• unlimited access with print and download
    $ 37 00
  • read full document, no print or download, expires after 72 hours
    $ 4 99
More info
Unlimited access including download and printing, plus availability for reading and annotating in your in your Udini library.
  • Access to this article in your Udini library for 72 hours from purchase.
  • The article will not be available for download or print.
  • Upgrade to the full version of this document at a reduced price.
  • Your trial access payment is credited when purchasing the full version.
Buy
Continue searching

ESL speakers' production of English lexical stress: The effect of variation in acoustic correlates on perceived intelligibility and nativeness

ProQuest Dissertations and Theses, 2009
Dissertation
Author: Paul Edmunds
Abstract:
Non-native speakers of English often experience problems in pronunciation as they are learning English, many such problems persisting even when the speaker has achieved a high degree of fluency. Research has shown that for a non-native speaker to sound most natural and intelligible in his or her second language, the speaker must acquire proper prosody, such as native-like speech rhythms (Tajima et al., 1997; Wenk, 1985; Wennerstrom, 2001). This dissertation investigates how native English and Spanish ESL (English as a Second Language) speakers compare in their production of three acoustic correlates of lexical stress in English, namely the relative durations of stressed and unstressed vowels and their relative intensities and fundamental frequency (F0) values. A set of three-syllable words, including cognates and non-cognates, was analyzed. The results from the production study were used to design a listening task that investigated how the ESL speakers' varying productions of these acoustic cues affected native English listeners' perception of their speech intelligibility and nativeness. The ESL speakers produced a wider range of values than did the native speakers for all three acoustic correlates. The ESL and native speakers differed statistically in their productions of Spanish/English cognates with different stress patterns in each language and often differed on non-cognates. The ESL group produced the most native-like patterns in cognates with the same stress pattern in each language. The stimuli for the listening task were words recorded by native English participants and subsequently modified to emulate the production of the acoustic correlates of lexical stress by the ESL speakers. Listeners' ratings of speech intelligibility were statistically higher for words in which intensity was increased on the vowel that is expected to receive lexical stress compared to the adjacent unstressed vowel. Increasing vowel duration on the unstressed vowel led to statistically lower ratings of both intelligibility and nativeness. This dissertation contributes to a small body of research regarding the production of prosody in second language speech. The results suggest that a speaker's prosody alone can influence a native listener's perception of speech intelligibility and nativeness.

ix TABLE OF CONTENTS LIST OF FIGURES.........................................................................................................xvi LIST OF TABLES..........................................................................................................xvii Chapter 1: Introduction........................................................................................................1 1.0 Overview...........................................................................................................1 1.1 Background and Motivation for the Study........................................................2 1.2 Need for the Study............................................................................................5 1.3 Research Questions and Hypotheses................................................................7 1.3.1 Research Questions for the Production study....................................8 1.3.2 Hypotheses for the Production study.................................................9 1.3.3 Research questions for Listening study...........................................10 1.3.4 Hypotheses for the Listening study..................................................11 1.4 A brief overview of second language teaching...............................................12 1.5 Organization of the Dissertation.....................................................................17 1.6 Summary.........................................................................................................18 Chapter 2: Literature Review.............................................................................................19 2.0 Overview of Prosody......................................................................................19 2.1 Pitch accent.....................................................................................................21 2.2 Previous studies of lexical stress in English...................................................22 2.2.1 Acoustic correlates of stress in English...........................................22 2.2.2 ESL speakers’ production of English lexical stress.........................24 2.3 Lexical stress in Spanish.................................................................................26 2.4 Language transfer............................................................................................27

x 2.5 Intelligibility...................................................................................................30 2.6 Accentedness...................................................................................................33 2.7 Nativeness.......................................................................................................36 2.8 Pronunciation research in applied linguistics.................................................40 2.9 Application: Pronunciation in ESL classroom................................................42 2.10 Summary.......................................................................................................45 Chapter 3: Methodology....................................................................................................47 3.0 Overview.........................................................................................................47 3.1 Production Study.............................................................................................47 3.1.1 Participants.......................................................................................47 3.1.2 Recording procedure........................................................................48 3.1.3 Materials..........................................................................................50 3.2 Analysis of the Recordings.............................................................................55 3.2.1 Labeling...........................................................................................55 3.2.2 Exclusions........................................................................................56 3.3 Measures.........................................................................................................57 3.3.1 Applying the measures to the data...................................................59 3.3.2 Test Statistics for Production Data..................................................60 3.4 Additional Data...............................................................................................61 3.4.1 Stress placement ratings...................................................................61 3.4.2 Pitch accent ratings..........................................................................62 3.4.3 Target Word Familiarity Scale.........................................................63 Part II: Manipulation and resynthesis of target words.......................................................64

xi 3.5 Overview.........................................................................................................64 3.5.1 Selection of words for manipulation................................................65 3.5.2 Choosing the speakers for manipulations........................................65 3.5.3 Selecting individual tokens to create manipulations........................66 3.5.4 Note on the words chosen for manipulations...................................67 3.6 Manipulations.................................................................................................68 3.6.1 Ranges for manipulation..................................................................69 3.6.2 Types of manipulations....................................................................69 3.6.3 Particulars of manipulations............................................................70 3.7 Values for manipulations................................................................................71 3.7.1 Manipulation of Duration................................................................71 3.7.2 Manipulation of Fundamental Frequency........................................72 3.7.3 Manipulation of Intensity.................................................................73 Part III: Intelligibility and Nativeness Listening Study.....................................................73 3.8 Overview.........................................................................................................73 3.8.1 Listeners...........................................................................................74 3.8.2 Rating Scale.....................................................................................74 3.8.3 Task..................................................................................................75 3.9 Test Statistics for Listening Study..................................................................77 Chapter 4: Results and Discussion from the Production Study.........................................78 4.0 Overview.........................................................................................................78 4.1 Organization of this chapter............................................................................80 4.2 Range of variation in productions...................................................................80

xii 4.3 Results for 12-word set...................................................................................82 4.3.1 Vowel Duration Ratio......................................................................82 4.3.2 Change in Intensity..........................................................................84 4.3.3 Change in Fundamental Frequency.................................................85 4.4 Results for 6-word set with trend for increase in F0 on SV by native speakers ...............................................................................................................................88 4.4.1 Vowel Duration Ratio......................................................................88 4.4.2 Change in Intensity..........................................................................89 4.4.3 Change in Fundamental Frequency.................................................90 4.5 Comparison of productions with correct stress placement by the ESL speakers ...............................................................................................................................91 4.5.1 Vowel Duration Ratio......................................................................94 4.5.2 Change in Intensity..........................................................................95 4.5.3 Change in Fundamental Frequency.................................................96 4.6 Results for 6-word set, words with a reliable increase in F0 on SV by native speakers, ESL speakers’ correct productions only...............................................97 4.6.1 Vowel Duration Ratio......................................................................97 4.6.2 Change in Intensity..........................................................................98 4.6.3 Change in Fundamental Frequency.................................................99 4.7 Discussion.....................................................................................................100 4.7.1 12-word set, all productions...........................................................100 4.7.2 6-word set, all productions.............................................................102 4.7.3 12-word set, correct productions only...........................................103

xiii 4.7.4 6-word set, correct productions only.............................................104 4.8 Summary.......................................................................................................105 4.8.1 Research question #1: “How does the production of lexical stress by Spanish ESL speakers compare prosodically to native English speakers?” .................................................................................................................105 4.8.2 Research Question #2: Does the cognate status of a word affect a Spanish ESL speaker’s production of lexical stress?..............................107 Chapter 5: Results and Discussion of Listening Task.....................................................109 5.0 Overview.......................................................................................................109 5.1 Results of the Intelligibility and Nativeness rating tasks..............................112 5.1.1 Intelligibility..................................................................................113 5.1.2 Nativeness......................................................................................114 5.2 Variation in the ratings by listeners..............................................................115 5.3 Correlation analysis of intelligibility and nativeness....................................117 5.4 Reliability analysis........................................................................................118 5.4.1 Inter-rater reliability.......................................................................118 5.4.2 Intra-rater reliability.......................................................................119 5.5 Discussion.....................................................................................................120 5.5.1 Intelligibility..................................................................................120 5.5.2 Nativeness......................................................................................123 5.5.3 Correlation of intelligibility and nativeness ratings.......................125 5.5.4 Inter-rater reliability.......................................................................126 5.5.5 Intra-rater reliability.......................................................................126

xiv 5.6 Summary.......................................................................................................127 Chapter 6: Conclusion......................................................................................................129 6.0 Overview.......................................................................................................129 6.1 Summary of the results.................................................................................129 6.1.1 Production study............................................................................129 6.1.2 Listening study...............................................................................131 6.2 Implications of the research findings............................................................132 6.3 Strengths and limitations of the study...........................................................136 6.4 Future Work..................................................................................................137 6.5 Closing Comments........................................................................................139 Appendices.......................................................................................................................141 Appendix A: Speaker demographics...............................................................................142 Appendix B: Language background questionnaire - Native English speakers................143 Appendix C: Language background questionnaire - Non-native English speakers........144 Appendix D: Non-cognate target words and sentence stimuli.........................................145 Appendix E: Cognate target words and sentence stimuli................................................148 Appendix F: ICC for Stress Placement Ratings...............................................................151 Appendix G: ICC for Pitch Accent Ratings.....................................................................155 Appendix H: Target Word Familiarity Scale...................................................................156 Appendix I: Parametric (t-test) statistical output for production study from SPSS.........159 Appendix J: Non-parametric (Mann-Whitney U test) statistical output for production study from SPSS..................................................................................................183 Appendix K: Variability in F0 production by native speakers in production study........187

xv Appendix L: Background questionnaire for listening experiment...................................188 Appendix M: Listener demographics...............................................................................189 Appendix N: Screen shots of rating scales for intelligibility and nativeness ratings.......190 Appendix O: Statistical output for one-way ANOVA for listening study from SPSS....191 Appendix P: Correlation analysis of intelligibility and nativeness..................................197 Appendix Q: Statistical output from ICC for Inter-rater reliability.................................201 Appendix R: Intra-rater Reliability..................................................................................203 References........................................................................................................................210

xvi LIST OF FIGURES Figure 3.1: Sample Text Grid displaying the four tiers labeled in the word bipolar.........56 Figure 4.1: Vowel Duration Ratio.....................................................................................83 Figure 4.2: Change in Intensity..........................................................................................84 Figure 4.3: Change in Fundamental Frequency.................................................................85 Figure 4.4: Vowel Duration Ratio. 6-word set with a reliable increase in F0 on the SV by native speakers.......................................................................................................88 Figure 4.5: Change in Intensity. 6-word set with a reliable increase in F0 on the SV by native speakers.......................................................................................................89 Figure 4.6: Change in Fundamental Frequency. 6-word set with a reliable increase in F0 on the SV by native speakers.................................................................................90 Figure 4.7: Vowel Duration Ratio, correct productions only............................................94 Figure 4.8: Change in Intensity, correct productions only.................................................95 Figure 4.9: Change in Fundamental Frequency, correct productions only........................96 Figure 4.10: Vowel Duration Ratio. 6-word set with a reliable increase in F0 on the SV by native speakers, correct productions only.........................................................97 Figure 4.11: Change in Intensity. 6-word set with a reliable increase in F0 on the SV by native speakers, correct productions only..............................................................98 Figure 4.12: Change in Fundamental Frequency. 6-word set with a reliable increase in F0 on the SV by native speakers, correct productions only........................................99 Figure 5.1: Ratings of Intelligibility and Nativeness.......................................................113

xvii LIST OF TABLES Table 3.1: Distribution of target words by syllable length and cognate status..................53 Table 3.2: 6 Target words chosen based on consistently increased F0 on SV by native speakers, by cognate status....................................................................................59 Table 3.3: Naming key for modified versions of one target word ....................................70 Table 4.1: 3-syllable target words with stress pattern and cognate status.........................78 Table 4.2: Range of variation in productions.....................................................................81 Table 4.3: Incorrect productions of target words by the ESL speakers.............................92 Table 4.4: Frequency of usage of target words by the ESL speakers................................93 Table 5.1: Seven conditions for stimuli in listening experiment.....................................110 Table 5.2: Range of ratings by listener............................................................................116 Table 5.3: Results of Pearson product-moment correlation analysis for intelligibility and nativeness.............................................................................................................118

1 Chapter 1: Introduction 1.0 Overview Non-native speakers of English often experience problems in pronunciation as they are learning English, many such problems persisting even when the speaker has achieved a high degree of fluency. Difficulties with pronunciation involve segmental errors (i.e. incorrect production of target consonants and vowels) and prosodic (or suprasegmental) errors (e.g. rhythm and stress). Much previous research has analyzed differences in the production of segmental information by non-native speakers compared to native speakers (e.g. Flege, 1984, 1993; Rogers & Dalby, 2005; Tarone, 2005). Studies regarding the perception of segmental structure depending on a speaker’s native language and cross-linguistically have also received a great amount of attention (cf. Grabe et al., 2003). Relatively little, however, is known about non-native speakers’ prosody when speaking in their second language. The present study investigates the production of one aspect of prosody – lexical stress – in the speech of Spanish ESL (English as a Second Language) speakers. Lexical stress refers to the assignment of stress to a specific syllable within a word. Lexical stress can be marked in various ways. Segmentally, lexical stress in English is associated with long or unreduced vowels (Ladefoged, 1993). Prosodically, lexical stress in English is cued by various acoustic dimensions including fundamental frequency (F0), intensity, and duration (Lehiste, 1996). Lexical stress in English can be used to distinguish between lexical items in some noun/verb pairs. For example, placing stress on the first syllable of the word record indicates that the word being pronounced is a noun, whereas placing stress on the second syllable (record) indicates that the word is a

2 verb. Although noun/verb pairs such as those in English do not exist in Spanish, lexical stress disambiguates the meaning of words in a similar way in Spanish among conjugated verb forms. For example, the word hablo means “I speak” while habló means “he/she spoke.” The correct placement of stress on a particular syllable is crucial in conveying the intended meaning of a word in languages that use stress distinctively such as English and Spanish. Consequently, this prosodic feature of speech can affect the perception of intelligibility, “the degree to which a speaker’s utterance is understood by a listener” (Tarone, 2005, p. 493). It can also impact the perception of nativeness, the degree to which a speaker sounds like a native speaker of a particular language. In this dissertation, the production of lexical stress by Spanish ESL speakers is compared to that of native English speakers and the degree to which variation in the acoustic correlates of lexical stress affects speech intelligibility and perceived nativeness is investigated. This dissertation is composed of six chapters. Chapter 1 gives a brief overview of lexical stress and the importance of producing it in a native-like way in order to ensure efficient communication between the speaker and his or her listener. This chapter also provides motivation for the study, the research questions and hypotheses, and a brief overview of second language teaching pedagogy with a focus on pronunciation teaching to frame this work in a broader, educational context. A synopsis of the following chapters is given below in Section 1.5. 1.1 Background and Motivation for the Study My interest for doing this study centered on English pronunciation comes from over ten years as an ESL instructor. I have witnessed the importance of pronunciation in

3 all core ESL classes (reading, writing, grammar, and speaking) at almost every level (i.e. from beginning to highly advanced students). In speaking classes, in particular, pronunciation instruction is an important component of the coursework, though pronunciation issues also come into play in reading, writing and grammar classes. Through casual observation, I came to notice that correct production of stress could be particularly difficult for Spanish ESL students and got into the habit of writing down words the students mispronounced during class. I would often transcribe these words phonetically (particularly if the pronunciation problem was segmental in nature), make a note when one syllable was substituted for another, or when patterns of stress placement did not sound as expected. Though not a systematic or scientific study, one thing became clear to me: mispronunciations due to incorrect placement of lexical stress were common, and these differing pronunciation patterns were very noticeable. In addition to noticing that misplacing lexical stress was a very salient error, I noticed that these errors often occurred on Spanish/English cognates such as the word catholic. Native Spanish speakers often pronounce the word catholic with stress on the second syllable (“ca-tho-lic”) [kæ.ˈӨa.lɪk] instead of with stress on the first syllable (“ca-tho-lic”) [ˈkæ-Өə-lɪk]. It seemed that word familiarity was not always related to these mispronunciations. For example, the ESL speakers seemed almost as likely to misplace stress on a very common word such as catholic as they were to do on a less common word like lunatic. These rather casual observations motivated me to think more deeply about how the cognate status of a word could affect a language learner’s pronunciation patterns. It also made me wonder just how much these divergent patterns

4 could affect a listener’s perception of the speaker’s speech intelligibility or if they could affect the perception of nativeness. Certainly, producing words that are easy to understand (that is, highly intelligible) is beneficial to enabling communication. For a non-native speaker, being intelligible is often the number one goal. Frustration ensues, for both the speaker and the listener, when one’s words are difficult (if not impossible) to understand. As a second language learner of Spanish, and a person who lived in a Spanish speaking country on different occasions, I know first-hand the feeling of not being understood, something my students here feel as non-native speakers of English in their host country. Besides feeling frustration myself, I know that my listeners were sometimes frustrated as they could not easily understand me. I always remember a friend who would tell me, “Habla mejor. No te entiendo” (Speak better. I don’t understand you). This request stayed with me as an ongoing motivation to produce speech that was as intelligible as possible. But being intelligible is only part of the picture. While living in Venezuela, my host family would call me a “loro” (parrot) as I made a habit out of repeating their words out loud – imitating their speech as closely as possible in order to practice my pronunciation. This exercise made me realize how important it can be (at least for some language learners) to sound as native as possible. These personal experiences as a language learner and language teacher have come together in this dissertation. I feel privileged to be able to work on a project such as this in which I was able to get specific insight into a question that has intrigued me about language learning, specifically how native and non-native speakers of English differ in terms of pronunciation and how those

5 differences can affect communication in terms of a listener’s perception of speech intelligibility and nativeness. 1.2 Need for the Study The present study first assesses production of lexical stress in English by native Spanish speakers and later investigates how differences in lexical stress production, compared to that of a native English speaker, can affect a speaker’s perceived intelligibility and degree of nativeness. The Spanish ESL speakers involved in this study are university students from South America. Studying the language of Spanish ESL speakers is important because they represent a growing sector of the population of the United States. According to the U.S. Census of 2000, there are more than 35.3 million Hispanic or Latinos residing in the United States, totaling 12.5% of the U.S. population. Of course, not all Hispanic people are ESL speakers. The U.S. Census defines Hispanic as “Persons of Hispanic origin, in particular, those who indicated that their origin was Mexican, Puerto Rican, Cuban, Central or South American, or some other Hispanic origin” ("United States Census 2000"). The number of Hispanic people living in the U.S. is increasing, with a projected 120.6 million Hispanics (approximately 24.4% of the total U.S. population) to reside in the U.S. by the year 2050. Close to home, the 2000 U.S. Census states that 42.1% of New Mexico’s population is Hispanic and some reports have stated that 50% of the residents of New Mexico speak Spanish. Assuming many Hispanic people have learned or are in the process of learning English as their second language, this dissertation is directly relevant not only for our community in New Mexico but for a substantial portion of the U.S. population.

6 As Wennerstrom (2001, p. 230) states, “For those nonnative speakers whose goal is to participate in English-language speech communities, an understanding of the English-specific aspects of prosody will be an enhancement.” This statement reflects a reality in the ESL classroom where pronunciation is a major element of instruction. For a non-native speaker to sound more natural in his or her second language, the speaker must acquire proper prosody, such as native-like speech rhythms, in order to be most intelligible (Tajima et al., 1997; Wenk, 1985; Wennerstrom, 2001). However, until recently instruction in pronunciation has largely been limited to practice in segmental aspects of speech. The focus has been on how to produce consonants and vowels correctly, often using drills involving minimal pairs. Although this type of segmental instruction continues to be both popular and important, there is a growing trend for instruction to focus on prosodic aspects of speech in order to help the student be more intelligible and sound more natural. Wennerstrom (2001) gives some evidence for this shift to working on prosodic aspects of pronunciation over segmental aspects when she explains how many of the most current text books, for example Well Said (Grant, 2000), have relegated specific instruction on segmental aspects to the appendices. Indeed, Well Said begins right away with practice in speech rhythms. As Wenk (1985, p. 170) puts it, “The development of a proper sense of the nature and operation of speech rhythms is possibly the single most important area of pronunciation study.” The necessity for a complete understanding of the speech of second language speakers extends beyond the ESL classroom. According to a recent newsletter from the National Student Speech Language Hearing Association (NSSLHA), “In national surveys, SLPs [Speech-Language Pathologists] have identified assessment of culturally

7 and linguistically diverse children as a top priority need for enhanced personnel preparation efforts and continuing education” ("¿Qué hay de nuevo? Learn what's new in serving those of diverse backgrounds", 2006, p. 3). Treatment for bilingual clients at speech clinics is in high demand. Thus, in addition to serving those who work in second language teaching, the present study hopes to be of service to those who work with non- native speakers of English in the field of communication science and disorders. The main desire is that the results of this dissertation further the understanding of prosody as it relates to second language speakers and language learning. It is hoped that the work here will help further our understanding of how lexical stress is realized by both native English and Spanish ESL speakers, and how variation in the acoustic cues related to lexical stress can improve or reduce a speaker’s level of intelligibility or the degree to which he or she sounds like a native speaker of English. 1.3 Research Questions and Hypotheses The two main goals of this dissertation were to 1) Compare the production of lexical stress in English by Spanish ESL speakers to that of native English speakers and 2) investigate the relation between native and non-native lexical stress production and listeners’ evaluations of speech intelligibility and nativeness. To investigate these two questions, the research was divided into three parts: a production study, modification and resynthesis of various target words based on observed patterns of production, and a listening study using the words generated with these modifications to be presented as stimuli to native English listeners who rated the words for intelligibility and nativeness. The research questions for the production and perception studies and their hypotheses, respectively, are explained below.

8 1.3.1 Research Questions for the Production study The first research question is: How does the production of lexical stress by Spanish ESL speakers compare to that of native English speakers? That is, is there a difference in the acoustic characteristics of vowel duration, intensity and fundamental frequency between stressed and unstressed syllables produced by native and non-native English speakers? Although other acoustic correlates of lexical stress in English have been documented, such as the contribution of spectral tilt and noise in the region of F3 (Okobi, 2006), the three acoustic correlates of lexical stress in English at the focus of this study (duration, intensity and fundamental frequency) are perhaps the best understood and most studied with regard to their production by native English speakers (e.g. Fry, 1955, 1958). For this reason, it is beneficial to select these three acoustic correlates of stress to compare the speech of non-native speakers of English to that of native English speakers. Pertinent to the first research question of how non-native speakers produce lexical stress in English is the issue of language transfer from a speaker’s first language (L1) to his or her second language (L2). Studies in bilingualism have shown that language transfer from L1 to L2 is common and that second language learners often experience varying degrees of transfer (Wenk, 1985). Transfer is a possible reason for mispronunciations in a speaker’s second language, and it is hypothesized below that cognates may be especially susceptible to effects of language transfer. This leads to the second research question of the production study: Does the cognate status of a word affect a Spanish ESL speaker’s production of lexical stress? For example, do Spanish ESL speakers pronounce cognates with differing stress patterns in

9 the two languages differently when speaking in English? Furthermore, do pronunciation errors occur more frequently in cognates with differing stress patterns in each language than cognates with the same stress patterns? And how do non-cognates fit into this pattern? To address these research questions, the production of three acoustic correlates of lexical stress were measured in the vowel that is expected to receive lexical stress and compared in an adjacent unstressed vowel. These acoustic measures were then compared between native and non-native speaker groups. The results of the production study and answers to these research questions are found in Chapter 4. 1.3.2 Hypotheses for the Production study First, ESL speakers, who have less knowledge of the English phonological system, may rely heavily on increasing fundamental frequency to mark stress. As increasing F0 is perhaps the most important of the three acoustic correlates in creating the perception of stress in Spanish (cf. Llisterri et al., 2003), language transfer of F0 patterns from Spanish to English may result. Secondly, language transfer may cause higher degrees of variability in English/Spanish cognates with lexical stress on different syllables in the two languages (DSCs) (e.g. lu-na-tic / lu-ná-tico). It is hypothesized that the productions of the DSCs will be the most divergent from the native speakers’ productions because negative transfer from the Spanish ESL speakers’ L1 will influence his or her production of the word English. Same-stress cognates (SSCs) (e.g. English ra-di-o / Spanish ra-di-o) should be produced more like the productions of the native English speakers as a positive transfer of patterns is expected. The production of non-cognates is hypothesized to be

10 more similar to that of the native speakers because their pronunciation may have been learned by rote, though more variation is predicted than in the SSCs. 1.3.3 Research questions for Listening study In the second part of this study, the consequences of variation in the production of lexical stress by the non-native speakers are examined. Stress patterns from non-native speakers’ productions were modeled by modifying and resynthesizing original recordings of native speakers in order to imitate prosodic characteristics of the non-native speakers. Modifying the speech of native English speakers ensured good segmental quality in the resynthesized tokens in order to avoid the effect of potential segmental errors by the ESL speakers. This allowed the focus of the investigation to remain strictly on prosodic aspects of the speech signal. This, in part, addresses a claim by Munro (2008, p. 203) that “volume, loudness of speech, voice quality, and clarity can affect judgments [but] may be very difficult or even impossible to assess … [though] they are clearly relevant to communication.” Manipulating the various acoustic correlates of stress individually makes it possible to investigate the extent that single acoustic dimensions affect the perception of intelligibility and nativeness of a particular speaker. The first research question of the listening study attempts to determine if changes to individual acoustic cues for lexical stress have an effect on a native English listener’s ratings of speech intelligibility and nativeness. For example, does producing the stressed vowel with longer duration than the unstressed vowel (as native speakers typically do) make a speaker sound more intelligible than if the unstressed vowel has increased duration? Similarly, does raising fundamental frequency on the stressed vowel make a speaker sound more intelligible than raising F0 on the unstressed vowel? And for these

11 patterns of production, do native English listeners rate the stimuli as sounding more native-like when vowel duration is lengthened or F0 is increased on the stressed vowel rather than on the unstressed vowel? The second main research question for the listening study is: Do ratings of speech intelligibility for the manipulated stimuli correlate positively with ratings for nativeness? In other words, does a speaker who sounds highly intelligible also sound highly native? The results of the listening study and answers to these research questions are given in Chapter 5. 1.3.4 Hypotheses for the Listening study It is hypothesized that increasing vowel duration, intensity and F0 on the stressed vowel will lead to higher ratings of intelligibility and nativeness than tokens having the same values for vowel duration, intensity and F0 on both the stressed and unstressed vowels, or tokens with less vowel duration, intensity and F0 on the stressed vowel than on the unstressed vowel. This hypothesis is based on the literature that has shown that increases in these variables are trademarks for marking prominence and lexical stress in English (Fry, 1955, 1958; Lehiste, 1996). An absence of correlation between speech intelligibility and nativeness is hypothesized as a speaker may sound highly intelligible though not necessarily native. The next section provides a brief overview of language teaching, in particular with respect to the teaching of pronunciation, in order to put the goals of this dissertation in a broader, educational context.

Full document contains 238 pages
Abstract: Non-native speakers of English often experience problems in pronunciation as they are learning English, many such problems persisting even when the speaker has achieved a high degree of fluency. Research has shown that for a non-native speaker to sound most natural and intelligible in his or her second language, the speaker must acquire proper prosody, such as native-like speech rhythms (Tajima et al., 1997; Wenk, 1985; Wennerstrom, 2001). This dissertation investigates how native English and Spanish ESL (English as a Second Language) speakers compare in their production of three acoustic correlates of lexical stress in English, namely the relative durations of stressed and unstressed vowels and their relative intensities and fundamental frequency (F0) values. A set of three-syllable words, including cognates and non-cognates, was analyzed. The results from the production study were used to design a listening task that investigated how the ESL speakers' varying productions of these acoustic cues affected native English listeners' perception of their speech intelligibility and nativeness. The ESL speakers produced a wider range of values than did the native speakers for all three acoustic correlates. The ESL and native speakers differed statistically in their productions of Spanish/English cognates with different stress patterns in each language and often differed on non-cognates. The ESL group produced the most native-like patterns in cognates with the same stress pattern in each language. The stimuli for the listening task were words recorded by native English participants and subsequently modified to emulate the production of the acoustic correlates of lexical stress by the ESL speakers. Listeners' ratings of speech intelligibility were statistically higher for words in which intensity was increased on the vowel that is expected to receive lexical stress compared to the adjacent unstressed vowel. Increasing vowel duration on the unstressed vowel led to statistically lower ratings of both intelligibility and nativeness. This dissertation contributes to a small body of research regarding the production of prosody in second language speech. The results suggest that a speaker's prosody alone can influence a native listener's perception of speech intelligibility and nativeness.