Citation
The Social Motivation of Retroflex Variation of Taiwan Mandarin in an Immigrant Setting

Material Information

Title:
The Social Motivation of Retroflex Variation of Taiwan Mandarin in an Immigrant Setting
Creator:
Lai, Yu Ning
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (195 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Linguistics
Committee Chair:
BLONDEAU,HELENE
Committee Co-Chair:
BOXER,DIANA
Committee Members:
MCLAUGHLIN,FIONA
FANG,ZHIHUI
Graduation Date:
8/6/2016

Subjects

Subjects / Keywords:
Dialects ( jstor )
Fricative consonants ( jstor )
Interviews ( jstor )
Language ( jstor )
Length of stay ( jstor )
Linguistics ( jstor )
Mandarins ( jstor )
Regional dialects ( jstor )
Social networking ( jstor )
Spoken communication ( jstor )
Linguistics -- Dissertations, Academic -- UF
chinese -- mandarin -- retroflex -- sociolinguistics -- taiwan -- variation
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Linguistics thesis, Ph.D.

Notes

Abstract:
This dissertation examines the social motivation for Taiwan Mandarin (hereafter TM) speakers choice of variation of the retroflex initials (zh) (ch) (sh) (r) when they are in an immigrant setting. It further investigates TM speakers linguistic behaviors when they come into contact with Putonghua (hereafter PTH) speakers in an immigrant community where there is no dominant dialect when TM and PTH speakers value their Mandarin as a part of their Taiwanese and Chinese identity respectively in their homeland. The results indicate that among all variants of the retroflex initials, the nonstandard palato-alveolar and the lateral are dominant in TM speakers speech. Multivariate analysis reveals that gender displays the strongest magnitude of effect affecting TM speakers linguistic choice of the variants of the retroflex initials. Female and male speakers preference for standard and nonstandard variants respectively suggests that Taiwanese men and womens language use is constrained by traditional gender role inside Taiwanese society. The preference for the nonstandard variant of the retroflex initials by speakers who have greater strength of network with PTH speakers was unexpected. This tendency suggests that their frequent contact with PTH speakers contributes to their sensitive to the linguistic differences between PTH and TM, and their sensitivity further determines their choice of the variants. In addition, speakers who have longer residence time in the US tend to favor the standard variant of the retroflex initials, which suggests that when an immigrant community values Chinese culture as the social standards, speakers choice of the variants would be affected regardless of their places of origin. The results also show the language use between generations reflects different stages of sociopolitical changes in Taiwanese society. The comparative results from the two types of interviews further reveal that the particular sociocultural ideologies and function roles behind the variants of the retroflex initials are assigned by TM speakers. These meanings and ideologies emerge during the group interview, and TM speakers may consciously or subconsciously apply them when coming in contact with PTH speakers in the immigrant community. The findings evidence that socio-psychological factor also plays an influential role affecting speakers linguistic choice. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2016.
Local:
Adviser: BLONDEAU,HELENE.
Local:
Co-adviser: BOXER,DIANA.
Statement of Responsibility:
by Yu Ning Lai.

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright Lai, Yu Ning. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Classification:
LD1780 2016 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

THE SOCIAL MOTIVATION OF RETROFLEX VARIA TION OF TAIWAN MANDA RIN IN AN IMMIGRANT SETTING By YU NING LAI 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 2016

PAGE 2

2016 Yu Ning Lai

PAGE 3

To my parents, for their love and support

PAGE 4

4 ACKNOWLEDGMENTS This dissertation would have been difficult to complete without support and help from many people. First of all, I would like to express my d eepest gratitude to my advisor, Dr. Hlne Blondeau, who has assisted me in coordinating this dissertation with her insightful comments and suggestions . With her invaluable guidance and encouragement throughout the researc h for and writing of this dissertation, I was able to accomplish it . I am also very grateful to my committee members Dr. Diana Boxer, Dr. Fiona McLaughlin, and Dr. Zhihui Fang , for their invaluable suggestions and support. I would like to thank all of the Taiwanese and Chinese participants in my dissertation research. Without their volunt eering , this research would not have been possible. I appreciate every friend who assist ed me in recruiting participants in Gainesville . I also would l ike to thank the faculty and staff of the Department of Linguistics and Chinese program at the University of Florida (UF) f or their help and encouragement through my academic st udies and in writing the dissertation . I sincerely thank my dearest friends in Taiwan and in the US for their genuine friendship and the support throughout these years of my studies. I am especially thank ful to Eunha Hwang and Orapat Pookkawes for their academic and emotional support at UF. Finally, I would like to express my sincere gratitude to my parents, Wen Pin Lai and SuChiu Hs iao , for their unconditional love and tremendous encouragement which have provided me with comfort and warmth through this long journey of academic striving and during the difficult times.

PAGE 5

5 TA BLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 8 LIST OF FIGURES ........................................................................................................ 10 ABSTRACT ................................................................................................................... 12 CHAPTER 1 INTRODUCTION .................................................................................................... 14 Sociolinguistic Background in Taiwan ..................................................................... 17 Language Ideology in Taiwan Mandarin ................................................................. 20 Research Questions ............................................................................................... 22 2 HISTORICAL AND LINGUISTIC BACKGROUND OF STANDARD MANDARIN, PUTONGHUA, TAIWAN MANDARIN ..................................................................... 25 Mandarin ................................................................................................................. 25 Linguistic F eatures in Standard Chinese ................................................................ 27 Mandarin in China: Putonghua (PTH) ..................................................................... 30 Mandarin in Taiwan: Taiwan Mandarin (TM) ........................................................... 31 Variants of Retroflex Initials in Taiwan Mandarin .................................................... 36 3 LITERATURE REVIEW .......................................................................................... 42 Dialect in Contact: Convergence and Divergence .................................................. 42 Research in Taiwan Mandarin ................................................................................ 51 Theoretical Framework ........................................................................................... 57 Variationist Sociolinguistics Approach .............................................................. 57 Order of Indexicality ......................................................................................... 60 Accomm odation Theory .................................................................................... 65 4 STUDY DESIGN ..................................................................................................... 69 The Linguistic Variables .......................................................................................... 69 Part icipants ............................................................................................................. 69 Criteria for Participants ..................................................................................... 70 Individual interview ..................................................................................... 70 Group interview .......................................................................................... 72 Recruitment of Participants .............................................................................. 72 Individual interview ..................................................................................... 72

PAGE 6

6 Group interview .......................................................................................... 72 Data Collect ion ....................................................................................................... 73 Data Analysis ................................................................................................... 75 Data Transcription ............................................................................................ 75 Data Extra ction and Coding .............................................................................. 75 The Extralinguistic Variables ................................................................................... 77 Generation ........................................................................................................ 77 Gender ............................................................................................................. 79 Social Network Strength ................................................................................... 80 Leng th of Stay in the US ................................................................................... 81 Statistical Methods .................................................................................................. 82 5 RESEARCH RESULTS AND INTERPRETATIONS ............................................... 86 Variation in Retroflex Initials (zh) (ch) (sh) .............................................................. 86 Overall Distribution of Retroflex Initials Variants ..................................................... 87 Correlation between Retroflex (zh) (ch) (sh) and the Extralinguistic Factors .......... 8 8 Strength of Social Network ............................................................................... 88 Generation ........................................................................................................ 91 Gender ............................................................................................................. 92 Length of Stay in the US ................................................................................... 95 Mult ivariate Regression Analysis ............................................................................ 98 Gender ............................................................................................................. 99 Generation ........................................................................................................ 99 Length of Stay in the US ................................................................................... 99 Strength of Social Network ............................................................................. 100 Group Interview .................................................................................................... 100 Distribution of Retroflex (zh) (ch) (sh) and the Extralinguistic Factors .................. 102 Generation ...................................................................................................... 102 Gender ........................................................................................................... 103 Strength of Social N etwork ............................................................................. 104 Length of Stay in the US ................................................................................. 106 6 RESEARCH RESULTS AND INTERPRETATIONS ............................................. 118 Variation in Retroflex Fricative Initial (r) ................................................................ 118 Overall Distribution of Retroflex Fricative (r) Variants ........................................... 118 Correlation between Retroflex Fricative (r) and the Extralinguistic Factors ........... 119 Strength of Social Network ............................................................................. 119 Generation ...................................................................................................... 120 Gender ........................................................................................................... 123 Length of Stay in the US ................................................................................. 124 Mu ltivariate Regression Analysis .......................................................................... 126 Gender ........................................................................................................... 127 Length of Stay in the US ................................................................................. 127 Ge neration ...................................................................................................... 128 Strength of Social Network ............................................................................. 128

PAGE 7

7 Group Interview .................................................................................................... 129 Distribution of Retroflex Fricative (r) and the Extralinguistic Factors ..................... 130 Generation ...................................................................................................... 130 Gender ........................................................................................................... 131 Strength of Social Network ............................................................................. 132 Length of Stay in the US ................................................................................. 134 7 DISCUSSION ....................................................................................................... 144 The Retroflex Initials (zh) (ch) (sh) ........................................................................ 144 The Retroflex Fricative (r) ..................................................................................... 149 Sociolinguistic Ideologies of Retroflex Initials (zh) (ch) (sh) (r) ............................. 153 8 CONCLUSION ...................................................................................................... 166 Revisiting Research Questions ............................................................................. 166 Broader Implications ............................................................................................. 171 Limitations of the Study ......................................................................................... 173 Suggestions for Further Research ........................................................................ 176 Closing .................................................................................................................. 177 APPENDIX A INFORMED CONSENT FORMS .......................................................................... 179 B QUESTIONNAIRE ................................................................................................ 182 LIST OF REFE RENCES ............................................................................................. 186 BIOGRAPHICAL SKETCH .......................................................................................... 195

PAGE 8

8 LIST OF TABLES Table page 2 1 Phonological differences between PTH and TM ................................................. 40 2 2 Syntactic differences between PTH and TM ....................................................... 40 2 3 Lexicon difference between PTH and TM ........................................................... 40 4 1 Participants’ demographics in individual interviews ............................................ 85 4 2 TM Participants’ demographics in group interviews ............................................ 85 5 1 Overall distribution of the variants of retroflex (zh) (ch) (sh) ............................. 111 5 2 Overall distribution of the variants of retroflex (zh) (ch) (sh) ............................. 111 5 3 Overall distribution of retroflex (zh) (ch) (sh) by strength of social network ...... 111 5 4 Distribution of retroflex (zh) (ch) (sh) by age ..................................................... 111 5 5 Distribution of retroflex (zh) (ch) (sh) by gender ............................................... 111 5 6 Distribution of retroflex (zh) (ch) (sh) by length of stay in the US ...................... 112 5 7 Multivariate analyses of the contribution of extralinguistic factors selected as significant to the probability of the retroflex (zh) (ch) (sh) ................................. 112 5 8 Overall distribution of retroflex (zh) (ch) (sh) by type of interview ..................... 113 5 9 Distribution of retroflex (zh) (ch) (sh) by age and type of interview ................... 113 5 10 Distribution of retroflex (zh) (ch) (sh) by gender and type of interview .............. 113 5 11 Distribution of retroflex (zh) (ch) (sh) by strength of social network and type of interview ........................................................................................................... 113 5 12 Distribution of retroflex (zh) (ch) (sh) by length of stay in the US and type of interview ........................................................................................................... 113 6 1 Overall distribution of the variants of retroflex fricative ( r ) ................................. 137 6 2 Overall distribution of [ ] and [l] by strength of social network .......................... 137 6 3 Overall distribution of [ ] and [l] by age ............................................................. 137 6 4 Overall distribution of [ ] and [l] by gender ....................................................... 137

PAGE 9

9 6 5 Overall distribution of [ ] and [l] by length of stay in the US .............................. 138 6 6 Multivariate analy ses of the contribution of extralinguistic factors selected as significant to the probability of the retroflex fricative (r) ..................................... 138 6 7 Overall distribution of retroflex fricative (r) by type of interview ......................... 139 6 8 Distribution of retroflex fricative (r) by age and type of interview ...................... 139 6 9 Distribution of retroflex fricative (r) by gender and type of interview ................. 139 6 10 Distribution of retroflex fricative (r) by strength of social network and type of interview ........................................................................................................... 139 6 11 Distribution of retroflex fricative (r) by length of stay in the US and type of interview ........................................................................................................... 139

PAGE 10

10 LIST OF FIGURES Figure page 2 1 Language distribution in Taiwan ......................................................................... 41 5 1 Overall distribution of retroflex (zh) (ch) (sh) by strength of social network ...... 114 5 2 Distribution of retroflex (zh) (ch) (sh) by age ..................................................... 114 5 3 Distribution of retroflex (zh) (ch) (sh) by gender ............................................... 114 5 4 Distribution of retroflex (zh) (ch) (s h) by length of stay in the US ...................... 115 5 5 Effect of gender on the production of full retroflex variants [ ], [h], [ ] ........... 115 5 6 Effect of age on the production of full retroflex variants [ ], [h], [] ................ 115 5 7 Effect of length of stay in the US on the production of full retroflex variants [ ], [h], [] ....................................................................................................... 116 5 9 Distribution of the variants of retroflex (zh) (ch) (sh) by generation and type of interview ........................................................................................................... 116 5 10 Distribution of the variants of retroflex (zh) (ch) (sh) by gender and type of interview ........................................................................................................... 117 5 11 Distribution of the variants of retroflex (zh) (ch) (sh) by strength of social network and type of interview ........................................................................... 117 5 12 Distribution of the variants of retroflex (zh) (ch) (sh) by length of stay in the US and type of interview ................................................................................... 117 6 1 Overall distribution of [ ] and [l] by strength of social network .......................... 140 6 2 Overall distribution of [ ] and [l] by age ............................................................. 140 6 3 Overall distribution of [ ] and [l] by gender ....................................................... 140 6 4 Overall distribution of [ ] and [l] by length of stay in the US .............................. 141 6 5 Effect of gender on speakers’ production of the retroflex fricative [ ] ............... 141 6 6 Effect of length of stay on the production of retroflex fricative [ ] ...................... 141 6 7 Effect of age on the production of retroflex fricative [ ] ..................................... 142 6 8 Effect of strength of social network on speakers’ production of retroflex fricative [ ] ........................................................................................................ 142

PAGE 11

11 6 9 Distribution of variants of the retroflex fricative (r) by generation and type of interview ........................................................................................................... 142 6 10 Distribution of variants of the retroflex fricative (r) by gender and type of interview ........................................................................................................... 143 6 11 Distribution of variants of the retroflex fricative (r) by strength of social network and type of interview ........................................................................... 143 6 12 Distribution of variants of the retroflex fricative (r) by length of stay in the US and type of interview ......................................................................................... 143

PAGE 12

12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THE SOCIAL MOTIVATION OF RETROFLEX VARIATION OF TAIWAN MANDARIN IN AN IMMIGRANT SETTING By Yu Ning Lai August 2016 Chair: Hlne Blondeau Major: Linguistics This dissertation examines the social motivation for Taiwan Mandarin (hereafter TM) speakers’ choice of variation of the retroflex initials (zh) (ch) (sh) (r) when they are in an immigrant setting. It further investigates TM speakers’ linguistic behaviors when they come into contact with Putonghua (hereafter PTH) speakers in an immigrant community where there is no dominant dialect when TM and PTH speakers value their Mandarin as a part of their Taiwanese and Chinese identity respectively in their homeland. The results indicate that amo ng all variants of the retroflex initials , the nonstandard palatoalveolar and the lateral are dominant in TM speakers’ speech. Multivariate analysis reveals that gender displays the strongest magnitude of effect affecting TM speakers’ linguistic choice of the variants of the retroflex initials . Female and male speakers’ preference for standard and nonstandard variants respectively suggests that Taiwanese men and women’s language use is constrained by traditional gender role inside Taiwanese society. The preference for the nonstandard variant of the retroflex initials by speakers who have greater strength of network with PTH speakers

PAGE 13

13 was unexpected. This tendency suggests that their frequent contact with PTH speakers contributes to their sensitive to the linguistic differences betw een PTH and TM, and their sensitivity further determines their choice of the variants. In addition, speakers who have longer residence time in the US tend to favor the standard variant of the retroflex initials , which suggests that w hen an immigrant community values Chinese culture as the social standards, speakers’ choice of the variants would be affected regardless of their places of origin. The results also show t he language use between generations re flects different stages of sociopolitical changes in Taiwanese society. The comparati ve results from the two types of interviews further reveal that the particular sociocultural ideologies and function roles behind the variants of the retroflex initials ar e assigned by TM speakers. These meanings and ideologies emerge during the group interview, and TM speakers may consciously or subconsciously apply them when coming in contact with PTH speakers in the immigrant community. Th e findings evidence that socio psychological factor also play s an influential role affecting speakers’ linguistic choice .

PAGE 14

14 CHAPTER 1 INTRODUCTION In the era of globalization, people use different transportation tools to travel around the world. The high mobility increases their opportunities to interact with people from different language backgrounds. As Weinreich (1966) suggests, the occurrence of language contact is triggered by the contact between individuals who use more than one language in response to their immediate social contexts , and the contact consequently leads to linguistic changes. Furthermore, that wh ich increases complexity in language contact is the interaction between speakers’ language use and a variety of social factors . People’s mobility triggers not only contact between speakers with different language backgrounds, but also contact between dialectal speakers. Peter Tru dgill (1986) proposes that contact between speakers who use varieties of the same language results in linguistic changes. Trud gill argues that al though dialect al speakers are mutual ly intellig ib le to some degree, linguistic differences between each variety of the same language can be found due to contact with regional dialects. When such a contact situation occurs, the degree of mutual intelligibility is affected, and comprehension difficulties between dialectal speakers occur . Mandarin dialect s1 are no exception. Mandarin in China (i.e. Putonghua) and Taiwan (i.e. Taiwan Mandarin) is based on the pronunciation of the Beijing dialect and on the vocabularies from northern Mandarin. Due to contact with regional dialects and sociopolitical changes in the two 1 ‘Mandarin dialect ’ in this study refers to a variety of Mandarin used in China and Taiwan as the official language respectively. The term “regional/local dialect” refers to Chinese fangyan ‘dialect’ including Wu, Gan, Xiang, Hakka (or Ke), Yue (or Cantonese), Min, and Mandarin.

PAGE 15

15 Chinese2 speaking regions, Mandarin in these two places gradually deviate from each other. Furthermore , because of the linguistic differences such as tonal patterns and lexicon resulting from contacts with local dialects and foreign languages, it i s not uncommon that speakers from the two regions may encounter comprehension difficulties from time to time when they communicate with each other. While studying abroad, I have had the chance to meet Mandarin speakers from China . W hen I talk to them , at first, both of us encounter some difficult y understanding each other due to the lexical differences and different tonal patterns. Take the term ‘ to print out ’ as an example. ‘ T o print out ’ is read as l i y n in Taiwan Mandarin (hereafter TM) and yn in Putonghua (hereafter PTH) . Therefore, when one of my Chinese friend s first said it to me, I did not understand what she meant and I asked her to explain it to me. The same misunderstanding occurred with my Chinese friends when I used a term that was common in Taiwan, but with which they were not familiar . In present day, speakers from the two regions have access to multimedia res ources such as Youtube and various Chinese websites, and notice the linguistic differences in t he two Mandarin speaking regions . However, when they have chance to communicate with each other, they still experience comprehension di fficulties and miscommunication. Moreover, i n addition to miscommunication, when I talk to my Chinese friends, I find that I am slightly influenced by their way of speaking from time to time , particularly 2 ‘Chinese’ is an English term referring to a language which belongs to the Sinitic subgroups of SinoTibetan languages. The term ‘Chinese’ can be interpreted differently depending on the context when translated into Chinese words (Sun 2006 ). For example, it can refer to the language , the language spoken by ethnic Hn people i.e. , or different Chinese dialects . In addition, the term ‘Chinese’ can also refer to the standard dialect i.e. ‘common language’ used in the People’s Republic . The term ‘Chinese’ here is used as a neutral term to refer to any particular kind of Chinese ‘dialect’ as a single cultural entity.

PAGE 16

16 by the use of the retroflex feature which is absent in TM because of the influence of the local dialect. The retroflex feature is the standard and prescribed form in Mandarin, and is acquired as the institutional form in schools in Taiwan. However, it is seldom practiced in daily TM speakers’ speech (Chung 2006) . As a native speaker of Taiwanese and Mandarin, the use of retroflex is also seldom practiced in my daily speech. However, when I talk to my Chinese friends, I find myself modifying the use of the retroflex feature when it appears frequently in their speech. After talking to them, t he modification in my speech and my awareness of the linguistic differences led me to beco me interested in how TM s peakers use language when they are in a foreign country , and when they have the chance to be in contact with PTH speakers in an immigrant community in which the population comprises speakers from different language and cultural backgrounds . Building upon the concept of dialect s in contact , the present study examines initial consonant s: retroflex (zh), (ch), (sh), and retroflex fricative (r) used by TM speaker s in both an immigrant setting , where the major population consists of students who come to the US to pursue their degree in a local university , and in the local immigrant community by the permanent residents who immigrate from Taiwan for job needs and have resided in the community for a long time. In addition, TM speakers may not have many opportunities to come into contact with PTH speakers in their home country , so t his study is also interested in investigating TM speakers’ linguistic choice of retroflex variants when they have chances to come in contact with PTH speakers while away from their homeland. These variables will be examined in terms of their correlation with social factors in the speech of TM speakers.

PAGE 17

17 Sociolinguistic B ackground in Taiwan In Taiwan, l anguage has be en used as a political means to cater to different ethnic groups under different political regimes. The implement ation of language policies in different political hegemonies in Taiwan ’s history —i.e., the Dutch (16241662), the Chinese (around the mid17th century), and the Japanese (18951945) is further connected with the development of identity —in particular , ethnic group identity . For example, during the Japanese colonization period, Japanese was promoted as the main language and Chinese dialects were forbidden in both public and private domains . A wareness of a Taiwanese identity first emerged when a Japanese sovereign used the language ( Japanese) to segregate the colonizers and the colonized. After returning to a Chinese regime led by Chinese N ationalists —“Kuomintang” (hereafter KMT ) —in 1945, Mandarin, the language associated with Chinese ideology, was promoted as the national language to e stablish full control over the different ethnic groups , and also to emphasize Chinese culture and identity. Local dialects such as Taiwanese, Hakka, and aboriginal languages were strictly forbidden in public domains such as educational institutions, go vernment, and media until the lift of martial law in 1987. The implement ation of Mandarinonly since 1945 had a significant impact on language shift toward Mandarin across generations . Sandel (2006) and his associates found different language attitudes toward their language choice of Mandarin and their mother tongue, Taiwanese, between three generations of the family. The researchers revealed that within three generations of the family, the young and the middleaged generation evidently tend to shift from their mother tongue, Taiwanese, to Mandarin. In particular, the middleaged generation was more varied in terms of their language choice of

PAGE 18

18 Mandarin and mother tongue, i.e. their language choice depended on which generat ion they interacted with. Furthermore, the promotion of Mandarin as the only national language resulted in diglossia in Taiwanese society ( Taso 1999) . Speaking standard Mandarin was indexed as the Mainlander3 identity and was associated with higher social status because the Mainlander identity represented a ruling class in Taiwanese society . On the other hand, speaking nonstandard Mandarin —i.e. Taiwaneseaccented Mandarin—or local dialects —i.e. Taiwanese—was indexed as Taiwanese ethnic identity and was linked with lower social status and a lower class (Liao 2008; Taso 1999; Tse 2000). The effect of diglossia still influence s TM speakers’ language attitudes d espite the fact that since 1987, local dialects a re no longer forbidden in public occasions . Mandarin still serves as the main instruction language in the educational system, and functions as a lingua franca among different ethnic groups ( Tse 2000 ) . In addition, Lee (1981) finds t hat Taiwanese people’s attitude toward Mandarin and local dialects is conflicted because the functional role of the two languages was different. Mandarin served the function of “national unity” ( 1981:124) , but at the same time, using Taiwanese or other local dialects could reinforce ethnic root s; thus the language policy made it impossible to eradicate the local dialects . The language attitude toward Mandarin, Taiwanese, and Taiwaneseaccented Mandarin mirrors the complex sociolinguistic situation in Taiwanese society. 3 Mainlander w ish ngr n or ‘ external province people’ refer s to Chinese immigrants who migrated to Taiwan with KMT after they were defeated by Chinese Communist s after the Chinese Civil War in 1945.

PAGE 19

19 KMT promoted a Mandarinonly policy to establish their political control and further push their China centered ideology to the people on the island. In the late 1980s, the liberalization of language policy and increasing voices for the support of local languages, culture, and identity led to the rise of localism and the promotion of local Taiwanese identity. However, the group identity of “being Chinese” versus “being Taiwanese” is debate d to this day . Native Taiwanese speakers comprise the largest population , but Taiwanese had not gained popular attention in the mainstream educational system until a new curriculum standard was implemented in 1993. The new curriculum emphasizes that all languages and dialects used in Taiwan are equal and all ethnic groups may be considered as “ Taiwanese” because people share the same historical experiences on the island. S hared affection for the history and social events on the island where people have lived for decades redefine s the meaning of “ be ing Taiwanese”. The new re defined identity of “ being Taiwanese” is not only linked with the place, but also serves as a symbol of localness and solidarity (Brubaker 2012; Tse 2000) . To the present day, language ideology plays a strong role in defining group and national identity inside Taiwanese society. A sensitive political climate regarding Taiwancentered identity versus Chinacentered identity is still in effect inside Taiwanese society due to the historical attachment to Chinese culture and identity, and complicated political relation s with Mainland China. TM gradually deviates from the standard Mandarin, which is based on Beijing pronunciation and is practiced in PTH in China. The most salient fea ture in TM is the lack of the full retroflex initials4 [ ], [h], [] , [ ] because these features are absent in 4 A Chinese syllable structure is divided into two parts: an initial i.e. an initial consonant and a final. The initial consonant is optional in a syllable. The final includes a medial vowel, a main vowel, and a final

PAGE 20

20 Taiwanese phonology. Although the full retroflex initials are taught in schools in Taiwan, they are seldom practiced in TM speakers’ everyday speech. Previous studies have pointed out that lack of the full retroflex initials is a distinctive feature in TM, and has been considered as nonstandard form and associated with negative connotations such as low prestige, unsophisti cation, and a peasant accent (Bara n 2014; Kubler 1985; Su 2008). However, recent studies show that the negative connotation of the non standard form is gradually diminish ing . The nonstandard form is widely accepted by people in Taiwan as a signification of localness. On the other hand, th e use of the full retroflex feature—the standard form —is linked with a negative evaluation as “off putting, affected and show off” (Chung 2006: 202). In Brubaker’s (2012: 133) study, his TM interviewees comment that the use of the full retroflex feature is “too formal”, “creates distance”, and has “a strong sense of otherness”. Some further in dicate that it is the mainland accent or Beijing accent which is di fferent from the Mandarin spoken in Taiwan. T he full retroflex feature is further associated with its geographic origin, Beijing. As Gal and Irvine (1995: 973) argue , “ linguistic features are seen as reflecting and expressing broader cultural images of p eople and activities”. Thus, when people in Taiwan detect the full retroflex feature, they connect it with social meaning s that are indexed by the people inside the society as well as its geographic origin. Language I deology in Taiwan Mandarin Trudgill (1986) notes that in contact situation s, speakers modify their language use to accommodate their interlocutors by eliminating or reducing the frequency of consonant (Chao 1968). In this study, t he term “initial” is used to refer to an initial consonant in the beginning of the syllable. Please see Chapter 2 for a detailed discussion.

PAGE 21

21 salient dialect features. And ultimately, the elimination and reduction leads to linguis tic convergence such as dialect leveling . However, convergence is not the only outcome when dialects come in contact. Schilling Estes’ (2002) study investigating two isolated communities in North Carolina finds that social factors such as community members ’ strong community identi fication , strong awareness of cultural distinctiveness, and attitudinal factors lead speakers to reinforce their salient dialectal features to retain their group identity. The strong attitude towards the community identity and str ong sense of cultural distinctiveness resist the occurrence of dialect leveling and trigger linguistic divergence. The findings from Schilling Estes’ study raise questions among scholars (Gal and Irvine 1995; Johnstone 2004, 2010) , such as how local linguistic features are linked with local identities , and whether there is a correlation between the use of particular local forms and speakers’ social identities . The formation of Taiwan Mandarin (TM) may provide an answer to these questions. TM is the linguistic outcome of the contact between Southern Min (hereafter Taiwanese5) and Mandarin based on Beijing pronunciations. T he influence of Taiwanese on TM is shown in phonology, lexicon, and syntax , such as the lack of retroflex features as initial consonants and lexicon borrowing . In addition, the influence of other languages such as Japanese and English , a number of Austronesian languages and Chinese dialect s—i.e., Hakka —is also found in TM (Kubler 1985) . These influences cause TM to drift away from the standard Mandarin and Mandarin in China. 5 Taiwanese is a subgroup of Southern Min ( cf . Northern Min), and very similar to Amoy. Taiwanese is a mixtu re of Zhangzhou and Quanzhou speech , which are the variants of Southern Min. P honological features of Southern Min remain , such as extensive tone sandhi, lack of t palatalization, and voiced obstruents. Frequent contact between the two speeches in Taiwan results in a mixture of them . This mixture has come to be considered as Taiwanese Southern Min or Taiwanese (Hong 1994; Hong 2010 as cited in Ratte 2011; Lai 2006).

PAGE 22

22 The linguistic features in TM lead several scholars to claim that TM may be associated with the identity of being local Taiw anese across the island because the particular linguistic features in TM are linked with the people, the place, and the social context in which TM is embedded (Chung 2006; Kuo 2003; Kubler 1985; Pi 1991; Tse 2000). The development of TM provides an exa mple for Michael Silverstein’s (2003) concept of indexical order , which explains the process of how particular dialect features are connected with social meanings embedded within a broader social context. Silverstein notes that w hen particular dialect features are ideologically connected with social meaning s within the society, speakers apply them to the shared language structure and use them appropriately in their speech in order to emphasize their social and group identity and the place they are orig inally from . As Gal and Irvine (1995: 973) note, “ linguistic features are seen as reflecting and expressing broader cultural images of people and activities ” . Thus, to interpret speakers ’ language use, it is necessary to understand speakers’ language ideologies which are socially constructed based on speakers ’ personal experience and their understanding of social values and cult ural norms emerging from the society. Research Questions Under a sensitive political climat e regarding Chinacentered versus Taiwancentered identity, examining TM speakers’ linguistic behaviors when they interact with PTH speakers in an immigrant setting can provide insights into whether TM speakers use their linguistic re sources to achieve convergent communication or to reinforce their linguistic norms and group identity to result in divergence. In this understanding, the present study investigate s TM speakers ’ linguistic behaviors in an immigrant setting and when they come in cont act with PTH speakers in an immig rant community where there

PAGE 23

23 is no dominant variety of Mandarin. As Blommaert (2010: 4) notes, “the mobility of people also involves the mobility of linguistic and sociolinguistic resources . ” In other words, when people move around, the speech patterns they use in the place of origin may spread to the place they migrate to. From this view, the immigrant setting is an ideal research site to examine if speakers ’ identity and language ideologies that are constructed in the place of origin would affect their linguistic behaviors in an immigrant context. Therefore, this study tries to answer the following questions: 1. How and to what extent does linguistic variation of the retroflex initials, namely (zh) (ch) (sh) and retroflex fricative (r), occur in TM speakers’ speech in an immigrant setting ? 2. To what extent, if at all, do TM speakers assimilate their speech to a standard variety of the retroflex feature or maintain their TM linguistic feature when they come into contact with PTH speakers in the local immigrant community? 3. How and to wh at extent do social network, generation, gender, and the length of stay in the US affect the variable patterns and frequency of the retroflex initials ? This dissertation is organized as follows. Chapter 2 presents the historical and l inguistic background of standard Mandarin, PTH , and TM. It provides a linguistic overview of Mandarin, PTH, and the development of TM. It also provides a detail ed description of linguistic charac teristics of the retroflex variables in TM. Chapter 3 reviews previous studies regarding dialect in contact, recent research relevant to the linguistic varieties in Taiwan Mandarin and sociolinguistic context s in Taiwan. It also presents the m ain theoretical framework and concepts for the analyses in later chapters . Chapter 4 present s the methodology adopted in this dissertation and reviews methods of data collection and data analysis . C hapter 5 presents the findings of the first linguistic var iable, retroflex initial s ( zh ) , ( ch ) , ( sh ) . It provides statistic results and the interpretation of retroflex initials regarding their relationship with f our extralinguistic

PAGE 24

24 factors: generation , gender, social network strength, and length of stay in the US, and findings from a statistic analysis. Chapter 6 presents the results of the second linguistic variable, retroflex fricative ( r ) as an initial consonant . It provides interpretation of the retroflex fricative (r) and its relationship with four extralinguistic factors: generation , gender, social network strength, and length of stay in the US, and findings from a statistic analysis. Chapter 7 summari z es the interpretation of two linguistic variables, retroflex initials ( zh) , ( ch) , ( sh) and retroflex fricative (r), and further discuss es th e sociolinguistic ideologies of the retroflex initials . Chapter 8 revisits research questions, and further discusses broader implications, the limitations of this study, and provides suggestions for future studies.

PAGE 25

25 CHAPTER 2 HISTORICAL AND L INGUISTIC BACKGROUND OF STANDARD MANDARIN, PUTONGHUA , T AIWAN MANDARIN Considering the linguistic differences between TM and PTH, t he present study explores TM speakers’ linguistic behaviors in an immigrant setting and when they have contact with PTH speakers in an immigrant communi ty where people’s high mobility trigger s the flow of linguistic recourses and the diffusion of different cultures. The study ex pands the original ideas of the influence of Taiwanese on Mandarin in Taiwan into TM as a variety of Mandarin and its association with language ideology and the rise of Taiwancentered identit y. In order to better understand the development of TM, a historical background and linguistic description of the standard M andarin, Mandarin in China, and Mandarin in Taiwan are presented in this chapter. Mandarin1 Chinese language consists of seven major fangyan ‘dialect’ including Mandarin, Wu, Gan, Xiang, Hakka (or Ke), Yue (or Cantonese), and Min. These dialects are different from each other in terms of phonology, syntax, and lexicon, and these differences lead to mutual unintelligibility. All of the dialect groups except Mandarin are categorized as Southern dialects. From a linguistic perspective, these dialects may be viewed as different ‘languages’ as we see the distinction between English and German. Although C hinese dialects show linguistic diff erences in various degrees, the speakers of different Chinese dialects use the same writ ten language i.e. Chinese characters and cultural heritage . The shared written language and cultural norms lead speakers of 1 The information provided in this paper on the history of Chinese dialectology, Mandarin, and Putonghua are drawn from a number of sources, namely Chao (1934), Coblin (2000), Li (2001) , Li (2006) , Pan (1994), Ramsey (1987), Shi (2004) , Swihart (2003), Yan (2006).

PAGE 26

26 different Chinese dialects t o consider themselves as a single cultural entity, and to believe that they ar e dialect speakers of a single Chinese language. With a large population of different dialect speakers, selecting a national standard language became an issue after overthrowing Qing Imperial government in 1911. And finally in 1932, Mandarin, based on the pronunciation of Beijing dialect , was adopted as the standard language for the people in China. Mandarin originally consisted of various northern dialects of Chinese, and is divided into three subgroups based on geographic distribution, including (1) Northern Mandarin, which is spoken in Northeast areas; (2) Northwestern Mandarin, which includes the Loess Plateau and extends to the west of Loess Plateau; and (3) Southwest Mandarin, which includes several southwestern provinces such as Sichuan, Kunming, Kweiyang, etc. In general, these varieties of Mandarin do not have major distinction s between each other in terms of linguistic structures. They have the same syntactic structures and very similar phonetic systems. For example, they have four vowels, up to six tonal categories, do not have initial voiced stops e.g., [b], [d], [g], [v], and [z], and do not hav e consonant endings e.g., [ m], [p], [t], and [k]. In addition, pronoun systems such as the third person pronoun ta2, possessive s like te or ti , or qualifying particle s like sh mo appear in all varieties of Mandari n. A lthough there are no significant linguistic differences among these varieties of Mandarin, some variations are still found due to contact with regional dialects. A ll varieties of Mandarin have four tone categories , but tonal patter n s are di fferent in some varieties . For example, the third and 2 The italic is pinyin. Pinyin ‘lit. spelling’ is the phonetic alphabet system presented by the Romanized alphabet , e.g. zh , ch , sh . It is widely used in Mainland China. Each symbol of pinyin represents a corresponding Chines e sound. In this dissertation, pinyin is presented in italic.

PAGE 27

27 fourth tones form a particular pitch p attern in Sichuan Mandarin, but the same pitch pattern is shown reversely in other variet ies of Mandarin. Due to contact with regional dialects, the differences between these varieties of Mandarin are detect able; however, these differences do not impede communication between speaker s of different varieties of Mandarin. Mandarin has served as the official language nhu ‘Official Speech’ since the 14th century. A s the language used by officials, Mandarin was not spread throughout China until the twentieth century. When Mandarin bec ame the nationwide spoken language in the early twentieth century , it ha d two name s: Guoyu ‘National Language’ and Putonghua ‘Common Speech’ . Now , because of sociohistorical reasons and political forces , the two terms are employed respectively in Taiwan and in China : Guoyu3 in Taiwan and Putonghua in China. TM in Taiwan and PTH in China are set as the linguistic standards based on the Beijing dialect and on Northern Mandarin in terms of grammar and vocabulary. T he influence from local dialects and foreign languages leads the development of TM and P TH to deviate from the standard Chinese4 and from each other. Linguistic F eatures in S tandard Chinese S tandard Chinese is based on Beijing dialect pronunciation. The basic syllable structure in s tandard Chinese is composed of an initial , a fin al and a tone. The initial is in the first consonant in a syllable and can be optional . T he final consists of three parts : a n optional medial (semivowel) presented as an M, a main vowel (headvowel) 3 In this dissertation , “Taiwan Mandarin” is used to refer to Guoyu as its English translation. 4 Here, standard Chinese refers to the language of Chinese encompassing both cultural norms and linguistic standards.

PAGE 28

28 presented as a V, and an optional syllabic ending which coul d be a vowel or two nasal consonants (alveolar nasal . Therefore, the standard s yllable structure in Chinese can be presented as (C) (M) V (V) (C) (Sun 20 06 : 35). In other words, the shortest syllable structure in Chinese can have only one vowel such as ‘goose’. The longest syllable structure can be CMVC as in qi ng ‘ strong’. The standard Chinese has twenty one initial consonants. Chinese does not have voiced stops /b, d, g/; instead, there is a contrast between voiceless aspirated and voiceless unaspirated stops and affricates , i.e., [ p, t, k ] vs. [ ph, th, kh] , and [ ts, t, t] vs. [ tsh, th, th] . Furthermore, retroflexes [ t , th, , ] are the most distinctive features in the Beijing dialect because they are not found in southern dialects , and also are not commonly used in some northern dialects. Retroflexes are viewed as a part of the standard Chinese , but not all of speakers of southern dialects , nor those of some northern dialects , are able to pronounce it properly. In most cases, in the speech of southern speakers of Mandarin, there is no distinction between retroflexes and dentals , as in /thun / ‘spring’ / thun/. In regard to the finals, there are thirty five finals in the standard Chinese , including six single vowels, namely a [a] , e [ ] , i [i] , o [o], u [u], [y] , and t he rest of the thirty finals include various combinations of a medial, mai n v owel, and a syllabic ending , e.g. uang as in ‘to load’ . As mentioned above, t he most distinctive feature in the Beijing dialect is the retroflex or ‘ retroflex sound’ . To produce the retroflex, the tip of the tongue needs to be brought up against the hard palate. Yet, the majority of Chinese speakers not born in Beijing are not able to pronounce it correctly. Speakers of southern dialects and some northern dialects are

PAGE 29

29 unable to distinguish the retroflex zh , ch , sh from the dental z , c, and s. Although the retroflex distinction is the part of the standard form, it is not often practically used among most PTH speakers . In addition, Chinese language is a tonal language, so it relies on different tonal patterns to distinguish meaning. There are four tones . The first tone is high and level, and its tone mark is presented using a horizontal line , e.g., or described as HH or 55. The second tone is high and rising, and its tone mark usually rises on the right , e.g. or described as MH or 35. The third tone is low, and its tone mark is in the middle, e.g. or described as MLH or 214. The fou rth tone is hig h and falling, and its tone mark usually falls from left to right , e.g. or de scribed as HL or 51. In standard Chinese, every syllable bears a tone. However, when a syllable is unstressed, the tone becomes weak or neutralized. For example, the question particle m a is a toneless syllable. The neutral tone is a salient feature in the standard Chinese, but it is not common in s outhern dialects. Therefore, speakers of s outhern dialects have difficulty when they distinguish the first tone from the neutral tone in actual speech. Another tonal feature in Chinese is tone sandhi. That is, tone changes its surface value when it appears together with other tones in a given environment. These changes add complexity to the tonal pattern in standard Chinese and other Chinese dialects. The typical example of tone sandhi in Chinese is that a third tone becomes a second tone when it is followed by another third tone, as in ‘office of president’ .

PAGE 30

30 Mandarin in China: Putonghua (PTH) PTH , the spoken standard Chinese, developed based on the Beijing dialect , which is a subgroup of Northern Mandar in. People who live in nort hern provinces of the Yangtze River are considered as standard PTH speakers. The term ‘ Putonghua’ was officially adopted for Mandarin in China in 1955, referring to ‘common speech’. The purpose of promoting PTH is to encourage everyone to speak a common language. U nlike mutual int elligibility among speakers of northern Mandarin dialec t s, speakers of southern dialects ( Wu, Gan, Xiang, etc.) are not mutually intelligible because of different tone patterns which are used to distinguish lexical meanings. T herefore, in order to communicate with speakers of oth er southern dialects and n orthern Mandarin, PTH is promoted as the national language. Speakers of s outhern dialects who use PTH as a second dialect become PTH bilinguals. Linguistic features in PTH preserve a large amount of features from the Beijing dialect . However, they are not completely identical. Some phonological features that appear in the Beijing dialect are excluded from PTH , such as specific syllables like i ‘catch’, variation between initial consonants such as /w/ as /v/, rhotacized words with stylistic implication s like /r/, and weak stress or neutral tone (Chen 1999, as cited in Li 2006). Despite these differences, Chinese linguists generally agree that the pronunciation of PTH is the same as the speech sounds of the Beijing dialect. Since 1949, PTH has been popularized as the national language throughout China. Meanwhile, the promotion of PTH does not attempt to replace local dialects; rather, the government encourages dialectal speakers to preserve their local dialects. As a result, various local dialects remain active and coexist with PTH, and the function of the local dialect is to project speakers’ local identity, family ties, and shared linguistic

PAGE 31

31 heritage (Dede 2006; Erabugh 1995; Gong, Chow and Ahlstrom 2011; Lehmann 1975; Li a nd Thomason 1979). The co existence of PTH and local dialects results in the occurrence of “dialect bilingualism” (Erbaugh 1995) among speakers of southern dialects. The interaction between local dialects and PTH results in cross dialectal variation. As Li and Thomason (1979) point out, contact between PTH and local dialects inevitably results in linguistic changes in both PTH and local dialects. For example, speakers’ local accent affects their pronunciation o f PTH, and new lexical items from PTH may be borrowed into their local dialects. They further note that encouraging people to us e local dialects encourages a positive attitude towards those dialects. This observation corresponds with Freeman and Heb ermann’s (1996) study , which s uggests that for a majority of Chinese, the use of local dialects reinforces their regional origins, local roots, and family ties. Therefore, it is impossible for Chinese people to give up their local dialect. C ontact between PTH and local dialects not only results in linguistic changes of local dialects , but also affects the way speakers of local dialects speak PTH, particularly for non nativ e speakers of northern dialects . Lehmann’s (1975) observation on the effect of local dialects on PTH reveals that speakers of southern dialects have difficulty distinguish ing retroflexes and dentals when they speak PTH. For example, zh nggu ‘China’ is pronounced as [t PTH; however, speakers of southern dialects pattern is found in sh u ‘speak’ [ w ] , in which the retroflex fricative [ ] is replaced with dental [s] in [sw ] by speakers of southern dialects. Mandarin in Taiwan : Taiwan Mandarin (TM) The population of Taiwan consists of f our major ethnic groups , and each group has different dialects/ languages. The Southern Min people occupy 73.3% of the

PAGE 32

32 population. They migrated from the coastal Fujian province in the southeast of China before 1895, and speak the Southern Min dialect of Chinese. Mainlanders occupy 13% of population. The Mainlanders are the Chinese immigrants who migrated to Taiwan after 1945, and speak Mandarin as their main language. T he Hakka people occupy 12% of the population. Th ey migrated from Guandong province in the south region of China, and their main Chinese dialect is Hakka. T hey migrated to Taiwan about the same time as Taiwanese people. The fourth ethnic group is Aborigine, who have inhabited Taiwan for thousands of years. The Aborigine consist of 10 tribes with various distinctive languages from the Austronesian language family . 2.1 presents the language distribution in Taiwan. Mandarin in Taiwan originally belong ed to Northern M andarin and has the same linguist ic features as Putonghua. However, it developed in its own path due to the linguistic influence s from local dialects such as Taiwanese and Hakka, foreign languages such as Japanese and English, aboriginal languages ( Austronesian languages ) , and social political and rapid economic changes in Taiwanese society after 1945. The implement of the Mandarinonly policy in 1945 resulted in diglossia. Mandarin served as the high language used in formal contexts like schools, and was associated with high prestige, legitimate practice, and standardization, whereas local dialects served as the low language used in informal contexts like family gathering s, and w ere linked with low prestige, homely names and localized objects. Until t he promotion of the “ Mother Tongue Movement” in 2001, local dialects began their revival path through education , political reforms, and media influences. Since then, the interaction between local dialects — particularly Tai wanese—and Mandarin spoken in Taiwan has become

PAGE 33

33 more intensive and has trigger ed various linguistic changes in both Taiwanese and Mandarin in Taiwan. The increasing popularity of Taiwanese and its influence on Mandarin in Taiwan ushered i n a new type of Mandarin variant that integrates the features of Taiwanese and Mandarin, and developed a new form of Mandarin particularly used in Taiwan ( Bar a n 2014; Brubaker 2012; Cheng 1985; Chung 2006; Her 2010; Kubler 1985; Li 1983; Li 1985; Pi 1991; Su 2008, 2012; Swihart 2003; Ts e 2000). Kubler ( 1985) and Li (1985) both point out that the linguistic features in TM have undergone several changes due to the great influence of Taiwanese . For example, t he absence of the non syllabic final r and retroflex initials z h , ch , sh , r is a salient feature in TM in comparison with the standard Mandarin and PTH . In addition, some tonal patterns in TM are different from those in PTH , such as the uncommon neutral tone in TM , whi ch i s widely used in PTH . Recent studies also provide evidence to demonstrate the particular linguistic features found in TM. Sanders and Uehara’s (2007) investigation o f vowel pronunciation in TM reveals that due to the merger s of / / and / / and of / u/, / u/, and /u/ into [ ], an unrounded monophthong, TM has a much smaller acoustic vowel space in comparison with acoustic space in PTH . Therefore, the researchers conclude that TM has a reduced vocalic inventory. Yang’s (2010) study on coda shift on TM has a different pattern from PTH , According to the tradi tional view, TM is the result of the influence of Taiwanese because the large population of native Taiwanese speakers carr y phonological features in their first dialect when they acquire Mandarin as a second language. Nonetheless, Kuo (2003) argues that the speakers’ first dialect seems to be oversimplified, as it

PAGE 34

34 triggers the merger from retroflex into dentals in TM as the explanatory tool . She proposes that TM is taking shape in a leveling process in which speakers level out the regionally marked features in the contact process. Kuo argues that w hen KMT retreated to Taiwan, the Mandarin population from Mainland China was composed of speakers from d ifferent dialectal backgrounds . Based on the historical records, she points out that the Mandarin population speaking Beijing Mandarin was a minority. When this wa s the case, the features of the Beijing dialect were likely to be excluded first because they were such a small minority. In comparison with the small population of Beijing Mandarin speakers , the majority of speakers were the speakers of Wu, Min, Yue, and Hakka, which only used dent al features. As a result, the use of dentals survived and spread among Mandarin speakers. Kuo further suggests that this leveling process had taken place before these speakers arrived in Taiwan, since they had served in military camp together for a lo ng time. T hese soldiers’ Mandarin was the result of a leveling process , so that when they communicated with Taiwanese speakers, the retroflex feature was no longer in their speech. Consequently , when native Taiwanese speakers learned Mandarin, the re troflex initials were absent in their Mandarin. K u o’s argument provides a promising explanation for the formation of TM and for some features that are preserved in TM but do not exist in Taiwanese, such as dental apical vowels . Additionally, Kuo’s argument echoes Cheng’s (1985) observation on the comparison among Taiwanese, Taiwan Mandarin, and Peking Mandarin (Beijing Mandarin) . He suggests that the features in TM are simpler and more regular than in the standard Mandarin. Moreover, based on his data, Cheng confirms that the effect of speakers’ first dialect is at work , as seen in some

PAGE 35

35 features of southern dialects. For example, the re is a tendency toward VO characteristics in TM . Cheng further points out that TM also has universal characteristics that exist in the Chinese language , such as adopting characteristics and zero markers for future action that are common to Chinese . In sum, w e have seen that P TH and TM originally shared the same linguist ic features , but due to contact with the local dialects , both PTH and TM under went several linguistic changes. Consequently , they both deviated from their decreed standard and drift ed away from each other. Moreover, the attitude towards the use of dialects is different in Mainland China and in Taiwan. In China, the use of local dialects is not stigmatized; rather, using local dialects is encouraged because it projects local identity and relates to cultural roots ( Dede 2006) . This posi tive attitude results in the occurr ence of interdialect al form s that not only promote the role of local dialect , but also show a positive attitude toward accomm odating the s tandard Mandarin. On the other hand, t he implementation of a Mandarinonly linguistic policy in Taiwan is reflected i n the language competence of young native Taiwanese generations after the 1950s. The young Taiwanese speakers acquire Mandarin as a second language at school while using Taiwanese as home language. Mandarin with a trace of Taiwanese phonological features is often associated with negative evaluations. Thus, it results in diglossia inside Taiwanese society. With sociopolitical changes and the rise of localism, Mandarin with Taiwanese phonological features symbolizes a new type of identit y from a mixture of Mandarin and Taiwanese (Hong 1994; Pi 1991 ). In addition to Taiwanese phonological characteristics, TM also has universal linguistic characteristics of the Chinese language, and these features lead some scholars to conclude that TM has been going through a

PAGE 36

36 leveling process , with the regionally marked features being leveled out. Therefore, TM becomes simpler and more regular, and develops its own characteristics (Her 2010; Kuo 2003). T able 2 1 to Table 23 present examples adopted from previous studies (e.g. Kubler 1985) regarding differences between PTH and TM. Variants of R etroflex I nitials in Taiwan Mandarin This section focuses on the main variants of the retroflex initials (zh), (ch), (sh), and (r) in T M in terms of the phonological features and the ex t ralinguistic factors related to the variants. The most distinctive feature in Beijing Mandarin is the retroflex initials zh [ ] , ch [ h] , sh [ ] , r [ ] and final retroflex – r [r] . However, because these features are absent in Taiwanese, it is difficult for native Taiwanese speakers to acquire or to pronounce them correctly , and therefore, hypercorrection form s are commonly found in TM speakers ’ speech (Chung 2006; Kubler 1985). With mixed l inguistic features from different dialects and languages in TM, scholars propose that TM is a new variety of Mandarin and is normatively accepted by people in Taiwan as a new “standard Mandarin” (Brubaker 2012; Kubler 1985; Li 1985; Pi 1991; Tse 2000). Chu ng (2006) argues that TM has relatively boarder range of realizing the retroflex initials (zh) (ch) (sh) . They can be pronounced by TM speakers with a complete retraction of the tongue moving from the back of the oral cavity ( [ ], [h], [ ] ) through the palatoalveolar area ( [t ], [t ], [ ] ) to the dental area ( [ts], [tsh], [s] ) . Following Chung’s observation, Brubaker (2012) further categori zes three variations as the full retroflex ([ ], [h], []), the palatealveolar ([t ], [t ], [ ]), and the full dental ([ts], [tsh], [s]) based on their place of articulation . The full retroflex initials ( zh ) , ( ch ) , ( sh ) are the standard and prescribed form taught in schools but seldom practiced in TM speakers’ daily speech. Chung (2006)

PAGE 37

37 finds that the use of the standard versus the nonstandard form is sensitive to TM speakers in terms of gender. F emale TM speakers tend to use the standard form when they are in formal settings , whereas male TM speakers do not show such a tendency. In addition to gender, Kubler (1985) notes that being able to speak proper Mandarin— for example, correctly produc ing the retroflex links with prestigious social status , while being unable to use it properly , i .e. Mandarin with Taiwanese phonological features —is associated with negative connotations such as unsophistication, low prestigious social status, etc. Some studies (e.g. Pi 1991) also s uggest that due to sociopolitical fact ors such as the unfavorable attitude towards Mainland China, young er generations tend to avoid using the retroflex feature. A s a result, they no longer distinguish retroflex features from nonretroflex ones. On the other hand, old er generations do not show a negative attitude towards retroflex features, so the distinction between retroflex and nonretroflex remains in their speech. Besides ( zh ) , ( ch ) , ( sh ) , the variation of the retroflex initial ( r ) is another distinctive feature in TM. It can be realized as two forms: retroflex fricative [ ] and lateral [l] before a vowel, in the speech of bilingual speakers of Mandarin and Taiwanese speakers. Liao (2010: 79) argues that the realization of (r ) as the l ateral [l] is considered as an “ethnic marker” because th e use of this phonological variant is associated with a particular ethnic group ( e.g. bilingual speakers of Taiwanese and Mandarin) , and region ( e.g. central Taiwan) . The standard form for ( r) should be realized as retroflex fricative [ ]. However, due to the lack of retroflex features in Taiwanese, native Taiwanese speakers are likely to substitute the retroflex fricative [ ] with the lateral [l] (Kubler 1985). Liao’s (2010) study invest igates the alternation between retroflex fricative [ ] and lateral [l] by

PAGE 38

38 comparing two groups of TM speakers in different regions in Taiwan. The results show that social factors such as political orientation, age, and gender play a role in the realization of ( r). I t is found that older speakers favor the realization of lateral [l] , while younger speakers disfavor it . Liao explains that this tendency is related to historical context in Taiwan. That is, older speakers were born in the period before the Ma ndarinonly policy had been implemented , so their dominant language was Taiwanese. During their schooling, they acquired Mandarin as the second language. Therefore, it is common that their Mandarin carries the trace of Taiwanese phonological features. On t he other hand, younger speakers were born in the era when the Mandarinonly policy was overwhelmingly implemented across the island. Thus, Mandarin became the main communication medium and their dominant language in daily speech. A s a result, most of them had poor Taiwanese proficiency , which contribute d to the lower rate of [l] realization. In sum, TM develops on the basis of the Beijing accent, and its grammar and vocabulary are align with n orthern Mandarin. However, the influence f rom foreign sources and sociopolitical factors leads TM to deviate from its decreed standard as well as PTH in terms of phonology, syntax, and lexicon. According to the traditional view, TM is the linguistic outcome due to the influence from Taiwanese phonological features when native Taiwanese speakers acquire Mandarin as a second language. Yet, Ku o’s argument regarding TM as the result of leveling provides a plausible explanation for the formation of TM and is able to explain some features that only exist in TM , but do not exist in Taiwanese . The development of TM could be explained from different perspectives. It cannot be denied that the linguistic features in TM lead it to become a

PAGE 39

39 new form of Mandarin. P revious studies also show that th e realization of retroflex variants in TM is closely correlated with extralinguistic factors such as age, gender, style, and social status. This chapter has provided the historical and linguistic background of Mandarin and its two varieties, PTH and TM. It has included a phonological description of standard Mandarin, PTH, TM, and the major retroflex variants of (zh), (ch), (sh), and (r) in TM and its relationship with extralinguistic factors. The next chapter includes a literature review relevant to dialect in contact, recent research in TM, and the major theoretical framework and concepts adopted for this study .

PAGE 40

40 Table 2 1. P honological differences between PTH and TM PTH TM a. Rhotacization in syllabus final ‘here’ zhe [t ] ~ [t ] b. /s/, /t/ and /t h/ are lenited as [ ] in word medially, and across word boundaries ’] [ts’], [t] [ts], [] [s], [ ] [dz] before a vowel or [] word finally b. No distinction between nasal finals. (Cheng 1985; Kubler 1985; Swihart 2003) Table 22. Syntactic differences between PTH and TM PTH TM a. the use of yong is used as adverbal expression, and does not serve as an auxiliary a. yong is an auxiliary to appear along with a nominalized verb with the marker de (Cheng 1985; Kubler 1985; Swihart 2003) Table 23. Lexicon difference between PTH and TM `PTH TM zi xing che ‘bicycle’ shang ren ‘businessman’ liang ‘measure for vehicles’ jiao ta ce ‘bicycle’ seng yi ren ‘businessman’ tai ‘measure for vehicles’ (Cheng 1985; Kubler 1985; Swihart 2003)

PAGE 41

41 Figure 2 1. Language distribution in Taiw an ( Source: In Chiung , Wi vun Taiffalo. 2001. Language and ethnic identity in Taiwan. In 7th Annual North American Taiwan Studies Conference, June 2001, 23 25. )

PAGE 42

42 CHAPTER 3 LITERATURE REVIEW This chapter lays out a summary of previous studies related to dialect s in contact and an overview of recent research on TM . The theoretical framework and conceptual tools are presented for the analyses of language variation and language ideologies in Taiwanese immigrant commu nit ies . Dialect in Contact : C onvergence and D ivergence The interplay of formal linguistic structures and socio cultural fact ors in contact situation s has enjoyed great attention in the field of sociolinguistics. For decades, studies have focused on the issue of language contact to explore linguistic outcomes among bilingual speakers in terms of their language performance and social cultural factors (Sankoff 2001; T homason and Kaufman 1988; Thomason 2001; Winford 2003). Nonetheless, Trudg ill (1986) argues that a contact situation occurs not only among speakers with no mutual intelligibility , but also among dialectal speakers with a certain degree of mutual intelligibility. Trudgill further states that when two dialects are in contact , a lthough regional or geographical factors may come into play, they do not impede comprehension because speakers already share some degree of mutual intelligibility by using varieties of the same language. This claim is supp orted by Shocky’s (1994) study , which investigates the pronunciation of intervocalic /t/ and /d/ among American English speakers who have resided in England for a few years . Shocky indicates that speakers generally agree that the mutual intelligibility between two English dialects facilitates their integration into local communit ies , and they do not need t o change thei r own exotic features because th ose features do not affect their comprehension of other English dialect s . Shocky also notes

PAGE 43

43 that long term accommodation and social network s are important factor s influenc ing speakers’ use of flap for intervocalic /t/s . Speakers further report that they like to receive positive evaluations from their local network when they use phonological features from the local dialect, and by doing so, they develop a sense of belonging. The se findings suggest that mutual intelligibility eases speakers ’ anxiety about miscommunication and also show that socio psychological factors may affect speakers ’ linguistic behaviors when dialects are in contact. Giles and Bourhis (1973) point out that speakers’ linguistic choice s depend on their interloc utors, topics, and settings. If speakers do not desire or acquire approval from their interlocutors, they may not adopt their interlocutors’ speech patterns. They modify their speech to disassociate themselves and create social distance from their interlocutors. Other studies demonstrate that factors such as speakers’ attitudes toward dialect s or social norm s also affect speakers ’ linguistic choice s when the dialects are in contact. For example, Zentella’s (1990) study investigates the use of lexical items from Spanish in a mixed Hispanic immigrant community in New York City , including speakers of four Spanish dialects: Puerto Ricans, Dominicans, Colombians, and Cubans. Zentella finds that owing to linguistic similarities between the four dialects, speakers become more aware of differences between the dialects, and this awareness leads them to be susceptible to the use of the lexical items in order to avoid taboo terms or homonyms in other dialect s, e.g. words such as bicho ‘insect,’ vente ‘come’, and papaya ‘papaya fruit’ are ho monyms in different dialects and refer to sex organs for different gender s. As a result, the speakers assimilate the regional terms of other dialects to avoid misunderstanding. It is also found that speakers’ attitude s determine their preference for

PAGE 44

44 certain dialect s. For instance, the image of Dominican Spanish is linked with poverty and lower educational level , and neg ative attitudes toward the se two Spanish dialects result in the rejection of the Dominican lexicon by speakers of other Spanish dialects inside the community. Despite the fact that the mutual intelligibility between dialects may reduce comprehension difficulties among speakers of different dialects, communication difficulties could happen when regional linguistic features are added to each dialect . For example, TM shares the same linguistic features with PTH in terms of identical syntact ic and syllable structures. However, the pronunciation of TM has undergone several changes due to the great influence of Taiwanese. Regional linguistic features add variations to a dialect and could result in comprehension difficulties. As Li (1985) su ggests, due to the rapid economic growth and innovation in contemporary science and technology both in Taiwan and in China, the use of youth slang, inconsistent pronunciation of the same character with different semantic interpretations, and simplificatio n of pronunciation, new Mandarin expressions are invented to pertain to socio economic changes in modern Taiwan and China. Consequently , some expressions appear in TM but are not popular in P TH , and vice versa. All these factors add linguistic variation to TM and PTH , a nd may contribute to comprehension difficulties between speakers of the two Mandarin dialect s. Because of t he influence of regional linguistic features on each dialect, dialectal speakers may also need to modify their speec h through accommodation process es . Due to the need to communicate with each other , and the desire to be intelligible, dialectal speakers tend to diminish dissimilarities between diverse dialects. L eveling is

PAGE 45

45 one of the strategies used to eliminate regionally marked features, and to make different dialects become more homogenous , i.e. lead to dialect convergence (Kerswill and Williams 2000; Torgersen and Kerswill 2004; Trudgill 1986). For example, T rudgill (1986) finds that most Norwegian varieties have two standard vowel systems: long ( /y:/, /i:/, / :/) and short vowels ( / / , / /, /u/) . However, the vowel system in Sogn dialects in the vicinity of Hyanger island is relatively dif ferent . Sogn dialects have long high vowels ( /y:, i:, : , u:/) as diphthongs. I t is found that the long vowel /o:/ in the original Hyanger dialect is realized as a diphthong / o / in Sogn dialects , but diphthongization of long vowels is only found in certain areas of west Norway. Trudgill explains that the realization of diphthong is regionally marked, so it disappears in the new Hyanger dialect due to leveling . Trudgill ( 2000) and his associates’ investigation o f the formation of Modern New Zealand English further suggest s that the leveling process usually takes a long time to complete. Trudgill and his associates claim that the formation of Modern New Zealand English is the result of dialect mixtur e and new dialect formation through contact among settlers from different British Isles ’ varieties of English. At the first stage of the leveling process, many common 19thcentury features found in British Isles English —e.g., the realization of /v/ as [w]—disappear in Modern New Zealand English. At th e first stage, regionally marked features are first eliminated, and then the unmarked features survive. At the later stage, the process of simplification diminishes those which show fewer irregularities. The development of Modern New Zealand English presents an example of new dialect formation, whose process takes several stages through several generations, and ultimately leads to homogenous language.

PAGE 46

46 Yet, dialect divergence would occur when speakers try to create social distance between themselves and interlocutors who do not use the same dialect. For example, Bourhis and Gile’s (1977) study on accent convergence and divergence among Welshborn adults shows that Welsh speakers diverge from speakers with Received Pronunciation (RP) by using a Welsh accented dialect and only converge with their RP speakers during business transactions. An interesting example from Stanford’s (2008) study also shows that when dialects are in contact, convergence is not necessarily a default result. Stanford inv estigates tone variants of immigrant Sui women in southwest China when they are married to husbands who are from different clans and have regional features in their own Sui dialect. The results show that most of the i mmigrant Sui women maintain the t one variants of their original dialect after having lived in their husband’ s clan for a long time. Most of the immigrant women report that they would be criticized or ridiculed if they did not use their original dialect, and the use of their original dialect is considered as a projection of their clan identity. Stanford further notes that long term contact does not necessarily result in dialect convergence when social factors such as avoiding negative criticism and maintaining clan identity come into play contributing to dialect divergence. Stanford (2008: 446) concludes that the case of Sui clan based exogamy shows that “an individual’s identification with his or her home clan outweighs the common human tendency to adapt to a new dialect region to some extent.” In other words, speakers’ individual and social identit ies are re inforced by the linguistic features they use. In addition to social identity, social factors such as speakers’ social bias and social norm s inside the community affect speakers’ accommodation patterns. For

PAGE 47

47 example, Bell’s (2010) investigation on N ew Zealand English (hereafter NZE) speakers’ accommodation patterns in response to an Australian talker finds that NZE speakers show some degree of accommodation to the Australian talker’ s vowel pattern. However, Bell argues that the NZE speakers’ convergence to the Australian talker’s vowel pattern does not occur automatically because vowel convergence does not apply to all Australian vowels. NZE speakers only adopt some Australian vowels and have different degrees of accommodation. Bell explains that t his is a result of social factors , such as NZE speakers’ social bias , which affect the degree of their accommodation. When NZE speakers hold a more positive attitude towards Australia, they are more likely and willing to accommodate. Bell argues that in the accommodation process, social factors play an important role in determining if speakers accommodate to their interlocutors , and to what degree. In a similar vein, the work of Otheguy, Zentella, and Livert (2007) also shows that speakers’ social bias affects their accommodation pattern. The researchers examine the use of overt subject personal pronouns in finite clauses among six Spanish dialect groups in a Spanish immigrant community in New York City . They find that the speakers of Spanish dialects incline to a given variety of the same language depending on the social value that emerges from local communities. Due to local prestige, longer residence in New York City, and positive social value in Caribbean Spanish1, accommodation to the Caribbean usage takes place instead of moving toward the prestigious Mainland Spanish. 1 Otheguy et al. classified six Spanish dialect groups based on two main dialect regions: Caribbean and Mainland. The speakers of Caribbean Spanish include Dominicans, Puerto Ricans, and Cubans. The speakers of Mainland Spanish include Colombians, Ecuadorians, and Mexicans.

PAGE 48

48 In addition to the above social factors, the social network serves as an influential factor in contact situations as well . Several studies have employed Milroy’s (1980) social network framework to examine the correlation between speakers’ social network structure s and their linguistic behaviors. As Romaine (1982) suggests , in a speech community, individual social network usually has a great impact on speakers’ language use because individual variation adds variability to it , and speakers may be involved in mor e than one membership or social group i nside the speech community . Additionally, Milroy and Gordon (2003) note that a speaker ’ s personal network is embedded within a w ider sociolinguistic context , so it can serve as a middle site between individual and community. In this sense, to understand the dynamic of individual variation on language changes within a speech community, personal social network provides researchers with “[] a means of capturing the dynamics underlying speakers’ interaction behaviors []” (Milroy 2002: 549). Bayley (2012) and his associates investigate the use of subject personal pronouns (henceforth SPP) in Spanish between highland dialect s ( e.g. Mexico ) and lowland dialects ( e.g. Puerto Rico ) . It is noted that highland Spanish speakers use fewer overt SPPs than lowland Spanish speakers. T h e researchers are interested in the us e of SPPs among Puerto Ric an Spanish speakers in comparison with the rate among Mexican Spanish speakers, and the correlation with the role of social network s and individual variations. Due to the majority popul ation’s influence , a convergence with Mexican Spanish is predicted. Nonetheless, the results show that although demographic imbalance and geographic proximity seem to play a role in triggering convergence, frequent contact in a variety of social contexts is the key factor leading to

PAGE 49

49 convergence. Fo r Puerto Rican speakers , when their social and professional social networks primarily consist of Mexican Spanish speakers, the rate of using overt SPP is similar to Mexican immigrants and Mexican American Spanish. On the other hand, when Puerto Ricans speakers ’ social networks mainly consist of other Puerto Rican speakers, the rate of using overt SPPs is higher than among those who frequently use Spanish with Mexic an Spanish speakers . The findings from Bailey’s study suggest that an individual speaker’s social network is one of crucial factors that constrain the occurrence of lingui stic va riations. A number of e arlier studies have employed Milroy’s (1980) social network framework to investigate speakers’ linguistic behaviors in a closeknit network community . For example, Cukor Avila’s (1997) study on AfricanAmerican Vernacular English (A AVE) in an East Central Texas town finds that each member develops different social networks based on the type of shared information from different social groups. Within each network, group members develop intimate relations hip s , share confidence by using intimate vernacular forms, and are more likely to retain local vernacular forms . T he earlier studies focus on the linguistic changes in closeknit communit ies ; however, several studies explore the correlation between speakers’ social net work and language change in a looseknit community , which consists of high mobile speakers with a lack of loyalty to the local vernacular form. For example, Hirano (2008) investigates the use of intervocalic (t) as in “ letter ” among native speakers of English from different English speaking counties in a highly mobile community in Japan where there is no dominant English dialect. The results show that speakers’ social network plays an important role in affecting their lingu istic preference in the looseknit

PAGE 50

50 community. Hirano finds that when speaker s have a strong social network with speakers from the same home country, they are more likely to remain loyal to t heir own variety of English . On the cont rary, the speakers who have a strong network with speakers using different English varieties tend to avoid using their home variants and accommodate to the speakers of th e other varieties . Hirano concludes that the direction of accommodation in an immigrant setting is determined by the speakers’ close network and frequency of interaction with their interlocutors. An overview of several studies in dialect contact has demonstrated that a co mplex variety of factors contributes to changes in dialectal speakers’ linguistic behaviors when they interact with each other. The correlation between linguistic variables and social factors is found in domestic migration as well as in migration to other countr ies . Migration to foreign countries provides speakers with an opportunity to come into intensive contact not only with speakers from other languages , but also with speakers from other varieties of the same language. Hock and Joseph (20 09) note that in an immigrant setting, the effect of local loyalties to the home dialect is likely to become minimal after a long length of residency in a foreign country. However, Hock and Joseph do not provide further explanation as to the effect s that local loyalties to a home dialect may have on speakers’ linguistic behaviors . Hock and Joseph provide a possible factor that may affect speakers’ linguistic behaviors in an immigrant setting. As traditional dialectology is concer ned with the distribution of linguistic features in a specific geographical place ( e.g. a country, a region, or a community ) , in the context of globalization, people move around frequently , and their mobility results in their linguistic features moving around as well. L anguage is viewed as local practice, and social

PAGE 51

51 meanings and functions of linguistic variables are given and valued by the speaker s in the local community. When they move arou nd, they carry these linguistic variables with social meanings and functions ideologically constructed from their local practice to the globalization context (Blommaert 2010; Johnstone 2004). If we would like to better understand how and to what extent dialectal speakers change or maintain their home d ialects in an immigrant setting in the context of globalization, more factors such as gender, generation , social network strength, and length of stay in a foreign country need to be examined in order to provide us with a more complete picture . Research in Taiwan Mandarin Earlier research (e.g. Kubler 1985; Li 1995) regarding TM has focused on linguistic features emerging out of contact between Mandarin and local dialects. Later studies (e.g. Kuo 2003 ) propose that due to the influence of Taiwanese and leveling, TM has become a distinctive variety of Mandarin that differs from Mandarin in China. As early studies focus on the influence of Taiwanese on Mandarin used in Taiwan, recent studies have shifted their focus to explore the correlation between the variety of TM features and their social meanings that possibly contribute to TM speakers’ linguistic behaviors. For example, Chung (2006) investigates the hypercorrection of retroflex initials concerning the type s of context s and the kind s of interlocutors in terms of gender, age, educational level, and home language use. Chung points out that although the standard retroflex initials, which are pronounced with complete retraction of the tongue moving from the back of the oral c avity, are taught in Taiwan schools and used in major media, they are seldom practiced in daily conversations. Furthermore, TM speakers are not familiar enough with the forms , and usually apply them in the wrong context.

PAGE 52

52 Particularly, the substitution of retroflex ed initials with dentals often occurs when the dentals are called for , e.g. [ s] in s ‘four’ is mis pronounced as [ ] ; as a result, the hypercorrect form [ i ] is formed. Although TM speakers are conscious of the use of retroflex initials in conversations, their self consciousness of the retroflexed initials often results in hypercorrection. Chung further suggests that the employment of the hypercorrect retroflex initials serves as a stylistic device when TM speakers are conscious of their speech in different contexts , such as formal versus inform al settings. T he use of hypercor rect forms in one’s speech is further associated with social background such as educational level, age, and soc ial class. Although Chung does not provide a quantitative analysis, her observation still provides valuable insights into the actual practice of the full retrof lex feature, such as [ ], [ h], [ ] in TM . Chung argues that TM has a relatively b roa der range of realizing the retroflex initials , in cluding the full retroflex form, the intermediate form ( [t ], [t ], [ ] ) , and the full dental form ( [ts], [tsh], [s]) . She further notes that the intermediate form in particular is commonly found in TM speakers’ causal speech. Yang’s (2008) study examines the alternation of (f) in the speech of native Taiwanese women who intermarried with native Mandarin men. The variation of [f] and [hw] is one of the salient linguistic features in the speech of native Taiwanese speakers when they speak Mandarin. It is postulated that due to the influence of Taiwanese, native Taiwanese speakers are more likely to produce [hw] in place of [f] when the standard variant [ f] is called for. The results show that social factors such as speakers’ language attitude towards their mother tongue and education level affect their production of [f] and [hw]. For example, when speakers have a positive language

PAGE 53

53 attitude towards their m other language and do not have a high education level, they show a higher frequency of [hw] when the standard variant [f] is called for. Yang further finds factors such as age, education level, and dialect background also play a role in production of hypercorrect forms of [hw] for [f]. Based on the findings, Yang concludes that intermarriage could be a factor that increases the occurrence of the variation between [hw] and [f]. H owever, Liao (2010) argues that other factors besides intermarriage may co me into play and trigger the occurrence of the variation of [hw] and [f]. Liao notes that native Taiwanese speakers without intermarriage could also produce the hypercorrect form due to the Mandarin only policy. The overwhelming implement ation of the policy since 1949 results in diglossia in Taiwanese society ; Mandarin i s the high language, while local dialects such as Taiwanese are the low language. Liao suggests that native Taiwanese speakers’ Mandarin could be affected by the phonol ogical features of their mother language ( i.e. Taiwanese) , and therefore, the variation of [f] and [hw] is found prevail ingly in native Taiwanese speakers’ speech. Many researchers have pointed out that due to the influence of Taiwanese, Mandarin in Taiw an is considered as a new and distinctive Mandarin that differs from Mandarin in China. Previous works such as Chung (2006) and Yang (2008) focus on the use of specific TM features in different social contexts. R ecent research ers are more concern ed about how language ideology and social identity are constructed through the use of different TM features. For example, Liao’s (2010) study examines the use of two particular regional TM features: T one 4 rising and the alternation between retroflex

PAGE 54

54 approxim ant [ 2] and lateral [l] in the speech of TM speakers from cent ral Taiwan when they migrate to northern Taiwan. The researcher is interested in ascertaining whether speakers’ linguistic choice is constrained by social factors such as social identity, la nguage attitude, social network, and ideology by comparing two groups of TM speakers in different sociopolitical context s. Liao finds that TM speakers who migrate to northern Taiwan do not maintain their home TM features due to constant exposure to northern TM features and their extensive social network with the locals. A t the same time, the findings show that certain speakers preserve home TM features , such as T4 rising , in order to project their nonlocal identity , although they have r esided in the northern region for a long time. In addition, the variation of retroflex approximant [ ] and lateral [l] is triggered by social factors rather than internal linguistic context. Liao points out that historically, the realization of [l] for ret roflex approximant [ ] has been considered as a stigmatized f or m due to its association with the nonstandard Mandarin, i.e. Taiwaneseaccented Mandarin. However, Liao finds that the negative attitude towards the realization of [l] seems to change. Liao explains that this may be due to the promotion of Taiwanese identity and local culture since 2000. The positive attitude towar ds local cultur e and emphasis of local identity may affect how TM speakers view and use [l] when the retroflex approximant [ ] is needed. Moreover, speakers with a positive attitude toward local norms tend to resist accommodation to the standard retroflex variant. The findings from Liao’s study suggest that TM speakers’ attitude towards local dialects, local 2 Liao argues that TM differs from standard Mandarin in terms of the degree of retroflex feature, so she adopts the retroflex approximant [ ] for the retroflex fricative [ ], the standard variant in standard Chinese.

PAGE 55

55 identity, and language ideology influence their linguistic choice, and their linguistic behaviors further reflect the changing sociocultural contexts inside Taiwanese society. Previous studies (e.g. Bara n 2007; Chung 2006; Tse 1998) have discussed the existence of the intermediate form in Mandari n spoken in Taiwan. As Tse (2000) suggests, in TM there is a retroflex f eature that is neither the full dental f eature nor the full retroflex feature. However, no empirical research was conducted regarding the presence of this intermediate form. Brubaker’s (2012) study is the first study that connects the varieties of TM retroflex feature with different sociocultural meanings. Brubaker uses the retroflex variable zh [ ] , ch [ h] , sh [ ] to examine how TM speakers evaluate the different sets of the TM retroflex feature. Following Chung’s (2006) observation, Brubaker notes that the retroflex feature in TM has a relatively broader range. I t can be realized from “the full retroflex, to an intermediate palatoalveolar articulation, and then to the fully dental form” (Chung 20 06: 200). The results show that the full retroflex feature, the standard form, is often linked with speakers’ socio economic status such as educational level, occupation, and income, and regional differences, e.g. northern Taiwan. Furthermore, Brubaker points out that TM speakers are sensitive to the use of the full retroflex feature. Although it is the prescriptive form taught in schools, the proper use of the retroflex feature seems to become a barrier t hat creates social distance among TM speakers, and is further linked with negative evaluations such as “unnatural, artificial and pretentious” (Brubaker 2012: 106). In addition, in the past, the use of the dentalized retroflex feature was associat ed with negative connotations, such as unsophistication and lower social status. Yet, Brubaker notes that although the dentalized retroflex feature is often associated with negative

PAGE 56

56 implication s, it is also used by some speakers to project an ethnic identity ( i.e. as Taiwanese) and to symbolize solidarity. In addition, the speakers in Brubaker’s study report the presence of the intermediate form , which is considered an “acceptable” and “typical” form in the speech of the majority of TM speakers. Brubaker further notes that the employment of the intermediate form ( i.e. the palatoalveolar feature) is not linked with any negative stigma , as he finds with the use of the full retroflex and the dentalized retroflex features. Despite the fact that the intermediate form is also conceived as the nonstandard form, it is the normative form that is widely accepted by TM speakers in contemporary Taiwanese society . Bru baker’s study demonstrates an ongoing change in Taiwan Mandarin in terms of its linguistic features , which gradually shift away from Beijing pronunciation, and of the ideology of what has been considered the standard Mandarin versus nonstandard form inside Taiwanese society . With the rise of Taiwan local identity and a strong sense of local culture, the sociocultural meanings associated with the standard form as well as the nonstandard form are a lso chang ing in Taiwanese soc iety . Research on Taiwan Mandarin has demonstrated that the influence of Taiwanese on TM results in a variety of specific TM linguistic features, and gradually deviates TM away from the standard Mandarin and Mandarin in China. Furthermore, in addition t o age and gender, TM speakers’ linguistic behavior are found to be closely linked with a diversity of social factors , such as language attitude, group identity, and language ideology. As Liao (2010 : 46) notes, TM speakers’ linguistic choice “reflects the conflicting ideologies in Taiwan and how identities can be constructed within discourse.” To better understand TM speakers’ linguistic choices, adopting multivariate analysis

PAGE 57

57 helps explain the correlation between linguistic variables and a variety of social factors, and explores whether the variations of linguistic variables are systematically modeled. Theoretical F ramework The v ariationist sociolinguistics approach is adopted as the main theoretical framework in this dissertation. The principle is presented for the analysis of language variation in TM speakers’ speech. In addition, the principles of order of indexicality and acco mmodation theory are also described for the analysis of language ideologies and language attitudes that affect TM speakers’ linguistic choices in an immigrant setting. Variationist S ociolinguistics A pproach The v ariationist sociolinguistics approach pr oposes that language variation and change can be systematically analyzed by using quantitative modeling. According to this approach, variations in language occur everywhere and are constrained by both linguistic and social factors , and speakers’ linguistic behaviors are the reflection of the underl ying grammatical system of these settings (Bayley 2002). Based on this framework, researchers’ primary objective is to seek the patterns of usage in terms of the frequency of r ecurrent features or cocoexistence of features in the context of a wi der sociocultural setting. T he employment of quantitative methods enables researchers to find the distribution of both linguistic and social patterns within a community (Poplack 1990; Tagliamonte 2012). Sankoff (1990) also notes that variationist methodology holds the advantage of using statistical tool s and quantitative data to analyze patterns systematically and provide researchers with objective results , in comparison with using intuitive judgment and anecdotal report s in other approaches. The v ariationist approach also takes social factors into account , since they could possibly influence speakers’ linguistic choice s and also tr igger language changes and

PAGE 58

58 variations . As Labov (1963: 275) points out, “social pressures are continually operating upon language, not from some remote point in the past, but as an imm i nent social for ce acting in the living present .” In other words, lang uage change and variation not only rely on the interpretation of internal linguistic changes , but also need to be explained by extra linguistic factors which potentially affect speakers’ linguistic choice s. The v a riationist approach has been applied to various research topics such as dialect contact in terms of short and long term accommodation in an immigrant setting i n various studies ( e.g. Chi 1991; Li 1991; Hirano 2008, 2010; Pan 1994; Williams 2011) ; social network ing in Hernndez (2009) , in view of the use of Spanish wordfinal nasal velarization between inand out group exchanges ; code switching in bilingual contexts i n Poplack (1990) ; the use of English loanwords in Canadian F rancophone neighborhoods i n Poplack, Sankoff, and Miller (1988) ; and style changes between Indian English and British English in Sharma’s study ( 2011). These studies have demonstrated that the sociolinguistic interview provides researchers with a platform to collect their data, and quantitative methods allow them to discover the systematic differences between speakers and the distribution of the interaction between social factors and linguistic variables. The major data collection technique in the variationist appro ach is the sociolinguistic interview. The design of the interview is to elicit interviewees’ vernacular speech from a natural conversation in which they are expected not to pay attention to their language use (Labov 1966, 1972) . T o elicit interviewees’ spontaneous and naturally unscripted conversational speech, researchers need to be cautio us about the design of the interview. To achieve th eir purpose , a careful ly p lanned interview is

PAGE 59

59 required. Before conducting a sociolinguistic interview, researchers usually preselect a set of specific topics, i.e. modules that include possible situations or interests which are pertinent to interviewees ’ everyday li v e s. Furthermore, u nlike discussion between friends, the purpose of a sociolinguistic interview is to allow researchers to collect as much spontaneous data as possible in order to attest to whether speakers’ language use presents a systematic pattern, and whether the distribution between linguistic variables and social factors is constrained by regular rules of language. In addition, i nterviewees are encouraged to narrate their personal experience, and the role of interviewers is to engage interviewees in the conversation, to keep the conversation flowing as freely as possible and to control their interviewees’ speech in order to gather the linguistic variables they need for their research (Gordon 2006; Milroy and Gordon 2003; Schiffrin 1994). Another characteristic of the variationist approach is to find a representative sample, but it can not be randomly selected. S peakers need to b e specifically chosen based on researchers’ prior observation, which enables them to select promising speakers in the community and avoid any biases toward a certain group (Chi 1991; Pan 1994). Furthermore, it is stressed that although language use shows variations, they are still rule governed. In order to find variations, eliciting vernacular is important because vernacular is viewed as natural speech to which speaker s pay minimal attention when they are encouraged to narrate their own stories. Therefore, the more vernacular speakers use, the more linguistic variables that researchers can capture. In order to analyze the factors that may significantly affect speakers’ choice s of linguistic variant forms, the employment of multivariate statistical testing serves as a supplement al tool to

PAGE 60

60 interpret patterns of variants, to discover systematic patterns of usage between different speakers, and to reveal a general patter n in speakers’ groups or communit ies (Bayley 2002; Tagliamonte 2012). However, Mi lroy and Gordon (2003) note that t he design of the sociolinguistic interview is not flawless. R esearchers encounter some problems , such as an unequal relationship between interviewers and interviewees, the formality of the interview, the rarity of some linguistic structures, and misleading hypercorrected forms. In addition, the sociolinguistic interview may encounter the observer’s paradox (Labov 1984) , which suggests that the interviewees’ language use is influence d when they are consciously aware that they are being audioor videorecorded during the interview. Thus, to diminish interviewees’ anxiety of being recorded, it is suggested that topics asked in the interview should be nonthreatening and include general interests related to the interviewees’ daily life in the community . Doing so allow s the interviewees to pay minimal attention to their language use when they narrate their personal experience and stories. Order of I ndexicality The rise of Taiwanese identity is ideologically correlated with the development of TM in the sociohistorical context in Taiwanese history. Michael Silverstein’s semiotic concept of “orders of indexical” present s a process of linguistic forms acquir ing social meanings w ithin a broader social context. Thus, this study employs the concept of indexicality to explain how linguistic variants are ideologically linked with particular social meanings in Taiwan, and how language use as local practice is associated w ith group identity in the context of globalization.

PAGE 61

61 Silverstein (2003) claims that “any n th order indexical presupposes that the context in which it is normatively used has a schematization of some particular sort, relative to which we can model the “appropriateness” of its usage in the context” (2003: 193). Silverstein further compares his orders of in d exicality with Labov’s (1972) taxonomy , which stresses the type of labels for linguistic forms correlating with soc ial meanings , i.e. indicators, marke rs , and stereotypes. The first order indexicality presupposes that a correlation between linguistic variables and social meanings occurs , but speakers are not aware of the correlation between indexical linguistic features and social meanings in their speec h. W ithin their social network, everyone uses them in the same way, so they do not notice the differences between themselves and speakers from other regions. A nd due to limited contact with speakers from other regions, the use of the features is strengthened within a close network. This concept is similar to Labov’s taxonomy of sociolinguistic ‘indicators’ which propose that “at this stage, speakers show zero degrees of social awareness, and are difficult to detect for both linguists and native speakers” (Labov 2001: 196). T he secondorder indexicality is recognized by speakers’ self awareness of first indexical features whi ch are related to certain social and historical meanings inside the community. At this stage, speakers begin to notice the differences of the first indexical features used by speakers from different regions or in different social contexts, and later, their linguistic behaviors reflect their awareness , such as hypercorrection or style shifting due to their ling uistic insecurity. Silverstein notes that the secondorder indexicality is similar to Labov’s taxonomy of sociolinguistic ‘markers’, as Labov (2001:

PAGE 62

62 196) suggests, which are “reflected in sharp social stratification of speech production, a steep slope of style shifting, and negative responses on subjective reaction tests. ” T he reinforcement of the indexical features as sociolinguistic marke rs ultimately leads to Labov’s sociolinguistic ‘stereotypes’, which “are markers that have tilted in the direction of ideological transparency, the stuff of conscious, valueladen, imitational inhabitance” ( 1972 as cited in Silverstein 2003: 220). Silverstein’s (2003: 220) third order indexicality which proposes “ n +1storder indexicality has become presupposing, in other words, in effect replacing an older n th order indexical presupposition” is linked with Labovian’ sociolinguistic stereotypes. Later studies such as Johnstone, Andrus and Danielson (2006) combine Micha el Silverstein’s semiotic concept of “orders of indexical” and Labov’s taxonomy, to explain the process of how linguistic variables are used to index social meanings. By combining two frameworks, Johnstone and her colleagues propose that the history of dialect enregistration in Pittsburghese is followed by connected orders , i.e. first order indexicality, secondorder indexicality, and thirdorder indexicality, and these orders of indexicality are in “dialectical r elationships with one another ” ( 2006: 84) . In other words, each order is a connected process; therefore, the occurrence of the first order indexicality triggers the second order indexicality to take place, and the third order indexicality proceeds to occur after the secondorder indexicality. A s seen in the case of enregistration of Pittsburghese, the process of speakers’ self conscious ness of the linguistic features in their speech reflects that a set of linguistic variables come to carry social meanings step by step. At first, Pitts burghers do not notice that linguistic features in their speech project their identit ies , and index localness until they experience social

PAGE 63

63 and geographical mobility driven by economic growth in the region. Additionally, in interviews, interviewees report how they come to be aware that the use of the specific linguistic features underlies some social meanings. T hey begin to notice certain linguistic variables in their speech, e.g . /aw/ monophthongization indexing social class and their local identity when they have been socially mobile, such as moving from working class childhood to middleclass adulthood, or geographically mobile, such as study ing out of town, and these local linguistic features are n oticed by speakers of other social classes or of different regions. Furthermore, the researchers indicate that once the connection between regional features with local identity is formed, the speakers are performing the third order indexicalty of original features. T he process of the enregistration of Pittsburghese presents that linguistic practices within a speech community shape speakers’ linguistic behaviors, and speakers’ dense network s strengthen these behaviors. I n addition, the enregist ration of Pittsburghese presents that individual speakers’ linguistic experience as well as the norm inside speakers’ speech communit ies can be reinforced by media and community support. Finally, the researchers conclude that t he enregist er ed process t akes place in order , and indexical orders allow us to see that dialect normative practices are connected by different levels instead of by a separated process. The study of Pittsburghese provides us with a detail ed description of dialect normative practices in progress from sociopsychological, sociodemographic and interactional perspectives . It further illustrates the change of speakers’ attitudes toward local linguistic features in an era of socioeconomic growth, and the linguistic variables come to link to social values under the effect of social and geographical mobility.

PAGE 64

64 As Blommaert (2010: 38) points out, “ordered indexicality within large stratified complexes in which some forms of semiosis are systematically perceived as valuable and some are not taken into account at all”. And the register within this systematic pattern is sensitive to outside social forces such as authority, evaluation, perception from others, and politics. For example, Liao’s (2008) study applies Silverstein’s order of indexicality to examine TM speakers’ perception of different TM phonological variants in terms of their indexical relationship with social meanings. After the Nationalist Party ( KMT ) retreated to Taiwan in 1949, their government aggressively promoted their main language, Mandarin, as lingua franca among speakers of different dialects and l anguages such as Taiwanese, Hakka, and native AustroPolynesian languages , to reinforce their political regime. At that time, Mandarin was considered as the high language while local dialects were the low language. After liberalization of language policy in 1987 and the rise of localism , TM speakers whose Mandarin carries the trace of Taiwanese phonological features are indexed as local identity. On the other hand, TM speakers whose Mandarin is not influenced by Taiwanese ( standard Mandarin) are indexed as Mainlander . Liao points out that this linguistic phenomenon is related to the first order indexicality. Since 1987, the democratization in politics has led to huge changes in Taiwanese society in terms of language policy, language use, and ideology. Due to sociopolitical changes in Taiwanese society, TM speakers’ perceptions and attitude s toward different languages on the island are also affected. Liao further indicates that TM speakers’ perceptions and attitudes toward different languages affect their presupposition of political parties. The image associated with KMT is not only the language that speakers use, but also the region: northern

PAGE 65

65 metropolitan areas. On the other hand, the image linked with the second largest political party, the Democratic Progressive Party (hereafter DPP) is the advocator of local Taiwanese dialect s and Taiwancentered identity, and the rural areas in the south. TM speakers ’ language preference indirectly projects their political inclination and region. TM speakers who use more standard Mandarin would be indexed as Mainlander, politically incline d towards KMT, and residing in the northern region. In contrast, TM speakers whose Mandarin carries the trace of Taiwanese phonological features would be indirectly indexed as local Taiwanese, politically aligned with the DDP, and resid ing in th e southern region. Liao notes that when TM speakers notice the different linguistic features, they impose social meanings according to their perception of local history and language ideology on these linguistic features, and further assign their presuppos ed ideology to other speakers. This stage, as Liao notes , is when the secondorder indexicality arises . Liao’s study exemplifies the social context in Taiwanese society , which has undergone socio political changes , to illustrate how linguistic features come to be assigned to social meanings. As Blommaert (2010: 37) claim s, “clustered and patterned language forms that index specific social personae and roles can be invoked to organize interactional practice, and have a prima facie s tability that can sometimes be used for typifying or stereotyping.” Accommodation T heory The concept of ‘Speech Accommodation Theory’ (SAT), first proposed by Giles and his colleagues in 1973, was designed to interpret the interpersonal accommodation pro cesses. Later, it was renamed ‘ Communication Accommodation Theory ’ (CAT) to encompass various topics , including speakers’ verbal and nonverbal behaviors. Two fundamental concepts , convergence and divergence, are applied to explain linguistic

PAGE 66

66 phenomen a when different dialect/language groups interact with each other in different contexts. Giles, Coupland and Coupland (1991: 27) argue that convergence “is a strategy of identification with the communication patterns of an individual internal to the int eraction.” Ultimately, through the process of convergence, linguistic dissimilarities are likely to be eliminated. As Trudgill (1986) notes, for the need to communicate with each other, be intelligible, and seek approval from their interlocutors, speakers try to decrease comprehension difficulties by reducing or eliminating dissimilarities. Giles et al. (1991) further note that social factors contribute to the occurrence of convergence, such as a desire to gain social approval from others, asymmetrical power relationships such as those between employers and employees or customers and sales persons, and social norms, etc. A number of s tudies of linguistic convergence have found that the elimination of salient dialect features (such as dialect leveling ) , koineization ( a new form that levels out minority and marked speech forms of local dialects ), and the emergence of a compromise variety (the interdialect ) occur when dialects are in contact, and during the accommodation process, social factors also interact to affect speakers’ linguistic behaviors (Anderson 2002; Britian and Trudgill 1999; Dye r 2002; Hinskens , 1998; Kerswill and Williams 2000; Kerswill W illiams 2002; Ntzel and Salmons 2011; Trudgill 1986; Trudgill, Gordon, Lewis, and MacLagan 2000) . Divergence occurs “when speakers accentuate speech and nonverbal differences between themselves and others” (Giles and Coupland 1991: 65). Giles exemplifi es Bourhis and Giles’ (1977) study to illustrate how Welsh born adult speakers use their linguistic resources ( i.e. Welsh accented dialect ) to diverge themselves from speakers with Received Pronunciation (RP ). The idea of divergence has been adopted

PAGE 67

67 as one of the theoretical concept s to explain speakers’ language use in contact situations and the correlation between linguistic phenomen a and social factors . Tong, Hong, and Chiu (1999) examine the language attitudes of people in Hong Kong toward bilingual codeswitching in terms of how they identify themselves in transitional Hong Kong one year before the hand over of sovereignty from the United Kingdom to China. The researchers point out that in transitional Hong Kong, Hong Kong people are divided into two main groups regarding their attitude towards the handover: separation and integration. People who identify themselves as Hongkongers, permanent resident s of Hong Kong, are considered to be ingroup members. On the other hand, people who identify themselves as Hong Kong Chinese ( both Hong Kong natives and Chinese Mainlanders ) are considered to be out group members. The dichotomy of Hong konger and Hong Kong Chinese is further linked with Hong Kong citizen s’ preference for separation and integration ( Hongkongers incline toward separation, while Hong Kong Chinese prefer integration) . The researchers find that Hongkonger s intentionally avoid anything related to strong Chinese identity. In addition, their attitude towards social identity affects their language use. T h at is, Hong Kong people with a strong Hongkonger identity incline to ward us ing Cantonese, the ingroup language, to communicate with Mainlander Chinese, the out group members. On the other hand, Hong Kong people who identify themselves as Chinese tend to use Putonghua and also expect other Hong Kong people w ill use it as an intragroup communication tool. The researchers point out t hat a strong identification as Hongkonger might interfer e with creating a harmonious intergroup relationship between Hongkongers and Hong Kong Chinese after the

PAGE 68

68 hangover . This study is a good example show ing how language attitude and social identity trigger the occurrence of speakers ’ divergent communication in order to maintain their social identity . As Bo r uhis (1979, as cited in Giles and Coupland 2001: 66) notes, “’ sp eech maintenance’ is a valued (and possibly conscious and even effortful) act of maintaining one’ s group identity. ” This chapter has provided a description of the theoretical framework adopted for this dissertation, namely variationist sociolinguis tics for the analysis of language variation in TM speakers ’ speech, and also presented an overview of the studies regarding dialect contact and recent research in Taiwan Mandarin as a basis for understanding speakers’ linguistic behaviors in contact situations, and language practice s and language ideologies that are constructed in Taiwanese society . The chapter also includes theoretical concepts : order of indexicality and accommodation theory for the analysis of language ideologies of TM speakers and language practice in an immigrant environment . The next chapter presents the methodology applied in this study.

PAGE 69

69 CHAPTER 4 STUDY DESIGN This chapter explains the study design of the current research and the main methodology principles adopted. The discussion begin s with a description of linguistic variables applied in this study. An explanation of the criteria for selecting the participants and their demographic information is also provided. D ata collection and procedur e are explained, as is the rationale for using the sociolinguistic interview a s a major data collection technique. E xtralinguistic variables adopted in this study are described, and f inally, statistic al methods are discussed . The Linguistic Variables Two linguistic variables are selected in this study. The first is retroflex initials ( zh ) , ( ch ) , ( sh ) and the second is retroflex fricative initial ( r ) . A basic Chinese sylla ble consist s of an initial consonant, final s, and a tone. Both variables function as initial consonant s, followed by finals. In standard Mandarin, retroflex initials are articulated with a complete retraction of the tongue, moving from the back of the oral cavity. D ue to the influence of local dialects, the realization of retroflex initials may vary. TM has a relatively b roa der range of realizing retroflex initials (zh) (ch) (sh) . The present study adopts Brubaker ’s (2012) three alternations , including the full r etroflex ([ ], [h], []), the palatealveolar ([t ], [t ], [ ]), and the full dental ([ts], [tsh], [s]). T he second variable retroflex initial (r) is realized as a retroflex fricative [ ] and lateral [l] in the speech of bilingual speakers of Mandarin and Taiwanese. Participants The participants in this study were recruited from a Taiwanese immigrant community in Gainesville, a college town in Nort h Florida. The population of the target

PAGE 70

70 immigrant community consists of students who come to the US to pursue their Master or PhD degree in University of Florida and the immigrants who have resided and settle d down with their family in the community for a long time. The average length of stay for student population is from 1 to 5 years, and for the immigr ants is usually over 5 years. The following section lays out general information regarding the criteria for selecting participants for individual as well as group interview s, and the procedure for r ecruiting p articipants for both types of interview s. Lastly, the major data collection technique adopted in the present study is presented . Criteria for Participants A total of 30 participants were selected for this study. Thirty participants were interviewed individually. Five participants were randomly selected from those 30 participants , and were interviewed together with their PTH colleague or friend. E ach interview began with the interviewer briefly explaining the purpose of the interview, and acknowledg ing that participants had right to withdraw from the study before they signed the consent form (Appendix A) . After each interview, as an expression of gratitude to the participants who voluntar i ly participated in the interview without compensation, they were each given a thank you gift. T he following subsections describe the recruitment of participants for individual interview s, then the recruitment of participants for group interview s. Individual interview The purpose of conducting individual interview s was to examine the variation of the retroflex variables (zh) (ch) (sh) (r ) in spontaneous speech with the researcher , who was also a TM speaker . Participants were required to meet certain demographic criteria. They were required to be born and rais ed in Taiwan, and their first language must not

PAGE 71

71 be Mandarin. The main communication medium in the family had to be Taiwanese or other dialects. P articipants who were bilingual Mandarin/ Taiwanese were preferred. However, participants who were speakers of other dialects such as Hakka, but were able to use Taiwanese to communicate with other Taiwanese speakers were also considered. A total of 30 participants were interviewed for this study , including 28 bilingual speakers of Mandarin and Taiwanese, and 2 female bilingual speakers of Mandarin and Hakka. Two male participants were excluded. The first male participant was born in Taiwan but moved to Japan at the age of three; thus, he acquir ed Japanese as his first language, Taiwanese as a second language. Japanese and Taiwanese were the main forms of communication in the family until he studied Mandarin at the age of seven when his family moved back to Taiwan. In Japan, Japanese was the main form s of daily communication for him, so his Taiwanese and Mandarin might be affected by Japanese . Due to these reasons , he was removed from the data. The second male participant was excluded due to poor sound quality of his recording . In addition, in his speech, the realization of retroflex variables (zh) (ch) (sh) (r ) did not show a broader range, as other TM speakers did. In his Mandar in, the alternation between full retroflex variants [ ], [ h], [] and dental variants [ts], [tsh], [s] was more salient than that between the full retroflex variants [ ], [h], [] and the palatoalveolar variants [t ], [t ], [ ]. This tendency sugge sts that his Mandarin might be under great interference from Taiwanese phonological features. Therefore, he was also removed from the data. The total number of participants for the quantitative analysis was 28 in individual interview s.

PAGE 72

72 G roup interview The purpose of conducting group interviews was to investigate the variation of the retroflex variables (zh) (ch) (sh) (r ) that TM speakers would produce when they had an interaction with PTH speakers. This study focused on the linguistic production of TM speakers when they interacted with PTH speakers. Hence, PTH speakers’ dialect background was not taken into consideration. Each group interview included 1 TM speaker, 1 PTH speaker ( the TM speaker ’s friend or colleague) , and the researcher. Recruitment of Participants Individual interview T he target immi grant community had a small population , with community members constantly being replaced by new arrivals. P articipants had to meet demographic criteria such as place of birth and dialectal background, so random sampling was not considered in this study. Employing judgment sampling was more appropriate for the purpose of this study because the researcher could focus on a smaller sample size and select speakers based on personal knowledge of the communities. In addition to judgment sampling, using a friendof a friend recommendation recruitment strategy helped extend the researcher’s existing network and select potential people who fulfilled the criteri a for the purpose of this study . The demographic information of the par ticipants in individual interview s is presented in Table 4 1 . Group interview After each individual interview, TM participants were asked if they had PTH speakers in their network. If they did, they were asked whether they could invite their PTH friends or colleagues to have a group interview. Five TM participants responded

PAGE 73

73 that their PTH friends/colleagues were willing to have a group interview. Thus, in this study, there were 5 group interviews. As mentioned earlier, discussion was limited to linguistic behaviors of TM speakers , so only TM participants’ demographic information is included in Table 4 2 . Data Collection I t is suggested that natural speech is the speech that speakers pay minimal attention to when they are encouraged to narrate their own stories. Therefore, the more spontaneous speech speakers use, the more linguistic variables the researchers can capture (Bayley 2002; Gordon 2006; Milroy and Gordon 2003; Schiffrin 1994; Tagliamonte 2012) . In this study , t he data was collected through sociolinguistic interviews to elicit TM speakers’ spontaneous speech. T he interviews lasted about an hour and fifteen minutes for individual interviews, and about an hour for group interviews. T o elicit spontaneous speech in individual interview s, a set of specific topics was preselected that included general interests and personal experience related to TM participants’ everyday li v e s. During the interview, they were encouraged to narrate their personal experience , such as their school or work life. The researcher’s role was to engage them in conversation, to keep the conversation flowing as freely as poss ible , and to control their speech in order to gather the linguistic variabl es needed for this study . However, it is inevitable that in sociolinguistic interview s we may encounter the observer’s paradox (Labov 1984) due to t he presence of the researcher in both individual and group interviews. To minimize the effect of the observer’s paradox, the topics of the interview questions focus on participants’ general interests or personal experience, which may reduce their anxiety when they know they are being recorded in

PAGE 74

74 the interview. There was also built in flexibility for the researcher to adjust the topics emerging from the conversation, to encourage the participants to be more engaged in the conversation and produce more vernacular speech. After the individual interview, TM participants were asked to complete a questionnaire (Appendix B) , which consisted of three parts: demographic information, ranking strength of social netw ork with their TM friends (or colleague) and PTH friends (or colleague) , and openended questions regarding language attitudes. Since another purpose of the individual interview was to investigate how TM speakers construct their identity through their linguistic preferences, questions such as language attitude towar ds Taiwanese, standard Mandarin, Putonghua, and Taiwan Mandarin, and the correlation between these languages/dialects and national identity as Taiwanese identity versus Chinese identity , were included. The main purpose of conducting group interview s was to observe the interaction between TM and PTH speakers in a group context . In p articular , we could examine the variations of the retroflex variables in the speech of TM speakers, and whether TM speakers would increase their use of the retroflex feature in group interview s. In order to elicit TM speakers’ spontaneous speech in group interviews, preselected nonsensitive , general topics were used, such as school, work or family life, educational systems in China and in Taiwan, sports, or personal experience abroad. Sensitive topics such as political issues between China and Taiwan were avoided because it might result in possible conflicts between TM and PTH participants. Each group interview included 1 TM speaker, 1 PTH speaker, and the researcher. The role of the researcher was to initiate the conversation, but not engage in the conversation.

PAGE 75

75 Data Analysis T his sub section presents data analysis procedures , including data transcription and how speech corpus was extracted, coded, and analyzed. Lastly , the statistical methods applied in this study are described . Data Transcription The length of both individual and group interviews was bet ween fifty minutes and one hour and fifteen minutes. The transcription and coding began about fifteen minutes after an interview began because this might be the appropriate time when a participant felt more relaxed and comfortable talking spontaneously. Each participant was labeled using the abbreviation of Participant (P ), gender (F;M), and number —for example, P F1 instead of a pseudonym. Data Extraction and Coding For retroflex initials (zh) (ch) (sh) (r) , each occurrence of words containing these variables was coded. Chinese is a language with a high frequency of homonyms. For example, a one syllable word with the same pronunciation can denote different meanings . [ ] ‘ to know ’ has the same pronunciation as [ ] ‘juice’. S uch cases were coded as two separate tokens because they denoted different lexical meanings. When two words had the same pr onunciation and denoted the same lexical meaning, the y were coded as one token. Additionally, the immediate linguistic context (prec eding and following word) of the retroflex initials (zh) (ch) (sh) (r) were taken into consideration. For example, r n [ n] ’ person; people’ can be combined with another character to be a compound word such as b nr n ‘ patient ’ and r nji ‘ another person’ . S uch cases were coded as separate tokens because the words preceding and

PAGE 76

76 following r n are different . And because words beginning with retroflex initials (zh) (ch) (sh) (r) are commonly found in Chinese, to limit the number of the occurrence of a word, the first five tokens from each occurrence of the one syllable word were coded, as well as words with more than two syllables if they appeared in the same variable context. The words containing the variables under investigation w ere highlighted in each participant’s transcript on Excel spreadsheets. The major approach to code linguistic variations in this study was impressionistic judgment , which relied on the researcher’s perception of linguistic variants. In order to have a reliability and accuracy check, a native TM speaker was hired as a co coder1 to code the data blindly. After coding was completed and the results saved, those results were erased, and t he Excel spreadsheets sent to the co coder along with sound files. After the cocoder completed her coding, her results were compared with the original impressionistic judgments . For each participant, the number of tokens for the (zh) (ch) (sh) variants was approximately 350 to 500 in an individual interview and 220 to 430 in a group interview. B ecause the total number of tokens for (zh) (ch) (sh) variants was over 10, 000 , the tokens were removed when the coding had a disagreement. The total number of tokens for (zh) (ch) (sh) variants was 12 , 718. Regarding the (r) variable, the number of tokens from each participant was approximately 80 to 100 in an individual interview and 50 to 100 in a group interview. When there was a disagreement, the sound file from each interview was re reviewed to re check the disagreement. If the 1 The cocoder majored in Foreign Language and Literature, and knowledge of linguistics was one of the main courses in her undergraduate program.

PAGE 77

77 disagreement still could not be settled, the tokens were removed. The total number of tokens for (r) was 1 , 428. The Extralinguistic Variables In this study, four extralinguistic factors —age, gender, social network strength , and length of stay in the US —are used in analyzing the effect of extralingusitic factors on the variation of retr oflex initials (zh) (ch) (sh) (r) in the speech of TM speakers in an immigrant setting and when they contact with PTH speakers in an immigrant community . Generation Age represents an individual’s life stage and experience of life span. As Eckert (1997: 166) points out, “to the extent that social and political events can affect the way people speak, age differences in variation can reflect social and political changes.” People from the same age range may experience the same major sociohistorical e vents , which possibly trigger linguistic changes in their speech patterns. Hence, in this study, grouping speakers by different generations under consideration for their shared experience in political and historical events may explain how linguistic norms undergo changes across different generations . The majority of members in the target immigrant community pursue their m aster or doctoral degrees after their graduation from undergraduate or m aster ’s programs in Taiwan. A university student in Taiwan normally graduates from undergraduate programs at the age of 22 years old or older , or finishes a M aster degree at the age of 25 years old or older. Thus, t he participants in the young generation in this study ar e age of 22 or older . The participants are classified into three generational groups as follows.

PAGE 78

78 Young generation : < 35 years old Middle aged generation: 36 45 years old Old er generation: > 46 years old The cl assification of three generational groups corresponds to different historical contexts in Taiwan. T his study examines whether different historical contexts affect the use of the varieti es of Mandarin and local dialect – standard Mandarin, Taiwan Mandarin, and Taiwanese. The young generation, ages 22 to 35, started school a few years before or after t he lifting of martial law in 1987. The participants in the young generation grew up in the period when local dialects were less restricted in public domains, and using local dialects were encouraged to present their ethnic root s. The participants aged 3645 are classified as the middleaged generation . T he participants in this generation were born before the liberation of the Mandarinonly policy in 1987. When they started schools, speaking local dialects was forbidden . I f students used local dialects in the classroom, they would be fi ned and asked to wear a tag with a caption of “Mandarin Only ” . T he participants in this generation grew up at the turning point when Taiwanese society was under social political changes. They not only witnessed but also participated in the process o f the social and political liberalization in Taiwanese society and enjoyed the liberty from socio political changes and the benefits from rapid economic growth. The participants older than 46 years belong to the old er generation. The participants in this generation w ere born after 1949 and grew up in the period when the Mandarinonly policy began and was implement ed overwhelmingly in all public domains and schools. The successful implementation of the Mandarin only policy gradually affected Taiwanes e people’s language attitude toward the varieties of Mandarin and dialects on the island. T he varieties of Mandarin, i.e. standard Mandarin

PAGE 79

79 and Mandarin with local dialect features , become “a social register identifying social status or geographical identity” (Li 1985: 126) , which inevitably results in diglossic situation in Taiwanese society. That is, the standard Mandarin was prestigiously valued, whereas Mandarin with a trace of local dialect features and local dialects were less valued. Gender Gender has been one of the most important extralinguistic variables in the field of sociolinguistics since Labov (1966) first point ed out that men and women present different linguistic behaviors in terms of the preference for standard and nonstandard forms. Trudgill’s (1972) study in Norwich further confirmed Labov’s observation. That is, women prefer standard forms more than men , who prefer nonstandard forms. W omen’s preference for the standard or prestigious forms is to manifest their social position in response to their lower social status in the society. Labov (1998) later notes that women have a tendency to adopt an innovative form quicker than men, and further lead in linguistic changes. Moreover , Milroy and Milroy (1997) argue that in addition to style and prest ige, identity a lso results in linguistic variations between two sexes. They find that when a linguistic variant is associate d with local identity inside a community, men prefer the variant that denotes solidarity and localness , while women prefer a supra local one. Earlier s tudies regarding the use of three linguistic varieties in Taiwanese society —standard Mandarin, Taiwan Mandarin, and Taiwanese— find that besides styl e and prestige, there is a significant correlation between gender and the three linguistic varieties (Brubaker 2012; Chung 2006; Liao 2010; Su 2011). The researchers indicate that TM speakers are sensitive to the varieties they use in different occasions because these varieties are usually linked with a particular kind of sociol inguistic status and can

PAGE 80

80 project their ethnic identity. Although TM speakers do not explicitly express the social meanings behind these varieties, these social m eanings and ideologies are found in their actual linguistic performance. In their findings, f emale TM speakers prefer standard variants because they would like to avoid using Mandarin with Taiwanese phonological features or Taiwanese due to their association with the negative social meanings in Taiwanese society. On the other hand, male TM speakers prefer Mandarin with Taiwanese phonological features or Taiwanese due to their relation with localness and solidarity. Thus, in the present study, gender is an important factor to be examined in terms of its correlation with the linguistic variables. Social Network Strength Social network strength has been considered as an influential extralinguistic factor affecting speakers’ linguistic behaviors. Milroy and Milroy (1978) point out that a community with a close knit social network tends to sustain homogene ous linguistic features and resist changes from the impact of outside network s in order to maintain local community norms. On the contrary, a community with a looseknit network tends to be open to change under the influence of an outside network. As a result, the transmission of innovated linguistic chang es and resources is often observed in this type of network. Unlike immobility and close knit social network s in Milory’s Belfast study, members in the target immigrant community are replaced constantly by new arrivals , so the network structure inside the community is considered a looseknit network. Due to the nature of the target community, it is not completely appropriate to adapt Milroy’s network measure. Thus, to measure an individual speaker’s strength of his/her social network in the target community, adapting Hirano’s (2008) network measure, which is

PAGE 81

81 concern ed with the degree of closeness and the frequency of contact, is more pert inent to the n ature of the community. The adapted measure is presented as follows. The participants are asked to name individuals in their personal network, define their relationship ( i.e. friends, office colleagues, acquaintances ) , and rank each individual according to the degree of closeness by using a numerical value ranging from 1 to 5. A score for each relationship is calculated based on the given rank order of closeness. For example, when a named individual is ranked as the first in the order, the highest score, 5 , is given. On the other hand, when an individual is ranked as the last, this individual is given 1 point. Furthermore, with regard to the frequency of meetings and telephone calls, the participants write down the frequency o f telephone calls within a week for each named individual. A score for the frequency of meetings and telephone calls ranges from 1 to 5. For example, when the frequency of meetings and telephone calls reaches over 5 times a week, the highest score, 5 , is g iven. On the other hand, when the meetings and telephone calls only occur once a week or never , the relationship gains the lowest score , 1. Length of Stay in the US Hirano (2008) suggests that long term accommodation can also be observed when dialectal speakers move to an immigrant setting in which they come into contact with speakers from other varieties of the same language. I t is suggested that in an immigrant setting, mobile speakers accommodate to non mobile speakers and acquire the host language as their stay extends , and there is no dominant dialect in this type of setting. Hence, the direction of accommodation cannot be determined. Previous studies (e.g. O theguy et al. 2007) show that social factors such as longer residence time and

PAGE 82

82 pre stigious status of one dialect may affect speakers’ attitude towards their own dialects and other varieties of the same language. Speakers ’ attitude towards certain dialects further affect s speakers’ linguistic behaviors. Hock and Jos eph (2009) point out that their loyalty to their homeland dialects diminish es gradually during the accommodation process. Therefore, the present study includes length of stay in the US as one of the extralinguistic variables , to examine whether different residence time in the US is a factor affecting their linguistic feature patterns . In th e present study, the participants are classified into four groups according to their length of stay2 in the US. Less than 6 mont hs 1 2 years 3 4 years Over 5 years Statistical Methods T h e major statistic tool in this study is Goldvarb X program , which is also referred as “variable rule program” (Tagliamonte 2012: 121). It is a software that uses logistic regression to provide a generalized linear model. The employment of Goldvarb X program can examine if the occurrence of the variants of (zh) (ch) (sh) (r) models in contextual conditioning . The design of Goldvarb is to capture th e variability of natural speech analyzes the results by using “ the three lines of evidence” (Poplack and Tagliamonte 2001 : 92; Tagliamonte 2012:122 ). First, the first evidence is st atistical significance , which demonstrates which factors are statistically significant at the 0 .05 level and which are not . 2 In general , Taiwanese students studying in the US usually take 1 to 2 year s to complete a Master ’s program, and above 4 years in a PhD program. For immigrants, they may have been in the US for over 6 years. Therefore, the shortest length of residence time in the US could be less than 6 months, and the longest length of residence time could be over 5 years.

PAGE 83

83 The second evidence is magnitude of effect , which shows the relative strength of each factor group. The order of relative strength of each factor group is determined by which factor group has the highest range of factor weights and which has the lowest ones. The factor group with the highest range suggests the strongest constraint, whereas the lowest range identifies the least constraint. The third evidence is the hierarchy of constraints , which presents the ra nking of factor weight of given factors within a factor group. The constraint hierarchy reveals “the detailed structure of relationship between variant and linguistic context, or the ‘grammar’ underlying the variable surface manifestation” (Poplack and Tag liamonte 2001: 94). It not only shows how much a given factor weights in the distribution of the dependent variables but also presents the interaction between each factor within a factor group. With the three lines of evidences, Goldvarb can provide researchers an effective argument to explain the relationship between variant and context. Also, i t is suggested that Goldvarb is the appropriate tool for the study of sociolinguistics and language variation (Guy 2010; Tagliamonte 2009). Therefore, t he em ployment of Goldvarb allowed this study to identify the significant factors that affected speakers’ linguistic choice of (zh) (ch) (sh) (r), and helped interpret the interactions of effects between different factor groups. This chapter has provided the study design for this research. It has described the two linguistic variables under examination and the four extralinguistic factors investigated in this study. Information regarding the sample and the participants w as also provided. A description of the data collection procedures for individual and group interviews and decisions regarding data transcription have been presented. Finally ,

PAGE 84

84 methodological procedures regarding extraction and coding, as well as the statistic al methods used , were discussed. In the next chapter, analysis of the distribution and variation of retroflex initials (zh) (ch) (sh) (r) is provided, in view of their correlation with extralinguistic factors .

PAGE 85

85 Table 4 1 . Participants’ demographics in individual interviews Generation N Gender N Strength of social network N Length of stay in the US N Young 16 Male 12 Strong 10 Less than 6 months 4 Middle aged 7 Female 16 Weak 8 1 2 years 4 Older 5 No Contact 10 3 4 years 4 Over 5 years 16 Table 42. TM Participants’ demographics in group interviews Generation N Gender N Strength of social network N Length of stay in the US N Young 3 Male 2 Strong 2 Less than 6 months 1 Middle aged 2 Female 3 Weak 2 1 2 years 1 No contact 1 Over 5 years 3

PAGE 86

86 CHAPTER 5 RESEARCH RESULTS AND INTERPRETATIONS This chapter reports a quantitative analysis from the perspective of variationist analysis in terms of the variation of the retroflex initials ( zh ) ( ch ) ( sh ) from both individual and group interviews. T he research hypotheses are addressed through out the chapter , followed by an analysis of the correlation between the retroflex initials (zh) (ch) (sh) and extralinguistic factors to support or reject t he present study’s hypotheses. The results from the individual interview are presented first, followed by the results from the group interview. Vari ation in Retroflex Initials ( zh ) (ch ) ( sh ) Previous studies have shown that the formation of Mandarin in Taiwan is the linguistic outcome of the heavy influence from Taiwanese and other local dialects and foreign languages. N ative Taiwanese speakers carry f eatures from their first dialect, which affects the way they speak Mandarin (Cheng 1985; Chung 2006; Her 2010; Kubler 1985; Li 1983; Li 1985). A typical feature in TM is lack of the retroflex feature zh [ ] , ch [ h], sh [ ], and r [ ] , which is the distinctive f eature in s tandard Chinese . Chung (2006) argued that TM has a relatively b r oader range of realizing the retroflex (zh) (ch) (sh) as initial consonants, i.e . initials. T he retroflex (zh) (ch) (sh) can be pronounced by TM speakers with co mplete retraction of the tongue , moving from the back of the oral c avity ( [ ], [h], [ ] ) through the palatoaveolar area ( [t ], [t ], [ ] ) or to the dental area ( [ts], [tsh], [s]) . Besides the different linguistic realization, the variations in producing the retroflex initials are also associated with covert prestige. The retroflex initials with full retraction of the tongue, i.e. the full retroflex feature, have been considered as a “canonical index” (Chung 2006:198). The full retroflex feature often

PAGE 87

87 serve s as a stylistic device in different contexts , such as academic conferences v ersus family gatherings, and is also associated with positive social evaluations like high socioeconomic background and educational level. The palatoal veolar feature i s widely accepted as the normative form and the denta l feature is considered as a s ymbol of solidarity in Taiwanese society (Brubaker 2012; Tse 2000). In light of the variations of the retroflex initials and their association with social meanings, the p resent study applies sociolinguistic variationist analysis to explore the correlation between the variants of the retroflex initials (zh) (ch) (sh) and TM speakers’ linguistic behaviors in an immigrant community. It further aims to uncover whether or not t he extralinguistic factors of strength of social network, generation, gender, and length of stay in the US play a role that affects the occurrence of the variants of the retroflex initials (zh) (ch) (sh) in TM speakers’ speech. Overall Distribution of Retroflex Initials Variants This section presents the results of the retroflex initials from 28 individual interviews. A total of 12,718 tokens have been extracted to examine TM speakers’ speech patterns with regard to the correlation between the distribution of the variants of the retroflex initials (zh), (ch), (sh) and four extralinguistic factors. Table 51 shows the overall distribution of the retroflex initials (zh) (ch) (sh). The production of the palatoalveolar variants had the highest frequency at 94. 7%, followed by the use of the full retroflex variants at 5.2% . The dental variants had the lowest frequency at 0.1%. It was initially hypothesized that the dental variants would be the second favored variant among TM speakers due to the lack of the retrof lex feature in native Taiwanese speakers ’ repertoire. Native Taiwanese speakers were commonly found to use t he dental variants in substitution for the full retroflex variants when they

PAGE 88

88 acquired Mandarin as their second language (Kubler 1985; Li 1985). Howe ver, as shown in Table 51, the dental variants had the least frequency in the present study. The small percentage of the dental variants resulted in knockout1 so the dental variants were re coded and placed in the group of palatoalveolar variants because the two features were in a similar natural class, and were also linked with similar social meanings in Taiwanese society . After recoding, the overall distribution of retroflex initials is presented in Table 52. It showed that the palatoalveolar variants had the highest frequency, 94.8%, in TM speakers’ speech. T he distribution of the retroflex initials (zh) (ch) (sh) was examined in terms of their correlation with four extralinguistic factors. T he pattern and the tendency are presented below. Strength of Social Network The framework of social network predicts that speakers’ social network affects their linguistic preferences. S peakers who have greater strength of social network with speaker s of other varieties tend to avoid using their own language varieties or to adopt the speech of others to reach convergent communication ( Milroy 1980) . Following this line of reasoning , this study had postulated that TM speakers with greater strength of social network with PTH speaker s would produce more variants of the full retroflex feature, the distinctive feature in PTH speakers’ speech, due to their frequent contact with them. On the other hand , it was predicted that TM speakers who had less strength 1 Knockout refers to zero percent value in statistical results in Goldvarb X program. It is suggested to recode the variant that shares a similar property (Tagliamonte 2006).

PAGE 89

89 of social network with PTH speakers would produce more variants of the palatoalveolar feature, which is the particular linguistic feature in TM . Table 53 presents the overall distribution of the retroflex (zh) (ch) (sh) regarding TM speakers’ strength of social network with PTH speaker s. As seen in Table 53, r egardless of the degree of social network strength with PTH speakers, a large p roportion of the palatoalveolar variants was found in TM speakers’ speech. T he speakers who had greater strength of network with PTH speakers produced a higher frequency of the palato alveolar variants, at 96.1%. In contrast, the speakers who had less strength of network with PTH speakers produced a lower frequency of the palatoalveolar variants, at 94%. Figure 51 illus trates that the palatoalveolar variants were the most dominant variants in TM speakers’ speech. Although the statistical results show that the use of palatoalveolar variants occupied the highest frequency and was the predominant variant in TM speakers’ s peech, the results from the frequency of the full retroflex variants could still provide us with some nuanced differences in regard to TM speakers’ use of the full retroflex variants between two groups. The speakers who had greater strength of network with PTH speakers produced t he lowest occurrence rate of using the full retroflex variants, 3.9%. On the other hand, t he speakers with less strength of network with PTH speakers demonstrated the highest frequency of the full retroflex variants at 6.0 %. The statistical difference regarding the use of the full retroflex variants between the two groups suggests that the speakers who had greater strength of network did not prefer the use of the full retroflex variants, whereas the speakers who had less strength of network did prefer the full retroflex variants.

PAGE 90

90 The findings did not conform to Hirano’s (2008) study , which investigated t he use of intervocalic (t) among English speakers from different Englishspeaking count r ies in an immigrant community in J apan . Her study revealed that speakers who had greater strength of network with speakers of other varieties tended to avoid usi ng their own language varieties to reach convergent communication. The results from the negative correlation between TM speakers’ network strength and their choice of variants of the retroflex (zh) (ch) (sh) was unexpected. This study initially postulated TM speakers’ strength of network with PTH speakers would have a hig her frequency of the full retroflex variants due to their frequent contact with PTH speakers. With their stronger network with PTH speakers, TM speakers’ preference for the use of the full retroflex should have been found from the results. This finding suggests that the use of the full retroflex feature indicates different communicative intentions between two groups, and may have some hidden values that are not expressed. Furthermore, TM speakers with greater strength of network with PTH speakers may be more aware of the values than the speakers who had less strength of network with PTH speakers. It is likely that the hidden values emerge through the interaction with PTH speakers and are only accessible to the speakers with greater strength of network PTH speakers. The different communicative patterns found between the two groups further suggests that the direction of dialect accommodation in an immigration setting may not be determined by the strength of TM speakers’ network nor the frequency of interaction with their interlocutors, but by other factors such as the sociopsychological factor.

PAGE 91

91 Generation It was hypothesized that a higher frequency of the full retroflex variants would be found in the young generation. TM speakers in this generation are mostly bilingual speakers of Taiwanese and Mandarin, use Mandarin as their main communication medium in daily conversation between peers, and m ay have less influence of Taiwanese phonological features than the older TM speakers (Liao 2010) . The middle aged generation would also produce a higher frequency of the full retroflex variants due to the fact that the speakers in this generation grew up in the period when Mandarinonly policy was implemented overwhelmingly in Taiwanese society . For the older generat ion, it was posited that the speakers in this generation would have a higher frequency of the palato alveolar variants and a lower frequency of the full retroflex variants due to the fact that they might have a heavier influence of Taiwanese than the other two generations. For statistical analysis, the middleaged generation was collapsed with the older generation. As seen in Table 54, the palatoalveolar variants were the dominant variant in all generations. The young generation had a lower frequency at 91.5%, while the older generation had a higher frequency at 98.6%. The statistical difference of the use of the palato alveolar variants between two generations shows that the influence of Taiwanese phonological features on the older generation remained and was greater than the young generation. Despite the skewed distribution between the palatoalveolar variants and the full retroflex variants in TM speakers’ speech, the statistical difference in the frequency of use of the full retroflex variants betw een the two generations could also provide subtle

PAGE 92

92 meanings with regard to TM speakers’ use of the full retroflex variants by different generations. The young generation had a higher tendency to use the full retroflex variants, at 8.5%, as was hypothesized. On the other hand, the older generation had a much lower rate of using the full retroflex variants, at 1.4%. The different rate of the full retroflex variants between two generations further confirmed that speakers of the two generations had different degrees of influence of Taiwanese on their use of variants of the retroflex (zh) (ch) (sh). The above findings thus support the hypothesis that a higher frequency of the full retroflex variants would be found in the younger generation. Figure 5 2 illustra tes the distribution of the retroflex initials by different generations. Gender Previous studies ( e.g. Labov 1990; Trudgill 1972) have suggested that women t end to favor standard and prestigious form s more than men in order to manifest their social position in response to their less powerful social position in society. In contrast, men usually obtain prestigious social status through their achievement s from their work performance; therefore, proper language use does not serve as a means to obtain their social position . Women in Taiwanese society, however, are required to behave in certain way s, including being polite and well behaved, and talking in proper ways (without cursing ) . On the other hand, men’s language use does not have many restrictions from social pressure (Su 2008). In this vein, the present study hypothesized that TM female speakers would produce a higher frequency of the standard and pre stigious form, i.e. the full retroflex variants; whereas , men would have a higher frequency of the nonstandard form i.e. the palatoalveolar variants.

PAGE 93

93 Table 55 and Figure 5 3 show the distribution of the retroflex initials in terms of gender. As seen in Table 55, male speakers had a higher rate of producing the palatoalveolar variants, 99.2%, than female speakers, 91.2%. The distribution between male and female showed that the use of the palato alveolar variants is predominant in the speech of both male and female speakers. Com paring with the high rate of the palatoalveolar variants, the use of the full retroflex variants showed a relatively small statistical result. For the use of the full retroflex variants, the occurrence rate in the female group was 8.8%, while 0.8% in male group. Although the production of the full retroflex variants had a much lower rate, the statistical difference between male and female speakers still provided nuanced details regarding the choice of variants between male and female speakers. A higher frequency of use of the full retroflex variants in TM female speakers’ speech indicates that female speakers tended to favor the full retroflex variants, whereas their male counterparts did not because they produced a much lower frequency of the full retroflex variants. The statistical differences in use of the full retroflex variants support the hypothesis that TM female speakers would prefer the full retroflex feature, the standard form. Furthermore, when looking at the use of the palatoalveolar variants, a higher freq uency found in male speakers suggests that they preferred the palatoalveolar variants, the nonstandard form. The above findings reveal that the use of the palatoalveolar and the full retroflex variants are favored differently by male and female speaker s. This favorable attitude could attribute to different degrees of awareness of social status between Taiwanese men and women. Previous studies (e.g. Lee 1981) suggest that the ability to speak standard Mandarin reflects one’s socioeconomic background and educational level in

PAGE 94

94 Taiwanese society. Particularly, the proper use of the full retroflex feature is usually associated with positive sociocultural evaluations of the speaker, such as being well educated, sophisticated, and presenting a higher social stat us. In contrast, being unable to use the full retroflex feature often links speakers with negative evaluations such as being unsophisticated and of a lower social class. As Eckert (1989) points out, when a linguistic variable denotes a stable social meaning which holds overt social values such as prestige and power in the society, females tend to be more conservative, producing a higher frequency of a certain variable which preserves the traditional social meanings. TM female speakers’ preference for the full retroflex variants echoed Eckert’s observation. In addition, when female speakers were asked about their language attitude toward the use of the full retroflex feature in the interview4, the majority reported that they were conscious that they retracte d their tongue back frequently to produce more full retroflex feature when talking to their PTH interlocutors. On the other hand, although some male speakers reported that they noticed the full retroflex feature in PTH speakers’ speech, they generally comm ented that they did not change the way they spoke , such as switching to use full retroflex feature, when talking to their PTH interlocutors , and also stated that there was no need to use the full retroflex feature to accommodat e to their PTH interlocutors. Both male and female speakers’ comments regarding their attitude towards the accommodative preference reflect their choice of 4 T he topics in the sociolinguistic interview focused on participants’ general interests and personal experience in order to elicit spontaneous speech. Besides general topics commonly used in sociolinguistic interviews, ques tions were included regarding their consciousness of the differences between PTH and TM to examine whether they were aware of a salient feature of PTH in comparison with TM.

PAGE 95

95 variants. That is, a high frequency of the full retroflex variants is found in female speakers’ speech, and a high frequency o f the palatoalveolar variants in male speakers’ speech. More recent studies (e.g. Brubaker 2012; Baran 2014) revealed that the attitude towards the use of the full retroflex feature has been changed inside Taiwanese society. They found that consistent us e of the full retroflex feature in speech is affiliated with negative sociocultural impressions. For example, speakers feel uncomfortable when the ir interlocutor s frequently use the full retroflex feature, and further think that their interlocutors are showy . In addition, the attitude towards the palatoalveolar variants (and dental variants) has also changed due to the rise of Taiwanese localism. The use of the palato alveolar feature (and dental variants) is s ignalin g group identity and Taiwanese identity. Although the use of both the full retroflex feature and palatoalveolar feature now denotes conflicting sociocultural meaning s, and TM speakers have ambivalent feelings about their use , previous studies in general found that TM female speakers favored and used the full retroflex feature more frequently than male speakers , and also avoided using colloquial forms when the immediate settings were perceived as formal situations. The palatoalveolar feature (and dental feature) was favored by TM male speakers (Brubaker 2012; Chung 2006; Lin 1988). In addit ion to the extralinguistic factors of social network strength, generation, and gender, length of stay in the US was also examined to determine if it affected TM speakers’ choice of the retroflex variants. Hock and Joseph (2009) pointed out that speakers maintain ed the ir local linguistic patterns when they first arrived in a new

PAGE 96

96 country. Also, their loyalt ies to the homeland dialect features became minimal after extension of the stay and intensive contact with speakers of other dialects and different la nguages. As a result, they avoided using homeland dialectal features when interacting with other dialect speakers to reach convergent communication. Following this reasoning , this study hypothesized that when TM speakers’ residence time in the US was extended, they would prefer the full retroflex feature, which was the prescribed form taught in schools and associated with Chinese value and culture. On the other hand, the speaker who had a shorter residence time in the US would produce more palatoalveolar v ariants, which was the nonprescribed form and associated with Taiwanese identity and Taiwanese local value (Brubaker 2012). Table 56 presents the overall distribution of the retroflex initials in regard to the length of stay in the US. As seen in Table 5 6 and Figure 54, the speakers who had longer residence time in the US (over 3 years) had a higher frequency of the full retroflex variants, 5.8%, while the speakers who had stayed in the US less than 2 years had a lower frequency, 4.4%. This supports th e hypothesis that TM speakers who had stayed in the US for a longer time would have a higher frequency of the full retroflex variants. Meanwhile, this group also had a lower frequency of the palatoalveolar variants, 94.4%. As seen in Figure 54, the production of the palatoalveolar variants in the group with a relatively shorter residence time in the US only exceeds the group with a longer residence time in the US by 1.2%. Although the statistical difference between the two groups is small, it still sugge st s that the speakers who have a shorter residence time in the US tend to favor the palatoalveolar feature.

PAGE 97

97 As hypothesized, a higher frequency of the full retroflex variants was found in the group with a longer residence time in the US . T he palatoalveol ar variants that presented hom eland linguistic features were found at the least frequency in th is group. This result indicates that the speakers who have a longer residence time in the US prefer the full retroflex feature. Furthermore, in the interviews, TM speakers were asked a question regarding the correlation between the length of stay in the US and their awareness of Taiwanese identity . T he speakers with longer resi dence time in the US generally stated that as their residence time in the US was ext ended, their sense of identifying themselves as Taiwanese became stronger. Their responses might lead us to assume that a higher frequency of the variants that represented Taiwanese local identity would be found in their speech. Yet, as Table 5 6 shows, this was not the case. S peakers who had stayed in the US for over 3 years had a higher frequency of the full retroflex variants instead of the palatoalveolar variants. A plausible explanation could be related to the nature of Chinese immigrant community. Skel don (2004) argued that inside the Chinese community, the places of origin among the community members were various and the members might perceive Chinese identity differently, but for self protection against the perceived hostility from the host country, t he solidarity was emphasized and the ideology of being “all Chinese” (2004: 45) prevailed. T he positive correlation between length of stay in the US and higher frequency of full retroflex variants in the present study suggest s that the same trend occurs in the target community . That is, TM speakers’ sense of belonging in the local Chinese immigrant community, which views Chinese culture as the social standards, affects their use of a certain linguistic pattern in their speech. A similar pattern is found in t he work of

PAGE 98

98 Otheguy (2007) and his associates, who investigated the accommodation phenomenon in a Spanish immigrant community comprising speakers of Spanish dialects in New York City. They found that in this community, the speakers were prone to use Car ibbean Spanish . W ith a longer residential history in the community, the Caribbean Spanish speakers enjoyed prestigious local status. The positive social value of Caribbean Spanish emerging from the community led speakers of other Spanish dialects in the co mmunity to accommodate to Caribbean Spanish instead of to the Mainland Spanish6. M ultivariate R egression A nalysis A multivariate regression analysis was performed to discover the significance of the specific factor group. Factor groups include one dependent variable and four independent variables: speakers’ strength of social network, generation, gender, and length of stay in the US. The results from the binominal onelevel test revealed that all four extralinguistic factors had statistical significance (p=0.05) affecting the speakers’ choice of the full retroflex variants. Table 57 presents the multivariate analys is of eac h extralinguistic factor . In terms of the range value, gender held the highest value, 68, whereas the strength of social network had the lowest value, 24. This result showed that among the factor groups affecting the speakers’ production of the f ull retroflex variants gender displayed the strongest magnitude of effect , while the strength of social network displayed the weakest magnitude of effect. The following r anking summarizes the order of each factor group by the values ( the most important factor is on the left ) . 6 See Chapter 3 for O theguy’s classification between six Spanish dialects of Latin/cent ra l America.

PAGE 99

99 Gender > Generation > Length of Stay in the US > Strength of Social Network Gender With a range of 68, gender displayed the strongest magnitude of effect . The constraint ranking of gender showed that female speakers tended to favor the full retroflex variants, with a factor weight of .82, while their male counterparts tended to disfavor them, with a factor weight of .14. This discrepancy is shown in Figure 55. This result support s the hypothesis of the present study that female speakers preferred the full retroflex in their speech. Generation Wi th a range of 66, generation displayed t he second strong est magni tude of effect . As shown in Figure 56, the young generation had a tendency to favor the full retroflex variants, with a factor weight of .81, while the older generation tended to disfavor the full retroflex variants, with a factor weight of .15. This result supports the hypothesis of the present study that the full retroflex variants were favored by the young generation. Furthermore, as seen in Table 54, the older generation produced a much lower percentage of the full retroflex variants (1.4%) than the young generation. This tendency suggests that the older generation may have a greater influence of Taiwanese on their use of the retroflex (zh) (ch) (sh). Figure 56 illustrates the effect of generation on the production of the full retroflex variants. Length of S tay in the US With a range of 44, length of stay in the US displayed a weak magnitude of effect affecting TM speakers’ production of the full retroflex variants. The constraint ranking showed that the speakers who had a longer residence time, over 3 years tended to favor the full retroflex variants, with a factor weight of .64. In contrast, the speakers

PAGE 100

100 whose residence time in the US was less than 2 years tended to disfavor the use of the full retroflex variants, with a factor weight of .20. Figure 57 illustrates the effect of length of stay in the US on the production of the full retroflex variants. Strength of Social Network With a range of 24, strength of social network displayed the weakest magnitude of effect affecting the production of the full retroflex variants. The constraint ranking showed that with a factor weight of . 65, the speakers who had greater strength of network with PTH speakers tended to favor the full retroflex variants. The speakers who had less strength of network with PTH speakers did not show the tendency to favor the full retroflex variants, with a factor weight of .41. Figure 58 presents the effect of strength of social network on speakers’ production of the full retro flex variants. The following section presents the results from the group interviews of 5 TM speakers in comparison with the results from their individual interview. Group Interview Five TM speakers (2M, 3F) were randomly selected from 30 individual interviews to participate in group interviews . Each group interview included 1 TM speaker, 1 PTH speaker (the TM speaker’s friend or colleague ) , and the researcher. The role of the researcher was to initiate questions but not engage in the conversation. The purpose of the group interview was to examine if TM speakers would increase the frequency of the full retroflex variants when talking to PTH speakers during the interview. For this purpose, the 5 speakers’ use of the retroflex (zh) (ch) (sh) were examined with regard to different types of interviews. The overall distribution of the variants of the retroflex (zh) (ch) (sh) in the group interview showed that there was a knockout in dental variants so they were recoded and collapsed with the palatoalveolar variants .

PAGE 101

101 The production of the retroflex (zh) (ch) (sh) variants from 5 TM speakers’ individual and group interviews was compared to examine whether frequency of the variation of the retroflex (zh) (ch) (sh) would be different. Table 58 presents the overall distribution of the retroflex initials by different types of interview. In the group interview, the production of the palatoalveolar v ariants had the highest frequency, 92.7%, while the occurrence rate of the full retroflex variants was at 7.3%. In the individual interview, the production of the palatoalveolar variants had the highest occurrence rates at 92.2%. The production of the ful l retroflex variants was 7.8%, in comparison with 7.3% in the group interview . The above results revealed that in the group interview, the frequency of the full retroflex variants was 0.5% decreased, whereas the occurrence rate of the palatoa l veolar variants was increased by 0.5% . Compared with the frequency of the full retroflex variants in the individual interview, the decrease of the full retroflex variants in the group interview suggests that the presence of PTH speaker seem to be a factor affecting TM speakers’ linguistic choice during the interview. The decrease of the full retroflex variants and the increase of the palatoalveolar variants in the group interview suggest that the two v ariants of the retroflex (zh) (ch) (sh) seem to be assigned particular social meanings by TM speakers. The meanings behind the variants are not expressed explicitly but when the results from the two types of interviews are compared, the meanings and ideologies emerge. In their interviews, TM speakers consider the palatoalveolar feature as a distinctive feature in their Mandarin, which is different from Mandarin in China i.e. PTH. Thus, for TM speakers, the palatoalveolar feature seems to be viewed as a T M feature

PAGE 102

102 and an ingroup speech feature. Conversely, according to their responses in the interview, the majority of TM speakers do not consider the full retroflex feature as a distinctive feature in their speech. Although they have to learn it in schools, it is seldom practiced in their daily speech. Therefore, the full retroflex feature seems to be viewed as a nonTM feature and an out group speech feature. In the group interview, the increase of the palatoalveolar variants, which carry local dialectal features and are affiliated with TM speakers’ place of origin, suggests that TM speakers did not accommodate to their PTH interlocutors. The following subsection provides the distribution of retroflex variants and the extralinguistic factors in the group interview in comparison with the individual interview. Distribution of R etroflex (zh) (ch) (sh) and the Extralinguistic Factors Generation The 5 TM speakers mainly included two generations: young (less than 35 years old) and middleaged (3645 years old). Table 59 shows the distribution of retroflex initials by these two generations in group and individual interviews. In the group interview, the young generation had a higher frequency of the full retroflex variants, 9.7%, whereas the middleaged generation had a lower frequency at 2.2%. A similar pattern was observed in the individual interview. The young generation had a higher frequency of the full retroflex variants at 12.6%, while the middleaged generation had a lower frequency of the palatoalveolar variants at 0.5%. Regarding the production of the palatoalveolar variants in the group interview, the young generati on showed a lower frequency at 90.3% in comparison with the frequency in the middleaged generation at 97.8%. In the individual interview, the young generation had a lower frequency at 87.4%, while the older generation had a higher frequency at 99.5%.

PAGE 103

103 Com paring the results from the two types of interviews showed that a decreasing rate of the full retroflex variants and an increasing rate of the palatoalveolar variants were found in the young generation. This suggests that in the group interview, the young generation may be less willing to accommodate to their PTH interlocutors, so they decrease the frequency of the full retroflex variants. Furthermore, the increase of the palatoalveolar variants suggests that the young generation is likely to use the palato alveolar feature to create inand out group speech in order to distance themselves from PTH speakers. A different picture was found in the middleaged generation. The decrease of the palatoalveolar variants and the increase of the full retroflex var ia nts in the middleaged generation suggest that the middleaged generation may be more willing to use the full retroflex feature when PTH speakers are present. Figure 59 illustrates the distribution of the variants of retroflex (zh) (ch) (sh) by generation and type of interview. Gender As seen in Table 510, in terms of the production of the full retroflex variants in the group interview, TM female speakers produced a higher frequency at 10.9%, while their male counterparts did not produce any full retroflex variants. A similar distribut ion was found in the individual interview, where TM female speakers had a higher frequency of the full retroflex variants at 13.9%, while TM male speakers had a much lower frequency at 0.2%. In regard to the production of the palatoalveolar variants in g roup interview, TM male speakers had the highest frequency at 100%, while the female speakers had a lower frequency at 89.1%. In individual interview, the male speakers produced a higher

PAGE 104

104 frequency of the palato alveolar variants at 99.8%, while the female speakers had a lower frequency at 86.1%. Putting the above results altogether reveals that in the group interview, the female speakers reduced the frequency of the full retroflex variants at 3%, while the male speakers reduced it at 0.2%. Meanwhile, both groups were found to increase the frequency of the palato alveolar variants at 3% and 0.2%. Particularly, there was zero frequency of the full retroflex variants in male speakers’ production. The findings indicate that the male speakers favor the palatoal veolar variants, whereas the female speakers favor the full retroflex variants. The results also reveal that both female and male speakers increase the frequency of the palatoalveolar variants and decrease the frequency of the full retroflex variants when PTH speakers are present in the group interview. This result seems to suggest that TM speakers use the palatoalveolar variants as an ingroup speech feature, while they use the full retroflex variants as an out group speech feature. Figure 510 presents the distribution of the variants of retroflex (zh) (ch) (sh) by gender and type of interview. Strength of Social Network The 5 TM speakers were initially divided into 3 groups according to their strength of social network with PTH speakers. To obtain the best results from statistical analysis, the Nocontact group was recoded and collapsed with the group of Weak network (les s strength of network). The distribution of the full retroflex variants and the palatoalveolar variants in terms of their correlation with speakers’ strength of social network in different type of interviews is shown in Table 511. In the group interview , the speakers who had less strength of network with PTH speakers produced a higher frequency of the full retroflex variants, 11.3%, while the

PAGE 105

105 speakers who had greater strength of social network with PTH speakers had zero frequency. In the individual inter view, the speakers who had greater strength of network with PTH speakers had a lower frequency of the full retroflex variants at 4.1% in comparison with a higher frequency, 9.8%, produced by the speakers with less strength of network with PTH speakers. In view of the production of the palatoalveolar variants in group interview, the speakers who had greater strength of network with PTH speakers had the highest frequency at 100%, while the speakers who had less strength of network with PTH speakers produced a lower frequency at 88.7%. In regard to the production of the palatoalveolar variants in the individual interview, the speakers who had greater strength of network had a higher frequency at 95.9%, while the speakers who had less strength of network had a lower frequency at 90.2%. The above results indicated that in the group interview, the speakers with greater strength of network with PTH speakers decreased the frequency of the full retroflex variants but increased the frequency of the palatoalveolar variants at 4.1%. On the other hand, the speakers with less strength of network with PTH speakers increased the frequency of the full retroflex variants at 1.5% when they had a decreasing frequency of the palatoalveolar variants. This is a rather surprisi ng result . The present study initially hypothesized that there was a positive correlation between strength of social network and the production of the full retroflex variants. It was assumed that TM speakers who had greater network strength with PTH speakers would be prone to and more willing to use the feature that is distinctive in PTH, i.e. the full retroflex feature, when they had an interview with PTH speakers . The results from the group interview showed that the

PAGE 106

106 speakers who had greater strength of social network with PTH speakers did not produce any full retroflex variants but increased the frequency of the palatoalveolar variants. On the other hand, the speakers who had less strength of network with PTH speakers increased the frequency of the full r etroflex variants in the group interview. The above findings reveal that speakers with greater strength of network with PTH speakers may be more aware of the linguistic differences between PTH and TM, and their awareness affects their preference for particular linguistic variants. The findings fur ther demonstrate that speakers’ social network strength may not be a determinant factor affecting speakers’ linguistic choice. Figure 511 illustrates the distribution of the variants of retroflex (zh) (ch) (sh) by strength of social network and type of interview. Length of Stay in the US The 5 TM speakers originally belonged to three groups according to length of stay in the US: over 5 years, 1 to 2 years, and less than 6 months. For analysis purposes, the groups of 12 years and less than 6 month residence time were collaps ed into the group of less than2 year residence time in the US. Table 512 presents the distribution of the full retroflex variants and the palatoalveolar variants with regard to length of stay in the US in different types of interviews. In the group interview, the speakers who had a longer length of stay in the US (over 5 years) produced a higher frequency of the full retroflex variants, 13.8%, while the speakers who had a shorter residence time in the US (less than 2 years) had zero frequency. In ter ms of the production of the palatoalveolar variants, the speakers with a longer length of stay in the US had a lower frequency at 86.2% in comparison with the 100% frequency produced by the speakers who had a shorter residence time in the US.

PAGE 107

107 In view of the production of the full retroflex variants and the palatoalveolar variants in the individual interview, the production of the full retroflex variants in the group which had a longer residence time in the US showed a higher frequency at 11.8% in compari son with the speakers who had a shorter residence time at 3.3%. In terms of the production of the palatoalveolar variants, the speakers who had a shorter residence time in the US showed a higher frequency at 96.7%, whereas the speakers who had a longer length of stay in the US had a lower frequency at 88.2%. The results showed that in the group interview, the speakers who had a longer residence time in the US increased the frequency of the full retroflex variants at 2%. This tendency suggests that the speakers with a longer residence time in the US may be apt to favor the full retroflex feature and to be more willing to adjust their speech pattern when PTH speakers are present. On the other hand, the speakers who have a shorter residence time in the US dec rease the frequency of the full retroflex variants. In other words, this group does not accommodate to their PTH interlocutors. Figure 512 demonstrates the distribution of the variants of retroflex (zh) (ch) (sh) by length of stay in the US and type of interview. The comparative results from the individual and group interviews reveal that the use of the palatoalveolar feature prevails in TM speakers’ speech, but nuanced statistical differences are still found between each factor within the factor groups. Furthermore, the comparative results also show that the use of the full retroflex feature and the palatoalveolar feature is assigned particular social meanings and ideologies by TM speakers. Previous studies (e.g. Baran 2014; Brubaker 2012) find that Man darin with the palatoalveolar feature (and full dental feature) has been regarded as a non-

PAGE 108

108 standard and stigmatized Mandarin in Taiwanese society, but due to the rise of Taiwanese localism and the complicated political climate between Taiwan and China, Ma ndarin with the palatoalveolar feature is attached positive social meanings, i.e. group solidarity and localness, by Taiwanese people. In contrast, the full retroflex feature, the standard form, links with Chinese culture and value. Before the liberalizat ion of language policy and the promotion of local dialects, Mandarin with the full retroflex feature was considered standard form, and enjoyed prestigious status. With the emphasis of locality, Mandarin with the full retroflex feature is attached negative connotations, such as showiness, in present day. The social meanings and ideologies behind the features become more salient and emerge in the group interview, and affect TM speakers’ choice of variants. TM speakers are particularly likely to use the palatoalveolar feature as an ingroup speech feature that connects them with local Taiwanese cultural values to signal their Taiwanese identity. On the other hand, the full retroflex feature is identified as an out group speech feature that is not a common feature in their daily speech and is associated with Chinese identity and values (Brubaker 2012). These social meanings and ideologies behind the variants of the retroflex (zh) (ch) (sh) are not overtly expressed but from the comparative results, they af fect TM speakers’ choice of variants of the retroflex (zh) (ch) (sh), and also provide evidence that TM speakers may consciously or subconsciously apply these ideologies to their language use in order to emphasize their identity as a member of Taiwanese co mmunity and to create distance between their PTH interlocutors when they are in the immigrant setting.

PAGE 109

109 In this chapter, the distribution of the retroflex (zh) (ch) (sh) has been presented to examine the hypothesis regarding TM speakers’ linguistic behavior s in an immigrant setting, and to what extent they assimilated or maintained their regional linguistic features when they came in contact with PTH speakers. The distribution of the retroflex (zh) (ch) (sh) variants —full retroflex and palatoalveolar —has be en presented. First, the palatoalveolar variants were predominant in TM speakers’ speech. This result corresponds to TM speakers’ reports on their positive attitude towards the palatoalveolar feature that is actually practiced in their speech. Second, th e correlation between the retroflex (zh) (ch) (sh) and four extralinguistic factors — strength of social network, generation, gender, and the length of stay in the US — has been tested. In addition, a multivariate analysis was performed and presented the stati stical differences, a constraint ranking, and a relative strength of each factor that affects the production of the full retroflex variants in 28 individual interviews . The results from the multivariate analysis indicated that all four extralinguistic fact ors were significant. Among four factors, gender displayed the strongest magnitude of effect determining the production of the full retroflex variants, while th e strength of social network displayed the weakest magnitude of effect . In addition, 5 TM speakers were randomly selected from individual interviews to conduct group interviews in order to examine whether the presence of PTH speakers would affect TM speakers’ linguistic choice of the variants of the retroflex initials (zh) (ch) (sh). The distribution of the full retroflex and the palatoalveolar variants in group interview in comparison with the results from individual interview has been presented. The results showed that TM speakers did not increase their use of the full retroflex

PAGE 110

110 feature when they had a group interview with PTH speakers. Additionally, the findings also indicated that the use of the full retroflex and the palatoalveolar variants are assigned ideologies and social meanings, i.e. ingroup and out group, by TM speakers. The next chapter will discuss the results of the retroflex fricative (r). The main findings and interpretation of (r) variation will be provided.

PAGE 111

111 Table 5 1. Overall distribution of the variants of retroflex (zh) (ch) (sh) Variants Number of t okens Frequency (%) Full retroflex [ Table 5 4 . Distribution of retroflex (zh) (ch) (sh) by a ge Age Full retroflex Palatoalveolar % N % N Young ( < 35 y ea rs old) 8.5 579 91.5 625 3 Old er 3 ( > 35 y ea rs old) 1. 4 84 98. 6 5802 Table 5 5 . Distribution of retroflex (zh) (ch) (sh) by g ender Gender Full retroflex Palatoalveolar % N % N Male 0.8 49 99. 2 571 9 Female 8.8 614 91. 2 633 6 1 As was explained in the Methodology chapter, the score for the strength of network was calculated by the following formula: rank order of closeness (score of frequency of meetings + score of telephone calls). The average score of the total social network strength was 9.75. The average score was used as a midpoint to divide TM speakers into three groups, including strong network strength (<9.75), weak network strength (>9.75), and no network strength (0). 2 No contact group was combined with Weak group. 3 The middleaged group was combined with the older group.

PAGE 112

112 Table 5 6 . Distribution of retroflex (zh) (ch) (sh) by length of stay in the US Length of stay Full retroflex Palato alveolar % N % N Less than 2 years 4 4.4 168 95.6 366 4 Over 3 years 5 5.8 430 94. 4 692 6 Table 5 7 . Multivariate analyses of the contribution of extralinguistic factors selected as significant to the probability of the retroflex (zh) (ch) (sh) Corrected mean Log likelihood .01 5 1981.665 Total N 12718 Group Factor weight Percent 6 N Gender Female .82 8. 8 614 Male . 14 0.8 49 Range 7 68 Age Young ( < 35 years old) .81 8. 5 579 Older ( > 35 years old) . 15 1.4 84 Range 66 Length of stay in the US Over 3 years . 64 5. 8 4 30 Less than 2 years . 20 4.4 168 Range 44 Strength of social network Strong .65 3.9 191 Weak .41 6 472 Range 24 4 The group of less than6 month residence time in the US was combined with the group of 12 year residence time in the US. 5 The group of over 5 year residence time in the US was collapsed with the group of 34 year residence time in the US. 6 A percentage of the full retroflex variants is presented as the application value. 7 Range indicates a relative strength of a factor group by subtracting the lowest factor weight from the highest factor weight.

PAGE 113

113 Table 5 8 . Overall distribution of retroflex (zh) (ch) (sh) by type of interview Type of interview Full retroflex Palato alveolar % N % N Individual interview 7.8 195 92. 2 231 8 Group interview 7.3 94 92 .7 1 190 Table 59. D istribution of retroflex (zh) (ch) (sh) by a ge and type of interview8 Age FR 1 % N F R 2 % N PA 1 % N PA 2 % N Young 12.6 190 9.7 85 87.4 1317 90.3 790 Middle aged 0.5 5 2.2 9 99.5 1001 97.8 400 Table 510. D istribution of retroflex (zh) (ch) (sh) by gender and type of i nterview Gender FR 1 % N F R 2 % N PA 1 % N PA 2 % N Female 13.9 193 10.9 94 86.1 1196 89.1 770 Male 0.2 2 0 0 99.8 1122 100 420 Table 511. D istribution of retroflex (zh) (ch) (sh) by strength of social network and type of interview Strength of social network FR 1 % N F R 2 % N PA 1 % N PA 2 % N Strong 4.1 38 0 0 95.9 879 100 452 Weak 9.8 157 11.3 94 90.2 1439 88.7 738 Table 512. D istribution of retroflex (zh) (ch) (sh) by length of stay in the US and type of interview Length of stay in the US FR 1 % N F R 2 % N PA 1 % N PA 2 % N Less than 2 years 3.3 40 0 0 96.7 1157 100 604 Over 5 years 11.8 155 13.8 94 88.2 1161 86.2 586 8 FR represents full retroflex variant and PA represents palatoalveolar variant. Number 1 attached to FR and PA respectively represents an individual interview. Number 2 represents a group interview.

PAGE 114

114 Figure 5 1. Overall distribution of retroflex (zh) (ch) (sh) by strength of social n etwork Figure 5 2. Distribution of retroflex (zh) (ch) (sh) by a ge Figu re 5 3. Distribution of retroflex (zh) (ch) (sh) by g ender 0.0% 50.0% 100.0% Strong Weak 3.9% 6.0% 96.1% 94.0% Full retroflex Palato-alveolar 0.0% 100.0% Young Older 8.5% 1.4% 91.5% 98.6% Full retroflex Palato-alveolar 0.0% 50.0% 100.0% Male Female 0.8% 8.8% 99.2% 91.2% Full retroflex Palato-alveolar

PAGE 115

115 Figure 5 4. Distribution of retroflex (zh) (ch) (sh) by length of stay in the US Figure 5 5. Effect of gender on the production of full retroflex variants [ ], [h], [] Figure 5 6. Effect of age on the production of full retroflex variants [ ], [h], [ ] 0.0% 100.0% > 2 years < 3 years 4.4% 5.6% 95.6% 94.4% Full retroflex Palato-alveolar 0 1 Female Male 0.82 0.14 0 1 Young Older 0.81 0.15

PAGE 116

116 Figure 5 7. Effect of length of stay in the US on the production of full retroflex variants [ ], [h], [] Figure 5 8. Effect of strength of social network on speakers’ production of full retroflex variants [ ], [ h], [ ] Figure 59. Distribution of the variants of retroflex (zh) (ch) (sh) by generation and type of interview 0 1 < 3 years > 2 years 0.64 0.20 0 1 Strong Weak 0.65 0.41 0.0% 50.0% 100.0% Young (under 35 years old) Middle-aged (36-45 years old) 12.6% 0.5% 9.7% 2.2% 87.4% 99.5% 90.3% 97.8% FR1 FR2 PA1 PA2

PAGE 117

117 Figure 510. Distribution of the variants of retroflex (zh) (ch) (sh) by gender and type of interview Figure 511. Distribution of the variants of retroflex (zh) (ch) (sh) by strength of social network and type of interview Figure 512. Distribution of the variants of retroflex (zh) (ch) (sh) by length of stay in the US and type of interview 0.0% 50.0% 100.0% Male Female 0.2% 13.9% 0.0% 10.9% 99.8% 86.1% 100.0% 89.1% FR1 FR2 PA1 PA2 0.0% 50.0% 100.0% Strong Weak 4.1% 9.8% 0.0% 11.3% 95.6% 90.2% 100.0% 88.7% FR1 FR2 PA1 PA2 0.0% 50.0% 100.0% Less than 2 years Over 5 years 3.3% 11.8% 0.0% 13.8% 96.7% 88.2% 100.0% 86.2% FR1 FR2 PA1 PA2

PAGE 118

118 CHAPTER 6 RESEARCH RESULTS AND INTERPRETATIONS Variation in Retroflex F ricative I nitial ( r ) This chapter presents the results of the retroflex fricative initial (r) with a total of 1,428 tokens. The retroflex fricative (r) is pronounced with turning up the tip of the tongue toward the hard palate. In TM, the retroflex fricative ( r ) can be realized as the retroflex fricative [ ] and the lateral [l]. Due to the lack of the retroflex fricative [ ] in Taiwanese, the substitution of the lateral [l] for the retroflex fricative [ ] has been commonly found among bilingual speakers of Taiwanese and Mandarin, and is viewed as an “ethnic marker”1. The following section presents the results of the quantitative analysis for individual intervi ews, followed by the results from group interviews . Overall Distribution of Retroflex F ricative ( r ) Variants Table 61 presents the overall distribution of the retroflex fricative [ ] and the lateral [l] in the speech of TM speakers. As Table 61 shows, the frequency of the retroflex fricative [ ] in the speech of TM speakers was at 41.2%, whereas the frequency of the lateral [l] was at 58.8%. The production of the lateral [l] in TM speakers’ speech exceeded the use of the retroflex fricative [ ] by 17.6%. The higher frequency of the lateral [l] suggests that although TM speakers acquire the retroflex fricative [ ] as the standard form in schools, and are consciously aware of its existence, the influence of Taiwanese still plays a role for the absence of the retroflex fricative [ ] in their natural speech production. The following subsections present the correlation between the retroflex fricative (r) and four extralinguistic factors . 1 See Chapter 3 for further information regardi ng the realization of (r) in Liao’s (2010) study.

PAGE 119

119 Correlation between Retroflex Fricative (r) and the Extralinguistic Factors The correlation between the retroflex fricative (r) and TM speakers’ social network with PTH speakers is presented first, followed by its correlation with three other extralinguistic factors . Strength of Social Network I t was hypothesi zed that TM speakers who had a strong social network tie with PTH speakers might produce a higher rate of the retroflex fricative [ ], which is the standard form and a distinctive feature in PTH speakers’ speech. Unlike the influence of Taiwanese in TM, PT H speakers’ speech still preserves the realization of the retroflex fricative [ ]. Based on the framework of social network, the present study proposed that TM speakers who had a strong network tie with PTH speakers would produce a higher frequency of the retroflex fricative [ ] and a lower frequency of the lateral [l] due to their frequent contact with PTH speakers. On the other hand, TM speakers who had less strength of network and no contact with PTH speakers might produce more lateral [l], which is comm only found in TM speakers’ speech. Table 62 presents the overall distribution of retroflex fricative (r) variants in view of strength of social network . As shown in Table 62, TM speakers who had less strength of network with PTH speakers produced a higher occurrence rate of the retroflex fricative [ ], 42.7%. On the other hand, the speakers who had greater strength of network with PTH speakers had a lower frequency of the retroflex fricative [ ], 38.9% . In terms of the frequency of the lateral [l], TM speakers who had greater strength of social network with PTH speakers had a higher rate of the lateral [l], 61.1%. On the other hand, the speakers who had less strength of social network with PTH speakers produced a lower rate of using the lateral [l] , 57.3%.

PAGE 120

120 The results do not support the hypothesis. The results showed that TM speakers who had less strength of social network with PTH speakers produced a higher frequency of the retroflex fricative [ ] than speakers who had greater strength of social network with PTH speakers. This finding suggests that TM speakers with a greater strength of social network with PTH speakers may be more sensitive to the linguistic differences between TM and PTH than the other group due to their frequent contact with PTH speaker s. Their stronger awareness of the linguistic differences between TM and PTH may also affect their choice of variants of the retroflex fricative. Figure 61 illustrates the distribution of the retroflex fricative [ ] and the lateral [l] in terms of TM speakers’ social network strength with PTH speakers . Generation It was hypothesized that the young generation would produce a higher frequency of the retroflex fricative [ ], while the older generation would have a higher frequency of the lateral [l] according to the previous findings (e.g. Liao 2010). Liao investigated the correlation between language ideology and linguistic behaviors among TM speakers in two local Taiwanese cities. She found a lower frequency of the lateral [l] in the speech of the young group and a higher fr equency of the lateral [l] in the older group. She argued that this phenomenon was due to the fact that the young speakers in general considered Mandarin as their first language, and their Mandarin had less influence from Taiwanese in comparison with the heavy influence of Taiwanese phonological features on the older speakers’ speech. Table 63 presents the overall distribution of the retroflex fricative [ ] and the lateral [l] with regard to the three generations analyzed. As seen in Table 63, the older generation had the highest rate of using the retroflex fricative [ ], 50.9%, followed by the

PAGE 121

121 young generation at 40.3%, and the middleaged generation at 32.5%. Regarding the production of the lateral [l], the middle aged generation had the highest frequency of the later al [l] at 67.5%, followed by the young g eneration at 59.7%. The lowest frequency of the later al [l] was found in the older g eneration at 49.1% . The above result s d o not support the hypothesis, which initially proposed that the young generation would produce the highest rate of the retroflex fricative [ ], while the older generation would produce the highest rate of the lateral [l]. As seen in Table 6 3 , the highest frequency of retroflex fricative [ ] was not found in the young gener ation, and the highest lateral [l] did not occur in the older generation. Instead, the middleaged group had the lowest frequency of the retroflex fricative [ ], 32.5% and the highest frequency of the lateral [l] , 67.5% . A plausible explanation regarding the results in the present study may be related to the development of sociopolitical changes and language attitudes in Taiwanese society from 1949 until present day. The speakers of the older generation were born before the Mandarinonly policy began to b e implemented in all public domains. At that time, Taiwanese and Japanese were still the main communication medium in all domains. However, when the speakers of the older generation in this study started schools, the Mandarinonly policy was implemented, and they thus acquired Mandarin as their second language. Due to the lack of the retroflex feature [ ] in Taiwanese phonology, the substitution of [l] for [ ] is commonly found in native Taiwanese speakers’ speech and speakers who use Taiwanese on a d aily basis (Kubler 1985). The implementation of the Mandarinonly policy resulted in diglossia, which valued Mandarin that carried all the standard form as high language, whereas Mandarin with trace of

PAGE 122

122 Taiwanese phonological features and local dialects as low language. The highest frequency of the retroflex fricative [ ] and the lowest frequency of the lateral [l] suggest that the impact of diglossia may contribute to this pattern in the older generation. The middleaged generation was born and grew up in a period when the Mandarinonly policy was implemented overwhelmingly until the language policy was liberalized in 1987. During this period of liberalization , Taiwanese society underwent huge sociopoli tical changes including encouraging speakers of local dialects to learn and use their dialects to protect their ethnic roots. Also, the government emphasized that all languages and dialects in Taiwan were equal and promoted Taiwanese localism and Taiwance ntered identity. The language use in middleaged generation reflected these socio political changes in Taiwanese society. The use of the standard retroflex fricative [ ] was associated with Chinese value. Due to the rise of Taiwanese localism, the standard form no longer enjoyed its prestigious status inside Taiwanese society. The lowest frequency of the standard retroflex fricative [ ] in middle aged generation reflected that TM speakers’ attitude towards Chinesecentered value was changing, while the highest frequency of the nonstandard lateral [l] reflected the rise of Taiwanese localism affecting TM speakers’ language use, such as use of the form that was linked with Taiwanese local culture and value. The young generation started school a few years before or after the lifting of martial law in 1987. The speakers in the young generation grew up in the period when using local dialects was less restricted in public domains. After liberalization of language policy in 1987, speakers of local dialects were encouraged to present their ethnic roots. The production of the lateral [l] in the young generation reflects that the emphasis of

PAGE 123

123 Taiwanese localness has taken effect in their language use. It also suggests that the use of the nonstandard lateral [l] for the standard retroflex fricative [ ] becomes more regular. In addition, despite the fact that Taiwanese was their home language, the majority of young people considered Mandarin as their first language and used it as their main communication tool. Their production of retroflex fricative [ ] reflected the trend that Mandarin gradually become the major communication medium in Taiwanese society. It further suggests that the influence of Taiwanese in young generation’s speech is gradually diminished. Figure 62 illustrates the distribution of [ ] and [l] in three generations . Gender Previous studies (e.g. Su 2008) pointed out that TM female speakers negatively valued Mandarin that carried a strong Taiwanese accent (i.e. lack of retroflex feature) because it was usually associated with age and lower education level. On the contrary, TM male speakers did not avoid using stigmatized Mandarin (i.e. Taiwaneseaccented Mandarin) because it was linked with positive social values such as masculinity , gr oup solidarity, and locality. In light of this reasoning, the present study hypothesized that TM female speakers would have a higher frequency of the retroflex fricative [ ], the standard form, whereas TM male speakers would have a higher frequency of the lateral [l], the nonstandard form . Table 64 shows that the female speakers produced a higher frequency of the retroflex fricative [ ] , 58.7%, whereas the male speakers had a lower frequency, 20.2%. In regard to the production of the lateral [l], the male speakers had a higher frequency, 79.8%, while the female speakers had a lower frequency, 41.3%. The results support the hypothesis that the female speakers would have a higher frequency of the retroflex fricative [ ], while the male speakers would have a higher

PAGE 124

124 frequency of the lateral [l]. As Table 64 shows, the frequency of the retroflex fricative [ ] in the female group, 58.7% exceeded the male group’s frequency, 20.2%, by 38.5%. The results support that the retroflex fricative [ ] was favored by TM female speakers, while TM male speakers preferred the lateral [l]. In regard to the production of the lateral [l], the male speakers had a higher frequency, 79.8%, which exceeded the frequency of the female speaker (41.3%) by 38.5%. T he higher frequency of the lateral [l] in TM male speakers’ speech suggests that the male speakers favored the lateral [l]. The distribution of the retroflex fricative and the lateral between male and female speakers reveals that the use of the two variants has hidden values that emerge when male and female speakers use the variants in their speech in certain contexts. Further discussion will be presented in next chapter. Figure 63 presents the illustration of the distribution of [ ] and [l] by gender . Length of Stay in the US Hock and Joseph (2009) suggested that in addition to linguistic leveling and convergence, speakers’ loyalty to their homeland was likely to become minimal as their stay extended in the immigrant setting due to the need for accommodating the host country and coming into contact with people from diverse language and dialect backgrounds. Hock and Joseph did not further discuss to what extent that the length of stay and loyalty to the homeland dialect would affect speakers’ linguistic choices as their stay extended. Therefore, the present study hypothesized that the length of stay in the US would affect the production of the retroflex fricative [ ] in the speech of TM speakers. That is, the longer they had resided in the US, the more they would favor the retrofl ex fricative [ ], the standard form. On the other hand, TM speakers who had a

PAGE 125

125 shorter residence time in the US would favor the lateral [l]. Table 65 presents the overall distribution of [ ] and [l] in regard to the length of stay in the US . As seen in Table 65 , TM speakers who had resided in the US for 3 4 years produced the highest rate of the retroflex fricative [ ], 55.9 %, followed by the speakers who had resided in the US for over 5 years, 40.7%. The speakers who had the shortest residence time in the US (less than 2 years) had the lowest frequency of the retroflex fricative [ ] at 36%. The results support the hypothesis, which proposed that the speakers who had a longer residence time in the US would produce a higher frequency of the retroflex fric ative [ ] . In view of the production of the lateral [l], the highest frequency was found in the speech of the speakers who had the shortest residence time in the US, 64%, followed by the speakers who had resided in the US for over 5 years, 59.3%. The lowest frequency of the lateral [l] was found in the speech of the speakers who had resided in the US for 34 years, 44.1% . The highest frequency of the lateral [l] was found in the speech of the speakers who had resided in the US less than 2 years, while a lower frequency was found in the speakers who had resided in the US for 34 years and over 5 years. This result supports t he hypothesis, which proposed that the speakers who had a shorter residence time in the US would have a higher frequency of the lateral [l]. In contrast, the speaker who had a longer residence time in the US would have a lower frequency of the lateral [l]. This result suggests that the speakers who have stayed in the US less than 2 years, tend to maintain use of the variant that carries the feature from their homeland dialect. In view of the production of the retroflex fricative [ ], the group which had stayed in the US over

PAGE 126

126 3 years in general had a higher frequency of the retroflex fricative [ ], while the group which had stayed in the US less than 2 years had a lower frequency. Furthermore, that the speakers who had resided in the US for 3 4 years had t he highest frequency of the retroflex fricative and the lowest frequency of the lateral. This distribution seems to suggest the 34 year residence time is a transitional point in which speakers’ language use undergoes changes after having resided in the im migrant setting for a while. Figure 6 4 illustrates the distribution of [ ] and [l] by length of stay in the US . Multivariate Regression Analysis The present study was designed to investigate the variation in TM speakers ’ speech in an immigrant setting and when they were in contact with PTH speakers . For this purpose, the occurrence rate of the variants of the retroflex fricative initial (r) was exa mined . The retroflex fricative ( r ) has two variants, [ ] and [l], which re present the standard form and the ethnic marker, respectively. To run the Goldvarb X program, factor group 1 is the dependent variants of the retroflex fricative [ ] and the lateral [l]. Factor group 2 to 5 are 4 independent variables, including speaker s’ strength of social network, generation, gender, and the length of stay in the US, examined in terms of their correlation with the dependent variable The results from binominal onelevel testing showed that all four extralinguistic factors had statistically significant (p=0.05) effects on the speakers’ choice of the retroflex fricative [ ] . Table 6 6 presents the multivariate analysis of each extralinguistic factor. As seen in Table 66, among four factor groups, gender had the highest range value at 41, whereas strength of social network had the lowest range value at 2. This result suggest ed that among the factor groups affecting the speakers’ choice of the retroflex fricative [ ] , gender displ ayed the s trongest magnitude of effect, while the

PAGE 127

127 strength of social network displayed the weakest magnitude of effect. The following r anking summarizes the order of each factor groups by the range values (the most important one is on the left) Gender > Le ngth of Stay in the US > Generation > Strength of Social Network The following subsections present the factor weight, constraint ranking, and relative strength of each factor group. Gender With a range of 4 1 , gender displayed the stronges t magnitude of effect affecting TM speakers’ production of the retroflex fricative [ ] . The constraint ranking of gender showed that female speakers tended to favor the retroflex fricative [ ], with a factor weight of .69, whi le their male counterparts tended to disfavor it, with a factor weight of .28. This result supports the hypothesis, which initially proposed that female speakers would prefer the retroflex feature [ ], the standard variant, in their speech, whereas their m ale counterparts would not show such a preference. Instead, the male speakers favored the lateral [l]. The discrepancy between gender group is presented in Figure 6 5 . Length of Stay in the US With a range of 26, length of stay in the US displayed the second strongest magnitude of effect . The constraint ranking showed that the speakers with a residence time in the US of 3 4 years tended to favor the retroflex fricative [ ] , with a factor weight of .68, followed by the speakers who had resided in the US over 5 years, with a factor weight of .50. The speakers who had resided in t he US less than 2 years tended to disfavor the retroflex fricative [ ] , with a factor weight of .42. In other words, speakers who have a longer residence time in the US prefer the retroflex fricative [ ], whereas

PAGE 128

128 speakers who have a shorter residence time in the US do not have such a preference. Figure 6 6 presents the effect of length of stay in the US on the production of the retroflex fricative [ ] . Generation Wi th a range of 13, generation display ed weak mag nitude of effect . The constraint ranking show ed that with a factor weight of .55, the older generation tended to favor the use of the retroflex fricative [ ], while the middleaged generation tended to disfavor it with a factor weight of .42. With a factor weight of .51, the young generation showed a tendency to favor the retroflex fricative [ ]. Figure 6 7 illustrates the effect of generation on speakers’ production of the retroflex fricative [ ]. The hypothesis proposed that the young generation speakers would favor the retroflex fricative [ ] , while the older generation would tend to disfavor it. Yet , as seen in Figure 6 7 , the hypothesis is partially supported. Both the young and the older generations had a tendency to favor the retroflex fricative [ ] , while the middleaged generation disfavo red it . Strength of Social Network With a range of 2 , th e strength of social network displayed the weakest magnitude of effect . W ith a factor weight of .5 1 , the speakers who had a less strength of social network with PTH speakers favored the retroflex fricative [ ] , while the speakers who had a greater strength of social network with PTH speakers disfavored it , with a factor weight of .49 . Figure 6 8 presents the effect of strength of social network on speakers’ production of the retroflex fricative [ ]. The hypothesis initially proposed that speakers who had a greater strength of network with PTH speakers would tend to favor the retroflex fricative [ ]. However, as seen in Figure 6 8 , the hypothesis is not supported.

PAGE 129

129 Group Interview This section presents the use of the retroflex fricative ( r ) in group interview with a total of 130 tokens. Five TM speakers (2M, 3F) were randomly selected from individual interviews to participate in the group interview. Ea ch group interview included 1 TM speaker, 1 PTH speaker (the TM speaker’s friend or colleague) , and the researcher. The role of the researcher was to initiate questions without engaging in the conversation. The purpose of the group interviews was to examin e the variation of the retroflex fricative ( r ) that 5 TM speakers would have when they had an interview with PTH speakers. The distribution of the two variants in the 5 speakers’ individual interview is also presented in comparison with the results from their group interview . It was hypothesized that TM speakers ’ speech might be affected by their PTH interlocutors, causing them to prefer the retroflex fricative [ ] and avoid using TM features, i.e. the lateral [l], when they had an interview with PTH speakers to reach convergent communication. The overall distribution of the retroflex fricative [ ] and the lateral [l] in the group interview showed that the frequency of the retroflex fricative [ ] was 33.1%, whereas the frequency of the lateral [l] was 66.9%. The distribution of the retroflex fricative [ ] and the lateral [l] in group and individual interviews is presented in Table 67 . It was initially hypothesized that a higher frequency of the retroflex fricative [ ] would be found in TM speakers’ speech when they had an interview with PTH speakers. The hypothesis is supported. As seen in Table 67, the frequency of the retroflex fricative [ ] was at 32.4%, whereas the f requency of the lateral [l] was at 67.6% in the individual interview. As compared with the individual interview, in group interview, the frequency of the retroflex fricative [ ] increased 0.7%. On the other hand, the frequency

PAGE 130

130 of the lateral [l] decreased by 0.7%. This result indicates that TM speakers were likely to increase the use of the retroflex fricative [ ] when PTH speaker were present in the group interview. In other words, the presence of PTH speakers may be a factor affecting TM speakers’ linguistic choice. The following subsection presents the correlation between retroflex fricative (r) and four extral inguistic factors in group interviews . Distribution of R etroflex Fricative ( r ) and the Extralinguistic Factors Generation The 5 TM speakers involved two generations: young (less than 35 years old) and middle aged (3645 years old). Table 68 shows the distribution of retroflex fricative (r) by the two generations in the group and individual interviews. In the group interview, the middleaged generation had a higher frequency of the retroflex fricative [ ] at 54.8%, while the young generation had a lower frequency at 26.3%. The results in indiv idual interviews indicated that the middle aged generation had a higher frequency of the retroflex fricative [ ] at 44.7%, whereas the young generation had a lower frequency at 26.8%. The production of the retroflex fricative [ ] in interviews revealed that the middleaged generation increased the frequency of the retroflex fricative at 10.1%, while the young generation decreased the frequency at 0.5% . In regard to the production of the lateral [l] in the group interview, the young generation had a higher frequency at 73.7%, while the middleaged generation had a frequency of 45.2%. A similar pattern was found in the individual interview. The young gener ation had a higher frequency of the lateral [l] at 73.2%, whereas the middleaged generation had a frequency of 55.3%. The results concerning the production of the lateral [l] in two types of interviews showed that in the group interview, the young

PAGE 131

131 generat ion increased the use of the lateral [l], while the middleaged generation decreased the frequency of the same variant . The results indicate that the young generation may be more aware of the presence of PTH speakers during the interview, so they produce more lateral [l] than the middle aged group. The use of the lateral [l] seems to be assigned social meanings and is used as an ingroup speech feature by TM speakers to distance themselves from their PTH interlocutors. On the other hand, the middleaged ge neration seems more willing to adjust their use of the retroflex fricative [ ] when having an interview with PTH speakers. Their willingness implies that speakers of this generation are less sensitive to the social meanings behind the variants than their y oung counterparts. The findings suggest that speakers’ awareness of the presence of PTH speakers may come into play, affecting their choice of variants. Figure 69 illustrates the distribution of variants of the retroflex fricative (r) by generation and ty pe of interview . Gender In the interview, the majority of female speak ers reported that they tended to use the retroflex feature that the PTH speakers used when talking to their PTH friends or colleagues . Based on this reasoning, it was hypothesized that female speakers were inclined to use more retroflex fricative [ ] when talking to their PTH interlocutor during the interview. As seen in Table 69, TM female speakers had a higher frequency of the retroflex fricative [ ] at 47.7%, while their male counterparts had a lower frequency of this variant at 2.4% in the group interv iew. The results from their individual interview also showed that female speakers had a higher frequency of the retroflex fricative [ ] at 55.7%, whereas male speakers had a lower frequency at 5.3%.

PAGE 132

132 Regarding the production of the later [l] in the group interview, the female speakers had a lower frequency at 52.3%, while the male speakers had a higher frequency at 97.6%. A similar tendency was found in the individual interview. The female speakers had a lower frequency of the lateral [l] at 44.3%, whereas t he male speakers had a higher frequency at 94.7%. The above results revealed that regardless of type of interview, the female speakers had a tendency to produce more of the retroflex fricative variant than their male counterparts. On the other hand, the m ale speakers tended to favor the lateral [l]. This finding supports the hypothesis. In addition, when comparing the results between individual and group interviews, female and male speakers are found to decrease the use of the retroflex fricative [ ] at 8% and 2.9% respectively. In contrast, the use of the lateral [l] was increased at 8% in the female group, and at 2.9% in the male group. In other words, both female and male speakers increased the use of the lateral [l] during the group interview. These res ults suggest that the retroflex fricative and the lateral [l] are attached social meanings by both female and male speakers. Although these meanings were not explicitly expressed, it is clear that both male and female speakers’ choice of variants is affect ed by the meanings and ideologies behind the variants. Figure 610 illustrates the distribution of variants of the retroflex fricative (r) by gender and type of interview . Strength of Social Network Following the framework of social network would predict that the speakers who have greater strength of network with PTH speakers would produce more retroflex fricative [ ] , which is the standard form and also th e distinctive feature in PTH speakers’ speech. Therefore, it was hypothesized that when PTH speakers were present during

PAGE 133

133 the interview, the speakers who had greater strength of network with PTH speakers would be more willing to use more retroflex fricative [ ]. As seen in Table 610, in the group interview, the speakers who had less strength of network (and no contact) with PTH speakers had a higher frequency of the retroflex fricative [ ] at 45.7%. On the other hand, the speakers who had greater strength o f network with PTH speakers had a much lower frequency of the retroflex fricative [ ] at 2.6%. Regarding the production of the lateral [l] in the group interview, the speakers who had greater strength of network with PTH speakers had a higher frequency of the lateral [l] at 97.4%, while the speakers who had less strength of network with PTH speakers had a lower frequency at 54.3%. A similar pattern was found in the results from the individual interview. The speakers who had less strength of network (and no contact) with PTH speakers had a higher frequency of the retroflex fricative [ ] at 42.4%. On the other hand, the speakers who had greater strength of network with PTH speakers produced a lower frequency of the retroflex fricative [ ] at 11.4%. Concerning the frequency of the lateral [l], the speakers who had greater strength of network with PTH speakers produced a higher frequency at 88.6%, whereas the speakers who had les s strength of network with PTH speakers had a lower frequency at 57.6%. Putting the results altogether revealed that the speakers who had greater strength of network with PTH speakers tended to have a lower frequency of the retroflex fricative [ ] and a hi gher frequency of the lateral [l] in comparison with the results in the group of speakers who had less strength of network with PTH speakers. The findings do not support the hypothesis, which initially proposed a positive correlation between the use of the retroflex fricative and the strength of social network with PTH speakers.

PAGE 134

134 The comparative results from two types of interviews found that the speakers who had greater strength of network with PTH speakers showed a greatly decreasing frequency i n the group interview ( 8.8%). On the other hand, the speakers who had less strength of network with PTH speakers showed a different pattern. This group increased the use of the retroflex fricative [ ] in the group interview. This implies that the use of the retroflex fricative [ ] and the lateral [l] has implicit social meanings and ideologies which become evident when TM speakers interact with PTH spe akers . In particular, the speakers who have contact with PTH speakers may be more aware of the linguistic differences between TM and PTH due to their frequent contact with PTH speakers, so they may be more sensitive to the sociocultural meanings behind the variants. In contrast, the speakers who do not have much contact experience with PTH speakers may not have strong awareness of the association between variants and the social meanings and ideologies behind them. Figure 611 illustrates the distribution of variants of the retroflex fricative (r) by strength of social network and type of interview . Length of Stay in the US It was suggested that a longer residence time in a foreign country would diminish speakers’ loyalty to speakers’ homeland dialect in order to accommodate to the foreign environment (Hook and Joseph 2009). The present study hypothesized that the speakers who had a longer residence time in the US would produce less homeland dialectal features to reach convergent communication, while the speakers who had a shorter residence time in the US would maintain homel and dialectal features. Table 611 presents the distribution of the retroflex fricative [ ] and the lateral [l]. As seen in Table 611, in the group interview, the speakers who had a longer residence time in the US (over 5 years) produced a higher frequency of the retroflex

PAGE 135

135 fricative [ ] at 70.7%. The speakers who had a shorter residence time in the US (less than 2 y ears) had a lower frequency of the retroflex fricative at 2.8%. With regard to the production of the lateral [l] in the group interview, the speakers who had a longer residence time in the US had a lower frequency of the lateral [l] at 29.3%, while the spe akers who had a shorter residence time in the US had a higher frequency of the lateral at 97.2%. In regard to the results in the individual interview, the sp eakers who had a longer residence time in the US had a higher frequency of the retroflex fricative [ ] at 52.5%. On the other hand, the speakers who had a shorter residence time had a lower frequency of the retroflex fricative at 12.3%. Regarding the use o f the lateral [l], the speakers who had a longer residence time in the US had a lower frequency of the lateral [l] at 47.5%, whereas the speakers who had a shorter residence time had a higher frequency at 87.7% . Putting the a bove results altogether reveals that in the group interview, the speakers who had a longer residence time in the US increased the use of the retroflex fricative [ ] at 18.2%. On the other hand, the speakers who had a shorter residence time in the US decreased the use of the retroflex fricative at 9.5%. This result suggests that when the residence time in the US is extended, the speakers tend to favor the retroflex fricative and are more willing to adjust their speech pattern when they interact with PTH speakers. On the other hand, the shorter the speaker’s residence time in the US, the more likely they are to maintain the speech pattern that they use in their place of origin. Figure 612 illustrates the distribution of variants of the retroflex fricative (r) by length of stay in the US and type of interview.

PAGE 136

136 This chapter has presented the distribution of the retroflex fricative initial (r) in terms of the correlation between (r) variants and four extralinguistic factors. The findings suggest that extralinguistic factors significantly affect TM speakers’ choices of the retroflex fricative (r) variants. The next chapter first summarizes th e major findings of the two linguistic variables, followed by a discussion with the excerpts from interviews and short questions from questionnaires regarding TM speakers’ language attitude, and the social meanings and ideologies behind the variants of the retroflex initials.

PAGE 137

137 Table 6 1 . Overall distribution of the variants of retroflex fricative ( r ) Variants Number of t okens Frequency (%) Retroflex fricative [ ] and [l] by strength of social n etwork Strength of social n etwork Retroflex fricative [ ] and [l] by age Age Retroflex fricative [ ] Lateral [l] % N % N Young ( < 35 y ea rs old) 40.3 325 59.7 482 Middle a ged (36 45 y ea rs old) 32.5 94 67.5 195 Old er ( > 46 y ea rs old) 50.9 169 49.1 163 Table 6 4 . Overall di stribution of [ ] and [l] by gender Gender Retroflex fricative [ ] Lateral [l] % N % N Male 20.2 131 79.8 518 Female 58.7 457 41.3 322 1 No contact group was collaborated with Weak group.

PAGE 138

138 Table 6 5. Overall di stribution of [ ] and [l] by l ength of stay in the US Length of Stay Retroflex fricative [ ] Lateral [l] % N % N Less than 2 years 2 3 6 156 64 277 3 4 years 55.9 100 44.1 79 Over 5 years 4 0.7 3 32 5 9.3 484 Table 6 6 . Multivariate analyses of the contribution of extralinguistic factors selected as significant to the probability of the retroflex fricative (r) Corrected mean Log like li hood .39 834. 847 Total N 1428 Group Factor weight Percent 3 N Gender Female . 69 58.7 457 Male . 28 20.2 1 31 Range 4 41 Length of stay in the US 3 4 years . 68 55.9 100 Over 5 years . 50 40.7 332 Less than 2 years .4 2 36 1 56 Range 26 Age Older . 55 50.9 169 Young . 51 40.3 325 Middle aged .42 32.5 94 Range 13 Strength of social network Weak .5 1 42.7 364 Strong .49 38.9 224 Range 2 2 T he group that had stayed in the US less than 6 months was combined with the group that had stayed in the US for 1 2 years. 3 The retroflex fricative variant is presented as the application value. 4 Range indicates a relative strength of a factor group by subtracting the lowest factor weight from the highest factor weight.

PAGE 139

139 Table 6 7 . Overall distribution of retroflex fricative ( r ) by type of interview Type of Interview Retroflex Fricative [ ] Lateral [l] % N % N Individual 32.4 79 67.6 165 Group 33.1 43 66.9 87 Table 68. D istribution of retroflex fricative (r) by age and type of i nterview5 Age RF1 % N RF2 % N L1 % N L2 % N Young 26.8 45 26.3 26 73.2 123 73.7 73 Middle aged 44.7 34 54.8 17 55.3 42 45.2 14 Table 69. D istribution of retroflex fricative (r) by gender and type of interview Gender RF1 % N RF2 % N L1 % N L2 % N Female 55.7 73 47.7 42 44.3 58 52.3 46 Male 5.3 6 2.4 1 94.7 107 97.6 41 Table 610. D istribution of retroflex fricative (r) b y strength of social network and type of interview Strength of social network RF1 % N RF2 % N L1 % N L2 % N Strong 11.4 9 2.6 1 88.6 70 97.4 37 Weak 42.4 70 45.7 42 57.6 95 54.3 50 Table 611. D istribution of retroflex fricative (r) b y length of stay in the US and type of i nterview Length of stay in the US RF1 % N RF2 % N L1 % N L2 % N L ess t han 2 years 12.3 15 2.8 2 87.7 107 97.2 70 Over 5 years 52.5 64 70.7 41 47.5 58 29.3 17 5 RF represents retroflex fricative and L represents lateral. Number 1 attached to RF and L respectively represents an individual interview. Number 2 represents a group interview.

PAGE 140

140 Figure 6 1 . Overall di stribution of [ ] and [l] by strength of social n etwork Figure 6 2. Overall di stribution of [ ] and [l] by a ge Figure 6 3 . Overall di stribution of [ ] and [l] by gender 0.0% 20.0% 40.0% 60.0% 80.0% STRONG WEAK 38.9% 42.7% 61.1% 57.3% Retroflex Fricative Lateral 0.0% 20.0% 40.0% 60.0% 80.0% YOUNG MIDDLE AGED OLD 40.3% 32.5% 50.9% 59.7% 67.5% 49.1% Retroflex Fricative Lateral 0.0% 50.0% 100.0% MALE FEMALE 20.2% 58.7% 79.8% 41.3% Retroflex Fricative Lateral

PAGE 141

141 Figure 6 4. Overall distribution of [ ] and [l] by length of stay in the US Figure 6 5 . Effect of gender on speakers’ production of the retroflex fricative [ ] Figure 6 6 . Effect of length of stay on the production of retroflex fricative [ ] 0.0% 20.0% 40.0% 60.0% 80.0% LESS THAN 2 YEARS 3 4 YEARS OVER 5 YEARS 36.0% 55.9% 40.7% 64.0% 44.1% 59.3% Retroflex Fricative Lateral 0 0.2 0.4 0.6 0.8 MALE FEMALE 0.28 0.69 0 0.2 0.4 0.6 0.8 LESS THAN 2 YEARS 3 4 YEARS OVER 5 YEARS 0.42 0.68 0.50

PAGE 142

142 Figure 6 7 . Effect of age on the production of retroflex fricative [ ] Figure 6 8 . Effect of strength of social network on speakers’ production of retroflex fricative [ ] Figure 69. Distribution of variants of the retroflex fricative (r) by generation and type of interview 0 0.2 0.4 0.6 YOUNG MIDDLE AGED OLD 0.51 0.42 0.55 0.48 0.50 0.52 STRONG WEAK 0.49 0.51 0.0% 50.0% 100.0% Young (under 35 years old) Middle-aged (36-45 years old) 26.8% 44.7% 26.3% 54.8% 73.2% 55.3% 73.7% 45.2% RF1 RF2 L1 L2

PAGE 143

143 Figure 610. Distribution of variants of the retroflex fricative (r) by gender and type of interview Figure 611. Distribution of variants of the retroflex fricative (r) by strength of social network and type of interview Figure 612. Distribution of variants of the retroflex fricative (r) by length of stay in the US and type of interview 0.0% 50.0% 100.0% Female Male 55.7% 5.3% 47.7% 2.4% 44.3% 94.7% 52.3% 97.6% RF1 RF2 L1 L2 0.0% 50.0% 100.0% Strong Weak 11.4% 42.4% 2.6% 45.7% 88.6% 57.6% 97.4% 97.6% RF1 RF2 L1 L2 0.0% 100.0% Under 2 years Over 5 years 12.3% 52.5% 2.8% 70.7% 87.7% 47.5% 97.2% 29.3% RF1 RF2 L1 L2

PAGE 144

144 CHAPTER 7 DISCUSSION This chapter first summarizes the major findings regarding the variations of the retroflex initials (zh) (ch) (sh) (r) and their correlation with four extralinguistic factors. It further summarizes TM speakers’ linguistic behaviors when they have group int erviews with PTH speakers. Next , excerpts from the interviews and short answers from the questionnaire are presented to discover social meanings and ideologies linked with the variations of the retroflex initials (zh) (ch) (sh) (r), and to provide insights into the use of the variants of the retroflex initials under dialectal contact in the context of globalization. The Retroflex Initials (zh) (ch) (sh) The distribution of the variants of the retroflex initials (zh) (ch) (sh) and four extralinguistic factors is presented. The realization of the retroflex (zh) (ch) (sh) is categorized as three major variants including the full retroflex, the palatoalveolar, and the full dental. The f ull retroflex is the standard and prescribed form, while the palatoalveolar and the full dental are the nonstandard forms. The results showed that the nonstandard palatoalveolar was the predominant variant in TM speakers’ speech, followed by the standar d full retroflex. The full dental had the least frequency in TM speakers’ speech. In this study, it was collapsed with the group of the palatoalveolar for statistical analysis purposes. The full retroflex feature, the prescribed form, is taught in schools but not commonly practiced in TM speakers’ daily speech even if speakers are aware of its presence in highly formal occasions. The findings echoed Brubaker’s (2012) study that the Mandarin that carries local dialectal features has become the normative for m that is actually practiced by TM speakers in their daily speech. The

PAGE 145

145 quantitative analysis showed that four extralinguistic factors had statistical significance for TM speakers’ choice of the retroflex (zh) (ch) (sh) variants. First, the results with re gard to the strength of social network were unexpected. Initially, the effect of TM speakers’ strong social network with PTH speakers on their production of the full retroflex feature were predicted to be found and present a positive correlation due to their frequent contact with PTH speakers. Yet, the speakers who had greater strength of network with PTH speakers did not show this tendency. Rather, they produced a higher frequency of the palatoalveolar variant which is TM feature, and a lower frequency of the full retroflex variant, which is a PTH feature. What is even more interesting is the pattern found in the results from the speakers who had less strength of network with PTH speakers. In this group, speakers produced a higher frequency of the full retroflex variant, and a lower frequency of the palatoalveolar variant. The different frequency of the full retroflex and the palatoalveolar feature between the two groups may be attributed to the different degree of awareness of linguistic differences b etween TM and PTH, and of speakers’ sensitivity to social meanings and ideologies behind the variants. Due to the frequent contact with PTH speakers, the speakers with greater network strength with PTH speakers seem to be more sensitive to the linguistic d ifferences between the two Mandarin dialects, and their sensitivity affects their choice of a particular variant. On the other hand, although the speakers who have less network strength with PTH speakers may notice the linguistic differences between TM and PTH, they may not be as sensitive as the speakers with greater strength of network with PTH speakers because of their infrequent interaction with PTH speakers.

PAGE 146

146 Furthermore, although the frequency of the palatoalveolar and the full retroflex features is highly skewed, the different frequency of the full retroflex found between two groups of speakers still provides us some evidence to explain TM speakers’ linguistic choice. As discussed earlier, the speakers who had greater strength of network with PTH spe akers may be more sensitive to the differences between TM and PTH due to their frequent interaction with PTH speakers. Their sensitivity seems to affect their use of the full retroflex feature which had a lower frequency in comparison with the speakers who did not have frequent contact with PTH speakers. In other words, different frequency found in the use of the full retroflex feature and the palatoalveolar feature seems to suggest that speakers assign particular social meanings and ideologies to the full retroflex feature and the palatoalveolar features respectively. The lower frequency of the full retroflex feature may indicate that the full retroflex feature is viewed as an out group speech feature, and the palatoalveolar feature as an ingroup featur e due to its higher frequency . Lastly, the findings also suggest that the social network strength is not a strong factor affecting TM speakers’ linguistic choice of the variants of the retroflex (zh) (ch) (sh) when other stronger factors come into play . The correlation between the retroflex (zh) (ch) (sh) and strength of social network reveal s that when the use of the variants that are affiliated with speakers’ sociocultural ideologies come into play, speakers’ linguistic choice of a particular variant can be affected in order to show their identification as a member of a particular social group. The results with regard to generation revealed that the palatoalveolar feature was favored by all generations. The young generation had a higher occurrence rate of

PAGE 147

147 the full retroflex feature in comparison with the older generation. On the other hand, the older generation produced a much higher frequency of the palatoalveolar feature and a much lower frequency of the full retroflex feature. The skewed distribution between the full retroflex and the palatoalveolar features in the speech of the older generation suggests that the influence of Taiwanese on the use of the palatoalveolar feature for the full retroflex feature in this generation seems to be greater than the young generation. In other words, when speakers are younger, the tendency of using the full retroflex feature is stronger . The findings in regard to gender showed that despite the fact that the palatoalveolar feature always has the highest percentage in all gender groups, the differences between male and female speakers are still presented. The statistical difference in the use of the full retroflex feature between male and female speakers reveals that the female speakers tend to favor the full retroflex feature, the standard form, and to disfavor the use of the palatoalveolar feature, the nonstandard form. A higher frequency of the palato alveolar feature in the speech of TM male speakers suggests that they prefer the nonstandard palatoalveolar feature. The results conform to previous studies (e.g. Su 2008) that the TM female speakers tend to favor the standard form in order to meet traditional social expectations as women in Taiwanese society, and to avoid using Mandarin with the palatoalveolar feature which is linked with negative social evaluations such as unsophistication, lack of manners, and lack of education. Contrary to negative social evaluations for female speakers’ use of the palatoalveolar feature, TM male speakers tend to favor the palatoalveolar feature which is associated with positive connotations such as masculinity and solidarity.

PAGE 148

148 Different connotations are attached to the full retroflex feature and the palatoalveolar feature when different gender group’s language use is evaluated in Taiwanese society. The findings concerning the length of stay in the US showed that when TM speakers extended their residence time in the US, they increased their use of the full retroflex feature, the standard form. On the other hand, when TM speakers had a shorter residence time in the US, they had a higher frequency of the palatoalveolar feature, the nonstandard form. The findings also reveal that speakers who have a shorter residence time in the US may tend to maintain the speech feature that is associated with their place of origin. In contrast, as the speaker’s length of stay extends, their preference changes in favor of the form that is accepted widely by the local immigrant community. Furthermore, the results from multivariate regression analysis conclud ed that gender displayed the strongest magnitude of effect affecting TM speakers’ linguistic choice of the full retroflex feature, whereas the strength of social network displayed the weakest magnitude of effect . In addition , the findings from the group interview demonstrated that the variants of the retroflex (z h) (ch) (sh) were assigned social meanings and ideologies, and used by TM speakers when they came into contact with PTH speakers. The different rate between the two types of interview reveals that these meanings and ideologies behind the variants are not overtly expressed, but rather emerge when an outside factor comes into play. In this study, the interaction with PTH speakers is the factor attributing to the emergence of these meanings and ideologies. Particularly, the full retroflex feature is associated with an out group speech feature, whereas the palatoalveolar feature with an in group speech feature.

PAGE 149

149 The Retroflex Fricative (r) The production of the retroflex fricative initial (r) in TM speakers’ speech can be realized as tw o variants : retroflex fricative [ . The realization of retroflex fricative [ is considered as the prescribed and standard form , whereas the lateral [l] is the nonstandard form and has been viewed as an ethnic marker. The alternation between retroflex fricative and lateral is commonly found in the speech of bilingual speakers of Mandarin and Taiwanese. The overall distribution of the retroflex fricative (r) variants showed that the lateral [l] was the dominant variant in the speech of TM speakers. Additionally, the multivariate analysis showed that all four extralinguistic factors had statistical significance for affecting TM speakers’ choice of the retroflex fricative [ The results in regard to the strength of social network revealed that the speakers who had greater strength of network with PTH speakers produced a lower frequency of the retroflex fricative [ , the standard form. It was originally postulated that the speakers who had greater strength of network with PTH speakers would produce more retroflex fricative [ and a lower frequency of the lateral [l] because the standard form is still preserved in PT H speakers’ speech, and TM speakers’ speech patterns would be affected by them due to their frequent contact with PTH speakers. This hypothesis was rejected. The results showed that the speakers who had less strength of network with PTH speakers had a higher frequency of the full retroflex fricative [ and a lower frequency of the lateral [l]. The findings did not conform to the framework of social network, which predicted a higher frequency of the retroflex fricative would be found in the speech of the sp eakers who had a greater strength of network with PTH speakers. The pattern of the retroflex fricative (r) in the present study indicates that the speakers

PAGE 150

150 who have greater strength of network with PTH speakers may be more sensitive to the use of the retro flex fricative [ of particular variants. Furthermore, the results also demonstrate that the variants carry particular social meanings and ideologies which are assigned by TM speakers who may consciously or subconsciously apply them to their actual language use. The findings regarding the distribution of the retroflex fricative [ and the lateral [l] in different generations revealed that the older generation had the highest rate of the retroflex fricative [ , followed by the young generation. In view of the production of the lateral [l], the middleaged generation had the highest frequency of the lateral [l], followed by the young generation. Furthermore, the middleaged generation had the highest frequency of the lateral [l] and the lowest frequency of the retr oflex fricative [ ]. The patterns between three generations suggests that speakers’ language use reflects different stages of sociopolitical changes in Taiwanese society. The speakers in the young generation started their schooling a few years before the l iberalization of language policy. Although using the local dialects was encouraged due to the rise of localism, Mandarin was still the institutional form and the lingua franca between different ethnic groups in Taiwanese society. Furthermore, the distribut ion of the use of the retroflex fricative [ aged generation presents a different picture. The speakers in this generation were born and grew up in the period that had endured the implementation of the MandarinOnly policy, the liberalization of language policy, and the rise of Taiwanese localism . T hese changes thus affect their linguistic choice of the two variants. Regarding the older generation, the speakers in this generation had the highest frequency of the retrof lex fricative [ and the lowest

PAGE 151

151 frequency of the lateral [l]. This generation grew up in the period when the Mandarinonly policy began to be implemented and to eventually have overwhelming success. The distribution of the two variants in the older generation suggests that diglossia seems to play a role affecting speakers’ choice of the two variants . In view of the correlation between the distribution of the retroflex fricative [ and the lateral [l] and gender, the results showed that the female speakers preferred the retroflex fricative [ suggests that the female speakers tend to use the standard form, the retroflex fricative [ rther suggests that Taiwanese cultural and traditional norms still regulate how women should behave to show good manners and to project their femininity with the proper language use. However, the social norms do not restrictively regulate men’s language us e, particularly, the use of the nonstandard form. Their preference for the nonstandard feature is not linked with negative connotations. Rather, when they use the Mandarin that carries nonstandard features, they mostly receive positive evaluations such as masculinity and their connection with Taiwanese local ity. The results regarding length of stay in the US revealed that the speakers who had stayed in the US for 34 years had the highest frequency of the retroflex fricative [ but had the lowest frequency of the lateral [l]. In contrast, the speakers who had stayed in the US less than 2 years had the highest frequency of the lateral [l] but had the lowest frequency of the retroflex fricative [ . The findings suggest that the l ength of residence time in the US affects speakers’ linguistic choice of the retroflex fricative (r). Furthermore, the results indicate that the speakers who have resided in the US for 34

PAGE 152

152 years may gradually develop a sense of belonging in the local immig rant community after having stayed for a period of time. Their choice of the variants may be affected by the social standards inside the community. In contrast, the speakers who have resided in the US less than 2 years may not develop the sense of belonging and may still have more connection with their place of origin. The results from the speakers who had resided in the US over 5 years showed an interesting pattern. The second highest frequency of the retroflex fricative [ and the second lowest lateral [ l] were found in this group. This finding seems to suggest that when speakers’ residence time in the US is extended, they prefer the feature that is affiliated with their place of origin. The multivariate regression analysis showed that gender displayed the strongest magnitude of effect affecting TM speakers’ linguistic choice of the variants of the retroflex fricative (r), whereas strength of social network displayed the weakest magnitude of effect . In the group int erview, the frequency of the retroflex fricative [ , the standard form, was slightly increased, whereas the frequency of the lateral [l], the nonstandard form, was slightly decreased. This result suggests that accommodating to PTH speakers is likely to t ake place during the interview. The different frequency of the var iants between two types of interviews reveals that TM speakers implicitly assign social meanings and ideologies to the variants of the retroflex fricative (r) . Particularly, the retroflex fr icative [ is indexed as an out group speech feature, whereas the lateral [l] as an in group speech feature. These meanings and ideologies emerge when TM speakers come into contact with PTH speakers. The above section summarizes the major findings regarding the correlation between the retroflex initials (zh) (ch) (sh) (r) and four extralinguistic factors, and TM

PAGE 153

153 speakers’ linguistic choice when they interact with PTH speakers in the group interview . The following section provides an indepth discus sion regarding the sociolinguistic status and ideologies associated with the retroflex initials. Sociolinguistic Ideologies of Retroflex I nitials (zh) (ch) (sh) (r) The findings from the group interview reveal that the variants of the retroflex initials (zh) (ch) (sh) (r) are linked with particular social meanings and ideologies, which are socially and culturally constructed within Taiwanese society, their place of origin. The ideologies behind the variants of the retroflex initials surface when they come into contact with PTH speakers. The findings further show that when TM speakers move to a foreign country, the ideologies and their attitude towards standard Chinese, PTH, TM, and Taiwanese play an important role affecting their linguistic choice in the i mmigrant environment. This section presents the excerpts1 from the interview and shorts answers to the questions in a questionnaire to show TM speakers’ viewpoints and attitude regarding the sociolinguistic status as well as ideologies of the variants of t he retroflex initials (zh) (ch) (sh) (r) in TM. For discussion purposes, the term ‘retroflex feature’ is used to refer to the full retroflex feature and the retroflex fricative feature, and the term ‘de retroflex feature’ refers to the palatoalveolar feat ure and the lateral feature. The results in Chapter 5 and 6 reveal that the use of the standard form of the retroflex initials do es not have a high frequency in TM speakers’ speech although the speakers acquire it as the prescribed form and are aware of their presence in highly formal occasions. F o r example, when asked what is considered as “standard Man darin”, 1 The excerpts, short answers and questions are translated from Mandarin to English by myself.

PAGE 154

154 PF216 (A: 29 SN: No contact LS: 34 years )3 immediately referred to the proper use of the standard retroflex feature and said: PF16: The so called “standard Mandarin” means that we are supposed to follow Bopomofo4. For example, in speech or recitation contest s, you are asked to pronounce 5 zh ch sh r correctly as the way it is spelled in Bopomofo . Lai: What is the standard way of pronouncing zh ch sh r ? PF16: It depends on the degree that you curl your tongue. It should pronounce with a er sound. PF1 6: Of course, I can b ecause of the . If I randomly hear a conversation between PTH speakers on the street , the first thing I notice is the ir usage of the , and then it is the words /terms that they use. For instance, they use i rn to refer to boyfriend or girlfriend []. Her reply reflects common impressions among TM speakers regarding PTH speakers’ speech. The majority of TM speakers ’ responses i n their questionnaire correspond to that of PT16. They report ed that when they interacted with their PTH friends and colleagues , the first distinctive feature they noticed was the use of the prescribed retroflex feature , and the next was the lexicon. Furthermore, w hen asked if they would change their speech pattern by increasing the use of the prescribed retroflex feature to accommodate to their PTH interlocutors, for instance, they generally reported that their speech pattern would be affected by their PTH interlocutors such as producing more prescribed retroflex feature, changing tone patterns and using lexicon after having 2 ‘P’ stands for Participant and ‘F’ for Female. 3 ‘A’ stands for Age, ‘SN’ for Social Network, ‘LS’ for Length of Stay 4 Bopomofo or zh y n “phonetic symbol” is the official phonetic notation system of Mandarin phonology adopted in Taiwan. 5 ‘ ’ refers to retroflexion

PAGE 155

155 talked to them for a long time. For example, PM6 (Age:28 SN: weak LS: less than 2 years) reported that he felt that he was subconsciously influenced by his PTH friends’ speech pattern: PM6: [] I would not change (the way I talk). But when I talk to PTH speakers for a long time, sometimes I am affected by them. If you often hang out with PTH speakers, you sound like them. Lai: Do you think that this tendency occurs unconsciously? PM6: Yes. It is unconscious . PM6’s responses suggest that human tendency of accommodation could take place. The statistical results regarding the retroflex (r) in the group interview reflect this tendency. Furthermore, TM male and female speakers’ language attitudes toward accommodation are different. For example, when asked if they would change their speech to accommodate to their PTH friends or colleagues , PM5 (Age:33 SN: Strong LS: over 5 years) stated: PM5: No. Not at all. I just maintai n the usual way, yes, just the way we usually speak . What the way I speak in daily life is the way it is . PM5’s attitude toward accommodating to PTH speakers exemplifies the majority of TM male speakers’ attitude, and a high frequency of the deretrofl ex features (the palatoalveolar and the lateral) found in male speakers’ quantitative results reflects their attitude. On the other hand, the female speakers generally agreed that they were able to control their use of the retroflex feature and used it more frequently when talking to their PTH friends or colleagues . For instance, PF9 (Age: 43 SN: weak SL: over 5 years) stated in her interview: PF9: I found myself to have that tendency. I do not like it, but I find that whenever I start to talk to them (PTH friends ), I have an intention to imitate their speech pattern. I am not sure whether I would like to become a part of their group or whatever reason b ecause not only me but also my friend

PAGE 156

156 shows such a tendency. But when I talk to them over an hour, I feel tired because I have to curl my tongue to imitate them all the time. Her comments reflect a general reaction from female speakers. Their higher frequency of using the standard retroflex feature corresponds with their comments. This tendency also indicates that TM female speakers seem to be more willing to adjust their speech pattern to accommodate to PTH speakers. It also shows that the female speakers prefer the standard retroflex feature as Su’s (2008) study suggests. That is, TM female speakers tend to avoid using Mandarin which carries the trace of Taiwanese phonological features, like the deretroflex feature , because Taiwaneseaccented Mandarin is usually associated with negative connotations in Taiwanese society. On the other hand, men are encouraged to use Mandarin which carries local Taiwanese features to display their masculinity, solidarity in the group, and their connection with locality. Su argued that Taiwanese society still regulates women to behave in certain ways such as being refined, and one way to perform their refinement is through proper language use. The gender ideology affects TM speakers’ attitude toward the use of the stan dard and nonstandard features. The above excerpts show how male and female TM speakers hold different attitudes toward accommodating to PTH speakers, and also reveal that the variants of the retroflex initials are linked with gender ideology, which requir es TM male and female speakers to behave in certain ways to conform to sociocultural norms. It was predicted that the distribution of the variants of the retroflex (zh) (ch) (sh) (r) would have a positive correlation with TM speakers’ strength of network with PTH speakers. It was suggested that when speakers’ social network is developed within a close knit community where the members participate in the same practice, they would

PAGE 157

157 be likely to share the same preference for a particular linguistic variable with their members (Eckert 2000; Milory 1980). A great number of studies (e.g. Cukor Avila 1997; Hirano 2010) have demonstrated the positive correlation between social network and speakers’ linguistic preference in a close knit community. The findings in the present study do not support the framework of social network . This study finds that TM speakers who have less strength of social network with PTH speakers tend to favor the retroflex feature, the distinctive feature in PTH speakers’ speech. On the other hand, the speakers who have greater streng th of social network with PTH speakers prefer the deretroflex feature, the distinctive feature in TM speakers’ speech. In questionnaire and interviews, TM speakers generally agree that the deretroflex feature is a distinctive feature in TM and also a maj or feature differing from PTH. As commented by PM2 (Age:33 SN: No contact LS: over 5 years), “TM has its own features and has been under great influence from Taiwanese, so in a sense it can represent Taiwan identity in addition to Taiwanese.” TM speakers with greater strength of network wit h PTH speakers may be more sensitive to the differences between PTH and TM because of their close relationship with PTH speakers. The excerpt from PF2 presents this tendency. With the network score 12, PF2 (Age: 24 SN: S trong LS: less than 6 months) was pl aced in the group of Strong Net work. When asked what she thought about TM and PTH, and w hether she would accommodate to her PTH friends by changing her speech pattern, she had the following responses: Question: Do you think whether TM and PTH are the same? In what way are they the same or different? PF2: No. they are not the same. The accent and some words are different.

PAGE 158

158 Question: Do you feel that you use TM differently when you talk to friends from Taiwan and from China? PF2: No. I stay the way I speak. I do not accommodate to them. PF2’s responses indicate that she is aware of linguistic differences between TM and PTH. Her responses correspond to the high frequency of the de retroflex feature (315 tokens of the palatoalveolar, and 3 tokens of the full dental; 48 tokens of the lateral), and the low frequency of the retroflex feature (0 token of the full retroflex; 1 token of the retroflex fricative) in her group interview. The low rate of the retroflex feature in PF2’s speech also suggests that when speakers hav e greater strength of network with PTH speakers, they may not share the general preference for the retroflex feature with their PTH interlocutors when other factors come into play affecting TM speakers’ linguistic choice. For example, during the interview, when asked if TM and PTH are different, PF7 (Age: 33 SN: No contact LS: over 5 years) replied: PF7: N ot exactly the same. The majority of language use and structure are the same so it does not cause any difficulties to communicate. When asked if she spoke TM differently when talking to friends from Taiwan and from Ch ina, she replied: PF7: I try to sta y the way I speak (TM) including intonation and words/terms (when I talk to friends from China). A ccording to the social network measure (see Chapter 4 for details), PF7’s s core for her network strength was zero so she was placed in the g roup of “No contact” but she had a relatively higher frequency of using the full retroflex (149 tokens) in comparison with other TM speakers who had no contact or have less strength of network with PTH speakers. For instance, PF9’s score of network strengt h was 6, so she was placed in the group of Weak network with PTH speakers. PF9 produced only 5

PAGE 159

159 tokens of the full retroflex feature. The major difference between the two speakers is that PF7 has been teaching Chinese as a second language in a local high sc hool for over 3 years. For her job needs, she is required to be able to use and teach the prescribed retroflex feature, as a result, a high frequency of the retroflex feature is shown in her data. In other words, when other factors come into play, even speakers who have strong social networks with their interlocutors may resist changing their choice of variants. This study initially hypothesized that the framework of social network would be a powerful factor to explore the correlation between TM speakers’ l inguistic choice and their interaction between PTH speakers as suggested by previous studies. However, the findings in this study show that the strength of social network is not the strongest factor determining the direction of accommodation and affecting speakers’ language use . The use of the retroflex and deretroflex features in different generations reflects different stages of the sociopolitical changes in Taiwanese society. At first, to eliminate the influence from Japanese sovereign before 1945 and to reinforce KMT political hegemony, the Mandarinonly policy was implemented overwhelmingly in Taiwan. PF12 (Age: 60 SN: Strong LS: over 5 years) grew up in a family in which Japanese and Taiwanese were the two mai n languages for daily communication. After 1945, KMT government required that every government employee use Mandarin as the main communication language. PF12 recalled that her parents, who only spoke Japanese and Taiwanese, were required to learn Mandarin when they were recruited as high school teachers. When asked where she learned Mandarin, she replied: PF12: “School, because everyone speaks Mandarin in school []. At that time, if you speak Taiwanese, you would be fined. This is very ridiculous!

PAGE 160

160 PF12’s reply illustrates the situation when the Mandarinonly policy began to be implemented in schools after 1945. Under K MT regime, learning Mandarin was mandatory and the local dialects such as Taiwanese were strictly forbidden in all public domains. She further commented: PF12: “ because we lived in the countryside, no one really cared about whether you spoke Taiwanese or Mandarin in schools. Unlike in metropolitan areas such as the capital, Taipei, the policy was strictly implemented. ” When asked how she defined standard and n onstandard Mandarin, she stated: PF12: The Mandarin we are speaking now is relatively standard. Lai: What is non standard Mandarin? PF12: The Mandarin is affected by Taiwanese. Taiwanese has 7 tones but Mandarin only has 4 tones. T onal patterns in Mandarin are not complicated as much as Taiwanese. PF12 ’s comments suggest that t he implement ation of the Mandarinonly policy has a certain impact on TM speakers’ language attitude toward the languages they used on a daily basis . Some speakers mentioned their experience regarding using Mandarin and Taiwanese in schools before the liberalization of language policy in 1987. For example, PM9 (Age:44 SN: Weak LS: over 5 years) in his interview mentioned: PM9: In fact, when we were kids, we [] were taught that speaking Taiwanese indicated lower social status. You must speak Mandarin. In Taipei, some Taipei born classmates at that time th ought that I spoke nonstandard Mandarin because I could not (properly) pronounce . But now, I feel that’s nons ense. Right? This is the way I am. There’s no reason for me to change. PF12 and PM9’s comments reveal the sociolinguistic status of standard Mandarin, nonstandard Mandarin, and Taiwanese, and the ideologies behind these languages in Taiwanese society. The proper use of the retroflex feature is viewed as a criterion for judging whether speakers can use “good Mandarin”, and is further

PAGE 161

161 associated with positive connotations such as high socioeconomic status and regional differences (Kubler 1985; Laio 2008; Tse 2000). The distribution of the variants of the retroflex fricative (r) further portrays the effect of socio political ch anges and language policy on TM speakers’ language use between three generations. The high frequency of the retroflex fricative [ ], the standard form, in the older generation suggests that the impact of diglossia on the use of the retroflex fricative (r) remains. When looking at the use of the retroflex (zh) (ch) (sh), the distribution of the variants tells a different story in the older generation. The high frequency of the palato alv eolar feature and the low frequency of the full retroflex feature are found in the older generation. The results found in the older generation demonstrate that the retroflex initials (zh) (ch) (sh) (r) have different sociolinguistic status. Brubaker (2012) points out that the palatoalveolar gradually becomes the normative form accepted widely by TM speakers as the standard form in Taiwanese society. Evidence from the skewed frequency of the palatoalveolar feature in all generations supports Brubaker’s claim. On the other hand, the distribution between the retroflex fricative [ ] and the lateral [l] seems to be less predictable. For the young generation, Mandarin has become the main communication tool between friends and family members although their par ents are Taiwanese native speakers. For example, when asked what language she first learned, Taipei born speaker PF1 (Age: 30 SN: No contact LS: 23 years) replied: PF1 : Mandar in. Lai : Can you speak Taiwanese? PF1: A little. Lai: Do your parents speak Taiwanese?

PAGE 162

162 PF1: Yes, they do. Lai: So at home do they usually speak Taiwanese or Mandarin when talking to you? PF1: Both Lai: If they talk to you in Taiwanese, will you respond in Mandarin? PF1: It depends. Sometimes I would respond in Taiwanese when I know how to say it in Taiwanese. The same situation also happens to other young speakers who were born in the south. For example, P M3 (Age:27 SN: Strong LS: 34 years) was born in Pingtung which is stereotypically considered as one of Taiwanesespeaking cities (Liao 2008). When aske d if he could speak Taiwanese, he replied : PM2: I have been us ing Mandarin since I was a kid. My family is somewhat particular . My parents t al k to each other in Taiwanese but when they talk to me or my elder brother, they use Mandarin. Lai: Why? PM2: hmmmI do not particularly ask them why. That is the way we use (the languages) in our family. Lai: So can you fully understand when people talk to you in Taiwanese? PM2: Yes, mostly. Lai: What about speaking Taiwanese? PM2: Not fluent. PF1 and PM2’s responses reflect that the young generation generally use Mandarin as their main communication medium to talk to their parents and peers. T his tendency corresponds to the quantitative results which show a high frequency of the retroflex feature. The preference for the retroflex feature in the young generation indicates that the influence of Taiwanese seems to be less than the older generation. In addition, among all generations, the use of the deretroflex feature is dominant. This

PAGE 163

163 tendency suggests that the use of the deretroflex feature has gradually become more normative than the retroflex feature in TM. In addition, the ideologies of the retroflex and deretroflex features are also changed within generations. When asked how they feel about hearing people speaking TM and PTH, the majority of speakers’ comment that they feel more comfortable and amicable being surrounded by the Mandarin that contains local Taiwanese features (i.e. deretroflex feature). Some speakers further indicate that the Mandarin with the deretroflex feature is accepted as the standard form in Taiwanese society. Using Mandarin carrying the deretroflex feature alongside other local dialects (Taiwanese) while being abroad make them feel intimate with their interlocutors if their Mandarin has TM feature, and further connected with their place of origin. The findings reveal that there is a positive correlation between the use of the retroflex feature and the length of stay in the US. The longer residence time in the US that TM speakers have, the more likely they favor the retroflex feature. On the contrary, the speakers who have a shorter residence time in the US would favor the deretroflex feature. This correlation suggests that TM speakers are likely to use the standard form, i.e. retroflex feature, accepted by the local community members as the normative form as means of identifying themselves as a member of the local Chinese immigrant community. This phenomenon reflects an ideology in Chinese immigrant community in which Chinese culture and the language are considered as social standards, so Chinese immigrants regardless of their place of origin use it as their identity when they need to present their identity outside the community. Silverstein (1998: 412) states that the standard variety of a language is

PAGE 164

164 “ at the neutral topand center of stratified society ; any deviations [ ] can only be interpreted as marked variants that index the demographic figuration of the producer as also being a corresponding ly removed one. ” Despite the fact that TM speakers may not consider that the retroflex feature is the feature in their Mandarin and feel strange if they hear their Taiwanese friends deliberately use or imitate this feature , for self protection in an immigrant environment, the use of the retroflex feature can emphasize their sense of belonging, and show their affiliation with the local Chinese community members by viewing Chinese culture and norms as social standards even if they are struggling to maintain their Taiwanese identity at the same time. When the population in the local immigrant community consists of Chinese and Taiwanese immigrants who share basic cultural norms , using the standard form to project the group identi ty and to underline the solidarity among community member seems to be a better way to protect themselves against the seemingly hostile host country . TM speakers identify the retroflex feature as a distinctive feature in PTH and a nonTM feature . The results from the correlation between the retrofl ex initials with different extralinguistic factors show that the use of the retroflex feature is associated with ideologies that are constructed according to the social norms in TM speakers’ place of origin, Taiwan. The social interpretations behind the variants of the retroflex (zh) (ch) (sh) (r) are not explicitly expressed. They emerge when they are applied in different social contexts. TM speakers may or may not be aware the variants of the retroflex (zh) (ch) (sh) (r) carry a pragmatic function or can project local identity, but they may consciously or subconsciously use them to decide whether they would assimilate or maintain their TM feature when they come into contact with PTH speakers.

PAGE 165

165 This section has discussed that the use of the retroflex and the deretroflex features is assigned social meanings and ideologies by TM speakers. The shared social interpretations behind the variants connect TM speakers with each other to form a unique membership in the Chinese and Taiwanese mixed immigrant community. Their membership with Chinese or with Taiwanes e community further affects their linguistic choice to create ingroup and out group communication. The next chapter includes revisiting research questions, broader implications, limitations of the present study, and suggestions for further research.

PAGE 166

166 C HAPTER 8 CONCLUSION The present study has shown the sociolinguistic variation of the retroflex initials (zh) (ch) (sh) (r) in TM speakers’ speech. The correlation between the retroflex initials and extralinguistic factors of social network strength, generation, gender, and length of stay in the US reveals that all four extralinguistic factors have statistical significance for affecting TM speakers’ linguistic choice of the variants of the retroflex initials. In addition, this study has analyzed the patterns of th e retroflex initials (zh) (ch) (sh) (r) in two different types of interviews. The findings reveal that the variants of the retroflex initials (zh) (ch) (sh) (r) are assigned particular social meanings and ideologies which are socioculturally constructed i n speakers’ place of origin. When TM speakers come into contact with PTH speakers in an immigrant setting, the ideologies and social interpretations emerge, and further affect TM speakers’ linguistic choice of the variants. This chapter closes with revisit ing research questions, and discusses broader implications, limitations of the study, and suggestions for further research. Revisiting Research Questions As Mandarin dialectal speakers from China and Taiwan view their homeland Mandarin as a part of Chinese and Taiwanese identity, the present study seeks to understand TM speakers’ linguistic behaviors in a foreign country and when they come into contact with PTH speakers in an immigrant community . For this purpose, it adopts a variationist approach to examin e the variations of the two linguistic variables: retroflex initials (zh) (ch) (sh) and retroflex fricative (r), and further investigates the correlation between the linguistic variables (zh) (ch) (sh) (r) and four extralinguistic variables: generation, gender, strength of social network, and length of stay in the US. In addition,

PAGE 167

167 in order to investigate to what extent that TM speakers maintain their TM feature or assimilate PTH speakers’ language use when they interact with PTH speakers in an immigrant set ting, the present study also compares 5 TM speakers’ linguistic behaviors in their individual and group interviews. The questionnaire and interview questions provide complementary information with regard to speakers’ language attitude toward standard Mandarin, PTH, TM, and local dialects. Their responses help us understand how the social meanings and ideologies of the linguistic variables, which are constructed according to the sociocultural norms in TM speakers’ place of origin, affect their linguistic be haviors when they meet PTH speakers in an immigrant setting. This study seeks to answer three research questions, which are repeated below, followed by a brief discussion in response to each question. 1. How and to what extent do linguistic variation of the retroflex initials, namely retroflex (zh) (ch) (sh) and retroflex fricative (r), occur in TM speakers’ speech in an immigrant setting? As presented in previous chapters, the variation of the retroflex (zh) (ch) (sh) includes the full retroflex ([ ], [h], []), the palato alveolar ([t ], [t ], [ ]), and the full dental ([ts], [tsh], [s]) . The results in Chapter 5 show that the predominant variant in TM speakers’ speech is the nonstandard palato alveolar despite the fact that the full retroflex is the standard and prescribed form that TM speakers learn as institutional form in schools. Furthermore, due to the influence of Taiwanese, the substitution of the full dental for the full retroflex i s commonly found in TM speakers’ speech, but the quantitative analysis shows that the full dental does not happen as frequently as initially predicted.

PAGE 168

168 The variation of the retroflex fricative (r) includes the retroflex fricative [ ] and the lateral [l]. The results in Chapter 6 show that the dominant variant is the nonstandard lateral [l]. The alternation between the retroflex fricative [ ] and the lateral [l] is commonly found in bilingual speakers of Taiwanese and Mandarin. In Chapter 5, a statistically skewed distribution of the palatoalveolar is found, showing that it always has more than 90% of frequency in comparison with the full retroflex which has a much lower frequency . This finding suggests that the use of the palatoalveolar for the full retroflex is stereotyped and overtly recognized by TM speakers. Unlike the skewed distribution in the variation of the retroflex (zh) (ch) (ch), the distribution of the retroflex fricative [ ] and the lateral [l] is not statistically extreme. Although the lateral feature is the domina nt variant, the lateral for the retroflex fricative behaves as a sociolinguistic marker that appears depending on different social factors and contexts. Furthermore, the correlation between social interpretations and the variants of the retroflex (zh) ( ch) (sh) (r) is constructed by TM speakers. They impose their perceptions of sociocultural norms and language ideologies from their place of origin to the variants of the retroflex (zh) (ch) (sh) (r). The process of linking the social meanings and ideolog ies with the retroflex (zh) (ch) (sh) (r) has been through different stages, and each stage pertains to the sociopolitical changes in TM speakers’ place of origin. Before the liberalization of the Mandarinonly policy, the standard forms of the full retro flex and the retroflex fricative enjoyed prestigious status and were indexed with positive connotations such as high socioeconomic class, sophistication, and high educational level. On the other hand, the nonstandard form, the palatoalveolar (and full dental) and the lateral was linked with negative social evaluations such as peasant

PAGE 169

169 accent and rudeness. Due to the rise of Taiwanese localism, the social meanings and ideologies attached to these variants have changed. The distribution of the retroflex initials demonstrates that TM speakers may have different degrees of awareness of the social meanings and ideologies behind the variants, which may directly or indirectly affect TM speakers’ choice of variants when they in an immigrant setting and when they come into contact with PTH speakers. 2. To what extent, if at all, do TM speakers assimilate their speech to a standard variety of the retroflex feature: (zh) (ch) (sh) and (r) or maintain TM linguistic feat ure when they come into contact with PTH speakers To examine whether TM speakers assimilate their speech to the standard form of the retroflex feature (the full retroflex feature and the retroflex fricative feat ure), 5 TM speakers are randomly selected to have an interview with their PTH friend or colleague. In terms of the distribution of the full retroflex and the palatoalveolar, the results indicate that in the group interview, the frequency of the full retro flex is decreased. In addition, although it does not have a greatly statistical difference, the use of the palatoalveolar is increased in the group interview. This distribution suggests that in the group interview, TM speakers maintain their speech patter n according to the increasing use of the palatoalveolar feature. With regard to the distribution of the retroflex fricative [ ] and the lateral [l] in the group interview, the frequency of the retroflex fricative [ ] is increased. The findings indicate that TM speakers increase the use the retroflex fricative [ ], the standard form, in the group interview. In other words, TM speakers ass imilate their speech to the standard variety of the retroflex fricative (r). On the other hand, the findings of the retroflex (zh) (ch) (sh) present a different pattern. As mentioned previously, the two

PAGE 170

170 linguistic variables have different sociolinguistic s tatus, in which the retroflex (zh) (ch) (sh) is a linguistic stereotype, while the retroflex fricative (r) is a linguistic marker. 3. How and to what extent do strength of social network, generation, gender, and the l ength of stay in the US affect the variable patterns and frequency of the retroflex initials (zh) (ch) (sh) (r)? The multivariate analyses reveal a correlation between the variable (zh) (ch) (sh) (r) and the extralinguistic factors of strength of social network, generation, gender, and the length of stay in the US. All four extralinguistic factors are significant factors affecting TM speakers’ production of variants of the retroflex (zh) (ch) (sh) (r). Regarding the correlation of the retroflex (zh) (ch) (sh) and four extralinguistic factors , gender di splays the strongest magnitude of effect , while t he strength of social network display s the weakest magnitude of effect affecting TM speakers’ production of the full retroflex feature. In terms of the correlation between the retroflex fricative (r) and four extralinguistic factors , gender displays the strongest magnitude of effect , while t he strength of social network displ ays the weakest mag nitude of effect affecting TM speakers’ production of the retr oflex fricative [ ]. The significance of the gender factor suggests that TM male and female speakers have different attitudes towards the use of the standard and the nonstandard form. TM female speakers tend to favor the standard form, while male speakers tend to favor the nonstandard form. Male and female speakers’ preference for the particular forms further reveals that Taiwanese traditional social norms still have effect on how men and women in Taiwanese society should use proper language to project and reinforce their gender image, i.e. femininity or masculinity . Furthermore, the result that social network strength displays the weakest magnitude of e ffect is unexpected. It is

PAGE 171

171 suggested that when speakers form a community in which they share a close knit network through their frequent interaction, they also construct a “community of practice” (Lave and Wenger 1991 as cited in Eckert 2000:35) in which they share social as well as linguistic norms. In this sense, TM speakers who have greater strength of n etwork with PTH speakers would share a general preference for the local linguistic form with PTH speakers in the local Chinese immigrant community. However, this is not the case in the present study. Instead, TM speakers who have less strength of network w ith PTH speakers tend to assimilate PTH speakers. The findings suggest that although speakers who have greater network strength with PTH speakers seem to build a close relationship with PTH speakers, they are also the ones who are more sensitive to the lin guistic differences between PTH and TM as well as the social meanings and ideologies behind the linguistic variables, and may be more careful about their choice of variants. Bro a der Implications To my knowledge, very few studies have explored the contact between the varieties of Mandarin. The present study is the first study that attempts to explore TM speakers’ linguistic behaviors in an immigrant setting and when they come into contact with PTH speakers. The present study has broader implications in three ways. First, as mentioned above, the present study has attempted to investigate TM speakers’ linguistic behaviors when they interact with PTH speakers. Under the sensitive political climate bet ween the Taiwanese identity and the Chinacentered culture, the exploration of TM speakers’ linguistic choice when they come into contact with PTH speakers in an immigrant setting has shed light on understanding the varieties of TM, and their correlation w ith extralinguistic factors. In addition, it has shown how TM

PAGE 172

172 speakers employ their linguistic resources when they come into contact with PTH speakers. It has further provided evidence that the variants of retroflex initials are assigned social meanings and ideologies , which affect TM speakers’ choice when they are in an immigrant setting and when they come into contact with PTH speakers in an immigrant community. Previous studies have shown that dialect contact results in linguistic changes in internal context to reach linguistic convergence. However, it should be noted that socio psychological factors also play an influential role affecting the directions of speakers’ accommodative intention. Second, the present study has contributed to the study of the varieties of Mandarin in the field of variationist sociolinguistics. To my knowledge, variationist sociolinguistic studies applied to Mandarin studies are still very limited, especially the investigation of contact between speakers from China and from other Chinese speaking regions such as Hong Kong, Singapore, and Malaysia, which comprise of a great number of speakers of Chinese dialects such as Mandarin, Cantonese, and other Chinese dialects. Due to contact with the different languages and a diversity of local dial ects, the Mandarin spoken in these regions may also carry linguistic features of the local dialects and languages. It may be interesting to explore the variations in the Mandarin in these regions, and the speakers’ linguistic behaviors when they come into contact with PTH speakers. The present study has shown that the influence of Taiwanese adds variations to the Mandarin used in Taiwan, and also revealed that the linguistic variables can be associated with particular social meanings and ideologies construc ted according to the social and cultural norms in speakers’ place of origin. These ideologies that speakers carry would also affect speakers’ linguistic behaviors in

PAGE 173

173 an immigrant setting. It will be a good idea to explore the varieties of Mandarin in the speech of the speakers from these regions and further investigate their linguistic behaviors when they contact with PTH speakers in the context of globalization. A comparative study of the speakers from these regions when they come into contact with PTH spe akers in the immigrant environment would provide a more complete understanding of the varieties of Mandarin. In addition, it would help both educators and learners of learning Chinese as foreign language to better recognize different varieties of Mandarin, and also to help learners be prepared for real conversations with a diversity of native speakers of Mandarin with different accents. Third, the present study has adopted the variationist approach as the main method of quantitative analysis. It has demonstrated how statistical results including multivariate analysis can provide a more comprehensive understanding in terms of the pattern of the retroflex feature (zh) (ch) (sh) (r) corresponding with external factors. Multivariate analysis is not widely applied to analyze the significance of external factors in Chinese studies. The present study includes quantitative and qualitative analyses to provide a detailed account of the pattern of the retroflex variables (zh) (ch) (sh) (r), the correlation between the linguistic variables and the ex ternal factors, and the ideologies behind the variants of the linguistic variables. The design of the present study can be the basis for further research of other linguistic variables in Mandarin, and the varieties of Mandarin used in different Chinese speaking regions to provide a more complete picture of linguistic variation and change in Mandarin and speakers’ language use . Limitations of the Study The present study focuses on TM speakers’ linguistic behaviors in an immigrant setting an d when they have an opportunity to interact with PTH speakers. It reveals the

PAGE 174

174 patterns of (zh) (ch) (sh) (r) variation are closely related to extralinguistic factors of gender based on the statistical results. Furthermore, the social meanings and language ideologies behind the variants of the retroflex (zh) (ch) (sh) (r) affect TM speakers’ language use when they have contact with PTH speakers. However, there are several limitations of the present study. The first limitation is lack of speakers who are younger than 22 years and older than 60 years old due to the nature of the target immigrant community. The community is located in a college town so the community members mainly consist of young students between 22 years old to 35 years old, which is categoriz ed as the young generation, and immigrants between 36 and 60 years old, which involves the middleaged generation and older generation. Lack of information from speakers who are under 22 years old and older than 60 years old in the present study may not pr ovide a complete picture for TM speakers’ linguistic behaviors across generational stratifications . The second limitation is that TM speakers ’ educational level and socioeconomic status are relatively homogenous in this study . Previous studies (e.g. Tse 2000) indicate that the use of the retroflex feature is associated with TM speakers ’ socioeconomic status. The target community is located in a college town in which the majority of TM speakers come to pursue their master or PhD degree after they graduate from undergraduate or master program , so the speakers generally have high educational level s in the present study. In addition, the majority of longterm TM residents and immigrants hold a PhD degree so most of them are the professors, researchers, or staffs in the university or work in biological lab in private companies . A comparative study that adds different educational levels as one of extralinguistic factors would show

PAGE 175

175 a more complete picture of the use of retroflex and deretroflex features in the speech of TM speakers. The third limitation is the design of the group interview. This study expected to obt ain more spontaneous speech from a natural interaction between TM and PTH speakers. Although the design of the group interview in the present study was able to obtain natural speech from the interaction between TM and PTH speakers, the influence of observe r ’ s paradox on both TM and PTH speakers was observed during the interview. For instance, one or two PTH speakers were not comfortable to be interviewed together with the researcher who they did not know before the interview and were aware of being recorded. In addition, each group interview includes 2 TM speakers and 1 PTH speaker. TM speakers sometimes signified the researcher to participate in their conversation so sometimes she participated in the conversation for a short time in order to keep the conver sation flowing. Unequal number of Taiwanese and Chinese participants in the group interview seems to affect TM speakers’ language use. To minimize the effect of observer paradox, it would have been better to have at least four speakers including 2 TM speak er and 2 PTH speakers who know each other in the group interview so it would be more like a natural conversation. Lastly, the forth limitation is the coding method. In the present study, t he coding method is the impressionistic judgment which depends on the researcher ’ s perception of the target linguistic variable. To secure the reliability of coding results , a co coder is hired to code the data blindly. When there are disagreement s regarding the perception of variants, they are deleted to obtain t he consistent results from the res earcher and the co coder. Although deleting conflicting judgments ensures the consistent results

PAGE 176

176 between the researcher and the cocoder , it may not be the best method. When a study contains a large data set such as the present study, deleting conflicting judgment may be a convenient method to reach consistent results between the researcher and the cocoder, but hiring a third cocoder to reexamine the conflicting coding results would be a better method. When the study has a smaller data set, it is suggested that employing spec trogram would be a better and an objective method to resolve the conflict judgments between the researcher and the cocoder . S uggestions for Further Research The first suggestion for further research is to include speakers with different generations in the group interview. In the present study, due to the nature of the target community, the generation in the group interview only include two generations, the young (under 35 years old) and the middleag ed (3645 years old). The data from younger speakers (less than 24 years old) and older (more than 60 years old) in the group interviews is not able to be obtained. Therefore, it would be a good idea if further research could collect the data from speakers who are under 24 years old and more than 60 years old in the group interview. Including a wider generational range can provide further research with more evidence regarding the generationstratified patterns of the retroflex initials when TM speakers interact with PTH speakers, and also examine if there is an ongoing change in regard to the ideologies behind the variation of the retroflex initials . The second suggestion is to include a PTH speaker as a cocoder. As implied in the limitation section above, a cocoder was hired to code the data blindly in order to secure the reliability of coding results. Since the cocoder in the present study is a TM speaker with linguistic background, an examination from PTH cocoder may provide

PAGE 177

177 different observations when the deretroflex feature is the distinctive feature in TM and stereotyped as TM feature. In addition, the data could also be analyzed by the two c o coders to examine whether the accommodation takes place bi directionally. The present study focuses on TM speakers’ linguistic choice in an immigrant setting. The data from PTH speakers is not examined throughout. It would be a good idea to investigate P TH speakers’ linguistic behaviors in an immigrant setting and when they come into contact with TM speakers . The third suggestion is to conduct similar research in a larger immigrant community in metropolitan cities where the population of immigrants consists of a diversity of Chinese dialect speakers. Previous studies such as Pan (1994) and Chi (1991) investigate Chinese immigrants’ accommodat ion pattern when they have contact with speakers from Taiwan and with different Chinese dialect speakers from China in China town in New York City and in Los Angeles. Their studies focus on the linguistic accommodation of Mandarin speakers from China. For my knowledge, the present study is the first study examining TM speakers’ linguistic behaviors in an immigrant community and can be the basis of further research in a larger Chinese and Taiwanese immigrant community. A similar research conducting in a mixed culture and language immigrant community in metropolitan cities would provide more insights and a complete picture regarding the variation of Mandarin and Mandarin speakers’ linguistic behaviors . Closing My experience in the use of the retroflex feature motivated me to start this dissertation. I was curious about whether other TM speakers would behave as I did, by adjusting their use of the retroflex feature, or whether they would maintain their use of TM feature when they are in an immigrant community and when they have opportunities

PAGE 178

178 to interact with PTH speakers. Due to struggle between the historical attachment to China for ethnic roots and the rise of Taiwanese localism for Taiwancentered identity, peo ple in Taiwan have an ambivalent attitude towards their identity of being Chinese or being Taiwanese. In Taiwan, people are highly sensitive to the political ideologies and ethnic identity. It is suggested that the political ideologies are correlated with people’s language use because their choice of linguistic varieties, e.g. Mandarin, Taiwanese, or Taiwaneseaccented Mandarin, provide the clue for people to identify people’s political inclination and their ethnic identity (Liao 2008). Due to the complex s ociolinguistic background in Taiwanese society, it would be interesting to explore whether there are any other ideologies that are socioculturally and politically constructed in Taiwanese society would affect TM speakers’ linguistic use when they are away from their home country and have opportunities to be in contact with PTH speakers in an immigrant community in which there is no dominant dialect. The findings of this dissertation shows that TM speakers’ linguistic patterns of the retroflex initials (zh) (ch) (sh) (r) are correlated with four extralinguistic factors. It further reveals that social meanings and ideologies behind the variants of the retroflex (zh) (ch) (sh) (r) are assigned by TM speakers according to the sociocultural norms in their pl ace of origin. Although these social meanings and ideologies are never overtly expressed, they emerge via the interplay between the linguistic variable and extralinguistic factors. This dissertation hope s to encourage future research ers to conduct a similar study in the speech of speakers of varieties of Mandarin in an immigrant setting and when they come into contact with PTH speakers, and to provide a more comprehensive account of varieties of Mandarin in contact in the context of globalization.

PAGE 179

179 APPENDIX A INFORMED CONSENT FORMS

PAGE 180

180

PAGE 181

181

PAGE 182

182 APPENDIX B QUESTIONNAIRE Thank you for participating in this research. This questionnaire comprises three sections: demographic questions, social network, and openended questions. You can fill out this questionnaire in English or in Mandarin Chinese. I. Demographic Questions ( 1. Gender a. Male [ ] b. Female ) [ ] 2. Date of Birth 0PF,!D. _______________ 3. What is your occupation? ? _______________ 4. What is your marital status? a. Single b. Married -* [ ] (Please continue 4.1 – 4.4) ZsN$N4$g 4.1 – 4.4 [ ] (Please continue 4.5 – 4.6) ZsN$N4$g 4.5 – 4.6 4.1 Do you have roommates? 0P6+L#s%vp Yes [ ] (Please continue 4.2 – 4.4) 4.2 – 4.4 No [ ] (Please continue Question 5) 5 4.2 Do you have roommates who are Putonghua speakers? Yes [ ] (Please continue 4.3) 4.3 No [ ] (Please continue Question 5) 5 4.3 How many? ? 4.4 How long have you lived together? _______________

PAGE 183

183 II. Social Network ( Name Relationship Kind of Mandarin a. b. c. d. e. 4.5 What language/dialect does your spouse speak at home? a. Mandarin &ZF [ ] b. Taiwanese #ZF [ ] c. English RZF [ ] c. Others !~ ___________ 4.6 If you have a child/children, what language/dialect do they speak at home? a. Mandarin &ZF [ ] b. Taiwanese #ZF [ ] c. English RZF [ ] d. Others !~ ____________ 5. How long have you been in the US? 0P&–O6&/-(.p a. Less than 6 months [ ] b. 1 2 years [ ] c. 3 4 years [ ] c. 5 years above [ ] 6. Are you willing to participate in a group interview for the second phase of this research? a. Yes [ ] b. No h\267 [ ] 7. Which of the following Mandarin that most of your friends speak? bF,6#s6ZR$’ I/p a. Guoyu [ ] b. Putonghua ’&F,6_Z [ ] c. Others !~ ______________ 8. Please name five people in your network and identify their Mandarin e.g., Guoyu or Putonghua , and the relationship between you and them e.g., family, friend, office colleague or acquaintance.

PAGE 184

184 9. Please rank (15) each individual above based on the closeness between you and him/her (1: least close; 5: closest). 8 (1 5) 1: ; 5: ______________ Most close 5 Somewhat close 4 Neutral 3 Not very close 2 Least close 1 ______________ Most close 5 Somewhat close 4 Neutral 3 Not very close 2 Least close 1 ______________ Most close 5 Somewhat close 4 Neutral 3 Not very close 2 Least close 1 ______________ Most close 5 Somewhat close 4 Neutral 3 Not very close 2 Least close 1 ______________ Most close 5 Somewhat close 4 Neutral 3 Not very close 2 Least close 1 10. Please write down the frequency of faceto face meetings for each individual in 8 within a week. 8 Name Frequency of Faceto Face Meeting a. b. c. d. e. 11. Please write down the frequency of conversation via telephone call for each individual in 8 within a week. 8 Name Frequency of Conversation via Telephone calls a. b. c. d. e.

PAGE 185

185 III. Open ended Questions ( 1. Do you think whether Guoyu and Putonghua are the same? In what way are they the same or different? Please give examples. 2. Do you feel more comfortable to hear people speaking Guoyu or Putonhua? Why? 3. Do you feel that you use Guoyu differently when you talk to friends from Taiwan and from China? Please give examples if differences exist. 4. Taiwanese and Guoyu , which do you think that presents Taiwan identity? Why? (If none of the above languages is applicable, in your opinion, what language/dialect presents Taiwan identity? Why?) / / Thank you for your cooperation!

PAGE 186

186 LIST OF REFERENCES Anderson, Bridget L. 2002. Dialect leveling and /ai/ monophthongization among African American Detroiters. Journal of Sociolinguistics 6: 86 98. Babel, Molly. 2010. Dialect divergence and convergence in New Zealand English. Language in Society 39: 437 4 56. Baran, Dominika. 2014. Linguistic practice and identity work: Variation in Taiwan Mandarin at a Taipei county high school. Journal of Sociolinguistics 18:3249. 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. Bayley, Robert, Norma L. Crdenas, Belinda Trevio Schouten, and Carlos Martin Vlez Salas. 2012. Spanish Dialect Contact in San Antonio, Texas: An Exploratory Study. In Kimberly Geeslin and Manuel Daz Campos (eds.) Selected Proceedings of the 14th Hispanic Linguistics Symposium . 4860. Somerville, MA: Cascadilla Proceedings Project. Bell, Allan. 1984. Language style as audience design. Language in Society 13: 145204. Blommaert, Jan. 2010. The Sociolinguistics of Globalization. Cambridge University Press. Bourhis, Richard Y. and Howard Giles. 1977. The language of intergroup distinctiveness. Language, ethnicity and intergroup relations 13 : 119136. Brubaker, Brain. 2012. The normative standard of Mandarin in Taiwan: An analysis of var iation in metapragmatic discourse. Unpublished PhD dissertation. Pittsburgh, Pennsylvania: University of Pittsburgh. Chao, Yuenren. 1943. Lang uages and dialects in China. The Geographical Journal 102: 6366. Cheng, Robert L. 1985. A comparison of Taiwanese, Taiwan Mandarin, and Peking Mandarin. Language 61 : 352 77 Chi, Hong. 1991. Social networks and linguistic accommodation of Mainland Chinese in an urban American Chinese community. Unpublished Ph D dissertation. Louisiana State University. Chiung, Wi vun Taiffalo. 2001. Language and ethnic identity in Taiwan. In 7th Annual North American Taiwan Studies Conference, June 2001, 2325

PAGE 187

187 Coblin, South W. 2000. A brief history of Mandarin. Journal of the American Oriental Society 120: 537552. Coupland, Nikolas. 1984. Accommodation at work: some phonological data and their implications. International Journal of Sociology of Language 46: 4970. Cukor Avila, Patricia. 1997. An ethnolinguistic approach to the study of rural Southern AAVE. In Cy nt hia Bernstein, Thomas Nunnally and Robin Sabino (eds.) Language variety in the South revisited. Tuscaloosa : University of Alabama Press. 447 462 . Chung, Karen Steffen . 2001. Some returned loans: Japanese loanwords in Taiwan Mandarin. Language Change in East Asia . Surrey: Curzon. 161179. Chung, Karen S. 2006. Hypercorrection in Taiwan Mandarin. Journal of Asian Pacific Communication 16: 197 214. Dede, Keith. 2006. Standard Chinese and the Xining dialect: the rise of an interdialect standard. Jour nal of Asian Pacific Communication 16: 221238. Dyer, Judy. 2002. ‘We all speak the same round here’: Dialect levelling in a Scottish English community. Journal of Sociolinguistics 6: 99116. Eckert, Penelope. 1989. The whole woman: Sex and gender differences in variation. Language Variation and Change 1: 245267. Eckert, Penelope. 1997. Age as a sociolinguistic var iable. In Florian Coulmas (ed.) The H andbook of S ociolinguistics . Malden: Bl ackwell Publisher. 151– 167 . Eckert, Penelope. 2000. Linguisti c Variation as Social Practice: The Linguistic Construction of Identity in Belten High . Malden, Massahchusetts: Blackwell. Erbaugh, Mary S. 1995. Southern Chinese Dialects as a Medium for Reconciliation within Greater China. Language in Society 24: 7994. Freeman, Norman H. and Gustav M. Habermann. 1996. In Michael H. Bond (ed.) Linguistic socialization: Chinese Perspective. 7992. Gal, Susan and Judith Irvine. 1995. The boundaries of languages and disciplines: How ideologies construct difference. Social Research 62: 965 1001. Giles, Howard, Donald M. Taylor and Richard Bourhis. 1973. Towards a theory of interpersonal accommodation through language: Some Canadian data. Language in Society 2: 177192.

PAGE 188

188 Giles, Howard, Nikolas Coupland and Justine Coupland. 1991. Accommodation theory: Communication, context, and consequence. Contexts of Accommodation. Cambridge & Paris: Cambridge University Press & Editions de la maison d es sciences de l AZhomme. 1 68. Gong, Yuanyuan, Chow Irene H. S and David Ahlstrom. 2011. Cultural diversity in China: Dialect, job embeddedness, and turnover. Asian Pacific Journal of Management 28: 221238. Gordon, M atthew. 2006. Interview with William Labov. Journal of English Linguistics 34: 332351. Guy, Gregory. 2009. GoldVarb: still the gold standard. Workshop at the 38th Conference on New Ways of Analyzing Variation. Hernndez, Jos Esteban. 2009. Measuring rates of wordfinal nasal velarization: the effect of dialect contact on ingroup and out group exchanges. Journal of Sociolinguistics 13: 583 612. Her, One Soon. 2010. Lun Taiwan Huayu de zaidihua (On the indigenization of Taiwan Mandarin) [In Chinese]. Journal of Macau Linguistics Association 35: 1929. Hinskens, Frans. 1998. Dialect leveling: a twodimensional process. Folia Linguistica 32: 3551. Hirano, Keiko. 2008. L1 dialect contact in an L2 setting: int ervocalic /t/ in the Anglophone community of Japan. Essex Graduate Student Papers in Language and Linguistics 10: 45 75. Hirano, Keiko. 2010. Modification in the pronunciation by native speakers of English as an effect of social networks with Japanese speakers of English. Paper presented at the 18th Sociolinguistic Symposium, 3rd September, Southampton, UK. Hock, Hans Henrich and Brian D. Joseph. 2009. Language history, language change, and language relationship: an introduction to historical and com parative linguistics . New York: Mouton de Gruyter. Hsieh, Shelley C. Y. 2005. Taiwanese Loanwords in Mandarin Chinese: Language interaction in Taiwan. Taiwan Papers 5: 1 20. Johnstone, Barbara. 2004. Place, globalization, and linguistic variation. In Fought, Carmen ( ed. ) Sociolinguistic variation: Critical reflections . Oxford University Press . 65 83.

PAGE 189

189 Johnstone, Barbara, Jennifer Andrus and Andrew E. Danielson. 2006. Mobility, indexicality, and the enregisterment of “Pittsburghese”. Journal of English Linguistics 34: 77104 . Johnstone, Barbara. 2010. Indexing the local. In Nikolas Coupland (ed.) The H andbook of Language and G lobalization. Oxford: Oxford University Press. 386405. Juan, YuTien. 2011. Chinese Teenagers Perceptions o f and Attitudes toward Taiwan Mandarin. Unpublished PhD Dissertation. University of Edinburgh. Kerswill, Paul and Ann Williams. 2000. Creating a new town koine: children and language change in Milton Keynes. Language in Society 29: 65 115. Kubler, Cornelius C. 1985. The influence of Southern Min on the Mandarin of Taiwan. Anthropological Linguistics 27: 15676. Kuo, YunHsuan. 2003. The ‘Founder Principle’ and the origins of linguistic change in contact situations: a sociohistorical linguistic investigation of retroflex initials & apical vowels in Taiwanese Mandarin. Essex Graduat e Student Papers in Language & Linguistics 5: 54 75. Lai, Xun hui. 2006. The role of Japanese loanwords in Taiwanese: review of education policy in Taiwan [in Chinese]. Unpublished MA thesis. Taipei, Taiwan: Soochow University. Labov, William. 1963. The social motivation of a sound change. Wor l d 19: 273309. Labov, William. 1966. The S ocial S tratification of English in New York City . Washi ngton DC: Center for Applied Linguistics. Labov, William. 1972. Language in the Inner City: Studies in the Black English Vernacular . Philadelphia: University of Pennsylvania Press. 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 class in the course of linguistic change. Language Variation and Change 2: 205254. Labov, William. 2001. Principles of Linguistic Change– Volume 2: Social Factors . Malden and Massachusetts: Blackwell. Lee, Yen Ling, 1981. A study on codeswitching in Taiwan. Studies in the linguistics science s 11: 121136.

PAGE 190

190 Lehmann, Winfred P. 1975. The common language and the language of everyday life. In Winfred P. Lehmann (ed.) Language and Linguistics in the People’s Republic of China . Austin and London: University of Texas Press. 1040. Li, Charles N. and Sandra A. Thompson. 1979. Chinese: dialect variation and language reform. In Timothy Shopen (ed.) Languages and Their Status . Cambridge, MA: Winthrop. 295335. Li, Chen C hing. 1983. The Sociolinguistic context of Mandarin in Taiwan: trends and developments. I n Chauncey Chu, W. South Coblin and Feng Fu Tsao (eds.) Papers from the Fourteenth International Conference on SinoTibetan Languages and Linguis tics . Taipei: Student Book Co. 259 278 . Li, David ChenChing . 1985. Problems and trends of standardization of Mandarin Chinese in Taiwan. Anthropological Linguistics 27: 122 140. Li, Ling. 1991. Language variation and change in Mandarin: Mainland Chinese speech community in the US. Unpublished MA Thesis. East Lansing, Michigan: Michigan State University. Liao, Silvie. 2008. A perceptual dialect study of Taiwan Mandarin: Language attitudes in the era of political Battle. In Marjorie K.M. and Hana Kang (eds.) Proceedings of the 20th North American Conference on Chinese Linguistics (NACCL20) . Columbus, Ohio: The Ohio State Univer sity. 391 408. Liao, Sze Wei. 2010. Identity, ideology, and language variation: A sociolinguistic study of Mandarin in central Taiwan. Unpublis hed PhD dissertation. Davis, California: University of California at Davis. Lin, YenHwei. Consonant variation in Taiwan Mandarin. In Kathleen Ferrara, Becky Brown, Keith Walters and John Baugh (eds.) Proceedings from the Sixteenth Annual Conference on New Ways of Analyzing Variation in Language. Austin, Texas: The University of Texas at Austin, Department of Linguistics. 200208. Milroy, James and Lesley Milroy. 1978. Belfast: Change and variation in an urban vernacular. Sociolinguistic patterns in B ritish English . 19 36. Milroy, Lesley. 1980. Language and Social Networks . Baltimore: University Park Press. Milroy, James and Lesley Milroy . 1997. Varieties and variation. In Florian Coulmas (ed.) The Handbook of Sociolinguistics . Oxford, UK: Blackwell Publishers . 47 64. Milroy, Lesley and Matthew Gordon. 2003. Sociolinguistics . Oxford: Blackwell. Nijman, Jan. 1999. Cultural globalization and the identity of place: The reconstruction of Amsterdam. Cultural Geographies 6: 146164.

PAGE 191

191 Ntzel, Da niel and Joseph Salmons. 2011. Language contact and new dialect formation: evidence from German in North America. Language and Linguistics Compass 5: 705717. Pan, Shiwen. 1994. A sociolinguistic study of Mainland Putonghua and Taiwan Guoyu in New York City. Unpublished PhD Dissertation. City University of New York. Paolillo, John C. 2002. Analyzing Linguistic Variation: Statistical Models and Methods . Stanford. Center for the Study of Language and Information. Pi, Tony. 1991. Retrogression of retro flexion in Mandarin speech. Toronto Working Papers in Linguistics 11: 85 90. Poplack, Shana. 1990. Variation theory and language contact: concept, methods and data. In Papers for the Workshop on Concepts, Methodology and Data, 12 13 January , Basel . Strasbourg: European Science Foundation. 3366. Otheg uy, Ricardo, Ana Celia Zentella and David Livert. 2007. Language and dialect contact in Spanish in New York: toward the formation of a speech community. Language 88: 770 802. Ramsey, Robert S. 1987. The Languages of China. Princeton University Press. Ratte, Alexander Takenobu. 2011. Contact induced phonological change in Taiwanese. Unpublished PhD thesis . Columbus, Ohio: The Ohio State University . Romaine, Suzanne (ed.). 1982. Sociolinguistic Variation in Speech Communities . London: Edward Arnold. Romaine, Suzanne. 1982. What is a speech commu nity? In Suzanne Romaine (ed.) Sociolinguistic Variation in Speech Communities . London: Edward Arnold. 13 24. Sanders, Robert and Satoshi Uehara. 2007. Gender, bilingualism and vowel inventory in Taiwan Mandarin. 15: 19 36. Sank off, Gillian. 2001. Linguistic outcomes of language c ontact. In Jack Chambers, Peter Trudgill and Natalie Schilling Estes (eds.) The Handbook of Sociolinguistics . Blackwell. 638668. Schiffrin, Deborah. 1994. Approaches to Discourse. Oxford, UK: B.Blackwell. Schilling Estes, Natalie. 2002. On the nature of isolated and post isolated dialects: Innovation, variation and differentiation. Journal of Sociolinguistics 6: 64 85.

PAGE 192

192 Shockey, Linda. 1984. All in a flap: long term accommodation in phonology. International Journal of the Sociology of Language 46: 87 95. Sun, Chaofen. 2006. Chinese: A Linguistic I ntroduction. Cambridge University Press. Poplack, Shana, David Sankoff and Christopher Miller. 1988. The social correlates and linguistic process of lexical borrowing and assimilation. Linguistics 26: 47 104. Po plack, Shana and Sali Tagliamonte. 2001. African American English in the Diaspora. Malden, Massachusetts: Blackwell. Sharma, Devyani. 2011. Style repertoire and social change in British Asian English. Journal of Sociolinguistics 15: 464492. Skeldon, Ronald. 2003. The Chinese diaspora or the Migration of Chinese peoples? In Laurence J.C. Ma and Carolyn Cartier (eds.) The Chinese Diaspora: Space, place, mobility, and identity . Rowman & Littlefield. 51 66. Silverstein, Michael. 1998. Contemporary transformations of local linguistic communities. Annual Review of Anthropology , 401 426 . Silverstein, Michael. 2003. Indexical order and the dialectics of sociolinguistic life. Language and Communication 23: 193229. Standford, James N. A sociotonetic analysis of Sui dialect contact. Language Variation and Change 20: 409450. Su, His Yao. 2008. What does it mean to be a girl with qizhi? Refinement, gender and language ideologies in contemporary Taiwan. Journal of Sociolinguistics 12: 334358. Su, His Yao. 2011. The meaning contestation of Tai: Language ideologies and the global local ambivalence. Concentric: Studies in Linguistics 37: 283308. Swihart, D AW. 2003. The two Mandarins: Putonghua and Guoyu. Journal of the Chinese Language Teachers Association, 38: 103 118. Tagliamonte, Sali A. 2009. Analysing Sociolinguistic Variation . New York: Cambridge University Press . Tagliamonte, Sali A. 2010. Using statistical tools to explain linguistic variation. Workshop at the 39th Conference on New Ways of Analyzing Variation. Tagliamonte, Sali A. 2012. Variationist Sociolinguistics: Change, Observation, Interpretation . Oxford: Wiley Blackwell.

PAGE 193

193 Thomason, Sarah G. and Terrence Kaufman. 1988. Language Contact, Creolization and Genetic Linguistics . Berkeley: University of California Press. Thomason, Sarah G. 2001. Language contact: an introduction . D.C.: Georgetown University Press. Tong, Yuk Yue, Ying Yi Hong , Sau Lai Lee and ChiYue Chiu. 1999. Language use as a carrier of social identity. International Journal of Intercultural Relations 23: 281296. Trudgill, Peter. 1972. Sex, covert prestige and linguistic change in the urban British English of Norwich. Language in Societ y 1: 179195. Sandel, Todd L., WenYu Chao and Chung Hui Liang. 2006. Language shift and language accommodation across family generat ions in Taiwan. Journal of Multilingual and Multicultural Development 27: 126 147. Trudgill, Peter. 1986. Dialects in Contact. Oxford: Blackwell. Trudgill, Peter, Elizabeth Gordon, Gillian Lewis and Margaret MacLagan. 2000. Determinism in new dialect formation and the genesis of New Zealand English. Journal of Linguistics 36: 299 318. Tse, John Kwock Ping. 2000. Language and a rising new identity in Taiwan. International journal of the sociology of language 143: 151164. van den Berg, Marinus E. 1992. Ethnolinguistic identities and accommodation across generation in Taiwan. Journal of Asian Pacific Communication 3: 145164. Yang, Karen ChenYi. 2008. Sociolinguistic variation of Mandarin Chinese /f/ and /hw/ in Taiwan. Unpublished MA thesis. Taichung, Taiwan: Providence University. Zentella, Ana Celia. 1990. Lexical leveling in four New York City Spanish dialects: linguistic and social factors. Hispania 73: 1094 1105. Zhang, Qing. 2005. A Chinese yuppie in Beijing: phonological variation and the construction of a new professional identity. Language in Society 34: 431466. Wei, Jennifer M. 2006. Language choice and ideology in multicultural Taiwan. Language and Linguistics 7: 87 107. Williams, Salla. 2011. The Change in the Pronunciation of /e /, /a / and /i:/ in the Speech of Australian English Speakers Living in Finland. Unpublished MA Thesis. University of Tampere. Weinreich, Uriel. 1966. Languages in contact . The Hague: Mouton

PAGE 194

194 Winford, Donald. 2003. An introduction to contact linguistics . Oxford: Blackwell Publishing.

PAGE 195

195 BIOGRAPHICAL SKETCH Yu Ning Lai wa s born and raised in ChaoZhou, Taiwan. She received her bachelor’s degree in the Department of Foreign Language and Literature at Chung Hua Univers ity, Hsin Chu, in 2001. Before completing her doctorate, she earned an M.A. in Teaching English to Speakers of Other Languages and Applied Linguistics from Southern Illinois University at Carbondale, Carbondale, in 2009. She received her Ph.D. in Linguistics from the University of Flor ida in the s ummer of 2016 .