1 A VARIATIONIST ACCOUNT OF VOICE ONSET TIME AMONG BILINGUAL WEST INDIANS IN PANAMA By DELANO SYDNEY LAMY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMEN TS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012
2 2012 Delano Sydney Lamy
3 P ara mi hijo, Delano Emmanuel por llegar en el momento oportuno a llenar mi vida de luz durante este proceso
4 ACKNOWLEDGMENTS I would first lik e to express my gratitude to the faculty of the Department of Spanish and Portuguese Studies and to my dissertation committee. I especially want to thank my advisor, J essi Aaron for introducing me to the world of variationist so ciolinguistics and function alist grammar, and for seeing the potential in me to be a good researcher. I truly appreciate the confidence you have had in me throughout these years. To Gillian Lord thank you for your valuable advice and for allowing me to gain so much teaching experie nce during my years at UF. To James Essegbey, thank you for encouraging me to delve further into the world of creolistics. To Hlne Blondeau, thank you for your thoughtful comments and having me see the big picture in this study. To Ana de Prada thank yo u for suggesting that I study voice onset time. Estabas ah desde el comienzo de este proyecto. To Jason Rothman, thank you for having work with you. To David Pharies, tha nk you so much for your constant praises. Although I may not have expressed it, your view of me as an academic and an intellectual means a lot to me. I also want to thank Sue Ollmann, Bobbye Pilkington, and Tania Fleming for all their work in Dauer 170. I would also like to thank Luca Montas, Moniqua Acosta Heyman, Alexander Torres, Valerie Trujillo, Meagan Day, Sara Zahler, Becky Halloran, Jake Firestine, Tiffany Judy, Anne Lingwall, Donna Gillespie, Elisabet Liminyana, Francesc Morales, Gerardo Muoz, an d Heather Kaiser for their support during my time at UF. You have been such great colleagues, and most importantly, friends. I also want to express my sincerest gratitude to Diego Pascual y Cabo por siempre estar pendiente y por darme nimo en los tiempos dif ciles. Desde que te conoc, has sido tan bueno co n migo, y te
5 lo agradezco mucho. To Jennifer Cabrelli Amaro thank you for helping me with the rater validation for this study, but even more for being such a great person and for teaching me Portuguese. I especially want to acknowledge my very close friends, Maria Fionda and Katherine Honea. You have been such a support to me and my family. It has been an amazing experience to study linguistics with you and to be friends outside the world of academia. I truly appreciate the friendship I have with both of you. I also want to acknowledge the families of the SPS graduate students, especially Chad Atkinson, Joo Felipe Amaro, Laurie Donnelly Pascual, and Don Trujillo. You were a support not only to me, but al so to Briane and Emmanuel. I must acknowledge Daniel Johnson for introducing me to Rbrul last year at SSS3 in Glasgow, and for being so available to me after that to help me with statistics. I truly appreciate your assistance in this project. I also want to recognize Jennifer Leeman, my first sociolinguistics profes s or who encouraged me to pursue an academic career in Hispanic Linguistics. I am so grateful that we crossed paths. A La familia Bailey, gracias por conseguirme muchas de la s entrevista s que hice en la C iudad de Panam Agradezco de todo corazn su ayuda y su apoyo A Raul Houlstan for your support in Bocas del Toro and for allowing me to use you as a contact to collect interviews in Bastimentos. Also, to La Sociedad de Amigos del Museo Afro A ntillano de Panama ( SAMAAP ) and Professor Ifeoma Kiddoe Nwankwo, principal investigator of Voices from Our America (VFOA), for their work in the West Indian community of Panama. Quiero agradecerle a mi compadre Luis Martnez, quien realmente es como un he rmano. T has sido un apoyo espiritual increble para m. Estar siempre agradecido
6 por tu amistad y tus consejos. Tambin quiero darles gracias a sus padres, Luis Martnez y la seora Marianela por ser como mis padres all en Panam. Me aceptaron en su ho gar y me trataron como un hijo ms. Los quiero mucho. Quiero reconocer tambin a mi amiga, Anamara Bejarano vila, quien siempre me apoya en todo y siempre est disponible para ayudarme en mis estudios. Agradezco mucho tu amistad. Tambin a "scar Lara Yej as, gracias por escribirme el programa. Â¡Me ahorraste tanto tiempo! I also want to acknowle dge my wonderful family: mis tas Celia y Jacqueline, my grandma Lynn, my father, my brothers and sister, and mis tos Ren y Claudio. I especially want to thank my cousin Oneal Bastidas y su esposa, Joanna por permitirme quedar en su casa cada vez que viajo a Panam Gracias a sus hijos tambin por tratarme no slo como un primo, pero tambin como un to. Los quiero de todo corazn. I want to t h an k my mother who h as supported me all my life. The only thing I can say is that you are truly a blessing from God. I want to acknowledge my brother, Daanyal, for showing me love and support. To Maple, thank you for everything you have done for me and my family during my col lege career. Lastly, but certainly not least, I want to thank my wife, Briane Accius and my son, Delano Emmanuel. You two are the reason why I have been able to persevere in this process. Thank you for putting up with me, Briane, through all the ups and d owns of this dissertation process. I thank God for your presence in my life. Delano Emmanuel, mi principito, gracias po r aguantar horas en la cuna mientras escriba esta tesis. Te amo, mi hijito. Todo esto, lo hice para ti.
7 TABLE OF CONTENTS pag e ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ .......... 10 LIST OF FIGURES ................................ ................................ ................................ ........ 12 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 1.1 Language Contact in Panama ................................ ................................ .......... 15 1.2 Pidgins and Creoles ................................ ................................ .......................... 20 1.2.1 Theories of Creole Genesis ................................ ................................ ..... 22 1.2.2 The Creole Continuum ................................ ................................ ............ 26 1.2.3 Features of Creoles ................................ ................................ ................. 27 1.3 Bilingualism and Language Attitudes among West Indians in Panama ............ 29 1.4 Creole English speaking West Indians in Bastimentos ................................ ..... 32 1.5 The Language Contact Study ................................ ................................ ........... 37 2 LITERATURE REVIEW ................................ ................................ .......................... 40 2.1 Language Contact and Contact induced Change ................................ ............. 40 2.1.1 Social Factors ................................ ................................ .......................... 42 2.1.2 Linguistic Predictors and Mechanisms ................................ .................... 45 2.1.3 Linguistic Outcomes of Language Contact ................................ .............. 51 2.2 The Comparative Variationist Enterprise ................................ .......................... 53 2.3 Sound Change Due to Language Contact ................................ ........................ 55 2.3.1 Voice Onset Time ................................ ................................ .................... 56 2.3.2 Spanish VOT Values ................................ ................................ ............... 60 2.3.3 English VOT Values ................................ ................................ ................ 60 2.3.4 Factors Con ditioning VOT ................................ ................................ ....... 61 2.3.5 English vs. Creole English VOT ................................ .............................. 68 2.4 Usage based Models of Language ................................ ................................ ... 69 2.4.1 The Usage Event ................................ ................................ ..................... 70 2.4.2 Cognitive Processes ................................ ................................ ................ 71 2.4.3 The Social Aspect of the Usage based Approach ................................ ... 78 2.4.4 Second Language Phonology ................................ ................................ 79 3 METHODOLOGY ................................ ................................ ................................ ... 87 3.1 Language Use and the Vernacular ................................ ................................ ... 87 3.2 Participants ................................ ................................ ................................ ....... 89
8 3.3 Research Questions ................................ ................................ ......................... 91 3.4 Hypotheses ................................ ................................ ................................ ....... 91 3.5 Instrumental Analysis ................................ ................................ ........................ 93 3.5.1 Data Extraction ................................ ................................ ........................ 93 3.5.2 VOT Measurements in Praat ................................ ................................ ... 94 3.5.3 Rater Validation ................................ ................................ ....................... 97 3.6 Variationist Approach ................................ ................................ ........................ 98 3.7 Linguistic Factors ................................ ................................ ............................ 101 3.7.1 Preceding Segment ................................ ................................ ............... 103 3.7.2 Following Vo wel ................................ ................................ .................... 104 3.7.3 Position in Word ................................ ................................ .................... 104 3.7.4 Syllable Stress ................................ ................................ ....................... 105 3.7 .5 Rate of Speech ................................ ................................ ...................... 105 3.7.6 Lexical Item, Word Class, Word Frequency ................................ .......... 106 3.8 Comparative Variationist Method ................................ ................................ .... 108 3.9 Summary ................................ ................................ ................................ ........ 109 4 A VARIATIONIST ACCOUNT OF MEAN VOT OF /T/ IN A CONTACT SITUATION ................................ ................................ ................................ ........... 110 4.1 Linear Regression Analyses ................................ ................................ ........... 110 4.1.1 Distribution of Mean VOT Duration ................................ ........................ 110 4.1.2 Conditioning Factors in Monolingual Spanish ................................ ........ 112 4.1.3 Conditioning Factors in Monolingual Creole English ............................. 114 4.1.4 Conditioning Factors in Bilingual Spanish ................................ ............. 117 4.1.5 Conditioning Factors in Bilingual Creole English ................................ ... 119 4.2 Modeling Sound Change in Bilingual Speech ................................ ................. 122 4.2.1 Bilingual Spanish Compared to Monolingual Spanish and Monolingual Creole English ................................ ................................ ............................. 123 4.2.2 Bilingual Creole English Compared to Monolingual Spa nish and Monolingual Creole English ................................ ................................ ......... 127 4.2.3 Bilingual Spanish compared to Bilingual Creole English ....................... 131 4.3 Summary ................................ ................................ ................................ ........ 135 5 DISCUSSION ................................ ................................ ................................ ....... 138 5.1 Mean VOT and Conditioning Factors ................................ .............................. 138 5.2 Creole E nglish VOT and Creole Genesis Theories ................................ ......... 142 5.3 Sound Change in the Bilingual West Indian Speech Community .................... 143 5.3.1 Emerging Ant illano Spanish (AS) and Antillano English (AE) ................ 146 5.3.2 Qualitative Explanations for the Emergence of AS and AE ................... 150 188.8.131.52 Con vergence moving towards Creole English ............................. 152 184.108.40.206 Convergence moving towards Spanish ................................ ........ 153 220.127.116.11 Divergence ................................ ................................ ................... 157 5.4 Summary ................................ ................................ ................................ ........ 158 6 CONCLUSIONS ................................ ................................ ................................ ... 160
9 6.1 Quantitative Analyses of Language Contact and Change ............................... 160 6.1.1 Voice Onset Time ................................ ................................ .................. 161 6.1.2 Internally Motivated Change vs. Contact induced Change .................... 161 6.1.3 The Sociophonetic Study ................................ ................................ ....... 162 6.2 Theoretical Implications ................................ ................................ .................. 165 6.3 Limitations a nd Further Research ................................ ................................ ... 171 6.4 Conclusion ................................ ................................ ................................ ...... 172 APPENDIX A MEAN VOT VALUES OF INDIVIDUAL SPEAKERS ................................ ............. 174 B LANGUAGE BACKGROUND QUESTIONNAIRE ................................ ................. 175 C LANGUAGE ATTITUDES SURVEY ................................ ................................ ..... 176 LIST OF REFERENCES ................................ ................................ ............................. 178 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 193
10 LIST OF TABLES Table page 1 1 Caribbean creoles ................................ ................................ ............................. 22 3 1 Participants ................................ ................................ ................................ ........ 90 3 2 Rate validation correlation results ................................ ................................ ...... 9 7 3 3 Some factor weights (probabilities) and the corresponding log odds .............. 101 3 4 Linguistic factors coded ................................ ................................ ................... 102 3 5 Comparison of four language modes ................................ ............................... 109 4 1 Mean VOT, standard deviation, and range for the monolingual and bilingual varieties ................................ ................................ ................................ ............ 111 4 2 Fact ors conditioning mean VOT in Monolingual Spanish ................................ 112 4 3 Factors conditioning mean VOT in Monolingual Creole English ...................... 115 4 4 Factors conditioning mean VOT in Bilingual Spanish ................................ ...... 118 4 5 Factors conditioning mean VOT in Bilingual Creole English ............................ 120 4 6 Comparison of conditioning factors: Monolingual Spanish vs. Bilingual Spanish ................................ ................................ ................................ ............ 124 4 7 Comparison of conditioning factors: Monolingual Creole English vs. Bilingual Spanish ................................ ................................ ................................ ............ 125 4 8 Monolingual varieties vs. Bilingual Spanish: internally motivated change in word class ................................ ................................ ................................ ......... 126 4 9 Monolingual varieties vs. Bilingual Sp anish: contact induced change in preceding segment ................................ ................................ ........................... 127 4 10 Monolingual varieties vs. Bilingual Spanish: contact induced change in position ................................ ................................ ................................ ............. 127 4 11 Comparison of conditioning factors: Monolingual Creole English vs. Bilingual Creole English ................................ ................................ ................................ .. 128 4 12 Comparison of conditioning factors: Monolingual Spanish vs. Bili ngual Creole English ................................ ................................ ................................ .. 129
11 4 13 Monolingual varieties vs. Bilingual Creole English: internally motivated change in vowel height ................................ ................................ ..................... 130 4 14 Monolingual varieties vs. Bilingual Creole English: internally motivated change in word class ................................ ................................ ........................ 131 4 15 Comparison between bilingual varieties ................................ .......................... 134 5 1 Comparison of mean VOT across varieties ................................ ..................... 138 5 2 Factors showing internally motivated change ................................ .................. 146 5 3 Factors showing contact induced change ................................ ....................... 147 5 4 Convergence between bilingual varieties in significance of rate of speech ..... 149 5 5 Convergence between bilingual varieties in preceding segment ..................... 149 5 6 Convergence between bilingual varieties in position ................................ ....... 149 5 7 Individual speaker as random effect in linear regression model ...................... 151
12 LIST OF FIGURES Figure page 2 1 A waveform showing voice onset time and voiceless closure interval of a word initial stop ................................ ................................ ................................ ... 57 2 2 A spectrographic and waveform analysis of the Spanish phrase tal vez ................................ ................................ ................................ ............... 58 3 1 Monolingual Bilingual Continuum ................................ ................................ ...... 91 3 2 Spectrograph and waveform land in monolingual Spanish ...... 94 3 3 ......... 95 3 4 all in bilingual Spanish .................... 95 3 5 ................ 96 4 1 Comparison of mean VOT across four varieties ................................ .............. 132 5 1 Assimilation and dissimation in bilinguals ................................ ........................ 140
13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy A VARIATIONIST ACCOUNT OF VOICE ONSET TIME AMO NG BILINGUAL WEST INDIANS IN PANAMA By Delano Sydney Lamy August 2012 Chair: Jessi Elana Aaron Major: Romance Languages The present study is concerned with language contact between Creole English and Spanish spoken by bilingual West Indians who live in Panama City, Panama. The goal of this study is to examine the speech patterns of monolinguals of Creole English and Spanish and Spanish Creole English bilinguals in the local communities of this region, by employing the comparative variationist method. This method allows us to tease apart internally motivated change from contact induced change (Poplack & Levey, 2010) occurring in bilingual speech A statistical comparison of the factors contributing to voice onset time (VOT) of the voiceless dental plosiv e /t/ in four speech modes address es the possibility of phonetic permeability due to contact in this region. A total of 2128 occurrences of /t/ were included in the analysis, with a sample extracted from monolingual Spanish, monolingual Creole English, and Creole English and Spanish of bilingual speakers. The VOT measurements were done using Praat (Boersma & Weenink, 2006 ). In order to pinpoint linguistic patterns that would aid in ascertaining variation and change with regard to V OT, a linear regression model with mixed effects w as incorporated using Rbrul (Johnson, 2009 ). Based on the hierarchies of the constraints that condit ion VOT duration, a series of inter speaker comparisons were
14 performed to determine whether or not bilingual speech had undergone change, and the putative source of the change. Also, an intra speaker comparison was performed to test Learning Model, in which he claims that early bilinguals commonly have two separate phonetic systems. Through this comparison, a Variationist Speech Learning Model (VSLM) was proposed. The inter speaker comparisons show ed that when bilinguals spoke Span ish, both internal and contact induced changes were evidenced. When bilinguals spoke Creole English, only internal change was noticed. The intra speaker comparison revealed that the bilingual West Indians had two separate phonetic systems; however, the var iationist methodology made visible signs of convergence. Interestingly, this convergence showed that bilinguals moved towards monolingual Creole English norms. A qualitative analysis of their sociolinguistic interviews revealed that factors such as languag e attitudes, language loyalty and maintenance, and social identity play an important role in this language contact situation.
15 CHAPTER 1 INTRODUCTION The present analysis is concerned with language contact and change in the speech of Spanish Creole E nglish bilinguals of West Indian origin in Panama City, Panama. We analyze the voiceless dental consonant stop /t/ in terms of a linguistic property defined as the time interval between the release of a stop and the onset of vocal cord vibration. This time interval differs between Spanish and English, thus it has been chosen because it provides an area at which the two languages do not match. Analyzing this area of the language pair allows us to find potential effects of language contact in this region. T his chapter gives an introduction to the language contact situation of Panama (Section 1.1). Later, we focus our attention on pidgins and creole languages (Section 1.2), as this analysis involves a dialect of Creole English spoken in Panama. In Section 1.3 we place Creole English in the context of language contact and bilingualism among West Indians in Panama. We also explore a specific variety of Creole English spoken in Bastimentos because it serves as a baseline with which we can compare the dialect spo ken in Panama City (Section 1.4). Lastly, in Section 1.5, a general description of the analysis, along with its goals, are presented. 1.1 Language Contact in Panama Immigrants of West Indian origin arrived in Panama to work on the three most important pro jects in the history of this country: the construction of the Panama Railroad and the first and second phases of the Panama Canal (Connif, 1985:3). Because of the location of Panama, it has been the most preferred area for travel between the Atlantic and P acific oceans. Transit between the oceans by way of Panama has been going on
16 since colonization by the Spanish (Diez Castillo, 1981). According to Diez Castillo (1981:19), black slaves were used to transport treasures and construction materials on the Cami no Real which connected Nombre de Dios on the Atlantic to Panama City on the Pacific. When the California Gold Rush began in the mid 1800s, many people started travelling from the east coast of the United States to the west. The quickest way to make this journey was through Panama; however, the only obstacle was the tropical forests. This trip, according to Lamaitre (1972:69), was extremely dangerous, and many travelers contracted yellow fever before arriving to their destination. The most logical action t o take was to construct a railroad, which began in 1850. The construction of the Panama Railroad attracted many immigrants from the English speaking islands of the Caribbean. In 1838 slavery had been abolished in the British West Indies, and many freedmen who were unable to find jobs in their homeland decided to seek work abroad, particularly in Panama. About 5,000 workers migrated to the Isthmus to labor in the railroad construction (Connif, 1985:17). In the late 1800s, the construction of the canal was i nitiated by the French under the command of Ferdinand de Lesseps. This project triggered another wave of West Indian immigration, but this time from French speaking islands such as Martinique and Guadeloupe (Diez Castillo, 1981; de Banville, 2005). About 5 0,000 workers went to Panama, some of whom returned to their islands after the construction of the canal failed; however, many decided to stay and raise families while they waited for another project (Connif, 1985:18).
17 After Panama won its independence fr om Colombia on November 3, 1903 with the help of the United States, the second phase of the Panama Canal began. This project triggered another wave of immigrants from the West Indies. It was estimated that about 150,000 persons arrived in Panama between 19 04 and 1914. Many remained after the canal was completed because of the many opportunities available in this country. According to Connif (1985:4), the West Indians, or antillanos became the largest group of immigrants in a country that had a total popula tion of 400,000 people. As the presence of West Indians in Panama became more pronounced, racism began to grow intensely. This was especially the case by the late 1920s when Panamanians of mestizo origin started to demand the deportation of West Indian imm igrants because of competition for Canal jobs (Connif, 1985:4). Race aversion came to a peak in 1941 with the promulgation of a new constitution that disenfranchised many West Indians (Connif, 1985:4). The Constitution of 1941 denied black West Indians cit izenship despite having been born in Panama (Constitution 1941, Title II). Bad race relations were also noticeable in the Canal Zone, which was governed by the United States. Americans implemented Jim Crow segregation, which was disguised under a gold sil ver system. The payroll was divided into two categories, gold roll and silver roll. Skilled workers with American citizenship were paid according to the gold roll standard, and unskilled workers, normally of West Indian descent, were subjected to the silve r roll standard (Connif, 1985:4 5). One of the participants of the present study spoke about this form of segregation: Y en ese tiempo, mucho antes de mi tiempo, yo no viv ese tiempo. Haba lo que llamaba dizque gold roll, silver roll porque haban fuente de agua
18 que pintaban de un color a otro, nada ms los blancos tomaban en ese fuente y t tenas otro los dems tienen que tomar. Hasta el servicio, t no podas usar el servicio de ellos. Todo eso ellos tenan, como lo que pas en Estados Unidos (Samuel, Antill ano S p a n i s h 6/5/2010) And in those times, way before my tim There was what they called dizque=quotative gold roll, silver roll because there were water fountains that they painted from one color to another, only t he white people drank from that fountain, and you had the other the rest all of that, like what happened in the United States (Samuel, Translation, 6/5/2010) Due to this extreme racism, West Indians began to unite despite their island differences. Some migrated to the U.S., which was enabled by the treaties of 1955 and 1977. However, those who stayed established British West Indian schools, churches, and businesses. This caused t he growth of a new subculture that incorporated traditions borrowed from England, the Caribbean, North America, and Panama (Connif, 1985:12). These establishments were usually run in English, which helped in the maintenance of the language in the following generations. Some sent their children to Panamanian schools to learn Spanish; however, this was not an option for most since access to education was limited (Connif, 1985:6). West Indians settled mainly in major areas along the Panama Canal. In the Atlan tic coast, many lived in the city of Coln, and on the Pacific, they settled in areas of Panama City such as El Chorrillo, San Miguel, Calidonia, and El Maraon also known as G uachapal (Thomas Brereton, 1992 :49 ). However, as land became available in subu rban areas, West Indians began moving to areas such as Rio Abajo, Sabanas Parque Lefevre, and Juan Diaz (Connif, 1985: 85) These areas are still populated by West Indians to this date.
19 West Indian Panamanians today are much more integrated into Panamania n society. Many have occupied major positions in government, and others have become grade school teachers and university professors. Organizations, such as t he Society of Friends of the West Indian Museum of Panama work arduously to spread acceptance and knowledge of the West Indian community. The West Indian Museum is located in the area of El Mara on and it was the source of some of the interviews collected for this project. Since 1941, Spanish has not only been the national language, but also the offi cial language of Panama ( Constitution of 1941, Title 1, Article 10 ) 1 Most of the population speaks the language as their mother tongue, and according to Alvarado de Ricord (1982:98), it has become part of the national identity of Panama in face of America n imperialism. However, because of the strong presence of the United States, the English language is considered to be a means of survival (Alvarado de Ricord, 1982:98). The West Indian speech community is the only group in the capital of Panama and the cit y of Coln besides American expatriates that speak English as a first language. However, the variety that West Indians refer to as English in this community is an English based creole that originates from their Jamaican and Barbadian ancestors. In order to have a better idea of this creole variety, we discuss some relevant themes related to pidgin and creole languages in Section 1.2. 1 A copy of Constitution 1941 can be found at Biblioteca Digital Panamea: http://www.binal.ac.pa/buscar/clconst.htm
20 1.2 Pidgins and Creoles began to attract att ention in the 1960s. Some definitions are negative, as in the following for creole languages: Creole languages result from the adaption of a language, especially some Indo European language, to the (so to speak) phonetic and grammatical genius of a race t hat is linguistically inferior. The resulting language is composite, truly mixed in its vocabulary, but its grammar is essentially Indo European, extremely simplified (Vinson, 1889:345 345, cited in DeGraff, 2003:393) More recently, however, definitions have become more neutral or positive, as in the following definition for pidgin languages: Pidgins are languages lexically derived from other languages, but which are structurally simplified, especially in their morphology. They come into being where peopl e need to communicate but do not have a language in common. Pidgins have not (or few) first language speakers, they are the subject of language learning, they have structural norms, they are used by two or more groups, and they are usually unintelligible f or speakers of the language from which the lexicon derives (Bakker, 1994:25) We define a pidgin in this project as a language that emerges in a contact situation where there is a necessity to communicate among various groups of people who do not have a c ommon language. Because human beings share the same cognitive abilities, pidgins are structured in a way that the speakers who create them are able to understand each other even when innovative constructions arise. A creole then emerges out of the pidgin l anguage during acquisition in later generations. In this process, frequent constructions become conventionalized, while new ones continue to be created and infrequent ones disappear.
21 Creoles differ from pidgins in that the former have native speakers and the latter do not (Muysken & Smith 1994). That is, the creole language is a nativized version of the pidgin. According to some theories, nativization of pidgins is instantaneous (Bickerton, 1988), and for others, it is a gradual process (McWhorter, 1992). Creoles have a donor language that is usually referred to as a lexifier or superstrate. This is the language from which the creole incorporates a great portion of the vocabulary (although vocabulary is also taken from the substrates), but maintains more or less the syntactic structure of the substrate(s). Various creoles are spoken in the Americas, mainly around the Caribbean basin. They make up part of the Atlantic creole group, which includes creoles spoken in West Africa. Caribbean creoles can be based on English, Spanish, Dutch, or French (Holm, 2000). However, they are structurally similar because their substrata are part of the Niger Congo language family. Table 1 1 is an outline of the creoles spoken in the Caribbean, based on Holm (1988:xx xxi). Mos t of the Caribbean creoles (with the exception of Costa Rican Creole English and some Panamanian varieties that are offshoots of those spoken in Jamaica and Barbados) arose in the context of plantations where African slaves purchased along the western coas t from Senegal to Angola were taken to work in large agriculture units of sugar cane, coffee and tobacco (Arends, 1994:15). These creoles are known as plantation creoles. Another category in which these creoles can fall is marronage creoles. Marronage refe rs to the type of creole spoken by slaves who escaped from their plantations and formed communities in isolated areas, for example, Palenquero in Colombia (Arends, 1994:16 17).
22 Table 1 1. Caribbean creoles Caribbean creoles English based Bahamian C E Leeward Islands CE Barbadian CE Miskito Coast CE Bay Islands E Ndyuka CE Belizean CE Panamanian CE Caymanian E Providencia and San Andrs CE Commonwealth Windward Islands CE Saman E Costa Rican CE Saramaccan CE Dutch Windward Islands CE Sranan CE Guyanese CE Trinidadian and Tobago CE Jamaican CE Virgin Islands CE Spanish based Dutch based Palenquero CS Berbice, Skepi CD Papiamento CS Negerhollands CD French based Grenada CF Guyanais CF Haitian CF Lesser Antillean CF Trinidadi an CF According to Arends (1994:16), it seems that there is some divergence in these varieties due to their isolation; however, since they originate from plantation creoles the structural differences are limited. There is much debate on the origin of c reole languages based on their similarities with their superstrates and substrates. In Section 1.2.1, we discuss some of the theories that have been developed to explain the formation of these languages. 1.2.1 Theories of Creole Genesis There are various t heories of creole genesis in creolistics literature. Some focus on the influences of the superstrate and others, on the influences of the substrate. 2 2 There are also theories of creole genesis that explain creolization as nativization of the pidgin. within one or two generations. The Gradualist Model and the Gradual Creolization Hypothesis propose
23 Those that focus on the European input, that is, the lexifier, are the monogenesis view, the European dial ect origin hypotheses, theories concerning the influence of the Atlantic slave trade, the mixed European source creole hypothesis, the foreign talk and baby talk view and the imperfect L2 learning hypothesis (den Besten, Muysken, & Smith, 1994). From the m onogenesis perspective, creoles around the world developed from one specific pidgin called West African Pidgin Portuguese (WAPP) spoken in the forts along the West African coast (Taylor, 1961; Thompson, 1961). Therefore, French based and English based creo les would have had to develop through a process called relexification in which there was a word for word replacement of the Portuguese vocabulary for the French or English vocabulary (den Besten, Muysken, & Smith, 1994:88 89). In another view, WAPP was der ived through relexification from Lingua Franca spoken in the Mediterranean. An additional monogenesis explanation is more restricted in that there was a West African Pidgin English and a West African Pidgin French from which English based and French based creoles derived (den Besten, Muysken, & Smith, 1994:88 89). The general European dialect (partial) origin hypothesis explains that creole languages derived from non standard varieties of the lexifiers as was spoken in the countries of origin of the Europea ns ( Chaudenso n 1979, 1992; den Besten, Muysken, & Smith, 1994:89; Muwfene, 1996:124). This hypothesis implies that the substrate languages had little to do with their development. Particularly for African American Vernacular English, there is much debate a round this idea. Poplack (2000) and that creolization is a gradual process that extends over a number of generations (Carden & Stewart, 1988; McWhoter, 1992; Arends, 1993).
24 Tagliamonte & Poplack (2001) used a comparative method to prove claims that the variety spoken in North America originated from British origin dialects, as opposed to ) review and arguments against Poplack (2000)). The theories concerning the influence of the Atlantic slave trade take into account influences from English in Africa, nautical language, and English in the Americas (Reinecke, 1937; Hall, 1966; Hancock, 198 6). The mixed European source creoles view hypothesizes that some creoles derived from more than one lexifier language, for example, Saramaccan and Papiamento (den Besten, Muysken, & Smith, 1994:94 95). The foreign talk and baby talk view suggests that cre oles derived out of accommodation to non native competence, which is characterized by slower speech, shorter and less complex sentences, the introduction of pauses, use of semantically unspecific terms, and repetitions. This view also includes behavior suc h as imitation of the speech of the non native speaker and telegraphic condensation (Hesseling, 1897, 1933; Bloomfield, 1933; den Besten, Muysken, & Smith, 1994:95 96). The imperfect L2 learning hypothesis claims that creoles were the result of speakers no t achieving dominance in the lexifier language (Coelho, 1880, cited in den Besten, Muysken, & Smith, 1994); therefore, the interlanguages became nativized. Most importantly, these superstratist views characterize creoles as simply being varieties that have evolved from their Chaudenson 1979, 1992, Muwfene, 1996:124). 3 3 These theories oppose substratist views that creoles are typologically distinct languages that can be characterized by the presence of three traits: 1) little or no inflectional affrication, 2) little or no use of tone to lexically contrast monosyllables or encode syntax and 3) semantically regular derivational affixation (McWhorter, 1998:798). The traits are in the language due to influence from the substrates.
25 Other theories that attempt to explain creole origin are those focusing on the non European input, that is, the subs trate languages. Depending on the sociocultural context of the region, substrate influence can be complex since it could involve several languages. This type of interference, which involves a higher, more prestigious language (the superstrate), and a lower less prestigious language (the substrate), can occur in situations of slave trade and colonialism, such as creoles spoken in the Americas (Arends, Kouwenberg, & Smith, 1994). In this view, the substrates had an important role in the formation of creoles. Substratum influence has been noticed in all aspects of language, including phonology, morphology, syntax, lexicon, and semantics. The sources particularly for Atlantic creoles were West African languages related to the Kwa branch of the Niger Congo fami ly (Alleyne, 1980; Boretzky, 1983; Holm, 1988; Mufwene, 1993). In creolistics, the proponents of substratum influence have found that because of this common origin, many creoles spoken today in the Caribbean have several commonalities regardless of their l exifiers (Holm, 2001:62 64). The varieties of Creole English spoken in Panama make up part of these Atlantic creoles because they originate from speakers who emigrated from the Caribbean; therefore, it is possible to trace their substratum influence to the same West African languages. One of the reasons why Creole English varieties in Panama are of interest in this study is because these dialects are in contact with a superstrate that is not a lexifier language. The question we could ask in this situation i s how these varieties would compare to standard varieties of English when taking into account the sociohistorical context of the region.
26 This question is especially important when we consider what has been termed by ch is discussed in Section 1.2.2. 1.2.2 The Creole Continuum I t is well argued that in English based creoles a creole continuum is apparent in se registers are normally referred to as acro lect (the most standard), mesolect (semi creole), and basilect (the most creole) (Bickerton, 1973 ; de Rooij, 1994 ). There is some opposition to the creole continuum where different registers of a creole in a single community are interpreted as a discrete d ifference; there is a separation between the acrolect and the mesolect/basilect (Rickford, 1987:16 20). The discrete diglossic model is supported in the literature based on various grounds: 1) linguists who are native members of a creole speaking community tend to perceive variation in discrete terms, 2) linguistic knowledge might not be spread evenly across the continuum, thus some communities may have speakers of only the basilect, or at least who have more control over it, 3) some have discovered signifi cant co occurring restrictions between features; as the frequency of a grammatical feature increases, there is a sharp decrease in another feature, and 4) there is a fundamental difference between the acrolect and the mesolect/basilect, and minimal differe nces between the mesolect and the basilect. It appears, however, that the specific context in which the creole is spoken determines the type of model that should be applied. When looking at cases of Haitian Creole or Martinique Creole, a discrete diglossic model is appropriate, but in the case of Jamaican or Guyanese Creole, the continuum model better represents the gradience (Rickford, 1987:19 22). Some research concerning the creole continuum suggests that basilect varieties that are in contact with the ir lexifier experience a unidirectional process called
27 progressively change the basilect grammar so that its output comes to resemble the output of an acrolectal grammar variety, which has been considered too strong of an argument based on evidence in different creole speaking communities (see Rooij, 1994 for arguments against the link between the creole continuum case in this respect because in this country the English based creole spoken is not necessarily in contact with its superstrate language, but rather with a non lexifier, Spanish. Thus, any acrolecta l or mesolectal characteristics that arise, particularly in the variety spoken in Panama City, should not be due to contact with a lexifier. In order to pinpoint these characteristics, it is important that we understand what features can be highlighted as being part of a creole. These features are presented in Section 1.2.3, paying special attention to English based creoles. 1.2.3 Features of Creoles Several features found in creoles have been pointed out in the literature (e.g. Alleyne, 1980; Akers, 1981 ; Holms, 1988; Thomas Brereton, 1992), and are related to different areas of the language such as the lexicon, syntax, and phonology. According to the view that one takes, one could see that the lexifier languages had a strong influence on creole languages in terms of vocabulary; however, the African substrates also contributed a portion of their lexicon. As we mentioned in Section 1.2.1, many of the substrates form part of the Niger Congo family. Terms such as you nyam d in Jamaican Creole English (Alleyne, 1980:110; Thomas Brereton, 1992:22). Also originating from this language family are terms such as big e y e
28 eye water literal translations of aa uku and aa mmiri (Alleyne, 1980:110). We see that in terms of syntax, creoles can show several differences from their lexifier and their substrates. One area of syntax where differences are apparent is tense and aspect. Creoles tend to mark tense and aspect with particles that precede the verb, such as di in (1), which indicates the past of a stative verb (Holms, 1988:156). Usually there is zero marking with action verbs to denote past (Holms, 1988:150) as in (2): 1) im di fiil baad 2) im go to Panama The markers doz and don are used to denote habitual and completed actions, respectively (Holmes, 1988:158 163), as in (3) and (4): 3) im doz caal alat 4) im don riich alredi Another are a of differences between creoles and their lexifiers and substrates is in phonology. Common processes that have been seen particularly in English based creoles are unstressed syllable deletion (e.g., memba deletion (e.g., gi: ) h deletion ( im av it before stressed vowels (e.g., heggs mentioned can be found in Panamanian varieties of English Creole since they are dialects spoken by those who se ancestors migrated from Caribbean nations, such as
29 Jamaica and Barbados. However, certain factors may have brought about differences between Panamanian varieties and other Caribbean creoles, one of which could be bilingualism. 1.3 Bilingualism and Lang uage Attitudes among West Indians in Panama Creole English is present in various areas of Panama. The main varieties are Panama City Creole English (PCCE), which is spoken in the capital. Also, there is Coln City Creole English (CCCE) spoken in the provin ce of Coln on the Atlantic end of the Panama Canal. Other varieties can be found in Bocas del Toro. Isla Coln Creole English (ICCE) is spoken on the main island of Coln, and Bastimentos Creole English (BCE) is spoken on Bastimentos Island. Many West Ind ians in Panama speak both Spanish and Creole English; however, in the metropolitan areas, that is Panama City and Coln city, bilingualism is more common in the older generation. All the speakers included in this study claimed to be balanced bilinguals who had learned Creole English at home and Spanish in their neighborhoods and in school at the age of 5. There are also some West Indians from the younger generation that speak Spanish and Creole English, which was the case for one of the interviewees. Becaus e of his young age, he was not included in the present analysis. 4 Attitudes towards Spanish and English are very positive in this community. Many West Indians consider the knowledge of these languages essential to be academically and professionally success ful. 4 An attempt was made to not include younger speakers in the analysis because it has been observed in liter ature concerning Language Variation and Change that adolescents tend to have more nonstandard linguistic features in their repertoire than standard ones. They use these features as a way to go against societal norms (Holmes, 1992). Once they have reached m iddle age, this behavior changes in that they grading (Labov, 1994). In order to have a more homogenous gr oup, only middle age speakers were included in this study.
30 Interestingly, the West Indian speech community does not explicitly mark a difference between English and Creole English when referring to the languages (Thomas Brereton, 1992). In actuality, West I ndians could take offense if their English is called household where Standard English was spoken. They are even proud of having taken English courses in school. It is more than likely that many are not even familiar with the ter 5 came up in several instances during a couple of the interviews conducted for this study. One speaker from Panama City made fun of the Creole English varieties spoken on Bastimentos Island, and even in the cit y of Coln, which happen to have more basilect characteristics. Also in the area of Bocas del Toro, speakers make a distinction between the variety spoken on Isla Coln 6 and the Bastimentos variety. A lady who helped collect interviews in this area explain ed that Bastimentos. Therefore, according to the comments that were encountered during the fieldwork for the present analysis, we could assume that not only are speakers aware of English, but that this dialect does not have much prestige in this region. Also, we can be sure that these varieties spoken in Panama are indeed English based creoles, as they ex hibit various features discussed in the literature. 5 Normally, this term is designated for the Bastimentos variety; however, it is also used in a pejor ative way to refer to all varieties of Creole English spoken in Panama. 6 This is the main island of Bocas del Toro, which is different form the city of Coln located on the Atlantic side of the Panama Canal.
31 Thomas Brereton (1992) highlighted several creole features that appeared in her study on PCCE, focusing heavily on syntactic features. One of the features that she presented was zero marking for past tens e in action verbs in (5) (Thomas Brereton, 1992:75): 5) Yes bekaz wen wii big mai mada sho os and tel os. She also explained that the preverbal marker di was used in this variety to mark anter ior/non anterior tense, such as in (6) (Thomas Brereton, 1992:81): 6) Wondering if a friend was awake at the time of the invasion: Ai tink shi woz op beka piipul kaal tu. Piipul staat kaalin evribadi. An piipul fram di Steits di noo about it aredi so piipul woz kaalin dem piipul dong ier. I think that she was awake, because people called too. People started calling everybody. And people from the United States knew about it already so people were calling their relatives here in Panama. Interestingly, only one speaker used the preverbal marker don to denote completedness (7); the most common form was actually the Spanish ya or preclausal, or postverbal in an ya (8) & (9) (Thomas Brereton, 1992:93 94): 7) Ai don wash dis maanin 8) Bikaa iiz aw ya is about a ier nou 9) If dem fiil laik, dem mek caafi and serv it an ya Addit ionally, doz and yuustu were used to mark habitual actions in the present and the past, respectively, (10) & (11) (Thomas Brereton, 1992:99 101):
32 10) Wat ai doz luk an is di uan dem giv sem aklak bout Paraso 11) Mai aant yuustu ponish mi My aunt used to punish me. Although these forms appear in other Caribbean creoles, they seem to be following slightly different patterns, according to Thomas Brereton (1992), either because of internal chang e or because of contact with Spanish; however, this was not quantitatively tested by the researcher. To test this claim, it is crucial that we turn to the Creole English variety spoken in Bastimentos, Bocas del Toro, as this is an area in Panama where Creo le English perhaps has less contact with Spanish than PCCE. The lack of evidence of Spanish influence in BCE could shed light on changes occurring in PCCE. Therefore, in Section 1.4, we discuss the situation of Creole English in Bastimentos and the charact eristics of this variety. 1.4 Creole English speaking West Indians in Bastimentos As we mentioned in Section 1.3, there is also an enclave of West Indians located in the province of Bocas del Toro. This group is made up of descendants of people who emigrat ed from various English speaking areas of the Caribbean and Central America. The first recorded influx was during the 1800s with the arrival of slave owners from the islands of San Andrs and Providence, Jamaica and the town of Bluefields, Nicaragua. This migration began because of the growing coconut and turtle business (Guerrn Montero, 2002:146). There were also fishermen from Corn Island, Nicaragua and San Andrs who settled specifically on the island of Bastimentos. According to Guerrn Montero (2002:1 46 147), after the abolition of slavery in Panama in 1852, slaves who
33 also had arrived with their slave owners became peasants and began to integrate into Panamanian society. The second influx of West Indians to Bocas del Toro, and later to Bastimentos, w as during the late 19th and early 20th centuries. Many were associated with the Methodist and Seventh Day Adventist churches, and others had migrated to Panama for the construction of the Panama Railroad and Canal (see discussion in Section 1.1). When thes e projects came to an end, many went searching for work on the banana plantations of Bocas del Toro (Guerrn Montero, 2002). A dialect of Creole English is spoken today in this region because of this migration. Although in this area there is Spanish in fo rmal settings, such as school (Snow, 2000; 2007), Creole English is the first and dominant language of Bastimentos, where it is spoken by both the younger and older generations. This dialect has evolved from contact among different dialects of English base d creoles and can be now considered Bastimentos Creole English (BCE), differing from the dialects of Panamanian Creole English spoken in the capital of Panama and the city of Coln (Aceto, 1998). BCE has attracted some attention in creolistics (e.g. Aceto 1995, 1998, 2001, 2002; Snow, 2000, 2003, 2007) in which the focus of analyses range from variation and change to sociocultural factors that play a role in the development and use of the variety in Bastimentos. Some general characteristics of BCE, which are also present in PCCE are absence of inflectional morphology (e.g., mi fada hous de as a copular form (e.g., shi de gud if yu stil iz di baas him iz mi fr en marking in past tense (e.g., mi trai it 7). Additionally, relativization is
34 marked by wat The mar king of future tense also has been described in detail by Aceto (1998). Aceto (1998) took a synchronic view of the use of forms that mark future in order to give a diachronic explanation of a new emerging form in BCE. The following is a list of these forms (listed synchronically and diachronically), which derived from the lexeme goin (Aceto, 1998:34): 12) goin 13) gowain 14) gwain 15) gwan 16) wan 17) wain 18) an 19) goinan 20) gwainan 21) gwanan 22) goan Examples 12 22 are forms specifically found in BCE that originated from goin + an Aceto (1 998) claimed that gwainan was a new form that had not been attested in any other varieties of Creole English. He explored the possibility of superstrate influences from British varieties spoken in areas such as Warwickshire, southeast Worcestershire, and O xfordshire. However, the form used in these areas were not employed as a future marker but rather a verb of movement. Due to the fact that this form was not found in other English based creole varieties, Aceto (1998) eliminated the argument of superstrate influence and attributed the presence of gwainan to an internally motivated change brought about by an analogical process.
35 First the form an appeared to arise from a course of change in the order of the list above, that is goin > gowain > gwain > wain > w an > an (Aceto, 1998:37). An was then influenced by certain structures in BCE, such as a gwain iit a gwain go iit a gwain go an iit hich an appeared to be working as a conjunction (Aceto, 1998:38). Thus, Aceto (1998:38) proposed a four part analogy that brought about the construction a gwain an iit : 23) 23) a gwain go iit : a gwain iit :: a gwain go an iit : x, x = a gwain an iit Later, a n was reattached to gwain also through a four part analogy (Aceto, 1998:40). 24) 24) go : goin :: gwain : x, x = gwainin Aceto (1998) suggested that goin and gwain speakers interpreted in as an ending that denoted pr ogressive, which could be added to any verb. This ending was realized as an and was attached to the form gwain that resulted in gwainan in order to mark future (Aceto, 1998:41). In another study, Aceto (1995) explored variation in a cryptolect spoken on t he island of Bastimentos called Gypsy In this secret language, the speakers created new syllables built around the consonant /g/, for example (Aceto, 1995:538): we im de weger higim igiz. According to Aceto (1995), if the phonological rule s of Gypsy were reversed, we would be left with wer him iz which reveals characteristics of a more mesolect variety of Creole, such as final r in wer h in him and the verbal form iz Aceto (1995) argued that these mesolectal forms did not appear because of decreolization, but rather because of the coexistence of several varieties of English based creoles since the 19 th and 20 th century in this area (see discussion in this section about history of Bocas del Toro and Bastimentos). The proposal of mesolecta l creole
36 influence was made because BCE is not in contact with metropolitan varieties of English, thus, there is no opportunity for decreolization (Aceto, 1995). Related to this proposal, Snow (2000) explained that ideas of the creole continuum are applie d to an extent to which decreolization is used as an explanation for language variation and change in situations where creole languages are not in contact with their lexifier (e.g. Anderson 1990; Garrett, 1990). He proposed that a discrete diglossic mode l be used to explain these situations, which characterizes the case of Creole English and Spanish in Bastimentos today. This model could also be applied to the case of PCCE and CCCE as they are also in contact mainly with Spanish. The diglossic model seems to be appropriate when considering the situation of Creole English versus Spanish in Panama; however, as we will see, a creole continuum could be apparent among the Creole English varieties. Panamanian varieties of Creole English differ from other Engli sh based creoles in various ways. One of the most obvious aspects is vocabulary. They share English words, but because of contact with Spanish, Creole English speakers in Panama tend to borrow Spanish words. This is especially common in PCCE and CCCE; howe ver, when speaking Creole English, the Spanish word normally is not phonologically integrated into Creole English. That is, the speakers living in the more metropolitan areas of Panama do not pronounce the Spanish word with Creole English phonology. Conver sely, in BCE and ICCE, Creole English phonology is indeed applied to the Spanish word. In an interview with one of the participants who spoke about the different churches on Bastimentos Island, she borrowed the Spanish word culto
37 however it was pronounced as though it were completely integrated into Creole English, [k h u t h The Panamanian varieties also differ slightly in syntax. It appears that there are more occurrences of standard subject pronouns in Panamanian dialects when compare d to other varieties, such as Jamaican Creole English. This would be more so in PCCE and CCCE than in ICCE and BCE. For example, the opposition of mi go vs. a go vs. I go is present in all of the Panamanian varieties; however, the frequency of the basilect al form mi is probably higher in the varieties of Bocas del Toro. Therefore, we see that Panamanian Creole English varieties differ from other Caribbean creoles, but more interestingly, there appears to be a continuum among the Panamanian creoles. The diff erences mentioned above and the putative continuum among the Panamanian varieties could be due to dialectal leveling that occurred during the influx of West Indians to Panama from various islands, which in turn could have caused internal changes in the lan guage. Another explanation for this situation could be related to language contact with Spanish, which could have brought about changes in the speech of West Indians living in Panama. The present analysis, thus, investigates possible changes in the West In dian speech community to find evidence of the potential effects of contact in both Spanish and Creole English. A general description of the study is provided in Section 1.5. 1.5 The Language Contact Study As in most areas of the world, language contact and bilingualism is common in Panama. Therefore, this region provides an ideal situation for analyzing effects of contact on linguistic systems. Various researchers have become interested in language contact in Panama, focusing mainly on variation and chan ge in Creole English and its
38 diglossic relationship with Spanish, particularly in Bastimentos (e.g. Aceto, 1995; 1998; Herzfeld, 1983; Snow, 2000) However, there is a lack of studies that focus on possible bidirectional effects in this contact situation. A study of this type would entail analyzing both Creole English and Spanish of this speech community to discover change and determine its possible source. Therefore, one of the goals of this study is to take into account the whole linguistic repertoire whe n studying contact induced change in bilingual West Indian speech. We also explore a specific hypothesis about bilingual phonology found in the Speech Learning Model developed by (Flege, 1987; 1995a). These two main goals are achieved by incorporating vari ationist methodology that utilizes naturally occurring speech in the community. In order to analyze change in a contact situation, it is crucial that we identify a conflict site. This is an area where the structures of a language pair do not coincide (Po plack & Meechan, 1998:132), that is, they conflict. Therefore, in this study, we focus on a linguistic property that appears to differ between Spanish and Creole English, voice onset time (VOT), specifically, of the voiceless dental plosive 7 VOT is an aco ustic property defined as the time interval between the release, or occlusion, of a consonant stop and the onset of vocal cord vibration (Lisker & Abramson, 1964 :387). This property differs across languages in terms of duration. Particularly in Spanish, VO T duration is no longer than 25 milliseconds (msec.), and in English, no less than 30 msec. (Lisker & Abramson, 1964). Thus, VOT duration provides an ideal site for analyzing potential effects of language contact in the bilingual West Indian speech communi ty in Panama. 7 Spanish and English differ in place of arti culation with regard to this consonant stop. In Spanish, the plosive is dental, and in English, it is alveolar. Place of articulation will not be considered as a factor in the present analysis as no study has reported on its possible effects.
39 This is the first study to use variationist methods to analyze VOT duration in monolinguals and bilinguals.
40 CHAPTER 2 LITERATURE REVIEW In this chapter, we review the literature concerned with various topics of interest to the present study. As language contact is the main subject, Sections 2.1 and 2.2 predictors and mechanisms that could bring about contact induced change, and the linguistic outcomes of contact. We mention in Chapter 1, that contact induced change is most adequately investigated in an area where a language pair conflicts, which is why we analyze VOT duration. This linguistic property is discussed in further detail in Section 2.3, focusing mainly on varieties of Spanish and English. Lastly, since language involves many cognitive processes, we must be able to explain these processes and how language could be represented in the minds of bilinguals. Therefore, in Section 2.4, we focus on how u sage based models of language can explain the phonology of bilinguals generally, and in terms of VOT duration. 2.1 Language Contact and C ontact i nduced C hange Language contact is a situation in which two or more languages coexist in the same place at the same time (Weinreich, 1966; Thomason, 2001; Matras, 2009). In this type of situation, bilingual or multilingual s peakers must be present for communication to occur, although these speakers need not be completely fluent in the languages involved (Thomason, 2001:1 2). Language contact and multilingualism are present in many societies worldwide. According to Matras (2009:48), multilingualism can occur in neighboring tribal communities in non urban societies, such as in the Amazon or in the Arnhem Land in north ern Australia. Multilingual speakers can be found also in border areas of industrialized societies, such as on the border between Germany and
41 Denmark. In this area, Frisian is spoken as a home language and Low German is the vernacular that is spoken among members of different rural communities Another example of a multilingual border area is the Brazilian Uruguayan border, where Portuol, or Fronterizo, is used in Rivera, Uruguay and Santana do Livramento, Brazil (Lipski, 2009). Multilingualism is also fo und in societies established through colonialism such as in Africa. Although language contact was already present before the arrival of Europeans to this continent, contact was intensified with the presence of European This not only brought about multilingualism, but also the creation of pidgins and creoles that are still spoken today (Thomason & Kaufmann, 1988). Another area where colonial multilingualism is evidenced is in Latin America and the Caribbean. Haiti has French as an off icial language but a French based creole is also spoken there as the everyday native language (Matras, 2009:49). Massive immigration to urban centers can also bring about multilingualism. Many researchers (e.g. Roca & Lipski, 1993; Silva Corvalan, 1994; Torres Cacoullos & Aaron, 2003; Otheguy, Zentella, & Livert, 2007 ) have analyzed language contact in the United States in areas such as Los Angeles, N ew York, New Mexico, and Miami Outside of the Unites States, the focus has been in areas such as Canada ( e.g. Poplack, 1993) and Panama ( e.g. Thomas Brereton, 1992) where migration driven multilingualism is also very common. B ilinguals not only come into contact with one another in the situations mentioned above but also with monolingual speakers, which pre ssures them to constantly work towards maintaining effective communication (Matras, 2009:39). According to Matras
42 (2009:39), there are two goals that bilinguals try to comply with the expectations of interlocutors in selecting struc tur and 2) they have a need to exhaust the full resources of their linguistic repertoire in order to Since both goals cannot be achieved at the same time, in certain situations, one may overr ide the other during communication. This is why bilinguals may mix their languages as a need to express themselves, but still be control over selection s which can be a ffected by cognitive, social, and linguistic factors (Matras, 2009:38 39). Another important topic in language contact is contact induced change, that is, a linguistic change that would have been less likely to occur outside of a contact situation (Thomas on, 2001:62). There can be direct changes, which originate from the source language, or indirect changes, which can consist of attrition or changes triggered by an early direct importation ( Thomason, 2001:62). With respect to phonology, contact induced cha separate their two phonological systems (Matras, 2009:222) or the borrowing and imposition between the systems (van Coetsem, 1988:3) In Sections 2.1.1 & 2.1.2 we highlight vario us so cial factors and linguistic predictors that can lead to contact induced change. 2.1.1 Social Factors Thomason (2001) discusses two main social factors that contribute to contact induced change. The first is intensity of contact. As Thomason (2001 ) points o ut, it is difficult to quantify this factor; however, there are three ways that can be useful in defining the intensity of a contact situation. First, the duration of the contact period could
43 bring about certain changes and a certain amount of change. If t wo groups of speakers have been in the same region for a long period of time, this will lead to more opportunities for language change due to contact ( Thomason, 2001:66). Second, the group of speakers with the smallest population more than likely will adop t features from Third, i f this dominant group is also socioeconomically more successful than the smaller one, it is even more likely that the smaller group will adopt features from the larger one (Thomason, 2001:66), as is the case with /u/ fronting in Chicano English (Fought, 1999). Another social factor leading to contact induced change is imperfect learning, or in some cases, the absence of imperfect learning. Thomason & Kaufman n (1988) and Thomason (2001) refer to this as borrowing versus substratum interference. In the former, there is absence of imperfect learning, where speakers maintain their native language but incorporate features from another. The first features normally borrowed are lexical items (Thomason & Kaufmann, 1988:37), as in New Mexican Spanish (Torres Cacoullos & Aaron, 2003:290): 1) Y o arreaba gatos, arreaba loaders y trocas y todo eso I drove tractors, I drove loaders and trucks and all that Borrowing spre ads through the native language as contact becomes more intense. Besides words, the speakers begin to incorporate phonological, phonetic, and syntactic elements. Matras (2009) claims in his discussion on contact induced changes in phonology that this is w hen bilingualism emerges. S ubstratum interference is the result of imperfect learning (Thomason & Kaufmann, 1988; Thomason, 2001). This is when a group of speakers begin to shift to a
44 target language and spread learning er rors to the rest of the members o f the group 1 It is claimed that this interference is evidenced first in the sound patterns and syntax (Thomason & Kaufmann, 1988:39). With regard to interference in sound patters, an ethnolect can emerge in which speakers speak the target language with a foreign accent (Matras, 2009:223 ). Substratum interference can yield situations where there is total shift to the L2, and possible language death if the L1 is not maintained through generations (Thomason, 2001) A three generation shift pattern has been ob served in various coun tries with high rates of immigration (e.g., Silva Corval n, 1994), although not always resulting in language death. Another result of substratum interference is the emergence of pidgins and creoles. According to van Coetsem (1988), w ho equates substratum interference with source language agentivity, pidgins are settled or socialized forms of interlanguages that emerge when a group is acquiring the target language. In this situation, speakers no longer seek to acquire the L2, but rathe r the pidgin. This has been the case in most, if not all, areas where pidgins and creoles are spoken. If we consider the example of Haiti, we see that there was a superstrate, French, and various substrate languages spoken by African slaves. Surely, in the initial stages of contact, the target language for the African slaves would be French, and perhaps many were able to acquire the L2. However, at some point, when a pidgin began to emerge among slaves who originated from different African tribes, thus spea king different language, a French based pidgin became the target language for these speakers. Their interlanguage would become 1 See discus sion on substratum interference in Introduction.
45 conventionalized and be dispersed among the group and spread to the younger generations. 2 2.1.2 Linguistic Predictors and Me chanisms Apart from the social factors, it is also important to mention the linguistic predictors of contact induced change. Thomason (2001:77) presents certain linguistic predictors of change when discussing the results of contact, and lat er presents mec hanism s for the process of contact induced change. One predictor of change is markedness which is claimed to be important mainly in shift induced, or substratum interference (Thomason, 2001:76). Linguistic forms that are simpler and more widely distribute d within a language and across languages are considered u nmarked (Moravcsik & Wirth, 1986). These unmarked forms normally have various subtypes. The opposition to the unmarked form is the marked one. Its distribution is restricted to more specific contexts thereby being less frequent. It has been noticed that unmarked forms can appear in a language with the absence of the marked forms. However, marked forms do not appear without the unmarked forms. This phenomenon is known as typological markedness, and an example is noticed regarding plosives, where we see that voiceless plosives are more widely distributed across languages than voiced plosives due to their simple articulation (Moravcsik & Wirth, 1986:2). There are languages in the world that have voiceles s plosives without voiced ones, but never voiced plosives without voiceless ones. Markedness appears to have an important role in second language phonology. In the Markedness Differential Hypothesis, Eckman (1977) states that typological markedness is esse 2 More detailed discussions on the sociocultural context of the formation of Haitian Creole can be found in DeGraff (1992), DeGraff (2001), and Lefebvre (1998).
46 hypothesis is that, where two languages differ, more marked forms are more difficult to learn than the more unmarked forms. Thomason (2001) also makes this claim and attributes conta ct induced change to the difficult learnability of marked forms, which can reveal an apparent simplification in the language. 3 only inquires about the areas of differences in the two languages, He developed the Structural C onformity Hypothesis, in which it is stated that learners perform better on less marked forms. 4 Differences in the degree of integration can also affect the outcome of contact induced change (Thomason, 2001:76). It is less likely that borrowing will occu r with features that are deeply embedded into the structure of the language Thomason (2001:77) states that deeply embedded features, such as inflectional morphology are usually incorporated last and only in more intense contact situations However, in in tense contact situations of pidgins and creoles, there are varieties that do exhibit features such as derivational and inflectional morphology (Roberts & Bresnan, 2008). Thus, it appears that other factors may override the degree of integration. Another predictor for change due to contact is typological distance bet ween the two languages involved (Thomason, 2001). Features and structures are more readily exchanged if the languages are more typologically similar. Thus, marked forms and deeply embedded stru ctures can be incorporated from one language to another if these languages have similar systems (Thomason, 2001:77). If this is the case, borrowing 3 Though the loss of marked forms can br ing about simplification, this does not insinuate that the language is simplified as a whole. Where there is simplification in one area, there could be complexities in another. 4 See discussion on L2 phonology in Section 2.4.
47 would be more likely to occur when it involves two dialects of a language as opposed to when there are two c ompletely different languages. As far as mechanism s for contact induced change, Thomason (2001) discusses the presence of code switching in the speech community. This mechanism has been ach of which internally consistent with the morphological, syntactic (and optionally, phonological) ode switching can be intersentential switching between sentences or it can occur intrasentent ially sw itching within a single utterance (Poplack, 1980). Various studies (e.g., Poplack, 1980; Torres Cacoullos & Aaron, 2003) have focused on the latter because the structure of intrasentential code switching can resemble borrowing. In the definition provided in Poplack (1993), code switching involves the integration of elements into the recipient language that maintain the morphological, syntactic, and sometimes phonological rules of the source language. Borrowing involves the adaption of words to the morphological, syntactic, and usually phonological rules of the recipient language. In this process, loanwords are distinguished from nonce borrowings in that the former involves widespread diffusion and the latter does not (Poplack, 1993:255 256). Becaus e code switching and borrowing look similar on the surface, Poplack (1993) and Poplack & Meechan (1998) propose the use of the comparative method to empirically distinguish the two processes. One important issue to point out in code switching is the option ality of phonological maintenance of the source language element. If in code switching it is not required that the foreign element follow the rules of the source language, this would mean that
48 phonetic features from both languages can appear in or around t he switch point. In a study of VOT duration, Bullock Toribio, Gonz lez, & Dalola (2006) analyze the effects of code switching in the production of stops by Spanish English bilinguals They discover that both native English speakers and native Spanish spe akers exhibited effects of the L2 immediately before the switch and at the switch point. However, they were able to recover and produce target stops after the switch (Bullock et al., 2006). Sancier and Fowler (1997) label this behavior as gestural drift, w hich can be brought about by code switching because, according to Grosjean (2001), a speaker of two languages is normally in bilingual mode. Code alternation is another mechanism for change in which each language is used in specific domains ( Thomason (2 001:137) This type of situation is also commonly known as diglossia (see Fishman, 1967; Appel & Muysken, 1987) It is assumed here that there is communication between bilinguals and monolinguals. Thomason (2001) states that it is difficult to find contact induced change in code alternation because it is not certain that the change occurred as a result of this mechanism or because of code switching. Therefore, most exam ples are anecdotal in nature (Thomason 2001: 137 139). Sayahi (2007) analyzes two situatio ns of diglossia, one that involves French and Tunisian Arabic, and another that involves Spanish and Northern Moroccan Arabic. He claims that contact induced change can occur in these situations in which interference is facilitated from the foreign languag e into the lower variety on both the lexical and structural level (Sayahi, 2007). P assive familiarity another mechanism for contact induced change, refers to the activation of foreign elements by speakers who understand the language but do not
49 actively s peak it (Thomason, 2001:139) It is diffi cult to find changes due to this mechanism because it does not involve active use of the languages in contact ; however, Thomason (2001:139 140) discusses anecdotal examples where she claims that change is clear. Ne gotiation 5 can also bring about change in language contact situations. This mechanism involves the approximation of the patterns of two languages, and i f the speakers are fully bilingual, Thomason (2001 :142 ) claims that this approximation will lead to conv ergence. This phenomenon was evidenced in various studies on voice onset time (VOT) values of stops among bilinguals (Caramazza, Zurif, Yeni Komshian, & Carbone, 1973; Yavas & Wildermuth, 2006; Fowler, Sramko, Ostry, Rowland, & Halle, 2008; MacLeod & Stoel Gammon, 2009), who exhibited intermediate VOT values in both languages. On the other extreme, if there are no bilinguals in the language contact situation, this may lead to the creation of pidgins (Thomason 2001:142). Also discussed as a mechanism of cont act induced change are second language acquisition (SLA) strategies (Thomason, 2001:146). The mechanism of negotiation can be considered one of these strategies employed by L2 speakers. Another strategy is the gap filling approach in which speakers use ma terial from the L1 when speaking the L2 (Thomason, 2001: 146 147). This is their way to make up for their lack of linguistic knowledge. An example of this is the aspiration of Spanish unaspirated stops by speakers of English, a language that has aspirated s tops. The strategy of overdifferentiation is also found in SLA. This is when speakers carry over a distinction 5 It must be noted that nego tiation in language contact situations can also be a social phenomenon. That is, speakers may use their bilingualism to negotiate their identity in a particular community. Thus, in this context, it is not a linguistic phenomenon and may involve more consci ous efforts by the bilinguals (see Pavlenk o & Blackledge, 2004 )
50 from their L1 to the L2 where it does not exist (Weinreich, 1966; Thomason, 2001). A clear example of this is the phonemic difference that is mad e between the voiced bilabial stop /b/ and the voiced labialdental stop /v/ in Spanish by English speakers in words such as botar votar allophonic. The opposite strategy is to ignore the distinc tion in the L2 because they do not exist in the L1 (Thomason, 2001:148). This is referred to as underdifferentiation / allophonically in English when they actually have a phon emic function in this language. B ilingual first language acquisition can also bring about changes (Thomason, 2001). This is when a person learns two or more languages from birth. 6 In this situation negotiation can occur (Thomason, 2001:148), and outcomes s uch as convergence can be observed. This has b een evidenced in studies on VOT duration by heritage speakers who exhibit compromise values in their production of stops (Caramazza et al., 1973, Yavas & Wildermuth, 2006). Finally speakers can make d eliberate decisions in contact situations, which can serve as a mechanism for change. The phon ological example eme that does Bantu sometimes introduce this sound into their speech deliberately to emphasize differentness when using Bantu words (Thomason, 2001:150). 6 There are two options in child bilingualism. Children can learn two languages at the same time or they can learn one language after the other. The first option is referred to as consecutive bilingualism and the second, successive bilingualism (Romaine, 1995).
51 Apart from the se linguistic predictors and mechanisms i t is important to recognize that the underlying strategy in most of the processes is that multilinguals tend to identify a feature from one language with a similar feature from the other. This has been referred to as interlingual identification ( Weinreich 1966). Therefore even though English consists of aspirated voiceless stops and Spanish, unaspirated voiceless stops, a bilingual of these two languages will more than likely cognitively connect these sounds due t o their shared place of articulation and lack of vocal cord vibration This, in turn, could lead to convergence and a possible diffusion of change. Interlingual identification seems to be the driving force, especially with regards to phonology, in mechanis ms leading to contact induced change. This change allows us to observe several outcomes of contact, some of which we discuss in Section 2.1.3. 2.1.3 Linguistic Outcomes of Language Contact The social and linguistic factors discussed above can bring about several types of linguistic outcomes in language contact, such as simplification, overgeneralization, transfer, and analysis (Silva Corvaln, 1990). Simplification is defined as the reduction of the linguistic inventory, in which certain forms are expanded to the contexts of competing forms. In her research on Chicano Spanish in Los Angeles, Silva Corvaln (1990) discovered that when comparing second and third generation bilinguals to first generation monolinguals, there was simplification and loss in the S panish verbal morphology of the bilinguals. That is, they featured loss of the subjunctive mood and the simplification of the pr eterit imperfect opposition (Silva Corvaln, 1990:167 ). Overgeneralization is related to simplification; however, acco rding to S ilva Corvaln (1990:163 ), the difference is that the extension of a form is not at the expense of a competing form, and the form can appear where ther e was not a for m before. This
52 linguistic outcome has been evidenced in several studies concerning pronomin al subject expression in Spanish where, in spite of being a pro drop language, bilinguals tend to produce pronominal subjects in Spanish in contexts where, according to prescriptive grammar, they would not normally appear (Flores Ferrn, 2004; Otheguy & Ze ntella, 2007; Orozco & Guy, 2008 ). Transfer is defined as the incorporation of features from one language to another. It can refer to borrowing, extension or reduction of function, higher frequency of use of a form, and the loss of a form 7 (Silva Corvaln 1990). Although, these characteristics seem similar to simplification and overgeneralization, the assumption is that the forms are related to corresponding, or seemingly corresponding, forms in another language. Silva Corvaln (1990) defines analysis as a process in which there is preference for analytical or periphrastic forms as opposed to synthetic ones. According to the researcher, Spanish English bilinguals living in Los Angeles featured changes in periphrastic and auxiliary constructions involving ser estar and ir Bilinguals were employing these constructions exclusively in certain contexts with specific lexical meanings that did not align with fir st generation monolinguals (Silva Corvaln, 1990:73 ). While these linguistic outcomes are common in s ituations of language contact and bilingualism, it is crucial that we recognize that these same outcomes, particularly, simplification, overgeneralization, and analysis, can occur in monolingual speech and in dialect contact That is, they can be the resul t of internal linguistic processes. Therefore, 7 Transfer has been a highly debatable topic in literature concerned with contact induced change. Researchers have argued that many of those that study language contact have ill define d the term borrowing (recipient language agentivity) and imposition (source language agentivity). This distinction is based on the notion of language dominance ; that is, borrowing occurs when the speaker, or agent, is RL dominant, and imposition occurs when the speaker is SL dominant. This framework has been considered to be the most comprehensive in this line of research (Winford, 2007 ).
53 it is paramount in a study of language contact that the researcher disentangle contact induced change from internally motivated change in order to make claims that any change evidenced in bilingual speech is i ndeed due to contact. One way to distinguish between these two types of change is through the comparative variationist metho d (Poplack & Tagliamonte, 2001) discussed in Section 2.2. 2.2 The Comparative Variationist Enterprise I n their discussion on the comparative variationist enterprise Poplack & Levey (2010) show that much evidence brought to bear on contact induced change diachronic or synchronic either fails to demonstrate that change has occurred, and/or if it has, that it is the product o f contact and not internal evolution ( Poplack & Levey, 2010:391). In order to make this distinction, they propose a variety of steps First, one must determine that variation exists because change cannot come about without variation ( Poplack & Levey, 2010 :394). Next, one must locate the earlier stage or precursor variety to make a comparison with the contact variety. If a precursor cannot be referenced, a surrogate will usually suffice ( Poplack & Levey, 2010:394 395). Another factor to keep in mind is to n ot trust a standard variety for comparison because variability in the vernacular speech usually involves a competition between standard and nonstandard forms. Standard language is invariant, and therefore will lead us to believe that change has occurred wh en compared to vernacular speech ( Poplack & Levey, 2010:395). Related to the aforementioned step, one must examine comparable modes. That is, spoken vernacular can only be compared to spoken vernacular, and contact varieties should be compared to non conta ct varieties. These comparisons should also be systematic ( Poplack & Levey, 2010:395).
54 With respect to social factors, the language being captured for comparison should reflec t the real life experience of bilingual s with their languages. Any innovations t hat are observed must be diffused across the community in order to be considered change (Poplack & Levey, 2010:396). Also, bilingualism should be long lasting and stable. It is not adequate to look for change in shifting populations. Moreover, L2 speakers are not good candidates for analyzing change as they may exhibit many acquisition errors depending on their proficiency level ( Poplack & Levey, 2010:396 397). It is also important to understand the status of the innovation in the language, its regularity its productivity, and the extent to which it is entrenched in the language. That is, synchronically speaking, we need to pinpoint patterns of use through statistical analysis ( Poplack & Levey, 2010:397). Next, we must det ermine the source of the change a nd be able to explain whether it i s contact induced or internally motivated ( Poplack &Levey, 2010:397). Popla ck & Levey (2010:398) present recognizing contact induced change: 1) the form must be a bsent in a pre contact or non contact variety, or 2) if present, not conditioned in the same way as in the source language, and 3) the behavior of the form must parallel that of a counterpart feature in the source. The previously mentioned conditions can b e accomplished through systematic quantitative comparison s in which a comparative variationist method (Poplack & Tagliamonte, 2001) is employed (Poplack & Levey, 2010:398 401), which is further discussed in Chapter 2. Recall that in order to adequately e xamine potential effects of contact within the comparative variationist enterprise, we must identify a conflict site (see Section 1.5). A
55 possible conflict site could be apparent in terms of phonetics and phonology; thus, we focus on sound change occurring in bilingual speakers living in Panama. We can consider VOT duration to be an ideal site for the analysis of sound change among bilinguals because it differs between Spanish and English. Therefore, in Section 2.3, we focus on this linguistic property and describe in detail the differences that arise in this language pair, while also discussing various factors that have been shown to affect VOT duration. 2.3 Sound C hange Due to Language Contact As mentioned in Chapter 1, the linguistic property on which w e focus in this study is VOT duration of the voiceless dental plosive. In order to accurately examine this acoustic property, it is essential that we incorporate instrumental techniques into the analysis. Instrumental phonetics, also known as acoustic phon etics, is the use of techniques, such as spectrographic analysis, to examine the physical properties of speech sounds (Gordon, 2007; Hay & Grager, 2007). Although other alternatives are available for sound analysis, such as auditory phonetics (Hay & Drager 2007), perceptions when looking at sound change. In this section, I focus on the use of instrument al analysis in sociolinguistics and later give a detailed discussion of V OT duration in monolingual and bilingual speech. Thomas (2002: 169) states that there are many issues for which instrumental techniques are more In the analysis of discrete variables, measurements tend to be straightforwa rd, such as /h/ dropping in English, [ht] vs. [t]. That is, the presence versus the absence of a sound is more noticeable in speech because of its binary nature (Gordon, 2007:19 21). In this type of study, an
56 auditory analysis could be sufficient; howeve r, when variables are continuous, the coding is much more complicated. This is the case with // raising in which there is a continuum between  and [i]. When examining this phenomenon, the most appropriate technique is instrumental analysis, because it allows for the capturing of different degrees of a change through precise measurements (Thomas, 2002; Gordon, 2007). It also assists in the analysis of properties that are context dependent (Thomas, 2002:169) such as voice onset time. The goal of an instru mental analysis is to capture variables of a continuous nature in a discrete way. Thomas (2002:169) has also noted that instrumental analysis is suitable for studies on ethnic identification that involve the ch synthesizers. It is also applicable in variati onist studies on production (Thomas, 2002: 170). This is usually referred to as techniques to examine speaker perception of phonetic boundaries based on voice onset time. It can be seen that this technique is useful and reliable when utilized correctly. It perception is excluded from the analysis th rough impartial measurements. Besides, it is easily perceptible. This phonetic property is discussed in Section 2.3.1, with particular attention on Spanish and English. 2.3.1 Voice Onset Time Lisker & Abramson (1964 :387 ) define voice onset time (VOT) as the duration of the time interval between the release of a stop and the glottal vibration, whether this onset occurs before the release or follows it. VOT is a universal property in languages but they can be organized differently. When discussing this property, the relevant
57 sounds are voiceless and voiced stops. The production of a stop involves certain sequences of articulatory and acoustic events (Stevens, 1998:324). Acc ording to :8 ) description, the passage of the pulmonic egressive airstream out of the oral cavity is completely impeded, thereby causing a build up of intra oral pressur e behind the constriction The stop closure consists of three stages: 1 ) complete closure leading to a build up of pressure, 2) rapid release of the air build up, and 3) complete aperture of the articulators right before voice onset (Docherty, 1992:8). When there is presence of glottal buzz during the closure, the stop is con sidered voiced. If there is absence of glottal buzz during this stage, the stop is categorized as voiceless (Lisker & Abramson, 1964: 384). Figure 2 1. A waveform showing voice onset time and voiceless closure interval o f a word initial stop [Reprinted with permission from Zampin i, Mary L., & Kerry P. Green. (2001) The voicing contrast in English and Spanish: The relationship between perception and production In J anet L. Nicol (ed.), One mind, two languages: Bilingual processing (Page 24, Figure 2.1) B oston: Blackwell. ] The two kinds of stops can be acoustically distinguished through a spectrographic or waveform analysis. Figure 2 1 is a simple image of a waveform exemplifying voiceless closure, release burst of the stop and the voice onset time (Zam pini & Green, 2001:24). Voiceless stops are normally represented as /p t k/, where /p/ symbolizes the voiceless bilabial, /t/, the voiceless dental/ alveolar, and /k/, the voiceless velar. The voiced stops are presented as /b d g/. /b/ is the voiced bilabi al, /d/, the voiced dental/ alveola r, and /g /, the voiced velar. Since in certain languages, such as English,
58 voiced stops can be produced with silent closure in initial position, we must rely on an additional attribute, aspiration, to distinguish between / p t k/ and /b d g/ (Lisker & release (Docherty, 1992:8), which is noticeable in a spectrographic analysis. Therefore, we recognize that there are two features that cor relate with aspiration and voicing, according to Lisker & Abramson (1964 :387 ), and in the spectrogram they are detected as the periodic pulsing at the frequency of the voice pitch and the noise in the freque ncy range of the high formants Figure 2 2. A spectrographic and waveform analysis o f the Spanish phrase tal vez Where there is absence of one feature, there is presence of the other. We use this pattern to pinpoint the stop and to measure the interval between the release of the burst and the onset of voicing, which has been labeled as voice onset time (Lisker &
59 Abramson, 1964). An example 8 of a true spectrographic analysis can be seen in Figure 2 2 in which the boundary marks the voice onset time of [t] (11 msec.) in tal vez maybe As stops are apparent in almost all languages, Lisker & Abramson (1964:388) present three groups of languages based on their number of stop categories: 1) two category languages: American English, Cantonese, Dutch, Hungarian, Puerto Rican Spanish, and Tamil, 2) th ree category languages: Korean, Eastern Armenian, and Thai, and 3) four category languages: Hindi and Marathi. Despite the number of categories, there are only four combinations of voicing and aspiration that emerge in measuring stops. The first is when v oicing precedes the release of the stop, which is labeled as lead. This would be a voiced stop that is unaspirated. The second is when the voicing follows the release, which would be a voiceless and unaspirated stop. This has been labeled as short lag. The third is when voice onset lags significantly behind the release. This is a voiceless aspirated stop, which is known as long lag. As mentioned previously, certain dialects of English can feature a voiced aspirated stop, which is also considered short lag. Thus, there are two combinations falling under the same option, voiceless unaspirated stops and voiced aspirated stops (Lisker & Abramson, 1964:389). In view of the aforementioned categories, voicing oppositions with regard to lead, short lag and long lag emerge. As the present paper is concerned with Spanish an d English, Sections 2.3.2 and 2.3.3 focus on the VOT measurements of these two languages and describe the voicing oppositions that come about. 8 The examples of spect rographs that appear in this paper are from my own data.
60 2.3.2 Spanish VOT Values In word initial position, the bilabial sound /b/ features a mean VOT value of 138 milliseconds (msec.), the dental sound /d/ is produced at a mean VOT value of 110 msec., and the velar sound /g/ has a mean VOT value of 108 (Lisker & Abramson, 1964:392). In intervocalic position, ho wever, these sounds undergo fricativization (Guitart, 2004), a process by which the sounds become approximates. This makes VOT measurements complicated and almost impossible as aspiration is not detected. The approximates are represented with the following occurs after the release. In word initial posi tion, the mean VOT value for the bilabial sound /p/ in Spanish is 4 msec., for the dental sound /t/ is 9 msec., and for the velar sound /k/, 29 msec. (Lisker & Abramson, 1964:392). Based on these measurements, it can be concluded that, in Spanish, there is a voicing opposition of short lag, where voicing occurs with or shortly after the stop release, and lead, where voicing occurs before the stop release. 2.3.3 English VOT Values In English, the voiced stops are presented with two value sets (Lisker & Abra mson, 1964:394). According to these researchers, certain dialects can have positive VOT values (aspiration) and negative VOT values (non aspiration) for these sounds. The bilabial stop /b/ features a mean VOT value of 1 msec. or 101 msec. The alveolar sto p /d/ is produced at a mean VOT value of 5 msec. or 102 msec. Lastly, the velar stop /g/ has a mean VOT val ue of 21 msec. or 88 msec. (Lisker & Abramson, 1964: 394). The voiceless bilabial stop /p/ in word initial position receives a mean VOT
61 value of 58 msec. The alveolar stop /t/ has a mean VOT value of 70 msec. The velar stop /k/ is produced at a mean VOT value of 80 msec. These values classify these sounds as aspirated. Said aspiration is represented in the following examples: [p h [t h h up]. (Carr, 1999; Guitart, 2004; Lisker and Abramson, 1964; Stockwell & Bowen, 1965). According to these measurements, English exhibits a voicing opposition of short lag, where voicing occurs with or shortly after the stop release, and long lag in which the voicing occurs well after the release of the stop. Another opposition is lead, where voicing occurs before the release, and long lag in some varieties. Although these mean VOT values have been evidenced in several varieties of Spanish and E nglish, it has been noticed that consonant stops can show a wide range of variability. Said variability can depend on several factors, whether they are linguistic or e xtra linguistic. In Section 2.3.4, we discuss some of the previous studies that test thes e factors with regard to VOT in both Spanish and English. Also, where it is possible, some results from analyses of Spanish English bilinguals are mentioned. 2.3.4 Factors Conditioning VOT There are several factors that can condition the values of VOT in S panish and English, exhibiting similar effects in both languages. One of these factors is place of articulation, which affects both voiceless and voiced stops. It has been observed that VOT values become greater as place of articulation moves from anterior to posterior. Jacques & Gurlekian (1992) found that in Buenos Aires Spanish the voiceless bilabial /p/ exhibited a shorter VOT value than the corresponding dental stop /t/, and the dental featured a shorter VOT value than the velar stop /k/. The order was the same for the voiced counterparts. The VOT value for the voiced bilabial /b/ was shorter than that of
62 the voiced dental /d/, which in turn was shorter than the voiced velar /g/. Poch (1984), Rosner et al. (2000), and Munday (2001) observed the same pat tern in European Spanish. In Munday (2001) speech style was taken into consideration and the investigator found that it did not have an effect on the order. Roldan & Soto Barba (1997) found the same correlation between VOT values and the place of articula tion. However, the direction was different for each category. For /p t k/, VOT increased as the place of articulation moved from anterior to posterior, and for /b d g/, VOT decreased in this direction. The variety analyzed in this study was Valdivian Spani sh from Chile. As for English, Morris, McCrea & Herring (2008) found this pattern among young men and women in Flo rida and Tennessee. Fowler, Sramko, Ostry, Rowland, & Halle (2008) discovered that in Canada, monolinguals exhibited the same patterns for pl ace of articulation for voiceless stops. Voiced stops were not examined in this study. In the studies mention ed above, place was a significant factor. As for Spanish English bilinguals, the same results were shown in Yavas & Wildermuth (2006) and Thornburg & Ryalls (1998). The phonological environment is said to affect the production of sounds. With regard to VOT, the main environments examined have been preceding and following segment. With regard to preceding segment, Munday (2001), who analyzed the spe ech of only two speakers, found that the differences among VOT duration in one speaker were highly significant for /t/. VOT values were longest when the preceding sound was a fricative in all three speaking styles. For /k/, there was high significance amon g VOT values only when reading sentences. A preceding stop caused VOT values to be longest in this style. In the speech of the other participant, significant values were found
63 in sentence and text reading, but not in conversation for /t/. When the precedin g segment was a fricative, the VOT value was longest. For /k/, significance was found in the conversation mode. VOT duration was greater when there was no preceding significa nt results for the voiceless bilabial /p/ in this study. Roldan & Soto Barba (1997) examined the differences in values in Valdivian Spanish when the preceding segment was zero or a nasal consonant. It was observed that /p t k/ did not vary in VOT values de pending on the context, and that the values of /b d g/, where the preceding segment was a nasal consonant, were shorter than when the preceding segment was zero. In English, Yoshioka, Lfqvist, & Hirose (1981) noticed a one peak pattern in spectrographic a nalyses in clusters of voiceless fricative + voiceless plosive. No studies have reported on Spanish English bilinguals with regard to preceding segment. The following segment has also been seen to affect the duration of VOT in stops. The segment type that has received most attention is the vowel. According to Lisker & Abramson (1964), vowel height does not condition stops with regard to VOT values. H owever, this has been refuted in several studies in which VOT values are longer with higher vowels. The effe Castilian Spanish with all but one stop, the voiceless velar. According to Rosner Lopez Bascuas, Garcia Albea, & Fahel. (2000), Valdivian Spanish featured longer VOT values for /p/ and /k/ when the following vowel was /o/ rather than /a/. The voiced sounds /b/ and /g/ had earlier voicing in this context. The values for /t/ and /d/ were not significant vowel heigh t. The VOT values are longer as the vowel became higher.
64 in Tennessee and Florida. VOT values were longer when followed by high vowels and shorter when followed by low vowels. The same vowel effect was seen in bilingual speech. Yavas & Wildermuth (2006) found that VOT values were greater when the following vowe l was high in the speech of native Spanish speakers learning English. Least common is a focus on the postvocalic consonant; that is, in a CVC syllable, the consonant that follows the vowel. It has been seen that the postvocalic consonant can condition the length of the initial stop. Port & Rotunno (1979) and Weismar (1979) supported this hypothesis in their study in which the former showed that VOT values of voiceless stops were shorter when the postvocalic segment was the consonant cluster /pt/ as opposed to /n/. The latter found that VOT values of voiceless stops were longer when the postvocalic consonant was voiced as opposed to voiceless. Syllable stress has been seen to affect VOT duration, although very few studies have focused on this factor. Casta eda (1986) and Dniz (2005) discovered that when voiceless plosives appeared in stressed syllables, their VOT values tended to be shorter in Peninsular Spanish and Canarian Spanish, respectively. Lisker & Abramson (1967 ) found that in monolingual varieties of English VOT duration was longer in tonic syllables, and shorter in atonic ones. Another factor conditioning VOT values of stops in Spanish and English is speech rate. Generally, researchers have hypothesized that a faster rate of speech will usually c ause VOT duration to be shorter. Schmidt & Flege (1996) supported this claim in
65 English VOT values. They found that English monolinguals produced the voiceless bilabial and dental with shorter VOT values at the fast rate of speech than at the normal rate. The Spanish monolinguals showed little effect of speech rate on /p/ that appeared in initial position of a sentence and on initial /t/ in the final word of a sentence. However, a significant effect in the opposite direction of English /p/ was evidence for Spanish /p/ (1999) study in which English voiced and voiceless stops are rate dependent. This was also the case for Spanish voiced stops, but not voiceless stops as was seen in Schmidt & Flege (1996). Schmidt & Flege (1996) and Magliore & Green (1999) also examined the effect of speaking rate on bilinguals. Schmidt & Flege (1996) observed that early Spanish English bilinguals produced English stops with similar VOT values as English monolinguals and exhibited the same speaking rate effect. Late bilinguals had a smaller effect of speech rate on their VOT values in English. They also showed a wider range of VOT values in English, thereby supporting the claim of a merged phonetic representation of stops. Schmidt & Flege (1996) did not examine the production of Spanish stops by bilinguals. Magliore & Green (1999), however, found that early bilinguals showed little effect of rate of speech on Spanish stops. These results filled the e (1999) analysis that early bilinguals have separate phonetic categories for English and Spanish. OT duration of stops. Due to the smaller vocal tract volume, it has been hypothesized that women would produce longer VOT values (Morris et al. 2008:309). However, according
66 both voiced and voiceless stops. Age has also been examined in the production of stops. Eilers Oller, & Benito (1984) carried out a longitudinal study of Spanish and English monolingual infants in Miami. They found that at age 1, infants had not gained th e capacity to systematically distinguish phonetic category boundaries for initial consonant stop voicing in their language where that difference exists. However, by age 2, both English and Spanish learners were able to make this distinction, and showed evi dence of approaching adult like VOT values (Eilers et al., 1984). For bilinguals, the age of acquisition of the second language is usually the focus. language phonetic differences are more li kely to be discerned as the age of acquisition decreases. 9 Thornburg & Ryalls (1998) found that early learners of English usual ly contrasted VOT values to a greater degree than later learners of English, thus (1995a) model. Early French English bilinguals are more capable of maintaining a separation of phonetic categories (Caramazza et al., 1973; Macleod & Stoel Gammon, 2008). However, interference was more evident from Canadian French to Canadian English, which Caramazza et al. (1973) c laim was due to these bilinguals learning English at a later age. The studies analyzing bilingual production of stops have shown that several factors play a role on VOT values and have shed light on how bilinguals mentally 9 See discussion on different hypotheses c oncerning L2 phonology in section 2.5.
67 se parate languages In the S LM, Flege (1995a) bases the merged phonetic representation hypot hesis on evidence of compromise VOT values that is, values that fall between the prototypical ranges for each language. This process is referred to as cross language assimilation, and is said to occur because bilinguals make an equivalence classification of similar sounds from both languages ( Flege, 1987; Flege, 1995b:101). If the L2 is learned at a later age, the development of new or similar L2 phones is blocked because older children and ad ults do not note the phonetic differences between L1 phones and L2 phones (Flege, 1987:50). Another type of process that can occur is cross language dissimilation, in which speakers deflect from a phonetic category in order to maintain a distinction. Tha t is, a Spanish English bilingual may produce extra long VOT values in English in order to distinguish from longer VOT values in Spanish. According to Flege (1995b:101), this is more common in bilinguals who acquire the L2 at an earlier age Since the L1 i s not completely developed in younger children, they do not solely depend on the L1 to categorize L2 phones (Flege, 1987). Perceptually, although they coexist in the bilingual mind, these L1 and L2 sound categories move away from each other, thus showing l ess convergence in their production (Guion, 2003; Flege, 2007; Simonet, 2010). Some limitations can be pointed out with regard to these studies on VOT duration. The first has to do with data collection. All of the studies, except for Munday (2001), analy ze VOT in a laboratory setting, where participants are asked to read lists of words or sentences. This is pointed out as a limitation, at least in the perspective of the variationist framework, since reading tasks do not represent real language use. Munday (2001) is the only study that incorporates conversation as one of the speaking styles
68 conditioning VOT duration; however, this analysis only consists of two speakers. It will be wor thwhile to see what the present analysis offer s when looking at natural sp eech among a greater quantity of participants. Another limitation is the practice of separating languages when analyzing bilingual speech. Magliore & Green (1999), for example, had bilinguals tested in two separate sessions for Spanish and English. The pa rticipants knew that both sessions had to do with the same analysis only at the end of the second session. The results showed that these bilinguals were able to maintain separate phonetic categories for each language. This procedure is considered a limitat ion to the study because in a natural environment, bilinguals constantly have both languages activated (Grosjean, 2001). Separating the languages does not parallel true language use. A study seeking to determine whether or not bilinguals can maintain separ ate phonetic categories can be more valid if this is supported in natural speech without the separation of language modes. This is one of the goals of this study 2.3.5 English vs. Creole English VOT Up to now the languages involved in th is discussion hav e been Spanish and English. This study, however, examines Panamanian varieties of Spanish and Creole English. The distinction between Creole English and Standard English varieties has not been made until now because in Pana ma, when referring to the former one would normally say In Chapter 1, we discussed the creole continuum on which the Panamanian Creole English varieties can be identified. Panama City Creole English (PCCE) has been characterized as a mesolect (Thomas Brereton, 1992) and Bastim entos Creole English (BCE), as a more basilectal variety. Therefore, it could be expected that VOT values in PCCE, and especially BCE would differ from the figures
69 reported for North American varieties of English in Section 2.3.3. Lack of studies on this particular phenomenon makes it impossible to compare the VOT values beforehand. This study serve s to fill this gap and could shed light on the extent to which creole varieties possibly differ from standard ones with regard to acoustic cues. At this point, the possibility that VOT patterns of Creole English are similar to those of Standard English has not been discarded, but nor is it assumed. One of the main goals of linguistic theories is to account for the representation of language in the minds of indivi duals. In order to reach this goal, many have focused on monolingual speakers because they do not exhibit external influences from another language. However, as stated in Section 2.1, language contact and multilingualism is a norm in almost all societies, thus focusing on only monolinguals does not completely explain how language is represented in the mind. A theory of linguistic should be able to account for situations of multilingualism so that we have complete story of the cognitive processes that are in volved in language development, variation and change. The present analysis is based on a usage based model of language in which frequency of use play a crucial role in the representation of language. In Section 2.4, we discuss this model and propose that l anguage contact and bilingualism in Panama be explained from this framework. 2 4 Usage based Models of L anguage T he analysis is concerned with how innovations and possible changes due to contact can be explained by usage based models of language. The con proposes a dynamic approach to explain the mental representation of language. According to Langacker (2000:3), this model should prove to be a unified account of a ll
70 aspects of language, rather than handling them in separate ways. The basic orientation of a usage based model of language is one that analyzes language from a bottom up approach, as opposed to the traditional top down approach (Langacker, 2000:2). The t op down approach refers to the idea that instantiations of language are produced or perceived based on generalizations. If these generalizations, or rules, cannot be applied to an idiosyncratic language event, then such event is assumed to be stor ed in the lexicon. Bybee (2001: 20) states that the emphasis of this type of model is the systematicity of language, where the focus is to reduce complexities by capturing the regularities and listing the idiosyncratic elements. This idea is referred to as the rule/ list fallacy ( Langacker, 1987; Langacker, 2000 ; Bybee, 2001 ), That is, a linguistic unit either has to be generated by a rule or stored in the lexicon. The bottom up approach differs in that language is considered massive and highly redundant, and a speak is redundancy (Langacker, 2000: 2). The bases of language in this approach are the actual instantiations of production and perception. Input and use help to establish an inventory of linguistic k nowledge through memory, which later leads to more general tendencies. Experience and memory is important in this model because predictable features are mapped onto mental representations of previous instantiations of the feature. When the speaker encounte rs an unpredictable feature, new memories are created and the feature is stored in the brain (By bee, 2001:20). 2.4.1 The Usage Event Usage based f a vocalization, in all its specificity,
71 specificities could include physical, social, cultural, and linguistic factors. B oth production and perception make up the usage event or what is also said to be an instantiation of language use. Langacker (2000) does not differentiate between what speakers produce and what they perceive because in comprehension, the hearer has to be able to interpret an utterance with its intended mean ing. In production, the speaker has to choose the correct linguistic structure that evokes the intended interpretation, and also be sure that it is understood (Langacker, 2000:10). Therefore the experience of the speaker/hearer consists of a constant infor mational loop where production affects perception and vice versa ( Volaitis & Miller, 1992; Miller, 1994; Miller, 2001; Pierrehumbert, 2001 ), and in which certain cognitive and linguistic processes are evidenced. Anything can be stored in memory and can hav e an impact on the linguistic system according to this type of model The impact is determined by repetition and frequency (Bybee, 2000 2001), which are driving forces in usage based models 2.4.2 Cognitive Processes There are various cognitive and psycho logical processes that are important in language organization and use. The first process to be discussed here is entrenchment which is the result of the repetition of a linguistic form. When a form is repeated at a high rate in a given context it can be r outinized and stored in a spe (Langacker, 2000: 3). Each instantiation of the form contributes to its strength and makes it easily accessible in real time. The higher the frequency of the form, the stronger it becomes in our memory, and the str onger it is, the more accessible it becomes, no matter how complex the utterance ( Nathan, 1998; Langacker, 2000; Pierrehumbert, 2001; Hopper & Traugott, 2003; Bybee, 2001; 2006; Bybee & Torres Cacoullos, 2008; Brown, 2009 ). Entrenchment can also cause the formation of units,
72 that is, complex structures produced without paying close attention to its various parts. Therefore some forms can have either non unit or unit status. An example of entrenchment can be the high frequency of words with /s/ lenition in S panish. If repeated enough, common words such as ms more and frequent phrases such as las cosas can appear with a weakened /s/ in all contexts. It would be concluded from this that the lenited forms of these structures have become entrenche d. Another cognitive process involved in language is abstraction This occurs when the possible differences of a structure exhibited in multiple experiences are overlooked and the inherent commonality is reinforced (Langacker 2000:4). Speakers have the ab ility to notice differences on various levels of granularity. This is what Langacker (2000) calls schematization a type of abstraction. In this process, a schema emerges when one fails to notice differences in the instantiations of a single structure. The speaker operates on a high level of granularity where there is less focus on details (Langacker, 2000:4). An example of this schematization can be seen in the length of aspiration in voiceless stops. A voiceless dental stop can exhibit variation in its VO T with each instantiation, but it is still represented as a voiceless dental stop in spite of this fine grained detail. Although the inherent commonality detected in language is very important for schematization, also apparent in cognition is the ability to detect differences between two structures. In this process, one structure is the standard and the other is the target (Langacker, 2000: 4). There are instances where the standard can represent an established unit and the target is a novel instantiation. Langacker (2000) call this categorization which in some cases, does not show any discrepancies between the
73 standard and the target form. When a target is produced, a variety of established units with similar characteristics can be activated. This set of units is called the activation set (Langacker, 2000:15). Eventually, one of the members of the activation set is chosen to categorize the target depending on the level of entrenchment, the contextual priming and the amount of overlap between the target and the potential categorizing structure. The winning member is considered the active structure It is also possible that deviations from the target occur, causing differences between the standard and the target. This is considered an extension (Langacker, 20 00: 4). That is, even though the target does not completely conform to the specifications of the standard it is still considered a good instantiation. Therefore in a language, such as Spanish, that permits VOT values ranging between 0 15 msec. for voicele ss dentals stops, a stop exhibiting a value of 7 msec. would activate other stops with a similar measurement, not going beyond the prototypical range. These stops would be the activation set. Depending on the level of entrenchment, the contextual priming, and the amount of overlap between the target sound and the potential categorizing sound, one of the stops would be selected to categorize the target. In the case that a stop with a value of 25 msec. is accepted by the speaker as being a good fit, this woul d be considered extension because it falls outside of the prototypical range. Another cognitive process to be considered is composition which is the integration of simple structures to produce more complex ones (Langacker, 2000:4). These complex structure s become their own units, and are no longer considered a combination of their components. An example of this process would be the combination
74 of k and l According to Langacker (2000 :4 ), the composite kl is no longer taken to be the combination of the two sounds, but rather an indepen dent structure in its own right. Also in usage based models of language, o ne kind of experience is able to be invoked by another (L angacker 2000:5). This process is called association. Within association, we can find symboliza tion where there is an association between conceptualizations that have certain observable entities in their mental representations. That is, certain gestures can be stored mentally and serve to elicit a certain sound. In the case of VOT the coarticulati on of the teeth, tongue, and glottis can conjure a certain VOT value. phonetics, it is explained that phonetic gestures are the basic units of the phonological system. Therefore, in the usage based approach, the exemplar clusters would include the gestures associated with the voiceless dental stop in either Spanish or English through symbolization The cognitive processes mentioned above are also apparent in the exemplar model of ph onology discussed by Pierrehumbert (2001). First, we find that categories are represented by clouds of tokens that have been stored through input and use. These clouds, which consist of memories, are organized and connected to each other in a network. Some connections are stronger than others; they depend on the association that one memory has with another and on the frequency of the memory. Also, the level of similarity that memories have with one another determines the distance bet ween them (Pierrehumbert 2001: 140). That is, memories of higher similarity are closer to each other in the network than those with lower similarity.
75 Another important aspect of the exemplar model is granularity. Pierrehumbert (2001) states that if differences between exemplars are too fine grained, they are Association strength, which was discussed by Langacker (2000), can also be found in the exemplar model. More frequent recent experiences esta blish exemplars with higher resting activation levels than exemplars with infrequent remote ex periences (Pierrehumbert, 2001: 141). New tokens are mapped based on its similarity with other tokens already stored in the system. If the stored tokens are highly activated, when a new token is encountered, the new token will receive the label of the highly activated ones. The exemplar model also involves prototype effect s That is, certain exemplars are considered better than others with respect to a category. Ac cording to Pierrehumbert (2001:143), new tokens are better examples of a category than previously experienced ones. Bybee (2006 :717 ) also states that exemplar representation exhibit prototype effects, in which members are central to the category. Miller (1 994) and Miller (2001) s of certain voiceless stops is affected by context. They examined how rate of speech which was determined by syllable strength, could shift the best exemplar range of VOT values. They reported that the best exemplar range was shifted toward longer VOT values for longer syllables (Miller, 1994: 276), thus showing that listeners can rate some exemplars to be better than others. Various researchers have discovered several linguistic proces ses in usage based model of language and phonology (Bybee, 2001; 2006; Langacker, 2000;
76 Pierrehumbert, 2001; Bybee & Torres Cacoullos, 2008; Brown, 2009), and all have shown that the overarching idea is repetition and frequency of use. According to Bybee ( 2001), language is a conventionalized behavior in which structures are shaped by repetition. This can lead to lexical strength; greetings become easily accessible because they are ritualized. Repetition also lead to reduction of forms such as how are you t o hi and going to to gonna It can also lead to reduction of meaning, as in the lost of the emphatic meaning of pas step in the French negation construction ne pas (Bybee, 2001:8 9). Repetition is intimately linked to frequency, in which various effect s are observed. One of the main effects of frequency is reduction, such as /s/ lenition (Brown, 2009), where higher frequent words exhibit more /s/ reduction than lower frequent words. Bybee, (2000b) also discovered that t/d deletion was more apparent in h ighly frequent words in the speech of Chicano speakers in Los Angeles. A significant effect of word frequency in t/d deletion was also seen in Gregory, Raymond, Bell, Fosler Lussier, & Jurafsky (1999). Reduction has also been observed in articulatory gestu res (Pagliuca & Mowrey, 1987; Bybee; 2001). Pagliuca & Mowrey (1987) propose the replacement of the terms lenition and assimilation for more general one s substantive reduction and temporal reduction The former refers to reduction of muscular gestures suc h as /s/ aspiration and the latter involves the compression of gestures such as regressive nasalization of vow els before nasal consonants (Pagliuca & Mowrey, 1987:460 462). Another effect of frequency is conservation. That is, highly frequent forms resist phonetic change or change due to analogy. Therefore, past tense forms of weep creep and leap can be regularized to weeped creeped and leaped respectively. However,
77 the past tense form of keep cannot be regularized to keeped because of its high frequen cy (Bybee, 2001:12). Clark & Watson (2011) also find a conservation effect in Liverpool English in which t to r ( t flapping) change was resisted in contexts of high token and type frequency due to the strengthen of schema as a result of a high rate of inst antiations. A c /s/ reduction in four dialects of Spanish. He finds that high ly frequent two word strings in high reduction dialects, such as Venezuelan and Puerto Rican Spanish show mor e /s/ maintenance when the following segment is a vowel than in low reduction dialects, such as Colombian and New Mexican Spanish. He attributes this to entrenchment (Brown, 2009:170). This would mean that two word strings consisting of a /s/ preceding a v owel are treated as units that, as discussed by Langacker (2000), are produced automatically and are easily accessible because of its high frequency. Pierrehumbert (2001 :150 ) also discusses entrenchment in the exemplar model and explains that this process hel ps to decrease variance Autonomization, th e third effect of frequency refers to the fact that highly frequent morphological complex words can become autonomous or independent, from their related forms. An example of this could be dislocate which i s more frequent than its base locate An even more extreme case of autonomy involves suppletion. A form can disconnect from its paradigm and join another due to its high frequency, for example, go and went (Bybee, 2001 2006). The latter was originally th e past tense of the verb wend We have seen that linguistic processes in the usage based approach are directly tied to the cognitiv e processes discussed especially that of entrenchment. Also, repetition and frequency lead to conventionalization, reduction conservation, and
78 autonomization (Pagliuca & Mowrey, 1987; Nathan, 1998; Bybee, 2001; 2006; Langacker, 2000; Pierrehumbert, 2001; Bybee & Torres Cacoullos, 2008; Brown, 2009 ). These processes could be evidenced in the present study, especially in regards to extension. I propose that changes occurring in the speech of bilinguals emerge by way of accepting new exemplars of the voiceless dental plosive that exhibit longer VOT durations than those of monolinguals speakers of Spanish and Creole English. Apart from the cognitive and linguistic processes, u sage based models also can account for the social aspect of language. 2.4.3 The Social Aspect of the Usage based Approach In the exemp lar model, Pierrehumbert (2001: 151) states that entrenchment can also involv e feedback, which affects how people adjust categories faster and to a greater degree. Speakers adapt their speech patterns based on their community through social accommodation which guides production. Exemplars are strengthened only when speech signals a re analyzable; ambiguous forms are not labeled. There is a constant feedback loop fr om production to classification which is evidenced in conversational int eractions (Pierrehumbert, 2001: 152). Thus, without social contact between individuals in a speech co mmunity, language would be static and would not exhibit dynamic processes and changes. We discussed in Section 2.1.3 that bilinguals can exhibit different outcomes due to language contact. Particularly in phonology, contact outcomes have been explained th rough several models concerned with similarities and dissimilarities and markedness factors in the L1 and the L2. These models of bilingual phonology happen to fit well within a usage based model of language. They appear to incorporate the notion of exempl ars in the classification of L1 and L2 phones during acquisition. In Section 2.4.4,
79 we discuss some of the hypotheses that have been made and a particular model that is relevant to the present analysis and the usage based approach. 2.4.4 Second Language Ph onology In the Contrastive Analysis Model, Lado (1957) posits that any areas of difference between an L1 and L2 will cause difficulty for the acquisition of the L2, and thus, transfer of the native sound system can ensue. The transfer involves production d istortions and perception blind spots. In the former, L2 learners tend to speak with a foreign accent and in the latter, learners miss differences that exist in the L2 (Lado, 1957:11). According to Lado (1957:11), when L2 learners hear the L2, they really hear their own language. He further claims that this transfer is usually in one direction, from the native language to the foreign language (Lado, 1957:11). The hypothesis is later updated by Oller & Ziahosseiny (1970:186) in which they propose a hierarchy of difficulty. L2 learners will find difficulty where there is minimal distinction in form or meaning in one or more systems. This hypothesis was supported by a study that they conducted among English language learners whose L1 used a non Roman script, su ch as Japanese. These learners had better abilities in spelling the English language than learners whose L1 used a Roman Script such as French. According to Major & Kim (1996), this idea of facilitation through the dissimilarities between the languages can be applied to phonology. L2 phones that have a similar sound in the L1 is more difficult to learn than a L2 phone that has a dissimilar sound in the L1. This is based on the idea that differences that are greater will have more saliency for the L2 learner than sounds that are similar. The learner is more likely to overlook minimal differences between the similar phones (Major & Kim, 1996:154). Wode (1983) states that L1 transfer only occurs when there are crucial similarities between the L1 form and the L2 form. If the L2
80 form does not have the similarity requirements, then it is acquired as if it were a process of L1 acquisition (Wode, 1983:185). Various researchers have presented and tested hypotheses concerning phonological similarity/dissimilarity. Fl ege (1987), in a merger hypothesis, claims that similar sounds are harder to learn because they are identified as being equivalent to a sound in the L1. This hypothesis is based on an equivalence classification that allows speakers to establish sound categ ories based on the phonetic information of the sound. If the L2 sound is similar to an L1 sound it will be classified as the L1 sound and will not be produced authentically. His study on English speaking learners of French confirmed this hypothesis. Experi enced L2 learners of French produced the French / / more authentically than the French /u/, which has a similar counterpart in English. hypothesize that the amount of L2 experience w ould not affect the L1 German experience would allow them to produce // authentically since a counterpart does not exist in German. They performed two experiments comparing the production of the English vowels by experienced learners and inexperienced learners who were native speakers of German. They were instructed to produce these vowels appearing in the context of /bVt/. The results of the comparison confirmed the hypothesis. The more experienced group did not produce the s than the less experienced group. With regard to //, which does not have a counterpart in German, inexperienced L2 learners produced this new vowel in the same way as the native English speakers. Major (1987 ) also discovered support for the
81 similarity/dissimilarity hypothesis in Brazilian Portuguese speakers learning English. The //. Despite the support found for the h ypothesis, Major & Kim (1996) point out that identifying what constitutes a more similar sound or a dissimilar sound is not always clear. They provide the examples of the German /x/ and English /k/, which seem to be more similar to each other due to place and manner of articulation, than German /s/ versus English / /, which also has the same type of similarity (Major & Kim, 1996:156). The question here is how it is determined that one pair shows more similarity than the other. In order to better account for this lack of precision, Eckman (1977) incorporates markedne ss to determine degree of difficulty. The Markedness Differential Hypothesis (MDH) states that less marked phenomena are acquired before more marked ones. Markedness Hypothes is (IDMH) which assumes that markedness relationships are not only found between L1 and L2, but also within the L2. Major & Faudree (1996) found evidence that supported claims related to markedness in the acquisition of voicing contrast by Korean learners of English. In Korean, there are only voiceless obstruents, and in English voiced obstruents in final position are more marked than in initial and medial position. Therefore, Korean learners of English should not acquire voicing contrast in final position until they have acquired the contrast in initial and medial position in English. Major & Faudree (1996) found that Korean learners showed 100% accuracy with the unmarked phenomenon (initial and medial position) and only 50% accuracy with the marked phenome non (final position).
82 Eckman (1991) later developed the Structural Conformity Hypothesis (SCH) based on results found in Farsi speakers learning English. These speakers exhibited word final devoicing when producing English in spite of the fact that Farsi has voicing contrast in final position along with initial and medial positions (Eckman, 1984). In the SCH, instead of considering areas of differences, it is claimed that interlanguages obey primary language universals. Therefore, as was seen in Farsi lear ners of English, the interlanguage can have voicing contrasts in initial and medial but not in final position, although the native language has this contrast in all three positions (Major & Kim, 1996:157). Major & Kim (1996) problematize the importance of similarity/dissimilarity and markedness for L2 acquisition. In the similarity/dissimilarity hypothesis, dissimilar phenomena are acquired more easily than similar ones, in spite of markedness. From the perspective of markedness, less marked forms are acqu ired more easily than more marked forms, regardless of similarities and differences. Major & Kim (1996:158) point out that similarity and markedness can therefore compete in L2 acquisition with certain phenomena. This seems to be the case for English speak ers learning Arabic voiceless velar fricative /x/ and voiced pharyngeal fr /x/; therefore English speakers s 2001:45). Major & Kim (1996:158) state that the solution would be to set a hierarchy in which one claim overrules the other. Anecdotally, they explain that English learners of Arabic actually have mo
83 discrepancy, it appears that with certain phones, markedness and similarity/dissimilarity can worked together to either benefit L2 learners or make learning the L2 more difficult, which means that they do not compete. It can aid in L2 acquisition if an L2 sound is both unmarked and dissimilar, or it can be harder, if an L2 sound is marked and has a similar counterpart in the L1 (Major & Kim, 1996:158). Therefore, based on the hypotheses p resented above, Major & Kim (1996) develop the Similarity Differential Rate Hypothesis (SDRH), which not only takes into account similarity and markedness, but also considers the rate at which L2 learners acquire sounds without making claims about ultimate competence. This hypothesis states that an L2 phone that is dissimilar to an L1 phone is acquired faster than an L2 phone that is similar to an L1 phone. Markedness is a mediating factor between similar and dissimilar phenomena. The rate of acquisition de creases as markedness increases given a similarity differential (Major & Kim, 1996). In order to test this hypothesis, along with the hypotheses of previous models, s the Moreover, fricatives. Therefore, it is expected that /z/ because s imilar sounds are more difficult to acquire than dissimilar sounds and m arked sounds are more difficult to acquire than unmarked sounds Also, it is expected that L2 acquisition of /z/ and that
84 L2 experience will enable Korean speakers to produce /z/ more authentically (Major & Kim, 1996:159 162). The analysis includes both experienced and inexperienced Korean learners of English and a group of native English speakers. Occurrence through a reading task of carrier phrases with words containing the sounds. The results reveal that both groups of Korean speakers had higher competence with the similar sound than with the dissimilar sound, which disconfirms hypothesis 1. This also disconfirms hypothesis 2 which states that advanced learners would have higher competence with dissimilar phenomena than with similar phenomena, and that beginning learners would have higher competence with similar phenomena than w ith dissimilar phenomena. The results, however, confirm the SDRH in that the differences between production of the similar sound and the dissimilar sound are greater for the advanced learners than for the beginning learners (Major & Kim, 1996:170). Therefo re from this analysis, we see that when complemented with rate of acquisition, markedness and similarity can be used to explain acquisition of L2 phonology without making claims of ultimate competence and allowing for variability in the acquisition of diff erent sounds. We notice that in order to test these hypotheses of similarity and markedness one must analyze pairs of sounds to describe the phonology of bilingual speakers. However, the present analysis only focuses on one sound that is similar between S panish and English and is unmarked, the voiceless dental plosive. A foreign accent, however, can be detected in the production of this sound when analyzing VOT duration. The most adequate model to incorporate in the analysis of the voiceless dental, theref ore, would
85 of acquisition plays a role in how bilinguals produce sounds that are similar in their two languages. Early bilinguals tend to maintain similar sounds separate according to languages based on the fact that the L2 phone is established independently of the L1 that is not fully developed at an early age. Late bilinguals tend to show convergence in their phonetic system because an equivalence classification blocks th e L2 sound category from being established. Late bilinguals seem to overlook important sensory information that distinguishes the L2 from the L1 (Flege, 1987, Flege, 1995a). Therefore, this study will examine the speech of early bilinguals of Spanish Creol e English from a variationist perspective to see if the Speech Learning Model can be further supported. perspective can have important implications for usage based models of langu age and phonology. A ccording to Flege (1995 a ), bilinguals can produce compromise VOT values That is, they show intermediate values that are the same in either language and different from prototypical ranges It is proposed here that this is a type of exte nsion which was discussed as a process with in the usage based approach. When interacting in their speech community, Spanish Creole English bilinguals could encounter a new token of /t/ that does not fall within the range for a given language, but rather in between the two. If bilingual speakers can accept this token it will be mapped near a /t/ with the most similar value. This would be the creation of a new cluster of compromise VOT values, which could eventually become a new sound category for bilingual s
86 could s trengthen with every compromise VOT that they encounter. An extreme case could be if speakers do not interact with any monolinguals. According to the exemplar model, the strongest clus ters would be those made up of intermediate VOT duration, and the monolingual like VOT would weaken and become less accessible for production. This process could lead to language change in this particular speech community as speakers participate in a feedb ack loop ( Pierrehumbert 2001). We see here that the usage based model of language, and particularly phonology and phonetics, can be a useful model in explaining language contact, variation, and change among bilingual speakers in Panama. This model fits we ll with the Speech Learning Model and the variationist framework adopted in this analysis since it assumes that probabilistic in nature. We see how this is operationalized in Chapter 3.
87 CHAPTER 3 METHODOLOGY 3.1 Language Use and the Vernacular L anguage use is the foundation for usage based models of language (Langacker, 2000:9) and for the sociolinguistic framework (Labov, 1972:184). Thus, the present study is based on so ciolinguistic interviews which are designed to capture the vernacular speec h or perhaps more appropriate, real language use of the participants. Labov (1972:208) describes vernacular speech as the style in which the minimum attention is giv en to the mon According to Labov (1972), wh en speaker s access their vernacular, more systematic speech emerges and the course of linguistic evolution is more apparent. However, some have suggested that the idea of the vernacular is an abstract object just as the standard language (Milroy & Gordon, 2003:50). The objective in sociolinguistic studies should not be to capture solely the (Sankoff, 1980:54) because characterist ics from both the vernacular and the standard language can appear in the linguistic production of a community (Milroy & Gordon, 2003). The ability to employ the vernacular and the standard language implies that speakers may not necessarily stop monitoring their speech, but rather actively manipulate their language and perform by accessing a wide range of styles according to their environment (Milroy, 1987). This, therefore, brings up another issue in language variation and change studies, the radox which refers to the fact that linguists attempt to observe how a speaker uses language when not being observed. This is a challenge for accessing spontaneous speech because the environment in
88 which speakers are being recorded causes them to shift ou t of their vernaculars. However, keeping in mind that the vernacular is not the only goal for linguistic analysis, there can be methods to overcome the paradox (Milroy 1987:59 60), one of which is the A friend of a friend solve problems through the help of friends, or even friends of friends due to temporary alliances (Boissevain, 1974:2; Milroy & Gordon, 2003 :75) Therefore, one can introduce th emselves by mentioning a common acquaintance in order to enter the community and be treated as an insider. This type of ap proac h was used to acquire participants during the fieldwork for the present analysis. All of the participants, in one way or another, were found through connections with family members or close friends. This first allowed for friendly interaction with the participant, which in turn made the environment a little more relaxed in which to conduct the interview. S everal interview modules de scribed by Labov (1966) were adopted, such as the American Language Survey Also, some questions were based on historical and current events particular to Panama. For example, several weeks before I carried out interviews in Bastimentos Island, a yo ung man had passed away in a boating accident. Since this event was so fresh in the minds of the residents, this particular topic came up in almost all of the interviews. This was also a good opportunity to bring up daily activities on the island such as boating and fishing. Such t opics aided in accessing the everyday speech of the informan ts and allowed for a longer interview, an additional requisite for a sociolinguistic study (Milroy, 1987:39). All inter views lasted at least 1.5 hours, which
89 was necessary not only for capturing phonetic patterns of everyday speech, but also for gaining an appropriate amount of data for a statistical analysis. The equipment used for the data collection included a Sony Digital Voice Recorder ICD SX7 50 and a Shure Microphone SM10A which were selected based on their high quality recordings. This is import ant in an acoustic analysis because spectrographic programs are sensitive to noise. Besides, since the goal was to capture natural speech s uch as their home s or even in a car, the best option was to use a head mounted microphone. Placement of the microphone in outh was also controlled as much as possible. 3.2 Participants This analysis utilized data from 15 spe akers living in different areas of Panama. There are 5 monolingual speakers of Panamanian Spanish from the capital of Panama. They live in areas such as Las Acacias, Las Cumbres, Pedregal and Tocumen. Additionally, 5 monolingual speakers of Bastimentos Cr eole English are included. These speakers live in the province of Bocas del Toro on the Island of Basitmentos. Lastly, there are 5 bilingual speakers of Panamanian Spanish and Panamanian Creole English spoken in Panama City 1 who live in areas such as Rio Abajo, Parque Lefebre, La Boca, Tocumen, and Chorrera. This final group constitutes two datasets in this analysis: bilingual Spanish and bilingual Creole English. Table 3 1 provides a list of the participants of the present analysis. 1 Although there was an attempt to only include bilingual speakers of the Creole English variety spoken in Panama Ci ty, one of the participants was originally from the city of Coln. Exceptional linguistic behavior from this speaker was not anticipated, however, because he had lived in Panama City for over 20 years.
90 Poplack & Levey (2010 ) state that monolingual varieties are necessary in the comparative method in order to determine whether or not change is occurring in a language contact situation. However, ideally, these monolinguals should be from a local region so that change is substa ntiated through the comparison of local norms (Torres Cacoullos & Aaron, 2003). Table 3 1. Participants Pseudonym sex place of origin Monolingual Spanish speakers Gloria F Tocumen Jacinta F Pedregal Lalo M Las Cumbres Laura F Las Acacias R icardo M Panam Centro Monolingual Creole English speakers Debby F Bastimentos Herlin F Bastimentos Janaira F Bastimentos Merilyn F Bastimentos Nadia F Bastimentos Spanish Creole English bilinguals Alberto M Rio Abajo Alicia F Par que Lefev re Dora F Juan Diaz Rogelio M Juan Diaz Samuel M La Boca This is slightly problematic for the present analysis because it is dif ficult to find true monolingual Creole English speakers in Panama since in certain domains one must be able to communicate in Spanish Although we categorize the Creole English speakers of Bastimentos as monolinguals, some of the residents in this area may have some level of Spanish. Therefore, for purposes of this study, we consider them dominant speakers of Creol e English. Figure 3 1 illustrates a monolingual bilingual continuum on which the participants of this analysis have been placed.
91 Figure 3 1. Monolingual Bilingual Continuum 3.3 Research Questions Based on the discussion of the previous literature, we p ose three research questions: 1) what is the distribution of mean VOT duration of the voiceless dental /t/ in the speech of Spanish and Creole English monolingual speakers and Spanish Creole English bilingual speakers, 2) what are the factors that signific antly condition VOT duration in monolingual Spanish, monolingual Creole English, bilingual Spanish and bilingual Creole English, and 3) based on the comparisons of factors contributing to VOT duration, is there evidence of change in bilingual speech, and i f so, what is the source of the change? 3.4 Hypotheses Hypothesis 1 is concerned with the overall mean VOT durations in the monolingual and bilingual speakers in this analysis. Based on previous studies that analyzed VOTs in varieties of Spanish and Engli sh, I hypothesize for Research Question 1 that monolingual speakers of Spanish will exhibit the shortest mean VOT for the voiceless dental plosive. Also, I expect monolingual speakers of Creole English to produce the voiceless dental plosive with a longer mean VOT than that of monolingual speakers of Spanish (Lisker & Abramson, 1964). As for bilinguals, I hypothesize that they will exhibit
92 mean VOTs in Spanish that fall between the prototypical values of the monolingual varieties of Spanish and Creole Engli sh. In Creole English, bilinguals will produce mean VOTs that are longer than the monolingual variety of Creole English (Caramazza et al., 1973; Flege, 1995a; Thornburg & Ryalls, 1998; Macleod & Stoel Gammon, 2008). Hypothesis 2 is concerned with the sta tistical analyses of the factors that condition VOT duration. I hypothesize for Research Question 2 that factors having to do with the phonetic environment in which /t/ is found would have a significant effect on VOT duration in all four varieties. I also hypothesize that factors related to prosody, such as syllable stress and rate of speech, will significantly affect VOT duration (Lisker & Abramson, 1967; Yoshioka, Lfqvist, & Hirose, 1981; Castaeda, 1986; Schmidt & Flege, 1996 ; Roldan & Soto Barba, 1997; Rosner et al., 2000; Munday, 2001; Morris et al., 2007; Yavas & Wildermuth, 2006). Hypothesis 3 is concerned with the comparisons of the statistical results across the four varieties (Research Question 3). These comparisons provide evidence of language change with regard to the phonetic system of bilinguals. When comparing bilingual Spanish to the monolingual varieties of Spanish and Creole English, I expect bilinguals to perform more similarly to monolingual speakers of Spanish. When comparing bilingual Creole English to the monolingual varieties, I expect bilinguals to assimilate Creole English to Spanish and not to Creole English. These hypotheses are based on the fact that Spanish is the dominant language of this speech community (see Thomason, 2001). Bilinguals would be more prone to having more Spanish like production in both languages, thus giving evidence of language change due to contact. Sections 3.5 3.8 explain the procedures carried out in order to test these hypotheses.
93 3.5 Instrumental Ana lysis As was mentioned in Section 2.3, VOT is considered to be a continuous variable that is context dependent and not easily perceptible. With this type of data, the use of spectrographic analyses is important in order to record numeric data (Thomas, 2002 ; Gordon, 2007). The instrumental analysis in the present study was carried out using Praat (Boersma & Weenink, 2006), a free software program for analyzing speech sounds. Sections 3.5.1 3.5.3 discuss the step performed so that the instrumental analysis was valid and reliable. 3.5.1 Data Extraction All words in a 10 minute segment containing a voiceless dental were extracted from the interviews conducted with monolinguals and bilinguals of Spanish and Creole English. Tokens were extracted anywhere after the 10 th minute of the interview. 2 The total number of tokens extracted from the 10 minute segments of all 15 participants is 2,128. There were three positions relevant to this analysis in which /t/ could have appeared: absolute word initial, prevocalic ( #__V); post consonantal, pre vocalic (C__V); and intervocalic (V__V). In the case that a word contained more than one /t/ (e.g. ToTalmenTe examples of the relevant positions with a phone tic transcription in parentheses of the syllable containing the voiceless dental: 1) Tea is good ([ti]) 2) H e is seve ntee n ([nti]) 3) I am L ati no ([ati]) 2 Each interview varied according to the context in w hich it was carried out. Therefore, it was not possible to always begin and end the token extraction at the same minute for all participants, as they spoke Spanish and Creole English at different times
94 3.5.2 VOT Measurements in Praat The acoustic measurements of the voiceless dentals were performed based on 1) the description of a syllable containing a plosive followed by a vowel (e.g. Docherty, 1992) and, 2) the explanations by other researchers that specialize in the acoustic analysis of consonant stops (e.g. Ladefoged, 2003). As I discussed in the Introductio n, the stop closure consists of three stages. The first stage, complete closure leading to the build up of pressure, is noticeable in a spectrograph by a flat line. The second stage, rapid release of the air build up, appears as a short periodic pulse in t he flat line. This stage is also referred to as the release burst. Figure 3 2. Spectrogra ph and waveform land in monolingual Spanish
95 Figure 3 3. Figure 3 4 all in bilingual Spanish
96 The third stage, complete aperture of the articulators, during which aspiration can be heard, is noticeable by noise in the frequency range of the high formants. Later, regularly spaced vertical striations indicate the onset of the following vowel, that is, voicing. Figures 3 2 3 5 exemplify these stages of /t/ in both Spanish and Creole English for all three relevant positions. Figure 3 5. Spectrograph bilingual Creole English Once these characteristics were detected for each voiceless dental in Praat, the VOT interval was highlighted and a measurement in seconds was provided by the program. These seconds were later converted to milliseconds. In the cas e that the characteristics are not easily detected in the spectrogram due to background noise, that particular /t/ was excluded from the analysis.
97 3.5.3 Rater Validation In order to provide valid and reliable results for the VOT durations measured by the investigator, a rater validation was carried out by another trained phonetician who is an L2 speaker of Spanish. Ten percent of each dataset was remeasured, and subsequently, a Pearson product moment correlation test was performed to determine a significa nt association between the first and second sets of measurements. This correlation test provides a negative coefficient if the association between the two measurements is negative; that is, there is no association. A positive coefficient is provided if the association is positive. Coefficients that are closer to 1 or +1 are considered strong associations. For the present validation, positive coefficients closer to +1 are desired. Table 3 2. Rate validation correlation results Dataset t (r) Monolingual Spanish 0.860957 Monolingual Creole English 0.9795382 Bilingual Spanish 0.9537144 Bilingual Creole English 0.984428 p <0.05 As seen in Table 3 2, the results of the test indicate that the two measurements obtained for each dataset have a significant strong association. The p value is less than 0.05, and each dataset has a positive coefficient close to +1 (monolingual Spanish = 0.861 3 monolingual Creole English = 0.98, bilingual Spanish = 0.954, bilingual Creole English = 0.984). On ce the instrumental analysis was completed, various steps developed within the variationist methodology were performed, which is explained in Sections 3.6 3.8. 3 Although this coefficient is not as high as the re st, it still indicates a positive association between the two sets of measurements.
98 3.6 Variationist Approach The variationist method is based on the claim that language featur es inherent variability (Weinreich et al., 1968:167; Labov, 1969:728). Due to contact between dialects and languages, speakers are provided with an inventory of alternatives that they access during discourse to express a certain idea. This means that there is a coexistence of various linguistic systems (Weinreich et al., 1968:159) in a community that a theory of language change must account for without isolating a subsystem. Several alternatives that usually have the same referential values or grammatical f unction can arise (Weinreich et al, 1968; Poplack & Tagliamonte, 2001:88). The choices between alternatives are not random, but rather constrained by factors that guide production. Weinreich et al. (1968:121) refer to this as structured heterogeneity The variationist framework incorporates the concept of the linguistic variable (Weinreich et al., 1968:167; Labov, 1972), which consists of a group of variants that alternate in speech based on meaning and social and stylistic functions (Weinreich et al., 1968 : 159 ; Poplack & Tagliamonte, 2001:88). Therefore, language is not static, but rather changes based on social structures. In language contact, the variants are forms that have gone through some type of change and that are in competition with each other. In order to analyze these contact variants through a variationist method, it is necessary to choose the appropriate community based on the intensity of contact, the length of contact, the status of the language in the community, and the size of the speaker p opulation (Poplack & Levey, 2010:399). These are the aforementioned social factors discussed by Thomason (2001). Next, we must take bilingual and monolingual samples of the community. In doing so, an attempt should be made to capture the vernacular, or eve ryday speech, of the speakers (Poplack & Levey, 2010:399). It is also
99 essential to extract all instantiations of the variants of interest, accounting for the context where they occur and where they could occur. This is referred to as the principle of accou ntability (Labov, 1972:72). This principle ensures that we deal with a full range of variability that is present in the language (Bayley, 2002:123). When analyzing phonetic variants of binary variables, such as h dropping ([h] vs. [ ]), this is simple, sin ce one can easily identify occurrences and zero realizations of [h]. This is also the case in a situation such as s lenition in Spanish, a variable that consists of three variants ([s], [h], and [ ]). However, in this analysis, fully accounting for VOT dur ation is different. We have adhered to the principle of accountability by randomly selecting a 10 minute segment after the 10 th minute of each interview, so as to not be bias to areas where longer or shorter VOT durations occur. Then, the voiceless dental plosive is extracted when occurring in phonetic environments where the literature has discussed that for, factors must then be chosen in order to operationalize hypotheses made about the variants, in this case, mean VOT duration. Both linguistic and extraling uistic factors are appropriate for testing (Poplack & Tagliamonte, 2001:91). The effect of the conditioning factors is then statistically analyzed. This is done through using a variable rule analysis, which is a multiple regression procedure that determin es which factors contribute statistically to the choice of one form over another (Poplack & Tagliamonte, 2001:92). This type of analysis, which is normally carried out in GoldVarb X (Sankoff, Tagliamonte & Smith, 2005) presents three bases for interpretin g
100 the results: statistical significance of effect, the magnitude of effect, and the hierarchy of constraints 4 (Poplack & Tagliamonte, 2001 :92 93 ). The statistical significance level, which is set at .05, tells us that there is a 5% chance that the results are random; the lower the percentage, the better. The magnitude of effect, which is determined by the range, shows us which factor group has a greater relative effect on the choice of the variant. The hierarchy of constraints is a more detailed view of ea ch factor being tested and is what we take as the underlying system of the language. It tells us the weight of each factor within a group. The greater the weight, the more likely it is that this form is occurring with a particular factor. The variable rule analysis orders these factors according to their strength, which enables us to make comparisons across forms, and in the case of contact induced change, across variety types (Poplack, 1993; Poplack & Tagliamonte, 2001; Poplack & Levey, 2010). Instead of employing GoldVarb X to perform the variable rule analysis, I chose to use Rbrul (Johnson, 2009), a package that can be loaded in the free statistical environment, R According to Johnson (2009:362), this package is comparable to een designed to replicate the functionalities and factor weight analysis is that Rbrul makes it possible to carry out linear regression with continuous responses, such as VO T. The program estimates the effects of factors on mean values (Johnson, 2009), which is not possible in GoldVarb, as values cannot be coded as numeric. The only option would be to group VOT values into categories, which would most likely result in arbitra ry groupings. 4
101 The linear regression carried out in Rbrul reports results in coefficients as opposed to the factor weights seen in GoldVarb. A coefficient of 0 presented in Rbrul as a neutral effect is the equivalent to a factor weight of .5 in GoldVarb. A positive coefficient shows a favoring effect, and a negative one, a disfavoring effect. Table 3 3, taken from Johnson (2009), shows the factor weight coefficient correspondence. Table 3 3. Some factor weights (probabilities) and the corresponding log odds (Johnson, 2009) Factor weight (probability) log odds .000 .100 2.197 .200 1.386 .300 0.847 .400 0.405 .500 0 .600 +0.405 .700 +0.847 .800 +1.386 .900 +2.197 1.000 + Note: Reprinted with permission from Johnson, Daniel Ezra. (2009). Getting off the GoldVarb standard: Introducing Rbrul for mixed effects variable rule analysis (Page 361, Figure 1). Language and Linguistics Compass 3:359 383. For the purposes of this study, the order of coefficients are taken to be the constraint hierarchy usually reported in GoldVarb, as it is the order of effect of the conditioning factors that contribute to VOT duration. To my knowledge this is the first analysis to replicate the comparative variationist method in a language contact study using Rbrul. 3.7 Linguistic Factors As discussed above, in order to pinpoi nt the variation patterns of linguistic forms, it is crucial that one determines the linguistic factors that condition the use of these forms. These factors are normally determined through a detailed investigation of previous studies dealing with the pheno menon. With regard to the voiceless dental plosive, the
102 analyses of these factors allow us to observe which constraints favor a longer or shorter VOT. Table 3 4. Linguistic factors coded Factor groups Preceding segment grater some coconut a nd boil ta [PART 004]) voiceless fricative (e.g. stu dies [PART 009]; but we had acces s to medical.. [PART 001]) vowel (e.g. o to [PART 002]; I'm a l ati na [PART 007]) other con sonants (e.g. He's seve nt een, i think, sixteen [PART 002]) Following vowel a e o u i Position in word word initial (e.g. tea [PART 001]) word internal (e.g. sometime [PART 009]) Syllable stress talking [PART 007]) certain [PART 004]) Rate of speech syllables per second Lexical item For example, Turkey besides your turkey and your ham and your Retire he retire here and then he went ART 004] Word class adjective (e.g. terrible [PART 007]) adverb (e.g. sometimes [PART 001]) infinitival marker (e.g. to live [PART 002]) noun (e.g. university [PART 004]) preposition (e.g. to [PART 001] verb (e.g. notice [PART 007]) Word frequency per 10,000 words After excluding examples of /t/ that were not easily detected in Praat (Section 3.5.2), all remaining 2,128 tokens were coded (in Excel) according to the factors hypothesized
103 to constrain VOT duration. The factors included in the analys es were preceding segment, position of /t/ in the word, following vowel, syllable stress, rate of speech, lexical type, word class, and word frequency (Table 3 4 ). These factors were chosen based on previous studies that analyzed the production of consonan t stops in monolingual speakers of Spanish and English, and bilinguals speakers of these languages (Lisker & Abramson, 1967; Yoshioka, Lfqvist, & Hirose, 1981; Castaeda, 1986; Schmidt & Flege, 1996 ; Roldan & Soto Barba, 1997; Rosner et al., 2000; Munday, 2001; Morris et al., 2007; Yavas & Wildermuth, 2006). Another source for some of these linguistic constraints was literature dealing with usage based models of language and phonology ( Bybee, 2000, 2001 ; Pierrehumbert, 2001 ; Brown, 2009 ). 3.7.1 Preceding Segment The first factor, preceding segment, was posited to have an effect on VOT duration based on previous analyses that have observed effects of the preceding segment on a consonant stop. With regard to varieties of English, very few studies have focuse d on the effects of preceding segment. However, it has been discussed that in English, when clusters of voiceless fricative + voiceless plosive are produced, there is generally a one peak pattern in the spectrogram because the plosive is unaspirated (Yoshi oka, Lfqvist, & Hirose, 1981). The plosive is aspirated less in this context due to laryngeal adjustments. Surprisingly, Munday (2001) discovered that when the preceding segment was a fricative VOT duration was significantly longer in different speech sty les of Spanish These results, thus, contradict claims of physiological reasons, such as laryngeal adjustments. Herrera (1997) and Dniz (2005) found that VOT values were greater when the voiceless dental was preceded by a consonant than when it was preced ed by a vowel in Canarian Spanish. However, Roldan & Soto Barba (1997) found
104 that the preceding segment had no effect on the VOT of voiceless stops when preceded by zero or a nasal consonant in Valdivian Spanish In the present study, it is hypothesized th at VOT duration would be shorter when preceded by voiceless fricatives voiceless fricative would favor a longer VOT duration in Spanish. 3.7.2 Following Vowel Following vowel height is heavily considered in previous studies that analyze plosives. Researchers have discovered that VOT duration increase as the following vowel rises. This has been evidenced in monolingual speakers of Spanish and English (Port & Rotunno, 1979; Weis mer, 1979; Castaeda, 1986; Rosner et al., 2000; Munday, 2001; Dniz, 2005; Morris et al., 2007) and bilingual speakers of these languages (Thornburg & Ryalls, 1998; Yavas & Wildermuth, 2006) In the present study, I hypothesized that vowel height would ha ve the same effect on VOT duration with regard to the voiceless dental in all language modes, and that the effect would be significant when compared with other factor groups. 3.7.3 Position in Word Position of a consonant stop in a word, to my knowledge, h as not been analyzed in most studies concerning VOT. The little attention paid to this factor is because most analyses have been concerned with only word initial stops in a laboratory setting. The present study was concerned with voiceless dental stops in both word initial and word internal position when speech was spontaneous. It was posited that the position of the /t/ would have an effect on VOT duration in all language modes.
105 3.7.4 Syllable Stress Very few studies have focused on syllable stress as a c onditioning factor on VOT duration. Castaeda (1986) and Dniz (2005) discovered that when voiceless plosives appeared in stressed syllables, their VOT values tended to be shorter in Peninsular Spanish and Canarian Spanish, respectively. Lisker & Abramson (1967 ) found that in monolingual varieties of English VOT duration was longer in tonic syllables, and shorter in atonic ones. Based on these results, I posited that in the monolingual variety of Spanish, atonic syllables would favor longer VOT duration, an d tonic syllables would favor shorter VOT duration. As for the monolingual variety of Creole English, I hypothesized that there would be an opposite effect; that is, atonic syllables would favor shorter VOT duration, and tonic syllables would favor longer VOT duration. 3.7.5 Rate of Speech Schmidt & Flege (1996) discovered that English monolinguals had significant differences between all three speaking rates (fast, normal, slow) and early bilinguals of Spanish and English only differed between fast and norm al/slow. The former exhibited shorter VOT values as speech rate increased in both sentence initial and sentence medial words. Spanish monolinguals showed an opposite effect on /p/ in sentence medial words only; t hat is, VOT values became shorter as speech rate decreased Kessinger & Blumstein (1998) also discovers that as speech rate increased, VOT values became shorter. It was posited that both monolingual and bilinguals would show the same results as monolinguals and bilinguals in Schmidt & Flege (1996) and Kessinger & Blumstein (1998) As rate of speech increases, VOT duration of the voiceless dental would become shorter.
106 3.7.6 Lexical Item, Word Class, Word Frequency Lexical item word class, and word frequency factors was also included in this study. Lexical item refers to the possible effect that an individual word has on the length of VOT. Word class is concerned with the category of the word, such as noun or preposition. Word frequency refers to the number of times the word occurred in the datasets These factors were chosen based on the literature regarding usage based models of language and phonology ( Bybee, 2001; 2006; Langacker, 2000 ; Pierrehumbert, 2001; Bybee & Torres Cacoullos, 2008; Brown 2009) They should sh ed light on the effects of freq uency. Bybee (2001) states that one of the effects of frequency is reduction. Pagliuca & Mowrey (1987) also shares this idea of temporal reduction in highly frequent words. When words are repeated, their length or segments tend to reduce. This reduction ca n be substantive where there is overlapping of sounds, such as the nasalization of vowels in vowel + nasal contexts. Reduction can also be temporal where the length becomes shorter ( Pagliuca et al. 1987 :460 462). A ccording to Bybee & Torres Cacoullos (200 8:400 402), high frequency words also tend to show more variation, and this variation is directly represented in an exemplar model because individual words have their own exemplar clusters. Also with frequency, exemplars can be strengthen, which can lead to entrenchment of a sound (Bybee, 2001; Pierrehumbert, 2 001; Bybee & Torres Cacoullos, 2008; Brown, 2009 ), thereby allowing a sound to be diffused throughout the lexicon. The phonological environment of the sound plays a role in this diffusion as well. I f a word appears more frequently in a conditioning environment for change, it leads to a more rapid diffusion of that change for that particular word. Eventually, that sound change will be apparent in the word even when it does not show up in the condition ing
107 environment (Bybee, 2001; Bybee & Torres Cacoullos, 2008). Based on these claims, it was expected that VOT would be shorter in highly frequent words in all language modes. Likewise, functional words (e.g. prepositions) would have voiceless dentals with a shorter VOT duration because they are more frequent than non functional words (e.g. nouns). Frequency was calculated (in Excel) by counting the instances of each word out of the total number of words in the corpora (Spanish: 13,681; English: 22,050), a nd then normalizing by per 10,000 words: 4) 22,050x = 140,000 x = 6.35 After all the tokens were coded according to the factors discussed above, the statistical analyses (Section 3.6) we re performed in order to determine the significance and the constraint hierarchies in each language mode. This factor ranking was considered to be the underlying system of the language in terms of VOT duration. When the factor groups were not determined to be significant, the rankings were taken as tendencies of the language. Once the results were organized for each language mode, comparisons of the constraint hierarchies across the four varieties were performed. Section 3.8 explains the method of compariso n.
108 3.8 Comparative Variationist Method The primary goals of the application of the comparative variationist method in a language contact study are: 1) to determine whether change is occurring in a bilingual speech community, and 2) to ascertain the putati ve source of the change (Poplack & Levey, 2010). These goals are accomplished through a series of comparisons of factor effect rankings between non contact and contact varieties. Therefore, in the present analysis, I took the results of the variable rule a nalyses for each language mode and compared them side by side. If a contact variety (e.g. bilingual Spanish) showed the same order of effect of conditioning factors as its non contact counterpart (e.g. monolingual Spanish), and not that of the source varie ty (e.g. monolingual Creole English), it was determined that no change had occurred in the Spanish production of bilinguals. If a contact variety did not show the same order of effect of conditioning factors of either of the other varieties, then it was de termined that an internal change was occurring in this variety. If a contact variety did not feature the same order of effect as the monolingual counterpart, but did have the same order as the source variety, then it was determined that a contact induced c hange was occurring in the bilingual variety. Another important comparison in this study is between contact varieties. This comparison allows us to see whether bilinguals have a merged phonetic system or two separate systems for each language with regard t o VOT duration. If the bilingual varieties shared the same order of effect, then it was determined that bilinguals had a merged phonetic system. However, if the order of effect was different for each variety, then it was determined that bilinguals had two separate phonetic systems for Spanish and Creole English. Table 3 5 gives an overview of the comparisons carried out in this analysis.
109 Table 3 5. Comparison of four language modes Comparisons Monolingual Spanish Bilingual Spanish Bilingual Creole Eng lish Monolingual Creole English Bilingual Creole English Bilingual Spanish Bilingual Spanish Bilingual Creole English 3.9 Summary The methodology described in this chapter was designed in order to discover possible outcomes due to contact bet ween Spanish and Creole English in the bilingual West Indian community of Panama. Certain hypotheses have been made based on previous studies concerning language contact, variation and change, and voice onset time in monolinguals and bilinguals. This is on e of the first studies to examine voice onset time using a variationist approach. In Chapter 4, we explain the results of the analyses as they pertain to the predictions presented in this chapter.
110 CHAPTER 4 A VARIATIONIST ACCOUNT OF MEAN VOT OF /T/ I N A C ONTACT SITUATION 4.1 Linear Regression Analyses Traditionally, literature that has analyzed the effect of certain factors on VOT values has presented means and has used t tests in order to show significant results (e.g. Schmidt & Flege (2006)). The data an alyzed are usually collected in a laboratory where participants read a list of words containing the target sound. As mentioned, the present study offers a variationist account of the voiceless dental when it is produced in naturally occurring speech. Follo wing the standard methods used in variationist work (see Section 3.6), a linear regression model with mixed effects is incorporated in order to determine which factors significantly affect mean VOT. These results, which consist of constraint hierarchies of the conditioning factors, are considered to be the underlying system of VOT duration of the voiceless dental plosive. In this chapter, we present the findings for each language mode (Sections 4.1.1 4.1.5). Subsequently, the results from the comparisons among the varieties are presented (Section 4.2). 4.1.1 Distribution of Mean VOT Duration Table 4 1 presents the overall mean VOT duration for all four varieties, which address Research Question 1 (Section 3.3) and Hypothesis 1 (Section 3.4). It can be see n that the mean VOT value for the voiceless dental in monolingual Spanish is 15.24 msec. This length is greater than that of the voiceless dental in Puerto Rican Spanish analyzed in Lisker & Abramson (1964) and in Argentine Spanish analyzed in Borzone de M anrique (1980). The mean VOT values for these varieties were 9 msec. and 15 msec., respectively. Conversely, the mean VOT of the voiceless dental in the present study is shorter when compared to Valdivian Spanish (16.4 msec.), Canarian Spanish
111 (16.6 msec.) and Castilian Spanish (19.75 msec.) (Poch Olive, 1985; Roldan & Soto Barba, 1997; Deniz, 2005). In the monolingual Creole English variety, the mean VOT for the voiceless dental is 28.43 msec, which is much smaller than the mean VOT observed in American E nglish analyzed in Lisker & Abramson (1964), which was 70 msec. The voiceless dental in bilingual Spanish exhibits a mean VOT of 21.15 msec. Lastly, in the bilingual Creole English variety, 34.20 msec. is the mean VOT for the voiceless dental. Table 4 1. M ean VOT standard deviation, and range for the monolingual and bilingual vari eties 1 Group Mean VOT (/t/) in msec. Standard deviation Range Monolingual Spanish 15.24 7.58 0 52 Monolingual Creole English 28.43 13.48 0 82 Bilingual Spanish 21.15 8.95 2 8 7 Bilingual Creole English 34.20 18.33 0 168 p <0.05 These results confirm the hypothesis that monolingual Spanish speakers would have the shortest VOT duration for the voiceless dental plosive. Also, we see that monolingual speakers of Creole Englis h produce this sound with a longer mean VOT than that of the monolingual speakers Spanish. For bilinguals, we notice that their Spanish VOT value falls between the monolingual values, and their Creole English value is greater than that of the monolingual C reole English speakers. In Sections 4.1.2 4.1.5, I discuss the underlying patterns made evident through the linear regression analyses. The results presented address Research Question 2 (Section 3.3) and Hypothesis 2 (Section 3.4), which address the con ditioning factors that contribute to VOT duration in the monolingual and bilingual varieties. 1 See Appendix A for individual speaker mean VOT duration.
112 4.1.2 Conditioning Factors in Monolingual Spanish Table 4 2 shows the significant and non significant factors that condition mean VOT of the voiceless dental in the monolingual Spanish variety. Table 4 2. Factors conditioning mean VOT in M onolingual Spanish Factor Coef N Vowel i 2.958 61 u 1.104 44 o 0.164 114 e 0.846 118 a 3.052 162 Preceding segment Zero 1.058 36 Voiceless fricative 0.819 73 Consonant 0.762 169 Vowel 1.115 221 Word class Preposition [1.174 ] 5 Verb [0.374 ] 172 Adverb [0.255 ] 53 Pronoun [0.154 ] 40 Noun [ 0.086 ] 126 Adjective [ 1.870 ] 103 Position Word internal [0.356 ] 321 Word initial [ 0.356 ] 178 Stress No [0.455 ] 314 Yes [ 0.455 ] 185 Rate of speech +1 [ 0.112 ] -Frequency +1 [ 0.037] -p <0.05, Log likelihood = 159.743 interce pt = 17.757 grand mean = 15.253. *Coefficients in brackets [ ] indicate that factors were not sig nificant in the statistical analysis
113 The two factors selected as significant in this variety are vowel height and preceding segment. We see in monolingual Spanish that VOT is longer with high vowels, which is indicated by positive coefficients (i = 2.95 8, u = 1.104), and shorter with low and mid vowels, indicated by negative coefficients (o = 0.164, e = 0.846, a = 3.052). These results are in line with previous analyses of the effect of vowel height on VOT duration in monolingual Spanish varieties (Ca staeda, 1986; Flege, 1995a, 1995b; Rosner et al., 2000; Munday, 2001; Deniz, 2005; Schmidt & Flege, 2006). Turning to preceding segment, it is observed that longer VOT duration is favored when the voiceless dental is phrase initial, that is, when nothing is before (1.058), or a voiceless fricative (0.819). Shorter VOT duration is favored with a preceding consonant ( 0.762) or vowel ( 1.115). We notice that VOT values are greater when the voiceless dental is preceded by a consonant than when it is preceded by a vowel. Although Roldan & Soto Barba (1997) did not discover this effect in Valdivian Spanish, Herrera (1997) and Dniz (2005) did in Canarian Spanish. The effect of preceding voiceless n Castilian Spanish; that is, voiceless fricatives favor a longer VOT duration. With regard to word class, although not significant, we see that the voiceless dentals in prepositions (1.174), verbs (0.568), and pronouns (0.259) have longer VOT duration, an d in adverbs ( 0.072), nouns ( 0.159), and adjectives ( 1.846), the VOT duration is shorter. We notice that there is no apparent functional/nonfunctional pattern in the effect of word class. In terms of position in the word, which is also non significant, we observe that when the voiceless dental is word internal, VOT tends to be longer (0.356) than when it is word initial ( 0.356).
114 Syllable stress also resulted as non significant in the linear regression; however, we observe a tendency for mean VOT to be longer in atonic syllables (0.455), and shorter in tonic syllables ( 0.455). These results correspond to those found in Castaeda (1986) and Dniz (2005). In terms of rate of speech, we notice that as speech rate increases by one unit, VOT duration becomes shorter ( 0.112). Schmidt & Flege (1996) discovered the same effect of speech rate in their analysis of the voiceless dental in monolingual Spanish speakers. Lastly, with regard to word frequency, which did not achieve significance, as frequency increases by one unit, VOT duration becomes shorter ( 0.037). 4.1.3 Conditioning Factors in Monolingual Creole English Table 4 3 presents the significant and non significant factors conditioning the mean VOT of the voiceless dental in the speech of monolingual Cr eole English speakers. The three factor groups that achieved significance were rate of speech, following vowel height, and preceding segment. With regard to rate of speech, it is seen that as speech rate increases by one unit, VOT duration becomes shorter ( 1.013). This effect was also found among monolingual speakers of English in Schmidt & Flege (1996). In terms of vowel height, we notice that longer VOT duration is favored with high vowels (i = 7.430, u = 6.124), and shorter VOT duration is favored with mid and low vowels (o = 2.071, e = 3.772, a = 7.711). These results parallel those seen in previous analyses of the effect of vowel height on VOT duration in monolingual varieties of Spanish and English (Port & Rotunno, 1979; Weismer, 1979; Castaeda, 1 986; Flege, 1995a; 1995b; Rosner et al., 2000; Munday, 2001; Deniz, 2005; Schmidt & Flege, 2006; Morris et al., 2007). Thus, with regard to vowel height, this particular variety of Creole English resembles other English varieties.
115 Table 4 3. Factors condit ioning mean VOT in Monolingual Creole English Factor Coef N Rate of speech +1 1.013 -Vowel i 7.430 46 u 6.124 185 o 2.071 38 e 3.772 177 a 7.711 142 Preceding segment Zero 1.527 8 Consonant 1.286 223 Vowel 0.491 199 Voicel ess fricative 3.304 158 Position Word initial [2.001] 329 Word internal [ 2.001] 259 Word class Pronoun [5.588] 45 Preposition [0.414] 48 Infinitival marker [ 0.110] 110 Adverb [ 0.281] 56 Noun [ 0.666] 178 Verb [ 1.129] 93 Adjectiv e [ 3.817] 58 Stress Yes [1.601] 197 No [ 1.601] 391 Frequency +1 [ 0.032] -p <0.05, Log likelihood = 225.891, intercept = 38.366, grand mean = 28.434 Turning to preceding segment, shorter VOT duration is favored when the voiceless de ntal is preceded by a voiceless fricative ( 3.304), as opposed to when it is phrase initial, or preceded by zero/pause (1.527), other consonants (1.286) or vowels (0.491). These results are in line with discussions concerning loss of aspiration in clusters of
116 voiceless fricative + voiceless plosive ( Yoshioka, Lfqvist, & Hirose, 1981 ). We also notice that other preceding consonants favor longer VOT duration and vowels favor shorter VOT duration. Although this has not been observed in varieties of English, H errera (1997) and Dniz (2005) found that VOT values are greater when the voiceless dental is preceded by a consonant than when preceded by a vowel in Canarian Spanish. Concerning position of the voiceless dental in the word, which was not significant, VO T duration is longer when the plosive is word initial (2.001), as opposed to word internal ( 2.001). Although not significant, we also notice certain tendencies with regard to word class. VOT values are longer in pronouns (5.588) and prepositions (0.414), and shorter in infinitival markers ( 0.110), adverbs ( 0.281), nouns ( 0.666), verbs ( 1.129), and adjectives ( 3.817). It appears that in this variety there is a tendency to have longer VOT duration in functional words and shorter VOT duration in lexical words. Since this factor group is not significant, these results could be due to frequency. Turning to syllable stress, which was also not significant, we notice that VOT values are longer in tonic syllables (1.601) and shorter in atonic syllables ( 1.601 ). Lisker & Abramson (1967) found similar results in monolingual varieties of English With regard to word frequency, which did not achieve significance, we notice that as frequency increases by one unit, VOT duration becomes shorter. The linear regressio n of mean VOT duration in monolingual Creole English reveals certain patterns in terms of several factor groups, and indicates general similarities with other varieties of English. In 4.1.4, I discuss the linear regression analyses of mean
117 VOT in the varie ties of Spanish and Creole English spoken by five bilinguals living in Panama City. 4.1.4 Conditioning Factors in Bilingual Spanish Table 4 4 presents the significant and non significant factors conditioning mean VOT of the voiceless dental in the bilingua l variety of Spanish. Two factor groups achieved significance in the linear regression, following vowel height and rate of speech. With regard to vowel height, it can be observed that longer VOT values are favored by high vowels (i = 4.527, u = 1.239), and shorter VOT values are favored by mid and low vowels (o = 1.119, e = 1.824, a = 2.841). This same effect of following vowel height has been noticed in the speech of monolingual and bilinguals (Castaeda, 1986; Flege, 1995a, 1995b; Thornburg & Ryalls, 1 998; Rosner et al., 2000; Munday, 2001; Dniz, 2005; Schmidt & Flege, 2006; Yavas & Wildermuth, 2006). In terms of rate of speech, as speech rate increases by one unit, shorter VOT values are favored in the bilingual variety of Spanish ( 0.372). Schmidt & Flege (1996) notice this effect in the speech of early bilinguals. As we turn to word class, which did not achieve significance, certain tendencies are observed with regard to VOT duration. We see that VOT values tend to be longer in pronouns (3.563), pre positions (1.475), and verbs (0.84 2), and shorter in adjectives ( 1.044 ), nouns ( 1. 510 ), and adverbs ( 3. 326 ). It appears that there could be effect of functional/lexical word effect; however, this factor is not significant. Therefore, it is not obvious w hat patterns this factor group reveals. Preceding segment also resulted as non significant in the linear regression; however, it can be observed that VOT duration seem to be longer when the voiceless
118 dental is phrase initial (0.855) or preceded by other c onsonants (0.116), and shorter when it is preceded by vowels ( 0.199) and voiceless fricatives ( 0.772). Table 4 4. Factors conditioning mean VOT in Bilingual Spanish Factor Coef N Vowel i 4.527 75 u 1.257 48 o 1.119 91 e 1.824 151 a 2.841 16 3 Rate of speech +1 0.372 -Word class Pronoun [3.563] 39 Preposition [1.475] 11 Verb [0.842] 147 Adjective [ 1.044] 88 Noun [ 1.510] 158 Adverb [ 3.326] 85 Preceding segment Zero [0.855] 27 Consonant [0.116] 200 Vowel [ 0 .199] 244 Voiceless fricative [ 0.772] 57 Position Word initial [0.06] 219 Word internal [ 0.06] 309 Stress No [0.527] 320 Yes [ 0.527] 208 Frequency +1 [ 0.018] -p <0.05, log likelihood = 180.235, intercept = 25.649, grand me an = 21.146 This effect of voiceless fricative is the opposite of what Munday (2001) found for Castilian Spanish, but follows the tendency found in English clusters of voiceless
119 fricative + voiceless plosive. The results for preceding segment also line u p with previous studies in that in monolingual varieties of Spanish, VOT values are greater when the voiceless dental is preceded by a consonant than when preceded by a vowel (Herrera, 1997; Deniz, 2005). In terms of position of the voiceless dental in the word, which is a non significant factor group, it can been seen that VOT duration tends to be longer when the sound is in word initial position (0.06), than when it is word internal ( 0.06). Syllable stress is also not significant in this analysis; ho wever, VOT values tend to be greater in atonic syllables (0.527), and smaller in tonic ones ( 0.527). Analyses of VOT duration in Spanish spoken by bilinguals have not taken syllable stress into account. However, Castaeda (1986) and Deniz (2005) found sim ilar results in monolingual Spanish speakers. Lastly, word frequency is also not significant; however there is a tendency for VOT duration to be shorter as frequency increases. The linear regression analysis of mean VOT duration in Spanish spoken by bilin gual West Indians reveals that this variety is generally similar to other varieties of Spanish analyzed in previous studies, with one exception (i.e. word position). In the 4.1.5, I explain the results of the linear regression analysis of mean VOT in the C reole English variety spoken by bilingual West Indians. 4.1.5 Conditioning Factors in Bilingual Creole English Table 4 5 presents the factors that condition the mean VOT duration of the bilingual variety of Creole English. Two factor groups were select ed as significant in this analysis, rate of speech and position. With regard to rate of speech, we observe that as speech rate increases by one unit, a shorter mean VOT is favored ( 1.56). This tendency has also been noted in the speech of early bilinguals (Schmidt & Flege, 1996).
120 Table 4 5. Factors conditioning mean VOT in Bilingual Creole English Factor Coef N Rate of speech +1 1.56 -Position Word initial 6.25 303 Word internal 6.25 210 Word class Adjective [7.947] 57 Prepositi on [0.906] 63 Infinitival marker [ 0.312] 74 Verb [ 1.156] 112 Pronoun [ 2.079] 17 Adverb [ 2.585] 34 Noun [ 2.720] 156 Vowel i [7.991] 31 u [0.646] 157 e [ 2.198] 172 o [ 3.163] 21 a [ 3.277] 132 Preceding segment Zero [1.642] 10 Consonant [0.547] 206 Vowel [0.383] 153 Voiceless fricative [ 2.572] 144 Stress Yes [2.223] 220 No [ 2.223] 293 Frequency +1 [0.003] -p <0.05, log likelihood = 210.278, intercept = 47.24, grand mean = 34.203 Turning to position, w e notice that greater VOT values are favored when the voiceless dental appears in word initial position (6.25), as opposed to when it is in word internal ( 6.25).
121 Although not significant in the linear regression analysis, word class shows certain tendenc ies with regard to VOT duration in the bilingual variety of Creole English. VOT duration seems to be longer in adjectives (7.947) and prepositions (0.906), and shorter in infinitival markers ( 0.312), verbs ( 1.156), pronouns ( 2.079), adverbs ( 2.585), an d nouns ( 2.720). In terms of following vowel height, VOT duration tends to be longer with high vowels (i = 7.991, u = 0.646), and shorter with mid and low vowels (e = 2.198, o = 3.163, a = 3.277). This pattern has also been attested in both monolingual and bilingual varieties of English and Spanish (Port & Rotunno, 1979; Weismer, 1979; Castaeda, 1986; Flege, 1995a; 1995b; Thornburg & Ryalls, 1998; Rosner et al., 2000; Munday, 2001; Deniz, 2005; Schmidt & Flege, 2006; Yavas & Wildermuth, 2006; Morris et al., 2007). With regard to preceding segment, VOT duration seem to be longer when the voiceless dental is phrase initial (1.642), preceded by other consonants (0.547), and preceded by vowels (0.383). VOT duration is shorter, however, when preceded by voic eless fricatives ( 2.572). Regarding syllable stress, it can be seen that VOT duration tends to be longer in tonic syllables (2.223) and shorter in atonic syllables ( 2.223). This factor has not been the focus of analyses concerning VOT in English spoken b y bilinguals; however, Lisker & Abramson (1967) found that in monolingual varieties of English VOT duration was longer in tonic syllab les, and shorter in atonic ones, thus, lining up with what has been discovered in the present study. In terms of word fre quency, although not significant, we notice that as frequency increases, VOT duration tends to be longer (0.003). We notice here that this direction of effect is opposite from that of the other three language modes; however, this could be
122 due to the number of tokens of /t/ in non frequent words as opposed to frequent ones. Notice that the coefficient is very close to 0. Here we see that the linear regression analysis allows us to observe the order of effect of the factors contributing to VOT duration in th e Creole English variety spoken by bilingual West Indians in Panama. Only two factors were selected as significant in this variety, position and rate of speech. In Section 3.2, I focus on the differences between monolingual and bilingual varieties of Spani sh and Creole English spoken in Panama, utilizing the comparative variationist method (Section 3.8). 4.2 Modeling Sound Change in Bilingual Speech This section is concerned with modeling sound change in bilingual speech, utilizing the comparative variati onist method established by Poplack & Tagliamonte (2001) in their work on African American English in the diaspora. This method has also been used in studies concerning language contact (e.g., Torres Cacoullos & Aaron, 2003; Blondeau & Nagy, 2008; Meyerhof f, 2009; Poplack & Levey, 2010). In the present study, I analyze the speech of bilinguals and compare it to that of monolinguals to discover evidence of sound change, and its possible source, based on the criteria discussed in Section 3.8. In this section, I present the results of the comparisons of the underlying patterns of VOT duration of the voiceless dental in the four language modes. In Section 4.2.1, I discuss how bilingual Spanish compares to the monolingual varieties of Spanish and Creole English. Later, in Section 4.2.2, I focus on the comparison of bilingual Creole English to the same monolingual varieties. Lastly, in Section 4.2.3, I compare bilingual Spanish to bilingual Creole English to find evidence of possible convergence in bilingual speech The results address Research Question 3 (Section
123 3.3) and Hypothesis 3 (Section 3.4) of the analysis, that is, do bilinguals exhibit change, and if so, is this change contact induced or internally motivated. 4.2.1 Bilingual Spanish Compared to Monoling ual Spanish and Monolingual Creole English Internally motivated change is determined when the bilingual variety does not follow its monolingual counterpart, or the source variety. That is, the bilingual variety does not resemble any of the languages. Conta ct induced change is determined when the bilingual variety does not follow its monolingual counterpart, but rather, the source variety. This is to say that the bilingual variety resembles only the source variety, thus showing that the change is due the ext ernal influences (see Section 3.8). Tables 4 6 and 4 7 compare bilingual Spanish with monolingual Spanish (the monolingual counterpart) and with monolingual Creole English (the source variety), where both internally motivated change and contact induced cha nge are apparent. Internally motivated change is noted in word class, in which bilingual Spanish does not line up with its monolingual counterpart or with the source variety. The change in constraint hierarchy is indicated by arrows pointing up or down. I n bilingual Spanish, VOT duration tends to be longer in pronouns than in prepositions or verbs, whereas in monolingual Spanish, VOT duration is longer in prepositions than in verbs, adverbs, or pronouns. When compared to monolingual Creole English, we obse rve that bilingual Spanish VOT duration is longer in adjectives than in adverbs or nouns, but in monolingual Creole English VOT duration is longer in adverbs and nouns than in adjectives. It is important to note that word class is not significant in either of the language modes. Since bilingual Spanish shows a change in constraint rankings when compared to both the monolingual counterpart and the source variety, this is evidence
124 of internal change in Spanish spoken by bilinguals. Table 4 8 indicates the cha nge in the constraint hierarchy of word class when bilingual Spanish is compared to the monolingual varieties. Table 4 6. Comparison of conditioning factors: Monolingual Spanish vs. Bilingual Spanish Monolingual Spanish Bilingual Spanish Factor Coef N Factor Coef N Vowel Vowel i 2.958 61 i 4.527 75 u 1.104 44 u 1.257 48 o 0.164 114 o 1.119 91 e 0.846 118 e 1.824 151 a 3.052 162 a 2.841 163 Preceding segment Preceding segment Zero 1.058 36 Zero [0.855] 27 Voiceless fricative 0.819 73 Voiceless fricative [ 0.772] 57 Consonant 0.762 169 Consonant [0.116] 200 Vowel 1.115 221 Vowel [ 0.199] 244 Word class Word class Preposition *[1.174] 5 Preposition [1.475] 11 Verb [0.374] 172 Verb [0.842] 147 Adverb [0. 255] 53 Adverb [ 3.326] 85 Pronoun [0.154] 40 Pronoun [3.563] 39 Noun [ 0.086] 126 Noun [ 1.510] 158 Adjective [ 1.870] 103 Adjective [ 1.044] 88 Position Position Word internal [0.356] 321 Word internal [ 0.06] 219 Word initial [ 0.356 ] 178 Word initial [0.06] 309 Stress Stress No [0.455] 314 No [0.527] 320 Yes [ 0.455] 185 Yes [ 0.527] 208 Rate of speech Rate of speech +1 [ 0.112] -+1 0.372 -Frequency Frequency +1 [ 0.037] -+1 [ 0.018] *Arrows indicate that the rankings change when compared to the monolingual variety.
125 Table 4 7. Comparison of conditioning factors: Monolingual Creole English vs. Bilingual Spanish Monolingual Creole English Bilingual Spanish Factor Coef N Factor Coe f N Rate of speech Rate of speech +1 1.013 -+1 0.372 -Vowel Vowel i 7.430 46 i 4.527 75 u 6.124 185 u 1.257 48 o 2.071 38 o 1.119 91 e 3.772 177 e 1.824 151 a 7.711 142 a 2.841 163 Preceding segment Preceding seg ment Zero 1.527 8 Zero [0.855] 27 Consonant 1.286 223 Consonant [0.116] 200 Vowel 0.491 199 Vowel [ 0.199] 244 Voiceless fricative 3.304 158 Voiceless fricative [ 0.772] 57 Position Position Word initial [2.001] 329 Word initial [0.06] 219 Word internal [ 2.001] 259 Word internal [ 0.06] 309 Word class Word class Pronoun [5.588] 45 Pronoun [3.563] 39 Preposition [0.414] 48 Preposition [1.475] 11 Adverb [ 0.281] 56 Adverb [ 3.326] 85 Noun [ 0.666] 178 Noun [ 1.510] 158 Verb [ 1.129] 93 Verb [0.842] 147 Adjective [ 3.817] 58 Adjective [ 1.044] 88 Stress Stress Yes [1.601] 197 Yes [ 0.527] 208 No [ 1.601] 391 No [0.527] 320 Frequency Frequency +1 [ 0.032] -+1 [ 0.018] -Contact induced change is noticed in preceding segment, position, and rate of speech; that is, bilingual Spanish is similar to the source variety and not its monolingual counterpart in terms of these factors. The preceding segment constraint hierarchy of bilingual Spanish parallels to that of monolingual Creole English and not to that of
126 monolingual Spanish. We notice that voiceless fricatives favor a longer VOT duration as opposed to other consonants and vowels in monolingual Spanish, however, in the bilingual variety, V OT duration tends to be shorter with voiceless fricatives and longer with other consonants and vowels. This order, zero > consonant > vowel > voiceless fricative, is also exhibited in monolingual Creole English. Table 4 8. Monolingual varieties vs. Bilingu al Spanish: internally motivated change in word class Monolingual varieties Bilingual Spanish Monolingual Spanish Preposition Pronoun Verb Preposition Adverb Verb Pronoun Adjective Noun Noun Adjective Adverb Monolingual Creole English Pr onoun Pronoun Preposition Preposition Adverb Verb Noun Adjective Verb Noun Adjective Adverb With regard to position, bilingual Spanish shares the same constraint hierarchy with monolingual Creole English. VOT duration is longer when the voiceless dental is in word initial position, and shorter in word internal position. In monolingual Spanish, the order is reversed. I should note that preceding segment is not significant in bilingual Spanish, and that position is not significant in any of the lang uage modes. In terms of rate of speech, monolingual Creole English and bilingual Spanish share this factor as significant, and it has the same direction of effect; as rate of speech increases, a shorter VOT duration is favored. In monolingual Spanish, rate of speech is not a significant factor, however, it does have the same direction of effect. Due to the
127 fact that bilingual Spanish resembles the source variety and not its monolingual counterpart in terms of preceding segment (Table 4 9), position (Table 4 10), and rate of speech, it can be concluded that contact induced change has occurred in the Creole English variety spoken by bilinguals. In 4.2.2, I compare bilingual Creole English to the monolingual varieties. Table 4 9. Monolingual varieties vs. Bilin gual Spanish: contact induced change in preceding segment Monolingual varieties Bilingual Spanish Monolingual Spanish Zero ( Zero ) Voiceless fricative ( Consonant ) Consonant ( Vowel ) Vowel ( Voiceless fricative ) Monolingual Creole English Zero ( Zero ) Consonant ( Consonant ) Vowel ( Vowel ) Voiceless fricative ( Voiceless fricative ) Table 4 10 Monolingual varieties vs. Bilingual Spanish: contact induced change in position Monolingual varieties Bilingual Spanish Monolingual Spanish ( Word int ernal ) ( Word initial ) ( Word initial ) ( Word internal ) Monolingual English ( Word initial ) ( Word initial ) ( Word internal ) ( Word internal ) 4.2.2 Bilingual Creole English Compared to Monolingual Spanish and Monolingual Creole English Tables 4 11 and 4 12 present the comparisons between bilingual Creole English and the monolingual varieties, where internally motivated change is apparent, and there is no evidence of contact induced change. That is, bilingual Creole English only exhibits change where it does not parallel its monolingual counterpart or the source variety. Contact Contact
128 Table 4 11 Comparison of conditioning factors: Monolingual Creole English vs. Bilingual Creole English Monolingual Creole English Bilingual Creole English Factor Coef N Factor Coef N Rate of speech Rate of speech +1 1.013 +1 1.56 Vowel Vowel i 7.430 46 i [7.991] 31 u 6.124 185 u [0.646] 157 o 2.071 38 o [ 3.163] 21 e 3.772 177 e [ 2.198] 172 a 7.711 142 a [ 3.277] 132 Preceding segment Preceding segment Zero 1.527 8 Zero [1.642] 10 Consonant 1.286 223 Consonant [0.547] 206 Vowel 0.491 199 Vowel [0.383] 153 Voiceless fricative 3.304 158 Voiceless fricative [ 2.572] 144 Position Position Word initial [2.001] 329 Word initial 6.2 5 303 Word internal [ 2.001] 259 Word internal 6.25 210 Word class Word class Pronoun [5.588] 45 Pronoun [ 2.079] 17 Preposition [0.414] 48 Preposition [0.906] 63 Infinitival marker [ 0.110] 110 Infinitival marker [ 0.312] 74 Adverb [ 0. 281] 56 Adverb [ 2.585] 34 Noun [ 0.666] 178 Noun [ 2.720] 156 Verb [ 1.129] 93 Verb [ 1.156] 112 Adjective [ 3.817] 58 Adjective [7.947] 57 Stress Stress Yes [1.601] 197 Yes [2.223] 220 No [ 1.601] 391 No [ 2.223] 293 Frequenc y Frequency +1 [ 0.032] +1 [0.003] -Internally motivated change is noticed in vowel height, word class, and word frequency. Although the expected effect was found for vowel height in bilingual Creole English, that is, VOT duration tends to be lon ger with high vowels, and shorter with low and mid
129 vowels, the constraint hierarchy differs from both monolingual varieties (Table 4 13). VOT duration is longer with /o/ than with /e/, whereas in the monolingual varieties, /e/ favors a longer VOT duration than /o/. Table 4 12 Comparison of conditioning factors: Monolingual Spanish vs. Bilingual Creole English Monolingual Spanish Bilingual Creole English Factor Coef N Factor Coef N Vowel Vowel i 2.958 61 i [7.991] 31 u 1.104 44 u [0.646] 157 o 0 .164 114 o [ 3.163] 21 e 0.846 118 e [ 2.198] 172 a 3.052 162 a [ 3.277] 132 Preceding segment Preceding segment Zero 1.058 36 Zero [1.642] 10 Voiceless fricative 0.819 73 Voiceless fricative [ 2.572] 144 Consonant 0.762 169 Consonant [0.547] 206 Vowel 1.115 221 Vowel [0.383] 153 Word class Word class Preposition *[1.174] 5 Preposition [0.906] 63 Verb [0.374] 172 Verb [ 1.156] 17 Adverb [0.255] 53 Adverb [ 2.585] 112 Pronoun [0.154] 40 Pronoun [ 2.079] 34 Noun [ 0 .086] 126 Noun [ 2.720] 156 Adjective [ 1.870] 103 Adjective [7.947] 57 Position Position Word internal [0.356] 321 Word internal 6.25 210 Word initial [ 0.356] 178 Word initial 6.25 303 Stress Stress No [0.455] 314 No [ 2.2 23] 293 Yes [ 0.455] 185 Yes [2.223] 220 Rate of speech Rate of speech +1 [ 0.112] -+1 1.56 -Frequency Frequency +1 [ 0.037] -+1 [0.003] -
130 Following vowel height is not significant in the bilingual variety of Creole English as opposed to monolingual Creole English and Spanish, thus further demonstrating an internal change because it does not look like any of the languages. In terms of word class (Table 4 14), bilingual Creole English does not line up with either monol ingual variety. When compared to monolingual Creole English, we notice that the only similarity in ranking is found with preposition and infinitival marker. When compared to monolingual Spanish, the hierarchies coincide only with pronouns. Word class did n ot achieve significance in any of the language modes. Table 4 13. Monolingual varieties vs. Bilingual Creole English: internally motivated change in vowel height Monolingual varieties Bilingual Creole English Monolingual Spanish i i u u o e e o a a Monolingual Creole English i i u u o e e o a a Turning to the effect of word frequency, we observe that bilingual Creole English differs from both monolingual varieties in that as frequency increases, VOT duration becomes longer; whereas in monolingual Creole English and Spanish, VOT duration becomes shorter as frequency increases. This factor also did not achieve significance in any of the language modes. As we mentioned above, there is no evidence of contact induced change. Bilingual Creole English does not parallel monolingual Spanish in any of the factors, except for
131 rate of speech. However, monolingual Creole English also has the same effect of rate of speech, thereby demonstrating that the similarity found between bilingual Creole Englis h and monolingual Spanish is not due to contact. Table 4 14 Monolingual varieties vs. Bilingual Creole English: internally motivated change in word class Monolingual varieties Bilingual Creole English Monolingual Creole English Adjective Pronoun Pr eposition Preposition Infinitival marker Infinitival marker Verb Adverb Pronoun Noun Adverb Verb Noun Adjective Monolingual Spanish Preposition Pronoun Verb Preposition Adverb Adverb Pronoun Noun Noun Verb Adjective Adjective 4.2.3 Bili ngual Spanish compared to Bilingual Creole English The comparison between bilingual varieties is performed to determine whether bilinguals have two separate phonetic systems or one merged system. In his Speech Learning Model (SLM), which is based on an equ ivalence classification ( Section 2.4 .4), Flege (1995a) discovers evidence of cross language assimilation in bilinguals. In this process, bilingual speakers produce compromise VOT durations that fall between the prototypical ranges of monolingual speakers. According to Flege, this would be evidence for a merged phonetic system. Flege (1995a) also discovers cross language dissimilation in which bilingual speakers display deflected VOT durations; that is, they deviate from a phonetic category in order to maint ain a phonetic contrast. He claims that
132 this process is evidence for two separate phonetic systems, and it is most common in early bilinguals or heritage speakers (Flege, 1995b:100 101). If we revisit Table 4 1, we notice that the mean VOT value obtained f or the monolingual speakers of Spanish is 15.24, while for the monolingual speakers of Creole English, it is 28.43. Thus, if the bilinguals produced Spanish stops with VOT values that we could consider to be a compromise, we would expect to observe a value of 21.84. In fact, bilingual speakers produced the Spanish voiceless dental with a VOT value of 21.15, thus showing that bilingual speakers have a compromise VOT value in Spanish. Regarding the production of the Creole English voiceless alveolar by biling uals, if the VOT values are considered compromises, the mean should also fall between monolingual Spanish and monolingual Creole English means. In Table 4 1, we observe that bilinguals produced the voiceless alveolar with a VOT value of 34.20. Thus, we see here that the mean is not compromised, but rather deflected; that is, it is even longer than the monolingual Creole English mean. This is better exemplified in Figure 4 1. Figure 4 1. Comparison of mean VOT across four varieties If we are working withi n the SLM, these values indicate that the Spanish Creole English bilinguals may have a separate phonetic system for each language and that
133 these speakers have moved away from monolingual norms. However, it is generally accepted that language change is grad ual; usage based models of language claim that change occurs in certain dimensions of the languages before others. Therefore, a claim that change has taken place based on mean values must be regarded with suspicion. Likewise, the idea that the two language s of bilinguals have completely separate systems is also dubious. Considering these suspicions, we should ask ourselves to what extent bilingual Spanish differs from bilingual Creole English, and whether the line is clean cut. A variationist model elucidat es the situation when we compare the underlying patterns of the factors that condition VOT duration in bilingual Spanish and bilingual Creole English. Table 4 15 presents the comparison of the underlying systems of bilingual Spanish and bilingual Creole E nglish. This comparison should uncover the similarities and the differences between the phonetic systems of the two languages spoken by a bilingual group. Generally, the results show that the constraints in each language are indeed different, thus, demonst rating two phonetic systems. However, the comparison also reveals evidence of convergence. Although we notice that bilingual Spanish and bilingual Creole English differ in the factors selected as significant in the linear regression analysis (in the forme r, vowel height; in the latter, word position), there is also one that is shared: rate of speech. Moreover, the direction of effect is the same in this factor: as rate of speech increases by one unit, VOT values increase. Thus, we have the first piece of e vidence demonstrating that the two phonetic systems are not completely separate, in spite of preliminary conclusions we may like to draw based on mean values alone.
134 Table 4 15 Comparison between bilingual varieties Bilingual Spanish Bilingual Creole En glish Factor Coef N Factor Coef N Vowel Vowel i 4.527 75 i [7.991] 31 u 1.257 48 u [0.646] 157 o 1.119 91 o [ 3.163] 21 e 1.824 151 e [ 2.198] 172 a 2.841 163 a [ 3.277] 132 Rate of speech Rate of speech +1 0.372 -+1 1.56 -Word class Word class Pronoun [3.563] 39 Pronoun [ 2.079] 17 Preposition [1.475] 11 Preposition [0.906] 63 Verb [0.842] 147 Verb [ 1.156] 112 Adjective [ 1.044] 88 Adjective [7.947] 57 Noun [ 1.510] 158 Noun [ 2.720] 156 Adverb [ 3. 326] 85 Adverb [ 2.585] 34 Preceding segment Preceding segment Zero [0.855] 27 Zero [1.642] 10 Consonant [0.116] 200 Consonant [0.547] 206 Vowel [ 0.199] 244 Vowel [0.383] 153 Voiceless fricative [ 0.772] 57 Voiceless fricative [ 2.572] 144 Position Position Word initial [0.06] 219 Word initial 6.25 210 Word internal [ 0.06] 309 Word internal 6.25 303 Stress Stress No [0.527] 320 No [ 2.223] 293 Yes [ 0.527] 208 Yes [2.223] 220 Frequency Frequency +1 [ 0.018] -+1 [0.003] -Another piece of evidence of convergence between bilingual Spanish and bilingual Creole English is found in preceding segment, where the tendency is the same: zero > consonant > vowel > voiceless fricative. It is also important to note that monolingual Creole English exhibits the same constraint hierarchy, and monolingual Spanish does
135 not (see Table 4 9); therefore, the convergence indicates that bilingual speakers are moving towards monolingual Creole English norms. Convergence is further supported in terms of this factor group due to the fact that preceding segment is not significant in the bilingual varieties but is in the monolingual ones. A final piece of evidence of convergence is found in word position, where VOT values te nd to be greater in word initial position and shorter in word internal in bilingual Spanish and bilingual Creole English. I should note here that monolingual Creole English shares the same tendency, and monolingual Spanish does not (see Table 4 10). Theref ore, the convergence shows that bilinguals are leaning towards monolingual Creole English norms, as was seen for preceding segment; however, bilingual Creole English is the only variety to have word position as a significant factor. 4.3 Summary In this cha pter, I give a variationist account of voice onset time of the voiceless dental in monolingual and bilingual varieties of Spanish and Creole English spoken in Panama. This method includes all the possible factors into a multiple linear regression in order to capture the underlying system of the voiceless dental in terms of VOT duration. This differs from traditional studies on voice onset time, in which researchers analyze the main effects of factors separately. Also, when analyzing the speech of bilingual s, previous studies do not focus on changes occurring in the grammar of bilingual speakers. In this chapter, I use the comparative variationist method to determine whether change has occurred in bilingual speech, and the purported source of that change. Th is method also allows us to see exactly where the change has occurred. When bilinguals speak Spanish, it is noticed that internally motivated change has occurred in their production of the voiceless dental,
136 particularly in word class. Internally motivated change is also discovered in the production of the voiceless dental in Creole English. The change is evinced in vowel height, word class, and word frequency. The comparative method also reveals evidence of contact induced change in bilingual speech, parti cularly when bilinguals speak Spanish. This change was observed in preceding segment, position, and rate of speech. The underlying system of bilinguals in Spanish is more similar to that of monolingual speakers of Creole English. Interestingly, contact ind uced change was not discovered in the Creole English of of Spanish. The speech of bilinguals can also be examined by comparing their two systems side by side; that is, bilingual Spanish versus bilingual Creole English. With this comparison, I offer an alternative approach rooted in variationist theory to explore how phonetic systems are represented in the minds of bilinguals. Traditional studies concerning VOT duration have employed techniques normally found in SLA research, and have based findings on mean values alone. The approach taken in the present analysis incorporates natural spontaneous speech, and hones in on the patterns among bilinguals concerning VOT duration of the voiceless dental. The results suggest that the mental representation of phonetic systems in bilinguals seems to be far more complex than mean VOT duration, and that in order to capture gradient differences, the probabilistic knowledge of bilingual speakers must be taken into account. Therefore, in spite of the fact that bilinguals seem to exhibit two separate phonetic systems, otherwise invisible evidence of convergence in multiple dimensions of the languages
137 have been brought to light through a var iationist approach. These results are further discussed in Chapter 5.
138 CHAPTER 5 DISCUSSION In this chapter, I revisit the research questions and hypotheses presented in Chapter 3. I discuss the mean VOT durations and the conditioning factors in monoling ual and bilingual varieties of Spanish and Creole English in Section 5.1. I then summarize the findings revealed in the series of comparisons and discuss the changes that have occurred in the bilingual speech community (Section 5.2). I also present qualita tive data drawn from the sociolinguistic interviews to explain why these changes could be taking place. These explanations may have important implications for sociophonetic research concerning language contact, variation and change. 5.1 Mean VOT and Condi tioning Factors Research Question 1 addresses the distribution of mean VOT duration for the voiceless dental plosive in the speech of monolingual and bilingual speakers of Spanish and Creole English. It was expected that the Spanish varieties would exhibit a shorter mean VOT duration than the Creole English varieties, regardless of the monolingual/bilingual distinction. The results, shown in Table 5 1, confirmed this hypothesis. Table 5 1. Comparison of mean VOT across varieties Spanish Creole English Monolingual 15.25 28.43 Bilingual 21.15 34.20 Both the monolingual and bi lingual varieties of Spanish did indeed have shorter mean VOT duration s than both the monolingual and bilingual varieties of Creole English. These results are in line with previou s literature concerning VOT duration, which notes that Spanish VOT is short lag and English VOT is long lag. In short lag VOT, voicing occurs with, or shortly after the release of the stop. In long lag VOT,
139 voicing occurs well after the release of the stop and normally, aspiration can be perceived during this interval (Lisker & Abramson, 1964). I also posited that bilinguals would exhibit mean VOTs in Spanish that would fall between the prototypical ranges of the monolingual varieties of Spanish and Creole English. In Creole English, bilinguals would produce longer mean VOTs than the monolingual variety of Creole English. Speech Learning Model in which he discovered processes such as cross language assimilation in bilinguals due to an equivalence classification of similar sounds in both languages (Flege, 1995b:101) A compromise VOT value is evidence for this process because both languages exhibit about the same mean VOT that falls between the monolingual varieti es. Flege (1995b) also discovered cross language dissimilation, in which speakers produce deflected, or extra long, VOTs in a language in order to distinguish from longer VOTs in another. This would be evidence for separate phonetic systems, which Flege (1 995b:101) claims is more common in early bilinguals. Separate phonetic systems were expected in the present analysis because the participants were early bilinguals of Spanish and Creole English. ation hypothesis, which is in line with the Speech Learning Model. In Figure 5 1, we see that bilinguals have compromise VOT values in Spanish, but their values in Creole English exceed those of monolingual Creole English speakers. Thus, bilinguals seem to show dissimilation in their phonetic system based on mean VOT values. The bilingual participants included in this analysis claimed to have learn Creole English in the home and Spanish by the age of 5; therefore, we can consider them to be early bilinguals According to Flege (1987)
140 and Flege (1995b), these bilinguals would have two systems because by the time they began to learn the L2, their L1 had not been fully developed already; therefore the process of equivalence classification did not block the deve lopment of the L2 phone, but rather aided in establishing it independently of the L1. This would explain why these bilinguals exhibit a different value for each language. If the voiceless dental of another group of bilinguals with a later age of acquisi tion were analyzed in terms of VOT duration, we would expect to see less divergence in their system. Late bilinguals show more convergence in their system because the equivalence classification process blocks the development of L2 phones that are similar t o those of the L1. They do not produce the L2 phone authentically because as they mature they rely less on the sensory information that is used to establish sound categories in a new language. Therefore, the information they perceive is filtered through th e L1 (Werker & Logan, 1985; Flege, 1987), thus looking more like the L1. Figure 5 1. Assimilation and dissimation in bilinguals
141 In Chapter 2, we saw that several studies have discovered various linguistic factors that condition VOT duration; however, these factors are generally analyzed separately such that we are not able to see which factors are stronger than others. Figure 4 1 does not tell the whole story about bilingual VOT duration; this acoustic feature is more complex. That is, the underlying p attering of the factors that condition VOT duration could line up or not when comparing monolingual varieties to bilingual varieties. In the present analysis, I took a variationist approach in which I incorporated linear regression models with mixed effect s to pinpoint the patterns of variation in the natural production of each variety. The findings of these analyses addressed Research Question 2 and Hypothesis 2. I was concerned with which factors would significantly condition VOT duration in the four lang uage modes and their constraint hierarchies. I expected to find patterns generally similar to those that have been attested in the previous literature; however, this innovative approach to examining the conditioning factors allows us to see a more detailed snapshot of the full system in one analysis for each language. This is important for a usage based model of phonology because in this framework interactions within a phonological system are inevitable. Sounds are organized in a network of clusters in whic h exemplars are activated based on factors such as frequency and association (Langacker, 2000; Pierrehumbert, 200; Bybee, 2001). In order to truly analyze the full system, one must attempt to consider as many factors as possible that play a role in the rea lization of sound categories. Therefore, the variationist methodology utilized in this study is perhaps one of the most important tools with regard to the analysis of VOT duration.
142 The results showed that when looking at the constraint hierarchies, Panaman ian Spanish and Panamanian Creole English had tendencies similar to other varieties of Spanish and English. For example, low vowels favor a shorter VOT duration and high vowels favor longer ones. Interestingly, in monolingual Spanish, a preceding voiceless of VOT duration in Castilian Spanish. This finding is surprising because it contradicts physiological claims for the opposite effect. Yoshioka et al. ( 1981 ) explain that voiceless fricative + voiceless plosive clusters general show a one peak pattern for the plosive because the fricative causes it to be unaspirated, thus exhibiting a shorter VOT duration. We see that this is not the case for monolingual Spanish speakers i n my analysis, but it is so for the monolingual Creole English speakers. 5.2 Creole English VOT and Creole Genesis Theories One of the questions in this study (Section 2.3.5) was how the Creole English varieties would compare to other varieties of English, as there are no studies that have analyzed VOT duration in English based creoles. This was an interesting question to pose because as we mentioned in Section 1.2.1 certain theories attribute the origin of creoles mainly to the dialects spoken by European colonizers (Mufwene, 1996). Poplack (2000) and Poplack & Tagliamonte (2001) made this hypothesis about Early African American English (AAE) and African American Vernacular English (AAVE). They utilized the comparative variationist method to test their pred ictions, comparing the mentioned varieties with other putative sister dialects, such as the speech of ex slaves living in the U.S. South obtained from old recordings, Saman English in the Dominican Republic, and African Nova Scotian English in Canada. The y also included two controls: Nova Scotian Vernacular English and an English variety spoken in the county of Devon
143 in southwest England. Striking parallels were found among the underlying grammars of all the varieties. From this, they were able to conclude that modern day AAVE is a variety derived from British origin English varieties and not English based creoles spoken in the Caribbean. It appears that in the present analysis we could contribute to the debate on the origin of creoles. The results in this study allows us to confirm that in terms of VOT duration of the voiceless dental plosive, Panama City Creole English and Bastimentos Creole English do not differ drastically from other English varieties, especially in the underlying patterns. First, we sa w that VOT duration was longer in the Creole English varieties than in the Spanish varieties. This was noted in previous analyses of other varieties of English (Section 2.3.3). Additionally, the underlying system of the conditioning factors resembled patte rns of the North American varieties reported on in Sections 2.3.4 and 3.7. These results appear to support claims in creolistics literature that creole languages are dialects that developed from English varieties spoken by European colonizers. 1 More acoust ic work on consonant stops is needed in order to further investigate these claims and find possible similarities and differences between English based creoles and their lexifiers, as this could have important implications for theories on creole genesis. 5.3 Sound Change in the Bilingual West Indian Speech Community In this section, I address Research Question 3 and Hypothesis 3. These address possible change occurring in the speech of bilingual West Indians, and whether this change is due to internal pro cesses or language contact. Change was thought to be 1 See Po plack & Tagliamonte (2001) for their work on African American English in the diaspora.
144 internally motivated if bilinguals did not follow any of the monolingual norms in any of their languages. For example, if bilingual Spanish patterned differently from monolingual Spanish and monolingual Creole English, this would be evidence for internal change. Contact induced change was identified if bilingual speech patterns paralleled the potential cross language source variety, but not its monolingual counterpart. That is, if bilingual Spanish resemb led monolingual Creole English, and not monolingual Spanish, this would be evidence for change due to contact. I posited that bilinguals would follow monolingual Spanish norms in both Spanish and Creole English. That is, I expected bilinguals to show conta ct induced change when producing Creole English, and to have patterns similar to those of monolingual Spanish speakers when producing Spanish. This hypothesis was based on the fact that the West Indian community has been present in Panama since the early 1 850s, and it is smaller in size compared to the monolingual Mestizo community. Also, historically, West Indians have not been as socioeconomically successful as Mestizos. 2 It is also important to note that Spanish is the dominant language in this region; t hus, I expected there to be a language shift towards Spanish. These factors are part of a general predictor for contact induced change, intensity of contact, as discussed in Thomason (2001). Another goal was to determine whether bilinguals had a merged pho netic system or if they maintained separate phonetic systems for each language. As mentioned previously, Flege (1995a) proposed that we determine the nature of the mental representation of bilinguals based on mean VOT durations. Traditional studies concern 2 Refer to historical background of bilingual West Indians in Panama in Chapter 1.
145 in second language acquisition (SLA) in order to identify the factors that condition VOT duration. These techniques include data elicitation through reading tasks, and difference s between monolinguals and bilinguals are determined based on mean VOT lengths. However, there are alternatives. In fact, various researchers have taken on the task of analyzing learner variation with Labovian variationist methods (e.g., Adamson, 1988; Bay ley, 1991, 1994, 1996; Young, 1991; Adamson et al., 1996; Flege, Munro, & MacKay, 1996; Regan, 1996; Geesling & Guijarro Fuentes, 2006). Flege et al. (1996), in particular, analyze VOT values in native Italian speakers learning English. Through multiple re gression analyses, they find that the age of second language (L2) learning is not the sole factor to condition the production of /p/ and /t/, but that factors such as language use also play an important role in the acquisition of L2 sounds. While their an alyses, in part, utilize variationist methods, the data are collected through reading tasks, and the multiple regression analyses are used to identify only external factors, not internal. To my knowledge, despite growing interest in the Labovian approach, this study is the first to truly employ Labovian variationist methods in the analysis of VOT duration in bilinguals. I utilized variationist methods to capture the sponta neous speech and a statistical analysis that allowed us to observe the constraint hierarchies of the conditioning factors in bilingual speech. I expected that early bilinguals would display separate phonetic systems for each language, thus showing evidence of cross
146 5 3 .1 Emerging Antillano Spanish (AS) and Antillano English (AE) The findings of the analysis suggest that the West Indian speech community could be in the initial stages of having n ew emerging varieties, which I will call Antillano Spanish (AS) and Antillano English (AE) 3 Table 5 2. Factors showing internally motivated change Monolingual Spanish Antillano Spanish Antillano English Monolingual Creole English word class Preposit ion Pronoun Adjective Pronoun Verb Preposition Preposition Preposition Adverb Verb Infinitival marker Verb Infinitival marker Adverb Pronoun Adjective Pronoun Noun Noun Noun Adverb Verb Adjective Adverb Noun Adjective vowel height i i i i u u u u o o e o e e o e a a a a word frequency Although not certain, the apparent emergence of these varieties is based on internally motivated and contact induced changes occurring in this speech community. Additionally, converg ence in bilingual speech suggests that bilingual West Indians are moving towards monolingual Creole English speaker norms. When Antillano Spanish 3 n Panama to refer to people of West Indian descent. I chose to use this term specifically in Spanish because it represents a cultural identity that consists of West Indian, Mestizo, and even North American customs and traditions. The meaning of the work re presents their West Indian identity, but the Spanish translation symbolizes the Panamanian influences in their daily lives.
147 (=bilingual Spanish) was compared to the monolingual varieties, internal change was observed only in word clas s (Table 5 2), and contact induced change, which is shown in Table 5 3, was seen in preceding segment, position, and rate of speech. When comparing Antillano English (=bilingual Creole English) to the monolingual varieties, internally motivated change (Tab le 5 2) was noticed in vowel height, word class, and word frequency. Contrary to the hypothesis, there was no evidence of contact induced change when bilinguals spoke Antillano English; it appears that a language shift is not occurring among the participan ts of this study. Table 5 3. Factors showing contact induced change Monolingual Spanish Antillano Spanish Antillano English Monolingual Creole English preceding segment Zero Zero Zero Zero Voiceless fricative Consonant Consonant Consonant Consonant Vo wel Vowel Vowel Vowel Voiceless fricative Voiceless fricative Voiceless fricative position Word internal Word initial Word initial Word initial Word initial Word internal Word internal Word internal rate of speech Not significant Significant Sign ificant Significant When performing the intra speaker comparison, that is, Antillano Spanish compared to Antillano English, the hypothesis was confirmed. We discovered that bilinguals generally maintain separate phonetic systems for each language; howeve r, there were signs of convergence. First, we saw that the languages are similar with regard to rate of speech (Table 5 4). This factor was significant in both varieties and had the same direction of effect. Convergence was also evident in preceding segmen t (Table 5 5) in that this factor group was not significant in either language and the
148 constraint hierarchy was the same. Finally, Antillano Spanish and Antillano English shared the same constraint hierarchy in word position (Table 5 6), thus another sign of convergence. We see that in spite of what the merger hypothesis (see Flege (1987) and Flege (1995)) predicts for early bilinguals, the participants in this analysis still approximate their languages in some areas. It appears that even though these bilin guals maintain separate phonetic systems because they were able to establish Spanish phones independently of Creole English phones, as they become older, the process of equivalence classification begins to cause new exemplars from Spanish and Creole Englis h to be identified as the same, thus convergence ensues. As bilinguals receive input from both monolingual and bilingual speakers, the new exemplars are strengthened, and therefore convergence spreads gradually throughout the systems. In this case, this ha s occurred in rate of speech, preceding segment, and position. What I propose therefore, is a Variationist Speech Learning Model (VSLM) in which convergence is predicted also for early bilinguals, but that begins at a later age. This hypothesis allows for more variability of outcomes in the speech of early bilinguals, as some may not show convergence but others could. On one hand, convergence could occur if bilinguals tend to mix their languages in their daily lives or use either or in the same contexts. It could also occur as a way for them to act out their identity through their speech patterns. On the other hand, the absence of convergence could be explained if bilinguals tend to assign certain domains to their two languages, thus always keeping them sepa rate. At first it seems that the VSLM does not account for the direction of convergence. It is possible that there can be convergence in which the language moves towards the
149 norms of one of the systems. That is, Spanish Creole English bilinguals could con verge their languages and resemble Spanish, or they can converge and move towards Creole English. These types of convergence could be due to external factors. It appears that we could incorporate the importance of external influences into the VSLM, which in this case is related to social factors. As noted in Chapter 4, the areas in which the Antillano varieties converged were where they became more similar to monolingual Creole English; therefore, with this convergence, we can ask the question of whether or not these bilinguals are establishing or acting out an Antillano identity in which they move towards monolingual Creole English norms. We could assume that this works as a mechanism to block possible language shift towards Spanish in these speakers. Thi s is further discussed in Section 5.3.2. Table 5 4. Convergence between bilingual varieties in significance of rate of speech Antillano Spanish Antillano English Significant Significant Table 5 5. Convergence between bilingual varieties in precedin g segment Antillano Spanish Antillano English Zero Zero Consonant Consonant Vowel Vowel Voiceless fricative Voiceless fricative Table 5 6. Convergence between bilingual varieties in position Antillano Spanish Antillano English Word initial Word i nitial Word internal Word internal Based on the analysis of linguistic factors, we saw that various changes have taken place in the West Indian speech community; some were due to internal processes and others, language contact. However, when approaching the topics of language
150 contact and bilingualism, it is paramount that we take into account possible social pressures leading to change. Since social factors are not quantitatively accounted for in the statistical analyses, in Section 5.3.2, I give qualita tive explanations on how certain social factors could bring about the possible emergence of Antillano Spanish and Antillano English. 5.3.2 Qualitative Explanations for the Emergence of AS and AE In the framework of language variation and change (LVC), the evolution of language has been attributed to the geography or the social setting in which language changes. Such external factors, along with internal ones, have been considered to be re can be many explanations for why certain forms die out, and why others enter or become more frequent in a language. Many researchers concerned with LVC take into account social factors, such as language attitudes (e.g. Hill & Hill, 1980; Appel & Muysken 2005), language loyalty (e.g. Fishman, 1964; Grosjean, 1982) and indexicality (e.g. Eckert, 2000; 2008; Hall Lew, 2009). Wolfram & Schilling Estes (2006) provide a list of several reasons for language evolution that include sociohistorical explanations, and Tagliamonte (2012) adds to this list sociocultural factors in which she mentions media influence and the nature of the social network. In this section, I explain the changes occurring in the West Indian speech community from a social perspective by inc orporating some statements extracted from my sociolinguistic interviews in which language attitudes and loyalty are manifested so as to index a certain West Indian persona. As a guide, I use the random effect results provided by the linear regression model s of the bilingual speakers (Table 5 7). The extraction from individual speaker
151 interviews will be organized as follows: convergence moving towards Creole English, convergence moving towards Spanish, and divergence. As I discuss in the methodology (Chapte r 3), the linear regression models were fitted with mixed effects. There were linguistic factors that were considered fixed effects, such as following vowel height. However, also included were factors considered random effects because they were sampled fro m a larger population, such as individual speaker or lexical item. For purposes of the qualitative explanations for language change, we focus on the individual speakers of the analysis. Each speaker is assigned a coefficient in the regression model, which allows you to observe how they vary from one another (Baayen, Davidson & Bates, 2008; Johnson, 2009; Drager & Hay, 2012). Therefore, in terms of VOT duration, we are able to determine the likelihood that a speaker will produce a longer VOT as opposed to a shorter one. Table 5 7 presents the results of individual speaker (bilinguals) as a random effect in each language. We are able to see that different types of behavior in these bilinguals, such as convergence moving towards Creole English (Section 18.104.22.168) convergence moving towards Spanish (Section 22.214.171.124), and divergence (Section 126.96.36.199). Table 5 7. Individual speaker as random effect in linear regression model Spanish Creole English Speaker (random) coef tokens Speaker (random) coef tokens Samuel 6.938 108 Alicia 3.641 124 Alicia 0.615 104 Samuel 2.993 115 Alberto 0.917 100 Alberto 2.588 101 Rogelio 1.154 110 Rogelio 1.194 102 Dora 4.251 106 Dora 8.028 71
152 188.8.131.52 Convergence moving towards Creole English Positive coefficients correspond to longer VOT durations and negative coefficients, to shorter VOT durations. Thus, we notice that Samuel is more likely to produce voiceless dental plosives with longer VOT durations in Spanish (6.938) and in Creole English (3.641). This can be interprete d to mean that he shows convergence in his phonetic systems, and the languages are moving towards Creole English because he favors longer VOTs. These results seem accurate when we look at the qualitative data. First of all, Samuel has a very positive attit ude towards what we referred to in the Interviewer: How important is S panish to you? Samuel: M ore or less important. Just to um... j ust to do the basic like to go to the to do you g um, if you have fi deal with the government or something. But, um, it not that important to me. It it's important but I could do without it. ( laughs ) Interviewer: nd how important is English to you? Samuel: V ery important. (Samuel, Antillano En glish, 6/5/2010) His attitudes towards these languages can also be gleaned from how he evaluates his own language skills: Interviewer: H ow well do you speak S panish? Samuel: N ot so well. Well, not perfectly, but very well. Interviewer: S o, very well or n ot so well ? Samuel: V ery well. Interviewer: A nd how well do you speak English? Samuel: V ery good. Rea d speak, write, very well. Perfectly. (Samuel, Antillano English, 6/5/2010)
153 We see here that Samuel is very confident of English and explains that he i s good at all aspects of using the language; however, he seems non committal when describing his Spanish. Samuel also shows preference for English in his language use, and explains that when he was growing up, he only spoke English with his parents. He als o states that he instilled the importance of English in his own children: Interviewer: and who do you normally speak Spanish to? friends? Samuel: um, neighbors or... um, people that round here. Yea, the young children, not my children, I speak English to them. I even even to my dogs and my ( laugh ) parrot I speak English to them ( laugh ) Interviewer: and so when you were growing up with you parents, you spoke usually spoke...? Samuel: English, they speak to me in English. Interviewer: um, do you ever s peak English to neighbors? Samuel: um, yes, I have one that come from Denver. I have next neighbor that she grow with me on the zone. I speak English with her. Yea, all the time. And another one that live across the road, the other street, live yea, I sp eak English with them. and that one of the things I instill in my kids. I told them bruntly, in this house everybody speak English. That language, Spanish, you will learn it outside or with you neighbors. And so said, they live today to realize what I wa s trying to tell them. Now it's a hassle for people to learn English in P anama, when they didn't want. So they they they is ahead of the game. (Samuel, Antillano English, 6/5/2010) ive data show for his use of Spanish and Creole English. 184.108.40.206 Convergence moving towards Spanish to produce shorter VOT durations in both Spanish ( 4.251, 1.154) and Creole English ( 8.028, 1.194). They seem to also be exhibiting a convergence in their phonetic
154 systems; however, contrary to Samuel, they are moving towards Spanish because they favor shorter VOTs. Their statements about language and culture also seem to corroborate these results. For example, Dora, although she has very positive attitudes towards Spanish and English, she seems to lean more towards Spanish, which is seen in her reasons for giving each language importance: Interviewer: How important is Spa nish to you? Dora: How? Interviewer: How important is Spanish to you? Dora: It is it is very important. It is very important because I live in a Latin country. Although they say they claim now that the English is the number one language because of the call centers I guess, you know? But I think that en Spanish is very important. Interviewer: And how well do you think that you speak Spanish? Dora: Interviewer: so perfectly or very well? Dora : I say very well, yea. Interviewer: and how important is English to you? Dora: I would say it is very important too, because um... sometimes on the street or in different area you can meet a person that is looking for an address or they want to know how to get certain place, and they don't speak Spanish. (Dora, Antillano English, 5/13/2010) Also, when asked about her linguistic skills in the language, she is more confident with Spanish than with English: Interviewer: How well do you speak English? Dora: umm... I would say not that well. Interviewer: So not so well?
155 Dora: Not so well. (Dora, Antillano English, 5/13/2010) However, contrary to the fact that she seems to prefer Spanish a litt le more than English, she uses English a lot in her family: Interviewer: do you get a lot of chance to speak English? Dora: Yea. yes. Yes, I do. Interviewer: Who do speak English with? Dora: My cousin. em, my sister, my son, eh... my nieces and nephew that live in the states. my husband, my mother in law, because she don't speak much Spanish, yea. and my father in law also, you know. He didn't speak much Spanish either. so, we speak more we got to speak more English with him, and my brother in law als o. He speaks Spanish, but usually we mother and my father. XXX although they speak they used to they speak Spanish too, but you know, when they got when they got sick and then I would have to go with them I spend time with them, you know I would speak more English. Definitely, the Spanish was out, and I strictly speak speak English with them. (Dora, Antillano English, 5/13/2010) Dora also identifies herself as West Indian, which seems to be very important when it Interviewer: How do you identify yourself as far as culture and race? especially now with the census going on? Dora: well, my husband say t Interviewer: (laugh) Dora: I me I don't fe I don't consider myself a Spanish, is just that I grew up among a lot of Spanish people. But deep down, my roots is from the West Indies, you know. So I consider myself black or from the West Indies, or whatever afroantillano, afroantillano descendiente, whatever they want to say, but I consider myself what Panamanian. (Dora, Antillano English, 5/13/2010)
156 We notice at the end of he r statement that she still continues to hold on to being a Panamanian, which is something that was not seen as much in previous generations (see Chapter 2). Interestingly, Rogelio is married to Dora, which could explain why they have the same tendencies. H owever, according to his qualitative data, he associates himself much more with West Indian culture and with people of West Indian descent. This also appears to corroborate the random effect results, since he shows longer VOT durations than Dora: Interv iewer: and so how do yo u usually identify yourself? Rogelio: (laugh ) I don't need to identify myself. I I project exactly what I am. I am... what should I say ah... black descendent from the West Indians. Alright, I project and ... most the people I k now, people I relate with, I have the same, you know, same the descendants, so. Okay I ... Interviewer: S o a lot of you r friends are also...? Rogelio: Yes, yes, definitely definitely definitely. (Rogelio Antillano English, 5/24 /2010) Rogelio: Bueno, en Col n, eh... ramos bastante, cmo se dice, cmo la mayora de la gente era de habla inglesa, la mayora de la msica que yo escuchaba, RIGHT? eh... era en ingls. Haba estaciones de radio, y todo lo que era SOUL y lo dems, eso eso estaba de moda en ese ti empo, okay, para nosotros, okay. Algunos que otros, eh... cmo dira yo, msica en espaol, alguna salsa, tal vez, en esos tiempos El Gran Combo, ya despus la cosa fue evolucionando, y habiendo (inc) lo que es... msica en espaol, no? pero primo primord ialmente, primera poca era era cuestiones en ingls. (Rogelio, Antillano Spanish, 5/24/2010) Translation: of the people were English speaking, the majority of the music we listened Soul, that El Gran Combo. After that, things beg an to change, having XXX what is music in Spanish, no? But primarily, at the beginning it was music in English.
157 220.127.116.11 Divergence Turning to Alicia and Alberto, we notice that they are more likely to produce longer VOTs when speaking C reole E nglish ( 0. 615, 0.917) and shorter VOTs when speaking Spanish (3.641, 2.588) They seem to be keeping both languages separate while still continuing to show bilingual norms; that is, they have compromise VOTs in Spanish and extra long VOTs in C reole E nglish (see App endix A). These results are interesting because these two speakers happen to work in call centers in Panama City where they use English all day long; however, they are also members of churches where only Spanish is used. It seems that because of this, thes e speakers are skilled at distinguishing Spanish from English in their daily usage; they have a domain where each one is used. They also consider English to be very important, which could be why they having generally longer VOTs than monolingual speakers: Interviewer: how important is Spanish to you? Do you think it's very important or just important or... Alicia: well right now the important language right now is English. It's bilingual but the basic right now is English. Spanish, it's good to speak it b ecause umm translation from English to Spanish is not the same and from Spanish to English is probably difficult. But the Spanish is necessary because the right now XXX is Spanish. Interviewer: do would you say very important or just important. Alicia: it's just important, not very important. Interviewer: How important is English to you? Alicia: Very important. (Alicia, Antillano English, 5/23/2010) Interviewer: and what about when you speak Spanish? Do people think you have an accent? Alberto: well, sometimes. Good n the job, people say: this gringo. Because they hear with my Spanish like it's a little
158 c ertain words I may say them wrong not because I didn't learn them before but you know, it's like, it's like, um, how should I say? It's like at the to say it right. So I may make a mistake and they may laugh at me and say: ah you a with U.S. I need to have a good English because if they monitoring me, they going to question my English, right. I need to have good English, ya. (Alberto, Antillano English, 5/30/2010) Another facto r worth mentioning is religion. One of the various differences found between West Indians and Mestizos in Panama is the affiliation with the Catholic Church. Mestizos normally associate with Catholicism, although recently, many are beginning to attend prot estant churches. West Indians, however, have deep roots in Protestantism, which was inherited from their Caribbean descendants. All but one of the bilingual speakers in this analysis attends a protestant church. Dora is the only Catholic in the group and a lso shows more of a preference for Spanish than the rest of speakers, which is reflected in her production of voiceless dentals. 5.4 Summary In this chapter, we offer a summary and discussion of the results revealed through several linear regression models fitted with mixed effects presented in Chapter 4. The analyses allowed us to confirm all of the hypotheses concerning mean VOT duration and conditioning factors that contribute to this acoustic cue in all four varieties. We also summarize and discuss the inter speaker and intra speaker comparisons. All hypotheses concerning change in the West Indian speech community were confirmed, except for Hypothesis 3. Despite the fact that bilinguals generally showed evidence of both internally motivated and contact induced change in their speech, contact induced change was not evident in their production of Creole English.
159 The intra speaker comparison confirmed the hypothesis that these early bilinguals would exhibit a separate phonetic system for each language but would also show convergence between Spanish and Creole English. Interestingly, as a whole, bilinguals seem to not only show convergence, but they converge towards monolingual Creole English norms, thus showing effects of possible identity marking. This beh avior has been explained with the proposal of the Variationist Speech Learning Model. These changes led us to ask whether there could be new emerging varieties among West Indians, which we could call Antillano Spanish and Antillano English. Qualitative da ta were extracted from the sociolinguistic interviews in order to explain why change is occurring in this language contact situation. While using the random effect results as a guide, we were able to show central themes, such as language attitudes, languag e loyalty and maintenance, and self identity. These themes appear to play an important role in this language contact situation and are manifested in the production of the voiceless dental plosive in the possibly new Antillano varieties.
160 CHAPTER 6 CONCLUSI ONS 6.1 Quantitative Analyses of Language Contact and Change The main goals of this study were to use variationist methodology to: 1) find evidence of change in a language contact situation and determine the possible source of such change, 2) analyze the s peech of bilinguals in terms of an acoustic cue referred Model. This particular method was employed because it allowed us to pinpoint the patterns of language variation and change in a speech community. In a usage based model of language, language is dynamic, massive, and highly redundant (Langacker, 2000). As speakers produce and perceive language, exemplars are stored in the mind and organized in relation to previous exempl ars. Thus, a variationist methodology is adequate for analyzing language because it sheds light on the probabilistic knowledge of speakers, which is normally spread throughout the community. Many studies have utilized this method to analyze the speech of monolingual communities; however, in order to fully understand the nature of linguistic variation and change, it is necessary that we examine the speech of multilingual communities. Analyses of bilinguals/multilinguals are paramount for any theory of lingu istic, since we know that multilingualism is a norm in most societies. Focusing on monolinguals only gives us part of the story. The present study was concerned with analyzing voice onset time (VOT) in the speech of Spanish Creole English bilinguals living in the Republic of Panama, particularly in Panama City.
161 6.1.1 Voice Onset Time VOT duration is an acoustic cue that is relevant when describing consonant stops (/p t k b d g/). It is defined as the time interval between the release of the stop and the on set of glottal vibration. The present analysis was concerned with the voiceless dental (or alveolar) plosive /t/. In Spanish, this sound exhibits a VOT that is categorized as short lag because voicing occurs shortly after the release of a stop. In English, the VOT of the voiceless alveolar plosive is long lag because voicing occurs well after the release of the stop. Therefore, VOT duration provided an ideal site to analyze potential effects of language contact in the West Indian speech community in Panama. 6.1.2 Internally Motivated Change vs. Contact induced Change When approaching the topic of language contact and change, it is paramount that we distinguish between the types of changes that can arise, that is, internally motivated or contact induced chan ge. In order to make this distinction in the West Indian speech community, the comparative variationist method established by Poplack & Tagliamonte (2001) was employed. This method has been further extended to contact situations in order to determine the s ource of linguistic change (e.g. Meyerhoff, 2009; Poplack & Level, 2010). The comparative variationist method is based on multivariate analyses that provide three baselines that can be used to compare languages and varieties: statistical significance, th e magnitude of effect of the factor groups, and the constraint hierarchy within the factor groups. This variable rule analysis is often carried out using GoldVarb X; however, it was decided to utilize Rbrul mainly because it allows for the analysis of nume ric data. Therefore, the statistical model employed in this study was a linear regression fitted with mixed effects. The order of the factors based on coefficients
162 provided by this model was taken to be the constraint hierarchy usually seen in GoldVarb. Th e constraint hierarchy is considered to be the underlying system of a language. In the comparative variationist method, in order to find evidence of change due to contact, we must compare bilingual varieties to forms of the languages that existed prior to the change. Therefore, in this analysis, I also included monolingual speakers of Spanish and Creole English who lived in Panama. I performed a series of comparisons of the constraint rankings between monolingual and bilingual varieties in terms of VOT dura tion. There were two types of comparisons: inter speaker and intra speaker. In the former, I compared the monolingual varieties to the bilingual varieties. In the latter, I compared bilingual Spanish to bilingual Creole English. 6.1.3 The Sociophonetic Stu dy In Chapter 4, the inter speaker comparisons showed that bilinguals exhibited evidence of both internal and contact induced change. Internal change was determined if the bilingual variety did not share the same patterns with any of the monolingual varie ties with regard to the constraint rankings. Contact induced change was determined if the bilingual variety exhibited the same patterns as the source variety, and not its monolingual counterpart. The comparative method revealed that when bilinguals spoke Spanish, internal change had occurred in terms of the effect of word class, and contact induced change had occurred in terms of the effect of preceding segment, position, and rate of speech. When bilinguals spoke Creole English, only internal change was no ticed in terms of the effect of vowel height. Therefore, the variationist method allowed us to see evidence of
163 both types of changes, and we were able to note exactly where these changes had taken place. The intra speaker comparison addressed claims made i Learning Model. In this model, Flege (1995a) finds evidence of cross language assimilation in which bilinguals have compromise VOT values. These are values that fall between the prototypical ranges of their two languages. According to Flege (1995b), this shows that bilinguals have a merged phonetic system. He also finds evidence of cross language dissimilation in which bilinguals have deflected VOT values; that is, they exhibit extra long values in one language in order to compensat e for longer values in the other. Flege (1995a) claims that this process shows evidence of two separate phonetic systems, and that it is most common among early bilinguals. That is, the earlier speakers learn a second language, the more they will keep them separate. as reading tasks in environments that do not allow for naturally occurring speech. Also, the findings are normally based on mean VOT values and standard deviation s. In the model, which entails an intra speaker comparison of the underlying systems of bilinguals. The results showed that the early Spanish Creole English bilinguals livi ng in Panama did indeed have a separate phonetic system for each language. However, the variationist methods used in the analysis revealed evidence of convergence that was not visible when looking solely at mean VOT values. Convergence was noticeable in th ree factors: rate of speech, preceding segment, and position. These results were
164 explained by proposing a Variationist Speech Learning Model that predicts that although bilinguals learn two languages from an early age, and in turn, establish separate phone tic systems, convergence could ensue at a later age. The VSLM also predicts that external influences could determine towards which language bilinguals will move. Interestingly, in this analysis, the bilingual varieties paralleled the monolingual Creole E nglish variety. These findings brought us to ask why bilinguals living in Panama would follow monolingual Creole English norms in a country where Spanish is the societally dominant language. In Chapter 5, I posed the question of whether there were new eme rging varieties, which I termed Antillano Spanish and Antillano English, in the West Indian speech community based on the results in Chapter 4. Considering social factors seen to affect languages in a contact situation, I further analyze the sociolinguisti c interviews in order to find qualitative explanations for linguistic change. I used the random effects results provided by the linear regression model in order to see the patterns of each bilingual speaker, and I explained the patterns in terms of factors such as language attitudes, language loyalty and maintenance, and social identity marking. I found that the speaker who showed positive attitudes towards Creole English and slight negative attitudes towards Spanish was most likely to show convergence in his languages which moved more towards monolingual Creole English norms. This speaker also identified strongly with his West Indian culture. The speakers who had equally positive attitudes towards Spanish and Creole English also showed convergence in thei r languages, but they moved more towards monolingual Spanish norms. These speakers were loyal in using Creole English but recognized that Spanish was very important in
165 their lives. They affirmed that they used both languages equally and in the same situati ons. Lastly, the speakers who seemed to keep both languages separate, producing Spanish with shorter VOT durations and Creole English with longer VOT durations, happened to separate the languages in their daily lives. They both work in places where they h ave to use English constantly, which aids in maintaining their skills in this language. Also, they are both members of a church where they always speak Spanish and are surrounded mainly by monolingual Spanish speakers. Therefore, the social factors conside red here seem to play an important role in the way the bilinguals use their languages in this particular language contact situation. 6.2 Theoretical Implications In usage based models of language, frequency and memory are the driving forces in language acq uisition, variation, and change. In terms of phonology and phonetics, sounds are stored in the mind through repetition, which establishes mental representations of sound categories (Pierrehumbert, 2001). This makes language a dynamic system and change in t his system is normally gradient. Likewise, linguistic knowledge is probabilistic, which means that language can be quantitatively analyzed (Weinreich et al., 1968; Tagliamonte, 2006), hence the variationist methodology. When language contact is involved, t his is just another factor that can push and pull on linguistic change and possibly bring about the emergence of new varieties. This emergence is gradual; thus, the most appropriate tool for analyzing linguistic outcomes of language contact is the variatio nist methodology. With this approach we a re able to pinpoint changes and take a snapshot of bilinguals' mental representation of their languages through statistical analyses. Also, since social experiences are stored in this
166 dynamic system as well change can be attributed not only to linguistic factors, but also to social ones. Some factors may override others in certain contact situations (Sankoff, 2002; Poplack & Levey, 2010), which is why the variationist methodology is important for these contexts. The results of this analysis suggest that these processes could be at work among bilingual West Indians in Panama, which could have important implications for literature concerned with language variation and change (LVC) and language contact and bilingualism. The findings of this study make important contributions to literature concerned with VOT duration. This acoustic property has an inherent variability; however, the production of consonant stops in terms of VOT is not haphazard, but rather systematic. LVC research is rooted in the idea of ordered variation, or structured heterogeneity. Thus, VOT duration should be examined within the framework of LVC. This type of examination includes all the possible factors that affect VOT duration into a single analysis. In this way, we are able to capture the full VOT system, as opposed to considering factors separately, since in real language use factors may interact in a phonetic system. As far as language contact and bilingualism is concerned, this study first makes a contribution in that it documents the speech of a community that has received very little attention in linguistics. The varieties of Creole English in Panama have been analyzed; however, no one has conducted a true variationist study of both the Spanish a nd Creole English of West Indian Panamanian. Furthermore, the analysis highlights the importance of distinguishing internally motivated change from contact induced change. Studies concerned with determining that contact induced change has occurred in a
167 spe ech community must be able to determine that any change reported was less likely to have occurred in a situation where there was no contact (Thomason, 2001; Poplack & Levey, 2010). According to Poplack & Tagliamonte (2001:3), an assessment of change requir es that we reference an earlier stage of the language. Since historical data for certain languages are difficult to obtain, which was the case for the present study, we must find a way to solve this matter for an efficient study of language change. The mos t appropriate solution is to incorporate precursor or surrogate varieties from local regions to find evidence that changes such as simplification and overgeneralization are not due to language drift, that is, internal processes. This has been implemented i n various LVC studies, such as in Poplack & Tagliamonte (2001) to argue the origin of AAVE and in Blondeau & Nagy (2008) to examine the variation of subordinate clause marking in French and English among Anglophone Montrealers. Secondly, this study could have consequences for contact research related to the emergence of ethnolects, or ethnic language varieties. Ethnolects have been defined as backgrounds attempt to speak the to this definition, dialects, such as Chicano English and Canadian Italian could be identified as ethnic varieties. These ethnolects have been viewed from two different perspectives: the shift perspective a nd the multidimensional perspective. Muysken (2010:8) describes that from the shift perspective ethnolects emerge because the ethnic group approximates their speech to the dominant national language of the region in which they live. The group acquires the L2, but transforms it by incorporating characteristics of the L1, which are associated with their ethnic identity.
168 In the multidimensional perspective, there are additional factors, the mutual convergence between the L1 and the L2 and simplification (Muys ken, 2010:8). In this perspective, the minority language is maintained but is also transformed, thereby causing an emergence of an ethnic variety of the mother tongue. The multidimensional perspective, according to Muysken (2010), views the ethnolect as a combination of several forms of language: a second language, a street language, a mixed language, and a transplanted variety. Additionally, they can have a combination of properties that make it distinct from these varieties, such as stability, signal of i dentity, specific to a particular ethnic group, particular pronunciation or intonation, non standard use of endings from the national language, etc. (Muysken, 2010:16). Interestingly, the combination of these properties and the varieties mentioned could be easily applied to the possible emerging varieties among bilingual West Indians in Panama. A further analysis of this speech community could serve as an important element to the roots of ethnolects 1 A third line of research to which the present analysis m akes a contribution is that of second language acquisition, particularly, the Speech Learning Model (Flege, 1995a). We have discussed throughout the project that the SLM is based on a merger hypothesis presented by Flege (1987). It is posited in this hypot hesis that older children and adults learning L2 phones that are similar to phones of their L1 would not produce equivalence classification that causes late bilinguals to ignore p honetic information 1 Roots of Ethnolects is a current research project that involves experimental comparative analyses of ethnolects in order to understand how they emerge, stabilize, and spread. The varieties being analyzed currently are young ethnolects of Dutch: http://www.rootsofethnolects.nl/web/
169 necessary to produce the L2 sound in a native like manner. The late bilinguals attempt to categorize, or map, new exemplars onto already existing exemplars from the L1, which causes convergence or cross language assimilation. Early bili nguals, in comparison, tend to not experience convergence of their two languages because L2 sound categories are established independently of the L1 due to the fact that the L1 is not yet fully developed. Therefore, early bilinguals normally exhibit cross language dissimilation in their speech (Flege, 1995b). Studies concerned with VOT duration among early and late bilinguals (e.g. Caramazza et al., 1973; Flege et al., 1996; Macleod & Stoel Gammon, 2008 ) have tion hypothesis, thus supporting the SLM. The present analysis also finds evidence of dissimilation among early bilingual West Indians in Panama. Previous studies based evidence of cross language dissimilation on mean VOT duration. The present analysis did the same; however, here the examination of bilingual VOT duration was taken further using Labovian variationist methodology. This methodology made signs of convergence visible in early bilinguals despite the claims made about dissimilation or divergence i n the speech of people who acquire the L2 at an early age. It appears that the equivalence classification mechanism cause early bilinguals to show convergence in their phonetic systems at a later age. Therefore, what is proposed here is a shift towards a V ariationist Speech Learning Model (VSLM) to capture convergence of bilingual speech that is otherwise invisible. Lastly, it is important that we highlight a contribution this analysis could make to another line of research regarding language contact, creol istics. In Chapter 2, we discussed English VOT values that have been established by measuring consonant
170 stops produced by speakers of North American English varieties. It was unsure how the Creole English participants living in Panama City, and especially those from Bastimentos, would compare to North American English speakers in terms of VOT duration. It has been noted in creolistics that phonetics and phonology in creole languages differ from their superstrates due to substratum interferences. Plag (2009: 5) we would expect in pidgin/creole terms the emergence of a variety that is at least phonetically, if not phonologically, clearly distinct from the superstrate varieties involved in the contact. peakers in Bastimentos, at least, should show a significantly different mean VOT value for the voiceless dental and the underlying patterning of the conditioning factors should differ from other non creole varieties. However, the analysis revealed that thi s variety did not appear to differ from other standard varieties of English. This dialect still exhibited a similar value to that of an English variety with which it did not have contact in spite of its coexistence with a non lexifier language, Spanish. 2 A lso, Panama City Creole English and Bastimentos Creole English had similar underlying constraints as the varieties reported on in this paper. We could conclude from these results that creoles are cases of regular language change and that they are offshoots of British origin dialects spoken by European colonizers. This analysis adds to the ongoing debate concerning theories of creole genesis (see Section 5.2). 2 See discussion of monolingual bilingual con tinuum in Chapter 2. In order to be sure of any claims that BCE does not differ from its superstrate in terms of VOT, this lexifier language must first be identified and then VOT measurements need to be collected. Additionally, a comparison among the under lying systems of Spanish (spoken by monolinguals and BCE speakers), BCE and the superstrate variety must be carried out (see Poplack & Levey (2010) and Chapter 1 for details on this method).
171 6.3 Limitations and Further Research Although this analysis of the phonetic systems of bilinguals was able to show and explain linguistic variation and change, there are still some limitations left to tackle in further research of the West Indian speech community in Panama. First, a larger sample of participants is needed in order to make more general statements about language change. The amount of data submitted for the statistical analysis was sufficient; however, it has been noted in other studies that frequency of a form can affect constraint hierarchies (Alim, 2004). Therefore, a greater amount of data could reveal more robust findings of the outcomes of language contact. Second, many studies concerned with language variation and change tend to include social factors into the statistical analysis to see if there are any significant effects of thes e constraints. In the present analysis, a quantitative analysis of social factors was not possible due to the small number of participants. The solution to this was to use the random effects results to observe individual behavior and corroborate this behav ior with qualitative data extracted from the sociolinguistic interviews, a language background questionnaire, and a language attitude survey. In further analyses, I hope to include additional bilingual speakers in order to analyze the social factors statis tically. Third, Poplack & Levey (2010) state that in order to determine that contact induced change has occurred in a speech community, one must include in the comparative method a precursor or a surrogate variety. In this analysis, I incorporate data from a so called monolingual variety of Creole English spoken in Bastimentos. However, speakers in this area of Panama still have access to Spanish in school and outside the Bastimentos Island. Although these are dominant speakers of Creole
172 English (see monoli ngual bilingual continuum in Chapter 3), their access to Spanish could be problematic for this study. In further analyses of language contact, it would be beneficial to collect data from a truly monolingual group of Creole English. Many of the West Indians who arrived in Panama in the mid 1800s were from Jamaica or Barbados; thus, Jamaican Creole English, or even Barbadian English, could be more adequate varieties to use as monolingual baselines. Lastly, in this study I chose to focus on one particular sound, the voiceless dental plosive, first, because it was the most frequent sound in the sociolinguistic interviews. This is paramount for variationist work because an adequate amount of data is needed for a statistical analysis. Secondly, from my experie nces in the speech community, it appears that the aspiration of this sound has sociolinguistic meaning, or at least, marks a speaker as being West Indian. When monolingual speakers of Spanish imitate the speech of West Indians, the voiceless dental is the main sound they use to mark the difference. However, in order to fully understand linguistic change in this speech community, it is important to examine other speech sounds. Investigating other sounds in a single contact situation could allow us to determi ne if all sounds are affected in a similar way at the same time, or if different sounds are vulnerable to different motives for change. This is important for usage based models of phonology because if language change is gradual then we should see that not all sounds follow similar patterns of change in a contact situation. 6.4 Conclusion This analysis has shed light on the dynamic nature of languages in a contact situation. I have employed variationist methodology, specifically, the comparative variationi st method, to analyze and explain certain linguistic outcomes in the West
173 Indian speech community in Panama. It is the first to analyze voice onset time with this approach. Therefore, it not only contribute s to the already existent literature of Spanish di alectology and creolistics but it also explain s the permeability of phonetic systems in a contact situation. Moreover, it add s to the debate on VOT duration among bilinguals and attempt s to support Flege variation theory, a Variationist Speech Learning Model.
174 APPENDIX A MEAN VOT VALUES OF I NDIVIDUAL SPEAKERS Group Mean VOT (/t/) in msec. Standard deviation Range Monolingual Spanish 15.24 7.58 0 52 PART 3 12.69 6.18 5 39 PART 6 15.08 4.59 7 29 PART 11 23.66 8.82 11 52 PART 12 11.09 5.07 2 30 PART 15 13.72 5.37 0 33 Monolingual Creole English 28.43 13.48 0 82 PART 5 31.28 14.94 0 70 PART 17 30.47 12.71 6 86 PART 18 22.34 13.32 0 68 PART 19 23.44 9.66 0 6 3 PART 20 35.31 13.75 0 83 Bilingual Spanish 21.15 8.95 2 87 PART 1 19.29 7.84 2 46 PART 2 20.49 8.10 4 47 PART 4 28.23 8.92 8 87 PART 7 16.38 7.16 6 63 PART 9 20.62 7.54 10 51 Bilingual Creole English 34.20 18.33 0 168 PART 1 34.18 18.84 3 113 PART 2 34.28 16.50 0 123 PART 4 40.63 15.13 17 113 PART 7 24.82 24.16 3 168 PART 9 33.58 15.86 10 108 p <0.05
175 APPENDIX B LANGUAGE BACKGROUND QUESTIONNAIRE Name: ________ Age: _________ Sex: ____ Male ____ Female Area of birth: _________ How m any years did you spend there? ____ Area of residence: _______ How many years have you lived there? ____ What is you highest level of education? ____ Primary ____ Secondary ____ University ____ Post grad Occupation: Languages that you speak: ___ ________ Languages that your parents/grandparents speak/spoke: __________ Where did you learn Spanish? ___ At home ___ In church ___ In school At what age did you learn Spanish? _____ Where did you learn English? ____ At home ____ In church ___ In school At what age did you learn English? _____ How often do you travel to an English speaking country? ____ Never ____ Once in a while (for vacation) ____ lived there for more than three years
176 APPENDIX C LANGUAGE ATTITUDES S URVEY How often do you speak Spanish? ____ Everyday ____ Once in a while ____ Never How often do you speak English? ____ Everyday ____ Once in a while ____ Never Where do you speak Spanish? Mark all that apply. ____ At work ____ At church ____ At home ___ In public areas (supermarket, school, government areas, etc.) Where do you speak English? Mark all that apply. ____ At work ____ At church ____ At home ____ in public areas (supermarket, school, government areas, etc.) Who do you speak Spanish t o? Mark all that apply. ____ to people at work ____ to friends ____ to neighbors ____ to children ____ to parents/grandparents ____ to strangers Who do you speak English to? Mark all that apply. ____ to people at work ____ to friends ____ to ne ighbors ____ to children ____ to parents/grandparents ____ to strangers How important is Spanish to you? ____ very important ____ important ____ more or less important ____ not important ____ definitely not important How well do you speak Spanish? __ __ Perfectly ____ Very well ____ Not so well ____ Not well at all How important is English to you? ____ very important ____ important ____ not so important ____ definitely not important
177 How well do you speak English? ____ Perfectly ____ Very well ____ Not so well ____ Not well at all Do you think English should be taught in school? ____ Yes ____ No How would you describe the English that you speak/spoke with your grandparents? ____ Standard/Proper English ____ Broken/Creole English
178 LI ST OF REFERENCES Abramson, A rthur S. & L eigh Lisker (1973) Voice timing perception in Spanish word initial stops. Journal of Phonetics 1 : 1 8. Aceto, Michael. (1995). Variation in a secret Creole language of Panama. Language in Society 24:537 560. Aceto, Michael. (1998). A new creole future tense marker emerges in the Panamanian West Indies. American Speech 73 :29 43. Aceto, Michael. (2001). The English creole of Panama. Quaderni del centro rsit di Sienna Siena: CISAI. 2:7 21. Aceto, Michael. (2002). Ethnic personal names and multiple identities in Anglophone Caribbean speech communities in Latin America. Language in Society 31:577 608. Adamson, Douglas. ( 1988 ) Variation theory and SLA W ashington, DC: Georgetown University Press. Adamson, D ouglas, Bonnie Fonseca Greber, Kuniyoshi Kataoka, Vincent Scardino, & Shoji Takano. (1996). Tense Marking in the English o f Spanish speaking Adolescents. In R. Bayley & D. Preston (eds.), Second Languag e Acquisition and Linguistic Variation pp. 121 134. Amsterdam: John Benjamins. Akers, Glen A. ( 1981 ). Phonological Variation in the Jamaican Continuum Ann Arbor, MI: Karoma. Alim, H. Samy. (2004). You know my steez: An ethnographic and sociolinguistic st udy of styleshifting in a Black American speech community Durham, NC: Duke University Press. Alleyne, Mervyn C. (1980). Comparative Afro American. Ann Arbor: Karoma. Alvarado de Ricord, Elsie. (1982). The impact of English in Panama. Word 33:92 108. And ersen, Roger W. (1990). Papiamentu tense aspect, with special attention to discourse. In John Singler (ed.), Pidgin and Creole Tense Mood Aspect Systems pp. 59 96 Amsterdam: John Benjamins Publishing Company Appel, Rene, & Pieter Muysken. ( 1987 ). Language contact and bilingualism. New York: Edward Arnold. Arends, Jacques. (1993). Towards a gradualist model of creolization. In Francis Bryne & John Holm (eds.), Atlantic Meets Pacific: A Global View of Pidginization and Creolization pp. 371 380. Amsterdam: J ohn Benjamins Publishing Company.
179 Arends, Jacques. (1994). The socio historical background of creoles. In Jacques Arends, Pieter Muysken, & Norval Smith (eds.), Pidgins and Creoles : An Introduction pp. 15 24. Amsterdam: John Benjamins Publishing Company Arends, Jacques, & Adrienne Bruyn. (1994). Gradualist and developmental hypotheses. In Jacques Arends, Pieter Muysken, & Norval Smith (eds.), Pidgin and Creoles: An introduction pp. 111 120. Amsterdam: John Benjamins Publishing Company. Arends, Jacques Silvia Kouwenberg & Norval Smith. (1994). Theories focusing on the non European input. In Jacques Arends, Pieter Muysken, & Norval Smith (eds.), Pidgin and Creoles: An introduction pp. 99 109. Amsterdam: John Benjamins Publishing Company. Baayen, R. H. D.J. Davidson, & D.M. Bates. (2008). Mixed effects modeling with crossed random effects for subjects and items. Journal of Memory and Language 59: 390 412. Bakker, Peter. (1994). Pidgins. In Jacques Arends, Pieter Muysken, & Norval Smith (eds.), Pidgins a nd Creoles pp. 24 39. Amsterdam: John Benjamins Publishing Company. Bayley, Robert. ( 1994 ) Interlanguage variation and the quantitative paradigm: Past tense marking in Chinese English. In Elaine Tarone, Susan M. Gass, & Andrew Cohen (eds.), Research me thodology in second language acquisition pp. 157 181. Hillsdale, NJ: Lawrence Erlbaum. Bayley, Robert. (2002). The Quantitative paradigm. In J. K. Chambers, Peter Trudgill, & Natalie Schilling Estes (eds.) Handbook of Language Variation and Change pp. 11 7 141 Oxford: Blackwell. Bickerton, Derek. ( 1973 ) The nature of a creole continuum. Language 49:640 669. Bickerton, Derek. (1980). Decreolisation and the creole continuum. In Valdman & Highfield (eds.) Theoretical Orientations in Creole Studies pp. 109 128 N ew York: Academic Press Bickerton, Derek. (1988). Creole languages and the bioprogram. In Newmeyer (ed.), Linguistics : The Cambridge survey pp. 267 284 Cambridge : Cambridge University Press. Blondeau, Hlne, & Naomi Nagy. ( 2008 ) Subordinate cla use marking in Montreal Anglophone French and English. In M. Meyerhoff & N. Nagy (eds.), Social Lives in Language. Sociolinguistics and Multilingual Speech Communities. Celebrating the work of Gillian Sankoff pp. 273 313. Philadelphia: John Benjamins. Bl oomfield, Leonard. (1933). Language London: Allen and Unwin.
180 Boersma, P aul, & D avid Weenink. (2006) Praat: doing phonetics by computer (Version 4.4.24) Retrieved from http://www.praat.org/ Boissevain, Jeremy. (1974). Friends of friends: Networks, manipu lators, and coalitions. Oxford: Blackwell. Boretzky, Norbert. (1983). Kreolsprachen, Substrate und Sprachwandel WiesbadenL Otto Harrassowitz. Borzone de Manrique, Ana M. ( 1980 ) Manual de fontica acstica Buenos Aires: Hachette. Broce, Marlene & Ren a Torres Cacoullos. (2002). Dialectologa urbana rural: la estratificacin social de (r) y (l) en Cocl, Panam Hispania 85:342 354. Brown, Earl. (2009). A usage based account of syllable and word final /s/ reduction in four dialects of Spanish. New Mex ico: The University of New Mexico, Ph.D. dissertation. Brown, Robert A. (1976). A sociolinguistic description of the Spanish of Panama City. Washington D.C.: Georgetown University, Ph.D. dissertation. Bullock, Barbara. E., Almeida J. Toribio, Veronica Gon zlez, & Amanda Dalola. ( 2006 ) Language dominance and performance outcomes in bilingual pronunciation. In M ary hristine Shea, J ohn Archibald (eds.) Proceedings of the 8 th Generative Approaches to Second Language Acquisition: The Banff Confe rence pp. 9 16. Somerville, MA: Cascadilla. Bybee, Joan L. (1994). A view of phonology from a cognitive and functional perspective. Cognitive Linguistics 5 285 305. Bybee, J oan L. ( 2000 ) The phonology of the lexicon: evidence from lexical diffus ion. In Michael Barlow & Suzanne Kemmer (eds.), Usage based models of language pp. 65 85. Stanford, CA US: CSLI Publications. Bybee, Joan. ( 2001 ) Phonology and language use. Cambridge: Cambridge University Press. Bybee, J oan. (2002) Word frequency and context o f use in the lexical diffusion of phonetically conditioned sound change. Language Variation and Change 14 : 261 290. Bybee, Joan. (2006). From usage to grammar: The mind's response to repetition. Language 82 : 711 733. Bybee, Joan. L., & Paul. J Hopper (eds.) (2001). Frequency and the emergence of linguistic structure (T ypological studies in language 45). Philadelphia: John Benjamins Publishing Company.
181 Bybee, Joan, & Rena Torres Cacoullos. ( 2008 ) Phonological and grammatical variation in exemplar models. Studies in Hispanic and Lusophone Linguistics 1: 399 413. Canfield, D. L incoln. ( 1981 ) Spanish Pronunciation in the Americas Chicago: University of Chicago Press. Caramazza, A., G. H. Yeni Komshian, E. B. Zurif, & E. Carbone. (1973). The acquisition of a new phonological contrast: The case of stop consonants in French English bilingua ls. Journal of the Acoustical Society of America 54 : 421 428. Carden, Guy, & William A. Stewart. (1988). Binding theory, bioprogram, and creolization: Evidence from Haitian Cre ole. Journal of Pidgin and Creole Languages 3:1 67. Castaeda, Mara Luisa. (1986). El V.O.T. de las ocl usivas sordas y sonoras espaolas. Estudios de fontica experimental 2.93 110. Cedergren, Henrietta. (1973). The Interplay of Social and Linguistic Fac tors in Panama. Ithaca: Cornell University, Ph.D. dissertation. Chaudenson, Robert. (1979). Les crole franais Evreux: Nathan. Chaudenson, Robert. (1992). Des les, des homes, des langues Paris Chomsky, Noam, & Morris Halle. ( 1968 ) The sound pattern of English. New York: Harper and Row. Clark, Lynn, & Kevin Watson. (2011). Testing claims of a usage based phonology with Liverpool English t to r English Language and Linguistics 15:523 547. Coelho, F rancisco A. (1880). Os dialectos romn icos ou neolatinos na Africa, Asia e Amrica. Bolletim da Sociedade de Geografia de Lisboa 2:129 196. Comrie, Bernard. (1976). Aspect Cambridge: Cambridge University Press. Comrie, Bernard. (1985). Tense Cambridge: Cambridge University Press. Conniff, Michael L. (1985). Black labor on a white canal: Panama, 1904 1981 Pitt La tin American series. Pittsburgh: University of Pittsburgh Press. Danesi, M arcel (1985). A glossary of lectal terms for the description of language variation. Language Problems and Language Planning 89 : 115 124. de Banville, Marc. (2005). de Rooij, Vincent. (1994). Variation. In Jacques Arends, Pieter Muysken, & Norval Smith (eds.), Pidgins and Creoles : An Introduction pp. 53 64. Amsterdam: John Benjamins Publishing Company.
182 DeGraff, Michel. (1992). Creole grammars and acquisition of syntax: the case of Haitian. University of Pennsylvania, Ph.D. dissertation. DeGraff, Michel. (2001). On the origin of Creo les: A cartesian critique of neo darwinian linguistics. Linguistics Typology 5:213 310. DeGraff, Michel. (2003). Againt creole exceptionalism. Language 79:391 410. den Besten, Hans, Pieter Muysken, & Norval Smith. (1994). Theories focusing on the Europea n unit In Jacques Arends, Pieter Muysken, & Norval Smith (eds.), Pidgins and Creoles: An Introduction pp. 87 98 Amsterdam: John Benjamins Publishing Company. Dniz, Magnolia Troya. ( 2005 ) El VOT de las oclusivas sordas en la norma culta de Las Palmas d e Gran Canaria. Boletn de Lingstica 24:92 107. Diez Castillo, L uis ( 1981 ) Panama: Imprenta Julio Mercado Rudas. Docherty, G erard. J. (1992). The timing of voicing in English obstruents. Berlin: Foris Drager, Katie and Jennifer Hay. ( 2012 ). Exploiting random intercepts: Two ca se studies in sociophonetics. Language Variation and Change 24 : 59 78. Eckert, Penelope. ( 2000 ) Linguistic variation as social practice Oxford: Blackwell. Eckert, Penelope. ( 2008 ) Variation and t he indexical field. Journal of Sociolinguistics 12.453 76. Eckman, Fred. (1977). Markedness and the contrastive analysis hypothesis. Language Learning 27:315 330. Eilers, R ebecca E., D. Kimbrough Oller, & Carmen. R. Benito Garcia. (1984) The acquisition of voicing contrasts in Spanish and English learning infants and children: A longitudinal study. Journal of Child Language 11 : 313 336. Fishman, Joshua A. (1964). Language maintenance and language shift as a field of i nquiry. Linguistics 9 : 32 70. Fishman, J oshua A. ( 1967 ). Bilingualism w ith and without diglossia; di glossia with and w ithout b ilingu alism. Journal of Social Issues 23 : 29 38. language: evidence for the effect of equivalence classification. Journal of Phonetics 15:47 65.
183 Fle ge, James E. (1995a). Second language speech learning: Theory, findings, and problems. In Winifred Strange (ed.), Speech perception and linguistic experience: Issues in cross language research pp. 233 277. Timonium: York Press. Flege, James E. (1995b). Th e phonetic study of bilingualism. In G. Bloothooft, V. Hazan, D. Huber & J. Llisterri (eds.), European studies in phonetics and speech communication pp. 98 103. Utrecht: OTS Publications. Flege, James E. Murray J. Munro, & Ian R. A. MacKay. (1996). Fa ctors affecting the production of consonants in a second language. In Robert Bayley & Dennis Preston (eds.), Second Language Acquisition and Linguistic Variation pp. 47 74 Amsterdam: John Benjamins. Flege, James E. (2007). Language contact in bilingualis m: Phonetic system interactions. In J. Cole, & J. Hualde (eds.), Laboratory phonology 9 pp. 353 381. Berlin: Moutonde Gruyter. Flores Ferrn, N ydia ( 2004 ) Spanish subject personal pronoun use in New York City Puerto Ricans: Can we rest the case of Engli sh contact? Language Variation and Change 16: 49 73. Fought, Carmen. (1999). A majority sound change in a minority community: / u/ fronting in Chicano English. Journal of Sociolinguistics 3 : 5 23. Fowler, Carol. A., Valery Sramko, David J. Ostry, Sarah A. Row land, & Pierre Halle. (2008). Cross language phonetic influences on the speech of French English bilinguals. Journal of Phonetics 36 : 649 663. Garrett, Paul. 1999. Language socialization, convergence, and shift in St. Lucia, West Indies. New York City: New York University, Ph.D. Dissertation. Geeslin, Kimberly, & Pedro Guijarro Fuentes. ( 2006 ) A Longitudinal Study of Copula Choice: Following Development in Variable Structures. In N. Sagarra & J. Toribio (eds.) Selected Proceedings of the 9th Hispanic Ling uistics Symposium pp. 144 156. Somerville, MA: Cascadilla Proceedings Project. Guion, Susan. (2003).The vowel systems of Quichua Spanish bilinguals. Phonetica 60:98 128. Givn, Talmy. (1995). Functionalism and grammar. Philadelphia: John Benjamins Publish ing Company. Gordon, Matthew. ( 2007 ) Techniques of analysis I: Phonological variation. In Car men Llamas, Louise Mullany, & Peter Stockwell (eds.) The routledge companion to sociolingustics pp. 19 27. New York: Routledge.
184 Gregory, Michelle L., William D. Raymond, Alan Bell, Eric Fosler Lussier, & Daniel Jurafsky. (1999). The effects of collocational strength and contextual predictability in lexical production. CLS 35:151 169. Grosjean, F ranois (1982). Life with two languages: An introduction to bilingu alism. Cambridge, Mass: Harvard University Press. Grosjean, Franois. ( 2001 ) In Janet L. Nicol (ed.) One mind, two languages: Bilingual language processing pp. 1 22. Oxford: Blackwell. Guerrn Montero, C arla M. (2002) Li ke an alien in we own land: International tourism, gender and identity in Afro Antillean Panama Oregon : Unive rsity of Oregon, Ph.D. dissertation. Guitart, Jorge M. ( 2004 ) Sonido y sentido Washington D.C.: Georgetown University Press. Hall, Robert A., Jr (1966). Pidgin and creole languages Ithaca: Cornell University Press. Hall Lew, Lauren. ( 2009 ) Ethnicity and Phonetic Variation in a San Francisco Neighborhood Stanford: Stanford University Ph.D. d issertation. Hancock, Ian F. (1986). The domestic hyp othesis, diffusion and componentiality: an account of Atlantic Anglophone creole origins. In Pieter Muysken & Norval Smith (eds.), Substrata versus universals in creole genesis pp. 71 102. Amsterdam, John Benjamins. Hay, Jennifer, & Katie D rager. ( 2007 ) Sociophonetics. The Annual Review of Anthropology 36: 89 104. Herrera, Juana. (1997). Es tudio acstico de /p, t, c k/ y /b, d, y, g/ en Gran Canaria. In Man uel Almeida & Josefa Dorta (eds.), Contribuciones al estudio de la ling stica hisp nica pp. 73 86. Tenerife: Montesinos. Herzfeld, Anita. (1983). The Creoles of Costa Rica and Panama I n John Holm (ed.), Central American English pp. 131 156 Heidelberg: Julius Groos Verlag. Hesseling, D. C. (1897). Dutch in South Africa. De Gids Jaargang 61:138 16 2. Hesseling. Dirk C. (1933). Hoe ontstond de eigenaardige vorm van het Kreools? Neophilologus 18:209 215. Holm, John. (1988 ). Pidgins and creoles Vol I. Cambridge: Cambridge University Press.
185 Holmes, Janet. (1992). An Introduction to sociolinguistics London and New York: Longman. Jacques, Benoit, & Jorge A. Gurlekian. (1992). Comparative study of four acoustic parameters of stops in Buenos Aires Spanish and Montreal French. [Etude compare de quatre paramtres acoustiques des occlusives en espagnol de Buenos Aires et en franais de Montral] Revue Qubcoise De Linguistique 22: 257 272. Johnson, Daniel Ezra. (2009). Getting off the GoldVarb standard: Introducing Rbrul for mixed effects variable rule analysis. Language and Linguistics Compass 3:359 383. J ohnson, Keith. (1997). Acoustic and auditory phonetics Cambridge, Mass: Blackwell Publishers. Kessinger, R.H. and Blumstein, S. E. (1998). Effects of speaking rate on voice onset time and vowel production: Some implications for perception studies, Journal of Phonetics 26:117 128. Labov, William. ( 1966 ) Social stratification of English in New York City Washington D.C.: Center for Applied Linguistics. Labov, William. (1969). Contraction, deletion, and inherent variability of the English copula. Language 4 5:715 762. Labov, William. ( 1972 ) Sociolinguistic Patterns. Philadelphia: The University of Pennsylvania Press. Labov, Willian. (1994). Principles of linguistic change: Volume I: Internal Factors Oxford: Blackwell. Ladefoged, Peter. (2003). Phonetic da ta analysis: An introduction to fieldwork and instrumental techniques Malden, MA: Blackwell Pub. Langacker, R onald W. (2000). A dynamic usage based model. In M ichael Barlow & S uzanne Kemmer (eds.), Usage based models of language pp. 1 63. Stanford, C alif : CSLI Publications, Center for the Study of Language and Information. Lefebvre, Claire. (1998). Creole genesis and the acquisition of gram m ar: the case of Haitian Creole. Cambridge: Cambridge University Press. Lemaitre, Eduardo. (1972). Popular. Lipski, John. (1998). Latin American Spanish: creolization and the African connection. PALARA 2:54 78.
186 Lipski, John M. (2009). Searching for the origins of Uruguayan Fronterizo dialects: Radical code Journal of Portuguese Linguistics 8:3 44. Lisker, L eigh & Arthur Abramson ( 1964 ) A cross language study of voicing in initial stops: Acoustical measurements. In R.J. Baken & R.G. Da niloff (eds.), Readings in the c linical spectrography of speech pp. 384 422. San Diego, CA: Singular Publishing Group, Inc. Lisker, L eigh & Arthur Abramson ( 1967 ) Some effects of context on voice onset time in English stops. Language and Speech 10 : 1 28 Ldtke, Helmut. (1995). On th e origin of Middle and Modern English. In Jacek Fisiak (ed.) Linguistic Change under Contact Conditions pp. 51 53. Berlin: Mouton de Gruyter. MacLeod, A ndrea & C arol Stoel Gammon (2009) The use of voice onset time by early bilinguals to distinguish ho morganic stops in Canadian English and Canadian French. Applied Psycholinguistics 30 :53 77. MacPherson, Seonaigh. (2005). Negotiating language contact and identity change in developing Tibetan English bilingualism. TESOL Quarterly 39:585 607. Magloire, Jo el, & Kerry P. Green. (1999). A cross language comparison of speaking rate effects on the production of voice onset time in English and Spanish. Phonetica 56 :158 185. Matras, Yaron. ( 2009 ) Language contact Cambridge: Cambridge University Press. McWhorter John. (1992). Substratal influence in Saramaccan serial verb constructions. Journal of Pidgin and Creole Languages 7:1 53. McWhorter, John. (1998). Identifying the creole prototype: Vindicating a typological class. Language 74:788 818. Meyerhoff, Miria m. ( 2009 ) Replication, transfer and calquing: Using variation as a tool in the study of language contact. Language Variation and Change 21: 1 21. Miller, Joanne L. (1994). On the internal structure of phonetic categories: A progress report. Cognition 50 : 2 71 285. Miller, Joanne L. (2001). Mapping from acoustic signal to phonetic category: Internal category structure, context eff ects and speeded categorization. Language and Cognitive Processes 16 :683 690. Milroy, James. (1992). Linguistic variation and chang e Oxford: Blackwell.
187 Milroy, Lesley. ( 1987 ) Observing and analyzing natural language: A critical account of sociolinguistic method. Oxford: Basil Blackwell. Milroy, Lesley, & Matthew Gordon. (2003). Sociolinguistics: Methods and interpretation. Oxford: Blackwell. Moravcsik, Edith, & Jessica Wirth. (1986). Markedness: An overview. In Fred R. Eckman, Edith A. Moravcsik, & Jessica Wirth (eds.), Markedness pp. 1 12. New York: Plenum Press. Morris, Richard J., Christopher R. McCrea, & Kaileen D. Herring. ( 2008). Voice onset time differences between adult males and females: Isolated syllables. Journal of Phonetics 36 : 308 317. Mufwene, Salikoko. (1993). Africanisms in Afro American language varieties Athens: University of Georgia Press. Muwfene, Salikoko. ( 1996). The Founder principle in creole genesis. Diachronica 13:83 134. Mufwene, Salikoko. (2001). The ecology of language evolution Cambridge: Cambridge University Press. Munday, Pilar. (2001). An acoustic study of Spanish voiceless stops in three speak ing sty les. New York: New York University, Ph.D. dissertation. Muysken Pieter. (2010) Ethnolects as a multidimensional phenomenon. In Muriel Norde, Bob de Jonge, & Cornelius Hasselblatt (eds.), Language Contact: New perspectives pp. 7 25. Amsterdam: Jo hn Benjamins Publishing Company. Muysken, Pieter, & Norval Smith. (1994). The study of pidgin and creole languages. In Jacques Arends, Pieter Muysken, & Norval Smith (eds.), Pidgins and Creoles : An Introduction pp. 8 14. Amsterdam: John Benjamins Publish ing Company. Nathan, Geoffrey. (1998). What functionalists can learn from formalists in phonology [What formalists can learn from formalists in phonology]. In Michael Darnell, Edith Moravcsik, Fredrick Newmeyer, Michael Noonan, & Kathleen Wheatley (eds.), Functionalism and Formalism in Linguistics: Volume 1 pp. 305 328. Amsterdam: John Benjamins Publishing Company. Otheguy, Ricardo, & Ana Celia Zentella. ( 2007 ) Apuntes preliminares sobre el contacto lingstico y dialectal en el uso pronominal del el es paol en Nueva York. In Kim Potowski and Richard Cameron (eds.), Spanish in contact: Policy, social and linguistic inquiries pp. 275 296. Philadelphia, PA: John Benjamins. Otheguy, Ricardo, Ana Celia Zentella, & David Livert. ( 2007 ) Language and dialect contact in Spanish in New York: Towards the formation of a speech community. Language 83: 1 33.
1 88 Orozco, Rafael & Gregory Guy. (2008). El uso variable de los pronombres sujetos: qu pasa en la costa Caribe colombiana? In Maurice Westmoreland & Juan Antonio Thomas (eds.) Selected proceedings of the 4 th workshop on Spanish sociolinguistics pp. 70 80. Somerville, MA: Cascadilla Proceedings Project. Pagliuca, William, & Richard Mowrey. ( 1987 ) Articulatory evolution. In Anna Giacalone Ramat, Onofrio Carruba & Giuliano Benini (eds.) Papers from the 7th International Conference on Historical Linguistics pp. 459 472. Amsterdam: John Benjamins. Pavlenko, Aneta, & Adrian Blackledge (eds.). (2004). Negotiation of identities in multilingual settings Clevedon: Multi lingual Matters. Pierrehumbert, Janet B. (2001). Exemplar dynamics: Word frequency, lenition and contrast. In Joan Bybee, & Paul Hopper (eds.), Frequency and the emergence of linguistic structure pp. 137 157. Amsterdam: John Benjamins. Plag, Ingo. (2009) Creoles as interlanguages: phonology. Journal of Pidgin s and Creole Languages 24:119 138. Poch Olive, D. (1985). Acoustic characterization of spanish occlusives: The problem of VOT. [Caractrisatio n acoustique des occlusives de l'espagnol: le problme d u V.O.T.] Revue de Phontique Applique 77 : 477 490. Poplack, Shana. ( 1980 ) espaol: toward a typology of code switching. Linguistics 18: 581 618. Poplack, Shana. (ed). (2000). The English history of A frican American English Oxford: Blackwell Poplack, Shana. (1993). Variatio n theory and language contact. In Dennis Preston (ed.), American Dialect Research: An Anthology Celebrating the 100th Anniversary of the American Dialect Society pp. 25 1 286. A msterdam: Benjamins. Poplack, Shana, & Stephen Levey. (2010). Contact induced grammatical change In Peter Auer & Jrgen Erich Schmidt (eds.), Language and Space An international handbook of linguistic variation: Volume 1 Theories and methods Berlin: Mouton de Gruyter. Poplack, Shana & Marjory Meechan. ( 1998 ) How languages fit together in code mixing. International Journal of Bilingualism 2: 127 138. s inflection in early Black English. Language Variation and Change 1:47 84. Poplack, S hana, & S ali Tagliamonte ( 2001 ). A frican American English in the d iaspora. Oxford: Blackwell.
189 Port, Robert F., & Rosemarie Rotunno. (1979) Relation between voice onset time and vowe l duration. Journal of the Acoustical Society of America 66 : 654 662. Regan, Vera. (1996). Variation in French interlanguage: A longitudinal study of sociolinguistic competence. In Robert Bayley & Dennis Preston (eds.), Second Language Acquisition and Lingu istic Variation pp. 177 202 Amsterdam: John Benjamins. Reinecke, John E. (1937). Marginal languages: a sociological survey of the creole languages and trade jargons Yale University, Ph.D. Dissertation. Rickford, John R. (1987). Dimensions of a creole c ontinuum: History, texts, & linguistic analysis of Guyanese Creole Stanford: Stanford University Press. Rickford, John R. (2006). Down for the count? The Creole origins hypothesis of AAVE at the hands of the Ottawa Circle, and their supporters. Journal o f Pidgins and Creole Languages 21:97 155. Robe, S tanley L ( 1960 ) The Spanish of rural Panama. Berkeley: University of California Press. Roca, A na & J ohn M. Lipski. (1993) Spanish in the United States: Linguistic contact and d iversity Studies in Anth ropological Linguistics 6 Berlin: Mouton de Gruyter. Roldan, Yasna, & Jaime Soto Barba. ( 1997 ) Voice onset times for /p t k/ and /b d g/ in Valdivian Spanish: An acoustic analysis. [El V.O.T. de /p t k/ y /b d g/ en el es pa ol de Valdivia: un anlisis ac stico] Estudios Filolgicos 32 : 27 33. Romaine, Suzanne. (1995). Bilingualism (2nd edition) Oxford: Blackwell. Rosner, Burton. S., Luis E. Lopez Bascuas, Jose E. Garcia Albea, & Richard P. Fahey. (2000). Voice on set times for Castilian Spanish initial st ops. Journal of Phonetics 28 : 217 224. Sancier, Michele L. & Carol A. Fowler. ( 1997 ). Gestural drift in a bilingual speaker of Brazilian Portuguese and English. Journal of Phonetics 25:421 436. Sankoff, David, Sali Tagliamonte, & Eric Smith. (2005) Goldvar b X: A variable rule application for Macintosh and Windows Department of Linguistics, University of Toronto. Sankoff, Gillian. (2002). Linguistic outcomes of language contact. In J. K. Chambers, Peter Trudgill, & Natalie Schilling Estes (eds.) Handbook of Language Variation and Change pp. 638 668 Oxford: Blackwell. Sayahi, Lofti. (2007). Diglossia and contact induced language change. International Journal of Multilingualism 4:38 51.
190 Schmidt, Anna Marie, & James E. Flege. ( 1996 ) Speaking rate effects in stops produced by Spanish and English monolinguals and Spanish/English b ilinguals. Phonetica 53: 162 179. Silva Corvaln, Carmen. (1990). Current issues in studies of language contact. Hispania 73:162 176. Silva Corvaln, C armen. (1994) Language Contact and Change: Spanish in Los Angeles OUP, Oxford. Simonet, Miquel. (2010). Dark and clear laterals in Catalan and Spanish: Interaction o phonetic categories in early bilinguals. Journal of Phonetics 38:663 678. Snow, David. (1997). Children's acquisition o f speech timing in English: A comparative study of voice onset time and final syllable vowel lengthening. Journal of Child Language 24 : 35 56. Snow, Peter. ( 2000 ) Language variation in Caribbean creole/non lexifier contact situation s: Continua or diglossia ? Texas Linguistics Forum 44: 148 162. Snow, Peter. (2003). Talking with tourists in a Panamanian creole village: An emerging site of production. Journal of Pidgin and Creole Languages 18:299 309. Snow, Peter. (2007). Vernacular shift: Language and the bui ld environment in Bastimentos, Panama. Identities: Global Studies in Power and Culture 14:161 182. Stevens, Kenneth N. (1998). Acoustic phonetics Cambridge, Mass: MIT Press. Stockwell, Robert P., & Jean Donald Bowe n ( 1965 ) The sounds of English and Sp anish Chicago: University of Chicago Press. Tagliamonte, Sali. (2006). Analyzing sociolinguistic variation Cambridge: Cambridge University Press. Tagliamonte, Sali. (2012). Variationist sociolinguistics: Change, observation, interpretation. Oxford: Wile y Blackwell. Taylor, Douglas R. (1961). New languages for old in the West Indies. Comparative Studies in Society and History 3:277 288. Thomas, Erik R. ( 2002 ) Instrumental analysis. In J. K. Chambers, Peter Trudgill, & Natalie Schilling Estes (eds.) Han dbook of Language Variation and Change pp. 168 200. Oxford: Blackwell. Thomas Brereton, Leticia. C. (1992). An exploration of Panamanian Creole English: Some syntactic, lexical and sociolinguistic features New York: New York University, Ph.D. Dissertatio n.
191 Thomason, Sarah G. (2001). Language contact: An introduction Washington, D.C.: Georgetown University Press. Thomason, Sarah G., & Terrence Kaufman. (1988). Language contact, creolization, and genetic linguistics. Berkeley: University of California Pres s. Thompson, Robert W. (1961). A note on some possible affinities between the creole dialects of the Old World and those of the New. In R.B. LePage (ed.), Proceedings of the 1959 conference on creole languages studies. Creole language studies 2 pp. 107 11 3. London: MacMillan. Thornburg, D ianne & J ohn H Ryalls. (1998) Voice onset time in Spanish English bilinguals: Early versus late learners of English. Journal of Communication Disorders 31: 215 229. Torres Cacoullos, Rena, & Jessica E. Aaron. ( 2003 ) Bar e English origin nouns in Spanish: Rates, constraints, and discourse functions. Language Variation and Change 15 : 289 328. van Coetsem, Frans. ( 1988 ) Loan phonology and the two transfer types in language contact Dordrecht: Foris Publications Holland. Vinson, Julien. (1889). Crole. In Adolphe Bertillon et al. (eds.), Dictionnaire de sciences anthropologiques pp. 345 347. Paris : Doin. Weinreich, Uriel. (1966). Languages in contact: Findings and problems The Hague: Mouton. Weinreich, Uriel, Willi am L abov, & Marvin I. Herzog. ( 1968 ) Empi rical Foundations for a theory of language change. In Winfred P. Lehmann & Yakov Malkiel (eds.), Directions for historical linguistics: A symposium pp. 95 189. Austin: University of Texas Press. Weismer, G ary. ( 1979 ) Sensitivity of VOT measures to certain segmental features in speech production. Journal of Phonetics 7 : 197 204. Winford, Donald. (2007). Some issues in the study of language contact. Journal of Language Contact Thema 1:22 39. Wolfram, Walt, & Natalie Schi lling Estes. (2006). American English Second edition. Malden and Oxford: Blackwell. Yavas, M ehmet & R enee Wildermuth. ( 2006 ) The effects of place of articulation and vowel height in the acquisition of English aspirated stops by Spanish speakers. IRAL 4 4: 251 263.
192 Yoshioka, Hirohide, Anders Lfqvist, & Hajime Hirose. (1981). Laryngeal adjustment in the production of consonant clusters and geminates in American English. Journal of the Acoustical Society of America 70:1615 1623. Young, Richard F. ( 1991 ) Va riation in interlanguage morphology New York: Peter Lang. Zampin i, Mary L., & Kerry P. Green. (2001) The voicing contrast in English and Spanish: The relationship between perception and production. In J anet L. Nicol (ed.), One mind, two languages: Biling ual processing pp. 23 48. Boston: Blackwell.
193 BIOGRAPHICAL SKETCH Delano S. Lamy is originally from Brooklyn, NY with strong West Indian roots in Panama and Trinidad and Tobago. In 2002, he received two Bachelor of Arts, one in international studies and a nother in Spanish from the University of South Florida in Tampa, Florida. In 2007, he came to the University of Florida to purse a Master of Arts and a Doctor of Philosophy in Spanish with a concentration in Hispanic linguistics. During his time at UF, he has had the opportunity to teach various Spanish language courses, as well as Introduction to Hispanic Linguistics. He has also served as a teaching assistant for Dr. Jessi Elana Aaron in the course Language Use and Language Change in the Spanish speaking World. Delano has completed his Doctor of Philosophy and will begin teaching at SUNY Geneseo in Geneseo, NY in the fall of 2012.