived by a listener. If the message intended by the talker has been correctly understood by the listener, it is considered to have been intelligible, regardless of the variation involved in the specific pronunciation of the talker. Although researchers have posited several levels to comprehension, intelligibility is the most basic of these levels.

Smith and Nelson (1985) suggested that there were three levels to comprehension: intelligibility, comprehensibility, and interpretability. The last level, however, has been difficult to operationalize in research and has therefore fallen out of use. Instead, Munro and Derwing’s (1995, 1997) construct of listening comprehension beginning with intelligibility (as the speech signal being received accurately by the listener) and followed by comprehensibility (wherein the meaning of the message has been understood) have been more widely accepted and used.

Finally, intelligibility studies use listener feedback to make recommendations about what should be prioritized in pronunciation teaching. The findings are used primarily to improve L2 production, although implications for L2 listening comprehension or perception may also be included. Listening comprehension studies, in

contrast, are focused on investigating the oral comprehension abilities of L2 listeners.

The findings are generally used to inform models of L2 perception at the theoretical level, and to inform the teaching of listening skills for L2 learners at the pedagogical level.

2.2.2 Theories of L2 Phonological Acquisition

Second language phonological acquisition theories have ranged from assuming that all errors resulted from first language transfer to assuming that none of them did.

When first language transfer fell out of favor, universal principles of language were assumed to affect language learning processes, whether a first or second language were being learned. Further research in second language speech perception and production has revealed interaction between cross-language transfer and universals, depending on the stage of learner development.

2.2.2.1 Cross-Language Transfer

Seminal studies on language transfer, language contact, or Contrastive Analysis

(CA) were conducted within the structuralist paradigm by Fries (1945), Weinrich (1953), and Lado (1957), respectively. With regard to language transfer in second language acquisition, Fries (1945) argued that “[t]he most efficient materials are those that are based upon a scientific description of the language to be learned, carefully compared with a parallel description of the native language of the learner” (p. 9).

In the related field of bilingualism, Weinrich (1953) studied the impact of two languages in contact, German and Swiss Romansh. He defined “contact” as occurring within the individual and found that cross-linguistic influence varied depending on the

social context in which the individual was communicating. Although Weinrich acknowledged that some sounds in the L2 exist in the L1 and therefore facilitate L2 acquisition (positive transfer), he was most concerned with interference (negative transfer) of the L1.

Lado (1957) is regarded as the founder of Contrastive Analysis, in that he developed specific techniques for comparison with which “we can predict and describe the patterns that will cause difficulty in learning, and those that will not cause difficulty, by comparing systematically the language and culture to be learned with the native language and culture of the student” (p. vii). Lado’s claim – that CA is a predictor of all learning difficulties in second language acquisition – was too strong, however, leading to the theory’s fall from grace beginning in the 1970s. Although Odlin (1989, 2003) noted that language transfer was less controversial for the linguistic subsystems of phonology and phonetics, CA lost credibility on the grounds that it failed to predict many student errors, most notably in morphology and syntax.

In response, a weak version of CA was postulated, since Lado’s strong version had been discredited (Wardhaugh, 1970). Wardhaugh’s (1970) weak version of CA explained student errors after the fact, while continuing to assume that these errors were all due to transfer. In turn, since the weak version of CA could not predict which errors were more likely to be transferred than others, Oller and Ziahosseiny (1970) developed a moderate version of CA, which incorporated degrees of similarity between the L1 and

L2. Although the main tenet of CA theory – that all errors are due to transfer – was

discredited, later studies in cross-linguistic research have come to acknowledge transfer to have substantial influence, even if it is not the only influence, on L2 learner errors.

More recent second language phonological research has included studies on both speech perception and speech production. While several studies have found training in one aspect to facilitate learning in the other (Rochet, 1995; Bradlow, Pisoni, Akahane-

Yamada, & Tokhura, 1996; Matthews, 1997; Wang, Jongman, & Sereno, 2003), this is not to suggest that perception and production are simply mirror images. Broselow and

Park (1995) posited a Split Parameter Setting Hypothesis to explain research data from second language learners that indicated perception and production functioning at different levels. They hypothesized that perception and production would both initially be governed by the L1; but, as learners moved through second language developmental stages, perception and production would split. The L2 would then begin to govern perception, while the L1 would continue to govern production, until L2 settings were finally attained for both.

A more nuanced model based on inter- and intra-dialectical differences suggests greater variation in production, while perception remains aligned with L1 norms. Davis and Beckman’s (1983) study of the production and identification of initial stops in

American and Indian English indicated that after significant experience with American

English, Indian speakers’ productions had begun to shift toward the VOT settings of the new dialect, but that identification of initial stops remained within the perception categories of their L1 dialect.

Studies focused on speech perception investigated whether L2 production errors were due to difficulty with phonetic realizations or whether they originated in the failure to perceive a new phonological category. Perceptual Magnet Theory, which Kuhl (1991) first posited as a perceptual magnet effect, and later developed into the Native Language

Magnet theory (Kuhl, 1993), suggested that the native language’s phonological categories act like magnets pulling L2 sounds toward them, thereby inhibiting the learner’s ability to perceive the new sounds.

In fact, McAllister (1997) suggested a ‘perceptual foreign accent’ in which the first language conditioned the sounds the learner was listening for, which potentially inhibited the sounds the listener was able to perceive. However, the Perceptual Magnet

Theory could not account for two types of data: one, the acquisition of L2 sounds which were not in the L1 inventory and two, the attrition or loss of the first language.

Lively and Pisoni (1997) replicated Kuhl’s (1991) experiment with adults. They found that while Kuhl’s findings with infants were still robust, her findings for the effect of a perceptual magnet on adult phonological categorization could not be replicated and was therefore inconclusive. While Lively and Pisoni’s (1997) subjects also provided a range of “goodness” ratings for each instance of the vowel /i/, there was no correlation between the instances that received the highest “goodness” ratings and low discrimination accuracy, which the perceptual magnet effect would predict.

In 1994, Best proposed her Perceptual Assimilation Model (PAM) and in 1995,

Flege posited the Speech Learning Model (SLM), both of which hypothesized that phonological space is shared among all languages learned. Therefore, if a new sound is

perceptually distant from another sound already learned, then it can be learned more easily than a sound which seems perceptually very close to another sound. For example,

American English speakers have neither the trilled /r/ of Spanish nor the tapped /r/ of

Japanese in their L1 phoneme categories, but can produce them at native-like levels when learning Spanish and Japanese. In contrast, L2 sounds which are perceptually close to L1 sounds, such as the well-researched ‘allophonic split’ (Eckman, Elreyes, & Iverson, 2001,

2003) of /r/ and /l/ in English for Japanese learners, are the most difficult to perceive, and thus, learners may not create a new phonological category for them.

Research on bilinguals indicated that language transfer was not unidirectional and that the L2 could in fact impact the L1. Even in cases where speech perception studies showed that bilinguals could distinguish between two sounds in their L1 and L2, which evidenced the creation of two distinct phonological categories, their spoken productions differed from those of monolinguals in order to maintain sufficient contrast among all the sounds they knew. In addition, some bilinguals exhibited a loss of their first language as their L2 usage eclipsed that of their L1 and became the more dominant language (Flege et al., 2002).

In order to account for these data from bilinguals through a lifespan of language learning, Best’s (1994) Perceptual Assimilation Model (PAM) hypothesized that perceptual distance between L1 and L2 sounds could be used to predict the extent to which a listener would successfully discriminate new sounds. According to PAM (Best,

1995, p. 193), a new sound could be categorized by a listener in three ways: (1) as an exemplar (good, acceptable, or notably deviant) of an existing L1 “gestural

constellation”; (2) as uncategorizable within existing L1 phonological categories and thereby assimilated to a newly created L2 category; or (3) as a non-speech sound. In this model, therefore, L1 categories were not as fixed as was assumed with the Perceptual

Magnet Theory (Best et al., 2001), but rather, they could be adapted in response to new language experience, or even lost altogether.

Flege’s research on L2 speech perception and production led to his development of the Speech Learning Model (SLM) in 1995, which shared Best’s (1994) hypothesis of a common phonological space for all known speech sounds. Flege’s (1995) SLM included seven hypotheses as elaborated below.

H1 Sounds in the L1 and L2 are related perceptually to one another at a

position-sensitive allophonic level, rather than at a more abstract

phonemic level.

H2 A new phonetic category can be established for an L2 sound that differs

phonetically from the closest L1 sound if bilinguals discern at least some

of the phonetic differences between the L1 and L2 sounds.

H3 The greater the perceived phonetic dissimilarity between an L2 sound and

the closest L1 sound, the more likely it is that phonetic differences

between the sounds will be discerned.

H4 The likelihood of phonetic differences between L1 and L2 sounds, and

between L2 sounds that are noncontrastive in the L1, being discerned

decreases as AOL increases.

H5 Category formation for an L2 sound may be blocked by the mechanism of

equivalence classification. When this happens, a single phonetic category

will be used to process perceptually linked L1 and L2 sounds (diaphones).

Eventually, the diaphones will resemble one another in production.

H6 The phonetic category established for L2 sounds by a bilingual may differ

from a monolingual’s if: 1) the bilingual’s category is “deflected” away

from an L1 category to maintain phonetic contrast between categories in a

common L1-L2 phonological space; or 2) the bilingual’s representation is

based on different features, or feature weights, than a monolingual’s.

H7 The production of a sound eventually corresponds to the properties

represented in its phonetic category representation. (p. 239)

Much of Flege’s research focused on the effect of age-related factors on second language speech production, such as the age of first instruction, the age of arrival, or length of residence, in response to the Critical Period Hypothesis (Scovel, 1969; Long,

1990). This notion of a critical age for language acquisition was first proposed in the area of first language acquisition research (Lenneberg, 1967). As a result of early studies with feral children without access to first language development during the critical period who could never reach normal adult proficiency levels (e.g., Victor and Genie), it was suggested that if a language had not been learned during this period, it could never be recovered.

Scovel (1969) and Long (1990) extended the Critical Period Hypothesis (CPH) to second language acquisition, justifying it with research in neuroscience, which suggested

that brain lateralization during puberty led to decreased neuronal plasticity. Flege (1988,

1993; Flege et al., 1997, 2001) and colleagues conducted several studies investigating the effect of age-related factors on second language phonological acquisition and found that in general, the earlier one learned a language the better. However, in contrast to what the

CPH would predict, Flege found that some adults could still perceive and produce new sounds with native-like accuracy and some children failed to do so.

In terms of their contributions to phonological theory, Best (1994) responded to the Perceptual Magnet Theory by noting that L1 phonological categories were not fixed and that language transfer was not unidirectional. Flege’s (1995) SLM responded to the

Contrastive Analysis and Critical Period hypotheses by accounting for L2 learners’ abilities to perceive and produce new sounds in the L2, at a variety of ages beyond that defined as critical. In general, early bilinguals perceived and produced sounds more like monolinguals than late bilinguals. However, late bilinguals still demonstrated changes in their L1-L2 phonetic categories as a result of increased language experience, and some early bilinguals failed to achieve native-like L2 production. If at least some of the phonetic differences between an L2 sound and an existing L1 sound could be perceived, then a learner could overcome ‘equivalence classification’ and create a new phonetic category. Furthermore, Flege’s SLM explained that bilinguals’ phonetic categories might differ from those of monolinguals for two reasons. One, because bilinguals needed to maintain contrast among many more phonetic categories within a single L1-L2 phonological space than monolinguals; and two, because bilinguals might create new categories based on different phonetic features, or feature weights, than monolinguals.

Although researchers agree that the L1-L2 perceptual ‘similarity’ of a sound significantly affects a learner’s ability to perceive and produce it, measures of similarity remain difficult to define. According to Strange and Shafer (2008, p. 172), four different techniques to establish similarity have been used in L2 perception research: (1) qualitative descriptions of articulatory-phonetic similarities; (2) qualitative perceptual comparisons, involving (narrow) L2 transcriptions; (3) acoustic comparisons of L1 and

L2 phones, including discriminant analysis; and (4) direct measures of perceived similarity, including listener categorization of L2 phones according to L1 phonetic categories. Since L2 perception involves both phonetic and more abstract, phonological processing, Strange and Shafer (2008) recommend that research findings be culled from a variety of paradigms which balance the need for experimental rigor with ecological validity.

2.2.2.2 Universals, Transfer, and their Interaction

In addition to cross-language transfer, universal principles of language were found to significantly influence second language acquisition and thus student errors.

Universals of language affect the sequence in which language features are acquired, independent of the specific L1 or L2 of the learner. In syntax structure acquisition, for example, Martohardjono and Flynn (1995) found similar sequences of acquisition for L2 learners with several different L1s, which L1 transfer could not explain. In phonology, voiced coda obstruents are marked, or less common across languages, and acquired later by both first- and second-language learners (Eckman & Iverson, 1994). These examples

of Universal Grammar and markedness suggest that some principles of language acquisition are universal, rather than language specific.

In Sound Pattern of English, Chomsky and Halle (1968) developed the theory of generative grammar, including the concept of Universal Grammar (UG). Chomsky’s

(1993) later Pisa lectures further elaborated the theory of UG to consist of principles and parameters. “Principles” refer to the core grammar found in all languages, while

“parameters” designate the specific settings for these universal elements of grammar. In his review of literature, Major (2001) identified several other universals besides UG which have been the subject of study, including learnability theory (Baker, 1979), markedness (Greenberg, 1966, 1978), underlying representations (Chomsky & Halle,

1968), rules and processes (Chomsky & Halle, 1968; Donegan & Stampe, 1979; Stampe,

1969, 1979), constraints (Prince & Smolensky, 1997, 2004), and stylistic variation

(Bayley & Preston, 1996; Tarone, 1988).

In response to Chomsky and Halle’s (1968) generative view of language structure, the field of linguistics shifted away from structuralism. Using a generative paradigm,

Michaels (1974) proposed markedness values for linguistic features. However, this paradigm was of more limited use for phonological applications. While generative theories of syntax are still robust, in terms of phonology, the focus on surface patterns of language was limiting. As Broselow, Chen, and Wang (1998) noted, the generative model cannot account for rules formulations that are not based on surface patterning, such as those based on typological markedness.

The principle of markedness, as conceived by Trubetskoy (1939) and Jakobson

(1941), was that linguistic representations in binary opposition (e.g., voiced and voiceless obstruents) were not equally distributed within a language or across languages. The more widely distributed ones were considered ‘unmarked,’ and those with more limited distribution were considered ‘marked.’ For example, voiceless obstruents are much more common in the world’s languages and, therefore, are considered ‘unmarked,’ while the rarer voiced obstruents are considered ‘marked.’

In 1977, Eckman proposed the Markedness Differential Hypothesis (MDH), which predicted increasing levels of difficulty for learning second language structures that were different from L1 structures as a function of their degree of markedness.

According to Eckman (2008), three issues arose which led him to propose a revised version of the MDH, which he termed the Structural Conformity Hypothesis (1991). The first issue was that ‘level of difficulty’ proved problematic to define and measure.

Second, the MDH did not predict SLA data which correlated with markedness but not with L1-L2 areas of difference. Third, while the MDH predicted difficulties with marked

L2 structures, it did not predict how a learner would choose to simplify or alter the structure. Therefore, the Structural Conformity Hypothesis stated that “The universal generalizations that hold for primary languages hold also for interlanguages,” which eliminated the L1-L2 contrastive requirement from the application of degree of markedness considerations (Eckman, 1991, p. 24).

A theory of Natural Phonology (NP) (Stampe, 1979) was proposed to address the universal constraints of the phonological system, or “the residue of a universal set of

processes reflecting all the language-innocent phonetic limitations of the infant”

(Donegan & Stampe, 1979, p. 127), upon which Dressler (1984) later expounded. Major

(1987a) observed that the advantage of NP was its accounting for diachronic, synchronic, and child language phenomena, as well as second language acquisition.

Subsequent research identified two areas of weakness of Natural Phonology. The first was that although NP could predict that a learner would apply a process, it could not predict which process (Kerschhofer-Puhalo, 1997). Second, Zborowska (1997) noted that natural processes were often absent in small children, or appeared only in an irregular way. As a result, a “weak” version of NP became more widely accepted, which was based on a model of self-organization that assumed an “interplay between genetic programming and selection and evaluation of postnatal information according to preferences for parallelism, frequency, and regularity of data” (Leather, 1999, p. 14).

Optimality Theory (OT) is similar to Natural Phonology, but replaces the classical generative phonology notion of a sequential derivation of ‘rules’ with ‘rankings of constraints’ (Prince & Smolensky, 1993, 1997, 2004). In OT, phonological systems result from rankings of innate and universal constraints. Therefore, the process of second language acquisition changes from the process of applying a series of rules (from L1 rule, to developmental rule, to L2 rule) to the process of applying a series of constraint rankings (from L1 rankings, to non L1 or non L2 rankings, to L2 rankings) (Major,

2008). Studies on second language English acquisition applying OT include Broselow,

Chen, and Wang’s (1998) study of coda obstruents, Hancin-Bhatt’s (2000) review of OT

in L2 phonology, Kim’s (2001) study on Korean English, and Bunta and Major’s (2004) study of English front vowels.

To address the methodological issue of measuring ‘level of difficulty,’ Major and

Kim (1996) recommended conceptualizing the ‘similarity’ and ‘dissimilarity’ of L2 sounds, as compared to L1 sounds, in terms of ‘rate of learning.’ They proposed the

Similarity Differential Rate Hypothesis (SDRH), which predicts faster rates of acquisition for L2 sounds perceived as phonetically ‘dissimilar’ from L1 sounds than for those perceived as ‘similar.’ In their data on Korean learners of English, this prediction was supported by the faster rate of acquisition for /z/, a dissimilar sound, than for /dʒ/, a

similar sound.

In his Ontogeny Phylogeny Model (OPM) (Major, 2001), revised from the earlier

Ontogeny Model (Major, 1987), Major incorporated the interaction between transfer and

universals in second language phonological acquisition. In the OPM, he hypothesized

that “Over time and as style becomes increasingly formal, L2 increases, L1 decreases,

and U [= the universal set of language properties] increases and then decreases. In

addition, the relative proportions of U and L1 depend on whether phenomena are normal

[= dissimilar and unmarked], similar, or marked” (Major, 2001, p. 157). Therefore,

research on second language phonology must be especially careful to take developmental

levels and speech styles into account, and the universal processes of language they

reflect, in addition to the phonological categories of the L1 and L2, their perceptual

similarity, and typological markedness.

The current study benefits from L2 phonological acquisition theories and research by holding developmental level and speech style constant to investigate L2 speech intelligibility. Developmental level is measured by oral English proficiency level and speech style is held to the formal task of reading scripted sentences. The researcher aimed to investigate speech intelligibility at a specific oral English proficiency level which will in turn inform models of L2 phonological acquisition by providing more detailed information about perception and production at this developmental level.

Further analyses on the Buckeye GTA Corpus will also shed more light on the respective contributions of the phonological inventories the L1, L2, and typological markedness at this oral proficiency level.

2.3 Factors Affecting Listening Comprehension and Intelligibility

While L2 phonological acquisition research focuses on learners as individuals and the factors that can affect their learning, intelligibility research incorporates the feedback of listeners, which then informs second language pedagogy. Although the research goals of these two SLA areas are different, many of the same factors are considered. The personal attributes of both talkers and listeners, such as age and experience with the second language, have been found to affect listening comprehension and intelligibility.

Also important are the phonetic-acoustic characteristics of the speech signal, the oral text or written materials, the task which elicited the talker’s speech sample, as well as the task of the listeners to register their comprehension. The following sections review the effects of these factors in previous research and provide a rationale for the choices made for

addressing each of these factors in the current study that quantifies these variables and determines to what extent they correlate with intelligibility.

2.3.1 Talker Factors that Affect L2 Speech Production

Flege’s (1988, 1993) research focused on perceptions of and factors that affect foreign accent. As researchers interested in the acquisition of second language phonetics,

Flege and colleagues (1995, 1997) investigated the degree to which native listeners perceived a foreign accent in English language learners’ speech. These researchers were interested in both the perception and production of L2 English and the factors that affect these, especially age of learning.

Flege examined Chinese learners of English to determine the effects of various factors on L2 perception and production. These factors included the following: age at which the learner begins L2 acquisition, L2 usage experience, and L1 usage.

Furthermore, he investigated whether nonnatives with a foreign accent can accurately evaluate another’s foreign accent, and whether removing pauses makes sentences sound less accented. His findings indicated that perception accuracy is affected by more experience with English, while production is not. In terms of judging accentedness, the highly experienced L2 English users performed similarly to native speakers. Since the highly experienced Chinese learners of English still had strong accents, it appeared that while exposure alone improved perception, pronunciation would not improve without formal instruction.

Conducting four experiments, Flege (1993) examined the effect of the age at which the Chinese learner of English begins L2 acquisition, specifically for the

perception and production of /t/ and /d/ in word-final position. He found that childhood learners of English closely resembled native speakers in all experiments, whereas adult learners of English differed from native speakers in both perception and production. This suggested that production limitations were related to a limitation in the perceptual representations of sounds in the L2.

In a study they conducted among Italian immigrants to Canada, Flege, Munro, and Mackay (1995) discovered a lot more variation in the effect of age on second language learning than had been previously noted by Flege (1988, 1993). They found that there was no identifiable threshold age for acquiring a native-like accent, but that age did account for 59% of the variance in the foreign accent ratings. In addition, language use and gender also accounted for a percentage of the variance in foreign accent ratings.

Flege, Bohn, and Jang (1997) examined the effect of experience on nonnative speakers’ (German, Spanish, Mandarin, and Korean) perception and production of

English vowels. They found that not only did the more experienced subjects perceive and produce English vowels better, but that both perception and production varied as a function of the phonological inventory of the L1.

Piske, MacKay, and Flege (2001) reviewed research on the factors found to affect foreign accent in a second language. These factors included the age of L2 learning

(AOL) or the age of arrival (AOA) in an L2-speaking country, length of residence (LOR) in an L2-speaking country, gender, formal L2 instruction, motivation, language learning aptitude, and first- and second-language use. In an experiment they conducted with

Italian-English bilinguals in Canada, Piske, MacKay, and Flege (2001) compared the relative effect of several of these factors on the degree of foreign accent in English.

In their experiment, Piske, MacKay, and Flege (2001) chose not to include three variables found in their review of literature: motivation, formal L2 instruction, and language learning aptitude. Motivation was not included both because the literature showed that it had little general effect and also because the subjects in their study had resided in Canada for an average of more than thirty-five years. Formal L2 instruction was not included in their study because in another study by Flege, MacKay, and Piske

(2002), this variable correlated with AOL, but was not an independent predictor of foreign accent in the L2. Language learning aptitude was not included in their study for two reasons. First, it was found to have little or no effect on foreign accent, and second, because it was difficult to interpret the defining nature of “mimicry ability,” which was a component within the language learning aptitude construct.

Piske, MacKay, and Flege’s (2001) findings indicated a strong correlation between L1 use and high degree of accent in L2, for both early and late bilinguals.

However, AOL was found to have a greater effect on foreign accent than L1 use, while gender did not show any significant effect or interaction with AOL. A surprising finding of Piske, MacKay, and Flege’s (2001) research was that although L1 usage was found to be significant, self-reported ability in the L1 was not. With regard to the effect of LOR,

Piske, MacKay, and Flege’s (2001) findings supported the view that “after L2 learners have spent a certain amount of time in a predominantly L2-speaking environment, LOR will cease to have a further ameliorative effect on L2 pronunciation” (p. 210).

2.3.2 Listener Factors that Affect L2 Speech Perception

In her review of research on listening comprehension and the factors that affect it,

Rubin (1994) identified five major factors to be considered: (1) text characteristics, (2) interlocutor characteristics, (3) task characteristics, (4) listener characteristics, and (5) process characteristics. Text characteristics included the variation of the listening text type and associated visual support. Interlocutor characteristics were those which related to variation in the speakers’ personal characteristics. Task characteristics were defined as differences in the purposes for listening and the responses elicited. Listener characteristics pertained to the variation in the listeners’ personal characteristics (e.g., gender and proficiency level). Process characteristics referred to the variation in the listener’s cognitive processes and in the nature of the interaction between talker and listener (e.g., input processing, learner strategies, and meaning negotiation).

The listener characteristics considered in the research reviewed by Rubin (1994) included age (Seright, 1985; Halladay, 1970), gender (Boyle, 1987) and gender bias

(Markham, 1988), and oral language proficiency level (Thompson, 1995; Dunkel,

Henning, & Chaudron, 1993). More recently, Kachi (2004) also found that L2 listeners’ oral proficiency levels had a significant effect on their listening comprehension.

Several studies have also investigated the effect of familiarity on intelligibility, although familiarity has proved difficult to operationalize, leading to several different definitions methodologically. One body of research has focused on familiarity with an accent. For example, it has been hypothesized that if a talker and listener share an L1, then this familiarity with their L1-accented English will translate into greater

intelligibility. Flowerdew (1994) reviewed several studies on listening comprehension, including Anderson-Hsieh and Koehler (1988), Bilbow (1989), Brown (1968), Ekong

(1982), and Richards (1983). In terms of the effect of accent familiarity on comprehension, the studies revealed not surprisingly, that unfamiliar accents can cause problems. These findings were confirmed by Eisenstein and Berkowitz (1981), who found that ESL learners in the US understood the General American English accent more easily than either a regional L1 English accent (New York) or foreign-accented L2

English.

Support for the hypothesis that listening comprehension is facilitated when both listener and speaker share the same accent (and L1) can be found in Wilcox (1978), in which Singapore listeners better understood their own accent, and in Brown (1968), in which West Africans better understood their own accent. However, other studies offered only partial support for this hypothesis. In Smith and Bisazza (1982), Japanese listeners better understood their accent in English, but subcontinental Indians understood the

American accent better than they did their own. This finding could be due to the fact that subcontinental India includes a wide range of L1 dialects that transfer to English in different ways, leading to a highly heterogeneous Indian English.

Spanish-, Chinese-, Japanese-, and American-accented English were studied and compared by Major, Fitzmaurice, Bunta, and Balasubramanian (2002), who found that while Spanish speakers better understood their own accent in English, Chinese listeners comprehended their own accent most poorly. The Chinese and Japanese listeners better comprehended lectures read by two Spanish speakers than lectures read by speakers with

their own accents. Major et al. (2002) speculated that shared prosody patterns (i.e., syllable-timed) improved comprehension, but this theory did not explain why their own accents, which were also syllable-timed, were less intelligible to them. A more plausible explanation might be that the Spanish speakers were more proficient speakers of English, since they were described by the researchers and raters as having “the least accent.”

This study was limited by the fact that the listeners heard different lectures and answered different comprehension questions and therefore, variation in results may have been due to topic familiarity or difficulty as well as question difficulty, rather than the talkers’ accents. In addition, the researchers did not test their subjects for oral English proficiency level, used only two talkers to represent each nonnative accent, and did not identify which “Chinese” languages or dialects were represented by talkers and listeners, so the L1 for this talker/listener group may not have been held constant. Finally, since the talkers’ speech was not transcribed or analyzed acoustically, the phonological features that comprised each accent were not inventoried in any detail. Major et al. (2002) concluded that while familiarity does not necessarily improve comprehension, it can be an aid.

Most of the research on the impact of accent familiarity on listening comprehension has been general in nature. Interest has been focused on whether one accent is more or less intelligible to different types of listeners, but few studies have examined which phonological features may have led to differences in intelligibility.

Pihko (1997) was one of the first researchers to test for not only the intelligibility of various accented Englishes, but also for the phonological factors which led to differences

in intelligibility scores. Pihko (1997) investigated the intelligibility of familiar L1

English, such as British, as compared to that of unfamiliar L1 English, such as American, and familiar L2 English, including Finnish- and German- accented English, as well as unfamiliar L2 English, such as Gambian-accented English, as determined by Finnish and

British listeners. Her intelligibility scoring for the partial dictation test had three levels, rather than two, and she did not statistically compare the two listening groups in terms of the populations that successfully transcribed the listening ‘gap statements.’

While familiarity has been investigated by many researchers for its impact on listening comprehension and intelligibility, one of the methodological problems has been that the variable has been variously defined as (1) familiarity with a specific accent

(Tauroza & Luk, 1997), (2) familiarity with the test speaker (Gass & Varonis, 1984), (3) familiarity with accents in general (Smith & Nelson, 1985; Smith, 1992), and (4) familiarity with the topic (Gass & Varonis, 1984; Schmidt-Rinehart, 1994).

2.3.3 Text Factors: Word Frequency or Word Familiarity

Researchers have identified the following text factors as significant for intelligibility and listening comprehension: text type (word, sentence, or passage), text length (for sentence and passage intelligibility), and word frequency or word familiarity.

Because the current study is focused on the academic context but sought to identify segmental production effects on intelligibility, sentence-level intelligibility was selected over word and passage intelligibility as being the most valid, yet reliable text for answering the research questions. Since the sentences in the Bamford-Kowal-Bench-

Revised Standard Sentence Lists were determined to not overtax the capacity of the

working memory of even very young listeners, this text factor was not reviewed.

Therefore, this review focuses on research examining the effects of listener word familiarity.

Researchers have found that words that are more frequently occurring in a language are also more intelligible. Nusbaum et al. (1984) proposed that since a listener’s personal vocabulary interacted with word frequency, a better measure would be word familiarity. Using the nearly 20,000 words contained in Webster’s Pocket

Dictionary, the researchers asked native English speakers to rate their familiarity with each word on a scale of increasing familiarity from 1 to 7, where 1 = “I don’t know this word” and 7 = “I know this word very well and could use it in a sentence.” These ratings are part of the Hoosier Mental Lexicon (HML) and measure a person’s subjective familiarity with a word. It should be noted that in SLA research, there is also objective familiarity rating, which is an objective count of the frequency of occurrence of each word in the text; in psycholinguistics research, this is known as frequency.

Bradlow and Pisoni (1999) examined the interaction between talker-, listener-, and item-related factors that influence speech perception. For their intelligibility test, they ensured that listeners would be familiar with the list of words on the test by conducting a 7-point Likert scale familiarity test with a group of students who were similar to the subjects in their study. Words which received a familiarity rating of 6.25 or higher on a scale of 1 (lowest) to 7 (highest) were then included in their listening test.

They found that word familiarity had a greater impact on listening comprehension for nonnative listeners than for native listeners.

Flege (personal communication, July 21, 2008) also recommended a 7-point

Likert scale for subjective word familiarity ratings, but noted that subjects had different definitions of what it meant “to know” a word. Therefore, he proposed using a two-step rating process in which the first step asked listeners whether they had ever seen or heard the word before, and if they had, they were then asked to rate their familiarity with the word on a scale of 1 to 7, where 1 = “I have very seldom seen or heard this word” and 7 =

I have very often seen or heard this word.” The current study adopts these end labels and this two-step subjective word familiarity rating process, but on a scale of 1 to 5, rather than 1 to 7, since the purpose of the rating is simply to confirm that all of the key words included in the listening text are highly familiar to the listeners, rather than to conduct a word familiarity rating test of several thousand words.

2.3.4 Task Factors

Different types of tasks have been found to affect the types of speech that talkers produce, as well as what listeners understand. Therefore, talker tasks ranging from formal to informal vary in terms of the attention required to produce both form and meaning. Listeners can also be asked to perform many different types of tasks depending on the researcher’s questions and objectives. Researchers should therefore be careful to choose a task that will help answer their questions, as well as account for the listeners’ potentially different abilities to perform the tasks they choose.

2.3.4.1 Talker Tasks

Research on L2 phonological acquisition has been concerned with the effect of different talker tasks on the speech production of second language learners. In this type

of research, therefore, the greater the variety of tasks that talkers perform, the better a representation the researchers can get of a given talker’s abilities in his/her second language. For example, Tarone (1979; 1983; 1988) found that the more formal a task, the more precise speakers attempted to make their speech. In contrast, she found that on more informal tasks, perhaps due to less pressure to perform, speakers were more accurate in their pronunciation. Elliot (1997), on the other hand, found that more formal tasks allowed his speakers to put more attention into their pronunciation and thus exercise greater control and demonstrate greater accuracy than on less formal, conversational tasks, which demanded more attention to meaning.

For the purposes of intelligibility, however, the use of unscripted speech introduces potentially confounding variables in the form of syntax, morphology, and lexicon. Scripted speech of various lengths, such as word, sentence, and passage, ensures that variation will be contained within the areas of phonology and phonetics. At the same time, formal and scripted speech is different from informal and spontaneous speech, but most notably in prosody. The current study therefore employs scripted carrier sentences, which are recorded as talkers read them.

2.3.4.2 Listener Tasks

Listeners have been asked to perform a variety of tasks in order to determine how intelligible a talker’s utterance was to them. Early research asked listeners to translate what they had heard into their L1. However, it was unclear whether this task measured the listeners’ intelligibility or simply their translation ability. Alternatively, some researchers have used written transcriptions to measure intelligibility, and finally, they

have also used oral repetitions of the talkers’ utterances. When comprehensibility is measured at the same time, then the task chosen for this measure may determine which task is chosen for the intelligibility measure. For example, if written transcriptions are used as the intelligibility measure, then subjective ratings of global comprehensibility may be combined (e.g., Munro & Derwing, 1995a). Alternatively, since global comprehensibility ratings have been found to correlate with processing or response times

(RTs), these may be used as the comprehensibility measure (Munro & Derwing, 1995b;

Weil, 2003).

2.3.5 Phonetic-Acoustic Factors

The phonetic-acoustic factors of the speech signal include the segments (e.g., in

English, the vowels and consonants), prosody (e.g., pitch accent and intonational phrase), and the speech rate (articulation rate and pause phenomena). In studies on the intelligibility of foreign-accented speech, variation in these factors has been measured by acoustic analyses, by subjective listener ratings of degree of accentedness, or by researcher phonetic transcriptions or other expert ratings. Researchers have used degree of accentedness in combination with intelligibility or comprehensibility ratings in order to establish gravity hierarchies of L2 speaker errors, which are important for pronunciation teaching applications. The section below reviews this research, as well as research that has attempted to determine the relative effects of segments and prosody on intelligibility and comprehensibility. Research on speech rate, which includes articulation rate and pause phenomena, concludes this section on the effect of phonetic- acoustic factors on intelligibility and comprehensibility.

2.3.5.1 Segmental and Prosodic Production

From the perspectives of contrastive analysis (CA) (Lado, 1957) and interlanguage studies (Selinker, 1972), researchers in the 1960s through the 1980s analyzed L2 learner errors in order to classify them and establish hierarchies according to the impact each type of error had on second language speech intelligibility. In 1967,

Corder’s article “The Significance of Learners’ Errors” marked the beginning of interlanguage studies, although it was not until 1972 that Selinker coined the term

“interlanguage” (IL). Selinker (1972) defined IL as the language learner’s working construct of the second language which evolved during the process of acquisition. When

Selinker (1992) revisited IL theory within the frame of SLA, he suggested that “we need an understanding of how to decide what is and is not IL fact and where IL variation, language transfer, fossilization and universal linguistic processes fit in” (p. 2).

One of these early CA error classification studies was conducted by Moulton

(1962), in which he offered four classifications of pronunciation errors for American learners of German: (1) phonemic, (2) phonetic, (3), allophonic, and (4) distributional, as well as specific drills to address each type of error. Later CA studies considered the relative effects on L2 intelligibility of different error types, such as vocabulary, grammar, pronunciation, and discourse. For American learners of German, Politzer (1978) ranked the error types that affected intelligibility most as follows: (1) vocabulary; (2) grammar; and (3) pronunciation. For the intelligibility of American learners of French, Ensz (1982) found grammar errors to be more significant than pronunciation errors. For American

learners of Spanish, Gynan (1985) found pronunciation to have a greater impact on intelligibility than grammar errors.

For Puerto Rican learners of English, Fayer and Krasinski (1987) found that pronunciation errors and hesitations resulted in both native (English L1) and nonnative

(Spanish L1) listener distraction and annoyance, but that the nonnative listeners were less tolerant overall than the native listeners. For Danish learners of English, Albrechtsen,

Henriksen, and Faerch (1980) found that accurate pronunciation and intonation correlated only slightly with intelligibility and that discourse-level errors had the most significant impact. Therefore, they concluded that it was not error type, but error frequency that impacted intelligibility the most.

In a related body of literature, researchers investigated whether segmental or prosodic pronunciation had a greater impact on intelligibility. Several researchers found that prosody was more important to intelligibility than segments (Anderson-Hsieh,

Johnson, & Koehler, 1992; Johansson, 1978; Palmer, 1973). Others came to the opposite conclusion, that segments were more important than prosody to intelligibility (Fayer &

Krasinski, 1987; Koster & Koet, 1993).

Within the research on segmental pronunciation, researchers have examined whether pronunciation of vowels or consonants was more important to intelligibility

(Albrechtsen, Henriksen, & Faerch, 1980; Gimson, 1970; Johansson, 1978; Schairer,

1992). For learners of English, Gimson (1970) found that accurate pronunciation of consonants was more important than vowels. For English-speaking learners of Spanish,

Schairer (1992) found that accurate pronunciation of vowels was more important than

consonants. Due to differences in first language, second language, and research methodology, the findings on the impact of specific pronunciation features on intelligibility have been mixed and inconclusive.

Munro and Derwing collaborated on several studies (1995a, 1995b, 1997). They focused on the intelligibility and comprehensibility of L2 speech, and to what extent the degree of foreign accent affected them. The variable, “perceived comprehensibility,” was measured as a global judgment rating, and the variable, “intelligibility,” was based on an accuracy score of written transcriptions. Munro and Derwing (1995a) investigated the accentedness, perceived comprehensibility, and intelligibility of Chinese-accented

English, and found that while there was a wide range among the accentedness judgements, the utterances tended to be highly intelligible and comprehensible. They concluded that a strong foreign accent does not necessarily reduce intelligibility and comprehensibility.

Again researching Chinese-accented English, Munro and Derwing (1995b) investigated processing time, accent, and comprehensibility ratings. The utterances that received low comprehensibility ratings also had longer processing times, revealing a direct, positive relationship between the two measures. The moderately and highly comprehensible utterances had significantly shorter processing times. However, the degree of accent was not found to be related to processing time. Derwing and Munro

(1997) found similar results when they studied Cantonese, Japanese, Polish, and Spanish speakers.

Rogers (1997) investigated word, sentence, and passage intelligibility of Chinese- accented English. Analyzing the English speech of two recently arrived graduate students who were native speakers of Mandarin Chinese, Rogers (1997) identified the specific sounds, or phones, which deviated from those of an L1 American English speaker. Using a word list that then focused on these potentially confusable phones, she tested the impact of segmental accuracy on word intelligibility for American listeners.

To investigate the intelligibility of longer speech, she had Mandarin talkers also read

English sentences and a passage, which were also rated by American English-speaking listeners.

Rogers (1997) found that sentence intelligibility was strongly correlated to word intelligibility, but that passage intelligibility was not. However, when the impact on intelligibility of the segments was analyzed in detail by consonants and vowels, she found that sentence and passage intelligibility were both strongly correlated with vowel accuracy, but not with consonant accuracy.

Using Chinese speakers who had been rated at two levels of intelligibility (high and low) in the Rogers (1997) experiment, Rogers, Dalby, and Nishi (2004) recorded sentences read by these speakers as well as L1 American English speakers. They presented the sentences in quiet, as well as in three increasing levels of background noise to L1 American English listeners. The results indicated that while the American English speakers were as intelligible with or without noise, the nonnative speakers were much more difficult to understand in noisy conditions. The low proficiency nonnative speakers became less intelligible in noisy conditions, though not significantly so. The more

surprising finding was that the high proficiency nonnative speakers, who were nearly as intelligible as native speakers in quiet conditions, became significantly less intelligible in noisy conditions.

Bent, Bradlow, and Smith (2007) investigated the effects of segmental errors on the intelligibility of Chinese-accented English. Intelligibility scores were calculated for each talker based on the number of key words that the native English speaking listeners correctly transcribed. Overall, vowel production accuracy correlated with intelligibility, while consonant production accuracy did not. However, when position-in-word was considered for consonants, only word-initial consonants correlated with intelligibility, whereas all other word positions did not, a finding which is consistent with spoken word recognition research on word onsets and “uniqueness points” (e.g., Billerey-Mosier,

2000; Marslen-Wilson & Zwitserlood, 1989; Radeau, Mousty, & Bertelson, 1989).

2.3.5.2 Speech Rate

One acoustic characteristic that is very relevant to the current study is speech rate.

Anderson-Hsieh and Koehler (1988) investigated the effect of both foreign accent and speech rate on native speaker comprehensibility scores. Three Chinese speakers with varying degrees of accentedness and one American speaker read passages at three different speaking rates. The researchers found that comprehension scores for all speakers corresponded to the degree of foreign accent and to speech rate. While comprehension scores decreased significantly from the regular to the fast speech rate, this decrease was most notable for the speaker with the strongest accent. The researchers concluded that speech rate may be more critical for speakers with more pronounced

accents. Alternatively, speech rate may be more critical for listeners with a difficult cognitive task to complete.

In Rubin’s (1994) review of L2 listening comprehension research, normal speech rates for different tasks by native speakers of English and the most comprehensible speech rates for L2 English learners were noted in two different studies. Tauroza and

Allison (1990) measured speech rates of British English speakers in words per minute

(wpm) for four types of speech: radio broadcasts, interviews, conversation, and lectures.

They found that average speech rates were 160-190 wpm for radio and interviews, 210 wpm for conversation and 140 wpm for lectures. In terms of the most comprehensible speech rates for L2 English learners, Griffiths (1992) found that the listener’s oral proficiency level and the talker’s speech rate both affected listening comprehension. L2

English learners at a low intermediate proficiency level best comprehended speech rates of 127 words per minute and experienced difficulty comprehending speech at rates of

185-250 words per minute.

The research reviewed in this section found both segmental and prosodic aspects of pronunciation, as well as speech rate, to have an effect on L2 speech intelligibility and listening comprehension. Therefore, the current study takes these factors into account. In order to establish a baseline of the extent to which segmental pronunciation affects L2 speech intelligibility, the current study includes this variable in the analysis model. In addition, although talker speech rate is not controlled, it is measured and analyzed for both the native and nonnative English talkers.

2.4 Research on Accommodations of the Native Speaker

Most research on second language acquisition has focused on the second language learner rather than on native speakers. However, native speakers in communication with nonnative speakers of English are able to adjust their speech and their listening perception in certain ways in order to accommodate different proficiency levels or unfamiliar accents. Some accommodation strategies have been found to be more successful than others and are reviewed below.

2.4.1 Speaking Accommodations

The tendency of native speakers of English to simplify their speech in certain ways when they are communicating with nonnative speakers has been well documented

(e.g., Ferguson, 1971). However, the extent to which “foreigner talk” simplifications are successful varies considerably. According to Haegeman (2002), native speakers’ foreigner talk simplifications involve syntax, lexicon, and phonology. Syntax simplifications include using more basic syntactic structures, such as deleting auxiliaries, pronouns, and articles. Lexical changes involve using an easier word in place of a more difficult one (e.g., car instead of vehicle, or house instead of residence). Phonological simplifications involve slower rates of speech and clearer articulation, which will be treated in greater detail below, since the focus of the current study is phonological.

2.4.1.1 Speech Rate

Since native speakers of any language tend to speak more quickly than do nonnative speakers, one of the more successful simplifications that native speakers employ to adapt their speech to the needs of nonnative listeners is to speak more slowly

(Hock, 1986). Several studies have found a relationship between rate of speech and listening comprehension, with nonnative listeners being impacted negatively by faster rates of speech (Conrad, 1989; Derwing & Munro, 2001; Griffiths, 1990, 1992; Munro &

Derwing, 1998).

Griffiths (1992) noted that while previous research (Conrad, 1989; Griffiths,

1990) had indicated that beyond a certain level higher rates of speech negatively impacted on comprehension, more details were needed about the nature of their relationship. For example, for what specific groups and on what specific texts is comprehension negatively affected by extreme speech rates (slow and fast)? Griffiths

(1992) thus created a greater range of speech rates for passages than those he had used in his 1990 study, in which he had found that speech rates faster than 200 wpm impaired comprehension for low intermediate Japanese learners of English. Natural speech at slow

(127 wpm), average (188 wpm), and fast (250 wpm) rates were recorded and presented to

Japanese listeners who had an average English proficiency level of low intermediate.

The findings confirmed that the slow rate of speech was significantly more comprehensible than either the average or the fast rates for these listeners. Although the average rate was more comprehensible than the fast rate, the difference was not significant.

Munro and Derwing (1998) conducted two experiments with Mandarin and

Canadian English speakers. They calculated speech rates as well as articulation rates, which were calculated by subtracting total pause times from the speech rates in syllables per second, rather than words per minute, since the latter measure had been critiqued as

being too imprecise (Griffiths, 1991; Hieke, 1985). In the first experiment, Munro and

Derwing (1998) asked talkers to speak at a normal and a slow rate. For the Mandarin talkers, the self-selected speech rates measured at 3.8 syll/sec for the normal rate and 2.9 syll/sec for the slow rate. The native English talkers slowed their speech to a lesser degree (69%; from 4.9 syll/sec to 3.4 syll/sec) than did the Mandarin talkers (76%; from

3.8 syll/sec to 2.9 syll/sec). Although similar pause times were added to the slow rates for both types of talker, the slower articulation rate for some of the Mandarin talkers included distorted syllables and may have confounded the impact of rate alone on comprehension. Native speaker listeners then rated the comprehensibility of all speaking conditions, finding the normal rates to be more comprehensible than the slower rates. In fact, native listeners rated the slow Mandarin speech to be more accented, as well as less comprehensible, than the normal speech condition (Munro & Derwing, 1998).

In the second experiment, Munro and Derwing (1998) synthetically manipulated the speech rates of the same Mandarin and Canadian English talkers from the first experiment, and ensured that no distortions were included. The researchers created three new sets of stimuli: one at the average native speaker rate (4.9 syll/sec), the second at the average Mandarin speaker rate (3.8 syll/sec), and the third at a rate 10% slower than the average Mandarin speaker rate (3.4 syll/sec). These new sets of stimuli, as well as a set containing the original unmodified normal rates of speech, were presented to native speaker listeners and rated as “too slow,” “just right,” and “too fast” on a scale from 1 to

9. They calculated that for native listeners, the preferred speech rate for L1 English was

4.7 syll/sec and for Mandarin-accented English it was 4.1 syll/sec. Therefore, although

native listeners preferred to hear Mandarin-accented speech at a slightly slower rate than

L1 English speech, the normal speech rate selected by the Mandarin talkers (3.8 syll/sec) was slower than the rate that native English listeners preferred.

For native English speakers in communication with nonnative listeners, therefore, the notion of “slow speech” may be significantly different. Although the native and

Mandarin talkers were given the same instructions, to slow their speech rate by half, the native speakers were less successful at it than were the nonnative speakers. Since the preferred rate of speech for native listeners was quite similar to their average speech rate, for native talkers to be more comprehensible to nonnative listeners, they may need to slow their speech into a range that they themselves would find irritating or less comprehensible, depending on their listeners’ English proficiency level. This may prove difficult to achieve, since Derwing (1990) found that native speakers’ estimations of their own speech rate could be inaccurate.

In order to determine the preferred speech rate for nonnative-, rather than native listeners, Derwing and Munro (2001) conducted a follow-up study. Using the same four sets of L1 English and Mandarin-accented English stimuli from their 1998 Experiment 2, they asked two groups of advanced nonnative listeners (Mandarin L1 and mixed L1s) to rate the speech as “too slow,” “just right,” and “too fast” on a scale from 1 to 9. They found that Mandarin listeners preferred both Mandarin and L1 English speech at a rate of

4.5 syll/sec, while the mixed L1 group of nonnative listeners preferred Mandarin- accented English at 4.1 syll/sec, and L1 English speech at 4.3 syll/sec. Not surprisingly, native listeners preferred faster speech rates (4.7 syll/sec) for L1 English speech and

slightly slower rates (4.1 syll/sec) for Mandarin-accented English. For all the nonnative listeners, unmodified normal speech rates were preferable to the slightly slowed rates and slightly speeded rates were preferable to the unmodified slow speech (Derwing & Munro,

2001).

The L1 of the listener, the L2 oral proficiency level, and individual differences all appeared to influence nonnative listeners’ speech rate preferences. For example,

Mandarin listeners preferred Mandarin-accented and L1 English at the same rate. In contrast, listeners from other L1 backgrounds preferred foreign-accented speech at a slower rate than L1 speech, as did native listeners. Compared with Zhao (1997), who included listeners with low proficiency levels and found that slower speech rates improved nonnative listeners’ comprehension, these findings with only high proficiency nonnative listeners suggest that speech rate preferences are influenced by proficiency level. Furthermore, since there was a wide range of preference ratings in the mixed L1 listener group, both L1 and individual differences appeared to play a role in speech rate preference (Derwing & Munro, 2001).

2.4.1.2 Pause Phenomena

Other researchers investigating speech rate have focused on pause phenomena rather than articulation rate. Griffiths (1990) noted that for teaching purposes, the speaking accommodations that native English speaking teachers can make to improve comprehension for their L2 English learners are in the areas of speech rate and pause phenomena. Pause phenomena include pause duration, distribution, and frequency. As with speech rate, the effectiveness of pausing as a strategy to improve L2 listening

comprehension depends on the listening task they are presented with and their oral proficiency level.

Blau (1990) investigated the effects of speech rate, syntactic simplicity, and pausing at natural boundaries on L2 English comprehension. First, to analyze the effects of speech rate and syntactic simplicity on two L1 listening groups, she asked Polish and

Puerto Rican ESL students to listen to recorded passages of English and answer multiple choice comprehension questions immediately afterward. Since neither reducing the rate of speech nor simplifying the sentence structure yielded significantly higher comprehension scores, she followed up with an investigation of the effect of pausing in terms of both duration and frequency.

Using a subset of the Polish and Puerto Rican ESL students from the first study as listeners, Blau (1990) presented the listeners with longer recordings in three conditions: at a normal speech rate (about 200 wpm), at a slowed speech rate (about 185 wpm), and at a normal speech rate with 3-second pauses inserted at some sentence, clause, and phrase boundaries (about 150 wpm). For both the Polish and the Puerto Rican listeners, the third condition was significantly more comprehensible than were the other two. For the Puerto Rican listeners, as their oral proficiency increased, so did the benefit of pausing, while the opposite relationship obtained for the Polish listeners. Since the

Polish listeners were English majors and may have been more proficient than the Puerto

Rican listeners, Blau (1990) speculated that both a certain threshold of proficiency was necessary to take advantage of the pauses, and also, that above a certain threshold of proficiency, increased pausing was no longer necessary for comprehension.

Blau (1990) therefore concluded that English language teaching professionals do not need to simplify their sentence structure in order to be more comprehensible to their students, and that slowing their rate of delivery will be beneficial only to students at the lowest levels of proficiency. In contrast, lengthening pauses at natural boundaries at normal rates of speech would be beneficial to most listeners, while the most proficient listeners would not need any special modifications to understand normal rates of speech.

Derwing (1990) investigated the effects of speech rate and pausing on communicative success between native talkers and nonnative listeners. The native talkers were presented with a short film which they were asked to narrate to two different listeners (native and nonnative) in separate sessions. The listeners’ comprehension was then tested with oral comprehension questions posed by an interviewer immediately afterwards.

Based on the results of the nonnative listeners’ comprehension questions, the native talkers were divided into two groups, successful and unsuccessful. The successful group’s listeners scored higher than 50% on the comprehension questions and the unsuccessful group’s listeners scored 50% or less. The recordings of the narrators of each of these two groups were then analyzed in terms of speech rate and pausing.

Surprisingly, only the unsuccessful narrators modified their speech rate. They increased their pausing, although they believed that they had slowed their articulation rate. In any case, the increased pausing strategy was not successful (Derwing, 1990).

Since these findings differed from previous studies which found both slower speech rates and pausing to be beneficial to comprehension, Derwing (1990) suggested

that the demands of the listening tasks (e.g., written sentence dictation vs. film narration) may have affected the results. In addition, after analyzing the narrations for both the nonnative and native listeners, a significant difference was found in the frequency of the lexical items used. When narrating to nonnative listeners, significantly more frequently occurring words were used than with the native listeners (Derwing, 1990). In Derwing

(1989), she reported that the unsuccessful narrators also included a lot more background detail than did the successful narrators. Therefore, the narrators made several adjustments in their speech (like lexical frequency and elaborative detail) that may have interacted with pausing and had an impact on nonnative listeners’ comprehension.

2.4.1.3 Clear Speech

Clear speech is elicited under research conditions by asking talkers to speak as if they were articulating for a hearing impaired or a nonnative listener. It involves slower speech rate, greater vowel contrasts, and higher pitch and volume, which are characteristics that have thus far proved consistent across languages with different vowel inventories (Smiljiac & Bradlow, 2007).

Bradlow and Bent (2002) investigated the extent to which clear speech proved to be a successful strategy for improving comprehension. They recorded two native English speaking talkers (one male, one female) who were asked to read 64 English sentences first in a conversational speech style, and next in a clear speech style. These two speech style conditions were then embedded in two different levels of white noise and presented to native and nonnative listeners (L1s = Bengali, Chinese, Hindi, Japanese, Korean,

Romanian, Slovakian, Spanish, and Thai), who were asked to transcribe each sentence

that they heard. Furthermore, all of the nonnative listeners returned to the lab 1-2 weeks later in order to be recorded reading the same set of 64 sentences.

Bradlow and Bent (2002) found that the native listeners showed a significant benefit from the clear speech style in contrast to the nonnative listeners who showed a relatively small benefit. The native listeners also showed a difference in comprehensibility due to the gender of the talker relative to both speech style and noise condition, which the nonnative listeners did not exhibit. In other words, native listeners derived a much greater benefit from the clear speech of the female talker than that of the male, and she remained more comprehensible than the male talker even in higher noise conditions. The nonnative listeners did not reflect this difference in comprehension based on gender. Surprisingly, the nonnative listeners who were most proficient in listening comprehension on the conversational speech sentences derived the least benefit from the clear speech. Lower proficiency nonnative listeners derived greater benefit from the clear speech enhancements, which was counterintuitive, since the most proficient listeners – i.e., the native listeners – derived the most benefit from the clear speech (Bradlow & Bent, 2002). In a later study, Bent, Bradlow, and Smith (2007) suggested that nonnative listeners require more of the speech signal to be intact before they can access contextual information from the sentence.

Haegeman (2002) noted that foreigner talk may also be employed in conversations without any native speakers. In the context of English as a Lingua Franca

(ELF), Haegeman recorded 49 business telephone calls between Flemish Dutch speakers of English and their international business partners at 7 different businesses in Flanders,

Belgium. She analyzed four of these phone conversations in order to determine the characteristics of foreigner talk displayed by nonnative speakers of English. She found that “[foreigner talk] is a display of a participant’s orientation to the co-participant’s perceived lower proficiency in the language” (p. 158). As a result, Haegeman (2002) suggested that adaptability (Verschueren, 1995) better described the interaction rather than negotiability in ELF communication, since negotiation is limited by the linguistic skills of the interlocutors and may not be possible in the case of some low proficiency nonnative speakers.

2.4.2 Listening Accommodations

English as a Lingua Franca and World Englishes research has been critiqued for neglecting the impact of context (Rajadurai, 2007). For example, one of the principal tenets of ELF is that interlocutors are equally responsible for ensuring successful communication. However, in certain contexts differences in status hold some interlocutors more accountable than others. In academia, teachers are more responsible than their students for ensuring successful communication in the classroom. Therefore,

ITAs are held accountable for classroom communication regardless of whether their students are native or nonnative speakers of English. In contrast, communication between international graduate students in their roles as students can legitimately assume interlocutors to carry an equal burden in conversations. In either case, listening comprehension skills can theoretically be improved through training for native and nonnative listeners. The following research reviews training focused on improving native

English speakers’ abilities to comprehend foreign-accented speech.

Rogers (1998) developed a short-term training program to improve the listening comprehension of native English speakers for Mandarin-accented speech. The training consisted of three one-hour sessions which provided listeners with both auditory and visual feedback. Pretests and posttests for both the training and the control groups of native English speaking listeners were compared to determine the level of improvement for the listening comprehension of words and sentences of two Chinese speakers. The training groups improved in both word and sentence comprehension, but gains in word comprehension were greater.

While the listening comprehension of a particular talker (talker-dependent) has been shown to improve as a result of greater exposure or familiarity, Clarke (2000) studied familiarity with a single talker to determine whether it could be generalized to other talkers with the same foreign accent (talker-independent). She trained native

English speaking listeners to be familiar with English sentences recorded by a Spanish L1 and a Chinese L1 talker. She then conducted a word intelligibility test in noise with carrier sentences recorded by both the familiar and new Spanish L1 and Chinese L1 talkers. Clarke (2000) found that the “familiar” talkers became more intelligible with training, but that the intelligibility gains did not transfer to the “new” talkers.

Using both regional and foreign accents in English, Kubota (2001) investigated whether American high school students would improve in attitude toward and comprehension of foreign-accented speech. The curriculum included the video

“American Tongues” which exposed students to a variety of regional dialects of

American English as well as excerpts from popular films depicting foreign-accented

English. At the end of the program, students completed a questionnaire which measured their attitude toward foreign-accented speech and their confidence levels in understanding it. In addition, they took a listening test of foreign-accented English, which they were asked to transcribe. Kubota (2001) found some improvements in attitudes toward foreign-accented speech and perceived confidence levels, but no significant improvements in listening comprehension as a result of training.

Focusing on native English-speaking social work students, Derwing, Rossiter, and

Munro (2002) investigated the relative effects of linguistic and cross-cultural awareness training on the comprehensibility of Vietnamese-accented English. One group of listeners received both linguistic and cross-cultural awareness training; a second group received cross-cultural awareness training only; and the third group received no training and served as a control. The training programs met once a week for eight weeks and pre- and post-tests and questionnaires gathered both qualitative and quantitative data.

Qualitative questionnaires measured the participants’ differences in attitude toward immigrants. The quantitative test included listening comprehension questions and sentence transcriptions.

The experimental groups showed significant improvement on the quantitative measures of listening comprehension and sentence transcription. In terms of their ability to successfully comprehend foreign-accented English, the group which received linguistic training showed substantial improvement, while the group which received only cross- cultural awareness training improved only moderately. The largest gain resulting from linguistic training was in the level of confidence to interact successfully that listeners

reported feeling when faced with foreign-accented English (Derwing, Rossiter, & Munro,

2002).

Similar to Clarke (2000), Bradlow and Bent (2003) investigated whether exposing native English listeners to Chinese-accented English would transfer to increased comprehensibility of an unfamiliar Chinese speaker of English. They conducted pre- and post-tests consisting of written transcriptions of Chinese-accented English sentences in noise. In one training condition, multiple Chinese-accented talkers were heard; in four training conditions, only one Chinese-accented talker was heard. Additionally, two control conditions were set, where one group received task training with native English speakers, and the other group received no training.

Bradlow and Bent’s (2003) findings indicated a significant task effect, since all post-tests revealed improvement as a result of either task or listening training.

Furthermore, subjects who received multiple-talker training improved in their ability to comprehend foreign-accented speech in general, not only of Chinese-accented English.

This improvement was even equivalent to that experienced by subjects who were trained on one Chinese talker, who was then also the talker on the post-test. In contrast to Clarke

(2000), Bradlow and Bent (2003) found that exposure to multiple foreign-accented talkers did result in talker-independent gains in listening comprehension.

Bradlow and Bent (2008) conducted two experiments on native speaker adaptations to foreign-accented English. In the first experiment, they examined the degree and rate to which native listeners would adapt to a single foreign-accented talker.

Using Chinese talkers at low, medium, and high levels of intelligibility as well as a

medium level Slovakian talker, the researchers asked the listeners to transcribe the 64 sentences they heard.

The findings showed that listeners adapted more quickly to the talkers at the higher levels of intelligibility, but that all listeners were able to adapt to and accurately comprehend the speech of each individual talker within the range of the 64 sentences.

That is, even talkers at the low and medium levels of intelligibility became more intelligible as the listeners’ exposure to them increased. This confirmed previous studies of talker-dependent comprehension gains for native listeners with sufficient familiarity

(Bradlow & Bent, 2008).

In the second experiment, Bradlow and Bent (2008) extended their study to talker-independent adaptation. Training conditions to which listeners were exposed included a single Chinese talker, multiple Chinese talkers, task training with American

English talkers, and a control group of listeners with no training. Training consisted of two consecutive days of sentence transcription followed by two posttests. One posttest was with the medium level Chinese talker; the second posttest was conducted with the medium level Slovakian talker.

The researchers found that even listeners who had only task training showed significantly higher comprehension scores than did the control group. The listeners with task training and those with single talker training showed improvement at equal levels.

The listeners who were exposed to multiple Chinese talkers exhibited the highest gains.

They also matched the performance levels of listeners who had received their task training from the same talker as the one in the post-test (Bradlow & Bent, 2008). The

findings from both experiments suggest that talker intelligibility level increases the rate of native listener adaptation and that talker-independent comprehension of a specific accented English can be achieved with exposure to multiple talkers.

The findings of the research reviewed above lend support to the idea that native speakers can adapt to communication with nonnative speakers. Whether native speakers will adapt is influenced by their attitude toward nonnative speakers and the context of communication. The following sections describe research into two different issues of context: first, purposes for communication, and next, the L1s of the interlocutors.

2.5 English for Specific Purposes and EAP

English language teaching has been divided into general and specific purposes.

Specific purposes have a practical or vocational application, such as business English, health care English, and English for Academic Purposes (EAP). Originally, EAP referred to English in all four language skills (reading, writing, listening, and speaking) as utilized in the university. Since the advent of training programs for International Teaching

Assistants (ITAs), listening and speaking skills became more closely associated with these ITA programs, while EAP became more narrowly defined as pertaining to only reading and writing skills. However, since ITA programs are focused on the oral English language skills needed for the specific purpose of teaching, the oral skills needed for studying at the university level are now the purview of no particular program.

As a result of being an outgrowth of EAP, ITA programs have continued to be influenced by the theoretical frameworks and research practices of EAP. At the same time, since ITA research is concerned with oral academic English, it has dominated this

area. On a much smaller scale, other researchers have continued to explore the oral

English needs of international graduate students. Therefore, both research on ITAs and international graduate students is included below.

2.5.1 International Teaching Assistants

Research on ITAs has been devoted to identifying the factors that affect the teaching and communication abilities of international graduate students who teach monolingual English-speaking undergraduates in US universities. Early research attributed communication breakdowns exclusively to ITA pronunciation problems. The research reviewed below reflects the trends in research and teaching that have moved away from a narrow focus on pronunciation toward a wider scope on the discourse level.

Today, ITA research focuses mainly on discourse analysis, and the teaching methods have shifted to Communicative Language Teaching, but the “foreign TA problem” persists.

The following researchers laid the groundwork for investigating international

Teaching Assistants and the development of the current ITA program curriculum: Bailey

(1984), Constantinides (1987), Nelson (1990), Smith (1994), Anderson-Hsieh, Johnson, and Koehler (1992), Orth (1982), Rubin and Smith (1990), Anderson-Hsieh and

Venkatagiri (1994), Hahn (1999), and Pickering (2000, 2001, 2004).

Early studies of the ‘foreign TA problem’ attributed most communication breakdowns to pronunciation problems (Heller, 1985; Nelson, 1991; Solomon, 1991).

However, by the mid 1980s researchers began to see the intercultural communication dynamic as more complex. For example, Bailey (1984) reported on the “foreign TA

problem” from the TAs’ point of view. The ITAs believed US undergraduates’ negative reactions to them was primarily due to academic laziness among undergraduates and a lack of maturity, rather than their own oral English proficiency. In fact, some studies began to examine the American undergraduate side of the communicative equation.

2.5.1.1 Undergraduate Attitudes toward ITAs

Several studies have shown that the intelligibility or comprehensibility of ITAs can be influenced by attitudes or stereotypes held by the listeners. Mestenhauser et al.

(1980) surveyed undergraduates about ITAs. Of their respondents, 43% reported that course quality was negatively affected by an ITA.

Rubin and Smith (1990) investigated the effects of accent, ethnicity, and lecture topic on the intelligibility of Chinese ITAs. They found that accent correlated negatively with teacher effectiveness ratings. Furthermore, there was an interaction between lecture topic and perceived ethnicity. Rubin (1992) presented American undergraduates with audio clips and photographs of speakers of English. The photographs included a mix of ethnicities. The researcher then solicited the undergraduates’ evaluations of the comprehensibility of the TA speech they had heard, as well as the TAs’ effectiveness as a teacher. Even when the audio clips were of native English speakers, the speech was evaluated as less comprehensible and the TAs’ teaching effectiveness was lower when the associated photograph was of a non-white talker.

In 1987, Constantinides noted that the “FTA problem” continued to be an issue of national concern in the US. She considered the problem to be a function of three interrelated issues at US research institutions: (1) large numbers of international graduate

students as compared to US graduate students; (2) low status of teaching as opposed to research; and (3) cultural chauvinism on the part of American undergraduates. She called for ITA language instruction to be situated in the teaching specific to US universities, including explicit teaching of the US educational system and the related attitudes and expectations of US undergraduates toward their classes and instructors.

Fox and Gay (1994) collected quantitative survey and qualitative interview data from university administrators, department chairs and supervisors, ITAs, native English- speaking TAs, and undergraduates at a Midwestern land-grant institution regarding their perceptions of ITA-related difficulties. They found that undergraduates had more negative attitudes toward ITAs than did administrators. The survey results reflected a normal distribution curve in undergraduate attitudes toward ITAs, as measured on a 5- point Likert scale. Students’ attitudes correlated with their GPAs, with more negative attitudes associated with lower GPAs. The surveys revealed that most undergraduates had direct experience with an ITA and that the courses they took from ITAs were concentrated at the beginning of their studies. The interviews with undergraduates corroborated the survey’s attitude ratings. Those students who had trouble with difficult academic material expressed the most frustration about ITAs’ limited English proficiency and a lack of recourse to resolve their problems.

Plakans (1997) investigated student factors that had an impact on attitudes toward

ITAs. For example, if students had experience abroad, they had more positive attitudes toward ITAs. Male students with lower GPAs and with majors outside liberal arts and sciences were most likely to have negative attitudes towards ITAs. Students with higher

GPAs reported having developed coping strategies, such as creating student study groups and preparing questions based on their textbook prior to attending recitations with an

ITA. Ironically, these students’ low expectations regarding the help they would receive from their ITA led them to take more responsibility for their learning.

Gill (1994) also examined the relationship between accent and perceptions of teachers and lecture comprehension, but had mixed results. While listeners preferred the

American accent, they perceived and comprehended British and Malaysian accents equally well. This suggested that Americans had more trouble comprehending unfamiliar accents in English, regardless of ethnicity. In addition, familiarity with the lecture topic also facilitated lecture comprehension. Cargile and Giles (1998) also found that familiarity with content interacted with perceptions of accent.

Orth (1982) compared teacher and student evaluations of ITA speech and intelligibility. Since no single TA received the same intelligibility ratings from both students and teachers, he suggested that extralinguistic features of ITA speech might account for the variation. Nelson (1990) reviewed some of the research on ITAs and observed that pronunciation issues had garnered the most attention, to the detriment of other important ITA concerns, namely effective teaching behavior and intercultural communication. She concluded that pronunciation had been assumed to be the only cause and therefore the only solution to the “FTA problem,” but that this assumption was in fact an oversimplification of a more complex dynamic. She noted that research had also indicated that American undergraduates preferred ITAs who used interactive and interpersonal teaching behavior and who discussed their native culture in class.

These studies on undergraduate attitudes towards ITA speech in the 1980s and

1990s contributed to a shift away from research on ITA pronunciation and intelligibility.

Since research found that undergraduate biases impacted their ability to comprehend

ITAs, and that motivated and successful undergraduates found ways to further their educational objectives regardless of whether they had an international or an American

TA, the importance of ITA pronunciation waned in the research community. Poor attitudes or low comprehension were, in some respects, dismissed as stereotypical reactions of low-performing students. On one hand, the relationship between American undergraduates and ITAs was acknowledged to be a complex interplay among many variables, both linguistic and cultural. On the other hand, both teaching and research took the “social” turn.

Throughout the 1990s, the rise of the Communicative Language Teaching model shifted ITA research attention away from pronunciation issues, except for a small but dedicated resistance (Morley, 1991). This shift in emphasis in ITA research on discourse mirrored that of the broader field of applied linguistics, indicated by Grabe (1990) in the volume of the Annual Review of Applied Linguistics devoted to discourse analysis:

“discourse analysis constitutes a, if not the, crucial research foundation for applied linguistics” (p. vii). This shift towards more discourse-oriented teaching and research led to the development of two different strands in ITA programming. ITA teaching included more discipline-specific materials and ITA speech research moved to a discursive perspective.

2.5.1.2 EAP and ITA Research

The theoretical concept of the ‘discourse community’ has informed the development of English for Specific Purposes in general, and English for Academic

Purposes (EAP) in particular. As the spoken branch of EAP, ITA programs and materials have been influenced accordingly. According to Casanave (1995), the discourse community theory was derived from Kuhn (1970), who described a discourse community as consisting of practitioners of a scientific specialty who share a language, beliefs, and practices. These practitioners’ work within the scientific community was facilitated by

“similar educations and professional initiations” and based upon their shared literature, its shared interpretations, and their shared professional goals (cited in Casanave, 1995, p.

87).

Swales (1990) expanded the concept beyond the scientific community. He defined a discourse community as a group of people who shared a set of social conventions that were directed towards some purpose. Swales (1990) suggested six criteria for determining a discourse community: (1) common goals, (2) participatory mechanisms, (3) information exchange, (4) community-specific genres, (5) a highly specialized terminology, and (6) a high general level of expertise. In order to acquire the knowledge bases upon which community membership was founded, formal or informal apprenticeship generally ensued – reminiscent of Kuhn’s (1970) ‘professional initiations.’

Bizzell (1982a, 1982b) and Bartholomae (1985) argued that this apprenticeship process was precisely what students were engaged in when entering academic disciplines.

In order to gain membership, students had to learn the genres and conventions commonly

employed by speakers and writers in their discipline. Berkenkotter and Huckin (1995) asserted that students must also become familiar with what Bazerman (1980, 1985) called the ‘conversations of the discipline,’ or the issues and problems current at any one time.

This familiarity was achieved through study and collaboration on research projects with experienced practicing scholars.

For international graduate students, this ‘professional initiation’ was complicated by the fact that they had to learn the terminology, genres, conventions, and cultures of their discipline in a second language. For some scholars in composition and rhetoric

(e.g., Spack, 1988), the split between ESL and the academic discipline proved especially problematic since ESL instructors generally lacked the discipline-specific knowledge necessary to train ITAs in the appropriate discourse of their community. Similarly, their home departments were not designed to teach ITAs how to improve their academic

English skills. However, even though ESL instructors did not hold advanced degrees in all the fields in which their students were apprenticing, attention could still be brought to the discourse differences and teaching styles that varied across disciplines. Discipline- specific materials for the ITA classroom began to be researched, developed, and incorporated into the ITA curriculum.

Byrd and Constantinides (1992) compiled and analyzed data on the language used by American professors, instructors, and TAs to teach mathematics. They combined the notion of pragmatics for accomplishing specific moves in the genre of academic lecturing with the specific vocabulary of the discipline of mathematics. They recommended the

development of new ITA teaching materials, which would include the most frequently used phrases and chunks of “mathematical” English.

Myers (1995) reported on the use of discipline-specific written materials to improve the oral skills of ITAs. She found that ITAs’ motivation and familiarity with the topic were high, and intelligibility and oral presentation skills both improved according to post-test SPEAK scores. At the same time, the initial investment of time and energy to record the ITAs’ readings of the written materials was significant, but could then be used again with other ITAs from the same discipline. Furthermore, for ESL instructors who were not experts in the same field, there was also a time investment incurred to check on the pronunciations of discipline-specific jargon with which they were unfamiliar.

Smith (1994) concurred with the use of discipline-specific materials in the ITA classroom and identified several areas important for the enhancement of ITA curricula, including but not limited to the following: (1) the incorporation of English for Specific

Purposes (ESP) methodology to direct ITAs in departmentally-based needs assessments;

(2) creating field- and classroom-specific materials for ITAs to use in practicing the spoken language; (3) practicing listening comprehension and interactive skills for the US university classroom; and (4) extending ITA development programs to meet the needs of nonnative English-speaking university faculty, graduate research assistants, undergraduates entering the job market, and nonnative English-speaking professionals in medical, technical, and business settings (pp. 53-54).

2.5.1.3 Discourse Analysis and ITA Intelligibility

As a result of the shift in ITA research to discourse analysis, ITA speech began to be investigated more globally. Accordingly, research and teaching in ITA pronunciation placed more emphasis on prosody than on segmentals. Anderson-Hsieh, Johnson, and

Koehler (1992) examined the impact of deviance in segmentals, prosody, and syllable structure on pronunciation evaluations of nonnative speakers of English from 11 different

L1s. They found that prosody was significantly related to global pronunciation ratings, but that the other two factors were not. Anderson-Hsieh and Venkatagiri (1994) focused only on aspects of prosody in Chinese-accented English for intermediate and high proficiency speakers. They found that the intermediate proficiency speakers were significantly different than high proficiency and native speakers. From this, they concluded that ITA intelligibility studies needed to evaluate the proficiency levels of subjects more carefully.

Hahn (1999, 2004) investigated the effect of lexical stress on the communicative effectiveness of ITAs. In her 1999 study she examined three different lexical stress patterns: native-like primary stress, nonnative primary stress, and no primary stress. She found that native listeners processed the information with native-like primary stress more easily and recalled more main ideas than with the other two nonnative stress patterns, and received higher ratings of communicative effectiveness. Between the two nonnative stress patterns, evaluative reactions to the second pattern were higher. However, information was processed more easily and slightly more main ideas were recalled from the third pattern, which had no primary stress.

Hahn noted that while the ITA field expressed overwhelming preference for prosody over segmentals in pronunciation instruction, there was little empirical evidence to support that preference. In her 2004 study of the same issues on ITA intelligibility,

Hahn found that when correct primary stress was used, it enhanced content recall and favorable evaluation of the speaker. In addition, processing time followed the same pattern of improvement as in her 1999 study, though not significantly so. Drawing on the findings of both of her studies, Hahn recommended including pronunciation teaching in the ITA curriculum, especially that of prosody.

Pickering (2000, 2001, 2004) also investigated the impact of prosody on ITA intelligibility, but focused on the structure and function of intonation rather than on word stress. In her 2001 study, she found that ITAs’ use of abruptly falling tones not only negatively impacted upon coherence, but also on rapport building with American undergraduates. As a result, ITAs with nonnative intonational patterns were judged to be not only less comprehensible, but also less inclusive or accessible as teachers.

Halleck and Moder (1995) also stressed the importance of proficiency level when evaluating the relative importance of linguistic and strategic competence to ITA speech.

Previous ITA research, such as that of Hoekje and Williams (1992) privileged sociolinguistic discourse and “compensatory” strategies in the ITA curriculum. Halleck and Moder found that while it was true that “compensatory” strategies were very helpful for improving the performance of more proficient ITAs, less proficient ITAs lacked the linguistic base necessary to take full advantage of more communicative “compensatory”

strategies. As a result, they emphasized the value of linguistic competence and its role in communicative competence, which had been neglected in favor of discourse orientations.

2.5.2 International Graduate Students

Very little research (Ostler, 1980; Mason, 1995; Ferris & Tagg, 1996a, 1996b;

Liu, 1996; Kim, 2006) has investigated international graduate students’ oral and aural

English needs beyond the classes they teach as ITAs. Ostler (1980), Mason (1995) and

Ferris & Tagg (1996a, 1996b) investigated the oral and aural needs of international graduate students from a variety of stakeholders’ perspectives.

Ostler (1980) conducted a survey at the University of Southern California (USC) among both undergraduate and graduate ESL students. First, she hoped to determine their needs across all four language skills: reading, writing, listening, and speaking.

Second, she asked them to assess their own oral/aural proficiency in a variety of settings and with a variety of interlocutors. Third, she evaluated their sentence combination and summarization skills.

Not surprisingly, she found that while undergraduate and graduate students shared the need to be able to read textbooks and take notes during lectures, they differed on every other academic need they were asked to rank. Undergraduates felt they needed preparation mainly for multiple-choice tests and lab reports, while graduate students wanted more practice reading academic journal articles and giving oral presentations.

Interestingly, the students rated themselves significantly more highly on oral/aural proficiency in settings that involved daily business transactions (grocery and bank clerks, waiters, etc.) than in settings with academic peers and professors, even in informal

conversation. Ostler (1980) suggested that the language required for these latter types of settings was less formulaic and practiced, and ESL curricula should therefore include more opportunities for creative language generation.

Mason (1995) interviewed both international graduate students and instructors at

Georgetown University. Her interviews with the international graduate students were intended to identify their “perceived degree of lecture comprehension” (p. 201). The interviews with the instructors found “a growing expectation of oral participation in the lecture hall” (p. 217). A comparison of the results of both sets of interviews suggested that graduate students needed both oral and aural skills development to function in their dual roles as students and instructors (p. 217).

In two articles discussing different aspects of the same research study, Ferris and

Tagg (1996a, 1996b) focused on the oral/aural skills required by subject-matter instructors at four different types of tertiary institutions. Their 1996a study resulted in six generalizations about the listening and speaking tasks that university instructors expect of their students. First, all instructors agreed that lecture notetaking was a very important academic skill. Second, levels of oral participation expected from students in class varied significantly by discipline, with business courses demanding the most and science courses the least. Third, although tasks such as debates, student-led discussions, and outside-class interactions with native speakers are all commonly included in current EAP textbooks and courses, they were very uncommon in any discipline. Fourth, class size had a significant effect on participation rates, with larger classes including very little.

Fifth, oral presentations were uncommon, except in business courses. Finally, when oral

presentations were assigned, students were often expected to work in pairs or groups, rather than alone, especially in business and engineering courses.

Ferris and Tagg (1996a) acknowledged, however, that little research, including their own, solicited the opinion of the international students themselves. In addition, based on the work of Flowerdew (1994) and Mason (1995), Ferris and Tagg (1995a) suggested that an instructor’s “willingness and ability to allow or promote in-class interaction” may be affected by whether the instructor is a native or nonnative speaker of

English (p. 35). The researchers added that although they did not include that factor in their survey, “several of our survey respondents commented that they themselves and many of their peers were nonnative speakers of English” (1995a, p. 35).

Ferris and Tagg (1995b) offered three recommendations for EAP programs: (1) to recognize the differences between undergraduate and graduate student needs, even within the same major, as well as the differences across majors for graduate students, (2) to include discipline-specific materials whenever possible, as well as offering some department-specific EAP courses, where feasible, and (3) to better prepare students for a notable shift toward greater oral participation in classes by exposing them to a variety of lecturers and delivery styles.

Liu (1996) researched international graduate students’ oral participation in their academic content courses. He found that sociocultural and affective factors inhibited class participation, while cognitive and pedagogical/environmental factors facilitated it.

The inhibiting sociocultural factors included showing respect for the teacher by not interrupting and expecting that small group discussions or talking to the professor after

class would compensate for the lack of in-class participation. Inhibitive affective factors included worrying about wasting other’s time or appearing too egocentric by asking questions in class. Cognitive factors that facilitated participation were obtaining information, ideas, and opinions of others, training one’s own thinking and ability to synthesize ideas and information, demonstrating one’s own knowledge and preparedness for class, and contributing to others’ understanding. Pedagogical and environmental factors that facilitated participation included seminar settings, inclusive teaching style, lively atmosphere, support for participation from the instructor and American peers, and one’s participation affecting one’s overall grade.

Kim (2006) surveyed the academic oral communication needs of East Asian international graduate students in the arts, humanities, social sciences, business, and education. She argued that despite a shift of emphasis in the academic classroom to greater oral participation (Lucas & Murray, 2002; Mason, 1995; Meyers & Jones, 1993), few studies have investigated students’ assessments of their own needs and concerns

(e.g., Liu, 2001). According to students’ perception of their oral communication needs, the tasks necessary to classroom success ranked in the following order, from the most important to the least: (1) giving formal oral presentations, (2) understanding lectures, (3) general listening comprehension, (4) class participation, (5) communication strategies with peers, and (6) communication strategies with professors. Although students ranked oral presentations as the most important skill to academic success, followed by lecture comprehension, they considered listening skills to be much more easily acquired than speaking skills.

These studies assessing the oral English needs of international graduate students outside their teaching duties have been broadly descriptive in nature. They have contributed toward an overall understanding of the tasks required of graduate students, and the listening and speaking skills necessary to complete these tasks, but they did not explore the specific oral/aural proficiency levels needed in any detail. Most graduate schools require a threshold level on the Test of English as a Foreign Language (TOEFL).

However, until 2007, this test did not evaluate spoken English proficiency and, not surprisingly, there was little correlation between a student’s oral English proficiency level and his/her TOEFL scores (Yule & Hoffman, 1990; Halleck & Moder, 1995). Even now that a speaking section is included on the TOEFL, most graduate schools do not have a threshold score for admissions based on this section.

At the same time, a threshold level of oral English proficiency is often required of

International Teaching Assistants in the US. English language teaching in China is focused mainly on reading and grammar skills, in order to facilitate student success on the TOEFL. In the US, graduate schools require specific overall TOEFL scores for admission, but not specific speaking section subscores. As a result, when Chinese graduate students seek teaching assistant positions in the US, they are largely unaware of and unprepared for the oral English proficiency level required of ITAs. Furthermore, the oral English proficiency level necessary for success in academic coursework and research has been largely unexamined, and therefore is undetermined in any specificity, to date.

2.6 World Englishes (WE) and English as a Lingua Franca (ELF)

According to the Institute for International Education (IIE), higher education is consistently one of the top exports of the United States. As increasing numbers of international students choose the US for postgraduate study, especially in disciplines such as business and engineering, standards of intelligibility need to consider the impact of the

L1s on communication between interlocutors. Research that investigates intelligibility between international users of English has developed largely in response to World

Englishes.

The research that has developed out of World Englishes (WE) theory can be categorized into two related branches: linguistic description of nativized Englishes and intelligibility studies, the latter of which has investigated the intelligibility of WE from the perspective of nonnative English listeners rather than only relying on the judgements of native speakers of English. Kachru (1985, 1986, 1997), Bamgbose (1998), Smith and

Nelson (1985), and Smith (1992) are among the pioneering scholars of World Englishes.

Another significant development in WE theory was to propose that English was no longer the exclusive property of its native speakers (Widdowson, 1994; Cook, 1999).

While WE has been focused primarily on theory development and descriptive research of nativized Englishes in the Outer Circle, English as a Lingua Franca (ELF) has focused on the use of English in the Expanding Circle and the practical applications of

WE theory to English Language Teaching (ELT). Since ELF theory is an outgrowth of

WE, the research that has developed out of ELF theory can also be categorized into two related branches: linguistic description based on ELF corpora and intelligibility studies.

Seidlhofer (2001b) and Mauranen (2003) have both begun to develop corpora in ELF interactions, Seidlhofer focusing on European English for a variety of purposes and

Mauranen focusing on ELF for academic purposes in a non-English speaking country.

The phonology of English as an International Language by Jenkins (2000) built on some previous research of ELF communication, drawing attention to the need for further investigation of the intelligibility of English as a Lingua Franca. The following research investigated the effects of speaker and listener variables on ELF intelligibility and comprehensibility. Anderson-Hsieh and Koehler (1988), Anderson-Hsieh, Johnson and

Koehler (1992), Munro and Derwing (1995a, 1995b), Derwing and Munro (1997), Van

Wijngaarden et al. (2002a, 2002b), Major et al. (2002), Bent and Bradlow (2003),

Meierkord (2004), Rogers, Dalby, and Nishi (2004), Field (2003; 2005), Date (2005),

Kirkpatrick and Saunders (2005), Deterding (2005), and Deterding and Kirkpatrick

(2006).

2.6.1 Terminology in WE Theory

Kachru (1985) argued for a socially realistic view of the interactional contexts of

World Englishes, drawing on the works of Halliday (1978), Labov (1966, 1972), and

Saville-Troike (1981). Prior to Kachru’s (1985) writings, the terminology used to classify countries in terms of their English usage was in reference to L1 English: English as a Native Language (ENL), English as a Second Language (ESL), or English as a

Foreign Language (EFL). In his 1985 treatise, Kachru pointed out that “the status of

English in the language policies of [ESL and EFL] countries changes from time to time.

What is an ESL region at one time may become an EFL region at another time, or vice

versa (Kachru, 1985, p. 14). Therefore, he recommended a new taxonomy which would reflect the types of spread, patterns of acquisition, and functional domains for English usage worldwide.

Kachru (1985) argued for countries to be conceived as situated within three concentric circles of English usage: Inner, Outer, and Expanding Circles, which are defined in terms of intranational (Inner, Outer) or international usage (Expanding). One of his main concerns was with establishing appropriate norms and standards for each

Circle of English usage. While native norms are appropriate for the Inner Circle, Kachru asserted that they do not respect the pragmatic sociolinguistic usage of English in the

Outer Circle. For this Circle, he suggested that local norms and standards of English were more appropriate. In fact, Kachru and Nelson (1996) recommended rejecting the concept of interlanguage, popularized by Selinker (1972, 1992), for the Outer Circle, since a native language variety of English as a target language is neither desirable nor accessible. They argued that interlanguage can only be valid if two crucial elements are present: “the desire of learners of English to emulate one or another Inner Circle English model, and the availability of such models in accessible materials, not only in the classroom but also in broader social and cultural interactions. Neither of these conditions can be shown to obtain in broad ways in the Outer Circle” (Kachru & Nelson, 1996, p.

80).

As for the Expanding Circle varieties of English, Kachru (1985) described them as having international applications rather than intranational, since English was neither a first nor a secondary official language. Although Kachru (1985) referred to Expanding

Circle varieties as “performance varieties,” he later revised his position in light of the substantial changes English had undergone in Asia during the intervening decade

(Kachru, 1997). He reviewed the case of Japanese English in the Expanding Circle according to the following categories: Historical, Functional, Formal, Attitudinal,

Pragmatic, and Acquisitional (Kachru, 1997, p. 70). He recognized that while Japanese

English had limited intranational applications, it did have significant intraregional applications in Asia. He concluded that “[w]hat we need is a shared designing by all the users of World Englishes, in our own way, with our own messages” (Kachru, 1997, p.

85). Kachru’s conception of World Englishes, then, continued to be both pluricentric and inclusive of a diversity of English varieties. Kachru (1997) considered the use of the term “international English” to be “misleading in more than one sense: It signals an international English in terms of acceptance, proficiency, functions, norms, and creativity. That is far from the reality” (pp. 214-215). Kachru (1997) preferred the term

“World Englishes,” also the title of the journal he edited, to characterize the world’s multiple uses of a variety of Englishes over the terms “English as an International

Language (EIL)”, “English as a lingua franca,” or “world language English.”

Bamgbose (1998), although comfortable calling the role of English worldwide an

“international lingua franca,” otherwise concurred with Kachru in his ‘pluricentric’ view of English since “the emergence of separate national norms presupposes a certain degree of divergence between varieties” (p. 11). For Bamgbose (1998), the imposition of an international standard could not succeed in a way that would sufficiently respect the

“pragmatic and creative” processes of English nativization on the local level (p. 1).

Instead, Bamgbose (1998) envisioned a more organic interaction of international varieties of English, unified by “the fact of common origin, and in the case of nonnative varieties, universal language learning strategies, and institutional context of language acquisition”

(p. 12). Bamgbose identified five factors for determining whether a linguistic innovation will become an accepted part of the local norm: demographics, geography, codification, authority, and acceptability, the last two being the most important. Demographics referred to the total number and social status of the speakers using the innovation.

Geography referred to how widespread the geographical distribution of the innovation was. Codification indicated that the innovation appeared in written form such as in formal grammars, dictionaries, and other reference works, while authority had to do with acceptance by the teaching and publishing circles. Finally, acceptability referred to acceptance by the majority of the speech community. Bamgbose (1998) concluded that

“[i]f an international standard does emerge, it will not be identical with any national variety, native or nonnative, because all the varieties would, in varying degrees, have contributed to it” (p. 12).

2.6.2 Terminology in ELF Theory

In her 2000 publication, Jenkins used the term “English as an International

Language” (EIL), as she believed the term would be more easily identifiable for scholars outside of Europe. Since 2000, however, the term “English as a Lingua Franca” (ELF) came into wider use, and she therefore opted for that term in subsequent publications.

She explained her preference for the term ELF over EIL:

ELF emphasizes the role of English in communication between speakers from

different L1s, i.e., the primary reason for learning English today; it suggests the

idea of community as opposed to alienness; it emphasizes that people have

something in common rather than their differences; it implies that ‘mixing’

languages is acceptable . . . and thus that there is nothing inherently wrong in

retaining certain characteristics of the L1, such as accent; finally, the Latin name

symbolically removes the ownership of English from the Anglos both to no one

and, in effect, to everyone. (Jenkins, 2000, p. 11)

In her support for ELF over EIL, Seidlhofer (2004) added that ELF “best signals that it is those nonnative users that provide the strongest momentum for the development of the language in its global uses as ‘agents of language change’” (p. 212).

As can be seen in the above discussion of terms, one of the first issues that arises in discussing ELF is whether or not native English speakers are included. Jenkins (2000) initially defined ELF as being used “in communication with other ‘non-native speakers,” rather than English used as “a foreign language in communication with its ‘native speakers’” (p. 1). The phonology of English as an International Language included a section on pronunciation learning for ‘native speakers’ of English who intended to use

English in international contexts (pp. 227-229), however, which implied that L1 English speakers were included in conceptions of ELF. In a later publication, Jenkins (2005) clarified that “[a]lthough native speakers of English are not excluded from [ELF] interactions, they are a very small minority, given that English is the first language of

100

only around 25 percent of the world’s English speakers” (Crystal, 2003, cited in Jenkins,

2005, p. 9).

Jenkins (2007) noted that while a lingua franca is generally “a contact language used among people who do not share a first language” (p. 1), the case of English as a

Lingua Franca is unique. According to Seidlhofer (2004), “ELF interactions often also include interlocutors from the Inner and Outer Circles”; however, the “nativeness criterion” does not apply to the concept of ELF (pp. 211-212). Jenkins (2007) affirmed

Seidlhofer’s (2004) distinction between ELF speech events and the development of ELF standards, as derived from corpus data. ELF speech events may or may not include Inner and Outer Circle English speakers, but if they do, they are not representative of an

English “reference point” (Jenkins, 2007, p. 3). ELF corpus data, on the other hand, are designed to describe interactions among Expanding Circle English speakers, and therefore are not concerned with native English speech. Jenkins’s (2000) project was to identify the core phonology necessary for ELF intelligibility; however, in order for her goal to be realized, significantly more research needs to be conducted.

2.6.3 Linguistic Description Research in WE

Since WE researchers reject native varieties of English as appropriate models for the Outer Circle, one branch of WE research has focused on linguistic descriptions of new or nativized varieties of English. One of the earliest examples of this type of linguistic description research in WE was Kachru’s (1965) description of Indian English.

Later, Pride (1982) classified and described the nativized varieties of English in Africa and Asia, such as in Cameroon, Nigeria, India, Sri Lanka, Malaysia, Singapore, and the

101

Philippines. Further work in WE can be characterized by the published descriptions of

Malaysian English (Wong, 1983), Philippine English (Gonzalez, 1983), Singapore

English (Tay & Gupta, 1983; Butler, 1997b), Indonesian English (Nababan, 1983), Thai

English (Sukwiwat, 1983), Nigerian English (Bamgbose, 1998), Hong Kong English

(Bolton, 2003), and Singapore English (Deterding & Low, 2005), among many others.

Many WE researchers have also concerned themselves with defining what criteria should be employed to define an English variety as new or nativized. Platt, Weber, and

Ho (1984) suggested several criteria for determining whether a variety of English had become nativized. Among these criteria was that the variety be used in the educational system as a medium of instruction, as well as taught as a subject. Furthermore, a nativized English should be used for a range of functions, including social, governmental, and media communication in both spoken and written forms. Finally, it should reflect some language features of the local region, including specific sounds, intonation patterns, sentence structures, words, or expressions (Platt, Weber, & Ho, 1984, pp. 2-3). Since most Outer Circle countries were once colonies of an Inner Circle country, many retained

English for official and institutional uses, and therefore could meet all of these criteria.

Asia, especially, is a region in transition in terms of English usage, including many Expanding Circle countries as well as former colonial Outer Circle and newly emerged Outer Circle countries. Llamzon (1983) identified four criteria specifically for

English usage in Southeast Asia. According to Llamzon, the nativization process of

English was defined by ecological, historical, sociolinguistic, and cultural features.

Ecological features were defined as comprising the nativized English’s linguistic

102

environment, which exhibited three characteristics: polyglossia, code switching and mixing, and lexical shifting. Historical features referred to the fact that the new variety of English had persisted long enough to have stabilized. Sociolinguistic features included the domains of use and the functions of language that speakers performed in culturally specific ways. Cultural features indicated the transmission of the culture through written literature in the new variety of English.

In the case of Chinese Englishes, Bolton (2003) added three more features to those identified by Llamzon (1983): linguistic, attitudinal, and political. Linguistic features entailed “the identification of sets of distinctive linguistic items typically associated with a new variety” (Bolton, 2003, p. 46). Attitudinal features played a role in determining the difference between items accepted by the local speech community as expressions of identity and those that were rejected as errors. Finally, political features were those associated with the government and educational institutions.

In documenting Southeast Asian Englishes, Butler (1997a) also recognized the importance of locally accepted linguistic features. In addition, she noted the importance of the history of that variety’s usage and its “key features of the physical and social environment” (p. 106). Finally, she identified a body of written work as an important feature of a nativized English, including not only literature, but also dictionaries and style guides “which show that people in the language community look to themselves, not to some outside authority, to decide what is right and wrong in terms of how they speak and write their English” (p. 106).

103

2.6.4 Linguistic Description Research in ELF

For her dissertation research, Jenkins (1995) recorded the interactions of international examinees for the Certificate in Advanced English (CAE) during their interview practice sessions prior to sitting for the exam. The first of two studies included a total of six interlocutors, of whom two were Japanese, three were Swiss-German, and one was Swiss-French (and fluent in French, German, and Italian). In the second study, which used similar tasks as the first but was not in preparation for the CAE, a total of eight interlocutors participated. Their nationalities were as follows: two Japanese, one

Korean, one Taiwanese, one German, one Italian, one Portuguese, and one Colombian.

In both studies, the researcher paired interlocutors with different L1 backgrounds to the greatest extent possible (in the Swiss-German / Swiss-French pair, the Swiss-

French interlocutor was also fluent in German). Jenkins (1995) transcribed and analyzed her participants’ speech data for points at which communication broke down and then noted which phonological features were present which may have led to these breakdowns. Jenkins (2000) further examined these phonological features in terms of

‘learnability,’ and ‘teachability’ to develop a Lingua Franca Core of features which she proposed would lead to improved intelligibility in ELF interactions.

Two large-scale corpus projects are dedicated specifically to collecting ELF interactions. Seidlhofer (2001b) built on Bamgbose’s (1998) example of codifying nonnative varieties, which would allow researchers to differentiate innovations from errors as determined by the local standard, and recommended establishing a database of

English as a Lingua Franca. Seidlhofer’s (2001b) Vienna-Oxford International Corpus of

104

English (VOICE) contains recordings of spoken ELF interactions among “fairly fluent adult speakers from a wide range of first language backgrounds, whose primary and secondary education and socialization did not take place in English” (p. 146). The speech events include “private and public dialogues, private and public group discussions and casual conversations, and one-to-one interviews” (p. 146). In their articles in English

Today, Modiano (2001), Seidlhofer (2001a) and Jenkins (2001), argued for the recognition of Euro-English as an accepted variety of English. Modiano noted that within the political entity of the European Union (EU), certain fossilized constructions of

English are commonplace and acceptable. Seidlhofer called for a systematic codification of Euro-English which could then lead to the development of dictionaries, grammars, and appropriate teaching materials. Finally, Jenkins observed a need for determining a core phonology which would ensure mutual intelligibility of the Englishes used in Europe.

In designing an academic ELF corpus, Mauranen (2003) followed Swales’s

(1990, 1998) theory of discourse community applied to the academic community – i.e.,

“a social formation with its own discourses that serve both as resources and as products of the community . . . [which] bring cohesion to the community and mark its identity” (p.

519). There are two corpora in the US which are comprised of academic spoken English:

The Michigan Corpus of Academic Spoken English (MICASE, Simpson, Briggs, Ovens,

& Swales, 1999) and the TOEFL 2000 Spoken and Written Academic Language (T2K-

SWAL) Corpus (Biber, Reppen, Clark, & Walter (2001). However, neither of these two corpora focuses on ELF interactions and both are in a native English-speaking country.

Therefore, Mauranen (2003) proposed compiling a corpus of ELF in Academic Settings

105

(ELFA) outside an L1 English country. She recorded academic speech acts at two universities in Finland and at international conferences where English is the medium of communication. She excluded native English speakers as much as possible, but coded them if they were present. Mauranen’s (2003) ELFA corpus will complement the other academic speech corpora by focusing on international student speech events and will also complement the other ELF corpus by restricting the social domain of interaction to the academic discourse community.

A smaller study analyzing syntactic variation in ELF speech was based on a corpus created by Meierkord (2004). The corpus consisted of 22 hours of naturally occurring informal conversations between 49 Outer and Expanding Circle interlocutors.

The Expanding Circle speakers were determined to be at two proficiency levels, the competent and the less competent, based on their grammatical accuracy. Utterances were coded into three categories: (1) “regular” if they conformed to an L1 variety of English grammar, (2) “marked” if they conformed to a nativized or L2 English variety, or (3)

“doubtful” if they did not fit either of the previous two categories.

Analyses of the utterances showed that 88% of all utterances were regular, 9% were marked, and 3% were doubtful. Surprisingly, 95% of the higher proficiency

Expanding Circle utterances and 94% of the Outer Circle utterances were regular, showing consistency to an L1 norm according to proficiency level, rather than Circle of

English usage. In contrast, 73% of lower proficiency Expanding Circle utterances were regular. Therefore, Meierkord (2004) concluded that ELF is a syntactically heterogeneous form of English with three identifying characteristics: (1) an

106

overwhelming correspondence to L1 English grammar, (2) nativized forms as well as L1 transfer and L2 developmental patterns, and (3) simplification, regularization, and leveling processes.

2.6.5 Intelligibility Studies in WE and ELF

Intelligibility has been investigated from the perspectives of World Englishes and

English as a Lingua Franca. Researchers have defined these perspectives as moving

“beyond the native speaker” (Cook, 1999) by determining intelligibility from the perspective of Outer Circle and Expanding Circle listeners. Smith and Nelson (1985) posited a model for listening comprehension consisting of three categories: (1) intelligibility, which is word/utterance recognition; (2) comprehensibility, which is word/utterance meaning (locutionary force); and (3) interpretability, which is the meaning behind word/utterance (illocutionary force).

Levis (2006) noted that in terms of research applications this third category of listening comprehension, interpretability, has been largely abandoned. More commonly at the present, intelligibility and comprehensibility have been defined in accordance with

Munro and Derwing (1995a, 1995b). Munro and Derwing define “intelligibility” as the

“extent to which a speaker’s message is actually understood by a listener” (1995a, p. 76), as measured by written transcriptions or oral repetitions; and “comprehensibility” as the degree of difficulty involved in processing the speaker’s message, as measured by subjective “perceived comprehensibility” ratings or listener processing times (1995b). In fact, Weil (2003) compared these two common measures of comprehensibility directly

107

and found a correlation between response latencies and subjective ratings of overall comprehensibility for native English-speaking listeners.

Unlike L2 listening comprehension studies, which are designed to gauge the comprehension or perception abilities of second language English learners as determined by native English speakers, WE and ELF intelligibility studies determine intelligibility from the perspective of nonnative English speakers. Smith and Nelson (1985) argued that native speakers were neither the sole judges of what is intelligible nor were they most intelligible to nonnative listeners.

In addition, Smith (1992) asserted that mutual intelligibility was not a realistic goal since it was not even a reality between speakers of native varieties of English. He contended that “[o]ur speech/writing in English needs to be intelligible only to those with whom we wish to communicate in English” (Smith, 1992, p. 75). Smith found that

British and American English speakers were neither the most easily understood, nor the best able to understand different varieties of English. Therefore, he concluded that being a native speaker was not as important to comprehension as listener proficiency level or familiarity with different varieties of English.

Smith’s (1992) research explored the concept of standard English and what it meant to use it in a global context. He observed that in his study, both native and nonnative listeners had considered “standard English” to be inclusive of several nonnative accents. Therefore, he concluded that the accented speech of educated nonnative speakers was accepted as being just as intelligible as the speech of native

108

speakers of English. In other words, he asserted that accented English, too, belonged in the category of “standard English.”

2.6.6 Listener Factors that Affect ELF Intelligibility

Several studies have investigated whether one’s own accent in English is easier to understand than other accents in English, or whether native varieties of English are the most intelligible. They wanted to determine whether listener variables such as familiarity with a particular accent, especially that of their own L1, would facilitate comprehension.

In a review of research on listening comprehension (Anderson-Hsieh & Koehler, 1988;

Bilbow, 1989; Brown, 1968; Ekong, 1982; Richards, 1983), Flowerdew (1994) found, not surprisingly, that unfamiliar accents could cause comprehension problems. Similarly,

Eisenstein and Berkowitz (1981) found that ESL learners understood standard English more easily than accented English with which they were unfamiliar (either foreign- or regional L1-accented).

Focusing on Chinese-accented English, Flege (1988) examined the age at which the learner began L2 acquisition, the influence of L2 experience, whether nonnatives with a foreign accent could accurately evaluate another’s foreign accent, and whether removing pauses made sentences sound less accented. He found that more experience with English affected perception, but not necessarily production, of English sounds.

Although highly experienced Chinese speakers of English judged accented English similarly to native speakers of English, they themselves had strong accents. Therefore,

Flege (1988) suggested that adult L2 learners in naturalistic settings of acquisition would not improve their pronunciation without formal instruction.

109

One of the problems with examining the effect of familiarity on L2 speech intelligibility has been that the variable has been variously defined by researchers as (1) familiarity with a specific accent (Tauroza & Luk, 1997), (2) familiarity with the test speaker (Gass & Varonis, 1984), (3) familiarity with accents in general (Smith & Nelson,

1985; Smith, 1992), and (4) familiarity with the topic (Gass & Varonis, 1984; Schmidt-

Rinehart, 1994). While a lack of familiarity has negative repercussions for the listener, there is somewhat less support in the literature for familiarity improving comprehension.

Flowerdew (1994) hypothesized that listening comprehension would be easier when both listener and speaker share the same accent (and L1), which is also defined as an interlanguage match (Bent & Bradlow, 2003). Support for this hypothesis can be found in Wilcox (1978), in which Singapore listeners better understood their own accent and in

Brown (1968), in which West Africans better understood their own accent. However, other studies offered only partial support for this hypothesis. In Smith and Bisazza

(1982), Japanese listeners better understood their accent in English, but subcontinental

Indians better understood the American accent, which may simply reflect that L1 was not sufficiently controlled, given the wide range of L1 dialects in India.

As part of the revision process for the TOEFL, Major, Fitzmaurice, Bunta, and

Balasubramanian (2005) investigated the effects on listening comprehension for ESL listeners of Southern American English (Georgia), African American Vernacular English

(Georgia, Pennsylvania), subcontinental Indian English (Hindi L1, Malayalam L1), and

Australian English (Australia) as compared to Standard American English (Pennsylvania,

Oregon, Michigan). They used two speakers for each accent, a male and a female, except

110

for SAE, for which they used four speakers. The speakers were recorded reading 12 different lectures in a variety of disciplines. These lectures were heard by 180 ESL listeners (36 Chinese L1, 30 Japanese L1, and 15 Spanish L1) who answered four multiple choice comprehension questions afterward, for a total of 48 test questions. The

ESL listeners’ comprehension test scores were analyzed using an ANOVA with repeated measures for two factors, listener (native speaker or nonnative speaker) and speaker dialect. ESL listeners scored significantly lower when hearing the African American,

Indian, and Australian dialects of English than when hearing Standard American English; but there was no significant difference between the Southern American dialect and

Standard American English.

Bent and Bradlow (2003) studied the influence of the L1 on the intelligibility of

English L2 speech as determined by listeners from the same and from different L1 backgrounds. Listeners were asked to transcribe English sentences read by high- and low proficiency Chinese and Koreans, as well as Americans. There were four listening groups: Chinese, Korean, American English, and a mixed group of nonnative English speakers. Not surprisingly, American English listeners found American English talkers to be most intelligible. Surprisingly, nonnative listeners found high proficiency nonnative talkers to be as intelligible as native talkers. If an L1 were shared by both talker and listener, then even a low proficiency talker was found to be as intelligible as the high proficiency nonnative talkers and native talkers. These findings suggested a benefit to an ‘interlanguage match’ between talkers and listeners, but the benefit was only substantial for low proficiency talkers.

111

The above studies on listener variables show slight and inconsistent benefits to having a familiarity with the speaker’s accent. As Hazan and Markham (2004) note, in regard to native English speaking listener judgements, differences in listener variables have less impact on comprehension than the acoustic-phonetic features of the speech itself. Munro (2005, cited in Munro et al., 2006) confirmed that comprehension is affected by (1) stimulus properties or the acoustic-phonetic characteristics of the speech signal and (2) listener factors, such as language experience and accent familiarity.

According to Munro, Derwing, and Morton (2006), if stimulus properties had the greatest effect on listening comprehension, then we could expect a high degree of correlation among listeners, regardless of their individual listener factors. If the reverse were true, then there should be a range of variation according to listener differences.

To test this hypothesis, Munro, Derwing, and Morton (2006) investigated the intelligibility, comprehensibility, and accentedness judgements of Cantonese, Japanese, and Mandarin listeners for accented English speech spoken by Cantonese, Japanese,

Polish, and Spanish talkers. As with previous research, they found the benefits of interlanguage match between talkers and listeners to be inconsistent since only Japanese listeners showed benefit from an interlanguage match. However, there were strong correlations among intelligibility, comprehensibility, and accentedness judgements across all listeners, regardless of L1, L2 experience, and familiarity with an accent. The researchers concluded that while listener factors do impact on comprehension, the overwhelming agreement among listeners in this study indicated a greater impact from stimulus factors, or the features of the speech itself.

112

One listener factor that does have a consistent impact on intelligibility is oral proficiency level. For example, Van Wijngaarden, Steeneken, and Houtgast (2002) quantified speech intelligibility in noise for nonnative listeners in selected populations.

They used nine trilingual Dutch listeners with high proficiency in German and English, eleven Dutch listeners with low proficiency in German and English, and two control groups of native German and native American English speaking listeners. Overall, they found that the nonnative listeners needed a 1-7 dB better speech-to-noise ratio than did the native listeners in order to achieve 50% sentence intelligibility. They found a significant effect from the oral proficiency level of the listeners, which could be predicted by linguistic entropy estimates based on a letter guessing task.

2.6.7 Phonetic-Acoustic Factors that Affect Intelligibility

If the features of the stimuli or the phonetic-acoustic factors of L2 speech affect intelligibility more strongly than listener factors, then for teaching purposes, it is important to determine what specific features have the greatest effect. Pihko (1997) investigated the phonetic factors which contribute to the intelligibility of international

English in the context of English as a foreign language in Finland. Her subjects, Finnish and British secondary students, found familiar native English standards, such as British, easier to understand than unfamiliar native English standards, such as American, and nonnative varieties, including Finnish, German, and Gambian English. Her scoring for the partial dictation test had three levels, rather than two, and she did not statistically compare the two listening groups in terms of the populations that successfully transcribed the listening ‘gap statements.’ Pihko (1997) was one of the first researchers to test for

113

not only the intelligibility of various accented Englishes but also for the phonetic factors which led to the differences in intelligibility scores.

In response to the need for more spoken data of international English of a high enough recording quality for phonetic analyses, Deterding and Low (2001, 2005) compiled the National Institute of Education Corpus of Spoken Singapore English

(NIECSSE). Preliminary studies by Date (2005) and Kirkpatrick and Saunders (2005) reported on the intelligibility of Singapore English to a Japanese listener and to international and domestic students at an Australian university, respectively.

Date (2005), a native L1 speaker of Japanese and an EFL teacher for 25 years, suggested segmental and suprasegmental features of Singapore English that might negatively impact intelligibility for Japanese listeners. Potential segmental problems included three aspects of consonant variation and two aspects of vowel duration. The three consonant variations included substitution of dental /t/ and /d/ for the interdental fricatives /Ө/ and /δ/, simplification of consonant clusters especially in final position, and word-final elision of plosives. The two aspects of vowel duration included the shortening of vowels in CVC contexts and the monophthongalization of the front and back mid- vowels. However, the examples given to illustrate vowel durations seemed instead to illustrate vowel quality differences; so perhaps both spectral and temporal differences were present. Potential suprasegmental problems included an absence of nuclear accented syllables, incorrect sentence stress placement, failure to de-stress old information, narrow pitch range, high level tone (especially sentence-final), and falling tone on the tag question, “right?” (Date, 2005, pp. 178-179).

114

Kirkpatrick and Saunders (2005) presented 1-2 minute excerpts of Singapore

English speech to Australian and international undergraduates. One group of students heard the excerpts twice, while another heard them only once. Using Smith’s (1992)

60% intelligibility score as a benchmark, they determined that, not surprisingly, the first group scored higher, with only two listeners scoring less than 60%, one Australian and one Bhutanian. Even among those who heard the excerpts only once, over 50% of the listeners scored more than 80%. Five out of 34 listeners scored less than 60%. They were Norwegian, Iraqi, Japanese, Taiwanese, and Chinese. Based on their findings,

Kirkpatrick and Saunders (2005) suggested that the low scorers may not have been as familiar with Singapore English. On the other hand, their scores could have been due to a low level of English proficiency. In conclusion, the researchers observed that Singapore

English was over 60% intelligible to a majority of listeners, regardless of L1.

Deterding (2005) reported on the intelligibility of a familiar L1 English variety,

British RP, and an unfamiliar regional L1 variety, Estuary English. Singapore English listeners had no problem transcribing British RP, but encountered difficulties in transcribing several features of Estuary English. The phonetic aspects of Estuary English that were shared by Singapore English, such as vocalization of dark /l/ and glottalization of /t/ in word-final position did not impede intelligibility. However, glottalization of /t/ in medial position, th-fronting, and close back vowel fronting all caused problems in intelligibility. Finally, the fast speech rate of Estuary English also caused problems. The speech rate of 8.6 syllables per second was much faster than the normal range for English of 3.3 to 5.9 syllables per second, as reported by Roach (1998). In conclusion, Deterding

115

(2005) observed that while successful listening comprehension does not depend on the intelligibility of each and every word, in these data intelligibility problems based on the failure of bottom-up processing led to regular gaps and mistranscriptions of substantial sections of the Estuary English speech.

Deterding and Kirkpatrick (2006) described the pronunciation features of the

English of the Association of Southeast Asian Nations (ASEAN), which includes ten countries where English is used as a Lingua Franca. The researchers considered four of the ASEAN countries, Brunai, Malaysia, the Philippines, and Singapore, to be in the

Outer Circle of English usage, since English is an official or semi-official language.

They considered the other six countries, Cambodia, Indonesia, Laos, Myanmar, Thailand, and Vietnam, to be part of the Expanding Circle. Recordings of semi-informal conversations were analyzed for the non-standard pronunciation features that they shared and the extent to which each of these features impacted on intelligibility. Features that occurred in the speech of speakers from at least four different countries were defined as shared. The phonetic features shared among ASEAN speakers included /Ө/ — > /t/ dental fricative substitution, reduced aspiration of initial plosives, monophthongal /eI/ and /əυ/, the insertion of the semi-vowel /w/ in triphthongs, lack of reduced vowels, stressed pronouns, and heavy end-stress. These non-L1 English features frequently appeared in the speech data and did not appear to impact negatively on intelligibility. In five cases, non-shared pronunciation differences led to breakdowns in communication, and interestingly, all five of those features were included in Jenkins (2000) Lingua Franca

Core.

116

Field’s (2005) research focused on the impact of lexical stress on intelligibility, and he noted that other researchers had oversimplified the construct of prosody. He asserted that prosody included a constellation of elements, such as lexical stress, intonation, and the relative duration of weak and strong syllables, which may contribute to intelligibility in different ways. Accordingly, he structured his research to examine

Cutler and Clifton’s (1984) correct and incorrect primary stress placement of bisyllabic words, to which he added a condition in which the stress was shifted, but did not change the vowel quality.

Overall, nonnative listeners had lower levels of identification than did native listeners, but the patterns of impact of each of the conditions of lexical stress were strikingly similar for each listening group. Field (2005) found that lexical stress shifts accompanied by vowel quality change were more intelligible than lexical stress shifts with no accompanying vowel quality changes. The researcher speculated that this was due to the vowel quality changes consisting of weak forms of the vowel becoming full forms, and therefore more closely associated with the orthography. Lexical stress shifts to the right were found to be more problematic than lexical shifts to the left. Field (2005) speculated that this might have been due to the common practice in English of shifting stress to the left for purposes of contrast. For example, “I said INform them, not REform them” or “Instead of CONtaining the rioters, they DEtained them.” As a result, English

L1 listeners may have been more tolerant of stress shifts in that direction.

The effects of misplaced lexical stress on intelligibility for native listeners was

19.78% and for nonnative listeners, it was 21.28%. Since Cutler and Carter (1987)

117

estimated that 40.59% of all words in English conversation are polysyllabic of this type,

Field (2005) concluded that the threat to intelligibility from misplaced lexical stress was of concern in conversational speech situations, though not of the highest priority. The effects of context and purpose may be significant, however, to the role of prosody in comprehensibility.

2.6.8 Research and Teaching Implications of WE and ELF

World Englishes research and theory has argued for shifting the locus of power to determine intelligibility standards from the Inner Circle to a more equal sharing between the Inner and Outer Circles, as warranted by their sociolinguistic needs. Building on the theories of World Englishes, English as a Lingua Franca has focused on documenting the

Expanding Circle’s English usage, as well as on changing pronunciation and listening comprehension teaching priorities to reflect this new global context. Since intelligibility continues to be a central concern for international communication, it is important to identify the central core of pronunciation features necessary for English to function effectively in the international context.

The focus of WE has been on legitimizing nativized Englishes, which has limited the teaching applications of WE to some extent. If Outer Circle speakers of English are not considered to be language learners within WE theory (Kachru & Nelson, 1996), then there is nothing to teach. Consequently, in L2 English classrooms, educated speakers of

Outer Circle Englishes may be used as pronunciation teaching models of the ‘idiolect,’ but Outer Circle English speakers are not considered subject to “improvement” toward an

Inner Circle English. WE therefore supports nonnative English teachers as valid models,

118

who constitute the majority of English Language Teaching professionals worldwide

(Matsuda & Matsuda, 2001), and rejects the ‘native speaker fallacy” that native speakers of English are best qualified to teach, regardless of their training or experience

(Phillipson, 1992). In addition, courses exposing native English speakers to a variety of

World Englishes, in order to improve their listening comprehension, have been developed and implemented (Kubota, 2001).

In contrast to the WE perspective of Outer Circle English speakers, ELF regards

Expanding Circle English speakers as language learners; but at the same time, it refuses to accept Inner Circle English as the ultimate goal. Instead, ELF aims to improve intelligibility among Expanding Circle English speakers as determined by their own needs. This goal of mutual intelligibility has led ELF researchers and teachers to pose two basic questions for the teaching of pronunciation and listening comprehension: What is the core phonology, or the phonetic-acoustic features necessary for English to function intelligibly as a Lingua Franca, rather than in the context of the Inner Circle? And how do we teach pronunciation and listening comprehension so that ELF interlocutors can communicate effectively with one another?

Interest in establishing a core phonology for intelligible English communication in a global context has spanned several decades. Hockett (1958) proposed establishing a common core of English phonology based on the sounds shared by all the native dialects of English, since these were assumed to be mutually intelligible. Three decades later, in

“Teaching Pronunciation: The Common Core,” Jenner (1989) revisited the notion of a common phonology derived from native dialects of English specifically for the purposes

119

of teaching pronunciation. For those learners who “for one reason or another, do not want or need to sound entirely native-like,” Jenner proposed to “establish what all native speakers of all native varieties have in common . . . which would offer the learner a guarantee of intelligibility and acceptability anywhere in the world” (p. 2).

Jenkins (2000, 2002) did not just describe, but sought to prescribe, a unified core of phonological features, or the Lingua Franca Core (LFC), for English as a Lingua

Franca. She suggested that her approach differed from previous attempts to establish a core of mutually intelligible English phonological features in two important respects.

First, she contended that her LFC was based on the empirical data of authentic spoken interactions. Second, her LFC also accounted for “how people actually respond, so correcting the assumption that intelligibility is a function of relative frequency in naturally occurring speech” (Jenkins, 2000, p. 131). Jenkins (2002) then presented a syllabus for pronunciation teaching based on the LFC.

The purposes of WE and ELF research have been different, although they draw on similar theories. World Englishes argues in large part that nativized varieties of English, i.e., those in the Outer Circle, are as stable and legitimate as native varieties of English, i.e., those in the Inner Circle. World Englishes theory, therefore, rejects some basic concepts in SLA theory and research, such as interlanguage and language proficiency level, since both concepts assume a native variety of English as a reference point and an objective. In fact, the very notion of “L2 language learner” is suspect in World Englishes theory.

120

ELF, in contrast, has a more difficult theoretical case to make because, by definition, there is no single variety of English as a Lingua Franca to be legitimized. In the Expanding Circle, each individual L2 user of English will be influenced by the features of their L1 and their L2 developmental level – in short, the factors that affect second language learners’ speech production. Therefore, it is not yet clear whether any core phonological features will emerge from ELF interactions, or whether specific phonetic features will be consistently intelligible to a majority of ELF interlocutors.

Thus, ELF research is in its infancy, requiring description and analyses from corpora of

ELF interactions in a variety of social contexts and functional domains. Information from these data may indicate directions for further research on intelligibility, but description alone is not enough for drawing reliable conclusions regarding pronunciation teaching priorities.

Thus far, ELF research has mainly consisted of linguistic descriptions and case studies, which are limited in terms of the conclusions that can be drawn about the intelligibility of groups of speakers. Descriptive research is limited to the case of the interlocutors involved. Information gleaned from where communication broke down may certainly indicate phonetic-acoustic features which warrant further exploration.

However, descriptions of limited numbers of speech acts with few interlocutors representing few L1s and functional contexts are not generalizable to ELF. Descriptions of specific cases do not tell us whether the communication breakdowns the interlocutors experienced would occur at levels higher than chance with other interlocutors from the same L1s, from different L1s, or in different contexts or speech acts. The need for more

121

generalizable conclusions about ELF intelligibility requires further empirical studies using psycholinguistic methods which can indicate whether specific phonetic features consistently lead to intelligibility, or the lack thereof, for listeners with a variety of L1s in a variety of contexts.

2.7 The Literature Gap

Within the academic context, the US university determines the threshold of oral

English proficiency that is “good enough” for direct teaching contact with undergraduates who are assumed to be monolingual American English speakers. This notion of “good enough” for applications within a given context has been called “satisficing” by Simon

(1957) and has been applied in the fields of rational decision making and industrial engineering. The researcher extrapolates this concept to the model of oral English proficiency used in the current study, that of “graduate TA certification,” because it has been determined to be a satisficing model of English pronunciation, rather than the optimal model (i.e., Standard American English), for the purposes of US undergraduate teaching.

L2 phonological research traditionally focuses on individual learners. The social implications of speech production or perception mainly fall outside its scope. Although this research does an excellent job describing L2 language learners’ speech productions, it does not consider the context in which these productions are acceptable or unacceptable; nor does it consider the listeners for whom these productions are good enough or not, nor the factors that may affect these judgements.

122

Research on ITAs, WE, and ELF consider the social context of international

English communication. However, since within the academic context teachers are more accountable for the success of classroom communication than are students, ITA research has focused on the improvements that ITAs can make to facilitate their classroom discourse with American undergraduates. Since most US undergraduates continue to be monolingual English speakers, ITA research has not traditionally concerned itself with

L2-L2 English communication, nor with what native speakers can do to improve intercultural communication.

WE and ELF research do consider the factors that affect English communication among international interlocutors. WE, however, does not consider Outer Circle speakers to be second language learners of English, and therefore rejects any notion of improving interlanguage communication, except perhaps on the part of Inner Circle interlocutors.

ELF research, on the other hand, does consider international interlocutors from the

Expanding Circle to be second language learners of English. ELF researchers, however, reject the notion that Inner Circle English users set the standards for English pronunciation. They prefer instead that Expanding Circle users determine their own standards of intelligibility, based on their communications with one another. ELF research has only recently begun to consider the impact of context on standards for intelligibility. Most of this research remains descriptive in nature, which provides detailed data on the pronunciations that lead to communication breakdowns within particular cases. What has yet to be examined is to what extent these pronunciations lead

123

to communication breakdowns at levels greater than chance for larger numbers of speakers from specific L1s.

The current study builds on the findings from all of these areas of research:

Second Language Acquisition, International Teaching Assistants, World Englishes, and

English as a Lingua Franca. This study endeavors to account for the factors that have been found to affect intelligibility in L2 phonological acquisition research, including talker and listener factors, as well as the factors incorporated in the speech signal, such as text, talker task, and phonetic-acoustic factors. Unlike L2 phonological acquisition research, however, the current study situates its investigation in the academic context by selecting subjects who are all graduate students at a US university. Furthermore, conducting an intelligibility study, it takes these listeners’ feedback into consideration in order to prioritize the pronunciation features of Chinese graduate students to improve their intelligibility for Indian, Chinese, Korean, and American listeners.

Although all of the international participants have been certified as graduate TAs, the current study focuses on identifying their oral English communication needs in the graduate school rather than in the classes they teach as TAs. In other words, unlike ITA research, the current study considers both L1-L2 and L2-L2 English communication needs. Finally, unlike previous research in ELF, the current study includes native speakers in its analyses, since this is an appropriate reflection of the academic context of the US graduate school.

124

2.8 Summary

This chapter reviewed research to date in various areas relating to L2 speech intelligibility. Among these is L2 phonological acquisition and how native speakers of

English have been utilized for research purposes, as well as the extent to which they can learn to adapt to nonnative interlocutors. Researchers have also focused on the influence of the academic context on ITA communication. First, in ITAs’ roles as teachers, their speech has been investigated in terms of its comprehensibility for monolingual English undergraduates. Little research has examined international graduate students’ oral and aural English needs in the US graduate school, a context which includes both native and nonnative interlocutors. Finally, since US graduate schools are becoming increasingly international, research on World Englishes and English as a Lingua Franca was considered. Each of these areas represents a stance taken on the investigation of second language learners’ use of English.

SLA research into L2 phonological acquisition helps us understand to what extent adults can learn to perceive and produce new sounds. Researchers in linguistics, psychology, and second language acquisition (SLA) have investigated the acquisition of

L2 phonology in terms of both perception and production of L2 speech. Research from these perspectives has led to the refinement of language learning models, which in turn inform the teaching of L2 listening comprehension and pronunciation. These language learning models help identify which L2 sounds may be acquired at faster or slower rates.

Therefore, SLA studies can contribute to pronunciation teaching first by assuring teachers

125

that adults can in fact learn a new language, and second, by providing insight into which sounds may require the greatest effort and investment of time to learn.

In the context of international academic communication, researchers have investigated the oral English needs of international graduate students from two perspectives, based on their roles as teachers and students. First, the intelligibility of

ITAs at US universities has been explored from the perspective of monolingual English undergraduate students. Since pronunciation had been cited as the most salient problem, early research on ITAs had focused on segmental production, but more recently, the focus has shifted to issues of prosody and other discourse-level issues. Second, a few surveys have described the speaking and listening needs of international graduate students within their home departments.

With the global spread of English, researchers of WE and ELF have approached intelligibility and comprehensibility from the perspective of the needs of other international users of English. As research into L2-L2 English communication grows, more empirical evidence may help to identify the core phonology necessary for intelligible communication in different contexts of use and for different listeners.

The current intelligibility study contributes to our understanding of how native and nonnative listeners perceive L2 English speech. Since all aspects of a foreign accent do not necessarily impact negatively on intelligibility, this study provides empirical data on which sounds are most important to the comprehension of which listeners. Grounded in SLA theories of phonological acquisition, the intelligibility data from the current study will help pronunciation teachers determine the sounds that will most enhance

126

communication – for specific purposes and specific listeners. Therefore, intelligibility studies such as this one advance the field of applied linguistics by bridging the gap between the theories of second language phonological acquisition and the practical applications of pronunciation teaching.

127

Chapter 3: Methods

3.1 Introduction

As the literature review revealed regarding quantitative intelligibility studies, the word recognition in noise paradigm has frequently been employed in psycholinguistics, its early applications being to evaluate telephone transmission quality. More recently, this paradigm has been effective in investigating phonological effects on lexical access and processing. By measurably and consistently degrading the speech signal across stimuli, it allows researchers both to avoid ceiling effects common in presenting stimuli in quiet and to investigate effects on intelligibility of various signal-to-noise ratios. Since the context of the current study is the academic classroom, which includes the ambient noise of heating/cooling systems, traffic noise, and background conversation, the addition of noise to the stimuli increased the ‘face’ validity of the study for communication in this context. The research questions focused on the effects of the segmental pronunciation of

Chinese-accented English on intelligibility for international and American listeners.

Therefore, the researcher selected a quasi-experimental counterbalanced design and the psycholinguistic paradigm of word recognition in noise.

This chapter elaborates on the design and location of the study, the procedural components of the experiment, and the methods of analysis. First, the description of the

128

selected research design is presented along with the rationale for choosing it. Second, the international student population at the campus under study is compared to that in the US overall, in order to justify the selection of the study’s institutional location. Next, the various elements that comprise the experiment are described, beginning with the stimuli which incorporate the materials, talkers, and talker recording procedure. The following sections describe the process of listener selection and the listener variables collected in the language background questionnaire. The chapter concludes with the experiment procedure and the data collected, including the measures of intelligibility and talker accuracy, and finally, the methods of analysis employed.

3.2 Research Design

In social sciences research with human subjects, true experimental designs – in which the researcher has full control over the scheduling of experimental stimuli (e.g., when and to whom, including the ability to randomize exposures) – are often not feasible

(Campbell & Stanley, 1968). Therefore, a quasi-experimental, specifically a counterbalanced, design was selected. Counterbalancing is especially effective where pretests are inappropriate, by posing a potential testing threat to internal validity, and in which it is not feasible to make the comparison groups equivalent prior to treatment through the process of random assignment. All respondents are therefore exposed to all treatments, or stimuli, in a rotational order. In the case of the current study, all six talkers were rotated in presentation to all four L1 listener groups.

The stimulus order of presentation, or trial, was held constant, so that all listening subjects heard all 60 sentences in the same order. Since there were six talkers and 360

129

total sentences recorded, the talker of each sentence was rotated through the 60-sentence block presented to each listener. In this way, each listener heard each sentence only once, but with 18 listeners in each L1 group, each of the 60 sentence items by a single talker received three separate observations within each L1 group. As shown in Table 3.1, within each of the four listener L1 groups, the first subject heard the first six sentences spoken by the talkers in the following order: AE1, AE2, AE3, ME1, ME2, and ME3. The second subject in each L1 group heard the first six sentences beginning with talker AE2 and similarly, the third subject began with hearing talker AE3, and so on, through each of the six talkers. Each talker cycle was rotated a ten times throughout the 60 sentences.

Sentence Sentence Sentence Sentence Sentence Sentence Item 1 Item 2 Item 3 Item 4 Item 5 Item 6

Subject 1 AE1 AE2 AE3 ME1 ME2 ME3

Subject 2 AE2 AE3 ME1 ME2 ME3 AE1

Subject 3 AE3 ME1 ME2 ME3 AE1 AE2

Subject 4 ME1 ME2 ME3 AE1 AE2 AE3

Subject 5 ME2 ME3 AE1 AE2 AE3 ME1

Subject 6 ME3 AE1 AE2 AE3 ME1 ME2

Table 3.1: Sentence item by listening subject rotation

130

3.3 Location of the Study

The location for the present study was a large public US research institution in the

Midwest. This institution and campus were chosen not only due to the convenience and access to students available to the researcher, who is a graduate student of the institution, but also because the university meets three important criteria necessary for such a study.

First, it has a significant number of international students, being recognized as among the top fifteen hosting institutions in the US for international students. Second, its international students are concentrated in many of the same departments listed as popular areas of study for international students in the US. And finally, the university under study was chosen because its international students reflect the nationalities, and native languages, most common on all US campuses today.

International student enrollment in American universities is tracked by the

Institute of International Education (IIE). The IIE publishes an annual report on international students in the US, entitled Open Doors. In the 2010 Open Doors Report, which describes the previous year’s enrollment, the university under study ranked fourteenth in the US in overall international student enrollment. Table 3.2 below shows the top fifteen US institutions hosting international students (Institute of International

Education, 2009).

131

Rank Institution International Total % of Total Students Enrollment

1 University of Southern 7,482 33,747 22.2% California 2 New York University 6,761 50,917 13.3%

3 Columbia University 6,685 25,414 26.3%

4 University of Illinois – 6,570 41,495 15.8% Urbana-Champaign 5 Purdue University – 6,136 40,090 15.3% Main Campus 6 University of Michigan – 5,790 41,028 14.1% Ann Arbor 7 University of Texas at 5,703 50,006 11.4% Austin 8 University of California – 5,590 38,263 14.6% Los Angeles 9 Boston University 5,037 31,766 15.9%

10 Michigan State University 4,757 46,648 10.2%

11 Harvard University 4,735 20,307 23.3%

12 University of Florida 4,731 51,413 9.2%

13 University of Pennsylvania 4,635 24,107 19.2%

14 The Ohio State University 4,583 61,658 7.4% – Main Campus 15 Indiana University – 4,565 40,354 11.3% Bloomington

Table 3.2: Top 15 US institutions hosting international students, 2008–2009

132

In terms of fields of study, international students at the campus under study are similar to other international students in the United States. Additionally, international students gravitate toward a selected few departments, increasing their likelihood of interacting with one another, as opposed to interacting with American students alone.

According to fifteenth-day enrollment statistics for Autumn Quarter, 2009, international students represented a majority of the graduate students in three departments and/or colleges: Biological Sciences, Engineering, and Pharmacy. Moreover, international graduate students (IGS) comprised nearly half the enrollment in three other departments and/or colleges: Food, Agricultural, and Environmental Sciences (FAES); Mathematical and Physical Sciences; and Veterinary Medicine (Enrollment, 2009). This high enrollment of international students is not new, but has been true in these departments for the last decade, if not longer.

Table 3.3 below shows the total number of IGSs enrolled in the six departments mentioned above and the percentage of the total enrollment that IGSs have represented in each department over the past ten years (Enrollment, 2000, 2001, 2002, 2003, 2004,

2005, 2006, 2007, 2008, 2009).

133

Food, Math & Agricultural, Engineering Biological Physical & Pharmacy Veterina Sciences Sci. Environmental ry Sciences Medicin e 2009 815 230 289 181 39 37 56.4% 51.1% 41.3% 45.0% 57.4% 37.1% 765 171 301 182 47 42 2008 57.6% 46.1% 44.3% 45.4% 57.3% 37.5% 731 182 304 181 54 37 2007 57.1% 47.2% 45.4% 43.9% 55.7% 35.6% 670 208 311 151 52 46 2006 56.5% 46.0% 46.3% 45.6% 59.8% 39.3% 717 213 327 143 62 50 2005 59.6% 48.1% 47.4% 45.4% 66.0% 43.5% 809 218 353 152 61 60 2004 62.7% 49.0% 48.6% 46.1% 61.6% 47.2% 834 204 330 164 54 61 2003 63.1% 47.6% 49.3% 43.4% 54.5% 47.7% 867 197 304 165 49 59 2002 67.4% 46.4% 48.3% 45.0% 55.1% 46.1% 859 186 311 162 50 56 2001 70.1% 43.8% 49.6% 45.0% 52.6% 45.5% 800 170 320 138 38 49 2000 69.1% 40.5% 50.0% 42.3% 41.3% 48.0%

Table 3.3: OSU graduate student enrollment by department, 2000–2009

134

Overall in the Graduate School under study, international students comprised a quarter of the total enrollment in 2009 (2,563 out of 10,385 total, or 24.7%), and during the past decade their enrollment has been as high as one third of total enrollment.

The fields of study most favored by international students at this university are also among the most popular for international students in the US overall. Nationally, the top two most popular fields of study are Business & Management and Engineering, with

Physical & Life Sciences fourth, and Mathematics & Computer Sciences, sixth.

2007–08 2008–09 2008–09 International International % of Total % Change Enrollment Enrollment TOTAL 623,805 671,616 100% 7.7

Business & 110,906 138,565 20.6% 24.9 Management Engineering 96,133 118,980 17.7% 23.8

Misc. Liberal Arts 61,304 73,011 10.9% 19.1

Physical & Life 52,867 61,699 9.2% 16.7 Sciences Social Sciences 49,375 57,348 8.5% 16.1

Mathematics & 46,313 56,367 8.4% 21.7 Computer Science

Table 3.4: Fields of study of international students in the US, 2008–2009

135

The top six fields of study among international students in the US in 2008–2009 and their percent of the total international student enrollment are tabulated above in Table 3.4

(Institute of International Education, 2009).

Country of Origin 2007–08 2008–09 % of Total % Change

Total 623,805 671,616 100% 7.7 International 1. India 94,563 103,260 15.4% 9.2

2. China 81,127 98,235 14.6% 21.1

3. South Korea 69,124 75,065 11.2% 8.6

4. Canada 29,051 29,697 4.4% 2.2

5. Japan 33,974 29,264 4.4% -13.9

6. Taiwan 29,001 28,065 4.2% -3.2

7. Mexico 14,837 14,850 2.2% 0.1

8. Turkey 12,030 13,263 2.0% 10.2

9. Vietnam 8,769 12,823 1.9% 46.2

10. Saudi Arabia 9,873 12,661 1.9% 28.2

Table 3.5: Countries of origin of international students in the US, 2008–2009

136

Finally, the leading countries sending international students to the university under study are also among the leading countries sending international students to the US overall. The top ten countries of origin for international students in the US are shown in

Table 3.5 above for 2007–2009 (Institute of International Education, 2009).

OSU OSU Country of Origin Au 2008 Au 2009 % Change

1. China 977 1,411 44.4

2. South Korea 731 706 –3.4

3. India 718 700 –2.5

4. Taiwan 229 216 –5.7

5. Canada 113 106 –6.2

6. Turkey 92 92 0

7. Indonesia 86 72 –16.3

8. Japan 84 68 –19.0

9. Malaysia 41 50 22.0

10. Germany 40 43 7.5

Table 3.6: Countries of origin of international students at OSU, 2008–2009

137

In Table 3.6 above, the data for Autumn Quarter enrollments in 2008 and 2009 are also provided, as well as the percent change from 2008 to 2009, so that these data can be compared to the total US data (Enrollment, 2008, 2009).

As can be seen from the previous tables of data, the campus selected is an appropriately representative academic institution in which to conduct research regarding international students in the US. First, it ranks among the top fifteen universities in the

US for international student enrollment. Second, its international students are concentrated in many of the same departments as other international students in the US.

Third, where Indian, Chinese (mainland), and Korean students rank first, second, and third in the US for total overall enrollment, they also rank among the top three in enrollment at this university, where Chinese (mainland) students are first, Koreans are second, and Indians third. Therefore, Chinese, Korean, and Indian graduate students are likely to interact with one another in the US, as well as on the campus selected, using

English as their medium of communication.

3.4 Stimuli

The oral sentence stimuli in this experiment were created by recording Chinese and American talkers reading written materials in English. The sections below describe the written materials, the talker selection process, talker variables and speech rates, and finally, the recording procedure that was followed.

3.4.1 Materials

The materials were taken from the Bamford-Kowal-Bench Standard Sentence

Test (BKB-ST), revised by the Cochlear Corporation for use with Americans (BKB-R).

138

The original BKB-ST was designed to evaluate the speech intelligibility curves of British

English speakers, based on a database of the speech of partially hearing children aged 8-

15 (Bamford & Wilson, 1979; Bench & Bamford, 1979).

LIST NUMBER MEAN SCORE LIST NUMBER MEAN SCORE

(%) (%)

1 59 11 56

2 42 12 45

3 55 13 51

4 56 14 57

5 54 15 64

6 55 16 54

7 64 17 68

8 61 18 52

9 66 19 59

10 66 20 71

21 60

Table 3.7: Mean percent correct scores for each BKB-ST list

139

There are 21 BKB lists, with 16 declarative sentences containing 3-4 key words each, for a total of 50 key words per list. Samples of sentences which contained 3-4 key words included “The dog came back,” “The book tells a story,” and “They washed in

cold water.” The sentence lists use simple and balanced syntactic and morphological

structures and vocabulary that is highly familiar to native English speaking children. See

Appendix L for a complete list of the materials used.

In order to evaluate the equalization of the sentence lists, Bamford and Wilson

(1979) calculated the mean intelligibility scores of each list for normally hearing children

(N = 11) using the scoring method of percentage of key words correctly repeated. These

mean percent correct intelligibility scores by list are reported in Table 3.7 above. The

results of an ANOVA revealed that there were significant differences between the means

for each list (F20, 200 = 4.68, P < 0.001). Four lists (2, 12, 17, and 20) had means which

deviated by more than 9% from the grand mean of 57.89%, and should be treated with

caution (Bamford & Wilson, 1979).

The materials in the current study consisted of four consecutive sentence lists, 7-

10, which were selected because they had the most similar mean intelligibility scores out

of all 21 lists and they did not deviate from the grand mean by more than 9%.

Furthermore, these lists were employed in studies by Bent and Bradlow (2003) and Bent,

Bradlow, and Smith (2007), who also investigated the intelligibility of foreign-accented

speech.

Talkers were recorded reading 67 sentences, beginning with three sentence

practice items followed by 64 sentence test items. To ensure a balanced research design,

140

60 of the 64 test items were used as stimuli. The stimuli were equalized by taking the root-mean-square amplitude and equating it across all digital speech files. The digitized recordings were then mixed with white noise, yielding a speech-plus-noise file with a +5 dB signal-to-noise ratio. Each of the stimulus files consisted of a 400-millisecond silent leader, followed by 500 ms of noise, followed by the speech-plus-noise file, and ending with 500 ms of noise only. Five sentence practice items were also created, using the four excluded test items and one practice item.

Although the key words were reported to be highly familiar for the American listeners (Bradlow & Pisoni, 1999), this same level of word familiarity could not be assumed for the international listeners. Word familiarity for Americans was confirmed by consulting the Hoosier Mental Lexicon (HML), which includes ratings of word familiarity for native speakers on a seven-point scale, with 1 indicating the lowest degree of familiarity and 7 indicating the highest (Nusbaum, Pisoni, & Davis, 1984). The root

(e.g., “climb” for “climbed” or “box” for “boxes”) or exact forms of 133 out of 146 key words (91%) were listed in the HML with an average familiarity rating of 6.9% (range

6.5-7.0). Since this experiment used international as well as American listeners, immediately following the listening session, subjective word familiarity ratings of words that listeners indicated as having “seen or heard” before were elicited on a 5-point Likert scale. The variable of listener word familiarity was then included in the analysis model.

3.4.2 Talkers

Six talkers were selected among graduate students enrolled during Autumn

Quarter, 2005 with the following two L1s: (1) northern Mandarin and (2) General

141

American English. To ensure that no single talker’s accent was used to represent an L1 as a whole, three talkers were selected from each L1. In order to hold gender and age constant, the talkers were all male and all graduate students.

To identify American English talkers who could participate in this study, the researcher contacted the Registrar’s Office of the Graduate School and obtained a list of the currently enrolled male American students with in-state residency status. The researcher sorted this list by permanent residence zip code and eliminated the students whose zip codes were not included among the 47 zip codes associated with the capital city area. From the remaining students, the researcher selected a random sample of three students whom she contacted to request their participation. Once the frame was reduced to include only those students within one of the 47 zip codes of Columbus, Ohio, the total number of students was 590. A Texas Instruments TI-83 Plus calculator was used to generate random integers. On the Excel spreadsheet, students were listed between the numbers of 2 and 591, which were entered as the range in the calculator. Since not all the

American students contacted were originally from that state, another random selection was made until a total of three speakers from the state had agreed to participate, ensuring that all of them spoke General American English from a region with similar dialects.

To develop a list of potential northern Mandarin talkers for this study, the researcher contacted the Spoken English Program (SEP), which keeps a database of international graduate students who have achieved “graduate TA certification.” For this oral English proficiency level, the SEP provided the researcher a list of all the male

Chinese students who had been certified as TAs in the previous two years. The

142

researcher crosschecked this list against the list of currently enrolled male graduate students from the People’s Republic of China provided by the Registrar’s Office. A random sample of three students was taken from among those Chinese students who were both TA certified and enrolled Autumn Quarter, 2005. The researcher then contacted these students to request their participation. Since not all the Chinese students contacted were from northern mainland China, another random selection was made until a total of three Chinese TAs from northern mainland China agreed to participate.

Each talker recording session consisted of a language background questionnaire and a speaking session. Prior to each session, the talkers filled out the informed consent form and the language background questionnaire. Completing these forms and the speaking session took a total of about 30 minutes. After each recording session, the talkers were paid $20.00 for their participation.

3.4.2.1 Talker Variables

All talkers were male graduate students and the Chinese talkers had achieved the oral English proficiency level of “graduate TA certification.” The talkers all self- reported having normal hearing and no speech impediments. The variables of the

Chinese talkers collected in the language background questionnaires included city and province of origin, TOEFL and SPEAK scores, major of study, age at testing, age of first

L2 instruction (AFI), length of L2 instruction (LOI), age of arrival (AOA) in the US, length of residence (LOR) in the US, amount of L1 use in the last 12 months, amount of

L2 use in the last 12 months, familiarity with accented English, familiarity with Chinese-

143

accented English, and other foreign languages studied. Table 3.8 below illustrates these variables for each of the three Chinese talkers.

Interestingly, although all three talkers performed well on the paper-based

TOEFL, their oral proficiency levels upon arriving in the US were relatively low, except for ME3. Their average TOEFL score was 620 out of 677 (range 603-637), while their average SPEAK score was 146 out of 300 (range 120-200). This low correlation between the TOEFL and the SPEAK is one of many reasons the iBT TOEFL, which includes a speaking section, was created. These talkers represented the most popular fields of study

(math, science, and engineering), reported similar levels of usage for their L1s and L2s, and similar levels of familiarity with foreign-accented and Chinese-accented English.

They reported similar background characteristics, such as age of first formal English instruction (mean = 12 years old), length of formal English instruction (13.8 years), age of arrival (23.6 years old), and length of residence (2 years). None reported being proficient in any foreign language other than English.

144

ME1 ME2 ME3

City, Province of Origin Yinchuan, Gaizhou, Lankao, Ningxia Liaoning Henan TOEFL 620/677 603/677 637/677

SPEAK 120/300 120/300 200/300

Major Organic Chemical Math Chemistry Engineering Age at Testing 27 27 24

AFI 12 yrs old 13 yrs old 11 yrs old

LOI 15 yrs 14 yrs 12.5 yrs

AOA 25 yrs old 24 yrs old 22 yrs old

LOR 2 yrs 3 yrs 1.5 yrs

L1 Usage 10 hrs/wk 20-30 hrs/wk 24 hrs/wk Last 12 Months L2 Usage 10 hrs/wk 4 hrs/wk 10 hr/wk Last 12 Months Familiarity with Accented English 2 hrs/wk 1 hr/wk 3 hrs/wk

Familiarity with Chinese-Accented 2 hrs/wk 2 hrs/wk 5 hrs/wk English Other Foreign Languages Studied N/A N/A Japanese, French

Table 3.8: Chinese talker variables

145

Table 3.9 below reports on the variables of the American talkers collected in the language background questionnaires. They include their city and state of origin, major of study, age at testing (mean = 31), familiarity with foreign- (5.6 hrs/wk) and Chinese- accented English (2 hrs/wk), foreign languages studied, time spent abroad in the last 12 months, total time spent abroad, and places visited abroad. The American talkers reported no significant experience abroad or with foreign languages.

AE1 AE2 AE3

City, State of Origin Cleveland, Ohio Jackson, Lakewood and Ohio Dayton, Ohio Major Statistics Occupational Architecture Therapy Age at Testing 28 30 35

Familiarity with Accented 6 hrs/wk 10 hrs/wk 1 hr/wk English Familiarity with Chinese- 5 hrs/wk 1 hr/wk 5 min/wk Accented English Foreign Languages N/A N/A French, Russian Studied Time Spent Abroad Last 0 0 0 12 Months Total Time Spent Abroad 0 2 wks 1 mo

Places Visited Abroad N/A Jamaica France, Canada

Table 3.9: American talker variables

146

3.4.2.2 Talker Speech Rates

Since speech rate was not controlled, a one-way ANOVA was conducted to determine the effect of talker on speech rate. Pairwise comparisons with Tukey 95% confidence intervals revealed a significant difference in speech rates between the native and nonnative talkers. Figure 3.1 below shows the dotplot of talker by speech rate in words per second (WPS).

Dotplot of Speech rate WPS vs talkers

AE1

AE2

AE3 talkers ME1

ME2

ME3 2.1 2.8 3.5 4.2 4.9 5.6 6.3 Speech rate WPS

Figure 3.1: Talker speech rate analyses measured in words per second (WPS)

147

The pairwise Tukey 95% confidence intervals around the differences in mean speech rates revealed that American talkers AE1 and AE2 had a mixed confidence interval, indicating that their speech rates were not significantly different (-0.420, 0.073).

American talker AE3 was significantly slower than AE1 and AE2, as shown by the negative confidence intervals with each [AE1 subtracted from AE3 (-1.422, -0.928); AE2 subtracted from AE3 (-1.248, -0.755)]. However, while he was the slowest speaker among the native talkers, AE3 was still significantly faster than all of the Mandarin talkers [AE3 subtracted from ME1 (-0.6941, -0.1991); AE3 subtracted from ME2

(-0.5057, -0.0126); AE3 subtracted from ME3 (-0.7234, -0.2303)], as were the other two

American talkers [AE1 subtracted from ME1 (-1.869, -1.374); AE1 subtracted from ME2

(-1.681, -1.188); AE1 subtracted from ME3 (-1.898, -1.405); AE2 subtracted from ME1

(-1.696, -1.201); AE2 subtracted from ME2 (-1.507, -1.014); AE2 subtracted from ME3

(-1.725, -1.232)]. The Mandarin talkers, in contrast, were not found to have significantly different speech rates from one another, with all mixed confidence intervals [ME1 subtracted from ME2 (-0.0600, 0.4350); ME1 subtracted from ME3 (-0.2778, 0.2173);

ME2 subtracted from ME3 (-0.4643, 0.02883)].

3.4.2.3 Talker Segmental Accuracy

Besides talker L1 and speech rate, talker pronunciation accuracy was also tabulated. Since the reference model of English for the context of the US university is

American English, for the purposes of this study, the American talkers’ segmental pronunciations of each key word were labeled as “accurate.” Each Chinese talker

148

recorded 60 sentences containing a total of 186 key words. Therefore, for these three talkers, a total of 558 key words were rated for accuracy on a binary scale by two raters.

The first rater was the researcher, who was an instructor in the ITA program for six years and is a native speaker of American English, with additional background in

Spanish, Japanese, and Italian. She labeled key words as either “accurate” or “not accurate” as compared to a canonical phonemic transcription in American English, given by the Carnegie Mellon University (CMU) Pronouncing Dictionary. If each of the phones (vowel and consonant sounds) in the talkers’ pronunciation matched all the phones in the canonical American English pronunciation, the researcher labeled the key word as “accurate.”

The second rater was an undergraduate research assistant in the Honors program majoring in Linguistics and Psychology and a native speaker of American English, with additional background in French and British English. She also labeled the accuracy of the key words in the 180 Chinese-accented sentences. An American undergraduate was selected as the second rater because the objective of the ITA program is to prepare ITAs to be successful interlocutors primarily with American undergraduates.

Talker accuracy ratings from both raters were collated into an Excel spreadsheet and a macro was written to compare them. The key word was labeled “not accurate” if either rater had entered a “0” for that word. Next, an R code was used to calculate inter- rater reliability as well as the overall percent correct means of key word accuracy by talker. Overall inter-rater reliability was 75% with a range of 69-84% across talkers.

149

Rater agreement was lowest for ME2 and highest for talker ME3, so that only the agreement rate for ME1 (73%) was comparable to the overall agreement rate (75%).

Talker: ME1 ME2 ME3

Rater 1 55.91% 58.60% 78.49% (Researcher) Rater 2 58.60% 71.5% 75.27% (Research Assistant) Aggregate Ratings 43.55% 49.46% 68.82%

Table 3.10: Mean percent correct talker key word accuracy ratings

As can be seen in Table 3.10 above, the mean percent key word accuracy ratings

of each talker varied by rater as follows. Although the mean accuracy ratings for ME1

(Rater 1 = 55.91%; Rater 2 = 58.60%) and ME3 (Rater 1 = 78.49%; Rater 2 = 75.27%)

were similar between the two raters, there was a notable difference in the mean accuracy

ratings for ME2 (Rater 1 = 58.60%; Rater 2 = 71.5%). Furthermore, although the mean

percent accuracy rates of the two raters were similar for ME1, the aggregate mean was

lower than these, due to some disagreement between the raters as to which words ME1

produced accurately.

150

Due to the relatively low inter-rater reliability overall, the researcher selected a conservative aggregate rating scale, which logged a key word as “not accurate” if either of the raters had labeled it that way. This rating scale acknowledged that at least one of the native English-speaking raters had heard some sound that did not agree with the canonical American English phonemic transcription. Following the collection of the intelligibility data, the key words that listeners did not accurately transcribe were analyzed further. The key words of each Chinese talker which were misidentified by

75% or more of all listeners were phonemically transcribed by the researcher and by a veteran professor of the ITA program. These key word transcriptions were subjected to segmental error analyses as well as categorized by the listeners’ L1s and are detailed later in the data analysis section.

3.4.3 Talker Recording Procedure

Recording was done during the 2005-06 academic year in a sound-attenuated booth in the Linguistics Department, using a Shure SM10-A head-mounted microphone.

The microphone was connected to a Symetrix SX302 Dual Microphone Pre-amplifier

(gain ~ 50 dB). The talkers’ utterances were digitally recorded using an Audigy 2 sound blaster card. The researcher used Cool Edit (Syntrillium Co.) software to save the sound files (Microsoft PCM.wav) on the computer hard drive, using a 22,050 kHz sampling rate with 16-bit resolution. These recordings were then archived onto CDs.

Each talker recorded the sentence lists during a 30-minute session with the researcher. Each sentence was read three times but only one of these was included in the experimental set of stimuli. The sentences chosen for the stimuli were generally the

151

third, or final iteration. The researcher found that although the American English talkers were told to speak the sentences as if they stood alone, in fact many of the first and second iterations of the sentence exhibited rising intonation in sentence final position as you would find in a list or series of tokens. Therefore, the final iteration was used for the stimuli to preserve falling sentence final intonation. However, if there were obvious disfluencies, or if the talker only repeated the sentence twice, then the second iteration was used. After the English sentences were recorded, the northern Mandarin talkers also recorded the same type and number of sentences, but this time from materials translated into Mandarin Chinese. These L1 recordings were made for the purpose of further analyses to be conducted post-doctorally, which cannot be reported here.

The talkers’ speech rate was not strictly controlled; however, the researcher provided a model of the practice sentences at a medium speech rate for the talkers to follow. In the literature, a medium speech rate was identified as ranging between 465 and 575 ms per word (Bradlow & Pisoni, 1999). Following Ortega-Llebaría (1997), word familiarity was ensured by first presenting talkers with the key word list, and then asking them to circle any words that were unfamiliar to them. They were then provided with a picture, definition, and oral pronunciation of the unfamiliar words they had identified.

3.5 Experiment Paradigm

In order to examine the effect of phonological production on intelligibility, the psycholinguistic experimental paradigm of word recognition in noise was selected. This paradigm has been used extensively, beginning in 1910 with telephone transmissions

152

(Campbell, 1910), to measure intelligibility. More recently, it has been used successfully to measure effects on the intelligibility of nonnative speakers of English (Lane, 1963;

Bent & Bradlow, 2003; Bent, Bradlow, & Smith, 2007).

There are very few disadvantages associated with this paradigm. Some of the potential artifacts are that the intelligibility score as measured by percent correct recognition includes both bottom-up processing based on the acoustic-phonetic information in the signal and top-down processing based on the listener’s lexical knowledge. In addition, white noise masks consonants more than vowels and intelligibility can vary by talker and key word. The advantages to this paradigm are that it is easy to use, and it allows the experimenter to control the acoustic-phonetic information and the degradation levels of the signal. It also allows investigation into the underlying perceptual processes used to recognize words from degraded or partial information, and allows several different dependent variables to provide converging evidence on the processes of word recognition and lexical access (Pisoni, 1996). Finally, white noise mimics some of the environmental noise typically encountered in academic settings such as the classroom.

3.5.1 Pilot Study

A pilot study was conducted in Summer Quarter, 2008 using the stimuli described above with one Chinese, one Korean, and two American listeners selected from among graduate student volunteers in the Linguistics Department. These pilot subjects provided feedback on the clarity of the language background questionnaire and of the researcher’s instructions and efficacy of the data logging settings of the E-Prime program (Psychology

153

Software Tools, Inc.). In response to this feedback, changes were made to the language background questionnaire, the procedural instructions read by the researcher, the instructions displayed by the E-Prime program, and the experiment’s data logging settings.

On the language background questionnaire, several new questions were added to elicit greater detail from subjects. First, a time grid (0-24 hrs/wk) was added from which subjects could choose, which would describe their L1 and L2 usage. Second, more options describing the communicative purposes and contexts in which they listened to and spoke English were offered. Next, the instructions read to each subject by the researcher included information that there would be no audible or written prompts between the time that they heard the sentence and the time that they chose to repeat it.

Finally, subjects were instructed to repeat the sentence, or any of its words and sounds, as soon as they had heard and understood the sentence to the best of their ability.

In the E-Prime program, changes were made both to the OralRep procedure block and the word familiarity (WordFam) rating procedure block. In the OralRep block, the data logging settings were changed from “Time Audit Only” to a customized set of data including RT, RT Time, Onset Time, Finish Time, Offset Time, and Start Time

(Schneider, 2002, p. 30). In the WordFam block, instead of a finite response time of two minutes, subjects were allowed infinite time to rate their familiarity with the words they had identified as “seen or heard.”

154

3.5.2 Listener Subject Selection

A total of 72 listening subjects was selected among graduate students enrolled during the 2008–2010 academic years at the main campus of a large, public Midwestern university from the following L1 groups: Hindi (n = 18), northern Mandarin (n = 18),

Korean (n = 18), and General American English (n = 18). The international students who were selected as listeners had achieved the same oral English proficiency level as the northern Mandarin talkers, that of “graduate TA certification,” which is defined as a

SPEAK test score of 230 or higher out of 300 or as an “unconditional pass” on the Mock

Teaching Test (MTT), which is a locally developed teaching performance test.

To identify American English listeners who could participate in this study, the researcher contacted the Registrar’s Office of the Graduate School and obtained a list of the American students enrolled in Autumn 2008 who had in-state residency status. The researcher then sorted this list by permanent residence zip codes and eliminated the students whose zip codes fell outside the 47 zip codes associated with the city of the main campus of the university. The talkers were also excluded from participating as listeners.

The population frame from which the random sample was taken could not be entirely cleaned, or limited to contain only General American English speakers, because this type of detailed linguistic data does not exist.

Therefore, the researcher selected a random sample of 200 students, 100 male and

100 female, whom she contacted by mail and email to request their participation, if they identified themselves as natives of Ohio who learned English as a first language, spoke it at home, and who had lived in the state between the ages of 2 and 10. A Texas

155

Instruments TI-83 Plus calculator was used to generate random integers. On the Excel spreadsheet, female students were listed between the numbers 2 and 1,540 and male students were listed between the numbers 1,541 and 2,567, which were entered as the ranges in the calculator for the two random integer generations. The 200 male and female students whose names corresponded to the 200 random integers generated were contacted via email for participation.

To develop a list of potential international listeners for this study, the researcher contacted the Spoken English Program (SEP), which keeps a database of international graduate students whom they have tested and/or certified as TAs. For the “graduate TA certification” proficiency level, the SEP provided the researcher with a list of graduate students certified as TAs between 2006 and 2009 who identified their native languages as

Hindi, Mandarin/Chinese, or Korean. The researcher then crosschecked this list against the list of currently enrolled graduate students from India, China, and Korea provided by the Registrar’s Office, eliminating those not currently enrolled. Letters were sent to a census of these international TAs requesting their participation if they identified themselves as native speakers of Korean, of Hindi from northern India, or of Mandarin from northern mainland China. Recruitment emails were also posted to the listservs of the Chinese Student and Scholars Association and the Korean Student Association by members of each of these organizations.

The subjects who satisfied the regional dialect and native language criteria of the experiment were scheduled on a first come, first serve basis. Data from subjects who identified themselves as not having normal hearing were eliminated. All subjects

156

completed an informed consent form, a language background questionnaire, a listening session, and a recording session. American listeners completed all sessions in approximately one hour. Due to the longer recording sessions in both their L1 and L2, the international listeners finished in approximately one hour and 30 minutes. All subjects were paid $20.00 for their participation.

3.5.3 Main Experiment Procedure

During the 2008–2010 academic years, listening subjects were tested one at a time in a sound-attenuated booth in the Psycholinguistics Laboratory in the Linguistics

Department. All 72 subjects were tested by the researcher, except during Winter Quarter

2009, when 15 subjects were tested by an undergraduate research assistant majoring in

Linguistics and Spanish, who was paid for her services. Both researchers used the same instruction sheet to read to subjects and followed the same procedural checklist to complete the experiment. The experiment consisted of three main parts: 1) Informed consent and language background questionnaire, 2) Listening session, and 3) Speaking session. In the first part, each participant was asked to complete an informed consent form and a language background questionnaire, which took a total of 15-20 minutes.

From the review of the literature, the following factors were found to affect intelligibility or listening comprehension: gender, current age, age of first instruction in the second-language (AFI), L1 model(s) of English studied, current L2 proficiency level, age of arrival (AOA) in an L2-speaking country, length of residence (LOR) in an L2- speaking country, amount of L1 and L2 usage, familiarity with a specific accent in

English, and familiarity with foreign-accented English in general. Therefore, the

157

participants were asked to report in the language background questionnaire on these factors.

The listening tasks were programmed with E-Prime psycholinguistic software, distributed by Psychology Software Tools, Inc. (PST). The oral repetition, written transcription, and word familiarity rating tasks were elicited in two procedure blocks:

OralRep and WordFam. In the OralRep block, listeners were told that all of the utterances they would hear would be real English sentences, spoken by Chinese and

American graduate students. Prior to each new OralRep trial, a black “+” appeared on the computer screen for 750 ms. This indicator was followed by a 500 ms pause and then the stimulus sentence was presented orally via headphones (Sony MDR-V600) at a comfortable listening level (65 dB SPL). A 500 ms pause also followed each trial. The listeners were instructed to repeat the sentence they heard as soon as they could identify it. If they did not understand the entire sentence, they were encouraged to repeat any words or sounds they did understand. Speed and accuracy were equally emphasized.

Two microphones on table stands, positioned 3-5 inches from the subject, were used to collect the oral responses. One microphone (Shure SM-48) was connected to a voice-activated timer in a PST Serial Response Box, which was used to measure response latencies, or reaction times (RTs). The second microphone (RadioShack Highball) was connected to a Symetrix SX302 Dual Microphone Pre-amplifier (gain ~ 50 dB) and then to a Marantz portable digital flash recorder (PMD670). Oral responses were first recorded digitally to a memory card and then transferred to a CD. After orally repeating each sentence, the listeners were instructed to type what they had just said into the

158

computer keyboard. The typed response was then presented on the computer screen to the subject, who was asked to confirm if the response was correct. The subject was instructed to press “Y” if yes, and was progressed to the next trial. If no, s/he was instructed to press “N” and was taken to a screen which allowed corrections and retyping of responses which were again presented for approval.

The second part of the listening session of the experiment consisted of a word familiarity rating task. The list of 146 unique key words, which were rated as highly familiar by native listeners, was combined with a total of 50 filler words of low (25) and medium (25) familiarity ratings according to native listeners (Lewellen et al., 1993), so that the full range of the familiarity scale would be reflected in the test items. The low and medium familiarity rated words were chosen from the data base compiled by Frisch

(2007), consisting of 200 words from the Hoosier Mental Lexicon (Nusbaum et al.,

1984). In Frisch (2007), these 200 words were rated by American undergraduates on average between 1.2 and 4.5 on a 7-point Likert scale. After sorting the word list by mean familiarity, the first 25 words and the last 25 words were chosen as fillers for the current word familiarity rating task. The only exception was the word “altho,” which was eliminated from the last 25 words because it did not appear in the Oxford English

Dictionary and the researcher thought that subjects might be confused, thinking it was a misspelling of the word “although.” The 25 low familiarity words chosen ranged from a mean of 1.2 to 1.5 on a 7-point scale. The 25 medium familiarity words chosen ranged from a mean of 3.5 to 4.5 on a 7-point scale.

159

Each of the 196 target words was written in all capital letters and presented to the listeners in random order on the computer screen. Each listener was asked whether s/he had ever seen or heard the word. If the listener entered “NO” (= 1) on the 5-button PST

SRBox, s/he progressed to the next word. If the listener entered “YES” (= 5), s/he progressed to a rating screen. At the rating screen, the subject was asked to rate the familiarity of the word on a scale of 1 to 5, where the end labels were 1 = “I very seldom see or hear this word” and 5 = “I very often see or hear this word.” These rating instructions and end label definitions were written on each screen above each word presented. With this method, listeners would not be asked to rate their familiarity with words that were unknown to them. Instead, each subject entered a familiarity rating only for each “known” word, which was clearly defined as one which had been “seen or heard.” As soon as a response was entered, the next trial was presented. The total time spent for each listening session ranged between 45 minutes to one hour, with 30-40 minutes spent on the oral repetition and typed transcription tasks and 15-20 minutes on the word familiarity rating test, including practice sessions prior to each of the two procedure blocks.

For the speaking session, each subject wore a Shure SM10-A head-mounted microphone, which was adjusted so that the microphone extended one inch from the left corner of the subject’s mouth. The microphone was connected to a Symetrix SX302

Dual Microphone Pre-amplifier (gain ~ 50 dB). The subjects read sentence lists 7, 8, 9, and 10 (totaling 64 sentences) from the BKB-R Standard Sentence Lists (Bamford &

Wilson, 1979; Bench & Bamford, 1979). They had previously heard 60 of the 64

160

sentences as stimuli during the listening session. In addition, the international subjects read translations of these materials in their L1s (Hindi, Mandarin, and Korean). The researcher digitally recorded all sentences with an Audigy 2 sound blaster card and used

Cool Edit (Syntrillium Co) software to create the sound files (Microsoft PCM.wav), with a 22,050 kHz sampling rate and 16-bit resolution. These were first saved to the computer hard drive and then to a secure server, and finally to an external hard drive. These recordings constitute the Buckeye GTA Corpus, a full analysis of which is beyond the scope of this dissertation study.

3.6 Data Collected

3.6.1 Listener Variables

In the language background questionnaire (LBQ) for the international listeners, the following variables were collected: age at testing, gender, familiarity with Chinese- accented English, familiarity with foreign-accented English, age L2 learning began or age of first instruction (AFI), L1 model(s) of English studied, age of arrival (AOA) in the US, length of residence (LOR) in the US, amount of L1 use, amount of L2 (L3, etc.) use,

SPEAK score, date of “TA certification,” and TOEFL score.

In the language background questionnaire for the American listeners, the following factors were covered: age at testing, gender, familiarity with Chinese-accented

English, familiarity with foreign-accented English, foreign languages studied, time spent abroad in the last 12 months, and total time spent abroad, including places visited.

161

3.6.2 Intelligibility Measure

For each trial of the oral sentence repetition and written transcription task, three responses were collected: orally repeated sentence, RT, and typed sentence, and only the last of these was analyzed. The typed sentence responses were saved on the laboratory booth computer (3.4 Ghz Pentium 4 with 512 MB of RAM) and then saved to an external hard drive and archived on CDs. The intelligibility scores were calculated using the second of these responses, the typed sentence. An R code was written to compare the key words in each listener’s typed response sentence to the target key words in the materials.

Key words were logged as “intelligible” if all characters in the response exactly matched those of the target key word.

One of the researcher’s assumptions was that participants had sufficient proficiency in English to successfully complete the tasks in the experiment. Furthermore, within the experiment procedure, after subjects typed their response, they were presented with it on the computer screen and asked if it were correct, which gave them the opportunity to proofread every sentence they had typed and rewrite it if needed. Any remaining typographical or spelling errors of key words were logged as “0” to indicate that they had not been accurately transcribed.

3.6.3 Word Familiarity Ratings

Although the key words for the BKB-R Standard Sentence Lists have already been rated as “highly familiar” to native speakers of English, word familiarity ratings were collected for each key word from each listener because this experiment used nonnative speakers of English as participants. For the key words that the listener

162

indicated s/he had seen or heard, a word familiarity rating was collected on a five-point

Likert scale, using a 5-button PST SRBox. Familiarity was rated on an increasing scale from 1 to 5, where 1 = “I very seldom see or hear this word” and 5 = “I very often see or hear this word.” The word familiarity ratings for the filler words were excluded from the analyses of the key words. The listener key word familiarity ratings were included in the data analysis model as an independent variable.

3.7 Data Analysis

A mixed-effects model with logistic regression was selected to analyze the factors affecting intelligibility. In a mixed-effects model, the treatment factors, or the repeatable factors, are represented with fixed-effects terms. The independent variables with levels randomly sampled from a larger population are represented with random-effects terms.

In the current study, the talker L1 (American English or Mandarin) and talker segmental accuracy, as well as the listener L1 (Mandarin , Korean, Hindi, and American English) and listener word familiarity ratings were treated in the model as fixed-effects terms. The random-effects terms captured individual talker, listener, and key word variation.

To provide more detail than previous intelligibility studies, further analyses were conducted on the key words which resulted in inaccurate listener transcriptions. The

Chinese-accented words which at least 75% of the listeners misidentified were transcribed and analyzed by their component segments in order to identify segmental error patterns which led to breakdowns in intelligibility. Each segment or phone in the key word was labeled as a consonant or a vowel and then the talker production error was identified and correlated to the segments involved. If a consonant led to the breakdown

163

in intelligibility, its position in the syllable and word (initial or final) was noted. If a vowel led to the breakdown in intelligibility, its phonetic environment – the phone immediately following the vowel – was also noted. Next, all key words that were inaccurately transcribed were sorted by listener L1, so that the patterns of vowel and consonant errors that caused the most problems for 50% or more of listeners in each L1 could also be identified.

3.8 Limitations

This study is limited by the institutional location, talker variables, laboratory setting, methods, and listener variables involved. The findings cannot be generalized beyond the US university under study because the sample was selected from graduate students of this institution and therefore the findings apply only to this population. In addition, the population studied may differ from other university populations in many ways – for instance, in its student body’s familiarity with either foreign-accented or

Chinese-accented English. The institutional location of the study included a student body of more than 50,000 and an international student body of more than 4,500, which would influence participant familiarity with foreign-accented English in general, and Chinese- accented English in particular.

The findings of this research were also limited by several talker factors, such as major of study, L1 and regional dialect, educational attainment level, age of first instruction (AFI), age of arrival (AOA) in the US, age at testing, length of residence

(LOR) in the US, L1 usage in the last 12 months, L2 usage in the last 12 months, oral

English proficiency level, and gender. Since major was not held constant, the findings

164

cannot account for any variations in talker pronunciation that may be influenced by majors of study. The results of the study reflect only the intelligibility of northern mainland Mandarin-accented English and the regional dialects represented by the three

Chinese talkers included in this study. Therefore, the findings cannot be generalized to other Mandarin languages or regional dialects in the population of graduate students in the campus under study.

In addition, the talkers all reflected high levels of educational attainment; they were all college graduates seeking advanced degrees in their chosen disciplines.

Therefore, the intelligibility of these talkers is not representative of Mandarin speakers with significantly different educational backgrounds. The Chinese talkers had an average

age of 12 (with a range of 11-13) when they began their first formal English instruction

(AFI). Therefore, the findings of this study cannot be generalized to Mandarin speakers

who began to study English at a much younger or much older age. The age of arrival in

the US and age at testing were not held constant to a single year and instead reflected an

age range (24-27) commonly found among graduate students. Therefore, the findings

cannot be generalized to nontraditional graduate students with substantially different ages

of arrival and ages at testing. The Chinese talkers had also been living in the US for

similar lengths of time (1.5-3 years), that is, the duration of their graduate studies to date.

As members of the same Chinese student community, living in the same city, and using

English primarily for their academic programs, they reported similar levels of L1 usage

(10-24 hrs/wk) and L2 usage (4-10 hrs/wk) in the last 12 months. The findings of this

165

study therefore cannot apply to other Chinese residents living in the US for different lengths of time and with different levels of L1 or L2 usage.

This study’s findings are also limited to Chinese speakers of English at the oral

English proficiency level of “graduate TA certification” because it examined a sample of a population of speakers that has been under-investigated to date. Therefore, the findings cannot be extended to include other graduate students who have not yet been certified to teach, or who have not demonstrated the equivalent of this proficiency level on any other measure. Since the existing literature revealed mixed results in terms of the effect of gender on intelligibility and listening comprehension, the current study held this factor constant by using only male talkers. As a result, the findings of the study are limited to male speakers.

This study was conducted in a laboratory, rather than a naturalistic setting, because the researcher sought to exclude other factors – both linguistic and extra- linguistic – which can affect communication. In order to focus on the effect of segmental, i.e., vowel and consonant, pronunciation on intelligibility, linguistic factors such as syntax, morphology, and lexicon were all held constant. The recorded stimuli allowed the researcher also to exclude extra-linguistic factors such as gestures, eye contact, facial expressions, or lip movement. Finally, topic familiarity was held constant by using simple sentences with highly familiar vocabulary. The multitude of variables at play in a naturalistic setting make conclusions about the role of a single variable impossible. In order to establish a baseline of the impact of segmental pronunciation on

166

intelligibility, the laboratory setting was chosen as most appropriate for answering the research questions, and the findings are therefore limited to this setting.

The methods used in this study, including the selected stimuli and task both limit the findings, as well. The stimuli were recordings of read, scripted sentences. In order to avoid the ceiling effect common with stimuli presented in quiet laboratory settings and to approximate the ambient noise levels of academic classrooms, the sentence stimuli were degraded with white noise at a +5 dB signal-to-noise ratio and presented at a comfortable listening level of 65 dB. Although key word transcription accuracy was used to calculate intelligibility, the key words were embedded in carrier sentences, which limited this research to sentence intelligibility rather than word or passage intelligibility. Word intelligibility was not used because of the greater ‘face’ validity of sentences for measuring human communication. Passage intelligibility was not selected because in longer passages of text, intelligibility can be affected by topic familiarity, greater contextual information, and prosodic factors such as intonation, rhythm, and stress, which were not within the scope of the study. The task listeners were asked to perform was a written transcription of the oral stimuli, as typed on a computer keyboard. As previous research has shown, listeners perform differently in different tasks, so the intelligibility findings of this research are limited to the written transcription of oral stimuli.

The findings cannot be generalized to listeners with significantly different educational backgrounds. In order to focus on the listener factor of L1, other listener factors such as educational level were held constant. All listeners were graduate students, so they were similar in educational attainment level. They had all completed

167

undergraduate degrees, and were currently pursuing graduate degrees at either the master’s or doctoral level.

To the greatest extent possible on the campus under study, and in order to be able to recruit sufficient numbers of subjects, the listeners represented as little regional dialect variation as possible in their L1s. The L1s and regional dialects represented by the listeners who participated in this study were Hindi from northern India, Mandarin from northern mainland China, Korean, and American English from Ohio, and the findings are limited to these.

3.9 Summary

This chapter elaborated on the research design, location, stimuli, methods, data collected, and analysis methods used in the current study. Since the use of a pretest and the random assignment of subjects to treatment and control groups were not feasible, a quasi-experimental, rather than a true experimental, design was employed. A counterbalanced design, which rotated six talkers through 60 stimuli items presented to

72 listeners in four listener L1 groups, was selected to control for as many threats to internal validity as possible while exposing listeners to the stimuli sentences only once.

The location of the large public US research institution was selected because of its substantial international student population and the degree to which this population mirrored that of the US in general both in terms of the students’ countries of origin and their fields of study.

The chapter then detailed the stimuli, which were composed of written materials read by Chinese and American talkers, and concluded with the talker recording

168

procedure. The materials consisted of 60 sentences taken from the Bamford-Kowal-

Bench Standard Sentence Lists, revised for American English. The talker variables collected in a language background questionnaire were reported, as well as the analyses of talker speech rates and the talker segmental accuracy ratings for key words. Finally, the procedure and equipment used in the linguistics laboratory to record the talkers individually was detailed.

The methods consisted of the experiment paradigm, the subject selection, and experiment procedures. The psycholinguistic paradigm of word recognition in noise was chosen due to its extensive history in measuring speech intelligibility. International subjects were selected from a census of Chinese, Indian, and Korean graduate students who had achieved TA certification, or its equivalent, within the previous two years. A random sample of Americans with in-state residency status living in the municipal area of the main campus was contacted. Subjects were selected from those Americans who learned English as a first language, spoke it at home, and had lived in the state between the ages of 2 and 10.

The main experiment consisted of three sections: the language background questionnaire, the listening session, and the recording session. The L1 and L2 speech samples collected in the recording session were used to compile the Buckeye GTA corpus and were not reported on for this study. After filling out the language background questionnaire, the subjects completed the listening session which included oral repetition and written transcription tasks, followed by a word familiarity rating task.

169

Intelligibility ratings for the key words in each of the carrier sentences were collected as a dichotomous measure according to whether the listener’s transcription of the key word exactly matched the key word of the talker’s materials. The independent variables collected were talker L1, talker segmental accuracy, listener L1, and listener word familiarity. A mixed-effects model with crossed random effects was used with logistic regression to determine the effects of these variables on intelligibility.

170

Chapter 4: Results

4.1 Introduction

This study investigated the intelligibility of Chinese graduate students as determined by their Indian, Chinese, Korean, and American peers. First, the researcher sought to identify the main effects and interaction of talker L1 and listener L1 on intelligibility. Next, she examined the extent to which the accuracy of the talker’s pronunciation affected intelligibility and how this varied by listener L1. Finally, the effects of listener word familiarity on intelligibility were considered.

A mixed-effects model with logistic regression was selected to analyze the factors affecting intelligibility. The talker L1 (American or Chinese) and segmental accuracy and the listener L1 (Chinese, Korean, Hindi, and American) and word familiarity were fixed effects. Random effects included in the model were individual talker, listener, and key word. Further analyses were conducted on the Chinese talkers’ key words in order to identify the segmental pronunciation of each talker, as well as the key words which were most frequently misidentified within each listener L1 group.

4.2 Effects of Talker and Listener L1

In order to compare the effects on intelligibility of talker L1 and listener L1, an R code was used to extract those data from the E-Prime files merging all listeners. Using

171

percent correct calculations of intelligibility, a barplot showing talker L1 and listener L1 was generated, which is displayed in Figure 4.1 below.

Figure 4.1: Word intelligibility means by talker L1 and listener L1

As can be seen in Figure 4.1, all listeners found the American talkers (mean =

84%) to be more intelligible overall than the Chinese talkers (mean = 73%), despite the fact that the speech rate analyses indicated the Americans to be significantly faster talkers than the Chinese. The Chinese talkers were most easily understood by the American 172

listeners (82%) and most difficult to understand for Korean listeners (59%). Listeners with American English and Hindi L1s appeared to be more successful understanding

American English (95% and 89%, respectively) than were those with Mandarin and

Korean L1s (79% and 74%, respectively). American listeners were most successful when hearing American English, performing nearly at ceiling (95%). Furthermore, when we eliminate American listeners from the mean intelligibility, we find that international listeners accurately identified 81% of the American English key words they heard in contrast to accurately identifying 70% of the Chinese-accented English key words.

An interaction between talker L1 and listener L1 can clearly be seen for the

Mandarin L1 listeners when listening to Mandarin L1 talkers. Since listeners with

Mandarin and Korean L1s performed nearly equivalently when listening to American

English (79% and 74%, respectively), they are apparently at equivalent levels of oral proficiency in English. However, when listening to Chinese-accented English, Chinese listeners (74%) showed a benefit that equaled the performance of the Indian listeners

(76%). This ‘interlanguage match’ benefit is clearly missing when examining the Korean listeners’ results with Chinese-accented English, which averaged only 59% correct.

To determine the statistical significance of the main effects of talker L1 and listener L1 on intelligibility, a series of mixed effects models with logistic regression were built. The base model included talker L1 and listener L1 as fixed effects, as well as the interaction when talker L1 and listener L1 are both Mandarin. Listener, key word, and individual talker were entered as random effects. Using the lme4 package, the equation entered in R is given below:

173

Model0 = lmer (KeyWordResponse ~ talkL1 + lisL1 + Mand + (1Ilistener) +

(1IKeyWord) + (1Italker), alldata, family = binomial)

where “KeyWordResponse” is the key word intelligibility response variable; “talkL1” and “lisL1” are the fixed effects of talker L1 and listener L1, respectively; “Mand” is the interaction variable when talker L1 and listener L1 = Mandarin; “(1Ilistener),”

“(1IKeyWord),” and “(1Italker)” are the random effects of individual listeners, key words, and talkers, respectively; “alldata” identifies the data frame which contained all of the variables being considered; and “family = binomial” identifies the response variable as binomial. Tables 4.1 and 4.2 below show the statistical output of the fixed and random effects of the mixed-effects base model.

174

Fixed Effects Estimate Std. Error z Value Pr(> IzI) Significance

(Intercept) 2.9917 0.2142 13.967 < 2e-16 0.001

TalkL1M -1.0910 0.2388 -4.569 4.91e-06 0.001

LisL1Hindi -0.5956 0.1431 -4.163 3.14e-05 0.001

LisL1Korean -1.6694 0.1407 -11.869 < 2e-16 0.001

LisL1Mand. -1.4940 0.1541 -9.694 < 2e-16 0.001

Mand. 0.7912 0.1059 7.469 8.09e-14 0.001

Table 4.1: Statistical significance of the fixed effects of the mixed-effects base model

Random Number of Effects Observations Name Variance Std. Dev. Groups Key Word 146 (Intercept) 0.944568 0.97189

Listener 72 (Intercept) 0.130165 0.36078

Talker 6 (Intercept) 0.080455 0.28365

Total 13,392

Table 4.2: Variance and standard deviation of random effects of the base model

175

4.2.1 Effects of Listener L1 vs. Individual Listening Subject

The analysis of the fixed effects revealed that the effects of talker L1 and listener

L1 on intelligibility were all statistically significant (α = 0.001). In order to visually display the size of the L1 fixed effects relative to the variance accounted for by the random effects of individual listeners, a barplot was created, which is displayed in Figure

4.2 below.

Figure 4.2: Listener L1 group effects vs. within-group variation, Model 0

176

As can be seen in Figure 4.2, the listener L1 group effects are substantially larger than the within-group individual deviations. First, the intercept is used as the baseline accuracy level, as it represents when talker L1 and listener L1 are both American English.

The deviation from that baseline is displayed in the listener L1 bars in the left panel of the figure, as well as the effect of the talker L1 being Mandarin and the interaction when both talker L1 and listener L1 were Mandarin. In the panel on the right, the within-group individual deviations from the group means are displayed. These are graphic representations of the previous table of statistical output from the base model, showing the relative sizes of the fixed effects of listener L1 and talker L1 as compared to the random effects of individual listeners.

4.2.2 Effects of Talker L1 vs. Individual Talker

Figure 4.3 below was created in order to display the size of the effects of the talker L1 group fixed effect in relation to the random effects of individual talkers within those L1 groups. The figure shows the large effect of talker L1 = Mandarin in the bar across the bottom as compared to the smaller random effect of individual talker variation.

The analysis of the talkers’ speech rates showed Americans to be significantly faster talkers than the Chinese. However, this had no adverse effect on their intelligibility; in fact, percent correct intelligibility results showed that American talkers were more intelligible overall. In addition, the talker speech rate analyses revealed no significant difference in speech rate among the Chinese talkers. Therefore, the talker speech rate variable was not included in the mixed-effects base model. However, the figure below does show a benefit for talker AE3 of speaking more slowly. It also

177

illustrates that listeners found talker AE2 to be more difficult to understand, possibly due to his tendency toward underarticulation, or mumbling. Finally, the higher initial oral

English proficiency level, as measured by the SPEAK test, of Chinese talker ME3 is also evident in overall intelligibility. However, as noted above, none of these individual variations is even half as large as the effect size of the talker L1 group as a whole.

Figure 4.3: Individual talker effects vs. talker L1 group effect, Model 0

178

4.2.3 Individual Key Word Random Effects

The third and final random effect accounted for in the mixed-effects base model is individual key word variation. To display this variation, a Pareto plot was created, which showed the random effect intercepts by key word. Figure 4.4 shows the deviations from mean key word accuracy of each key word, represented by the bars in increasing order.

As shown in Figure 4.4 below, the panel to the left begins with the key words which deviated negatively the most from the mean, that is, those which were the most frequently misidentified. The panel in the center contains those words which were identified most frequently at the mean. The panel on the right shows the words which were identified at rates above the mean.

179

Figure 4.4: Individual key word random effects, Model 0

180

4.3 Effects of Talker Segmental Accuracy

To analyze the effects of talker segmental accuracy, two more mixed-effects models were created, which used a subset of the Mandarin talkers only, since all of the

American talkers had been labeled “accurate” as native speakers of the reference model of English. Both models maintained all of the random effects of the base model, which were listener, talker, and key word, varying only in terms of the fixed effect of talker segmental accuracy. In order to determine the goodness of fit of each of the models to the data, the Akaike's Information Criterion (AIC) values were compared. Since AIC penalizes for increasing numbers of parameters included in the model, it discourages overfitting. The model with the lowest AIC number is the one which best fits the data.

Model 1 included listener L1 as the only fixed effect, which generated an AIC value of 11,751. Model 2 added talker segmental accuracy as a second fixed effect and carried an AIC value of 11,750. Although this is a lower AIC value than that for Model

1, it is only negligibly better. This result indicates that talker segmental accuracy, as measured by phonemic accuracy judgments, does not significantly increase predictive power over that accounted for by the random effect of individual talker.

4.4 Effects of Listener Word Familiarity

In order to examine the effect of listener word familiarity on intelligibility, this fixed effect term was added to a third mixed-effects model. The AICs of Model 0 and

Model 3 were compared to determine goodness of fit. The AIC value for the base model was 11,693 and for Model 3, which included word familiarity, it was 11,580. This difference was sufficiently lower to justify accepting word familiarity as a predictor of

181

intelligibility, even with the variation of individual listening subjects and key words accounted for as random effects. The statistical output of the mixed-effects Model 3, with the fixed effect of word familiarity rating, is reported in Tables 4.3 and 4.4.

Fixed Effects Estimate Std. Error z Value Pr(> IzI) Significance

(Intercept) 2.37179 0.28888 8.210 < 2e-16 0.001

TalkL1M -1.10556 0.23512 -4.702 2.58e-06 0.001

LisL1Hindi -0.60844 0.14589 -4.171 3.04e-05 0.001

LisL1Korean -1.66968 0.14344 -11.640 < 2e-16 0.001

LisL1Mand. -1.46265 0.15696 -9.319 < 2e-16 0.001

Mand. 0.78848 0.10658 7.398 1.38e-13 0.001

Word Fam. 0.13383 0.04187 3.196 0.00139 0.01 Rating

Table 4.3: Statistical significance of the fixed effects of Model 3

182

Random Number of Effects Observations Name Variance Std. Dev. Groups Key Word 145 (Intercept) 0.913513 0.95578

Listener 72 (Intercept) 0.136699 0.36973

Talker 6 (Intercept) 0.077781 0.27889

Total 13,306

Table 4.4: Variance and standard deviation of random effects of Model 3

As can be seen in Table 4.3, the coefficient for word familiarity as a fixed effect is significant at the 0.01 level and is positive, above and beyond the effect of listener L1.

Table 4.4 shows that the variance in the random effects of key word (0.913513) is slightly smaller than the variance shown in the base model (0.944568) for that variable, as reported in Table 4.2. In addition, the size of the listener L1 effects were slightly altered by adding listener word familiarity to the model. These differences between models, as a result of the addition of word familiarity as a fixed effect, led the researcher to examine in greater detail the relationships between word familiarity and listener L1, as well as comparing how word familiarity impacted the key word random effects in Model

3 versus in Model 0.

183

4.4.1 Word Familiarity Ratings by Listener L1

Since a listener’s familiarity with a key word was a significant predictor of intelligibility, histograms were created to show the distribution of word familiarity ratings by listener L1, as displayed in Figure 4.5. Although familiarity was rated on a 5-point

Likert scale, listeners only rated their familiarity with known words – i.e., those they identified as previously “seen or heard.” The researcher assigned a total of six possible categories to the word familiarity variable by logging unknown words as category 0. If the listener indicated that s/he had “seen or heard” the word, then s/he rated his/her familiarity with the word on a scale of 1-5, where 1 = “very seldom seen or heard” and 5

= “very often seen or heard.” As the figure illustrates, the majority of the key words were highly familiar to all listeners, with only Mandarin listeners indicating that some key words were unknown to them – i.e., that they had not “seen or heard” those words before.

184

Figure 4.5: Word familiarity rating categories by listener L1

185

4.4.2 Key Word Effects of Model 0 versus Model 3

To compare the sizes of the fixed effects of listener L1 to the random effects of key word in the base model and Model 3 with word familiarity, Figure 4.6 below was created, displaying the barplot of listener L1 alongside the boxplot of key word effects.

As can be seen, comparing Model 0 on the left to Model 3 on the right, the fixed effect of listener L1 remains quite large. As the whiskers on the boxplots demonstrate, the addition of listener word familiarity reduces the range of the variation of key word effects very little.

Figure 4.6: Listener L1 effects and key word effects, Model 0 vs. Model 3

186

In order to examine the relative individual key word effects in greater detail, after including word familiarity as a fixed effect, the researcher created a key word effects

Pareto plot for Model 3, which can be compared to that of Model 0 in Figure 4.4. As in

Figure 4.4, Figure 4.7 below shows the words which were accurately transcribed at rates below the mean in the panel on the left, near the mean in the center panel, and above the mean in the panel on the right. Comparing the key word effects between the models reveals only minor differences in position for some key words. For example, after word familiarity is added to the model, the word “boxes” becomes the second most frequently mistranscribed word, and “buckets” moves to third, but “saucepan” remains the most frequently mistranscribed word in both models. Similarly, “house,” “cold,” and “family” remain the most frequently accurately transcribed words in both models.

And finally, using simple linear regression, the researcher calculated the correlation between the two key word effects in Model 0 and Model 3, which is displayed in Figure 4.8. The Pearson correlation coefficient was 0.99975, which indicates a strong positive linear relationship. The amount of variation in word familiarity was quite small, as most key words were highly familiar to listeners, so adding that factor to Model 3 reduced the variance in the key word effects very little. This slight change in variance between key word effects in the models is shown in the boxplots in Figure 4.6 above and confirmed by the strong correlation.

187

Figure 4.7: Individual key word effects, Model 3

188

Figure 4.8: Correlation of key word effects, Model 0 vs. Model 3

189

4.5 Key Word Analyses

4.5.1 Segmental Error Patterns of Mandarin L1 Talkers

The Chinese-accented key words which listeners misidentified were subjected to further analyses. Since some of the key words occurred in more than one sentence, a pivot table by sentence and key word was created in Excel to examine the error rates for each key word token. Those key words which at least 75% of all listeners did not identify accurately were reviewed by two transcribers so that segmental error patterns which caused consistent problems for the Chinese talkers could be identified.

The first transcriber was the researcher, who was an instructor in the ITA program for six years and is a native speaker of American English. The second transcriber was a veteran professor in the ITA program with a Ph.D. in Linguistics, who is also a native speaker of American English. The transcriptions were compared and those few which differed were heard again by the two transcribers together, and agreement was reached on all transcriptions. The disagreements primarily consisted of the symbol used to represent an ambiguous sound – or a Chinese-accented sound which was not a good exemplar of an

American English phoneme category. For example, the word “table” was transcribed as

/tiybow/ and /tiybɔ/ by the first and second transcriber, respectively. Both transcribers agreed that the final /l/ had been vocalized, but disagreed slightly in how to represent the back vowel produced. After a second listening together, the two transcribers agreed on the second transcription due to the short vowel duration.

The key words read by Chinese talkers that at least 75% of listeners mistranscribed were sorted in decreasing order of error frequency. The transcription of

190

the talker’s key word was compared to a canonical transcription in American English

(CMU), and then the phonemic error type (V = vowel; C = consonant; V,C = both) was identified. Tables 4.5 through 4.7 show these data for Chinese talkers ME1, ME2, and

ME3, respectively. An asterisk was used to indicate that a segmental phoneme boundary had not been crossed, as the transcription shows, suggesting that a prosodic issue had led to the breakdown in intelligibility. Transcriptions in these tables are given in Arpabet, which is a machine-readable phonetic alphabet based on the International Phonetic

Alphabet (IPA) and used by the Carnegie Mellon Pronouncing Dictionary to represent the phonemes of American English. A correspondence key between the IPA and the CMU phonemic symbols are given in Appendix M.

191

Error Error Transcription CMU Transcription Key Word Count % Type

buckets 100% V B AA1 K IH0 T S B AH1 K AH0 T S saucepan 100% * S AO2 S P AE1 N S AO1 S P AE2 N table 100% V, C T IY1 B AO0 T EY1 B AH0 L climbed 92% V K L AE1 M D K L AY1 M D clock 92% V K L AO1 K K L AA1 K five 92% V F AE1 V F AY1 V ground 92% V G R AO1 N D AH G R AW1 N D hold 92% C HH OW1 Ø D HH OW1 L D ladder 92% C, V L AE1 N D AH0 L AE1 D ER0 letter 92% V L AE1 D AH0 L EH1 T ER0 stood 92% V, C S T UW1 T S T UH1 D cheese 83% * CH IY1 Z CH IY1 Z legs 83% V L AE1 G AH0 Z L EH1 G Z she 83% V SH IH1 SH IY1 bus 75% V B AA1 S B AH1 S followed 75% V F AO1 L OW2 D F AA1 L OW0 D jug 75% V JH AA1 G JH AH1 G stopped 75% * S T AA1 P T S T AA1 P T they 75% V DH AH1 DH EY1 writes 75% C ZH R AY1 T S R AY1 T S

Table 4.5: Key words of ME1 mistranscribed by 75% or more listeners

192

Key Word Error Error Transcription CMU Transcription Count % Type

boy 100% V B AY1 B OY1

bread 100% V B R AY1 D B R EH1 D

good 100% V G UW1 D G UH1 D

helping 100% C, C HH EH1 W P IY0 N HH EH1 L P IH0 NG

jug 100% * JH AH1 G JH AH1 G

washed 100% C W AA1 S T W AA1 SH T

bicycles 92% V, C B AE1 S IH0 K OW2 Ø B AY1 S IH0 K AH0 L Z

buckets 92% C B AH1 K IH0 Ø Z B AH1 K AH0 T S

fruit 92% V F R UH1 T F R UW1 T

broom 83% * B R UW1 M B R UW1 M

five 83% V, C F AW1 Ø F AY1 V

men 83% V, C M AE1 Ø M EH1 N

milk 83% C M IH1 W K M IH1 L K

shoe 83% * SH UW1 SH UW1

writes 83% V R AE1 T S R AY1 T S

ground 75% V G AH0 R AA1 N D G R AW1 N D

looked 75% V L UW1 K T L UH1 K T

path 75% V P AA1 TH P AE1 TH

riding 75% V, C R EH1 D IY0 N R AY1 D IH0 NG

tells 75% C T EH1 W Z T EH1 L Z

used 75% C Y UW1 S T Y UW1 Z D

Table 4.6: Key words of ME2 mistranscribed by 75% or more listeners

193

Key Word Error Error Transcription CMU Transcription Count % Type

climbed 100% C K L AY1 M T K L AY1 M D

legs 100% V, C L AE1 K S L EH1 G Z

three 100% * TH R IY1 TH R IY1

jug 92% C JH AH1 K JH AH1 G

ladder 92% V L EH1 D ER0 L AE1 D ER0

book 75% V B AO1 K B UH1 K

hold 75% V HH OW1 Ø D HH OW1 L D

pond 75% C, C P AA1 Ø Ø P AA1 N D

Table 4.7: Key words of ME3 mistranscribed by 75% or more listeners

Interestingly, among the key words which were most frequently misidentified by listeners from all L1 groups, there were many more vowel errors (30) than consonant

errors (21), mainly due to talker ME1 (15V, 5C). Vowel errors were the most

problematic, as well as the most frequent. Key words with the monophthong vowels /æ/,

/^/, /ε/, /ʊ/, and /u/ were all reported to have high error ratings. For example, words like

“ladder,” “path,” “buckets,” “bus,” “letter,” “legs,” “looked,” “stood,” and “fruit” were

frequently mistranscribed (75-100%). The productions of the diphthong vowels (/ai/, /oi/,

194

/au/) by ME1 and ME2 caused problems for listeners. For example, substituting the monophthong /ae/ for the diphthong /ai/ in “five,” (92%) “bicycles,” (92%) and “writes”

(83%) frequently led to misidentification. Substituting one diphthong for another also led to problems, such as /ai/ → /au/ in “five” (83%) and /oi/ → /ai/ in “boy,” which was the most problematic, as it led to 100% of listeners misidentifying the word.

The vowel errors for all of the Chinese talkers were tabulated in the vowel error chart displayed in Table 4.8. The errors are categorized by the features of the target vowel, with the arrows indicating the substitutions that the talkers made. The most frequent errors were due to the substitution of the monophthong /æ/ for both the monophthong /ε/→/æ/ and for the diphthong /ai/→/æ/. This suggests that the Chinese talkers perceived the monophthong and the diphthong to be similar in their L2 phonological space.

The next most frequent vowel errors involved the mid central vowel /^/→/a/ and the high lax back vowel /ʊ/→/u/. Also as indicated in the following error chart, the

Chinese talkers often confused /I/→/i/ and simplified the r-colored central vowel

/ɚ/→/ə/, especially in word-final position. The substitution of /a/→/ɔ/ suggests that they accurately perceived the backness of the vowel, but were unable to lower it sufficiently.

Finally, the Chinese talkers reduced unaccented syllables slightly differently than those transcribed in the canonical pronunciation of American English /ə/→/I/, but the researcher noted that these variations also occur in L1 dialects.

195

Vowel Errors

Front Central Back

Tense Lax Tense Lax Tense Lax

High /i/→/I/ /I/→/i/ /ɚ/→/ə/ /u/→/ʊ/ /ʊ/→/u/ (2x) (2x) (3x) /ʊ/→/ɔ/

Mid /e/→/i/ /ε/→/æ/ /^/→/a/ /ə/→/I/ /oi/→/ai/ /e/→/^/ (4x) (3x) (2x) /ε/→/ai/

Low /æ/→/ε/ /ai/→/æ/ /a/→/ɔ/ /æ/→/^/ (4x) (2x) /ai/→/ε/ /ai/→/aʊ/ /aʊ/→/ɔ/ /aʊ/→/a/

Table 4.8: Vowel errors in mistranscribed words spoken by Chinese talkers

196

Among the consonant errors, the key words with word final consonants and consonant clusters were frequently misidentified. For example, words ending in consonants like “five,” “table,” “men,” and “jug” were frequently misheard (83-100%).

Consonant clusters in word final position, such as /Vnd/ in “pond,” /Vmd/ in “climbed,”

/Vgz/ in “legs,” /Vld/ in “hold,” and /Vlz/ in “bicycles” were simplified, devoiced, or elided by the Chinese talkers, leading to key words not being recognized by the listeners

75-100% of the time. As can be seen in Table 4.7, Talker ME3 devoiced the final /d/ in

“climbed” and elided the final /nd/ in “pond” while nasalizing the preceding vowel and adding a glottal stop. Neither strategy proved to be successful, with 100% of all listeners mishearing ME3’s pronunciation of “climbed” and 75% mishearing “pond.” In pronouncing the word initial consonant cluster /gr/ in “ground,” talker ME2 used schwa epenthesis instead of simplifying the cluster, which was the strategy applied to word final clusters. This epenthesis led to misperceptions of the word for 75% of the listeners.

ME1’s pronunciation of the words “ladder” was misheard by 92% of listeners, as he added an /n/ before the /d/, so that it was easily confused with “lander.” However, this seems more an idiosyncrasy of the speaker, rather than a common feature of Chinese- accented English, since neither of the other two speakers exhibited this pattern, and ME1 also pronounced the word “suddenly” with the same insertion of /n/ before the /d/.

197

Consonant Errors

Error Single Consonants Consonant Clusters Category

Word- Word- Word- Word- Word- Initial Medial Final Initial Final

ŋ Place of / /→/n/ /ʃt/→/st/ Articulation (2x)

Manner of /r/→/ʒr/ Articulation

/zd/→/st/ /l/→/ɔ/ (4x) Voicing /d/→/t/ (2x) /ts/→/z/ /g/→/k/ /gz/→/ks/

Deletion /l/→/ø/ /ld/→/ød/ /v/→/ø/ (2x) /n/→/ø/ /lz/→/ø/ /nd/→/ø/

Epenthesis /d/→/nd/ /d/→/də/ /gr/→/gər/

Table 4.9: Consonant errors in mistranscribed words spoken by Chinese talkers

198

The consonant errors for all Chinese talkers were combined and displayed in

Table 4.9. They are grouped according to error category, position in word, and whether they occurred singly or in a cluster. The most common error for single consonants was that of voicing and occurred most frequently in word-final position. Final consonants were frequently devoiced or elided, which proved difficult for listeners to understand. In addition, /l/ in word-final or syllable coda position was vocalized or elided if preceded by an /o/. Consonant clusters in word-final position also proved problematic for Chinese talkers to pronounce. Devoicing, simplification, and elision were common strategies, not all of which were intelligible to the listeners.

To determine which of these pronunciation variations were intelligible, or problematic, for each L1 listener group, the key words which were mistranscribed by more than half of the group are analyzed in the following section.

199

4.5.2 Intelligibility Results by Listener L1

Listener L1 was found to be a significant predictor of key word intelligibility. In order to identify which key words caused the most problems for each listener group, the key words which were incorrectly identified by more than half the listeners were tabulated by L1. The tables below show the percentage (56-100%) of listeners within each of the four L1 listener groups (American English, Hindi, Mandarin, and Korean) who mistranscribed the key word.

Key Word Intelligibility Key Word Intelligibility (% Incorrect) (% Incorrect) boxes 89% buckets 67%

jug 89% fruit 67%

match 89% hold 67%

bus 78% five 56%

climbed 78% he 56%

ladder 78% saucepan 56%

legs 78%

Table 4.10: American L1 listener percent mistranscriptions of key words

200

For American L1 listeners, vowel pronunciation led to the most misperceptions.

The monophthong vowels /^/ and /æ/ caused problems in the words “buckets” (67%),

“bus” (78%), “jug” (89%), “saucepan” (56%), “ladder” (78%), and “match” (89%). The diphthong /ai/ caused problems in the words “five” (56%) and “climbed” (78%). The only consonant pattern which led to some difficulty were consonant clusters in syllable coda position. For example, 67-89% of American L1 listeners could not identify

“boxes,” “climbed,” “buckets,” and “hold.”

201

Key Word Intelligibility Key Word Intelligibility (% Incorrect) (% Incorrect) buckets 100% story 67%

hold 89% towel 67%

jug 89% washed 67%

five 78% book 56%

front 78% bus 56%

stood 78% clock 56%

tells 78% dish 56%

climbed 67% men 56%

ladder 67% she 56%

legs 67% they 56%

looked 67% three 56%

saucepan 67% writes 56%

Table 4.11: Hindi L1 listener percent mistranscriptions of key words

202

For Hindi L1 listeners, the vowels that caused the most trouble also included /æ/ and the diphthong /ai/, as for the American L1 listeners. However, unlike the American

L1 listeners, the Hindi L1 listeners frequently did not recognize key words with the mid vowels /^/ and /ε/, as well as the high unrounded back vowel /ʊ/. For example, words such as “buckets” (100%), “tells” (78%), and “stood” (78%) were correctly identified only 0-22% of the time by Hindi L1 listeners.

Consonant clusters, primarily those in syllable coda position, caused problems in words such as “hold” (89%), “front” (78%), “tells” (78%), “looked” (67%), “washed”

(67%), “legs” (67%), and “writes” (56%). Most frequently, consonant clusters which included /l/ caused problems in words like “hold” (89%) and “tells” (78%). Words ending in single consonants , especially sonorants, also proved problematic. For example, “saucepan” (67%), “towel” (67%), and “men” (56%), were accurately transcribed only 33-44% of the time by Hindi L1 listeners.

203

Key Word Intelligibility Key Word Intelligibility (% Incorrect) (% Incorrect) saucepan 100% cup 56%

looked 89% dropped 56%

broom 78% fell 56%

buckets 78% fruit 56%

five 78% he 56%

jug 78% helping 56%

shoe 78% hold 56%

used 78% ladder 56%

climbed 67% legs 56%

dancing 67% men 56%

drank 67% path 56%

janitor 67% stood 56%

tells 67% story 56%

writes 67% they 56%

cheese 56% washed 56%

Table 4.12: Mandarin L1 listener percent mistranscriptions of key words

204

As can be seen in Table 4.12, the majority of the key words which caused problems for Mandarin listeners were those which included the mispronounced vowels shown in the vowel error chart of the previous section (Table 4.8). For example, words with the monophthong vowels /æ/, /ε/, /^/, and /u/ were most frequently misidentified, as well as words with the diphthong /ai/.

Key words that included consonant clusters were highly likely to be misperceived, especially those with final consonant clusters or with both word-initial and word-final consonant clusters. For example, only 11-44% of Mandarin L1 listeners could identify the following key words containing consonant clusters: “buckets,” “broom,” “drank,”

“climbed,” “looked,” “used,” “dropped,” “fruit,” “tells,” “washed,” “writes,” “stood,” and

“story.” To some extent, /l/ in syllable coda position also negatively affected intelligibility for Mandarin listeners. Therefore, only 33-44% of Mandarin L1 listeners correctly identified key words such as “tells” (67%), “fell” (56%), “helping” (56%), and

“hold” (56%).

205

Key Intelligibility Key Intelligibility Key Intelligibility Word (% Incorrect) Word (% Incorrect) Word (% Incorrect)

buckets 100% packed 78% wrong 67%

climbed 100% used 78% dinner 56%

jug 100% writes 78% boy 56%

shoe 100% bicycles 67% dancing 56%

stood 100% bouncing 67% dirty 56%

broom 89% boxes 67% door 56%

cup 89% bus 67% fell 56%

fruit 89% came 67% funny 56%

ladder 89% climbing 67% hard 56%

story 89% five 67% high 56%

tells 89% front 67% home 56%

book 78% ground 67% hot 56%

cheese 78% he 67% letter 56%

hold 78% saucepan 67% listening 56%

janitor 78% shirts 67% nose 56%

legs 78% suddenly 67% plate 56%

looked 78% tree 67% riding 56%

Table 4.13: Korean L1 listener percent mistranscriptions of key words

206

As can be seen in Table 4.13, Korean L1 listeners had the most difficulty correctly identifying the Chinese-accented key words that they heard. Like other listeners, Korean listeners frequently misidentified key words containing the front vowels

/æ/ and /ε/, as well as the central vowel /^/, and back vowels /u/ and /ʊ/. Korean L1 listeners also shared difficulty in perceiving key words with the diphthong /ai/, such as in words “climbed” (100%) and “bicycles” (67%). Unlike other listeners, Koreans frequently misidentified key words containing /ə/ and /ɚ/ as well as the high front vowels, with slightly more trouble identifying words with /I/ than with /i/. For example,

Korean L1 listeners had difficulty perceiving words containing /I/ as in “janitor” (78%),

“bicycles” (67%), “listening” (56%), and “dinner” (56%); and /i/ as in “cheese” (78%) and “tree” (67%).

As with other listeners, Koreans found consonant clusters in word-final position difficult to understand, but they had significant difficulty with consonant clusters in word-initial position. For example, the following key words containing initial consonant clusters were only correctly identified 0-44% of the time by Korean L1 listeners:

“climbed” (100%), “climbing” (67%), “ground” (67%), “stood” (100%), “story” (89%),

“stopped” (56%), “tree” (67%), “fruit” (89%), “front” (67%), “broom” (89%), and

“plate” (56%). Korean L1 listeners also had difficulty identifying Chinese-accented words beginning with /b/ (8 out of 55 key words, or 15%), /h/ (6/55 key words, or 11%), and /f/ (5/55 key words, or 9%).

Since several of the most frequently misperceived words occurred in the same stimulus sentence, a mixed-effects model was created to identify which sentences were 207

outliers. In this model, which was that used as the base model plus word familiarity, the random effect of key word was replaced with the random effect of trial. Trial was used because each stimulus sentence was presented in the same order, 1-60, which corresponded to the trial number. Three sentences were identified as highly frequently missed: #18 – “The match boxes are empty,” #19 – “He climbed his ladder,” and #59 –

“The buckets hold water.”

4.6 Summary

This chapter presented the statistical analyses of the fixed effects of talker L1,

listener L1, talker segmental pronunciation accuracy, and listener word familiarity on

intelligibility. Using a mixed-effects model with logistic regression, the random effects

of subject, talker, and key word could also be accounted for and compared to the size of

the fixed effects.

The findings indicated that talker L1 (α = 0.001), listener L1 (α = 0.001), and

listener word familiarity (α = 0.01) were all significant predictors of intelligibility. In

addition, the interaction when talker L1 and listener L1 were the same was also

significant (α = 0.001). In comparison, although there was some individual variation

within the talker and listener L1 groups, the size of the effect was small relative to the

fixed effects between groups. The individual variation by key word was also relatively

small and was accounted for as a random effect in the model. Among the fixed effects,

only talker segmental accuracy was not found to be significant over and above individual

talker variation.

208

The Chinese-accented key words which were mistranscribed by 75% or more of the listeners were analyzed for segmental patterns of the talkers. Next, the key words which were misidentified by more than half of the listeners in a listener L1 group were also examined and categorized for each L1. American listeners found key words with the monophthong vowels /æ/ and /^/, as well as the diphthong /ai/ to be the most difficult to understand. Hindi L1 listeners also found those vowels to be problematic, as well as the front mid vowel /ε/ and the high back vowel /ʊ/. In addition, consonant clusters including /l/, especially in final position, were frequently inaccurately transcribed.

Mandarin L1 listeners had the most trouble understanding words with consonant clusters, in either word-initial or word-final position. They also frequently misidentified key words with monophthong vowels /æ/, /ε/, /^/, and /u/ and the diphthong /ai/.

Overall, Korean listeners had the most difficulty understanding Chinese-accented

English. As with the Mandarin L1 listeners, Koreans frequently misperceived consonant clusters; like the Hindi L1 listeners, they found central vowels /^/, /ə/, and /ɚ/ difficult to identify; and as with the American listeners, they misunderstood /æ/ and /ai/. Unlike the other listeners, Korean listeners misidentified key words containing high front vowels, both /i/ and /I/.

209

Chapter 5: Discussion and Conclusions

5.1 Introduction

Research on second language (L2) speech has recently begun to consider the needs of L2-L2 English communication; however, much remains to be investigated regarding the impact of talker L1, listener L1, and listener word familiarity on L2 English intelligibility. To this end, the current study measured the intelligibility of Chinese- accented English to native and nonnative listeners. It focused on an oral English proficiency level which is the goal of many international students in the US, that of

“graduate TA certification,” rather than the more commonly investigated proficiency level of “graduate school admission.” Finally, it focused on L2 English pronunciation needs within a specific context, that of the US university.

Therefore, this study contributes to L2 research by investigating nonnative

English speech production at an oral proficiency level which has been determined to be sufficient for the specific purposes of teaching native English-speaking students.

However, until this study, the intelligibility of foreign-accented English to nonnative listeners in the context of the international academic community had not been evaluated.

210

5.2 Effects of Talker and Listener L1

One of the strengths of using a mixed-effects model with logistic regression was to be able to analyze the predictive ability of the fixed effects, while capturing the random variation of individual talkers, listeners, and key words in the random effects.

For example, the effect of random variations of individual talkers within the native and nonnative L1 talker groups could be accounted for as a random effect, allowing the fixed effect of the talker being either a native or a nonnative speaker of English to be more clearly evaluated. The large group effect of talker L1 relative to the individual talker variation made it clear that being a native speaker of English had a significant positive effect on intelligibility regardless of any of the individual talker’s proclivities toward mumbling, for instance.

Similarly, the listeners’ native language also proved to be a significant predictor of their success in perceiving native and nonnative English, regardless of the variation found among individual listeners within that language group. This finding confirms that the listener’s native language has an important role to play in theories of intelligibility, as well as in the application of standards for the intelligibility of international English.

Furthermore, the analyses affirmed the strength of the interaction when the L1 of the listener matched that of the talker.

This L1-L1 match benefit meant that when oral English proficiency level was equal, as demonstrated by the equivalent performance of Mandarin L1 and Korean L1 listeners on the American English speech tokens, Mandarin L1 listeners found Mandarin-

211

accented English to be equivalently intelligible to American English, whereas the Korean

L1 listeners found Mandarin-accented English to be significantly less intelligible.

This interaction between talker L1 and listener L1 is an important finding for international graduate students, who will have to acknowledge that they cannot be accurate judges of the intelligibility of their own speech to other international students from the Expanding Circle. In fact, standards of English intelligibility among Expanding

Circle speakers may tend even more conservatively toward native-speaker norms, except when an L1 is shared, than for interlocutors from the Inner and Outer Circle.

5.3 Effects of Talker Segmental Accuracy

Interestingly, while segmental differences were found in production and perception due to L1, no significant effect on intelligibility was found for the segmental accuracy judgements of the key words spoken by the Chinese talkers. The model included the random effects of individual talkers, above which the talker segmental accuracy ratings were not large enough to be significant. Since the group effect of the talker’s native language was a significant predictor of intelligibility, then there are three possible interpretations for why talker segmental accuracy was not significant, all of which may be operating.

First, that there were only three Chinese talkers, which may have been too few to show an effect above that captured by the individual talker variation. Second, that the measure of talker segmental accuracy may not have been sufficiently precise. Two native

English-speaking raters listened to the key words in the sentence items in quiet, rather than in noise, and then judged them as either “accurate” or “not accurate.” The

212

instructions for rating were to compare the talker’s segmental pronunciation to a phonemic transcription of the same word in American English. However, one of the raters had considerably less phonetic training and may have judged a word to be inaccurate based on a subjective scale of acceptability, rather than on the phones produced by the talker. A final explanation for why segmental accuracy was not significant could be that segmental pronunciation represents too small a portion of the full range of phonetic-acoustic factors comprising an accent.

5.4 Effects of Listener Word Familiarity

A listener’s familiarity with a word significantly improved his/her likelihood of finding that word to be intelligible. Since the model also included individual listener and key word variation as random effects, word familiarity was significant above and beyond this individual variation. Although the majority of the key words were highly familiar to listeners, low familiarity was still a predictor of low intelligibility.

Once word familiarity was accounted for as a fixed effect, the changes in both listener L1 and key word effects were examined in greater detail. The size of the fixed effects of listener L1 remained substantial, compared to the variation in word familiarity.

Correlation analyses revealed that the key word effects in both models were strongly correlated. So, the range of the key word variation was reduced only slightly by adding word familiarity to the model. This finding indicates that some words remained more difficult to recognize than others, due to factors other than listener word familiarity, and this random variation is important to capture in the data analysis model.

213

Since word familiarity was a significant predictor of intelligibility, it is important to test for this factor when using subjects who are nonnative speakers of English. All of the key words in the sentence lists had previously been rated as highly familiar to native

English speakers. However, as the results of the word familiarity test revealed, words which were highly familiar to native English speakers were not necessarily familiar to nonnative speakers. Therefore, holding word familiarity constant is unlikely with subjects who are both native and nonnative speakers. Including word familiarity as a fixed effect in the mixed-effects model produced a better fit to the data, and its inclusion allows researchers to account for its effect on speech intelligibility, which has frequently been found to be significant.

In practical terms, this finding suggests that listeners can improve their ability to comprehend English speech, regardless of the talker’s L1, by increasing their vocabulary.

In the academic context, since many key terms in math, science, and engineering will be unfamiliar to listeners, it would be helpful to provide students with audio textbooks or other digital resources so that they can learn key terms with two modalities of input, one written and one oral.

5.5 Key Word Analyses

Analyzing the unintelligible key words of the Mandarin L1 talkers helped to identify patterns in their segmental pronunciation, which may have led to frequent misidentification. For example, in syllable coda position, the Chinese talkers tended to simplify consonant clusters, vocalize /l/, and elide final consonants, or substitute them

214

with glottal stops. As for vowels, they produced diphthongs /ai/ and /au/ as monophthong

/a/ or substituted /æ/ for /ai/, confused /ε/ with /æ/, and substituted /a/ for /ʌ/.

Comparing these pronunciation variations to the intelligibility results revealed that not all of these variations had an equally negative impact on intelligibility for all listeners, but was affected by the listener’s L1. For American L1 listeners, the Chinese talkers’ pronunciations of consonants were not particularly problematic, as compared to the vowels /æ/, /ʌ/, and /ai/, which resulted in the highest error frequencies. In contrast, for Hindi L1 listeners, the consonant cluster simplification and vocalization of /l/ in syllable coda position had a negative impact on intelligibility. Like the Americans, the

Hindi L1 listeners had difficulty understanding /æ/, /ʌ/, and /ai/, but they additionally could not successfully perceive many key words with the vowels /ε/ and /ʊ/. In addition to the problematic consonant clusters and vowels mentioned above, Mandarin L1 listeners had high error rates with key words containing the high back vowel /uw/.

Final consonant cluster simplification proved difficult for Korean L1 listeners to understand, but word-initial clusters posed more difficulty. Like the other listeners,

Korean listeners frequently misidentified key words containing the front vowels /æ/ and

/ε/, as well as the central vowels /^/, /ə/, and /ɚ/ and the diphthong /ai/. Unlike listeners with different L1s, Koreans frequently misidentified key words containing the high front vowels /i/ and /I/.

215

Knowing which types of sounds are most problematic for which listener L1 could help English language teachers tailor their pronunciation curricula to the needs of specific listeners. For example, Chinese speakers preparing to present academic papers to

American English speaking audiences could practice /ae/ and /ai/ to have the most beneficial impact on their intelligibility. Although /l/ vocalization is an acceptable allophone in American English and was therefore intelligible to American listeners, Hindi listeners considered it frequently unintelligible. Therefore, some more intelligible pronunciation of final /l/ would need to be employed for these listeners. For Korean listeners, more careful attention to initial consonant and consonant cluster pronunciation might enhance intelligibility the most.

Since some of the most frequently misperceived words occurred in the same sentence, the sentences which were outliers were identified for further consideration:

#18 – “The match boxes are empty,” #19 – “He climbed his ladder,” and #59 – “The buckets hold water.” Notably, “match” and “boxes,” “buckets” and “hold” were

frequently mistranscribed as a unit. This finding suggests that segmental variation alone

does not account for the low intelligibility ratings of these key words. Further research

should explore in greater phonetic-acoustic detail how these sentences were produced, as

well as how listeners transcribed them to shed light on specifically what words they

heard.

5.6 Recommendations for Further Research

Further research needs to address other types of institutional locations, especially

smaller institutions, where students with different levels of familiarity with foreign-

216

accented speech could be investigated. Since this study focused on graduate students, other studies should investigate the effects on intelligibility of different academic levels, such as that of undergraduates or of professors. Other languages and dialects for both talkers and listeners, such as Taiwanese Mandarin or Cantonese, could provide quite a more comprehensive view of Chinese-accented English. Other L2 English oral proficiency levels should be investigated as well, in order to shed light on how pronunciation changes over time and with increasing levels of language experience. In general, data on a greater number of speakers, as well as both male and female speakers, within an L1 would yield more reliable findings.

Other studies employing different materials and tasks can provide useful comparisons to the current study’s findings. For example, a full range of materials would yield data on the intelligibility of words and phrases to passages or longer speech samples. Settings other than the laboratory, such as the classroom or other naturalistic settings could provide converging data toward a clearer picture of how these research contexts affect intelligibility findings.

5.7 Conclusions

5.7.1 Contributions to Second Language Speech Research

Research in speech perception has contributed to the development of phonological theory. Similarly, second language speech intelligibility research contributes to theories of L2 phonological acquisition by providing detailed information on second language speech production at specific developmental or proficiency levels, but also for how that production is perceived by listeners with different L1s. The current study identifies the

217

phonological features of Chinese-accented English at the oral proficiency level of

“graduate TA certification,” as well as its intelligibility for the wide variety of international listeners found in US graduate schools and the international academic community. Using a mixed-effects logistic regression model to analyze intelligibility data proved to be a powerful tool, due to the model’s ability to account for random effects and fixed effects separately, simultaneously, and in relation to one another.

The findings confirmed the significant effects of a listener’s word familiarity and

L1 on intelligibility, especially when the L1 is shared with the talker. This interlanguage match benefit at this proficiency level is in keeping with the findings of Bent and

Bradlow (2003). In fact, the large size of the talker and listener L1 effects indicates that further research is warranted into the specific aspects of the accent and their relative effects on intelligibility.

Finally, the finding that talker segmental accuracy was not a significant predictor of intelligibility raises important questions for research and theory. As suggested by

Ladefoged (2005), while vowels and consonants are convenient units for writing in alphabetic languages, they may not be the units in which the human brain processes speech. Therefore, for research purposes, they may not be ideal units of analysis for examining human communication, as also implied by the current study’s findings.

Instead of analyzing segmental and prosodic pronunciation as if they occurred in isolation, it may be more useful to develop a measure of pronunciation that is more ecologically valid. It would only follow then, that language teaching would also need to

218

move beyond the limits of articulatory phonetics, requiring of classroom teachers a more holistic approach to pronunciation teaching.

5.7.2 Contributions to English Language Teaching

The Chinese talkers were all certified as TAs, but tests of oral English proficiency evaluate many factors besides pronunciation. Therefore, measuring the intelligibility of these TAs to American listeners helps identify what the acceptable level of intelligibility is for teaching purposes at a US university. As the results of this study showed,

American listeners understood Chinese-accented English at an average rate of 82%, implying that this rate of intelligibility may be the benchmark for direct teaching contact with American students. In contrast, international listeners found Chinese-accented

English to be 69% intelligible on average, suggesting that the pronunciation “acceptable” for native listeners is not necessarily sufficient for intelligible communication with nonnative listeners in the academic context.

These findings indicate that Chinese speakers of English need to adapt their pronunciation in specific ways to increase their intelligibility for their international peers in graduate school. Likewise, in preparation for studying in the US, international students will need to improve their ability to comprehend foreign-accented English, especially those accents most commonly found in the US university. This puts the burden on English pronunciation teachers to prepare their students for the diversity of

English accents they will encounter.

The finding that an interlanguage match between talkers and listeners was a benefit to intelligibility is also important for language teachers. American English was

219

more intelligible to all listeners, but Mandarin-accented English was equally intelligible to Mandarin L1 listeners. Since many English language teachers are either native speakers or nonnative speakers who share their students’ native language, this bodes well for the intelligibility of English within the language classroom. However, it suggests that a special effort must be made to train international speakers and listeners of English to accommodate a wider variety of accented English so they will be prepared to meet the linguistic challenges they will face outside the language classroom.

In the context of English Language Teaching at the college level in China, textbook publishers and teachers can both make greater efforts to develop digital materials with a representative array of accents in English for international usage. In addition, Chinese universities could partner with US institutions in order to connect their

English language classes directly. Doing so via Skype, videoconferencing, or other computer-assisted technologies will help Chinese students develop the linguistic skills they will need to accommodate different accents as they pair up with ESL students with different L1s for practice with English conversation and presentation.

Foreign-accented speech presents a challenge for intelligibility, as evidenced by the findings in this study. However, considering that international students will continue to study, conduct research, and be employed by other international users of English, there are significant professional benefits to becoming proficient in comprehending a variety of accents in English. Therefore, within English-speaking countries, where language classes include students with many different L1s, language teachers can take advantage of the accent diversity in the classroom. They can make a conscious effort to put students

220

together in different L1-L1 pairs for information gap exercises, giving and receiving instructions, and micro-teaching. For presentation purposes, students should know how to prioritize different aspects of their English pronunciation to enhance intelligibility for listeners with different L1s.

In addition to the impact of the native language, one of the avenues to improving a listener’s comprehension skills is through increasing his/her vocabulary, since the findings indicated that word familiarity was a significant positive predictor of intelligibility. Specifically, because this study was conducted in a US university, achieving high word familiarity with discipline-specific English would seem to be beneficial for both Americans and international listeners. Arguably, American students who find their Chinese TAs difficult to understand can improve their perception of

Chinese-accented English by becoming more familiar with the vocabulary most likely to be used in their particular field. Math, science, and engineering textbook publishers in particular would benefit both American and international students by offering audio book versions of their publications, as well as audio resources pronouncing key terms in a variety of accents in English. Receiving input in two modalities, oral and visual, has also been found to enhance vocabulary learning (Wong, 2001).

Besides focusing on increasing student vocabulary, US universities can proactively address the other half of the ITA-undergraduate communication issue by providing linguistics workshops during orientation for new students. Especially nonlanguage majors need to learn the basic features of L2 speech to be more familiar with and prepared for the various accents they will hear in their classrooms. In addition,

221

websites can be developed to provide American undergraduates with further audio and video samples, as well as various listening exercises for practice. Recent research showing multi-speaker exposure can translate into gains in understanding an accent in general (Bradlow & Bent, 2008) shows promise for applications with university populations.

Finally, as undergraduates become increasingly international as well, ITA programs in the US would also benefit by incorporating more materials with a variety of accents in English. These resources could be combined with discipline-specific vocabulary building, using commonly employed academic “lexical bundles” (Biber &

Barbieri, 2007) in pronunciation exercises focusing on prosody, aligning tune with text for academic communication purposes. However, since ITA programs are not primarily charged with improving intelligible communication among international graduate students even as their numbers continue to increase, perhaps it is time for US universities to expand the educational offerings of English for Academic Purposes, and fund them appropriately, so that they can address these needs. Since research conducted by international graduate students attracts substantial external funding and since they contribute higher out-of-state tuition to the university as well as low-cost teaching labor, it is in everyone’s best interest to provide crucial language training which focuses on their diverse oral English research and study needs. Indeed, US universities have a responsibility to better facilitate the intercultural communication that has become a reality of higher education.

222

References

Adamson, B. (2002). Barbarian as a foreign language: English in China’s schools.

World Englishes, 21(2), 231-243.

Akaike, H. (1974). A new look at the statistical model identification. IEEE Transactions

on Automatic Control, 19(6), 716–723.

Albrechtsen, D., Henriksen, B., & Faerch, C. (1980). Native speaker reactions to

learners’ spoken interlanguage. Language Learning, 30, 365-396.

Anderson-Hsieh, J., & Koehler, K. (1988). The effect of foreign accent and speaking rate

on native speaker comprehension. Language Learning, 38(4), 561-613.

Anderson-Hsieh, J., & Venkatagiri, H. (1994). Syllable duration and pausing in the

speech of Chinese ESL speakers. TESOL Quarterly, 28, 807-812.

Anderson-Hsieh, J., Johnson, R., & Koehler, K. (1992). The relationship between native

speaker judgments of nonnative pronunciation and deviance in segmentals,

prosody, and syllable structure. Language Learning, 42, 529-555.

Bailey, K. (1984). The “Foreign TA Problem.” In K.M. Bailey, F. Pialorsi, & J.

Zukowski-Faust (Eds.), Foreign Teaching Assistants in US Universities (pp. 3-

15). Washington, DC: National Association for Foreign Student Affairs.

Baker, C.L. (1979). Syntactic theory and the projection problem. Linguistic Inquiry, 10,

533-581.

223

Bamford, J., & Wilson, I. (1979). Methodological considerations and practical aspects of

the BKB sentence lists. In J. Bench & J. Bamford (Eds.), Speech-hearing tests

and the spoken language of hearing-impaired children (pp. 147-187). London:

Academic Press.

Bamgbose, A. (1998). Torn between the norms: Innovations in world Englishes. World

Englishes, 17(1), 1-14.

Bartholomae, D. (1985). Inventing the university. In M. Rose (Ed.), When a writer can’t

write (pp. 134-165). New York: Guilford Press.

Bayley, R., & Preston, D.R. (Eds.) (1996). Second language acquisition and linguistic

variation. Amsterdam: John Benjamins.

Bazerman, C. (1980). A relationship between reading and writing: The conversational

model. College English, 41, 656-661.

Bazerman, C. (1985). Physicists reading physics: Schema-laden purposes and purpose-

laden schema. Written Communication, 2, 3-23.

Beckman, M.E. (1996). The parsing of prosody. Language and Cognitive Processes,

11(1), 17-68.

Bench, J., & Bamford, J. (Eds.) (1979). Speech-hearing tests and the spoken language of

hearing-impaired children. London: Academic Press.

Bent, T., & Bradlow, A.R. (2003). The interlanguage speech intelligibility benefit.

Journal of the Acoustical Society of America, 114(3), 1600-1610.

224

Bent, T., Bradlow, A.R., & Smith, B. (2007). Phonemic errors in different word

positions and their effects on intelligibility of non-native speech: All’s well that

begins well. In O.-S. Bohn & M. J. Murray J. (Eds.), LanguageU experience in

second language speech learning: In honor of James Emil Flege (pp. 331-348).

Amsterdam: John Benjamins.

Berkenkotter, C., & Huckin, T. (1995). Genre knowledge in disciplinary communication:

cognition/culture/power. Hillsdale, NJ: Lawrence Erlbaum.

Best, C.T. (1994). The emergence of native-language phonological influences in infants:

A Perceptual Assimilation Model. In J.C. Goodman & H.C. Nusbaum (Eds.), The

development of speech perception: The transition from speech sounds to spoken

words (pp. 167-224). Cambridge, MA: MIT Press.

Best, C.T. (1995). A direct realist view of cross-language speech perception. In W.

Strange (Ed.), Speech perception and linguistic experience: Issues in cross-

language research (pp. 171-204). Timonium, MD: York Press.

Best, C.T., McRoberts,G., & Goodell, E. (2001). Discrimination of non-native consonant

contrasts varying in perceptual assimilation to the listener’s native phonological

system. Journal of the Acoustical Society of America, 109(2), 775-794.

Biber, D., & Barbieri, F. (2007). Lexical bundles in university spoken and written

registers. English for Specific Purposes, 26(3), 263-286.

225

Biber, D., Reppen, R., Clark, V., & Walter, J. (2001). Representing spoken language in

university settings: The design and construction of the spoken component of the

T2K-SWAL Corpus. In R.C. Simpson & J.M. Swales (Eds.), Corpus linguistics

on North America (pp. 48-57). Ann Arbor: University of Michigan Press.

Bilbow, G.T. (1989). Towards an understanding of overseas students’ difficulties in

lectures: A phenomenographic approach. Journal of Further and Higher

Education, 13(2), 85-99.

Billerey-Mosier, R. (2000). Segment realization in relation to a carrier word’s

uniqueness point. Journal of the Acoustical Society of America, 108(5), 2479-

2480.

Bizzell, P. (1982a). Cognition, convention, and certainty: What we need to know about

writing. PRE/TEXT, 3, 213-243.

Bizzell, P. (1982b). College composition: Initiation into the academic discourse

community. Curriculum Inquiry, 12, 191-207.

Blau, E.K. (1990). The effect of syntax, speed, and pauses on listening comprehension.

TESOL Quarterly, 24, 746- 753.

Bolton, K. (2003). Chinese Englishes: A socioliguistic history. Cambridge, UK:

Cambridge University Press.

Boyle, J.P. (1987). Sex differences in listening vocabulary. Language Learning, 37,

273-284.

Bradlow, A. R., & Bent, T. (2002). The clear speech effect for non-native listeners.

Journal of the Acoustical Society of America, 112, 272-284.

226

Bradlow, A.R., & Bent, T. (2003). Listener adaptation to foreign-accented English. In

M.J. Solo, D. Recasens, & J. Romero (Eds.), Proceedings of the XVth

International Congress of Phonetic Sciences (pp. 2881-2884). Barcelona, Spain.

Bradlow, A.R., & Bent, T. (2008). Perceptual adaptation to non-native speech.

Cognition, 106, 707-729.

Bradlow, A.R., & Pisoni, D.B. (1999). Recognition of spoken words by native and non-

native listerners: Talker-, listener-, and item-related factors. Journal of the

Acoustical Society of America, 106(4), 2074-2085.

Broselow, E., & Park, H-B. (1995). Mora conservation in second language prosody. In

J. Archibald (Ed.), Phonological acquisition and phonological theory (pp. 151-

168). Hillsdale, NJ: Lawrence Erlbaum.

Broselow, E., Chen, S-I., & Wang,C. (1998). The emergence of the unmarked in second

language phonology. Studies in Second Language Acquisition, 20, 261-280.

Brown, A. (1991). Pronunciation models. Singapore: Singapore University Press.

Brown, K. (1968). Intelligibility. In A. Davies (Ed.), Language testing symposium (pp.

180-191). Oxford: Oxford University Press.

Bunta, F. & Major, R.C. (2004). An optimality theoretic account of Hungarian ESL

learners’ acquisition of /ε/ and /æ/. International Review of Applied Linguistics,

42, 277-298.

Butler, S. (1997a). Corpus of English in Southeast Asia: implications for a regional

dictionary. In M.L.S. Bautista (Ed.), English as an Asian language: The

Philippine context. Manila: Macquarie Library.

227

Butler, S. (1997b). World English in the Asian context: Why a dictionary is important. In

L.E. Smith & M.L. Forman (Eds.), World Englishes 2000 (pp. 90-125).

Honolulu: University of Hawai’i.

Byrd, P. (1992). Funding for graduate students: An update on assistantships. Advising

Quarterly, 21(1), 12-14.

Byrd, P., & Constantinides, J.C. (1992). The language of teaching mathematics:

Implications for training ITAs. TESOL Quarterly, 26(1), 163-167.

Campbell, G.A. (1910). Telephonic intelligibility. Philosophical Magazine, 19, 152-

159.

Campbell, D.T., & Stanley, J.C. (1968). Experimental and quasi-experimental designs

for research. Boston: Houghton-Mifflin.

Canale, M. & Swain, M. (1980). Theoretical bases of communicative approaches to

second-language teaching and testing. Applied Linguistics, 1(1), 1-47.

Cargile, A.C., & Giles, H. (1998). Language attitudes towards varieties of English: An

American-Japanese context. Journal of Applied Communication Research, 26,

338-356.

Casanave, C.P. (1995). Local interactions: Constructing contexts for composing in a

graduate sociology program. In D. Belcher & G. Braine (Eds.), Academic writing

in a second language (pp. 83-110). Norwood, NJ: Ablex.

Cheng, Z. (2002). English departments in Chinese universities: Purpose and function.

World Englishes, 21(2), 257-267.

228

Chomsky, N. (1993). Lectures on government and binding: The Pisa lectures (7th ed.).

Berlin, New York: Mouton de Gruyter.

Chomsky, N., & Halle, M. (1968). The sound pattern of English. New York: Harper &

Row.

Clarke, C.M. (2000). Perceptual adjustment for foreign-accented English. Journal of the

Acoustical Society of America, 107(5), 2856.

CMU Pronouncing Dictionary. (n.d.) Retrieved from http://www.speech.cs.cmu.edu/cgi-

bin/cmudict

Conrad, L. (1989). The effects of time-compressed speech on native and EFL listening

comprehension. Studies in Second Language Acquisition, 11, 1-6.

Constantinides, J.C. (1987). The “Foreign TA Problem – an Update. NAFSA Newsletter,

38(5): 3-6.

Cook, V. (1999). Going beyond the native speaker in language teaching. TESOL

Quarterly, 33, 185-209.

Corder, S.P. (1967). The significance of learners’ errors. International Review of

Applied Linguistics, 5(4), 161-170.

Couper, G. (2003). The value of an explicit pronunciation syllabus in ESOL teaching.

Prospect, 18(3), 53-70.

Crystal, D. (2003). English as a global language (2nd ed.). Cambridge: Cambridge

University Press.

229

Date, T. (2005). The intelligibility of Singapore English from a Japanese perspective. In

D. Deterding, A. Brown, & E.L. Low (Eds.), English in Singapore: Phonetic

research on corpus (pp. 173-183). Singapore: McGraw-Hill Education.

Davis, K., & Beckman, M.E. (1983). Production and perception of the voicing contrast

in Indian and American English. Working Papers of the Cornell Phonetics

Laboratory, 1, 77-90.

Derwing, T.M. (1990). Speech rate is no simple matter: Rate adjustment and NS-NNS

communicative success. Studies in Second Language Acquisition, 12, 303-313.

Derwing, T.M., & Munro, M.J. (1997). Accent, intelligibility, and comprehensibility:

Evidence from four L1s. Studies in Second Language Acquisition, 20, 1-16.

Derwing, T.M., & Munro, M.J. (2001). What speaking rate do nonnative listeners

prefer? Applied Linguistics, 22, 324-337.

Derwing, T.M., & Munro, M.J. (2005). Second language accent and pronunciation

teaching: A research-based approach. TESOL Quarterly, 39, 379-397.

Derwing, T.M., & Rossiter, M.J. (2003). The effects of pronunciation instruction on the

accuracy, fluency, and complexity of L2 accented speech. Applied Language

Learning, 13, 1-18.

Derwing, T.M., Munro, M.J., & Wiebe, G. (1997). Pronunciation instruction for

fossilized learners: Can it help? Applied Language Learning, 8, 217-235.

Derwing, T.M., Munro, M.J., & Wiebe, G. (1998). Evidence in favor of a broad

framework for pronunciation instruction. Language Learning, 48, 393-410.

230

Derwing, T.M., Rossiter, M.J., & Munro, M.J. (2002). Teaching native speakers to listen

to foreign-accented speech. Journal of Multilingual and Multicultural

Development, 23(4), 245-259.

Deterding, D. (2005). Listening to Estuary English in Singapore. TESOL Quarterly, 39,

425-440.

Deterding, D., & Kirkpatrick, A. (2006). Emerging South-East Asian Englishes in

intelligibility. World Englishes, 25(3/4), 391-409.

Deterding, D. & Low, E.L. (2001) The NIE corpus of spoken Singapore English

(NIECSSE). SAAL Quarterly, 56, 2-5.

Deterding, D. & Low, E.L. (2005). The NIE corpus of spoken Singapore English. In D.

Deterding, A. Brown, & E.L. Low (Eds.), English in Singapore: Phonetic

research on corpus (pp. 1-13). Singapore: McGraw-Hill Education.

Donegan, P., & Stampe, C. (1979). The study of Natural Phonology. In D.A. Dinnsen

(Ed.), Current approaches to phonological theory (pp. 126-173). Bloomington:

Indiana University Press.

Dunkel, P., Henning, G., & Chaudron, C. (1993). The assessment of a listening

comprehension construct: A tentative model for test specifications and

development. Modern Language Journal, 77, 180-191.

Eckman, F.R. (2008). Typological markedness and second language phonology. In

J.G.H. Edwards & M.L. Zampini (Eds.), Phonology and second language

acquisition (pp. 95-116). Amsterdam: John Benjamins.

231

Eckman, F.R., & Iverson, G. (1994). Pronunciation difficulties in ESL: Coda consonants

in English interlanguage. In M. Yavas (Ed.), First and second language

phonology (pp. 251-265). San Diego: Singular.

Eisenstein, M.R., & Berkowitz, D. (1981). The effect of phonological variation on adult

learner comprehension. Studies in Second Language Acquisition, 4, 75-80.

Ekong, P. (1982). On the use of an indigenous model for teaching English in Nigeria.

World Language English, 1, 87-92.

Elliot, A.R. (1997). On the teaching and acquisition of pronunciation within a

communicative approach. Hispania, 80(1), 95-108.

Enrollment of international students by curriculum. (2000). Highlights of fifteenth day

enrollment for the autumn quarter 2000, The Ohio State University. Retrieved

from http://www.ureg.ohio-state.edu/ourweb/srs/srscontent/intadobe.html

Enrollment of international students by curriculum. (2001). Highlights of fifteenth day

enrollment for the autumn quarter 2001, The Ohio State University. Retrieved

from http://www.ureg.ohio-state.edu/ourweb/srs/srscontent/intadobe.html

Enrollment of international students by curriculum. (2002). Highlights of fifteenth day

enrollment for the autumn quarter 2002, The Ohio State University. Retrieved

from http://www.ureg.ohio-state.edu/ourweb/srs/srscontent/intadobe.html

Enrollment of international students by curriculum. (2003). Highlights of fifteenth day

enrollment for the autumn quarter 2003, The Ohio State University. Retrieved

from http://www.ureg.ohio-state.edu/ourweb/srs/srscontent/intadobe.html

232

Enrollment of international students by curriculum. (2004). Highlights of fifteenth day

enrollment for the autumn quarter 2004, The Ohio State University. Retrieved

from http://www.ureg.ohio-state.edu/ourweb/srs/srscontent/intadobe.html

Ensz, K.Y. (1982). French attitudes toward typical speech errors of American speakers

of French. Modern Language Journal, 64, 210-215.

Fayer, J.M., & Krasinski, E. (1987). Native and nonnative judgments of intelligibility

and irritation. Language Learning, 37, 313-326.

Ferguson, C. (1971). Absence of copula and the notion of simplicity: A study of normal

speech, baby talk, foreigner talk and pidgins. In D. Hymes (Ed.), Pidginization

and creolization of languages (pp. 141-151). Cambridge: Cambridge University

Press.

Ferris, D., & Tagg, T. (1996a). Academic oral communication needs of EAP Learners:

What subject-matter instructors actually require. TESOL Quarterly, 30, 31-58.

Ferris, D., & Tagg, T. (1996b). Academic listening/speaking tasks for ESL students:

Problems, suggestions, and implications. TESOL Quarterly, 30, 297-320.

Field, J. (2005). Intelligibility and the listener: The role of lexical stress. TESOL

Quarterly, 39, 399-423.

Firth, A. (1996). The discursive accomplishment of normality: On “lingua franca”

English and conversational analysis. Journal of Pragmatics, 26, 237-259.

Flege, J.E. (1988). Factors affecting degree of perceived foreign accent in English

sentences. Journal of the Acoustical Society of America, 84(1), 70-79.

233

Flege, J.E. (1993). Production and perception of a novel, second-language phonetic

contrast. Journal of the Acoustical Society of America, 93(3), 1589-1608.

Flege, J.E. (1995). 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). Timonium, MD: York Press.

Flege, J.E., Bohn, O. & Jang, S. (1997). Effects of experience on non-native speakers’

production and perception of English vowels. Journal of Phonetics, 25, 437-470.

Flege, J.E., MacKay, I.R.A., & Piske, T. (2002). Assessing bilingual dominance.

Applied Psycholinguistics, 23(4), 567-598.

Flege, J.E., Munro, M.J., & MacKay, I.R.A. (1995). Factors affecting strength of

perceived foreign accent in a second language. Journal of the Acoustical Society

of America, 97(5), 3125-3134.

Flowerdew, J. (1994). Research of relevance to second language lecture comprehension

– An overview. In J. Flowerdew (Ed.), Academic listening: Research

perspectives (pp. 7-29). New York: Cambridge University Press.

Fox, W.S., & Gay, G. (1994). Functions and effects of international teaching assistants.

Review of Higher Education, 18(1), 1-24.

Fries, C.C. (1945). Teaching and learning English as a foreign language. Ann Arbor:

University of Michigan Press.

Gass, S., & Varonis, E.M. (1984). The effect of familiarity on nonnative speech.

Language Learning, 34, 65-89.

234

Gill, M.M. (1994). Accent and stereotypes: Their effect on perceptions of teachers and

lecture comprehension. Journal of Applied Communication Research, 22, 348-

361.

Gimson, A.C. (1970). An introduction to the pronunciation of English (2nd ed.). London:

E. Arnold.

Gonzalez, A.B. (1983). In R.B. Noss (Ed.), Varieties of English in Southeast Asia (pp.

150-172). Singapore: Regional Language Center.

Gonzalez, A.B. (1997). Philippine English: A variety in search of legitimation. In W.

Schneider (Ed.), Englishes around the World (pp. 205-212). Amsterdam: John

Benjamins.

Grabe, W. (1990). Foreword. Annual Review of Applied Linguistics, 11, vii-xi.

Graddol, D. (1997). The future of English. London: The British Council.

Greenbaum, S. (1991). ICE: The International Corpus of English. English Today, 28, 3-

Greenberg, J. (1966). Language universals. The Hague: Mouton.

Greenberg, J. (1978). Some generalizations concerning initial and final consonant

clusters. In J. Greenberg, C. Ferguson, & E. Moravcsik (Eds.), Universals of

human language (Vol. 2, pp. 243-279). Palo Alto, CA: Stanford University Press.

Griffiths, R. (1990). Speech rate and NNS comprehension. Language Learning, 40(3),

311-336.

Griffiths, R. (1991). Pausological research in an L2 context: A rationale, and review of

selected studies. Applied Linguistics, 12, 345-364.

235

Griffiths, R. (1992). Speech rate and listening comprehension: Further evidence of the

relationship. TESOL Quarterly, 26, 385-390.

Gynan, S.N. (1985). Comprehension, irritation, and error hierarchies. Hispania, 68, 160-

165.

Haegeman, P. (2002). Foreigner talk in lingua franca business telephone calls. In K.

Knapp & C. Meierkord (Eds.), Lingua franca communication (pp. 135-162).

Frankfurt: Peter Lang.

Hahn, L.D. (1999). Native speakers’ reactions to non-native stress in English discourse.

Unpublished doctoral dissertation, University of Illinois, Urbana-Champaign.

Hahn, L.D. (2004). Primary stress and intelligibility: research to motivate the teaching of

suprasegmentals, TESOL Quarterly, 38, 201–223.

Halladay, L. (1970). A study of the effects of age on achievement in adults studying

English in an intensive course. Unpublished doctoral dissertation, University of

Michigan, Ann Arbor.

Halleck, G.B., & Moder, C.L. (1995). Testing language and teaching skills of

international teaching assistants: The limits of compensatory strategies. TESOL

Quarterly, 29, 733-758.

Halliday, M.A.K. (1978). Language as a social semiotic: The social interpretation of

language and meaning. London: Edward Arnold.

Hancin-Bhatt, B. (2000). Optimality in second language phonology: Codas in Thai ESL.

Second Language Research, 16, 201-232.

236

Hazan, V., & Markham, D. (2004). Acoustic-phonetic correlates of talker intelligibility

for adults and children. Journal of the Acoustical Society of America, 116, 3108-

3118.

Heller, S. (1985). Colleges try tests and training to make sure foreign TA’s can be

understood. Chronicle of Higher Education, 31, 32-33.

Hieke, A. (1985). A componential approach to oral fluency evaluation. Modern

Language Journal, 69, 135-142.

Hock, H.H. (1986). Principles of historical linguistics. Berlin: Mouton de Gruyter.

Hockett, C.F. (1958). A course in modern linguistics. New York: Macmillan.

Hoekje, B., & Williams, J. (1992). Communicative competence and the dilemma of

international teaching assistant education. TESOL Quarterly, 26, 243-269.

Institute of International Education. (2008). Open Doors Report 2008. Retrieved from

http://opendoors.iienetwork.org

Institute of International Education. (2009). Open Doors Report 2009. Retrieved from

http://opendoors.iienetwork.org

Jenkins, J. (1995). Variation in phonological error in interlanguage talk. Unpublished

doctoral dissertation, University of London, Institute of Education, London.

Jenkins, J. (2000). The phonology of English as an International Language. Oxford:

Oxford University Press.

Jenkins, J. (2001). Euro-English accents. English Today 68, 17(4), 16-19.

237

Jenkins, J. (2002). A sociolinguistically based, empirically researched pronunciation

syllabus for English as an International Langauge. Applied Linguistics, 23(1), 83-

103.

Jenkins, J. (2005). The ABC of ELT … ‘ELF’. IATEFL Issues, 182, 9.

Jenkins, J. (2007). English as a Lingua Franca: Attitude and identity. Oxford: Oxford

University Press.

Jenner, B. (1989). Teaching pronunciation: The common core. Speak Out!: Newsletter

of the IATEFL Pronunciation SIG, 4, 2-4.

Johansson, S. (1978). Studies of error gravity: Native reactions to errors produced by

Swedish learners of English. Gothenburg Studies in English, 44, 1-138.

Kachi, R. (2004). Factors predicting native and nonnative listeners’ evaluative reactions

to Japanese English. Unpublished doctoral dissertation, The Ohio State

University, Columbus.

Kachru, B.B. (1965). The Indianness in Indian English. Word, 21, 391-410.

Kachru, B.B. (1985). Standards, codification and sociolinguistic realism: The English

language in the Outer Circle. In R. Quirk & H.G. Widdowson (Eds.), English in

the world: Teaching and learning the language and literatures (pp. 11-30).

Cambridge: Cambridge University Press.

Kachru, B.B. (1986). The alchemy of English: The spread, functions, and models of non-

native Englishes. Oxford: Pergamon Press.

238

Kachru, B.B. (1997). Past perfect: The other side of English in Asia. In L.E. Smith &

M.L. Forman (Eds.), World Englishes 2000 (pp. 68-89). Honolulu: University of

Hawai’i.

Kachru, B.B., & Nelson, C.L. (1996). World Englishes. In S.L. McKay & N.H.

Hornberger (Eds.), Sociolinguistics and Language Teaching (pp. 71-102).

Cambridge: Cambridge University Press.

Kim, H-K. (2001). The interlanguage phonology of Korean learners of English: A

computational implementation of Optimality Theoretic constraints. Unpublished

doctoral dissertation, Georgetown University, Washington, DC.

Kim, S. (2006). Academic oral communication needs of East Asian international graduate

students in non-science and non-engineering fields. English for Specific Purposes,

25, 479-489.

Kirkpatrick, A., & Saunders, N. (2005). The intelligibility of Singaporean English: A

case study in an Australian university. In D. Deterding, A. Brown, & E.L. Low

(Eds.), English in Singapore: Phonetic research on corpus (pp. 153-162).

Singapore: McGraw-Hill Education.

Koster , C.J., & Koet, T. (1993). The evaluation of accent in the English of Dutchmen.

Language Learning, 43, 69-92.

Kubota, R. (2001). Teaching World Englishes to native speakers of English: A pilot

project in a high school class. World Englishes, 20 (1), 47-64.

Kucera, F., & Francis, W. (1967). Computational Analysis of Present Day American

English. Providence, RI: Brown University Press.

239

Kuhl, P.K. (1991). Human adults and human infants show a “perceptual magnet effect”:

for the prototypes of speech categories, monkeys do not. Perception &

Psychophysics, 50, 93-107.

Kuhl, P.K. (1993). Innate predispositions and the effects of experience in speech

perception: The native magnet theory. In B. de Boysson-Bardies, S. de Schonen,

P. Jusczyk, P. McNeilage, & J. Morton (Eds.), Developmental neurocognition:

Speech and face processing in the first year of life (pp. 259-274). Dordrecht:

Kluwer.

Kuhn, T.S. (1970). The structure of scientific revolutions (2nd ed.). Chicago: University

of Chicago Press.

Labov, W. (1966). The social stratification of English in New York City. Washington

DC: Center for Applied Linguistics.

Labov, W. (1972). Language in the inner city: Studies in the Black English vernacular.

Philadelphia: University of Pennsylvania Press.

Ladefoged, P. (1993). A course in phonetics (3rd ed.). New York: Harcourt Brace

College Publishers.

Lado, R. (1957). Linguistics across cultures. Ann Arbor: University of Michigan Press.

Lam, A. (2002). English in education in China: Policy changes and learners’

experiences. World Englishes, 21(2), 245-256.

Lane, H. (1963). Foreign accent and speech distortion. Journal of the Accoustical

Society of America, 35, 451-453.

240

Leather, J. (1999). Second-language speech research: An introduction. In J. Leather

(Ed.), Phonological issues in language learning (pp. 1-56). Oxford: Blackwell.

Lenneberg, E. (1967). Biological foundations of language. New York: Wiley and Sons.

Levis, J.M. (2006). Pronunciation and the assessment of spoken language. In R. Hughes

(Ed.), Spoken English, TESOL and applied linguistics (pp. 245-270). New York:

Palgrave Macmillan.

Lewellen, M.J., Goldinger, S.D., Pisoni, D.B., & Greene, B. G. (1993). Lexical

familiarity and processing efficiency: Individual differences in naming, lexical

decision and semantic categorization. Journal of Experimental Psychology:

General, 122, 316-330.

Liu, J. (1996). Perceptions of selected international graduate students towards oral

classroom participation in their academic content courses in a U.S. university.

Unpublished doctoral dissertation, The Ohio State University, Columbus.

Liu, J. (2001). Asian students’ classroom communication patterns in U.S. universities:

An emic perspective. Westport, CT: Ablex.

Lively, S.E. & Pisoni, D.B. (1997). On prototypes and phonetic categories: A critical

assessment of the perceptual magnet effect in speech perception. Journal of

Experimental Psychology, 23(6), 1665-1679.

Llamzon, T.A. (1983). Essential features of new varieties of English. In R.B.Noss (Ed.),

Varieties of English in Southeast Asia (pp. 92-109). Singapore: Regional

Language Centre.

241

Long, M.H. (1990). Maturational constraints on language development. Studies in

Second Language Acquisition, 12, 251-285.

Lucas, C.J., & Murray, J.M. (2002). New faculty: A practical guide for academic

beginners. New York: MacMillan.

Macdonald, D., Yule, G., & Powers, M. (1994). Attempts to improve English L2

pronunciation: The variable effects of different types of instruction. Language

Learning, 44, 75-100.

Major, R.C. (1987). The Natural Phonology of second language acquisition. In A.James

& J. Leather (Eds.), Sound patterns in second language acquisition (pp. 207-224).

Dordrecht: Foris.

Major, R.C. (2001). Foreign accent: The ontogeny and phylogeny of second language

phonology. Mahwah, NJ: Lawrence Erlbaum.

Major, R.C. (2008). Transfer in second language phonology: A review. In J.G.H.

Edwards & M.L. Zampini (Eds.), Phonology and second language acquisition

(pp. 63-94). Amsterdam: John Benjamins.

Major, R.C., & Kim, E. (1996). The Similarity Differential Rate Hypothesis. Language

Learning, 46, 465-496.

Major, R.C., Fitzmaurice, S.F., Bunta, F., & Balasubramanian, C. (2002). The effects of

nonnative accents on listening comprehension: Implications for ESL assessment.

TESOL Quarterly, 36(2), 173-190.

242

Major, R.C., Fitzmaurice, S.M., Bunta, F., & Balasubramanian, C. (2005). Testing the

effects of regional, ethnic, and international dialects of English on listening

comprehension. Language Learning, 55, 37-69.

Markham, P. (1988). Gender differences and the perceived expertness of the speaker as

factors in ESL listening recall. TESOL Quarterly, 22, 397-406.

Martohardjono, G., & Flynn, S. (1995). Language transfer: What do we really mean? In

L. Eubank, L. Selinker, & M. Sharwood Smith (Eds.), The current state of

interlanguage: Studies in honor of William E. Rutherford (pp. 205-218).

Amsterdam: John Benjamins.

Marslen-Wilson, W., & Zwitserlood, P. (1989). Accessing spoken words: The

importance of word onsets. Journal of Experimental Psychology: Human

Perception and Performance, 15(3), 576-585.

Mason, A. (1995). By dint of: Student and lecturer perceptions of lecture comprehension

strategies in first-term graduate study. In J. Flowerdew (Ed.), Academic listening:

Research perspectives (pp. 199-218). Cambridge: Cambridge University Press.

Matsuda, A., & Matsuda, P. K. (2001). Autonomy and collaboration in teacher

education: Journal sharing among native and nonnative English-speaking

teachers. The CATESOL Journal, 13(1), 109-121.

Mauranen, A. (2003). The corpus of English as Lingua Franca in academic settings.

TESOL Quarterly, 37, 513-527.

243

McAllister, R. (1997). Perceptual foreign accent: L2 users’ comprehension ability. In A.

James & J. Leather (Eds.), Second-language speech: Structure and process (pp.

119-132). Berlin: Mouton de Gruyter.

Meierkord., C. (2004). Syntactic variation in interactions across international Englishes.

English World-Wide, 25(1), 109-132.

Mestenhauser, J., Perry, W., Paige, M., Landa, M., Brutsch, S., Dege, D., Doyle, K.,

Gillette, S., Hughes, G., Judy, R., Keye, Z., Murphy, K., Smith, J., Vanderluis, K.,

& Wendt, J. (1980). Report of a special course for foreign student teaching

assistants to improve their classroom effectiveness. Minneapolis: University of

Minnesota.

Meyers, C., & Jones, T.B. (1993). Promoting active learning: Strategies for the college

classroom. San Francisco: Jossey-Bass.

Michaels, D. (1974). Sound replacements and phonological systems. Linguistics, 176,

69-81.

Modiano, M. (2001). A new variety of English. English Today 68, 17(4), 13-14.

Morley, J. (1991). The pronunciation component in Teaching English to Speakers of

Other Languages. TESOL Quarterly, 25, 481-520.

Moulton, W.G. (1962). Toward a classification of pronunciation errors. Modern

Language Journal, 46(3), 101-109.

MRC Psycholinguistic Database. (n.d.) Retrieved from

http://www.psy.uwa.edu.au/mrcdatabase/uwa_mrc.htm

244

Munro, M.J. (1998). The effects of noise on the intelligibility of foreign-accented

speech. Studies in Second Language Acquisition, 20, 139-154.

Munro, M.J. (2008). Foreign accent and speech intelligibility. In J.G.H. Edwards &

M.L. Zampini (Eds.), Phonology and second language acquisition (pp. 193-218).

Amsterdam: John Benjamins.

Munro, M.J., & Derwing, T.M. (1995a). Foreign accent, comprehensibility, and

intelligibility in the speech of second language learners. Language Learning,

45(1), 73-97.

Munro, M.J., & Derwing, T.M. (1995b). Processing time, accent, and comprehensibility

in the perception of native and foreign-accented speech. Language and Speech,

38(3), 289-306.

Munro, M.J., & Derwing, T.M. (1998). The effects of speaking rate on listener

evaluations of native and foreign-accented speech. Language Learning, 48(2),

159-182.

Munro, M.J., Derwing, T.M., & Morton, S.L. (2006). The mutual intelligibility of L2

speech. Studies in Second Language Acquisition, 28, 111-131.

Myers, S.A. (1995). Using written text to teach oral skills: An ITA traning class using

field-specific materials. English for Specific Purposes, 14(3), 231-245.

Nababan, P.W.J. (1983). In R.B. Noss (Ed.), Varieties of English in Southeast Asia (pp.

113-124). Singapore: Regional Language Center.

Nelson, G.L. (1990, March). International teaching assistants: A review of research.

Paper presented at the 24th Annual TESOL Convention, San Francisco, CA.

245

Nelson, G.L. (1991). Effective teaching behavior for international teaching assistants. In

J. D. Nyquist, R. D. Abbott, D. H. Wulff, & J. Sprague (Eds.), Preparing the

professorate of tomorrow to teach (pp. 427-434). Dubuque, IA: Kendall Hunt.

Nusbaum, H.C., Pisoni, D.B., & Davis, C.K. (1984). Sizing up the Hoosier mental

lexicon: Measuring the familiarity of 20,000 words. In H.C. Nusbaum, D.B.

Pisoni, & C.K. Davis (Eds.), Research in speech perception, progress report 10

(pp. 357-376). Bloomington, IN: Indiana University.

Odlin,T. (1989). Language transfer. Cambridge: Cambridge University Press.

Odlin,T. (2003). Cross-linguistic influence. In C.J. Doughty & M.H. Long (Eds.), The

handbook of second language acquisition (pp. 436-486). Malden, MA:

Blackwell.

Ohio Revised Code 3345.281 (1986). Program to assess oral English language

proficiency of teaching assistants providing classroom instruction. Retrieved

from http://codes.ohio.gov/orc/3345

Oller, J.W., & Ziahosseiny, S.M. (1970). The contrastive analysis hypothesis and

spelling errors. Language Learning, 20, 183-189.

Ortega-Llebaría, M. (1997). An explanatory intelligibility test for Spanish-accented

English. Unpublished doctoral dissertation, Indiana University, Bloomington.

Orth, J.L. (1982). Evaluational reactions to spoken language: A dilemma for the teaching

and testing of speaking proficiency. Journal of the Linguistic Association of the

Southwest, 5, 216-233.

246

Ostler, S.E. (1980). A survey of academic needs for advanced ESL. TESOL Quarterly,

14, 489-502.

Palmer, L.A. (1973). A preliminary report on a study of the linguistic correlates of raters’

subjective judgements of non-native English speech. In R. W. Shuy & R. W.

Fasold (Eds.), Language attitudes: Current trends and prospects (pp. 41-59).

Washington, D.C.: Georgetown University Press.

Perlmutter, M. (1989). Intelligibility rating of L2 speech pre- and post-intervention.

Perceptual and Motor Skills, 68, 515-521.

Phillipson, R. (1992). Linguistic imperialism. Oxford: Oxford University Press.

Pickering, L. (2000). The analysis of prosodic systems in the classroom discourse of

native speaker and nonnative speaker teaching assistants. Unpublished doctoral

dissertation, University of Florida, Gainesville.

Pickering, L. (2001). The role of tone choice in improving ITA communication in the

classroom. TESOL Quarterly, 35, 233-255.

Pickering, L. (2004). The structure and function of intonational paragraphs in native and

nonnative speaker instructional discourse. English for Specific Purposes, 23, 19-

43.

Pihko, M. (1997). "His English sounded strange": The intelligibility of native and non-

native English pronunciation to Finnish learners of English. Jyväskylä, Finland:

Center for Applied Language Studies.

Piske, T., MacKay, I.R.A., & Flege, J.E. (2001). Factors affecting degree of foreign

accent in an L2: A review. Journal of Phonetics, 29, 191-215.

247

Pisoni, D.B. (1996). Word identification in noise. Language and Cognitive Processes,

11(6), 681-687.

Plakans, B.S. (1997). Undergraduates’ experiences with and attitudes toward ITAs.

TESOL Quarterly, 31, 95-119.

Platt, J., Weber, H., & Ho, M.L. (1984). The New Englishes. London: Routledge &

Kegan Paul.

Politzer, R.L. (1978). Errors of English speakers of German as perceived and evaluated

by German natives. Modern Language Journal, 62, 253-261.

Pride, J. (1982). New Englishes. Rowley, MA: Newbury House.

Prince, A., & Smolensky, P. (1997). Optimality: From neural networks to Universal

Grammar. Science, 275, 1604-1610.

Prince, A., & Smolensky, P. (2004). Optimality Theory: Constraint interaction in

generative grammar. Malden, MA: Blackwell

Radeau, M., Mousty, P., & Bertelson, P. (1989). The effect of the uniqueness point in

spoken-word recognition. Psychological Research, 51(3), 123-128.

Rajadurai, J. (2007). Intelligibility studies: A consideration of empirical and ideological

issues. World Englishes, 26, 87-98.

Richards, J.C. (1983). Listening comprehension: Approach, design, procedure. TESOL

Quarterly, 17, 219-239.

Rochet, B. (1995). Perception and production of L2 speech sounds by adults. In W.

Strange (Ed.), Speech perception and linguistic experience: Theoretical and

methodological issues (pp. 379-410). Timonium, MD: York Press.

248

Rogers, C.L. (1997). Intelligibility of Chinese-accented English. Unpublished doctoral

dissertation, Indiana University, Bloomington.

Rogers, C.L. (1998). Effects of listener training on intelligibility of Chinese-accented

English. Journal of the Acoustical Society of America, 103(5), 3090.

Rogers, C.L., Dalby, J., & Nishi, K. (2004). Effects of noise and proficiency on

intelligibility of Chinese-accented English. Language and Speech, 47(2), 139-

154.

Rubin, D.L. (1992). Nonlanguage factors affecting undergraduates' judgments of non-

native English speaking teaching assistants. Research in Higher Education, 33,

511-531.

Rubin, D.L., & Smith, K.A. (1990). Effects of accent, ethnicity, and lecture topic on

undergraduates’ perceptions of nonnative English-speaking teaching assistants.

International Journal of Intercultural Relations, 14(3), 337-353.

Rubin, J. (1994). A review of second language listening comprehension research.

Modern Language Journal, 78(2), 199-221.

Saville-Troike, M. (1981). Ethnography of communication: An introduction. Oxford:

Basil Blackwell.

Schairer, K. (1992). Native speaker reaction to non-native speech. Modern Language

Journal, 76, 309-319.

Schmidt-Rinehart, B. (1994). The effects of topic familiarity on second language

listening comprehension. Modern Language Journal, 78, 179-189.

249

Schneider, W., Eschman, A., & Zuccolotto, A. (2002). E-Prime Reference Guide.

Pittsburgh: Psychology Software Tools, Inc.

Scovel, T. (1969). Foreign accents, language acquisition and cerebral dominance.

Language Learning, 19, 245-253.

Seidlhofer, B. (2001a). Toward making ‘Euro-English’ a linguistic reality. English

Today 68, 17(4), 14-16.

Seidlhofer, B. (2001b). Closing a conceptual gap: The case for a description of English

as a lingua franca. International Journal of Applied Linguistics, 11(2), 133-158.

Seidlhofer, B. (2004). Research perspectives on teaching English as a Lingua Franca.

ARAL, 24, 209-239.

Selinker, L. (1972). Interlanguage. IRAL, 10, 201-231.

Selinker, L. (1992). Rediscovering interlanguage. London: Longman.

Seright, L. (1985). Age and aural comprehension achievement in Francophone adults.

TESOL Quarterly, 19, 455-473.

Simon, H.K. (1957). Models of Man: Social and rational. New York: Wiley.

Simpson, R.C., Briggs, S.L., Ovens, J., & Swales, J.M. (1999). The Michigan Corpus of

Academic Spoken English. Ann Arbor: The Regents of the University of

Michigan.

Smith, J. (1994). Enhancing curricula for ITA development. In C.G. Madden & C.L.

Myers (Eds.), Discourse and performance of International Teaching Assistants

(pp. 52-62), Alexandria, VA: TESOL.

250

Smith, L.E. (1992). Spread of English and issues of intelligibility. In B.B. Kachru (Ed.),

The other tongue: English across cultures (2nd ed., pp. 75-90). Urbana:

University of Illinois Press.

Smith, L.E., & Bisazza, J.A. (1982). The comprehensibility of three varieties of English

for college students in seven countries. Language Learning, 32, 259-269.

Smith, L.E., & Nelson, C. (1985). International intelligibility of English: Directions and

resources. World Englishes, 4(3), 333-342.

Solomon, B.B. (1991). Teaching an increasingly diverse undergraduate population. In J.

D. Nyquist, R. D. Abbott, D. H. Wulff, & J. Sprague (Eds.), Preparing the

professorate of tomorrow to teach (pp. 55-64). Dubuque, IA: Kendall Hunt.

Spack, R. (1988). Initiating ESL students into the academic discourse community: How

far should we go? TESOL Quarterly, 22, 29-51.

Stampe, D. (1969). The acquisition of phonetic representation. Papers from the Fifth

Regional Meeting of the Chicago Linguistic Society, 443-453.

Stampe, D. (1979). A dissertation in Natural Phonology. New York: Garland.

Strange, W., & Shafer, V.L. (2008). Speech perception in second language learners: The

re-education of selective perception. In J.G.H. Edwards & M.L. Zampini (Eds.),

Phonology and second language acquisition (pp. 153-192). Amsterdam: John

Benjamins.

Sukwiwat, M. (1983). In R.B. Noss (Ed.), Varieties of English in Southeast Asia (pp.

190-210). Singapore: Regional Language Center.

251

Swales, J.M. (1990). Genre analysis: English in academic and research settings.

Cambridge: Cambridge University Press.

Tarone, E. (1979). Interlanguage as chameleon. Language Learning, 29, 181-191.

Tarone, E. (1983). On the variability of interlanguage systems. Applied Linguistics, 4,

141-163.

Tarone, E. (1988). Variation in interlanguage. London: Edward Arnold.

Tauroza, S., & Allison, D. (1990). Speech rates in British English. Applied Linguistics,

11, 90-105.

Tauroza, S., & Luk, J. (1997). Accent and second language listening comprehension.

RELC Journal, 28, 54-71.

Tay, M.W.J., & Gupta, A.F. (1983). In R.B. Noss (Ed.), Varieties of English in Southeast

Asia (pp. 173-189). Singapore: Regional Language Center.

Thompson, I. (1995). Assessment of second/foreign language listening comprehension.

In D.J. Mendelsohn & J. Rubin (Eds.), A guide for the teaching of second

language listening (pp. 31-58). San Diego: Dominie Press.

Van Wijngaarden, S.J., Steeneken, H.J.M., & Houtgast, T. (2002). Quantifying the

intelligibility of speech in noise for non-native listeners. Journal of the

Acoustical Society of America, 111(4), 1906-1916.

Verschueren, J. (1995). The pragmatic perspective. In J. Verschueren, J.O. Ostman, & J.

Blommaert (Eds.), Handbook of pragmatics: Manual (pp. 1-19). Amsterdam:

John Benjamins.

252

Wang, X., & Munro, M.J. (2004). Computer-based training for learning English vowel

contrasts. System, 32, 539-552.

Wang, Y. (1999). ‘College English’ in China. English Today 57, 15(1), 45-51.

Wang, Y., Jongman, A., & Sereno, J. (2003). Acoustic and perceptual evaluation of

Mandarin tone productions before and after training. Journal of the Acoustical

Society of America, 113, 1033-1043.

Wardhaugh, R. (1970). The contrastive analysis hypothesis. TESOL Quarterly, 4, 123-

130.

Weil, S.A. (2003). The impact of perceptual dissimilarity on the perception of foreign

accented speech. Unpublished doctoral dissertation, The Ohio State University,

Columbus.

Weinrich, U. (1953/1968). Languages in contact. The Hague: Mouton.

West, M. (1953). General service list of English words. Harlow: Longman.

Widdowson, H.G. (1994). The ownership of English. TESOL Quarterly, 28, 377-389.

Wilcox, G.K. (1978). The effect of accent on listening comprehension: A Singapore

study. English Language Teaching Journal, 32, 118-127.

Wong, I.F.H. (1983). In R.B. Noss (Ed.), Varieties of English in Southeast Asia (pp. 125-

149). Singapore: Regional Language Center.

Wong, W. (2001). Modality and attention to meaning and form in the input. Studies in

Second Language Acquisition, 23, 345-368.

253

Xi, X. (2007). Validating the Use of TOEFL® iBT Speaking Section Scores for ITA

Screening and Setting Standards for ITAs. Language Assessment Quarterly, 4(4),

318-351.

Yano, Y. (2001). World Englishes in 2000 and beyond. World Englishes, 20, 119-131.

Zhao, Y. (1997). The effects of listeners’ control of speech rate on second language

comprehension. Applied Linguistics, 18, 49-68.

254

APPENDIX A: QUESTIONNAIRE FOR AMERICAN SUBJECTS

255

QUESTIONNAIRE FOR AMERICAN SUBJECTS

[Subject Number ______]

Please answer the following questions about yourself, your language experiences, and your language use. Check (√) the box, or fill in the space provided.

(1) Do you have any speech impediments? NO YES (2) Would you describe your hearing as normal? NO YES (3) Date of birth: ______(4) Current age: ______years (5) Sex: MALE FEMALE (6) Major(s) of study ______(7) Are you a native speaker of American English? (i.e., was English your first language learned and/or did you live in the United States between the ages of 2 and 10?) NO YES (8) In which city(ies) and state(s) did you live between the ages of 2-10? City ______State ______City ______State ______(9) Which city and state are your parents from? Mother: City ______State ______Father: City ______State ______(10) If you were not raised by both parents, with whom did you grow up and which city and state were they from? Caregiver(s) ______(Please specify relationship) City ______State ______

256

(11) In a typical week during the past 12 months, in what contexts did you hear Chinese-accented English and for how many hours in each context? _____ In class (as a student) _____ In class or lab (as a TA) _____ Outside class with native English-speaking professors/students _____ Socially with native English-speaking friends _____ Outside class with nonnative English-speaking profs/students _____ Socially with nonnative English-speaking friends _____ Business transactions (e.g., banking, ordering in a restaurant) _____ Medical or dental appointments _____ Other activities (Please specify)

(12) In a typical week during the past 12 months, for how many hours did you hear Chinese-accented English spoken, in total, by international students, faculty, or others on or off campus? ______hours minutes (13) In a typical week during the past 12 months, with what frequency did you hear Chinese-accented English spoken by international students, faculty, or others on or off campus? _____ 7 days/week, many times/day _____ 4 days/week, once a day _____ 7 days/week, a 2-3 times/day _____ 3 days/week, many times/day _____ 7 days/week, once a day _____ 3 days/week, a 2-3 times/day _____ 6 days/week, many times/day _____ 3 days/week, once a day _____ 6 days/week, a 2-3 times/day _____ 2 days/week, many times/day _____ 6 days/week, once a day _____ 2 days/week, a 2-3 times/day _____ 5 days/week, many times/day _____ 2 days/week, once a day _____ 5 days/week, a 2-3 times/day _____ 1 days/week, many times/day _____ 5 days/week, once a day _____ 1 days/week, a 2-3 times/day _____ 4 days/week, many times/day _____ 1 day/week, once a day _____ 4 days/week, a 2-3 times/day _____ Never

257

(14) In a typical week during the past 12 months, in what contexts did you hear accented English (EXCLUDING Chinese-accented Engish) and for how many hours in each context? _____ In class (as a student) _____ In class or lab (as a TA) _____ Outside class with native English-speaking professors/students _____ Socially with native English-speaking friends _____ Outside class with nonnative English-speaking profs/students _____ Socially with nonnative English-speaking friends _____ Business transactions (e.g., banking, ordering in a restaurant) _____ Medical or dental appointments _____ Other activities (Please specify)

(15) In a typical week during the past 12 months, for how many hours did you hear Chinese-accented English spoken, in total, by international students, faculty, or others on or off campus? ______hours minutes (16) In a typical week during the past 12 months, with what frequency did you hear Chinese-accented English spoken by international students, faculty, or others on or off campus? _____ 7 days/week, many times/day _____ 3 days/week, many times/day _____ 7 days/week, a 2-3 times/day _____ 3 days/week, a 2-3 times/day _____ 7 days/week, once a day _____ 3 days/week, once a day _____ 6 days/week, many times/day _____ 2 days/week, many times/day _____ 6 days/week, a 2-3 times/day _____ 2 days/week, a 2-3 times/day _____ 6 days/week, once a day _____ 2 days/week, once a day _____ 5 days/week, many times/day _____ 1 days/week, many times/day _____ 5 days/week, a 2-3 times/day _____ 1 days/week, a 2-3 times/day _____ 5 days/week, once a day _____ 1 day/week, once a day _____ 4 days/week, many times/day _____ Never _____ 4 days/week, a 2-3 times/day _____ 4 days/week, once a day

258

(17) In a typical week during the past 12 months, for how many hours did you hear foreign-accented English, spoken by international students, faculty, or others on or off campus? (NOT including Chinese-accented English) ______hours minutes (18) Please give further details regarding your exposure to accented English if the two questions above did not fully capture your experiences. ______

(19) Are you proficient, or fluent, in any language(s) besides English? NO YES (Please specify) (20) If you answered “yes” to the previous question, for how many hours do you speak this other language in a typical week? ______hours

(21) How much time have you spent abroad in the past 12 months? ______months weeks

(22) Where was this time spent? City ______Country ______(23) How much time have you spent abroad in your life altogether? ______years months

(24) Where was this time spent? City ______Country ______City ______Country ______

This concludes the questionnaire. Thank you very much for your participation!

259

APPENDIX B: QUESTIONNAIRE FOR INTERNATIONAL SUBJECTS

260

QUESTIONNAIRE FOR INTERNATIONAL SUBJECTS

[Subject Number ______]

Please answer the following questions about yourself, your language experiences, and your language use. Check (√) the box or fill in the space provided, as indicated.

(1) Do you have any speech impediments or difficulty pronouncing your native language? NO YES (2) Would you describe your hearing as normal? NO YES (3) Date of birth: ______Current age: ______years (4) Sex: MALE FEMALE (5) Major(s) of study ______(6) At what age did you first start studying English? _____ years old (7) How long have you been studying English altogether? _____ years _____ months (8) What were your English teachers’ native languages? School level (e.g., elem.) Teacher’s native language ______

(9) If you have ever taught in English, please describe the extent and nature of your teaching experiences below: Total months/years Description of teaching duties

(10) What dialect of English did you study in school? Check all that apply. American English British English Other English (please specify) ______261

(11) If you checked more than one box, please describe these educational experiences in detail below. ______(12) At what age did you first arrive in the United States? _____ years old (13) How long have you been in the US or other English-speaking countries altogether? ______years months (14) In what city, state/province, and country did you live between the ages of 2 and 10? City ______State/Province ______Country ______(15) What city, state/province, and country are your parents from? Mother: City______State/Province ______Country ______Father: City______State/Province ______Country ______(16) If you did not live with both parents between the ages of 2 and 10, with whom did you live? Caregiver(s) ______(Please specify relationship) (17) If you identified a caregiver in the previous question, where was your caregiver from? City ______State/Province ______Country ______(18) What is your native language? Please choose ONE. Mandarin Chinese Korean Hindi Other (Please specify) For questions 19-25, please insert the letter of the category that best describes the hours you have spent during a typical week in the past 12 months: A) 0-1.99 hrs/wk D) 6-7.99 hrs/wk G) 12-13.99 hrs/wk B) 2-3.99 hrs/wk E) 8-9.99 hrs/wk H) 14-15.99 hrs/wk C) 4-5.99 hrs/wk F) 10-11.99 hrs/wk I) 16-17.99 hrs/wk* *If 18 hrs/wk or more, please specify the exact number of hours spent.

262

(19) _____How many total hours did you speak English? _____ In class (as a student) _____ In class or lab (as a TA) _____ Outside class with native English-speaking professors/students _____ Socially with native English-speaking friends _____ Outside class with nonnative English-speaking profs/students _____ Socially with nonnative English-speaking friends _____ Business transactions (e.g., banking, ordering in a restaurant) _____ Medical or dental appointments _____ Other activities (Please specify) (20) _____How many total hours did you listen to English? _____ Television _____ Movies/videos _____ Music/radio _____ Internet news or entertainment _____ Plays/public events _____ Academic classes _____ Other activities (Please specify) (21) _____How many total hours did you speak/listen in your native language? _____ Outside class with professors/students in your field _____ Socially with friends here in Columbus _____ With family members here in Columbus _____ On the phone or internet with friends (outside Col.) _____ On the phone or internet with family members (outside Col.) _____ Other activities (Please specify) (22) _____If you are proficient, or fluent, in any other language(s) besides your native language and English, how many hours did you speak it/them? Other language(s) (Please specify)

263

(23) _____ How many hours did you hear foreign-accented English (NOT including Chinese-accented English, spoken on or off campus)?

(24) _____ How many hours did you hear Chinese-accented English (spoken on or off campus)?

(25) Please give further details regarding your exposure to accented English if the two questions above did not fully capture your experiences. ______

(26) Please provide your best TOEFL exam score and circle its format ______CBT, PBT, or iBT

(27) If you are NOT certified to teach at the Ohio State University, please provide your iBT TOEFL speaking section score. ______

(28) If you ARE certified to teach at the Ohio State University, please provide the score and test date of your most recent SPEAK test. ______Score Date (or Qtr/Yr)

(29) If you have passed the Mock Teaching Test, when did you pass? ______Date (or Qtr/Yr)

This concludes the questionnaire. Thank you very much for your participation!

264

Appendix C: Recruitment Letter to Mandarin Talkers

Jocelyn Hardman, Ph.D. candidate Foreign and Second Language Education The Ohio State University (h) 937-767-2200 Email: [email protected]

Dear OSU Student:

I am a doctoral student in Foreign and Second Language Education and am writing to seek your help in my research. Mainly, I need to randomly select native speakers of Beijing Mandarin Chinese willing to read a short sentence list in English and in Chinese, which I will record.

Are you from Beijing, Hebei province, or northern China? If not, would you please let me know so I can remove you from my contact list? If you are from this area, would you be willing to participate in my research? It will take only one hour of your time, for which you will receive $20.00. In either case, please fill out the enclosed form and return it to me in the campus mail envelope provided. Please accept the enclosed gift certificate as a token of my appreciation for reading this letter, considering my request, and returning the response form.

For my dissertation research, I am studying the English pronunciation of Chinese TAs and investigating their intelligibility to other graduate students, including Chinese, Indian, Korean, and American students. The goal of my research is to develop an interactive pronunciation computer program for English language teaching purposes to advance pronunciation teaching methods and better prepare international students to become qualified as TAs at American universities.

Again, please fill out and return the enclosed form to me, or respond via email, at your earliest convenience.

Thank you for your time, and I look forward to meeting you!

265

Appendix D: Recruitment Letter to American English Talkers

Jocelyn Hardman, Ph.D. candidate Foreign and Second Language Education The Ohio State University (h) 937-767-2200 Email: [email protected]

Dear OSU Student:

I am a doctoral student in Foreign and Second Language Education and am writing to seek your help in my research. Mainly, I need to randomly select native speakers willing to read a short sentence list in English which I will record.

Are you a native of the greater Columbus area or of central Ohio? If not, would you please let me know so I can remove you from my contact list? If you are from this area, would you be willing to participate in my research? It will take only one hour of your time, for which you will receive $20.00. In either case, please fill out the enclosed form and return it to me in the campus mail envelope provided. Please accept the enclosed gift certificate as a token of my appreciation for reading this letter, considering my request, and returning the response form.

For my dissertation research, I am studying the English pronunciation of Chinese and American graduate students and investigating their intelligibility to other graduate students, including Chinese, Indian, Korean, and American students. The goal of my research is to develop an interactive pronunciation computer program for English language teaching purposes to advance pronunciation teaching methods and better prepare international students to become qualified as TAs at American universities.

Again, please fill out and return the enclosed form to me, or respond via email, at your earliest convenience.

Thank you for your time, and I look forward to meeting you!

266

Appendix E: Recruitment Letter to Mandarin Listeners

Jocelyn Hardman, Ph.D. candidate Foreign and Second Language Education The Ohio State University (h) 937-767-2200 Email: [email protected] Dear OSU Student:

I am a doctoral student in Foreign and Second Language Education and am writing to seek your help in my research. Mainly, I need to randomly select native speakers of Beijing Mandarin Chinese willing to listen to and repeat a short sentence list in English, and then read sentences in English and Chinese, which I will record.

For my dissertation research, I am studying the English pronunciation of Chinese and American graduate students and investigating their intelligibility to other graduate students, including Chinese, Indian, Korean, and American students. My goal is to provide an analysis of TA-certified Chinese-accented English, as well as rate its intelligibility and comprehensibility to many international and American listeners. With the results of this study, English language teachers can improve their pronunciation teaching methods and better prepare Chinese students not only to qualify as TAs at American universities, but also to be more intelligible to their classmates in graduate school.

Again, please fill out and return the enclosed form to me at your earliest convenience.

Thank you for your time, and I look forward to meeting you!

267

Appendix F: Recruitment Letter to Hindi Listeners

Jocelyn Hardman, Ph.D. candidate Foreign and Second Language Education The Ohio State University (h) 937-767-2200 Email: [email protected] Dear OSU Student:

I am a doctoral student in Foreign and Second Language Education and am writing to seek your help in my research. Mainly, I need to randomly select native speakers of Hindi willing to listen to and repeat a short sentence list in English, and then read sentences in English and Hindi, which I will record.

Are you a native speaker of Hindi? If not, would you please let me know so I can remove you from my contact list? If you are a native Hindi speaker, would you be willing to participate in my research? It will take only two hours of your time, for which you will receive $20.00. Please fill out the enclosed form and return it to me in the envelope provided, or respond via email if you prefer.

For my dissertation research, I am studying the English pronunciation of Chinese and American graduate students and investigating their intelligibility to other graduate students, including Chinese, Indian, Korean, and American students. My goal is to provide an analysis of TA-certified Chinese-accented English, as well as rate its intelligibility and comprehensibility to many international and American listeners. With the results of this study, English language teachers can improve their pronunciation teaching methods and better prepare Chinese students not only to qualify as TAs at American universities, but also to be more intelligible to their classmates in graduate school.

Again, please fill out and return the enclosed form to me at your earliest convenience.

Thank you for your time, and I look forward to meeting you!

268

Appendix G: Recruitment Letter to Korean Listeners

Jocelyn Hardman, Ph.D. candidate Foreign and Second Language Education The Ohio State University (h) 937-767-2200 Email: [email protected] Dear OSU Student:

I am a doctoral student in Foreign and Second Language Education and am writing to seek your help in my research. Mainly, I need to randomly select native speakers of Korean willing to listen to and repeat a short sentence list in English, and then read sentences in English and Korean, which I will record.

Are you from the Seoul region of South Korea? If not, would you please let me know so I can remove you from my contact list? If you are from this area, would you be willing to participate in my research? It will take only two hours of your time, for which you will receive $20.00. Please fill out the enclosed form and return it to me in the envelope provided.

For my dissertation research, I am studying the English pronunciation of Chinese and American graduate students and investigating their intelligibility to other graduate students, including Chinese, Indian, Korean, and American students. My goal is to provide an analysis of TA-certified Chinese-accented English, as well as rate its intelligibility and comprehensibility to many international and American listeners. With the results of this study, English language teachers can improve their pronunciation teaching methods and better prepare Chinese students not only to qualify as TAs at American universities, but also to be more intelligible to their classmates in graduate school.

Again, please fill out and return the enclosed form to me at your earliest convenience.

Thank you for your time, and I look forward to meeting you!

269

Appendix H: Recruitment Letter to American English Listeners

Jocelyn Hardman, Ph.D. candidate Foreign and Second Language Education The Ohio State University (h) 937-767-2200 Email: [email protected] Dear OSU Student:

I am a doctoral student in Foreign and Second Language Education and am writing to seek your help in my research. Mainly, I need to randomly select native speakers of English willing to listen to and repeat a short sentence list in English, and then read sentences in English, which I will record.

For my dissertation research, I am studying the English pronunciation of Chinese and American graduate students and investigating their intelligibility to other graduate students, including Chinese, Indian, Korean, and American students. My goal is to provide an analysis of TA-certified Chinese-accented English, as well as rate its intelligibility and comprehensibility to many international and American listeners. With the results of this study, English language teachers can improve their pronunciation teaching methods and better prepare Chinese students not only to qualify as TAs at American universities, but also to be more intelligible to their classmates in graduate school.

Again, please fill out and return the enclosed form to me at your earliest convenience.

Thank you for your time, and I look forward to meeting you!

270

Appendix I: Recruitment Response Form

Jocelyn Hardman, Ph.D. candidate Foreign and Second Language Education The Ohio State University (h) 937-767-2200 Email: [email protected]

I AM FROM BEIJING OR NORTHERN CHINA ⁭ YES ⁭ NO

I AM FROM COLUMBUS, OR CENTRAL OHIO ⁭ YES ⁭ NO

I AM FROM KOREA ⁭ YES ⁭ NO

I AM A NATIVE SPEAKER OF HINDI ⁭ ⁭ YES ⁭ NO

If you answered YES to one of the above questions, please fill out the following.

⁭ YES, I WILL PARTICIPATE IN YOUR STUDY

⁭ NO, I CANNOT PARTICIPATE IN YOUR STUDY

If you have any questions, please feel free to contact me: Email: [email protected] Home phone: 1-937-767-2200

271

APPENDIX J: INFORMED CONSENT FORM – TALKERS

272

INFORMED CONSENT FORM – TALKERS

Protocol #2005E0397

Research Study Title: The intelligibility and comprehensibility of Chinese-accented English to international and domestic students at a US university.

Purpose of the Study: The purpose of this study is to determine the intelligibility and comprehensibility of Chinese-accented English to Indian, Chinese, Korean, and American graduate students.

Benefits of the Study: This study will provide a model of TA-certified Chinese-accented English, which can be used for teaching purposes in interactive pronunciation computer programs. With the help of this model, English language teachers can improve their pronunciation teaching methods and better prepare Chinese students to qualify as TAs at American universities.

Procedure: If you agree to participate in this study, you will be asked to: (1) Complete a brief background questionnaire (30 minutes); (2) Attend one recording session (30 minutes) in which you will (a) read and record a list of English sentences; (b) read and record a list of sentences in your native language, if other than English.

Payment: Upon completion of the questionnaire and the recording session, you will receive $20.00 for your participation.

Withdrawal: You may withdraw your consent and discontinue participation in the study at any time.

Confidentiality: Your identity will be kept confidential. In published reports of this study, participants will be listed by number and native language only, not by name or any other identifier. After the study is complete, all recordings will be archived on CDs and kept in a locked metal cabinet in the researcher’s office. No one but the researchers listed below and any future research collaborators will have access to the data. If you have any further questions or concerns, please feel free to contact the researchers:

273

Co-Researcher: Principal Researcher: Jocelyn Hardman Keiko Samimy Ph.D. Candidate Associate Professor Spoken English Program Foreign/Second Language Education 75 Arps Hall/1945 N. High St. 232 Arps Hall/1945 N. High St. The Ohio State University The Ohio State University Columbus, OH 43210 Columbus, OH 43210 Home phone: (937) 767-2200 Office phone: (614) 292-7597 Email: [email protected] Email: [email protected]

274

CONSENT FOR PARTICIPATION IN RESEARCH

I consent to participating in the research entitled: The intelligibility and comprehensibility of Chinese-accented English to international and domestic students at a US university.

Dr. Keiko Samimy, Principal Investigator, or her authorized representative, Jocelyn Hardman, has explained to me the purpose of the study, the procedures to be followed, the expected duration, and payment for my participation. The benefits of the study have also been described.

I understand that no identifying information will be used in publications of this research, and my identity will be kept confidential. My recordings may be used for future research purposes.

My recordings may be used in conferences or public presentations of the research. If you would like to opt out of this, please check the following box: .

I acknowledge that I have had the opportunity to obtain additional information regarding the study and that any questions I have raised have been answered to my full satisfaction. Furthermore, I understand that I am free to withdraw my consent at any time and to discontinue participation in the study without penalty or prejudice.

Finally, I acknowledge that I have read and fully understand the consent form. I sign it freely and voluntarily. A copy has been given to me.

Date: ______Date: ______

Signed: ______Signed: ______

Researcher Participant

275

APPENDIX K: INFORMED CONSENT FORM – LISTENERS

276

INFORMED CONSENT FORM – LISTENERS

Protocol # 2005E0397

Research Study Title: The intelligibility and comprehensibility of Chinese-accented English to international and domestic students at a US university.

Purpose of the Study: The purpose of this study is to determine the intelligibility and comprehensibility of Chinese-accented English to Indian, Chinese, Korean, and American graduate students.

Benefits of the Study: This study will provide an analysis of TA-certified Chinese-accented English, as well as rate its intelligibility and comprehensibility to many international and American listeners. With the results of this study, English language teachers can improve their pronunciation teaching methods and better prepare Chinese students not only to qualify as TAs at American universities, but also to be more intelligible to their classmates in graduate school.

Procedure: If you agree to participate in this study, you will be asked to: (3) Complete a brief background questionnaire prior to the listening and recording sessions (about 15 minutes); (4) Participate in one listening session (about 45 minutes) in which you will (a) listen to a list of English sentences spoken by a variety of international and American graduate students; (b) orally repeat each sentence you hear; (c) confirm what you said by writing the sentence; (d) rate the familiarity of a list of English words. (5) Participate in one recording session (about 30 minutes) in which you will (a) read and record a list of English sentences; (b) read and record a list of sentences in your native language, if other than English.

Payment: Upon completion of the questionnaire, the listening session, and the recording session, you will receive $20.00 for your participation.

Withdrawal: You may withdraw your consent and discontinue participation in the study at any time.

277

Confidentiality: Your identity will be kept confidential. In published reports of this study, participants will be listed by native language, gender, and subject number only, but never by name or any other identifier. After the study is complete, all audio recordings will be archived on CDs and kept in a locked metal cabinet in the researcher’s office. No one but the researchers listed below and any approved future research collaborators will have access to the data. If you have any further questions or concerns, please feel free to contact the researchers:

278

CONSENT FOR PARTICIPATION IN RESEARCH

I consent to participating in the research entitled: The intelligibility and comprehensibility of Chinese-accented English to international and domestic students at a US university.

Dr. Keiko Samimy, Principal Investigator, or her authorized representative, has explained to me the purpose of the study, the procedures to be followed, the expected duration, and payment for my participation. The benefits of the study have also been described.

I understand that no identifying information will be used in publications of this research, and my identity will be kept confidential.

AUDIO RELEASE During the course of this experiment, I understand that my voice will be recorded for the “BuckeyeGTA Speech Corpus,” which will be used by the current researchers for present and future research studies.

In the future, I will also allow my audio recordings to be used for the purposes below. Speech files in the BuckeyeGTA corpus will only be labeled by native language, gender, and subject number – never by name. At no time will other identifying information be revealed and my identity will always be kept confidential.

A) For further research purposes by other researchers, but only with the approval of the current researchers. YES NO B) For conferences or public presentations of research. YES NO C) For educational purposes. YES NO

Finally, I acknowledge that I have read and fully understand the consent form. I sign it freely and voluntarily. A copy has been given to me.

Date: ______Date: ______

Signed: ______Signed: ______

Researcher Participant

279

APPENDIX L: MATERIALS LISTS

280

COCHLEAR CORPORATION

Revised Bamford-Kowal-Bench (BKB) Standard Sentence Test Randomization 1 - List 7 Answer Key

# Key TEST ITEMS Words

1. The children dropped the bag. 3

2. The dog came back. 3

3. The floor looked clean. 3

4. She found her purse. 3

5. The fruit is on the ground. 3

6. Mother got a saucepan. 3

7. They washed in cold water. 4

8. The young people are dancing. 3

9. The bus left early. 3

10. They had two empty bottles. 4

11. The ball is bouncing very high. 3

12. Father forgot the bread. 3

13. The girl has a picture book. 3

14. The orange was very sweet. 3

15. He is holding his nose. 3

16. The new road is on the map. 3

281

COCHLEAR CORPORATION

Revised Bamford-Kowal-Bench (BKB) Standard Sentence Test Randomization 1 - List 8 Answer Key

# Key TEST ITEMS Words

1. The boy forgot his book. 3

2. A friend came for lunch. 3

3. The match boxes are empty. 3

4. He climbed his ladder. 3

5. The family bought a house. 3

6. The jug is on the shelf. 3

7. The ball broke the window. 3

8. They are shopping for cheese. 3

9. The pond water is dirty. 3

10. They heard a funny noise. 4

11. The police are clearing the road. 3

12. The bus stopped suddenly. 3

13. She writes to her brother. 3

14. The football player lost a shoe. 4

15. The three girls are listening. 3

16. The coat is on a chair. 3

282

COCHLEAR CORPORATION

Revised Bamford-Kowal-Bench (BKB) Standard Sentence Test Randomization 1 - List 9 Answer Key

# Key TEST ITEMS Words

1. The book tells a story. 3

2. The young boy left home. 4

3. They are climbing the tree. 3

4. She stood near her window. 3

5. The table has three legs. 3

6. A letter fell on the floor. 3

7. The five men are working. 3

8. He listened to his father. 4

9. The shoes were very dirty. 3

10. They went on a vacation. 3

11. The baby broke his cup. 3

12. The lady packed her bag. 3

13. The dinner plate is hot. 3

14. The train is moving fast. 3

15. The child drank some milk. 3

16. The car hit a wall. 3

283

COCHLEAR CORPORATION

Revised Bamford-Kowal-Bench (BKB) Standard Sentence Test Randomization 1 - List 10 Answer Key

# Key TEST ITEMS Words

1. A dish towel is by the sink. 3

2. The janitor used a broom. 3

3. She looked in her mirror. 4

4. The good boy is helping. 3

5. They followed the path. 3

6. The kitchen clock was wrong. 3

7. The dog jumped on the chair. 4

8. Someone is crossing the road. 3

9. The mailman brought a letter. 3

10. They are riding their bicycles. 3

11. He broke his leg. 3

12. The milk was by the front door. 3

13. The shirts are hanging in the closet. 3

14. The ground was very hard. 3

15. The buckets hold water. 3

16. The chicken laid some eggs. 3

284

APPENDIX M: IPA – CMU CORRESPONDENCE KEY

285

IPA – CMU CORRESPONDENCE KEY

IPA Phone CMU Phone Word Word Orthography Transcription 1 a AA odd AA D 2 æ AE at AE T 3 AH hut HH AH T ə (w/ lexical stress /ʌ/) 4 AO ought AO T ɔ 5 au AW cow K AW 6 ai AY hide HH AY D 7 b B be B IY 8 CH cheese CH IY Z tʃ 9 d D dee D IY 10 δ DH thee DH IY 11 ε EH Ed EH D 12 ER hurt HH ER T ɚ 13 ey EY ate EY T 14 f F fee F IY 15 g G green G R IY N 16 h HH he HH IY 17 I IH it IH T 18 iy IY eat IY T 19 dʒ JH gee JH IY 20 k K key K IY 21 l L lee L IY 22 m M me M IY 23 n N knee N IY 24 ŋ NG ping P IH NG 25 ow OW oat OW T 26 oi OY toy T OY 27 p P pea P IY 28 R read R IY D r (for Eng) /ɹ/ 29 s S sea S IY 30 SH she SH IY ʃ 31 t T tea T IY 32 θ TH theta TH EY T AH 33 UH hood HH UH D ʊ 34 uw UW two T UW 35 v V vee V IY 36 w W we W IY 37 y (for Eng) /j/ Y yield Y IY L D 38 z Z zee Z IY 39 ʒ ZH seizure S IY ZH ER

286