1 SOCIAL AND STYLISTIC VARIATION OF (R) IN BANGKOK THAI By ORAPAT POOKKAWES A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2014
2 Â© 2014 Orapat Pookkawes
3 To my dearest family , for their endless love
4 ACKNOWLEDGMENTS This dissertation could have been very difficult to accomplish without the assistance of various people. First and foremost, I would like to express my deep gratitude to my advisor, Dr. HÃ©lÃ¨ne Blondeau , for her insightful suggestions and encouragement . Her guidance was invaluable and her support tremendous. I would also like to sincerely thank my committee members, Dr. Diana Boxer, Dr. Ra tree Wayland, and Dr. John Krigbaum , for their valuable advice and support. I would like to express my great appreciation to the University of the Thai Chamber of Commerce (UTCC), in B angkok , which gave me the opportunity to pursue my doctorate. A special thank you goes to Dr. Dhanate Vespada, the dean of the School of Humanities at UTCC, who has always supp orted and encouraged me. I thank my colleagues and the staff of the School of Humaniti es at UTCC for assisting me with recruiting the participa nts and other important things. I also wish to thank the people who willingl y participated in my research and the co coders for the work they have done for me. I am especially grateful to Dr. Donruet hai Khaodeedech and the faculty and staff of the Department of Linguistics at the University of Florida (UF), and to friends from the Thai student association at UF for everything they have helped me with from the beginning of my doctoral studies. In addi tion , I am thankful to my dearest friends in Thailand and Norway, Wilasinee Darachai, Krongkaew Trailertmanee, Krittaporn Termvanich, and Neti Phongsuchart for being supportive and helpful with academic resources and advice. I appreciate the academic and emotional support from Eunha Hwang and Yu Ning Lai
5 at UF . I also thank Raj B arua and other proofreaders for their helpful comments and proofreading. Finally, I am indebted to my beloved parents, Noppanan and Suebsai Pookkawes, who always have confidenc e in me. I could never have completed my degree without their unconditional love, support, patience, and encouragement in every way possible. I thank my entire family for always being there for me and my late sister who is always in my heart .
6 TABLE OF C ONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ .......... 10 LIST OF FIGURES ................................ ................................ ................................ ........ 11 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION AND BACKGROUND INFORMATION ................................ ....... 15 Language Variation in the Thai Language ................................ .............................. 16 Purpose and Rationale of the Study ................................ ................................ ....... 16 Research Questions ................................ ................................ ............................... 21 Hypotheses ................................ ................................ ................................ ............. 22 Organization of the Dissertation ................................ ................................ .............. 23 Overview of the Thai Language ................................ ................................ .............. 24 Overview of the City of Bangkok ................................ ................................ ............. 25 Description of Thai Consonant Inventory ................................ ................................ 27 Major Variants of (R) in Bangkok Thai ................................ ................................ .... 27 2 THEORETICAL FRAMEWORK ................................ ................................ .............. 35 The Study of Language Variation: Variationist Sociolinguistics .............................. 35 Quantitative Paradigm ................................ ................................ ...................... 35 The Linguistic Variable ................................ ................................ ..................... 39 The Principle of Accountability ................................ ................................ ......... 43 Circumscribing the Variable Context ................................ ................................ 44 Quantitative Analysis ................................ ................................ ........................ 45 Statistical Analysis ................................ ................................ ............................ 45 The Speech Community ................................ ................................ ................... 47 Extralinguistic Factors and Language Variation ................................ ...................... 49 Soc ial Variation ................................ ................................ ................................ 49 Age ................................ ................................ ................................ ............ 49 Gender ................................ ................................ ................................ ....... 52 Stylistic Variation ................................ ................................ .............................. 56 3 LITERATURE REVIEW ................................ ................................ .......................... 62 The Sociolinguistic Variable (R) ................................ ................................ .............. 62 Variation of (R) in English ................................ ................................ ................. 63 American English ................................ ................................ ....................... 63 Other varieties of English ................................ ................................ ................. 70
7 Variation of (R) in French ................................ ................................ ................. 72 Variation of (R) in Dutch ................................ ................................ ................... 74 Variation of (R) in Mexican Spanish ................................ ................................ . 76 The Sociolinguistic Variable (R) in Thai ................................ ................................ .. 77 4 METHODOLOGY ................................ ................................ ................................ ... 85 Linguistic Contexts Applied in the Study ................................ ................................ . 85 Extralinguistic Factors Applied in the Study ................................ ............................ 85 Age ................................ ................................ ................................ ................... 86 Gender ................................ ................................ ................................ ............. 87 Style ................................ ................................ ................................ ................. 87 Education ................................ ................................ ................................ ......... 88 Participants ................................ ................................ ................................ ............. 88 Criteria for Participants ................................ ................................ ..................... 88 Recruitment of Participants ................................ ................................ .............. 89 Data Collecti on ................................ ................................ ................................ ....... 90 The Sociolinguistic Interview ................................ ................................ ............ 91 Equipment ................................ ................................ ................................ ........ 93 Procedure ................................ ................................ ................................ ......... 94 Data Analysis ................................ ................................ ................................ .......... 98 Data Transcription ................................ ................................ ............................ 98 Data Extraction and Codi ng ................................ ................................ .............. 98 Statistical Methods ................................ ................................ ......................... 100 5 VARIATION IN THE INITIAL R: RESULTS AND INTERPRETATION .................. 105 The Initial (R) ................................ ................................ ................................ ........ 105 The Four Variants of the Initial (R) ................................ ................................ ........ 106 Overall Distribution of the Initial (R) Variants ................................ ........................ 107 Distribution of the Initial (R) Variants by the Extralinguistic Factors ...................... 107 Age ................................ ................................ ................................ ................. 107 Gender ................................ ................................ ................................ ........... 111 Education ................................ ................................ ................................ ....... 113 Style ................................ ................................ ................................ ............... 115 Relationship of the Extralinguistic Factors ................................ ............................ 116 Age and Style ................................ ................................ ................................ . 117 Age and Gender ................................ ................................ ............................. 117 Gender and Style ................................ ................................ ........................... 119 Multivariate Analysis of the Initial (R) ................................ ................................ .... 119 Factor Groups Used in the Study ................................ ................................ ... 120 Significant Factor Groups ................................ ................................ ............... 121 Style ................................ ................................ ................................ ......... 122 Age ................................ ................................ ................................ .......... 123 Gender ................................ ................................ ................................ ..... 124
8 6 VARIATION IN THE CONSONANT CLUSTER R: RESULTS AND INTERPRETATION ................................ ................................ ............................... 134 The Consonant Cluster (R) ................................ ................................ ................... 134 The Four Variants of the Consonant Cluster (R) ................................ ................... 134 Overall Distribution of the Consonant Clusters (R) Variants ................................ . 135 Distribution of the Consonant Cluster (R) Variants by the Extralinguistic Factors ................................ ................................ ................................ .................. 136 Age ................................ ................................ ................................ ................. 136 Gender ................................ ................................ ................................ ........... 139 Education ................................ ................................ ................................ ....... 141 Style ................................ ................................ ................................ ............... 142 Relationship of the Extralinguistic Factors ................................ ............................ 144 Age and Style ................................ ................................ ................................ . 144 Age and Gender ................................ ................................ ............................. 145 Gender and Style ................................ ................................ ........................... 146 Multivariate Analysis of the Consonant Cluster (R) ................................ ............... 148 Factor Gr oups Used in the Study ................................ ................................ ... 148 Significant Factor Groups ................................ ................................ ............... 149 Style ................................ ................................ ................................ ......... 150 Education ................................ ................................ ................................ . 151 Gender ................................ ................................ ................................ ..... 152 Age ................................ ................................ ................................ .......... 152 7 DISCUSSION AND CONCLUSION ................................ ................................ ...... 162 The Initial (R) ................................ ................................ ................................ ........ 1 62 The Consonant Cluster (R) ................................ ................................ ................... 166 The Status of the Variable (R) in Bangkok Thai ................................ .................... 170 Broader Implications ................................ ................................ ............................. 176 Limitations of the Study ................................ ................................ ......................... 179 Suggestions for Future Research ................................ ................................ ......... 180 Closing ................................ ................................ ................................ .................. 182 APPENDIX A INFORMED CONSENT FORM S ................................ ................................ .......... 183 English Version ................................ ................................ ................................ ..... 184 Thai Version ................................ ................................ ................................ .......... 186 B THE SOCIOLINGUISTIC INTERVIEW TOPICS ................................ ................... 187 C READING PASSAGES ................................ ................................ ......................... 188 Reading Passage 1 ................................ ................................ .............................. 188 Reading Passag e 2 ................................ ................................ .............................. 189
9 D QUESTIONNAIRE ................................ ................................ ................................ 190 LIST OF REFERENCES ................................ ................................ ............................. 191 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 204
10 LIST OF TABLES Table page 1 1 The Thai consonant phonemes ................................ ................................ .......... 31 1 2 The ini ti al consonant clusters in Thai ................................ ................................ .. 31 1 3 The Thai consonan ts in a syllable final position ................................ ................. 31 4 1 Demographic information of particip ants ................................ .......................... 103 4 2 Coding factor groups of the initial (r) on Goldvarb X ................................ ......... 103 4 3 Coding factor groups of the consonant cluster (r) on Goldvarb X ..................... 103 5 1 Overall distribution of the four variants of the initial (r) ................................ ...... 126 5 2 Distribution of the four variants of the initial (r) by age ................................ ...... 126 5 3 Distribution of the four variants of the initial (r) by gender ................................ 126 5 4 Distribution of the four vari ants of the initial (r) by education ............................ 126 5 5 Distribution of the four variants of the initial (r) by speech style ........................ 126 5 6 List o f the dependent and independent factor groups of the initial (r) ............... 127 5 7 Multivariate analysis of the contribution of the extralinguistic factors selected as significance to the probability of the initial (r) ................................ ............... 127 6 1 Overall distribution of the four variants of the consonant cluster (r) .................. 154 6 2 Distribution of the four variants of the consonant cluster (r) by age .................. 154 6 3 Distribution of the four variants of the consonant cluster (r) by gender ............. 154 6 4 Distribution of the four variants of the consonant cluster (r) by education ........ 154 6 5 Distribution of the four variants of the consonant cluster (r) by speech style .... 154 6 6 List of the dependent and independent factor groups of the consonant cluster (r) ................................ ................................ ................................ .......... 155 6 7 Multivariate analysis of the contribution of the ext ralinguistic factors selected as significant to the probability of the consonant cluster (r) ............................. 155
11 LIST OF FIGURES Figure page 1 1 Map of Thailand wi th the four major regions ................................ ....................... 32 1 2 Map of Thailand illustratin g the Bangkok metropolitan area ............................... 33 1 3 Variants of the /r/ realizati on in terms of phonological and extralinguistic factors ................................ ................................ ................................ ................. 34 4 1 Proportion of education level of the participants ................................ ............... 104 5 1 Overall distribution of the four variants of the initial (r) ................................ ...... 128 5 2 Distribution of the four variants of the initial (r) by age ................................ ...... 128 5 3 Distribution of the four variants of the initial (r) by gender ................................ 129 5 4 Distribution of the four variants of the initial (r) by education ............................ 129 5 5 Distribution of the four variants of the initial (r) by speech style ........................ 130 5 6 Distribution of the four variants of the initial (r) by age and speech style .......... 130 5 7 Distribution of the four variants of the initial (r) by age and gender ................... 131 5 8 Distribution of the four variants of the initial (r) by gend er and speech style ..... 131 5 9 Effect of speech style on the production of the lateral of the initial (r) ............... 132 5 10 Effect of age on the production of the lateral in the initial (r) ............................. 132 6 1 Overall distribution of the (r) va riants of the consonant cluster (r) .................... 156 6 2 Distribution of the four va riants of the consonant cluster (r) by age .................. 156 6 3 Distribution of the four variants of the consonant cluster (r) by gender ............. 157 6 4 Distribution of the four variants of the consonant cluster (r) by education ........ 157 6 5 Distribution of the four va riants of the consonant cluster (r) by spe ech style .... 158 6 6 Distribution of the four va riants of the consonant cluster (r) by age and speech style ................................ ................................ ................................ ...... 158 6 7 Distribution of the four va riants of the consonant cluster (r) by age and gender ................................ ................................ ................................ .............. 159
12 6 8 Distribution of the four va riants of the consonant cluster (r) by gender and speech style ................................ ................................ ................................ ...... 159 6 9 Effect of speech style on the zero realization of the consonant cluster (r) ........ 160 6 10 Effect of education on the zero realization of the consonant cl uster (r) ............ 160 6 11 Effect of gender on the zero realization of the consonant cluster (r) ................. 161 6 12 Effect of age on the zero real ization of the consonant cluster (r) ...................... 161
13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philos ophy SOCIAL AND STYLISTIC VARIATION OF (R) IN BANGKOK THAI By Orapat Pookkawes August 2014 Chair: HÃ©lÃ¨ne Blondeau Major: Linguistics This dissertation investigates the variable (r) in two linguistic contexts: the onset and initial consonant clusters. It further addresses the correlation of the variable (r) with age, gender, education, and style to demonstrate language behavior s of native speakers in relation to extralinguistic factors . Speech d ata from 30 native speakers of Bangkok Thai were collected in an interview and short passage reading, followed by a debriefing questionnaire to obtain their language attitude s . Res ults indicate that, among the potential variants, the nonstandard lateral is the most favored in onset (r), followed by the standard flap, the standard trill, and the nonstandard ap proximant. In the clusters, speakers tend to omit (r) as a second unit at the highest rate , followed by using the standard flap, the nonstandard lateral, and the standard trill. A nalys e s suggest that social a nd stylistic factor s tend to affect the choice of (r) variants in both linguistic contexts. A r egular increase of the lateral for initial (r) and the (r) omission in the clusters among younger speakers seem to suggest an ongoing change of (r). Also, men te nd to use the lateral, whereas women seem to use the flap and trill more frequently in onset (r) . In contrast, a high rate of (r) deletion of the clusters among women s uggests that they are not always sensitive to the standard language .
14 S peakers with a hig her level of education tend to preserve their standard (r) variants more frequently than s peakers with a lower level of education in both contexts . Clear cut differen t usage of (r) variants in the onset and cluster (r) in different speaking situation s show s that style seems to be a n important factor that connect s self awareness of using standard language in formal speech. Findings and the opinions from the questionnaire indicate that (r) in both linguistic contexts seem s to funct ion as a sociolinguistic marker , since there is social and stylistic interpretation overtly attached. This dissertation helps illustrate the current (r) variation among native speakers of Bangkok Thai and suggest s the direction of change. The study also h as imp lications for language teaching and research on language variation and change.
15 CHAPTER 1 INTRODUCTION AND BACKGROUND INFORMATION Change is a cons tant phenomenon in the world. L anguage is also chan ging continuously , and t here is no change witho ut variation. Guy ( 2003 ) suggests that l inguistic variation and linguistic change are considered two different aspects of the same phenomenon . Language change occurs in the context of linguistic variation , and linguistic variation reflects change and varia tion in society (Coates 2004). Research on language variation bring s us a clear perspective on language change by examining the actual speech production that reflects language behavior s of speakers in speech communities . Every language changes periodicall y and i t is unlikely that any communities change their language features abruptly without having a period when an old variant and a new variant co occur . Both old and new forms represent the linguistic variables in the usage of individual s in the transitio nal generations , occurring as ongoing changes in the community . The Thai language i s no exception as well . Bangkok Thai changes over a period of time and socially varies within the speech community. At the present time, t he language is greatly dissimilar with respect to the phonological, morphological, semantic , and discourse levels from the Thai language of many decade s ago (Hudak 2009). The present dissertation investigates the variable (r) in Bangkok Thai, focusing on the onset position and initial con sonant clusters, and the correlation of the variable (r) with social and stylistic factors used by native speakers of Bangkok Thai.
16 Language Variation in the Thai Language The examination of language variation and language change bridge s the present and th e past. One of the major objectives in the study of language variation is to investigate changes at the present time in order to uncover language phenomena that occurred in the past. The literature of sociolinguistic variatio n in Thailand has also indicate s ch anges of different levels of the language, such as phonological and tonal differences , as important evidence of language variation and change in different dialects of the Thai language, such as Bangkok (r): Beebe (1974), Chunsivimol (1993); Bangkok (k h ): Chalermsanyakorn (2000); Chiengmai unaspirated stops: Tan w attananun ( 1982 ) ; Suphanburi high falling tone: Kobsirikarn ( 1991 ) ; : Sapproong (19 93 ) ; Nakhon Si Homkaew ( 1997 ) . The findings have revealed that the linguistic variabl es of different dialects of Thai have their variants constrained by extralinguistic factors , such as age, gender, education level , occupation, social class, and speech styles. Purpose and Rationale of the Study The present research aims to demonstrate a current language situation and language behavior by exploring the use of the initial (r) and the consonant cluster (r) in the speech production of native speakers of Bangkok Thai. The examination is related to the extralinguistic factors in an effort to di scover social and stylistic constraints on language variation in Bangkok Thai. This study is also complemented by metalinguistic his study explores the sociolinguistic status 1 of t he linguistic variable (r) as a means of 1 See Chapter 2 for further information of the sociolinguistic status of the linguistic vari able (Labov 1972a).
17 understanding the interaction between language use and society in the speech community of Bangkok. There are several reasons why the present study concentrat es on an examination of the variable (r). First, the sound /r/ is the center of attention among researchers because of its richness in variation. It is suggested that the majority of world languages have at least one /r/ phoneme (Maddieson 1984). Ladefoged and Maddieson (1996) state that r sound or rhoticity 2 is defined as rather heterogeneous and somewhat arbitrary. This might be due to the fa ct that different type s of rhotic sounds are considered to be greatly varied in both places and manners of articulation . According to a phonetic perspective, a r hotic or r l ike sound can be realized in many different ways , ranging from vocalic to fricative (Wiese 2001). Lindau (1978) notes that it is difficult to phonetically characterize a single articulatory to rhotic consonants . However, /r/ consonants have been suggested to behave in similar phonological ways and have certain similar phonological features in different languages (Weise 2001). For example, in languages with consonant clusters, r sounds are likely to occur close to a syllable nucleus, and the postvocalic /r/ is likely to become a vowel or disappear altogether, such as in the Southern British English post alveolar approximant (Lindau 1985:157). In addition, Ladefoged and Maddieson (1996) suggest that /r/ tends to have syllabic variants, or merge with contiguou s vowels. This process is useful for analyzing phonetic differences between dialects of the same language . A broad natural class of /r/ consonants implie s potential variation within the class of rhotics , making the /r/ sounds a significant area for examin ation in the fields of 2 A character in an orthographic system derived from the Greek letter rho ise 2001:339).
18 sociolinguistics and language variation. Scobbie (2006:337) also suggest s that it is common to find sociolinguistic conditioning of variation in a rhotic sound within a language . I t is observed that many r sounds are often members of the same phoneme 3 when one type of a rhotic alternates with other rhotics in many languages . For instance, /r/ in Palauan , the language in the Austronesian family, is generally realized as a tap in intervocalic and postvocalic contexts, but an approximant in word initial position (Inouye 1995). In Huasa, the African language , /r/ is realized as a tap or approximant between vowels and as a trill before a consonant or in an initial position (Lindau 1985). In addition , it is noticeable in previous studies th at the variab le (r) is sensitive and subject to variation in the sociolinguistic configuration in many languages of the world. The present study investigates the (r) variation in Bangkok Thai to reveal the patterns of the (r) usage among native speakers at the present time. In terms of sociolinguistic orientation, it is evident that the variable (r) has been examined by renowned researchers in many languages and different speech communities, such as New York English: Labov (1966), Mather (2012) ; Montreal Fr ench: Clermont and Cedergren (1979) , G. Sankoff and Blondeau (2007, 2013 ); Mexican Spanish: Rissel (1989); Khmer dialects in northeastern Thai: Premsrirat (1995) ; Scottish English: Romaine (1978) ; and Middlesbrough English: Llamas (2001). Research on the v ariable (r) in different languages shows that (r) carries social and stylistic significance, and suggests the process of change in progress. A further overview of the literature of the variable (r) in different languages regarding social and stylistic fact ors is addressed in Chapter 3. 3 However, there are languages in which r sounds belong to different phonemes. For instance, t wo phonemes of /r/ in Spanish are distinguished from each other, according to the presence and absence of trilling . For example, is d i fferentiated from tense] in a medial position. The [tense] feature is not specified for initial /r/ in Spanish ( Cressey 1978).
19 Next , the present research is anticipated to fill the following major gap s in the literature of Thai language variation research , since ther e are limitations in previous studies of the variable (r) in Thai that can be furthe r investigated. The variable (r) has been one of the linguistic variable s of central interest in the study of the Thai language . Research o n language variation has demonstrate d the association of the variable (r) with res pect to internal and external facto rs on the basis of the speech production of native speakers. For instance, Beebe (1974) suggests a relationship between the social al consonant cluster (r). Chunsuvimol (1993) has found social variation of Thai (r) and English (r) with regard to sex, job level, and English proficiency of native speakers of Ba ngkok Thai. In addition, Pulsup (1993) has examined the stylistic variation of (r) used by young native speakers of Bangkok Thai. It is evident that adolescents vary their (r) production according to the formality of styles. Nonetheless, different limitati ons occurred in previous research, such as data collection procedures, methodology, and characteristics of native speakers (e.g., imbalanced age ranges, imbalanced total number of speakers i n each age and gender group ). Such limitations motivate the presen t study to be more carefully designed for the exploration of the variable (r) in terms of the methodology and selection of the extralinguistic factors. Three successive generational age groups , gender, and speec h styles are selected as primary independent variables, whereas education levels of the speakers serve as a complementary variable to examine the relationship between such external factors and the (r) variation by integrating the interspeaker and intraspeaker dimensions.
20 Although some previous studie s on the Thai (r) have adopted the Labovian t heor etical framework to examine language variation, more studies are necessary to demonstrate the current patterns of (r) variation by using more recent approaches of variationist sociolinguistics. The variation ist framework has been greatly developed in different aspects in the past f ew decades. For instance , in terms of statistical examination, the findings in the previous literature of (r) variation in Thai mostly use a statistical test , such as chi square, to look at the effect of only one factor at a time. More recently developed statistical programs in variationist analysis, such as Goldvarb and other VARBRUL programs that adopt a multifactorial approach , were infrequently applied to ascertain the statistica l signifi cance of the variables under examination. In contrast to the previous studies , th e present research applies the chi square test and the novel statistical method , namely Goldvarb X, to highlight the significance of specific linguistic and extraling uistic factor groups that influence the choice of (r) variants. Lastly, few studies in the literature have described the particular status of the distribution of the variable (r) in the speech co mmunity of Bangkok. The present study offers a discussion of a current sociolinguistic status of the variable (r) in Bangkok Thai on the basis of and language attitudes , complemented by the metalinguistic commentaries of the participants in the form of responses to a short questionnaire. The aforementioned points lead the present research to investigate the variable (r) with regard to the following research questions in order to achieve the three main goals of this study.
21 Research Questions Three major goals are ad dressed in the present research. The first goal is to explore general patterns of the sociolinguistic variable (r) in two linguistic contexts in the speech of native speakers of Bangkok Thai; namely, onset and consonant cluster s. The second goal is to disc over the correlation between the linguistic variable (r) and the extralinguistic variables in order to see whether the use of the variable (r) from native speakers of different ages, gender, and education levels are similar to or different from each other in terms of inter individual differences. In addition, the study explores how participants use the variable (r) in different speech styles to explore how intra individual variation comes into play. The last goal is to explore the relationship among all the extralinguistic variables to determine their relative importance in terms of the choice of the (r) variants in both contexts. In ot her words, the present study examine s the following questions: 1. How and t o what extent does the (r) variation occur in Ba ngkok Thai? What are the (r) variants used by native speakers? What is the distribution and frequency of each variant? How do native speakers use the variable (r) in two linguistic contexts ? 2. How and to what extent do age, gender, education, and speech style affect the patterns and the frequency of the (r) variants used in the two linguistic contexts? How do younger, middle aged , and older speakers use the variable (r)? How do female and male speakers use the variable (r)? How do speakers with lower a nd higher levels of education use the variable (r)? How do native speakers use the variable (r) in informal and formal s peaking setting s?
22 3. What is the relationship among different extralinguistic factors and the use of the (r) variants? Ho w do female and male speakers in different age groups vary their use of the variable (r) in informal and more formal speech styles? Hypotheses Corresponding to the research questions above, it is hypothesized that the social and stylistic factors affect the choice of the variants of (r) in the initial position and the consonant clusters used by native speakers of Bangkok Thai. First, age is hypothesized to affect a choice of the (r) variants. The older speakers m ight conserve their standard speech more t han the speakers in the other two age groups, whereas t he younger speakers might use the nonstandard variants the most frequently among the three groups of speakers . Second, it is expected that women might use standard forms of (r) more frequently than men , since female speakers tend to favor the use of standard language , and they are more sensitive to the prestige forms than male speakers ( Trudgill 1972; Labov 2001; Wolfram and Fasold 1974 ) . Third, it is hypothesized that the speakers with a higher level o f education would show a higher rate of the standard variants of (r) than a group of speakers with a lower level of education. In contrast, people with a lower level of education would use the nonstandard variants at a higher proportion. Lastly, speakers a re hypothesized to be more careful of their (r) production by using standard (r) variants more frequently when the degree of formality increases . Chapter 5 and Chapter 6 present the major findings of the variable (r) in the initial position and the consona nt clusters that support and reject these hypotheses according to the extralinguistic factors under examination.
23 Organization of the Dissertation The present dissertation consists of seven chapters. Chapter 2 provides the theoretical framework, variati onist sociolinguistics, adopted in this study. It presents the major principles and methodology of language variation study. It also provides details of social and stylistic dimensions in order to demonstrate the correlation of such factors in language var iation. Chapter 3 reviews the literature relevant to the sociolinguistic variable (r). It presents the findings of previous studies of (r) variation in different languages as well as research on the variable (r) in the Thai language. Chapter 4 presents the methodology applied in the present research and reviews methods of data collection and data analysis. It also contains a description of the linguistic and extralinguistic factors under examination. Chapter 5 demonstrates the major findings of the variable (r) in the first linguistic context analyzed: the initial position. It also provides an interpretation of the initial (r) with regard to age, gender, education, and style, the relationship among the extralinguistic factors, and statistical results. Chapte r 6 presents the key findings of the variable (r) in the second linguistic context analyzed: the consonant clusters. It also addresses an interpretation of the consonant cluster (r) according to social and stylistic factors, the relationship among extralin guistic factors and findings from a statistical analysis. Lastly, Chapter 7 provides a summary of the variable (r) in the two linguistic contexts, interprets the results, and discusses the current sociolinguistic status of the variable (r) in Bangkok Thai. In addition, it addresses broader implications for the field of variationist study and pedagogy, mentions limitations of the present study, and provides suggestions for future research. The following sections provide an overview of the Thai language and the city of Bangkok, a description of the Thai consonant inventory, and the major variants of (r)
24 found in Bangkok Thai in order for the reader to understand the general background of the Thai language and Bangkok . Overview of the Thai L anguage According to the 2013 census 4 , the Thai language is spoken by approximately three quarters of the population of 68 million in Thailand. It is an uninflected, mostly monosyllabic , and tonal language in the Tai Kadai language family (Slayden 2010). Thai is morphologic al isolating with a small number of free derivational morphemes used to create new words (Phootirat 2012). Th e Thai language does not rely highly on inflectional morphemes to give meaning to a word. Numerous Thai words were based on and borrowed from other languages, such as Pali, Sanskrit, and Mon Khmer . The Thai writing system , called the Sukhothai Script, was first introduced by King Ramkhamhaeng in AD 1283 during the Sukhothai period (Slayden 2010). Th is script was used until AD 1357. In the same year, invented during the reign of King Li Thai, the grandson of King Ramkhamhaeng (Ronnakiat 1997). The King Li Thai script changed through time, and it was used until the early period of the Ayuddhaya Kingdom . In 1 680, the new Kin g Narai script was introduced in . The script has been preserved and developed in the Thai writing system up to the present time (Ronnakiat 1997). Standard Thai is the national and official language of Thailand. It is the prestigious variety of the Thai language that children learn at school , and it is the medium of mass media, internal government affairs, and high prestige cultural activities (Huebner 2006). Additionally , t he standard Thai is used as the lingua franca amon g 4 http://worldpopulationreview.com/countries/thailand population/
25 native speakers of Thai from different parts of the country, and it is the language that is taugh t to non native speakers as a second or foreign language (Prasithrathsint 1996). In addition to Standard Thai, four regional dialects are broadly divided base d on geographical areas of the country : Northern Thai, Northeastern Thai, Central Thai, and Southern Th ai, as illustrated in Figure 1 1 . Although the four regional dialects are closely related to each other, they have distinct features in terms of vocabula ry, tones and phonemic inventory. For instance , the word in Thai uses different terms in the four regional dialects: Northern Thai: [ ] 5 ; Central Thai: [ p h Ã»:t ] 6 ; Northeastern Thai: [wÃ¢w] 7 ; and Southern Thai: [ ] . Overview of the City of Bangkok Bangkok is the capital city of Thailand. The city occupies approximately 605 square miles in the Chao Phraya River delta, located in the central part of Thailand. Bangkok is the most populous city of the country according to the 2013 census 8 . The greater Bangkok area is more populated than other urban areas in the country; 14 million people reside in the Bangkok m etropolitan region 9 . Figu re 1 2 present s a map of Thailand with five nearby cities Nontaburi , Pathumthani, Samutprakarn, Samutsakhorn, and Nakhonpathom area . Bangkok is the home of native Bangkok r esidents and people from other parts of 5 The phonetic transcriptions of Thai examples use the International Phonetic Alphabe t (IPA) symbols. 6 A superscript [ h ] is not an independent segment. It is treated as an aspiration feature co occurring with the first consonant that is always a voiceless stop. 7 Diacritical marks indicate five tones in Thai phonetics: none mid, Ã low , Ã¢ falling, Ã¡ high, and rising. 8 The Institute of Population and Social Research, Mahidol University , Bangkok. 9 http://worldpopulationreview.com/countries/thai land population/
26 the country and the world, such as Chinese, Japanese, Indians, Americans, Europeans, and immigrants from neighboring countries (e.g., Cambodians and Burmese 10 ) who come to Bangkok for higher levels of education, busin ess, and job opportunities . Bangkok Thai is a major dialect spoken in Bangkok and other contiguous areas of the city , as shown in Figure 1 2 . It refers to the speech of all native speakers of Bangkok (Beebe 1974). It is noted that Bangkok Thai and Standar d Thai are not the to be used most in Bangkok, it is easy to assume that it is the same as Bangkok Thai ords, Bangkok Thai is a geographical dialect which contrasts with other geographical dialects. However, some researchers, such as Bickner an d Hudak (1990) , consider Bangkok Thai a standard variety, used in academia and news broadcasting nationwide. In addi tion to Bangkok Thai, other varieties of languages are found in different ethnic communities in Bangkok, such as Japanese around the Phromp hong neighborhood , and Hindis in Phahurat, the Indian community. Teochew Chinese is another dialect spoken by the lar gest Chinese population in Chinatown (Panyaathisin 2013) . English is one of the dominant languages used in the country. It is the school requirement for children to learn at an elementary level . pecialized among Western expatriates, and is used by native speakers of Thai with high English proficiency. English is also the language on store signs in many places that foreigners frequent (Smalley 1994). 10 http://en.wikipedia.org/wiki/Thailand#Geography
27 Description of Thai Consonant Inventory Standard Thai consists of 21 consonant phonemes that contrast in the word initial position with 44 graphemes corresponding to the phonemes in the writing system. Table 1 1 pr esents the Thai consonant phonemes and their representations. In terms o f the structural phonotactics, the model of the Thai syllable is C 1 (C 2 )V 1 (V 2 ) T (C 3 ) (Iwasaki and Ingkaphirom 2005 , cited in Slayden 2009). The units shown in parentheses suggest that t 1 2 represent onset consonant segments, which are allowed maximally to two consonants in a word initial position. The combination of initial consonant clusters is limited to only three voiceless stop s: / p /, / t /, and / k /. The liquid or glide consonants : /r/, /l/, and /w/ function as a second unit in order to make 11 possible initial consonant clusters , as presented in Table 1 3 final position. Only on e consonant is allowed in the coda position , and it is restricted to eight consonants, as shown in Table 1 1 The vocalic nucleus can be long or short. If it is long, a final consonant is not required, whereas a final consonant is obligatory 2 is a diphthong vowel, occurring with the vowel /a/ in open and closed syllables , such as /, / /, and / one of the five tones : mid, low, falling, high, or rising. Major Variants of ( R ) in Bangkok Thai This section provides information specifically on the major variants of (r) occurred in Bangkok Thai from the literature. The phonetic properties and extralin guistic factors related to the (r) variants are also addressed.
28 Researchers found different (r) variants in previous studies of (r) in Thai . For instance, Harris (1972) suggests four variants of /r/ production that are commonly used in the speech of nativ e speakers. These variants share the same point of articulation, but they are different with regard to the manner of articulation (trill, tap, lateral, and approximant). Palakornkul (1972) characterizes /r/ in Thai as a trill or an approximant, whereas Tin gsabadh and Abramson (1993) and Harris (1996) suggest that /r/ is a trill or a flap. Phonetically, a trill and a tap are different in their phonetic properties. That is, a trill is made with a few contacts of the tongue tip against the alveolar area to mak e multiple periods of vibration. A tap, in contrast, is made with a brief contact between the tongue and the alveolar area (Ladefoged 1971; Ladefoged and Maddieson 1996). Additionally, many linguists suggest a distinction between a tap and a flap 11 . Ladefog ed (1971), Catford (1988), and Ladefoged and Maddieson (1996) point out that the direction of movement differentiates a flap from a tap. That is, a flap is made with a brief contact between the articulators when the tongue f the moving it forward so that it strikes the upper surface of the vocal tract (Ladefoged and Maddieson 1996:231) . A tap, on the other hand, is the sound made when the tongue moves directly to contact However, many linguists , such as Lindau ( 1985) , do not note any differences between a tap and a flap , and term both sounds a flap with a phonetic symbol , [ 11 I use the term flap t hroughout the present study.
29 Harris (1972:22) additionally suggests the devoiced variants of /r/ occurring after aspirated voiceless stops: /p h / and /k h / of initial consonant clusters. Only a devoiced tap [ ] and a devoiced lateral [ ] are found in casual speech, while a devoice d trill [r ] and a ] are infrequently used. In terms of the social significance of the sound /r/, previous studies (e.g., Beebe 1974; Treyakul 1986; Chunsuvimol 1993; Harris 1996) have suggested that a trill [r] and standard and prestigious variants of (r) in the onset . A flap and a trill [r] as a second unit of the initial consonant clusters are standard in Bangkok Thai. The production of these (r) variants is also affected by linguistic and extralinguistic facto rs. Speakers freque ntly use a flap in everyday conversation while a trill is infrequent in casual utterances. A trill is mainly found in emphatic speech and in more formal speaking settings, such as formal broadcasting and instruction at schools (Beebe 197 4; Treyakul 1986; Harris 1996). Other nonstandard variants of the initial (r) found in the literature (e.g. , Harris 1972; Beebe 1974; Chunsuvimol 1993) are a lateral [l] and an approximant is considered an atypical variant because it is used in the speech of speakers with a high fluenc y of English (Bickner and Hudak 1993) . The zero realization or null variant [Ã¸] is stigmatized, occurring in the consonant clusters when speakers omit (r) in the second segment of the clusters (Chunsuvimol 1993). Figure 1 3 presents the variants of the (r) realization, possibly occurring in the ini tial position and the consonant clusters according to the phonological and extralinguistic factors. This chapter has provided the purpose and rationale of the present research. It includes a background of the Thai language, demographic information , and the
30 languages used in Bangkok. A phonological description of the Thai consonant inventory and the major variants of (r) found in Bangkok Thai from the literature are additionally addressed. Next chapter discusses the theoretical framework adopted in the present study.
31 Table 1 1. The Thai consonant phonemes* Bilabial Labio Dental Alveolar Post Alveolar Palatal Velar Glottal Aspirated stop p h t h k h Unaspirated stop p, b , t, d , c k Nasal m n Fricative f s h Trill r Lateral Approximant l Semivowel w j * Adapted from Slayden (2009) and Phootirat (2012). Table 1 2. The initial consonant clusters in Thai* Bilabial Alveolar Velar Aspirated voiceless stop p h r t h r** k h r p h l k h l k h w Una spirated voiceless stop pr tr kr pl kl kw * Adapted from Beebe (1974) and Phootirat (2012). ** Beebe (1974:53) suggests that the t h r cluster occurs only in a few words that are derived from Sanskrit. Table 1 3. The Thai cons onants in a syllable final position * Bilabial Alveolar Palatal Velar Stop p t k Nasal m n Semivowel w j * Adapted from Beebe (1974).
32 Figure 1 1. Map of Thailand with the four major regions from http://www.thailanguagehut.com/vocabulary thai/lear n thai telling/
33 Figure 1 2. Map of Thailand illustrating the Bangkok m etropolitan area from http://en.wikipedia.org/wiki/Bangkok_Metropolitan_Region
34 [ everyday speech [r]: a voiced apical alveolar trill, standard variant, mainly occurs in emphatic speech and more formal speaking occasions [ l]: a voiced apical alveolar lateral, nonstandard variant, commonly found in casual speech a voiced apical alveolar a pproximant , nonstandard variant, occurr ing in the speech of speakers with high English proficiency ]: a d evoiced apical alveolar flap voiceless variants , [r ] : a d evoiced apical alveolar trill additionally found [ ] : a devoiced apical a lveolar lateral after an initial ]: a d evoiced apical alveolar approximant aspirated voiceless stop s in consonant clusters [Ã¸]: a zero realization or null variant, nonstandard variant, found in consonant cluster (r) in casual speech Figure 1 3. Variants of the /r/ realization in terms of phonological and extralinguistic f actors . Adapted from Harris ( 1972 ) and Phootirat (2012 ). /r/
35 CHAPTER 2 THEORET ICAL FRAMEWORK V ariationist sociolinguistics approach is the theoretical framework adopted in th is dissertation . The key principles and methodology of language variation study from variationist scholars are presented. In addition, the extralinguistic factors of social and stylistic dimensions are provided in order to demonstrate the correlation of th ese factors in language variation. The Study of Language Variation: V ariationist S ociolinguistics Labov. His research presents the association of language variation, such as phonological forms with their social meaning. Unlike other linguistic perspectives which state that language is homogeneous (e.g., Generative Grammar) , Labov asserts that variation in language can be systematically explained by the influence of external fa ctors such as age, gender, s ocioeconomics, and speech style. For data collection, Labov ( 1963, 1966 ) developed sets of language variatio n method s , such as the sociolinguistic interview, rapid and anonymous survey , written questionnaire, sampling strateg ies , and quantitative analysis . These data collection technique s have been adopted as a model of variation examination for numerous sociolinguistic studies in different large urban settings, such as New York City: Labov ( 1966 ), Fowler ( 1986 ), Mather ( 2012 ); N orwich: Trudgill ( 1974 ); Montreal: Sankoff and Sankoff ( 197 3 ) ; Sydney: Horvath (1985) . Quantitative Paradigm The q uantitative paradigm, an impo rtant approach in variationist socioli nguistics, suggests that variability is a fundamental part of linguistic c ompetence. An important
36 perspective of of human beings to accept, preserve , This approach is derived from the model proposed by Weinre ich, Labov, and He rzog (1968:188) , who note that involves change ; but all change involves It is stated that language is order ly heterogeneous, containing inherent variables influenc ed by the co variation of linguistic and extralinguistic features. This account is different from that of Chomsky (1965) in his G enerative Grammar. Chomsky views language as a n entity with homogeneous structures. perspective (1965:3 ), he states that : Linguistic theory is concerned primarily with an ideal speaker listener , in a completely homogeneo us speech community, who knows its (the speech irrelevant conditio ns as memory limitations, di stra ctions, shifts of attention and interest, and errors in applying his knowledge of this language in actual performance. The q uantitative approach is elaborated beyond what is explained and included in a generative framework. The majority competence is considered var ative paradigm. The variable rule was developed in thorough exploration of the variability of nonstandard Black English (NNE) 1 , such as the final t, d deletion, copula, and an auxiliary be . Labov (1972a) states that there is inherent variability in language that occurs at different levels of human languages (e.g. , phonology, m or phology, syntax, semantic, discourse ), in every dialect, in every variety of a language , and in every speaker (Tagliamonte 2012). The variation in the language does not random ly occur . Rather, it is 1 NNE refers to N onstandar
37 systematic and socially conditioned (structured variabi lity (Cameron 2009)) . The evidence is proven by various studies at different levels of the linguistic structure , such as studies on phonological variation of the suffix ing in English ( Fischer 1958; Labov 1966; Trudgill 1972 ; Chamber 2002b ), and final t , d deletion in English (Guy 1980 ; Tagliamonte and Temple 2005 ). Morpho syntactic research includes the study of nonstandard morphological and syntactic features of English such as the nonstandard s, never, do and auxiliary have in the speech of working cla ss adolescents (Cheshire 1982), and the variable of third person plural in French verbs (Nadasdi 2001). A syntactic variable study is, for example, the object clitic placement in Cypriot Greek (Pappas 2008 , cited in Walker 2010). S ome s tudies on language v ar iation in discourse are the pragmatic expressions of you know and I mean in English ( Ã– stman 1981; Holmes 1986), pre verbal do in Somerset English 2 (Jones and Tagliamonte 2004), and be like of quotative verbs in Canadian English ( Tagliamonte and Hudson 19 99; Tagliamonte 2005). In addition , empirical studies of language change , using the quantitative principle , have demonstrated that linguistic change has its origin in synchronic variation or in the heterogeneity of the speech community. In other words, syn chronic variation reflects diachronic change . Bailey (2002: 312) states that of language change, the study of change in progress, forms one of the cornerstones of The synchron ic approach wa s introduced t York City (1966). The model enables researchers to investigate the progress of 2 Somerset is a variety of English spoken in Southwest England.
38 language changes at the time it occurs. The concept of apparent time 3 as a surrogate of real time analysis has been corroborated in various studies of language variation and change, such as Bickerton (1973) , Kay (1975) , Bailey et al. (1991), and Labov (1994). In the quantitative paradigm, variants of a linguistic vari able are considered from the production of different ways of saying the same underlying form. The selection of one variant over another is systematically chosen by internal and external factors. Young and Bayley (1996:253) note, in their terminology , principle of multiple th multiple contextual factors (e.g., linguistic environment and social phenomena ) . T he frequency of each occurred variant is analyzed according to different linguistic and social factors. Linguistic or i nternal factors are determined by particular linguistic contexts , co occurring with the forms of a linguistic variable under examination . Social or e xternal factors, on the other hand, commonly incl ude education background, socioeconomic status , and speech style . Bayley ( 2002 ) also suggests that an important theoretical principle underlying the quantitative study of language variati on. It is claimed that l anguage change can be quantified with the pattern of variation. That is, quantitative statements are identified about language use . Analysts can count the frequency of a particular occurrence of a variant over others. It also allows researchers to analyze according to its different linguistic and social factors, and determine a proportion of the total number of occurrences on which form could have occurred, even if it did not. This way, researchers are able to see the patterns and te ndencies from the 3 Apparent time construct is a form of synchronic analysis , w here language is analyzed for a particular point in time .
39 data and determine the characteristics of particular factors that influence the c hoice of variants selected in different environments. Also, the use of quantitative evidence enables researchers to discover which linguistic and/or social f actors favor or disfavor the occurrence of each variant . By using a statistical software program , such as VABRUL and G oldvarb, r esearchers are able to describe and postulate the linguistic phenomena, such as the regular property of language variation, chan ges in progress , and linguistic systems. The Linguistic Variable Labov (1972a) divides the linguistic variables into three categories: indicators, markers and stereotypes according to a level of sociolinguistic salience. An indicator is a linguistic variab le that shows social variation, but that has no style shifting across speech settings. For instance, a vowel (a:) in words, such as and in Norwich English (Trudgill 1974) is not affected by speech styles. That is, Norwich speakers continue us ing a back vowel of (a:) 4 regardless of speech styles of different speaking situations. Another example of indicators is different vowels in words, such as and in American English. Whether the speakers are able to distinguish the vowels of s uch words does not affect the social interpretation (Wardhaugh 2006:148). In addition, indicators are not stratified by the ages of speakers, thus there is no interpretation of change in progress (Tagliamonte 2012). A marker, on the other hand, carries soc ial and stylistic significance. That is, markers vary with respect to the social factors of speakers, such as age, gender, and social class. Also, markers show variation within individual speakers according to the formality of a speaking situation. 4 The standard variety of British English uses a front (a:) vowel (Tagliamonte 2012).
40 Markers have the overt social interpretation attached to them. Classic sociolinguistic variables, such as (ing), (r), and (th) in New York English (Labov 1966) are considered markers because they mark with some social characteristics and styles of speaking situat ions. For example, speakers in New York City are aware that the r less sound in a word, such as asso ciates with the speech of lower class speakers. Therefore, New York speakers may alternate their production of the (r) according to different situatio ns. A stereotype is a highly recognized characteristic feature with obvious social significance of the speech of speakers in a particular group. Additionally, linguistic be havior. Stereotypes receive great attention, and speakers are great ly conscious of a highly stigmatized feature of stereotypes. Also, stereotypes are generally a topic of discussion for members of the community in which they are known, and also for outside rs (Tagliamonte 2012). Examples of local stereotypes include the stereotypical Brooklynese pronunciation of for 3 3rd Street in New York City, and cah and pahk for and park (Labov 1972a:248). Variationist analysts are interested in the inherent variability of language. In a linguistic a linguistic variable. For conversation; as Sapir (1921:147) not es, whereas D. Sankoff (1988a:142) states that linguistic variables must be alternatives within the same grammatical system which have the same referential meaning in
41 running discourse. In his more recent work, L abov (2004:7) explains that variable i s the first and also the last procedure in the analysis of variation. It begins with the overt manifestations: with phonemes we hear allophones; with variables, we hear The linguistic variable is the choice of variant that individual speakers make alternating these variants in their conversation. Variants are the actual speech production of each linguistic variable, and they must occur in the same context. For year , I visited my family in three c ities: Chicago for about two weeks, Atlanta for approximately three weeks, and Miami for like alternation of three variants of approximation: about , approximately , and like . A selection of one variant over anot her must be systematic (structured heterogeneity ) and involve patterns of linguistic and social factors. The variable comprises different levels of linguistic features, such as phonetic, phonological , morphological, syntactic , and there are at least two or more variants. In the level of phonology, the linguistic variable is straightforward, in which variants may alternate by one or more phonological features . A classic example of the phonological variable in English is, for example, the variable (ing) . The variable (ing) occurs at word final, consist ing of two variants of the velar [ ] and the words , such as talking versus talkin . However , when the focus is on other levels of language above phonology , such as syntax and semantics, questi ons arise whether or not it i s two different ways of saying the same things . Fo r
42 instance, the presence and absence of the suffix ( ly ) 5 , such as eat quick eat quickly , the use of different intensifiers, such as ver versus really 6 and syntactic structures, such as tense and aspect particles . In these cases, alternativ e variants are not exactly equivalent in meaning. According to these questions , the definition of a linguistic variable is challenged whether or not it is valid for a syntactic variable analysis (e.g., Lavandera 1978; Romaine 1981). D. Sankoff and Thibault (1981) study introduces a weak complementarity concept to account for the problematic cases of the variable s in syntax. It is stated that the linguistic variable can be identified by its distribution across the speech community. It is unnecessary to be semantically equivalent. Rather, functional equivalence or discourse equivalence is considered more relevant. D. Sankoff and Thibault say is that the proposed variants can serve one, or more generally, s imilar discourse Labov (2004) also states that with the exception of weak equivalences, there are clear cases of co 1996) studies on the development of the American English strong obligatio n system. In addition, in his later work, D. Sankoff (1988a:153) further points out that: T wo different lexical items or structures can almost always have some usages or contexts in which they have different meanings, or functions, and it is even claimed by some that this differences, though it may be subtle, is always pertinent whe never one of the forms is used. The concept of the linguistic variable, however, has been analyzed more thoroughly when more studies of the linguistic variables at the levels be yond phonology are investigated . Therefore, in order to examine the linguistic variable, two or more variants 5 Further discussion can be found in Tagliamonte ( 2012 ) . 6 F urther discussion can be found in Tagliamonte ( 2012 ) .
43 are required to have a common underlying form. Tagliamonte (2012) further suggests that the form and function correspondence may not be maintained because variants involve in the same change may not mean preci sely the same thing. More importantly, if these variants are members of the same set of grammar structure s in the speech community, the patterns can be observed. Moreover, the decision on all possible variants and their contexts in which they occur must be accountable. This is where the principle of accountability, a fundamental of language variation analysis, comes into play. The Principle of Accountability In examining language variation, th e principle of accountability should be taken into account. It is an important methodological procedure to appropriately and accurately analyze factors influencing a choice of variant (Labov 1966:49, 1972a:72). The principle of accountability requires tha t not only the target variants of the variable we are investigating , but also all other relevant occurrences in the system with which that form varies must be included. The importance of accountability is that researchers must take into consideration what to count, and calculate the distribution of counts in order to have an accurate interpretation of results. It is necessary to understand what variant is affected by what type of environment compared to another, and how. For instance , Tagliamonte (2005) ana lyzes a quotative verb, be like in Canadian English. She suggests that the nature of the quotation is one of the influences to the selection of be like be like it is necessary to determine how q uotative type influences the choice of be like by counting all the quotative types such as direct speech , indirect speech, gesture , and count the frequency of be like occurrence for each type of quotative occurs. The results show
44 that be like is sele cted with indirect speech more than other types of quotative. This way of counting provides the accountability of the results; as Labov (1972c: 94) notes, (a member o f a set of alternative ways of saying the same thing ) should be reporte d with the proportion of cases in whic h the form did occur in the related environment, compared to the total number of cases in which it might have contrast, if a researcher counts the number of be like tokens, and divides the counts accord ing to types of quotative in the data, the results show only the number of be like tokens occurring for each of the different quotative types. This way of counting demonstrates that direct speech has the highest counts for the use of be like . Nonethele ss, it is unable to reveal the pattern of be like , since not all of the quotative types are includ ed in the count, and it leads to incorrect interpretation. Circumscribing the V ariable C ontext Before an analysis of linguistic variation begins, it is cru cial to identify or circumscribe the variable context in Poplack and Tagliamonte ( 1989: 6 0 ) term where the v ariants possibly occur and specify where the variable might have occurred but ( Weiner and Labov 1983:33). That is, researchers need to know what the choices of variants are, although one of the choices is nothing at all (Tagliamonte 2009). An analyst must determine all the potential forms that are counted as variants of the vari able where they have equivalent meaning, and exclude the forms that are invariable. Categorical contexts that do not vary are also removed from the analysis. Once the variables and their environments are defined, researchers can ascertain both internal lin guistic factors, such as the linguistic contexts and the external factors, such as age, gender, education, etc. that may influence choices of variant. the
45 negative do in Buckie, the northeastern Scot dialect, demonstrates the disti nction of the categorical and variable contexts. The researcher states that the categorical context of a third person pronoun, where negative do is rarely absent, is explained by syntactic constraints, whereas the variable contexts of the first and secon d person pronouns is due to lexical, frequency, and processing constraints. Quantitative Analysis Quantitative analysis has been a central practice of language variation studies in the past few decades. Quantitative analysis reveals the behavior of the l inguistic variables corresponding to a set of internal (grammatical) or external (social) factors. The analysis examines an individual instance of linguistic forms in the context of any level of the grammar. The focus is on the repeated variable a speaker produces in the course of linguistic performance (Poplack and Tagliamonte 2001). Quantitative practice allows analysts to investigate the tendencies from the frequency of the occurrence of a particular variant over others and interpret how the alternative of the linguistic variable is influenc ed by particular factors in a specific environment in which it appears. In addition , these linguistic variables are considered as representative s of the underlying variable ch community (Tagliamonte 2007) . In addition, studies that apply quantitative analysis to examine synchronic variation in language are able to reveal change in progress (Coates 2004). Statistical Analysis Variationists apply multivariate statistical testin g as a standard methodology in variation analysis. software program, VARBRUL, wa s developed by Cedegren and D. Sankoff (1974) to account for analytical difficulties regarding intuitive judgment s. VARBRUL allows
46 researchers to perform a multiple regression analysis , which is a statistical tool that can model binary variables with the effect of multiple factors (D. Sankoff 1988b). Once the linguistics and social factors are determined, and all tok ens of the variable are extracted, each linguistic and social factor is assigned a specific code, and submitted to a statistical tool in order to analyze the factors that may significantly influence a choice of variant forms. The later software version, Go ldvarb X, was developed by D. Sankoff and his colleagues (D. Sankoff , Tagliamonte, and Smith 2005) to measure th e effects of linguistic and extralinguistic factors regarding a selection of variant forms by using a weight range from 0 to 1.0. The multiple r egression method in the Goldvarb program provides three levels of evidence that should be used together for interpret ing the results of an analysis of the variable (Tagliamonte 2009) . First, Goldvarb reveals the statistical significance of a specific facto r group. Statistical significance is important , since the significance of a particular factor group may affect the interpretation of the results. It is noted that different raw percentages only indicate the distribution of each variant, correlating with po tential factors it is influenced by. These differences in frequency do not indicate whether or not they are statistically significant. Second, th is statistical software shows constraint ranking. It provides the direction of specific factor groups presente d in factor weights . A factor weight favors the occurrence of a variant when it is measured greater than 0.5 or close to 1.0, whereas a factor weight indicates a disfavor ed effect when it is less than 0.5. A factor weight is considered neutral when it stan ds at 0.5. It neither favors nor disfavors a variant.
47 Last, the program demonstrates the relative strength of a particular factor group relating to other factor groups by comparing the range of factor weights within a factor group, and the order of select ion of factors. A greater range of a factor group identifies a stronger constraint of variant choice more than a smaller range of a factor group. The stronger effect of the factor is notable in accounting for the variability. By running the test, analyst s are able to use statistic al evidence to present t he characteristic s of the variation : whether the linguistic or social factor s favor or disfavor the occurrence of a specific variant , the statistical significance of each factor group , and the relative str ength of the factor groups o f the variable under examination . In addition, the results enable researchers to explain language variation phenomena, such as an ongoing change in the language. The S peech C ommunity The speech community is a main locale in ord er to examine the patterns of variability in language use in the variationist paradigm. Researchers have been offered a definition of speech community. For instance, Bloomfield (1933:42) describes that a speech community is a group of people who interact In his cut off criterion, Bloomfield suggests that a group or community is defined not only by what it is provided but by what it is not. Also, he states that the speech varieties employed within a speech community form a system becau se they are related to a shared set of social (Bloomfield 1933:161) . human aggregate characterized by regular and frequent interaction by means of a shared body of verbal signs and set off from similar aggregates by significant differences in language usage . It is stated that members of the speech community must share a set of
48 grammatical rules, and regular relationships between language use and social structure. In addition , rieties employed within a speech community form a system because they are related to a shared set of social norms However, it may overlap language boundarie s. F or example, speakers of Czech, Austrian , German, and Hungarian m ay share n orms for speech acts and conversational interac tion without sharing any common language. Gumperz states that speakers from these central Euro Sprachbund , a speech area, not a speech community (Wardhaugh 2006). La bov (1972a on the social meanings and evaluations of the variants chosen. The observation of language variation is from everyday speech that speakers in the community use in conversing or writing documents, such as advertisements, articles, and newspapers. This investigation of the pattern of the linguistic variables reflects how the speech community is organized. That is, it shows what variants are preferable, and this phenomenon mirrors the speech behavior of speakers as members and reveals social norms, such as the social evaluation of variants between prestigious and stigmatized forms and their status (i.e. , the insiders ) in the speech community. Empirical studies of language variation in urban settings demonstrate the relations of the linguistic variable with nonlinguistic variables of the social context of the speakers, the listener or the setting (Labov 1972a:237). Fo seminal studies in the 1960s on and the study of social stratification of English in New York City (1966) are influential in further research
49 in different urban cities, such as Norwich ( Trudgill 1972 ), Montreal (D. Sankoff et al. 1976), Glasgow (Macaulay 1976), Belfast (Milroy and Milroy 1978 ), Ottawa Hull (Poplack 1989), Tokyo (Matsud a 1995), Cairo (Haeri 1996) . These examples of research provide insight into the social and linguistic mechanism s and motivation of language variation and change in different speech communities around the world . Extralinguistic F actors and L anguage V ariation From relevant previous studies in the literatur e , this section provides a description of extralinguistic factor s: social factors of age and gender ; and stylistic factors, such as the formality of an interaction and style shifting. The observation of speech data in language variation studies take s social and stylistic dimensions into account , since these extralingui stic factors potentially influence language variation and chan ge . The interrelationship between the interspeaker and intraspeaker perspectives has been observed in sociolinguistic studies (e,g., Labov 1966; Trudgill 1974 ; Romaine 1978 , 1982; Cheshire 1982) . The same linguistic variable functions concurrently in both known phrase relates that it may be difficult to differentiate . Social Variatio n Age observable that people of different age use language differently. It is noticeable that utterances of older speakers are dissimilar to those of younger people. An important ana lytical technique for a variation study is the construct of apparent time. early works (1963, 1966), one of his methodological development sets is the apparent time construction to observe the progress of linguistic changes as they were occurrin g.
50 The concept of the apparent time hypothesis enables researchers to establish the basis for a synchronic approach to a language change. Labov states that when social and stylistic factors are combined in the study, linguistic differences of people from d ifferent generations (apparent time differences) would demonstrate actual diachronic developments in the language (real time language change). Vineyard (1963), for instance, Labov describes that an increase in the use of centraliza tion of (ay) and (aw) among different age groups of respondents reflects a diachronic increase in the use of these linguistic forms of the islanders. However, the expansion of centralized forms was not the largest in the youngest groups of speakers. The yo unges t islanders (14 30 ) rarely used the centralized forms because the forms were due to a pressure to leave the island for a better job opportunity. The older age groups of Y ankee fishermen (31 45 ) began to use the centralization in their speech to mainta in the island identity from being pressured to leave the island. Then the centralization spread to other groups of islanders, and it is marked as the island identity. Also, the use of the centralized onsets (ay) and (aw) has a high index of positive orient Vineyard. Apparent time construction is a major technique in variationist studies to interpret a language change in comparison of generational differences at a single point in time. It is a surrogate for real time examination at different points in history (Bailey 2002). The apparent time is used to examine how a language change may have occurred in t he past , and it enables investigators to study the history of a linguistic process from the
51 viewpoint of the present time. This assumption i period hypothesis , izes after reaching late adolescence, and it remains stable throughout her or his lif e. G. Sankoff (2006) suggests that a gradual increase or decrease of frequency in the use of linguistic in progress. An S curve is characterized as a classic pattern of change in progress when analyzed in apparent time. That is, one form of language is slowly replaced by a new form at the beginning of a change. It then progresses rapidly during the mid cour se, and slowly weakens in the final stage of the change. Also, a monotonic slope by age 7 is considered presenting a regular increase in the use of linguistic features by speakers in one age group to the next, and it leads to linguistic change for the com munity (Labov 1994:84) . Apparent time studies additionally allow data collection in similar settings and circumstances, thus eliminating the comparability problem (Chamber 2002a). However, different linguistic features among different generations can be potentially indicated as an age grading phenomenon . Age grading involves a pattern that reoccurs with each successive generation. This means that individuals change their linguistic behavior at different stages in their lives. This pattern continues repeat ing, but the linguistic system in the community does not undergo a change (Labov 1994). Change is only localized to the linguistic behavior of a specific age group. Certain features may occur or disappear when a speaker becomes older. The typical example o f age grading is during the transition of adolescence to adulthood. Holmes (1992:184, 186) suggests that nonstandard , slang , swear words , or innovative 7 A monotonic slope by age shows a steady increase or decrease in proportion of use of linguistic features from one age group to the next (Tagliamonte 2012).
52 forms tend to reach a peak during adolescent ages. This situation may be due to a pressure from a peer group, in which younger speakers refuse to correspond to the iddle aged stage, they are more likely to and child rearing. In older age, when m ost people are retired with fewer social pressures, the favoring of non prestigious forms may reoccur (Labov 1994:73). Gender Gender is another social factor involved in language variation. It is a powerful differentiatio n factor in almost every study of s ociolinguistics. In the literature, differences. Gender is different from biological sex of femaleness or maleness , in that the term is used to descr ibe research that refers to social and cultural conditions , such as the expectations and opportunities that correspond to people of being women or men in society . Analysts have discussed how gender is related to socialization as an important factor in the explanation of language var iation and language change. In many speech communities, women and men favor different patterns of using and interpreting language through their childhood when they develop diff erent styles of interaction in different spe aking situations. Also , gender diffe rences in linguistic behavior reflect the differential distribution of power in society. In communities where men have more power in society, their language is the norms that control s interaction, whereas women with less power are likely to be more linguis tically polite than men (Holmes 1995 ). Labov ( 1990, 2001 ) proposes two general principles about gender and language variation : that
53 men prefer non standard forms than women in stable sociolinguistic stratification (Principle I ), while women accept the incom ing prestige forms in change from above more than men (Principle Ia), and women serve as the innovators in linguistic change from below (Principle II). However, there is a factor that may not involve prestige or consciousness of speech. Milroy and Milroy ( 1997 ) point out that women are not always conservative. Rather, men use localized variants more frequently, consistently, and more genuinely as identity based social meaning in the community, whereas women pre fer supra local variants in speech. Various s ociolinguistic and variation studies have shown differen ces in the speech of women and men in ongoing forms within different speech communities. The in Norwich, presen t a common correlation pattern that women appear to favor the forms that contain standard or prestigious features in their speech more consistently than norms. The results can be observed from the different degrees of use. That is, both women and men use the same set of variants, but in different quantities. well known survey of New York City English (1966) illustrates that gender differences play a significant role in the process of language change. One of the variables under examination i s the variable (r). New York City is historically nonrhotic, where people prefer the r less variant after vowels , in words such as . The findings support gen eral principles of human language behavior that women are more conscious of the nonstandard value of r less form and shift towards the prestige r ful form more frequently than men.
54 Cameron (2006) suggest s that women and men are instrumental in different k inds of changes. Women lead in change from above when they are consciously aware of the social meaning of the incoming prestige forms, such as rhoticity in New York City English . Men, on the other hand, lead in change from below when they are unconscio us ly aware of the adoption of particular non prestige pronunciation to express their identity and solidarity. For instance, the shift to centralized diphthongs i s led by men o (Labov 1963) . Studies in later years focus on subgroups withi n broad social categories of age, education, sex, and so on . For example, the i nfluential language and gender studies from Eckert (1988, 1989) demonstrate an insight understanding of the dynamics of sociolinguistic variation. Eckert examine s phonological v ariation among American high school students in Detroit as a community of practice. The researcher correlates phonological variation with the social categories that are relevant to adolescents, rather than using only the convention social demographic categ ories. Female and male adolescent s are categorized between Jocks and Burnouts 8 . Through the long term participant observation and interviews, Eckert claims that these categories function as better predictors of phonological variation in the realization of specific vowels than the status. The differences of the identity of jocks and burnouts are also significant through their choices of phonological variable s . For instance, the variable (uh) is backing ; ( the cut caught file 8 The jocks represent a middle class culture, actively engaging in both academic and extra curr icul ar activities to pursue higher education. The burnouts represent a working class culture, engaging in the
55 may sound like foil ) . The results reveal a complex correlation between gender and social categories of the jocks and burnout s. In general, Eckert suggests that in both groups, girls are more advanced than boys in produc ing the innovative variants as a group membership marker. Precisely, burnout girls are more advanced and innovative, leading in new vernacular forms, w hereas joc k girls conserve the old variants. Eckert points out that the sex of the speakers is not directly associated with linguistic behavior. Rather , it reflects complex social practice. Eckert burnouts is highly symbolic . They assert themselves through their appearance and language more than the boys. The girls exhibit greater use of new variants that mark a group identity to establish membership and status in a more extreme way than the male jocks and burnouts do. Previ ous sociolinguistic studies are in agreement that women and men differ in their speech. It is noticed that women and men have been found leading in sound changes of different variables. Women are conservative in some aspects of their speech and innovative in other s . This situation also holds true for men. Whether it is females or males who initiate a language change is not the main concern. Females and males are linguistically sensitive to language change in di fferent circumstances. The reasons why and how women and men innovate and initiate a linguistic change or conserve the old linguistic features are more important. In addition, in order to understand the mechanism of language change and linguistic behavior, it would be better to consider women and men a s members o f particular speech communit ies or social network s , where their linguistic choices can be different according to the different social contexts (e.g., J. Milroy 1980; Nichols 1998; Chamber 2003 ; Coates 2004 ) .
56 Styl istic Variation Style is define d as the linguistic repertoire of individual speakers, in which they have an intuitive sense to speak in different ways. Individuals constantly change how they communicate depending on different settings. Different ways of speaking also convey different so cial meanings (Bell 2007). People may use certain linguistic features at a casual dinner with friends, while they may adjust their language at a formal conference with supervisors. Style is an extralinguistic factor with regard to language variation. Styli stic variation is a focus of language studies because it helps examine the self consciousness of the speakers. It yields the findings of style shifting according to the formality of speech. Style shifting i s broadened in the field of variationist socioling uistic studies, such as when a speaker uses an informal variant in a way corresponding to speaking style. Fischer (1958), for example , analyze d stylistic and social variation of the variable (ing) of three to ten year old children. The results from one of the ten year old children demonstrate d his stylistic variation, in which he use d a more standard variant of ing in a formal setting, and mostly chose a c olloquia l variant in in a less formal situation. A number of theoretical frameworks have been prop osed in order to explain the role and importance of style in sociolinguistics. One of the primary important variables study in New York City. In his proach, Labov considered formality informality style shifting as an outcome of the amount of attention speakers give to their speech in a specific situation. The occurrence of nonstandard linguistic features lessens when the formality increases. Individual speakers adjust or change their speech according to their perception and understanding of language
57 norms in the speech community. In order to investigate attention to speech and style shifting, Labov recorded a number of interviews with individual speaker s to collect different types of speech style. Five different speech styles casual speech, careful speech, reading aloud a short passage, reading aloud a list of words, and reading minimal pairs were designed to increase the degree of attention to the speec h. The results suggested that the pronunciation of a standard variant, such as ing , in words such as playing and talkin g occurred more frequently than a nonstandard variant when the style became more formal. However, many researchers raised ques tions regard ing the attention to speech approach. For instance, Romaine (1980, cited in Cheshire 1982), stated that speakers in her study did not show fewer usages of nonstandard features when attention in reading was directly concentrated on speech. Other language style models were developed to describe stylistic variations. For between the individu al speaker and interlocutor in the social context. That is, individuals speech to be more like that of the listener in order to be accepted by that listener. In other wo determines the style. In some cases, the speakers may diverge from the speech of the listeners to differentiate themselves, or preserve their identity from other people, su ch as Wolfram and Schilling Although the researchers found that style was mainly related to the amount of attention paid to speech, they also suggested that when interacting with outsider, some island
58 residents highly spoke less formal linguistic variants related to the local dialect in order to emphasize their local identity 9 . Alan Bell (1984) supported the accommodation theory, stating that the approach was comparable to stylistic variation approach within the field of sociolinguistics. Bell reason individuals shift their speech style is because they are responding to their present or might be intentionally or accidentally listening to the conversation. However, the shifting or convergence is mainly focused on the specific interlocutor or audience present in the interaction. The audience (1984) study on the speech style of the radio newsreaders in Wellington, New Zealand. He compared the same newscasters reading the same news articles at the same studios for two radio stations. The results showed t hat style was changed as a result of a different group of audience. That is, the newscasters shifted their style consistently when moving from the National Program (YA station) to the Community Network (ZB station). In his later study, Bell (2 001) examined the effect of gender and ethnicity on the speech of four early to mid twenties New Zealanders. The gender differences and ethnic identities (Maori versus Pakeha 10 ) were manipulated to examine the stylistic variation. Each of the speakers was interviewed th ree times according to same gender, cross gender, same ethnicity, and cross ethnicity. The results revealed that the Maori marked features, such as the pragmatic particle , a realization of / r / as a flap, 9 W olfram and Schilling Estes term this occurrence which some local residents ( unlike their everyday speech ) to tourist s . 10 Pakeha refers to most of the population of New Zealande rs of European origin , descendants of the British who colonized the country in the 19 th century (Bell 2001:149).
59 non aspiration of 11 , and a retentio n of that differentiated Maori English from Pakeha English occurred in the conversations. Maori speakers chose these marked features when having an interview with the Maori interviewer more often than when talking to the Pakeha interviewer. The results supported audience design model , since the linguistic features the speakers designed were sensitive and mainly responding to the ethnicity of the interlocutor . However, Maori speakers tended to use clusters of Maori features when talking about thei r family or culture to highlight the Maoriness , even though the listener was Pakeha . The association between social and stylistic variations of sociolinguistic variables has been demonstrated in a wide range of studies in many speech communities. Many type s of features, ranging from phonological to pragmatic variation, have been taken into consideration. One important finding was that men shifted their speech style less, while women shifted their speech style more (e.g., Labov 1966, 2006; Eckert 2000). An i social class. It was evident that linguistic change in New York City wa s associated with women in different speech styles. For instance, speakers in other social classes v ar ied their pronunciation of the variable (oh), the mid back rounded vowel, in words such as talk and dog , very little from casual to formal contexts, w hereas lower middle clas s women varied greatly with style shifting. The results suggested hypercorre ction behavior. Labov indicated that lower middle class women ha d linguistic insecurity, caused by the borderline of their position between the middle and working class. Thus, 11 In words such as Karekare and Piripiri
60 they pa id more attention when style shift ed towards formality as an attempt to b e able to speak better or more correctly . Chesh ire (1982) investigate d gender differentiation and stylistic variation in the vernacular forms a s linguistic markers of working class adolescents in Reading. Cheshire analyzed stylistic variation when girls and boys ha d an interview with their teachers at school to examine how the more formal school context interact ed with their speech. Some more complex factors involve d in stylistic variation were apparent when she focused on the linguistic behaviors of indi vidual speakers. Girls and boys in her study show ed different patterns of speech in more formal style s nonstandard forms decrease d significant ly more than those of the boys. Cheshire suggested that it may have been due to a loose knit . They ha d less pressure than the boys . The girls were likely to have social mobility and shift ed towards the standard English forms in the school environment . For the boys, there were cases where some of them retained thei r vernacular forms. Age was called into play to explain the linguistic behaviors in the school setting . Cheshire point ed out , , that some younger boys ha ve not yet acquired the ability of style shifting until the age of 14. However, she state d that it was not certain that age wa s a relevant factor in her study , since it was suggested in other studie s , such as Reid (1978), that younger children were sensitive to a speech style (Cheshire 1982:160). S ome other boys d id not use any nonstandard forms at all , which may have been because there were no other members of their peer group present in the classroom. Thus, the boys wer e influenced to shift their style according to the pressure of the school norms. In addition, t he relationship with the teacher wa s another
61 crucial factor affect ing speech pattern . I f the boys ha d a good relations hip with the teacher, they tended to adapt their speech to meet the standard norms. When the boys ha d an unpleasant relationship with the teacher, they seemed to maintain their use of the vernacular or increase it. This chapter has provided a description of key concepts in the theoretical framework of variationist sociolinguistics, the ma jor methodological analysis adopted by the present research. The motivation of this study is to attem pt to analyze the language data and describe the patterning and the stratification of the variants of two linguistic variables, the initial (r) and the initial consonant clusters (r) in Bangkok Thai , as well as their correlation to external factors. Therefore, social and stylistic factors are presented in order to demonstrate the effects of these factors on language variation . The next chapter reviews previous studies of the variable (r) in different languages and its correlation with the extralinguistic factor s.
62 CHAPTER 3 LITERATURE REVIEW This chapter provides a summary of relevant previous studies in the literature that have examined the sociolinguistic variable (r). In addition, the variation of (r) in English and other languages is presented, as well as an overview of empirical research on the variable (r) in Thailand in relation to extralinguistic constraints on language variation. The S ociolinguistic V ariable ( R ) The linguistic variable (r) is appropriate for examining variation in language , since it occurs ubiquitously in everyday speech. Also, (r) can be proven to be a variable of some non referential information. Lavandera ( 1978 ) states that the variable (r) has social and stylistic or oth er significance , and it undergoes language change in many speech communities . The patterns of (r) vary in different languages around the world. For instance , Received Pronunciation (RP), one of the varieties of English spoken in England, is non rhotic or r less, whereas Scottish English is rhotic or r ful, the same as many regions in North America. Sociolinguistic and variation studies of the variable (r) focus on phonetic variation in /r/, and other elements conditioned by /r/ and their interaction. Numero us studies have shown that the investigation of the variable (r) reveals its role in many aspects of sociolinguistic variation studies. As mentioned in Chapter 1, t he sociolinguistic variable (r) is subject to var iation, and it enables researchers to quant ify patterns of language variation. In the following sections, research on the sociolinguistic variable (r) in English and other different languages is reviewed.
63 Variation of ( R ) in English American English Unlike the early perspective s of researchers, su ch as Hubbell (1950) who state s department store study proves that r pronunciation c an be systematically examined. In addition , social and stylistic factors can be relevant to the explanation of the variation. Lab ov investigate d patterns of variation in the use of postvocalic /r/ in New York English according to l inguistic context, speech style and social class associated with salespeople from three department stores in Manh attan and S. Klein. The stores represent ed the highest, middle and lowest social stratification ranking of the stores. Labov adopted the method of rapid and anonymous surveys to aradox. The researcher naturally in a role of a customer, asking for the location of fourth floor . This phrase contained two examples of postvocalic /r/, a potential phonetic environment of r lessness. In order to elicit a more careful speech style, the salesperson ly the first time. Since the employees were not aware they were being examined, Labov could not record their responses at the store once h e was out of sight of the salespeople. The results from 264 informants demonstrate d that the rate of r pr onunciation increase d socially and stylistically. The overall stratification of (r) suggested that s alespeople in different department stores interacted with their customers according to ful the most, s howing that the
64 overt prestige form in New York City favored rhoticity. In contrast, salespeople in S. Klein used /r/ the least because they adjusted their speech to working class customers . Also, African American employees show ed lower use of the r ful va riant than the white informants. In terms of stylistic variation, the occurrence of r less decreased in careful tended to pronounce the final (r) close to the higher occurre nce of those in Saks. Labov stated in his later study (1972 a, 2006) that the r pronunciation was the norm that most contrast, there was a shift from the r less to the r ful between casual and emphatic pronunciation, but it was much less marked, suggesting that Saks employees had more linguistic security (Labov 2006:48). S. Klein employees showed the same pattern of stylistic variation of (r). The r pronunciation increased wh en the context became more careful. This pattern indicated that r ful variant is an appropriate pronunciation for emphatic speech since it was, at least, shared by speakers in all three stores. In his seminal study of the social stratification of New York City English, Labov (1966) conduct ed extensive interviews with New York speakers o n the Lower East Side t o investigate the presence and the a bsence of postvocalic /r/ and four other variables. Four types of activities of cont inuous speech in the interview short passage reading, word lists , and minimal pair word epresent ed an increase in formality , and the focus was on language itself. In terms of linguistic differentiation , the patterning of (r) show ed that there wa s a fine rather than sharp strat ification of the variable . Additionally , the differences between groups of speakers were not categorical because no group of speakers demonstra t ed a complete presence or absence of postvocalic /r/. At the level
65 of casual speech, only upper middle class spe akers demonstrate d a significant degree of r pronunciation, showing that the postvocalic /r/ functions as a marker of the highest ranking status group. In other words, t he variable (r) mark ed not only the social status of the speakers but also the style. T he lower middle class show ed an extensive increase of (r) usage when moving from informal to more formal styles of word list and minimal p air reading . The production of /r/ among lower middle class speakers overt ook the upper middle class average, resultin g in a crossover pattern differentiating the (r) variable from other stable variable s , such as (th). The high tendency of r pronunciation in lower middle class speakers reflect ed a hypercorrection phenomenon when they consider ed pronouncing (r) correct and appropriate for formal styles. This result indicated linguistic insecurity and social aspirations of lower middle class speakers. In short , they attempt ed to improve their speech by using the r ful form that they did not usually speak in their casual spee ch to separate th emselves from the working class and bec o me more like the upper middle class when being observed or evaluated (Labov 1966, 1972 a ; Mesthrie et al . 2000). This situation led Labov to conclude that norms involving the use of r pronunciation we re undergoing language change . Fowler (1986) replicate d the change in the variable (r) production in New York City . The result s suggested that the sociolinguistic structur e of the speech community was more stable than anticipated. Despite the pressure of the new r pronouncing norm, the speech of New Yorker s was changing slowly. The hypercorrection habits of the lower ha d not shown any sudden advance of r pronunciation as a whole ( cited in Labov 1994:91). The findings from
66 retained the patterns of age distribution. That is, Saks salespeople show ed a negative correlation with age , while demonstrate d a positive correlation, in which it was the younger upper middle class speakers who were acquiring the r pronunciation norm. However, the lower middle class speakers d id not increase the rate of r pro duction until moving to middle age when they were more aware of a new norm, their social contact and social awareness expand ing as they gr e w older (Bailey 200 2 :327 328). A recent study of a department store study of the social stratification of (r). Mather (201 2) co nducted a rapid and anonymous survey in 2009 to collect spontaneous speech data from 169 salespeople in four 1 . The ccurrences of /r/ as [r 1] for the (r) variants and [r 0] for the vocalized variant in pre obstruent and word final positions. Based on social variables of 35, 36 55, and 56 70), gender (70% women and 30% men), ethnicity (African Amer ican, White, Hispanic, and Asian), and social class (upper middle, lower middle, and lower class 2 ), this study aimed to present an ongoing change of the /r/ production of speakers in the New York speech community, and compared the findings with previous de partment store surveys from Labov (1966) and Fowler (1986). The distribution of /r/ patterns echoed the two previous studies regarding social, stylistic, and phonological variables of word final and pre obstruent [ ]. However, s study, there was a 1 substituted for the original store, S. Klein, which was out of business in 1970s. 2 r middle class, , Basement represented working class stores.
67 and Saks (from 29% to 54%), whereas there was no such change in the lowest ranking department stores. Mather reported that /r/ changed to almost co mplete r ful in Saks, the highest ranking store. Also, the change occu rred more frequently in the pre obstruent fourth than in the word final floor important difference in the way that younger speakers fa vored the r ful variant in both phonological contexts, whereas the older speakers preferred the r less pronunciation more frequently upper middle class speakers favored rhoticity mo re than lower middle class speakers, same as they did in both previous studies . In addition, the distribution of the (r) in c African American speakers use d r less more than the whites , especially in the pre obstruent position. However, the general pattern of stylistic and social differentiation of four stores show ed that African American speakers were moving to conform to the integration in the New York City speech community. The department store studies o f (r) variation from Labov (196 6 ) , Fowler (1986) , and Mather (2012) are instances of the apparent time hypothesis that possibly explains linguistic change in the progress of different generations . Notably, t he ir variation studies offer insight into social and stylistic patterns and language change. A comparison of owler provides significant real time evidence of phonetic change in progress in New York City . The variation of (r) appears in other parts of the United S tates as well. Boston, for example, is well known for its r dropping, in which an underlying representation of /r/ is
68 vocalized in a syllable coda. tuner sounds identical to tuna . Irwin and Nagy (2007) examined the vocalization of /r/ of white Bostonian speakers in careful speech. The analysts conducted a quantitative analysis of the use of the variable (r) from 24 local residents of Boston of different age groups, sexes, occupations, and education levels. The researchers recorde d the speech data from 12 female and 12 male Bostonians who were asked to read a three The story contained 224 function and lexical r words, such as and ; postvocalic /r/ in word final, such as in ; morpheme final but word internal, such as in pairs ; and morpheme internal, such as in . The results suggested that Bostonian speakers were more likely to pronounc e word final /r/ and delete non final /r/. No difference in /r/ deletion occurre d between morpheme final but word internal and morpheme internal contexts. This result which the speakers pronounced word floor word in . Also, lexical words favored r pronunciation more than function words. In terms of social factors, it was evident that for both female and male speakers, a higher rate of r production occurred in the youngest group 3 . In older generations, w omen showed a higher factor weight than men (0.39 versus 0.21) in a multivariate analysis, suggesting that women in this age group began the change earlier. In addition, the results showed that r vocalization was undergoing a change. That is, younger speak ers deleted /r/ less frequently than the older speakers, even though different education levels and income were taken into co nsideration. Speakers with post graduate 3 39), middle aged (40 69), and older (70 89).
69 degree produced [r] more frequently than those with college and high school education. T he researchers stated that post graduate speakers exhibited a higher frequency of the pronunciation of (r) because of a greater contact with speakers of rhotic dialects and more formal speech (Irwin and Nagy 2007:144 145). Levy (2010) undertook his study on B ostonian r dropping by employing speakers, varied by age, sex, race, social class, and neighborhood of employment. Levy used his natural rhotic /r/ pronunciation as a subu rban native of Boston to ask for the name of the closest subway station to go to a football game, and the expected . His findings showed that lower class male speakers and employees in neighborhoods with a strong local identity greatly dropped their final /r/. In addition, sex played an important role in the apparent stability of (r). Divergent patterns between females and males were found with regard to age. Change in apparent time can be viewed in the speech of female speakers. The r d ropping decreased gradually among younger female speakers, whereas male speakers did not show a similar change in apparent time. Males in their 20s favored r dropping the most. The same divergent pattern also appeared when taking social class composite 4 in to account. The gradient stratification was exhibited only by the males, while the females showed sharp stratification, in which /r/ was completely vocalized among outdoor and private employees. However, civil servants completely produced r pronunciation i n their speech, and the rest of the social class composites occasionally dropped /r/. The divergent patterns associated with female and male Bostonians in r pronunciation by 4 Levy categorized four groups of social class : o utdo or and p rivate ( e.g., v endors, s calpers), c ivil servant ( e.g., poli ce officers, subway operators), i ndoor p rivate ( e.g., clerks), and Red Sox fans .
70 social class and age suggested a competition of variants of (r) between different sexes. That is, females favored the prestige form of r production, whereas males retained their r dropping as the local prestige form. Levy further suggested that his results had not consistently followed the pattern of New York City (Labov 1966) or other parts of the country, which may have been due to different social patterns in past difficult times in Boston . That is, Boston star ted forced bussi ng as a way of desegregating public schools during the Boston Bussing Crisis from 1974 to 1988 (Levy 2010:9). Vocalized /r/ and rhotic / r/ interacted as a result of interaction among children from different areas of the city, different ethnicities, and different social classes who were studying in the same school s during this period. Other v arieties of English S ociolinguistic studies on the variable (r) reflect how language is changing phonetically and socially. For instance, Romaine (1978) investigated postvocalic /r/ in Scottish English among speakers in Edinburgh in order to test the claim from Grant (1914) th at Scottish speakers have a monolithic r pronunciation speech community. A trill [r] is considered a social norm in Scottish English which can be weakened to a flap the working class in two speech styles of a casual interview and a passage reading. Three phonological variants of [ ], [ ] , and [ Ã¸ ] were detected . The frequency of the more prestig ious [r] decreased in passage reading s compared to interviews . findin onstandard variants of [ ] and [ Ã¸ ] increased when the degree of formality wa s higher. Romaine stated her results according to Trudgill (1974) that the uncommon pattern accounted for the l ow conscious awareness of the sound. The [ ] and [Ã¸] variants were considered
71 prestige forms by male speakers. However, these variants were against the norm of the area. Additionally, it showed an ongoing linguistic change which moved in a different direct ion from an expected r pronunciation. Romaine (1978:156) commented accepting r lessness their usage happens to coincide with a much lar ger national norm , it was evident that the production of nonr hot icity of Scottish male speakers wa s a local develo pment and not the prestige norm of southern England. demonstrated the process of a linguistic change. Llamas studied the distribution of (r) variants in an intervocalic position, and suggest ed the process of accent leveling and diffusion. These two processes functioned simultaneously in the change in progress of (r) in Middlesb rough, a highly populated north eastern city of England. The significant reductio n in the use of a localized [ s observed as a process of accent leveling, whereas the process of diffusion wa s indicated by the production of labiodental approximant [ ] among young speakers and adolescents in the sample. Llamas point ed out that the degree of salience and the cove rt prestige of a variant did not necessarily predict which variant w ould be diffused. That is, even the labiodental approximant [ ] variant that ha d a low salient status and no covert prestige c ould be the spreading feature. The results also indicate d that females le d in the leveling out of variants, while males le d in the diffusion of new variants into the vernacular. Stu art Smith (2003) analyzed rhotic pronunciation among middle class and working class English speakers in Glasgow, the largest city in Scot land. The Glasgow accent is stigmatized by Scots and others due to its strong Scot tish vernacular character (Scobbie 2006:340). As a result, Glaswegians produce a wide range of
72 variation. Common variants are central approximants, retroflex approximants, al veolar tap, retro flex tap, and vowel. The study clearly showed that there ha d been a loss of the postvocalic phonetic of /r/ and phonetic variation in the realization of /r/. Stuart Smith examined /r/ in all positions in the word list data and sample s of s pontaneous speech. The main characteristics of young working class women and young working class men were the avoidance of the retroflex approximant and a great production of vocalization for coda /r/. The finding also suggest ed that while nonrhoticity ha d been the main pronunciation in the United Kingdom for centuries, rhoticity seem ed to have a secure and prestigious position in Scottish . Variation of ( R ) in French Clermont and Cedergren (1979) documented generational change in progress of the distribut ion of the alveolar to the innovative velar fricative or uvular (r) in Montreal French. Based on the 1971 data 5 from 120 speakers, the study was able to trace a survey back to the late 1940s that clearly demonstrated that alveolar [r] was the dominant vari ant. The finding corroborated the apparent time interpretation of the 1971 data, which exhibited the classic S curve pattern of change in progress. The researchers found that the innovative uvular [R] was increasingly more common among speakers under 35 ye ars old. Cedergren (1987:49, cited in G. S ankoff and Blondeau 2007) also suggested that according to a multiple regression with the rate of [R] chosen ... followed by socia 5 In 1971, Cedergren, D. Sankoff , and G. Sankoff recorded native speakers of Francophone Montreal as an atte mpt to conduct the original Montreal French study (Sankoff and Sankoff 1973). In random sampling, 120 speakers varied by social class, sex , and age. Speaker 1956.
73 on the 1963 Bibeau Dugas corpus, Cedergren further stated that these apparent time inferences showed that change had already started at least 10 years earlier. Th is trend comparison betwe en the 1963 and 1971 corpora corroborated the existence of change in progress at the level of community. L ongitudinal or real time studies of the (r) variable enable researchers to investigate the same or different speakers at a different time to obtain a sense of how the language or dialect is changing over time. For example, G. Sankoff and Blondeau (2007) raised the question with regard to the concept that the linguistic systems are completely stable in an adult life. The study combine d trend and panel st udies to examine a change in apical [r] to dorsal [R] in Montreal French across the individual life span in the 1970s and 1980s . The researchers examined whether or not the speakers changed their linguistic behavior during their lifetimes. The 1984 panel st udy r einterviewed some of the same speakers fro m the data in 1971, and compared the 1984 corpus to the 1971 corpus. This real time study represent ed a real time assessment of the community after a thirteen year interval. It indicate d that there wa s a gre at shift toward the replacement of apical [r] to dorsal [ R ] during the p eriod of 1971 to1984. T he majority of speakers in 1971 ha d increased their use of [R] to the categorical or near categorical production of [R] by 1984. The researchers suggested that s ome speakers changed their linguistic behavior throughout their adulthood. However, this only occurred with the speakers who showed the variability in the 1971 data. In addition, the speakers who changed their linguistic behavior always changed in the same direction of change in progress in the 1984 data. G. Sankoff and Blondeau concluded that the apparent time construction may underestimate the rate of change
74 since the apparent time interpretation did not consider a category that they called Variation of ( R ) in Dutch Verstraeten and Van de Velde (2001) analyzed the socio geographical variation pattern of /r/ in standard Dutch. The analysts observed that there were variants of /r/ in the Dutch language. The Netherlands and Flanders 6 were selected because there were cities to cover the main dialect regions with different degrees of ongoing changes in standard Dutch. The researchers conducted a formal reading experiment and a conversational interview with 160 Dutch language teachers, stratif ied for two communities of the Netherlands and Flanders, four regions in the Netherlands and Flanders of core areas, an intermediate zone, the first peripheral area, and the second peripheral area with female and male speakers in two age groups of young (2 2 40) and middle aged (45 60). The language teachers were chosen as the subjects in this study because they spoke standard Dutch on a daily basis as professional language users, that of broadcasters whose speech was mainly examined in other variation studies in standard Dutch pronunciation. In task one, the subjects were instructed to pronounce the /r/ in three different types of initial, internal and final position in the carrier sentences 7 , representing intraspeaker variation. The variable (r) in the carrier sentences was designed in 6 Flanders is a Flemish Dutch speaking region in the Southe rn part of the Netherlands. 7 A carrier sentence is used to present test words . T he target words in the carrier sentences do not contain any lexical meaning.
75 controlled segmental and rhythmic environments to elicit the wel l articulated standard Dutch realization of the variable (r). In task two, the subjects were assigned to have an interview, representing interspeaker variation. The subjects in the Netherlands had an interview with a young Dutch male interviewer, speaking modern northern standard Dutch, whereas those in Flanders had an interview with a young Flemish female interviewer, speaking modern southern standard Dutch. The results demonstrated variation of (r) in Dutch. Even in a formal reading task, eleven variants of (r) occurred. More variants were found in the speech of the speakers in the Netherlands than those in Flanders (10 versus 7). An alveolar trill was dominant in both geographical groups of speakers. The intraspeaker variation was mainly limited to manner o f articulation. O nly a small number of speakers demonstrated variation in the place of articulation of /r/, such as showing both front and back realization. Most speakers used front realizati on of (r) (e.g., alveolar trill, alveolar trill fricative) more frequently than the back realization (e.g., uvular trill, uvular fricative). The index score of place of articulation of (r) had a significant effect according to community and region factors, while age and sex showed no significant differences. Verstraet en and Van de Velde suggested that alveolar trill fricative was a new variant found in the speech of speakers in both the Netherlands and Flanders, and these variants had never been evidenced before in the standard language. In addition, the newer variants of alveolar tap and retroflex flap were considered to be important variants in conversational speech. However, Verstraeten and Van de Velde stated that these variants were not yet established as a norm in the language, since they were rare in the study. T his study
76 was presented as a starting point of further analysis on the geographical and social stratification of (r) in Dutch to reveal patterns of (r) variation in the Dutch language area. Variation of ( R ) in Mexican Spanish Rissel (1989) examined the ex tralinguistic factors affecting the assibilation of /r/ in the variety of Spanish spoken in San Luis Potosi, Mexico among young speakers. Harris (1969:48, cited in Rissel 1989) note d that assibilation is the loss of the feature anterior and the additio n of strident . It is widely use d in Spanish speaking areas, such as northern and western Argentina, Chile, and highland Mexico. The standard Spanish flap or trill /r/ can be assibilated in word initial position, in the onset of a syllable after a conson ant, at the end of a syllable, and in the group of /tr/. The researcher mentioned that assibilation of r sounds in Mexico occurred first in the speech of women in the middle and upper social classes. Based on previous studies on Spanish devoicing of fricat ive /z/ in BahÃa Blanca, Argentina (Fontanella 1979), Rissel included four extralinguistic factors of style (conversational speech and reading), age (12 22), sociocultural level 8 and attitude toward sex role differentiation 9 in her study. The results demon strated that sex and sociocultural level affected the assibilation separately examined female and male speech, the assibilation was a phenomenon characteristic of the mi ddle and lower sociocultural groups of women and of the middle 8 eaker attended . They included high end private school, mid range private school, and federal school. Group s 1 3 of sociocultural level s were assigned, comparable to higher, middle , and lower class. 9 Attitude toward sex role differentiation included three ratings. R ating 1: nontraditional attitude, women shoul d work outside after marriage. R ating 2: intermediate attitude, women should work outside only in the case of financial necessity. R ating 3: traditional attitude, wom e n should stay home and take care of children.
77 and upper groups of men. In terms of style, the difference between conversational and reading style declined among female speakers in the sociocultural group 2. In contrast, the degree of assib ilation of final /r/ slightly increased in reading style among women in sociocultural group 3. This pattern suggested that assibilation of final /r/ carried certain (e. g., Labov 1972a; Fontanella 1979) that when the variable that carried a local prestige began in the speech of the middle and upper class women, it would then spread to the speech of young women in the lower sociocultural group due to their sensitivity to t he prestige variant. Additionally, Rissel pointed out that young men in the lower sociocultural group resisted the final /r/ assibilation because it was recognized as the speech of women. Therefore, it was unlikely to be integrated into speech of working c lass men, as the study of Trudgill (1983) evidenced that low assibilation was a characteristic of working class speech which carried status, prestige and solidarity in the group. The interaction between sex and attitudes toward traditional sex roles demons trated opposite progression of the assibilation of final /r/ between female and male speakers in a traditional attitude group 3. That is, the assibilation occurred greatly e assibilation is closely associated with what is considered to be appropriate male and (Rissel 1989:282). The Sociolinguistic Variable ( R ) in Thai In this section, an overview of sociolinguistic and language variation studies of the Thai variable (r) is provided. Two linguistic contexts are identified where the correlation will be demonstrated with extralinguistic factors, such as age, gender, occupation, education level, ethnicity, and style in onset (r) and (r) consonant clusters.
78 Stud ies on sociolinguistics and language variation in the Thai language became a cent ral point of interest in the 1970s when analysts observed that the variable (r) in Thai tended to involve social factors (Prasithrathsint 1988) . R esearchers attempt ed to examin e variation at phono logical, syntactic, and lexical levels in different dialects of Thai. Different methodologies have been adopted to analyze and reveal the linguistic variation in different social contexts. Beebe study is o ne of the most well known studies in language variation in Thai. Based on the Labovian paradigm , Beebe investigate d the relationship between social factors and the pronunciation of the Bangkok Thai consonant clusters in which the consonant /r/, /l/, and /w/ can co occur. By a dopting an interview as the main data collection technique, the researcher found that the pronunciation of the consonant clusters under investigation var ied with regard to the social factors of social class 10 , education level 11 and age of the speakers. Beebe suggested five variants in the clusters are found in her data: the full retained cluster of a flap/tap, the reduced variants of r deletion of the second member of the cluster, the lateral, the trill, and the continuant retroflex. Quantitative anal ysis support ed her results that speakers of a higher lev el of occupation and education produced the /r/ clusters more frequently than those with a lower level of such social factors. Also, t here was a correlation between age and the full retained cluster p roduction. That is, the group of older speakers used more full clusters than the group of younger speakers. The older groups additionally show a higher 10 the highest to the lowest are professional (e.g. doctor, professor), managerial and business professional (e.g., lawyer, engineer), semi prof essional (e.g., accountant, nurse), semi skilled workers (e.g., foreman, clerk), and unskilled laborers and servants (e.g., driver, waiter, messenger). 11 Four education levels are university, vocational, secondary school , and primary school.
79 production of the lateral than the younger group. However, study ha d a limitation in terms of ag e range s of the speakers. Beebe classifie d the participants into two broad age groups of 35 and under and 36 and over . Th us , the study automatically exclude d the data from the younger and older speakers to demonstrate further evidence of apparent time pers pective from the other age groups. Based on her exploratory survey, gender was excluded as Beebe noted that it did no t show a significant effect on consonant cluster variation. However, previous variation studies indicated that women and m en significantly d id not conform to the same patterns of language use (e.g., Labov 1966 ; Cheshire 1982 ; Eckert 1989 ; Chunsuvimo l 1993; Nichols 1998 ; Coates 2004). Therefore, the present research aims to examine whether or not its findings in terms of gender correspond to . Stylistic variation is also taken into account for language variation studies in Thai. Several previous studies focus ed on the variable (r) and its rela tionship with speech style . Treyakul (1986) investigate d variation of (r) and (l) among female and male Bangkok radio broadcasters , taking education and four different speech styles of interviews, news announcements, passage reading, and minimal pair reading into account. The researcher found that (r) and (l) in an onset position and clusters var ied according to speech styles. In a formal style of minimal pair reading, speakers fully retain ed a trill and a lateral. The tap and the trill were more commonly prefer red than the approximant and the lateral in a news report passage . In an informal style of interview, in contrast, all participants favored the lateral much more than other variants of tap, approximant, and trill . For consonant clusters , the results were similar to those of the initial (r). That is, in most formal speech the trill wa s preferr ed . In general, the s tandard
80 tap and trill wer e mo re preferable than the lateral, the approximant, and /r/ dropping. In the least formal speech style, the speakers tend ed to omit the /r/ sound from the cluster s, specifically after an aspirated consonant. In addition, variation studies indicate that other social variables influence the variation in the production of (r). For example, Chunsuvimo l (1993) conducted a study on social variation of (r) in Thai and in English among 58 employees of three first class hotels in Bangkok with regard to social factors of sex , job level , and English language background. The researcher found that the lateral occur red greatly in a prevocalic position, wh ile speakers omitted (r) in Thai consonant clusters. The approximant was found most frequently in English (r) in both linguistic contexts of prevocalic and clusters . w omen were more conscious than men o f using the prestige trill in Thai and the standard approximant in English . This f inding supports the hypothesis in the present study that gender may influence the selection of (r) variation. In terms of job level, Chunsuvimol stated that since t he prestigious variants of Thai and English (r) are different, the examination of whether or not high English proficiency affected the patterns of (r) variation in Thai would contribute to a better demonstration of the (r) variant usage among native speake rs of Thai whose job was highly relevant to English. It was evident that speakers with a h igher job level preserved the standard forms more than speakers with a lower job level. Speakers with a good English language background preferred more prestigious va riants than those with less English language proficiency . In addition to the initial (r), Chunsuvimol investigate d the
81 production of the consonant cluster of Thai /r/ during the interviews. The results suggested that the (r) deletion wa s common ly found in female and male speech, but male speakers tend ed to drop the (r) in a higher proportion than female speakers. Chunsuvimol noted that the social factors were indicators of the favored variants of the Thai variable (r). However, his data collection method may allow only a limited amount of data to be obtained. In order to collect the speech data, the researcher conduct ed a 15 minute interview with each participant , followed by an interview in English with a native speaker of English for the same duration . H owever, d ue to the fact that people tend to monitor their speech when they are being observed, relatively short interviews especially in different language s may be insufficient to elicit the natural speech data to represent an actual speech production in b oth Thai and English . Therefore, the sociolinguistic interview technique adopted in the present study enables the collection of a variety of conversations to obtain relatively more natural language data for the analysis. Further, a stylistic variation study is analyzed to test the claim that the second element in Thai consonant clusters would be completely dropped especially in the young generation even in the higher register ( Dhebbayasuwan 1975 ). For instance, Pulsup (1993) examined the /r/ and /l/ clu sters in different speech styles, ranging from informal to formal situations with female high school students. The results in her study opposed Dhebbayasuwan assumption , since it was suggested that adolescent students retained their pronunciation of the consonant clusters /r/ and /l/ in different proportion according to the degree of formality in the situation. The (r) omission occurred in less formal speaking situations of casual speech, interviews, and a passage
82 reading. The teenagers in the study retai ned a standard trill for the sound /r/ and a lateral for the sound /l/ in more formal speech contexts of word list and minimal pair readings. Ethnicity is included as one of social factors potentially correlated to the variation of the variable (r) in Th ai as well. Panyaathisin (2013) conducted a comparative study of the variable (r) in initial consonant clusters used by different ethnic groups in Hao Lam Phong , the Chinese speech community in Bangkok. The researcher investigated whether or not native spe akers of Thai and speakers of Teochew 12 Thai produced variants of the consonant cluster (r) in the same or different direction. Teochew Thai speakers refer to the descendents of the Teochew Chinese immigrants residing in Thailand. The second generation spea kers of Teochew Thai are generally bilingual as they acquire a good Bangkok Thai linguistic competence. They are also able to speak Teochew with their families. The third and later generations are more likely to have a lower Teochew linguistic competence, whereas they fully acquire Bangkok Thai because they were born and raised in Bangkok. A total of 15 female and male speakers of Teochew Thai, ranging in age from 16 to 60, and 15 female and male speakers of Bangkok Thai, ranging in age from 15 to 40 were analyzed. Panyaathisin noted that the numbers of age and gender were unequal interview, a passage examine initial consonant clusters (r) vari ation. 12 Taochew or C haozhou is a Chinese dialect mainly spoken in a Chinese speech community in Thailand.
83 both groups of speakers retain ed their use of the prestige trill in more careful speech of a passage reading, and the nonstandard or vernacular forms were selected in a less careful context of an interview. It is noticeable that the prestigious trill is completely absent in an interview. In the view of the present research, the reason is perhaps due to the length of the interview, which is only three to five minutes for each p articipant, a very brief opportunity to elicit natural speech data and an occurrence of the variable (r) and its variants. Moreover, Panyaathisin found that native speakers of Thai used the trill less frequently than Teochew Thai speakers. However, thi s fi nding was not statistically significant, suggesting that ethnicity of the speakers did not influence the selection of the (r) variants. The researcher explained that the higher level of education background, the older age range of speakers, and stronger co mmunity bonding of Teochew Thai speakers may be the underlying reasons accounting for the higher production of the standard trill. In addition, a group of native speakers of Thai were more geographically mobile. Unlike the Teochew Thai speakers who were na tive Bangkokians, some participants in a group of native speakers of Thai originally came from other parts of the country. Therefore, there were cases of regional dialect interference and less community cohesiveness. However, both native speakers of Thai a nd speakers of Teochew Thai demonstrated a similar pattern in the ir use of the variable (r) according to different speech styles in this community. This chapter has provided a review of (r) introduced as the sociolinguistic variable in different dialects of English and other languages. In addition, studies on the variable (r) in Thai literature, involving social and stylistic factors conditioning the
84 language variation, has been discussed. In the next chapter, the methodology of the present study is outlin ed.
85 CHAPTER 4 METHODOLOGY The method of collecting data for the present research, including a description of linguistic and extralingui sitc factors used in this study are provided in this chapter. The criteria of participants and their demographic cha racteristics are described. Additionally , the sociolinguistic interview is introduced as a major data collection technique. The explanation of d ata collection procedures, data analysis, and statistical methods are includ ed. Linguistic Contexts Applied in the Study Two linguistic contexts of the surface phonetic variable (r ) are under examination in this study. First, a single initial consonant (r), in words such as [rÃ¡k] (love) and [ri:an] (study). Second, five consonants, /p/, /p h /, /t/, /k/, and /k h /, th at (r) can co occur with, and form the (r) consonant clusters, in words such as [prÃ ma:n] (estimate), [ p h rÃk] (chili pepper), [ (straight), [ h Ãª:p ] (Bangkok), and [ k h raj] (who) are included. There are advantages of selecting an initial position under investigation. First, the variable (r) in word initial position and the consonant clusters (r) are prevalent in Thai, and t hey show that there is variation in the language according to the literature ( e.g., Beebe 1974 , Chunsivimol 1993, Phootirat 2012 ). Second, the variable (r) in such contexts is easily identified and distinguished in audio recordings. Both linguistic context s include one syllable word, first syllable of two syllable word, first syllable of three syllable word, and first syllable of more syllable word. Extralinguistic Factors Applied in the Study Previous studies (e.g., Labo v 1963; Trudgill 1972, 1998; Milroy and Milroy 1978; Cheshire 1982 ) suggest that linguistic and social factors are correlated in the process of
86 language change. Language change occurs when there is a new linguistic form used by groups of people within a speech community. Stylistic factor an d social factors, such as the variant that the speaker favors. In th e present study, the extralinguistic factors of age, gender, and speech style s are taken into considerat ion to discover the relationship an d effects of linguistic and non linguistic factors on the sociolinguistic variable (r) in Bangkok Thai. Age Age is the basis of the apparent time hypothesis that possibly indicates a linguistic change in progress. Speaker s of one age group may linguistically behave different from people of other age groups. This may be due to different contexts that people from different age groups, along with different social status. The participants are classified into the following thre e generational age groups: Younger group: 18 25 Middle aged group: 35 45 Older group: 50 65 The age categorizing represe nts a relatively clear cut way of three successive generations of younger , middle aged , and older speakers in the Bangkok speech community. The age of 18 25 appropriately represents language use of the younger generation. Eighteen is the lawful age of an adult in Thailand. By this age, people are viewed as having their first language linguistically well developed. Although 18 year old people are considered young, they are expected to be responsible for their life. Twenty five is the age at which most people have already earned their first degree, and perhaps
87 ir working life. The age of 35 45 represents the middle aged generation. Most people in this age range have a stable career and their own family. The age of 50 65 belongs to the older group. People have settled down with a family. Their working life slows down, and most of them are retired when they turn 60. Gender sex and their social behavior of being women and men. Gender is potentially associated with language variation. Re searchers have discussed how gender is related to socialization as an important factor in the explanation of language variation and change. Numerous studies suggest that women and men have different model s of language use. For instance, w York City English (1966) evidences that women are more innovative than men when women favor more standard forms, responding to overt prestige connected to the forms. presents reversal findings. The resul ts show that the centralization of two diphthongs (ay) as in mice mouse island tradition. Style Two speech styles, informal and formal, are applied in the present study. A lengthy conversa tion in the sociolinguistic interview represents a relatively casual and informal style, whereas two reading aloud passages represent more formal speech. Stylistic variation is an important extralinguistic factor to examine self consciousness of the speake rs. It yields the finding of style shifting according to the informality and formality of style. Previous research has shown that speakers tend to use a standard
88 variant in a more formal setting (Cheshire 1982). In addition, the study of stylistic variatio n reveals linguistic behaviors, such as hypercorrection and linguistic insecurity of the speakers (Labov 1966). Education Education is not considered a primary factor to examine (r) variation in this study. However, education is included as a secondary so cial factor that may affect the pattern of the variable (r). Although education level is not a m ain requirement for potential participants, they are expected to at least have basic literacy skills because reading and writing abilities are required in the p from a vocational diploma to a doctoral degree , as illustrated in Figure 4 1. Most of them Parti cipants Criteria for Participants The criteria for participants were that they were native speakers of Bangkok Thai who were born and raised in Bangkok and used Bangkok Thai as the main form of communication in their family. However, native speakers of Th ai who moved to Bangkok from Chiengmai, a northern province of Thailand, to Bangkok when the participant was three years old. Participants who moved to other parts of the country or resided abroad before returning to Bangkok were excluded, since other dialects or foreign languages may have affected their current Bangkok Thai pronunciation. In addition, participants were required to have been living in Bangkok or its metrop olitan area where Bangkok Thai is the main language used. A total of 24 participants have their residence in
89 dif ferent neighborhoods of Bangkok and five of them reside in the vicinity of Bangkok , as shown on a map of Thailand in Chapter 1 ( Figure 1 2 ) . In addition, it is noted that only the potential participants , whose age fit into one of the three age categories, were qualified for further data collection . Recruitment of Participants It should be noted that perfect representatives of the population were not sought for the purpose of this study. That kind of data collection would require much more time to obtain a much larger sample. Additionally, recruiting native speakers from different neighborhoods of Bangkok can be difficult. Recall that Bangkok is a populous and diverse capital city where people come from different parts of the country for business and employment. There is no assurance that residents in each area are native speakers of Bangkok Thai. In addition, opportunities to gain access to current census data and other information about the native Bangkokians and their residence in different areas of the city remain limited. A stratified sample of native speakers of Thai was constructed in order to obtain a workable number of participants, satisfyi ng the specific requirements to examine the variable (r) under examination . Some responses to a request for participants for the current research were from a note posted on a bulletin board at my workplace in Bangkok. Participants were also recruited throu gh friends of friends in my social networks in the city. On some occasions, several participants were recruited through my colleagues when I had a chance to befriend them. Potential participants were observed in a group interaction (e.g. , at lunch) with th eir friends. This observation helped me to
90 more conversational and relaxing when I was no longer a complete stranger to the participants. A total of 33 participants fini shed the tasks. However, three of them were excluded from the analysis. Due to lack of an appropriate amount of data obtained, two participants were removed from an analysis. Only 25 30 tokens were elicited from each of these participants, while there were generally approximately 90 100 tokens from other participants. Other tokens found were not feasible for data analysis because they were not clear in the recordings. One more participant was also excluded due to the erally stammered in a passage reading section, and the data was not clear enough for further analysis 1 . As a result, an equal number of five female and five male participants were tabulated in each age cohort, yielding a total number of 30 participants sat isfying the requirements. The demographic information of the participants is presented in Table 4 1. Data Collection The data was collected between May and July 2012. After a participant agreed to participate in this study, I asked that participant to s elect a place where she or he was comfortable doing an interview and other tasks. Only request for each participant was that the interview took place in a quiet room 2 because it was crucial to have clear recordings of stretches of speech to distinguish the phonological variants. Only a few participants were willing to do the procedure at their home; this was because I knew 1 Although the data obtained was unfeasible, I allowed th is participant to complete the procedure because embarrassing the p articipant was not my intention . 2 I had to as k participants to set the air condition fan to a low level , since I had to rerecord some aloud sections as a result of the loud noise of the air conditioning. Turning off the air conditioning was almost impossible; the room temperatur e would have been too high for comfort.
91 close to any family members. If participants co uld not find an appropriate location, or if they were students, the recordings were made in a reserved room at my workplace. In most cases, the participants preferred to do the tasks at their workplace before their work hours or during their lunch breaks. Although the procedure took place without for permission to spend time at the workplace. Once an appointment with each participant was scheduled, I went to her of his office to perform a procedure in a quiet meeting room. The Sociolinguistic Interview In th e present study, the sociolinguistic interview was adopted as a major data collection technique. The sociolinguistic interview unavoidably encounters the s paradox. However, it is noted that I did not claim to collect completely unselfconscious, natural or vernacular speech in the present research. Rather, I concentrated on gathering relatively spontaneous and natural speech in the sense that Wolfson (1976: 208) notes. That is, the data are appropriate to the occasion. They are appropriate to the interview which is a normal speech event in the society. The sociolinguistic interview, pioneered by William Labov (e.g., 1966, 1972a, 1984) , is the heart of langua ge variation studies. This technique is a conversational interview designed to keep rapport at a moderate degree, and the interviewer is able to control and remove information that cannot be coded in an analysis later. The sociolinguistic interview was dev eloped to elicit a large quantity of casual speech which represents how people naturally speak in everyday conversation when their language is
92 (1972b:112), it is stated that the most important speech for linguistic theory comes from a relatively unobserved and natural conversation. The sociolinguistic interview 1984). That is, when speakers know they are being audio or vi deo recorded in the interview, many of them may shift even slightly from their casual conversation. However, the only way to examine how people naturally talk and do activities in their everyday lives when they are not being observed is to observe them. Th us, attention has been given to carefully designing the types of questions asked in the interview. Wolfram (2010) notes that it is crucial in the sociolinguistic interview to elicit spontaneous speech that taps into personal and community interests, and is non threatening to the participants. Topics of conversation used in the sociolinguistic interview are focused on general interests of community participants, in an attempt to minimize the attention paid to speech and to the fact that the participants are b eing recorded as part of a study. The interview modules are categorized into common areas of topics to encourage lengthy conversation by the interviewees. Topics of the sociolinguistic interview are flexible and loosely constructed with open ended question s. The interviewer can arrange, rearrange, and move naturally to another topic, focusing In his Lower East Side Stud in danger, since they would disregard the fact that they were being recorded and forget to monitor their speech. Howe ver, danger of death is an ineffective topic in some
93 N orthern of does not excite people in this locality, but creates an unpleasant and terrified experience (c ited in Milroy and Gordon 2003: 65 66). Rather, other questions were gathers good data when asking the in terviewee about ghost stories. Despite selecting effective questions, Labov (1984:40) suggests that the basic counter strategy of the position of lower authority than about their participants and how they talk from the questions asked. The sociolinguistic interview has been an important technique in variationist fieldwork. It has been developing beyond the basic interview for a few decades. A high volume of speech data and a high quality of recordings continue to be the advantages of th is data collection technique. A large amount of valuable data has been obtained by the sociolinguistic interview, encouraging the growth of research in the field of language variation. Equipment The procedure was recorded using a Sony ICD PX720 digital voice recorder with the Olympus ME 52W, a background noise canceling microphone . The recording equipment was highly functional to ob tain good quality speech data. It enabled repeated sound file analysis for more accurate identification of phonological conditioning of the variable (r). In addition, a long cord on the microphone allowed the voice recorder to be kept out of the participan being recorded.
94 Procedure In the present study, the data collection procedure was di vided into two structured tasks: the interview and the reading aloud passages, representing informal and formal speech styles, along with a debriefing session and a short questionnaire to Stylistic variation was included to verify the hypothesis that attention to speech is an important constra int on language variation (Labov 1966; Meyerhoff 2006). At the beginning of the procedure, each participant read and signed Thai and English versions of the informed consent form approved by Institutional Review Board of the University of Florida ( Appendi x A). I generally explained that the procedure related to my research and would last approximately one and a half to two hours. The participant understood that all tasks were audio recorded. I also informed the participant that her or his participation was voluntary and she or he had the right to withdraw from personal information was removed from the recordings and transcripts. Then, each participant was asked to clip a microphone to her or his shirt a little ways from her or his mouth. The participant spoke several sentences , and I played back the recording in order to confirm that the sound recorded was loud and clear enough for further examination . In task one, I co nducted an interview with each participant in a quiet environment. The interview lasted approximately one hour and fifteen minutes. I initiated the interview with self introduction, and proceeded to a conversation by asking what I should call the participa nt during the talk. It is common to address an interlocutor by using a kinship term plus name or nickname in Thailand (e.g., [ P Ã Mon ] (Aunt Mon) ).
95 This step was meant to relax the participant and create a more casual atmosphere between acquaintances, rathe r than a formal setting between an interviewer and an interviewee. For 10 to 15 minutes after that, demographic questions were mainly asked. In some cases, background information was asked at the end of the conversation. navoidable, a range of topics for the interviews, shown in Appendix B, was designed to involve the participants in more emotions and interests and minimize self consciousness in their speech. Such topics eral, work, family, childhood memories, leisure activities, current local events, the national economy, weather, natural disasters 3 , and world events. Other unlisted topics that the participants preferred and that drew their interest were also asked. I exp ressed my enthusiasm in listening to what each participant was talking about. For instance, one of the participants had a passion for sports. Thus, I asked sets of open ended questions regarding the sports he played. This topic continued for longer than 20 minutes. Topic observation also encouraged less voluble participants to talk more on topics they were interested in, and reduced the 20 minutes of an interview, which result ed in a more natural and pleasant conversation. Once an interview section was completed, I paused a voice recorder and the participant took a short break before task two began. At the end of an interview, the participant was able to unpin a microphone temp orarily if necessary. In that case, the participant was asked to clip it back on for the next task. 3 T he worst flooding in decades hit Thailand in 2011. M ost Bangkok residents experienced this crisis. I obtain ed a flow of emotional speech using this topic.
96 In task two, each participant read aloud two short narrative passages representing a formal speech style. The reading passages were designed to examine whe ther formality of speech affects the use of (r) variants. The passages were adapted from an online newspaper and a food and travel column , and contained the variable (r) in word initial and consonant clusters positions ( Appendix C). Natural disaster and fo od and travel were the chosen topics of the two passages. The natural disaster was a current big issue, and food and travel were common topics talked about in the country. Also, these topics were comparable to the topics in the interview; thus, participant s were already familiar with the contents. I acknowledge that reading is a qualitatively different activity than a conversation. However, reading is an artificial strategy that allows me to control the attention the participants pay to their speech. Accor ding to Labov (1966, 1972 a ), reading short narratives is an experimen tal task which produces another more formal and less regular style. Therefore, the two distinct forms of a casual interview and short narrative readings were developed to elicit the data from a big contrast of style shifting. Word list reading was not included in this research because participants from my previous studies tended to be aware of what was being ex amined. As a result, they were excessively conscious of their speech. At the be ginning of a reading task, each participant had approximately five minutes to skim through the readings and ask questions in case further explanation was necessary. I asked the participant to read aloud two readings continuously with a short pause before s tarting the second passage. Once the participant was ready, I started recording again. While the participant was reading aloud, I stayed in the same room
97 with her or him, listening quietly. The reading task lasted approximately 10 to 15 minutes. In some ca ses, the participant was asked to reread the passages when there was an unexpected interruption, such as sneezing, coughing, or a noise from outside. After the task was completed, I stopped the recorder and told the participant to unclip her or his microph one. The recording equipment was taken away, and the participant prepared for the final task. Finally, I conducted a debriefing session with each participant by playing back part of her or his interview recording in order to verify the naturalness of the speech. The participant was asked to provide information about whether or not her or his speech was close to that in an everyday c onversation. There was no case in which the participant reported that her or his recorded speech was artificial. Also, each pa rticipant completed a short questionnaire, as shown in Appendix D, after listening to the two short passages. The first passage was a conversation from a talk show, representing an informal speech style, and the second passage was a news report, representi ng a more formal speech style. All of the participants completed the questionnaire within 10 minutes. The short questionnaire contained questions regarding language attitudes specific to the (r) pronunciation in Bangkok Thai in order to examine whether or not the participants were conscious of a salient feature of the variable (r) in Bangkok Thai. In addition, metalinguistic data from the questionnaire complements the present study. At the end of each debriefing session, I expressed my gratitude to the pa rticipant. After the interview was completed, refreshments were provided to the participant for voluntarily participating in the study without compensation. If the procedure was conducted at their residence, the participants received a thank you gift.
98 Data Analysis This section provides information for the data analysis procedures . It explains the data transcription and how the data collected from the participants were extracted, coded, and analyzed. The statistical methods are also described. Data Transcr iption Number, not pseudonym, is used to refer to the participants in the present study (for example, P1), and their personal information was removed from transcription. The data from each interview were transcribed the way standard Thai words are ordinari ly precisely from the recordings, including colloquial language and speech errors. Basic transcription conventions by Tagliamonte (2009) were applied in transcription. For rea ding transcription, it is noted that although all participants read the same passages aloud, the total number of appropriate tokens varies. That is, some participants skipped some target words that contained the variable (r) or did not read the words clear ly. Data Extraction and Coding I listened to each sound file and highlighted only the words containing the variable (r) that the participants spoke clearly. To circumscribe the variable contexts, I identified the variable context of the (r) variants. The sentences where the variable (r) situated in the word initial and initial (r) consonant cluster positions were extracted and coded on Excel spreadsheets. According to Guy (1980, cited in Milroy and Gordon 2003), 30 tokens per variable is a reasonable obje ctive, and is generally a dividing line in statistics between large and small samples. For each participant, the number of tokens was approximately 90 100 in an interview and 70 in a reading section. The coding began about 20 minutes after an interview sta rted. This was an appropriate time
99 when a participant was assumed to be more relaxed and able to converse more spontaneously. The desired number of tokens was mostly located within 45 minutes. There were some cases, however, where an appropriate number of tokens were collected within 60 minutes. The number of occurrences of a word being counted was limited. There are several reasons why a particular word might be more common in an interview. For example, when the participant told a story about cars, the wo rd car or [rÃ³t] in Thai, containing an initial (r) consonant, was more likely to occur frequently in a conversation. According to Wolfram (1969) and Gordon (2001), counting the first three tokens of each transcription i s based on a lexically varied sample. Unclear contexts of (r) variant pronunciation were omitted from coding. Exclusion also included English loanwords and English words containing the variable (r) when the participants mixed their use of Thai with English in an interview. Additionally , it should be noted that different variants of (r) could be found in Thai , such as the voiceless counterparts of the initial consonant cluster . However, such variants occ ur relatively infrequently in the language (Harris 1972), and they can be difficult to differentiate in the process of data coding . In order to facilitate further analysis, the main focus for the present research is to examine the potential variants of (r) in the onset and consonant cluster contexts that are more likely to reveal the influence of social conditions on them.
100 For a reliability and accuracy check, a co coder 4 was employed in coding the data. When each transcription was completed, it was sent t o a co coder with a sound file enclosed. Words containing the variable (r) under investigation were highlighted on sound file and continued coding a distribution of the major (r) variants found on Excel scheduled to discuss the coding. When our coding did not match, in most cases, the co coder and I relistened to the sound file togeth er and settled the disagreement. When an agreement was not unanimous, the third person 5 listened to the speech and noted what variant it was. The co coder and I then checked whether it matched our coding and chose an agreeable variant . In a few cases in wh ich the coding was disagreed upon, unclear variants (1.38% of the total number of onset (r) and 1.51% of the total number of cluster (r)) were removed from the analysis. Statistical Methods In the present research, the effect of the linguistic factors of all tokens and extralinguistic factors of all participants wer e analyzed using Goldvarb, a VARBRUL software, which calculate s the probabilistic weight of different coded conditions to perform statistical analysis and display the significance of the factor s that influenced the selection of the variants of (r). In variation ist analysis, the linguistic variable that has the selection of one variant over the other is called the dependent variable. One of its variant s is considered the 4 The co coder is a native speaker of Thai and a graduate student in the department of linguistics. 5 The third person is a native speaker of Thai and a Ph.D. candidate in the department of linguistics.
101 application rule in Gol dvarb . In the present study , the lateral and the zero realization variants of (r) are chosen as the application rule s of the variable (r) in the on set and consonant cluster respectively . Other variants are considered the nonapplication rule s . The independe nt variable s are the features of the linguistic and extralinguistic context which have potential noticeable effect s on the choice of one variant over the other, and they are coded into factors or factor groups (Tagliamonte 2009:131). Each factor group cons ists of a set of factors which are independent of each other. It is required that there are no interactions among factor groups. In cases where there is an interaction of the factor groups, data recoding is required for adjustment (Paolillo 200 2 ; Tagliamon te 2009 ). Table 4 2 and Table 4 3 present the linguistic and extralinguistic coding factors used within each initial (r) and consonant cluster (r) factor group in the present study . A factor group of education is additionally included as a secondary social factor to examine whether or not it affects the production of (r) variants. Recall from Chapter 2 that the probability weight in variable rule analysis is measure d on a scale from 0 to 1.0. A weight greater than 0.5 suggests that a particular factor favo rs t he rule application, whereas a weight less than 0.5 disfavor s the rule application . A weight of 0.5 neither favor s nor disfavor s it. Each factor of a significant factor group is computed a probability weight based on the distribution of tokens of the v aria ble in the data. In addition, t he subtraction of the lowest factor weight from the highest factor weight in a factor group indicates the range, a relative strength of the factor group. T he step wise proc ess of the multiple regression analysis in the Go ldvarb program automatically indicate s the factor groups that have statistically significant
102 effects on the variable rule application. The first step is finding the group that suggests the most significant change to the model when it is included from the r est of the group. The first regression, called the step up procedure , includes factor group s one at a time and check s each new model with the previous added group. This way, the likelihood ratio statistic increases as significantly as possible. The program retains the most significant groups and continues to add more factor group s until a model cannot be statistical ly significantly improved. Then, the step down process begins to remove factor group s one at a time in o rder to test whether they improve o r wor sen the log likelihood. When no more factor groups without a significant effect on the likelihood of the model can be eliminated, the step down process stops. Generally , the best fit of the model in the step up and step down analysis should end with identi cal number s of significant factor groups (Tagliamonte 2009). T he chi square test is also employed in this research to examine whether or not each fact or within a factor group has a significant relationship with the variable (r) at a 95% confidence interva l, reflecting a significance level of 0.05 (Paolillo 2002). This chapter has described the linguistic contexts and extralinguistic factors applied in the present study . Information regarding participants, a description of the data collection technique: the sociolinguistic interview, equipment, procedure, data transcription, extraction and coding , and statistical method s are presented. The next chapter provides the data analysis and interpretation of the major findings of the variable (r) in the initial posi tion.
103 Table 4 1. Demographic information of participants Characteristics Groups Number of Participants Age 18 25 10 35 45 10 50 65 10 Gender female 15 male 15 Table 4 2. Coding factor group s of the initial (r) on Goldvarb X Factor group Factor Initial (r) r (a trill [r]) t ( a ) l ( a lateral [l] ) a ( ) Age 1 (18 25) 2 (35 45 3 (50 65) Gender f (female) m (male) Style I (informal) g (formal) Education d (vocational diploma) s (mas p (doctoral degree) Table 4 3. Coding factor group s of the consonant cluster (r) on Goldvarb X Factor group Factor Consonant cluster (r) r ( a trill [r] ) t ( ) l ( a lateral [l] ) n ( a zero re alization or a null [Ã¸] ) Age 1 (18 25) 2 (35 45 3 (50 65) Gender f (female) m (male) Style I (informal) g (formal) Education d (vocational diploma) p (doctoral degree)
104 Figure 4 1. Proportion of education level of the participants Note: The number of people in each education level is in parentheses.
105 CHAPTER 5 VARIATION IN THE INITIAL R: RESULTS AND INTERPRETATION This chapter presents the results of quantitative analysis of the usage of the variable (r) in the linguistic context of the word initial pos ition. The variationist analysis of the reali zation of the initial (r) and the distribution of each variant observed in the study are discussed as w ell as the interpretation. The current patterns of the production of the (r) variants in the initial position were examined to discover how the variable (r) plays a role in the language at the present time from the perspective of variationist sociolinguis tics. Age, gender, and speech styles were analyzed to demonstrate the correlation between the (r) variants and the extra linguistic factors. In addition, the education level of the speaker was included in the examination as previous studies (e.g., Beebe 197 4; Chunsivimol 1993) suggest that education tended to have an effect on the use of the variable (r) in different linguistic contexts, such as the onset (r) and consonant clusters. The r esearch questions and hypotheses are also addressed throughout the chap ter. The Initial ( R ) This section discusses the results of the variable (r) in th e initial position . A total of 3,353 tokens of the initial (r) were examined in th e present study. The analysis demonstrate d the patterns of variants elicited in the whole d ata to a nswer the research question, regarding to what extent the (r) variation is occurring in Bangkok Thai. The variants of the initial (r) are first presented, followed by an analysis of the distribution of the occurring (r) variants from the data. The extralinguistic factors of age, gender, education , and style were analyzed to answer whether or not and how social and stylistic factors affect the occurrences of the (r) variants among native speakers of
106 Bangkok Thai. Last ly, a multivariate analysis is p resented to measure the statistical significance of the factors that influence the variation patterns in the initial (r), to support or reject the hypotheses of the present research. The Four Variants of the Initial ( R ) In the variable context correspond ing to the pronunciation of the initial (r), the four major prestige forms in the word initial position (Harris 1972 ; Tingsabadh and Abramson 1993) . However, the trill is infrequent ly used in colloquial speech. It is generally found in emphatic speech and more formal situations, such as a formal announcement or an academic context (Beebe 19 74; Treyakul 1986 ; Phootirat 2012). According to Treyakul (1986) , the trill was most frequently used in the most formal style of minimal pair reading, and the tap was most preferable in a formal style of radio news announcement and in a more formal style o f passage reading. In contrast, the lateral is the variant commonly found in conversational speech ( e.g., Chunsuvimol 19 93 ; Panyaathisin 2013), although the lateral is considered nonstandard in the initial (r). Beebe (1974:232) states that the lateral is t he lowest status variant for the variable (r) when (r) occurs alone. The last variant found in the data is the approximant. Beebe (1974:119) states that native speakers of Thai who att ended Western own ed school s in Thailand or graduated from abroad tended to use the approximant in their speech. It is also suggest ed that the approximant occurs in the speech of native speakers of Thai with a high fluency in English, especially when their occupations are related to an English speaking environment (e.g., Ch unsuvimol 1993; Phootirat 2012) .
107 Overall D istribution of the Initial ( R ) Variants Table 5 1 and Figure 5 1 display the overall distribution of the four variants of the , ta regardless of the extralinguisti c factors. Figure 5 1 illustrates that the lateral wa s the most dominant variant in the initial position. The highest frequency of 55.9% suggest ed the preference of the lateral, as it was ubiquitously found among the spea kers. The second dominant variant wa s the flap, accounting for 33.6% of all occurrences, followed by the infrequent variant of the trill for 7.2%. The approximant occurred in the smallest proportion of 3.3%. Distribution of the Initial (R) V ariants by the Extra linguistic Factors In this section, the distributional patterns of the four variants are presented according to each extralinguistic factor . The four variants exhibit ed different patterns and tendencies along the social factors of age, gender, and a stylistic factor. Recall that education wa s examined as a secondary social factor to demonstrate the pattern of the variants used by speakers of different education levels. Age It wa s hypothesized that age influences the choice of the word initial (r) var iants. The older speakers might be more conservative in their speech than the younger speakers. The high rate of the standard fl ap and trill usage would occur more frequently than in the other two age groups, whereas t he younger group of speakers would use the nonstandard lateral the most frequently among the three groups of speakers . Table 5 2 demonstrat es according to the three generational groups of speakers. T he differences in the use of the initial (r) variants among speakers in different age groups were statistically
108 significant 2 = 72.6 , d f = 6, p < 0.001. The largest proportion of the variant wa s the nonstandard lateral. Speakers of all age groups extensively used the lateral, especially the younger speakers at 62.8%. The middle aged people used it at a proportion of 57.4%, followed by the older group of speakers at 47.2%. Though the older speakers selected the lateral at the lowest rate , the figure show ed that they tended to use the lateral at a high proportion when compared to the other (r) variants. The second dominant variant wa s the fl ap. It was mostly used by the older speakers at 38.4%, followed by the middle aged and younger speakers, selecting the flap at 32.5% and 30.2% , respectively. The trill and the approximant show ed a minimal preference among the speakers of all age groups. T he older speakers, however, used the trill most frequently at 10.4%, wh ereas the middle age d and young er speakers chose it at only 6.0% and 5.3% , respectively. The infrequent approximant was mostly employed by the middle aged speakers at 4.1%, and the olde r speakers at 4.0% . The younger speakers rarely used the approximant at a rate of 1.7%. A similar rate of the approximant occurrence among the middle aged and older speakers might be due to an exposure to English in their professions. Some speakers in thes e age groups work ed at a university or international organization where both Thai and English were mediums of communication. Th us , the production of the approximant in a foreign language might affect their initial (r) pronunciation in Thai, since the appro ximant i s not native Bangkok Thai (Beebe 1980:433). Furthermore, this finding correspond s (1993) study of the Thai and English onset (r) pronunciation of hotel employees. His findings suggest ed that the interaction in Engl is h with the hotel guests affected the
109 researcher. Table 5 2 additionally suggest s that there wa s a positive correlation between age and the production of the standard trill and flap. That i s , the older gro up of speakers ha d a higher proportion of the standard variant s than the other two age groups . This correlation is illustrated in Figure 5 2 as well with regard to the overall distribution of the (r) variants used by the younger, middle aged, and older spe akers. The results indicat ed that the lateral and the flap were the most preferr ed variants, whereas the trill and the approximant were infrequent among the speakers in all age groups. The nonstandard lateral show ed the reverse patterns of the selection ac cording to age of the speakers. The selection of the lateral showed a negative correlation to age , since the lateral was the most dominant variant used by the younger speakers, followed by the middle aged people . The older speakers used it the least freque ntly. The pattern of the standard trill was similar to the flap, in that the older speakers dominated the trill usage, followed by the middle aged people, and the younger speakers infrequently used the trill. The approximant show ed a slightly different pat tern . The middle aged speakers used the approximant slightl y more than the older speakers wh ile the younger speakers rarely used it. Figure 5 2 confirm s that the hypothesis holds true that the older speakers were more likely to retain their production of t he standard and prestigious flap and trill more than the other gr oups of speakers. As shown in Table 5 2 and Figure 5 2, the older speakers tended to be more conservative in their production of the standard flap . The trill show ed a similar pattern of occur rence in the groups of middle aged and older
110 speakers , although the trill was found minimally selecte d. A possible explanation for why the middle aged and older speakers 1 preserved their standard initial (r) production is conservatism . Since age marks an i norms (Eckert 1997), it is possible that the middle aged and older speakers ha d high production of the standard variant due to pressure for use of the standard language in their workplace. In addition, it i s suggested in the literature (e .g., Labov 1966; Trudgill 1974) that adults tend to conserve their use of the standard forms of language in the linguistic market 2 (D. Sankoff and Laberge 19 78) more than the younger speakers. In addition, adults are expe cted to exhibit appropriate behavior to enforce the normative timing in life events. This expectation can affect linguistic behavior, such as conservatism in adu lthood or the vernacular in pre adolescence (Eckert 1997). The younger speakers , on the other ha nd, are more likely to favor the nonstandard or the vernacular forms . In the present study, the speakers in the younger group are late adolescents in a younger generation who tend ed to employ their use of the no nstandard lateral in constructing identity wh ich seemed to be different from other age groups. Based on th e apparent time hypothesis, it i s also possible to hypothesize that the patterns of the (r) variants used by the three age groups suggest a generational change in progress of the (r) variants i n an onset position . In the apparent time perspective, it is assumed that people of different ages retain the speech patterns of their formative years ( Chambers 2002a). Thus , speech differences among people of different ages mirror the di fferences in the w ay people used their language in thos e years. It i s 1 In the present study, speakers in the older group are still working even though the y are over 60 . There is o nly one unemployed woman in this age group . 2 community.
111 notic eable that the standard and prestig ious flap and trill demonstrate a gradual increase when the age of the speakers i s older. These two standard variants of the initial (r) were more preferable among the older and middle aged people. In contrast, there i s a regular increase of the lateral when the age of the speakers i s younger. It indicate s that the younger speakers extensively used the lateral at the highest rate , while the middle aged and older spea kers showed a lower rate of lateral production. It i s suggested that the lateral show s its ongoing change in the pronunciation of the onset (r). Even though the lateral i s found outnumbered the other variants, the change might be considered at an initial stage because it i s apparen t that the standard flap remain s preferable at a high rate of occurrence among the speakers in all age groups . The results in Table 5 2 also indicate that the older speakers used the lateral approximately 10% less than the middle aged speakers, and 15% less than the younger speaker s . Notably, age show s an effect on the use of the variants of the initial (r). A high rate of the lateral suggest s the direction of change in progress. At the present time, it i s likely that the variati on of the onset (r) is progressing to the lateral pronunciation in a dimension of the age of speakers. Gender In the present study, gender was examined to discover variation across the relationship between gender differences and the patterns of the initi al (r) variants. The analysis examine d whether female or male speakers le d the use of the specific (r) variants in a word initial position. Women were hypothesized to use standard form s of the fl ap and trill more frequently than men, as previous studies (L abov 1972a, 2001; Trudgill 1972; Wolfram and Fasold 1974 ) suggest ed that female speakers favor ed
112 standard language and were more sensitive to the prestigious forms than male speakers. Table 5 3 provide s the overall distribution of the lateral, the flap, th e trill and, the approximant variants selected by female and male speakers. The chi square test revealed that gender play ed a significant role in the production of the (r) variants between female and male speakers 2 = 9.50, df = 3, p < 0.01. The figures of the lateral show ed the largest proportion of all variants used by two groups of gender . However, m en used the lateral at a higher frequency than women (58.4% versus 53.3%). The second dominant variant wa s the s tandard flap. Women used the flap more frequently than men at 35.4% and 31.9%, respectively. The trill and the approximant were infrequently produced by both gender groups of speakers. However, women chose the prestigious trill more often than men (8.0% ve rsus 6.5%), whereas both women and men used the approximant at a similar low rate of 3.3% and 3.2%, respectively. D ifferent degrees of the use of (r) variants between women and men are observe d from the results in Figure 5 3 . The nonstandard lateral wa s t he only variant the male speakers selected more frequently than women. It wa s also evident that women used the nonstandard lateral at the highest proportion among the four variants of the initial (r). In other words, both women and men used the same set of variants, but in different quantities. However, w omen tended to use the standard and prestigious flap and trill at a higher rate than men. The findings support the hypothesis that females preserved their choice of the standard for m s , considered better as corresponding to the social norm s , more than males. A common correlation pattern reflect ed that women seem ed to prefer the standard or prestigious forms in their speech more consistently
113 than men. In addition, t he nts in this study conform ed to previous studies, such as Trudgill (1974) , which suggest s that the standard and prestigious forms of language tend to be more powerful for women. Education In the initial study design step, education was not included as one of the main recruitment criteria. However, it was likely that the education level of the speakers could affect the choice of the (r) variants in the present research . As mentioned in the onal diploma, a 3 . For analytical purposes , the education classification use d two new categories . That i s, a vocational diploma wa s al degree wa s merg degree. It wa s hypothesized that the speakers with a higher level of education would maintain their standard pronunciation in the onset (r) more frequently than the group of people with a lower level of education. The speakers with a low er level of education, in contrast, would use the nonstandard variants at a higher proportion. V ariation in the distr ibution of all the (r) variants by education i s shown in T able 5 4. Table 5 4 and Figure 5 4 illustrate the distribution of the four varian ts of (r) of different education levels. It appear ed that there wa s a significant difference between the choice of the (r) variants and the different education level s of the speaker s 2 = 69.1 , df = 3, p < 0.001. Although, t he lateral was most frequently used by all speakers, overriding the other v ariants, t 3 Most of the participants earn There is only one partic ipant who receives a vocational diploma, and another o ne holds a doctora l degree.
114 degree or lower education was higher than that among the spe degree or higher education (58.6% versus 50.3%). The standard flap show ed a similarity in proportion of usage between both groups of speakers at 33.9% for the degree speakers. The pattern suggest ed that both groups of speakers preserved their use of the standard form, although the rate was less than the nonstandard lateral. Unlike the st andard flap, the standard trill rarely occurr ed, showing a low frequency of the overall distribution. However, the results show ed versus 5.5%). The approximant was also infrequently used , indicating a low rate of only 2.0% among the lower degree holders, wh ereas the higher degree speakers used it more often at 5.8%. However, this finding showed that the speakers with a higher level of education employed the use of the foreign variant of approximant. Th e reason might be due to their education environment, which allowed them to be in contact with English sources, such as foreign professors, studying materials , and international conferences, or lower education. The results with regard to education level of the speakers turn ed out to partly support the hypothesis in terms of the production of the standard variants. Only the standard trill demonstrate d a positive correlation with education. Th e speakers in the group with higher education conserved their use of t he standard trill more than the speakers with lower education. In contrast, the use of the standard flap in both groups of speakers appear ed to be nearly identical, in which the speakers lower degree used the fl ap slightly more frequently than the speakers with
115 higher degree . In addition, the results show ed that the lateral wa s negatively correlated with the education level of the speakers. The finding s upport ed the hypothesis that the speakers with a lower level of education seemed to use the lateral at a higher rate than the speakers with a higher level of education. Style In addition to classical social factors, such as age, gender and education , speec h style wa s taken into account for the research question of whether or not the formality of style affect ed the variation of the four variants observed in the present study. The speakers were hypothesized to be more careful of their production of the onset (r) when the degree of formality increase d . The standard flap and trill would occur more frequently in the passage readings, representing a relatively formal speech style, whereas the nonstandard lateral would appear more prevalently in the interview, repr esenting a relatively more informal speech style. Table 5 5 indicate s the potential influence of formality of style on the selection of the variants among the speakers. The chi square test suggested a significant differen ce between the selection of the (r ) variants and speech styles 2 = 693 , df = 3, p < 0.001. In the interview, the lateral wa s the most dominant among the four variants of (r) . It wa s commonly used by the speakers at 74.8% . The second preferred variant wa s the flap , which wa s selected at a much lower proportion of 17.9%. The approximant was infrequently used at 5.1%, and the trill occurred in only 2.1% of the overall distribution. It wa s noticeable that the speakers infrequently used the trill in the interview. This pattern support ed previo us studies , such as Beebe ( 1974 ), Treyakul ( 1986 ) , which stated that the standard and prestigious trill wa s more typical ly found in formal settings and emphatic speech.
116 Figure 5 5 present s a sharp contrast between the production of the lateral and the flap in the passage reading. The choice of the lateral dramatically decreased from 74.8% to 34.6%, whereas the reverse pattern occurred in the flap, which the speakers used it at the highest proportion of 51.3%. The trill also show ed an identical pattern to th e flap, in that the speakers increased their trill production in the passage reading, though the trill did not occur frequently (from 2.1% to 12.9%). In contrast choice of the approximant declined from 5.1% to 1.1%. An increase of the standa rd flap and trill in a formal style can be explained that the speakers were aware of the significance of the initial (r). Although the nonstandard variant was employed prevalently in informal speech, the speakers were more conscious that standard pronuncia tion was more appropriate when style switche d from informal to more formal. This pattern confirm ed the hypothesis that the speakers were more likely to increase their use of the standard flap and trill when the style became formal. In addition, these findi ngs correspond broadcasters 4 , in that the speakers preferred the nonstandard lateral in an informal style of interview, and the standard flap was mainly used in a relatively formal style of passage readings. Relationship of the E xtralinguistic F actors In this section, the relationship of the primary extralinguistic factors in this research was examined to answer the research question of how women and men of different ages var ied their pr oduction of the (r) variants in different speech styles . Figure 5 6 to Figure 5 8 illustrate the four variants of the initial (r) with regard to age and style , 4 Treyakul (1986) categorized four types of speech styles: an interview, news announcement, passage reading, and minimal pair reading.
117 age and gender, and gender and style in order to unfold the relationship between such external f actors and the occurrence of the (r) variants in the present study. Age and Style Figure 5 6 present s different patterns of the distribution of the variants by the speakers in different age groups in informal and formal styles . In an informal situation, th e speakers used the four variants at a similar rate of occurrence. The speakers highly employed the lateral and the fl ap in their speech wh ile the trill and the approximant were rarely selected. In a formal style, the speakers in all age group s reduc ed the ir use of the nonstandard lateral, and greatly employed the standard flap. The approximant occurred infrequently in all age groups, especially among the younger speakers, since it disappeared in a formal style. The results emphasize d the important roles of age and style in the production of standard forms, in that the older speakers were most likely to be conscious of using standard language. The older speakers switched from the nonstandard lateral to the standard flap most noticeably, followed by the middl e aged people, and the younger speakers changed to use the flap least frequently. The standard trill also show ed a similar pattern in the group of the older speakers when the style turn ed formal. That i s, the older speakers employed the selection o f the tr ill at the highest rate wh ile the younger and middle aged speakers use d the trill at the same frequency. Age and Gender Figure 5 7 illustrate s that women and men in different age groups varied their production of the four variants. The younger male speake rs highly preferred the lateral variant (68%), w hereas the older female speakers used the lateral the least (44%). The flap and trill show ed a similar pattern among the female and male speakers in different
118 ages. The older female speakers preserved their s tandard flap at the highest rate of 41%, followed by the older male speakers at 35%. The younger females used the standard flap more frequently than the males in the same age group (34% versus 27%). The middle aged males, on the other hand, used the flap (33%) and trill (7%) slightly At reason why middle the same age group more often than the middle aged female speakers (flap 32% and trill 5%), showing the slight preference for the stand ard variants among the males . A possible reason why may be due to the profession of the men in this age group, showing its effec t on their linguistic behavior, according to the linguistic market concept . That is, some male speakers in the middle aged group work in an academic setting, where the sp eakers think it is more proper to use the standard forms of language. The different language choices of middle aged men is further discussed in Chapter 7, after examining the questionnaire that provides the speaker use of the variable (r). The approximant minimally occurred in all age groups. The older male speakers and middle aged females used the approx imant at the highest rate of 6%, whereas women and men in other age groups infrequ ently employed the approximant. It is suggested that the females in the older group seemed to be highly conscious of using st andard pronunciation. T h ey frequently selected the flap and trill, but used the lateral at the lowest rate among women and men in a ll age groups. In contrast, t he younger male speakers were the least sensitive to the standard status of the (r) variants, as they used the nonstandard lateral the most frequently of the female and male speakers in all age groups. This pattern supports the observation that the nonstandard lateral is the norm of
119 the you nger generation, and the male s seemed to prefer the nonstandard (r) variant more than females , except the middle aged male speakers . Gender and Style Figure 5 8 demonstrate s the distribution o f the four variants with regard to gender and style. It i s indicated that men used the lateral most frequently than women in informal speech (72% versus 78%) . When focusing on the flap production of the male speakers, a notable example of self awareness of standard language us age wa s observed . The differentiation of the lateral between informal and formal styles of the male speakers wa s sharper than that of the female speakers. Although men most frequently used the lateral in an informal setting (78%), they switched their (r) variant to the style became more formal. In contrast , the female speakers demonstrated their self consciousness by using the standard and prestigious tr ill at a higher rate than the male speakers in a more formal situation. Although the flap and trill are both standard in the onset (r), the trill also carries an overt prestige, as it is preferred in a formal context and a high register of Thai (Harris 197 2). W support ed the hypothesis that women tend ed to prefer a standard and prestige form of language. However, the male speakers in the present study also showed their self awareness and a conscious attempt at using the standard language according to the formality of the situation. Therefore , it i s likely that the formality of style affected the selection of the variant s of the initial (r) more noticeabl y than gender differences did . Multivariate Analysis of the Initial ( R ) In this section, the results of a statistical analysis of the production of the initial (r) variants with regard to different factor groups are displayed . It is noted that showing only
120 the proportion of the distribution of each variant cor relating with the potential factors does not show whether or not each factor group is statistically significant. Thus, a multi variate regression analysis wa s performed to reveal the significance of specific factor groups. In addition, a factor weight and a relative strength of each factor group provide s statistical evidence of the characteristics of the initial (r) variation. Factor G roups U sed in the S tud y One dependent and four independent variables were included to perform the statistical analysis. Table 5 6 displays a list of the factors in each factor group used in the present study. The code i s assigned to each factor for running the Goldvarb X program. Factor group 1 is the four dependent variants of the initial (r), namely the lateral [l], the flap [ Factor group 2 5 are the independent variables corresponding to the extralinguistic factors considered in the present research. Factor group 2 represents the age of the speakers, con sisting of three groups of younger, middle aged , and older. Factor group 3 corresponds to the gender differences of female and male. Factor group 4 corresponds to style, consisting of a binary differentiation between informal and formal situations. Factor group 5 represents the education of the speakers. The education levels are classified according to a binary distinction: one degree degree. Before running the Goldvarb , it is important to find the best fit to account for the analysis. A factor group of the linguistic context of the onset (r) and four extralinguistic factor groups age, gender, style, and education were examin ed to figure out the potential gener al pattern of the initial (r) . The binomial one step test was performed to present the application and compare the probabilities for each factor across an
121 independent run . In addition, the binomial one step was tested to examine which factor group may not be well represented. The resu lt suggest ed that a factor group of education should be removed due to its insufficiency in a number of speakers across age groups. After removing the education factor group, the test was performed again. The result indicated that this run was a better fit of a model to the data analysis . After the discovery that eliminating the education factor group gave a model the best account for data analysis, the next step was the ex amination of the weight of each factor in the factor groups in order to reveal the s ignificance of the factors with regard to the choice of the variants of the onset (r). By measuring a relative strength, the ranges are compare d for each of the factor groups. The strongest factor group is ranked at the top of the table. After that, the f actor groups are presented in decreasing order of the strength that affect s the occurrence of the lateral. Within each factor group, the factor with the highest weight appear s first, followed by the factors ordered in decreasing order of strength. Signifi cant Factor Groups Table 5 7 suggests the three extralinguistic factor groups that contribute statistically significant effects to the choice of the lateral, used as the application rule in Goldvarb. T he following order of the choices indicates the relati ve importance of the factor groups, with the most significant factor group showing on the left: Style > Age > Gender It is evident that the largest effect is contributed by speech style, overriding the effect of age and gender. Not only is style shown as statistically significant, but the relative strength of its effect, with the widest range of 42, also stands out from tha t of other factor groups . Age i s the second strongest factor group with a range of 18, and gender
122 show s a narrow range of 6, suggesting that th e gender differences demonstrate the smallest effect on the production of the lateral in the onset (r). The following, Figure 5 9 to Figure 5 11, present th e effects on the realization of the lateral in the initial (r) position within each signific ant factor group. The factor groups are ordered from the most to the least significant . Style Figure 5 9 presents the effect of style on the realization of the initial (r) as the lateral. The results from a constraint ranking show that the lateral i s appa rently favored among the speakers in an informal style with a large factor weight of 0.70. In contrast, the lateral i s a disfavored variant with a small factor weight of 0.28 in a formal style. It can be seen that style shows a clear cut distinction of the lateral realization between informal and formal contexts. The preference of the lateral in an informal situation indicates a change of the social status of the lateral variant in the initial (r), since the lateral is dominantly used in conversational spee ch regardless of its nonstandard feature. The notion that a nonstandard or stigmatized form is considered incorrect and should not be chosen as a representation (Prasithrathsint 1989) is possibly changing with the high occurrence of the lateral in the ons et (r) . Wolfram and Fasold (1974) suggest that the status of the prestigious and stigmatized variant s can be unstable. In addition, it is suggested that the socially preferred variant possibly lose its prestige status when an innovative variant shows its p reference among speakers. The new variant will eventually replace the old variant. Also, t his change was evidenced in previous studies, such as the r ful pronunciation in New York City English (Labov 1966,
123 1972 a Edinburgh (Romaine 1978), and the labiodental approximant [ ] in British English in Middlesbrough (Llamas 2001) 5 . Previous research and the discussion above suggest that the lateral variant in the initial (r) is initiating a change of its social status, since it is widely used in colloquial speech. It is noticed that the nonstandard statu s of the lateral is likely to become the acceptable norm in informal settings in the future. The findings in the present study also indicate that the variable (r) in the onset position varies systematically in terms of the formality of style. However, a dramatic decrease of the lateral and an increase of flap and trill production in a formal style show using the standard varia nts remains to be observed. T he speakers are conscious that the standard forms of language are more appropriate, thus the prestige value of the standard flap and trill is preserved exclusively in a relatively formal context. Age Figure 5 10 illustrates th e effect of the age of the speakers and the realization of the lateral. The ranking of constraints shows that the younger and middle aged speakers favor the use of the lateral slightly above a neutral point with a factor weight of 0.58 and 0.52 , respect ively. is distinguished from the other groups, in which they disfavor the lateral , showing a factor weight of 0.40. The observed pattern demonstrates that the older speakers seem to be more linguistically sensi tive by highly producing the standard and prestigious flap and trill ( Figure 5 2). In addition, there is a negative correlation between age and the production of the lateral. That is, the yo unger speakers use the lateral at a higher rate 5 F urther information is provided in Ch apter 3 .
124 than the middle ag ed and older speakers. This pattern suggests a process of change in progress of the variable (r) in the initial position from three successive generations in this study. Gender Figure 5 11 presents the small effect of gender on the production of the later al. The results show the dissimilar preference for the nonstandard lateral between women an d men. The male speakers favor the lateral with a factor weight of 0.53 , while women disfavor it, showing a factor weigh t of 0.47. The results in the present study a lso support the claim that women have a tendency to preserve t heir standard variant of the initial (r) more than men. In this chapter, th e research questions and hypotheses with regard to the initial (r) have been answered and tested. First, the distributi on of t he four variants of the initial the nonstandard lateral variant was mostly preferred among native speakers, followed by the standard and prestigious flap and trill, and the approximant was found exc lusively among the speakers whose occupation was related to English. Second, t he results have demonstrated the correlation between the extralinguistic factors of age, gender , education, and style , and variation of the four variants of the onset (r ). In add ition, the chi square test reveal ed that the choice s of the (r) variants and the extralinguistic factor s were significantly different . Third , the present study has demonstrated a number of differences in the way that women and men of different age s varied their production of the (r) variants with regard to the formality of style. Lastly , a multivariate analysis was performed and illustrate d the statistical significance, a constraint ranking , and a relative strength of each factor group affecting the product ion of the (r) variants. The next
125 chapter will discuss the results of the second linguistic context of the consonant cluster (r). The key findings and interpretation of the consonant cluster (r) variation will be provided.
126 Table 5 1. Overall distribution of the four variants of the initial (r) Variant Number of tokens Frequency (%) Lateral [l] 1874 55.9 1128 33.6 Trill [r] 242 7.2 109 3.3 Total 33 5 3 100 Table 5 2. D istribution of the four variants of the initial (r) by age % N % N % N % N 18 25 62.8 709 30.2 341 5.3 60 1.7 19 35 45 57 .4 647 32.5 366 6.0 68 4.1 46 50 65 47.2 518 38.4 421 10.4 114 4.0 44 Table 5 3. Distribution of the four variants of the initial (r) by gender Gender Lateral [l] Approximant % N % N % N % N Female 53.3 885 35.4 588 8.0 132 3.3 55 Male 58.4 989 31.9 540 6.5 110 3.2 54 Table 5 4. D istribution of the four variants of the initial (r) by education % N % N L ateral [l] 58.6 1321 50.3 553 Flap 33.9 763 33.2 365 Trill [r] 5.5 125 10.6 117 5.8 64 Table 5 5. D istribution of the four variants of the initial (r) by speech style % N % N % N % N Informal 74.8 1327 17.9 318 2.1 38 5.1 91 Formal 34.6 547 51.3 810 12. 9 204 1.1 18
127 Table 5 6. List of the dependent and independent factor groups of the initial (r) Factor group Factor 1. Initial (r) variants t (flap) r (trill) l (lateral) a (approximant) 2. Age 1 (younger speakers) 2 (middle aged speakers) 3 (older speakers) 3. Gender f (female) m (male) 4. Style I (informal) g (formal) 5. Education Table 5 7. Multivariate analysis of the contribution of the extralinguistic facto rs selected as significant to the probability of the initial (r) Corrected mean .57 Total Number 3353 Group Factor Factor weight Percent 1 N Style Informal (interview) .70 74.8 1327 Formal (reading) .28 34.6 547 Range 2 42 Age 18 25 .58 62.8 709 35 45 .52 57.4 647 50 65 .4 0 47.2 518 Range 18 Gender Male .5 3 58.4 989 Female .4 7 53.3 885 Range 6 Note: Cross tabulation of the factor groups indicate s that each factor group is indepen d e nt. There is no potential interaction between the factor groups. 1 A percentage of the lateral is presented as the application value. 2 The r ange indicates a relative strength of the factor group. It is the subtraction of the lowest factor weight from the highest factor weight in a factor group.
128 Figure 5 1. Overall distribution of the four variants of the initial (r) Figure 5 2. D istribution of the four variants of th e initial (r) by age
129 Figure 5 3. D istribution of the four variants of the initial (r) by gender Figure 5 4. D istribution of the four variants of the initial (r) by education
130 Figure 5 5. D istribution of the four variants of the initial (r) by speech style Figure 5 6. D istribution of the four variants of the initial (r) by age and speech style 1 1 In Figure 5 6, n umber 1 attached to each variant re presents an informal style. N umber 2 represents a formal style. The number labeled above each bar is in percentage terms .
131 Figure 5 7. D istribution of the four variants of the initial (r) by age and gende r 2 Figure 5 8. Distribution of the four variants of the initial (r) by gender and speech style 3 2 The n umber labeled above each bar is in percentage terms . 3 In Figure 5 8, n umber 1 attached to each variant represents an informal s tyle. N umber 2 represents a formal style. The n umber labeled above each bar is in percentage terms .
132 Figure 5 9. Effect of speech style on the production of the lateral variant of the initial (r) Figure 5 10. Effect of age on the production of the lateral variant of the initial (r)
133 Figure 5 11. Effect of gend er on the production of the lateral variant of the initial (r)
134 CHAPTER 6 VARIATION IN THE CONSONANT CLUSTER R: RESULTS AND INTERPRETATION In this chapter, the variationist analysis of the second linguistic context , the consonant cluster (r) , was exa mined in order to explore the patterns of the (r) variants in the second unit of the clusters. The quantitative analysis demonstrate d the findings of the (r) variants occurred in the data, along with the interpretation. The extralinguistic factors of age, gender, education, and style were taken into account to analyze the correlation between the linguistic and extralinguistic factors. The C onsonant Cluster ( R ) The data presents the patterns of the variable (r) tokens of the consonant clu sters. A total of 2,215 tokens of the (r) variants were encoded for this variable context . The examination provides the extent of the (r) variation occu r r ing in the consonant cluster context . The cluster (r) variants and their distribution are first explained. Additionally, the roles of age, gender, education, and style were analyz ed and discussed to answer the research questions of whether or not and how these extralinguistic factors affect the distribution of the variants found in the present research , and to support or re ject the hypotheses of this study. The relationship of the extralinguistic factors i s also demonstrated for a better understanding of such factors and the occurrence of the (r) variants. Finally, a multivariate analysis provides a more accurate picture of the exact contribution of each factor gro up to the patterns of the consonant cluster (r) variation. The Four Variants of the C onsonant C luster ( R ) The f our major variants occur in the variable context of t he consonant cluster (r). They consist of a flap null or a zero realization [ Ã¸ ] . The
135 first variant is the flap realization, such as for [ Ã¢ :n ] (go home), and for [ ] (announce). The flap is standard and prestige. It is considered correct when pronou ncing the (r) variant of the Thai consonant clusters (Beebe 1974; Chunsuvimol 1993; Phootirat 2012). The second variant is the lateral, such as kl for [ ] (a bag) and p h l for [ p h lÃk ] (chili pepper). The lateral is assumed to have undergone liquid tran sformation (Beebe 1974:45), in which the second unit of (r) in a cluster is realized or transformed into [l]. The lateral has the in between status of being nonstandard, and it is more preferable than the null variant in the clusters (Beebe 1974:233). The third variant is the trill, such as kr for [ krÃ cÃ² k ] (a mirror) and pr for [ prÃ tu: ] (a door). The trill clusters are also standard and prestige in the language. However, the trill is uncommonly found in everyday speech. Its occurrence mostly occurs in forma l passage reading and formal public speech. The fourth variant is the null or the zero realization, suc h as [ ] for [ ] (straight) and [ p h Ã³rwÃ¢: ] for [ p h rÃ³rwÃ¢: ] (because). The non occurrence of (r) refers to the phonetic realization of the variable (r), in which the second unit of the consonant clusters (r) i s omitted . The zero realization of (r) is typically found in conversational speech, although it is a stigmatized variant in the clusters (Beebe 1974; Treyakul 1986; Chunsivimol 1993). Overall D istri bution o f the Consonant Clusters ( R ) Variants Table 6 1and Figure 6 1 indicate the overall distribution of the (r) variants of the elicited in the data regardless of the extralinguistic factors. The selection of the zero realization or the null variant dominate d th e other three variants. It show ed the largest proportion of 50.8% of the overall distribution . It can possibly be explained by the idea that the stigmatized status of the (r) omission of the clusters is changing. The status of
136 the zero realization in the c lusters has become more socially acceptable among the speakers at the present time. The second most dominant variant wa s the standard flap, occurring in 40.8% of the overall distribution, whereas the lateral and the trill were infrequently used, showing a frequency of only 4.1% and 2.3%, respectively. Distribution of the Consonant Cluster (R) Variants by the Extra linguistic Factors The distributional patterns of the four variants are presented by the extralinguistic factors. The variants demon strate the v ariation patterns and tendencies with the different external factors of age, gender, education, and style. Age It wa s hypothesized that age affect ed the selection of the consonant cluster (r) variants. The older speakers would conserve standard variants in their speech more than the younger age groups. The older speakers would employ the standard flap and trill more frequently than t he other two groups of speakers, while t he younger group of speakers would use the nonstandard null variant the mos t often among the three groups of speakers . Table 6 2 displays the overall distribution of the (r) variants chosen by speakers from three different age groups of younger, middle aged, and older. When taking age into consideration , there was a significant d ifference between the choice of the cluster (r) variants and the age of the speakers 2 = 69.7, df = 6, p < 0.001. T he (r) omission in the second unit of the clusters wa s the most frequently used by all speakers. The younger speakers omitted the (r) at th e highest rate of 58.9%, followed by the middle aged speakers (43.0%). The older speakers dropped the (r) variant the least often (41.1%). The second dominant variant wa s the flap. It show ed the reverse pattern of usage, in which the older speakers produce d the flap at the highest proportion of 52.9%,
137 followed by the middle aged speakers at 40.8%. The younger speakers used it the least frequently at 34.1%. The other two variants of clusters (r) show ed a much lower production. The speakers in all groups infr equently used the lateral and the trill . The younger speakers chose the lateral most often at a rate of 5.8%, whereas the middle aged and the older speakers selected the lateral at 4.0% and 2.7%, respectively. The least occurring variant of the trill was r arely used. The older speakers mostly selected the trill at 3.3%, followed by the middle aged speakers at 2.2%, and the younger speakers at only 1.3%. Table 6 2 indicates the distinct patterns of the distribution of the standard and nonstandard variants. There wa s a positive correlation between age and the use of the standard tap and trill. That i s, the older group of speakers ha d a higher production of the standard variants than the other two age groups. In contrast , the non occurrence of the (r) variant i n the clusters and lateral production demonstrate d the reverse patterns, showing a negative correlation to age. The younger speakers frequently used the zero realization and the lateral at the highest rate, foll owed by the middle aged people and the older speakers used these variants the least often. The correlations were also illustrated in Figure 6 2 with regard to the patterns of the four variants used by the younger, middle aged , and older speakers. It i s evident that the older speakers employed the st andard flap most frequently among the four variants, while the mi ddle aged and younger speakers dropped the (r) variant in the clusters most frequently. The distribution of the flap strongly support ed the hypothesis that the older people seemed to be more conservative in their use of the standard form of language than the other age groups. The middle aged speakers used the flap more
138 often than the younger speakers as well. In contrast, the high production of the zero realization and the lateral among the yo unger speakers suggests the preference of the nonstandard variants among the younger generation . In addition, the findings in the present study correspond to those of Beebe (1974) in the case of the production of the nonstandard variants. That i s, the rate of r reduction in the older groups was lower than that in the younger group. It is possible to hypothesize that due to social expectations for appropriate behavior and pressure for use of the standard language for their professions, the older and middle aged speakers preserved their standard variants of the flap and trill more frequently than the younger speakers. In addition, the pattern of the (r) distribution among speakers in different groups can be further explained by the apparent time hypothesis. L abov (1994) suggests that the language of individual speakers is stable across their lifespan once they have acquired it as children. Therefore, the high production of the standard flap in the older group suggest s that the older speakers acquired the flap as the variant of the cluster (r), and ha ve been using it up to the present time. The (r) omission, thus affected the (r) variation of the older speakers the least. The production of the fl ap (40.8 % ) and the r d ropping (43 %) among the mi ddle aged speakers occurred in similar proportions , showing that the middle aged speakers still preserved the use of the standard variant, but they also welcomed the nonstandard null variant. The younger speakers, on the other hand, employed high use of the (r) omission in t he cluster context, suggesting that there is an ongoing change process in the pronunciation of the consonant cluster (r). A major contrast between the zero realization and flap production among the younger speakers indica t e s that the change is
139 progressing to r reduction . However , the change is likely to progress in a n initial phase because the fl ap is still occu r r ing in a high frequency among the speakers in the middle aged and older groups. Gender To analyze the gender dimension of the variation, the resu lts present ed the correlation between the variants of (r) and the binary division between female and male speakers. The present study examine d the patterns of each (r) variant of the consonant clusters used by women and men. It wa s hypothesized that female speakers would use the standard flap and trill more frequently than male speakers. Table 6 3 presents the overall distribution of the zero realization, the flap, the lateral, and the trill used by women and men . It show ed that both female and male speake rs dropped their use of the (r) variant as a second unit of the clusters at the highest rate . It appeared that the use of consonant cluster (r) variants and two groups of gender were statistically significantly different , 2 = 17.47, df = 3, p < 0.001. The female speakers omitted the (r) more frequently than men ( 55.1% versus 46.4% ) . The second dominant variant wa s the standard flap. Men use d the flap at a higher proportion than female speakers (46.3% versus 39.3%). Unlike the zero realization and the flap, the lateral and the trill occur red in a relatively small frequency for both groups of speakers. However, the female and male speakers ha d different patterns of these infrequent variants. M en differed from women in their p roduction of the lateral. Male speakers used the lateral at a higher rate of 4.8%, whereas female speakers selected this variant at a rate of 3.4%. The least frequently used variant wa s the trill, which women and men chose at a low rate of 2.2% and 2.5%, r espectively.
140 In F igure 6 3 , the different degree s of (r) variant production between female and male speakers are observed. The high rate of the (r) deletion in the clusters among female speakers in the present research reject ed the hypothesis that women pr eserved their use of the standard variant more than men. The results suggest ed that among the four variants , w omen dominated the (r) deletion more often than men. In contrast, men used the standard flap and the nonstandard lateral at a higher proportion th an women. However, the nonstandard status of the lateral is more preferable than that of the zero realization according to Beebe (1974). Additionally, the (r) omission in the consonant cluster is considered stigmatized in Bangkok Thai (Chunsuvimol 1993). I t wa s also evident that the rate of the (r) dropping and flap production of the male speakers were identical at a rate of 46.4% and 46.3%, respectively. The producti on of the standard trill and nonstandard lateral occurred in a low rate for both women and men. It is observed that the m ale speakers h ad a balanced pattern between the production of the standard flap and the zero realization. It i s assumed that the higher production of standard flap among the male speakers i s due to their concern with correct pronunciation. However, it might not be a completely conscious effort to pronounce the clusters (r) more correctly because men showed a similar tendency to drop the (r) variant. In contrast, female speakers show ed an obvious distinction between the (r) dro pping and flap production. That i s, women omitted the (r) at a much higher rate than they used the flap , suggesting that it is not always the case that women invariably prefer the standard language. This phenomenon reflects the current speech behavior of f emale speakers and reveals that the stigmatized (r) omission is gaining status as being acceptable in the Bangkok speech community.
141 Education As mentioned that the education level of the speakers was included to analyze the patterns of the (r) variant in t he consonant cluster context , since it seemed to show its effect to the production of (r) variants . For analytical purposes, education was ree or lower level of education and education. It wa s hypothesized that the speakers with a higher level of education would show a higher rate of the standard variants of the clusters (r) than a group of people with a lower level of education. The speakers with a lower level of education, on the other hand, would use the nonstandard variants at a higher proportion. The distr ibution of the four variants by education is presented in T able 6 4. Table 6 4 and Figure 6 4 illustrate variation on the distribution of the zero realization, the flap, t he lateral, and the trill produced by the speakers of different education levels. The chi square test revealed that there was a significant difference between the selection of the (r) variants and the education level of the speakers , 2 = 71 , df = 3, p < 0.001. The speakers showed different patterns of their (r) dropping and the fl ap . and lower degree speakers omitted the (r) of the clusters at the highest rate of 57.2%, ed it at 39.1%. The use of the standard flap show ed a reverse pattern. T he speakers with a or higher degree chose the standard fl ap at a higher frequency of 52.9%, and the speakers wit the flap at a lower rate of 37.2%. The lateral and the trill presented a low occurrence among the speakers in both groups . The speakers with a higher level of education used the lateral relatively more frequently t han the speakers with a lower level of education ( 4.5% versus 3.9% ) . Unlike the
142 standard fl ap, the standard trill occurred infrequently , comprising a low percentage of the overall distribution. However, the results indicate d that the group of the higher degree speakers used the trill twice as frequen tly as the group of the lower degree speakers (3.6% versus 1.6%). A positive correlation between the level of education and the use of the (r) variants of the clusters support ed the hypothesis. The speakers in the group with a higher level of education preserved their use of the standard flap and trill more frequently than the speakers in the group with a lower level of education. A negative correlation between the education level and the (r) omission additionally confirm ed the grees dropped the (r) variant at a Nonetheless , the result of lateral production disagreed with the hypothesis, since the speakers with a higher level of education used the nonstandard lateral sli ghtly more often than the speakers with a lower level of education. Style In addition to the social factors of age, gender, and education, speech style was included to account for the research question of whether or not the formality of style affect ed t he variation of the four variants of the consonant clusters (r) . It wa s hypothesized that the speakers would be more careful of their choice of the (r) variants when the degree of formality increased. The standard flap would occur most frequently in the pa ssage reading, representing a relatively formal speech style. In contrast, the (r) deletion and the lateral would be found prevalen tly in the interview, representing a relatively more informal speech style. Table 6 5 presents the overall distribution of th e four variants by speech styles.
143 Table 6 5 suggests the potential influence of style on the occurrence of the (r) variants in the cluster context among the speakers. The chi square test indicates that the use of the (r) variants and the speech styles are significantly different 2 = 308.9, df = 3, p < 0.001. It i s evident that the speakers dropped the (r) variant at a very high proportion of 62.4% in a relatively informal style of interview. When the context change d to more formal passage reading , the sp eakers dramatically reduced the (r) omission to 27.8% . The rate of flap production also increase d to 66.6% in the passage reading , whereas the tap production decreased to 30.8% in the informal conversation . The lateral and the trill variants showed a small proportion of occurrence in both speech style s . However, the results suggest ed that style affect ed the production of the lateral, in which the speakers lessened their lateral usage from 5.6% in an informal style to 1.2% in a formal style. The use of the t rill showed a similar pattern to the flap, in which the rate of the trill increased from 1.2% in the interview to 4.4% in the passage reading. In addition, high occurrence of the (r) deletion in the present study co rroborat ed (1986) findings tha t the (r) dropping was the most dominant variant of the consonant clusters in informal speech among native speakers of Bangkok Thai (62.36%), whereas the lateral was most frequently found in the more formal style of the passage reading (60.71%). Figure 6 5 illustrates a sharp contrast between the (r) omission and the use of the fl ap in a formal context . It i s observed that the difference in patterning of the (r) deletion and t he flap usage wa s constrained by different speech styles. Th e finding s also suppor t ed the hypothesis that when style became formal, the speakers use d t he standard flap m ore frequently , and the rate of the (r) dropping occurred at the highest
144 proportion in conversational speech . In addition , a n increase of the standard fl ap and trill in a formal style wa s suggested as be ing due to concern s about correctness of pronunciation in formal situations . Even though the r lessness occurred extensively in an informal style, the speakers made a conscious effort to increase flap production i n formal situation s. The flap variant carrie s the standard value in the pronunciation of consonant cluster (r); thus, it i s considered more appropriate ly selected . A h igh proportion of the (r) omission additionally confirm ed the hypothesis that the (r) var iant was commonly dropped when speakers were unconscious of their speech in an informal situation. Relationship of the Extralinguistic Factors In this section, the relationship of the major extralinguistic factors in the present study was examined with reg ard to the formality of style associated with the social factors of age and gender. In other words, the examination aims to answer the research questions of how women and men of different ages var ied their production of the (r) variants of the clusters in different styles . The following, Figure 6 6 to Figure 6 8, demonstrate the (r) variants in the cluster context according to the extralinguistic factors of age and style , age and gender, and gender and style to give a better understanding of the occurrence of the (r) variants and the relationship of such factors . Age and Style Figure 6 6 indicates the distribution of the (r) variants by the speakers in different age groups in informal and formal speech styles. It i s evident that speakers of all age groups va ried their use of the (r) variants with regard to the formality of style. In an informal situation, the speakers in different age range s used the four variants at an almost identical rate of occurrence. That i s, the speakers frequently omitted the (r) and
145 us ed the fl ap in their speech, while the lateral and the trill rarely occurred in all age groups. For instance, the trill disappeared in informal style among the younger speakers. When moving from an informal to a formal style, the (r) omission decreased d ramatically, whereas the selection of the standard flap increased sharply in the formal style of the reading among the speakers in all age groups. The rate of the lateral decreased, while trill production increased to the highest rate among the older speak ers in a more formal context. In addition, style shows a greater stratification among the older speakers than among other age groups . The sensitivity to the standard status of the (r) variants of the older speakers resulted in the highest rate of flap pro duction and the lowest rate of the (r) omission in both informal and formal styles. On the other hand, the younger speakers omitted the (r) variant most frequently and used the flap least often in both styles. The rate of (r) deletion and flap production a mong the middle aged speakers occurred similarly to those of the younger speakers in informal speech (65% and 67%, respectively). However, the middle aged speakers selected the flap at a much higher rate than the younger speakers in a formal setting (67% v ersus 55%). The younger speakers used the lateral at the highest rate of 8% in informal speech. The speakers of all age groups rarely chose the lateral in the formal style. The trill occurred at a minimal rate in both styles in all age groups. However, the rate of trill slightly increased in formal speech, in which the older speakers used the trill most frequently at a proportion of 6%, followed by the middle aged and younger speakers at 4% and 3%, respectively. Age and Gender As illustrated in Figure 6 3 , in the previous section, that female speakers dropped the (r) variant more frequently than men. In this section, Figure 6 7 presents the pattern
146 of the (r) variants among female and male speakers with regard to age differences . It was to examine whether wo men and men in different age groups show ed similar or different patterns of the (r) variants, especially the (r) omission. It was evident that women in all age groups omitted the (r) variants more often than men in their groups. The younger female speakers showed the highest rate of the (r) deletion (62%), followed by the middle aged women (59%) and older women (44%). The (r) omission among the male speakers suggest ed a negative correlation in terms of age differences. The younger males dropped the (r) more frequently than middle aged men and older men (55%, 47%, and 38%, respectively). Flap production showed a different pattern. The results indicate that the older speakers used the flap most frequently at a rate of 56% for males and 50% for females. Women a nd men in the younger group had a similar average rate of flap production (34%). Flap usage in the middle aged group showed a sharp differentiation between the female and male speakers. The male speakers used the flap at a relatively higher proportion than female speakers (48% versus 34%). It is suggested that the (r) omission is the variant of the younger females, younger males, and the middle aged females, whereas the flap is the norm of the older women, older men, and middle aged men. The lateral was inf requently used among speakers of all age and gender groups. The younger males showed the highest use of the lateral (10%), followed by the middle aged women (6%). The speakers in other groups rarely selected the lateral. The trill shows a similarly low occ urrence between w omen and men in all age groups. Gender and Style Figure 6 8 illustrates the distribution of the four variants of the clusters (r) in terms of gender and style. Th e pattern of the four variants suggest s that st yle play s an
147 important role in the selection of the variant s among the female and male speakers. The results regarding gender and style show ed an identical pattern of flap and trill production to those of age and style, in which women and men used the flap and the trill more frequently when style became more formal. It was observed that female and male speakers used different speech styles according to the formality of speaking situations. The most noticeable pattern wa s that women and men used the flap more frequently and lessened thei r (r) deletion in the clusters when style wa s more formal. The production of the lateral also decreased between women and men when the style turn ed formal. In addition, the infrequent trill present ed a similar pattern to the use of the flap, in which the p roportion of the trill slightly incre ased in a more formal setting in both gender groups . Whe n focusing on gender differentiation , men use d the fl ap more frequently than women in both informal (34% versus 27%) and formal (70% versus 63%) contexts. On the other hand , women drop ped the (r) in the clusters at a higher frequency than men in both informal (67% versus 58%) and formal (32% versus 23%) contexts. Women and men decreased their nonstandard lateral production when style turned more formal. The male sp eakers used the lateral at a slightly higher rate than the female speakers in both styles. The rate of the trill occurrence of women and men were similar in an informal style. However, the female speakers showed a slightly higher proportion of trill produc tion than male speakers in formal speech (5% versus 4%). It is evident from the results that men tended to b e more self conscious of using the standar d variants in the clusters than women in a more formal context. Men switched from the (r) omission to the standard flap at sharply different rates when the style turned more formal. The
14 8 ed their self awareness of using the standard form of language, although the pattern wa s not apparent lik suggest ed that formality of style i s an important factor for the use of the standard variants between women and men. Multivariate Analysis of the Consonant Cluster ( R ) In this sectio n, the statistical results of the (r) variants of the consonant clusters are presented with regard to the different factor groups. As mentioned in the previous chapter , examin ing on ly the distribution of each variant in relation to the potential factors is insufficient to see whether or not the factor groups under the study are statistically significant. M ulti variate regression analysis wa s performed to reveal the factor groups that were statistical significant to the use of the (r) variants of the clusters . A factor weight and a relative strength of each significant factor group also provide s statistic al evidence of the characteristics of the (r) realization in the consonant cluster context. Factor Groups Used in the Study One dependent and four independent variables were taken into account to perform the statistical analysis. Table 6 6 demonstrates a list of the factors in each factor group with the code assigned to each factor for running the Goldvarb X program. Factor group 1 represents the dependent var iants of th e consonant clusters (r): the zero res ent research. Recall that f actor group 2 5 are the independent variables corresponding to the extralinguistic factors considere d in the present study. Factor group 2 corresponds to the , consisting of three groups of younger, middle aged, and older. Factor group 3 represents the gender differences of female and male. Factor group 4
149 corresponds to speech styles, compri sing a binary distinction between informal and formal styles. Factor group 5 represents the education background of the speakers. The education levels are operationalized according to a binary distinction: one group of a degree and one degree. The binomial one step analysis present ed the application and compare d the probabilities for each factor across an independent run before running the statistical test. The results suggest ed that all the factor groups f itted t o account for the analysis. After that, a multivariate analysis was performed to examine the statistical sig nificance of each factor group. Significant Factor Groups Table 6 7 presents the four factor groups that suggest statistically significant e ffects to the zero realization of the cluster (r ), used as the application rule i n the multivariate analysis. The order of the factors i s ranked from the most to the least significant groups that affect the (r) dropping . A mong the four significant factor g roups that account for the cluster variation, the following order of the choice s show the relative importance of the factor groups , with t he most significant factor located on the left: Style > Education > Gender > Age The statistical analysis provide s a m ore detailed pattern of the variants in the cluster. For instance, the overall distribution noticeably s how s that the older speakers favored the flap more frequently than speakers in other groups. However, the statistical results suggest that age i s not th e strongest constraint affecting the selection of the (r) variants. When running a statistical examination, it i s observed in Table 6 7 that speech style i s the strongest factor group, overriding the factor groups of education, gender, and age. The largest range of 37, belonging to style, indicate s that it most significantly affects an
150 occurrence of the zero realization in the cluster (r). Education i s the second strongest factor group , showing a wide range of 20. Gender i s ranked third with a range of 14. Beebe (1974:27) noted that gender 1 had no significant effect on consonant cluster variation in her study. S tatistical evidence in the present research , however, suggest s that gender differences significant ly a ffect the use of the consonant cluster (r) . Age appears to be the weakest constraint in the consonant cluster context, indicating the smallest range of 9 in the relative ranking . Figure 6 9 to Figure 6 12 demonstrate the effect on the zero realization within each significant factor group. The factor w ith the highest weight i s presented first, followed by the factor group s ordered in decreasing order of strength. Style Figure 6 9 displays the effect of the factor group of style on the realization of the (r) deletion in the clusters. The results from the constraint ranking suggest that the zero realization i s noticeably favored among the speakers in informal speech with a high factor weight of 0.63. In a formal setting, the (r) omission i s disfavored, showing a low factor weight of 0.26. It i s noticed tha t the rate of zero realization in an informal speech style i s distinct from the rate of such realization in a more formal speech style. Also, i t i s evident that speakers commonly dropped the (r) in the second unit of the clusters in an informal speech styl e, although the (r) omission carrie s nonstandard and stigmatized features. The preference of the (r) deletion in informal speech suggest s a change in progress of the social status of the zero realizatio n in the cluster context. As mentioned earlier, the st andard or prestige value of one variant can possibly change over time. It i s 1 Beebe used the term sex in her study.
151 suggested in the present study that the status of the (r) deletion in the clusters seem s to be more accept able to the speakers in an informal situation. A similar high occurrence of the zero realization i s also evidenced in previous studies (e.g. , Beebe 1974; Chunsivimol 1993). In addition, the (r) dropping occur s at a relatively high rate among speakers, especially in an informal setting (Treyakul 1986). Neve rtheless, the speakers disfavor the (r) deletion in a more formal situation. The d istribution of the (r) omission insignificantly shows a relative strength of only 0.26. It can be observed that style i s a key factor that connect s aware ness, since the sp eakers greatly reduce the (r) omission in formal speech. Therefore, the prestige value of the standard variants i s preserved in a more formal speaking setting . Education Figure 6 10 presents the effect of the education background of the speakers , correspo nding to the distribution of the zero realization. Although education is considered a complementary extralinguistic factor in the present study, it does affect the consonant cluster (r) variation. The findings indicate that the group of speakers with a bac showing a lower factor weight of 0.37. It is suggested that speakers with a higher le vel of education seem to preserve their standard production of the cluster more than those with a lower level of education . The possible reason may be due to more difficulty in pronouncing the consonant clusters. That is, it may require more effort to arti culate a sequence of consonants used as onset in the word compared with pronouncing a singl e initial consonant . Thus, speakers with hi gher education , who are likely to receive longer and more appropriate l anguage training at school , may be able to maintain their
152 standard (r) variant of the clusters more significant ly than their counterpart s . Also, s peakers with a higher level of education may not frequently drop the (r) of the consonant clusters in their academic setting, since they are more accustomed to u sing the standard forms in their academic writing, reading, and presentation s . Gender Figure 6 11 presents the effect of gender on the (r) dropping in the clusters. The results indicate the different preference of the zero realization between the female and male speakers. Women favor the (r) omission with a factor weight of 0.57, whereas men disfavor such a variant with a lower factor weight of 0.43. A high occurrence of the (r) deletion in the cluster among the female speakers support s the suggestion in t he present study that the nonstandard status of the zero realization becomes more neutral . In addition, it is noticeable that the pattern of women The results indicate that f linguistic choices do not necessarily conform to the standard form of language , and males do not always prefer the vernacular . The data use of the nonstandard (r) variant opp oses a typical concept that women are more sensitive to the p restige variant s , as suggested in earlier studies (e.g., L a bov ( 19 66 ); Trudgill (197 2 ) ; Chunsuvimol (1993 ) . Age Figure 6 12 presents a range of effects of the age of the speakers on the zero realization. The younger and middle aged speakers prefer to drop the (r) of the clusters with an identical factor weight of 0.53, whereas the older speakers disfavored such a variant, showing a lower factor weight of 0.44. The results indicate d that the older speake rs preserve their use of the standard forms more than the other two age groups.
153 This pattern suggest s that the pronunciation of the cluster (r) is in the process of change . The (r) deletion realization shows its direction of an ongoing change not only among the younger speakers , but also among the middle aged people . This chapter has addressed the research questions and tested the hypotheses with respect to the consonant cluster (r) . First, the distributi on of the four variants of the cluster (r) namely the zero realization, the fl ap, the trill, and the latera l has been discussed. In general , the (r) omission i s the most favored among the speakers, followed by the standard fl ap . The nonstandard lateral and the standard tril l are disfavored, as th ey rarely occur . Second, the results have shown the re are correlat ion s between the extralinguistic factors of age, gender , style , and education and variation of the four variants of the cluster (r) . T he statistic al results indicate the significant difference s between the (r) variants and each extralinguistic factor. In a ddition, t he key findings with regard to such ex tralinguistic factors and th eir interpretation s have been discussed . Lastly, the present research has demonstrated the relationship between the extralinguistic factors under examination. This study has provid ed the differen t patterns of female and male speakers of different age s, varying their use of the variants according to the formality of speech style. Additionally , a m ultivariate analysis was performed and presented the statistical significance, a constra int ranking , and a relative strength of the factor group s, affecting the production of the (r) variants. The next chapter provides a conclusion of the present research and discusses the status of the variable (r) in Bangkok Thai on the basis of the perspec tive of the native speakers.
154 Table 6 1. Overall distribution of the four va riants of the consonant cluster (r) Variant Number of tokens Frequency (%) Z ero realization [Ã¸] 1125 50.8 F 948 42.8 L ateral [l] 91 4.1 Trill [r] 51 2.3 Total 2215 100 Table 6 2. Distribution of the four va riants of the consonant cluster (r) by age Age Zero realization [Ã¸] Lateral [l] Trill [r] % N % N % N % N 18 25 58.9 408 34.1 236 5.8 40 1.3 9 35 45 43.0 407 40.8 313 4.0 31 2.2 17 50 65 41.1 310 52.9 399 2.7 20 3.3 25 Table 6 3. Distribution of the four va riants of the consonant cluster (r) by gender Gender Zero realization [Ã¸] Lateral [l] Trill [r] % N % N % N % N F emale 55.1 614 39.3 438 3.4 38 2.2 24 Male 46.4 511 46.3 510 4.8 53 2.5 27 Table 6 4 . Distribution of the four variants of the consonant cluster (r) by education % N % N Z ero realization [Ã¸] 57.2 818 39.1 307 F 37.2 532 52.9 416 L ateral [l] 3.9 56 4.5 351 T ri ll [r] 1.6 23 3.6 28 Table 6 5. Distribution of the four variants of the consonant cluster (r) by speech style Style Zero realization [Ã¸] Lateral [l] Trill [r] % N % N % N % N Informal 62.4 919 30.8 454 5.6 82 1 .2 18 Formal 27.8 206 66.6 494 1.2 9 4.4 33
155 Table 6 6. List of the dependent and independent factor groups of the consonant c luste r (r) Factor group Factor 1. Cluster (r) variants t (flap) r (trill) l (lateral) n (null/zero realization) 2. Age 1 (younger speakers) 2 (middle aged speakers) 3 (older speakers) 3 . Gender f (female) m (male) 4. Style i (informal) g (formal) 5. Education Table 6 7. Multivariate analysis of the contribution of the extralinguistic factors selected as si gnificant to the probability of the consonant cluster (r) Cor rected mean .50 Total Number 2215 Group Factor Factor weight Percent 1 N Style In formal (interview) .63 62.4 919 Formal (readi ngs) .26 27.8 206 Range 37 .57 57.2 818 M .37 39.1 307 Range 20 Gender Female .57 55.1 614 Male .43 46.4 511 Range 14 Age 18 25 .53 58.9 408 35 45 .53 53.0 407 50 6 5 .44 41.1 310 Range 9 Note: Cross tabulation of the factor groups indicate s that each factor group is independent. There is no potential interaction between the factor groups. 1 A percentage of the zero realization is presented as the applicat ion value.
156 Figure 6 1 . Overall distribution of the four va riants of the consonant cluster (r) Figure 6 2. Distribution of the four va riants of the consonant clust er (r) by age
157 Figure 6 3. Distribution of the four variants of the consonant cluster (r) by gender Figure 6 4. Distribution of the four variants of the consonant cluster (r) by education
158 Figure 6 5. Distribution of the four va riants of the consonant cluster (r) by speech style Figure 6 6. Distribution of the four variants of the co nsonant cluster (r) by age and speech style 1 1 Figure 6 6: Number 1 attached to each variant represents an informal style of an interview. Number 2 represents a formal style of readings. The number lab eled above each bar is in percentage terms.
159 Figure 6 7. Distribution of the four va riants of the consonant cluster (r) by age and gender 1 Figure 6 8. D istribution of the four variants of the consonant cluster (r) by gender and speech style 2 1 The number labeled above each bar is pres ented in percentage terms . 2 Figure 6 8: Number 1 attached to each variant represents an informal style of an interview. Number 2 represe nts a formal style of readings. The number labeled above each bar is in percentage terms.
160 Figure 6 9. Effect of speech style on the zero realization of the consonant cluster (r) Figure 6 10. Effect of education on the zero realization of the consonant cluster (r)
161 Figure 6 11. Effect of gender on the zero realization of the consonant cluster (r) Figure 6 12. Effect of age on the zero realization of the consonant cluster (r)
162 CHAPTER 7 DISCUSSION AND CONCLUSION The pr esent research has shown a significance of the variable (r) with regard to its sociolilnguistic variation in the Thai language. This study has analyzed the patterns of the variable (r) in Bangkok Thai in two different linguistic contexts: the onset positio n and (r) as a second member of the consonant clusters. The findings reveal the correlation between the variable (r) and the extralinguistic factors of ag e, gender, education, and style, s uggesting that such internal and external factors affect the choice of a particular (r) variant among native speakers of Bangkok Thai. In addition , this study reflects the speech behavior of speakers as members o f the Bangkok speech community and suggests an ongoing change of the (r) variants and the social evaluation of t he current status of the variable (r) in both linguistic contexts . This chapter provides a summa ry of the major findings of the (r) variation in the initial position and the consonant clusters, followed by a discussion of the current sociolinguistic status of the variable (r) in Bangkok Thai according to the findings in the present research. This conclusion also sheds light on the status of the variable (r) from the perspective of native speakers who were debriefed at the end of the interviews. Finally, the chapter closes with broader implications, limitations of the current study, and suggestions for future research. The Initial ( R ) From the data obtained from native speakers, the present study has found the distribution of the four major variants of the in lateral [l], and the Among the four variants, the flap and the trill are considered standard and prestigious in the language, while the lateral is nonstandard.
163 The fourth variant, the a pproximant, is also nons tandard. It is used the least frequently among the native speakers. The findings suggest the general pattern of (r) variation in an onset position, in which the nonstandard lateral was the most dominant variant, followed by the stan dard fl ap, the standard trill, and the nonstandard approximant . The order is shown from the least to the most preferable (r) variants in the initial position as [r] [l]. In addition, quantitative analysis of th e production of the four variants suggests the five ma jor findings with regard to the social factors of age, gender, and education , as well as a stylistic factor . Statistical evidence also indicates that these external factors significantly affect the use the initial (r) variants among native speakers. First, the s peakers in all age groups extensively used the lateral and th e fl ap, whereas t he speakers infrequently selected the trill and the approximant. The younger speakers highly favored the lateral. In contrast, the older speakers were most conservative in their production of the standard fl ap and trill, in which it can be a result of a social expectation to exhibit an appropriate linguistic behavior. Also, it i s assumed that an occurrence of the approximan t among the middle aged and older speakers was due to the English language environment at their workplace. Based on app arent time hypothesis, there was a regular increase of the lateral usage when the age of the speakers was younger. A generational change in progress of the (r) variants suggest s that the increase of the lateral variant show s a pattern of ongoing change in the pronunciation of the onset (r). Second, t he findings with respect to gender differentiation show ed that t he most prefer able variant among the male speakers was the lateral, whereas women used the standard and prestigious flap and trill more frequentl y. Females and males used the
164 approximant at an almost identical low rate of occurrence. The findings were evidence to support previous studies that women seemed to retain the ir use of the standard forms, corresponding to social norm s more consistent ly tha n men. Third, i n terms of education, the speakers with a higher level of education preserved their use of t he standard trill more than the speakers with a lower level of education. In contrast, the production of the standard fl ap in both groups of speakers showed a similar tendency. Th at is, th e speakers with the or lower degree used the fl ap slightly more often than the speakers with the or higher degree. For the nonstandard lateral, the group with the or lower degree speaker s selected the lateral at a higher rate than the group of speakers with the or higher degree. . This may be due to more exposure to other languages, such as English, in their education and professional environment, which ha d an influence on their use of the approximant in Thai (r). Fourth, style show ed an obvious effect on the production of the in itial (r). The nonstandard lateral wa s the norm in informal speech, whereas the standard flap and trill were used at a very high proportion in a more formal context . The rate of the flap and the trill increased in a more formal style as a result of the spe awareness of the significance of the initial (r) . The speakers were conscious that such standard variants were considered more appropriate in more formal settings. For the approximant, the speakers used it more often in conversational speech th an in more formal situations.
165 Lastly , the findings indicate that women and men in different age groups varied their variants of (r) in different speech styles. In terms of age and style, the findings show that the speakers in all age cohorts reduced their use of the nonstandard lateral, and greatly employed the use of standard flap in a more formal style. However, the older speakers were most conscious of using standard variants of (r). It was noticeable that the older speakers switched from the nonstandar d lateral to the standard flap. The younger speakers, in contrast, used the flap least frequently in a more formal style. The older speakers also selected the trill most frequently, whereas the younger and the middle aged speakers used it at the same propo rtion. The findings with regard to age and gender suggest ed that the younger male speakers highly used the lateral, whereas the older female speakers chose the lateral the least frequently. The standard flap and trill show ed a similar pattern among the fe male and male speakers in different age groups. The older female speakers were likely to retain their standard flap at the highest rate, followed by the older male speakers. The younger females used the flap more frequently than the younger males. In contr ast, the middle aged females selected the standard variants slightly less often than the middle aged male speakers. The approximant was infrequently used by the speakers in all age groups. It is suggested that the older female speakers were most conscious of using the standard pronunciation as they highly used the flap and the trill, and selected the lateral at the lowest rate among the female and male speakers in all ages. The younger male speakers showed the least sensitivity to the standard status of the (r) variants, since they used the lateral most frequently among all the speakers in the three groups.
166 In terms of gender and style, both women and men showed their self consciousness of the standard language according to the formality of the situation . T he results suggest that formality overrides the gender factor in selecting the (r) variants. However, the females and males showed their preferences of the (r) variant differently. The male speakers changed their use of the nonstandard lateral to the stand ard fl ap at a very high proportion when the style was more formal, w hereas women preferred to use the standard trill more than men. In addition, the statistical results of the contribution of the extralinguistic factors indicate d the three significant fac tor groups of style, age, and gender that account for the (r) variation in the onset . The speakers seemed to adhere to the norms of the society that consider standard speech more appropriate for a relatively more formal speaking situation. Speech style has the most significant influence on the selection of different variants of the (r), with the largest range of the factor weight, followed by The C onsonant Cluster ( R ) The f our major variants of (r) occurred in the consonant cluster context. They [ Ã¸ ] . The flap and trill are considered standard and prestige . The lateral is nonstandard, and the deletion of (r) in the second unit of the cluster is considered nonstandard and stigmatized. The findings show ed a pattern of (r) variation in the consonant cluster context: the (r) omission was most frequently used among the native speakers of Bangkok Thai, followed by the flap, the lateral, and the trill, illustrati ng the order from the least to the most preferable (r) variants of the consonant cluster (r) as follows: [r] [l] [Ã¸]. In addition, quantitative analysis reveals the five key findings regarding the correlation of
167 the four variants with the extralinguistic factors of age, gender, education , and speech style. Additionally, statistical results show that such exter nal factors pertaining to the speakers significantly affect their use of the consonant cluster (r) variants. First, speakers of different ages, in general, removed the (r) of the cluster most frequently in their speech . A negative correlation between the ( r) deletion and the age of the speakers showed that the younger speakers dropped the (r) most often, followed by the middle aged and older speakers . Also, there wa s a positive correlation between age and the production of the flap and trill, in which the o lder speakers used the flap at the highest rate , followed by the middle aged and younger speakers. The patterns of the (r) deletion among the older and younger speakers indicate d that the variable (r) in the cluster context is undergoing change. T he direct ion of change is moving to the non occurrence of (r) of the clusters . The change, however, seem s to be at a n initial stage, since the standard fl ap remains preferable among the middle aged and older speakers. Second, in terms of gender, women and men showed different degree s of (r) variant production. W omen employed higher production of the nonstandard (r) variant . T hat is, women tended to omit the (r) in the second unit of a cluster more frequently than men, while men used the flap and the lateral more ofte n . The trill occurred at a similarly low frequency in both gender groups . However, women used the trill slightly more frequently than men. The high rate of the (r) omission among female speakers suggests that the nonstandard status of this variant becomes more prominent among the native speakers . The results of the present study contribute to the field of language and gender by suggesting a current pattern of language use of women and men. The
168 study shows that women are less sensitive to nonstandard languag e, at least in the consonant cluster context. They did not typically conform to the standard variants of (r) Third, the finding s with regard to education show ed a positive correlation between education level and the use of the (r) variants . The s peakers with a higher level of education employed their use of the standard fl ap and trill more frequently than the speaker s with a lower level of education. However, the higher level of education speakers used the nonstandard lateral more frequently than the l ower level of education speakers. In addition, a negative correlat ion between the education of the speakers and the (r ) omission showed that lower degree speakers omitted the (r) of the cluster at a higher proportion than the group of the higher degree speakers. Fourth, the findings regarding stylistic variation show ed that pa tterns of the zero realization and the flap were explicitly governed by speech styles. There wa s a sharp difference in the use of (r) clusters in informal and formal styles. The (r) dropping was more prominent in conversational speech . When style turned fo rmal, the speakers consciously switched from omitting the (r) to using the standard flap. Style wa s suggested to be a key factor that connect ed awareness of using standard language in formal speech. The nonstandard lateral and the standar d trill showed a similar pattern, in which the speakers reduced the use of the lateral in an informal context, and increased the trill production in a more formal style.
169 Lastly, the findings suggest ed that female and male speakers of different age s varied their choice of the (r) variants in different speech style s . The (r) omission decreased dramatically, w hile the standard fl ap increased sharply in a more formal style among the speakers in all age groups . The rate of the lateral also decreased, while the trill production increased to the highest rate, especially among the older speakers. The speakers of all age groups rarely selected the lateral in a formal style. The older speakers tended to be most sensitive t o the standard variants. Th ey used the standa rd flap most frequently and omitted the (r) the least often in both informal and formal styles. The younger speakers showed a reverse pattern, in which they used the standard flap least often and dropped the (r) variants at the highest rate in both speech styles. The middle aged speakers also used the flap at a much higher rate than the younger speakers in a more formal context. In terms of age and gender, the results indicate d that women in all age groups dropped the (r) more often than men in their group s. The younger female speakers showed the highest rate of (r) deletion, followed by the middle aged and older women. The younger males dropped the (r) more frequently than the middle aged men and older men. Flap production of the middle aged speakers showe d a sharp contrast between female and male speakers. The male speakers seemed to favor the flap much more than women. These patterns suggest ed that the (r) omission wa s preferable among younger females, younger males, and middle aged females, wh ile the fla p wa s the norm of the older women, older men, and middle aged men. The lateral and the trill occurred infrequently among speakers of all age and gender groups.
170 With respect to gender differences and style , i t wa s suggested that women and men were more con scious of using different standard variants in a more formal situation . That is, men favored fl ap production, whereas women preferred to use the trill. In fact, m ales used the fl ap more frequently than females in both informal and formal speech settings. I n contrast, the female speakers used the trill slightly more freque ntly than the male speakers in formal speech. W omen also deleted the (r) of the clusters more often than men in both styles , while the male speakers used the lateral slightly more often tha n the female speakers in both styles. Add itionally, the statistical tests further revealed the four significant social and stylistic factor groups affecting the use of the (r) variants of the clusters. The results indicate d that s tyle ha d the relatively s trongest effect , showing its most significan t feature on an occurrence of the cluster (r) variants . The following strong effects are the factor group s of education level , gender , and age. As shown in the sections above, the variable (r) in the onset and c luster positions suggests the particular type of the linguistic variable, showing its sociolinguistic status in Bangkok Thai. Further discussion is provided in the following section. The Status of the Variable ( R ) in Bangkok Thai On the basis of the findi a short questionnaire 1 , this section discusses the current sociolinguistic s tatus of the initial (r) and the consonant cluster (r) in Bangkok Thai. The questionnaire provides additional information the (r) 1 Information in the question naire shows that all participants can explain how to pronounce the standard /r/ in Bangkok Thai (a fl ap and a trill).
171 pronunciation in Bangkok Thai . All t he excerpts of the participants included in this section were translated from Thai into English. As mentioned in Chapter 1 and Chapter 2 with respect to the st atus of the variable (r), the aforementioned findings of the realization of (r) observed in the onset position and the consonant clusters seem to suggest that the variable (r) functions as a sociolinguistic marker in Bangkok Thai. The variables (r) in the two linguistic contexts ha ve their social interpretation overtly attached. Th at is, th ey show their marking with both social characteristics and stylistic differentiation of their distribution. There are several reasons to consider the status of the variab le (r) in Bangkok Thai as a marker. First, it is noticeable that the standard variants of (r) mark with a more formal speech style. It shows that the standard flap has a style salient meaning. In other words, style is an important feature that links with s ocial awareness of using the standard variant of (r) in the onset and clusters. As discussed in the previous chapters, speakers significantly switched to the standard flap at a very high rate when style turned more formal because of their self consciousnes s. Also, an increase of the flap in a more formal speech situation suggests that the variable (r) has a significant feature in Bangkok Thai. Evidence from the participants corroborates that native speakers acknowledge that different variants of (r) should be selected according to the formality of the speech situations . Speakers are aware of the nonstandard features attache d to the lateral in the initial (r) and the zero realization in the (r) clusters , diffusing in the speech community . Even th ough the lev el of awareness of the nonstandard lateral and (r) omission is high, it is unlikely to perceive the variable (r) in Bangkok Thai as a sociolinguistic
172 stereotype. That is, the use of such nonstandard (r) varia nts in both linguistic contexts do es not seem to be actively avoided in casual speech. In other words, their social significance is not great enough to be a stereotype that highly affects conscious speech behavior. Additionally, the findings in Chapter 5 and Chapter 6 s uggest that native speak ers frequently used such variants especially in conversational speech. Excerpt (1) to Excerpt (5) show that the female and male participants of different ages, gender , and education levels seem to agree that the lateral in the onset (r) and the (r) deletio n of the clusters are acceptable variants in casual speech. For example, P7 2 said that he did not use the (r) very often in the clusters. However, the participants realize that the standard (r) in both linguistic contexts is a better choice in general and in a more formal situation, such as a formal public performance because the standard forms are considered more correct. In addition , P19 , a male participant in the middle aged group commented that the standard (r) may not be regularly used in casual talk. T he participant further noted in relation to his profession that he and his colleagues clearly pronounce the standard (r), since they are university lecturers. This is evidence that the speaker knows that he is expected to speak appropriately because the s tandard language is the medium of instruction in an academic environment . opinion provides additional interesting information , s uggesting that occupation may His co mment additionally supports the findings in Chap ter 5 ( Figure 5 7 ) , in which the middle aged male speakers used the standard flap and trill in the initial (r) at a slightly higher rate than middle aged female speakers. 2 P stands for Participant .
173 3 ) Question: In your opinion, how do native speakers of Bangkok Thai pronounce (r) in an onset position and consonant clusters? Do speakers use the same or different (r) in different speaking situations? P1: In an informal conversation, Thai people use [l] for the sound /r/. But in formal speech, people pronounce the /r/ sound clearly and correctly. Excerpt (2): P7 (male, 19, undergraduate student) P7: The (r) pronunciation of the clusters is infrequently in everyday speech, and I do not use it often. However, speakers should pronounce cluster (r) in a proper way in public. P14: Now, speakers infrequently pronounce standard initial (r) and clusters in conversational speech. Speakers are more careful of their speech when speaking in public or in news announcement. They are more likely to pronounce the (r) more clearly and appropriately. Excerpt (4 ): P19 (m P19: In everyday speech, the use of (r) may be omitted, but in my job, other lecture r s and I pronoun ce the (r) very clearly. The clear (r) pronunciation is also found among people working in mass media. P21: Speakers emphasize their standard initial (r) and cluster (r) in a formal situation, such as p ublic speech or news broadcasting, while speakers are not serious to do so in a conversation. Next , it is suggested that the nonstandard (r) va riants in the onset and cluster contexts associates with the age of the speakers. That is, the younger speakers extensively used the lateral in the former linguistic context, and they omitted the (r) in the latter one. Comments from the participants in different age groups also indicate that younger speakers tended to employ their use of the lateral for the initial (r), and dropped the (r) for the clusters. Even the younger speaker, P4, states that the younger 3 Gender, a ge, and an education level of the participants show in parentheses.
174 opinions regarding the (r) pronunciation at the present time are shown in Excerpt (6) to Excerpt (8). Excerpt (6): P4 (female, 20, undergraduate student) Question: What do you think about the pronunciation of an initial (r) and consonant cluster (r) of native speakers of Bangkok Thai nowadays? P4: The (r) pronunciatio n tends to decrease because teenagers do not think of pronouncing the (r) in both contexts, or they forgot to pronounce bec ause they were not instilled the correct pronunciation when they were younger . Excerpt (7): P11: Young speakers do not care to correctly pronounce an initial (r) and consonant cluster (r). They pronounce [l] for the sound / r/ . This pattern also occurs in their writing when they misspell words containing the consonant (r). Excerpt (8): P20 (male, 39, P20: The (r) of the clusters is disappearing especially among the younger generation. They delete the (r) in a cluster. Lastly, statistical results indicate that the omission of the (r) in the consonant clusters marks with the spe ech of female speakers with a lower level of education. This group of women dropped the (r) more frequently than their counterparts, females with a higher level of education, and males with a higher level of education ( Figure 6 10 and Figure 6 11). Therefo re , it is suggested that the (r) deletion or the zero realization has a significant feature for gender and education in the consonant cluster context. Additionally, the findings in the previous studies (e.g., Beebe 1974; Treyakul 1986; Chunsuvimol 1993) and the present research suggest that the nonstandard status of the lateral in the onset position and the nonstandard and stigmatized status of the (r)
175 omission in the consonant clusters are changing. Extensive use of the lateral and the (r) dropping indic ates that the status of these variants is more acceptable in casual speech. In short, their nonstandard and stigmatized features are more neutral. It is likely that the lateral and the (r) deletion are in the process of becoming the norms of the initial (r ) and the consonant cluster (r) in colloquial speech. Nonetheless, the progress of a change of (r) in both linguistic contexts is considered in its early stage. Even though native speakers favor the lateral , the second dominant variant in an initial positi on is the standard flap, followed by the standard trill, and the approximant ( Figure 5 1). For the consonant clusters, the (r) omission is the most preferable variant, while the second favored variant also appears to be the standard flap, followe d by the l ateral and the trill ( Figure 6 1). It shows that speakers maintain their consistent production of the flap, regardless of the extralinguistic factors. In addition to evidence of using the standard flap found in the present study , the participants express that the standard (r) pronunciation should be preserved in the language . Their opinions reflect that native speakers have the concepts of correctness and appropriateness attached to the and older participants also stated , as shown in Excerpt (9) to Excerpt (11), that the standard pronunciation of (r) should be practiced accordingly, especially among the younger generation, in order to pro nounce the (r) in a more proper way. Excerpt (9): P3 (female, 22, undergraduate student) Question: How would the initial and cluster (r) pronunciation in Bangkok Thai be in the future?
176 P3: If we do not seriously train or practice young children to pronounce (r) correctly, speakers would not pay attention to pronounce it appropriately. Also, many people have social values to pronounce (r) in Thai similarly to (r) in English which may affect how to pronounce (r) in the future. Excerpt (10): P7 (ma le, 19, undergraduate student) P7: The (r) of the clusters tend to disappear, and we should find the way to preserve it , including the initial (r). Excerpt (11): P25 (female, 56, Ph.D.) P25: There should be a national campaign to promote the importance of pronouncing an initial (r) and the clusters (r) correctly. Public people, such as television personality and famous people should be a role model to use the language properly. This section discusses the sociolinguistic status of the variable (r) in th e two linguistic contexts and suggests that the variable (r) functions as a sociolinguistic marker in Bangkok Thai, as supported by attitudes of the participants in a questionnaire and the findings in the present study . It is further suggested that the non standard and stigmatized status of the lateral in an onset position and the zero realization in the clusters becomes more acceptable in co lloquial speech. However, the participants expressed that the standard variants of the flap and trill should be conser ved in Bangkok Thai. The following sections provide broader implications, limitations of this study, and suggestions for future research. Broader Implications The findings of the present research have implications for the field of variationist study, th e situation of language change, and pedagogy. First, the present study has theoretical and methodological implications for the field of variationist sociolinguistics. This study provides further evidence of phonological variation, and corroborates that
177 in herent variability in language can occur at different levels of languages (Labov 1972a; Bod, Hay , and Jannedy 2003 ) . Second, the bridge of an examination between the linguistic and extralinguistic aspects in this research suggests that linguistic variation in the language is associated with the effect s of the external factors, and it can be systematically explained. Third , with respect to the methodological aspect, the present study has suppor ted that the sociolinguistic interview is an effec tive data colle ction technique for obtaining useful data to uncover the patterns of the variable (r). Also, the technique can be adapted to develop other studies on different levels of language variation and change in Bangkok Thai and other dialects of the Thai language . Next , as suggested in Chapter 1 , a multivariate analysis is not widely to present statistical evidence in many studies on variation in Thai; the use of quantitative evidence in the present research, therefore, serves as a good example of employing a curre nt statistical tool with a multifactorial approach , as well as the chi square test to reveal the patterns of the variable (r) corresponding to the external factors. Lastly, the findings of the present research can be the basis for a further investigation o f (r) variation and other linguistic variables in order to present a current language use and predict a direction of language change in the future. Another implication is the empirical findings in this research . They indicate the situation of the (r) varia tion in the speech of native speakers at the present time. As discussed in the previous chapters and the section s above, it is evident that there is an ongoing change of the variable (r) in the two contexts of the initial position and (r) in the second uni t of clusters associated with the extralinguistic factors. The dominant use of the lateral for the onset (r) and the (r) omission for the cluster (r) in a colloquial context
178 support previous literature in the past forty years (since Beebe 1974) that the no rm of the (r) pronunciation is changing. For instance, the rate of standard flap production in the initial position decreases in casual speech, but the flap is preserved in more formal contexts. Additionally, the current situation of (r) usage suggests a d irection of change in progress. It is possible that replacement of the lateral variant for the initial (r) in casual speech will occur more commonly in the future. However, it is suggested that the merger of two different phonemes of /r/ and /l/ in Thai s eems unlikely to occur in the initial (r), at least in the near future. This is due to the fact that phoneme /r/ and /l/ represent sounds of different written alphabetical symbols in the rule of Thai orthography . repres ents /l/ in the Thai writing system. Figure 5 5 also supports this assumption by indicating that the speakers used the standard flap and trill at a rate of over 60%, while the production of the lateral decreased to 34% in the reading passage . Lastly, in t erms of pedagogical implications for language education, the findings provide useful information for teaching the Thai language. It can be seen from the findings that the variables (r) in th e two linguistic contexts are changing. The lateral for the onset (r) and the (r) omission of the consonant clusters become more commonly used in conversational speech. However, based on opinions of na tive speakers in the present study , the standard pronunciation of (r) should be preserved to symbolize the uniqueness of the Thai language . It is, therefore, suggested that Thai language instructors may want to use extra language learning activities and courses to encourage younger students to develop proper pronunciation skills. For older students, Thai language instructors may cooperate with English language teachers to provide activities
179 in order to promote the appropriate pronouncing of the standard (r) in Thai in the initial position and consonant clusters. It would also allow students to learn how to speak Thai and Engl ish appropriately in different speech contexts. Limitations of the Study The present research reveals the patterns of (r) variation with respect to social and stylistic variation. However, there are several limitations that should be addressed. First, ther e is lack of speech data from retirees. Labov (1994) suggests that when people are retired, they may have fewer social pressures from their workplace; thus, their favoring of non prestigious forms may reoccur. The present study expected to obtain the speech of some retired speakers from the older group (50 65 years old). However, due to time constraints of the data collection, retirees who fit the participant recruitment criteria were not located. Nine out of ten participants in the older group are still wor king, and the one remaining participant is a housewife. Therefore, the speech of retired speakers that may have been informative to compare with the patterns of (r) production in the other groups of speakers is restricted. Second, the variety of linguisti c contexts is limited. The variable (r) was not analyzed when located in other positions, such as in the second or third syllable of a word. The linguistic contexts under examination are only the onset position of (r) and the consonant clusters. Different (r) variants may have been found if other lin guistic contexts were included. Third, there is limited information to be gained from a short questionnaire. An open the (r) sound and the use of (r) in different speech situations. Th erefore viewpoints with respect to gender differences and education were not obtained.
180 Fortunately, in the questionnaire, most speakers provided comments about the age of the speakers in rela tion to the production of the variable (r). Comments on women and for further analysis. Lastly, due to lack of access to the current socio economic census of native residents of Bangkok, th e present study did not include social stratification or occupation to differentiate the social class of the participants. It would be more informative to include such social factors of the participants to further examine and interpret their effects on the (r) variation. Suggestions for Future Research The present study has examined the variable (r) in Bangkok Thai with regard to certain extralinguistic dimensions. There are some suggestions that would be interesting to explore in future research. The fir st recommendation implies the limitations of the study. An examination of the speech patterns of retired speakers and children would be intriguing. The examination could investigate the (r) production of the retirees with regard to different social factors , such as gender, social c lass, and level of education. A comparative study of the retired speakers and working speakers could also be interesting, investigating the similarities and differences of language behaviors between two groups of speakers. The st udy of the variable (r) among children would Second, the study of (r) variation with respect to other social factors a nd linguistic environments would provide insightful findings. As noted, education i s taken into account as a complementary factor in this research. More thorough examination of the education level of the speakers and other social factors, such as occupatio n, would be
181 interesting. It is also production of the variable (r). In addition, different linguistic variables, such as stress, pause, following vowels, and a syllable position wou ld be good examples of internal factors to analyze to determine whether and how they affect the use of the variable (r) in Thai. Third, the current study aim s to examine the major (r) variants that are potentially affected by extralinguistic factors. Thus, it exclude s some rarely occurring (r) variants . Further analysis could investigate such variants to examine their distribution and whether or not they suggest a relation ship to external factors . Fourth , it is suggested that markers are sociolinguistic va riables that indicate both social and stylistic variation (Coupland 2007). Future research could include social stratification as one of the external factors, as well as stylistic differentiation, to further explore the variable (r), and in order to test t he suggestion in the present study that the variable (r) acts as a sociolinguistic marker in Bangkok Thai. Finally, the findings in this research would be useful for a longitudinal study of the sociolinguistic variable (r) in the future. As shown in previo us longitudinal studies (e.g., Fowler 1986; Cedergren 1987; Yaeger Dror 1994; Sankoff and Blondeau 2007), analysts would be able to investigate the same or different speakers at different time to explore the relation between change over the course of a lif e time and generational change . Researchers could replicate the study by revisiting the Bangkok speech community in the next 10 y ears, or later, in order to examine whether or not native speakers change their linguistic behaviors during their lifetimes. The findings would also provide real time evidence of the change situation of the variable (r) in Bangkok Thai.
182 Closing This dissertation has provided the patterns of the variable (r) in the onse t and the initial consonant clusters. The aspects of social and stylistic variation have been taken into consideration to reveal the correlation between linguistic and extralinguistic factors on language variation. Ideally, the present research of the variation in the realization of the sociolinguistic variable (r) in Bangkok Thai will shed light on an understanding of language variation and change and their relationship with extralinguistic factors. This dissertation will encourage continued quantitative analysis research on language variation of the Thai language, an d will contribute to our understanding of variationist sociolinguistics becom ing more prominent in Thailand .
183 APPENDIX A INFORMED CONSENT FORM S The following are informed consent forms, English and Thai versions. It is noted that a debriefing session was included in the study after approval from the University of Florida Institutional Review Board (IRB). Thus, this session was not shown as part of the tasks in the informed consent forms. However, each participant was informed of a debriefing questionnaire after a second task was completed. In addition, the IRB has been informed of this debriefing section. The IRB acknowledged and approved this additional procedure.
184 English Ve rsion
186 Thai Ve rsion
187 APPENDIX B T HE SOCIOLINGUISTIC INTERVIEW TOPICS A se t of topics in the sociolinguistic interview was designed to elicit everyday speech related to the general experiences of living in the city of Ban gkok. In order to maintain the natural flow of the interview, the order of topics and the structure of the in terview were loosely constructed. The following topics were asked to the participants in Bangkok Thai. Some topics were not asked to all participants. Only certain topics that drew attention were chosen. Demographics information Childhood memories Family a nd friends Life in Bangkok School and Study Work experience Future plan Leisure activities: traveling, sports Natural disaster: big flood in Thailand, tsunami in Japan, etc. Weather Traffic Ente rtainment Nation and world current issues Economics The devel opment of the city of Bangkok World events
188 APPENDIX C READING PASSAGES Reading Passage 1
189 Reading Passage 2
190 APPENDIX D QUESTIONNAIRE 2 1 . 1 2 __________ _______________________________________________________________________ _________________________________________________________________________________ _________________________________________________________________________________ 2 . 1 2 _________________________________________________________________________________ _________________________________________________________________________________ _________________________________________________________________________________ 3 . _________________________________________________________________________________ ________________________________________________________________________________ _________________________________________________________________________________ 4 . _________________________________________________________________________________ ________________________________________________________________________________ _________________________________________________________________________________ 5 . _________________________________________________________________ ________________ ________________________________________________________________________________ _________________________________________________________________________________
191 LIST OF REFERENCES Bailey, Guy. 2002. Real and apparent time. In Ja ck K. Ch ambers, Peter Trudgill and Natalie Schilling Estes (eds.) The Handbook of Language Variation and C hange . Oxford: Blackwell Publisher. 312 332 . Bailey, Guy, Tom W ikle, Jan Tillery and Lori Sand. 1991. The apparent time construct. Language Variation and C hange 3(3): 241 264. Bayley, Robert. 2002. The quantitative paradigm. In J ack K. Chambers, Peter Trudgill and Natalie Schilling Estes (eds.) The Handbook of Language Variation and C hange . Oxford: Blackwell Publisher. 117 141. Beebe, Leslie M. 1974. Soc ially conditioned variation in Bangkok Thai. PhD dissertation. Ann Arbor, Michigan : Uni versity of Michigan . Bell, Allan. 1984. Language style as audience design. Language in Society 13(2) : 145 204. Bell, Allan. 2001. Back in style: Reworking audience design. In Penelope E ckert and John R. Rickford (eds.) Style and S ociolinguistic V ariation . Cambridge: Cambridge University Press. 139 169. Bell, Allan. 2007. Style and the linguistic repertoire. In Carmen Llamas, Louise Mullany and Peter Stockwell (ed s.) The Routledge C ompanion to S ociolinguistics . London and New York: Routledge. 93 100. Bickerton, Derek. 1973. Quantitative versus dynamic paradigms: the case of Montreal que. In Charles J.N. Bailey and Roger W. Shuy (eds.) New W ays of A nalyzing V ariat ion in English . Washington: Georgetown University Press. 23 43 . Bickner, Robert J. and Journal of South Asian Literature 25(1): 163 175. Bloomfield, Leonard. 1933. Language . New York: Holt. Bod, R ens, Jennifer Hay and Stefanie Jannedy. 2003. Probabilistic L inguistics. Cambridge and Massachusetts : MIT Press. Cameron, Deborah. 200 6. Gender. In Keith Brown (ed.) Encyclopedia of Languages and L inguistics . Oxford: Elsevier. 733 739. Cameron, Debora h. 200 9. Demythologizing sociolinguistics. In Nikolas Coupland and Adam Jaworski (eds.) The New Sociolinguistics Reader . New York: Palgrave Macmillan. 106 118. Catford, John, C. 1988. A Practical Introduction to P honetics . Oxford: Clarendon Press.
192 Cederg ren, Henrietta.1987. The spread of language change: Verifying inferences of linguistic diffusion. In Peter H. Lowenberg (ed.) Language Spread and L anguage Policy: Issues, I mplications, and C ase S tudies . Washington DC: Georgetown University Press. 45 60 . C edergren, Henrietta and David Sankoff. 1974 . Variable rules: Performance as a statistical reflection of competence. Language 50(2): 333 355. Chalermsanyakorn, Suntaree. 2000. Variation of (kh) in Bangkok Thai by phonetic . Unpublished m aster thesis. Bangkok : Chulalongkorn University . Chambers , Jack K. 2002a. Patterns of variation including change. In Jack K. Chambers, Peter Trudgill and Natalie Schilling Estes (eds.) The H andbook of L anguage V ariation and C hange . Oxford : Blackwell Publisher. 349 372 . Chambers , Jack K. 2002b . Dynamics of dialect convergence . Journal of Sociolinguistics 6(1) : 117 130. Chambers , Jack K. 200 3 . Sociolinguistic Theory . Oxford and Malden : Blackwell. Cheshire, Jenny. 1982. Linguistic variati on and social function. In Suzanne Romaine (ed.) Sociolinguistic Variation in Speech C ommunities . London: Edward Arnold. 153 166 . Cheshire, Jenny. 2002. Sex and gender in variationist research. In Jack K. Chambers, Peter Trudgill and Nathalie Schilling Estes (eds.) The Handbook of L anguage V ariation and C hange . Malden : Blackwell. 423 443. Chomsky, Noam. 1965. Aspects of the Theory of Syntax . Cambridge and Massachusetts : MIT Press. Chunsuvimol, Boonruang. 1993. Relationship between the social variation of (r) in Thai and (r) in English in the speech of Bangkok Thai speakers. Unpublished PhD dissertation. Bangkok : Chula longkorn University . Clermont, Jean and l'air. I n Pierrette Thibault ( ed.) Le FranÃ§ais P arlÃ©: Etudes Sociolinguistiques . Edmonton: Linguistic Research. 13 28. Coates, Jennifer. 2004. Women, Men, and L anguage. London: Pearson Longman . Coupland, Nikolas. 2007. Style: Language Variation and I dentity . Cambridge and New York: Cambridge University Press.
193 Cressey, William W. 1978. Spanish P honology and M orphology: A G enerative V iew . Washington: Georgetown University Press. Dhebbayasuwan, M.L. Boonlua. 1975 . The F orgotten Thai L anguage . Bangkok : Teacher Training Department . Eckert, Penelope. 1988. Adolescent social structure and the spread of linguistic change. Language in Society 17(2): 183 207. Eckert, Penelope. 1989. The whole women: sex and gender differences in variation. Language Variation and Change 1 : 245 267. Eck ert, Penelope. 1997. Age as a sociolinguistic var iable. In Florian Coulmas (ed.) The H andbook of S ociolinguistics . Malden: Blackwell Publisher. 151 167 . Eckert, Penelope. 2000. Language V ariation and S ocial P ractice . Oxford and Malden: Blackwell. Fisch er, John L. 1958. Social influences on the choice of a linguistic variant. Word 14: 47 56. Fontanella de Weinberg and Maria Beatriz.1979. DinÃ¡mica S ocial de un Cambio L ingÃ¼Ãstico. Mexico: UNAM. Fowler, Joy. 1986. The social stratification of (r) in New York City department stores, 24 U npublished manuscript. New York : New York University Giles, Howard and Peter F. Powesland. 1975. Speech S tyle and S ocial E valuation . London: Academic Press. Gordon, Matthew. J. 2001. Small T own V alues and B ig C ity V owels: A study of the Northern Cities Shift in Michigan . The America n Dialect Society 84. Durham , North Carolina : Duke University Press. Grant, William. 1914. The P ronunciation of English in Scotland . Cambridge: Cambridge University Press. Gumperz, John Joseph. 1968. The speech com munity. In David L. Sills (ed.) Intern ational E ncyclopedia of the S ocial S ciences . New York: Macmillan. 381 386. Gumperz, John Joseph. 1971. Language in S ocial G roups . Stanford : Stanford University Press.
194 Guy, Gregory. R.1980. Variation in the group and individual: the case of final stop deletion. In Will iam Labov (ed.) Locating Language in T ime and S pace . New York: Academic Press. 1 36. Guy, Gregory. R. 2003. Variationist approaches to p honological change. In Brian D. Joseph and Richard D. Janda (eds.) The H andbook of Historical L ingui stics , Malden and Massachusetts : Blackwell. 369 400. Haeri, Niloofar. 1996. The S ociolinguistic M arket of Cairo: Gender, C lass and E ducation . London: Kegan Paul International. Harris, James. 1969. Spanish Phonology . Cambridge and Massachusetts : MIT Pre ss. Harris, Jimmy G. 1972. Phonetic notes on some Siamese consonants. In Jimmy G. Harris and Richard B. Noss (eds.) Tai P honetics and P honology . Bangkok: Central Institute of English Language. 8 22. Harris, Jimmy G. 1996. The c onsonant sounds of 17th ce ntury Siamese. Mon K hmer Studies Journal 21 : 1 17. Herman, David. 1999. Toward a socionarratology : New ways of analyzing natural language na rratives. In David Herman (ed.) Narratologies: New P erspectives on N arrative A nalysis . Columbus: Ohio State Un iversity Press. 218 246. Holmes, Janet. 1986. Function of you know Language in Society 15(1) : 1 21. Holmes, Janet. 1992. An introduction to S ociolinguistics. London: Longman. Holmes, Janet. 1995. Women, Men and P oliteness . London: Longman. Homkaew, Suppharerk . 1997. Variation of ( ) in Nakhon Si Thammarat Thai by age and language attitude . Unpublished m aster thesis. Bangkok : Chulalongkorn University . Horvath , Barbara. 1985. Variation in Australian English: The S ociole cts of Sydney . Cambridge: Cambridge University Press. Hubbell, Alan Forbes. 1950. The P ronunciation of English in New York City: Consonants and V owels Hudak, Thomas J. 2009. Some Historical Background of Thai Language. Last accessed 23 May 2014 at http://thaiarc.tu.ac.th/thai/thai.htm
195 Huebner, Thom. 2006. codemixing and language change . International Journal of Multilingualism 3 : 31 51. Inouye, Susan B. 1995. Trills, taps and stops in contrast and variation. PhD d issertation. Los Angeles, California : Unive rsity of California . Irwin, Patricia and Naomi Nagy. 2007. Bostonians /r/ speaking: A quantitative look at (R) in Boston. University of Pennsylvania W orking P apers in Linguistics 1 3(2): 135 147. Iwasaki, Shoichi and Preeya Ingkaphirom. 2005. A R eference G rammar of Thai . Cambridge: Cambridge University Press. Jones, Megan and Sali A. Tagliamonte. 2004. F rom Somerset to SamanÃ¡: Pre verbal did in the voyage of English. Language Variation and Change 16(2): 93 126. Kay, Paul. 1975. Synchronic variability and diachronic change in basic color terms. Language in Society 4(3) : 257 270. Kobsirikarn, Rataya. 199 1. Variation of the high falling tone in Suphanburi Thai by some social variables. Bangkok : Chulalongkorn University. Labov, William. 1963. The social motivation of a sound change. Word 19: 273 309. Labov, William. 1965. Sta ges in the acquisition of Standard English. In Roger W. Shuy (ed.) Social D ialects and Language L earning . Proceedin gs of the Bloomington, Indiana C onference 1964. National Council of Teachers of English. Labov, William. 1966. The S ocial S tratification o f English in New York City . Washington DC : Center for Applied Linguistics. Labov, William. 1969. Contraction, deletion, and inherent variability of the English copula. Language 45(4): 715 762. Labov, William. 1972a. Sociolinguistic P atterns . Philadelph ia: University of Pennsylvania Press. Labov, William. 1972b. Some principles of linguistic methodology. Language in Society 1: 97 120. Labov, William. 1972c. Language in the Inner C ity: Studies in the Black English V ernacular . Philadelphia: University of Pennsylvania Press.
196 Labov, William. 1984. Field methods of project on linguistic change and variation. In John Baugh and Joel Sherzer (eds.) Language in U se: Readings in S ociolinguistics . Englewood Cliffs and New Jersey : Prentice Hall. 23 53 . Labov, William. 1990. The intersection of sex and social cl ass in the course of linguistic change. Language Variation and Change 2: 205 254. Labov, William. 1994. Principle of L inguistic C hange Volume 1 : Internal F actors . Cambridge and Oxford: Blackwell. Labo v, William. 2001. Principles of L inguistic C hange Volume 2 : Social F actors . Malden and Massachusetts : Blackwell. Labov, William. 2004. Quantitative an alysis of linguistic variation. In Ammon, Ulrich, Norbert Dittmar, Klaus J. Mattheier and Peter Trudgil l (eds.) HSK S ociolinguistics/ Sociolinguisik Vol umn 1. Berlin: de Gruyter. 6 21. Labov, William. 2006. The Social Stratification of English in New York City . Cambridge: Cambridge University Press. Ladefoged, Peter. 1971. Preliminaries to Li nguistic P ho netics . Chicago: University of Chicago Press. Ladefoged, Peter and Ian Maddieson. 1996. The S ounds of the W L anguages. Cambridge and Massachusetts : Blackwell. Lavandera, Beatriz R. 1978. Where does the sociolinguistic variable stop? Language in Society 7 : 171 182. Lehmann, Winfred P. and Yakov Malkiel. 1968. Directions for H istorical L inguistics. Austin: University of Texas Press. Lenneberg, Eric Heinz. 1967. Biological F oundations of L anguage. New York: John Wiley and S on. Levy, Joshua. 2 010. The social stratification of (r) in Boston. Toronto Working Papers in Linguistics 33 : 1 10. Lindau, Mona. 1978 . Vowel features. Language 54(3): 541 563. Lindau , Mona. 1985. The story of /r/. In Vict oria A. Fromkin (ed.) Phonetic L inguistic: E ssays in H onor of Peter Ladefoged . Orlanda: Academic Press. 157 168.
197 Llamas, Carmen. 2001. The sociolinguistic profiling of (r) in Middlesbrough English. In Hans Van de Velde and Roeland van Hout (eds.) atics: Sociolinguistic, P honetic, and P honological C haracteristics of /r/ . Etudes and Travaux. Institut des Langues Vivantes et de PhonÃ©tique : UniversitÃ© Libre de Bruxelles. 123 139. Macaulay, Ronald K. S. 1976. Social class and language in Glasgow. Language in Society 5 : 173 188. Maddieson, Ian. 1984 . Patterns of S ounds . Cambridge: Cambridge University Press. Mather, Patrick department store study revisited. Journal of English Linguistics 40(4) : 338 356. Matsuda, Kenjiro. 1995 . Variable zero marking of (o) in Tokyo Japanese. PhD dissertation . Philadelphia : University of Pennsylvania. Mesthrie, Rajend, Joan Swann, Andrea Deumert and William L. Leap. 2000. Introducing S ociolinguistics. Edinburgh: Edinburgh University Press . Me yerhoff, Miriam. 2006. Introducing S ociolinguistics . New York: Routledge. Milroy, James. 1980 . Lexical alt ernation and the history of English: Evidence from an urban vernacular. In Elizabeth C. T raugott , Rebecca Labrum , and Susan C. Shepherd (eds.) Paper from the 4th I nternational C onference on Historical Linguistics . Amsterdam: John Benjamins. 355 362. Milroy, James and Lesley Milroy. 1978. Belfast; change and variation in an urban vern acular. In Peter Trudgill (ed.) Sociolinguistic P atterns in British English . London: Arnold. 19 36. Milroy, James and Lesley Milroy. 1997. Varieties and vari ation. In Florian Coulmas (ed.) The H andbook of S ociolinguistics . Cambridge: Blackwell. 47 64. Milroy, Lesley and Matthew J. Gordon. 2003. Sociolinguistics: Method and I nterpretation . Malden and Massachusetts : Blackwell. Myhill, John 1995. Change and continuity in the functions of the American English modals. Linguistics 33: 147 211. Myhill, John 1996. The development of the American English strong obligation s ystem. American Speech 71: 339 388. Nadasdi, Terry. 2001. Agreeing To Disagree: Variable Subject Verb Agreement in Immersion French. Canadian Journal of Applied Linguistics 4: 87 101.
198 Nichols, Patricia C. 1998. Black women in the rural south: conserva tive and innovative. In Jennifer Coates (ed.) Language and G ender: A Reader. Malden , Massachusetts : Blackwell. 55 63. stman, Jan Ola. 1981 . You know: A D iscourse F unctional A pproach . Amsterdam: John Benjamins. Palakornkul, Angkab . 1972. A sociolinguis tic study of pronominal strategy in Spoken Bangkok Thai. PhD dissertation . Austin, Texas : University of Texas. Panyaathisin, Kosin. 2013. The study of variable (r) in Bangkok Thai consonant clusters by Teochew Thais and native Thais in Hao Lam Phong Ch inese community of Bangkok, Thailand. Essex Graduate Student Pape rs in Language and Linguistics On L ine 14 : 55 77. Paolillo, John C. 2002. Analyzing Linguistic Variation: Statistical Models and Methods . Stanford. Center for the Study of Language and Information . Pappas, Panayiotis A. 2008. Object clitic placement in Cypriot Gree k: Results from a variationist analysis. Paper presented at 38th New Ways of Analyzing Variation , Rice University, Houston, Texas . Phootirat, Parichart. 2012. Register variation in Thai English interphonology: The contrast of /r/ and /l/. PhD dissertati on . Milwaukee , Wisconsin : The University of Wisconsin Milwaukee . Poplack, Shana. 1989. The care and handling of a mega corpus: The Ottawa Hull French project. In Ralph W. Fasold and Deborah Schiffrin (eds.) Language Change and Variation. Philadelphia: Jo hn Benjamins. 411 451. Poplack, Shana and Verbal s inflection in early Black English. Language Variation and Change 1(1): 47 84. Poplack, Shana and Sali A. Tagliamonte. 2001. African America n English in the Diaspora: Tense and A spect . Malden: Blackwell. Prasithrathsint, Amara. 1988 . Sociolinguistic research on Thailand languages. Language Sciences 10(2) : 263 272. Prasithrathsint, Amara. 1989 . Definitions of T erms in S ociolinguistics . Ban gkok: Chulalongkorn University Press.
199 Prasithrathsint, Amara. 1996. Stylistic Differentiation of /kaan/ and /khwaam/ Nominalization in Standard Thai. Proceedings of International Symposium on Language and Linguistics 4 . Institute of Language and Cult ure for Rural Development , Mahidol University . 1206 1216 . Premsrirat, Suwilai. 1995. Phonetic variation of final trill and final palatals in Khmer dialects of Thailand. Mon Khmer Studies Journal 24 : 1 26. Pulsup, Suthida.1993. Stylistic variation of (r) and (I) in Bangkok Thai consonant clusters among mathayom IV girl students at Sriayudhaya School. Unpublished thesis . Bangkok : Chulalongkorn University. Reid, Euan. 1978. Social and stylistic variation in the speech of children: some evidence from Edi nburgh. In Peter Trudgill (ed.) Sociolinguistic P atterns in British English . London: Edward Arnold. 158 171. Rissel, Dorothy A. 1989. Sex, attitudes, and the assibilation of /r/ among young people in San Luis PotosÃ, Mexico. Language Variation a nd Change 1: 269 283. Romaine, Suzanne. 1978. Postvocalic /r/ in Scottish English: sound change in progress? In Peter Trudgill (ed.) Sociolinguistic P atterns in British English . London. Edward Arnold. 144 157. Romaine, Suzanne. 1980. Stylistic variation and evaluative reactions to speech: problems in the investigation of linguistic attitudes in Scotland. Language and Speech 23(3): 213 232. Romaine, Suzanne. 1981. On the probl a reply to Beatrice Lavandera and William Labov. Wo rking Papers in Sociolinguistics 82 . Austin, Texas: Southwest Educational Developmental Laboratory. Romaine, Suzanne (ed.). 1982. Sociolinguistic V ariation in S peech C ommunities . London: Edward Arnold. Ronnakiat , Nantana . 1997. The Thai writing System. Last accessed 25 May 2014 at http://thaiarc.tu.ac.th/thai/thaiwrt.htm Sankoff, David. 1988a. Sociolinguistics and syntactic variation. In Frederick J. Newmeyer (ed.) Linguistics: The Cambridge S urvey V olume 4 , Language: The S ocio C ultural C ontex t. Cambridge: Cambridge University Press. 140 161. Sankoff, David. 1988b. Variable rules. In Ulrich Ammon, Norbert Ditt mar and Klaus J. Mattheier (eds.) Sociolinguistics: An I nternational H andbook of the S cience of L anguage and S ociety 2 . Berlin: Mouton de Gruyter. 984 997.
200 Sankoff, David and Gillian Sankoff. 1973. Sample survey methods and computer assisted analysis in the study of grammatical va riation. In Regna Darnell (ed.) Canadian L anguages in th eir S ocial C ontext . Edmonton, Alberta: Linguistic Research. 7 64. Sankoff, David, Gillian Sankoff, Suzanne Laberge and Marjorie Topham. 1976. variation grammaticale. Cahiers d e L 6: 85 125. Sankoff, David and Suzanne Laberge. 1978. The linguistic market and the statistical explanation of variability. In David Sankoff (ed.) Linguistic V ariation: Models and M ethods . New York: Academic Press. 239 250. Sankoff, David and Pierrette Thibault. 1981. Weak complementarity: tense and aspect in Montreal French. In Brenda Johns and David R. Strong (eds.) Syntactic Change 25 . Ann Arbor: University of Michigan. 205 216. Sankoff, David, Sali A. Tagliam onte and Eric Smith. 2005. Goldvarb X: A V ariable R ule A pplication for Macintosh and Windows . Toronto, Ontario: University of Toronto. Sankoff, Gillian. 2006. Apparent time and real time. In Keith Brown (ed.) Elsevier Encyclopedia of Language and Lingui stics 2 . Amsterdam: Elsevier. 110 116. Sankoff, Gillian and HÃ©lÃ¨ne Blondeau. 2007 . Language change across the lifespan: /r/ in Montreal French. Language 83(3): 560 588. Sankoff, Gillian and HÃ©lÃ¨ne Blondeau. 2013 R] alternatio n in Montreal French: An exploration of stylistic conditioning in a sound change in progress. In Lorenzo Spreafico and Alessandro Vietti (eds.) Rhotics, N ew D ata, and P erspectives . Bolzano: Bozen Bolzano University Press. 249 265. Sapir, Edward. 1921. La nguage: An introduction to the S tudy of S peech . New York: Harcourt, Brace. Sapproong, Paleerat. 1993. Social variation of ( ) in the urban Songkhla dialect. thesis . Bangkok : Chulalongkorn University. Scobbie, James M. 2006. (R) as a variable. In Keith Brown (ed.) Encyclopedia of Language and Linguistics . Oxford: Elsevier. 337 344. Slayden, Glenn. 2009. Central Thai phonology. Last accessed 5 March 2014 at http://www.thai language.com/resources/slayden thai phonology.pdf Slayden, Glenn. 2010. Overview of the Thai language. Last accessed 5 March 2014 at http://www.thai language.com/ref/Overvie w
201 Smalley, William A.1994. Linguistic D iversity and N ational U nity: Language E cology in Thailand . Chicago: University of Chicago Press. Smith, Jennifer. 2001. You Ã¸ : Negative do in Buckie Scots. English World Wide 21(2) : 231 259. Stuart Smith, Jane. 2003. The phonology of mo dern urban Scots. In John Corbett, Derrick McClure and Jane Stuart Smith (eds.) The Edinburg C ompanion to Scots . Edinburg: Edinburg University Press. 110 137. Tagliamonte, Sali A. 2005. So who? Like how? Just what? Discourse markers in the conversations of young Canadians. Journal of Pragmatics 37 (special issue): 1896 1915. Tagliamonte, Sali A. 2007. Quantitative analysis. In Robert Bayley and Ceil Lucas (eds.) Sociolinguistic V ariation: Theories, M ethods and A pplications . New York: Cambridge Universi ty Press. 190 214. Tagliamonte, Sali A. 2009. Analysing S ociolinguistic V ariation . New York: Cambridge University Press. Tagliamonte, Sali A. 2012. Variationist Sociolinguistics: Change, O bservation, I nterpretation . Oxford: Wiley Blackwell. Tagliamon te, Sali A. and Rachel Hudson. 1999. Be like et al. Beyond America: The quotative system in British and Canadian youth. Journal of Sociolinguistics 3(2) : 147 172. Tagliamonte, Sali A. and Rosalind Temple. 2005. (t,d) in British English . Language Variation and Change 17: 281 302 . Tanwattananun, Penporn. 1982. The change of unaspirated stops in Chiengmai dialect and some social variables . thesis . Bangkok : Chulalongkorn University. Tingsabadh, M.R . Kalaya and Arthur S. Abramson. 1993. Illustrations of the IPA: Thai. Journal of the International Phonetic Association 23(1) : 24 28. Treyakul, Saengchant. 1986. the pronunciation of Bangk ok F.M. radio newscasters. thesis Bangkok : Chulalongkorn University. Trudgill, Peter. 1972. Sex, covert prestige and linguistic change in the urban British English of Norwich. Language in Society 1(2): 179 195.
202 Trudgill, Peter 1974 . The S ocial D ifferentiation of English in Norwich . Cambridge: Cambridge University Press. Trudgill, Peter 1978. Sociolinguistic P atterns in British English . London: Edward Arnold. Trudgill, Peter 1983. Sociolinguistics: An introduction to L anguage and S ociety . Harmondsworth: Penguin. Trudgill, Peter. 1998. Sex and covert pre stige. In Jennifer Coates (ed.) Language and G ender: A R ea der. Malden and Massachusetts : Blackwell. 21 28. Verstraeten, Bart and Hans Van de Velde. 2001. Socio geographical vari ation of /r/ in standard Dutch. In Hans Van de Velde and Roeland van Hout (eds.) atics: Sociolinguistic, P honetic, and P honological C haracteristics o f /r/. Etudes and Travaux. Institut des Langues Vivantes et de PhonÃ©tique : UniversitÃ© Libre de Bruxelle s. 45 61. Walker, James A. 2010. Variation in Linguistic S ystems . New York: Routledge. Wardhaugh, Ronald. 2006. An introduction to S ociolinguistics . Malden and Oxford: Blackwell . Weiner, Judith and William Labov. 1983. Constraints on the agentless pass ive. Journal of Linguistics 19(1): 29 58. Weinreich, Uriel, William Labov and Marvin Herzog. 1968. Empirical foundations for a theory of language change. In Winfred P. Lehmann and Yakov Malkiel (eds.) Directions for Historical L inguistics . Austin: Unive rsity of Texas Press. 95 188. Weise, Richard. 2001. The phonology of /r/. In T Alan Hall (ed.) Distinctive Feature Theory . Berlin and New York: Mouton de Gruyter. 335 368. Wolfram, Walt. 1969. A S ociolinguistic D escription of Detroit N egro S peech . Washi ngton DC: Center for Applied Linguistics. Wolfram, Walt. 2010. Fieldwork methods in language variation. In Ruth Wodak, Barbara Johnstone and Paul Kerswill (eds.) The SAGE H andbook of S ociolinguistics . London: SAGE Publications. 296 311. Wolfram, Walt an d Ralph W. Fasold. 1974. The S tudy of S ocial D ialects in American English . Englewood Cliffs, New Jersey : Prentice Hall . Wolfram, Walt and Natalie Schilling Estes. 1998. American English . Malden and Massachusetts : Blackwell .
203 Wolfson, Nessa. 1976. Speec h events and natural speech: Some implications for sociolinguistic methodology. Language in Society 5: 189 209. Yaeger Dror, Malcah. 1994. Sound change in Montreal French. In Patricia Keating (ed.) Phonological S tructur e and Phonetic Form. Papers in La boratory P honology 3. Cambridge: Cambridge University Press. 267 292. Young, Richard and Robert Bayley.1996. VARBRUL analysis for second language acquisition research. In Robert Bayley and Dennis R. Preston (eds.) Second L anguage A cquisition and L inguist ic V ariation . Amsterdam: John Benjamins. 253 306.
204 BIOGRAPHICAL SKETCH Orapat Pookkawes was born and raised in Bangkok, Thailand . She received her in English at Srinakharinwirot University , Bangkok , in 2001 . After that, s he graduated fr om Chulalongkorn University , Bangkok, with an M.A. in linguistics in 2003 . She earned her Ph.D. in linguistics from the University of Florida in the summ er 2014.