|UFDC Home||myUFDC Home | Help|
Links for objects in Zoe Ziliak (3128 8108) dissertation Object 3 1 Title: Continuum of Job~Jab Stimuli Permanent Link: http://ufdc.ufl.edu//IR00001187/00001 Object 3 2 Title: Continuum of Bad~Bed Stimuli Permanent Link: http://ufdc.ufl.edu//IR00001188/00001 Object 3 3 Title: Continuum of Dud~Dead Stimuli Permanent Link: http://ufdc.ufl.edu//IR00001189/00001 Object G 1 Title: Zoe Ziliak Dissertation Chapter 5 Production Results Mixed Effects Regressions Permanent Link: http://ufdc.ufl.edu//IR000 01186/00001
1 THE RELATIONSHIP BETWEEN PERCEPTION AND PRODUCTION IN ADULT By ZOE LYNN ZILIAK 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 2012
2 2012 Zoe Lynn Ziliak
3 To my parents
4 ACKNOWLEDGMENTS First and foremost, I thank my brother Zach and his wife, Lisa, for hosting me (rent free) for three months while I collected data in Chicago. They opened their home to me and made me feel welcome, even while (unbeknownst to me) they were expecting my god daughter and niece, Jillian. I realize how stressful having a three month houseguest is, but they were unendingly gracious. I also thank Zach for writing several Praat scripts and Excel macros for me and for lending his advice on some analyses. I thank my whole family, Mom, Dad, Zach, and Zeke, for love and support from birth through the completion of this degree. You have all encouraged me in every way possible. I could never have completed this work without the support of my husband, Dan Michel. Ove r the course of this degree, he has gone from acquaintance to boyfriend to fianc to husband, and at all these steps, he has been a rock of support, convincing me I could continue even when I was most disheartened. Next, I thank my committee members. Hl ne Blondeau first taught me about sociolinguistics, and she has always been enthusiastic about my ideas and scholarship. She has been tremendously supportive during all my years of analyzing data. Hans Van de Velde got me interested in sociophonetics and suggested the readings that led to my topic. He is always appropriately demanding and very giving of his time. Ratree Wayland has been both patient and generous with her time. She had many great suggestions on my methodology and perception stimulus cre contagious enthusiasm for my research is greatly appreciated. I am also grateful to her for forcing me to live up to my potential by picking a challenging and worthwhile
5 research topic. Finally, Jessi Aaron is always helpfully nitpicky about both my methodology and my word choices. This dissertation research was supported by a Sigma Xi Grant in Aid of Research and a National Science Foundation Doctoral Dissertation Improvement Grant, Award ID 0843403. Linda Isaacs at the Univ helped me with the preparation of my NSF budget; I would never have gotten the grant without her help. Walt Wolfram and Natalie Schilling Estes allowed me to use their diagram of the NCS in my NSF applica tion. Jolee Gibbs did the extensive work of managing my grant and processing all my receipts for reimbursement. Ann Schulte, J.J. (I lost his last name!), Alex Arnesen, and Sherri Cwik gave their time to make recordings for me. Nancy Schroering and Marci a Zehr provided copies of the Guess Who? game. Mary Ann Hayes at the Dubois County Historical Society helped me get permission to donate some of my interviews to the society and was very enthusiastic about the project. Sander van der Harst provided advice on vowel measurements and vowel normalization and provided some Praat scripts. Bartek Plichta provided advice on how to record properly for my perception stimuli, and Christian Koops actually created the continua. Lucien Carroll wrote the script to allow me to create my vowel charts (and Dan Michel modified the script as needed). Gabe Doyle wrote the macros to normalize my vowels, and he also provided much help with statistics. He really should have let me pay him. Emily Morgan allowed me to ambush her with statistics questions. Finally, I am grateful to the 90 people who agreed to participate in my experiment.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ .......... 11 LIST OF FIGURES ................................ ................................ ................................ ........ 15 LIST OF OBJECTS ................................ ................................ ................................ ....... 22 LIST OF ABBREVIATIONS ................................ ................................ ........................... 23 ABSTRACT ................................ ................................ ................................ ................... 24 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 26 ................................ 26 Gender and Sex in Sociolinguistic Research ................................ .......................... 27 Research Questions ................................ ................................ ............................... 28 Study Overview and Hypotheses ................................ ................................ ............ 28 Organization of the Dissertation ................................ ................................ .............. 30 2 LITERATURE REVIEW AND BACKGROUND INFORMATION ............................. 31 Comparisons of Perception and Production in Language Change ......................... 31 Language Change across the Lifespan ................................ ................................ .. 32 Acquisition of a Second Dialect ................................ ................................ .............. 39 Acquisition of a Second Dialect in Childhood or Youth ................................ ..... 39 Acquisition of a Second Dialect in Adulthood ................................ ................... 43 Gender and Sex in Sociolinguistic Research ................................ .......................... 56 American Regional Dialects ................................ ................................ .................... 59 The North ................................ ................................ ................................ ......... 59 The South ................................ ................................ ................................ ......... 61 The Midland ................................ ................................ ................................ ...... 63 3 METHODOLOGY ................................ ................................ ................................ ... 68 Vowels of Interest in the Study ................................ ................................ ............... 6 8 Predictor Variables Used in the Study ................................ ................................ .... 68 .................... 69 Jasper, Indiana ................................ ................................ ................................ 69 Location of and demographic information for Jasper ................................ 69 The debatable dialect of Southern Indiana ................................ ................ 70 Chicago, Illinois ................................ ................................ ................................ 71
7 Demographic information ................................ ................................ ........... 71 The Northern Cities Shift in Chicago ................................ .......................... 71 Participants ................................ ................................ ................................ ............. 72 Criteria for Inclusion ................................ ................................ ......................... 72 Requirements for Lifelong Jasper participants: ................................ .......... 72 Require ments for Mobile group participants: ................................ .............. 72 Requirements for Lifelong Chicago participants: ................................ ........ 72 Recruitment of Participants ................................ ................................ .............. 74 Total Participants and Exclusions ................................ ................................ ..... 75 Location within Chicago of Mobile and Lifelong Chicago Participants .............. 75 ................................ ................................ .. 77 Experimental Design and Procedure ................................ ................................ ...... 79 Equipment ................................ ................................ ................................ ........ 79 Materials ................................ ................................ ................................ ........... 80 The Extended Personality Attributes Questio nnaire ................................ ... 80 Recorded instructions ................................ ................................ ................ 84 Reading passage ................................ ................................ ....................... 85 Perception stimuli ................................ ................................ ....................... 86 Discussion topics ................................ ................................ ....................... 87 Procedure ................................ ................................ ................................ ......... 88 4 PER CEPTION ................................ ................................ ................................ ...... 102 Reaction Times ................................ ................................ ................................ ..... 102 Effects of Residence Group and Related Factors on Reaction Time ............. 103 Effects of Sex and Gender on Reaction Time ................................ ................ 105 Differences in Perception ................................ ................................ ...................... 108 Method of Statistical Analysis ................................ ................................ ......... 108 Measurement of formant values ................................ ............................... 109 Method of determining 50% crossover point using difference between F2 and F1 as a predictor ................................ ................................ ....... 110 Method of determining 50% crossover points using F1 and F2 as predictors ................................ ................................ .............................. 110 Statistical Com parisons of 50% Crossover Points ................................ .......... 114 Results of the Job~Jab continuum ................................ ........................... 115 Results of the Bad~Be d continuum ................................ .......................... 119 Results of the Dead~Dud continuum ................................ ....................... 121 Perception of individual Mobile participants ................................ ............. 122 Conclusions a bout the Malleability of Phonetic Perception in Adulthood .............. 126 Implications for Production Investigation ................................ ............................... 127 Discussion of Gender vs. Sex in Phonetic Perception ................................ .......... 127 5 PRODUCTION ................................ ................................ ................................ ...... 138 Interview Transcription ................................ ................................ .......................... 138 Method of Analysis ................................ ................................ ............................... 138 Extraction of Target Vowels from th e Interviews ................................ ............ 138
8 Segmentation of Vowels ................................ ................................ ................. 143 Praat settings ................................ ................................ ........................... 143 Beginning of vowels ................................ ................................ ................. 144 End of vowels ................................ ................................ ........................... 145 Measurement of Formant Values from Segmented Vowels ........................... 146 Determining the number of coefficients for LPC analysis ......................... 147 Extraction and checking of values ................................ ............................ 149 Normalization of Vowel Measurements ................................ .......................... 151 Statistical Analysis of Vowels ................................ ................................ ......... 153 Results ................................ ................................ ................................ .................. 157 Variation in Production of // ................................ ................................ ......... 158 Analysis of F1 ................................ ................................ .......................... 158 Analysis of F2 ................................ ................................ .......................... 162 Analysis of F3 ................................ ................................ .......................... 167 ................................ ................................ ........... 168 Analysis of F1 ................................ ................................ .......................... 168 Analysis of F2 ................................ ................................ .......................... 171 Analysis of F3 ................................ ................................ .......................... 174 Variati ................................ ................................ ........... 176 Analysis of F1 ................................ ................................ .......................... 176 Analysis of F2 ................................ ................................ .......................... 179 Analysis of F3 ................................ ................................ .......................... 182 Conclusions about the Malleability of Phonetic Production in Adulthood .............. 184 Relative Importance of Gender and Sex in Phonetic Production .......................... 186 6 COMPARISON OF PERCEPTION AND PRODUCTION ................................ ...... 218 Comparison of Perception and Production of NCS / / Raising ............................ 218 Statistical Comparison ................................ ................................ .................... 218 ................................ .............................. 219 Comparison of Perception and Production of NCS / / Backing ............................ 226 Statistical Comparison ................................ ................................ .................... 226 ................................ .............................. 227 ................................ ........ 230 7 CONCLUSIONS ................................ ................................ ................................ ... 252 The Research Questions Revisited ................................ ................................ ...... 252 How and to What Extent Does Language Change with Exposure to a New Dialect in Adulthood? ................................ ................................ .................. 252 When an Adult Moves to a New Region and Acquires a New Dialect, does Production or Perception Change First? Which Changes More? Are these Relationships Consistent across Individuals? ................................ .... 253 Which is a Better Predictor of Adult Language Change in the Individual, Age of First Exposure to the New Dialect or Length of Exposure? ..................... 254 Is Gender or Binary Biological Sex a Better Predictor of Language Change in Adulthood? ................................ ................................ .............................. 255
9 Limitations of the Study ................................ ................................ ......................... 255 Suggestions for Further Research ................................ ................................ ........ 257 Broader Implications ................................ ................................ ............................. 258 Closing ................................ ................................ ................................ .................. 260 APPENDIX A PARTICIPANTS ................................ ................................ ................................ .... 261 B TEXT OF RECORDED INSTRUCTIONS ................................ ............................. 273 Instructions for the EPAQ ................................ ................................ ..................... 273 Instructions for the Reading Task ................................ ................................ ......... 274 Instructions for the Perception Experiment ................................ ........................... 27 4 C TEXT OF REA DING PASSAGE ................................ ................................ ........... 276 D INTERVIEW/CONVERSATION TOPICS AND DEMOGRAPHIC QUESTIONS ... 277 Discussion Topics for All Groups ................................ ................................ .......... 277 How did you Meet your Husband/Wife? ................................ ......................... 277 Danger of Death ................................ ................................ ............................. 277 September 11 ................................ ................................ ................................ 277 Memories of High School, Middle School ................................ ....................... 277 Dreams ................................ ................................ ................................ ........... 278 Work ................................ ................................ ................................ ............... 278 Economy ................................ ................................ ................................ ........ 278 Discussion Topics for Lifelong Jasper Participants ................................ ............... 278 Thoughts about Jasper ................................ ................................ ................... 278 Construction in Jasper ................................ ................................ .................... 279 Sports ................................ ................................ ................................ ............. 279 Closings ................................ ................................ ................................ .......... 279 Discussion Topics for Mobile and Lifelong Chicago Participants .......................... 279 Questions Asked of All Lifelong Jasper Participants ................................ ............. 280 Questions Asked of All Mobile Participants ................................ ........................... 280 Questions Asked of All Lifelong Chicago Participants ................................ .......... 281 E WHICH PARTICIPANTS PARTICIPATED TOGETHE R OR ALONE ................... 282 F / BACKING COMPARED TO THEIR SOCIAL CHARACTERISTICS ................................ ....... 284 G PRODUCTION RESULTS MIXED EFFECTS REGRESSIONS ........................ 288 H FURTHER VOWEL CHARTS ................................ ................................ ............... 289 I CHARTS COMPARING PERCEPTION AND PRODUCTION FOR MOBILE AND LIFELONG JASPER PARTICIPANTS ................................ .......................... 322
10 LIST OF REFERENCES ................................ ................................ ............................. 323 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 329
11 LIST OF TABLES Table page 3 1 Community areas/neigh borhoods and suburbs in which Mobile participants lived at the time of the interview ................................ ................................ ......... 97 3 2 Community areas/neighborhoods and suburbs in which Mobile participants had lived most (the plurality) of their time in Chicago, if different from where they lived at the time of the interview ................................ ................................ .. 98 3 3 Community areas/neighborhoods and suburbs in which Lifelong Chicago participants lived at the time of the interview ................................ ...................... 99 3 4 Community areas/neighborhoods and suburbs in which Lifelong Chicago participants had lived most (the plurality) of their lives, if different from where they lived at the time of the interview ................................ ................................ 100 4 1 Perception stimuli not receiving a response, tallied by continuum and residence group ................................ ................................ ................................ 129 4 2 F1 and F2 values of perception stimuli, measured in Barks ............................. 130 4 3 Mean perceptual crossover points (in Barks) for Job~Jab continuum by residence group and perception measure ................................ ........................ 131 4 4 Mean perceptual crossover points (in Barks) for Bad~Bed continuum by residence group and perception measure ................................ ........................ 131 4 5 Mean perceptual crossover points (in Barks) for Dead~Dud continuum by residence group and perception measure ................................ ........................ 132 4 6 Results of multiple regression for Job~Jab continuum, F1 crossover points .... 132 4 7 Results of multiple regression for Job~Jab continuum, F2 crossover points .... 132 4 8 Results of multiple regression for Job~Jab continuum, F2F1d crossover points ................................ ................................ ................................ ................ 133 4 9 Results of multiple regression for Job~Jab continuum including interaction of sex and residence group, F1 crossover points ................................ ................. 133 4 10 Results of multiple regression for Job~Jab continuum including interaction of sex and residence group, F2 crossover points ................................ ................. 133 4 11 Results of multiple regression for Job~Jab continuum including interaction of sex and residence group, F2F1d crossover points ................................ ........... 134 4 12 Results of multiple regression for Bad~Bed continuum, F1 crossover points ... 134
12 4 13 Results of multiple regression for Bad~Bed continuum, F2 crossover points ... 134 4 14 Results of multiple regression for Bad~Bed continuum, F2F1d crossover points ................................ ................................ ................................ ................ 135 4 15 Results of multiple regression for Bad~Bed continuum including interaction of residence group and YOB, F1 crossover points ................................ ............... 135 4 16 Results of multiple regression for Bad~Bed continuum including interaction of residence group and YOB, F2 crossover points ................................ ............... 135 4 17 Results of multiple regression for Bad~Bed continuum including interaction between residence group and YOB, F2F1d crossover points .......................... 136 4 18 Results of multiple regression for Dead~Dud continuum, F1 crossover points 136 4 19 Results of multiple regression for Dead~Dud continuum, F2 crossover points 136 4 20 Results of multiple regression for Dead~Dud continuum, F2F1d crossover points ................................ ................................ ................................ ................ 137 5 1 Number of tokens, by participant and vowel, extracted from the first 15 minutes of interviews or conversations and total number of tokens extracted, when fewer than 15 ................................ ................................ .......................... 189 5 2 Number of participants by sex found to be best analyzed at each number of formants for LPC analysis ................................ ................................ ................ 194 5 3 Social factors included in mixed effects regression analysis, by normalization method and vowel category ................................ ................................ .............. 195 6 1 / raising ........................... 237 6 2 Relationships between perception and production in language change in adulthood for Mobile participants ................................ ................................ ...... 237 6 3 / raising ................................ ................................ ................................ ............... 238 6 4 Counts of agreement between production and perception of / / for Lifelong Jasper participants ................................ ................................ ........................... 239 6 5 production of / / separated by sex ................................ ................................ ... 239 6 6 production of / / separated by masculinity scores ................................ ............ 239
13 6 7 production of / / separated by femininity scores ................................ .............. 240 6 8 production of / / separated by year of birth ................................ ...................... 240 6 9 perception and production of // separated by sex ................................ .......... 240 6 10 / raising, based on assumptions about their original relationship. ................................ ................... 241 6 11 Relationships between perception and production in language change in adulthood for / / raising for Mobile participant s, revised ................................ .. 242 6 12 / backing .......................... 242 6 13 Relationships between perception and production in language change of / / in adulthood for Mobile participants ................................ ................................ .. 243 6 14 / .. 244 6 15 Count of agreement between production and perception of / / for Lifelong Jasper participants ................................ ................................ ........................... 245 6 16 production of / / separated by sex ................................ ................................ .... 245 6 17 ................................ ............ 245 6 18 ................................ ............... 246 6 19 production of / / separated by year of birth ................................ ....................... 246 6 20 /, assuming production was originally more Chicago like than perception ................................ ............. 247 6 21 Relationships between perception and production in language change in adulthood for / / backing for Mobile participants, revised ................................ 248 6 22 Mobile participants tallied by relationship of perception and production in both / / raising and / / backing ................................ ................................ ................ 248 6 23 Mobile participants tallied by relationship of perception and production in both // backing, revised ................................ ................................ .. 248
14 6 24 acquisition of a D2, separated by sex ................................ ............................... 249 6 25 acquisition of a D2, separated by total masculinity ................................ ........... 249 6 26 acquisition of a D2, separated by positive masculinity ................................ ...... 249 6 27 acquisition of a D2, separated by negative masculinity ................................ .... 250 6 28 acquisition of a D2, separated by total femininity ................................ .............. 250 6 29 acquisition of a D2, separated by positive femininity ................................ ........ 250 6 30 acquisition of a D2, separated by negative femininity ................................ ....... 250 6 31 acquisition of a D2, separated by years lived in Chicago ................................ 251 6 32 acquisition of a D2, separated by education ................................ ..................... 251 A 1 Lifelong Jasper participants ................................ ................................ .............. 262 A 2 Lifelong Chicago participants ................................ ................................ ............ 265 A 3 Mobile participants ................................ ................................ ............................ 269 E 1 Participants who completed the experiment alone and in pairs, and the relationship of those who participated in pairs ................................ .................. 282 F 1 Perceptual boundaries on the Dead~Dud continuum for individual Mobile participants, with means for the Lifelong groups ................................ ............... 285 F 2 Mobile group members separated by perception style, with the Chicago neighborhoods in which they had lived and worked ................................ ......... 286 F 3 Mobile group members separated by perception style, with other social characteristics ................................ ................................ ................................ ... 287
15 LIST OF FIGURES Figure page 2 1 The Northern Cities Shift ................................ ................................ .................... 67 2 2 The Southern Shift ................................ ................................ .............................. 67 3 1 Maps showing Jasper, Indiana, within Dubois County, and Dubois County within the state of Indiana. ................................ ................................ ................. 96 3 2 Presentation order for perception test ................................ .............................. 101 4 1 Hypothetical example of responses to Job~Jab continuum showing a 50% crossover point between stimuli 5 and 6 ................................ ........................... 129 4 2 Logistic regression curve with two predictors. F1 is graphed on the x axis, F2 on the z axis, and probability on the y axis. ................................ ...................... 130 4 3 Intersection of the combinations of F1 and F2 (in Barks) that yield vowels that a listener would hypothetically hear as 50% Dead and 50% Dud (solid curve) and set of formant values for the stimuli actually heard (dotted line) ................ 131 5 1 An initial vowel boundary not placed at a zero cro ssing on the waveform ........ 190 5 2 Initial vowel boundary placed at zero crossing after end of stop burst because formants appeared during burst in the word dad ................................ 190 5 3 Initial vowel boundary placed at zero crossing near onset of full formants after an aspirated stop in the word past ................................ ............................ 191 5 4 Initial vowel boundary marked at the zero crossing between the end of the fricative and the beginning of the full formant structure in the word fast ........... 191 5 5 Initial vowel boundary marked at the zero crossing after the pulse with upward sloping formants in the word next ................................ ........................ 192 5 6 Final vowel boundary marked at zero crossing near end of full formant structure in the word back ................................ ................................ ................ 192 5 7 Final vowel boundary marked at the end of the formant structure and the beginning of the fricative, well after the end of voicing, in the word gas .......... 193 5 8 Final vowel boundary marked at the zero crossing near the disappea rance of formants in the word love ................................ ................................ ................. 193 5 9 Final vowel boundary marked at point of noticeable change in vowel quality due to lack of audible/visible fricative in the word other ................................ .... 194
16 5 10 Improperly tracked formants in the word back. ................................ ................ 195 5 11 Vowel space for the Lifelong Jasper group at T1. ................................ ............. 196 5 12 Vowel space for the Lifelong Jasper group at T2 ................................ .............. 197 5 13 Vowel spa ce for the Lifelong Jasper group at T3 ................................ .............. 197 5 14 Vowel space for the Lifelong Jasper group at T4 ................................ .............. 198 5 15 Vowel space for the Lifelong Jasper group at T5 ................................ .............. 198 5 16 Vowel space for the Mobile group at T1 ................................ ........................... 199 5 17 Vowel space for the Mobile group at T2 ................................ ........................... 199 5 18 Vowel space for the Mobile group at T3 ................................ ........................... 200 5 19 Vowel space for the Mobile group at T4 ................................ ........................... 200 5 20 Vowel space for the Mobile group at T5 ................................ ........................... 201 5 21 Vowel space for the Lifelong Chicago group at T1 ................................ ........... 201 5 22 Vowel space for the Lifelong Chicago group at T2 ................................ ........... 202 5 23 Vowel space for the Lifelong Chicago group at T3 ................................ ........... 202 5 24 Vowel space for the Lifelong Chicago group at T4 ................................ ........... 203 5 25 Vowel space for the Lifelong Chicago group at T5 ................................ ........... 203 5 26 Scatter plot of Lobanov normalized F1 val ues for // for the Lifelong Jasper participants. The lines connect the mean values at each time point ................ 204 5 27 Scatter p lot of Lobanov normalized F1 values for // for Mobile participants at each time point ................................ ................................ ............................. 204 5 28 Scatter plot of Lobanov normalized F1 values of // for Lifelong Chicago participants at each time point ................................ ................................ .......... 205 5 29 Scatter plot of Lobanov normalized F2 values for // for the Lifelong Jasper group at each time point ................................ ................................ ................... 205 5 30 Scatter plot of L obanov normalized F2 values of // for the Mobile group at each time point ................................ ................................ ................................ 206 5 31 Scatter plot of Lobanov normalized F2 values of // for the Lifelong Chicago group at each time point ................................ ................................ ................... 206
17 5 32 Scatter plot showing the Lobanov normalized F3 values for // for the Lifelong Jasper participants at each time point ................................ ................. 207 5 33 Scatter plot showing the Lo banov normalized F3 values for // for the Mobile participants at each time point ................................ ................................ .......... 207 5 34 Scatter plot showing the Lobano v normalized F3 values for // for the Lifelong Chicago participants at each time point ................................ .............. 208 5 35 Scatter plot showing the Lobanov Jasper participants at each time point ................................ .............................. 208 5 36 Scatter plot showing the Lobanov participants at each time point ................................ ................................ .......... 209 5 37 Scatter plot showing the Lobanov Chicago participants at each time point ................................ ............................ 209 5 38 Scatter plot showing Lobanov Jasper group at each time point ................................ ................................ ....... 210 5 39 Scatter plot showing Lobanov participants at each time point ................................ ................................ .......... 210 5 40 Scatter plot showing Lobanov Chicago participants at each time point ................................ ............................ 211 5 41 Scatter plot showing Lobanov Jasper group at each time point ................................ ................................ ....... 211 5 42 Scatter plot showing Lobanov group at each time point ................................ ................................ ................... 212 5 43 Scatter plot showing Lobanov Chicago group at each time point ................................ ................................ ..... 212 5 44 Scatter plot showing the Lobanov Jasper group at each time point ................................ ................................ ....... 213 5 45 Scatter plot showing the Lobanov group at each time point ................................ ................................ ................... 213 5 46 Scatter plot showing Lobanov Chicago group at each time point ................................ ................................ ..... 214 5 47 Scatter plot showing the Lobanov Jasper group at each time point ................................ ................................ ....... 214
18 5 48 Scatter plot showing the Lobanov group at each time point ................................ ................................ ................... 215 5 49 Scatter plot showing the Lobanov Chicago group at each time point ................................ ................................ ..... 215 5 50 Scatter plot showing the Lobanov Jasper group at each time point ................................ ................................ ....... 216 5 51 Scatter plot showing the Lobanov group at each time point ................................ ................................ ................... 216 5 52 Scatter plot showing the Lobanov Chicago group at each time point. ................................ ................................ .. 217 6 1 Hypothetical example of comparison of production and perception showing either perception and production equal ly shifted or no shift .............................. 235 6 2 Hypothetical example of comparison of perception and production showing perception shifting ahead of production ................................ ............................ 235 6 3 Hypothetical example of comparison of perception and production showing production shifting ahead of perception ................................ ............................ 236 H 1 Vowel space for Lifelong Jasper participants at T1 average value shown ................................ ................................ ........................ 290 H 2 Vowel space for Lifelong Jasper participants at T2 averages only ................ 291 H 3 Vowel space for Lifelong Jasper participants at T3 averages only ................ 291 H 4 Vowel space for Lifelong Jasper participants a t T4 averages only ................ 292 H 5 Vowel space for Lifelong Jasper participants at T5 averages only ................ 292 H 6 Vowel space for Mobile participants at T1 averages only .............................. 293 H 7 Vowel space for the Mobile group at T2 averages only ................................ 293 H 8 Vowel space for Mobile participants at T3 averages only .............................. 294 H 9 Vowel space for Mobile participants at T4 averages only .............................. 294 H 10 Vowel space for Mobile participants at T5 averages only .............................. 295 H 11 Vowel space for Lifelong Chicago participants at T1 averages only ............ 295 H 12 Vowel space for Lifelong Chicago participants at T2 averages only .............. 296
19 H 13 Vowel space for Lifelong Chicago part icipants at T3 averages only .............. 297 H 14 Vowel space for the Lifelong Chicago participants at T4 averages only ........ 298 H 15 Vowel space for the Lifelong Chicago participants at T5 averages only ........ 298 H 16 Vowel space for the nine oldest Lifelong Jasper participants (YOB 1949 1972) at T1 (A = //, E = / /, U = / /) ................................ ................................ 299 H 17 Vowel space for the nine oldest Lifelong Jasper participants at T2 .................. 299 H 18 Vowel space for the nine oldest Lifelong Jasper participants at T3 .................. 300 H 19 Vowel space for the nine oldest Lifelong Jasper participants at T4 .................. 3 00 H 20 Vowel space for the nin e oldest Lifelong Jasper participants at T5 .................. 301 H 21 Vowel space for the six Lifelong Jasper participants in the middle age range (YOB 1974 1980) at T1 ................................ ................................ .................... 301 H 22 Vowel space for the six Lifelong Jasper participants in the middle age range at T2 ................................ ................................ ................................ ................. 302 H 23 Vowel space for the six Lifelong Jasper participants in the middle age range at T3 ................................ ................................ ................................ ................. 302 H 24 Vowel space for the six Lifelong Jasper participants in the middle age range at T4 ................................ ................................ ................................ ................. 303 H 25 Vowel space for the six Lifelong Jasper participants in the middle age range at T5 ................................ ................................ ................................ ................. 303 H 26 Vowel space for the eleven yo ungest Lifelong Jasper participants (YOB 1981 1989) at T1 ................................ ................................ .............................. 304 H 27 Vowel space for the eleven youngest Lifelong Jasper par ticipants at T2 ......... 304 H 28 Vowel space for the eleven youngest Lifelong Jasper participants at T3 ......... 305 H 29 Vowel space for the eleven youngest Lifelong Jasper participants at T4 ......... 305 H 30 Vowel space for the eleven youngest Lifelong Jasper participants at T5 ......... 306 H 31 Vowel space for the nine Mobile participants who have spent the least amount of time in Chicago (1 year 3.5 years) at T1 ................................ ....... 306 H 3 2 Vowel space for the nine Mobile participants who have spent the least amount of time in Chicago at T2 ................................ ................................ ....... 307
20 H 33 Vowel space for the nine Mobile participants who have spent the least amount of time in Chicago at T3 ................................ ................................ ....... 307 H 34 Vowel space for the nine Mobile participants who have spent the least amount of time in Chicago at T4 ................................ ................................ ....... 308 H 35 Vowel space for the nine Mobile participants who have spent the least amount of time in Chicago at T5 ................................ ................................ ....... 308 H 36 Vowel space for the nine Mobile participants who have spent the middle range of time (3.75 years 7.5 years) in Chicago at T1 ................................ ... 309 H 37 Vowel space for the nine Mobile participants who have spent the middle range of time in Chicago at T2 ................................ ................................ .......... 309 H 38 Vowel space for the nine Mobile participants who have spent the middle range of time in Chicago at T3 ................................ ................................ .......... 310 H 39 Vowel space for the nine Mobile participants who have spent the middle amount of time in Chicago at T4 ................................ ................................ ....... 310 H 40 Vowel space for the nine Mobile participants who have spent the middle amount of time in Chicago at T5 ................................ ................................ ....... 311 H 41 Vowel space for the eight Mobile participants who have spent the most time (13.5 years 33.5 years) in Chicago at T1 ................................ ....................... 311 H 42 Vowel space for the eight Mobile partic ipants who have spent the most time in Chicago at T2 ................................ ................................ ............................... 312 H 43 Vowel space for the eight Mobile participants who have spen t the most time in Chicago at T3 ................................ ................................ ............................... 312 H 44 Vowel space for the eight Mobile participants who have spent the most time in C hicago at T4 ................................ ................................ ............................... 313 H 45 Vowel space for the eight participants who have spent the most time in Chicago at T5 ................................ ................................ ................................ ... 313 H 46 Vowel space for the eight oldest Lifelong Chicago parti cipants (YOB 1953 1965) at T1. ................................ ................................ ................................ ..... 314 H 47 Vowel space for the eight oldest Lifelong Chicago participants at T2 ............... 314 H 48 Vowel space for the eight oldest Lifelong Chicago participants at T3 ............... 315 H 49 V owel space for the eight oldest Lifelong Chicago participants at T4 ............... 315 H 50 Vowel space for the eight oldest Lifelong Chi cago participants at T5 ............... 316
21 H 51 Vowel space for the eight Lifelong Chicago participants in the middle age range (YOB 1966 1979) at T1. ................................ ................................ ......... 316 H 52 Vowel space for the eight Lifelong Chicago participants in the middle age group at T2 ................................ ................................ ................................ ....... 317 H 53 Vowel space for the eight Lifelong Chicago participants in the middle age group at T3 ................................ ................................ ................................ ....... 317 H 54 Vowel space for the eight Lifelong Chicago participants in the middle age group at T4 ................................ ................................ ................................ ....... 318 H 55 Vowel space for the eight Lifelong Chicago participants in the middle age group at T5 ................................ ................................ ................................ ....... 318 H 56 Vowel space for the eight youngest Lifelong Chicago participants (YOB 1980 1987) at T1. / / and / / are noticeably backed ................................ ......... 319 H 57 Vowel space for the eight youngest Lifelong Chicago participants at T2 .......... 319 H 58 Vowel space for the eight youngest Lifelong Chicago parti cipants at T3 .......... 320 H 59 Vowel space for the eight youngest Lifelong Chicago participants at T4 .......... 320
22 LIST OF OBJECTS Object page 3 1 Continuum of Job~Jab stimuli (.wav file 396 KB) ................................ ............... 87 3 2 Continuum of Bad~Bed stimuli (.wav file 336 KB) ................................ .............. 87 3 3 Continuum of Dud~Dead stimuli (.wav file 344 KB) ................................ ............ 87 G 1 Mixed Effects Regression Results ................................ ................................ .... 288 I 1 Charts comparing perception and pro duction for Mobile and Lifelong Jasper participants. ................................ ................................ ................................ ...... 322
23 LIST OF ABBREVIATION S AAC Age of arrival in Chicago BSRI Bem Sex Role Inventory D2 Second dialect EPAQ Extended Personality Attributes Questionnaire F1, F2, F3 First format, second form ant, third formant F+, F Ftotal Positive femininity, negative femininity, total femininity L1, L2 First language, second language M+, M Mtotal Positive masculinity, negative masculinity, total masculinity NCS Northern Cities Shift PLC Percentage of lif e lived in Chicago SSB Southern Standard British T1, T2, T3, T4, T5 First through fifth measurement time points (for production data) UMC Upper middle class YLC Number of years lived in Chicago YOB Year of birth
24 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 RELATIONSHIP BETWEEN PERCEPTION AND PRODUCTION IN ADULT By Zoe Lynn Ziliak December 2012 Chair: Hlne Blondeau Major: Linguistics perception and production, specifically upon exposure to a new dialect in adulthood. It further addresses t he relative importance of binary biological sex and socially Perception and production data was collected from lifelong residents of southern Indiana, lifelong residents of Chicago and individuals raised in southern Indiana who moved to Chicago in adulthood (the Mobile group). The Mobile participants were shifting from their native Lower Midland or Upper South dialect and adopting the Northern Cities Shift (NCS) characteristic of Chicago. A gender identity test, the Extended Personality Attributes Questionnaire, was also administered to each participant. Results indicate that both perception and production are malleable in adulthood, but production may be more subject to change th an perception. However, this relationship is not constant across individuals: some Mobile group members changed production more than perception, but others changed perception more, and still others had shifted the two equally or not at all. Analyses sugg est that socially constructed
25 femininity level may be more important than her ma sculinity in determining behavior. This study has implications for the research areas of language change across the lifespan, gender and language, second dialect acquisition, second language acquisition, and American dialectology.
26 CHAPTER 1 INTRODUCTION perception and production. The research has implications for the fields of sociophonetics, general sociolinguistics, and second language acquisition. The study illuminates some p rocesses of second dialect acquisition and, by extension, language acquisition. It also adds to the literature on language change across the lifespan, and the results of the methodological question, regarding the importance of including socially construct ed gender into sociolinguistic studies, will affect the design of future studies in the field. automatically ac quire new language characteristics disappears, or is greatly reduced, by seems to disappear after this age [puberty], and foreign languages have to be taught and learned t also expect that after puberty, a speaker would need to exert effort to acquire new changes in her own native language, rather than just picking the changes up through exposure. More r language system can change across the lifespan (e.g., Sankoff and Blondeau 2007; Harrington, Palethorpe, and Watson 2005), well past the purported end of the critical period, and probably without much conscious effort on the part of the speaker. One possible trigger for language change in adulthood is a move to a new dialect region.
27 When someone moves to a new dialect region in adulthood, he may adjust his language system to reflect that of the speakers he is now exposed to in his new home. This study examines evidence of phonetic change in individuals (the Mobile group) who grew up in (perhaps) the Upper South dialect region of the United States and moved in adulthood to a city that exhi bits the Northern Cities Shift. Specifically, I look for and/or production. While studies comparing perception and production have often found that perception changes first (e.g., Janson and Schulman 1983), at least one study (Janson 1983) has shown that it is production that leads, at least in the case of community change. One major goal of this study is to discover why one group of individuals might change their perceptio n before their production, while others would undergo the process in reverse. Further, I examine the relative importance of age of first exposure and duration of exposure to a new dialect for predicting how completely an adult will adopt a new dialect. continue to change across the lifespan, it may nonetheless be the case that younger is address es the question of whose language will have changed more, a 24 year old first exposed to the new dialect at age 21, or a 52 year old first exposed at age 42 (that is, early but brief exposure vs. late but long). Gender and Sex in Sociolinguistic Research A secondary goal is to demonstrate the importance of integrating modern definitions of gender into sociolinguistic research. While researchers studying language
28 change often mention the idea of gender as a social construct, in practice, many (though not all) still base their conclusions on binary biological sex, with the term gender whitewashed over it. In this study, I differentiate the effects of binary biological sex from those of socially constructed gender in an attempt to encourage other researcher s to incorporate both measures into their methodology. Research Questions The research questions addressed in this study are as follows: How and to what extent does language change with exposure to a new dialect in adulthood? What is the relationship betwe en perception and production when an adult moves to a new region and acquires a new dialect? o Does production or perception change first? o Is production or perception more malleable? (Which changes to a greater extent?) o Is this relationship consistent across individuals? Which is a better predictor of adult language change in the individual, age of first exposure to the new dialect or length of exposure to the new dialect? Is gender or binary biological sex a better predictor of language change in adulthood? (Or are they equally good?) Study Overview and Hypotheses In the present study, I collected data from three groups: the Lifelong Jasper group, the Lifelong Chicago group, and the Mobile group. The Lifelong Jasper participants had lived all (or nearly all ) their lives in Jasper, Indiana, a small town in the southwestern portion of the state, which may be in the Lower North or Upper South dialect region. (See Chapter 2 for more details.) The Lifelong Chicago participants had likewise lived all (or nearly all) of their lives in Chicago, Illinois, or one of its close suburbs. Chicago is
29 one of the cities that define the dialect region with the Northern Cities Shift. The Mobile group was the experimental group in this study; Mobile participants had been rai sed in upon exposure to the new dialect in adulthood. I expected to find that at leas t some of the Mobile participants would indeed have adopted the Northern Cities Shift in perception and/or production, and the data bore out this expectation. I did not make predictions regarding whether production or perception would either change first or be more malleable in the individual, or whether all individuals would have the same pattern. Later chapters will show that in fact, the pattern was not consistent across individuals, but overall, production seems to be more malleable than perception in adulthood. I suspected that length of exposure to the new dialect might actually trump age of first exposure in determining whether a Mobile individual adopted the new dialect, and Chapter 5 presents convincing evidence in support of this hypothesis. I also expected that socially constructed gender would be a better predictor of language behavior than binary biological sex. Chapter 4 shows that this expectation was not borne out for phonetic perception, but Chapter 5 shows that both gender and sex appe ar to influence adult malleability of phonetic production. I generally expected more masculine participants, regardless of their sex, to generally behave more like previous sociolinguistic generalizations have indicated males would behave, while more fem inine participants would generally exhibit behavior like that associated with females. (That is, I hypothesized that, because gender is
30 correlated with sex, previous findings of the relationship between sex and language were actually muted reflections of the relationship between gender and language.) However, because it is possible for an individual to be both highly masculine and highly feminine, I knew that the relationship must be more complex than I have just indicated. I did not make predictions rega rding this complexity, preferring to view this experiment as exploratory research on the effects of gender on language. Later chapters will show that femininity actually appears to influence adult second dialect acquisition more than masculinity. Organiza tion of the Dissertation Chapter 2 presents a review of the relevant literature, and the data collection methodology is presented in Chapter 3. In Chapters 4 and 5, I describe Mobile production, respectively, and in Chapter 6, I compare the acquisition of these two aspects of phonetic competence. Finally, in Chapter 7 I draw conclusions, discuss the significance of the study, and give suggestions for further research.
31 CHAPTER 2 LITERATURE REVIEW AND BACKGROUND INFORMATION This chapter reviews the literature relevant to the current study. The chapter begins with a presentation of the two studies of language change that initially inspired the current research. I then move to a discussion of various aspects of language change across the lifespan, followed by a review of the literature on gender, sex, an d language variation. I close the chapter with an overview of the regional American dialects relevant to this study. Comparisons of Perception and Production in Language Change The direct inspiration for this research came from a pair of studies focusing on the relationship between perception and production in generational language change. Janson (1983) investigated the relationship between perception and production in the backing of /a:/ in Sweden. He created an artificial continuum of tokens from /a:/ to /o:/ dramatically as their production. Janson concluded, for an individual in a situation of change, perception seems to lag behind production emphasis mine ) Janson argued that in cases of generational change, it makes sense for perception to change more slowly than production because, while the younger generation may produce words differently than the older generation, they still need to be able to understand the older speakers (31). Much as teenagers might like to, they cannot simply lose the ability to understand their elders. 1 1 ve changed.
32 Yet Janson himself fou nd evidence that may contradict his conclusion that perception lags behind production. As of the early 1980s, Stockholm speakers merged short /e/ and / / in production, while speakers of the Lycksele Swedish dialect kept them distinct. Using both synthesi zed and natural data, Janson and Schulman (1983) tested the ability of listeners from both regions to perceive the distinction. Surprisingly, while Lycksele participants did consistently produce the distinction, they were unable to perceive it. This is a case of perception preceding production in individuals participating in language change. (That is, the Stockholm merger was spreading to Lycksele in perception before it spread in production.) Thus, the relationship between perception and production in sound change must be more complex than Janson (1983) asserted. The contradictory results of these two studies inspired me to look further into the phonological system changes. What would cause perception to lead in one case and production to lead in another? Does it depend on the individual, on the external language or community situation, or both? However, while Janson (1983) and Janson and Schulman (1983) focused on individ uals whose (possibly stable) phonetic systems reflected an ongoing language change in their communities, I chose to focus on individuals undergoing change across the lifespan. Language Change across the Lifespan As mentioned in Chapter 1, the critical peri od hypothesis (Lenneberg 1967) indicates that once a speaker reaches puberty, her language system is mostly stable. However, a growing body of literature shows that, on the contrary, speakers can change their phonetic and phonological systems well into ad ulthood. For example, Sankoff and
33 adoption of an ongoing change in the language community, the switch from apical to dorsal pronunciations of /r/. Sankoff and Blondeau analyzed the speech of 32 Montreal ranged from 15 to 56 in 1971. 10 of the 32 speakers used apical [r] nearly categorically in both interviews, and another 10 used dorsal [R] nearly categorically in both. However, nine of the remaining 12 participants significantly increased their use of dorsal [R] from the first interview to the second; this group included three speakers who were at least 40 years old at the time of the firs t interview. Further, two of these nine speakers had been near categorical [r] users in 1971. Thus, this study provides evidence that adults can change their phonology to align with an ongoing change in the community. Not all studies have found such mall eability across the lifespan, however. MacKenzie and Sankoff (2010) analyzed the speech of a subsample of six of the speakers analyzed in Sankoff and Blondeau (2007). MacKenzie and Sankoff e incoming community change in eight long vowels. They found that the six speakers had an average of 3.7 of the eight vowels that were stable (showed no significant change in production) from 1971 to 1984. (The six speakers ranged from zero to seven stab le vowels.) On the other hand, only an average of 1.7 (ranging from zero to three) of the 8 vowels showed significant changes in the same direction as the community overall. The rest of the vowels changed significantly, but in different directions from t he community changes. For example, the Montreal community as a whole lowered and backed / :/ from 1971 to 1984, but speaker Louise C. significantly raised her / :/ across this same time period.
34 Additionally, although Claire R. significantly lowered thre e of her vowels to follow the community trend of lowering, she actually significantly lowered all eight of the analyzed vowels over the course of the study, indicating that the changes may be due to aging related physiological changes, not adopting the com munity trends. The directions from the community as a whole may also be due to physiological changes. In short, MacKenzie and Sankoff found that while their participants did shift some of their vowels in the direction of wider community changes, the plurality of their vowels remained stable across the 13 years of the study, and several others shifted in directions unrelated to community change. This study thus mostly demonst rated a lack of (nonphysiological) malleability in phonetic production across the lifespan. For the most part, participants did not adopt ongoing community changes as they aged. Note that Sankoff and Blondeau (2007) tested adoption of a change that speakers in the community are aware of; the participants who shifted from apical to dorsal /r/ over the course of their lives probably did so intentionally, consciously matching their o wn speech to the ongoing community change. In contrast, the vowel changes examined by MacKenzie and Sankoff (2007) are mostly (or completely) below the level of consciousness. If the participants had changed their vowel production to follow the community shift, they would presumably have done so unintentionally. Based on these studies, one might think that it is possible for individuals to adopt changes above but not below the level of consciousness in adulthood. generalization to be false.
35 community language changes (generally below the level of consciousness) has been carried out by Harrington and colleagues (Harrington, Palethorpe, and Watson 2000a, 2000b, and 2005; Harrington 2006). Every year since 1952, Queen Elizabeth II has delivered a radio address at Christmas; Harrington et al. have used the recordings of these annua They are interested both in ways that she has changed her speech to adopt ongoing changes in the community and in how speech changes over time due to aging. Harrington et al. (20 00a) found that for 10 of the 11 vowels they investigated, the southern British (SSB) norms of the 1980s, but they never reached the new SSB norms. Harrington et al. (2000 b) had similar findings, with the added note that some of the vowels over time, not only followed the movements of these vowels in SSB, but also contradicted findings fro is, some aging studies (e.g., Rastatter, McGuire, Kalinowski, and Stuart 1997) have found that women tend to have lower F1 values as they age. Thus, the changes in the on over the years seemed to be due to her adjusting to shifts in SSB, not to aging. / and found that these phonemes had similarly moved from the 1950s to the 1980s to follo w changes in SSB. However, while the location to her monophthongs had changed over time, the shape of the trajectory had not, at least for / /. Finally, from the 1950s to the 1980s, SSB has undergone tensing of both
36 th e KIT vowel and the unstressed vowel in the set happY. Both these vowels are closer to FLEECE than they once were. Harrington (2006) showed that production had changed accordingly for KIT and to a lesser extent for happY. These studies show that an adult can shift her phonetic production to follow ongoing changes in the community, though she may not reach the new production norms. The studies described above were at least partially designed to test the apparent time hypothesis (Labov 1963), which states that synchronic data from speakers of different age groups can be used to simulate diachronic data. For example, speech samples taken in 2000 from speakers born in 1940, 1960, and 1980 should reflect the language system of the community at ab out 1950, 1970, and 1990, respectively. However, the results of the studies reviewed here show that the apparent time hypothesis, while perhaps still a useful tool for sociolinguists, was originally stated too space have implications for studies that view diachronic change in terms of apparent, by underestimating the influences of c (74 speakers change in the direction of change in progress in their adult lives, apparent time underestimates (582, emphasis theirs). Bowie (2005) found mixed evidence of language change vs. stability over the lifespan. He used existing recordings of five Latter Day Saints clergy members made twenty years apart; each subject was recorded at least twice. Bowie m easured these
37 their adult lives. Three variables showed at least some change across the lifespan, however. For the feel fill merger, two of the five speakers distinguished the word classes more as they aged. (The other three had no change.) Bowie says that this could be taken as evidence either of lack of language stability in a dulthood or of age grading; perhaps speakers generally begin to distinguish the vowels more as they age. For pre nasal / / raising, three of the speakers again showed no change across the lifespan, but one speaker significantly increased this feature as h e aged, while the last significantly decreased it. Since the two unstable speakers changed in the opposite direction, this is not a case of age grading. Bowie suggests that earlier in life, each speaker switched between two variants (raised and unraised), and that later in life, each settled into a pattern of using only one variant or the other. Finally, for /u/ fronting, one speaker showed no change over his adult life, one decreased /u/ fronting, two increased it and one, who was recorded at three tim e points, first dramatically increased his /u/ fronting, and then returned to his original level! Bowie notes that when sociolinguistic panel studies measure individuals at exactly two times, they may be missing these reversals of direction. Wagner (2012 ) provided a good summary of studies of language change across language systems may be stable, but the community as a whole is undergoing language change) from age grading (in which the community may be stable, but an the community and individual are unstable, adopting an incoming linguistic change
38 usually abruptly 2 ). In her review o f real time studies of language change across the lifespan, she noted (p. 377) Firstly, panel studies suggest that in the majority of generational community changes in progress, some individuals make post adolescent adjustments to their speech in the direc majority of individuals remain stable after adolescence, providing support for the apparent time construct. Thirdly, some individuals exhibit retrograde behavior during a community change, favoring conservative forms as they age and retreating from innovative ones. Because the current study uses data collected only once for each participant, I do not have real time evidence of change for any individual, so I will not be able to determine, in particular, if any pa rticipants are exhibiting retrograde movement. A participant who appears to have retained her native language system upon exposure to a new dialect system may actually have retreated that without using lon gitudinal data. ability to change their phonetic and/or phonological systems across the lifespan. Sankoff and Blondeau (2007) showed how adults can dramatically change thei r place of articulation of a phoneme, and Harrington and colleagues have shown that one speaker can follow community changes through gradual phonetic shift over the course of decades. On the other hand, the speakers examined in Bowie (2005) were stable ac ross their lifespan for most of the variables he studied, and MacKenzie and Sankoff 2 Note that Wagner is referring to language change at all levels, not just in phonetics. An example of an after that person dies (due to communi ty taboos).
39 p language systems are truly stable, nor that all adults will shift their production in the direction of community change. Acquisition of a Second Dialect While language c hange in the individual can be triggered by language change in the home community, it can also result when the individual moves to a new dialect region. The literature on second dialect acquisition can be divided into studies on youth and studies on adult s. Acquisition of a Second Dialect in Childhood or Youth study of the acquisition of the Philadelphia accent by children born out of state (to parents who were also born outside acquisition of five phonetic and three phonological features of the Philadelphia sound system. She found that the out of state born children had at least partially acquired almost all of the variables. Unsurprisingly, there was a negative correlation between age of arrival in the new dialect region and acquisition; the older the children were when they had moved to the Philadelphia area, the fewer Philadelphia dialect features they had acquired. The one phonological feature that proved most difficult to acquire was the complicated tense/lax / / system. In Philadelphia, / / is tense in closed syllables before certain nasals and voiceless fricatives, as well as in some lexical exceptions such as mad bad and glad ; / / is lax elsewhere. The out of state born children came from regions with four different / / patterns: those from the Northern Cities had only tense / /, others had only lax / /, and a third group had tense / / always and only before
40 nasals. Finally, the children originally from New York City had a complicated tense/lax distribution somewhat similar to that of Philadelphia, but different enough that they would have to adjust to master the Philadelphian accent. Almost none of the chil dren accurately acquired this feature. In fact, even children born in the Philadelphia region whose parents were born in other regions did not master this system. Payne concludes that newcomer children acquire phonetic variants in a new dialect with rela tive ease, but acquiring phonological variants is much more difficult. Chambers (1992) interviewed six native Canadian English speaking youths two years after they first moved to South Eastern England. At the time they moved, their ages ranged from seven to fifteen. Chambers measured their acquisition of both lexical and phonological features of South Eastern English (SEE) and derived eight principles of second dialect acquisition (by children and youths). Some of these principles deal with acquisition o phonological rule, such as the North American English flapping rule, than a more complicated system such as the tense lax / / system described by Payne. Further, Chambers found that acquisition of phonological rules manifests itself word by word, not categorically. That is, chi ldren acquire new rules (such as r dropping) as changes to individual lexical items, e.g., first girl later first and still later teacher not as an integrated deletion rule that they apply across the board in all relevant contexts. (However, note that Labov  asserts that in cases of community language change, lexical diffusion of this sort only occurs in cases in which the community members are
41 aware of the change, and there are fairly dramatic phonetic or phonological differences between the olde which of course is a large assumption then it makes sense that Chambers found lexical diffusion in the acquisition of r dropping, which is a rule that the children would have been aware of, and which leads to a dramatic difference in pronunciation. In contrast, the vowel shifts studied in the present research are less phonetically dramatic and may, at lea st in one case, be below the level of consciousness, so we may not expect lexically gradual adoption of these shifts. 3 ) complex rules and new phonemes splits the population into early acquirers and late off in ability to acquire such features of a second dialect. The children who moved to South Eastern England at age who were older when they immigrated made almost no progress in acquiring them. That is, Chambers found an abrupt loss of ability to acquire second dialect features, not a gradual decline with advancing age of first exposure. The present study investigates phonetic (not phonological, as Chambers investigated) system. Finally, Chambers found that phonological rules are acquired (or abando ned) 3 not for individual lexemes. However, if participants were shifted in some lexemes and not others, then their average formant values should still be shifted enough to show a statistically significant change. Further investigations could look at the effect of individual lexemes.
42 the children he studied were mostly successful at losing their native flapping rule; orthography clearly shows a
43 did not fully acquire the British English system to the point of categorically using British variants, even after five years of exposure To summarize, children who move to a new dialect region have been found to be capable of successfully adopting the new dialect, as least to the point of sounding as if they were from the area, even if they do not master all the linguistic constraints on variants. Children who move to the new region at a younger age acquire the second dialect more successfully than those who arrive when they are older. Finally, so called simple phonological rules (e.g., those without apparent exceptions) are acquired mor e easily than more complex rules, such as tense/lax / / conditioning, that depend upon the environment in which the phoneme appears or force learners to memorize lexical exceptions. Acquisition of a Second Dialect in Adulthood The studies above showed that children can acquire a second dialect quite successfully, if not perfectly. Adults may face a greater challenge in acquiring a new dialect, given that their language systems are less malleable than those of children (though not, as we have seen, complete ly inflexible.) On the other hand, adults acquiring a new dialect may have the advantage of greater metalinguistic awareness than children have; adults moving to a new dialect may be better able to recognize not only the differing dialectal features thems elves, but also their social significance and the advantages or disadvantages the speakers may gain/suffer by adopting the new may be different than those exhibited by children. Trudgill (1986) described adult second dialect acquisition from the point of view of accommodation, pointing out that adults may adopt features of a new dialect specifically
44 to avoid being misunderstood. (Young children would presumably be less aware of this problem and less likely to accommodate for this reason.) Two examples from his own (natively British English) speech are the acquisition of American English flapping to prevent listeners from thinking he was asking for pizza when he was intr oducing himself as Peter and the switch from [a] to [ ] in words like glass and bathroom to allow waiters, concierges, etc., to understand his requests better (23). Trudgill noted, however, that he had been unable to acquire the flapping rule completely. Bortoni (1991) investigated dialect leveling by Brazilians who had moved from other areas of the country to Brasilia at age 17 or older and those who had moved to Brasilia as children. At the time of the study, the city Brasilia was only 28 years old and had not developed its own distinctive dialect features (though Bortoni noted that some may have been emergent). Thus, the linguistic task for immigrants to the area was less to adopt the dialect of the new region and more to lose the distinctive features that marked them as coming from their home (perhaps less prestigious) region. Bortoni but that even for adults, age of arrival in the new city played a role, with those who migrated as younger adults losing their native dialects more completely than those who arrived in middle age. Further, the social network of the individual played a role. Those adults who interacted largely with Brasilia natives adapted to phonological features more completely, while even children who immigrated before the age of six but interacted mostly with immigrants from their own native dialect region adults who arrive in a new
45 dialect region at a relatively young age (probably about age 30) and socialize with natives of the new region may acquire a second dialect (nearly) completely. Conn and Horesh (2002) investigated acquisition of the Philadelphia phonetic and phonological system by two speakers who had moved there (separately) from Michigan in adulthood. They specifically looked at one phonetic feature, /o/ fronting, and one phonological feature, the tense/lax ( ) system. In Michigan, the speaker s would not have been exposed to /o/ fronting, and their / / would have been uniformly tense. Each speaker had lived at least 15 years in Philadelphia. One speaker, called Fred in the study, had not changed his / / to reflect the Philadelphian system, bu t he had adopted /o/ fronting. The other speaker, Loretta, had not changed her /o/, but she had changed / / so that it neither was uniformly tense as in Michigan, nor reflected the Philadelphian system. This is initially puzzling, but Conn and Horesh exp lain that what is being called ) system is really the system exhibited by white Americans, who tend to have a different ( ) system than white sp eakers, and who do not participate in /o/ fronting. Loretta may have in fact adjusted her speech to reflect the dialect she was regularly exposed to in Philadelphia. This study provides some further und system in adulthood, though it does not clarify the relative ease of adopting a phonological or phonetic feature. While Payne (1980), Chambers (1992), and Conn and Horesh (2002) focused on the effect of the complexity of the phonological or phonetic ru les to be acquired, Auer and colleagues (e.g., Auer ; Auer, Barden, and Grosskopf ) investigated adaptation to a second dialect through the lens of salience. This group of studies
46 focuses on native speakers of the low prestige Upper Saxonian V ernacular (Leipzig/Dresden region in the former East Germany) dialect of German who moved to two different higher prestige dialect regions in former West Germany for work after German Reunification. The participants in the study should have been highly m otivated to adapt to the new dialects due to the strong stigmatization of their native vernacular. Participants were interviewed several times over two years to provide a longitudinal acquisition of the non standard features of their native dialect. dialect contact, more salient features of a dialect will be gained/lost more easily than less salient ones. Schirmunski proposed six criteria that could be used to determine whe ther a dialect feature should be considered salient. Auer and colleagues focused most on the issue of lexicalization: Schirmunski predicted that features that are part of regular, Neo Grammarian sound change will be less salient (and acquired more slowly) while those that are lexically specific / lexically diffused will be more salient. In fact, although Auer (1997) found that participants acquired lexically specific features of their new dialects more quickly than less salient features, Auer et al. (199 8) found that lexically specific features of the native dialect were less prone to being lost upon exposure to the new dialect. Thus, the effect of lexical diffusion on dialect acquisition is more complex than Schirmunski predicted. However, overall Auer and colleagues have found that more salient features of a new dialect are acquired more easily, and more
47 salient features of native dialects are lost more easily in cases of exposure to a (more prestigious) second dialect in adulthood. 928/1929, 1930) six criteria for salience, the one most with either/or options, su ch as the [r]/[R] production of French /r/ mentioned above, are more salient.) The present study investigates acquisition of a vowel chain shift, the Northern Cities Shift. Native speakers of the target (Chicago) dialect have productions that vary widely from standard (non shifted) to extremely shifted, along with many points in between. The variables under investigation thus have intermediate forms and should theoretically be less salient, and less easily acquired, than, for example, the difference betw een a stop and fricative production of /g/ in German. The participants in the current study may not be acquiring any phonetically dichotomous variables that could be used for comparison, though, because as shown below, the Northern dialect region is mostl y than .) In a dissertation that served as a partial model for the present study, Bowie (2000) investigated the changes in perception and production undergone by natives of his hometown of Waldorf, Maryland, after they moved to other areas of the country in of a documented with both word lists and sociolinguistic interviews, and their perception was measured with commutation tests, in which the participants were first recorded re ading
48 word lists of minimal pairs with vowels that may be merged pre laterally (e.g., pull vs. pole vs. pool ); later, the recordings were played for the participants, who were asked to identify which word they had pronounced. 5 In perception, the youngest lifelong Waldorfians (born after 1979) generally had pull, pole, and pool all merged, though older listeners distinguished pool and pole in perception and only merged pull with both pool and pole In contrast, the vast majority of lifelong Waldorfians co ntrasted all these words in production, while only a single speaker merged all three. (She was the only one to merge pool and pole .) All of the Waldorf exiles distinguished pole and pool in perception. All of those who had been out of Waldorf for less than five years merged pull and pool in perception, while over half of those who had been out of Waldorf longer perceived these vowels as distinct. Finally, only one exile, who had been out of Waldorf for 13 years, distinguished pull and pole in perception. Generally, the longer a listener had been away from Waldorf, the more of these vowels she perceived as distinct. However, there was no such progressive pattern no detectable pattern at all, really in the production data. When compar Waldorfians who were in the same age cohort as the exiles, interesting patterns appeared. Within this restricted group, all lifelong Waldorfians merged pull and pole in perception, all bu t one merged pull and pool and over one fourth merged pool and pole In contrast, one of the exiles distinguished pull and pole over one third distinguished pull and pool and all of them distinguished pole and pool in perception. The pattern is 5 present (That is, participants in my study were assumed to distingu ish each pair of phonemes. The question was where the line between the two phonemes fell.)
49 less p rofound for production, but generally a higher percentage of exiles than lifelong Waldorfians distinguish each of the vowel pairs in production. Bowie notes that it appears that some of the Waldorf exiles have been able to reverse their mergers in percept ion, but mergers in production were more difficult to reverse. 6 Also, the easiest change to reverse was the merger of pole and pool which was still an ongoing change in progress in Waldorf at the time the exiles grew up, whereas the two mergers that were reverse when they left Waldorf. Maine to Arizona and even one for awhile to Ontario, so w hile they would all have been moving away from the same native dialect, they were acquiring very different second only their loss of a native dialect. The participa nts were probably exposed in their new homes to varying degrees of distinction of the merged vowels in question, though, which makes his conclusions regarding the speakers who did not demerge perhaps questionable. 7 That is, a speaker could maintain a nati ve vowel merger either because the vowels are difficult to demerge or because the same vowels are merged in their new homes as well. The present study measures language change in a group that moved 6 As Bowie notes, this is rather surprising considering that the perception tests involved participants listening to their own production This probably indicates was simply inadequate, and the speakers were distinguishing the pairs in production in a way that his measurements did not detect. Bowie does not indicate at what point he measured the formant values for the vow els, though it seems he measured them at only one point per vowel. Perhaps the participants were distinguishing the vowels in production either at different time points in the vowels, or through another dimension, such as duration. 7 Of course, the comple mentary criticism could be lodged against Payne, for comparing children coming from very different dialect areas, although she did outline the various native / / systems of the children.
50 not only from the same city, but also to the same metropo litan area, and who thus are exposed to the same new dialect features. broadcasts to document language change, Sankoff (2004) used footage from Michael Up documentary serie s to investigate language change across the lifespan in two speakers who were exposed to new dialects in adulthood. These two speakers, Nicholas and Neil, both lived in the North of England when first filmed at age 7 and remained there until at least age 16, but by age 35, their lives had taken quite different paths: Nicholas had earned a PhD in physics from Oxford and become a professor in the United States, while Neil had dropped out of university after one semester and by adulthood was chronically unemp loyed and living in Scotland, though he had also spent time living in London and Wales. Short U. For the former, while Northern English speakers (as Nicholas and Neil nativel y were) pronounce path, grass, pat, and grab with the same short [a], Standard Southern British (SSB) speakers pronounce path and grass :]. Both groups of speakers are well aware of this dialectal difference and often parody it when imitating speakers from the other region. In contrast, Northern English speakers are generally unaware that their Short U system is different from that of SSB speakers. Northern English does not have / / in its vowel inventory, so buck and book are homophones, [b k]. SSB divides the / / and / / classes of words the same way American English
51 (Broad A) that they were conscious of, and thus might either intentionally a dopt or reject, and another (Short U) that they would presumably adopt only subconsciously. 8 Nicholas, who would have been exposed to SSB for years at Oxford but then moved to the U.S., had not adopted the Southern British Broad A; whether this was due t o rejection of the Southern British standard or accommodation to Standard American is uncertain. Neil, however, had almost entirely adopted the Broad A system. Notably, neither speaker had any cases of hypercorrection; they never used Broad A when an SSB speaker would use short [a]. Both speakers, however, had almost completely 9 adopted the Southern British (or American) Short U pattern by age 35. This is remarkable because they had to master a phonemic split, correctly memorizing which of the words the y grew up pronouncing with [ ] must be moved into the / / class. Further, once again, neither speaker had a single token of hypercorrection; all words that Southern British and American speakers would produce with [ ] had remained in these / c that most individuals are less mobile and also undergo less linguistic change across the and phonemic systems upon exposure to new dialect systems in adulthood. Evans and Iverson (2007) similarly investigated changes in Short U and Broad A in adulthood, but they analyzed changes in perception as well as production. 23 natives of a small town in Leicestershire, who should have had the Northern English Short U and Broad A pattern natively, participated in experiments before starting university and 8 It is also possible that they became consciously aware of this regional variation once they were more regularly exposed to the SSB system. 9 The exception seems to be in excited or emotional speech, when the two men occasionally produced a / / class word with [ ] as they did as boys.
52 again three months, one year, and two years after beginning their studies. 10 (The participants attende d different universities, but all should have been exposed to Standard Southern British at school.) Production was measured through impressionistic ratings and acoustic measurements of target words in a reading passage, Arthur the Rat and in frame senten ces. Trained phoneticians rated most of the speakers as impressionistically sounding more southern in the later recordings, though the changes in the ratings were usually fairly small. (That is, the accents sounded slightly, not dramatically, more southe rn.) Acoustic measurements showed that the participants had altered their production of bath toward a Broad A production, closer to their vowel in bard Their vowels in bud and cud had also moved towards a more central, southern style production. Howeve r, while Sankoff (2004) found no instances of hypercorrection in the Up could so that the word remained homophonous with cud That is, unlike the Up speakers, these university stu dents did not acquire the phonemic split; instead, they uniformly shifted the locus of their single / / phoneme. For perception, the participants heard computer generated versions of CVC target words inserted into carrier sentences read by someone who soun ded like either a native Northern English or native SSB speaker. The participants rated the goodness/appropriateness of each exemplar of the target word, and the computer an optimum exemplar was developed. Overall, for all the vowels studied, the authors found that participants did not significantly change their ideal exemplars over time; their 10 Only 19 of the participants were able to be contacted two years after starting university. The other four were thus only tested at the first three time points.
53 changes in production were not matched with changes in perception. Nor did the y choose different ideal exemplars based on the accent heard in the carrier sentence; oddly, they did not choose more northern style ideal exemplars when the speaker sounded northern, or vice versa. The only significant effect found in the data was an eff more southern sounding ideal exemplars for bud and cud Overall, Evans and Iverson (2007) shows that adults exposed to a second dialect may change their production (albeit slightly) but perhaps not their perception of D2 features. Note also that they found that participants were not able to master the phonemic split that Sankoff (2004) found the Up speakers to have adopted. 11 Nycz (2011) investigated both the acquisition of a phonemic split and the loss of a phonological rule upon exposure to a second dialect in adulthood. Nycz interviewed 17 native Canadians who had moved to the greater New York City region in adulthood. Canadian English exhibits the low back merger, in wh ich cot and caught are homophonous, and the famous Canadian raising, in which /ai/ and /au/ are raised to [ i] and [ Americans use when imitating Canadians. (Nycz measured only /au/ raising, not /ai/ raising, in this study.) New York English has neither the low back merger nor Canadian raising. 11 duction only impressionistically, while Evans and Iverson took acoustic measurements. Perhaps an acoustic measurement of the Up series would / class had in fact been centralized. (Of course, the recording q uality of the documentaries may make such an analysis impossible.)
54 A minimal pair task (reading minimal pairs and then reporting whether the words rhymed 12 ) showed that all speakers had retained the low back merger in both perception and production, but analysis of word list and sociolinguistic interview data showed that 1 1 of the 17 speakers actually distinguished /a/ and / / in production in F1 maximum.) Additionally, she found that word frequency affected acquisition of the split: more frequent lexemes were more distinct than less frequent ones. Nycz notes that it is unusual to find a group of speakers who distinguish a pair of phonemes in a casual interview situation but not in a minimal pairs task. She suggests that while her partici pants actually produced somewhat New York like low back vowels in normal speech, they retained their native dialect norms as their production targets, thus leading to the merged production in the minimal pair task. Nycz analyzed Canadian raising only in p roduction, using natural tokens from her sociolinguistic interviews. (She again measured each vowel only at one point, the nuclear F1 maximum.) She found that seven of her 17 participants still raised /au/ before voiceless consonants in all relevant word s, while another eight had apparently lost the general raising rule, but still raised in the word about and sometimes out i.e., the two words most stereotypically associated with Canadian raising. (The other two speakers had no significant differences ac ross the word classes and thus may have completely lost their Canadian raising, but in one of these cases, the paucity of extractable tokens may have prevented a real difference from being statistically significant.) The findings for Canadian raising cont radict those for the low back merger 12 Like Bowie (2000), Nycz (2011) tested perception quite differently than I do in this dissertation.
55 in that, while the participants had acquired the New York like low back distinction most completely for the most frequent lexical items, for Canadian raising, it was the two most frequent lexical items, out and about t hat exhibited the strongest retention of the native dialect feature. Moreover, Nycz was surprised to find a negative correlation between the low back merger and Canadian raising: participants who more fully distinguished the low back vowels also showed hi gher levels of Canadian raising. That is, those who seemed to have the greatest acquisition of one second dialect feature seemed to have acquired another one the least. Nycz proposes to explain this seeming contradiction with the concept of linguistic sen sitivity which includes both higher level conversational skills (e.g., tact) and sensitivity to lower level linguistic variation. She suggests (p. 204), Individuals with higher degrees of L[inguistic] S[ensitivity] would be predicted to more readily acqu ire the [low back] distinction as a result of interaction with [native New York] speakers, and to more readily exhibit Canadian Raising in the context of an interview that focuses on Canadian identity. Speakers with lower degrees of LS would lag in the ac quisition of the low back vowel distinction, but simultaneously be less likely to activate their raised Canadian vowels in the interview setting. This conjecture implies that those speakers who distinguished the low back vowels but maintained Canadian rais ing in the interview would not actually raise /au/ (or would raise it less) in everyday conversation; they only raised it when told to think about Canadian identity. At any rate, Nycz has further shown that adults who move to a new dialect region can adju st their phonemic systems to be more like, if not exactly like, that of natives of their new home. In summary, acquisition of a second dialect is often less complete in adulthood than in childhood, but the authors cited here have found evidence that adults can indeed adjust their language systems with exposure to a new dialect region (or, in the
56 cases of Bortoni and Bowie, we might say lack of exposure to their native dialect). We features, and that adults are more likely than children to exhibit such conscious accommodation. Gender and Sex in Sociolinguistic Research A secondary research focus of thi s dissertation is the relative importance of binary biological sex and socially constructed gender to language variation and change. Since early in the history of sociolinguistic research (e.g., Trudgill 1972), sociolinguists have viewed sex as an importa nt factor in language change. Labov (1990) summarized the sex related findings to that date and drew the following generalizations: Principle I: In stable sociolinguistic stratification, [males] use a higher frequency of nonstandard forms than [females] ( 210). Principle Ia: In change from above, [females] favor the incoming prestige form more than [males] (213). Principle II: In change from below, [females] are most often the innovators (215). 13 studies that investigated biological sex, not gender. These three principles have rightfully been important to the last twenty years of sociolinguistic res earch, but one wonders what different principles might reveal themselves if gender as well as sex were regularly incorporated into sociolinguistic studies. (Labov himself was aware of this ex is an appropriate 13 speakers adopt changes both below the level of consciousness (i.e., that speakers do not realize they are adopting) and those that originate in the lower socioeconomic classes.
57 category to explain linguistic behavior. It follows that an intervening variable must be formulated in terms of distinct cultural roles assumed by male and female members of Matsumoto and Britain (2003) criticized one principles based specifically on the difference between sex and gender. Principles I and Ia are sometimes viewed as paradoxical (Labov 1990) because in one case, females use more of the standard form, and in the other cas e, they use the (necessarily nonstandard) incoming form. Intuitively, one would expect a given group to favor either the standard or the nonstandard form consistently. Matsumoto and Britain, however, argue that this apparent paradox actually arises from standard sociolinguistic methodology rather than from reality. They studied a Japanese Palauan bilingual community in Palau. Instead of aggregating results at a macro level across all females and males, a methodology that they claim creates the appearanc e of the paradox above, Matsumoto and Britain practiced participant observation and conducted ethnographic interviews in the community. This led them to divide the sex groups up into various gender groups based on community defined social groups. They fou nd that some women behaved conservatively, favoring the standard, while another group gro ups, rather than being willing to divide sex groups into smaller, gender based ones. One study that still relied upon sex differences but also dug somewhat deeper into socially constructed gender is Fought (1999). Fought interviewed Latino teenagers and young adults in western Los Angeles and analyzed their use of /u/ fronting, an ongoing
58 change in the wider, dominant community. Some of her participants were active in local gangs, others were not gang members themselves but maintained close ties to gang members, and still others had no gang affiliation. Further, Fought was surprised to f ind that several of her female participants also identified as moms and were identified socially as part of the mom group. Fought found that different factors determined /u/ fronting among females and males. For the females, the most important factor was gang affiliation; females with no gang affiliation all fronted /u/. For those who were in or affiliated with gangs, the next most important determining factor was social class; middle class females with gang affiliations fronted /u/ more than working cla ss gang affiliated females. For males, on the other hand, social class was the most important factor. Working class males, regardless of gang affiliation, did not front /u/. As I said before, this study still relies upon binary biological sex as a facto r, but Fought also investigated the importance of socially constructed, community specific groups. Cheshire (2002) reviewed sociolinguistic studies that have attempted to operationalize gender rather than (or in addition to) sex but concludes that the resu lts (1988) in depth ethnographic study of high school cliques in a suburb of Detroit and and language use, but the gender roles expressed by their participants were deeply dependent upon the specific social circumstances of their exact communities. These studies cannot be used to draw conclusions about the role of gender in language change in general. What is needed is a method for integrating gender into variationist research in a way that will allow comparisons across speech communities. Luckily,
59 psychologists have developed tests of gender designed to have results that are generalizable a cross various communities. Bittson (2007) administered the Bem Sex Role Inventory (BSRI, Bem 1974) to undergraduates at the University of Florida. The BSRI characterizes a respondent as one of four gender types: feminine, masculine, androgynous (high iden tification with both masculine and feminine gender traits), and undifferentiated (low identification with both groups of gender traits). Using the BSRI, Bittson compared the utility of binary sex and gender as predictors of use of the standard [ I ] and nonstandard [ I n] variants of the variable ( ing) (as in going vs. ) in sociolinguistic interviews with her participants. Interestingly, she found that all biological females used all or predominantly the standard variant, while biological males varie d. However, all participants who used exclusively the nonstandard variant had masculine gender. This small study does not give a clear picture of the practical differences between sex and gender for sociolinguistic studies, but it does suggest that diffe rentiating the two is worthwhile. Administering standardized gender tests that have been validated by psychologists will allow sociolinguists to integrate gender into variationist research with results that will be generalizable across speech communities. the inclusion of a standardized gender test in the present study. American Regional Dialects I present here an overview of the American dialect regions relevant to this study: the North, the South, and the Midla nd. The North The North is defined by the Northern Cities Shift (NCS; Labov, Yaeger, and Steiner 1972). (This shift is also known as the Northern Cities Chain Shift.) While the
60 NCS is, as the name implies, most advanced in the major urban centers of th e North, area extending westward from the White Mountains, covering western New England, New York State, the Northern Tier of counties in Pennsylvania, northern Ohio, Ind iana (Labov 1991: 14). The NCS consists of a chain shift 14 of five vowels and an independent shift of a sixth. In the chain shift, / / falls toward /a/. /a/ in turn shi fts forward toward //, and // raises toward either /e/ (Labov 1991, Wolfram and Schilling Estes 1998) or / / (Eckert 1991). / / is backed toward / /, and / / is backed toward the spot vacated by / /. Finally, / I / is shifted toward the back. This is no t part of the chain shift but rather an independent vowel drift. Additionally, as in other American dialects, /u/ is being fronted. (According to Fridland, Bartlett, and Kreuz (2004), / / is fronted as well.) Because this fronting of /u/ is found across A merica, it does not differentiate the North from other dialect areas and is not considered part of the Northern Cities Shift. The NCS is illustrated in Figure 2 1, adapted from Wolfram and Schilling Estes (1998). The present study focuses on the acquisiti on of the NCS in Chicago. McCarthy (2011) investigated the current state of the NCS in Chicago specifically. One aim was to determine whether the NCS was still advancing rapidly in Chicago, or if it might have stabilized or begun to reverse. I should no of analysis: first, like many sociolinguists, she measured the vowel formants at only one time point per vowel. This was the point of inflection for F2 for obvious diphthongs, and 14 Stockwell and Minkova (1997) dispute the claim that it is a chain shift, arguing instead that it may be a series of independently motivated vowel drifts.
61 usually the maximu m F1 or the Second, she measured only the formants that she expected might show change in progress, thus preventing herself from being surprised by a vowel shift on an unexpected axis. She measured F1 and F2 of / /, / /, / /, and / I /, but only F2 of /a/, / /, and /a current state of the NCS in Chicago. McCarthy recorded word lists read by 36 native Chicagoans born between 1927 and 1988. She found that Chicago speakers clearly participate in the NCS; their / / is raised and fronted, and / / is shifted backward. On the other hand, their /a/ is less fronted than would be expected based on NCS studies in other cities, and / / is not lowered and fronted (moved toward /a/) as it is in other NCS cities. In terms of change over time, she found the following: / / is stable /a/ may be moving farther back, i.e. reversing the NCS / / may be lowering / / and / I / are moving back and lowering / / is moving back In short, McCarthy found that Chicago speakers participate in the NCS. Of the vowels I investigated in my analysis, she found that the shifts in / / and / / are still advancing in Chicago, while / / is stable and /a/ may actually be reversing the shift. The South The South is defined by another set of vowel shifts, the Southern Shift. Unlike the NCS, the Southern Shift is not a single chain shift but a collection of multiple systematic vowel shifts. Labov (1991) states that this pattern is
62 (22). The Southern Shift combines one shift in the system of front vowels and another in the back vowel system. In the front, the lax vowels / I / and / / are made tense by raising and fronting, while the tense vowels /i/ and /e/ are centralized and become lax. In essence, the pairs /i/ ~ / I / and /e/ ~ / / exchange places. In the back, the tense vowels /u/ and /o/ are both shifted toward the front. ( Fridland et al. 2004 again state that / / is fronted as well.) The fronting of the back vowels is more recent than the exchange of tense and lax vowels in the front (Fridland et al. 2004). The Southern Shift is illustrated in Figure 2 2, adapted from Wolf ram and Schilling Estes (1998). The most famous vocalic identifier of the Southern dialect, though, is not part of the Southern Shift. This is, of course, /ai/ monophthongization. Labov, Ash, and Boberg (2006) map this feature from the Ohio River south t o north Florida and west into southern Oklahoma and most of Texas. (There are also isolated tokens in Pennsylvania.) The conditioning environment for /ai/ monophthongization is not uniform across the South, however (Labov et al. 2006). In most areas, /a i/ is monophthongized (almost) exclusively before voiced segments and prepausally, while in Appalachia and northern Texas it also appears before voiceless segments. As their respective shifts become more advanced, the Northern and Southern dialects are div erging (Labov 1991). At least by the late 1980s, these vocalic differences were great enough to cause comprehension difficulties in cross dialectal go, Birmingham, and Philadelphia. Listeners from
63 each city heard isolated words (e.g., block ), short phrases ( on one block ), and then the entire sentences or phrases ( senior citizens living on one block ) pronounced by speakers advanced in either the NCS o r the Southern Shift. (That is, the isolated words and short phrases were extracted from the complete sentences, so listeners heard the same target word token three times.) Labov (1989) showed that even in the sentence context, Birmingham listeners had t rouble understanding words spoken by Chicagoans. Labov and Ash (1997) found that Chicago listeners had the same problem understanding Birmingham speech. The Midland The third dialect region relevant to this study is the Midland. There is, in fact, a lo ngstanding debate over the existence of the Midland dialect. 15 Some dialectologists (e.g., Kurath 1949, Johnson 1994, Murray 1993) have felt that there is adequate evidence for the Midland to be called a separate dialect, while others (e.g., Houck 1967, Ca rver 1987, Davis and Houck 1992) have argued that the northern section of the so called Midland is actually the Lower (or Inner) North, and the southern section is the Upper (or Inner) South. Even some who are in favor of the Midland dialect (e.g., Murray and Simon 2006) admit that many of its defining features are actually shared with either the North or the South. If the Midland dialect does exist, it apparently is defined by more grammatical than phonological features. Murray and Simon (2006) list seve nteen grammatical features that they say occur only or predominantly in the Midland. Some of these are the needs + past participle construction, as in My car needs washed or These towels need folded ; 15 See Murray and Simon (2006) for a history of this debate.
64 the phrase all the + comparative adjective, as in ; and positive anymore as in We eat a lot of rice anymore There are also a number of lexical items confined to the Midland, such as pop for soda and mango for green pepper. There are relatively few phonological features that differentiate the Midland from other American dialects. One is /l/ vocalization (e.g. pronouncing gold as [go:ud]), which Dodsworth (2005) says is most prevalent in that region. Another trait exhibited in more and more of the Midland region is the low back merger. In the low back merger, / / and /a/ are merged so that caught and cot are homophonous. The quality of the merged vowel varies by region. This merger is expanding (Labov 1991), and interestin gly, there is evidence that the merger is not the result of a single innovation. Rather, the merger arose independently in several locations (Herold 1990, Majors 2005) and is spreading from all these centers of innovation. (Note that speakers in Western states also exhibit the low back merger; this characteristic is not limited to the Midland.) Thomas (2001) found evidence of /ai/ monophthongization in the Midland, but this of course does not differentiate it from the South. Thomas specifically found tha t /ai/ monophthongization was more common in the South Midland than in the North Midland North and South Midland should be reclassified as part of the North and South respectively. measured, i.e., how people actually speak differently. The dialectolo gy of the ear, in
65 users hear themselves as speaking differently than people in what dialectologists consider to be other dialect regions? To test dialectology of th e ear, Preston made recordings of speakers whose homes ranged from central Michigan to southern Alabama. He then had listeners from Michigan and Southern Indiana listen to these clips and try to place the speakers on a map. Statistically, the listeners f rom Indiana treated the recordings made from northern Michigan all the way to Southern Indiana as the same. They distinguished recordings from Kentucky and Tennessee, but they treated recordings from northern and southern Alabama as the same. The Michiga n listeners, on the other hand, treated recordings made from Michigan only through n orthern Indiana as the same. They distinguished the recordings from central Indiana, Southern Indiana, and Kentucky, but heard recordings from Tennessee and Alabama as the same. Preston feels that his data show that Hoosiers hear themselves as the same as those farther north, while Michiganders perceptually link Indiana speakers with Southerners. I disagree with his interpretation, however. I agree that Hoosiers do not pe rceive much difference between their speech and that of Michiganders, but to me the data show that Michiganders hear three separate dialects: a North, a South, and a middle region (maybe a Midland) into which Indiana falls. That is, they can consistently hear a difference between speakers from Indiana and those from farther south. Preston notes that Michiganders distinguish the accents exhibited by those in Indiana from those from regions farther south, but he still thinks they hear all these regions as t he South.
66 To me, if they are able to distinguish these regions consistently, then it is probable that they are hearing different dialects. Preston concludes from production data that in terms of the dialectology of the mouth, there is a North, a South, a Indianapolis. In terms of dialectology of the ear, Preston finds that there is only a North and a South, but where the S outh begins depends on the geographic origin of the listener. As I said before, I feel that his perception data provide possible evidence for a perceived Midland. Midl and, it is distinguished mostly by grammatical and lexical features. There are few phonological features unique to this region. The only vocalic feature that may distinguish the Midland from the North and the South is the low back merger, which is also e xhibited in the West. This chapter began with a review of studies of language change across the (1983) studies of the relationship between perception and production in languag e systems in adulthood. The sections on second dialect acquisition showed that complete ) phonetic or phonological systems is rar e not only for adults, but also for children. I then noted important generalizations that have been drawn about the relationship between sex and language change and briefly reviewed sociolinguistic studies whose
67 authors have attempted to integrate gender as well as sex into their research. In of a different standardized gender test in the current study. Finally, I gave an overview of the relevant regional dialects of American English. The next chapter describes the design and data collection for the current study. /i/ /u/ / I / / / /e/ /o/ / / / / / / // /a/ Figure 2 1. The Northern Cities Shift /i/ /u/ / I / / / /e/ /o/ / / / / / / // /a/ Figure 2 2. The Southern Shift
68 CHAPTER 3 METHODOLOGY In Chapter 1, I explained that for this study, I collected data from three groups: the Lifelong Jasper group, who had lived all or most of their lives in Jasper, Indiana; the Lifelong Chicago group, who had lived all or most of their lives in Chicago; and the Mobile group, who grew up in Jasper and moved to Chicago in adulthood. My goal was (NCS) in phonetic perception and/or production. In this chapter, I provide specifics about the methodology of the experiment. Vowels of Interest in the Study NCS vowel changes: /a/ fronting, / / raising, and / / backing. The perception test therefore m /, / /, and / / (/ / was included to allow comparison between / / and the vowel it was moving towards in the NCS), and production data was designed to elicit tokens of these four vowe ls. Predictor Variables Used in the Study The following predictor variables were used in analyzing variation in perception and production in this study: residence group (Lifelong Jasper, Mobile, or Lifelong Chicago); year of birth (YOB); sex; gender, measu red along various continua as number of years lived in Chicago (YLC), and percentage of life spent in Chicago (PLC). YOB was included to give a picture of change over time in Jasper and Chicago. Studies reviewed in Chapter 2 found that the apparent time data underestimates the speed of
69 change in a community, but such data can still indicate the direction of a change. Sex and gender were included to address the research qu estion of whether binary biological sex or socially constructed gender is a more reliable predictor of linguistic behavior. AAC, YLC, and PLC were included to test whether the age of first exposure to a new dialect or the extent of exposure (as raw years or as a percentage of life) is a better determinant of acquisition of a new dialect in adulthood. Jasper, Indiana Location of and demographic information for Jasper Two thirds of the parti cipants in this study were born and raised in Jasper, Indiana, a small town in Dubois County in the southwestern part of the state, about 45 miles (as the crow flies) northeast of Evansville and 65 miles west northwest of Louisville, Kentucky. 1 The maps i n Figure 3 1 show the location of Jasper within Dubois County and Dubois County within Indiana. Jasper is my hometown. For the purposes of this According to the US Census Bu 15,038. The population was over 90% non Hispanic white and 7.7% Hispanic or Latino. (No other racial or ethnic category tracked by the census reached 1% of the population.) I mention this simply because it e xplains why I chose to have all white participants (except for one Lifelong Chicago participant of Asian descent). 1 Most J to Evansville to go shopping, for example, but going to Louisville is generally considered a more special event.
70 The debatable dialect of Southern Indiana There is actually no consensus in the literature as to what dialect is present in Southern Indiana (including Jasper). Part of the problem is the ongoing debate over whether the Midland dialect exists (as described in Chapter 2), and part is simply an issue of how far north the South extends. Clopper and Pisoni (2004, 2006) include the region surround ing Jasper in the South Midland. Carver (1987), who argued against the existence of the Midland, placed Jasper in the Hoosier Apex, a subsection of his Lower North. In contrast, Preston (2003), who does believe in the Midland, places the Midland so far n orth that Jasper (and a fair amount of the rest of Indiana) lies in the Upper South. If Preston (2003) is correct and Jasper lies in the Upper South, then speakers from Jasper should exhibit no Northern Cities Shift and no low back merger, but the presen ce of the Southern Shift. Clopper and Pisoni (2004, 2006) would instead predict the presence of the low back merger but no Southern Shift or NCS, and Carver (1987) would predict the presence of the low back merger 2 and perhaps a not at all advanced form o f the NCS. The purpose of this study was not to settle the debate of where the boundaries of American dialects lie. I did not examine all the vowels that would have been necessary to clarify which dialect native Jasperites use. For my purposes, though, it was most important that Jasper speakers simply have a different dialect (or phonetic system, at least) than Chicago speakers, and later chapters will show that this criterion was largely met. 2 Since Carver did not delineate a Midland the low back merger is found in his Upper South and Lower North.
71 To address the dialect boundary debate a bit, though, my findings do not indicate an incipient NCS in Jasper. Additionally, although I did not measure this specifically, it specifically measure the vowels in the Southern S hift, either, but several (though far from all) of the Lifelong Jasper participants spoke with a noticeable, albeit mild, Southern accent. I believe that based on grammatical criteria, Jasper lies in the Midland, while based on phonetics and phonology, it lies in the Upper South. Chicago, Illinois Demographic information 31.7% non Hispanic white, 28.9% Hispanic or Latino, 32.9% black, and 5.5% Asian. Maps of Chicago neighborhoods are too complex to reproduce legibly here. However, was available at http://en.wikipedia.org/wiki/Chicago_neighborhoods This page also has links to maps of the neighborhoods and community areas. The Northern Cities Shift in Chicago Whereas Jasper could lie in several different dialect regions, Chicago is clearly in the North. Indeed, it is one of the cities that define the Northern Cities Shift. Chapter 2 n Chicago.
72 Participants Criteria for Inclusion Participants in this study constituted three groups: Lifelong Jasper participants, Mobile participants, and Lifelong Chicago participants. Throughout the dissertation, participants are referred to by number (e.g., Participant 5), not by name. Requirements for Lifelong Jasper participants: Born in Jasper or moved there by age 5 Then lived in Jasper continuously until the end of high school Lived no more than 6 years outside of Jasper after high school Residin g in Jasper at the time of the study Never lived in Chicago or a close suburb No history of speech problems, hearing problems, or brain injury Requirements for Mobile group participants: Born in Jasper or moved there by age 5 Then lived in Jasper continuo usly until the end of high school Moved to Chicago or a close suburb at some point after high school Residing in Chicago or a close suburb at the time of the study Had lived in Chicago or a close suburb for at least one year at the time of the study No his tory of speech problems, hearing problems, or brain injury Requirements for Lifelong Chicago participants: Born in Chicago or a close suburb 3 or moved there by age 5 Then lived in Chicago or a close suburb continuously until the end of high school Lived no more than 6 years outside of Chicago or a close suburb after high school Residing in Chicago or a close suburb at the time of the study Never lived in Dubois County, Indiana No history of speech problems, hearing problems, or brain injury Obviously, it was preferable for participants to have moved to Jasper or Chicago well before the age of 5, and in fact, 74 of the 76 participants were born in their 3 also accept people from other towns in Dubois County outside of Jasper. The reason is th at I was able to get enough people from Jasper with little trouble.
73 respective cities. (Lifelong Jasper participants 1 and 3 moved to Jasper at ages 2 and 1, respectively. ) Similarly, it would seem to be preferable for the lifelong Jasper and lifelong Chicago participants to have spent well fewer than six years living outside of their respective cities in adulthood. However, allowing six years outside meant that participan ts could have left to pursue higher education before returning to live in their hometowns. Because many of the Mobile group participants were highly educated (12 Jasper and Chicago participants with college degrees whenever possible. (That is, I wanted to make sure that my Lifelong Jasper participants were the type of people whom the Mobile participants would probably have interacted with regularly in Jasper, so that ch of my Lifelong Jasper participants.) For most people, getting higher education meant moving away from their hometown for several years (though many came home during th e summers). 73 of the 76 participants were monolingual native speakers of English. The other three (all Lifelong Chicagoans) were natively bilingual: one was natively bilingual in English and Spanish, one was bilingual in Lithuanian and English, and a thi rd had known some Korean as a child but said she had forgotten most or all of it. All three of these speakers sounded like native speakers of American English when speaking English. In addition to the requirements listed above, Lifelong Jasper participant s were matched with the Mobile participants as follows: for each Mobile participant, I recruited one Lifelong Jasper participant of the same sex who was born no more than three years
74 before or three years after the Mobile participant. For example, Partici pant 12, a Mobile female born in 1978, was matched with Participant 62, a Lifelong Jasper female born in 1976. Matching participants in this way further increased the likelihood that the production and perception patterns demonstrated by the Lifelong Jasp er participants would be representative of those the Mobile participants would have had before moving to Chicago. Lifelong Chicago participants were not matched for sex and age with Mobile participants; instead, whenever possible, I found Lifelong Chicag oans that the Mobile group actually associated with, i.e. whose phonetics they were actually exposed to. Actual exposure trumped superficial similarity in this case. Recruitment of Participants Mobile participants were recruited by word of mouth and throug h radio announcements. One of my brothers now lives in Chicago (though he did not participate in the experiment), and he was able to put me in touch with numerous potential Mobile group participants. I also knew some from my high school class, and variou s people I talked to in Jasper had a cousin, friend, etc., who had moved to stations, WITZ. The radio station read my request, in which I explained that I was a Jasper na tive working on a PhD project in linguistics and that I needed the help of people raised in Jasper who had moved to Chicago as adults. A fair number of people called me with suggestions after hearing this radio announcement. Lifelong Jasper participants, like Mobile participants, were recruited using word of schools, and at the local community college.
75 Lifelong Chicago participants were mostly recruited with the hel p of Mobile group participants. Each Mobile group participant was asked to suggest at least one Lifelong Chicago participant, though some were unable to do so. Those Lifelong Chicago participants not referred by Mobile group members were mostly friends, family, or colleagues of my two brothers or their wives. 4 Total Participants and Exclusions A total of 90 participants completed the experiment. However, of these, only 76 were retained for data analysis: 26 Lifelong Jasper (13 female), 26 Mobile (13 femal e), and 24 Lifelong Chicago (17 female). Two Lifelong Jasper participants were excluded simply because they were extras; they were ultimately not needed to match Mobile participants. The remaining 12 participants were excluded because I discovered that t hey did not actually qualify for the experiment. For example, one Mobile participant originally told me she was from Jasper but turned out to be from Otwell, Indiana. Similarly, one Lifelong Chicago participant turned out to have lived nine years in Denv er as an adult. Such participants were necessarily excluded from the analysis. Location within Chicago of Mobile and Lifelong Chicago Participants My original intention was to require Lifelong Chicago and Mobile participants to live within the bounds of C hicago proper. However, none of my early Mobile participants knew anyone who had lived their entire lives (minus a few years) within Chicago itself. One Mobile participant informed me that everyone either grows up in the suburbs and then moves to the cit y or grows up in the city and then moves to the suburbs. It later also became apparent that I would be unable to recruit enough Mobile 4 friends and family there.
76 participants if I only accepted participants who lived in Chicago proper. In order to recruit enough participants, I ha d to expand the criteria to include people who lived or had lived in close suburbs of Chicago. (It should be noted that many participants who llows: No farther north than where I 94 and I 294 merge (so Northbrook was a close suburb, but Lincolnshire was not) No farther west than I 355 (so Downers Grove was a close suburb, but Schaumburg was not) No farther south than where State Highway 83 inte rsects with I 57 or I 94 (so Riverdale was a close suburb, but Harvey was not) No farther east than Lake Michigan The community areas/neighborhoods and suburbs in which Mobile participants lived at the time of the interview are listed in Table 3 1. The community areas/neighborhoods and suburbs in which they had lived most of their time in Chicago, if different from where they lived at the time of the interview, are presented in Table 3 2. These tables show that Mobile participants had generally settled on the North Side or in the Central area of the city. None of them lived on the South Side, and only one even lived in a southwestern suburb. (South Siders are generally known for having stronger/more stereotypical Chicago accents. This could be because as McCarthy  notes, South Side neighborhoods have a much lower rate of in migration than North Side neighborhoods 5 .) 5 Of course, my own Mobile participants helped incre ase the in migration rates in the North Side neighborhoods!
77 The community areas/neighborhoods and suburbs in which Lifelong Chicago participants lived at the time of the interview are liste d in Table 3 3. The community areas/neighborhoods and suburbs in which they had lived most of their lives, if different from where they lived at the time of the interview, are presented in Table 3 4. These tables show that while most of the Lifelong Ch icagoans, like the Mobile participants, lived on the North Side, in northern suburbs, or in the Central region of the city, several also lived or had grown up on the South Side or in southern suburbs. Thus, while the Mobile group mostly lived on the North Side, some of them were also exposed to South Side accents from people they regularly interacted with. presented in Appendix A. The Lifelong J 1989, with a mean of 1974. 6 a mean of 1972. The age ranges for the thr ee groups are comparable. I also 7 of slightly ged from one year of college to a 6 Note that for these measures, only descriptive and not inferential statistics are necessary. It would not make any sense to infer that, e.g., the true mean age of all lifelong Chicagoans is gre ater than the mean age of all lifelong Jasperites. While it might be possible that one group overall truly is more educated than another, this is not important for the purposes of my study. Here, I am only interested in comparing the descriptive statisti cs to determine whether the characteristics of my specific participants are comparable. 7 or who had dropped out of college were assigned a score of 12 plus the number of years of college completed.
78 from a high school diploma to a D.D.S. (dentistry degree) (average of half a year less aged about a year more education than either of the two Lifelong groups. Lifelong Jasperites had spent anywhere from 0 to 5.8 years living outside of Jasper, with a mean of 2.93 years. The Mobile group members had spent between 0 and 13 years living in pl aces other than Jasper and Chicago, with a mean of 5.06 years. The Lifelong Chicagoans had lived outside of Chicago for between 0 and 4.5 years, with a mean of 1.93 years. The Mobile group had arrived in Chicago between age 18 and 32, with a mean of 24. 03, and they had spent between 1 and 33.5 years in Chicago, with a mean of 10.49 years, but a median of only 4.25 years. (For other measures presented here, the means and medians were much closer together.) As described below, the Extended Personal Attrib utes Questionnaire provides scores for Positive and Negative Masculinity and Femininity; higher scores indicate and Total Masculinity score ranges were (14 31), (8 17), and (28 47), respectively, with Negative, and Total Masculinity score ranges were (15 31), (4 22), and (24 52), respectively, with respective means of 23.96, 12.58, an d 36.5. The Lifelong 28), (5 18), and (20 44), with respective means of 22.04, 11.33, and 33.4. The Mobile group members tended to be
79 slightly more masculine than the other two groups, but it is diffic ult to tell if the difference is enough to be important. ranges were (17 29), (6 20), and (26 42), respectively, with respective means of 23.12, 11.77, and 34.88. Mobile part ranges were (12 30), (1 17), and (18 47), respectively, with respective means of 24, 29), (4 19), and (21 46), resp ectively, with respective means of 25, 10.67, and 35.58. The femininity ratings for the three groups were comparable. The three participant groups were fairly well matched on the social characteristics of YOB and femininity. The Mobile group members were generally more educated and slightly more masculine than the other two groups. Lifelong Jasper participants had spent more time on average living outside of their hometown than Lifelong Chicagoans had. None of the differences reported here, however, wer e dramatic enough to say that the three groups were patently not comparable with each other in terms of social characteristics. Experimental Design and Procedure Equipment All recordings were made using a Tascam DR 1 portable digital recorder with SP BMC 3 binaural omnidirectional microphones from the Sound Professionals. The recorder was set to a sampling rate of 44.1 kHz and a file type of 24 bit .wav. The microphones have a frequency response of 20 20,000 Hz and a signal to noise ratio of 62 dB. Data was analyzed using Praat and Audacity software.
80 Perception items were presented using E Prime 2.0 software on a Dell Inspiron 600m laptop. Participants listened to the stimuli and instructions through Sennheiser HD 555 Audiophile headphones. A copy of Mil Guess Who? board game was also used. In this two player game, there are twenty four characters with distinctive appearances (facial hair, hats, glasses, etc.). Each player has a board with pictures of the twenty four characters on individua l tiles that can be flipped down. Each player draws a card with one of the characters on it. Players then take turns asking yes or no questions and eliminating person have fac down all the characters who did have facial hair. The first player to figure out which character the other player has is the winner. This game was chosen because many of the featu res that players can ask about (a hat a mustache red vowels of interest and would be stressed when the players asked about them. Materials The Extended Personality Attributes Questionnaire The Extended Personality Attributes Questionna ire (EPAQ; Spence, Helmreich, EPAQ consists of 32 items constituting four scales: Masculine Positive (M+), Masculine Negative (M ), Feminine Positive (F+), and Feminine Neg ative (F ). M+ items measure self confident ). M items measure possession of socially undesirable qualities that ar rogant ). F+ items measure possession of socially desirable traits that indicate a feeling of communion (e.g. helpful
81 to others ). When they developed the EPAQ, Spence et al. intended to find 8 F items that would measure possession of socially undesirable such traits (e.g., servile) and 4 traits indicating verbal passive aggression (e.g., nagging ). Each of the 32 items consists of a scale be tween two extremes (e.g., not at all arrogant to very arrogant ). Between these two extremes is a scale of the letters A through E. The participant circles the letter that indicates where she feels she falls on the scale. The responses are then scored wi th A getting 0 points, B=1, C=2, D=3, and E=4. (Nine items are scored backwards, with A getting 4 points and E getting 0.) The scores of the items on each of the four gender scales are added together, so that the participant receives one total score out of 32 for F+ items, one score out of 32 for F items, and so on. The scores for M+ and M can be combined into a Total Masculinity score (MTotal), and the Femininity scores can be combined likewise (FTotal). As with the Bem Sex Role Inventory described in Chapter 2, the EPAQ can be used in such a way as to assign respondents to one of four gender categories: masculine, feminine, androgynous, or undifferentiated. Androgynous participants are those who score highly on the masculine and feminine scales, whil e undifferentiated participants are those who score highly on neither the masculine nor the feminine scale. However, the authors warn that the dividing lines between these categories can be somewhat misleading (for example, a differing response on a singl e test item could lead to one person being labeled masculine and another undifferentiated), so it is preferable to view the results on a continuum; for example, someone who scores 28 on the M+
83 Although the EPAQ was designed to test socially co nstructed gender, binary and MTota score ranges were (13 28), (4 20), and (20 45), respectively, with respective means of 22.21, 10.33, and 32.53. For males, the score rang es were (12 31), (7 22), and (27 52), with respective means of 23.79, 13.82, and 37.61. Independent samples t tests showed that the difference between males and females for negative masculinity and total masculinity was highly significant (p < .001 for bo th). The average male is more negatively masculine and has a greater total masculinity than the average female. The difference by sex for positive masculinity was marginally significant (p = .092); we can be fairly confident that the true mean positive m asculinity of males is greater than that of females. and FTotal score ranges were (13 30), (5 19), and (21 47), respectively, with respective means of 25.14, 12.05, and 37.19, respectively. For males, these score ranges were (12 29), (1 20), and (18 42), respectively, with respectiv e means of 22.12, 10.76, and 32.88. Independent samples t tests showed that the difference between males and females was highly significant for positive femininity and total femininity (p = .001 for both). In both of these cases, we can be very confident that the true mean femininity score for females is higher than that for males. However, the difference by sex for negative femininity did not approach significance (p = .125). Paired samples t is significantly higher than their total femininity score (p = .013, MTotal mean 37.61, FTotal their total masculinity score (p = .002, FTotal mean 37.19, MTotal mean 3 2.53). Males
84 thus also tend to be more masculine than feminine, and females tend to be more feminine than masculine. In sum, although gender is socially constructed and supposedly separate from sex, it is clear from this data that females tend to be more feminine than males, and males tend to be more masculine than females. Males also tend to be more masculine than feminine, and females tend to be more feminine than masculine. Recorded instructions Two sets of recorded instructions for the EPAQ, the readi ng passage, and the perception test were recorded. One set was read by a 62 year old female who has lived in Jasper since 1979. She was born and raised in Dubois, IN, which is about 8 miles northeast of Jasper and, like Jasper, in Dubois County. She wen t to high school at a boarding school in Ferdinand, IN, which is 15 miles south of Jasper and also in Dubois County. (Her family still lived in Dubois while she was in high school, however, and she returned home for summers and some weekends.) After high school, she attended college in Indianapolis for four years, returning to Dubois for all summers and holidays, then spent three years living in Evansville, IN, approximately one and a half years in Goodland, IN, approximately one year in Berne, IN, just ov er three years in Paducah, Kentucky, three months in Oakland City, IN, and two years in Patoka, IN, before moving in June of 1979 to Jasper, where she has lived ever since. 8 At least to a casual listener, this speaker sounded like a native of Jasper. 8 I t ried to find a reader who would have met the criteria to participate in the experiment as a Lifelong Jasper participant, but I could not find anyone who both had lived all her life in Jasper and was able to read aloud fluently and with expression. The Lif elong Jasper readers I tried simply read word for word without intonation, such that participants who heard them would not have been able to understand the instructions.
85 The second set of instructions was recorded by a 44 year old female who was born in Chicago. At the age of six she moved to Hickory Hills, a close suburb southwest of Chicago, 9 and lived there until age 19. From then until age 25, she split her time between Westmont, a suburb a bit east of (i.e. closer to the city than) Downers Grove, and the Chicago neighborhood of Garfield Ridge. From age 25 to age 33, she lived in LaGrange, a close western suburb, and then moved to LaGrange Park, just north of LaGrange, u ntil age 42. Since turning 42 in 2007 until the time of the recording in early 2009, she lived in Lemont, a slightly more distant southwest suburb between Downers Grove and Orland Park. Since age 18, however, she has always worked in downtown Chicago. T his speaker had a noticeable but not extreme Northern Cities accent. The text of the instructions can be found in Appendix B. Reading passage conversation/interview, there was al ways the possibility that a given participant might not have enough stressed tokens of, say, [ ]. Therefore, I developed a reading passage to use as a backup in case the conversations and interviews did not yield enough tokens of the target vowels. As described in Chapter 5, I ended up not needing to extract any vowels from the reading passage. (However, this data could prove useful in further research.) The reading passage can be found in Appendix C. The reading passage was designed to force the part icipants to produce four tokens of each target vowel in 9 Distances from Chicago are more difficult to estimate because it of course depends w hich part of Chicago you measure from. Jasper has a small enough footprint that distances from the center of town can be easily computed.
86 stressed position, once each before a voiced stop, a voiceless stop, [n], and [l]. (Chapter 5 will show that I ended up not extracting vowels before nasals or [l].) In addition to the 16 target words intentionally placed in stressed position, several other words containing the target vowels also appeared in stressed position in the passage. Perception s timuli Stimuli for the perception test were created by modifying natural speech to create three 10 step continua, one Job~Jab (testing /a/ fronting), one Bad~Bed (testing / / raising), and one Dead~Dud (testing / / backing). The natural tokens were provided by a 62 year old male native speaker of American English. He was born and raised in Evansville IN, which is about 45 miles southwest of Jasper. After completing high school, he lived in West Lafayette, Indiana, for two and a quarter years. He then returned to Evansville for two months, then joined the Navy and spent three and a half years shuttl ing among Navy bases (mostly in Jacksonville, Florida) and overseas in the Western Pacific. He then lived two more months in Evansville, then six months in West Lafayette, and then two months in Evansville again. He then spent approximately one and a hal f years in Goodland, IN, approximately one year in Berne, IN, just over three years in Paducah, Kentucky, three months in Oakland City, IN, and two years in Patoka, IN. From 1979 until the time of the recording in late 2008, he lived in Jasper but worked in Princeton, IN, which is approximately 35 miles west of Jasper. 10 The recordings were made in a quiet room. The speaker stood and read a list of words into the microphone. The list included the six target words interspersed with 10 My original intention was to use a speaker who had never lived in Indiana or the Chicago area. I did make two sets of recordings with such a speaker, but technical problems prevented these recordings Gainesville, FL, for Jasper, so I had to use a speaker who was avail able there. I can only reiterate that this was not my original intention.
87 several distracter item s. Several takes were recorded. On some takes, the speaker said the words individually, with no frame sentence. On other takes, he used the frame row. That is, in some Cape. Cape. Cape cape again. Say cape again. Say cape recordings tended to be clearer and have more consistent pronunciation. The base tokens of the continua were therefore taken from these frameless takes. The continua were created by Christian Koops, then a graduate student at Rice University, using Praat software. He took natural tokens of job, bad, and dud and manipulated their F1, F2, and F3 so that the formants gradually became more like those of natural tokens of jab, bed, and dead The three ten step continua are presented in Objects 3 1 through 3 3. Object 3 1. Continuum of Job~Jab stimuli (.wav file 396 KB) Object 3 2. Continuum of Bad~Bed stimuli ( .wav file 336 KB) Object 3 3. Continuum of D ud~Dead stimuli (.wav file 344 KB) In addition to these test stimuli, natural tokens of cape and keep were used in the practice session of the perception test. Discussion topics The interviews and discussions wer e casual conversations that rambled to whatever topics interested the participants. To help prompt less talkative participants, lists of possible topics were developed. These included such topics as local events, the economic crisis, local and national p olitics, and the like. The lists ended with forth. These demographic questions were asked of all participants. The lists of possible
88 topics and demographic questions can be found in Appendix D. Of course, many topics not found on these lists were also covered in the sociolinguistic interviews. Procedure Data was collected from February to July of 2009. The experiment was run in one of various quiet places. Most of t experiment. On a few occasions, the participants preferred to meet at their place of Participants were run either indivi dually or in pairs. Ideally, I would have liked to have each Mobile participant complete the experiment paired with a Lifelong Chicago friend, relative, or acquaintance. Completing the experiment with a Lifelong Chicagoan was expected to bring out the Mo participants could find a Lifelong Chicagoan to complete the experiment with, so some Mobile participants completed the experiment either alone or (in one case) with another Mobile participant, and som e Lifelong Chicagoans participated alone. Lifelong Jasper participants were generally expected to complete the experiment alone. A list of who participated with whom is provided in Appendix E. When a participant was run alone, she would obviously do the entire experiment straight through. When they were run in pairs, I would begin with just one participant present. (I would tell the other participant in advance that he could arrive half an hour later if he liked.) This first participant would complete the EPAQ, the reading passage, and the perception experiment. I would then ask the first participant to leave the room, and the second participant would complete the same three tasks. Then the two would complete the interview together.
89 At the beginning o f the experiment, the participants read and signed the informed consent form. Following this, they donned the headphones and listened to the instructions for the EPAQ, which they were told was a personality questionnaire. Lifelong Jasper participants alw ays heard the Jasper speaker reading instructions, and Lifelong Chicago and Mobile participants always heard the Chicago speaker. (Each participant heard the same speaker for all the instructions.) Hearing the Lifelong Chicago speaker reading the instruc tions was intended to activate the NCS in Mobile talking with a Lifelong Chicago friend than as they would when talking to a fellow Jasper native like myself. Because this is a study of how far Mobile participants are able to shift their speech in a new dialect region, it was deemed best to (subtly) prompt them to respond in their most shifted style. Before hearing these instructions, participants were told that they could ask me to adjust the volume if need be. After the first set of instructions was complete, participants completed the EPAQ. During this time, I made it a point to wander to another part of the room or at least engross myself in preparing the recording equ ipment. I never looked over their shoulder as they were completing the questionnaire. (I always stayed within earshot in case they had questions 11 of course.) Once they were finished, I immediately placed the completed EPAQ into a folder without looking at any of their responses. This behavior was intended to put them at ease and make them more willing to supply honest answers to the test items listing negative traits 11 The most common question was what servile meant. Enough people requested the definition of this term that I wonder how many of my participants just guessed on this item because they were ashamed to ask. If someone were to prepare a revised version of the EPAQ, I would suggest that they replace this negative femininity item.
90 such as arrogant (Putting the test away without looking like it was supposed to make them feel more at ease about having given honest answers.) Next, the participants again donned the headphones and listened to the instructions for the reading passage. I then pinned both microphones to their clothing, one in the center of their shirt sev eral inches below their mouth, and the other farther away on the shoulder. The participants would say a sentence or two to test the sound. I would listen to the test recording, and once I verified that the sound was acceptable, 12 I would hand the particip ants the reading passage. In accordance with the instructions, participants would first read through the paragraph in their heads and then be recorded reading it aloud. After the participant recorded the reading passage, I unpinned the microphones and a sked them to put on the headphones again. They then listened to the instructions for the perception test (which was called a listening exercise in the instructions). The instructions emphasized that there would be no right or wrong answers, and that they should just respond with their gut instinct. Participants were also instructed to respond as quickly as possible without sacrificing accuracy. The perception test was presented using E Prime software. For each section of the perception test, participants were given two word choices and corresponding keyboard keys on my laptop, always C and M. They saw the word choices and the corresponding keyb oard keys on the screen in front of them. Each time a word was played, they would press the C or M key to indicate which of the two words they heard. 12 I sometimes had to ask participants to turn off a fan, computer, or other noise source before giving the m the reading passage. On a few occasions, the recorder temporarily malfunctioned, and I had to ensure that it was functioning properly again before we could commence with the recording.
91 Participants had 3 seconds to respond before E Prime would play the next stimulus. Once they responded, E Prime would wait 1 second before playing the next stimulus. The perception test began with a set of instructions describing the procedure and telling the participants that they would begin with a practice session. The practice session used natural utte rances (one utterance each) of cape and keep Each of these two utterances was played five times. Participants responded C if they heard cape and M if they heard keep Following the practice session, the participants read another set of instructions tel ling them what their next word choices would be and reminding them that there were no right or wrong answers. The three test continua were presented in separate blocks. There were two different experiment formats, with the block orders counter balanced. One format presented the Job~Jab continuum (responding C for Jab ) first followed by Dead~Dud (C for Dead ), and then Bad~Bed (C for Bad ); the other presented them in the opposite order. (Response keys were the same for each format.) In each block, the p articipants heard a total of thirty stimuli, that is, each of the ten steps on the continuum three times. E Prime first presented all ten steps once in random order; it then reset and played all ten a second time in random order; and likewise a third time After each of the first two blocks of thirty stimuli, the participants were given the option to take a break. (In practice, only one participant took the breaks. Each block was quite brief.) Before each of the blocks, participants read a new set of i nstructions telling them what word choices they would have. The presentation order for the perception test is shown in Figure 3 2. As I did during the EPAQ, I made it a point to wander or look away from the participants as they completed the perception te st, again staying within earshot.
92 Because I could actually hear the stimuli, though muted, coming out through the headphones, I did not want to see which responses they chose and inadvertently react. I also felt that remaining near them might make them f eel more like they were being tested and cause them to worry more about picking the (nonexistent) right answer. At the end of the perception test, I asked participants if they had any thoughts on where the speaker they had heard was from. I expected that listeners would either think the speaker was from (roughly) the Chicago area or Southern Indiana, and that their responses to the perception test would vary based on where they thought the speaker was from (Niedzielski 1999). In fact, many participants re ported not having any thoughts on where the speaker might be from, and to my surprise, several actually thought that the three continua featured three different speakers, possibly from three different regions. Further, to my greater surprise, a few partic ipants thought the speaker was a non native English speaker (e.g., from Sierra Leone or the Middle East), or at least not from the US (Australian). In total, the responses to this question were far too varied for me to run statistical tests on the relatio nship between perceptual barriers and where the listeners thought the speaker was from. It is worth noting that none of the speakers reported realizing that the stimuli were artificially created, and when I later explained to some of them that the words h ad been made using a computer, they expressed surprise. After the perception test, we moved on to the interview/discussion session. Participants who completed the experiment alone had a sociolinguistic interview with me, while those who participated in pai rs had a casual conversation. I would first attach the microphones (one microphone per person, placed in the center of the shirt a few
93 inches below the mouth) and test the sound again. Once I verified that the sound was working, I would commence with the interview or discussion. Almost all interviews and discussions began with a game of Guess Who ? 13 (In two cases, I simply forgot to bring the game along. In one other case, the participant seemed somewhat annoyed with the experiment and impatient to retur n to his work, so I decided to skip the game.) Many participants were already familiar with the rules of the game. When the game was new to a participant, I would explain the rules to them, or, player in the game, I generally tried to ask bad questions to increase the chance that the participant would win. (I figured a tiny ego boost could onl y make them happier to participate in the experiment. Some participants proved to be very competitive and wearing a hat ead, I tried to ask mostly Following the round of Guess Who? conversation turned to whatev er topic the participants were interested in. In the socolinguistic interviews I conducted with participants completing the experiment alone, I of course tried to find topics that they 13 Guess Who? was originally used for the first few participants beca use Dr. Ratree Wayland thought that an information gap activity might provide enough tokens of the target vowels and eliminate the need for a longer interview. This was a good suggestion. Had I created an original information gap activity like those ofte n used in second language acquisition experiments, however, it would have been difficult for me to feign ignorance of the solution in all the interviews I conducted myself. I instead used Guess Who? cipant had. It turned out that this game did not yield enough tokens of the target vowels, but participants seemed to enjoy it, and it served as a good ice breaker, so I continued to use it with later participants.
94 could talk about at length without my interrupting them. When two par ticipants had a casual conversation, I let them chat about whatever they wanted with as little input from me as possible. When there was a lull in the conversation, I would suggest a topic based on the list I had prepared in advance or based on something the participants had mentioned earlier. (Ideally, I would not have participated in the conversations between two participants until I began asking demographic questions at the end. In practice, a few pairs of participants usually the married couples 14 say to either other, so I would have to jump in and ask them questions.) Once I felt that enough time had passed for adequate data to have been collected, I asked each participant the demographic questions. When participants completed the experiment as time when I was asking the demographic questions. This ensured that my questions were clearly audible on the recording. The game plus the inte rview or discussion generally lasted about one hour, though the duration ranged from about 50 minutes to an hour and a half. Finally, Lifelong Jasper and Mobile group participants (and Lifelong Chicagoans who had discussions with Mobile group members) we re asked whether they would like to donate their recording to the Dubois County Museum in Jasper. The donated interviews were intended to serve as a historical snapshot of what life was like for current and former Dubois County residents in 2009. While p resenting them with this consent form, I emphasized that the donation was optional and that it was perfectly fine if they chose not to donate. I stressed that I would not benefit if they donated their 14 Tip for future researchers: married co instead.
95 discussion to the museum and that I could use their da ta either way. 15 I also offered to delete parts of the recording from the donated version and told them that if they did donate their recording, it would be stored only under their participant number(s) and not by name. In total, 17 Lifelong Jasper partic ipants, 16 Mobile participants, and eight Lifelong Chicago participants chose to donate their interviews to the museum. 16 Finally, participants received payment for their time. Most participants took less than two hours and received $20. Those participant s who took over two hours received $30. (Only one participant absolutely refused payment. Quite a few others initially told me to keep my money, but once I assured them that the money was from a government grant and they could essentially consider it a t ax refund, most were eager to take whatever I offered.) This chapter presented the data collection methodology. In Chapter 4, I explain how I analyzed the perception data and present my perception results. 15 I actually also asked participants who would be excluded from the study if they wanted to donate their interviews. Therefore, some of the interviews housed at the Dubois County Mus eum were not analyzed for this study. 16 The public may listen to these donated recordings and/or read the transcriptions by making an appointment with the Dubois County Museum. The museum is located at 2704 Newton Street, Jasper, IN 47546, and the phone n umber is (812) 634 7733. I would recommend asking to speak with Mrs. Mary Ann Hayes, who was instrumental in setting up the interview collection.
96 Figure 3 1. Maps showing Jasper, Indiana, wit hin Dubois County, and Dubois County within the state of Indiana. Maps taken from Wikipedia entry on Jasper, Indiana, on June 17, 2012. Maps were created by Wikipedia editor Arkyan.
97 Table 3 1. Community areas/neighborhoods and suburbs in which Mobile participants lived at the time of the interview Community area (Neighborhood in parentheses) or s uburb Participants Far North Side Uptown 9 Lincoln Square 22 Edgewater (Edgewater) 30 North Side Lakeview 12, 18, 36 Lakeview (Boystown) 6 Lincoln Park 8, 24, 28, 71 Logan Square (Bucktown) 26 North Center (Northcenter) 33 Central Near North Side (Gold Coast) 16, 19 Near North Side (River North) 37, 38 The Loop 32, 70 Northern s uburbs Winnetka 76 Wilmette 78 Western s uburbs Downers Grove 10, 39 Brookfield 46 Elmhurst 73 Southwestern s uburbs Orland Park 42 If no neighborhood is listed in parentheses after the community area, the participant did not specify the neighborhood, or the community area is not divided into smaller neighborhoods.
98 Table 3 2. Community areas/neighborhoods and suburbs in which Mobile participants had lived most (the plurality) of their time in Chicago, if different f rom where they lived at the time of the interview Community area (Neighborhood) or s uburb Participants Far North Side Edgewater (Andersonville) 30 North Side Lakeview 33, 71 Lakeview (Wrigleyville) 32 (tied with Lincoln Park) Lincoln Park 32 (tied with Wrigleyville) Central Near North Side (Gold Coast) 8, 38 Northern s uburbs Kennilworth 76 Northwest ern s uburbs Schaumburg 46
99 Table 3 3. Community areas/neighborhoods and suburbs in which Lifelong Chicago participants lived at the time of the interview Community a rea (Neighbo rhood) or s uburb Participants Far North Side Norwood Park (Oriole Park) 83 Albany Park (Mayfair) 84 Northwest Side Portage Park (Six Corners) 11 North Side Lakeview (Wrigleyville) 15, 34, 35 Lincoln Park 79 Logan Square (Bucktown) 27 North Center (Northcenter) 23 Avondale 80 Central Near North Side (Gold Coast) 17 Near North Side (River North) 41 Near North Side (Streeterville) 44 South Side Bridgeport 20 Far Southwest Side Mount Greenwood 47, 48 Northern Suburbs Winnetka 77 Lake Villa 7 (very short time Boystown immediately prior) Evanston 29 Northwestern s uburbs Arlington Heights 81 Park Ridge 85 Western s uburbs Brookfield 14 Southwestern s uburbs Oak Forest 31 Tinley Park 43
100 Table 3 4. Community areas/neighborhoods and suburbs in which Lifelong Chicago participants had lived most (the plurality) of their lives, if different from where they lived at the time of the interview Community area (Neighborhood) or s uburb Participant s North Side Lakeview 80 Lincoln Park 31 South Side Bridgeport 14 Southwest Side Chicago Lawn (Marquette Park) 11 McKinley Park 44 New City (Back of the Yards) 47 (unclear if more time Ashburn or Back of the Yards) New City (Canaryville) 48 Far Southwest Side Ashburn 47 (unclear if more time Ashburn or Back of the Yards) Northern suburbs Kennilworth 77 Vernon Hills 7 Glenview 35 Northbrook 41 Wheeling 79 Northwestern suburbs Mount Prospect 81, 84 Western suburbs Naperville 17 17 River Forest 27 Clarendon Hills 83 Southwestern suburbs Crestwood 34 Country Club Hills 43 17 Naperville is too far away from Chicago to meet my criteria to be a close suburb. There were two early Lif elong Chicago participants who lived a bit farther from Chicago than I later came to allow. Although I participated with a Mobile participant, and I felt that it was most important to include people whose accents the Mobile participants were actually exposed to.
101 Welcome screen Instructions for practice block Practice block Instructions for Block 1 Block 1 ( Job~Jab for one order, Bad~Bed for the other) Optional Break Instructions for Block 2 Block 2 ( Dead~Dud ) Optional Break Instructions for Block 3 Block 3 ( Bad~Bed or Job~Jab ) Thank you screen Figure 3 2. Presentation order for perception test
102 CHAPTER 4 PERCEPTION Ultimately, this study compares acquisition of second dialect in perception and production. This chapter reports the results of the perception experiment alone, that is, o, Job~Jab Bad~Bed and Dead~Dud continua was measured, and statistical tests were conducted to see if the three residence groups perceived each continuum differently, and if perception was further affected by other social variables. I first present the findings for variation in reactio n times, followed by the comparison of perceptual boundaries for the three continua. Reaction Times stimuli. The E Prime program used to present the perception experiment me asured the number of milliseconds that passed between the time when the audio file was initiated and the time when the participant pressed a response key (either C or M, to indicate the word they thought they heard). Mobile group members, unlike Lifelong Jasper and Lifelong Chicago participants, had had intensive exposure to two different dialects. I expected that this might make the perception task more difficult for them because, particularly for the stimuli in the middle of the continua, they would hav e to decide whether a stimulus was, e.g., similar to a Jasper like speaker saying dud or a Chicago like speaker saying dead That is, they would have to process the influence of two competing sets of exemplars (Nosofsky 1986) before deciding how to respon d. The two Lifelong groups should not have had this conflict. Therefore, I expected that the Mobile
103 group would respond more slowly than the Lifelong Jasper and Lifelong Chicago groups. The results below present findings related to this hypothesis. Whe n interpreting the following results, one should note that reaction times were only measured for those stimuli to which the participants responded. If respondents hesitated 3 seconds (3000 ms) before responding, no answer or reaction time was logged, and the next stimulus was played. However, out of the 6,840 stimuli heard by the 76 participants, only 14 (0.2%) did not receive a response. I therefore find it unlikely that the missing reaction times (however long they would have been) for these few stimuli would have an effect on the overall findings. The number of stimuli to which there was no response is reported by continuum and residence group in Table 4 1. Effects of Residence Group and Related Factors on Reaction Time Because the Mobile participants had to contend with the influence of two competing dialects when deciding which words they had heard, it was hypothesized that the Mobile group might have longer reaction times on the perception test than the Lifelong Jasper or Lifelong Chicago groups did. Statistical tests did not support this hypothesis, however. One way ANOVAs with residence group as the independent variable revealed no significant differences in mean reaction times for the Bad~Bed continuum (df Between Groups = 2, Within Groups = 73, F = 1.106, p = .336), the Dead~Dud continuum (F = 1.029, p = .362), or the Job~Jab 1 continuum (F = 1.191, p = .310). 1 A one way ANOVA revealed a main effect of Version for the Job~Jab continuum. For this continuum, mean reaction time was about 71 ms slower for Version 1 than for Version 2 (1095 ms vs. 1024 ms). An equal number of Mobile and Lifelong Jasper participants heard each version, so their results are not affected by this finding. However, through my own error, 11 Lifelong Chicago participants completed Version 1, and 13 completed Version 2. Therefore, the results for group could be slightly skewed for this continuum. (Version did not significantly affect either of the other continua; Bad~Bed p = .918, Dead~Dud p = .175.)
104 Nor was there a significant correlation between the number of years the Mobile group members had spent in Chicago and mean reaction times. If Mobile group members adjusted more fully to the Northern Cities Shift (NCS) the longer they lived in Chicago (i.e. became more monodialectal), we would expect a negative correlation between years spent in Chicago and reaction time. If, on the other ha nd, they retained their Jasper influences while adding in the influence of the Chicago dialect, this increased confusion should be reflected in a positive correlation between length of Chicago residence and reaction time. 2 In fact, for all three continua, the correlations were very weakly negative, but none of these correlations approached significance ( Bad~Bed r = .103, p = .615; Dead~Dud r = .129, p = .529; Job~Jab r = .053, p = .796). Nor were these correlations significant once age of arrival in Ch icago was controlled for (partial correlation Bad~Bed r = .111, p = .597; Dead~Dud r = .134, p = .524; Job~Jab r = .069, p = .744). Following Nycz (2011), I also tested the correlation between reaction time and the percentage of a Mobile group partici example, Participant 6 was 27 years old at the time of the interview and had lived 3.75 years in Chicago, meaning she had spent approximately 13.9% of her life living in Chicago.) As with the number of yea rs spent in Chicago, one could expect that having thus reduce reaction times), or make someone more confused (and increase reaction times). However, percent of life spent in Chicago (PLC) was not significantly correlated 2 We might also expect reaction time to initially increase as Mobile members first became more exposed to the NCS, but later decrease once they became more firmly monodialectal again. A graph of the data did not suggest such a relationship, however.
105 with mean reaction times for any of the three continua ( Bad~Bed r = .072, p = .728; Dead~Dud r = .141, p = .492; Job~Jab r = .053, p = .799). In summary, the expected effect of residence group on reaction time was not found. 3 Despite the (assumed) conflicting influence of two dialects, the Mobile group time affected by the number of years they had lived in Chi cago or the percent of their life they had spent there. This could suggest that having had intensive input from two different dialects does not make perception more difficult. Perhaps once a listener who has lived in two dialect regions (thinks she) know s where a speaker is from, she can still easily determine what words the speaker is saying. Effects of Sex and Gender on Reaction Time The results when one contrasts sex and gender are more interesting, however. Note that no predictions were originally ma de regarding differences in reaction times by sex or by gender. Two way ANOVAs using residence group, sex, and their interaction as independent variables again revealed no main effect of residence group and also revealed no main effect of sex (df = 1, Bad ~Bed p = .516, Dead~Dud p = .268, Job~Jab p = .974). Nor was there significant interaction between residence Group and sex (df = 2, Bad~Bed p = .776, Dead~Dud p = .699, Job~Jab p = .773). However, there were significant correlations between gender and rea ction time. 3 I al so considered the possibility that there was a difference by group only for the stimuli that were more difficult to classify, but that the rapid responses to stimuli near the ends of the continuum were masking these effects. I therefore ran additional test s based on the response times only for those stimuli for which at least 6 participants had inconsistent responses (e.g., two Bad and one Bed response). This included steps 3 through 10 for Bad~Bed steps 3 through 5 for Dead~Dud and steps 5 through 7 for Job~Jab However, the results for this restricted set of stimuli were still nonsignificant by residence group.
106 (FTotal) score on the gender test and their reaction times for the Dead~Dud continuum (r = .296, p = .009) and the Job~Jab continuum (r = .258, p = .024). (The correlation between FTotal and reaction time for the Bad~Bed continuum was positive but nonsignificant, r = .128, p = .269.) There were also a significant negative correlation MTotal) scores and their reaction times for the Dead~Dud continuum (r = .240, p = .037) and marginally significant negative correlations between MTotal and reaction times for the Bad~Bed continuum (r = .197, p = .089) and the Job~Jab continuum (r = .212 p = .066). However, the marginal significance of the correlation between MTotal and reaction time to the Bad~Bed continuum disappeared once a single outlier was removed (new p = .209). 4 Overall, there is a tendency for reaction time to increase as a pa To determine whether socially desirable or social undesirable traits were more closely correlated with reaction times, correlations were run between reac tion time and the individual positive femininity (F+), negative femininity (F ), positive masculinity (M+), and negative masculinity (M ) scores for those continua for which the correlation with the FTotal or MTotal scores was at least marginally significa nt. For the femininity correlations, only positive femininity proved to be significantly correlated with reaction times ( Dead~Dud F+ r = .273, p = .017; F p = .138; Job~Jab F+ r = .260, p = .023; F p = .281). In contrast, for the Masculine correlations it was only negative masculinity that correlated at least marginally significantly with reaction times ( Dead~Dud M r = .236, p 4 time for the second slowest participant. Removing this outlier did not affect the significance of other tests reported earlier. For example, the main effect of Residence Group on reaction time for Bad~Bed was still nonsignificant (p = .363).
107 = .04; M+ p = .235; Job~Jab M r = .219, p = .058; M+ p = .339). We can conclude that there are slight tendencies for react ion times to increase as socially desirable feminine traits increase, and for reaction time to decrease as socially undesirable masculine traits increase. Why might this be? Let us first consider the results for femininity scores. Recall that F+ items on communion (e.g., helpfulness), while half of the F items measure unmitigated communion (e.g., spinelessness) and half measure passive aggression (e.g., tendency to nag). If anything, then, we might expect a positive correlation between the F scores and reaction times; a spineless person might be expected to hesitate more before making a selection. Yet this is not what was found. It was F+ that correlated positively with reaction times. It is not immedi ately apparent why a person who is kind and aware of the feelings of others would take longer to react on the perception test. One could long enough to consider how other s might respond to a given stimulus, or to try to figure out how the experimenter wanted them to respond. That seems rather like reading too much into the correlation, however. The correlation between positive femininity and reaction times is difficult t o explain. Turning now to the negative correlations between negative masculinity and towards agency (e.g., making decisions easily), while M items measure unmitigated agency ( e.g., greed). It is somewhat intuitive that a person more able to take action should be able to respond more quickly in a decision making task. We would
108 particularly expect a negative correlation between reaction time and positive masculinity, considerin g that one of the M+ test items is ability to make decisions easily! However, it is M that correlates negatively with reaction time. It also makes sense that an overly competitive, boastful person might feel compelled to respond to stimuli as quickly as possible, perhaps to impress the experimenter with how quickly they could explain than its correlation with positive femininity. These findings should demonstrate t o linguists how separating sex from socially constructed gender can unveil interesting patterns. Binary biological sex had no were significantly correlated with t heir reaction times. Differences in Perception While differences in reaction times would have been interesting, the real focus of the experiment was, of course, the differences in perception by residence group. This section presents the results of the i perception. (As with the reaction times, the .2% of stimuli to which participants did not respond are not expected to have meaningfully skewed the perceptual boundary results.) Method of Statistical A nalysis Before the results are presented, the method of analysis needs to be explained. As is typical in phonetic perception experiments of this type, I compared the average 50% crossover point in perception for each group. The 50% crossover point is the spot on the continuum at which the individual switched from hearing a majority of one word to hearing a majority of the other, that is, the point at which the probability that they
109 would hear each word was .5. For example, if a participant listening to t he Job~Jab continuum heard Job twice and Jab once for stimulus 5, and Job once and Jab twice for continuum. An example using hypothetical responses is shown in Figure 4 1. The method of precisely determining the 50% crossover points is presented below. Measurement of formant v alues Naturally, the actual crossover point must be based on acoustic measurements boundaries lay, I first had to find the acoustic qualities of the perception stimuli. S everal algorithm used to create the stimuli was designed to give each vowel a steady state [Christian Koops p.c.].) Using Praat software, I found the average of the F1 and F2 over the steady state of each vowel. This steady state was generally the second half of the r each continuum, the duration of the measured steady state was kept constant across all stimuli. These values were used for the perception statistics. Measurements were originally made in Hertz but then were converted to Barks (Zwicker 1961) using the f ormula presented in Traunmller (1990). Barks are scaled to how the human ear perceives changes in pitch. For example, a change from 2 to 3 Barks (a difference of roughly 93 Hz) creates the same difference in perceived pitch as a change from 7 to 8 Barks (roughly 149 Hz). Because Barks reflect human perception better than Hertz, it is more appropriate to test differences in perception using the Barks measure. The
110 statistical results reported here are based on measurements in Barks. The F1 and F2 values of the perception stimuli are presented in Table 4 2. Method of d etermining 50% crossover point using difference b etween F2 and F1 as a predictor One acoustic cue to vowel quality is the distance between the first and second formants (Fahey, Diehl, and Tr aunmller 1996). Therefore, differences in the three predictor of their responses. For example, the first step on the Job~Jab continuum had an F1 of 7.49 Barks and an F2 of 10.76 Barks. The distance between F1 and F2 was thus 10.76 7.49 = 3.27 Barks. The distances calculated this way were used as the single predictor variable for logistic regressions run for each participant for each continuum. Based on the logistic reg crossover point, or perceptual boundary, for the distance between F1 and F2 (hereafter F2F1d). These crossover points were then used for the multiple regressions (described below) that revealed what fact boundaries. Method of d et ermining 50% crossover p oints using F1 and F2 as predictors In addition to F2F1d, I used the actual F1 and F2 values (in Barks) as predictor variables for the continuum respons es. Calculating the 50% crossover points for these two predictors was significantly more complicated than calculating the crossover points for the single F2F1d predictor. One approach, to keep the math simple, would have been to simply run one set of log istic regressions using only F1 as a predictor and another set using only F2. However, the resultant F1 and F2 crossover points would not have been related to each other. I would have found the 50% crossover point for F1
111 assuming no influence of F2 on th e responses, and vice versa. It was necessary to use both predictors to create a single logistic regression curve for each participant for each continuum. This ensured that the resultant F1 and F2 values would be mathematically related to each other. Th e influence of each formant would be calculated taking into account the influence of the other. However, when two predictor variables are used in the logistic regression equation, finding the 50% crossover point becomes more difficult. When using a single predictor, the probability is graphed on the y axis, the predictor value is on the x axis, and the resultant curve is two dimensional. The curve intersects the .5 probability line alues are graphed on the z axis, and the resulting curve is three dimensional: a curved plane rather than a curved line. This curve intersects the .5 probability plane at an infinite number of points. This is illustrated in Figure 4 2, which shows one pa regression curve indicating the probability that this listener would hear Bed for a given combination of F1 and F2 values. When using two predictors as shown in Figure 4 2, there is an infinite number of combinations of F1 and F2 values that describe vowels which the participant is predicted to hear as each target word 50% of the time. In Figure 4 2, this infinite set of values is represented by the line at which the graph changes colors, i.e. the line at which the curve crosses t he y = .5 plane. Some of these combinations are clearly nonsensical; for example, a vowel with an F1 of 5 Hz and an F2 of 2500 Hz 5 is not possible, even if mathematically it seems that such a vowel would yield a 50/50 5 As stated before, all my calculations were actually based on Barks, not Hertz. I prefer to give hypothetical examples using Hertz simply because they are more familiar to most linguists.
112 response rate for a certain participa nt. A combination in which one or both of the formant values are negative is likewise impossible. Yet there are still an infinite number of plausible combinations of F1 and F2 values that lie on the .5 probability line. The task is to determine which of these combinations it is most appropriate to select as the The optimal formant combination was that which intersected the line of formant combinations that the participants actually heard in the stimuli. Figure 4 3 show s a two dimensional graph with F1 values on the x axis and F2 values (in Barks) on the y axis. On this graph, the solid curve indicates the possible F1 and F2 values that would supposedly yield a .5 response probability for the participant (i.e., equivale nt to the curve where the curved plane changed colors at the .5 probability line in Figure 4 2). The dotted line indicates the actual F1 and F2 values of the stimuli that the participant heard for the continuum ( Dead~ Dud in this figure). The intersection of the curve and crossover point for each continuum. This method worked well for all 76 participants for the Dead~Dud continuum. It also worked for 75 out of 76 participan ts for Job~Jab and 72 out of 76 for Bad~Bed. On the Job~Jab intersected the line of values actually heard at two points rather than one. Because these two points were extremely close toge ther, however, I simply took the average of the two points. The Bad~Bed continuum also had one participant, Lifelong Jasper Participant 61, whose curve intersected the line of stimuli formant values in two places. In this case, one intersection seemed mo re valid than the other. That is, the regression
113 curve did not fit the data particularly well, so the data were visually inspected, and one of the two intersections seemed much closer to where an ideal logistic regression curve would show the 50% crossover point. This was the intersection chosen for this participant. There were also three participants, Participants 2 (Lifelong Jasper), 28 (Mobile), and 56 (Lifelong Jasper), whose Bad~Bed curves never intersected the line of stimuli formant values. These statistical calculations of F2 and F1 crossovers for the Bad~Bed continuum. (They were included, however, in the statistics using the difference between F2 and F1 as a predictor.) One outlier was als o excluded from the Job~Jab continuum. Lifelong Chicago data was eliminated from furt her analyses. Unfortunately, there is no obvious explanation of why this person would be an outlier on this continuum. I did not note any procedural irregularities when she participated, nor is there anything strikingly different about her social charact eristics. She does happen to be the oldest Lifelong Chicago participant, but only by one year. In production, she has a noticeable NCS, but many 6 Mean crossover points for the three residence groups a nd three perception measures for the Job~Jab Bad~Bed and Dead~Dud continua are given in Tables 4 3 through 4 5. 6 Participant 23 was never close to being the most extreme Lifelong Chicago participant in terms of production of / / ba cking or / / raising. Nor was she close to the most extreme for / / backing though she did have the second lowest / / among Lifelong Chicago speakers. Of course, it is in /a/ fronting that her perception is exceptional, but I did not end up measuring /a / fronting in production, so I cannot comment on whether her /a/ fronting production is unusual.
114 Statistical Comparisons of 50% Crossover Points The factors I used to predict the variation in perception were as follows: residence group, ye ar of birth (YOB), sex, total masculinity, total femininity, education, age of arrival in Chicago (AAC; Mobile group only), number of years lived in Chicago (YLC; Mobile group only), and percent of life spent in Chicago (PLC, Mobile group only). To begin, I ran Pearson correlations between the scalar variables and crossover points and ANOVAs using the nominal variables to compare crossover points. These simple statistical tests were run first not controlling for any other variables (that is, one predictor only) and then controlling for residence group only. 7 (These tests were run on all three crossover values for all three perception continua. For example, I found the correlation between YOB and the crossover points for F1, F2, and F2F1d for Job~Jab, Dea d~Dud and Bad~Bed .) However, it is more informative to see what factors remain significant when controlling for the others, so I then ran multiple regressions with all the predictor variables together. In the preliminary ANOVAs and correlations, four f actors were never significant AAC. 8 (It is not surprising that AAC was never significant simply because there was very little variability in this factor. 18 of the 26 Mobile participants arrived in Chicago rarely significant, and never once other factors were controlled for. PLC was also rarely 7 Except when residence group was the predictor, of course. 8 Notice that throughout this dissertation, I report marginally significant (.05
115 marginally significant, and these effects likewise disappeared once other variables were included in the regression model. To keep from over modeling my data, I finally ran multiple regressions including only the factors that had sometimes been significant in the previous tests, including previou s multiple regressions: residence group, YOB, and sex. It is these results that are presented below. Results of the Job~Jab c ontinuum The results of the multiple regressions for the Job~Jab continuum are presented in Tables 4 6 through 4 8. The tables sh ow that for the Job~Jab continuum, residence group (tested through dummy variables Jasper and Chicago ) was never a significant (or found no evidence in regional var iation in perception of NCS /a/ fronting. This is discussed further below. Nor was YOB ever a significant factor affecting perception of this continuum. For all three measures of perceptual boundaries, however, sex was a significant or marginally signifi cant predictor of participant responses. Females heard more Job than males did; that is, their perceptual boundaries were closer to the Jab end of the continuum, i.e., their perception of /a/ was more fronted. A 95% confidence interval for the true diffe rence between females and males in the F2F1d Job~Jab crossover point was (.028, .239). The true mean difference in the crossover point for males and females could be as little as .028 or as much as .239 difference in Bark distance between F1 and F2. At t he Job end of the continuum, the difference between F1 and F2 was 3.27 Barks. At the Jab end, the difference was 6.72 Barks, so the range of difference across the continuum was 6.72 3.27 = 3.45 Barks. The true difference
116 between males and females for p erception of this continuum could thus be as little as .028/3.45 = .81% of the distance across the continuum, or as much as .239/3.45 = 6.93% of the distance. It is thus possible that sex actually accounts for a practically significant difference in perce ption of /a/ fronting. On the other hand, it is also plausible that sex could account for only a tiny, practically nonsignificant difference in the perception of this variable. To see whether the effects of sex varied by residence group, I ran further mul tiple regressions including interaction terms for sex and residence group. The results of these regressions are given in Tables 4 9 through 4 11. In these regressions, the interaction between sex and the Lifelong Jasper group is nonsignificant, but the in teraction between sex and the Lifelong Chicago group is either significant or marginally significant. However, the change in effect is not consistent. For the F1 and F2 crossover points, the sex effect is reversed for Lifelong Chicago listeners: while females overall hear more Job than males do, Lifelong Chicago females hear more Jab ess /a/ fronting in perception. However, for the F2F1d scale, the effect of sex is strengthened for Lifelong Chicagoans: Lifelong Chicago females hear more Job (have more /a/ fronting in perception) than Lifelong Chicago males even moreso than females in general hear more Job than males. Which of these findings is to be trusted? It is worth noting that the interaction of Chicago by sex for the F1 and F2 scales is only marginally significant, whereas the for the F2F1d scale it is statistically significant. While the p = .05 cutoff point standard in
117 the social sciences is of course arbitrary, we can say that we are more confident in the interaction effect for the F2F1d scale than for the other two scales. It is therefore more likely that the differences be tween females and males are in the same direction but strengthened for Lifelong Chicago listeners as compared to the other two groups. This is an interesting finding in light of traditional reports of the production of males and females in sound changes. Labov (1990) states that in cases of language change, females produce more nonstandard (incoming) forms than males, but once a sociolinguistic variable becomes stable, males use the nonstandard form more frequently than females. According to Labov, once a change becomes advanced, sex differentiation is greatest in the upper middle class (UMC), with UMC females retreating from the nonstandard form. NCS /a/ fronting is a well advanced shift; it has been documented for at least 40 years (Labov, Yaeger, and S teiner 1972). (Note that Pederson 1965 found no evidence of what is now called the NCS in Chicago, so it is probably safe to say that the shift is fewer than 50 years old in Chicago.) Many of my participants would probably be classified in the upper clas s, UMC, or middle middle to exhibit less of the nonstandard form, at least in production. In fact, Lifelong Chicago does. (Their /a/ is more fronted in perception.) Indeed, recall that McCarthy (2011) found that /a/ fronting is actually retreating in Chicago. In that case, /a/ backing, so to speak, is an incoming change in Chicago, an d we would likewise expect females to be leading in this change and thus have perception with a more backed /a/. As stated above, this is not the case.
118 There are several possible explanations for this apparent paradox. One possibility is that perception does not align with production as closely as we would assume. (See perception is more nonstandard. Alternatively, the results may be related to the fact that the speaker of the perception stimuli was a male. Perhaps at least some of the females expected a male to exhibit a strong NCS and responded accordingly, while the males may no t have had such firm expectations of how another male would speak. Given that previous generalizations about the effects of sex on language variation have been drawn primarily from phonetic production, any ideas about the effects of perception are current ly purely speculation. Further studies should investigate the interaction of sex with phonetic perception as well as production. I return now to the question of why no regional differences in perception of /a/ fronting were found in this study. Multiple regression revealed no effect of residence group on perception of the Job~Jab continuum. Why not? There are several possible explanations. First, most of my participants were highly educated. Sociolinguistic differences tend to be muted in groups with higher education; for example, sex differences in language are more pronounced in less educated groups than highly educated ones (Hans Van de Velde p.c.). Therefore, it is possible that differences in perception of this vowel shift would have emerged had my participants not been educated beyond high school, but that regional differences in perception of these vowels do not exist among
119 college e ducated speakers. 9 If this were the case, however, we would not expect to have found perceptual differences for the other two continua, either. However, I now suspect that the reason I did not find regional variation in perception of this vowel is that t (2011) found that (in production), Chicagoans front /a/ less dramatically than speakers from other Northern Cities, and further that /a/ fronting is retreating in Chicago. Thus, Chicago /a/ may sim ply not be that much farther forward than Jasper /a/. I may have found no regional difference because there is no regional difference. Alternatively, there may be a very small amount of regional variation in perception (and production) of this vowel sm all enough that the high education level of my speakers could prevent this small difference from appearing in my data. Results of the Bad~Bed c ontinuum Results of the multiple regressions for the perceptual boundaries of the Bad~Bed continuum are presented in Tables 4 12 through 4 14. Residence group is a significant predictor of the perceptual boundaries for F1 and is marginally significant for F2 and the difference between F2 and F1. For all three ignificantly or marginally Mobile and Lifelong Jasper groups does not approach significance. For this measure, the mean crossover points were nearly the same for th e Lifelong Jasper (5.289 Barks) and Mobile (5.298) groups but closer to the Bed end of the 9 Of course, Education level was never a significant predictor of perceptual boundaries, but this could be because so many of the p articipants were clustered together with a high level of education. Had I recruited an equal number of college educated and non college educated participants, a significant influence of Education on responses may have been found.
120 continuum (5.15) for the Lifelong Chicago group. There is evidence that Chicagoans, as predicted, hear more Bad than either of the other two groups. There is also evidence that Mobile group members would not shift their perception to be Chicago like; it is to that of Lifelong Jasperites, but it is much less plausible that it would be identical to that of Lifelong Chicagoans. This finding can perhaps be taken as evidence that a) regional variation in the perception of // raising exists, and b) upon exposure to a new dialect in adulthood, a shift in perception is not to be e xpected. (The most restricted interpretation would be to say that a shift in perception will not necessarily occur, a finding which should be unsurprising.) Sex is never a significant predictor of the perceptual crossover point for the Bad~ Bed continuum YOB, on the other hand, is significant for the F2 crossover point and marginally significant for F1. (It does not approach significance, however, for the difference between F2 and F1.) As participants got younger, they heard more Bed To see if this was true of all residence groups equally, I tested the interaction between residence group and YOB. 10 Tables 4 15 through 4 17 show multiple regressions that test the interactions between residence group and YOB for this continuum. The interaction between residence group and YOB was never significant. We have no evidence that the effect of YOB varies a ccording to where the listener has lived (or which dialect he has been exposed to). This finding indicates a general trend of 10 For these regressions o nly, I recentered the YOB data to have a mean of zero (by subtracting the mean YOB). Unlike the others, these regressions would not work correctly in PASW until I recentered the data.
121 decreased / / raising in perception across the Midwest. Further, the standardized coefficients in Tables 4 12 through 4 14 show that this age effect is comparatively strong, sometimes even stronger than the effect of being a Lifelong Chicagoan. Results of the Dead~Dud c ontinuum As with the Job~Jab continuum, the Dead~Dud continuum shows no evidence of regional variation in percepti on based on the multiple regressions. 11 Tables 4 18 through 4 20 show that none of the predictors included in the model residence group, YOB, or sex ever approached significance. (Even if the least significant factor, sex, is removed from the model, t he other predictors still remain nonsignificant.) However, Table 4 for the Dead~Dud continuum were nearly halfway between the means for the Lifelong Jasper and Lifelong Chicago groups. Among my p perceptual boundaries were, as predicted, between those of the Jasper and Chicago significantly differed from that of either the Jasper or Chicago groups. For all three perceptual boundary measures, the difference between Lifelong Jasper and Mobile perception and the difference between Lifelong Chicago and Mobile were nonsignificant. Thus, it is statistically plausible that the average person exposed to a new dialect in adulthood completely shifts his perception to be like that of the residents of the new dialect region, but it is also 11 In fact, there is an important difference between these two continua. Even in a one way ANOVA, residence group was never a significant predictor of perception of the Job~Jab continuum. In contrast, before other factors were controlled for, residence group was a marginally significant predictor of perception of the Dead~Du d continuum (df Between Groups = 2, Within Groups = 73; F1 F = 3.046, p = .054; F2 F = 3.046, p = .054; F2F1d F = 2.783, p = .068). Post hoc Tukey tests consistently showed significant differences in perception between the Lifelong Jasper and Lifelong Chi cago groups, but not between the Mobile group and either other group. Thus, it appears that regional differences in perception may really exist for / / backing, whereas there is no such evidence for /a/ fronting.
122 statistically plausible that the average mobile person does not shift his perception at all, instead maintaining his native dialect style perception. This finding is not informative enough to allow me to draw conclusions about the malleability of phonetic perception in adulthood.. I therefore explored individual Mobile Perception of i n dividual Mobile p articipants Dead~Dud continuum and their social characteristics is presented in Appendix F. Individual Mobile res were compared to the mean responses for the Lifelong Jasper and Lifelong Chicago groups. Ten Mobile Chicagoans or indicative of more extreme / / backing than the Chicago mean. The Lifelong Jasper participants or indicative of even less / / backing than the Jasper mean. This seems to indicate that there are ten Mobile participants who have sh ifted their perception since exposure to a new dialect in adulthood, while the remaining 16 have maintained their native perception patterns in the face of exposure to a new dialect. The caveat is that some individual Lifelong Jasper participants had perc eptual boundaries close to the average for Lifelong Chicagoans, and vice versa. For example, Lifelong Jasper Participant 5 had a perceptual boundary for F2F1d of 6.26, which is much farther back in the mouth than the Lifelong Chicago mean of 6.45. Thus, it is not certain that all ten of the apparently shifted Mobile group members have actually changed their perception in adulthood. Some of them may already have had Chicago like perception when they lived in Jasper. Still, it seems unlikely that well ove r a third of
123 the Mobile participants would have happened to have native perceptual boundaries that were much farther back than the mean for Lifelong Jasperites. It is probable that at least some, and perhaps all, of these participants have truly shifted t heir perception in adulthood. Longitudinal data would be needed to confirm that the participants have actually shifted their perception. The social characteristics of the shifted and unshifted groups of Mobile participants were compared in an attempt to f ind patterns that might help predict whether an individual would shift her perception upon adult exposure to a new dialect. Unfortunately, no strong patterns are apparent. The mean education level of the two groups is practically identical, with each gro up averaging a little over half a year beyond amount of time living somewhere other than Jasper or Chicago (mean of the shifted group 5.18 years, unshifted 5.29 years). It would seem reasonable that those with a more positive attitude towards Chicago would be more likely to have shifted, but in fact, almost all of my Mobile participants loved Chicago and said they intended to stay there for some time. Only Participant 46 said he back. (His perception was, in fact, shifted, but one person does not constitute a pattern. Other participants who said that they currently preferred Chicago but could imagine moving back to Jasper eventually were split on whether they had shifted their perception.) One factor on which the shifted and unshifted groups do vary is the number of Chicago for a mean of 13.1 and medi an of 5.5 years at the time of the study, while the
124 unshifted ones had only lived there for a mean of 8.9 and median of 4.25 years. One could thus say that the longer someone has been in contact with her new dialect, the more likely she is to have shifted her perception accordingly. This would be logical, but there are two pieces of contradictory evidence to consider. First, recall that the number of years the Mobile Group members had lived in Chicago (YLC) was never a significant (or even marginally sig nificant) predictor of perceptual boundaries. Moreover, there are many individual exceptions to the trend. Most strikingly, Participant 73 is the Mobile member who has spent the most time in Chicago, 33.5 years, yet he still has Jasper like perception. On the other hand, Participant 18 has only lived in Chicago for 1.5 years but already has Chicago like perception. While duration of stay in the new dialect community may play a role in determining who will have shifted perception after adult exposure to a new dialect, it is clearly far from definitive. Chicago (PLC) initially seems to be different for the two groups. Those with consistently Chicago like perception had spe nt a mean of about 28.4% of their lives in Chicago, while the mean for those who were consistently Jasper like was only 21.6%/. However, the medians for the groups are almost identical 16.4% and 16.1%, respectively and there are again many individual exceptions to the pattern. Another possible determinant that I investigated was the area of Chicago to which people had moved. It seemed plausible that, because the Chicago dialect is not uniform across all neighborhoods, Mobile participants who had moved to some neighborhoods might have been more likely to have shifted their perception than those who had moved to others. In fact, though, as shown in chapter 3, the majority of my
125 Mobile participants had moved to the same few neighborhoods. Many moved to Lakeview, Lincoln Park, the Gold Coast, or Downtown all North Side neighborhoods while no one moved deep in the South Side to, e.g., Back of the Yards. There is simply not enough variation in the neighborhoods to which people moved to determine the ef fects of neighborhood on perception shift. Finally, as a proxy for measuring how much contact the Mobile members had with the Northern Cities Shift, I looked at whether they were able to find a Lifelong Chicagoan to participate with. For example, Particip ant 6 was able to complete the experiment with Lifelong Chicagoan Participant 7 (her friend), but Participant 9 had to Lifelong Chicagoan to do the experiment with her. It thus seems possible that Participant 6 would have more exposure to the NCS than Participant 9. One might thus though, Participant 9 had Chicago like perception for this continuum, while Participant like. Overall, only 45% of those Mobile participants with Chicago like perception had found a Lifelong Chicagoan to interview with, while 53% of those with Jasper like perception had done so. (If someone interviewed with someone they thought qualified as a Lifelong Chicagoan whom I later had to exclude, I counted them as half of a person interviewing with a Chicagoan.) Counting the other way, 34.6% of those who had interviewed with a Lifelong Chicagoan had shifted perception, but 41.7% of those who had to interview alone were also shifted. Either more exposure to the NCS actually makes a Mobile person less likely to shift his perception, or the
126 ability to find a Lifelong Chicagoan to interview with is not a good proxy for exposure to the new dialect. To summarize, there is moderately reliable evidence that some adult second dialect learners will shift their perception, but it is clear that some others will not. Further research, preferably wit h longitudinal data, is needed to determine which social and/or linguistic factors influence which individuals do and do not shift their perception upon adult exposure to a second dialect. Conclusions a bout the Malleability of Phonetic Perception in Adulth ood address only malleability of perception. Can adults exposed to a new dialect adjust their perception to match or approximate the perception of native speaker hearers of their new dialect? The results of this perception experiment suggest that adjustment of perception in adulthood is possible but relatively rare. For the Bad~Bed continuum (tes ting // raising), statistical evidence suggests that someone moving from the Upper South or Lower North dialect region to one of the Northern Cities would not adjust his perception. It is statistically plausible that the average perception of a speaker h earer of the native dialect and of mobile people would be identical, and it is not statistically plausible that the average perception of mobile people and speaker hearers of the target dialect are the same. (That is, we cannot reject the null hypothesis that the true mean perception for Lifelong Jasper and Mobile speaker hearers is the same, but we can reject the null hypothesis that the true mean perception for Mobile and Lifelong Chicago speaker hearers is the same.) Statistical evidence shows that mobi le Americans do not adopt NCS perception of // raising.
127 For the Dead~Dud continuum (testing / / backing), the picture is not as clear. It is statistically plausible that the average mobile person would completely adopt the NCS perception pattern, but it is also statistically plausible that the average mobile person would completely maintain her na tive dialect perception. (That is, we were able to reject neither of the two null hypotheses.) However, investigation of individual Mobile their perception to align with that of the Lifelong Chicagoans. I therefore tentatively conclude that phonetic perception can remain malleable in adulthood, but that the malleability is limited. Further, for some individuals, adult phonetic perception seems to be fixed. 12 Longitudinal studies of phonetic perception are needed to verify that some hearers can in fact change phonetic perception in adulthood. Implications for Production Investigation Because there was no evidence of regional variation overall in perception of /a/ fronting, I decided not to include this variable in my production analysis. In my data, / / raising and / / backing were shown to be perceived differently by Lifelong Chicagoans and Lifelong Jasperites, so I included those two variables in my product ion analysis to enable a comparison between production and perception. Production tokens of only //, / /, and / / were extracted for comparison with perception of // raising and / / backing. Discussion of Gender vs. Sex in Phonetic Perception One method ological aim of this study is to investigate the utility of including gender as well as sex in studies of sociolinguistic variation. First, let me return to the 12 This is particularly noticeable for Part icipants 10 and 73, each of whom has lived in Chicago for over 30 years and yet has maintained Jasper like perception of / / backing.
128 discussion of response time. Participant sex did not appear to affect their responses times a t all. On the other hand, gender scores were correlated with response times: response time increased with increasing positive femininity and decreased with increasing negative masculinity. Thus, gender appears to play a bigger role in determining reacti on times than sex does. However, reaction time is not really a part of language. A gender based difference in non linguistic behavior does not mean that gender is important to sociolinguistic studies. Gender scores were never significantly correlated w ith actual phonetic perception, while sex was a significant predictor of responses to the Job~Jab continuum. This tentatively indicates that sex is a more important factor than gender in sociolinguistic variation, at least for perception. However, Chapte r 5 will show that gender does have an effect on phonetic production. This chapter has reported the analysis and results of an experiment testing perception of NCS /a/ fronting, // raising, and / / backing. Contrary to expectations, residence group did n perception of /a/ fronting; this could be because /a/ fronting is less advanced in Chicago than in other Northern Cities and in fact is retreating over time. Regional variation in // ra ising and / / backing did appear in this study, and results suggest that a shift in phonetic perception during adulthood is possible but relatively uncommon. Variation in production of these two vowel shifts will be described in Chapter 5.
129 Table 4 1. P erception stimuli not receiving a response, tallied by continuum and residence group Residence group Bad~Bed Dead~Dud Job~Jab Total Lifelong Jasper 1 0 2 3 Mobile 3 0 1 4 Chicago 0 1 6 7 Total 4 1 9 14 Figure 4 1. Hypothetical example of responses to Job~Jab continuum showing a 50% crossover point between stimuli 5 and 6 0 0.2 0.4 0.6 0.8 1 1.2 1 2 3 4 5 6 7 8 9 10 Proportion Jab responses Stimulus
130 Table 4 2. F1 and F2 values of perception stimuli, measured in Barks Stimulus F1 F2 Stimulus F1 F2 Stimulus F1 F2 Job~Jab 1 7.49 10.76 Bad~Bed 1 6.19 12.49 Dead~Dud 1 5.54 11.09 2 7.47 11.23 2 6.03 12.49 13 2 5.47 11.28 3 7.45 11.17 14 3 5.88 12.50 3 5.39 11.46 4 7.37 11.65 4 5.74 12.50 4 5.33 11.63 5 7.23 12.03 5 5.61 12.51 5 5.27 11.81 6 6.85 12.30 6 5.45 12.51 6 5.20 11.98 7 6.81 12.56 7 5.27 12.52 7 5.12 12.13 8 6.61 12.67 8 5.07 12.52 8 5.04 12.27 9 6.43 12.76 9 4.83 12.52 9 4.95 12.42 Job~Jab 10 6.26 12.98 Bad~Bed 10 4.62 12.53 Dead~Dud 10 4.85 12.58 Figure 4 2 Logistic regression curve with two predictors. F1 is graphed on the x axis, F2 on the z axis, and probability on the y axis. 13 None of these F2 values are actually identical. They only appear the same when rounded to 2 decimal places. 14 This valu make the stimuli. However, the numbering of the stimuli from 1 to 10 is arbitrary, so one could just as well switch the numbering of stimuli 3 and 4, and the F2 values would then be in order. No statistics were run using the stimulus numbers.
131 Figure 4 3. Intersection of the combinations of F1 and F2 (in Barks) that yield vowels that a listener would hypothetically hear as 50% Dead and 50% Dud (solid curve) and set of formant values for the stimuli actually heard (dotted line) Table 4 3. Mean perceptual crossover points (in Barks) for Job~Jab continuum by residence group and perception measure Residence group F1 (lower value is farther forward [and higher] in vowelspace) F2 (higher value is farther forward) F2F1d (higher value is farther forward) Lifelong Jasper 6.88 12.33 5.44 Mobile 6.87 12.37 5.48 Lifelong Chicago 6.87 12.36 5.47 Table 4 4. Mean perceptual crossover points (in Barks) for Bad~Bed continuum by residence group and perception measure Residence group F1 (lower value is higher in vowelspace) F2 (higher value is higher [and farther forward]) F2F1d (higher value is higher ) Lifelong Jasper 5.29 12.52 7.27 Mobile 5.30 12.51 7.25 Lifelong Chicago 5.15 12.52 7.38 10.5 11 11.5 12 12.5 13 4.4 4.6 4.8 5 5.2 5.4 F1 F2 Actual F1 & F2 combos heard Individual Listener's 50% combos of F1 & F2
132 Table 4 5. Mean perceptual crossover points (in Barks) for Dead~Dud continuum by residence group and perception measure Residence group F1 (lower value is farther forward [and higher] in vowelspace) F2 (higher value is farther forward) F2F1d (higher value is farther forward) Lifelong Jasper 5.27 11.82 6.55 Mobile 5.28 11.77 6.49 Lifelong Chicago 5.29 11.74 6.45 Table 4 6. Results of multiple regression for Job~Jab continuum, F1 crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 9.098 1.925 4.727 .000 Jasper .009 .026 .048 .367 .715 Chicago .014 .027 .069 .521 .604 YOB .001 .001 .130 1.121 .266 Sex .045 .022 .237 2.011 .048 a. Dependent Variable: F1 Cross Job Jab Table 4 7. Results of multiple regression for Job~Jab continuum, F2 crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 8.473 2.676 3.166 .002 Jasper .040 .036 .145 1.118 .268 Chicago .016 .037 .055 .422 .675 YOB .002 .001 .163 1.419 .160 Sex .058 .031 .221 1.894 .062 a. Dependent Variable: F2 Cross Job Jab
133 Table 4 8. Results of multiple regression for Job~Jab continuum, F2F1d crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 1.115 4.589 .243 .809 Jasper .039 .061 .081 .632 .529 Chicago .035 .064 .070 .537 .593 YOB .003 .002 .158 1.391 .169 Sex .133 .053 .290 2.522 .014 a. Dependent Variable: F2F1DiffCrossJobJab Table 4 9. Results of multiple regression for Job~Jab continuum including interaction of sex and residence group, F1 crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 8.885 1.901 4.674 .000 Jasper .093 .080 .472 1.157 .251 Chicago .189 .092 .933 2.047 .045 Sex .005 .036 .027 .141 .888 YOB .001 .001 .122 1.064 .291 JasperSexInteract .056 .051 .459 1.096 .277 ChicagoSexInteract .109 .055 .958 1.987 .051 a. Dependent Variable: F1 Cross Job Jab Table 4 10. Results of multiple regression for Job~Jab continuum including interaction of sex and residence group, F2 crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 8.758 2.649 3.306 .002 Jasper .152 .112 .552 1.364 .177 Chicago .250 .129 .879 1.945 .056 Sex .008 .050 .031 .166 .869 YOB .002 .001 .155 1.365 .177 JasperSexInteract .075 .071 .440 1.061 .292 ChicagoSexInteract .146 .076 .913 1.911 .060 a. Dependent Variable: F2 Cross Job Jab
134 Table 4 11. Results of multiple regression for Job~Jab continuum including interaction of sex and residence group, F2F1d crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) .579 4.517 .128 .898 Jasper .180 .191 .376 .947 .347 Chicago .479 .220 .967 2.181 .033 Sex .025 .085 .054 .291 .772 YOB .003 .002 .148 1.332 .187 JasperSexInteract .094 .120 .319 .784 .436 ChicagoSexInteract .274 .130 .988 2.106 .039 a. Dependent Variable: F2F1DiffCrossJobJab Table 4 12. Results of multiple regression for Bad~Bed continuum, F1 crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 2.071 4.400 .471 .639 Jasper .009 .060 .020 .156 .876 Chicago .134 .061 .290 2.209 .031 YOB .004 .002 .192 1.684 .097 Sex .038 .051 .086 .746 .458 a. Dependent Variable: F1 cross Bad Bed Table 4 13. Results of multiple regression for Bad~Bed continuum, F2 crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 12.711 .086 147.860 .000 Jasper .000 .001 .042 .324 .747 Chicago .002 .001 .221 1.699 .094 YOB .000 .000 .260 2.301 .024 Sex .002 .001 .190 1.659 .102 a. Dependent Variable: F2 Cross Bad Bed
135 Table 4 14. Results of multiple regression for Bad~Bed continuum, F2F1d crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 12.736 4.387 2.903 .005 Jasper .012 .060 .027 .208 .836 Chicago .112 .062 .240 1.807 .075 YOB .003 .002 .144 1.255 .214 Sex .022 .051 .049 .423 .673 a. Dependent Variable: F2F1DiffCrossBadBed Table 4 15. Results of multiple regression for Bad~Bed continuum including interaction of residence group and YOB, F1 crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 5.353 .088 60.733 .000 Jasper .011 .060 .024 .186 .853 Chicago .139 .061 .301 2.284 .026 Sex .038 .051 .085 .735 .465 YOB .005 .004 .262 1.374 .174 JasperYOBinteract .001 .005 .041 .267 .790 ChicagoYOBinteract .006 .005 .163 1.035 .304 a. Dependent Variable: F1 cross Bad Bed Table 4 16 Results of multiple regression for Bad~Bed cont inuum including interaction of residence g roup and YOB, F2 crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 12.513 .002 7194.267 .000 Jasper .000 .001 .043 .330 .743 Chicago .002 .001 .226 1.716 .091 Sex .002 .001 .189 1.636 .107 YOB .000 .000 .310 1.627 .109 JasperYOBinteract 3.387E 6 .000 .005 .032 .975 ChicagoYOBinteract 6.193E 5 .000 .092 .585 .560 a. Dependent Variable: F2 Cross Bad Bed
136 Table 4 17. Results of multiple regression for Bad~Bed continuum including interaction between residence group and YOB, F2F1d crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 7.223 .088 82.076 .000 Jasper .013 .060 .029 .219 .828 Chicago .116 .063 .248 1.850 .069 Sex .021 .052 .049 .416 .679 YOB .003 .004 .159 .812 .419 JasperYOBinteract .002 .005 .064 .398 .692 ChicagoYOBinteract .003 .005 .094 .593 .555 a. Dependent Variable: F2F1DiffCrossBadBed Table 4 18. Results of multiple regression for Dead~Dud continuum, F1 crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 6.545 .856 7.644 .000 Jasper .018 .012 .197 1.526 .131 Chicago .010 .012 .112 .849 .399 YOB .001 .000 .167 1.471 .146 Sex .000 .010 .004 .034 .973 a. Dependent Variable: F1 Cross Dead Dud Table 4 19 Results of mu ltiple regression for Dead~Dud continuum, F2 crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 8.220 2.377 3.458 .001 Jasper .048 .032 .191 1.481 .143 Chicago .030 .034 .118 .901 .371 YOB .002 .001 .169 1.489 .141 Sex .001 .028 .004 .038 .970 a. Dependent Variable: F2 Cross Dead Dud
137 Table 4 20 Results of multiple regression for Dead~Dud continuum, F2F1d crossover points Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 1.681 3.177 .529 .598 Jasper .065 .043 .194 1.502 .138 Chicago .035 .045 .103 .785 .435 YOB .002 .002 .172 1.507 .136 Sex .007 .037 .021 .179 .859 a. Dependent Variable: F2F1DiffCrossDeadDud
138 CHAPTER 5 PRODUCTION The last chapter presented results of the perception experiment alone, without comparison to production. Similarly, this chapter presents only production. I begin by describing the methods of interview transcription and vowel extraction and analysis, acquisition of a second dialect in adulthood, as well as the relative importance of sex and gender in malleability of production. Interview Transcription Before I could is olate individual vowels for analysis, the interviews had to be transcribed. The 62 interviews and conversations were transcribed using Express Scribe software. For the transcription conventions, I adapted the transcription method used for the Anglo Montr ealer Corpus (e.g, Blondeau and Nagy 2008). Method of Analysis Extraction of Target Vowels from the Interviews /, / /, and / /. For each participant, I extracted 15 stresse d tokens of each of the three vowels of interest from the interview or conversation in which they took part. Because I expected participants to be more aware of the microphone and monitor their speech more toward the beginning of the interview, I began ex tracting vowels 15 minutes into the recording t o find less monitored vowels and then extracted the first 15 stressed tokens of each vowel (that met criteria described below) following the 15 minute mark in the interview.
139 Stress was determined through my ow n native speaker intuition. When deciding thumb was that if I had to listen to a phrase several times to decide whether a vowel was stressed, I should treat the vowel as u nstressed. By following this rule, I was able to extract only those vowels that were clearly stressed. Unfortunately, some participants did not have 15 stressed tokens of all three vowels in the portion of the interview that followed the 15 minute point. In some cases, this was due to one participant in a 2 participant conversation dominating the recording, while in others, a talkative participant simply happened not to use a lot of words with the vowel I needed. When this happened, I went back and extr acted vowels from before the 15 minut e mark of the interview. In these cases, I extracted the latest vowels first and worked my way toward the start of the interview. (That is, I would extract a vowel 14 minutes into the interview before one 13 minutes i n.) Unfortunately, this was necessary for 22 out of the 76 participants. For eight participants, even after I extracted vowels from the first fifteen minutes of the re cording, at least one vowel of interest still d id not have 15 stressed tokens In this case, I was forced simply to analyze the vowels that I was able to extract from the interview. 1 Table 5 1 lists participants whose vowels had to be extracted before the 15 minute mark of their interview/conversation, how many pre 15 minute tokens had to be extracted, and how many tokens of the vowel (if fewer than 15) were found in total. 1 Had I ever found fewer than 10 stressed tokens of a certain vowel in an interview, I planned to resort to extracting enough vowels from the reading passage to have a total of 10. Luckily, this was never necessary.
140 Overall 2.37% of / / tokens, 2.57% of / / tokens, and 3.2% of / / tokens were extracted from the first 15 minutes of interviews and conversations. Being stressed was o nly the first of several criteria that qualified a vowel for extraction. Second, only vowels in one or two syllable words were extracted; vowels in words of three or more syllables were expected to have shorter duration and thus yield less information ab separate one or two syllable words that were actually part of longer, set phrases. I therefore did not extract e.g., the word from the phrase or sudden from a ll of a sudden I assumed that speakers treated such phrases as single, multi syllable phonological units. Third, the vowels had to be followed by an obstruent. Vowels followed by liquids were excluded for ease of segmentation, those followed by nasals e xcluded to avoid the analysis problems caused by antiformants in nasalized vowels (Johnson 2003: Ch. 9) Because vowels before voiced and voiceless obstruents are often produced different ly (House and Fairbanks 1953), I made sure to extract at least three tokens followed by voiced obstruents and three followed by voiceless obstruents for each vowel and each participant. 2 3 2 Because I could not be sure how a flapped alveolar consonant was stored phonemically, I did not count vowels followed by flaps ( letter, ladder ) toward the voiced or the voiceless total. Have in the phrase have to that I could not feel confident that the
141 I further excluded repeat words; no participant could have the same word extracted more than once. Included in this ban were homopho nes (that is, what a standard dictionary would describe as homophones); I never extracted both red and read (or both last last continue participant. 4 Similarly, I excluded other forms of words that had already b een extracted, so no participant could have both pass and passes blood and bloody or seven and seventh Nor would I extract, for example, both back and background from the same participant. However, I would extract both cab and cabinet or let and letter since these words are not semantically related. I also would extract all of have, has, and had or both better and best from the same participant because the shared roots for these words are relatively difficult to identify, and an uneducated or ver y young English speaker might not realize they were related. For example, if a participant had mentioned a brother or friend named Greg I would not have extracted that as a (such as Brett Favre) and well known people or places in the area (such as referring to a barbershop in Jasper). I also extracted the name of my brother Zach as tokens of // since man y of my participants either were recruited by Zach himself or even given my relatively lax criteria, I suspect I would almost never have found 15 had I aimed to find at leas t two tokens preceded by liquids, at least two preceded by nasals, etc. 4 This exclusion could have caused me to miss tracking some variation in production. It is possible that, for example, red but not read could be a lexical exception to phonological co nditioning of / / production. In some American English dialects, can the noun is pronounced with a tense //, while can the modal auxiliary has a lax // (Labov et al. 1972). My extraction method would not capture such variation. I made this decision because it can be difficult to determine whether two words constitute a pair of homophones or simply two different meanings of the same word. It was more efficient to exclude all homophones as well as repeat words.
142 As special exclusions, I did not extract catch, nor just in the sense of people (including me) store catch would have had trouble deciding which vowel to count catch as a token of. Similarly, while speakers may pronounce just ssing it (or when it h it is generally pronounced [ I st ]. Just was therefore also excluded from analysis. I further excluded stressed vowels that occurred when a speaker was imitating Also excluded were vowels of exaggeratedly long duration. For example, if a speaker said not simply but vowel would be excluded. Finally, I excluded words used in (or similar to those used in) the perception experiment when it was clear that the participants were referring to the experiment. The referred to the choice of dead or dud in the perception experiment. Such jokes were fairly common, and in no cases were words from such jokes extracted for analysis of production. 47: Duh : [overlap] or dead ? [laugh] making me duh dead [both laugh] 47: [overlap] Duh ? Or dead ? 48: [overlap] Or dead duh ? Uh:
143 dead dad Segmentation of Vowels Once all the vowels were extracted, I segmented the spectrograms to mark where the v owel began and ended in the sound clip. I segmented these sound clips as consistently as possible, given the varied amounts of echo and background noise in the different recording locations. All boundaries were checked at least twice. One challenge when marking the end of the vowels in particular was differentiating speech from echo. Many of the recordings were made in carpeted rooms with relatively little echo, but some of the recordings have slight to noticeable echo throughout the interview. In these cases, echoes of the vowels sometimes still appeared fairly clearly in the spectrograms and waveforms. I did my best to differentiate speech from echo, but I am certain that I did not do so with 100% accuracy. Most of the boundaries were marked at a zero crossing in the waveform. However, sometimes waveforms strayed below or above the zero line for several glottal pulses; the nearest zero crossing might then be quite far from the proper boundary marking. In such cases, I made an exception and set the bo undary away from a zero crossing. An example is show n in Figure 5 1 in whi ch the initial boundary on the v owel tier is set an entire pulse before the nearest zero crossing. Praat s ettings Praat spectrogram settings were adjusted from their standard settings to enhance the clarity of the formants. The view range and dynamic range were left at the standard settings of 0 5000 Hz and 50 dB, respectively, but the window length was adjusted to .009 seconds. (All advanced spectrogram settings were left at their standards.)
144 Beginning of v owels The cues to the beginning of the vowel depended, of course, on the preceding segment. When the vowel followed an unaspirated stop or a pause (glottal stop), the boundary was placed at the zero cr ossing closest to the appearance of the full formant structure following the stop burst. If the full formant structure appeared during or before the stop burst, the boundary was placed at the zero crossing following the end of the stop burst, as shown i n Figure 5 2 Following an aspirated stop, there was usually a clear boundary between the aspiration, which sometimes showed faint formants, and the full formant structure of the unaspirated vowel. In such cases, the vowel onset was marked at the zero cross ing nea rest this boundary, as shown in Figure 5 3 If this boundary was not clear, the vowel onset was marked at the earliest zero crossing at which formants were clearly visible and no aspiration was audible. The same method of marking the boundary was used when the vowel followed /h/. When the vowel followed a voiceless fricative other than /h/, the boundary was placed at the zero crossing closest to the onset of the full formant structure following the voiceles s noise of the fricative, as in Figure 5 4 This boundary was usually quite clear. Relatively few of the extracted vowels followed voiced fricatives. When this did occur, the boundary was placed at the first zero crossing after the appearance of the full formant structure at which the fricatio n noise was no longer audible. Boundaries following affricates, whether voiced or voiceless, were marked at the zero crossing nearest the appearance of the full formant structure after the end of the audible frication. When the vowel followed a nasal, there was usually a dramatic change between faint and dark formants as the nasal moved into the vowel. In the transition zone, there
145 would usually be a single glottal pulse with dark formants that sloped steeply upward. T he boundary was marked at the zero crossing at the start of the glottal pulse following this transitional pulse, as in Figure 5 5 In the absence of such a transitional pulse, the boundary was marked at the zero crossing nearest the start of the darker fo rmants. Predictably, the most difficult onset boundaries to mark were those for vowels following liquids and glides. For these boundaries, I relied more upon my ears than my eyes. The boundary was placed at the zero crossing at which the liquid or glide ceased to be clearly audible. End of v owels Before an affricate or stop (including a glottal stop, as in button ), the boundary was generally marked at the zero crossing near the disappearance of the full forman t structure of the vowel, as in Figure 5 6 H owever, if the following consonant was clearly audible before this point, the boundary would be marked at the zero crossing right before the consonant became audible. Marking the boundaries between vowels and following fr icatives was more challenging Ma ny participants tended to devoice their vowels well before the onset of a voiceless fricative. I decided not to take into consideration whether the vowel was voiced or voiceless and simply mark the boundary at the end of the clearly visible formants, as i n Figure 5 7 However, if the devoicing caused Praat to dramatically misidentify the formant placement at the end of the vowel, I marked the boundary before the formant identification became markedly inaccurate. (This was not common, however.) The method of marking the boundary between a vowel and a voiced fricative depended on which fricative was involved. When a vowel transitioned into a /v/, as in
146 Figure 5 8 there was often a lightening or complete disappearance of the formant structure. The boundar y was marked at the zero crossing nearest this disappearance, or at the zero crossing b efore the [v] became audible. In contrast, t he formant structure fr equently continued well into a [ z], so th e boundary between a vowel and /z/ was marked at the zero cr ossing immediately before the [z] became audible. Th e boundary between a vowel and / / was the most difficult to mark because / / often looked like a sonorant on the spectrogram, as in Figure 5 9 and it was often impossible to find a portion of the sound clip where the fricative was at all audible. (That is, it seems as though medial / / is often actually deleted in casual speech.) When possible, I placed the boundary at the zero crossing before the fricative became audible; otherwise, I placed it at th e zero crossing before the vowel quality changed dramatically enough to make it clear that the speaker had begun producing the next vowel. Finally, the boundary before a flap was marked at the zero crossing before the flap became audible. Measurement of Formant Values from Segmented Vowels Once all vowels were segmented, the formant values of the vowels had to be measured I chose to measure F1, F2, and F3. F1 and F2, which correlate with vowel height and frontness, respectively, are of course most freq uently used to determine vowel quality. However, F3 can also differentiate vowels, par ticularly in terms of rounding (van der Harst 2011: Ch. 8) and R coloring (Ladefoged and Maddieson 1996), the latter of which I did not expect to find (frequently) in th e data because I did not examine vowels with following /r/. F3 was therefore extracted in case it would prove useful in differentiating vowels that were audibly different but appeared in the same location on an F1 x F2 plot. I did not attempt t o extract F0 because many of the vowels
147 in the dataset were produced with creaky voice, so Praat would be unable to find a pitch track for them. Other vowels, particularly those before voiceless fricatives, were actually devoiced partway through, m aking pitch trac king impossible and meaningless. Determining the n umbe r of coefficients for LPC analysis Praat determines the location of formants using Linear Predictive Coding (LPC) and t he results of LPC analysis vary dramatically according to how many coefficients are included in the equation. (Basically, for every two coefficients used, one formant appears in the results.) Not all speakers or vowel tokens are optimally measured using the same number of coefficients, and the consequences of using a suboptimal number of coefficients can be great. Using too few coefficients can lead Praat to ignore clearly visible formants, while using too many can lead the analysis to invent a formant t hat each token, or at least each speaker. There is a trade off between work time efficiency and accuracy in determining how many formants to look for (how many co efficients to use) (See van der Harst 2011 : Ch. 4 for an overview.) The quickest but least accurate method is to simply use a default number of formants, probably five, to use for all vowels. A still fast but somewhat more accurate method is to use dif ferent defaults for each sex, such as four formants for all females and five for all males. At the other end of the spectrum, the method that yields the most accurate results but consumes the most time is to determine the optimal number of formants to fin d for each individual vowel. For this analysis, I took a middl e road intended to balance the demands of time and accuracy For each participant, I inspected nine vowels, three tokens of each target
148 phoneme 5 and determined the best number of formants to u se for each of these nine vowels. The number of formants that was best for the majority (or plurality) of these nine vowels was used for all the vowels of that participant. 6 In this way, I only had to inspect 684 vowels, but the number of formants could be optimized for each participant. For each vowel, I first inspected the spectrogram with no formant tracking so that I and tested different numbers of formants/coeffici ents to see which led to the most to test more or fewer to get the most accurate tracking. When determining which setting yielded the most accurate results, I considered the tracking of only the first three formants, with more weight placed on the first two than the third. That is, if the five formant setting led to accurate formant tracking of F1 and F2 but mediocre tracking of F3, and the six formant setting improved the tracking of F3 but had poorer tracking of F1, then the vowel was determined to be optimally analyzed with five formants. Table 5 2 indicates how many participants of each sex were best analyzed with each number of formants in Praat. 5 I simply used the first three words al phabetically for each phoneme. For example, for Participant 22, I inspected added, attic, back, bedroom, best, bet, couple, covers, and One advantage of this selection method is that the vowels I inspected are likely to have been spread out over the course of the interview. 6 For example, four of the nine vowels I looked at for Participant 57 were best analyzed with six formants, three with seven formants, and one with five formants, and one was analyzed equally well with six or seven formants. Therefore, all vowels for this speaker were analyzed with six formants.
149 Extraction and checking of v alues Formants were extracted using a Praat script ori ginally by Mietta Lennes and modified by Dan Michel. The script extracted the values in Hertz for F1, F 2, and F3 1/6, 2/6, 3/6, 4/6, and 5/6 of the way through each vowel (hereafter referred to as T1 through T5) (Measuring the vowels at five points along their duration as opposed to simply their midpoints or perhaps two points for diphthongs gave me a mor e complete picture of the three this way, each individual vowel yielded fifteen measurements (3 formants x 5 time points). When running the script, I adjusted the number of formants the script would look for according to the results discussed in the previous section. The time step was set to .00625 s, the window length to .025 s, the maximum forman t value to 5500 Hz, Unfortunately, of course, not all the formant measurements produced by this script that match what the human eye can see to be correct. Measurements were less accurate when the participants used breathy voice or had unusually high pitched voices. to possible errors by flagging two ki nds of outliers: outliers in the Hertz value of the formants themselves, and outliers in the difference in values between two adjacent measurements (van der Harst 2011). For example, for Participant 1, the mean value for F1 at T1 for the // tokens was 68 6 Hz, and the standard deviation was 74 Hz. I flagged each measurement that was more than two standard deviations from the mean, i.e., less than 538 Hz or more than 834 Hz. Also for Participant 1, the mean difference
150 between F1 at T1 and T2 for // token s was 31 Hz (i.e., F1 tended to drop about 31 Hz between T1 and T2), and the standard deviation for this difference was 45 Hz. I flagged the measurements at T1 and T2 for any difference that was more than 2 standard deviations from the mean, i.e. when t he formant dropped more than 121 Hz or rose more than 59 Hz from T1 to T2. The flagged measurements were expec ted to be the result of However, it quickly became clear that many measurements not flagged in this way were also inaccurate. Therefore, for every vowel, I looked at the spectrogram with appeared likely to be inaccurate. I then measured by hand the measurements that appeared inaccurate. Be cause hand measurements in Praat cannot be perfectly accurate, I only changed the record of the formant value if what I measured was at least measurement of 1704 Hz was in accurate but measured the formant myself at 1746, I due to a small slip of my hand.) There were four main types of inaccurate measurements that I found. First, Praat occas ionally simply measures an individual point incorrectly and will produce a red formant dot on its own in the middle of nowhere, or miss a spot in an otherwise correctly measured formant. Second, some vowels (particularly but not exclusively those with bre athy voice) simply seemed impossible for Praat to analyze; the sea of red formant dots in this case would have no discernible pattern. (For such vowels, I occasionally had to correct all fifteen measurements by hand.) Third, Praat would occasionally
151 dete ct a phantom formant w here none actually existed (see Figure 5 10 ). In such a case, if, for example, the phantom formant was between the real F1 and F2, the measurements reported for F3 would often actually be the correct measurements for F2, and I could find the correct F3 measurements by asking Praat for the values for F4. Fourth, in rare instances, the reverse of this situation occurred, and Praat would simply ignore a formant that was really there. In these cases, if, for example, the ignored formant was F2, the measurements given for F2 would often be accurate measurements for F3, 7 and I would need to measure all of F2 by hand. In the end, a fair portion of my formant measurements had to be done by hand. It may therefore have saved time if I had ini tially followed van der Harst (2011) and found the optimal number of formants to tell Praat to find for each individual vowel token, measurements visually, but I might have ended u p needing to measure fewer formants by hand. Normalization of Vowel Measurements Before vowel production could be compared across individuals and residence groups, the vowel measurements had to be normalized so that no differences in formant values actu ally caused by physical differences in the participants would masquerade as sociolinguistic variation. However, many methods of vowel normalization not only eliminate non linguistic variation due to physical differences, but also mask the sociolinguistic variation that is vital to t his study. Van der Harst (2011: Ch. 4) tested 17 7 Van der Harst (2011) found that flagging potentially inaccurate measurements based on outliers in the differences between adjacent measurements could be more helpful than flagging based on outliers of the measuren ts themselves. However, note that in the third and fourth cases described here, the differences between adjacent measurements would not be unusually large. It is therefore more reliable to search for outliers both in raw measurement values and in the dif ference between adjacent measurements.
152 different vowel normalization methods to determine which could most reliably retain sociolinguistic variation while still eliminating variation that was not of interest to lingui sts. He found that the three normalization methods best suited to analyses of score transformation (henceforth score transformation applied to Bark values (henceforth LobBark), and Ne mean model (henceforth Nearey). I therefore applied these three normalization methods to my par values. (Running statistics on the output of all three normalization methods was intended to be a way of ve rifying the validity of my results. If a variable was found to be a significant predictor of formant values using all three normalization methods, I felt more confident that that variable actually affected vowel production than I was if the relationship w as found to be significant only for the Nearey normalized values, but appeared nonsignificant using Lobanov and LobBark.) However, there were two slight modifications that were necessary for the normalization methods to work with my data. First, I needed the comparisons of the values across the five time points to have meaning. Had I simply normalized the T1 values with reference to T1 norms, T2 with reference to T2, etc., then the normalized values for each time point would be unrelated to each other. Instead, again following norms. For example, in the Lobanov transformation, instead of taking the observed T1 F1 value, subtracting the mean T1 F1 value for the speaker, and dividing the remainder by the standard deviation for T1 of F1 for the speaker, I took the observed T1 F1 value, subtracted the mean T3 F1 value for the speaker, and divided by the standard deviation
153 of T3 of F1 for the speaker. This allowed me to com pare the movement of the vowel across time points, even after normalization. Secondly, the normalization methods all required taking the mean value of a given work well than others would have had skewed means if I had used this method; the vowel categories with instead of taking the simple mean of all the vowel tokens for each speaker, I instead first took the mean value for each vowel category, and then averaged the three resulting values. For example, the mean F1 values at T3 for Participant 1 were 733 Hz for //, These two modifications apply to all three normalization methods. Once the normalized values were obtained, I analyzed each set of normalized values statistically and compared the results. Statistical Analysis of Vowels The goal of the statistical analysis was to determine the effect of residence g roup and other social factors on the production of the three vowel categories at the five time points while controlling for non social linguistic factors (such as following consonant) that were expected to affect the production as well. At the same time, I did not want to over model my data by including too many predictive factor s. I therefore began, as I had with the perception data, by measuring effects of individual factors on production using ANOVAs, t tests, and Pearson correlations. To d o this, I first had to take the
154 averages then used PASW software to find the effects of v arious sociolinguistic factors on these averages, for all five time points of all three formants on all three vowel categories (for all three normaliz ation methods). The effect of residence g roup was tested with ANOVAs (with post hoc tests to determine wh ich of the groups were significantly different f rom each other); the effect of s ex was tested wi th t tests; and the effects of year of birth (YOB), education, number of years lived in Chi cago (YLC, Mobile group only), age of a rrival in Chicago (AAC, Mobile group only), percentage of life spent in Chicago (PLC, Mobile group only), total m ascul inity, and total f emininity were tested with Pearson corr elations. The effects of YOB, education, total masculinity, and total f emininity were also tested with partia l correlations controlling for residence g roup. For each vowel category, I then noted which of the social factors had been significant predictors of the averaged formant values for at least one of the time points. These social factors were then included i n mixed effects regressions. The social factors included for each vowel and nor malization method are listed in Table 5 3 Mixed effects regressions using Participant as a random factor and the appropriate social and linguistic factors as fixed effects wer e conducted using the lmer(4) library in the statistical package R. (These regressions were obviously run using all the individual data points, not the averaged values for each participant.) Unfortunately, R requires values in all cells for every data p oint, so the social fa ctors YLC, AAC, and PLC, which had no values for the Jasper or Chicago groups, could never be included in
155 these regressions. The linguistic factors included were Before 15 8 (whether the vowel was extracted from the first 15 minutes o f the interview), Number of Syllables (in the target word, always either 1 or 2), Voicing Before and Voicing After (whether the segment preceding/following the vowel was voiced), Place Before and Place After (place of articulation of the segment preceding/ following the vowel), and Manner Before and Manner After (manner of articulation of the segment preceding/following the vowel). However, multicollinearity effects sometimes prevented all these linguistic factors from being included. Specifically, I had t o exclude Manner Before for all // analyses, and T5 of F3 ) I ran an initial mixed effects regression using all the appropri ate social factors, all the linguistic factors not eliminated by multicollinearity, and the random factor of Participant. Then, in order to have final statistics that did not over model my data and also would be easier to interpret, I ran a new mixed effe cts regression eliminating the linguistic factors that did not reach significance (|t|>1.96) 9 and social factors that did not reach marginal significance (|t|>1.645) in the original mixed effects regression. (I set a lower threshold for the social factors because they were of theoretical interest to this investigation, whereas I was only including the linguistic factors as controls to ensure that no apparent sociolinguistic effects were actually due to, e.g., one residence group having more velar final tok ens than another.) However, the revis ed regressions always 8 9 For mixed effects regression, R reports t values for each factor but neither p values nor degrees of freedom. Given that I had a large number of tokens for each vowel, I assumed that the t distribution of the results would be fairly close to the z distribution. This assumption means that I will have a slight bias tow ards finding too many factors significant.
156 included residence g roup, since that was the factor in which I was most interested, and always included Participant as a random factor. When either total m asculi nity or total femininity ( M/F Tota l ) was at least marginally significant in the initial mixed effects regression, I would begin by running the revised mixed ef fects regression with the M/F Total score still used as a fixed effect. Then, I would run the revised regression again, this time s ubstituting the positive M/F score for the t otal one; I wou ld then run it again using the n egative M/F score instead. If neither the positive nor the n egative gender score alone was at least marginally significant, I would accept the revised regression us ing the t otal gender score as in the original. How ever, if either (and only) the positive or n egative gender score was at least marginally significant in the revised regression, I would accept that version of the r egression, finding, e.g., that p ositive f emininity had a more significant effect on the production at that point than negative f em ininity. Finally, if both the positive and n egative gender scores were at least marginally significant in their individual revised regressions, I ran one more revised regression this time including both the positive and the n egative gender scores as fixed effects. 10 Thus, the order of mixed effects regressions for each point was as follows: 1) all appropriate social factors, all possible linguistic factors, and Partici pant; 2) only social factors that had reached marginal significance (plus always residence g roup), only linguistic factors that had reached significance, and Participant; 3) (if applicable) repeating regression 2 but substituting positive gender for t otal gender; 4) (if applicable) 10 On very few occasions, both Total Masculinity and Total Femininity were at least marginally significant in the initial mixed effects regression. In this case, once I had eliminated the non significant factors f rom the first run, I first substituted in Positive and then Negative Masculinity while keeping Total Femininity in the model. Based on these runs, I decided which Masculinity score to maintain in the model, and then substituted in Positive and then Negati ve Femininity as indicated above.
157 repeating regression 2 but substituting negative gender for t otal gender; 5) (if applicable) repeating r egression 2 but including both positive and negative gender in place of t otal gender. When necessary, I ran further regressio ns including interaction terms, as described in the Results section below. The results of the final mixed effects regression for each vowel category, time point, formant, and normalization method can be found in Appendix G. Results Figures 5 11 through 5 2 5 sake, I use the Lobanov normalized values for all charts in this section.) These charts eye the statistical results described below indicate which apparent differences would reliably show up in other groups besides my specific participants. Perhaps the most striking thing to note about these vowel charts is the Lifelong at the following time points. This underscores the point that measuring vowels at only their midpoints, or even the point of inflection for known diphthongs, can cause major regional differences in production to be missed. While I believe that it is important to investigate vowel charts showing each vowel individually, as I present them here, I also recognize that these charts a re somewhat difficult to read, so I am including similar charts with the data made somewhat easier to read in Appendix H.
158 I now present trends revealed by the statistical analyses. Those wishing to know specific statistical results for each time point are again referred to Appendix G In this discussion, I refer to trends revealed by examining the results of statistical analyses on all three normalization methods Variation in Production of // Analysis of F1 Figures 5 26 through 5 28 chart the Lobanov no rmalized values for F1 of // by residence g roup over time, with the mean values at each time point connected by lines. These figures make it easier to compare the changes in each formant over time than the vowel charts in Figures 5 11 through 5 25 It c an be seen that for all three groups, the normalized F1 value starts comparatively low (higher vowel) and then rises (the vowel becomes lower) in the middle, and then F1 falls again. However, for the Lifelong Jasper and Mobile groups, the F1 peaks (the vo wel reaches its lowest point) around the not reached until T4. Note also that the Lifelong Chicago group is the only one to have the mean normalized F1 value start be low 0 at T1, meaning their // starts out in a Residence g roup. For al l three normalization methods, residence g roup was a significant or marginally significant predictor of F1 values at all five time points. For all was significantly (or marginally significantly) lower than that of the Mobile group (i.e., the as higher at all time points). Additionally, for all of T1 higher than that of the Mobile group (i.e., the Lifelong Jasper group had a lower //).
159 We can thus be c onfident that an average Mobile group member would have an F1 for // that is between that of average Lifelong Chicago and Lifelong Jasper residents. That is it is plausible neither that the average Mobile person would maintain her native Jasper like // height, nor that she would completely adopt the higher Lifelong Chicago //. We have evidence of movement toward the norm of a new dialect in adulthood, without full acquisition. Age of arrival in Chicago y ears lived in Chicago and p ercentage of life s pent in Chicago. 11 Do speakers who move to the new dialect area at a young age adopt the second dialect (D2) more completely or is it more important that they live many years or a large percentage of their life in the new regio n? For //, age of a rrival i n Chicago (AAC) was a significant or marginally significant predictor of F1 only for T5, never for T1 through T4. (At T5, the older the speaker was when she arrived in Chicago, the higher her F1 value. That is, speakers who arrive when they are younger are more likely to adopt the raised // at the en d of the vowel.) In contrast, number of years l ived in Chicago (YLC) was an (at least marginally) significant predicto r of F1 at T1, T2, and T5, and percent of life s pent in Chicago (PLC) was at least margi nally significant for T1 and T2 (and T5 for LobBark). 12 At T1 and T2, the more years/greater percentage of his life a Mobile speaker has spent in Chicago, the lower his F1 value. That is, speakers who have spent more years/a larger proportion of their lif e in Chicago 11 average formant value for each of the three vowels of interest at each time point, not with all ~15 tokens per vowel per participant. 12 YLC and PLC often patterned together in these correlations because the two measures are themselves strongly positively correlat ed (r = .987, p<.001). In contrast, AAC is not correlated with either of the other two measures (with YLC, r = .059, p = .775; with PLC, r = .066, p = .748).
160 are more likely to have adopted the NCS raised // at the start of the vowel. However, at T5, the more years a speaker has lived in Chicago, the higher his F1 value; people who have lived longer in Chicago actually retreat from the raised // at the tail of the word. (This could indicate that Mobile speakers are adopting a more extremely diphthongized form of // than Lifelong Chicagoans actually have. Perhaps they imitate the stereotype rather than the actual input they receive.) The relev ant correlation results are shown in Appendix G. Implications of these findings for the relative importance of these three measures are discussed below. YOB. Additionally, for T1 through T3, YOB was consistently significant: controlling for other factors, the younger a speaker is, the lower the first half of his // is (that is, F1 is higher at those points). Note that this tendency disappears for T4 and T5. (Only Nearey shows that the younger a speaker is, the higher the tail end T5 of his // is.) I ran mixed effect regressions that in cluded the interaction between residence g roup and YOB on T1 through T3, but this interaction never approached significance. There is no evidence that age had a different effect on F1 for different residence g roups. This would seem to indicate a general Midwestern trend for the beginning of // to be lowering over time. 13 Education. For T2 through T4, e ducation was also a significant predictor of F1: the more educated the speaker, the lower her //. This tendency was only present in 13 I am taking an apparent time interpretation of YOB here, i.e., assuming that differences by a ge indicate change over time. There is also the possibility of age grading, of course. Age grading seems more likely among variables of which the community is aware and which are assigned values in the linguistic marketplace (Wagner 2012). Since / / rai sing certainly has social significance in Chicago, it might be reasonable to take the effect of YOB as indicative of age grading, at least if the effect had been stronger for Lifelong Chicagoans than for other groups. Since the effect appears to be equal across residence groups, though, change in progress may be more likely. Of course, the same decreased / / raising in younger speakers could indicate age grading in Chicago and change in progress in Jasper!
161 the middle section of the vowel, not at either end. This finding makes sense if we assume that on average, more educated speakers avoid the stereotypical NCS raised //. Mixed effect regressions that included the interaction between resi dence group and e ducation showed that the interaction was never significant at T3. On the other hand, at T2, Lobanov and LobBark showed a marginally s ignificant interaction between e ducation and the Lifelong Jasper Group (nonsignificant for Nearey), and a t T4, Nearey showed a marginally s ignificant interaction between e ducation and the Lifelong Chicago group (nonsignificant for Lobanov or LobBark). At T2, for Mobile and Lifelong Chicago speakers, the higher the education, the lower the //. However, for Lifelong Jasper people, the effect is reversed: // raises with increased education. Perhaps Lifelong Jasper speakers actually emulate the NCS raising, or perhaps Jasper speakers tend to have a low // overall, and educated speakers move toward the stand ard. The caveat is of course that this interaction was only marginally significant, and only for one of the three time points at which e ducation was a significant predictor of F1, so we should not read too much into this finding. At T4, for both Lifelong Chicago speakers and others, // becomes lower as education increases, but the effect of education is stronger (more steeply sloped) for Lifelong Chicagoans than for others. More educated Lifelong Chicagoans may have even stronger diphthongization of / / than their less educated peers. The same caveat against reading too much into a marginally significant finding at one time point applies, though. Sex and g ender. Neither sex nor g ender was ever a significant predictor of F1 of // for Lobanov o r LobBark. For Nearey, though, s ex was mar ginally significant at T1,
162 and positive m asculinity was significant at T3. Controlling for other factors, at T1, females had lower F1 (higher //) than males did, and at T3, the more positively masculine someone was, the lo wer their F1. That is, there is some slight evidence that females had more NCS like production than males, but that more masculine people also were more NCS li ke. At T3, the interaction between residence group and positive m asculinity was not significant At T1 however, the interaction between residence group and s ex was significant for Nearey 14 : the difference between males and females was greater for the Lifelong Chicago group than for the other two residence groups. That is, Lifelong Chicago females tend to outstrip males in NCS like production even moreso than females from other residence groups outstrip their male counterparts. Recall that McCarthy (2011) found that // was stable in Chicago, not currently undergoing change. // raising should ther efore be a stable variable. 15 using a more nonstandard form tha n Chicago males. (Note, however, that these conclusions are based on findings each at a single time point for a single normalization method .) The i mplications of the findings on sex and g ender are discussed below. Analysis of F2 Figures 5 29 through 5 31 sho w scatter plots of the Lobanov normalized F2 values for each group over time. It can be seen that both the Lifelong Jasper and Mobile 14 I only tested interactions for the normalizatio n methods for which the main effects were significant. Thus, in this case, I only tested the interaction between residence group and sex for Nearey, not for Lobanov or LobBark. 15 At least in terms of the NCS. The findings for YOB above show that / / is u ndergoing change, but apparently equally so in Jasper and in Chicago.
163 as the vowel progresses. The Lifel back in the mouth as the vowel progresses, but the movement is much more dramatic. Note also that, unlike for the Lifelong Jasper and Mobile groups, the Lifelong Chicago T1 is greater Residence g roup. As with F1, residence g roup was a significant or marginally significant predictor of F2 in almost all cases. Only for T5 of Lobanov was the Lifelong Chicago group not significantly different from the Mobile group. Overall, we again have evidence of partial adult acquisition of a new dialect in production. Controlling for other s F2 for // is significantly higher (i.e. the vowel is farther forward) than that of the Lifelong Jasper group, but their F2 is also significantly lower (vowel farther back) than that of the Lifelong Chicago group. It therefore seems that the average Mob ile speaker would change his production of // to be farther forward, more like that of a Lifelong Chicagoan, but the acquisition would only be partial. The possible exception to this is T5, where the Lobanov analysis makes it appear plausible that the ac quisition of the Lifelong Chicago production could be complete. AAC, YLC, and PLC AAC was never a significant predictor of F2 of //. Conversely, YLC and PLC were both significant predictors for all of T1 through T4. As YLC and PLC increase, the F2 val ue increases. (Note that the correlation was quite strong for T1 through T3, for which r was generally around .7, whereas for T4 it was less than .5.) That is, Mobile speakers who have lived more years or a greater percentage
164 of their lives in Chicago ar e more likely to have adopted a Lifelong Chicago like fronted //. Education and YOB. Education was ne ver a significant predictor of F 2 value for //. YOB, on the other hand, was a significant or marginally significant predictor of F2 value in all cases f or T1 through T4 (never for T5). Controlling for other factors, the younger a speaker is, the farther back the majority of her // is. This matches the findings for F1, in that there seems to be a general Midwestern tendency for // to be both less front ed and less raised over time. Put together, this could indicate that Midwesterners are retreating from the NCS. Yet for F2, unlike for F1, t he interaction between YOB and residence g roup was usually at least marginally significant. For LobBark, the inte raction between YOB and the Lifelong Jasper group was at least marginally significant for all of T1 through T4, and the interaction between YOB and the Lifelong Chicago group was marginally significant for T3. For Lobanov, the interaction between YOB and Lifelong Jasper was at least marginally significant for T1 through T3, and for Nearey, it was marginally significant for T1 and T2. In every case, the interaction showed that for the Lifelong Jasper group, younger speakers still have lower F2 than older s peakers, but the slope is less steep; that is, the effect of YOB is not as strong. This could simply be because the Lifelong Jasper group has less fronted // overall, so there is little fronting for the younger generation to be retreating from. For T3 o f LobBark, the slope for Lifelong Chicagoan YOB was likewise in the same direction, but less steep. For this point, at least, it seems that age is also less important for Chicagoans than it is for Mobile group members.
165 Thus it seems that the effect of a ge is strongest for Mobile speakers, followed closely by Lifelong Chicagoans, and weakest for Lifelong Jasperites. ( Recall that Perhaps the apparent stronger effect of age on Mobile group members is actually an artifact of younger Mobile group members having generally spent fewer years in Chicago. ) Thus, while there seems to be a general Midwestern trend of lowering of / / over time, the backing of / / over time is not gen eral for the region. It appears that / / is (2011) finding that / / is stable in Chicago. The present study shows evidence that / / fronting is retreating in Chicago over time in production. While it is not certain, the difference in findings might be due to the fact that McCarthy measured her vowels at only one time point; perhaps if she had measured the formants across the vowel trajectory, she would also have found evid ence of decreased / / fronting in her younger speakers. (McCarthy also had more speakers with no college education than I did, so the differing results could also be due to differences according to education level. I found no evidence of an effect of edu cation for F2 of / /, but this could be due to the fact that so many of my participants were clustered at the higher end of the education scale.) Sex. Sex was a significant predictor of F2 of // at the end of the vowel. Specifically, at T4 and T5, for al l three normalization methods, females had lower F2 than males. The findings here seem to contradict those for F1 (if the findings for F1 were reliable). For F2, females had less NCS like production than males (at least at the end of the vowel) while fo r F1, females had more NCS like pr oduction. Indeed, in
166 terms of s ex, the findings for F2 would seem to fall more than the findings for F1 did. The discussion of YOB showed that / / is moving back over time in Chicago. Sin ce this is not a case of stable variation, and as far as I can tell this is a change below the level of consciousness, Principle II applies: females should be leading in / / backing, the retreat from NCS behavior. This is indeed what the statistics show. However, there is some slight evidence that the effect of sex may not be as important in Chicago as it is in Jasper. At T4, the interaction between residence group and s ex was marginally significant for Nearey: the difference between males and females i n F2 was greater for the Lifelong Jasper group than for the two others. T he possibility like behavior was most pronounced for the Lif elong Jasper group is telling In Jasper, the fronting of // would not be expected to carry mu ch social significance one way or the other, yet it may be in Jasper that the sex differences are most pronounced. In Chicago, where fronted // should be a strong marker of NCS/Chicago identity, the difference in production for females and males may be less dramatic Ho wever, the interaction between sex and r es idence group did not approach significance at T5 for Nearey, nor for either T4 or T5 for the other two normalization methods. Therefore, overall it seems reasonable to assume that the effect of s ex is steady across residence g roups, meaning that Chicago females do indeed front // less than Chicago males, and Mobile males acquire // fronting more completely than Mobile females stronger / / fronti ng, and Mobile females acquire the weaker / / fronting of Lifelong Chicago females.)
167 Gender. Gender specifically f emininity was likewise frequently a significant predictor of F2 values in the second half of //. At T3 and T4 controlling for other fa ctors, for all three normalization methods, score was, the lower his F2, and at T3 for Lobanov and LobBark, the more positively f emin ine someone was, the lower his F2. For Lobanov and Lo bBark, the interaction betwe en residence group and positive f emininity was significant at T3: the effect of positive f emininity was greater for the Lifelong Jasper group than for the Mobile or Lifelong Chicago groups Increased negative femininity also predicted lower F2 at T5 for Lo banov and LobBark, but the effect was nonsignificant at T5 for Nea rey. The interactions between negative femininity and residence g roup were nonsignificant at T4 and T5. Generally, it seems that the more feminine a speaker is, particularly the more negat ively feminine he is, the less he participates in // fronting. T3 would seem to indicate that this effect is stronger for Jasper, where, as I said above, // fronting should not have much social significance. Analysis of F3 Figures 5 32 through 5 34 sho w scatter plots of the Lobanov normalized F3 values for // for each of the three residence groups, with lines connecting the mean values at each time point. Each of the three groups shows a slight fall in F3 from T1 to T2, and F3 remains fairly steady or drifts downwards thereafter. Residence g roup AAC, YLC, and PLC R elatively few social factors were significant predic tors of F3 for //. Residence g roup was never a significant predictor for F3 at T3 through T5. At T1, all three normalization methods showed a significant difference between the Lifelong Chicago and Mobile groups, and at T2, Nearey (alone) showed a marginally significant difference be tween the same two groups While this
168 evidence could be used to argue that acquisition of F3 in adulthood is unlikely, in fact we simply have little evidence even of differences between Lifelong Jasperites and Lifelong Chicagoans in F3. Perhaps the Mobile group has no real differences to acquire here. (YLC and PLC were sometimes marginally signific ant predictors of the Mobile 4, but since there is no evidence of dialectal difference for them actually to be acquiring, I think this is just a fluke.) YOB and e ducation. YOB and e ducation were also only rarely significant predictors of F 3. For all three normalization methods, YOB was a significant predictor of F3 at T4: the younger the participant, the lower the F3 at this point, controlling for other factors. All three nor malization methods also showed e ducation to be a significant pre dictor of F3 at T5: controlling for other factors, the more educated the speaker, the higher the F3. While these findings could indicate wider sociolinguistic trends, I do not wish to read too much into them. Sex and g ender. Gender was never a significa nt predictor of F3 of //. Sex was only significant at T3, and only for Nearey. At this point, controlling for other factors, females had a lower F3 than males. (I did not test for inter action betweensex and residence group because residence g roup was n ot a significant predictor of F3.) Analysis of F1 Figures 5 35 through 5 37 chart the Lobanov residence group over time. Lines connect the mean values at each time point. For all three gro reaching its lowest point (highest F1) at T3.
169 Residence g roup. As with //, r e sidence g roup was always a significant s F1 values were always significantly groups are less consistent, though. A t At T1 and T5, however, wh ile Lobanov and LobBark still show at least values, Nearey shows no significant difference between them We can thus be confident that t he average Mobile speaker would th us partially acquire the Lifelong Chicago like lowering in the mid beginning and the end of the vowel would remain like that of Lifelong Jasper speakers. We again have evidence for partial acquisition adulthood. This is somewhat were expected in F2, not F1. AAC, YLC, and PLC. While YLC and PLC were often significant predictors of formant values for // those two measures were never significantly correlated with the of / / On the other hand AAC is a significant predictor of F1 at T 4 and T5. The older a speaker was when she arrive d in Chicago, the higher her F1 value, i.e. the lower her at these two points. This is strange because it indicates that speakers who arrive when they are older are more able to acquire the second dialect
170 counterintuitive, and I will show below that this significance of AAC does not appear for statistical error, in which a significant difference is found when no difference actually exists. Education and YOB. On the other hand, YOB was consistently at least marginally significant at T4 and T5, and was at least marginally significant at T3 for Lobanov and LobBark as well. At all th T he interaction between YOB and residence g roup was never significant. Therefore, this correlation seems to be indicative of a wider Midwestern trend toward lowering th e / is getting lower over time in Chicago.) Sex and g ender. Once other factors were controll ed for, g ender was never a nificant for all three normalization methods at T2, and for Lobanov and LobBark at T3. At these points, Females had a higher F1 (lower / /) than males; that is, their F1 production was more NCS like. The interaction of sex and residence g roup was never s ignificant. McCarthy this vow el are still progressing. N one of my Lifelong Chicago or Mobile participants seemed aware of this ongoing change; this appears to be a change below the level of the case here: females from all groups (thus including Lif elong Chicagoans) are more
171 bile females acquire the new dialect more completely than Mobile ma les (Or, again, Mobile females acquire Lifelong /, and Mobi le males acquire Lifelong Chicago females to tease these possibilities apart.) Analysis of F2 Figur es 5 38 through 5 40 show scatter plots of the Lobanov normalized F2 values for th e Lifelong Jasper group, stays mostly steady with a slight fall at the end for the Mobile group, and drifts slowly upward until a slight fall from T4 to T5 for the Lifelong Chicago group. The movement is slight for all three groups, however. More notable is that the mean value of F2 at all time points is less for Lifelong Chicago than it is for Chicago group than it is for the Lifelong Jasper group. Residence Group Residence g roup was often but not always a significant predictor of the Lifelong Chicago F2 values were significantly lower than those of the Mobile group (i.e. Lifelong back). At T4, Lifel ong Chicago was significantly different from the Mobile group according to Lobanov and LobBark, but the difference was nonsignificant for Nearey. At T5, Lobanov showed Lifelong Chicago to be marginally significantly different from the Mobile group, but th e difference was nonsignificant for both LobBark and Nearey.
172 never significant a t T3 through T5. (Post hoc Tukey tests from the ANOVAs run on factors, the Lifelong Chicago and Lifelong Jasper groups differed significantly at all five time points. / / is farther back in Chicago than in Jasper. ) At the first four time points, then, we can be confident that an average Mobile speaker would not acquire Lifelong Chicago maintai n his native Jasper like production. At T5, it is plausible either that the speaker would completely maintain his J asper like production, or that he would completely acquire the Lifelong Chicago ma intena nce of native dialect production upon exposure to a new dialect in adulthood, i.e., lack of malleability of production. AAC, YLC, and PLC. AAC was never a significant predictor of the Mobile F2 values. N or were YLC and PLC ever significant for LobBark. For Lobanov and Nearey, however, YLC and PLC were marginally significant predictors of F2 at T3 and T4, and for Lobanov alone, YLC was marginal ly significant at T5 as well. At these points, as the number of years or percentage of life lived in Chicago increased, the F2 values decreased. That is, Mobile speakers who have lived more years or a greater Lifelong Chicagoans. (The correlations are not particularly strong, with r ranging from about .26 to about .39.) Thus, while the effects of residence group above showed that overall a Mobile group member is likely to maintain native dialect production
173 there seems to be a sligh t tendency to adopt the second dialect (D2) production upon increased exposure to the D2 Education and YOB. and then only for Nearey. At this point, more educated speakers had higher F2 valu es, but this finding is not corroborated by Lobanov or LobB ark. (The interaction between residence group and e ducation was nonsignificant.) YOB was a significant predictor of F2 at T4 and T5 for all three normalization methods, and marginally significant at T3 for Lobanov. The younger the speaker, the higher her F2 value in the second half of the vowel The inte raction between residence group and YOB was never significant. This would seem to indicate a wider Midwestern trend toward fronting is surprising including in Chicago according to McCarthy (2011). Since I only found marginally significant evidence of fronting at T3, and this marginal significance only appeared for one no rmalization method, one could possibly say that McCarthy is right that / / is moving farther back at This seems somewhat unlikely, however. My best guess is that there actually is an interaction between YOB and residence group, and / / is moving backward in Chicago and forward in Jasper, but the difference was not strong enough to reach statistical significance in my data. Sex and g ender. Gender was never a significant /, and s ex was only significant for Lobanov, and then only at T2 and T3. At these points, controlling for other factors, females had lower F2 than males; that is, they had more NCS like produc tion. The interaction between residence group and s ex never
174 approached significance. These findings should again be taken lightly because they showed up only for one normalization method. However, to th e extent that they are reliable and if we assume that / / backing is advancing in Chicago the findings given that females are leading in a change below the level of consciousness, and they indicate again that Mobile females would acquire the new dialect more completely than Mobile mal es (or that Mobile individual s match their production to that of the same sex group in acquiring a D2). Analysis of F3 Figures 5 41 through 5 43 show scatter plots of the Lobanov normalized F3 values roup. Lines connect the mean values at each time point. The values for Lifelong Jasper rise from T1 to T2 and then fall for the rest of the vowel. and then falls to T5. The mean F3 values for the Lifelong Chicago group rise a bit from T1 to T3 and then fall for the rest of the vowel. None of these changes are particularly dramatic, however. Residence g roup. Unsurprisingly, residence g roup was never a significant predictor of F3 at T3 through T5. However, at T1 and T2, the values were significantly lower than those of the Mobile group for all three normalization from each other. The ANOVAs run on the participan had shown a significant difference by Residence Group at T2, and post hoc Tukey tests showed that the Lifelong Chicago and Lifelong Jasper groups were significantly different from each other at this point. It appears that there are regiona and Mobile speakers would tend not to acquire the Lifelong Chicagoan production.
175 AAC, YLC, and PLC. AAC was never significantly correlated with the Mobile Lobanov and marginally significant for Nearey, and PLC was marginally significant for LobBark and Lobanov. The more years or larger percentage of their life Mobile speakers have spent in Chicago, the lower their F3 is at this point. Thus, there is some evidence that Mobile speakers can move toward acquiring a Lifelong Chicagoan F3 at this point, given enough exposure. YOB and e ducation. /. Only for Nearey, e ducation was marginally significant at T4; at this point, controlling for other factors, speakers with more education had a higher F3. (The interaction between edu cation and residence g roup did not approach significance.) Since this effect only appeared at one time point and only for one normalization method, I am again inclined to consider this a Type I error. Sex and g ender. At T1, s ex was significant for Lobano v and marginally significant for LobBark. Controlling for other factors, females had a higher F3 at this point, i.e., they had less NCS like production. If the regional variation in F3 is stable (since there is no effect of YOB), then this finding would given that females are using the NCS like, and thus more standard, form. For Lobanov, however, the interaction between residence group and s ex was margina lly significant: the effect of s ex was much smaller, in fact almost none xistent, for the Lifelong Chicago group. Thus it may be that only the Mobile and Lifelong Jasper females had less NCS like production at this point. This, then, does not support Principle I, but it does not directly contradict it, either, since the Lifel
176 NCS like than Lifelong Chicago males. In terms of gender, At T4, negative f emininity was significant for Nearey; more negatively feminine speakers had h igher F3 at this point. Since residence g roup was not a signi ficant predictor of F3 at T4, i t is impossible to say whether n egativ e f emininity increased or decreased NCS like production. Analysis of F1 Figures 5 44 through 5 46 chart the Lobanov norm resid ence g roup over time. Lines connect the mean value at each time point. The Lifelong Jasper and Mobile groups show a rise in F1 (lowering vowel) from T1 to T3 and T2 a nd falls thereafter. Residence g roup. At all five time points, residence g roup was a significant significant or marginally significant for Nearey. At all five time points Li felong Chicago lowered their Lifelong Chicago ans, but they have not completed the acquisition. Once again, we have evidence for malleability of production in adulthood, without complete acquisition of a second dialect. AAC, YLC, and PLC was marginally significant at T2 and T5 for Nearey, and significant at T5 for Lobanov and LobBark. At T5, the older the speaker was when he arrived in Chicago, the higher is. (These correlations are of moderate strength, with r
177 ranging from .370 to .440.) This is again contrary to the expectation that those who are exposed to a second dialect at a younger age should be more adept at second dialect acquisition; here, it is is more Chicago like. In contrast, at T2 for Nearey, the older the speaker was when he .348). This is mo re in line with what would be expected, but the correlation only appeared in one of the three normalization methods, so I do not feel confident that the correlation is reliable. YLC and PLC were never significantly correlated with F1 for Nearey, but they w ere both marginally significant or significant predictors of F1 at T4 and T5 for Lobanov and LobBark. The more years or greater percentage of her life a Mobile speaker has spent is at these points. Speake rs who have spent more time in Chicago have more Chicago support for the idea that with greater exposure to a second dialect, more complete D2 production acquisition is possible. Education Controlling for other factors, Education was consistently a significant or marginally significant predictor of F1 At all five time points, the more educated a speaker is, the lower his F1 (the higher his makes sense if we again assume that more educated speakers might be more resistant to the stigmatized N CS. The interaction between education and residence g roup was significant only for Nearey at T5 At this point, Lifelong Chicagoans hav e a strong er effect of e ducation than the other two groups; having more education makes Chicagoans retreat more sharply from the NCS. This likewise could indicate that
178 educated members of the group that may be most aware of the stigma associated with the NCS are most likely to avoid this nonstandard production. However, this finding should be taken with a grain of salt, since it appeared only for one normalization method and only at one time point, and even then was only marginally significant. YOB. Cont rolling for other factors, YOB was a significant predictor of F1 at T3 through T5. At these points, the younger a speaker is, the lower her F1 is. This would suggest a Midwestern trend of retreating from the NCS (or simply raising of the second half of / 16 T he interaction between YOB and residence g roup was never significant. Sex and g ender. Neither gender nor s ex was ever a significant predictor of F1 of ever, s ex was significant at T4 and T5, and p ositive m asculinity was marginally significant at T2. At T4 and T5, controlling for other factors, females had lower F1 than males, i.e., they had less NCS like production. In neither ca se was the interaction between residence group and s ex significant. If / / is generally raising, and if, as I believe, this is a change from below, we would expect ple II to apply, and females do indeed have higher / / (are more advanced in raising) than males. However note that these findings for sex only held f or one of the three normali zation methods. Returning to g ender, at T2 for Nearey, the more positively masculine a speaker was, the higher (and more NCS lik e) his F1. The interaction of residence group and positive m asculinity did not approach significan ce. It seems that males and more masculine speakers may be trailing raising 16 McCarthy (2011) did not measure F1 of / /, so comparison with her findings is not possible here.
179 Analysis of F2 Figures 5 47 through 5 49 present scatter plots of the Lobanov normal ized F2 roup. Lines connect the mean values at each time rest of the vowel. (In other words, the vowel moves farther back and then farther 2 values fall only from T1 to T2 and then rise for the rest of the vowel. Residence g roup. Similar ly mostly shows a lack of malleability of production in adulthood. At all five time points for all t significantly lower than those of the Mobile group (except only marginally significant at T1 for Lobanov). We can conclude that the average Mobile speaker would not fully adopt Chicag o that he would change from his native Jasper like production. The difference between Lifelong Jasper and Mobile F2 values was never significant at T3 through T5, and significant only for Lobanov at T2. At T1, the difference was significant for Lobanov and marginally significant for LobBark and Nearey. (Note that post hoc Tukey tests in the were signi ficantly different from each other at all five time points, so there is indeed a difference that the Mobile group could be acquiring.) If Mobile speakers move toward Chicago AAC, YLC, and PLC. AAC was never a significant pre contras t, YLC and PLC were significantly or marginally significantly correlated with F2 at all five time points for all three normalization methods. As the length of time or
180 percentage of life spent in Chicago increased, F2 values decreased. (The correlations were moderate or moderately strong, with r values ranging from around .38 to around .75.) Thus, although in general Mobile group members do not adopt the Chicagolike F2 values, th ere is evidence that those who have been there longer will begin to back Education and YOB. Education was never a significant predictor of F2 for LobBark or Nearey, but it was marginally significant at T4 and T5 for Lobanov. At these farther forward). This is similar to my other findings for e ducation above in that more educated speakers show less evidence of the NCS. (The interaction of educ ation and residence g roup was never significant.) YOB was significant or marginally significant at all five time points for all three normalization methods. At all five points, younger speakers tend to have higher F2 value s. The interaction of YOB and r esidence g roup was never significant. This would indicate a general Midwestern trend toward fronting of / /. In contrast, McCarthy (2011) found that / / was undergoing increased backing in findings with F2 of / / also apply here. Sex. for both Lobanov and LobBark; s ex was also significant or marginally significant at all time points except T1 for Nearey. Controlling for o ther factors, at all time points, females had like. The interaction between sex and residence g roup was only significant (or marginally so) at T2, for LobBark and Nearey. At this poin t, the effect of s ex was stronger for the Lifelong
181 Jasper group than for the other two. At least at this point, Lifelong Jasper females were less NCS like than males in their production to an even greater extent than the Mobile or Lifelong Chicago females supports Principle II: fe males are leading in the fronting which my data shows to be an ongoing change across the Midwest Gender. Gender was also often a significant predictor of F2. At T1, for Lobanov and Neare y, the more negatively feminine a speaker was, the higher her F2 values (i.e. less NCS like her production) was The increased F2 with increasing negative f emininity was also significant at T2 for LobBark and Nearey, while Lobanov showed that increased ne gative masculinity and positive f emininity both led to higher F2 values at T2. At T3, all three normalization methods showed that inc reased total f emininity led to higher F 2 values, and at T4, increased total femininity for Lobanov and positive f emininity for LobBark led to higher F2 values. Generally speaking, more feminine speakers tended to have less NCS like production. At T2, the interaction between negative femininity and residence g roup was marginally signifi cant for Nearey: the effect of negativ e f emininity was weaker for the Lifelo ng Chicago group. Also at T2, positive f emininity interacted significantly with residence g r oup for Lobanov: the effect of positive f emininity was stronger for the Lifelong Jasper group. At T3, the interaction betwee n residence group and total f emininity was marginally significant for both LobBark and Nearey. According to Lo bBark, the effect of total f emininity was stronger for Lifelong Jasper, and according to Nearey, the effect was actually reversed fo r Lifelong Ch icago increased total f emininity led to a lower F2. At T4, the interaction between residence group and t otal
182 femininity or positive f emininity was significant for Lobanov and LobBar k, respectively: the effect of total or positive f emininity was stronger for the Lifelong Jasper group than for the other two. It seems that the apparent reduction in NCS like production with increased femininity is due mostly to the Lifelong Jasper group, and there is some slight evidence that the effec t is reversed for the Lifelong Chicago group. Thus, Lifelong Jasperites, who would not be expected to participate in the NCS anyway, show a tiny bit of evidence that i ncreased femininity may lead to increased adoption of the NCS by Lifelong Chicagoan speakers. Analysis of F3 Figures 5 50 through 5 52 present scatter plots of the Lobanov normal ized F3 roup. Lines connect the m ean values at each time point. Figure 5 52 excludes one major outlier at T5 (normalized value of about 82) that must have been due to a typo. Unfortunately, I discovered this error too late to correct it in my analysis. 17 The Lifelong Jasper and Lifelong Chicago groups show a gradual rise in F3 from T1 to T4, followed by a slight fall to T5. (The Lifelong Chicago group seems to have a dramatic rise from T4 to T5, but this is due to the mistaken outlier. Excluding this outlier, the mean falls from T4 to T5.) The Mobile group stays mostly steady across the five time points. Residence g roup. Residence g led 17 Note that this outlier does not skew the overall normalized values; recall that all v alues were normalized with reference to the means and standard deviations at T3, so the other normalized T5 values for the T3 would have been greatl y skewed, and all the normalized values would have been inaccurate.
183 significant differen ces by residence g roup, either, so there is simply no evidence of phonetic production in adulthood. 18 Education and YOB. Interestingly, both e ducation and YOB were freq uently significant for this formant. Education was a significant or marginally significant predictor of F3 at T1 through T4 for all three normalization methods. Controlling for other factors, at these points, the more education a speaker has, the lower h er F3. YOB was at least marginally significant at T1, T3, and T4 for all three normalization methods, and at T5 for LobBark and Nearey. Controlling for other factors, at these points, the younger a speaker is, the higher her F3 value. 19 Sex and g ender. / (though s ex, it is worth noting, never was). At T5, controlling for other factors, positive and total m asculinity were significant for Lobano v and Nearey, respectively: as positive or total m asculinity increased, F3 at the end of the vowel decreased. However, the outlier mentioned earlier was at T5, and the participant with the outlier had the lowest overall total masculinity and second lowest positive m asculinity scores. I therefore reran the statistics at this point without t he outlier, and the effects of m asculinity disappeared. 20 Femininity seemed to have a real ef fect on F3, however. Positive f emininity was at least marginally significant at T2 for all three normalization methods, 18 Unsurprisingly, the correlations between AAC, YLC, and PLC and F3 were always nonsignificant as well. 19 I tested interactions between residence group and YOB, and residence group and edu cation, for Nearey only, and the interactions never approached significance. 20 The only other notable change in significance when I eliminated the outlier was that YOB was no longer marginally significant at T5 for LobBark.
184 and marginally significant for LobBark at T1 as well: as F+ increased, F3 early in the vowel decreased. F was also a significant predictor of F3 for all three normalization methods at T4: as F increased, F3 at that point decreased. Generally, it seems tha t Conclusions a bout the Malleability of Phonetic Production in Adulthood In this chapter, I address the question of how malleable phonetic production is in adulthood. (Chapter 4 dealt with the same ques tion for phonetic perception.) Overall, this chapter has provided evidence for malleability of production in adulthood. The Mobile group showed movement toward Chicago like production of both F1 and F2 of //; in both cases it is clear that the Mobile gr the Lifelong Jasper speakers, but their production also does not show as strong a of movement for the Mobile group. It was found that the average Mobile speaker would adjust her production of F1 to be closer to that of a Lifelong Chicagoan, but it is were similar: there is evidence that the average Mobile speaker would adjust his F1 to be more Chicago like, though still without full acquisition, but it is plausible that his F2 values would remain completely Jasper like. Overall, it seems that phonetic production is malleable in adulthoo d. Upon exposure to a new dialect, an adult may adjust her production to be more like that of her new peers, though it appears that nativelike attainment is unlikely. Rather, the average Mobile speaker is likely to adjust production to be more like that native speakers, but not completely adopt the new norm. These findings are similar to those of Harrington et al. (2000a, 2000b, and 2005), who found that Queen Elizabeth
185 in Southern Standard British (SSB), but she has not fully adjusted to the new SSB norms. The current data suggest that F1 may be more malleable in adulthood than F2, but this is not certain. Instead, for example, it may be that speakers are more likely to change their production in the dimension whose regional variation is less obvious. That is, the main regional differences in both / / and / / were expected to be found in F2 because the NCS typically involves backing of these vowel s However, this data showed that these vowels are also lower for Chicagoans than for Jasperites Perhaps Mobile group members who did not want to adopt the NCS had noticed the backing element and rejected it in their own production, but they acquired the subtler lo wering of the vowel without realizing it. This is just one possible alternative explanation for the greater malleability of F1 than F2 in this data set. One of the most surprising results of this study was that age of a rrival in Chicago (AAC) was so rarely dialect. This is surprising in light of the primacy that linguists have long given to age o f exposure to a second language (Lenneberg 1967) and, indeed, a second dialect. For example, recall from Chapter 2 that Bortoni (1991) found that even among adults, younger age of arrival in Brasilia predicted greater loss of stigmatized non Brasilia dialect features. One would thus expect that age of first exposure to the Chicago dialect would b e a significant predictor of Jasper to new dialect. Not only was this rarely the case, however, but when AAC was a significant predictor of their acquisition, the correlation was sometimes in the opposite direct ion from that predicted: those who arrived in Chicago later had adopted more
186 Chicago like productions. While this latter fact may be due to statistical anomalies, overall AAC was a less reliable predictor of second dialect acquisition than the literature would have led one to expect. Instead, in this study, the number of years lived in Chicago (YLC) and the percentage of life spent in Chicago (PLC) were more reliable predictors of adult second dialect acquisition in production. Apparently extended exposu re to a second dialect can trump age of first exposure in allowing adults to acquire a new d Relative Importance of Gender and Sex in Phonetic Production One goal of this study is to investig ate the relative importance of sex and g ender to sociolinguistics, not just in terms of second dialect acquisition, but to the field in general. Th is chapter has shown that both sex and g ender affect (or, at least, can be used to predict) phonetic production of vowels involved in regional variation and/or change. In line with decade s of sociolinguistic research, s ex was often a significant predictor of production. The findings for sex here generally principles at least when sex was a significant predictor at all More inte restingly, g ender was often found to be a significant predictor of production of these vowels. Perhaps the most noteworthy finding is that f emininity apparently has a more reli able effect on production than masculinity. While m asculinity occasionally app eared to have an effect on production, its effects were sporadic and inconsistent enough that the significant findings may have simply been statistical flukes (Type I errors). In contrast, when f emininity was a significant predictor of production of a giv en vowel formant, the significance was generally consistent and systematic. Overall, more feminine speakers exhibited less NCS like production. This effect was often strongest for the Lifelong Jasper group, which complicates interpretation, given
187 that we like production. As speculation, perhaps less feminine speakers are somehow more open to influence of non standard production from outside their home region. Alternatively, perhaps more feminin e Lifelong Jasper speakers are actually leading changes in progress in Southern Indiana that happen to involve movement in the opposite direction of the NCS. At this point, I can only speculate as to what generalizations may eventually be drawn regard ing the comparative effects of sex and g ender. Howe ver, this study has shown that sex and g ender can have separate and even contradictory effects on phonetic production. Given this finding, I suggest that sociolinguists administer the EPAQ to their particip ants in future studies. As more such studies are cond ucted, the distinction between sex and g ender effects can be further teased apart, and eventually gener alizations about the effect of g ender on sociolinguistic variation may be drawn. Based on the pres ent stud y, I tentatively conclude that f emininity has a gre ater effect on production than m asculinity. Further studies are needed to confirm this finding, however. In this chapter, I investigated the malleability of phonetic production in adulthood by mea like production of three raise and front her // so that it would be closer to the production typical of a L ifelong back in accordance with the NCS. The acquisition of NCS like vowel height would be
188 to overall, I found evidence of mild malleability of phonetic production in adulthood. In duction of a vowel, while in others, they may partially but probably not completely adopt a production typical of their new dialect region. This chapter also showed that gender as well as s ex can be an important predictor of production in cases of soci olinguistic variation. The present study makes it appear that f emininity plays a greater role in production than m asculinity, but further studies are needed to confirm this finding
189 Table 5 1 Number of tokens, by participant and vowel, extracted from t he first 15 minutes of interviews or conversations and total number of tokens extracted, when fewer than 15 Participant Vowel Tokens extracted from first 15 minutes of recording Total tokens extracted (if fewer than 15) Participant Vowel Tokens extracted from first 15 minutes of recording Total tokens extracted (if fewer than 15) 10 / / 2 53 / / 1 10 / / 1 61 // 4 14 11 // 4 61 / / 3 11 / / 4 61 / / 7 14 16 / / 1 68 / / 0 13 17 / / 6 73 // 5 27 / / 1 73 / / 6 12 28 / / 1 13 73 / / 6 28 / / 0 11 76 // 5 30 / / 2 12 76 / / 5 31 / / 1 76 / / 2 11 32 / / 1 81 // 5 35 / / 1 87 // 1 35 / / 1 87 / / 2 13 37 / / 1 88 / / 0 11 37 / / 1 88 / / 3 12 38 / / 3 89 // 1 42 // 2 42 / / 1 42 / / 2 Total // 27 Total / / 29 Total / / 36
190 Figure 5 1. An initial vowel boundary not placed at a zero crossing on the waveform Figure 5 2. Initial vowel boundary placed at zero crossing after end of stop burst because formants appeared during burst in the word dad
191 Figure 5 3. Initial vowel boundary placed at zero crossing near onset of full formants after an aspirated stop in the word pa st Figure 5 4. Initial vowel boundary marked at the zero crossing between the end of the fricative and the beginning of the full formant structure in the word fast
192 Figure 5 5. Initial vowel boundary marked at the zero crossing after the pulse with upward sloping formants in the word next Figure 5 6. Final vowel boundary marked at zero crossing near end of full formant structure in the word back
193 Figure 5 7. Final vowel bo undary marked at the end of the formant structure and the beginning of the fricative, well after the end of voicing, in the word gas (The blue line indicates presence of voicing.) Figure 5 8. Final vowel boundary marked at the zero crossing near the d isappearance of formants in the word love
194 Figure 5 9. Final vowel boundary marked at point of noticeable change in vowel quality due to lack of audible/visible fricative in the word other Table 5 2. Number of participants by sex found to be best analy zed at each number of formants for LPC analysis # of f ormants for a nalysis # of f emales # of m ales Total # of p articipants 4 7 7 5 35 18 53 6 1 15 16 Total 43 33 76
195 Figure 5 10. Improperly tracked formants in the word back. The red dots show that Praat is tracking an imaginary formant between the real F1 and F2 Table 5 3. Social factors included in mixed effects regression analysis, by normalization method and vowel categ ory Normalization m ethod Vowel c ategory Social facto rs i ncluded Lobanov // Residence group, sex, YOB, education, total m asculinit y, total f emininity Residence group, sex, YOB, total masculinity, total f emininity Residence group, sex, YOB, education, total masculinity, total f emininity LobBark // Residence group, sex, YOB, education, total masculinity, total f emininity Residence group, sex, YOB, total f emininity Residence group, sex, YOB, education, total f emininity Nearey // Residence group, sex, YOB, education, total masculinit y, total f emininity Residence group, sex, YOB, education, total f emininity Residence group, sex, YOB, education, total masculinity, total f emininity
196 Figure 5 11 /) 21 21 In all these figures, the axes ied for this small of an error. Please
197 Figure 5 12. Vowel space for the Lifelong Jasper group at T2 Figure 5 13. Vowel space for the Lifelong Jasper group at T3
198 Figure 5 14. Vowel space for the Lifelong Jasper group at T4 Figure 5 15. Vowel space for the Lifelong Jaspe r group at T5
199 Figure 5 16. Vowel space for the Mobile group at T1 Figure 5 17. Vowel space for the Mobile group at T2
200 Figure 5 18. Vowel space for the Mobile group at T3 Figure 5 19. Vowel space for the Mobile group at T4
201 Figure 5 20. Vowel space for the Mobile group at T5 Figure 5 21. Vowel space for the Lifelong Chicago group at T1
202 Figure 5 22. Vowel space for the Lifelong Chicago group at T2 Figure 5 23. Vowel space for the Lifelong Chicago group at T3
203 Figure 5 24. Vowel space for the Lifelong Chicago group at T4 Figure 5 25. Vowel space for the Lifelong Chicago group at T5
204 Figure 5 26. Scatter plot of Lobanov normalized F1 values for // for the Lifelong Jasper participants. The lines connect the mean values at each time point Figure 5 27. Scatter plot of Lobanov normalized F1 values for // for Mobile participants at each time point
205 Figure 5 28. Scatter plot of Lobanov normalized F1 values of // fo r Lifelong Chicago participants at each time point Figure 5 29. Scatter plot of Lobanov normalized F2 values for // f or the Lifelong Jasper group at e ach time point
206 Figure 5 30. Scatter plot of Lobanov normalized F2 values of // for the Mobile group at each time point Figure 5 31 Scatter plot of Lobanov normalized F2 values of // for the Lifelong Chicago group at each time point
207 Figure 5 32. S catter plot showing the Lobanov normalized F3 values for // for the Lifelong Jasper participants at each time point Figure 5 33. S catter plot showing the Lobanov normalized F3 values for // for the Mobile participants at each time point
208 Figure 5 34. S catter plot showing the Lobanov nor m alized F3 values for // for th e Lifelong Chicago participants at each time poi nt Figure 5 35. Scatter plot showing the Lobanov Jasper participants at each time point
209 Figure 5 36. Scatter plot showing the Lobanov participants at each time point Figure 5 37. Scatter plot showing the Lobanov Chicago participants at each time point
2 10 Figure 5 38. Scatter plot showing Lobanov Jasper group at each time point Figure 5 39. Scatter plot showing Lobanov participants at each time point
211 Figure 5 40. Scatter plot showing Lobanov Chicago participants at each time point Figure 5 41. Scatter plot showing Lobanov Jasper group at each time point
212 Figure 5 42. Scatter plot showing Lobanov group at each time point Figure 5 43. Sc atter plot showing Lobanov C hicago group at each time point
213 Figure 5 44. Scatter plot showing the Lobanov Lifelong Jasper group at each time point Figure 5 45. Scatter plot showing the Lobanov Mobile group at each time point
214 Figure 5 46. Scatter plot showing Lobanov C hicago group at each time point Figure 5 47. Scatter plot showing the Lobanov Lifelong Jasper group at each time point
215 Figure 5 48. Scatter plot showing the Lobanov Mobile group at each time point Figure 5 49. Scatter plot showing the Lobanov Lifelong C hicago group at each time point
216 Figure 5 50 Scatter plot showing the Lobanov / for the Lifelong Jasper group at each time point Figure 5 51. Scatter plot showing t he Lobanov Mobile group at each time point
217 Figure 5 52. Scatter plot showing the Lobanov Lifelong Chicago group at each time point. One outlier at T5 is not shown on the chart bu t is calculated in the mean
218 CHAPTER 6 COMPARISON OF PERCEPTION AND PRODUCTION Chapter 4 presented the results of the perception experiment alone, and Chapter 5 presented the production results from the sociolinguistic interviews. However, the crux of this study lies in the comparison of perception and production. Which is more mal leable in adulthood, perception or production? This chapter addresses the main research question of the study. Recall from Chapter 4 that the algorithm used to create the perception stimuli in Praat was designed to give each vowel a steady state; this ste ady state is what was measured for the perception comparisons reported in Chapter 4. The steady state was located mostly over the second half of the synthesized vowel. Because the measurements for the perception stimuli came from the second half of the v owel, the comparison of perception to production was carried out with reference only to the production findings at time points T3 and T4. 1 Comparison of Perception and Production of NCS / / Raising Statistical Comparison Overall, Chapter 4 revealed no mall eability in adulthood for perception of Northern Cities Shift (NCS) / / raising. For F1, F2, and the difference between F2 and ignificantly different from the Lifelong Jasper / that a 1 I initially included T5 as well, but analyzing results at this time point visually proved difficult because e end of the vowel, presumably due to coarticulatory effects. (In Chapter 5, coarticulatory effects were controlled for by including the place of articulation of the following segment in the statistical tests. This type of control is simply more difficul t in an analysis based on visual inspection of the data as performed in this chapter.)
219 Mobile listener could adapt to, but that the average Mobile listener would plausibly retain her native Jasper like perc eption. (It is not plausible that she would adopt Chicago like perception.) / raising, on the other hand, established that production is malleable in adulthood. At T3 and T4, it was shown that the average Mobile speaker would ad just his production in both F1 and F2 to be more like that of a Lifelong Chicagoan, but would not fully acquire the NCS production. In summary, for NCS / / raising, this study has found evidence of malleability in adulthood for production but not percept ion. This vowel, at least, makes it appear that adults can generally adopt (partially) the production of a new dialect in adulthood, but their perception will tend to remain first dialect like. It follows that for an individual speaker, production would acquisition of a second dialect in adulthood. This finding supports some of the results cited in Chapter 2. For example, Janson ed of a change than their perception had, and Evans and Iverson (2007) found that their participants changed their production but not their perception of phonetic and phonological features when they moved to a new dialect region. Behavior While the above generalizations are useful for determining whether production or perception is more malleable overall in adulthood, they do not address another of my research questions, whether the relationship between perception and production i s shift in perception and production of / /.
220 perceptual boundary from the perceptio n experiment on the same vowel space. To do this, I used the LobBark method to normalize the endpoints of my perception continua. I with these normalized values. For e xample, the raw F1 values for the endpoints of the Bad~Bed continuum were 6.19 and 4.62 Barks, respectively. The LobBark normalized values for these endpoints were .963 and crossover point value was 5.4 Barks, wh ich is (5.4 6.19)/(4.62 6.19) = 50.32% of the way between the F1 values for the Bad and Bed F1 value for the perception crossover point was therefore 50.32% of the way between the normalized endpoint values, i.e., .1 21. Using these normalized perception values, I normalized production tokens and perception crossover point. Once the normalized perception and production were graphed on the same vowel charts, I cou Examples of the possible comparison results are shown with hypothetical (very clean) data in Figures 6 1 through 6 3. In Figure 6 s placed between his production of / / and / /. In this example, the perception and production can be assumed to either not be shifted or to have shifted about equally. (If the two were shifted equally, the space between the / / production tokens and the perceptual boundary would presumably be smaller than the space between the perceptual boundary and the / / production tokens, if the perceptual boundary was originally
221 F igure 6 / production, while the / / production is still quite separated from / / production. In this example, the Finally, in Fig ure 6 3, the production of / / is much closer to the production of / /, and the perceptual boundary is in the midst of the / / production. In this case, the However, there are two complications to the analysis. The first is that, of course, real data is never this clean. In reality, the production of the two vowels overlapped more, and many participants had two or three outliers in production that made the analysis more difficult. More importa ntly, the second complication is that / / is actually a moving target. The examples above assumed that only / / was shifting in production. In fact, Chapter 5 showed that at T3 and T4 (the time points investigated here), the / production is undergoing lowering. Thus, their / / is actually moving towards their / / in height. Thus, in a case like Figure 6 2, it could be the case that the perceptual boundary has not shifted, but the production of / / has Because of this complication, it was generally difficult to determine whether perception had actually shifted ahead of production, or whether perception and production had shifted equally/not at all, and / / had simply lowered. ction comparisons at T3 and T4 are shown in Appendix I. perception and production of / / raising is reported in Table 6 1.
222 The varied relationships between perception and production of / / raising shown by Mobile participants undergoing language change in adulthood are summarized in Table 6 2. Three assumptions made in this analysis must be acknowledged. The first assumption is that those Mobile participants with production and/or perception that is now at least somewhat Chicago like originally had production/perception that was similar to the average for Lifelong Jasper participants. In reality, of course, it is possible that at least some of the Mobile participants actually had Chicago like perception and production even before they moved to Chicago. However, for the purposes of this study, there is no way to avoid relying upon this assumption. The only solution would, of course, be to collect longitudinal data. This is a necessary step for further resea rch but not feasible for the current investigation. The second assumption is that Mobile participants whose production and/or perception has shifted have changed their language systems to be more rather than less Chicago like. That is, I assume that if a more Chicago like than her production, she has shifted her perception to be more Chicago like, rather than shifting her production to be more Jasper like. It is possible that there are exceptions to this assumption; perhaps a participant disliked the Chicago accent so much as to adjust her production away from the Chicago norm. However, I note that 25 of my 26 Mobile participants expressed (often extremely) positive feelings toward Chicago and generally seemed to pre fer life in Chicago to life in Jasper, at least for the next few years. If they embrace the Chicago accent as part of embracing the
223 Chicago lifestyle as a whole, then it is reasonable to assume that they would not shift their speech away from the Chicago norm. However, as with the first assumption, I would need longitudinal data in order to test this assumption. perceptual boundary was between (or on the border of) their production of / / and / /. (That is, their perception and production boundaries matched.) Unlike the first two assumptions described above, this assumption could be indirectly tested, by comparing the perception and production of Lifelong Jasper participants. If t he individual Lifelong originally lain on the production border as well. As I had with the M obile participants, I individually investigated the perceptual boundaries and production of each Lifelong Jasper participant. 2 The results are summarized in Table 6 3. Table 6 4 summarizes the apparent relationships between perception and production of / / raising for the Lifelong Jasper participants. This table makes it clear that my third assumption is violated. Over two thirds of the Lifelong Jasper participants do not have a perceptual boundary that lies near their production boundary for / / and / /. I therefore cannot assume that individual Mobile participants started out with their perception and production boundaries in similar positions, either. Thus, when I say that, e.g., Mobile Participant 42 shows perception shifted more than production in a cquisition 2 Chapter 4 explained that I was unable to find perceptual boundaries for Lifelong Jasper Participants 2 and 56. Those participants are therefore excluded from this ana lysis.
224 of a new dialect, it is possible instead that he simply originally had a perceptual boundary that was raised above his production boundary for the two vowels. My next step was thus to see whether there were any patterns that would allow me to p production for these vowels would be. To test this, I compared various social characteristics of the participants to their production/perception results. Results of these com parisons are presented in Tables 6 5 through 6 8. Table 6 8 shows that YOB can be helpful in predicting the relationship of perception and production of / / and / / for Lifelong Jasper participants. Those born before 1980 tend to have perception that is m ore Chicago like than their production, while those born later are more likely to have either matched perception and production boundaries or production that is more Chicago like than their perception. At first blush, both sex and positive femininity seem like they could be useful for differentiating the latter two possibilities in younger participants. Because sex is present at birth (when an infant is first exposed to language) but gender presumably develops later, 3 I decided that sex was the better opt ion, as shown in Table 6 9. Based on this comparison, I re perception and production upon exposure to the NCS in adulthood, this time assuming that Mobile participants born before 1980 had started out with pe rception that was more Chicago like than their production, male Mobile participants born in 1980 or later had 3 That is, I could be confident that my Mobile participants had had the same sex at the time they lived in Jasper, but their gender might have adjusted since they moved to Chicago. Following this path of reasoning could lead to another entire analy sis of the effects of gender, given that it is possible that investigation of this possibility is beyond the scope of this work.
225 started out with production more Chicago like than their perception, and female Mobile participants born in 1980 or later had started out with th eir perceptual and production boundaries matched. Given the many exceptions to this generalization among the Lifelong Jasper participants, of course, it is quite likely (almost certain) that these assumptions are wrong in at least a few cases, but what I present here constitutes my best guess as to how perception and production were originally related in the Mobile 10. The revised summary of which Mobile participants have which relati onship between perception and production is presented in Table 6 11. In Chapter 1, I posed the questions of whether production or perception would change first when an adult adopted a new dialect, which would change to a greater extent, and whether this re lationship is consistent across individuals. Technically, my data cannot tell me which of the two changed first in a given individual. That is, even if the data collected at a single point in time shows that production has shifted more than perception in one possible that his perception actually changed a bit first, but that once his production started changing, it overtook perception and ended up being more Chicago like. Once again, I would need longitudinal data to actu ally prove which aspect of the phonetic system changed first for a given individual. In contrast, my synchronic data can help me answer the second question, whether perception or production changes more upon adult exposure to a second dialect. T he curre nt data suggest that it is more common for production to change more than for perception to do so ; production seems to be more malleable in adulthood. However, in answer to the third question above it is apparent that this relationship is not consistent
226 across individuals. The question of what might deter mine which of the categories in Table 6 11 an i ndividual will fall into is addressed at the end of this chapter. Comparison of Perception and Production of NCS / / Backing Statistical Comparison In Chapt er 4, t / backing did not allow me to determine the malleability of adult perception for this vowel The multiple regressions showed no evidence of regional variation in perception on this continuum. On the o ther hand, ANOVAs had shown that the difference in perception by Lifelong plausible that the averag e Mobile listener would completely retain her native Jasper like perception, or that she would completely adopt Chicago like perception. Unlike the perception statistics, the production results allowed me to draw the conclusion that at least in F2, prod uction of this vowel was overall not very malleable in adulthood. At T3 and T4, the average Mobile speaker would definitely not acquire Chicago like production of F1; that is, the difference between Mobile and Lifelong Chicago production was statistically significant at these time points. The difference between Mobile and Lifelong Jasper F1 production was also statistically significant at T3 and T4. For F2, the formant in which a greater regional variation was expected, the significantly different from that of the Lifelong Chicago but not the Lifelong Jasper group at T3 and T4. Thus, while the average Mobile like in terms of vowel height, we have no evidence that it would change in terms of frontness. Given that this shift mostly involves backing of the vowel (Eckert 1991) it seems that this vowel overall
227 shows a lack of malleability of production in adulthood. However, the inconclusive perception results above prevent comparison of perception and production based on statistics. As with the investigation of / / raising, I examined each individual Mobile f / / backing. Chapter 5 showed that the Mobile group seems to be changing the height of / / more than its frontness, but the perception experiment tested mostly perception of F2 differences, so this comparison will necessarily focus more on the adoption of / / backing, rather than / / lowering. Vowel charts showing each Mobile and Lifelong / backing can be found in Appendix I. carried out in the same way as the comparison for / / raising above. LobBark normalized production data and perception crossover points were graphed on the same vowel chart, and conclusions were drawn based on the placement of the perception crossover in relation to the two groups of vowel tokens. Happily, this analysis was slightly less complicated than the analysis of / / raising because / / is not undergoing fronting. Recall that I stated above that the analysis of / / raising was complicated by the f act that / / was a moving target; while Mobile participants raised / / toward / /, they also lowered / / toward / /, and I had to take this into account when inspecting the vowel charts. On the other hand, while Mobile participants lower / / in production this change
228 does not move / / toward / /. This made the analysis of this vowel shift somewhat more straightforward, more like the hypothetical examples given in Figures 6 1 through 6 3. red in Table 6 12 4 The varied relationships between perception and production shown by Mobile participants undergoing language change in adulthood for / / are summarized in Table 6 13. The same three assumptions as were made for the analysis of / / raisin g were also made here. First, I assume that Mobile participants whose perception and/or production is now Chicago like was once Jasper like. Second, I assume that Mobile participants have only adjusted their language systems to make them more Chicago lik boundaries for this pair of vowels were originally located near their production boundaries. This last assumption is testable. As I had with the Mobile participants, I indi vidually investigated the perceptual and production boundaries of each Lifelong Jasper participant. The results are presented in full in Table 6 14 and summarized in Table 6 15. Table 6 15 makes it clear that, as it was with / /~/ /, my assumption that Mo bile participants would have started with matched production and perception boundaries between / / and / production/perception relationship to their social characteristics to see if there w ere 4 uction data for the / /~/ / comparison was less unruly than that for the / /~/ / comparison. Consequently, I am more confident of the conclusions drawn in Table 6 12, and it is
229 have been like. Results of these comparisons are given in Tables 6 16 through 6 19. Tables 6 16 through 6 19 show that there are almost no patterns to be found in Li or gender. Nor does YOB help differentiate; for both the younger and older groups, the plurality or majority of participants had production that was more Chicago li ke than their perception. Because social characteristics do not appear to be useful for determining was more Chicago like phonetic systems. Table 6 20 backing in light of this assumption. The revi sed summary of which Mobile participants have which relationship between perception and production is presented in Table 6 21. As with / / raising, the data for acquisition of NCS / / backing in adulthood shows that the relationship of perception and production in adult second dialect acquisition is not consistent across individuals. Unlike for the other vowel shift, however, the plurality or majority of Mobile participants had perception shifted more than production in acquisition of / / backing. Th e findings for this shift, then, contradict many of the findings cited in Chapter 2. There is thus no straightforward generalization that can be made regarding the relative malleability of phonetic perception and production in adulthood. While it would b e desirable to be able to say definitively that perception is
230 more malleable than production in adulthood, or vice versa, the present data show that such a neat conclusion would be inadequate. For th e close of this chapter, I wish to address the question of what qualities make it more likely for an individual to change perception or production to a greater extent, to shift them an equal amount, or not to change his phonetic system at all when exposed to a second dialect in adulthood. Many of the Mobile participants had different results for the two vowel continua presented in this chapter. For example, Participant 6 appeared to have shifted her perception and production equally for / / raising, but perception had shifted more than production for / / backing. For this analysis, I chose not to include these participants with conflicting results. There are two reasons for this decision: First, it is simply not legitimate to draw a conclusion that a pe rson with a certain social characteristics is likely to change perception more than production in 50% of the changes s he is exposed to and it would be even less legitimate to draw a more specific conclusion, saying that people with certain characteristics will have greater malleability of perception in cases of / / raising specifically and greater malleability of production in cases of / / backing specifically. The former conclusion simply seems indeterminate and unlikely, and the latter is specific to on e new dialect (the NCS), not generalizable, and thus not useful to linguists in general. (It is conceivable that certain people are likely to behave in x manner when acquiring changes primarily involving F1, and in y manner when acquiring changes primaril y involving F2, but one would need to examine many more than two vowel shifts to draw such conclusions.) The second reason I am not including participants with conflicting results is simply that the results may not be completely
231 accurate. Recall that for / / inconsistent enough that I was unable to feel confident of the conclusions I was drawing Consequently, it is possib le (though not certain) that the seemingly contradictory results are due to misanalysis of the data for one of the vowel shifts. When the results for the two vowel shifts match, I feel more confident in the accuracy of my analysis. 5 I therefore began by t allying the participants whose results were the same for both vowel shifts analyzed above. Happily, 15 of the 26 Mobile participants at least possibly match. These results are presented in Table 6 22. For the purposes of this analysis, I treated those pa rticipants whose relationships between perception and production for the two vowel shifts may match as matching. The exception to this is Participant 10, who could actually match in two different ways: no shift or production shifted more than perception. Because of this conflict, I dropped Participant 10 from this analysis. This left only one participant, Participant 8, in the no shift category, so I collapsed the unshifted and equally shifted participants into a single category. The revised categories and membership are shown in Table 6 23. I then investigated whether various sociolinguistic factors could be used to predict new dialect. I first investigated whether sex and/or gender would be a good predictor of this relationship, as shown in Tables 6 24 through 6 30. 5 Of c ourse it is possible that for some participants, I found matching results incorrectly. I would tend to really acquiring the NCS, not individual vowel changes), though, so I feel justified in having some confidence in those results that show consistent behavior by a given participant.
232 lated originally it does not make sense to use this variable to predict what a given person will do upon exposure to a new dialect in adulthood. Using apparent time, YOB is used as a proxy m is expected to reflect the system of the community as a whole at a given time; apparent time does not assume that people of different ages have different language systems because either evolution or fads have caused them to actually approach language dif ferently than their single generation, but the human faculty for acquiring language presumably does not. Thus, it would not make sense to assume that globally, people bor n before 1980 would tend to shift perception more upon exposure to a new dialect in adulthood, while those born later would shift production more. Instead of YOB, it might have been reasonable to examine Age of Arrival in Chicago (AAC). However, previous c hapters have shown that AAC is less important to D2 acquisition than the Number of Years Lived in Chicago (YLC), so I used the latter variable instead, as shown in Table 6 31. Finally, I considered whether education was a good predictor of the relationshi p between perception and production. This comparison is shown in Table 6 32. 6 6 I originally intended to run Chi square statistical analyses to determine which of these factors had significant effect s on the relationship between perception and production, but in each case, all of the cells had an expected count of less than 5, which makes the Chi square test inappropriate (Preacher 2001). Further, I artificially created the divisions among gender sco res (e.g., grouping those with Positive to run statistics. Consequently, I must rely on visual inspection of the trends to draw tentative conclusions about what social factors may affect how an adult acquires a new phonetic system.
233 The trends in this data do not indicate that sex is a reliable predictor of the relationship between perception and production when an adult acquires a D2. The males are nearl y equally distributed among not shifting/shifting equally, shifting perception more, and shifting production more. The majority of females shifted their perception more, but the distribution is not skewed strongly enough to call this a reliable trend. G ender may be a more reliable predictor. All three masculinity measures show that less masculine Mobile participants did not fall into the no shift/shifted equally group; either production or perception was shifted more. The negative masculinity tally sug gests that less negatively masculine individuals will specifically tend to shift production more than perception, while more negatively masculine individuals will tend to either shift perception more than production or shift them equally or not at all. Of the three femininity scores, positive femininity appears to be the best predictor of behavior: less positively feminine individuals tend to shift production more than perception, while more positively feminine people tend to either shift perception more o r shift the two aspects equally or not at all. YLC can in fact be used as a proxy to determine which changes first upon exposure to a D2, perception or production. Too few participants have lived in Chicago a long time for me to draw firm conclusions, but the trends in this data indicate the possibility that when someone first moves to a new dialect area, her perception will shift more than her production, but that once she has lived in the new area longer, her production will catch up to and possibly over take her perception. That is, the majority of those Mobile participants who have lived fewer than five years in Chicago have shifted
234 perception more than production, but that is true for none of those who have lived more than ten years in the city. Perha ps once the former group has lived in Chicago longer, their production will have shifted at least as much as their perception. be more likely than those with more education to shift their production more than their perception. In summary, gender seems to be a better predictor than sex of how an adult will negative masculinity and posit ive femininity may be the two best gender measures for this prediction. Education may prove to be a useful predictor, with the less educated 7 having a greater tendency to shift production more than their perception. To draw conclusions more certainly, a data set with more than 14 people would be necessary. This chapter has compared the perception and production of the Mobile group. I have shown that the relationship of perception and production upon exposure to a new dialect is not consistent across indi viduals, and I have made preliminary investigations into what social factors might be linked to what patterns of change in adulthood. In the final chapter, I will propose answers to my research questions and make suggestions for further research. 7 large; my Mobile group is simply very highly educated on average.
235 Figur e 6 1. Hypothetical example of comparison of production and perception showing either perception and production equally shifted or no shift (A = / /, E = / /, P = perceptual crossover point for Bad~Bed continuum) Figure 6 2. Hypothetical example of co mparison of perception and production showing perception shifting ahead of production
236 Figure 6 3. Hypothetical example of comparison of perception and production showing production shifting ahead of perception
237 Table 6 perception and production of / / raising 8 Mobile p articipant Relationship of perception and production 6 Shifted equally 8 No shift 9 Shifted equally, or perception shifted more than production 10 Shifted equally, or perception shifted more than production 12 No shift, or perception shifted more than production 16 Production shifted more than perception 18 Perception shifted more than production, or no shift 19 Perception shifted more than production 22 No shift 24 Production shifted more than perception 26 Production shifted more than perception 28 Perception shifted more than production 30 No shift, or shifted equally 32 Perception shifted more than production 33 Shifted equally 36 Production shifted more than perception 37 Shifted equally 38 Production shifted more than perception 39 Production shifted more than perception 42 Perception shifted more than production 46 Production shifted more than perception 70 No shift, or perception shifted more than production 71 No shift, or shifted equally 73 Production shifted more than perception 76 Shifted equally 78 Shifted equally, or production shifted more than perception Table 6 2. Relationships between perception and production in language change in adulthood for Mobile participants Relationship between perception and production Mobile participants Number of participants No change in adulthood 8, 22, possibly 12, possibly 18, possibly 30, possibly 70, possibly 71 2 7 Perception and production shifted equally 6, 33, 37, 76, possibly 9, possibly 10, possibly 30, possibly 71, possibly 78 4 9 Perception shifted more than production 19, 28, 32, 42, possibly 9, possibly 10, possibly 12, possibly 18, possibly 70 4 9 Production shifted more than perception 16, 24, 26, 36, 38, 39, 46, 73, possibly 78 8 9 8 The m essiness of real data made these conclusions comparatively difficult to draw. In the original on, but I decided to take those out for ease of reading. However, please interpret these conclusions as tentative; for many of the participants, the relationship between perception and production was not as easy to define as this table would lead one to b elieve.
238 Table 6 / / raising Lifelong Jasper participant Does perceptual boundary lie (mostly) on border between production of // and / /? If not, what would relationship appear to be? 1 N Perception shifted more than production 9 3 N Perception shifted more than production 4 N Perception shifted more than production 5 N Production shifted more than perception 49 N Perception shifted more than production 50 Y 51 N Production shifted more than perception 53 Y 54 N Perception shifted more than production 55 N Perception shifted more than production 57 N Production shifted more than perception 59 N Perception shifted more than production 61 N Perception shifted more than production 62 N Production shifted more than perception 63 N Production shifted more than perception 64 N Perception shifted more than production 67 N Perception shifted more than production 68 Y 69 N Production shifted more than perception 86 Y 9 I am not saying that Lifelong Jasperites are actually adopting the NCS. I simply mean that if these were Mobile participants instead, I would characterize their acquisition in the manner listed here. These phrases are meant only to describe how the perceptual boundaries and production tokens lie on the vowel charts.
239 Table 6 3. Continued Lifelong Jasper participant Does perceptual boundary lie (mostly) on border between production of / / and / /? If not, what would relationship appear to be? 87 N Perception shifted more than production 88 N Production shifted more than perception 89 Y 90 Y Table 6 4. Counts of agreement between production and perception of / / for Lifelong Jasper participants Agreement between perception and production Count Yes 6 No Perception more Chicago like than production 11 No Production more Chicago like than perception 7 Table 6 production of / / separated by sex Relationship between perception and production Males Females Perception and production boundaries matched 1 5 Production more Chicago like than perception 5 2 Perception more Chicago like than production 6 5 Table 6 perception and production of / / separated by masculinity scores Relationship between perception and production Total masc. 28 33 Total masc. 34 47 Positive masc. 14 23 Positive masc. 24 31 Negative masc. 8 11 Negative masc. 12 17 Perception and production boundaries matched 5 1 5 1 6 0 Production more Chicago like than perception 2 5 2 5 3 4 Perception more Chicago like than production 5 6 5 6 4 7
240 Table 6 7 elationship between perception and production of / / separated by femininity s cores Relationship between perception and production Total fem. 26 35 Total fem. 36 42 Positive fem. 17 22 Positive fem. 23 29 Negative fem. 6 11 Negative fem. 12 20 Perception and production boundaries matched 2 4 0 6 3 3 Production more Chicago like than perception 3 4 4 3 4 3 Perception more Chicago like than production 6 5 6 5 5 6 Table 6 production of / / separated by year of birth Relationship between perception and production YOB < 1980 YOB 1980 Perception and production boundaries matched 2 4 Production more Chicago like than perception 3 4 Perception more Chicago like than production 8 3 Table 6 9. between perception and production of // separated by sex Relationship between perception and production YOB 1980+, male YOB 1980+, female Perception and production boundaries matched 1 3 Production more Chicago like than perception 3 1 Perception more Chicago like than production 0 3
241 Table 6 / raising, based on assumptions about their original relationship. Mobile participant Presumed original relationship between perception and production Relationship between perception and production in adopting new dialect 6 Boundaries matched Shifted equally 8 Perception more Chicago like than production No shift, or shifted equally 9 Boundaries matched Shifted equally, or perception shifted more than production 10 Perception more Chicago like than production No shift, or production shifted more than perception 12 Perception more Chicago like than production No shift 16 Production more Chicago like than perception No shift, or production shifted more than perception 18 Boundaries matched Perception shifted more than production, or no shift 19 Boundaries matched Perception shifted more than production 22 Perception more Chicago like than production Production shifted more than perception 24 Perception more Chicago like than production Production shifted more than perception 26 Boundaries matched Production shifted more than perception 28 Production more Chicago like than perception Perception shifted more than production 30 Perception more Chicago like than production Production shifted more than perception, or no shift 32 Perception more Chicago like than production Production shifted m ore than perception 33 Perception more Chicago like than production Production shifted more than perception 36 Boundaries matched Production shifted more than perception 37 Perception more Chicago like than production Production shifted more than perception 38 Boundaries matched Production shifted more than perception 39 Perception more Chicago like than production Shifted equally, or production shifted more than perception 42 Perception more Chicago like than production No shift, or perception shifted more than production 46 Production more Chicago like than perception Shifted equally 70 Boundaries matched No shift, or perception shifted more than production 71 Production more Chicago like than perception Perception shifted more than producti on 73 Perception more Chicago like than production Production shifted more than perception 76 Perception more Chicago like than production Production shifted more than perception 78 Perception more Chicago like than production Production shifted more than perception
242 Table 6 11. Relationships between perception and production in language change in adulthood for / / raising for Mobile participants, revised Relationship between perception and production Mobile participants Number of participants Perhaps no shift 12, possibly 8, possibly 10, possibly 16, possibly 18, possibly 30, possibly 42, possibly 70 1 8 Perhaps equally shifted 6, 46, possibly 8, possibly 9, possibly 39 2 5 Perception shifted more than production 19, 28, 71, possibly 9, possibly 18, possibly 42, possibly 70 3 7 Production shifted more than perception 22, 24, 26, 32, 33, 36, 37, 38, 73, 76, 78, possibly 10, possibly 16, possibly 30 11 14 Table 6 / backing Mobile participant Relationship of perception and production 6 Production shifted more than perception, or no shift 8 Production shifted more than perception 9 No shift, or perception shifted more than production 10 Production shifted more than perception, or no shift 12 Production shifted more than perception, or no shift 16 No shift, or shifted equally 18 Perception shifted more than production 19 Production shifted more than perception, or shifted equally 22 Shifted equally, or no shift 24 Production shifted more than perception 26 Production shifted more than perception 28 Production shifted more than perception 30 No shift 32 Production shifted more than perception, or shifted equally 33 Production shifted more than perception 36 Shifted equally, or production shifted more than perception 37 Shifted equally 38 Production shifted more than perception 39 No shift 42 No shift 46 Production shifted more than perception 70 Shifted equally 71 No shift 73 Production shifted more than perception 76 Production shifted more than perception 78 Shifted equally
243 Table 6 13. Relationships between perception and production in language change of / / in adulthood for Mobile participants Relationship between perception and production Mobile participants Number of participants No change in adulthood 30, 39, 42, 71, possibly 6, possibly 9, possibly 10, possibly 12, possibly 16, possibly 22, 4 10 Perception and Production shifted equally 37, 70, 78, possibly 16, possibly 19, possibly 22, possibly 32, possibly 36 3 8 Perception shifted more than production 18, possibly 9 1 2 Production shifted more than perception 8, 24, 26, 28, 33, 38, 46, 73, 76, possibly 6, possibly 10, possibly 12, possibly 19, possibly 32, possibly 36 9 15
244 Table 6 / / Lifelong Jasper participant Does perceptual boundary lie (mostly) on border between production of / / and / /? If not, what would relationship appear to be? 1 N Perception shifted more than production 2 N Production shifted more than perception 3 N Merger of / / and / / 10 4 N Shifted equally 5 N Perception shifted more than production 49 N Perception shifted more than production 50 N Production shifted more than perception 51 N Production shifted more than perception 53 N Shifted equally 54 N Production shifted more than perception 55 N Production shifted more than perception 56 Y 57 N Production shifted more than perception 59 N Production shifted more than perception 61 N Production shifted more than perception 62 N Merger of / / and / / 63 N Production shifted more than perception 64 N Production shifted more than perception 67 Y 68 N Production shifted more than perception 69 Y 86 N Perception shifted more than production 87 Y 88 N Production shifted more than perception 89 Y 90 Y 10 Paired sample t normalized F2 values (at all 5 time points) showed the difference in frontness of / / and / / to be highly significant (p<.0005). The differences were also highly significant when I restricted the analyses to only the older speakers (born before 1980) or only the younger speakers (born 1980 or later). Thus, while individual Lifelong Jasperites may show signs of merging these two vowels in production, the community as a whole distinguishes them in production. Therefore, I feel confident that my Mobile participants began with these two vowels distinguished in production. They are not having to a cquire a new vowel distinction upon moving to Chicago. (The mean difference in F2 value was around 1.9, which should be of practical as well as statistical significance.) However, independent sample t showed that the difference in F2 values for the two vowels was never significant at T1 T5 for either of these two participants actually merge / / and / tests for Participant 90, and the difference in F2 values was highly significant (p<.005) at all 5 time points.
245 Table 6 15. Count of agreement between production and perception of / / for Lifelong Jasper participants Agreement between perception and production Count Yes 6 No Perception more Chicago like than production 4 No Production more Chicago like than perception 12 No Perception and production shifted equally 2 Possible merger 2 Table 6 relationship between perception and production of / / separated by sex Relationship between perception and production Males Females Perception and production boundaries matched 4 1 Production more Chicago like than perception 7 6 Perception more Chicago like than production 2 2 Perception and production shifted equally 0 2 Merger 0 2 Table 6 17. Relationship between perception and production Total masc. 28 33 Total masc. 34 47 Positive masc. 14 23 Positive masc. 24 31 Negative masc. 8 11 Negative masc. 12 17 Perception and production boundaries matched 1 4 3 2 0 5 Production more Chicago like than perception 7 6 7 6 7 6 Perception more Chicago like than production 4 0 3 1 4 0 Perception and production shifted equally 0 2 0 2 1 1 Merger 1 1 1 1 2 0
246 Table 6 18 p roduction of / separated by femininity s cores Relationship between perception and production Total fem. 26 35 Total fem. 36 42 Positive fem. 17 22 Positive fem. 23 29 Negative fem. 6 11 Negative fem. 12 20 Perception and production boundaries matched 3 2 4 1 2 3 Production more Chicago like than perception 6 7 5 8 8 5 Perception more Chicago like than production 2 2 1 3 1 3 Perception and production shifted equally 1 1 1 1 1 1 Merger 0 2 0 2 1 1 Table 6 19. Tally of Lifelong Jasper production of / / separated by year of birth Relationship between perception and production YOB < 1980 YOB 1980 Perception and production boundaries matched 3 2 Production more Chicago like than perception 8 5 Perception more Chicago like than production 0 4 Perception and production equally shifted 0 2 Merger 2 0
247 Table 6 /, assuming production was originally more Chicago like than perception Mobile participant Relationship between perception and production in adopting new dialect 6 Perception shifted more than production 8 No shift, or production shifted more than perception 9 Perception shifted more than production 10 No shift, or production shifted more than perception 12 Shifted equally, or perception shifted more than production 16 Perception shifted more than production 18 Perception shifted more than production 19 No shift, or production shifted more than perception 22 Perception shifted more than production 24 Production shifted more than perception, or no shift 26 No shift 28 Shifted equally, or perception shifted more than production 30 Perception shifted more than production 32 Perception shifted more than prod uction, or shifted equally 33 Production shifted more than perception, or shifted equally 36 Perception shifted more than production 37 Perception shifted more than production 38 Shifted equally, or production shifted more than perception 39 Shifted equally, or perception shifted more than production 42 Perception shifted more than production 46 Shifted equally, or perception shifted more than production 70 Perception shifted more than production 71 Perception shifted more than production 73 Shifted equally, or production shifted more than perception 76 Production shifted more than perception, or shifted equally 78 Perception shifted more than production
248 Table 6 21. Relationships between perception and production in language change in adulthood for / / backing for Mobile participants, revised Relationship between perception and production Mobile participants Number of participants Perhaps no shift 26, possibly 8, possibly 10, possibly 19, possibly 24 1 5 Perhaps equally shifted Possibly 12, possibly 28, possibly 32, possibly 33, possibly 38, possibly 39, possibly 46, possibly 73, possibly 76 0 9 Perception shifted more than production 6, 9, 16, 18, 22, 30, 36, 37, 42, 70, 71, 78, possibly 12, possibly 28, possibly 32, possibly 3 9, possibly 46 12 17 Production shifted more than perception Possibly 8, possibly 10, possibly 19, possibly 24, possibly 33, possibly 38, possibly 73, possibly 76 0 8 Table 6 22. Mobile participants tallied by relationship of perception and production in both / / raising and / / backing Relationship of perception and production Mobile participants Number of participants No shift Possibly 8, possibly 10 0 2 Shifted equally Possibly 39, possibly 46 0 2 Perception shifted more than production 71, possibly 9, possibly 18, possibly 28, possibly 42, possibly 70 1 6 Production shifted more than perception Possibly 10, possibly 24, possibly 33, possibly 38, possibly 73, possibly 76 0 6 Table 6 23. Mobile participants tallied by relationship of perception and production in both // backing, revised Relationship of perception and production Mobile participants Number of participants No shift or shifted equally 8, 39, 46 3 Perception shifted more than production 9, 18, 28, 42, 70, 71 6 Production shifted more than perception 24, 33, 38, 73, 76 5
249 Table 6 = production shifted more than perception) Count Relationship Sex Crosstabulation Sex Total F M Relationship No or equal 1 2 3 Perc Mor 4 2 6 Prod Mor 2 3 5 Total 7 7 14 Table 6 acquisition of a D2, separated by total masculinity Count Relationship Mtotalrange Crosstabulation Mtotalrange Total 27 34 37 50 Relationship No or equal 0 3 3 Perc Mor 3 3 6 Prod Mor 3 2 5 Total 6 8 14 Table 6 acquisition of a D2, separated by positive masculinity Count Relationship Mposrange Crosstabulation Mposrange Total 15 21 24 30 Relationship No or equal 0 3 3 Perc Mor 2 4 6 Prod Mor 3 2 5 Total 5 9 14
250 Table 6 acquisition of a D2, separated by negative masculinity Count Relationship Mnegrange Crosstabulation Mnegrange Total 13 22 8 11 Relationship No or equal 3 0 3 Perc Mor 4 2 6 Prod Mor 1 4 5 Total 8 6 14 Table 6 acquisition of a D2, separated by total femininity Count Relationship Ftotalrange Crosstabulation Ftotalrange Total 20 34 37 47 Relationship No or equal 1 2 3 Perc Mor 2 4 6 Prod Mor 3 2 5 Total 6 8 14 Table 6 production in adult acquisition of a D2, separated by positive femininity Count Relationship Fposrange Crosstabulation Fposrange Total 12 22 23 30 Relationship No or equal 0 3 3 Perc Mor 2 4 6 Prod Mor 5 0 5 Total 7 7 14 Table 6 30. Mobile acquisition of a D2, separated by negative femininity Count Relationship Fnegrange Crosstabulation Fnegrange Total 12 17 7 10 Relationship No or equal 1 2 3 Perc Mor 4 2 6 Prod Mor 3 2 5 Total 8 6 14
251 Table 6 acquisition of a D2, separated by years lived in Chicago Count Relationship NoYrsChicagorange Crosstabulation NoYrsChicagorange Total 0 5 11+ 6 10 Relationship No or equal 1 2 0 3 Perc Mor 5 0 1 6 Prod Mor 2 2 1 5 Total 8 4 2 14 Table 6 acquisition of a D2, separated by education (13 = in first year of college, 16 = Count Relationship Education Crosstabulation Education Total 13 16 17 18 Relationship No or equal 0 1 0 2 3 Perc Mor 1 2 2 1 6 Prod Mor 0 4 0 1 5 Total 1 7 2 4 14
252 CHAPTER 7 CONCLUSIONS system in perception and production. Specifically, I compared the perception and production of three groups: the Lifelong Jasper group, who had lived (nearly) all their lives in Jasper, Indiana; the Lifelong Chicago group, who had lived (nearly) all their lives in Chicago, Illinois; and the Mobile group, who grew up in Jasper and moved to Chicago in adulthood. I compared the (NCS), and to deduce whether perception or production was more malleable in adulthood. Using the Extended Persona l Attributes Questionnaire (EPAQ), I also investigated the relative importance of sex and gender to language change in the individual. The Research Questions Revisited In Chapter 1, I presented the research questions driving this study. I now repeat these questions and provide some answers and discussion. How and to What Extent Does Language Change with Exposure to a New Dialect in Adulthood? Chapter 4 indicated that perception, on the whole, is not particularly malleable in adulthood, though some individu al adults do shift their perception upon exposure to a new dialect. The Bad~Bed continuum provided evidence for a lack of malleability in adulthood, while the Dead~Dud continuum was ambiguous; statistics showed that the Mobile group could either have rema ined completely Jasper like in their perception, or completely adopted the Chicago norm. Investigation of individual hearers indicated that
253 the majority of Mobile participants had retained Jasper like perception, but others had shifted to become (more) Ch icago like. Chapter 5 presented evidence of mild or moderate malleability of phonetic production in adulthood. For all three vowels measured (/ /, / /, and / /), Mobile speakers were found to partially but not completely adopt Chicago like production. Th at is, they moved their production in the direction of, but did not reach, the Chicago norm. change upon exposure to a new dialect in adulthood. The examinations of perception a nd production individually indicate that production may shift more than perception. However, Chapter 6 showed that this is only a general tendency and not true for all individuals. When an Adult Moves to a New Region and Acquires a New Dialect, does Produ ction or Perception Change First? Which Changes More? Are these Relationships Consistent across Individuals? The third of these questions should be answered first: Chapter 6 showed that the relationship between perception and production upon exposure to a new dialect in adulthood is not and perception showed that some individuals shifted perception more than production, others shifted production more than perception, and still others ei ther shifted them equally or did not adopt the NCS at all. Thus, to answer the second question, while previous chapters had suggested that production may generally shift more than this is not always the case. Finally, for the first question: Chapter 6 provided slight evidence that perception may tend to change first upon exposure to a new dialect in adulthood, but that
254 production may then change more later. That is, participants who had spent less time in Chicago tended to have changed their perception more than their production, while those who had lived there longer tended to have shifted production more or changed both aspects equally. However, longitudinal data is necessary t o confirm which actually changes first. (It may well be the case that the Mobile participants who had spent the least amount of time in Chicago happened to be prone to changing perception more overall.) Which is a Better Predictor of Adult Language Change in the Individual, Age of First Exposure to the New Dialect or Length of Exposure? This study provided clear evidence that length of exposure to a new dialect is more important than age of first exposure in determining whether (or to what extent) an indiv Chicago (AAC) was never a significant predictor of change in perception, while both number of years lived in Chicago (YLC) and percentage of life spent in Chicago (PLC) wer e occasionally significant predictors of perception, albeit not once other factors were controlled for. For the production analysis, YLC and PLC were significant predictors (including after controlling for other factors) much more often than AAC was. It seems that the primacy linguists have given to age of exposure to a second dialect (D2) or language (L2) is somewhat misplaced. In fact, this factor did not play much of a role in this study. At least in terms of acquisition of a phonetic system, durati on of exposure is more important. This should be heartening news for adult second language learners, who may have feared that they had no hope of ever acquiring the sound system of their L2. The present study indicates that with
255 perseverance, they should vowels more like a native speaker would. Is Gender or Binary Biological Sex a Better Predictor of Language Change in Adulthood? The results on this question were mixed. Chapter 4 indicated that sex w as a better predictor of perceptual change in adulthood than gender, though even sex was rarely significant. Chapter 5, on the other hand, showed that both of these measures ed that gender is probably a better predictor than sex of the relationship between perception and production when an adult acquires a D2. Perhaps the most interesting finding related to this question was that femininity seems to relate to phonetic product ion more strongly than masculinity does. It may be linguistic behavior. While this study has not definitively ranked the importance of sex and gender for language, it has shown that gender plays a role in language behavior, and that standardized tests of gender, with scores generalizable across many different communities, can be useful in drawing conclusions about linguistic behavior. Limitations of the Study The greatest l imitation of this study is the lack of longitudinal data. This prevented me from satisfactorily answering my question of whether perception or production changes first upon exposure to a D2 in adulthood. More importantly, not having documentation of Mobi of my conclusions on speculation regarding their previous relationship between
256 perception and production. It would be much better to have snapshots of the Mobile r a period of years so that I could be confident that the conclusions I am drawing about whether perception or production has changed more are accurate. Another limitation is the relative homogeneity of my participant pool, especially the Mobile participan ts. In particular, most of my participants were highly educated, and my Mobile participants were more highly educated than the other two groups. All of my of the s tudy, while seven Lifelong Chicagoans and eight Lifelong Jasperites had informative to compare the behavior of the highly educated Mobile participants to mobile people who had less education. (However, it may simply be that people who do not pursue higher education are also unlikely to move from a small town to a big city. It is unclear whether the homogeneity of my Mobile group is due to my inability to access potential Mobi le participants with less education, or due to their nonexistence.) The study is similarly limited by the relatively short amount of time that most of the Mobile participants had spent in Chicago. 15 of the 26 Mobile participants had lived 5 years or less in Chicago, though there were also five who had lived there for more than 30 years. It would have been preferable to have had more participants who had lived a moderate and large number of years in Chicago, so that the effect of YLC and PLC could be bett er measured. A fourth limitation of the study is that the perception stimuli varied only along a single line, rather than as a matrix. More information could have been gleaned if, e.g,
257 the Dead~Dud test matrix had had three stimuli that all had F1 of 5.47 Barks but had F2 of 11.09, 11.28, and 11.46 Barks, respectively; three stimuli with F1 of 5.39 Barks but F2 of 11.28, 11.46, and 11.63 Barks, respectively; etc. This would have allowed me to better differentiate the effects of F1 and F2 on perception of each pair of vowels. Since each F1 value was paired with exactly one F2 value on my continua, it was impossible to tell whether listeners were relying upon vowel height or frontness more in deciding which vowel they heard. Finally, the conclusions about g ender are limited by the fact that the gender test used in this study was published in 1979, thirty years before I collected the data for this study. It is safe to say that American gender norms changed from 1979 to 2009. It would be preferable to find a standardized gender test that was developed more recently to ensure that the results reflect 21 st century gender norms. Suggestions for Further Research The most important suggestion for further research is that longitudinal data be collected from individ uals acquiring a second dialect in adulthood. This will allow direct the researchers to accurately document the extent to which perception and production have changed over time for each person. The current study relied upon assumptions the need to rely on assumptions. As implied by the limitations section above, I also suggest that future research use perception matrices rather than perception continua for vowels, with F1 and F2 varied crossover points more precisely and differentiate the effects of F1 and F2 on perception.
258 Further, I suggest that more sociolinguists administer the EPAQ (or a more modern gender test) to their participants. It takes less than ten minutes (usually less than five) for someone to complete this questionnaire, and this stu dy has shown that gender, as measured by the EPAQ, is a good predictor of linguistic behavior. The present study has suggested that femininity may be a more important predictor than masculinity. Further research can confirm (or disprove) this finding. Fi nally, it would be a good idea to conduct a similar (but longitudinal) study collecting perception and production data from individuals acquiring an incoming change in their own native community, as opposed to a new dialect. Such a study would be challeng ing to plan simply because it is generally difficult to identify which incipient changes will take root in a community, but if this could be achieved, then conclusions about the roles of perception and production in ongoing language change in a community c ould be drawn. Broader Implications In Chapter 1, I stated that the present study had implications for the fields of sociophonetics, language change across the lifespan, and second language acquisition. The implications for sociophonetics may be more methodological than theoretical. First, the investigation of production data in this study demonstrates the importance of measuring vowel formant trajectories, rather than simply measuring formants at the inve stigation in this study indicated that it is possible for a vowel shift to be acquired, e.g., more completely at the offset of the vowel than the onset. Measuring vowel formants at fewer points in the vowel can cause researchers to miss changes that actua
259 production and perception. Several other methods of comparison were tested before I decided that the method used here, a holistic inspection of production tokens and per ceptual boundaries graphed on the same chart, was the most effective. Future researchers may adopt and refine this method of comparison. Next, this study has given further evidence for language change across the lifespan. While this study is not based on longitudinal data, I concluded that some Mobile individuals had changed their phonetic systems in adulthood in both perception and production. To my knowledge, previous studies of language change across the lifespan have focused only on production. This study has thus broadened the scope of lifespan change research by showing that perception can also shift in adulthood, though it may be less prone to change than production. The findings for production support previous conclusions that production can shi ft in the direction of a community change provides further evidence that, as Sankoff and Blondeau (2007) noted, studies based on apparent time underestimate the rate of change in a community. Finally, for second language acquisition, perhaps the most striking finding of this study was that while perception may change first in adult second dialect acquisition, it appears that production ultimately changes more This conclusion was of course drawn based on acquisition of a vowel shift; it may not hold for cases of vowel merger or split. The analogous situation in terms of second language acquisition, then, would be acquisition of a vowel in the L2 that the learner men tally equates with a similar vowel in his L1. For example, an L1 English speaker learning L2 German may equate German / u/ with English /au/. In the acquisition of German / u/, then, we might expect the L2
260 duction to ultimately match that of a native Closing This research has shown that perception and production can both be malleable in adulthood. It has further shown that the relationship between perception a nd production in adult acquisition of a second dialect is not consistent across individuals. The relationship between perception and production in language change in the individual is more complex than it may initially appear, and further studies should p lumb the causes of this complexity.
261 APPENDIX A PARTICIPANTS
262 Table A 1. Lifelong Jasper participants Participant YOB Sex M+ M MT otal F+ F FT otal Education Occupation Age of arrival in Jasper 1 1980 M 24 9 33 21 7 28 Bachelor's Computer scientist 2 (from Vincennes, IN) 2 1982 M 20 15 35 19 10 29 Bachelor's Pilot Birth 3 1971 F 21 10 31 29 13 42 One semester college yoga instructor 1 (from Paoli, IN) 4 1981 F 27 8 35 26 10 36 Bachelor's Marketer Birth 5 1980 M 20 8 28 29 13 42 Bachelor's Tutor, remedial teacher Birth 49 1983 F 21 11 32 23 12 35 Master's Service writer at car dealership Birth 50 1981 F 27 15 42 24 7 31 Master's high school principal Birth 51 1981 F 20 11 31 26 12 38 Bachelor's customer service rep Birth 53 1986 F 24 13 37 20 14 34 Bachelor's unemployed (special ed teacher) Birth 54 1981 F 20 8 28 25 8 33 Some college Health aid? Birth 55 1981 F 23 8 31 29 11 40 Bachelor's Dietician Birth 56 1955 M 24 13 37 19 19 38 High school Truck driver Birth 57 1974 M 26 8 34 23 15 38 Master's special ed teacher Birth 59 1971 M 31 16 47 19 9 28 Master's high school teacher Birth 61 1957 F 25 9 34 29 8 37 Associate's dental hygienist Birth 62 1976 F 26 11 37 27 11 38 Master's social worker (behavior health therapist) Birth 63 1972 M 22 17 39 19 12 31 Associate's small business owner (printing) Birth 64 1958 F 24 8 32 25 11 36 Master's high school teacher Birth 67 1974 M 23 15 38 20 6 26 Master's high school teacher Birth 68 1977 M 24 8 32 22 9 31 Master's high school teacher Birth 69 1959 M 14 17 31 19 20 39 some college Volunteer Firefighter & ? Birth 86 1989 F 21 10 31 24 18 42 In college (1 yr complete) student (pharmacy) Birth 87 1952 F 17 13 30 23 13 36 Master's human resources Birth 88 1949 M 16 12 28 23 15 38 some college (little over a year) industrial engineer Birth 89 1982 M 23 14 37 21 14 35 High school butcher Birth 90 1983 M 25 14 39 17 9 26 Bachelor's Industrial designer? Birth
263 Table A 1. Continued Participant Lived outside Jasper Parents' place of birth Parents' education 1 4.5 years Terre Haute (ISU) but returned in summers Mount Carmel, Illinois Dad four year degree 2 2 years Vincennes (VU), 3 years West Lafayette (Purdue), came back all but one summer, one summer in Evansville Mom Montgomery, Dad Brookville 2 year degrees 3 1 semester Williamsburg KY, 1 year Owensboro Mom Kansas, Dad Meringo Both college 4 4 years Indianapolis for college (came home first summer) Mom Washington, IN, Dad Jasper Father associate's degree, Mother high school 5 3 1/2 years in Bloomington Mom Loogootee, Dad Jasper Mom some college, Dad associate's degree 49 Evansville 4 yrs (home all summers) Mom St. Anthony, Dad Jasper Both high school 50 Muncie 1 yr, Jasper for summer Pensacola FL 1 yr, Muncie 2 yrs (Jasper 1 more summer, part of last school year), back to Muncie (months?), Indianapolis Aug '04 May '05, Fishers, IN, May '05 May '06 Both Jasper Mom Master's, Dad Bachelor's 51 Evansville 9 mos, Jasper summer, Evansvi lle 1 semester, Jasper Dec. July, North Dakota 10 mos, Jasper month or two, Evansville August 2003 December 2006 Dad Huntingburg, Mom Jasper Dad high school, Mom some college 53 Dad Jasper, Mom Huntingburg Both high school (Mom insurance certification) 54 Cincinnati <1 yr Dad Jasper, Mom Chillicothe, Ohio Mom Master's, Dad some college + police academy 55 Terre Haute 4 yrs (Jasper 2 summers), back to Jasper after graduating Both Jasper Both high school 56 Both Jasper Both less than high school 57 Vincennes 18 22 (home summers), Princeton 1.5 2 yrs Both Jasper Both high school 59 Murry, KY, 5 yrs (home all summers) Dad Jasper, Mom Mitchell Mom associate's, Dad Master's 61 9 mos, Bloomington, 2 yrs Indy minus summer Dad Jasper, Mom Gentryville both high school 62 West Lafayette 1 yr (home summer), Evansville 3 yrs (home summers) back to Jasper 1998 Mom Jasper, Dad Pike County Mom high school, Dad GED 63 Vincennes 2 yrs (home summer) (age 19 and 20) Both Jasper Both less than high school 64 Muncie 3.5 yrs (home 2 summers) Dad Ireland, IN, Mom Jasper Mom high school, Dad master's 67 Evansville 3 yrs (home one summer), Bloomington 2 yrs (including summers) Dad Corydon, Mom Evansville Dad bachelor's, Mom master's 68 Muncie fall 95 spring 99 (Jasper all summers), Greenwood August 99 until June 2002 Dad Jasper, Mom Huntingburg both high school 69 one summer near Culver Indiana Dad Wickliffe, Mom Jasper Dad high school, Mom some college 86 9 mos West Lafayette Both Jasper Mom bachelor's, Dad associate's 87 had an apartment in Louisville 5 yrs (2002 2007) but came home every weekend Dad Jasper, Mom St. Henry Mom GED, Dad high school 88 4 mos in Chicago area?, Pensacola 9 mos, Japan 2.5 yrs Mom Schnellville, Dad Ireland, IN Both less than high school 89 Mom Jasper, Dad Huntingburg both high school 90 West Lafayette four years minus summers, later West Lafayette one more semester Mom Ireland, IN, Dad Ferdinand Dad associate's, Mom high school
264 Table A 1. Continued Participant Parents moved to Jasper Siblings (Age and where live) 1 late 20s No 2 early 20s 28 year sister in Newburgh 3 early 20s 31 y o sister in Fayetteville, Arkansas, 24 y o brother in Jasper 4 Mom 18 Sister in Tampa 5 early 20s 30 year old sister in Indianapolis 49 Mom 21 or 22 23 sis in Jasper 50 25 bro Atlanta, GA; 23 sis Louisville; 20 sis Muncie 51 Dad mid 20s 31 bro Atlanta, GA 53 Mom 30 or 31 35 half sis Dubois, 32 half bro Ireland, IN; 27 half sis Jasper 54 Mom early teens 25 bro Bloomington, 20 sis Evansville, 17 bro Jasper 55 30 bro Florida 56 72 sis Los Angeles, 68 sis Jasper, 66 sis Ireland, IN; 64 sis Bean Blossom, IN; 61 bro ?; 60 bro Wilmington, DE; 59 sis (says where but couldn't hear); 58 sis Chicago; 57 sis Sacramento, CA; 56 bro Hawaii; 55 sis Albuquerque, NM; 55 bro Chicago (ages are estimates; says he doesn't know siblings) 57 31 sis Jasper 59 Mom 20 34 sis Jasper 61 Mom 20 55 sis New Lenox, Illinois (Chicago suburb), 53 sis Lakewood, Colorado; 50 sis Golden, Colorado; 47 sis Jasper; 38 bro Thornton, Colorado 62 Dad 23 ish 28 sis Jasper 63 42 sis Jasper, 41 bro Jasper, 39 Jasper, 37 Jasper 64 53 sis Carmel, 49 bro Jasper, 47 bro Kentucky, 41 sis Newburgh, 38 bro Jasper 67 22 or 23 31 bro Denver, CO; 30 sis Greenville, TN 68 Mom mid thirties 34 sis Jasper, 39 bro Jasper, 42 sis Ireland, IN; 44 sis Racine, Wisconsin 69 Dad 44 63 half bro Jasper, 65 half sis Huntingburg 86 13 sis in Jasper 87 Mom early 20s 58 bro east of Indy, 54 bro Colorado Springs, 51 sis between Jasper and Evansville (couldn't think of town name), 49 sis in Birdseye, 47 sis in Otwell, bro Bretzville, 46 sis Velpen 88 mid twenties 61 bro in Jasper, 58 bro in Newburgh, 54 sis Spencer County, 52 sis Jasper, 48 bro in Ireland, IN 89 Dad small child 2 33 sises (twins) in Huntingburg 90 Dad 21 22, Mom? twin brother in Jasper
265 Table A 2. Lifelong Chicago participants Participant YOB Sex M+ M MT otal F+ F FT otal Education Occupation Age of arrival in Chicago 7 1984 F 20 12 32 29 13 42 College unemployed, former auditor Birth (suburbs) 11 1966 M 22 17 39 20 4 24 JD lawyer Birth 14 1959 F 28 5 33 27 8 35 some college legal assistant Birth 15 1953 M 26 12 38 22 10 32 working on JD business owner? Birth 17 1983 F 27 16 43 24 11 35 Bachelor's portfolio analyst Birth (Naperville) 20 1984 F 27 7 34 28 12 40 Bachelor's Birth 23 1956 F 23 6 29 28 12 40 Bachelor's accountant? Birth 27 1983 M 23 14 37 28 11 39 Bachelor's construction manager? Birth (Naperville) 29 1984 M 12 18 30 27 9 36 Bachelor's Birth (Evanston) 31 1957 F 28 13 41 27 7 34 HS Birth 34 1976 M 25 12 37 28 12 40 Bachelor's Birth (Oaklawn) 35 1980 F 22 12 34 21 12 33 Bachelor's fund raiser, fomer marketer Birth 41 1978 F 13 7 20 28 17 45 JD lawyer Birth (Northbrook) 43 1977 F 20 5 25 26 5 31 High School nanny Birth 44 1970 F 20 7 27 27 10 37 Bachelor's Birth 47 1960 F 19 8 27 29 14 43 Associate's legal assistant? Birth 48 1969 M 24 13 37 24 9 33 some college Birth 77 1987 F 23 13 36 18 12 30 Bachelor's volunteer for Olympic bid, will be consultant Birth 79 1979 F 22 16 38 20 13 33 Master's Birth 80 1986 F 16 11 27 27 19 46 high school optician Birth 81 1965 F 27 17 44 13 8 21 Bachelor's Birth 83 1954 M 22 11 33 24 9 33 DDS dentist Birth 84 1973 F 22 12 34 25 9 34 Bachelor's actor plus day job Birth 85 1964 F 18 8 26 28 10 38 some college small business owner Birth
266 Table A 2. Continued Participant Lived outside Chicago Parents' place of birth 7 Urbana Champaign four years, came home every summer Mom Tennessee, Dad Niles, IL, Stepmom maybe Buffalo Grove, IL 11 Four yrs Urbana Champaign (came back one summer) Mom Germany (but raised Lithuania), Dad Lithuania 14 2 mos. San Diego, 2 wks a year in Idaho Mom Galva City (?), IL; Dad Mexico 15 Peoria 3 yrs, Champagne few mos, 3 mos Lima, Peru Both Chicago (Dad Peterson Park, Mom Englewood) 17 Bloomington, IN, 4 yrs minus 2 summers; studied abroad one summer where? Dad Western Springs? Mom Arlington Heights? 20 Mom Bridgeport, Dad Hometown 23 Both Chicago (North Center) 27 West Lafayette 4 yrs. (home 3 summers) Both Chicago (Mom Humboldt Park, Dad by Niles) 29 U. of Mississippi one semester, Bloomington, IN, 4 yrs (Chicago one summer) Both New York (State) 31 San Antonio 6 mos. Mom Chicago (607 W Wrightwood), Dad Minnesota 34 South Bend 4 yrs (2 summers home, 1 summer London) 6 mos. Travel to Boston, less time in Conn., Kansas City Mom Indiana, Dad Northern Suburbs Chicago (Morton Grove) 35 1 yr Minneapolis Both Korea 41 New Orleans 1997 2001 (1 summer Northbrook, 2 summers DC) Dad Chicago (Rodgers Park), Mom St. Louis 43 Dad Baltimore, MD, Mom Bridgeport 44 Decatur, IL 1 yr Both Chicago (McKinley Park) 47 Both Chicago (Back of the Yards) 48 Both Chicago (Mom 16th & California) 77 South Bend 4 yrs (home to Kennilworth or Winnetka in summers) Mom Jasper, Dad Charleston, WV 79 Champaign 4 yrs (Wheeling all summers) Both Chicago (Dad near Rogers Park, Mom north) 80 Mom Mexico, Dad El Salvador 81 Dekalb IL four yrs (back to Mount Prospect all summers) Both Chicago (Dad west side, Mom northwest side) 83 Champaign 1 yr, (summer in Wisconsin) Macomb 1 yr Mom Ontario, Canada, Dad Chicago (Wrigleyville) 84 Schaumburg 2 yrs, Barrington 3 months, backpack in Europe 3 mos, New Zealand 2 yrs Dad Chicago, Mom Mason City, Iowa 85 Both Chicago (Taylor Street)
267 Table A 2. Continued Participant Parents' education Parents moved to Chicago Siblings (Age and where live) 7 Both college Dad birth (in suburb), Mom around age 5 28 bro in Rockford, IL, 26 sis in Seattle, 23 step sis in Grayslake, IL, 21 step bro at EIU, 16 half sis in Vernon Hills 11 Dad 6th grade, Mom Licensed Professional Nurse Mom 26+, Dad 28+ 14 Mom 8th grade; Dad none Both late 30s 63 bro in MD/DC; 6 sis in Chicago: 65, 64, 45 ; 43 sis in Ann Arbor 15 Both Master's 54 bro Chicago 17 Mom Bachelor's, Dad MBA 23 sis in Evanston, 18 and 16 step bros in Naperville (joined family when participant in college) 20 Both some college 27 bro in Chicago, 19 bro in Bridgeport 23 Both high school 55 sis Oak Park (Chicago), 50 bro North Center 27 Mom Bachelor's, Dad high school 24 sis in River Forest 29 Dad JD, Mom Bachelor's Mom 3, Dad 7 (grew up Beverly) 29 sis Lincoln Park, 39 sis Evanston 31 Mom some college, Dad? Dad? 50 sis Berwyn, 48 sis Oak Forest 34 Mom HS, Dad Bachelor's Mom ? 34 sis Rockport, IL 35 Dad Bachelor's, Mom Pharm.D. Both early 20s 32 sis Chicago 41 Mom Bachelor's, Dad Bachelor's Mom early 20s 28 bro Chicago, 25 sis New Orleans 43 Mom HS, Dad some HS Dad 6 mos. 29 bro Country Club Hills, 21 sis West Lafayette 44 Both HS 47 bro Lewiston, Minn; 46 LaGrange Park (suburb); 32 bro Boulder, CO; 28 sis Palis Hills (suburb) 47 Mom HS, Dad less 58 bro Woodridge, 52 bro Chicago 48 Mom HS, Dad less 41 bro, 43 bro, 44 bro, 45 bro, sisters in Naples, FL, and Georgia, sis Chicago, 77 Mom bachelor's, Dad JD 22 (both) 25 sis in Lakeview, 18 sis in Winnetka 79 Dad optometrist degree, Mom master's 35 sis in Vernon Hills 80 Dad master's, Mom some college Mom 13 or 14, Dad 17 (came to Lakeview) 27 sis in Old Irving Chicago; 25 sis Avendale 81 Dad high school, Mom Associate's 44 bro Mount Prospect and Chicago, 42 bro Arlington Heights 83 Mom some college, Dad bachelor's Mom mid 20s 56 sis country of Turkey, 51 sis Denver 84 Both bachelor's Mom 23 ish 46 bro in Berwyn, 45 sis in Houston, 42 bro in Elmhurst, 38 sis in Chicago 85 Both high school 50 sis in Kerry IL, 45 sis in Kerry, 47 bro in Mount Prospect, 37 bro in Oswego
268 Table A 2. Continued Participant Where in Chicago area lived (bold = outside of Chicago/ close suburb range) Where worked Other comments 7 Evanston, Rogers Park until age 2, Vernon Hills until age 6, Lake Forest one year, back to Vernon Hills until 18, elsewhere college, then Lakeview for two years, Boystown for less than one year, then Lake Villa, IL, to present 11 Marquette Park grew up, after college Lincoln Square until 2007, now Portage Park/Six Points Spoke Lithuanian with parents growing up? 14 Back of the Yards, Bridgeport (from 6th grade until age 38), Brookfield downtown/ Loop Spoke Spanish with her father, English with mother (mother responded in Span); older siblings born in Mexico 15 Englewood to 6, South Shore to 1967, Jeffrey Manor, after college 6 7 yrs Hyde Park, since Wrigleyville 17 Naperville til 2008 except college (plus one summer); intersection of Gold Coast/Lincoln Park/Old Town since August 2008 Northbrook, Naperville 20 Bridgeport til 1.5?, Hometown til 5, Lockport til 9, Bridgeport Back of the Yards, Loop 23 North Center til 30, Lakeview 1990 2001, North Center again downtown, (traveled in suburbs & out of town), River North 27 Naperville birth 2, River Forest age 2 to 18, again one yr after college, West Loop 6 mos, River Forest 1 yr., Bucktown 2 yrs Loop 29 Evanston always except indeterminate amount of time in Rogers Park for college Rodgers Park, Evanston 31 607 W Wrightwood (Lincoln Park) to 4, same neighborhood until Oak Forest 1995 State Farm, Grant Hospital, Children's hospital 34 Oaklawn, Crestwood, Tinley Park just summers, Wrigleyville 3 yrs, Lakeview 4 yrs, Wrigleyville 4 yrs? Downtown, travel to Northern suburbs 35 Chicago (North Side) to 5, Glenview to 12 13, Northbrook to 18, Evanston to 22, Lincoln Park 1 yr, Bucktown 1 yr, Lakeview (Wrigleyville) 2 yrs Loop 4 yrs, Evanston <1 yr Grew up bilingual Korean 41 Northbrook til 18, again after college 1.5 yrs, Lakeview <2 yrs, River North? (Superior) 4 5 yrs Glenview, Loop 43 Bridgeport to 12 13, Country Club Hills (iffy) 16 yrs, Glenwood (iffy) 8 mos., Tinley Park 4 yrs Orland Park, Tinley Park, Matsen (?) 44 McKinley Park to 18, again 19 25, Lakeview to 2006, Streeterville Loop, McKinl ey Park (HS), Lincoln Park, Streeterville 47 Back of the Yards to 18, Ashburn to early 20s, Evergreen Park 2 yrs, Scottsdale, Ashburn, Beverly, Mount Greenwood 2 yrs Loop (Mich. Ave. as teen) 48 Canaryville to 18, Cicero 1 yr, Lyons, IL (?) 6 yrs, Brookfield 2 yrs, Downers Grove 5 yrs, Canaryville 6 mos., Beverly 3 yrs, Mount Greenwood 2 yrs Everywhere, Bensonville? 77 Lincoln Park first 5 months, Wilmette until age 6, Kennilworth until 18, Winnetka since age 20 (except when at college) South Loop the Loop 79 Wheeling until 18, Arlington Heights briefly, Gold Coast 1 yr, Wrigleyville 3 yrs, Lakeview 3 mos, 2 yrs Lincoln Park Schaumburg, Glenview, Wheeling, Buffalo Grove 80 Lakeview until 14, Avondale, Rogers Park 2 mos, Portage Park briefly, Roscoe Village until back to Avondale for 1 yr downtown/West Loop, Lincoln Park Mom is actually Greek (but knows Spanish), parents spoke Spanish to her growing up, but now participant has forgotten much of her Spanish (was always English dominant) 81 Unkn own Chicago neighborhood (northwest side) 6 mos, Mount Prospect until end of high school, Mount Prospect again until age 23 (one year?), Arlington Heights since age 23 (11 years) Schaumburg, Arlington Heights, Rosemont 83 Park Ridge until age 1, Clarendon Hills until age 18, Lagrange 1 summer, The Loop six or eight years (except one summer LaGrange), Wrigleyville 7 or 8 years, Wheeling 4 yrs, Oriole Park/Norwood Park 11 years Des Plaines 84 Mount Prospect until end of first year of college, summ er in Rogers Park, Mount Prospect 1 yr, Lincoln Park one year (or nine months), Mount Prospect not long, Andersonville 1 yr, Ravenswood 1 yr, Mount Prospect briefly, Andersonville 2 and a half years, Mayfair one and a half years Downtown 4 or 5 yrs, Uptown 1.5 yrs, plus Andersonville, St. Charles, plus shows in other places 85 Wrigleyville until age 19, Elmwood Park 1 yr, Addison 5 yrs, Park Ridge ever since Bensenville, Downtown, from home
269 Table A 3. Mobile participants Participant YOB Sex M+ M Mtotal F+ F Ftotal Education Occupation Age of arrival in Jasper 6 1982 F 20 5 25 26 13 39 MBA internal auditor and credit analyst Birth 8 1971 M 30 13 43 24 10 34 Masters Public Policy, MBA financial adviser? Birth 9 1982 F 15 13 28 20 10 30 working on Masters director of arts programs Birth 10 1954 M 31 21 52 17 1 18 JD lawyer Birth 12 1978 F 20 4 24 28 13 41 JD lawyer Birth 16 1982 M 29 19 48 19 11 30 Bachelor's accountant Birth 18 1984 F 25 14 39 30 15 45 Bachelor's (working on Master's) student (counseling) Birth 19 1983 F 25 13 38 26 16 42 Bachelor's unemployed (graphic designer) Birth 22 1968 M 30 12 42 25 11 36 PhD chemist? Birth 24 1975 M 20 17 37 22 15 37 Bachelor's stock broker? Birth 26 1982 F 21 7 28 25 11 36 Bachelor's construction manager thing? Birth 28 1984 M 25 19 44 28 13 41 Bachelor's sales Birth 30 1962 M 23 14 37 21 16 37 Bachelor's medical researcher? Birth 32 1979 M 23 15 38 23 11 34 Bachelor's (working on MBA) architecture Birth 33 1976 M 19 8 27 21 7 28 Master's bank mathematical model maker? Birth 36 1982 F 25 20 45 18 16 34 Bachelor's carpet marketer Birth 37 1979 M 28 7 35 29 7 36 Bachelor's (working on Master's) helps run family shipping business Birth 38 1980 F 24 10 34 22 12 34 Bachelor's nurse Birth 39 1955 F 24 13 37 29 10 39 Master's audiologist/systems analyst Birth 42 1972 F 24 9 33 25 13 38 Bachelor's pharmaceutical rep Birth 46 1983 M 25 15 40 23 16 39 Bachelor's accountant Birth 70 1989 F 19 9 28 30 17 47 working on Bachelor's (1 yr complete) student (fashion) Birth 71 1982 M 28 22 50 12 8 20 MBA unemployed (acquisition and merger specialist) Birth 73 1950 M 28 11 39 22 7 29 Bachelor's business owner (design) Birth 76 1956 F 21 11 32 22 16 38 Bachelor's retired CPA Birth 78 1955 F 21 6 27 25 16 41 Bachelor's retired (accountant) Birth
270 Table A 3. Continued Participant Lived elsewhere Age of arrival in Chicago How long lived in Chicago 6 Bloomington age 18 23, went back to Jasper first three summers, one summer in Chicago One summer, then 23 3.75 yrs 8 College in West Point, NY (one summer Los Alamos, NM; other summers NY; one month Georgia) 4 mos. Indy, one year Germany, 6 mos. Of 2001 in Houston 24 except 6 mos. In Houston 13.5 yrs 9 3 mos Bloomington; 2 3 years Dayton, OH; 3 yrs Indy 24 3 yrs 10 5.5 yrs Lafayette (home for summers) 23? (1978) 31 yrs 12 4 yrs college in South Bend (Notre Dame) (back to Jasper for summers except one); one semester in London; 23? Fall 2001 7.5 yrs 16 Bloomington age 18 23, went back to Jasper one summer; one year suburb of DC; 24 2.5 yrs 18 Bloomington age 18 22 minus summers and one semester, one semester Australia 22 1.5 yrs 19 Bloomington 4 yrs (17 22) Jasper 2 summers, one summer study abroad, 2 yrs DC 24 1.67 yrs 22 Lafayette 2 y rs (Jasper both summers); Hamburg, Germany 1.5 yrs; Lafayette 2 more years (Chicago that summer); 4 mos. NYC, 1 mo. Houston, 3 mos. Boulder, CO 23? (except one summer before) 18 yrs 24 Bloomington 4.5 years (back to Jasper 2 summers), one summer Chicago, Stamford, Conn./NYC summer, San Francisco 9 years Most recently 32 3.5 yrs 26 West Lafayette 4 yrs (one summer Jasper, 2 summers San Diego), San Diego 4 mos. 23? (2005) 4 yrs 28 Bloomington, IN, 4 yrs (including summers) 23 1.8 yrs 30 West Lafayette 2 yrs (Jasper summer), back to Jasper 1 yr, Terre Haute 2 yrs (Jasper one summer?), back to Jasper 2.5 yrs 27 19.25 yrs 32 Vincennes 2 yrs (back to Jasper 2 summers?) 3 yrs Carbondale Illinois (including summers), Bloomington one summer, Hypoluxo FL one and a half yrs 27 3 yrs 33 South Bend 4 yrs (2 summers in Jasper), one summer West Lafayette, back to Jasper one summer; Stanford, CA, almost 5 yrs 27 6 yrs 36 Lafayette 5 yrs (1 summer Jasper, 2 summers Chicago, one half Chicago half Indy? Check) 23 but two summers before 4.5 yrs 37 Charleston, IL 4 yrs (1 summer Jasper, 1 summer Indianapolis), Milwaukee commute one year, then same travel as #38) 22 or 23 for 2+ yrs, again 27? 5 yrs 38 West Lafayette 4 yrs (Jasper 2 summers, 1 summer Chicago), 3 mos LA, 3 mos San Fran, 3 mos Boston, 6 mos. NYC, Houston 3 mos 22 for 2 yrs, again 26? 4 yrs 39 West Lafayette 1 yr, Bloomington 3.5 yrs (Jasper summers except one?, one semester VUJC), Michigan City IN 1 yr 1979 (24?) 30 yrs 42 West Lafayette 1991 1995 (1 summer Jasper, 1 summer Missouri, 1 summer Indy), Louisville 1 yr, Indy 6 yrs 2002 (30?) 7 yrs 46 West Lafayette 4 yrs (Jasper 3 summers) 22 4 yrs 70 18 1 yr 71 Bloomington 5 yrs (3 summers Jasper, one summer Indianapolis), St. Louis 7 mos., New York 2+ years 25? 1.33 yrs 73 Carbondale IL 18 23 (home all summers but last), Boston 2 yrs, Chicago one summer 25 (except one summer before) 33.5 yrs 76 South Bend 4 yrs (home summers) 22 30.8 yrs 78 South Bend 4 yrs (home summers) 22 31.75 yrs
271 Table A 3. Continued Participant Parents' place of birth Parents moved to Jasper Parents' education Siblings (Age and where live) 6 Both Jasper Both Bachelor's 22, 20 sisters both in Bloomington, IN 8 Both Jasper Mom 2 yrs of college, dad 2 yrs college early, then Bachelor's and Masters later on 36 bro in Piedmont, CA; 39 sis in Bloomington 9 Mom Jasper (Dad gone) Mom Associates, currently getting Bachelors 31 bro in Jasper 10 Both Jasper Both probably less than high school bro and 2 sis in Jasper; 10 other siblings, 1 Hawaii, 3 CA, 1 Maryland, sis in Chicago, one in Bean Blossom, IN; 1 Louisville, 2 Albuquerque 12 Dad Ireland, IN; mom Jasper Dad early 20s? Mom some college, Dad high school sis 27 in St. Louis, bro 23 in Chicago 16 Both Jasper both high school 24 sis in Chicago 18 Both Jasper Dad Bachelor's, Mom Associate's 22 bro in Bloomington 19 Mom Baltimore, Dad Jasper, stepdad Holland, IN; stepmom Jasper Mom 6 Mom, stepmom, stepdad bachelor's, Dad associates 17 sis, 13 half bro?, 9 half bro?, 7 half bro? In Jasper 22 Both Jasper Both high school 42 sis in Ireland, IN 24 Both Jasper Mom Master's, Dad optometry degree 30 bro in Indianapolis 26 Mom Washington, IN, Dad Epsom, IN Mom 19 or 20, Dad 29 ish Both high school (Dad GED) 32 sis in Bend, Ore. 28 Mom Clinton, IN, Dad Connorsville, IN mid 20s? Mom Bachelor's, Dad Master's 28 bro Chicago, 19 bro Jasper 30 Dad Ireland, IN, Mom Detroit Mom 8 or 10 Dad HS, Mom less 56 bro Cincinnati, 55 bro Jasper, 54 Evansville, 51 bro Jasper, 49 sis Jasper, 40 bro Alaska 32 Dad Ireland, IN, Mom Jasper Dad teenager both high school 28 sis Ireland, IN 33 Mom Jasper, Dad Queens NYC Dad mid 20s Both Bachelor's 38 bro Indianapolis, 35 bro Valporaiso, IN; 29 bro NYC 36 Mom Oakland, CA, raised NYC; Dad Jasper Mom 21 or 22 Mom dental hyg. Associates? Dad Bachelor's 40 sis Indianapolis, 28 bro New York, 24 bro Jasper 37 Both Jasper Mom Bachelor's Dad Master's 26 bro Jasper, 24 sis Chicago, 20 sis Bloomington 38 Both Jasper Both high school 31 sis Charlotte, 22 bro Jasper 39 Mom Staten Island, Dad Henryville, IN Mom 20, Dad 25 ish Mom Master's, Dad PhD 50 sis Jasper, 52 bro Nashville, TN & Bloomington, IN 42 Both Jasper Mom Master's, Dad some college 40 bro Columbus, OH; 32 sis Westfield, IN 46 Mom Loogootee, IN, Dad Jasper Mom 21? Both HS 27 sis Huntingburg, 26 bro Jasper, 24 bro Jasper 70 Both Jasper Dad certification after high school, Mom bachelor's 23 bro in Jasper, 17 bro in Jasper 71 Both Jasper both high school 41 half bro in Jasper 73 Both Jasper both high school 57 bro in Jasper, 55 sis Cincinnati, 52 sis Denver 76 Both Jasper Dad bachelor's, Mom high school 58 sis Celestine, 55 bro Indianapolis, 45 bro Indianapolis 78 Both Jasper Mom high school, Dad some college 72 bro in Jasper, 70 bro in Jasper, 67 sis in Jasper, 65 bro in Jasper, 51 bro in Carmel, IN
272 Table A 3. Continued Participant Prefers Jasper or Chicago Where in Chicago lived Where worked How often go back to Jasper 6 Chicago one summer Gold Coast, 1 yr Lakeview, 2 yrs Boystown 4 5 times a year 8 Chicago Hyde Park 2 yrs, Gold Coast 11 yrs, Lincoln Park 2 yrs Hyde Park, Gold Coast, Loop 2 times a year 9 Lakeview 1.5 yrs, Uptown 1.5 yrs 10 Both Woodridge 4 5 yrs Downers Grove 25 yrs Downtown 12 Chicago 1 yr Hyde Park, Lakeview/Lincoln Park 2 yrs, River North 2 yrs, Lakeview since 2006 16 Chicago Gold Coast 5 7 times a year 18 Chicago Lakeview South Loop, Wicker Park, Loyola University 3 tmes a year 19 Both (probably Chicago) Gold Coast the Loop 22 Both (probably Chicago) Ravenswood (7 yrs), Lincoln Square (11 yrs) Glenview 24 Both (more Jasper?) Lincoln Park 1 mo., Evanston 2 mos., Lincoln Park 3.5 yrs. Loop 26 Chicago Lakeview, Streeterville, Bucktown Loop, South Loop, River North 28 Jasper Lincoln Park Downtown 30 Chicago Rodgers Park, Edgewater, Andersonville 7 yrs, Ravenswood 1 yr, Lakeview 2 yrs, Oak Park 3 yrs, Edgewater 3 yrs Lincoln Park 32 Chicago Evanston 8 mos, Wrigleyville 1 yr, Lincoln Park 1 yr, The Loop 7 mos Downtown/The Loop 1 2 times a year 33 Chicago ish Lakeview 5 yrs, North Center 8 mos. Downtown, also from home 36 Both (right now Chicago) Streeterville 1 summer, Lakeview since River North, Loop 37 Chicago? River North 1 yr, Wrigleyville 3 yrs, River North 2 yrs River North 12 times a year 38 Both Wrigleyville 1 summer, Gold Coast 2 yrs, River North since Lincoln Park, Hyde Park, Wilmette?, UIC 12 times a year 39 Chicago Woodridge 4 5 yrs, Downers Grove 25 ish yrs Downtown, Aurora 42 Both (prob Chicago?) West Dundee 1 yr, River North 1yr, Orland Park 5 yrs All over suburbs 1 2 times a year 46 Jasper Tinley Park 2 mos, Schaumburg 1 yr, Brookfield <1 yr Loop 70 Chicago West Loop Downtown, Gold Coast, Lincoln Park 3 4 times a year 71 Chicago Lincoln Park 1 month, Lakeview 1 yr, Lincoln Park 3 months The Loop 8 9 times a year 73 Chicago Glen Ellyn 1 yr, Elmhurst since 1977 downtown, "about every suburb," Oakbrook 25 yrs 2 3 times a year 76 Chicago Arlington Heights 2 yrs, 1 yr Wrigleyville, 6 yrs Lincoln Park (DePaul), Wilmette, Kennilworth, Winnetka (see Participant 77) 2 ish times a year 78 Chicago Mount Prospect 2 yrs, Arlington Heights 2 yrs, Lincoln Park 4 yrs, Wilmette since 1985 all over (as auditor, based downtown), Deerfield, Lincolnshire, River Woods once a year?
273 APPENDIX B T EXT OF RECORDED INSTRUCTIONS As described in Chapter 2, two sets of recorded instructions (one read by a Southern Indiana speaker and the other read by an NCS speaker) were used in this experiment. The text of the recorded instructions is presented below. Instructions for the EPAQ Thank you for agreeing to participate in this experiment! The first portion of the experiment should take approximately one hour, perhaps less. The second portion should take 1 to 2 hours. 1 The first portion will involve a personality questionnaire, a reading task, and a listening exercise. The second portion will involve a casual conversation, possibly preceded by a brief game. F or the first part of this experiment, if you are willing, you are going to take a short personality questionnaire like you might find in a magazine. The questionnaire consists of 32 items. For each item, you will see a pair of opposing characteristics, w ith the letters A, B, C, D, and E printed as a scale in between. Each pair describes of characteristics, it is impossible to be both at the same time, such as not at all artistic and very artistic. The letters form a scale between the two extremes. You are to choose a letter which describes where you fall on the scale. Using the example of not at all artistic vs. very artistic, if you thought you had no artis tic ability, you would choose A. If you thought you were pretty good at art, you might choose D. If you thought you were only medium, you might choose C, and so forth. 1 In reality, most participants took less than two hours for t he entire experiment.
274 It is important that you circle the letter that corresponds to how you think you reall y are not to how you wish you were. 2 Please do not feel that you have to hide your faults. Remember that all responses will remain confidential. Nothing you circle will be reported individually, and your overall results will only be paired with your pa rticipant number, never with your name. Please try to be honest and accurate. Instructions for the Reading Task Next, if you are willing, you will be recorded reading a short passage aloud. The passage is only one paragraph long and tells a story. When you get the paper with the reading passage on it, please look through it before you begin reading it aloud. This will read it aloud. Please read the passage at a normal speaking rate. That is, you would use if you were reading the story on a local 3 radio station. Instructions for the Perception Experiment you hear certain words. In the exercise, you will hear a man saying words one at a time. Each time he says a wor d, you will choose which of two possible words he said. You will see the two word choices and corresponding keyboard keys on the screen in 2 3 Encouraging the participants to imagine reading on a local radio station was intended to encourage them to use a local accent when reading.
275 There will be a practice session followed by three blocks of thirty words. In each block, you will have the same two word choices for all thirty items. For example, you can take breaks in between each block. peopl e will hear the same words differently. You should just respond with what you think you hear. al think you hear. Sometimes you may have trouble deciding which of the two words you heard. In this case, just go with your gut instinct. Please respond as quickly a s possible. Do not feel you have to respond so quickly that you make mistakes, though. Once you respond, there will be a one second pause before the next word is played. However, 4 if you do not respond within three seconds, the next word will be played, and no response will be logged for the previous word. Again, please go with your first instinct and respond as quickly as possible. 4 The NCS reader left out the word however
276 APPENDIX C T EXT OF READING PASSAGE Each participant was recorded reading the passage below. The exception is t hat Ron rather than John I changed the name to John for the remaining 73 participants because I realized determining the starting point of the vowel would be easier following an affricate th an following an /r/. In the reading passage below, words intentionally placed in the reading passage as stressed tokens of the NCS target vowels are underlined. Other words that could also provide tokens of the NCS vowels are in italics. (Participants r ead a version in plain text with no indication of which words I was interested in, of course.) Reading Passage My friend John sure has a strange idea of fun One day he called me over to his house to see a hole that he had dug I asked what was so exciti ng about a hole, but he tell He said he had a great plan yet. To me the thought of a hole seemed pretty dull but John is my best pal so I went When I got to John house, he was playing with his old red truck trucks but I guess John not Then he ran the truck over to his big hole and let it fall that asked laughing I could only nod let the cat out of the bag on that one.
277 APPENDIX D INTERVIEW/CONVERSATION TOPICS AND DEMOGRAPHIC QUESTIONS These topics were brought up if there was a lull in the conversation during the sociolinguistic interview or partner conversation. As marked, certain questions were asked of all participants. Discussion Topics for All Groups How did you Meet your Husband/W ife? How soon did you know s/he was the one? How did he propose to you / How did you propose to her? What did your parents think when they met him/her? Danger of Death Have you ever been in a situation where you thought you really might die? Where you September 11 Where were you when you found out about the September 11 attacks? Memories of High School, Middle School What kind of stuff did you do in high school? Were you in any clubs, on any teams? Who did you hang out with? Do you still do stuff with them? Where are they now? Were there any teachers you especially hated? Is there a Do you still have dreams about being back in school?
278 Dreams Do you have any recurring dreams? Work Where are you working now? How long have you worked there? Do you like it? Do you get along with your co workers? Boss? Are there any real jerks that you work with? Any memorable clients / customers / events? Economy How are you weathering the recession? What about your parents? Do you How do you feel about the bail outs? Do you think the American car makers deserved the bail out? Do you think Obama will be able to put us back on track? Discussion Topics for Lifelong Jasper Participants Thoughts about Jaspe r Do you like living in Jasper? What kind of things do you do for fun now? Do you think Jasper has changed during your lifetime? How?
279 Do you see Jasper differently now than you did in high school? Did going to colle ge change your opinion? Do you plan to stay here a long time? Construction in Jasper What do you think about the plans for a new library? The Catholic high school? The train to French Lick? Sports How do you think Jasper will do in the tournament this year? How has the baseball team done since I graduated? Closings It seems like a lot of stores and restaurants are closing around here. e. Anybody else? Discussion Topics for Mobile and Lifelong Chicago Participants Life in Chicago Do you like living in Chicago? Have you ever thought about moving somewhere else? What parts of Chicago should I be sure to see before I leave?
280 What do you think about the whole Blagojevich thing? What abo ut the new governor, Quinn? Do you think Burris will be a good senator? Will he be respected? Did you go down to Grant Park on election night? Do you hope the Olympics come to Chicago in 2016? Questions Asked of All Lifelong Jasper Participants When were you born? Were you born in Jasper? When did you move here? Have you lived anywhere outside of Jasper? What about college? How many years have you lived outside Jasper? So can you take me through your life, where all you lived when? Where were your parent s born? About how old were they when they moved to Jasper? How much education did your parents have? Do you have any siblings? How old are they? Where do they live? Questions Asked of All Mobile Participants What made you decide to move to Chicago? Did you specifically want to leave Jasper? Which do you like better, Jasper or Chicago? What makes you like it better here/there? Do you plan to live a long time in Chicago? If not, where else might you like to live? Have you ever thought about moving back to Jasper? How often do you go back? Does your family still live in Jasper? When were you born?
281 Were you born in Jasper? When did you move there? Have you li ved anywhere else besides Jasper and Chicago? What about college? How many years did you live these other places? neighborhoods in Chicago? What Chicago neighborhoods and suburbs have you worked in? Where were your parents born? About how old were they when they moved to Jasper? How much education did your parents have? Do you have any siblings? How old are they? Where do they live ? Questions Asked of All Lifelong Chicago Participants When were you born? Were you born in Chicago? When did you move here? Which neighborhoods have you lived in? When? Which neighborhoods have you worked in? When? Have you lived anywhere outside of Ch icago? What about college? How many years have you lived outside Chicago? Where were your parents born? About how old were they when they moved to Chic ago? Which neighborhoods did they grow up in? How much education did your parents have? Do you have any siblings? How old are they? Where do they live?
282 APPENDIX E WHICH PARTICIPANTS PARTICIPATED TOGETHER OR ALONE Table E 1 shows which participants completed the experiment as a pair, and the participant completed the experiment without a partner. Table E 1. Participants who completed the experiment alone and in pairs, and the relationship of those who participated in pairs (J = Lifelong Jasper, M = Mobile, C = Lifelong Chicago) Participant Participated with Relationship J1 J2 J3 J4 J5 M6 C7 Friends M8 M9 M10 C11 Employer/employee (at law firm) M12 C14 C13 (excluded) Co workers (at law firm) C15 M16 C17 Boyfriend/girlfriend M18 M19 C20 Friends M22 C23 Friends M24 C25 (excluded) Friends M26 C27 Girlfriend/boyfriend M28 C29 Friends M30 C31 Friends M32 M33 C34 Friends C35 M36 Friends M37 M38 Married M39 C40 (excluded) Friends C41 M42 C43 Employer/employee (mom/nanny) C44 M46 C47 C48 Married J49 J50 J51
283 Table E 1. Continued Participant Participated with Relationship J53 J54 J55 J56 J57 J59 J61 J60 (excluded) Married J62 J63 J64 J67 J68 J69 M70 M71 M73 M76 C77 Mother/daughter M78 C79 C80 C81 C83 C84 C85 J86 J87 J88 Friends J89 J90
284 APPENDIX F INDIVIDUAL MOBILE / BACKING COMPARED TO THEIR SOCIAL CHARACTERISTICS of / / backing and compares it to their social characteristics. Table F 1 shows each Mobil Dead~Dud continuum, compared to the mean perceptual boundaries for the Lifelong Chicago and Lifelong Jasper groups. Tables F 2 and F 3 separate the Mobile participants into groups whose perception w as consistently Chicago like and consistently Jasper like and then list their social like; it was Jasper like on the F1 and F2 scales, but between Jasper and Chicago like for the F2F1d scale. periment with someone they thought qualified as a Lifelong Chicagoan who later had to be excluded from the experiment because s/he did not meet the eligibility criteria.
285 Table F 1. Perceptual boundaries on the Dead~Dud continuum for individual Mob ile participants, with means for the Lifelong groups Group or participant Crossover point for d ifference between F1 and F2 Group or participant Crossover p oint for F1 Group or participant Crossover p oint for F2 Participant 22 6.18 Participant 22 5.3746 Participant 22 11.5036 Participant 39 6.26 Participant 39 5.3409 Participant 39 11.599 Participant 28 6.31 Participant 76 5.3328 Participant 76 11.622 Participant 18 6.34 Participant 38 5.3298 Participant 38 11.6305 Participant 38 6.34 Participant 28 5.328 Participant 28 11.6359 Participant 76 6.34 Participant 18 5.3211 Participant 18 11.6566 Participant 9 6.42 Participant 9 5.3001 Participant 9 11.7197 Participant 32 6.42 Participant 32 5.3001 Participant 32 11.7197 Participant 42 6.42 Participant 78 5.3001 Participant 78 11.7197 Participant 78 6.42 Participant 42 5.2999 Participant 42 11.7202 Lifelong Chicago mean 6.4492 Lifelong Chicago mean 5.2939 Lifelong Chicago mean 11.7349 Participant 71 6.5 Participant 73 5.2716 Participant 73 11.8051 Participant 73 6.53 Participant 16 5.2716 Participant 16 11.8051 Participant 16 6.53 Participant 12 5.2716 Participant 12 11.8051 Participant 12 6.53 Participant 10 5.2674 Lifelong Jasper mean 11.8158 Participant 10 6.55 Participant 6 5.2674 Participant 10 11.8164 Participant 6 6.55 Participant 19 5.2674 Participant 6 11.8164 Participant 19 6.55 Participant 8 5.2674 Participant 19 11.8164 Participant 8 6.55 Participant 30 5.2674 Participant 8 11.8164 Participant 30 6.55 Participant 36 5.2674 Participant 30 11.8164 Participant 36 6.55 Participant 46 5.2674 Participant 36 11.8164 Participant 46 6.55 Lifelong Jasper mean 5.2649 Participant 46 11.8164 Lifelong Jasper mean 6.5519 Participant 71 5.2517 Participant 71 11.8546 Participant 24 6.66 Participant 24 5.2351 Participant 24 11.8947 Participant 26 6.66 Participant 26 5.2351 Participant 26 11.8947 Participant 33 6.66 Participant 33 5.2351 Participant 33 11.8947 Participant 37 6.66 Participant 37 5.2351 Participant 37 11.8947 Participant 70 6.66 Participant 70 5.2351 Participant 70 11.8947
286 Table F 2. Mobile group members separated by perception style, with the Chicago neighborhoods in which they had lived and worked Mobile group Neighborhoods lived in Neighborhoods worked in Consistently Chicago like 9 Lakeview, Uptown 18 Lakeview South Loop, Wicker Park, Loyola U. 22 Ravenswood, Lincoln Square Glenview 28 Lincoln Park Downtown 32 Evanston, Wrigleyville, Lincoln Park, the Loop Downtown/The Loop 38 Wrigleyville, Gold Coast, River North Lincoln Park, Hyde Park, Wilmette, UIC 39 Woodridge, Downers Grove Downtown, Aurora 42 West Dundee, River North, Orland Park All over suburbs 76 Arlington Heights, Wrigleyville, Lincoln Park, Wilmette, Kennilworth, Winnetka 78 Mount Prospect, Arlington Heights, Lincoln Park, Wilmette All over, Deerfield, Lincolnshire, River Woods Consistently Jasper like 6 Gold Coast, Lakeview, Boystown 8 Hyde Park, Gold Coast, Lincoln Park Hyde Park, Gold Coast, Loop 10 Woodridge, Downers Grove Downtown 12 Hyde Park, Lincoln Park, River North, Lakeview 16 Gold Coast 19 Gold Coast The Loop 24 Evanston, Lincoln Park The Loop 26 Lakeview, Streeterville, Bucktown Loop, South Loop, River North 30 Rogers Park, Andersonville, Ravenswood, Lakeview, Oak Park, Edgewater Lincoln Park 33 Lakeview, North Center Downtown 36 Streeterville, Lakeview River North, Loop 37 Wrigleyville, River North River North 46 Tinley Park, Schaumburg, Brookfield The Loop 70 West Loop Downtown, Gold Coast, Lincoln Park 71 Lincoln Park, Lakeview The Loop 73 Glen Ellyn, Elmhurst Downtown, suburbs, Oakbrook
287 Table F 3. Mobile group members separated by perception style, with other social characteristics Mobile group Education How many years lived elsewhere Years lived in Chicago Age of arrival in Chicago % of life spent in Chicago Interviewed with Chicagoan Consistently Chicago like 9 17 6 3 24 11.1 N 18 17 3 1.5 23 6.3 N 22 18 5.5 18 23 43.9 Y 28 16 4 1.8 23 7.2 Y 32 17 6.25 3 27 10 N 38 16 4.75 4 22 13.8 N 39 18 5 30 24 55.6 Sort of 42 16 10.75 7 30 18.9 Y 76 16 3.25 30.8 22 58.1 Y 78 16 3.25 31.75 22 58.8 N Mean: 16.7 5.18 13.09 24 28.37 45% Median: 16.5 4.88 5.5 23 16.35 Consistently Jasper like 6 18 4.25 3.75 23 13.9 Y 8 18 5.8 13.5 24 35.5 N 10 18 4.5 31 23 56.4 Y 12 18 3.5 7.5 23 24.2 N 16 16 5.75 2.5 24 9.3 Y 19 16 5.5 1.67 24 6.4 Y 24 16 13.25 3.5 32 10.3 Sort of 26 16 4.1 4 23 14.8 Y 30 16 3.5 19.25 27 41 Y 33 18 8.75 6 27 18.2 Y 36 16 4.125 4.5 23 16.7 Y 37 17 5.25 5 22 17.2 N 46 16 3.25 4 22 15.4 N 70 13 0 1 18 5 N 71 18 6.8 1.33 25 4.9 N 73 16 6.25 33.5 25 56.8 N Mean: 16.63 5.29 8.88 24.06 21.63 53% Median: 16 4.88 4.25 23.5 16.05
288 APPENDIX G PRODUCTION RESULTS MIXED EFFECTS REGRESSIONS This appendix presents the results of mixed effects regressions run in the R statistical package on the production data. As explained in Chapter 5, I ran several rounds of mixed effects regressions on each combination of time poi nt, formant, and normalization method. Here, for each combination I present the final mixed effects regression run that did not include interaction terms, as well as later runs testing interaction. Table G 1 explains what each dummy variable (for categor ical predictor variables) in the mixed effects regression runs means and also presents the default values for each of these categorical variables. I then present the results using the Nearey, then the Lobanov, and finally the LobBark normalized values. I n the results, statistically significant predictor variables (|t|>1.96) are in bold. Marginally significant social variables (|t|>1.65) are underlined. Following the results of the mixed effects regressions, I present the results for correlations with th in Chicago (AAC), Years Lived in Chicago (YLC), and Percentage of Life Spent in Chicago (PLC). The data for this Appendix can be accessed through the link below. Object G 1. Mixed Effects Regression Results
289 APPENDIX H FURTHER VOWEL CHARTS The vowel charts at the start of the results section of Chapter 5 mapped the vowel space for each residence group using every individual vowel. Since those charts are a bit difficult to read, this appendix shows similar charts with t he information made clearer in two ways. First, Figures H 1 through H value for each vowel at each time point. As a second way of displaying the data more clearly than in a scatter plot of all the vowels of each residence group at once, I then once again present all the individual vowels of each speaker, but this time with each residence group divided further into three groups. Figures H 16 through H 30 show the vowels of the oldest, middle age range, and youngest Lifelong Jasper participants; Figures H 31 through H 45 show the vowels of the Mobile participants who have spent the least, mid range, and most time in Chicago; and Figures H 46 through H 60 show the vowels of the oldest, middle age range, and youngest Lifelong Chicago participants.
290 Figure H 1. Vowel space for Lifelong Jasper participants at T1 average value s hown
291 Figure H 2. Vowel space for Lifelong Jasper participants at T2 averages only Figure H 3. Vowel space for Lifelong Jasper participants at T3 averages only
292 Figure H 4. Vowel space for Lifelong Jasper participants at T4 averages only Figure H 5. Vowel space for Lifelong Jasper participants at T5 averages only
293 Figure H 6. Vowel space for Mobile participants at T1 averages only (Note the greater variation, especially in //, than was present for the Li felong Jasper group) Figure H 7. Vowel space for the Mobile group at T2 averages only
294 Figure H 8. Vowel space for Mobile participants at T3 averages only (Note that measuring only at the midpoint would have hidden the initial raising and fronting of // shown by some Mobile participants) Figure H 9. Vowel space for Mobile participants at T4 a verages only
29 5 Figure H 10. Vowel space for Mobile participants at T5 averages only Figure H 11. Vowel space for Lifelong Chicago participants at T1 averages only (Extreme raising and fronting of // as well as backing of / / and / / are apparent)
296 Figure H 12. Vowel space for Lifelong Chicago participants at T2 averages only
297 Figure H 13. Vowel space for Lifelong Chicago participants at T3 averages only (As with the Mobile group, it is apparent that the most extreme regional variation
298 Figure H 14. Vowel space for the Lifelong Chicago participants at T4 averages only Figure H 15. Vowel space for the Lifelong Chicago participants at T5 averages only
299 Figure H 16. Vowel space for the nine oldest Lifelong Jasper participants (YOB 1949 1972) at T1 (A = //, E = / /, U = / /) Figure H 17. Vowel space for the nine oldest Lifelong Jasper participants at T2
300 Figure H 18. Vowel space for the nine oldest Lifelong Jasper participants at T3 Figure H 19. Vowel space for the nine oldest Lifelong Jasper participants at T4
301 Figure H 20. Vowel space for the nine oldest Lifelong Jasper participants at T5 Figure H 21. Vo wel space for the six Lifelong Jasper participants in the middle age range (YOB 1974 1980) at T1
302 Figure H 22. Vowel space for the six Lifelong Jasper participants in the middle age range at T2 Figure H 23. Vowel space for the six Lifelong Jasper p articipants in the middle age range at T3
303 Figure H 24. Vowel space for the six Lifelong Jasper participants in the middle age range at T4 Figure H 25. Vowel space for the six Lifelong Jasper participants in the middle age range at T5
304 Figure H 26. Vowel space for the eleven youngest Lifelong Jasper participants (YOB 1981 1989) at T1 Figure H 27. Vowel space for the eleven youngest Lifelong Jasper participants at T2
305 Figure H 28. Vowel space for the eleven youngest Lifelong Jasper participants at T3 Figure H 29. Vowel space for the eleven youngest Lifelong Jasper participants at T4
306 Figure H 30. Vowel space for the eleven youngest Lifelong Jasper participants at T5 Fig ure H 31. Vowel space for the nine Mobile participants who have spent the least amount of time in Chicago (1 year 3.5 years) at T1
307 Figure H 32. Vowel space for the nine Mobile participants who have spent the least amount of time in Chicago at T2 Figure H 33. Vowel space for the nine Mobile participants who have spent the least amount of time in Chicago at T3
308 Figure H 34. Vowel space for the nine Mobile participants who have spent the least amount of time in Chicago at T4 Figure H 35. Vo wel space for the nine Mobile participants who have spent the least amount of time in Chicago at T5
309 Figure H 36. Vowel space for the nine Mobile participants who have spent the middle range of time (3.75 years 7.5 years) in Chicago at T1 Figure H 3 7. Vowel space for the nine Mobile participants who have spent the middle range of time in Chicago at T2
310 Figure H 38. Vowel space for the nine Mobile participants who have spent the middle range of time in Chicago at T3 Figure H 39. Vowel space for the nine Mobile participants who have spent the middle amount of time in Chicago at T4
311 Figure H 40. Vowel space for the nine Mobile participants who have spent the middle amount of time in Chicago at T5 Figure H 41. Vowel space for the eight Mobile participants who have spent the most time (13.5 years 33.5 years) in Chicago at T1 (Note the raising and fronting of // that was less noticeable for the other two subgroups of Mobile participants)
312 Figure H 42. Vowel space for the eight Mobile partic ipants who have spent the most time in Chicago at T2 Figure H 43. Vowel space for the eight Mobile participants who have spent the most time in Chicago at T3
313 Figure H 44. Vowel space for the eight Mobile participants who have spent the most time in Chicago at T4 Figure H 45. Vowel space for the eight participants who have spent the most time in Chicago at T5
314 Figure H 46. Vowel space for the eight oldest Lifelong Chicago participants (YOB 1953 1965) at T1. Raising and fronting of // as well a s backing and lowering of both / / and / / are apparent. Figure H 47. Vowel space for the eight oldest Lifelong Chicago participants at T2
315 Figure H 48. Vowel space for the eight oldest Lifelong Chicago participants at T3 Figure H 49. Vowel space for the eight oldest Lifelong Chicago participants at T4
316 Figure H 50. Vowel space for the eight oldest Lifelong Chicago participants at T5 Figure H 51. Vowel space for the eight Lifelong Chicago participants in the middle age range (YOB 1966 1979) at T1. // is again raised and fronted, and / / is lowered, but the backing of / / and backing and lowering of / / are less striking than they were for the older group.
317 Figure H 52. Vowel space for the eight Lifelong Chicago participants in the middle age group at T2 Figure H 53. Vowel space for the eight Lifelong Chicago participants in the middle age group at T3
318 Figure H 54. Vowel space for the eight Lifelong Chicago partici pants in the middle age group at T4 Figure H 55. Vowel space for the eight Lifelong Chicago participants in the middle age group at T5
319 Figure H 56. Vowel space for the eight youngest Lifelong Chicago participants (YOB 1980 1987) at T1. / / and / / a re noticeably backed Figure H 57. Vowel space for the eight youngest Lifelong Chicago participants at T2
320 Figure H 58. Vowel space for the eight youngest Lifelong Chicago participants at T3 Figure H 59. Vowel space for the eight youngest Lifelong Chicago participants at T4
321 Figure H 60. Vowel space for the eight youngest Lifelong Chicago participants at T5
322 APPENDIX I CHARTS COMPARING PERCEPTION AND PRODUCTION FOR MOBILE AND LIFELONG JASPER PARTICIPANTS This appendix presents vowel charts charting Mobile and Lifelong Jasper /~/ / and / /~/ /) along with their perceptual production tokens for the vowels in question (not the average values for several participants). On the charts, the production tokens are represented by A for / /, E for / /, and U for / /. On the graphs of / /~/ /, the perceptual crossover is represente d by P. On the graphs of / /~/ /, the perceptual crossover is represented by X. For each participant and vowel pair, there are two charts, one each showing production and perception at T3 and T4, the two time points used for the perception/production comp arison. (The perceptual crossover point is the same in the two graphs.) The charts for this appendix can be accessed through the link below. Object I 1. Charts comparing perception and production for Mobile and Lifelong Jasper participants.
323 LIS T OF REFERENCES Auer, Peter. 1997. Fhrt Dialektabbau zur Strkung oder Schwchung der Standardvariett? Zwei phonologische Fallstudien. Standardisierung und Destandardisierung ed. by Klaus J. Mattheier and Edgar Radtke, 129 62. Frankfurt: Lang. Auer, Pet er; Birgit Barden; and Beate Grosskopf. 1998. Subjective and objective term dialect accommodation. Journal of Sociolinguistics 2.163 87. Bakan, David. 1966. The duality of human existence Chicago: Rand McNally. Bem, S andra L 1974. The measurement of psychological androgyny Journal of Consulting and Clinical Psychology 42 .155 62. Bittson, Genevieve. 2007 Gender, sex, and the sociolinguistic variable [ ing]. Gainesville, FL: University of Florida thesis. Blondea u, Hlne and Naomi Nagy. 2008. Subordinate clause marking in Montreal Anglophone French and English. Social lives in language Sociolinguistics and multilingual speech communities ed. by Miriam Meyerhoff and Naomi Nagy, 273 314. Amsterdam: John Benjamin s. Bowie, David. 2000. The effect of geographic mobility on the retention of a local dialect. Philadelphia, PA: University of Pennsylvania dissertation. Bowie, David. 2005. Language change over the lifespan: A test of the apparent time construct. Universit y of Pennsylvania Working Papers in Linguistics 11.45 58. Carver, Craig M. 1987 American regional dialects: A word g eography. Ann Arbor, MI: University of Michigan Press. Chambers, J.K. 1992. Dialect acquisition. Language 68.673 705. Cheshire Jenny. 2002 Sex and gender in variationist research. The h andbook of language variation and c hange ed. by J.K. Chambers, Peter Trudgill, and Natalie Schilling Estes, 423 43. (Blackwell handbooks in linguistics.) Malden, MA: Bla ckwell Publishers Inc. Clopper, Cynth ia G. and David B. Pisoni. 2004. Some acoustic cues for the perceptual categorization of American English regional dialects. Journal of Phonetics 32. 111 40. Clopper, Cynthia G. and David B. Pisoni. 2006. Effects of region of origin and geographic mobility on perceptual dialect categorization. Language Variation and Change 18.193 221.
324 Conn, Jeff and Uri Horesh. 2002. Assessing the acquisition of dialect variables by migrant adults in Philadelphia: A case study. University of Pennsylvania Working Papers in Linguistics 8.47 57. Davis, Lawrence M. and Charles L. Houck. 1992. Is there a Midland dialect area? again. American Speech 67.61 70. Dodsworth, Robin. 2005. Attribute networking: A technique for modeling social perceptions. Journal of Sociolinguistics 9.2 25 53. Eckert, Penelope. 1988. Adolescent social structure and the spread of linguistic change. Language in Society 17.183 207. Eckert, Penelope. 1991. Social polarization and the choice of linguistic variants. New ways of analyzing sound change ed. by Penelope Eckert, 213 32. New York, NY: Academic Press, Inc. Evans, Bronwen G. and Paul Iverson. 2007. Plasticity in vowel perception and production: A study of accent change in young adults. Journal of the Acoustical Society of America 121.3814 26. Fahey, Richard P.; Randy L. Diehl; and Hartmut Traunmller. 1996. Perception of back vowels: Effects of varying F 1 F 0 Bark distance. Journal of the Acoustical Society of America 99.2350 57. Fought, Carmen. 1999. A majority sound change in a minority community: /u/ fronting in Chicano English. Journal of Sociolinguistics 3.5 23. Fridland, Valerie; Kathryn Bartlett; and Roger Kreuz. 2004. Do you hear what I hear? Experimental measurement of the perceptual salience of acoustically manipulated vowel variants by So uthern speakers in Memphis, TN. Language Variation and Change 16.1 16. Christmas broadcasts. Journal of Phonetics 34.439 57. Harrington, J onathan; Sallyanne Palethorpe; and Catherine I. Watson. 2000a Does the Nature 408 .927 28. Harrington, J onathan; Sallyanne Palethorpe; and Catherine I. Watson. 2000b. Monophthongal vowel changes in received pronunc iation: An acoustic analysis of Christmas broadcasts. Journal of the International Phonetic Association 30. 63 78.
325 Harrington Jonathan; Sallyanne Palethorpe; and Catherine Watson. 2005 Deepening annual Christmas br oadcasts. A figure of speech: A f estschrift for John Laver ed. by William J. Hardcastle, and Janet Mackenzie Beck, 227 61 Mahwah, NJ: Lawrence Erlbaum. Herold, Ruth. 1990. Mechanisms of merger: The implementation and distribution of the low back merger i n eastern Pennsylvania. Philadelphia, PA: University of Pennsylvania dissertation. Houck, Charles. 1967. A computerized statistical methodology for linguistic geography: A pilot study. Folia Linguistica 1.80 95. House, Arthur S. and Grant Fairbanks. 1953. The influence of consonant environment upon the secondary acoustical characteristics of vowels. Journal of the Acoustical Society of America 25.105 13. Janson, T ore 1983. Sound change in perception and production. Language 59.18 34. Janson, Tore, and Rich ard Schulman. 1983. Non distinctive features and their use. Journal of Linguistics 19.321 36. Johnson, Ellen. 1994. Yet again: The Midland dialect. American Speech 69.419 30. Johnson, Keith. 2003. Acoustic & Auditory Phonetics 2 nd edition. Malden, MA: Blackwell Publishers Inc. Kurath, Hans. 1949. A word geography of the eastern United States. Ann Arbor, MI: University of Michigan Press. Labov, William. 1963. The social motivation of a sound change. Word 19. 273 309. Labov, William. 1989. The limitations of context. Chicago Linguistics Society 25. 171 200. Labov, William. 1990. The intersection of sex and social class in the course of linguistic change. Language Variation and Change 2.205 54. Labov, William. 1991. The three dialects of English. New ways of analyzing sound change ed. by Penelope Eckert, 1 44. New York NY: Academic Press, Inc. Labov, William. 1994. Principles of linguistic change. Vol 1: Internal factors. Oxford: Blackwell. L abov, William, and Sharon Ash.1997 Understanding Birmingham. Lan guage variety in the South revisited ed. by Cynthia Bernstein, Thomas Nunnally, and Robin Sabino, 508 73. Tuscaloosa, AL: University of Alabama Press.
326 Labov, William; Sharon Ash; and Charles Boberg. 2006. The atlas of north american English. Berlin: Mouto n de Gruyter. Labov, William; Malcar Yaeger; and Richard Steiner. 1972. A quantitative study of sound change in progress. Philadelphia, PA: U.S. Regional Survey. Ladefoged, Peter and Ian Maddieson. 1996. anguages. Oxford: Blackwe ll Publishing Ltd. Lenneberg, Eric H. 1967. Biological foundations of Language. New York, NY: Wiley. Lobanov, B.M. 1 971. Classification of Russian vowels spoken by d ifferent s peakers. Journal of the Acoustical Society of America 49.606 08. MacKenzie, Laure l and Gillian Sankoff. 2010. A quantitative analysis of diphthongization in Montreal French. University of Pennsylvania Working Papers in Linguistics 15. No page numbers. Majors, Tivoli. 2005. Low back vowel merger in Missouri speech: Acoustic description and explanation. American Speech 80.165 79. Matsumoto, Kazuko and David Britain. 2003. Investigating the sociolinguistic gender paradox in a multilingual community: A case study from the Republic of Palau. International Journal of Bilingualism 7.127 52. McCarthy, Corrine. 2011. The Northern Cities Shift in Chicago. Journal of English Linguistics 39.166 87. Murray, Thomas E. 1993. Positive anymore in the Midwest. Variation and transition in the American Midwest ed. by Timothy C. Fraze r, 173 86 Tuscaloosa, AL: University of Alabama Press. Murray, Thomas E. and Beth Lee Simon. 2006. What is dialect? Revisiting the Midland. Language variation and change in the American Midland: A new look at ed. by Thomas E. Murray a nd Beth Lee Simon, 1 30. Philadelphia, PA: John Benjamins Publishing Company. Nearey, T.M. 1978. Phonetic f eatur e systems for v owels. Bloomington, IN: Indiana University Linguistics Club. Niedzielski, Nancy. 1999. The effect of social information on the pe rception of sociolinguistic variables. Journal of Language and Social Psychology 18.62 65. Nosofsky, Robert M. 1986. Attention, similarity, and the identification categorization relationship. Journal of Experimental Psychology 115.39 57. Nycz, Jennifer R. 2011. Second dialect acquisition: Implications for theories of phonological representation New York, NY: New York University dissertation.
327 P ayne, Arvilla 1980. Factors controlling the acquisition of the Philadelphia dialect by out of state children. Loca ting language in time and s pace ed. by William Labov, 143 78. New York NY : Academic Press. Pederson, Lee. 1965. The pronunciation of English in metropolitan Chicago Tuscaloosa, AL: University of Alabama Press for the American Dialect Society. Preacher, K. J. 2001. Calculation for the chi square test: An interactive calculation tool for chi square tests of goodness of fit and independence. Online: http://quantpsy.org Preston, Dennis R. 2003. Where are the dialects of American English at anyhow? American Speech 78.235 54. Rastatter, M.P.; R. McGuire; J. Kalinowski; and A. Stuart. 1997. Formant frequency characteristics of elderly speakers in contextual speech. Folia Pho niatrica 49.1 8. Sankoff, Gillian. 2004. Adolescents, young adults, and the critical period: Two case Socolinguistic v ariation: Critical r eflections ed. by Carmen Fought, 121 39. (Oxford Studies in Sociolinguistics.) New York, NY: Oxford University Press. Sankoff, Gillian and Hlne Blondeau. 2007. Longitudinal change across the lifespan: /r/ in Montreal French. Language 83.560 88. Schirmunski, Victor. 1928/1929. Die schwbischen Mundarten in Transkaukasien und Sdukraine. Teuthoniste 5.38 60 (part I), 151 71 (part II). Schirmunski, Victor. 1930. Sprachgeschichte und Siedlungsmundarten. Germanistich Romanistische Monatsschrift 18.113 22, 171 88. Spence, Janet T.; Robert L. Helmreich; and Carole K. Holahan. 1979. Negative and positive components of psychological masculinity and femininity and their relationships to self reports of neurotic and acting out behaviors. Journal of Personality and Social Psychology 37.1673 82. Stockwell, Robert, and Donka Minkova. 1997. On drifts an d s hifts. Studia Anglica Posnaniensia 31. 283 303. a second dialect of the same language. Language in Society 36.649 75. Thomas, Erik R. 2001. An acoustic analysis of vowel variation in New World English. Durham, NC: Duke University Press. Traunmller, Hartmut. 1990. Analytical expressions for the tonotopic sensory scale. Journal of the Acoustical Society of America 88.97 100.
328 Trudgill, P eter 1972 Sex, covert prestige and linguistic change in the urban British English of Norwich. Language in Society 1 : 179 95. Trudgill, Peter. 1986. Dialects in Contact. (Language in Society.) Oxford: Basil Blackwell Ltd. U.S. Census Bureau. 2012a. State and county quickfacts: Jasper (city) Indiana. Online: http://quickfacts.census.gov U.S. Census Bureau. 2012b. State and county quickfacts: Chicago (city), Illinois. Online: http://quickfacts.census.gov van der Harst, Sander. 2011. The vowel space paradox: A sociophonetic study of Dutch. Utrecht: Radboud Universiteit Nijmegen dissertation. Wagner, Suzanne Evans. 2012. Age grading in sociolinguistic theory. Language and Ling uistics Compass 6.371 82. Wolfram, Walt and Natalie Shilling Estes. 1998. American English: Dialects and Variation Malden, MA: Blackwell Publishers Inc. Zwicker, E. 1961. Subdivision of the audible frequency range into critical bands (Frequenzgruppen). Jo urnal of the Acoustical Society of America 33.248.
329 BIOGRAPHICAL SKETCH Zoe Lynn Ziliak was raised in Jasper, Indiana. She graduated from the University of Oklahoma with summa cum laude B.A.s in English literature and German. Prior to completing her doc torate, she earned an M.A. in linguistics from the University of Florida She is currently the graduate fellowship advisor in the Office of Graduate Studies at the University of California, San Diego. She lives in San Diego with her husband, Dan Michel.